Razor Financial Principals

Version 3.2
Razor Financial Principals

Contents
Chapter 1 Introduction ................................................................... 14

1.1 RAZOR Financial Principles Document ............................................ 14

1.2 Credit Risk ............................................................................ 14
1.3 Market Risk ........................................................................... 15

Chapter 2 Monte Carlo Simulation ..................................................... 17

2.1 Introduction .......................................................................... 17

2.2 Specification of the Monte Carlo Simulation Process ........................... 18
2.2.1 Historical Statistical Simulation Parameters ....................... 18
2.2.2 Lognormal model ....................................................... 18
2.2.3 Normal Mean-Reverting model ....................................... 19
2.2.4 Lognormal Mean-Reverting model ................................... 22
2.3 Random Number Generation ....................................................... 22
2.4 Correlation Matrix and Eigenvalues Analysis ..................................... 22
2.4.1 Lognormal Model ....................................................... 22
2.4.2 Mean-Reverting models ................................................ 23
2.4.3 Correlation Matrix ...................................................... 23
2.4.4 Eigenvalues and Eigenvectors ........................................ 24
2.4.5 Principal Component Analysis ........................................ 25
2.4.6 Daily Simulations ....................................................... 26
2.5 Generation of the Monte Carlo Simulation ....................................... 27
2.6 Equity Simulation .................................................................... 27
2.6.1 Introduction ............................................................. 27
2.6.2 Model Specification .................................................... 28
2.6.3 Simulation Algorithm................................................... 29
2.6.4 Parameter Estimation for Secondary Equities ...................... 31

Chapter 3 Credit Exposure Calculation ............................................... 37

3.1 Pre-Settlement Risk ................................................................. 37

3.1.1 Calculation of Credit Exposure ....................................... 37
3.1.2 Treatment of Credit Nodes ........................................... 37
3.1.3 Economic Offsetting ................................................... 37
3.1.4 Netting Agreements .................................................... 37
3.1.5 Collateral Agreements ................................................. 38
3.2 Credit Exposure Measures .......................................................... 38
3.2.1 Introduction ............................................................. 38

©2009 Razor Risk Technologies Page 2 of 373

Razor Financial Principals

3.2.2 Key Measures ............................................................ 38

3.3 Settlement Risk ...................................................................... 40
3.3.1 Transaction/Cashflow Identification ................................ 40
3.3.2 Exposure Duration ...................................................... 40
3.3.3 Settlement Exposure Calculation .................................... 41
3.3.4 Settlement Netting ..................................................... 42
3.3.5 Spread Risk .............................................................. 42

Chapter 4 Credit Risk Measurement ................................................... 43

4.1 Modelling Credit Transition Losses ................................................ 44

4.1.1 Credit Rating Transition Matrices .................................... 44
4.1.2 Simulation of Credit Rating Changes ................................ 47
4.1.3 Credit Transition Co-Movement ...................................... 48
4.1.4 Generation of Credit Rating Changes ................................ 48
4.1.5 Credit Spread Risk Loss ................................................ 49
4.1.6 Default Risk Loss........................................................ 49
4.2 Credit Value-at-Risk ................................................................. 49
4.3 Economic Capital .................................................................... 50
4.3.1 Present Value of Losses................................................ 51
4.3.2 Aggregation of Losses to Final Credit Node ......................... 51
4.3.3 Default Treatment ..................................................... 51

Chapter 5 Market Risk Measurement .................................................. 53

5.1 Value-at-Risk (VaR) .................................................................. 53

5.2 Historical Simulation ................................................................ 54
5.2.1 Filtered Historical VaR Simulation ................................... 55
5.3 Monte Carlo Simulation ............................................................. 56
5.4 Analysis of VaR ....................................................................... 56
5.4.1 Partial Risk .............................................................. 57
5.4.2 Incremental VaR ........................................................ 57
5.4.3 Marginal VaR ............................................................ 57

Chapter 6 Market Rates, Prices and Curves.......................................... 58

6.1 The Market - Rates and Prices ..................................................... 58

6.2 Interest Rate Curves ................................................................. 58
6.2.1 Interest Rates and Security Prices ................................... 58
6.2.2 Market specific interest rate curves ................................. 60
6.2.3 Spread Yield Curve ..................................................... 61
6.2.4 Composite Yield Curve ................................................. 61

©2009 Razor Risk Technologies Page 3 of 373

Razor Financial Principals

6.2.5 Adjustment Factors Applied to Yield Curves ....................... 62

6.2.6 Negative Forward Rates ............................................... 62
6.3 Yield Curve Construction ........................................................... 63
6.3.1 Bootstrapping ........................................................... 63
6.3.2 Bootstrapping the Yield Curve Using Different Securities ........ 63
6.3.3 Interpolation Methods ................................................. 71
6.3.4 Zero Treatment ......................................................... 84
6.3.5 Summary of Bootstrapping Formulas ................................ 86
6.3.6 Demonstration of Bootstrapping ..................................... 87
6.4 CDS Spread Curves ................................................................... 90
6.5 FX Market Data ....................................................................... 90
6.6 Equity Market Data .................................................................. 90
6.7 Commodity Forward Curves ........................................................ 90
6.8 Volatility Market Data ............................................................... 90
6.8.1 Interest Rate Volatility ................................................ 90
6.8.2 Equity Volatility ........................................................ 91
6.8.3 FX Volatility ............................................................. 91
6.8.4 Commodity Volatility .................................................. 91
6.9 Calendars.............................................................................. 92

Chapter 7 RAZOR’S Product Support .................................................. 94

7.1 Design Objectives in RAZOR ’S Product Support................................. 94

7.2 Product Representation ............................................................. 94
7.2.1 The FinMark Schema ................................................... 94
7.3 Adding New Products ................................................................ 97
7.4 Pricing Contexts ...................................................................... 98
7.5 Transitioning .......................................................................... 98
7.5.1 Market Transitioning ................................................... 98
7.5.2 Trade Transitioning .................................................... 99

Chapter 8 Common XML Structures ................................................... 100

Chapter 9 FX and FX Derivatives ...................................................... 104

9.1 FX Forwards ......................................................................... 104

9.1.1 Description of Instrument ........................................... 104
9.1.2 XML Representation .................................................. 104
9.1.3 Pricing .................................................................. 105
9.1.4 Currency Calculations................................................ 105
9.2 FX Vanilla Options ................................................................. 106

©2009 Razor Risk Technologies Page 4 of 373

Razor Financial Principals

9.2.1 Description of Instrument ........................................... 106

9.2.2 XML Representation .................................................. 106
9.2.3 Pricing .................................................................. 108
9.3 FX Single Barrier Options ......................................................... 109
9.3.1 Description of Instrument ........................................... 109
9.3.2 XML Representation .................................................. 109
9.3.3 Pricing .................................................................. 112
9.4 FX Double Barrier Option ......................................................... 114
9.4.1 Description of Instrument ........................................... 114
9.4.2 XML Representation .................................................. 114
9.4.3 Pricing .................................................................. 116
9.5 FX Digital Option ................................................................... 117
9.5.1 Description of Instrument ........................................... 117
9.5.2 XML Representation .................................................. 118
9.5.3 Pricing .................................................................. 120
9.6 FX Digital Barrier Option .......................................................... 120
9.6.1 Description of Instrument ........................................... 120
9.6.2 XML Representation .................................................. 120
9.6.3 Pricing .................................................................. 122
9.7 FX Double Digital Option .......................................................... 123
9.7.1 Description of Instrument ........................................... 123
9.7.2 XML Representation .................................................. 123
9.7.3 Pricing .................................................................. 127
9.8 FX Average Rate Options .......................................................... 128
9.8.1 Description of Instrument ........................................... 128
9.8.2 XML Representation .................................................. 128
9.8.3 Pricing .................................................................. 132
9.9 FX Fade-In Option .................................................................. 133
9.9.1 Contract Definition ................................................... 133
9.9.2 Model Specification .................................................. 134
9.9.3 Pricing Formulas ...................................................... 135

Chapter 10 Base Metals ................................................................... 136

10.1 Metal Forwards ..................................................................... 136

10.1.1 Description of Instrument ........................................... 136
10.1.2 XML Representation .................................................. 136
10.1.3 Pricing .................................................................. 137
10.2 Metal Vanilla Options .............................................................. 137
10.2.1 Description of Instrument ........................................... 137
10.2.2 XML Representation .................................................. 137

©2009 Razor Risk Technologies Page 5 of 373

Razor Financial Principals

10.2.3 Pricing .................................................................. 138

10.3 Metal Single Barrier Options...................................................... 139
10.3.1 Description of Instrument ........................................... 139
10.3.2 XML Representation .................................................. 139
10.3.3 Pricing .................................................................. 140
10.4 Metal Double Barrier Option ..................................................... 141
10.4.1 Description of Instrument ........................................... 141
10.4.2 XML Representation .................................................. 141
10.4.3 Pricing .................................................................. 142

Chapter 11 Commodities and Energy .................................................. 144

11.1 Commodity Forwards .............................................................. 144

11.1.1 Description of Instrument ........................................... 144
11.1.2 XML Representation .................................................. 144
11.1.3 Pricing .................................................................. 145
11.2 Electricity Futures ................................................................. 145
11.2.1 Introduction ........................................................... 145
11.2.2 Definition .............................................................. 145
11.2.3 Valuation .............................................................. 145
11.3 Commodity Vanilla Options ....................................................... 145
11.3.1 Description of Instrument ........................................... 145
11.3.2 XML Representation .................................................. 146
11.3.3 Pricing .................................................................. 146
11.4 Commodity Single Barrier Options ............................................... 147
11.4.1 Description of Instrument ........................................... 147
11.4.2 XML Representation .................................................. 147
11.4.3 Pricing .................................................................. 148
11.5 Commodity Double Barrier Option............................................... 150
11.5.1 Description of Instrument ........................................... 150
11.5.2 XML Representation .................................................. 150
11.5.3 Pricing .................................................................. 151
11.6 Commodity Average Rate Swap .................................................. 152
11.6.1 Description of Instrument ........................................... 152
11.6.2 XML Representation .................................................. 152
11.6.3 Pricing .................................................................. 155
11.7 Commodity Average Rate Options ............................................... 156
11.7.1 Description of Instrument ........................................... 156
11.7.2 XML Presentation ..................................................... 156
11.7.3 Pricing .................................................................. 158

©2009 Razor Risk Technologies Page 6 of 373

Razor Financial Principals

Chapter 12 Interest Rate Derivatives .................................................. 160

12.1 Forward Rate Agreements ........................................................ 160

12.1.1 Description of Instrument ........................................... 160
12.1.2 XML Representation .................................................. 160
12.1.3 Pricing .................................................................. 160
12.2 Interest Rate and Cross Currency Swaps........................................ 161
12.2.1 Description of Instrument ........................................... 161
12.2.2 XML Representation .................................................. 161
12.2.3 Pricing .................................................................. 165
12.2.4 Other Variations of Swaps ........................................... 165
12.3 Interest Rate Futures .............................................................. 166
12.3.1 Description of Instrument ........................................... 166
12.3.2 XML Representation .................................................. 166
12.3.3 Pricing .................................................................. 175
12.4 Bond Options, Options on Bond Futures, and Options on Money Market or Bill
Futures .............................................................................. 177
12.4.1 Description of Instrument ........................................... 177
12.4.2 XML Representation .................................................. 177
12.4.3 European Exercise Pricing ........................................... 180
12.4.4 American Exercise Pricing ........................................... 181
12.5 Caps and Floors .................................................................... 182
12.5.1 Description of Instrument ........................................... 182
12.5.2 XML Representation .................................................. 182
12.5.3 Pricing .................................................................. 183
12.6 Collars ............................................................................... 184
12.6.1 Description of Instrument ........................................... 184
12.6.2 XML Representation .................................................. 184
12.6.3 Pricing .................................................................. 184
12.7 Swaption ............................................................................ 185
12.7.1 Description of Instrument ........................................... 185
12.7.2 XML Representation .................................................. 185
12.7.3 Pricing .................................................................. 188
12.7.4 Calculating Greeks ................................................... 189
12.8 European Swaption ................................................................ 192
12.8.1 Contract Definition ................................................... 192
12.8.2 Pricing Formulas ...................................................... 193
12.9 Bermudan Swaption ............................................................... 193
12.9.1 Contract Definition ................................................... 193
12.9.2 Pricing .................................................................. 194
12.10 Constant Maturity Swap ........................................................... 196

©2009 Razor Risk Technologies Page 7 of 373

Razor Financial Principals

12.10.1 Description of Instrument ........................................... 196

12.10.2 XML Representation .................................................. 196
12.10.3 Pricing .................................................................. 196
12.11 Power Reverse Dual Currency Swap ............................................. 198
12.11.1 Pricing FX-Linked Coupon Leg ...................................... 198
12.11.2 Pricing Floating Leg .................................................. 200
12.12 Cancellable Yield Curve Basket Swap with Floor .............................. 201
12.12.1 Pricing Option Leg .................................................... 201
12.12.2 Pricing Floating Fix Leg .............................................. 203
12.13 Cancellable Swap with Equity Trigger .......................................... 204
12.13.1 Pricing Fixed Leg ..................................................... 204
12.13.2 Pricing Floating Leg .................................................. 205
12.13.3 Cancellable Swap ..................................................... 205
12.13.4 Cancellable Swap with Equity Trigger ............................. 205
12.14 Cancellable Reverse Interest Rate Swap ....................................... 205
12.14.1 Pricing Reverse Floating Side ....................................... 205
12.14.2 Reverse Floating Side with Floor ................................... 206
12.14.3 Pricing Fixed-Floating Leg........................................... 206
12.15 Interest Rate Cliquet .............................................................. 207
12.15.1 Payoff of Interest Rate Cliquet ..................................... 207
12.15.2 Pricing .................................................................. 207
12.16 Range Accrual Swap ............................................................... 208
12.16.1 Pricing Range Accrual Note Leg .................................... 208
12.16.2 Pricing Floating-Fix Leg.............................................. 211
12.17 Constant Maturity Cap and Floor ................................................ 211
12.17.1 Pricing .................................................................. 211
12.18 Early Exercisable Interest Rate Derivatives .................................... 212
12.19 Bermudan Swaption – Geske’s Approximation ................................. 213
12.20 Numerical Interest Rate Models ................................................. 214
12.20.1 Hull-White One Factor Interest Rate Tree Model (Equal Time
Step).................................................................... 215
12.20.2 Hull-White One Factor Interest Rate Tree Model and Calibration
Method (Unequal Time Step) ....................................... 218

Chapter 13 Commercial Lending ........................................................ 222

13.1 Collateral ........................................................................... 222

13.2 Cash Advance Facility ............................................................. 222
13.2.1 Cash Advance Facility XML Representation ....................... 222
13.2.2 Cash Advance Facility Credit Issues................................ 224
13.2.3 Cash Advance Facility Pricing....................................... 224
©2009 Razor Risk Technologies Page 8 of 373
Razor Financial Principals

13.3 Generic Interest Rate ............................................................. 224

13.3.1 Description of Instrument ........................................... 224
13.3.2 XML Representation .................................................. 224

Chapter 14 Securities...................................................................... 225

14.1 Promissory Note .................................................................... 225

14.1.1 Description of Instrument ........................................... 225
14.1.2 XML Representation .................................................. 225
14.2 Commercial Bill .................................................................... 225
14.2.1 Description of Instrument ........................................... 225
14.2.2 XML Representation .................................................. 225
14.2.3 Pricing .................................................................. 226
14.3 Certificate of Deposit ............................................................. 226
14.3.1 Description of Instrument ........................................... 226
14.3.2 XML Representation .................................................. 226
14.3.3 Pricing .................................................................. 226
14.4 Commercial Paper ................................................................. 226
14.4.1 Description of Instrument ........................................... 226
14.4.2 XML Representation .................................................. 227
14.4.3 Credit Implications ................................................... 227
14.4.4 Pricing .................................................................. 227
14.5 Bank Accepted Bill ................................................................. 227
14.5.1 Description of Instrument ........................................... 227
14.5.2 XML Representation .................................................. 227
14.5.3 Pricing .................................................................. 227
14.6 Bonds ................................................................................ 228
14.6.1 Description of Instrument ........................................... 228
14.6.2 XML Representation .................................................. 228
14.6.3 Credit Implications ................................................... 229
14.6.4 Pricing .................................................................. 229
14.6.5 Spread Calculation ................................................... 230
14.7 Bonds – Capital Indexed ........................................................... 230
14.7.1 Description of Instrument ........................................... 230
14.7.2 XML Representation .................................................. 230
14.7.3 Credit Implications ................................................... 231
14.7.4 Pricing .................................................................. 231
14.8 Bonds – Indexed Annuity .......................................................... 232
14.8.1 Description of Instrument ........................................... 232
14.8.2 XML Representation .................................................. 232
14.8.3 Credit Implications ................................................... 232

©2009 Razor Risk Technologies Page 9 of 373

Razor Financial Principals

14.8.4 Pricing .................................................................. 232

14.9 Inflation Linked Bond with Indexation Lag ..................................... 233
14.9.1 UK ILG Eight-Month Indexation Lag ................................ 233
14.9.2 Inflation-Linked Bonds with Three-Month Indexation Lag
(“Canadian Model”) .................................................. 236
14.10 Perpetuity Bond .................................................................... 239
14.11 Cash Bond ........................................................................... 240
14.12 Floating Rate Notes ................................................................ 242
14.12.1 Description of Instrument ........................................... 242
14.12.2 XML Representation .................................................. 242
14.12.3 Pricing .................................................................. 242
14.13 Repurchase Agreements .......................................................... 244
14.13.1 Description of Instrument ........................................... 244
14.13.2 XML Representation .................................................. 244
14.13.3 Credit Implications ................................................... 244
14.13.4 Pricing .................................................................. 245
14.14 General Collateralised Repo...................................................... 245
14.14.1 Sterling GC Trades ................................................... 246
14.14.2 Euro GC Trades ....................................................... 246
14.15 Bill Index Deposits ................................................................. 247
14.15.1 Guaranteed Bill Index Deposits ..................................... 247
14.15.2 Protected Bill Index Deposits ....................................... 247

Chapter 15 Equities 254

15.1 Ordinary Shares .................................................................... 254

15.1.1 Description of Instrument ........................................... 254
15.1.2 XML Representation .................................................. 254
15.1.3 Credit Implications ................................................... 254
15.1.4 Pricing .................................................................. 254
15.2 Equity Forwards .................................................................... 254
15.2.1 Basket Equity Forward ............................................... 254
15.2.2 Foreign Equity Forward .............................................. 257
15.3 Equity Options ...................................................................... 261
15.3.1 Description of Instrument ........................................... 261
15.3.2 XML Representation .................................................. 261
15.3.3 Pricing .................................................................. 262
15.3.4 Equity Option Greeks ................................................ 269
15.4 Equity Index Options .............................................................. 271
15.4.1 Description of Instrument ........................................... 271
15.4.2 XML Representation .................................................. 271
©2009 Razor Risk Technologies Page 10 of 373
Razor Financial Principals

15.4.3 Pricing .................................................................. 271

15.5 Equity Basket Option .............................................................. 271
15.5.1 Contract Definition ................................................... 271
15.5.2 Valuation 1 ............................................................ 271
15.5.3 Valuation 2 ............................................................ 273
15.5.4 Equity Basket Option with Quanto Feature ....................... 275
15.5.5 Summary ............................................................... 275
15.6 Equity Futures ...................................................................... 275
15.6.1 Description of Instrument ........................................... 275
15.6.2 XML Representation .................................................. 275
15.6.3 Pricing .................................................................. 276
15.7 Asian Options: Introduction ...................................................... 277
15.7.1 Types of Averaging and Fixing ...................................... 278
15.7.2 Geometric Average Asian Option Pricing Models................. 279
15.7.3 Arithmetic Average Asian Option Pricing Models (Continuous
Averaging) ............................................................. 281
15.7.4 Arithmetic Average Asian Option Pricing Models (Discrete
Averaging, Equal Fixing Weighting and Constant Underlying
Variables) .............................................................. 290
15.7.5 Arithmetic Average Asian Option Pricing Models (Discrete
Averaging, Variable Fixing Weighting and Time-Dependent
Underlying Parameters) ............................................. 297
15.7.6 References ............................................................ 320
15.7.7 Appendix (Proofs of the Identities for Double Average Rate
Options): ............................................................... 322
15.8 Equity Swap ......................................................................... 324
15.8.1 Contract Definition ................................................... 324
15.8.2 Definition for Notations ............................................. 324
15.8.3 Valuation .............................................................. 325

Chapter 16 Credit Derivatives ........................................................... 329

16.1 Credit Default Swaps .............................................................. 329

16.1.1 Description of Instrument ........................................... 329
16.1.2 XML Representation .................................................. 329
16.1.3 Credit Implications ................................................... 331
16.1.4 CDS Pricing ............................................................ 331
16.1.5 Configuration of CDS Spread Curve ................................ 337
16.1.6 Pricing Parameters ................................................... 337
16.2 Collateralised Debt Obligation ................................................... 338
16.2.1 Description of Instrument ........................................... 338
16.2.2 XML Representation .................................................. 338
16.2.3 Credit Implications ................................................... 340

©2009 Razor Risk Technologies Page 11 of 373

Razor Financial Principals

16.2.4 CDO Pricing ............................................................ 341

16.2.5 Pricing Parameters ................................................... 344

Chapter 17 Pricing 345

17.1 Distribution of Forward Rates .................................................... 345

17.1.1 Definition .............................................................. 345
17.1.2 Pricing .................................................................. 346
17.1.3 Forward LIBOR Rate .................................................. 346
17.1.4 Forward CMS Rate .................................................... 346
17.2 Greeks Computation ............................................................... 346
17.2.1 Introduction ........................................................... 346
17.2.2 Definition .............................................................. 346
17.2.3 Generalised Black-Scholes Formula ................................ 347
17.2.4 Greeks .................................................................. 347
17.3 Interest Rate Risk Analytics ...................................................... 348
17.3.1 Valuation for Interest Rate Securities ............................. 350
17.3.2 Security Interest Rate Risk Analytics .............................. 351
17.3.3 Portfolio Risk Analytics .............................................. 358
17.4 Yield and Price Volatility ......................................................... 358
17.5 Day Count Conventions ............................................................ 359
17.6 Roll Conventions ................................................................... 360
17.7 Business Day Conventions ......................................................... 361
17.8 Interpolation Methods ............................................................. 362
17.8.1 Linear Interpolation .................................................. 362
17.8.2 Log-Linear Interpolation ............................................. 362
17.8.3 Cubic Interpolation ................................................... 362
17.9 Cumulative Normal Distribution Function ...................................... 363
17.10 Cumulative Bivariate Normal Distribution Function .......................... 363
17.10.1 Special Conditions .................................................... 365
17.11 Reiner and Rubinstein Single Barrier Model .................................... 366
17.12 Finite Difference Methods ........................................................ 368
17.12.1 Introduction ........................................................... 368
17.12.2 The Time Axis ......................................................... 369
17.12.3 The Price Axis ......................................................... 369
17.12.4 The Grid ............................................................... 370
17.13 Implicit/Explicit Finite Difference Methods .................................... 370
17.13.1 The Forward Difference Approximation ........................... 370
17.13.2 The Backward Difference Approximation ......................... 370
17.13.3 Averaging the Forward and Backward Approximations ......... 370

©2009 Razor Risk Technologies Page 12 of 373

Razor Financial Principals

Concept Index372

©2009 Razor Risk Technologies Page 13 of 373

Chapter 1
Introduction

1.1 RAZOR Financial Principles Document

This document examines the fundamental financial concepts RAZOR utilizes in
determining financial risk. RAZOR is an enterprise wide fully integrated credit
and market risk management system.

1.2 Credit Risk

Credit risk is the risk of a counterparty not fully meeting their financial
obligations. In attempting to manage this risk the probability, magnitude, and
possible offsetting effects must be estimated.
Traditionally, credit exposure arose from lending activities, and was measured
as simply the book value of all outstanding obligations from a counterparty.
Although book value is a fairly poor measure of credit exposure — the market
value of an obligation often diverges significantly from its book value — it does
have the advantage of being a fairly simple and consistent measure that can
provide a reasonable sense of the credit exposure to a counterparty.
Unfortunately the notion of book value starts to break down when examining
pre-settlement risk on many derivative instruments. Par swaps and bonds have
a market value of zero when they are first booked. Many other derivatives have
a credit exposure far higher than their book value. A more accurate,
probabilistic measure for potential credit exposure is required. The moment a
transaction is committed, its market value or exposure changes as time
progresses and market rates change. If this market value is in profit then there
exists a credit risk exposure, as the unrealised profit will be lost upon the
party defaulting. The problem is that only the current value is known, and
what exposure the trade will obtain during its future life.
RAZOR solves this problem by taking the current market rates and predicting
what they may be tomorrow, and measuring the exposure of the trade under
these new conditions. These predictions are produced statistically to simulate
the way market rates move in the market place. Historical data is examined to
determine the behaviour of each rate and how that rate interacts with all the
other rates. This information is used as parameters in the statistical simulation
of future market rates called market scenarios.
This process is repeated for the next day and so on until the trade expires. The
results are market values of the trade, for all the days in the duration of the
trade, which is called an exposure profile. The problem with this approach is
that a single statistical prediction of future market rate is unlikely to be
correct so this whole process is repeated many times, with the statistically
generated future markets being different each time. This produces a
distribution of possible values for each future date. Applying a confidence level
to this distribution will select a single value for each future date and therefore
produce an exposure profile. To save time only a sample, defined by the user,
of future exposures are calculated.
Exposure profiles for each trade are calculated, and then combined together in
portfolios, taking into account netting agreements, collateralisation
Razor Financial Principals

agreements and guarantees. Path-dependant trades are transitioned

appropriately as price barrier levels are encountered, or payments are rolled
off. In this way, RAZOR is able to accurately determine the sensitivities of the
portfolio to different rate and price movements, and determine the spectrum
of possible losses given specific rate movement scenarios.
RAZOR allows the risk manager to quantify and analyse credit exposures based
on the rules set by the business. Each trade is mapped into various portfolios
depending on the characteristics of the trade, the counterparty hierarchy, or
the internal business unit managing the deal.
RAZOR allows the risk manager to set and manage credit limits and also to
track the utilization of the credit limits. Different types of limits thresholds
may be set and alerts generated if threshold levels are breached.
Settlement risk is the risk that when an exchange of assets is supposed to
happen upon settlement of a trade, one of the counterparties fails to meet
their obligations. Settlement risk usually creates very large credit exposures
which last for brief periods of time. RAZOR can calculate the risk exposures
arising from settlement risk and spread the risk across days depending on the
characteristics of the trade.

1.3 Market Risk

Market risk is the risk of a decrease in value of a portfolio of investments,
trades, and positions due to changes in the market factors which determine
the market value of the portfolio.
The standard market risk factors or drivers which influence market risk are:
Equity risk, or the risk of a decrease in value due to changes in stock prices
Interest rate risk, or the risk of a decrease in value due to changes in interest
rates
Currency risk, or the risk of a decrease in value due to changes in foreign
exchange rates
Commodity risk, or the risk of a decrease in value due to changes in commodity
prices (i.e. grains, metals, etc.)
Spread risk, or the risk of a decrease in value due to changes in spreads
between interest rates
Volatility risk, or the risk of a decrease in value due to changes in implied
volatilities, as used in derivative/option revaluations
Market risk is typically measured using a Value at Risk methodology. Market
risk can also be contrasted with Specific risk, which measures the risk of a
decrease in ones investment due to a change in a specific industry or sector, as
opposed to a market-wide move.
Razor provides support for the modelling of all market risk factors for both
Historical VaR, and Monte Carlo VaR. Historical VaR for a particular time
horizon is derived from a distribution of portfolio values generated from
scenarios created by applying historical changes to the current market rates.
Monte Carlo VaR for a particular time horizon is derived from a distribution of
portfolio values generated from scenarios created by modelling the evolution
of current market rates using the stochastic processes driving the market risk
factors as implied by the historical data.

©2009 Razor Risk Technologies Page 15 of 373

Razor Financial Principals

Further market risk analytic measures to understand fully the dynamics of

market risk include Partial Risk Analysis, Market Risk Factor Sensitivity
Analysis, Scenario Analysis, Stress Testing, and Back Testing. These are all
supported by RAZOR.

©2009 Razor Risk Technologies Page 16 of 373

Chapter 2
Monte Carlo Simulation

2.1 Introduction
RAZOR employs a Monte Carlo simulation approach to predict credit exposure
distributions at predefined future dates (credit node). Exposure is determined
by sampling results from the exposure distributions based upon the required
confidence interval.
The essence of the Monte Carlo process is the stochastic simulation of market
rates. To produce the future market scenarios, RAZOR uses the form of
parametric simulations that assumes the normality or log normality of future
distributions of the market rates. The future values of the market rates
representing financial markets snapshots are simulated according to a specified
stochastic model.
Currently lognormal mean reverting, normal mean reverting and lognormal are
supported.
The stochastic models may be specifically associated with each individual rate
class, or sub class of rate classes, or simply defaulted globally to the one
model which is applied to all rate classes. In other words the user may override
the default global simulation model with a specific simulation model for
individual asset types or rate classes.
Currently all market rates which must be simulated, must not change state
moving from one day to the next. In other words the term to maturity must be
relative and not an explicit date such as would be the case for specific security
or futures for example. The historical time series of data used to generate the
stochastic parameters for a specific rate must also be constant in state
throughout the time series.
The statistical parameters of these models (e.g. volatilities, level and speed of
mean reversions, correlations, etc.) are either specified explicitly or derived
from historical time series.
The statistical parameters of simulation are pre-calculated using these past
prices of some historical sampling period. To produce realistic future scenarios
outcomes these statistical parameters can be reviewed and adjusted for all or
selected set of market rates.
The same price history can be also used for generation of meaningful
correlation matrices that reflects interdependency among market rates and
should be used to achieve the co-integration of simulated scenarios. It is
assumed that the correlations of market rates historical price changes are
stable.
Thus, the length of the historical sampling periods is important for both the
generation of parametric coefficients of stochastic simulations and for building
of non-degenerated correlation matrices. Ideally the number of price records
of the historical time series should be at least the same as the number of
simulated market rates. However, even this condition can not guarantee
positive semi-definite feature of correlation matrix when some of the price
series may be very close to linear dependency. In these situations some fast
Razor Financial Principals

methods of matrix decomposition, as, for example, Cholesky decomposition,

fail to perform. RAZOR uses the standard iterative QL algorithm for derivation
of the principal components from the correlation matrix.

2.2 Specification of the Monte Carlo Simulation Process

Results from the Monte Carlo simulation procedure critically depend on the
models used to describe the relevant market price process. Currently there are
three models implemented in RAZOR.
Lognormal Model
This model assumes that the evolution of market rates is governed by:

σ2
d ln( R t ) = (a − )dt + σdWt
2
where dWt is the Wiener process.
The model results in a lognormal distribution of the underlying rates.
Normal Mean-Reverting model
Market rates are simulated according to the following formula:
dR t = (θ − aR t ) dt + σdWt
The use of this model will produce normally distributed samples of possible
simulated market rates outcomes.
Lognormal Mean-Reverting model
The process of market rate simulations is described by:
d (ln R t ) = (θ − a ln( R t )) dt + σdWt
The distribution of the resulting simulated market rates will be lognormal.
The last two models have the attraction that they capture the stylised fact
that some market rates tend to revert to their mean. The use of these models
makes consistent the dispersion of the simulated values with one’s expectation
of “likely” values over a given time horizon: in particular rates should not be
allowed to become negative, or to assume “implausibly” large values.

2.2.1 Historical Statistical Simulation Parameters

The Monte Carlo Simulations of market rates according to the selected model
of simulation requires the generation of the appropriate stochastic differential
equation parameters.
In RAZOR the statistical parameters are calculated from the historical
observations of corresponding market rates.

2.2.2 Lognormal model

There are two parameters to be estimated: lognormal drift a and volatility σ .
In the calculation of these values it is assumed the historical observation
period is continuous, that is there are no missing observations for the period,
with there being 252 observations per year.

©2009 Razor Risk Technologies Page 18 of 373

Razor Financial Principals

Let’s assume that we have one year’s history of observations of market rate R
.
R = r− 251 , r− 250 , r− 249 ,..., r−1 , r0
Define:
∆ 1 = ln r− 250 − ln r− 251
∆ 2 = ln r− 249 − ln r− 250
.
.
.
∆ 251 = ln r0 − ln r−1
mean:
1 N −1
∆= ∑ ∆i
N − 1 i =1
variance:
1 N −1
σ2= ∑ (∆ i − ∆ ) 2
N − 2 i =1
Because we consider daily perturbations, on business days only, introduce:
1
δ= Where M = 252
M
The statistical parameters can be derived from the given data as:

σ2
σ = σ2 M = - annualised standard deviation
δ
∆ σ2
a= + - drift
δ 2

2.2.3 Normal Mean-Reverting model

Parameters to be estimated for this model are the drift parameter θ , speed of
mean reversion a and volatility σ .
As for the Lognormal model it is assumed the historical observation period is
continuous, with 252 observations per year.
Historically for these R :
R = r− 251 , r− 250 , r− 249, r− 248 ,..., r−1 , r0
Calculate:

©2009 Razor Risk Technologies Page 19 of 373

Razor Financial Principals

∆ 1 = r− 250 − r− 251
∆ 2 = r− 249 − r− 250
.
.
.
∆ 251 = r0 − r−1
mean:
1 N −1
∆= ∑ ∆i
N − 1 i =1
variance:
1 N −1
σ2 = ∑ (∆ i − ∆ ) 2
N − 2 i =1
Let’s consider:
r~1 = r− 251 , r~2 = r− 250 ,..., ~
r251 = r−1 , ~
r252 = r0
Parameters of the model can be estimated using the Method of Maximum
Likelihood which maximises the maximum likelihood estimator as a function of
the probability that the selected set of fitted parameters is “most likely” to be
correct.
The likelihood function in this case is:
L(θ , a, σ , ∆ 1 ...∆ 251 , r1 ...r251 = f (∆ 1 , r1 ) f ( ∆ 2 , r2 )... f ( ∆ 251 , r251 )
where

−
(∆i −(θ −ari )δ )2
1
f (∆i , ri ) = e 2σ 2δ
2πδσ

The maximum of this likelihood estimator coincides with the maximum of its
logarithm, so after maximising function
251
1 251
(∆ i − (θ − ari )δ ) 2
F = ln L = ∑ ln −∑
i =1 2πδσ i =1 2σ 2δ
the desired parameters can be calculated using expressions:
N −1 N −1 2 N −1 N −1

θ=
∑ i =1
∆ i ∑i =1 r~i − ∑i =1 ∆ i r~i ∑i =1 ~
ri

δ  (N − 1)∑i =1 r~i 2 −

N −1
(∑ r~ ) 
N −1
i =1 i
2

N −1 N −1 N −1

a=
∑ i =1
∆ i ∑i =1 ~
ri − ( N − 1)∑i =1 ∆ i r~i

δ  (N − 1)∑i =1 ~

N −1 2
ri − (∑ ~r ) 
N −1
i =1 i
2

and

©2009 Razor Risk Technologies Page 20 of 373

Razor Financial Principals

N −1

∑ (∆ − (θ − a~r )δ )
1
σ= 2

( N − 1)δ
i i
i =1

Proof:
For better readability, we replace all ~
ri by ri .

The log-likelihood function is given by:

251
1 251
(∆ i − (θ − ari )δ ) 2
F = ln L = ∑ ln −∑ .
i =1 2πδσ i =1 2σ 2δ

To find the maximum likelihood estimators:

∂F (∆ − (θ − ari )δ )
=∑ i
∂θ σ2
∂F (∆ − (θ − ari )δ )ri
=∑ i
∂a σ2

∂F ∂F
Set and to 0, we obtain:
∂θ ∂a

∂F (∆ − (θ − ari )δ )
=∑ i =0 … (1)
∂θ σ2
∂F (∆ − (θ − ari )δ )ri
=∑ i =0 … (2)
∂a σ2

To estimate θ , from (1), we obtain:

(N − 1)θδ − ∑ ∆ i
a= sub in (2),
δ ∑ ri

θ=
∑∆ ∑r − ∑∆ r ∑r . 2

δ ((N − 1)∑ r − (∑ r ) )
i i i i i
2 2
i i

To estimate a , from (1), we obtain:

θ=
∑ ∆ i + aδ ∑ ri sub in (2),
(N − 1)δ
∑ ∆ i ∑ ri − (N − 1)∑ ∆ i ri .
a=
(
δ ( N − 1)∑ ri 2 − (∑ ri )
2
)
Volatility can be obtained similarly by differentiating with respect to
volatility.

©2009 Razor Risk Technologies Page 21 of 373

Razor Financial Principals

2.2.4 Lognormal Mean-Reverting model

The set of parameters for the lognormal mean reverting model of simulation
can be obtained using the same procedure and expressions as for normal mean
reverting model with the exception that the input vector of historical market
rate observations:
R = r− 251 , r− 250 , r− 249 ,..., r−1 , r0
should be replaced by logs of the corresponding values:
R = ln r− 251 , ln r− 250 , ln r− 249 ,..., ln r−1 , ln r0

2.3 Random Number Generation

The Monte Carlo method relies on a random draw from a standardised (i.e.
N (0,1) normal distribution based on the generation of uniformly distributed
pseudo random numbers.
RAZOR uses the Box-Muller algorithm for generating random deviates with a
Gaussian distribution and ran2 procedure of long period random number
generation for creating a uniform random deviate.

2.4 Correlation Matrix and Eigenvalues Analysis

In RAZOR the future values of the market rates are simulated according to the
chosen stochastic model with statistical parameters derived from historical
time series.
The simulation procedure should take into account the historical
interdependency amongst selected market rates. This is achieved using the
correlation information of historical shifts around assumed parametric drifts
across the whole set of rates.
To build the correlation matrix, define new time series of normal
perturbations:

2.4.1 Lognormal Model

∆1 − ( a − σ2 )δ
2

Z1 = ,
σ δ
∆ 2 − ( a − σ2 )δ
2

Z2 = ,
σ δ
M
∆ 251 − ( a − σ2 )δ
2

Z 251 =
σ δ

©2009 Razor Risk Technologies Page 22 of 373

Razor Financial Principals

2.4.2 Mean-Reverting models

∆1 − a( θa − r%1 )δ
Z1 = ,
σ δ
∆ 2 − a( θa − r%2 )δ
Z2 = ,
σ δ
M
∆ 251 − a( θa − r%251 )δ
Z 251 =
σ δ

2.4.3 Correlation Matrix

The correlation matrix V = {υ k ,l } is defined for K market rates, where

k = 1,..., K , and l = 1,..., K , and where for each market rate the time series is
represented by {Z i }, i = 1,..., 251 .

The correlation co-efficient υ k ,l between two market rates k and l for the
time series Z k = {Z ik } and Z l = {Z il } is defined by

cov ( Z k , Z l )
υ k ,l = , where
σZ σZ k l

( )
cov Z k , Z l is the covariance defined by

cov(Z k , Z l ) =
1 N −1 k
(
∑ Z i − Z k Z il − Z l , and where
N − 2 i =1
)( )
σ Z is the standard deviation defined by
k

σZ =k
1 N −1 k
∑ Zi − Z k
N − 2 i =1
( ) , and where Z
2
k
is the mean of Z k for k ’th market

rate.

In Razor the correlation matrix is calculated directly from

cov ( Z k , Z l )
υ k ,l = , ie
σZ σZ k l

∑(Z )( Z − Z )
N −1

i
k
− Zk i
l l

υ k ,l = i =1

∑(Z ) ∑(Z − Z )
N −1 2 N −1 2

i
k
− Zk i
l l

i =1 i =1

Note that since by definition as the sample size increase, the sample series
{Z i } will approach a N(0,1) distribution, with mean Z of 0, and standard

©2009 Razor Risk Technologies Page 23 of 373

Razor Financial Principals

deviation σ Z of 1, and so with increasing sample size the correlation will

approach
1 N −1 k l
υ k ,l → ∑ Zi Zi
N − 2 i =1

This correlation matrix

V = {υ k ,l }
should be used for the transformation of the generated random vector of
independent normal components to the random vector of correlated normal
components.

2.4.4 Eigenvalues and Eigenvectors

Let G = [ g1 ,K , g k ] be a vector of independent Gaussian deviates. In order to
generate the vector of correlated Gaussian deviates with the given correlation
features, this vector G should be multiplied by the square root of the given
correlation matrix V .
There are various methods of finding a square root of a matrix, although not all
of them are equally robust. The numerical difficulty is that the correlation
matrix will be of a significant dimension and very possibly not well defined and
singular. Some fast methods like Cholesky decomposition are not always able
to perform these calculations, as a precondition is that the input correlation
matrix is positive semi-definite.
To find the square root of the correlation matrix V Razor uses principal value
decomposition in the eigenvectors space with the consequent representation of
it as:
V = B ΛB T
where B is the matrix of eigenvectors:

 β1,1 β 2,1 L β K ,1 

β1,2 β 2,2 L β K ,2 
B=
M M O M 
 
 β1, K β 2, K L β K , K 

(each column is an eigenvector) and Λ is the diagonal matrix

where λi are the elements of the main diagonal, and λ = [ λ1 ,K , λk ] is
the vector of eigenvalues.
Householder transformation and QL algorithm are used to perform this
decomposition. This approach has proven to be practically more robust.

Ref:
“Numerical Recipes in C”, Second Edition.

©2009 Razor Risk Technologies Page 24 of 373

Razor Financial Principals

Householder transformation , p470-p475.

QL algorithm, p478-p481.

We know define:
T
β j =  β j ,1 , β j ,2 ,K , β j , K 

= the eigenvector corresponding to the j th eigenvalue λ j .

It can be shown V = BΛBT and another way of writing V is:

K
V = ∑ λ j β j βTj .
j =1

2.4.5 Principal Component Analysis

For simulation purposes, RAZOR uses only dominant components of the
K
obtained eigenvalues, i.e. we approximate V = ∑λ β β
j =1
j j
T
j as:

K
V≈ ∑ λ j β j βTj where λ = [ λ1 ,..., λK ] are sorted in ascending order here. To
j = K −n
determine the ‘dominant’ eigenvalues (i.e. to find n ), we calculate the ratio
of convergence or level of representation of the square root matrix calculated
using selected components to the square root matrix derived from all
components. The ratio Rd n is defined as:

∑ λk
K

Rd n = k = K −n
, for n = 0,..., K − 1 , and note that λ = [ λ1 ,..., λK ] are sorted in
∑ λ
K
k =1 k
ascending order.

It appears that for the correlation noise matrices prepared on two years
observation history of approximately 1000 market rates, only a few principal
components are required (very often less than 20) in order to represent 95%
(this percentage is configurable) of the correlation matrix transformational
behaviour. In other words, there will be a significant reduction in the
dimensionality of the problem of the correlation noise building if only principal
components are chosen to represent the required level of correlation
significance.

Example:
If we have three asset classes and we obtained the covariance matrix:

1 0.5 0.9 
V = 0.5 1 0.3 .
0.9 0.3 1 
©2009 Razor Risk Technologies Page 25 of 373
Razor Financial Principals

After decomposition,
V
= BΛBT
0.73 0.17 0.66  0.07 0 0  0.73 -0.18 -0.65
= -0.18 -0.88 0.44 0 0.75 0  0.17 -0.88 0.44  .
  
-0.65 0.44 0.61  0 0 2.17  0.66 0.44 0.61 

∑ λk
K

Recall Rd n = k = K −n
.
∑ λk
K
k =1

If we set the confidence interval to be 95%, then we can find that n = 2 where
Rd 2
2.17 + 0.75
=
2.17 + 0.75 + 0.07
= 97.66% .
Then
K
V≈ ∑
j = K −n
λ j β j βTj

 0.66  0.17 
= 2.17 ⋅  0.44  [ 0.66 0.44 0.61] + 0.75 ⋅  −0.88 [ 0.17 -0.88 0.44]
 

 0.61  0.44 
0.97 0.52 0.93 
= 0.52 1.00 0.29  which is a very good approximation for V .
0.93 0.29 0.95 
It is computationally important to include only the dominant eigenvalues. For
example in daily simulations, if we have 1000 asset classes but only 20
dominant eigenvalues, then instead of looping through all 1000 eigenvalues and
eigenvectors, we only need to loop through 20 eigenvalues and eigenvectors.
It is also possible to filter the eigenvalues using a fixed threshold instead of a
confidence interval.

2.4.6 Daily Simulations

The formulae of discrete daily simulations could be presented as:
For the lognormal simulation model:
σ i2
Ri1 = Ri0 exp(aiδ − 2 δ + σ i δ ( β1,i λ1ϖ 11 + β 2,i λ2ϖ 21 + ... + β19,i λ19ϖ 191 ))
Ri1+1 = Ri0+1 exp(ai +1δ − σ2i+1 δ + σ i +1 δ ( β1,i +1 λ1ϖ 11 + β 2,i +1 λ2ϖ 21 + ... + β19,i +1 λ19ϖ 19
2
1
))

©2009 Razor Risk Technologies Page 26 of 373

Razor Financial Principals

and for time t and k market rate:

Rkt = Rkt −1 exp(a k δ − σ2k δ + σ k δ ( β 1, k λ1ϖ 1t + β 2, k λ 2 ϖ 2t + ... + β 19, k λ19 ϖ 19t ))

2

For mean reverting models:

R1j = R 0j + (θ j − a j R 0j )δ + σ j δ ( β1, j λ1ϖ 11 + β 2, j λ2ϖ 21 + ... + β19, j λ19ϖ 191 )

R1j +1 = R 0j +1 (θ j +1 − a j +1 R 0j +1 )δ + σ j +1 δ (β1, j +1 λ1ϖ 11 + β 2, j +1 λ2ϖ 21 + ... + β19, j +1 λ19ϖ 19
1
)

and for time t and k market rate:

Rkt = Rkt −1 + (θ k − ak Rkt −1 )δ + σ k δ ( β1,k λ1ϖ 1t + β 2,k λ2ϖ 2t + ... + β19, k λ19ϖ 19t )

2.5 Generation of the Monte Carlo Simulation

The Monte Carlo simulation can be pre-generated and stored for use in later
risk analysis jobs. This provides additional flexibility is scheduling operational
procedures, and allows multiple jobs to utilise the same simulation reducing
required processing resources.
The Razor simulation produces a daily simulated value for each rate defined
within the market. When the future simulated tenor corresponds to a credit
node the ratio of the simulated rate to the most recent observed market rate
is stored to a memory mapped file. For the purposes of simulation business
days are included, assuming 365 days per year. For example a node at 1 week
would require daily rates be generated for 7 days with the 7th rate begin
recorded for that node. This process is repeated so that the ratio file contains
a ratio for each market rate, path, and credit node.
Within the simulation the ratios for the required path, credit node, and market
rates are multiplied with the current base market values to obtain the
perturbed market rate values. Whilst the number of paths and credit nodes are
defined within the system configuration, the market rates required are
determined by performing a rate dependency analysis for the trades to be
priced for that scenario.
The ratio file can be overridden by defining a market perturbation process
within a user definable scenario adapter, to perturb the market during the risk
processing.

2.6 Equity Simulation

Note:
Simulation method for base equities coincides with the method to
simulate log-normal rates in Razor Financial Principle section 2.2.2.

2.6.1 Introduction
This section discusses on the generation of future equity prices using
Monte-Carlo simulation. Equities are categorised into two fundamental
types: base equity and secondary equity. Base equities are broad-market
©2009 Razor Risk Technologies Page 27 of 373
Razor Financial Principals

indices like S&P 500 and secondary equities are those equities that are
not broad enough to represent any particular market but whose
movement is correlated closely with a particular base equity.

2.6.2 Model Specification

Base Equities
Base equities are broad-market indices. Common base equities are like
S&P500, DJIA, etc. Base equities represent a particular section of the
market thus it is linked to the whole market movement closely. If each of
the base equities moves closely with the overall market, it is apparent
that all base equities should be correlated to each other. We are now to
construct the base equity model mathematically.

Define:
Si ( t ) = price of the i-th base equity at time t.
µi = drift coefficient of the i-th base equity.
σi = volatility of the i-th base equity.
β = Principal decomposition of the covariance matrix.
β ij = element of matrix β .
W j (t ) = Brownian motion for base equity j at time t.
N = the set of base indices {1, 2,..., n} .

The stochastic differential equation (SDE) representing the base equity is:
n
dSi ( t ) = µi Si dt + Si ∑ β ij dW j ( t ) .
j =1

Note that if there is only one base index, then β11 = σ 1 .

Apply Ito’s Lemma to solve the SDE and obtain:

 1  n 
S i (t ) = S i (0 ) exp  µ i − σ i2 dt + ∑ β ij dW j  .
 2  j =1 
Secondary Equities
Secondary equities are equities that are not broad enough to represent
any particular market but whose movement is linked closely with a
particular base equity. The secondary equities have two major sources of
risks, i.e. the risk associated with the overall market and the risk
associated with the linked base equity. We are now to construct the
secondary equity model mathematically.

Define:
Ski ( t ) = price of secondary equity k linked to base equity i at time t.
µ i
k = drift coefficient of the secondary equity k linked to base
equity i.

©2009 Razor Risk Technologies Page 28 of 373

Razor Financial Principals

β ki = coefficient relating secondary equity k to base equity i.

βk = coefficient specific to secondary index k.
W (t ) =
k
i
Brownian motion for secondary equity k linked base equity i
at time t.

The stochastic differential equation (SDE) for secondary equity k linked to

base equity i is:
dS ki ( t )  dS ( t ) 
= µ ki dt + β ki  i − µi dt  + β k dWki ( t )
Sk ( t )
i
 Si ( t ) 
 n 
= µ ki dt + β ki  ∑ βij dW j ( t )  + β k dWki ( t )
 j =1 

Solve the SDE to obtain:

β ki
 S (t )     1  
S (t ) = S (0) i  exp  µ ki − σ ki 2 dt − β ki  µ i − σ i2 dt + β k dWki (t )
i i 1
 S i (0) 
k k
 2   2  

where
σ ki 2 = β ki 2σ i2 + β k2 .

2.6.3 Simulation Algorithm

Base Equity

Step 1: To obtain µi and σ i

Obtain historical prices say 252 days for base index i and calculate the
historical daily log returns:
∆ i (1) = log Si ( −250 ) − log Si ( −251)
∆ i ( 2 ) = log Si ( −249 ) − log Si ( −250 )
M
∆ i ( 251) = log Si ( 0 ) − log Si ( −1)

Then we calculate
M N
∆i = ∑ ∆i ( j )
N − 1 j =1
M N
∑ ( ∆i ( j ) − ∆i )
2
σ i2 =
N − 1 j =1

where N is the number of log returns (251 number of returns in this

case) and M is the day convention per year (e.g. 365 if convention is 365
days per year).

©2009 Razor Risk Technologies Page 29 of 373

Razor Financial Principals

σ i2
µi = ∆ i +
2
σi = σ i
2

Step 2: To obtain µi and σ i for all base indices

Follow step 1 to calculate µi and σ i for each of the base index i = 1, 2,..., n
.

Step 3: To obtain covariance matrix for base indices.

Use the historical prices for each of the index i to generate the
covariance matrix V for all base indices.

Step 4: To obtain β ij
β ij is defined as the element of matrix β and β is defined as the
decomposition of covariance matrix V , i.e. V = β T β . We can obtain β
from V using principal decomposition algorithm.

Step 5: To generate random vectors from standard normal distribution.

For each base equity i, we need to generate φtil from N(0,1) for each of
the future simulating time point tl . Thus if we want to simulate stock
path at the future time point t1 , t2 ,..., tn , we need to generate a random
( )
vector φ i = φti1 , φti2 ,..., φtin . We do the same for each of the base equity
i = 1, 2,..., n .

Step 6: To simulate the base equity price path

We can simulate the future base equity price path using the formula:
 1  n 
S i (t m+1 ) = S i (t m ) exp  µ i − σ i2 ∆t + ∑ β ijφ tmj ∆t  .
 2  j =1 

Secondary Equity
Step 1: To obtain µ ki , β ki and β k
We obtain µ ki , β ki and β k using linear regression. Refer to the next
section for the details of using linear regression to obtain the secondary
equity parameters.

Step 2: To generate random vector from standard normal distribution

For secondary equity i, we need to generate ζ til from N(0,1) for each of
the future simulating time point tl . Thus if we want to simulate stock
path at the future time point t1 , t2 ,..., tn , we need to generate a random
( )
vector ζ i = ζ ti1 , ζ ti2 ,..., ζ tin .

Step 3: To simulate the secondary equity price path

©2009 Razor Risk Technologies Page 30 of 373
Razor Financial Principals

We can simulate the future secondary equity price path using the
formula:
β ki
 S (t ) 
S ( t ) = S ( tm )  i m +1  ×
i i

 Si ( tm ) 
k k

 1 1   1  
exp  µki − β ki 2σ ki 2 − β k2  dt − β ki  µi − σ i2  dt + β k ζ tikm ∆t 
 2 2   2  

Si ( tm +1 )
Note that can be obtained from the already simulated base
Si ( t m )
equity prices.

2.6.4 Parameter Estimation for Secondary Equities

Model Explanation

The equation governing the price of the secondary equity index is:

∆S ki (t m )  i 
= µ ki ∆t + β ki ∑ β ijφ tmj ∆t  + β k γ tkm ∆t .
S k (t m )
i
 j =1 

The left-hand-side component

S ki (t m ) is the price of the k-th secondary equity index linked to the i-th
base equity at time t m . Historical prices of the secondary equity index are
used in linear regression, thus S ki (t m ) is known and observable in linear
regression.

∆S ki (t m ) is defined as S ki (t m +1 ) − S ki (t m ) . It is the change of secondary equity

index price in one day (if we use daily data). Since all the historical prices
of the secondary equity index are observable, the changes in the prices of
the secondary equity index are also observable.

Thus the left-hand-side of the equation is completely observable and

∆S i (t )
known. The left-hand-side i k m is actually the historical daily return
S k (tm )
of the secondary index. We denote the left-hand-side constant as Y .

For example, if we have historical secondary index prices for six days,
then we will have five historical daily returns and we will have five values
for Y .

The µ ki ∆t component

©2009 Razor Risk Technologies Page 31 of 373

Razor Financial Principals

∆t is the change in time. If the historical data used in linear regression

1
are daily data, then ∆t = (if the day convention is 365 days per year).
365
Since we know the frequency of the historical data, thus ∆t is known too.

µ ki is the deterministic drift for secondary index k and it is an unknown

constant we want to find using linear regression.

Thus only µ ki is unknown in the µ ki ∆t component.

The µ ki ∆t consists of two known constants and one unknown constant,

thus µ ki ∆t is still a constant. We denote µ ki ∆t as α 0 .

 i 
The β ki ∑ β ij φ tmj ∆t  component
 j =1 

β ki is the correlation coefficient relating secondary index k to diffusions

in the i-th base index. It is an unknown constant we want to estimate
using linear regression and we denote it α 1 .

β ij is the principal decomposition of the covariance matrix and it was

determined when estimating the base index. Thus at the time when we
want to estimate the secondary equity index, β ij is a known value
already.

∆t is a known value as we mentioned before.

φt j is a standard normal variable N (0,1) . However, it has already been

m

generated when we estimated the base index. Thus φtmj is already known
when we try to estimate the secondary equity index.

 i 
The term ∑ β ijφ tmj ∆t  is a constant and observable from the base index.
 j =1 
 i 
We denote ∑ β ij φtmj ∆t  as X 1 . Note that X 1 is a constant because it is
 j =1 
observable from the process when we estimated the base index. When
doing linear regression, if we have five historical returns, i.e. five Y ,
then we need to get the corresponding five values of X 1 from the base
index.
X 1 can be recovered from the historical prices of the base index by the
following formula:

©2009 Razor Risk Technologies Page 32 of 373

Razor Financial Principals

 S (t )   1 
X 1 = ln i m +1  −  µ i − σ i2 ∆t .
 S i (t m )   2 
i
Note the summation ∑β φ
j =1
j
ij t m ∆t over j is based on the number of base

indexes. If there are 10 base indexes, we need to sum the j from 1 to 10.
10
Then the term becomes ∑β φ
j
j
ij t m ∆t . However, we do not need to

calculate the summation explicitly for regression purposes because they

are already included in the X 1 term. We do need to calculate the
summation term later for simulation.

The β k γ tkm ∆t component

β k is a coefficient unique to secondary index k and it is an unknown
constant we want to determine using linear regression. We denote it as
α2 .
∆t is a known constant as we mentioned previously.

γ tk is standard normal variable N (0,1) . It is random and different every

m

time we generate it. Thus it is not a constant.

In the term γ tkm ∆t , only γ tkm is not a constant, we denote γ tkm ∆t as X 2 .

Note that X 2 is not a constant because γ tkm is not a constant. When doing
linear regression, if we have five historical returns, i.e. five Y , then we
need to generate five γ tkm values from the standard normal distribution
randomly and we will obtain five X 2 .

To summarise, the equation for the secondary equity index can be re-
written as:

Y = α 0 + α1 X 1 + α 2 X 2 .
We can determine α 0 , α 1 and α 2 using linear regression.
Note that α 0 = µ ki ∆t and ∆t is a known constant. Thus once we obtain α 0
α0
from regression, we can calculate µ ki as . Also α 1 = β ki and α 2 = β k ,
∆t
thus if we know a1 and α 2 , we automatically know β ki and β k .

Linear Regression
To carry out linear regression, we write the linear regression equation as:

Y = α 0 + α 1 X 1 + α 2 X 2 + ε , where ε ~ N (0,1) .

Our aim is to minimise the sum of the square term ε .

©2009 Razor Risk Technologies Page 33 of 373

Razor Financial Principals

Example of linear regression

We have the equation for the secondary equity index:

∆S ki (t m )  i 
= µ ∆ + β ∑ β ijφ tm ∆t  + β k γ tm ∆t .
i i j k
t
S k (t m )
i k k
 j =1 

We assume in this example that all the known constants are equal to 1
and we have yearly historical secondary index data, thus both β ij and ∆t
are all equal to 1. We also assume that there is only one base index.

Thus the equation becomes:

∆S ki (t m )
= µ ki + β ki φtmj + β k γ tkm .
S k (t m )
i

Assume the annual historical prices for secondary equity index in the past
six years are:

Year Price
2001 10
2002 20
2003 30
2004 40
2005 50
2006 60

∆S ki (t m )
Then the vector of Y = becomes:
S ki (t m )
 20 − 10 
 10 
 
 30 − 20  1 
 20   
  0.5 
40 − 30
Y =  = 0.33 .
 30   
 50 − 40  0.25
  0.2 
 40   
 60 − 50 
 
 50 

α 0 = µ ki is constant.

α 1 = β ki is a constant and X 1 = φtmj . There are five values for Y , thus we

need to have five values for X 1 . The five values of X 1 should have
already been generated in the base index thus they are observable and

©2009 Razor Risk Technologies Page 34 of 373

Razor Financial Principals

can be recovered by the formula we mentioned previously. For example,

if the base index price for 2001 and 2002 are 30 and 40 respectively and
µi and σ i from base index estimation are 0.01 and 0.05, then the first
 40   1 
element of X1 is ln  − 1 ⋅ 0.01 − ⋅1 ⋅ 0.05 2  ⋅1 = 0.28 . If the five values of
 30   2 
X 1 in the base index are 0.28, 0.16, 0.81, 0.53 and 0.65 then the vector
of X 1 becomes:

0.28
0.16
 
X 1 = 0.81 .
 
0.53
0.65

α 2 = β k is a constant and X 2 = γ tk , γ tk ~ N (0,1) . Again, we need to

m m

generate five values for X 2 . We generate five values randomly from the
standard normal distribution. If the five random numbers we generated
are 0.1,0.9,0.35,0.65 and 0.95 then the vector of X 2 becomes:

0.1 
0.9 
 
X 2 = 0.35 .
 
0.65
0.95

The regression equation is thus:

1  1 0.28  0.1 

0.5  1 0.16  0.9 
      
0.33 = α 0 1 + α1 0.81  + α 2 0.35 + ε
      
0.25 1 0.53  0.65
0.2  1 0.65  0.95

1  1 0.3  0.1 

0.5  1 0.1  0.9 
       
ε = 0.33 − α 0 1 + α1 0.8  + α 2 0.35
       
0.25 1 0.5  0.65 
0.2  1 0.6  0.95 

©2009 Razor Risk Technologies Page 35 of 373

Razor Financial Principals

To find α 0 , α 1 and α 2 that minimise ε 2 = ε T ε , we can use the linear

least square algorithm.

©2009 Razor Risk Technologies Page 36 of 373

Chapter 3
Credit Exposure Calculation

3.1 Pre-Settlement Risk

Unlike settlement risk, which entails exposure equal to a counterparty’s gross
obligation, pre-settlement risk entails exposure equal to a counterparty’s net
obligation on that contract. For example, suppose an institution enters into a
forward contract to exchange $1M Australian dollars for $0.6M US dollars in
three months. Settlement risk exposes the institution to a possible loss of $1M
AUD. Pre-settlement risk exposes the institution to just the difference in
market value between the US and Australian dollar payments.

3.1.1 Calculation of Credit Exposure

The credit exposure is calculated by valuing each trade according to the
current market rates in the simulation. Each trade’s exposure is then
integrated into the portfolios that that trade impacts.

3.1.2 Treatment of Credit Nodes

Credit nodes are the points in time on which exposures are calculated. On each
credit node, the market rates are rolled forward to that particular day, and
the various yield curves, exchange rates and volatility surfaces are generated.

3.1.3 Economic Offsetting

Individual trades in a portfolio can be quite risky in isolation, but their addition
in a portfolio can serve to reduce the aggregate risk if the correlation of losses
between these credits and the rest of the portfolio is less than fully correlated.

3.1.4 Netting Agreements

The netting of cash flows or obligations is a means of reducing the credit
exposure to a counterparty:
Payment netting reduces settlement risk and makes the payment processing
more efficient: If counterparties are to exchange multiple cash flows during a
given day, they can agree to net those cash flows to one payment per
currency.
Closeout netting reduces pre-settlement risk: If counterparties have multiple
offsetting obligations to one another — for example, multiple interest rate
swaps or foreign exchange forward contracts — they can agree to net those
obligations. In the event that a counterparty defaults, or some other
termination event occurs, the outstanding contracts are all terminated. They
are marked to market and settled with a net payment. This technique
eliminates “cherry picking” whereby a defaulting counterparty fails to make
payment on its obligations, but is legally entitled to collect on the obligations
owed to it.
With a bilateral netting agreement, two counterparties agree to net with one
another. They sign a master agreement specifying the types of netting to be
performed as well as the existing and future contracts, which will be affected.
Razor Financial Principals

3.1.5 Collateral Agreements

A counterparty will enter into a collateral agreement in order to enhance their
credit quality. A collateral agreement obligates the counterparty to margins in
cash with the dealer to cover costs in the event of default.

3.2 Credit Exposure Measures

3.2.1 Introduction
In July 2005, the Basel Committee on Banking Supervision (BCBS), as
established by the Bank of International Settlements (BIS) published a
major amendment to the Basel 2 framework in its application to a bank’s
Trading Book. This amendment describes changes relating to Credit
Exposure of the Trading Book, treatment of Guarantees, and changes to
the regulatory specific risk calculation.

3.2.2 Key Measures

Within the Basel regulatory framework there are various models for the
calculation of exposure to Counterparty Credit Risk (CCR). This Exposure
at Default (EAD) can be measured using:

1. Current Exposure Method (CEM) – MtM+Addon

2. Standardised Method (SM) –Under SM, the exposure amount represents the
product of:
(i) the larger of the net current market value or a “supervisory EPE”,
times
(ii) a scaling factor, termed beta.
3. Internal Model Method (IMM) – Potential Future Exposure (PFE), Expected
Exposure (EE), Expected Positive Exposure (EPE) – all derived from the
distribution of counterparty exposures as generated from the stochastic
simulation process as described elsewhere in the Financial Principles.

In support for the Internal Model the following additional measures are
provided in Razor to support the aforementioned Trading Book
Amendments:

1. Expected Exposure (EE)

EE is the probability-weighted average exposure estimated to exist on
a future date.
2. Expected Positive Exposure (EPE)
EPE is the time-weighted average of individual expected exposures
estimated for a given forecasting horizon:
min (1 year , maturity )
EPE = ∑
k =1
EEtk × ∆tk .

3. Effective Expected Exposure (EEE)

©2009 Razor Risk Technologies Page 38 of 373

Razor Financial Principals

EEE is defined recursively as EEEtk = max EEEtk −1 , EEtk ( ) where exposure

is measured at future dates t1 , t2 , t3 ,... , and EEEt0 equals current
exposure.
4. Effective Expected Positive Exposure (EEPE)
EEPE is the average EEE during during the first year of future
exposure. If all contracts in the netting set mature before one year,
EPE is the average of EE until all contracts in he netting set mature.
EEPE is computed as a weighted average of EEE:
min (1 year , maturity )
EEPE = ∑
k =1
EEEtk × ∆tk .

The effective EPE estimate corresponds with the exposure that can be
calculated in the current economic climate. However in order to capture
a deteriorating economic climate the final number must be scaled by a
regulatory defined factor called “Alpha”, this value is currently set at
1.4.

The following diagram illustrates different credit exposure measures:

Credit Exposure Measures
1

0.9

0.8 MPE
PFE
0.7
EE
EEE
Dollar in Millions

0.6
EEPE
0.5 EPE

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time

Note that MPE is the maximum peak exposure and PFE is the maximum
exposure estimated to occur on a future date at a high level of statistical
confidence.

Reference

©2009 Razor Risk Technologies Page 39 of 373

Razor Financial Principals

Basel Committee on Banking Supervision (July 2005), ‘The Application of

Basel 2 to Trading Activities and the Treatment of Double Default
Effects’, Bank for International Settlements.

3.3 Settlement Risk

Settlement Risk is the exposure a bank faces to the loss of the full principal
value of a deal where the counterparty defaults in the payment of their part of
an exchange.
Settlement Risk is primarily incurred by foreign exchange transactions and
cross-currency swaps, where the bank pays away their side of the deal and
does not receive the counterparty’s matching payment.
In order to recognise the settlement risk inherent in a portfolio of deals, the
system must perform the following steps:
Identify the transactions/cashflows within the portfolio that generate
settlement risk
Identify any Settlement Netting agreements that cover transactions within the
portfolio, condense the netting agreements to a reduced set of net
cashflows.
For each exchange, determine the duration of the settlement exposure
Combine the transactions to form a settlement exposure profile
For each date, check the peak exposure from the profile to determine the
utilisation
Check the utilisation against limits to determine breaches.

3.3.1 Transaction/Cashflow Identification

Only those transactions that generate settlement risk against a portfolio should
be processed.
All cashflows that represent a matched exchange generate settlement risk, eg
FX, Cross Currency Swaps and Bond transactions.
Some transactions, e.g. swaps, can generate multiple settlement risks on
multiple dates.
An ‘exchange’ occurs, where both parties make payments to each other, where
these payments are intended, but are not guaranteed, to be simultaneous.

3.3.2 Exposure Duration

It is the nature of settlement risk, that exposure is derived from the mismatch
between the point at which the bank is irrevocably committed to making a
payment, and the point at which the bank can recognise that the
counterparty’s payment has been received.
Deals/Cashflows pass through the following states as part of the settlement
process:
Revocable - The bank’s payment instruction for the sold currency either has not
been issued or may be unilaterally cancelled without the consent of the
bank’s counterparty or any other intermediary. The bank faces no current

©2009 Razor Risk Technologies Page 40 of 373

Razor Financial Principals

settlement exposure for this trade. During this period the trade represents
a pre-settlement exposure.
Irrevocable - The bank’s payment instruction for the sold currency can no longer
be cancelled unilaterally either because it has been finally processed by
the relevant payments system or because some other factor (e.g. internal
procedures, correspondent banking arrangements, local payments system
rules, laws) makes cancellation dependent upon the consent of the
counterparty or another intermediary; the final receipt of the bought
currency is not yet due. In this case, the bought amount is clearly at risk.
Uncertain - The bank’s payment instruction for the sold currency can no longer
be cancelled unilaterally; receipt of the bought currency is due, but the
bank does not yet know whether it has received these funds with finality.
In normal circumstances, the bank expects to have received the funds on
time. However, since it is possible that the bought currency was not
received when due (e.g. owing to an error or to a technical or financial
failure of the counterparty or some other intermediary), the bought
amount might, in fact, still be at risk.
Failed - The bank has established that it did not receive the bought currency
from its counterparty. In this case the bought amount is overdue and
remains clearly at risk. Failed trades can be converted to a cash exposure
and should appear effectively as a Loan in the Pre-Settlement exposure to
the Counterparty.
Settled - The bank knows that it has received the bought currency with finality.
From a settlement risk perspective the trade is considered settled and the
bought amount is no longer at risk.
The key states for settlement risk measurement purposes are Irrevocable and
Uncertain.

3.3.3 Settlement Exposure Calculation

Settlement Date Exposure
The system calculates the settlement risk by determining the cashflows that
fall due on each forward date.
In the event that the exposures are being calculated on a gross basis, or if the
trades are not covered by a netting agreement, the settlement exposure is
calculated as the sum of the receipts that fall due on that date.
For example the following trades are being processed and all settle on the
same date:

Trade REF Direction Currency AUD

Pay NZD 85
T1
Recv AUD 95
Pay USD 100
T2
Recv NZD 90
Pay AUD 50
T3
Recv USD 45

These trades generate the following exposures:

©2009 Razor Risk Technologies Page 41 of 373

Razor Financial Principals

Trade REF Currency AUD

T1 AUD 95
T2 NZD 90
T3 USD 45

These trades will arise at a gross settlement exposure of 230.

3.3.4 Settlement Netting

Where netting is involved, the system must divide the cashflows into netting
pools organised by agreement (multiple netting agreements may be in
operation, for example and country settlement limit) and by currency.
Within each pool the payments and receipts are offset together to arise at net
payment or receipt for a counterparty/currency combination. The system sums
all net receipts to arise at the settlement exposure for the date in question.
For example if the trades encountered in Table 1 above are covered by a
netting agreement, the processing will be as follows:
The cashflows are divided into currency pools:

Currency Cashflows Position

AUD Pay 50 Recv 95 Recv 45
USD Pay 100 Recv 45 Pay 55
NZD Pay 85 Recv 90 Recv 5

The net settlement date exposure for these trades will be 50 (AUD 45 + NZD 5).
The system supports the processing of settlement exposure with and without
settlement netting enabled for different portfolio types.

3.3.5 Spread Risk

The concept of spread risk means that the system is enabled to take into
account the Irrevocable and Uncertain periods defined in the section “Exposure
Duration” above.
The system supports the processing of settlement exposures with or without
spread risk for different portfolio types.
Currently the spread risk is specified by currency, in terms of Receipt
Identification; the amount of time needed to identify that the receipt of the
counterparty’s payment has indeed occurred, and Cancellation Deadline; the
amount of time before settlement occurs needed to cancel the payment to the
counterparty.

©2009 Razor Risk Technologies Page 42 of 373

Razor Financial Principals

Chapter 4
Credit Risk Measurement
Credit risk arises from the changes in credit quality of a counterparty to a
trade. It is made up of default risk and downgrade credit spread risk. Default
risk is measuring the risk of failure of a counterparty meeting its financial
obligations, and what level of recovery is likely in the event of default.
Downgrade risk is the risk that a counterparty's credit rating falls and therefore
the value of the positions with that counterparty may also fall.
The traditional approaches to the management of credit risk by setting limits
to measures such as counterparty credit exposure are described in Chapter 3.
Additional to the measurement of counterparty credit exposure, which
measure current and potential future exposure assuming default, it is useful to
have more sophisticated measures of credit risk which adequately model credit
losses.
A credit risk model calculates credit losses by deriving at each credit or time
node a distribution of potential credit losses. These models not only consider
the changes in the market risk factors, but also credit rating migration and
rates of recovery if default occurs for a counterparty or an issuer.
Factors which more commonly determine portfolio credit risk include,
At a trade level include but are not limited to such factors as:
Current and future potential exposure as defined in 3.1.1, and Recovery rates,
which are a measure of the amount which may still be recovered from a
defaulting counterparty.
2. At a counterparty level the factors are:
Credit migration probabilities, that is the probability of a counterparty credit
rating transitioning to a lower credit rating and the subsequent change in
credit exposure to that counterparty,
Default probabilities, that is the probability a counterparty will default, and
the subsequent loss incurred by the counterparty failing to meet the
obligations associated with the trade.
The default probability may be embedded in the credit migration probabilities
by specifying a probability to transition to the default rating. Alternatively
default risk can be separated from credit risk by only specifying probabilities to
maintain the current rating or transition to default.
Recovery rates, which are a measure of the amount which may still be
recovered from the defaulting counterparty,
At the counterparty level the aggregation of exposures will take into
consideration offsetting, netting agreements and collateral as described in
3.1.3, 3.1.4, 3.1.5,
3. At a portfolio level such factors as joint credit migration probability
distribution, which recognize the level of co-movement between credit events
occurring for different counterparties.

©2009 Razor Risk Technologies Page 43 of 373

Razor Financial Principals

4.1 Modelling Credit Transition Losses

In addition to simulating the market rates changes the credit rating of
counterparties are also allowed to vary stochastically, independent of the
market simulation.
The Credit Rating Transition process evolves according to the probability
distribution specified via a Credit Rating Transition Matrix (CRTM). The CRTM
specifies probabilities of transitioning from an initial credit rating to other
ratings defined in the rating scheme. This distribution can vary over time by
defining multiple matrices for differing tenors.
The Credit Spread Risk can be modelled by letting the pricing model be
dependent on the simulated credit rating.
Similarly by analysing distribution of the exposure for default in the events of a
transition to the default rating the Default Risk can be determined.

4.1.1 Credit Rating Transition Matrices

The transition matrices may include any type of rating (i.e. provided by
external ratings models used by S&P, Moodys, Fitch, or internal ratings models)
and should support any level of rating granularity desired by the user. The
only constraint is that each credit rating in the matrix must be set up as a
separately simulated pricing curve, and for a given matrix the same ratings
must be used across all transition tenors.
The credit rating transition probabilities are derived from the analysis of
historical data and subsequent derivation into an external and/or internal
ratings model.
The migrations are specified as the % chance for an issuer of a given rating to
migrate to each of the other potential ratings within the matrix. Hence, this
matrix specifies the probabilities pij of transition from rating state i to rating
state j (including default) over a fixed time horizon τ .
Example transition matrix (annual):
Transitioned Rating
AAA AA A BBB BB B CCC D
92.97
AAA % 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00%
91.81
AA 2.60% % 4.78% 0.60% 0.06% 0.12% 0.03% 0.00%
91.68
A 0.70% 2.27% % 4.49% 0.56% 0.25% 0.01% 0.04%
Initial Rating

87.87
BBB 0.07% 0.20% 5.60% % 4.83% 1.02% 0.17% 0.24%
81.48
BB 0.00% 0.10% 0.61% 7.79% % 7.90% 1.11% 1.01%
82.80
B 0.00% 0.10% 0.28% 0.46% 6.95% % 3.96% 5.45%
12.32 60.44 23.69
CCC 0.00% 0.00% 0.37% 0.75% 2.43% % % %

©2009 Razor Risk Technologies Page 44 of 373

Razor Financial Principals

Therefore the % chance that a AAA rated issuer will stay AAA in a year’s time is
92.97%, while the chance that it will migrate to AA within a year is 6.43%.
This matrix can be simplified by just specifying the probabilities of default
from each credit rating which simulates default events but ignores
transitioning between other rating values.

Transition Matrix Tenor

During simulation, the tenor between credit nodes will not match the
frequency of the rating transition matrix; i.e. if an annual credit transition
matrix is used but the credit time nodes are weekly, using the annual matrix
would drastically overstate the transitioning.
Any given transition matrix may be specified at a number of different tenors.
For example for a US Industrials transition matrix, one matrix may be specified
with tenor 1W, another for 1M, another for 6M etc. During simulation of credit
transitions, the time between the current time node and the next simulated
node will determine the tenor used. When the time between credit nodes falls
between two transition matrices, the transition percentages will be
interpolated between the two matrices to determine the appropriate transition
values.
Linear interpolation is performed on each CRTM element. Hence, to
interpolate between CRTM at tenors t1 and t2 to tenor t, each element in the
interpolated matrix is given by:
t − t1
r= (r2 − r1 ) + r1
t 2 − t1
Where:
r1 element in CRTM1
t1 tenor of CRTM1
r2 corresponding element in CRTM2
t2 tenor of CRTM2
r corresponding element in interpolated CRTM
t tenor of interpolated CRTM

If only a single CRTM is defined then the migration process is assumed to be

time independent and no interpolation is performed.
The Process of Transition Probability Interpolation
The process of interpolation in some more detail, can be described in the
context of simulation of credit rating migration for each credit party. Each
credit rating migration probability matrix of a specific term represents the
probability of migrating from the start credit rating to the end credit rating
over that term, so if there are two matrices one for 1 week, and one for 1
month, these represent the probability of migrating over a 1 week and a 1
month period with respect to the current point – irrespective if this current
valuation time node is today or at some future credit node point.
In the transition/migration to the current credit node from the previous credit
node the term for this period is calculated, and then the probabilities for this

©2009 Razor Risk Technologies Page 45 of 373

Razor Financial Principals

calculated term is linearly interpolated from the set of provided term based
probabilities.
1. If there is only one provided matrix then the probabilities are constant with
respect to term.
2. Similarly if the calculated term required is shorter in duration compared
with the first sorted matrix, or longer in duration compared with the last
sorted matrix, then again the probabilities are flat with respect to term, and
the closest (respectively the first and last) matrix will be used.
3. If the duration of the term being transitioned from one credit node to the
next lies between two termed based credit rating migration probability
matrices then it is a straight line time based linear interpolation to derive the
probabilities to use.
So for example if a 1 month and 3 month credit rating migration probability
matrix are provided, and all the credit nodes are 2 months apart then the
probabilities used for each migration of 2 months duration in the simulation
will be those same linearly interpolated (half way) between the 1 month and 3
month matrices.
As another example if using the same matrices as above and just 3 credit nodes
of 1 month apart, then the probabilities for each credit rating migration of 1
month duration in the future will be those from the 1 month credit rating
migration probability matrix.
As another example if using the same matrices as above and just 2 credit
nodes, at 1 month and 4 months, then the transition to the 1 month credit
node will use the 1 month probabilities as supplied, and the transitioning from
1 month node to 4 month credit node will use the 3 month probabilities as
supplied.
Naturally as transitioning along a path the credit party credit rating is
migrating according to the probabilities and the simulation going from one
credit node to the next along the path.

Entity – Transition Rating Mapping

Adding a separate mapping to indicate the rating used to drive the migration
enables gives greater control over the migration; they don’t necessarily need
to use the actual S&P etc rating of a given issuer.

Entity – Transition Matrix Mapping

The ability to map an entity to a particular transition matrix allows the
simulation to capture different transition volatilities across industry, country,
or other attributes. For example it is possible to use one transition matrix for
US software companies, and another for US Utilities. Adding a transition
matrix field to the entity definition allows for this mapping and provides
complete flexibility to map any issuer to any matrix.

©2009 Razor Risk Technologies Page 46 of 373

Razor Financial Principals

Entities Transition Matrices (by tenor)

Daimler Chrysler

Rating AAA AA A BBB BB B CCC D Rating AAA AA A BBB BB B CCC D Rating AAA AA A BBB BB B CCC D
AAA 92.97% 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00% AAA 92.97% 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00% Rating AAA AA A BBB BB B CCC D AAA 92.97% 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00%
AA AA AAA 92.97% 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00% AA
2.60% 91.81% 4.78% 0.60% 0.06% 0.12% 0.03% 0.00% 2.60% 91.81% 4.78% 0.60% 0.06% 0.12% 0.03% 0.00% 2.60% 91.81% 4.78% 0.60% 0.06% 0.12% 0.03% 0.00%
AA 2.60% 91.81% 4.78% 0.60% 0.06% 0.12% 0.03% 0.00%
A 0.70% 2.27% 91.68% 4.49% 0.56% 0.25% 0.01% 0.04% A 0.70% 2.27% 91.68% 4.49% 0.56% 0.25% 0.01% 0.04% A 0.70% 2.27% 91.68% 4.49% 0.56% 0.25% 0.01% 0.04%
BBB 0.07% 0.20% 5.60% 87.87% 4.83% 1.02% 0.17% 0.24% BBB 0.07% 0.20% 5.60% 87.87% 4.83% 1.02% 0.17% 0.24% A 0.70% 2.27% 91.68% 4.49% 0.56% 0.25% 0.01% 0.04% BBB 0.07% 0.20% 5.60% 87.87% 4.83% 1.02% 0.17% 0.24%
BB 0.00% 0.10% 0.61% 7.79% 81.48% 7.90% 1.11% 1.01% BB 0.00% 0.10% 0.61% 7.79% 81.48% 7.90% 1.11% 1.01% BBB 0.07% 0.20% 5.60% 87.87% 4.83% 1.02% 0.17% 0.24% BB 0.00% 0.10% 0.61% 7.79% 81.48% 7.90% 1.11% 1.01%

Ford B
CCC
0.00%
0.00%
0.10%
0.00%
0.28%
0.37%
0.46%
0.75%
6.95%
2.43%
82.80%
12.32%
3.96%
60.44%
5.45%
23.69%
B
CCC
0.00%
0.00%
0.10%
0.00%
0.28%
0.37%
0.46%
0.75%
6.95%
2.43%
82.80%
12.32%
3.96%
60.44%
5.45%
23.69%
BB
B
CCC
0.00%
0.00%
0.00%
0.10%
0.10%
0.00%
0.61%
0.28%
0.37%
7.79%
0.46%
0.75%
81.48%
6.95%
2.43%
7.90%
82.80%
12.32%
1.11%
3.96%
60.44%
1.01%
5.45%
23.69%
B
CCC
0.00%
0.00%
0.10%
0.00%
0.28%
0.37%
0.46%
0.75%
6.95%
2.43%
82.80%
12.32%
3.96%
60.44%
5.45%
23.69%

Toyota
Rating AAA AA A BBB BB B CCC D Rating AAA AA A BBB BB B CCC D Rating AAA AA A BBB BB B CCC D
AAA 92.97% 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00% AAA 92.97% 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00% Rating AAA AA A BBB BB B CCC D AAA 92.97% 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00%
AA AA AAA 92.97% 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00% AA
2.60% 91.81% 4.78% 0.60% 0.06% 0.12% 0.03% 0.00% 2.60% 91.81% 4.78% 0.60% 0.06% 0.12% 0.03% 0.00% 2.60% 91.81% 4.78% 0.60% 0.06% 0.12% 0.03% 0.00%
AA 2.60% 91.81% 4.78% 0.60% 0.06% 0.12% 0.03% 0.00%
A 0.70% 2.27% 91.68% 4.49% 0.56% 0.25% 0.01% 0.04% A 0.70% 2.27% 91.68% 4.49% 0.56% 0.25% 0.01% 0.04% A 0.70% 2.27% 91.68% 4.49% 0.56% 0.25% 0.01% 0.04%
BBB 0.07% 0.20% 5.60% 87.87% 4.83% 1.02% 0.17% 0.24% BBB 0.07% 0.20% 5.60% 87.87% 4.83% 1.02% 0.17% 0.24% A 0.70% 2.27% 91.68% 4.49% 0.56% 0.25% 0.01% 0.04% BBB 0.07% 0.20% 5.60% 87.87% 4.83% 1.02% 0.17% 0.24%
BB 0.00% 0.10% 0.61% 7.79% 81.48% 7.90% 1.11% 1.01% BB 0.00% 0.10% 0.61% 7.79% 81.48% 7.90% 1.11% 1.01% BBB 0.07% 0.20% 5.60% 87.87% 4.83% 1.02% 0.17% 0.24% BB 0.00% 0.10% 0.61% 7.79% 81.48% 7.90% 1.11% 1.01%
B B BB 0.00% 0.10% 0.61% 7.79% 81.48% 7.90% 1.11% 1.01% B
0.00% 0.10% 0.28% 0.46% 6.95% 82.80% 3.96% 5.45% 0.00% 0.10% 0.28% 0.46% 6.95% 82.80% 3.96% 5.45% 0.00% 0.10% 0.28% 0.46% 6.95% 82.80% 3.96% 5.45%
B 0.00% 0.10% 0.28% 0.46% 6.95% 82.80% 3.96% 5.45%
CCC 0.00% 0.00% 0.37% 0.75% 2.43% 12.32% 60.44% 23.69% CCC 0.00% 0.00% 0.37% 0.75% 2.43% 12.32% 60.44% 23.69% CCC 0.00% 0.00% 0.37% 0.75% 2.43% 12.32% 60.44% 23.69%
CCC 0.00% 0.00% 0.37% 0.75% 2.43% 12.32% 60.44% 23.69%

British Petroleum

JP Morgan
Rating AAA AA A BBB BB B CCC D Rating AAA AA A BBB BB B CCC D Rating AAA AA A BBB BB B CCC D
AAA 92.97% 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00% AAA 92.97% 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00% Rating AAA AA A BBB BB B CCC D AAA 92.97% 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00%
AA AA AAA 92.97% 6.43% 0.48% 0.08% 0.04% 0.00% 0.00% 0.00% AA
2.60% 91.81% 4.78% 0.60% 0.06% 0.12% 0.03% 0.00% 2.60% 91.81% 4.78% 0.60% 0.06% 0.12% 0.03% 0.00% 2.60% 91.81% 4.78% 0.60% 0.06% 0.12% 0.03% 0.00%
AA 2.60% 91.81% 4.78% 0.60% 0.06% 0.12% 0.03% 0.00%
A 0.70% 2.27% 91.68% 4.49% 0.56% 0.25% 0.01% 0.04% A 0.70% 2.27% 91.68% 4.49% 0.56% 0.25% 0.01% 0.04% A 0.70% 2.27% 91.68% 4.49% 0.56% 0.25% 0.01% 0.04%
BBB 0.07% 0.20% 5.60% 87.87% 4.83% 1.02% 0.17% 0.24% BBB 0.07% 0.20% 5.60% 87.87% 4.83% 1.02% 0.17% 0.24% A 0.70% 2.27% 91.68% 4.49% 0.56% 0.25% 0.01% 0.04% BBB 0.07% 0.20% 5.60% 87.87% 4.83% 1.02% 0.17% 0.24%
BB 0.00% 0.10% 0.61% 7.79% 81.48% 7.90% 1.11% 1.01% BB 0.00% 0.10% 0.61% 7.79% 81.48% 7.90% 1.11% 1.01% BBB 0.07% 0.20% 5.60% 87.87% 4.83% 1.02% 0.17% 0.24% BB 0.00% 0.10% 0.61% 7.79% 81.48% 7.90% 1.11% 1.01%
B B BB 0.00% 0.10% 0.61% 7.79% 81.48% 7.90% 1.11% 1.01% B
0.00% 0.10% 0.28% 0.46% 6.95% 82.80% 3.96% 5.45% 0.00% 0.10% 0.28% 0.46% 6.95% 82.80% 3.96% 5.45% 0.00% 0.10% 0.28% 0.46% 6.95% 82.80% 3.96% 5.45%
B 0.00% 0.10% 0.28% 0.46% 6.95% 82.80% 3.96% 5.45%
CCC 0.00% 0.00% 0.37% 0.75% 2.43% 12.32% 60.44% 23.69% CCC 0.00% 0.00% 0.37% 0.75% 2.43% 12.32% 60.44% 23.69% CCC 0.00% 0.00% 0.37% 0.75% 2.43% 12.32% 60.44% 23.69%
CCC 0.00% 0.00% 0.37% 0.75% 2.43% 12.32% 60.44% 23.69%

Barclays

Hence in summary, CRTM are defined by tenor and interpolated to the current
time node. Multiple CRTM structures can be defined relating to different
segments of the market.
Configuration determines to which CRTM structure a trade is assigned.
Typically this would be determined by any combination of trade attributes such
as industry, location, and so on.

4.1.2 Simulation of Credit Rating Changes

The CRTM defines the credit transition probabilities. This section describes
how they are used within the credit simulation. For each credit party within a
simulation scenario a correlated uniform random number is drawn. This
random number is then used to index into the CRTM to determine the
transitioned credit rating, using the row of the CRTM for the current credit
rating of that counterparty. For instance using the previous CRTM example with
a current rating of A the probabilities of transition are:
New Probability Cumulative
Rating Probability
AAA 0.70% 0.70%
AA 2.27% 2.97%
A 91.68% 94.65%
BBB 4.49% 99.14%
BB 0.56% 99.70%
B 0.25% 99.95%
CCC 0.01% 99.96%
D 0.04% 100.00%
Hence a simulation value of 0 - 0.7% (ie the correlated uniform random
number) would indicate a new rating of AAA, 0.7% - 2.97% would transition to
AA, and so on to 99..96% - 100% for a default event. This perturbed credit
rating is passed to the trade pricing routines for all trades with that
counterparty in the scenario. Note that as in this example it is typical that the

©2009 Razor Risk Technologies Page 47 of 373

Razor Financial Principals

most likely outcome is the transitioned credit rating will be the same as the
current rating.
This perturbed rating is used as the initial rating for the next time node in the
simulation. At the beginning of each path the credit rating is reset to its initial
condition. In this manor a credit rating can transition up or down during the
simulation. Transitioning to the default rating causes a default event to be
raised and no further transitioning can occur for the remainder of the path.

4.1.3 Credit Transition Co-Movement

The process described thus far fails to take into account co-movement of
rating transition between counterparties.
This is required to accommodate the observation that credit rating transitions
and defaults are not independent events, as confirmed from statistical analysis
of historical observations. Many simple examples can illustrate this point, for
example a severe drought is likely to have a similar financial impact on similar
counterparties from the agricultural industry.
In razor this co-dependence is captured by a correlation matrix of the
counterparties, to produce correlated uniform random numbers to index into
the CRTM.
As part of the simulation of credit migration, the correlation matrix would be
decomposed using the same eigenvalue algorithm as is currently utilized for
market drivers (including performing Principal Component Analysis on the
correlation matrix) and then multiplied against a uniform random number
distribution to obtain correlated random numbers. This is similar to the
current implementation for simulation of market rates, with the exception
being that when simulating credit migration it is not necessary to transform
the uniform distribution to a N(0,1) distribution using a cumulative distribution
function.

Result of Uniform
Matrix Random
Decomposition Numbers Correlated random numbers
z1
b11 0 0 b11 * z 1
b21 b22 0
z2 = b 21 * z 1 + b 22 * z 2
b31 b32 b33 z3 b 31 * z 1 + b 32 * z 2 + b 33 * z 3

This yields a set of correlated random numbers on the interval (0,1) for each
credit entity, and for each scenario in a path.

4.1.4 Generation of Credit Rating Changes

As described in section 2.5 for market simulation the credit rating simulation
can be pre-generated, separately from the market. For a Credit Rating
Simulation a uniform random method is required with the correlation matrix
defined for all counterparties being simulated. This will result in a memory
mapped file containing a random number for each counterparty and scenario to
determine the perturbed credit rating.
As for a market simulation this pre-generated file can be overridden to
simulate transitions during the pricing phase.

©2009 Razor Risk Technologies Page 48 of 373

Razor Financial Principals

Note: Within RAZOR it is possible to set up multiple credit loss models running
in parallel where each model is using data from a different external or internal
ratings model, and hence it is possible to compare measures such as expected
loss, and economic capital as derived from each model.
The Credit Rating simulation has been described it remains to detail the
calculation of Credit Losses.

4.1.5 Credit Spread Risk Loss

To measure the credit spread risk the credit rating can be used to map to the
discount curve with the corresponding credit spread. The client can configure
any combination of credit rating, and other attributes such as industry, country
to map to any discount curve defined within the market. These discount curves
can in turn be defined as spread curves on top of risk free curve to reflect the
credit spread premium. The resulting simulated pricing of a trade will now
reflect the adjusted credit spread for the counterparty rating. In general all
pricing models will have access to the simulated credit rating and can
therefore make use of it in pricing.

4.1.6 Default Risk Loss

In the case of a default event Razor can be configured to calculate a Loss
Given Default (LGD). This value is calculated by pricing the trade at default,
called the Exposure at Default (EAD) using the simulated market for the
corresponding scenario, and a recovery rate (RR) applied to determine the loss
suffered.
LGD = EAD * (1 – RR)
The recovery rate in Razor is currently defined as a parameter at the trade
level.
From these results a distribution of losses is built up for each simulation path,
and various statistical analyses can be performed on this including Expected
Loss, and Capital. The aggregation of these losses will be examined in relation
to Credit Value-at-Risk and Economic Capital requirements.

4.2 Credit Value-at-Risk

Credit VaR is the simulated losses to a portfolio due to both market movements
and changes to the credit quality of underlying entity. The principles of credit
rating migration and default have been described in Chapter 4, and the general
market risk VaR methodology is described in 5.3.
To capture the market, credit spread and default risk a Credit VaR simulation
must stochastically vary market rates and credit ratings. The credit simulation
must then impact the credit spread used in pricing and generate LGD in the
case of default events. Any credit mitigation effects such as netting/collateral
agreements or right to break agreements are taken into account in aggregating
exposure in a portfolio.
Then similar to market risk VaR, Credit VaR measures the worst loss that can
be expected over the specified time horizon with the degree of certainty
defined by the confidence level. It is the extension of VaR to also take account
of the effects of credit risk on the portfolio value, effects such as a change in
credit rating causing a revaluation of credit sensitive instruments in the
portfolio such as corporate and semi-government bonds, interest rate swaps,

©2009 Razor Risk Technologies Page 49 of 373

Razor Financial Principals

and credit derivatives. It is normally performed using similar time horizons as

used for VaR.

4.3 Economic Capital

Economic capital is an estimate of capital needed by a financial organization to
manage their own risk and allow the allocation of the cost of maintaining
regulatory capital to different organizational units. Economic capital differs
from "Regulatory Capital". Economic capital is an internal allocation against
risk while "Regulatory Capital" is mandated by financial regulators.
Economic Capital is calculated from the modelling of credit losses as generally
described in, but is typically over a 1 year time horizon, but may involve any
number of interim credit nodes.
For Economic capital only losses due to default are considered not credit
spread risk. Hence, the CRTM need only specify the probability of default, and
exclude other rating transition probabilities, although this is not required.
Market rates are still simulated. Moreover, trade exposure is only computed
when there is a default event, with pricing disabled in non-default scenarios
and do not contribute to the exposure.
The resulting distribution of default exposures are aggregated taking into
account any credit mitigation effects such as netting/collateral agreements or
right to break agreements.
The aggregation of results in Razor is derived from this distribution of credit
losses, and is equal to the maximum unexpected loss greater than the
expected (mean) loss to a specified confidence level.
This can be represented by the following diagrammatic distribution of losses

Distribution of Credit Losses

Mean Loss
Probability

N’th percentile

Present Value 0
of Loss Economic Capital

Unexpected Loss Expected Loss

Currently RAZOR does not support a distribution of credit losses for a particular
node which have been present valued back to the current date. The present
value of the simulated credit losses back to the start date can be supported by
©2009 Razor Risk Technologies Page 50 of 373
Razor Financial Principals

a simple method with a small enhancement in RAZOR. This simplistic approach

will use rates from the base (or current) market. Alternatively a more
sophisticated model could be introduced which would derive the discount
factor to use for present value from the simulated rates along each path, and
this discount factor is used with the specific loss generated along this path.
The distribution of credit losses as used for calculation of economic capital is
derived from those losses where the credit entity has transitioned to default,
and the credit exposure aggregated for that portfolio is positive. In other
words an actual loss has been simulated. The value for that actual loss is
calculated from the positive credit exposure less any amount recovered (as
defined by the recovery rate). In all other instances the credit loss is set to
zero. In RAZOR the recovery rates are assumed to be static, and may be set at
any level, down to the individual trade level. As a future extension to RAZOR
the recovery rates will be stochastically modelled.

4.3.1 Present Value of Losses

The credit loss may be present valued at the credit node where loss occurs, by
discounting to the current value date from the date of the credit node where
the loss is incurred. This is a configurable setting which may be turned on or
off. The curve to use for discounting is specified as part of the configuration.
<economicCapital>
<presentValue />
<presentValueCurve>RF</presentValueCurve>
</economicCapital>

4.3.2 Aggregation of Losses to Final Credit Node

The credit losses may be aggregated along the simulation path to the final
credit node of the simulation. This is a configurable setting which may be
turned on or off.
<economicCapital>
<aggregate />
</economicCapital>

4.3.3 Default Treatment

There are various other configurable settings for default treatment using the
following xml under the simulatedRiskConfig (or distributedRiskConfig).

<defaultTreatment>
<valueAtDefault>false</valueAtDefault>
<valueAfterDefault>false</valueAfterDefault>

<pushValueAtDefaultForward>true</pushValueAtDefaultForward>
©2009 Razor Risk Technologies Page 51 of 373
Razor Financial Principals

</defaultTreatment>

1. valueAtDefault will value on the first node after the default event and then
assume 0 value for the remainder of the path.
2. valueAfterDefault will value on all nodes after the default event.
3. pushValueAtDefaultForward will value on the first node after the default
event and then use that same value for the remainder of the path.

All options default to false.

©2009 Razor Risk Technologies Page 52 of 373

Razor Financial Principals

Chapter 5
Market Risk Measurement
Market risk is the risk of a decrease in value of a portfolio of investments,
trades, and positions due to changes in the value of the market risk factors
which determine the market value of the portfolio.
Measuring this risk provides financial institutions with the knowledge to
properly manage the risk of portfolio losses, and the understanding to the
composition of this risk, and enables the setting aside of regulatory capital to
protect them against these potential portfolio losses.
There are various methods for measuring this risk, including the more
traditional measures of sensitivity such as interest rate sensitivity such as
PV01, basis point sensitivity, duration, convexity, and option based sensitivity
measures (the ‘Greeks’) such as delta, gamma, rho, vega, theta.
Market risk is also typically measured using a Value at Risk methodology. Razor
provides support for both the Historical, and the Monte Carlo VaR
methodologies

5.1 Value-at-Risk (VaR)

VaR, under the assumption of prevailing normal market conditions, is a
probabilistic bound of market losses. It is the worst-case loss expected over a
defined period of time, t (the holding period) within the defined level of
certainty, or probability defined by the confidence interval c. In other words
larger losses are still possible, but with a lower probability than the confidence
interval.
There are various non-simulation based statistical VaR methodologies,
including RiskMetrics, and Parametric VaR which use variance-covariance data,
with delta, and extension to gamma approximations.
The various simulation based VaR methodologies are scenario based methods,
whereby the revaluation is performed for and using each generated market
scenario, and the resultant values are compared to the current value. All these
methods derive the statistical analysis from a distribution of changes in value,
created from the difference between each of the simulated calculated values
and the current value.
Each simulated calculated value is calculated from a revaluation using a
specific simulated market scenario for that value. The simulation generates a
distribution of market scenarios, by generating a new set of market risk factors
for each of the scenarios.
Portfolio VaR is the same conceptually as that for an individual trade. The
portfolio VaR is derived from the current portfolio value, and a distribution of
portfolio values, where each portfolio value is the sum of the values for each
trade valued with the specific generated scenario.
The simulation methodologies differ in how they generate the distribution of
scenarios used.

©2009 Razor Risk Technologies Page 53 of 373

Razor Financial Principals

More formally, the VaR for a particular holding period may be derived from a
probability density function fτ (w) for the continuous random variable W of the
change in portfolio value with values w , for that holding period τ .
The probability p of a change in portfolio value w being lower than a level V
V
is ∫ fτ ( w)dw .
−∞

V
p = P(W < V ) = ∫ fτ ( w)dw .
−∞

Therefore the probability of a change in portfolio value not being less than this
value V is 1 − p . Hence 1 − p represents the confidence level c .
∞
c = 1 − p = P(W > V ) = ∫ fτ ( w) dw
V

In other words, with confidence level c the worst change in portfolio value
time horizon τ is V . Hence V is the value for VaR, and it is reported as a
positive number.
So for example, over a time horizon of 1 day, if 5% of the changes in portfolio
value are less than -10,000 (ie more negative than -10,000), then 95% of the
changes in portfolio value are greater than or equal to -10,000 (ie more
positive than -10,000). In other words if 5% of the portfolio losses are greater
than a loss of 10,000, then 95% must be less than or equal to 10,000. Hence to
a confidence level of 95% the largest likely loss will be 10,000, and this is the
reported VaR.

5.2 Historical Simulation

Historical simulation is a scenario-based approach to measuring market risk,
where the distribution of values is derived from revaluations using the historic
simulated scenarios S tτ .
Using the notation, where
S is a scenario, which contains market risk values Fi ,
Fi (t ) is the value of a market risk factor i observed at time t ,

S 0 is the current market scenario,

Fi (0) are the current market risk values Fi (0) in scenario S 0 ,

then scenario S tτ is defined by the application to the current market scenario

S 0 , containing the current market risk values Fi (0) , the historic changes in
value for these market risk factors from time t to time t + τ , as obtained from
the historic time series of market risk factor values.

©2009 Razor Risk Technologies Page 54 of 373

Razor Financial Principals

Hence for Stτ , each market risk factor value Fiτ (t ) is derived as
Fiτ (t ) = Fi (0) + ∆τ Fi (t ) , where ∆τ Fi (t ) = Fi (t + τ ) − Fi (t ) is the change in value
of a market risk factor i observed from time t to time t + τ .
This change ∆τ Fi (t ) can be set for each class of risk factors to be absolute (
Fi (t + τ ) − Fi (t ) , as above), relative ( Fi (t + τ ) / Fi (t ) ), or zero (for example as
use in partial risk as described below).
Historical simulation has the advantage of using actual real historic changes in
market risk factors to generate the distribution of values, and hence there are
no assumptions with respect to the distribution of risk factor values, the
volatility/correlations are as they actually occurred, and takes account of
outliers, and fat tails. Some of the disadvantages to be aware of include that
trends reflected in the historical data do not necessarily reflect current market
conditions, and also since it is possible for outliers to distort the VaR, careful
analysis of the outliers may be required.
An adequate VaR number depends on the availability of an adequate set of
historic data. It is recommended that at least 1 year of historic data is used.
The regulatory framework of the Basle Internal Models Approach requires
Historic VaR to be calculated for 1 day and 10 day holding periods with a 99%
confidence level. The 10 day VaR may be calculated as an approximation from
the 1 day VaR ( VaR10 day = 10 * VaR1day ), but ideally should be directly
calculated using the same historic simulation methodology as the 1 day VaR.

5.2.1 Filtered Historical VaR Simulation

Razor provides an additional volatility weighting method, called the filtered
HSVaR method, to adjust each scenario return value based on its history. The
procedures are as follow:
For a set of returns r, first calculate the set of volatilities σ2 then the set of
weightings w. The new modified return r* is r multiplied by w.
σ t2 = ασ t2−1 + (1 − α )rt 2−1
σt +σ N
w=
2σ t
rt is the return for scenario t
σ t2 is the volatility for scenario t
wt is the weighting for scenario t
rt* is the modified return for scenario t
α is the constant decay factor
σ N is the last in the sequence t = 1..N
The decay factor constant can be defined by the user, and if not specified,
Razor will assume a value of 0.97.

©2009 Razor Risk Technologies Page 55 of 373

Razor Financial Principals

5.3 Monte Carlo Simulation

Monte Carlo simulation is a scenario-based approach to measuring market risk,
where the distribution of values is derived from revaluations using the Monte
Carlo simulated scenarios S τ .

Monte Carlo simulation uses the generalized Monte Carlo simulation framework
as described in detail in Chapter 2, to generate the set of simulated changes in
the risk factors which are then used to create the scenarios S τj , for time
horizon τ .
The market risk factors are stochastically modelled forward in time τ
according to a set of correlated diffusion processes. The returns for each
market risk factor are assumed to have a specific probability distribution, such
as normal, or lognormal, and evolve using a stochastic process defined by
volatility and with or without mean reversion. From the set of market risk
factors and the probability distributions, it is necessary to estimate or define
the stochastic parameters for each of the risk factors and the correlations
between all of the risk factors. This can be by various methods, such as
deriving from a historical set of values for the market risk factors or defining
externally by any other mechanism, or from some combination of these
methods.
From this input the Monte Carlo simulation process generates m simulated
relative changes or returns S∆τ j Fi , in the n market risk factors Fi , over the
time horizon τ , by correlated random sampling.

Hence, in a similar fashion as for Historic VaR, for scenario S τj , each market
risk factor value Fiτ ( j ) is defined as the value of the i’th risk factor in the j’th
scenario for time horizon τ , and is derived as Fiτ ( j ) = Fi (0) * S∆τ j Fi .

Advantages of the Monte Carlo simulation methodology include the ability to

apply a theoretical distribution framework, and have a very large number of
distributions without the need for historical data of the same size.
Disadvantages include the high level of computational processing, and that
outliers are not included, and the assumed distributions do not adequately
describe the real world.

5.4 Analysis of VaR

To assist in understanding the important factors which influence the value of
the overall portfolio VaR (as calculated using either Historical or Monte Carlo
VaR), it is useful to analyse the composition of the portfolio VaR, by measuring
the contributions made by market risk factors, either single, multiple
collections, or general classes, and also by measuring the contributions made
by single trade, set of trades, or sub-portfolio.
The general classes of risk factors include equity stock prices, interest rates,
foreign exchange rates, commodity prices/rates, credit spreads, and implied
market volatilities. It is possible to define any collection of risk factors and/or
set of trades and hence to measure the contribution made by this collection to
the portfolio VaR

©2009 Razor Risk Technologies Page 56 of 373

Razor Financial Principals

5.4.1 Partial Risk

Partial risk measures the effect on portfolio VaR due to the volatility of the
individual defined set of market risk factors, where this set is composed of
either a single, a collection, or a class of market risk factors.
The portfolio VaR due to the volatility of the specific set of risk factors is
calculated by performing a VaR calculation where the volatility of all other risk
factors is set to zero. This will then measure the VaR due to that specific set of
risk factors.
A typical use of this would be to measure the VaR due to each individual
market risk class, and within the foreign exchange rate risk class measure the
VaR due to the individual common cross rates. The sum of these partial
portfolio VaR values can the be compared to the overall portfolio VaR to
measure the portfolio effect due to the correlation in movements between all
of the individual risk factors.

5.4.2 Incremental VaR

Incremental VaR measures the incremental effect on portfolio VaR due to a
change in the portfolio composition through for example the introduction of a
new set of trades, or due to the inclusion of the effects of the volatility of a
single risk factor or a class of risk factors. It is calculated from the difference
in the overall portfolio VaR before and after the change.
So for example, the incremental VaR for a portfolio due to the introduction of
a set of trades is equal to the portfolio VaR after the set of trades is included
in the portfolio less the portfolio VaR before the trades were added.
As another example, the incremental VaR for a portfolio due to the volatility of
the interest rate risk factors is equal to the overall portfolio VaR less the
portfolio VaR calculated by a partial risk calculation as described above which
includes all risk factors except the interest rate risk factors. This is the same
effect as setting the volatility for the interest rate factors to zero.

5.4.3 Marginal VaR

Marginal VaR is similar to incremental VaR, but measures the changes in
portfolio VaR due to more subtle changes in risk factors and portfolio
composition, rather than a straight out inclusion/exclusion. Examples of such
changes would be the change in portfolio VaR due to small change in positions
within the portfolio, small changes in the current value of individual risk
factors, and changes in volatility of a risk factor.

©2009 Razor Risk Technologies Page 57 of 373

Razor Financial Principals

Chapter 6
Market Rates, Prices and Curves
The various pricing routines in RAZOR make use of both observed and implied
security prices and theoretical structures that describe the characteristics of
the financial market at a point in time. This information is encapsulated by an
object in RAZOR called a Market.

6.1 The Market - Rates and Prices

The market object contains a set of rates and prices for various assets, yield
curves, forward curves, and volatility surfaces implied from those rates and
prices. The market object is arranged into a set of term structures, where each
term structure is defined for a specific asset as a collection of related rate
classes by term (relative term or discrete date), and where each rate class is a
single observable element of market data. For a more complete description of
these definitions and the relationship between each of these components
please refer to the Razor technical documentation.
Rates and prices in RAZOR are uniquely identified in a market by a rate class
identifier. Each rate class is differentiated by the elements that make that
type of rate unique — for example, a one month BAB rate class identifies that
the rate is based on a one month BAB future. The rate identifier by contrast,
identifies the particular instance of the rate.

6.2 Interest Rate Curves

6.2.1 Interest Rates and Security Prices
Zero Rate
Zero rate is simply the rate of return on a zero-coupon bond or any non-coupon
bearing instruments, e.g. cash deposits, bills or any single cashflow.
Z t →t1 = the annualised zero rate at time t1 .

d t →t1 = t1 − t = number of days from t to t1 .

D = number of days in a year (depends on the day convention used, see
reference)
P(t , t1 ) = the price of a zero-coupon bond at time t maturing at time t1 .
F = the face value of the zero-coupon bond.

The price of a zero-coupon bond can be determined as:

 
 
P(t , t1 ) = F 
1 
 d t →t1 
 1 + Z t →t1 ⋅ 
 D 

©2009 Razor Risk Technologies Page 58 of 373

Razor Financial Principals

If the zero-coupon bond price is observed, the zero rate can be uniquely
determined as:
 F  D
Z t → t1 =  − 1  ⋅
 P (t , t 1 )  d t → t1

When F = 1 , the price of the zero-coupon bond P(t , t1 ) is called the discount
factor at time t1 and we denote it as D(t , t1 ) , i.e.

D(t , t1 ) =
1
and equivalently
d t →t1
1 + Z t →t1
D
 1  D
Z t →t1 =  − 1 ⋅
(
 D t , t1 )  d t →t1

Forward Rate
Forward rate is the interest rate implied by current zero rates for a specified
future time period.
Define:
f (t , t1 , t 2 )
= the forward rate
= the implied annualised interest rate for future time period t1 to t 2 at
current time t .

The forward rate must mathematically satisfy the following equality:

 d  d   d 
1 + Z t →t1 ⋅ t →t1 1 + f (t , t1 , t 2 ) ⋅ t1 →t2  = 1 + Z t → t 2 ⋅ t → t 2 
 D   D   D 
  
The reason for the above equality to hold is explained by the arbitrage-free
theory:
If LHS > RHS
Short $1 of a t 2 -year zero-coupon bond.

Use the shorted $1 to invest in a t1 -year zero-coupon bond and lock in a

forward interest rate to invest for the period t1 to t 2 .

At the end of time t 2 , we will have the invested amount equal to the LHS and
we need to return the shorted amount with interest equal to the RHS. Since
LHS > RHS, an arbitrage profit of LHS – RHS is earned.
If LHS < RHS
Short $1 of a t1 -year zero-coupon bond and lock in a forward interest rate to
pay for the period t1 to t 2 .

©2009 Razor Risk Technologies Page 59 of 373

Razor Financial Principals

Use the shorted $1 to invest in a t 2 -year zero-coupon bond.

At the end of time t 2 , we will have the invested amount equal to the RHS and
we need to return the shorted amount with interest equal to the LHS. Since
LHS < RHS, an arbitrage profit of RHS – LHS is earned.
In an efficient market, investors will exploit the arbitrage opportunities. This
will force any arbitrage opportunity to be eliminated as soon as they arise and
fail to exist in the market. This means that the above mathematical equality
has to hold, i.e. LHS = RHS, in an efficient and arbitrage-free market.
The above mathematical equality can also be written as:
−1
 d t →t 
D(t , t 2 ) = D(t , t1 )1 + f (t , t1 , t 2 ) ⋅ 1 2 

 D 

Thus if the zero rates (or the price of the zero-coupon bonds) are observed,
then the forward rate can be uniquely determined as:

 d t →t 2   dt →t1   D
f ( t , t1 , t2 ) = 1 + Z t →t2 ⋅  / 1 + Z t →t1 ⋅  − 1 ⋅
 D   D   dt1 →t2
 D ( t , t1 )  D
= −1 ⋅
 D ( t , t )  d
 2  t1 →t2
Swap Rate
Swap rate is the fixed rate in an interest rate swap that causes the swap to
have a value of zero.
It is important to note that on the floating side of the swap, the present value
of all cash flows must equal to the principal amount. Thus for the interest rate
swap to have a value of zero, the present value of all the cash flows on the
fixed side of the swap must also equal to the principal amount. It implies that
the fixed rate on the swap is equivalent to the yield-to-maturity of a bond that
equates the present value of all cashflows of the bond to the face value of the
bond (i.e. the par yield). As a result, it is convenient to think the swap rate as
a par yield rate.

Bond Yield-to-Maturity Rate and Bond Price (Currently Under Development)

Bill Discount Rate (Currently Under Development)

6.2.2 Market specific interest rate curves

Market specific interest rate curves include Swap, Treasury, Corporate, Minor
Market (CP, BA, CD, Prime, Tbill, Fedfund, etc)), BMA, OIS, cross currency
curves, 1M Libor, 3M Libor Curves.
Specific interest rate curves in the market are built from collections of specific
interest rate market rates and security prices.
Currently Razor supports the configuration of any specific curves as long as
these curves are composed of rates/prices which Razor’s bootstrapping

©2009 Razor Risk Technologies Page 60 of 373

Razor Financial Principals

supports. This includes cash deposit, cash rates, swap rates, and futures
points. Currently security (bond) yield to maturity rates and security (bond)
prices are not supported in Razor, and there is no support for convexity
adjustments. Inflation Curves are currently not supported.
Specific market curves which contain rates/prices which are not directly
supported by Razor’s current bootstrapping function can still be supported by a
conversion of these points to points in the form above. This will involve some
integration layer / ETL processing.

6.2.3 Spread Yield Curve

This curve is defined by the standard bootstrapping as described in detail
above for simple yield curves, where each of the input points on this curve is
first derived from the corresponding reference curve input point plus the
spread point. The dates for the resultant set of input points for the spread
yield curve matches the dates for the reference curve input points.

For each input rate rtir , at time t i , on the reference yield curve r ,the
resultant input rate rtic at time t i , on the spread yield curve c is defined by

rtic = rtir + sti , where sti , is the spread rate at time t i

If no spread rate sti exists at time t i , then the spread rate to use is derived by
linear interpolation between the previous and next spread rates. If there is no
previous spread rate, then linear interpolate between the next spread rate and
a spread rate of zero at start of curve, time t . If there is no next spread rate,
then rtic = rtir .

6.2.4 Composite Yield Curve

This curve is defined as a function of two bootstrapped yield curves. The
discount factor Dc (t ,t1 ) ,for a specific date t1 on the composite curve c is
calculated from the corresponding respective discount factors from each of the
two bootstrapped yield curves, where each of these discount factors is
converted into a zero rate, the zero rates are summed, and the result Z tc→t1 is
then converted back to a discount factor. Hence

Dc (t , t1 ) =
1
d t →t1
1 + Z tc→t1 ⋅
Dc

Z tc→t1 = Z t1→t1 + Z t2→t1

 1  D
Z t1→t1 =  − 1 ⋅ c , the zero rate at time t1 from first yield curve
D (
 1 1t , t )  d t →t1
 1  D
Z t2→t1 =  − 1 ⋅ c , the zero rate at time t1 from second yield curve
 D2 (t , t1 )  d t →t1
Dc , the number of days in year, as defined on composite curve

©2009 Razor Risk Technologies Page 61 of 373

Razor Financial Principals

d t →t1 , number of days from time t to time t1

t , current time
t1 , time for which discount factor required from composite curve

6.2.5 Adjustment Factors Applied to Yield Curves

The yield curve framework in Razor supports an interface which allows for
customization of the bootstrapped curve, to adjust the curve by an additive
factor and/or a multiplicative factor.
The additive factor is applied before the multiplicative factor. The default
additive factor is 0.0, and the default multiplicative factor is 1.0. The factors
however are only applied if explicitly supplied.
The adjusted discount factor is calculated from the unadjusted discount factor
D(t , t1 ) , as provided from the bootstrapped curve as follows.

 1  D
Z t →t1 =  − 1 ⋅ , unadjusted zero rate
 D(t , t1 )  d t →t1
( )
Z tA→t1 = Z t →t1 + A * M , adjusted zero rate

D A (t , t1 ) =
1
, adjusted discount factor
d t →t1
1+ Z A
t →t1 ⋅
Dc

6.2.6 Negative Forward Rates

To eliminate the creation of simulated yield curves which may contain negative
forward rates and hence cause certain trades to fail to price during a
simulation, Razor applies the following method to remove the existence of
negative forward rate in all yield curves when they occur under simulation.
Post the bootstrapping of all simulated yield curves or scenarios, the pricing
server validates each yield curve (including the base scenario) by iterating
through the ordered set of discount factors, checking each discount factor
against the previous good discount factor from the term structure, and if it is
greater then removing it. From the resultant yield curve the normal
interpolation and extrapolation will then operate on the remaining discount
factors
Since the base scenario is representative of the current market yield curve, if
validation detects negative forward rates then the yield curve will be rejected
rather than be modified and any trade which uses this curve for valuation will
fail to find the curve and hence be rejected as failed to price.
There is a detailed log of all analysis and actions for any yield curve (base or
simulated) that contains negative forward rates (ie increasing discount
factors).

©2009 Razor Risk Technologies Page 62 of 373

Razor Financial Principals

6.3 Yield Curve Construction

By definition, yield curve is the curve that depicts the relationship between
zero rates and their maturities. Thus the objective of the yield curve
construction routines is to determine the values of zero coupon bond rates for
any maturity forward in time given the observed market prices of interest-
based securities.
The price of market securities embodies several considerations for a market
participant. When an interest rate security is exchanged the buyer of the
security is giving up an amount of equity in order to gain a certain return. In
general, the longer the cash is tied up in the security, the greater the return
the investor will require from the security.
Term structure analysis examines market securities to imply a curve that
exhibits an equilibrium time value of money. The process of determining the
attributes of this curve from market securities is known as “bootstrapping”, as
the process uses the value of near-dated securities to help determine the
theoretical value of further dated zero coupon bonds.

6.3.1 Bootstrapping
In RAZOR, the bootstrapping process is driven by a bootstrapping
configuration. The bootstrapping configuration determines what assets to use
to generate the yield curve. Assets are referenced by an asset class, an
identifier that represents the type of security used at a particular point on the
yield curve - for example; a one month bank accepted bill rate.
Different types of securities have different cash flow profiles and different rate
calculation methods. These differences need to be taken into account when
determining the theoretical value of a zero coupon bond at the specific
maturity.

6.3.2 Bootstrapping the Yield Curve Using Different Securities

Cash Deposits
Cash deposits are short-term non-coupon-bearing instruments issued to raise
short term capital.
The method to determine the price of a bill is the same as the zero-coupon
bond. We simply discount the face value by the corresponding zero rate, i.e.

 
 
P(t , t1 ) = F 
1 
 d t → d1 
 1 + Z t →t1 ⋅ 
 D 
It is very straightforward to bootstrap the zero rates if we are given the prices
of the bills. The zero rates can be found using the formula

©2009 Razor Risk Technologies Page 63 of 373

Razor Financial Principals

Z t → t1 =
(F − P (t , t 1 ))D .
P (t , t 1 )d t → t1

Example:
If we have the price for a 30-day bill with face value 100 equal to 99 (assuming
360 days in a year), the annualised 30-day zero rate is:
360(100 − 99 )
Z t →t1 = = 12.12% .
30 × 99

While it is very simple to determine zero rates using bills, it should be realised
that bills have very short-maturities. To construct a yield curve with maturities
up to say 40 years, it is obvious that we do not have bills with equivalent time-
to-maturities.

Bill Futures
A bill futures contract is a futures contract on bills. It obliges the buyer/seller
of the futures contract to buy/sell the underlying bill at a predetermined
delivery price in the specified future time.

Define:
B(t , t1 , t 2 ) = the futures price of a bill futures contract that begins at time t1
and matures at time t 2

100 − B(t , t1 , t 2 )
B(t , t1 , t 2 ) is quoted in a way such that is the annualised
100
interest rate applied to the period t1 to t 2 at the current time t .

100 − B(t , t1 , t 2 )
Thus it is intuitively clear that is just the annualised forward
100
rate applied to the period t1 to t 2 from the perspective of current time t .
Mathematically, we can write:

100 − B(t , t1 , t 2 )
f (t , t1 , t 2 ) =
100

Thus if we are given the bill futures price, we can uniquely determine the
forward rate and then bootstrap the zero rate for t 2 from this forward rate
and the zero rate we bootstrapped or given at time t1 .
Example:
Assume there are 360 days in a year and given the 180-day discount factor is
0.9730. We have a 90-day bill futures contract that begins in 180 days and the

©2009 Razor Risk Technologies Page 64 of 373

Razor Financial Principals

rate of the bill futures contract is 95. We want to determine the 270-day zero
rate, i.e. we want to calculate Z 270 .
t→
360

Answer:

 180 270 
We are given B t , ,  = 95 . We can use the forward formula
 360 360 
100 − B(t , t1 , t 2 )
f (t , t1 , t 2 ) =
100
to calculate the annualised forward rate implied for the period from 180 days
to 270 days, i.e.

 180 270  100 − 95

f  t, , = = 0.05
 360 360  100
 180 270 
We have calculated f  t , ,  and we are given the 180-day discount
 360 360 
 180 
factor D t ,  = 0.9730 , we can use the formula
 360 
−1
 d t →t 
D(t , t1 ) ⋅ 1 + f (t , t1 , t 2 ) ⋅ 1 2 
 = D(t , t 2 )
 D 
 270 
to work out the 270-day discount factor D t ,  as:
 360 
 270 
D t , 
 360 
−1
 d 180 270 
 180    180 270  360 → 360 
= D t ,  ⋅ 1 + f  t , , ⋅ 
 360    360 360  360 
 
−1
 270 − 180 
= 0.9730 × 1 + 0.05 ⋅ 
 360 
= 0.9610
The 270-day zero rate can be obtained directly from the formula

 1  D
Z t →t1 =  − 1 ⋅ so that
 D (t , t1 )  d t →t1
Z 270
t→
360

 
 
=  1
− 1 ⋅
360
  270   d
 D t ,   t → 360
270

  360  
©2009 Razor Risk Technologies Page 65 of 373
Razor Financial Principals

 1  360
= − 1 ⋅
 0 .9610  270
= 5.41%

Swaps
A swap contract is a contract that exchanges a fixed rate of interest on a
certain notional principal for a floating rate of interest on the same notional
principal.
The swap rate is the fixed rate in a swap that causes the swap to have a value
of zero. As we have discussed in section 6.1.2, the swap rate can be thought as
a par yield (i.e. the yield-to-maturity that makes the present value of all cash
flows of a bond equal to the face value of the bond).
The construction of a yield curve usually requires a very long maturity term.
Bills and bill futures generally do not have the equivalent maturities. However,
swaps can have very long maturity term (e.g. 40 years). It is thus very
important to find a method to construct the yield curve using swaps.

Define:
rs (c, N ) = the annualised swap rate of a swap contract that exchanges
interest payment c times a year with maturity term N
F = the principal amount under the swap

By the definition of a swap rate, the following mathematical equality must be

satisfied:
rs (c, N )  1  r (c, N )  2  r (c, N )  cN   cN 
F=F D t ,  + F s D t ,  + ... + F s D t ,  + FD  t , 
c   c c  c  c  c   c 

The above can be simplified to

rs (c, N ) cN −1  i 
1− ∑ D t , 
 c
D(t , N ) =
c i =1
rs (c, N )
1+
c
Using the above formula, we can bootstrap the yield curve based on the swap
contract.
Example:
Assume there are 30 days in a month and 360 days in a year. We have a 1 year
swap contract with the annualised swap rate 6% that exchanges interest
payment quarterly. We are given the 3-month, 6-month and 9-month discount
factors equal to 0.9851, 0.9698 and 0.9557 respectively. We need to bootstrap
the 1 year zero rate using the information given above.
Answer:

©2009 Razor Risk Technologies Page 66 of 373

Razor Financial Principals

From the question, we are given

rs (4,1) = 0.06

 90   1
D t ,  = D t ,  = 0.9851
 360   4
 180   2
D t ,  = D t ,  = 0.9698
 360   4
 270   3
D t ,  = D t ,  = 0.9557
 360   4
Using the formula
rs ( c, N ) cN −1  i 
1− ∑ D  t, 
 c
D (t, N ) =
c i =1
rs ( c, N )
1+
c

rs ( 4,1) 3
1− ∑ D (t, i )
D ( t ,1) =
4 i =1
r ( 4,1)
1+ s
4
r ( 4,1)
1− s
4
( D ( t ,1) + D ( t , 2 ) + D ( t ,3) )
=
r ( 4,1)
1+ s
4
0.06
1− ( 0.9851 + 0.9698 + 0.9557 )
= 4
0.06
1+
4
= 0.9422

Using the zero rate formula

 1  D
Z t →t1 =  − 1 ⋅
 D (t , t 1 )  d t →t1
The 1 year zero rate is:

 1  360
Z t →1 =  − 1  ⋅
 D ( t ,1)  d t →1
 1  360
= − 1 ⋅
 0.9422  360
= 6.13%

©2009 Razor Risk Technologies Page 67 of 373

Razor Financial Principals

Effective Swap Rate

As described above, when bootstrapping the next zero rate or discount factor
at t k from a swap rate at t k of frequency c, then the solution requires the
(t k c − 1) number of zero rates or discount factors for times t 1 to t
k−
1 . If all of
c c
these required discount factors are readily available as they have all been
previously directly bootstrapped, or may be derived by interpolation (support
different types of interpolation) as they fall before the term of the last zero
rate bootstrapped t i , then deriving at t k is straight forward.

For the case where the term of the last known zero bootstrapped t i does not
cover the terms for all required zero rates, then we need to derive in order
these interim zero rates first.
Assume t j is the term of the next interim zero rate which is required to be
bootstrapped but for which there is no direct swap rate of frequency c
available, t i is the term for the most recently bootstrapped zero rate,
t j −1 = t j − 1 / c , t j −1 ≤ t i , and t i < t j < t k .

To bootstrap the zero rate at t j , we firstly need to derive a swap rate at this
term. This is achieved by interpolating between the next available swap rate
at t k , and an effective swap rate of the same frequency derived at t i , or is
this t j −1 .

Once we find the effective swap rate at time t i , with the given swap rate at
time t k , we linearly interpolate the swap rate at time t j . After obtaining the
swap rate at time t j , the discount factor and zero rate at time t j can be
evaluated as we have done in the previous section.
Define Effective Swap Rate
An effective swap rate is a swap rate that is implied by the discount factors
An effective swap rate of maturity ti, and frequency c, can be derived from an
existing bootstrapped yield curve, or known series of zero rates or discount
factors for terms t1 to tn, spanning ti.
It is very simple to calculate the effective swap rate:
Recall:
rs (c, N ) cN −1  i 
1− ∑ D t , 
 c
D(t , N ) =
c i =1
rs (c, N )
1+
c
Rearrange the formula and we can calculate the effective swap rate as:
c(1 − D(t , N ))
rs (c, N ) = cN
 i
∑i =1
D t , 
 c

©2009 Razor Risk Technologies Page 68 of 373

Razor Financial Principals

Example:
Assume there are 30 days in a month and 360 days in a year. We are given the
3-month, 6-month and 9-month and 12-month discount factors equal to 0.9851,
0.9698, 0.9557 and 0.9422 respectively. We want to find the effective swap
rate for a 1-year swap contract with quarterly exchanges of interest payments.
Answer:
We are given

 90   1
D t ,  = D t ,  = 0.9851
 360   4
 180   2
D t ,  = D t ,  = 0.9698
 360   4
 270   3
D t ,  = D t ,  = 0.9557
 360   4
 360 
D t ,  = D(t ,1) = 0.9422
 360 
Using the effective swap rate formula
c(1 − D(t , N ))
rs (c, N ) = cN
 i
∑i =1
D t , 
 c

4 (1 − D ( t ,1) )
rs ( 4,1) = 4
 i
∑ D  t , 4 
i

4 (1 − D ( t ,1) )
=
 1  2  3
D  t ,  + D  t ,  + D  t ,  + D ( t ,1)
 4  4  4
4 (1 − 0.9422 )
=
0.9851 + 0.9698 + 0.9557 + 0.9422
= 6%
Thus the 1-year effective swap rate is 6%.
The effective swap rate formula is very useful to bootstrap zero rates.
Consider the following situation:
We want to find the 1.25-year zero rate. We have a 3-year swap with quarterly
interest exchange. We also have the 90-days, 180-days, 270-days and 1-year
discount factors obtained from other short term instruments.
From the formula we had before, to find the 1.25-year discount factor, we will
need a swap contract with maturity at 1.25 years. However, we only have a 3-
year swap contract. We even cannot interpolate the 1.25-year swap rate
because we do not have any other swap contract besides the 3-year swap. The
©2009 Razor Risk Technologies Page 69 of 373
Razor Financial Principals

method on effective swap rate is now useful for us to calculate an effective 1-

year swap rate with the corresponding discount factors. We then are able to
interpolate the 1.25-year swap rate and apply our usual formula to find the
1.25-year discount factor. Once we know the discount factor, the zero rate is
obtained easily.
Example:
Assume there are 30 days in a month and 360 days in a year. We have a 3-year
swap contract with the annualised swap rate 5.82% that exchanges interest
payments quarterly. We are given the 3-month, 6-month, 9-month and 12-
month discount factors equal to 0.9851, 0.9698, 0.9557 and 0.9422
respectively. We need to bootstrap the 1.25-year zero rate using the
information given.
Answer:
We can use the discount factors to calculate the 1-year effective swap rate
and then interpolate the 1.25-year swap rate with the 1-year effective swap
rate and the 3-year swap rate.
Given:

 90   1
D t ,  = D t ,  = 0.9851
 360   4
 180   2
D t ,  = D t ,  = 0.9698
 360   4
 270   3
D t ,  = D t ,  = 0.9557
 360   4
 360 
D t ,  = D(t ,1) = 0.9422
 360 
rs (4,3) = 0.0582
The 1-year effective swap rate can be obtained from the formula
c(1 − D(t , N ))
rs (c, N ) = cN
 i
∑i =1
D t , 
 c
and we have calculated in the previous example that the 1-year effective swap
rate rs (4,1) is 6%.
Using linear interpolation (for details on linear interpolation, please refer to
the next section):
rs ( 4,3) − rs ( 4,1)
rs ( 4,1.25) = rs ( 4,1) + (1.25 − 1)
3 −1
0.0582 − 0.06
= 0.06 + (1.25 − 1)
3 −1
= 5.98%
With the interpolated 1.25-year swap rate and the corresponding discount
factors, we can find the 1.25 discount factor using the formula:

©2009 Razor Risk Technologies Page 70 of 373

Razor Financial Principals

rs (c, N ) cN −1  i 
1− ∑ D t , 
 c
D(t , N ) =
c i =1
rs (c, N )
1+
c
We can now find D(t ,1.25) as:

rs ( 4,1.25 ) 4  i 
1−
4
∑ D  t, 
 4
D ( t ,1.25 ) = i =1
rs ( 4, N )
1+
4
rs ( 4,1.25) )   1   2  3 
1−  D  t , 4  + D  t , 4  + D  t , 4  + D ( t ,1) 
4        
=
rs ( 4,1.25 )
1+
4
0.0598
1− ( 0.9851 + 0.9698 + 0.9557 + 0.9422 )
= 4
0.0598
1+
4
= 0.9285
It is straightforward to find the 1.25-year zero rate Z t →1.25 :

 1  D
Z t →t1 =  − 1 ⋅
 D(t , t1 )  d t →t1
 1  360
Z t →1.25 =  − 1 ⋅
 D ( t ,1.25 )  dt →1.25
 1  360
= − 1 ⋅
 0.9285  360 (1.25 )
= 6.16%

6.3.3 Interpolation Methods

From previous sections, we can see that our rates are all given or calculated at
discrete points in time. It is often possible that we require rate that lies
between two discrete points. Thus we need some methods to interpolate the
required rate using rates on the discrete points.

Linear Interpolation
This is the simplest way to interpolate the required rate. If the rate at time t i
is ri and the rate at time t j is r j and we want to work out the rate at a time
between t i and t j , i.e. to calculate rα at time tα for some α ∈ (i, j ) . Linear
interpolation assumes there is a linear relationship between ri and r j during

©2009 Razor Risk Technologies Page 71 of 373

Razor Financial Principals

the time from t i to t j . Thus we can calculate any rate rα at time tα using
linear interpolation formula:
r j − ri
rα = ri + (tα − ti ) .
t j − ti
Rate here is a very general definition. It can be spot rate, discount rate, log-
rate, etc.
We will now discuss four options that are based on linear interpolation.
Linear Spot
This is the situation where we have got two spot rates Z t →t a and Z t →tb at time
t a and t b respectively. We need to find the spot Z t →ti at time t i for some
i ∈ (a, b ) using linear interpolation.
Example:
The 3-month spot rate is 6.01% and the 6-month spot rate is 6.11%. Calculate
the 5-month spot rate using the concept of linear spot.
Answer:
Z 3 = 0.0601
t→
12

Z 6 = 0.0611
t→
12

Using the linear interpolation formula:

Z −Z
t→
6
t→
3
 
Z 5 =Z 3 + 12 12
t 5 −t 3 
t→ t→ 6 3  12 12 
12 12 −
12 12
0.0611 − 0.0601 5 3
= 0.0601 + ( − )
6 3 12 12
−
12 12
= 6.08%

Linear Discount Factor

This is similar to linear spot except linear discount factor is to linearly
interpolate discount factors rather than spot rate.
Example:
The 3-month discount factor is 0.9852 and the 6-month discount factor is
0.9704. Calculate the 5-month discount factor using linear discount factor.
Answer:

 3
D t ,  = 0.9852
 12 

©2009 Razor Risk Technologies Page 72 of 373

Razor Financial Principals

 6
D t ,  = 0.9704
 12 
Using linear interpolation formula:

 5
D t , 
 12 
 6  3
D t ,  − D t , 
 3  12   12  ⋅  5 − 3 
= D t ,  +  
 12  6
−
3  12 12 
12 12
0.9704 − 0.9852  5 3 
= 0.9852 + ⋅ − 
6 3
−  12 12 
12 12
= 0.9753

Log Linear Spot

This is similar to linear spot. However, instead of linearly interpolate spot rate,
we will linearly interpolate the log of the spot rate and then obtain the spot
rate by calculating the exponential of the linearly interpolated log-spot rate.
Example:
Calculate the 5-month spot rate from the information in the example of linear
spot using log linear spot method.
Answer:
Z 3 = 0.0601
t→
12

Z 6 = 0.0611
t→
12

Use log linear spot with the linear interpolation formula:

 
log Z 5 
 t →12 
   
log Z 6  − log Z 3 
   t →
12   t →
12   5 3
= log Z 3  + ⋅ − 
 t→
12 
6
−
3  12 12 
12 12
log(0.0611) − log(0.0601)  5 3
= log(0.0601) + ⋅ − 
6 3
−  12 12 
12 12
= -2.80074

©2009 Razor Risk Technologies Page 73 of 373

Razor Financial Principals

Z 5
t→
12

  
= exp log Z 5  
 t→ 
  12  

= exp(− 2.80074)
= 6.08%

Log Linear Discount Factor

This is similar to log linear discount factor except that we need to linearly
interpolate the log-discount factor instead of the discount factor. We can
obtain the discount factor by calculating the exponential of the linearly
interpolated log discount factor.
Example:
Assuming the information in the example of linear discount factor, calculate
the 5-month discount factor using log linear discount factor.
Answer:

 3
D t ,  = 0.9852
 12 
 6
D t ,  = 0.9704
 12 
Use log linear discount factor with the linear interpolation formula:

  5 
log D t ,  
  12  
  6    3 
log D t ,   − log D t ,  
  3    12     12   ⋅  5 − 3 
= log D t ,   +  
  12   6
−
3  12 12 
12 12
log(0.9704) − log(0.9852)  5 3 
= log(0.9852) + ⋅ − 
6 3
−  12 12 
12 12
= −0.025
 5
D t , 
 12 
   5 
= exp log D t ,   
   12   
= exp(− 0.025)
= 0.9753
©2009 Razor Risk Technologies Page 74 of 373
Razor Financial Principals

Exponential Discount Factor Interpolation

Introduction
Exponential interpolation is one of the methods to interpolate discount
factors. This section explains the methodology of exponential
interpolation. We also show that exponentially interpolating discount
factors is the same as linearly interpolating continuous compounding
rates.

Notation
ti = time point i .
ri = continuous compounding rate at ti .
dfi = discount factor at ti .

Implementation
If we have got discount factors at t1 and t2 , i.e. df1 and df 2
respectively, then we can interpolate the discount factor dfi at time
t1 < ti < t2 as:
ti t2 − ti ti ti −t1
× ×
t1 t2 −t1 t2 t2 −t1
dfi = df1 × df 2 .

Example
If the discount factor at year 1 is 0.9900 and the discount factor at year
2 is 0.9800 then the discount factor at year 1.5 can be interpolated as:
1.5 2 −1.5 1.5 1.5−1
× ×
df = 0.9900 1 2 −1
× 0.9800 2 2 −1
≈ 0.9850.

Razor supports this directly and also via linearly interpolating continuous
compounding rate.
We now show that to exponentially interpolate discount factors is the
same as linearly interpolate continuous compounding rate.
We have the formulas:
df i = e − ri ti .
1
ri = − ln dfi .
ti
ti t2 − ti ti ti −t1
× ×
t2 −t1 t2 −t1 1
We substitute dfi = df1t1 × df 2 t2 into ri = − ln dfi , then we get
ti
1
ri = − ln dfi
ti
ti t2 −ti ti ti − t1
1  × × 
= − ln  df1t1 t2 −t1 × df 2 t2 t2 −t1 
ti  


©2009 Razor Risk Technologies Page 75 of 373

Razor Financial Principals

1   t1 × t2 −t1   tti × tti −−tt1  

ti t2 −ti

= − ln  df1  + ln  df 2 2 2 1  
ti   




 

1   − r1×ti × t2 −t1   − r2 ×ti × tti −−tt1  
t2 − ti

= − ln  e  + ln  e 2 1

ti   




 

1 t −t t −t 
= −  −r1 × ti × 2 i − r2 × ti × i 1 
ti  t2 − t1 t2 − t1 
t −t t −t
= r1 × 2 i + r2 × i 1
t2 − t1 t2 − t1
rt − rt + r t − r t
= 12 1i 2i 21
t2 − t1
rt − rt + rt − rt + r t − r t
= 12 1i 11 11 2 i 21
t2 − t1
r1 ( t2 − t1 ) t2 + ( r2 − r1 )( ti − t1 )
=
t2 − t1
( r2 − r1 )
= r1 + ( ti − t1 ) .
t2 − t1

Cubic Spline Interpolation

Given a piecewise continuous function that is twice differentiable

y = f ( x ) , x ∈ we can interpolate a point x̂ by defining a set of piecewise
smooth continuous polynomials of degree 3 on intervals of f ( x ) or in
tabular form yi = f ( x0 K xn ) where an interval is defined as
[ xi , xi+1 ] i = [0, n − 1] . It is advantageous to use sets of lower order
polynomials to estimate f ( x ) to avoid oscillation errors on the edges of
f ( x ) that occur when using higher order polynomials. This is known as
Runge’s phenomenon.

A spline polynomial segment over the interval [ xi , xi +1 ] is of the form:

Si ( x ) = ai (∆x)3 + bi (∆x)2 + ci (∆x) + di for ∆x ∈ [ xi , xi +1 ] (0.1)

Therefore, the spline polynomial S ( x ) is the set of Si ( x ) segments

i = [0, n − 1] . Since there are 4 coefficients a, b, c, d and n segments, there
are 4n parameters that are needed to fully define S ( x ) . Since it is
required that the spline function be piecewise continuous, the following
condition must hold:
Si ( xi ) = yi = Si +1 ( xi +1 ) (0.2)

©2009 Razor Risk Technologies Page 76 of 373

Razor Financial Principals

In order to ensure smoothness, the first and second derivatives of the

spline segments must also be continuous:
Si′ ( xi ) = Si′+1 ( xi +1 ) , Si′′( xi ) = Si′′+1 ( xi +1 ) (0.3)

Here we will consider the “Natural Spline” with boundary conditions

S0′′ ( x0 ) = 0 and Sn′′ ( xn ) = 0 this will result in a linear interpolation off the
edges.

We start by rewriting Eq. (0.1) where we substitute Si ( x ) = yi as:

∆x = x − xi
(0.4)
yi = di since ∆x = xi − xi = 0

yi +1 = ai ∆x 3 + bi ∆x 2 + ci ∆x + d i (0.5)

Since we require the first and second derivatives to be smooth where the
spline segments join we differentiate Eq. (0.4) twice:
y ′ = 3ai ∆x 2 + 2bi ∆x + ci (0.6)

y′′ = 6ai ∆x + 2bi (0.7)

We determine the parameters ai , bi , ci , di by rewriting the equations in

terms of the of the second derivatives starting from Eq. (0.7):
Si′′ = 6ai ∆xi + 2bi
= 2bi
Si′′+1 = 6ai ∆xi + 2bi
We can now define coefficients ai , bi as:
Si′′ ∆S ′′
bi = , ai = (0.8)
2 6 ∆x
Since we now have ai , bi , di (from Eq. (0.4)) we can substitute these into
Eq. (0.5) in order to solve for ci :
∆S ′′ 3 Si′′ 2
yi +1 = ∆x + ∆x + ci ∆x + yi
6∆x 2
∆y ∆x ( 2Si′′+ Si′′+1 )
ci = −
∆x 6
Based on Eq (0.3) we must ensure the first derivative at each i th interval is
that same such that Eq. (0.6) can be rewritten with x = xi as:

y ′ = 3ai ∆x 2 + 2bi ∆x + ci = ci since ∆x = ( xi − xi )

We also need to look at the previous interval ∆x−1 = ( xi − xi −1 ) which from

Eq. (0.6) becomes:
y ′ = 3ai −1∆x−12 + 2bi −1∆x−1 + ci −1

©2009 Razor Risk Technologies Page 77 of 373

Razor Financial Principals

Setting these two equations for interval i and i − 1 equal to each other and
substituting for ai , bi , ci , di in terms of

∆y ∆x ( 2 Si′′+ Si′′+1 )
yi′ = −
∆x 6
 ∆S−′′1   Si′′−1  ∆y−1 ∆x−1 ( 2Si′′−1 + Si′′)
= 3  ∆x−1 + 2 
2
+ −
 6∆x−1   2  ∆x−1 6
This equation is now simplified as:
 ∆y ∆y−1 
∆x−1Si′′−1 + (2∆x−1 + ∆x) + ∆xSi′′+1 = 6  − 
 ∆x ∆x−1 
And finally substituting ∆x = h, S ′′ = S and writing in a vector form:

 y − y y − yi −1 
hi −1S i −1 + (2hi −1 + hi ) + hi Si +1 = 6  i +1 i − i  (0.9)
 hi hi −1 

Next apply the interval i = [1, n ] to Eq. (0.9) to generate a system of

equations in matrix form:

 y2 − y1 y1 − y0 
 − 
h1 h0
2(h +
0 0 1 1
h h )h  1 
S  
     y3 − y2 y2 − y1 
h12(h1 + h2 )h2 S −
   2  = 6 h2 h1 
 O  M   
    M 
 hn−3 2(hn−3 + hn−2 )hn−2  Sn−2   
 yn−1 − yn−2 − yn−2 − yn−3 
 hn−2 hn−3 
(0.10)
We do not solve for either S 0 or S n−1 since these end-point second
derivatives are bound at 0 (natural spline). Since this is a symmetric,
tridiagonal matrix we use a sparse matrix algorithm for reduction and
back-substitution to solve for Si . From Eq. (0.8) the values for ai , bi , ci , di
are:
Si +1 − Si S y − y 2h ( S + S i +1 )
ai = , bi = i , ci = i +1 i − i i , di = yi (0.11)
6hi 2 hi 6

Example (Interpolate point .4 years)

N=5 Time X (in years) Discount Factor Y
0 0 1
1 .25 0.99635611
2 .5 0.99111989
3 .75 0.98683149
4 1 0.970428

©2009 Razor Risk Technologies Page 78 of 373

Razor Financial Principals

 1 .25 0   S1  -0.0382 
.25 1 .25 S  =  0.0227 
  2  
 0 .25 1   S 3  -0.2908 
A−1b = S
 1.0714 -0.2857 0.0714  -0.0382  -0.0682
-0.2857 1.1429 -0.2857   0.0227  =  0.1200 
    
 0.0714 -0.2857 1.0714  -0.2908  -0.3208
 0 
-0.0682 
 
S =  0.1200 
 
-0.3208
 0 

From equations (0.11) we derive the coefficients for the cubic spline
polynomial

-0.0455  0   0.0033   1 
 0.1255   -0.0341 -0.0078   
a= , b =  , c =   , d = 0.9964 
-0.2938  0.0600   0.0139   0.9911
       
 0.2138  -0.1604   0.0158   0.9868 

Since the point we are interpolating xˆ = .4 is on the second interval the

spline polynomial from Eq. (0.1):

h = xˆ − x1
yˆ = a1h3 + b1h 2 + c1h + d
= 0.9948

Extrapolation Methods
Suppose we have N number of rates in our data sample.
Define:
t (k ) = the time of the k th rate in our sample such that
t (1) < t ( 2 ) < t (3 ) < ... < t ( N )
r ( k ) = the rate at time t (k )
We have demonstrated in the previous section that we can find any rate r j at
( )
time t j for any t j ∈ t (1) , t ( N ) . For example, for some i such that
(i ) (i +1)
t < tj < t , we can use the linear interpolation formula

©2009 Razor Risk Technologies Page 79 of 373

Razor Financial Principals

r (i +1) − r (i )
r j = r (i ) + (
⋅ t j − t (i ) )
t (i +1) − t (i )
to find r j .

However, sometimes we need to find r j for t j < t (1) or t j > t ( N ) and it is

particularly true at the beginning or the end of the yield curve where we do
not have two rates to interpolate the rate in between. Under this situation, we
will need extrapolation techniques to help find the rates. Again as in
interpolation, r j here has a very general definition, it can be zero rates,
discount factors, etc.

Last Zero
This is a very simple and straightforward extrapolation method to extrapolate
zero rates Z t →t j for some t j < t (1) or t j > t ( N ) . This method is saying that we
use the boundary zero rate as an approximation for any zero rate beyond the
boundary rate. Thus, using last zero method, Z t →t j = r (1) for any t j < t (1) and
Z t →t j = r ( N ) for any t j > t ( N ) .
Example:
The 3-month zero rate Z 3 is 6.01% and the 40-year zero rate Z t → 40 is 5.75%.
t→
12
Calculate the 1-month zero rate and 41-year zero rate using last zero method.
Answer:

r (1) = Z 3 = 6.01%
t→
12

3
t (1) =
12
r ( N ) = Z t → 40 = 5.75%

t ( N ) = 40
1
Since t 1 = < t (1) , using last zero method, we then obtain:
12
12

Z 1
t→
12

= r (1)
= 6.01%
Also t 41 = 41 > t ( N ) , using last zero method, we then obtain:

Z t →41

= r (N )
= 5.75%
©2009 Razor Risk Technologies Page 80 of 373
Razor Financial Principals

Linear Zero

It is another method to extrapolate the zero rates Z t →t j for some t j < t (1) and
t j > t ( N ) . This method works as follows:
(1)
For t j < t , we assume Z t →t j lies on the straight line produced by the two
( ) ( )
points t (1) , r (1) and t ( 2 ) , r ( 2 ) . Mathematically, Z t →t j can be written as:

r (2 ) − r (1)
Z t →t j = r (1) +
t (2 ) − t (1)
(
⋅ t j − t (1) )
For t j > t ( N ) , we assume Z t →t j lies on the straight line produced by the two
( ) ( )
points t ( N −1) , r ( N −1) and t ( N ) , r ( N ) . Mathematically, Z t →t j can be written as:

r ( N ) − r ( N −1)
Z t →t j = r ( N −1) +
t ( N ) − t ( N −1)
(
⋅ t j − t ( N −1) )
Example:
Given the 3-month zero rate is 6.01% and 6-month zero rate is 6.11%. Calculate
the 1-month zero rate using linear zero.
Answer:
r (1) = 0.0601
3
t (1) =
12
r (2 ) = 0.0611
6
t (2 ) =
12
1
Since t j = < t (1) , we use linear zero formula to extrapolate Z 1 :
12 t→
12

Z 1
t→
12

0.0611 − 0.0601  1 3 
= 0.0601 + ⋅ − 
6 3
−  12 12 
12 12
= 5.94%
Example:
Given the 39.5-year zero rate is 5.75% and 40-year zero rate is 5.35%. Calculate
the 41-year zero rate using linear zero.
Answer:
r ( N −1) = 0.0575

©2009 Razor Risk Technologies Page 81 of 373

Razor Financial Principals

t ( N −1) = 39.5
r ( N ) = 0.0535
t ( N ) = 40
Since t j = 41 > t ( N ) , we use linear zero formula to extrapolate Z t →41 :

Z t →41
0.0535 − 0.0575
= 0.0575 + ⋅ (41 − 39.5)
40 − 39.5
= 4.55%

Linear Discount Factor

This is exactly the same as linear zero except that we are extrapolating
discount factors instead of zero rates. If we are given the discount rate
r (1) , r (2 ) , r ( N −1) and r ( N )

For t j < t (1)

r ( 2 ) − r (1)
D (t , t j ) = r (1) + (2 )
t −t
(
(1) ⋅ t j − t
(1)
)
and

for t j > t ( N )

r ( N ) − r ( N −1)
D (t , t j ) = r ( N −1) +
t ( N ) − t ( N −1)
(
⋅ t j − t ( N −1) )
Example:
Given the 3-month discount factor is 0.9850 and the 6-month discount factor is
0.9698, calculate the 1-month discount factor using linear discount factor
method.
Answer:
r (1) = 0.9850
3
t (1) =
12
r (2 ) = 0.9698
6
t (2 ) =
12
1
Since t j = < t (1) , we use linear discount factor formula to extrapolate
12
 1
D t ,  :
 12 
 1
D t , 
 12 
©2009 Razor Risk Technologies Page 82 of 373
Razor Financial Principals

0.9698 − 0.9850  1 3
= 0.9850 + ⋅ − 
6 3
−  12 12 
12 12
= 0.9951
Example:
Given the 39.5-year discount factor is 0.1062 and the 40-year discount factor is
0.1032, calculate the 41-year discount factor using linear discount factor
method.
Answer:
r ( N −1) = 0.1062
t ( N −1) = 39.5
r ( N ) = 0.1032
t ( N ) = 40
Since t j = 41 > t ( N ) , we use linear discount factor formula to extrapolate
D(t ,41) :
D(t ,41)
0.1032 − 0.1062
= 0.1062 + ⋅ (41 − 39.5)
40 − 39.5
= 0.09720

Linear Zero-Zero
Linear zero-zero is an extrapolation method to extrapolate zero rate for some
t j < t (1) . This method assumes the zero rate r0 at time 0 ( t 0 )is 0 and all the
zero rates Z t ←t j for some time t 0 < t j < t (1) will lie on the straight line of the
( )
two points (t 0 , r0 ) and t (1) , r (1) . Note that (t 0 , r0 ) = (0,0 ) by assumption.

This allows us to calculate the zero rates Z t →t j using linear zero-zero

extrapolation formula:

r (1)
Z t →t j = ×tj
t (1)
Example:
The 3-month zero rate is 6.01%. Find the 1-month zero rate using linear zero-
zero method.
Answer:
r (1) = 0.0601
3
t (1) =
12

©2009 Razor Risk Technologies Page 83 of 373

Razor Financial Principals

1
Since t 1 = < t (1) , we need to extrapolate the data to find Z 1 . Using
12 t→
12 12
linear-zero-zero formula:
Z 1
t→
12

0.0601 1
= ×
3 12
12
= 2.00%

Linear Discount Factor-One (Currently Under Development)

6.3.4 Zero Treatment

Zero treatment can be specified at the curve and/or asset level. At the curve
level the zero mode is an input to the interpolation and extrapolation in the
curve building process.

At the asset level the cash deposit allows rate inputs to be specified as zeros.
The modes available are:
Simple Zero
Compounded Zero
Continously-Compounded Zero
Cash Deposit
These are similar to the input zero methods at the curve level which are
explained below. Hence a set of cash deposit assets with a zero method of
continuously-compounded selected, is in affect a zero curve with no
bootstrapping required.

When the interpolation and extrapolation method selected requires the

computation of the zero rate (i.e. linear zero, log linear zero, last zero or
linear zero zero), Razor users can further specify the definition of the zero rate
in a number of ways. We currently support the following modes: simple,
compounding, continuously-compounding, and cash-deposit. Default setting is
cash-deposit if user specification is omitted.

Using similar denotations as above, that is,

d t →t1 = t1 − t = number of days from t to t1
D = number of days in a year
P(t ,t1 ) = the price of a zero-coupon bond at time t maturing at time t1
F = the face value of the zero-coupon bond,
D(t , t1 ) = the discount factor at time t1
©2009 Razor Risk Technologies Page 84 of 373
Razor Financial Principals

we provide the following different definitions of the zero rate:

Simple Zero

Let us denote the annualised simple zero by Z tS→t . The relationship between
1

the future and present values is given by

 
 
P(t , t1 ) = F   = F ⋅ D (t , t1 )
1
 d t →t1 
 1 + Z t→t1 ⋅
S

 D 
and the conversions between D(t ,t1 ) and Z t →t1 can be stated as:
−1
 d t →t1   1  D
D(t , t1 ) = 1 + Z tS→t ⋅  and Z tS→t =  − 1
 D   D(t , t1 )  d t →t1
1 1

Compounded Zero
For compounded zero, users can specify different compounding frequencies
(number of times compounding is performed in a year). For example, annual
compounding is represented by setting the frequency to “1” (once a year),
semi-annual to “2”, quarter to “4”, etc. We provide the following derivations
for different compounding frequency cases.

Annual Compounding

Denoting the annualised compounded zero rate by Z tC→t , The relationship

1

between the present and future values is given by

( )
d t →t1
P(t , t1 ) = F 1 + Z tC→t1 = F ⋅ D(t , t1 )
−
D

and the conversions between the discount factor and the zero rate are

( )
dt →t
− 1

D(t , t1 ) = 1 + Z t→t and Z tC→t = D(t , t1 )

D
D
−
C
d t →t1 −1
1 1

Semi-Annual Compounding
The relationship between the present and future values is given by
2 d t →t1
−
 ZC  D
P(t , t1 ) = F 1 + t→t1  = F ⋅ D (t , t1 )
 2 
 
and the conversions between the discount factor and the zero rate are

©2009 Razor Risk Technologies Page 85 of 373

Razor Financial Principals

2 d t →t
− 1
D

 Z tC→t  
D(t , t1 ) = 1 + and Z tC→t = 2 D(t , t1 ) − 1
D
1  −
2 d t →t1
 2  1
 
 

The Generalised Case

Generalising from above, the relationship between the present and future
values for compounding frequency K is given by
Kd t →t1
−
 ZC  D
P(t , t1 ) = F 1 + t→t1  = F ⋅ D (t , t1 )
 K 
 
and the conversions between the discount factor and the zero rate are
Kd t →t
− 1
D

 Z C
 
D(t , t1 ) = 1 + t→t1 and Z tC→t = K  D(t , t1 ) − 1
D
 −
Kd t →t1
 K  1
 
 

Continuously-Compounded Zero
For continuous-compounding, we have the following relationships

 d t →t1 
P(t , t1 ) = F exp − Z te→t1 ⋅  = F ⋅ D(t , t1 )
 D 
 d t →t1 
D(t , t1 ) = exp − Z te→t1 ⋅ log(D(t , t1 ))
D
 and Z te→t = −
 D 
1
d t →t1

where Z te→t represents the continuously-compounded zero rate.

1

Cash-Deposit
d t →t1 d t →t1
When ≤ 1 , it is treated as a simple zero, and when > 1 , it is
D D
treated as an annually-compounded zero.

6.3.5 Summary of Bootstrapping Formulas

Number Description Formula

1 Zero rate from price  F  D

of non-coupon bearing Z t → t1 =  − 1  ⋅
 P (t , t 1 )  d t → t1
instrument

©2009 Razor Risk Technologies Page 86 of 373

Razor Financial Principals

2 Discount factor from

D(t , t1 ) =
1
zero rate d t →t1
1 + Z t →t1
D

3 Zero rate from  1  D

discount factor Z t →t1 =  − 1 ⋅
 D (t , t 1 )  d t →t1

4 Discount factor from −1

 d t →t 
forward rate and D(t , t 2 ) = D(t , t1 )1 + f (t , t1 , t 2 ) ⋅ 1 2 

previous  D 
corresponding
discount factor

5 Forward rate from  D(t , t1 )  D

discount factors f (t , t1 , t 2 ) =  − 1 ⋅
 D(t , t 2 )  d t1 →t2

6 Discount factor from rs (c, N ) cN −1  i 

swap rate and 1− ∑ D t , 
 c
D(t , N ) =
previous discount c i =1

factors rs (c, N )
1+
c

7 Effective swap rate c(1 − D(t , N ))

from discount factors rs (c, N ) = cN
 i
∑i =1
D t , 
 c

8 Linear Interpolation r j − ri
rα = ri + (tα − ti )
t j − ti

6.3.6 Demonstration of Bootstrapping

Sample Curve

Contracts Available

Contract Number Contract Type Contract Maturity

1 Cash Deposit 1-month

2 Cash Deposit 3-month

3 Bill Futures starting in 3 3-month

months

4 Bill Futures starting in 6 3-month

months

©2009 Razor Risk Technologies Page 87 of 373

Razor Financial Principals

5 Swap with quarterly 1-year

payments

6 Swap with quarterly 3-year

payments

7 Swap with semi-annual 5-year

payments

8 Swap with semi-annual 10-year

payments

Bootstrapping Method

Time Contract Used Method to obtain zero rates

1-month Contract (1) Formula (1)

3-month Contract (2) Formula (1)

6-month Contract (3) Formula (4) then (3)

9-month Contract (4) Formula (4) then (3)

1-year Contract (5) Formula (6) then (3)

1.25-year Contract (6) Linear interpolate the 1-

year swap rate and the 3-
year swap rate with formula
(8) to find the 1.25-year
swap rate. Use formula (6)
to bootstrap the 1.25-year
discount factor and convert
to 1.25-year zero rate by
formula (3).

3-year Contract (6) Formula (6) then (3)

3.5-year Contract (7) Use formula (7) to find the

3-year effective swap rate
with semi-annual payments.
Linear interpolate the 5-
year swap rate and 3-year
effective swap rate with
formula (8) to find the 3.5-
year swap rate. Use formula
(6) to bootstrap the 3.5-year
discount factor and convert
to 3.5-year zero rate by
formula (3).

©2009 Razor Risk Technologies Page 88 of 373

Razor Financial Principals

5-year Contract (7) Formula (6) then (3)

5.5-year Contract (8) Linear Interpolate the 5-

year swap rate and the 10-
year swap rate with formula
(8) to find 5.5-year swap
rate. Use formula (6) to
bootstrap the 5.5-year
discount factor and convert
to 5.5-year zero rate by
formula (3).

10-year Contract (8) Formula (6) then (3)

Up to the moment, we assume each coupon payment period is equal. In reality,

payment periods may not be equal for each period i (due to day adjustments
and day count convention). Razor also caters for this case. A new formula is
used for unequal coupon payment period. For equal coupon payment periods,
the formula is:
rs (c, N ) cN −1  i 
1− ∑ D t , 
 c
D(t , N ) =
c i =1
rs (c, N )
1+
c
where c is the number of payments per year, and
rs (c, N ) is the annualised swap rate.
To cater for unequal coupon payment period, the formula becomes:

 N −1 
D(t , N ) = 1 − ∑ Couponi ⋅ D(t , i )  / (1 + Coupon N )
 i =1 
where Couponi = rs ⋅ DCFi , and
DCFi is the year count fraction for period i based on day count
convention of the swap.

Similarly, instead of defining effective swap rate (implied by discount factors)

using same assumption in the following formula:
c(1 − D(t , N ))
rs (c, N ) = cN
 i
∑i =1
D t , 
 c
the year count fraction for period i should also be used:
1 − D (t , N )
Coupon N = N .
∑ D(t , i ) ⋅ DCFi
i =1

In addition, when linearly interpolating between two swap rates, the year count
fractions (instead of number of calendar days) should be used:

©2009 Razor Risk Technologies Page 89 of 373

Razor Financial Principals

r j − ri
rα = ri + (DCFα − DCFi )
DCF j − DCFi
where DCF i is the year count fraction from t 0 to t i

6.4 CDS Spread Curves

For details of the support for CDS spread curves, please refer to section 16.1.5
.

6.5 FX Market Data

Razor supports a term structure of FX spot and FX forward rates.

6.6 Equity Market Data

Razor supports Equity/Stock spot prices and Equity Index spot, additionally
with a term structure of either continuously compounding forecast dividend
yield, or discrete dividend payments.

6.7 Commodity Forward Curves

Razor supports Commodity Forward curves. So for example ('NGO Spot', ‘NGO
1M', ‘NGO 3M', ‘NGO 6M', ‘NGO 1Y', ‘NGO 2Y', ‘NGO 3Y’)

6.8 Volatility Market Data

The basic building block for implied volatilities in Razor is the termstructure of
implied volatilities for different option expiry dates. A volatility value for a
specified expiry date can be obtained from this curve. If the requested date
does not exist as a volatility data point on this term structure, then the
volatility is derived.
If the expiry date for the required volatility is between two existing expiry
dates on the volatility curve then the required volatility is derived by linear
interpolation. Otherwise if the expiry date is before the start date, or after the
end date then straight line extrapolation is used.
A set of these term structures will represent the volatility surface for a specific
asset.

6.8.1 Interest Rate Volatility

For interest rate volatilities, Razor supports an implied volatility surface, by
option expiry date, and underlying term to maturity. In other words the
volatility surface is represented as a set of volatility term structures, one for
each underlying term to maturity.
The volatilities are implied from the market option prices. Currently to support
volatilities for different strikes or deltas (degree of moneyness) requires the
user to define a volatility surface for each strike or delta, and then a specific
surface with the closest strike or delta will be associated with the trade.
Hence currently there is no interpolation across the strike dimension.
©2009 Razor Risk Technologies Page 90 of 373
Razor Financial Principals

If the underlying term to maturity for which a volatility is required does not
match a known one on the surface, then linear interpolation is employed
between volatilities found from term structures for known underlying terms to
maturity, which span the underlying term to maturity of the option for which
an implied volatility is required. Otherwise straight line extrapolation is used
beyond either edge of the surface.
The interest rate volatility surface as described above is currently
implemented and used by the swaption and bond option pricing routines.
All other interest rate option products (including caps and floors) currently use
the volatility surface defined by the option expiry date and delta, as described
below. Currently only the volatility for the ATM delta is used.

6.8.2 Equity Volatility

For equity volatilities, Razor supports a term structure of implied equity
volatilities by option expiry date, and strike. Alternatively the volatility data
may be provided by option expiry date and delta. In other words the volatility
surface is represented as a set of volatility term structures, one for each strike
or delta.
If the strike or delta does not match, then linear interpolation is used between
volatilities for known strikes or deltas which span the strike or delta for which
an implied volatility is required. Otherwise straight line extrapolation is used
beyond either edge of the surface.
Note that even though volatilities for different deltas may be provided,
currently for this method only volatility for the ‘at-the-money’ delta is used by
the pricing / revaluation.
At present, local volatility modelling is not currently supported in Razor.
Currently only the vanilla equity option and vanilla index option (European and
American) use volatilities for specific expiry and strike, on the basis that
volatility data is provided in this form. Otherwise only a single ATM volatility
can be used as provided specifically for that trade or product by configuration.

6.8.3 FX Volatility
Treatment of FX volatilities is as above for Equity volatility.
Parametric stochastic volatility (Heston), construction from risk reversal and
broker strangle spreads for 10% and 25% deltas and ATM volatility are not
currently supported.
Currently within each of the fx option models, only a single ATM volatility can
be used as provided specific for that trade or product by configuration.

6.8.4 Commodity Volatility

Treatment of Commodity volatilities is as above for Equity volatility.
Currently only the vanilla commodity option uses volatilities for specific expiry
and strike, if volatility data is provided in this form. Otherwise only a single
ATM volatility can be used as provided specific for that trade or product by
configuration.

©2009 Razor Risk Technologies Page 91 of 373

Razor Financial Principals

6.9 Calendars
There are various named calendars in RAZOR. Calendars contain date events,
which define events such as holidays. There should be a calendar defined for
every currency code, as these are used in the date roll calculations.

©2009 Razor Risk Technologies Page 92 of 373

Chapter 7
RAZOR’S Product Support

7.1 Design Objectives in RAZOR ’S Product Support

RAZOR’s product support is designed with a number of goals in mind:
Completeness of representation: Financial products differ greatly in
characteristics even within the same product class. Any representation of
the product should strive to embody as much information about the
characteristics of the financial product as possible.
Flexibility: Innovation is rife within the financial industry. RAZOR’s product
support must be flexible enough to handle new products as they come
along.
Accessibility: RAZOR’s representation of the product should, as much as is
possible, be based around an open standard, to make collaboration with
other systems easier and to minimise vendor-specific risk. The pricing
strategy of the products should be able to incorporate user models easily.

7.2 Product Representation

FPML (Financial products Markup Language) is the industry-standard protocol
for complex financial products. It is based on XML (Extensible Markup
Language), the standard meta-language for describing data shared between
applications. All categories of privately negotiated derivatives will eventually
be incorporated into the standard. Version 1.0 of FPML covers interest rate
swaps and Forward Rate Agreements (FRA’s). Version 2.0 extends the interest
rate product coverage to the most common option products, including caps,
floors, swaptions, and cancellable and extendible swaps. Version 3.0 covers
different asset classes, this version includes the interest rate work of version
2.0 and additionally covers FX and Equity Derivatives.
The standard, which is freely licensed, is intended to automate the flow of
information across the entire derivatives partner and client network,
independent of the underlying software or hardware infrastructure supporting
the activities related to these transactions.
RAZOR uses FPML version 3.0 Markup when that product is covered by the
standard. Further information on FPML can be found on the FPML website.

7.2.1 The FinMark Schema

FINMARK is Razor’s representation of an organisation’s deals. It encapsulates
FPML, but adds additional elements in order to better support the risk
management process.
Razor Financial Principals

Schema Diagram

FinMark Schema
Name: Type Occurs Size Description
An identifier that allows us to run
scheme
1..1 20 multiple trade sets in the one
string
database.
deal
0..n The deal
Deal
ccyPair
0..n
CurrencyPair
party
0..n
Party
whatIfDealScenarios
0..1
WhatIfDealScenarios

Where a deal is represented as:

Deal Schema
Name: Type Occurs Size Description
sysId
1..1
int
dealHeaderDealId
1..1 20 The unique identifier for the deal
string
This is used to identify the type of
dealHeaderDealType
1..1 20 structure for structured deals – e.g.
string
CONDOR
dealHeaderDealDate
1..1 The date the deal was traded
date
dealHeaderStatus
1..1 20 The current status of the deal
string
trade
0..n
Trade

And the Trade is:

Trade Schema
Name: Type Occurs Size Description
sysId
1..1
int
tradeHeaderTradeId This identifier uniquely identifies the
1..1 30
string trade
tradeHeaderTradeDate The date that this particular trade
1..1
date was done

©2009 Razor Risk Technologies Page 95 of 373

Razor Financial Principals

Trade Schema
Name: Type Occurs Size Description
This code identifies the trade type
tradeHeaderTradeType
1..1 20 and is used to map to the
string
appropriate pricing model
This identifier provides the
tradeHeaderCreditLine
0..1 20 capability to link trades with specific
string
lines of credit
tradeHeaderDealer Identifies the dealer who did the
0..1 30
string trade
tradeHeaderCounterparty Identifies the counterparty of the
0..1 20
string trade
tradeHeaderInternalUnit Identifies the internal unit who is
1..1 20
string responsible for the trade
tradeHeaderBuySell Indicates whether the internal unit is
1..1 4
string long or short the trade
tradeHeaderStatus Indicates the current status of this
1..1 20
string particular trade
guarantees Guarantees provide information
0..1
Guarantees about any guarantors in the deal
collateralAgreementId ID of the collateral agreement if the
0..1 20
string trade has been lodged against
Defines a schedule for allowing the
rightToBreakSchedule
0..1 bank to break the contract
RightToBreakSchedule
extensions
extensions Extensions allow clients to extend
0..1
TradeExtensions the trade with additional data
A product that can representing
productGenericInterestRate
1..1 generic interest rate payments or
GenericInterestRate
cashflows
productCashPayment A product that can represent single
1..1
CashPayment cash payments
productCreditLine A product that represents a line of
1..1
CreditLine credit
productLetterOfCredit A product that represents a letter of
1..1
LetterOfCredit credit
productFxleg
1..1 A product that represents an FX deal
fpmlFXLeg
productFxSimpleOption A product that represents an FX
1..1
fpmlFXOptionLeg Vanilla Option deal
productFxBarrierOption A product that represents an FX
1..1
fpmlFXBarrierOption Barrier Option deal
productFxDigitalOption A product that represents an FX
1..1
fpmlFXDigitalOption Digital Option deal
©2009 Razor Risk Technologies Page 96 of 373
Razor Financial Principals

Trade Schema
Name: Type Occurs Size Description
productFxAverageRateOption A product that represents an FX
1..1
fpmlFXAverageRateOption Average Rate deal
A product that represents an FX
productFxOption
1..1 vanilla option or a vanilla option
FXOption
with single or double barriers
productSwap A product that represents an interest
1..1
fpmlSwap rate swap
productFra A product that represents an interest
1..1
fpmlFRA rate FRA
productEquityOption A product that represents an equity
1..1
fpmlEquityOption option
productCapFloor A product that represents an interest
1..1
fpmlCapFloor rate cap or floor
productSwaption A product that represents an interest
1..1
fpmlSwaption rate
productBondOption A product that represents a bond
1..1
BondOption option
productEquity product that represents a held or
1..1
Equity forward equity
productCreditDefaultSwap A product that represents an credit
1..1
fpmlCreditDefaultSwap default swap
productBond
1..1 A product that represents a bond
fpmlBond
otherPartyPayment 0..n
Any additional payments made to
OtherPartyPayment the counterparty

7.3 Adding New Products

In the back-end products are mapped to pricing models via mapping rules
specified in XML. Products are also mapped to input screens on the front- end
using mapping rules.
Adding pricing support for new products into RAZOR involves adding a DLL into
the Master/bin directory and also the Slave/bin directory on the master
computer and all its slave computers. This DLL prices the product and also
contains functions to extract information from the product that the RAZOR
engine needs in order to incorporate the product in the risk management
process.
The XML representation of the product can be incorporated in one of two ways:
The easiest way to represent a new product is to choose a product schema
that’s similar to the one being represented. Any information not
contained in the existing schema can be added using the product
extension schema.

©2009 Razor Risk Technologies Page 97 of 373

Razor Financial Principals

The other way to add a new product representation to RAZOR is to create an

XML Schema for the product, and then to generate the necessary support
files to link into RAZOR using the metatype framework.

7.4 Pricing Contexts

When evaluating the impact a financial product has on credit risk, there may
be several different exposures depending on the context you are looking at the
product at. As an example; a bond can have two exposures - the exposure to
the counterparty who issued the bond, and an exposure to the seller of the
bond. When determining the exposure of the bond the particular pricing
context must be taken into account.

7.5 Transitioning
7.5.1 Market Transitioning
The pricing adapter for a trade may need to transition certain rates between
different node dates. An example for when this needs to occur is on swaps. If a
rate set occurs between two node dates, the swap pricing model linear
interpolates between the rate that occurred on the last node date and the rate
that occurs on the current node date to determine the implied rate occurring
at the rate set date.
Obviously, this means that the pricing model has to have the information
available for the relevant rates on previous node dates. The pricing adapter
achieves this through the use of the market transition cache parameter on the
pricing request structure. The pricing adapter is responsible for allocating an
area of memory that it can use as a cache, in order to keep any historical
market information around that it needs to fulfil pricing requests.

Historical Data processing between time nodes

Any processing on the current time node which requires historical data for
market data events which have occurred in the past on dates between the
current time node and the previous is derived by linear interpolation between
the cached observed values saved from the last time node and current values
at the current time node.
The processing at the current time node will be a function of cached historical
market data, linearly interpolated missing historical market data, and the
current market data.
So for example for an average rate option, the market transition cache will
contain the set of rate fixings which have already occurred, or a current
average representative of these fixings. At each time node this cache will be
updated if there is a fixing on the current time node, and also with any fixings
that should have occurred since the last time node. These fixings will be
derived by linear interpolation between current observations and those
observations from the last time node.
This processing is specific and customized for each product supported so to
fully support the transitioning of the product forward through time along a
path of the simulation process. This also supports the situation during
simulation where the expiry date falls between two time nodes, and hence
there is the need to determine whether the option was exercised at expiry.

©2009 Razor Risk Technologies Page 98 of 373

Razor Financial Principals

Rate Resets between time nodes

The process for supporting rate resets as transitioning through time along a
path in the simulation process is as follows. The full set of floating flow
cashflows are initialized from the current market on the valuation date and
then cached, so that it may be updated during the transition along a path. For
each forward time node transitioning along a path, the full set of floating
cashflows are reprocessed and updated where needed, to effect the
transitioning along the path.
For each floating flow cashflow ci we have the fixing date t f i , the start date t si ,
and the end date t ei .

The initial set of floating flow cashflows.

The floating flow cashflows are initialized from today’s market (ie time 0, t 0 )
and as has been already been fixed with the trade. All previous fixings and
possibly also todays should already be set and marked as fixed.
a. If today is a fixing date and the floating rate is not already set then it is
observed and set from today’s market, and marked as fixed. It is set to the
forward simple rate from the start date t si to the end date t ei .

b. All other fixings which occur in the future will be forward forecast from
today’s market, with each flows observation date set to today’s date. They are
not marked as fixed. The rate is set to the forward simple rate from the start
date t si to the end date t ei .

c. For any previous fixings which have not been observed as yet, then it is set
to an observation from today’s market, and marked as fixed. It is set to the
( )
forward simple rate from the start date t 0 + t si − t f i to the end date
( )
t 0 + t ei − t f i . This represents the best estimate that can be made from today
as to what the fixing should have been.
The set of floating flow cashflows at a forward time node
The market scenario is simulated and built with respect to the time node date
t j . All floating rate flows which are already fixed are skipped. Those floating
flow rates which are not already fixed are processed as outlined in the cases
above (except t 0 is replaced with t j ), but with one additional case to be
handled.
For any previous fixings which have been previously observed as forecast in the
future, but for which the fixing date now lays between last time node and the
current time node, then the rate will be calculated as follows. It is the rate
which is linearly interpolated between the last observed rate, and the forecast
observed rate from the current time node – ie the forward simple rate from the
( ) ( )
start date t j + t si − t f i to the end date t j + t ei − t f i . The observed date is set
to the fixing date t f i , and the cashflow is market as fixed.

7.5.2 Trade Transitioning

Path-dependent trades may need to transition into different kinds of products
if, for example, barrier levels are breached. The pricing adapter may make use
of a trade transition cache in order to store the state of trades during a path.
©2009 Razor Risk Technologies Page 99 of 373
Chapter 8
Common XML Structures

fpmlInterestRateStream Schema
Name: Type Occurs Size Description
payerPartyReference
1..1 The identifier of the paying party
fpmlPartyReference
receiverPartyReference
1..1 The identifier of the receiving party
fpmlPartyReference
calculationPeriodDates Indicates the schedule that the
1..1
fpmlCalculationPeriodDates floating rate calculations occurs
paymentDates Indicates the schedule that the date
1..1
fpmlPaymentDates payments occur
resetDates This structure indicates when
1..1
fpmlResetDates floating rate resets occur
calculationPeriodAmount This structure indicates how the
1..1
fpmlCalculationPeriodAmount amounts to be paid is determined
This structure indicates how the
stubCalculationPeriodAmount
0..1 stub period amounts to be paid are
fpmlStubCalculationPeriodAmount
determined
principalExchanges Determines if and when the
0..1
fpmlPrincipalExchanges exchange
cashflows This type gives us the fixed cash
0..1
fpmlCashflows flows represented by the product

The fpmlCalculationPeriodDates structure allows us to generate the schedule

for working out when floating rate calculations are made.

fpmlCalculationPeriodDates Schema
Name: Type Occurs Size Description
effectiveDate
1..1
fpmlUnadjustedDate
terminationDate
1..1
fpmlUnadjustedDate
calculationPeriodDatesAdjustments
1..1
fpmlDateAdjustments
calculationPeriodFrequency
0..1
fpmlCalculationPeriodFrequency
id
1..1 50
string

The fpmlPaymentDates structure allows us to generate the dates payments

occur on.
Razor Financial Principals

fpmlPaymentDates Schema
Name: Type Occurs Size Description
calculationPeriodDatesReference
0..1
fpmlRef
paymentFrequency
1..1
fpmlTenor
payRelativeTo
0..1
fpmlDateRelativeTo
paymentDatesAdjustments
0..1
fpmlDateAdjustments

The fpmlResetDates indicate when floating rate resets are done.

fpmlResetDates Schema
Name: Type Occurs Size Description
calculationPeriodDatesReference
1..1
fpmlRef
resetRelativeTo
1..1
fpmlDateRelativeTo
fixingDates
1..1
fpmlRelativeDateOffset
resetFrequency
1..1
fpmlTenor
resetDatesAdjustments
1..1
fpmlDateAdjustments

The fpmlCalculationPeriodAmount element determines how the amounts to be

paid are calculated.

fpmlCalculationPeriodAmount Schema
Name: Type Occurs Size Description
calculation
1..1
fpmlCalculation
knownAmountSchedule
1..1
fpmlAmountSchedule

The fpmlStubCalculationPeriodAmount structure determines how the amounts

paid on stub periods are calculated.

fpmlStubCalculationPeriodAmount Schema
Name: Type Occurs Size Description
calculationPeriodDatesReference
1..1
fpmlRef
initialStub
0..1
fpmlStub

©2009 Razor Risk Technologies Page 101 of 373

Razor Financial Principals

fpmlStubCalculationPeriodAmount Schema
Name: Type Occurs Size Description
finalStub
0..1
fpmlStub

The fpmlCashflows type gives us the fixed cash flows represented by the
product.

fpmlCashflows Schema
Name: Type Occurs Size Description
Indicates whether the cashflows could
cashflowsMatchParameters be regenerated from the parametric
0..1
boolean information without any loss of
information
principalExchange The initial, intermediate and final
0..n
fpmlPrincipalExchange principal exchange amounts
The adjusted payment date and
paymentCalculationPeriod associated calculation period
0..n
fpmlPaymentCalculationPeriod parameters required to calculate the
payment amount
The fpmlPrincipalExchanges structure determines if and when the exchange of
principal is done. Note that any exchange of principal could have settlement
risk implications.

Generic Interest Rate Example

<genericInterestRate>
<interestRateStream>
<payerPartyReference href="#ITE"/>
<receiverPartyReference href="#BAN"/>
<cashflows>
<paymentCalculationPeriod>
<adjustedPaymentDate>2002-09-30</adjustedPaymentDate>
<fixedPaymentAmount>
<currency>AUD</currency>
<amount>1000000.000000</amount>
</fixedPaymentAmount>
</paymentCalculationPeriod>
</cashflows>
</interestRateStream>
<interestRateStream>
<payerPartyReference href="#BAN"/>
<receiverPartyReference href="#ITE"/>
<cashflows>
<paymentCalculationPeriod>
<adjustedPaymentDate>2002-09-30</adjustedPaymentDate>
<calculationPeriod>
<adjustedStartDate>2002-09-30</adjustedStartDate>
<adjustedEndDate>2002-12-30</adjustedEndDate>
<notionalAmount>

©2009 Razor Risk Technologies Page 102 of 373

Razor Financial Principals

<currency>AUD</currency>
<amount>1000000.000000</amount>
</notionalAmount>
<floatingRateDefinition/>
</calculationPeriod>
</paymentCalculationPeriod>
<paymentCalculationPeriod>
<adjustedPaymentDate>2002-12-30</adjustedPaymentDate>
<calculationPeriod>
<adjustedStartDate>2002-12-30</adjustedStartDate>
<adjustedEndDate>2003-03-30</adjustedEndDate>
<notionalAmount>
<currency>AUD</currency>
<amount>1000000.000000</amount>
</notionalAmount>
<floatingRateDefinition/>
</calculationPeriod>
</paymentCalculationPeriod>
<paymentCalculationPeriod>
<adjustedPaymentDate>2003-03-30</adjustedPaymentDate>
<calculationPeriod>
<adjustedStartDate>2003-03-30</adjustedStartDate>
<adjustedEndDate>2003-06-30</adjustedEndDate>
<notionalAmount>
<currency>AUD</currency>
<amount>1000000.000000</amount>
</notionalAmount>
<floatingRateDefinition/>
</calculationPeriod>
</paymentCalculationPeriod>
<paymentCalculationPeriod>
<adjustedPaymentDate>2003-06-30</adjustedPaymentDate>
<calculationPeriod>
<adjustedStartDate>2003-06-30</adjustedStartDate>
<adjustedEndDate>2003-09-30</adjustedEndDate>
<notionalAmount>
<currency>AUD</currency>
<amount>1000000.000000</amount>
</notionalAmount>
<floatingRateDefinition/>
</calculationPeriod>
</paymentCalculationPeriod>
<paymentCalculationPeriod>
<adjustedPaymentDate>2003-09-30</adjustedPaymentDate>
<fixedPaymentAmount>
<currency>AUD</currency>
<amount>1000000.000000</amount>
</fixedPaymentAmount>
</paymentCalculationPeriod>
</cashflows>
</interestRateStream>
</genericInterestRate>

©2009 Razor Risk Technologies Page 103 of 373

Chapter 9
FX and FX Derivatives

9.1 FX Forwards
9.1.1 Description of Instrument
Spot and Forward FX transactions involve the exchange of cash in one currency
for cash of another currency. With a spot FX transaction the money is
exchanged on the “spot” date, which is determined by the market convention
for the two currencies being exchanged. Normally the convention is that the
spot date is two business, or “good”, days forward from today, and must be
business days in each currency’s home country.
An FX Swap is a combination of two FX Forwards; one being the reverse of the
other, but being forward in time. This is represented in RAZOR as being a deal
containing the two FX Forward products.

9.1.2 XML Representation

fpmlFXLeg Schema
Name: Type Occurs Size Description
productType
1..1 50 Indicates the type of product
string
This is the first of the two currency flows
exchangedCurrency1
1..1 that define a single leg of a standard
fpmlCurrencyFlow
foreign exchange transaction.
This is the second of the two currency
exchangedCurrency2 flows that define a single leg of a
1..1
fpmlCurrencyFlow standard foreign exchange
transaction.
valueDate The date on which both currencies traded
1..1
date will settle.
exchangeRate The rate of exchange between the two
1..1
fpmlFXRate currencies.
Used to describe a particular type of FX
nonDeliverableForward
0..1 forward transaction that is settled in a
fpmlFXCashSettlement
single currency.

FX Forward Example
<fxleg>
<exchangedCurrency1>
<payerPartyReference href = "ITE"/>
<receiverPartyReference href = "BAN"/>
<paymentAmount>
<currency>AUD</currency>
<amount>10000000</amount>
Razor Financial Principals

</paymentAmount>
</exchangedCurrency1>
<exchangedCurrency2>
<payerPartyReference href = "BAN"/>
<receiverPartyReference href = "ITE"/>
<paymentAmount>
<currency>USD</currency>
<amount>5600000</amount>
</paymentAmount>
</exchangedCurrency2>
<valueDate>2003-12-21</valueDate>
<exchangeRate>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<rate>0.5600</rate>
</exchangeRate>
</fxleg>

9.1.3 Pricing

9.1.4 Currency Calculations

Exchange rates are the rate at which you exchange one unit of one currency
for another currency. For example, an exchange rate of AUD/USD 0.5406
means that we are exchanging Australian dollars for US dollars and that we are
getting 0.5406 cents of one currency for one unit of the other currency. The
question becomes - which currency are we getting 0.5406 cents for 1 dollar of
the other currency?
The answer is that this depends on the quote convention. The quote
convention describes the market convention for which currency is the
0.5406
numerator in the fraction , and which is the denominator. The numerator
1
and denominator currency is specified for each unique currency pair.
Interest rate parity describes the relationships of interest rates in different
currencies with the behaviour of exchange rates. This mechanism com- pares
investing in the risk-free rate locally with exchanging the money into a foreign
currency and investing offshore. The investor should be able to achieve the
same rate of return by exchanging money to the foreign exchange rate and
investing offshore and then exchanging the money back into local terms, as he
is to investing locally at the risk-free rate or an arbitrage opportunity will
exist.
FX spot and forward deals are agreements to exchange amounts denominated
in differing currencies at an agreed exchange rate. FX physicals can be priced
in two ways:
The value of the deal can be seen as the difference in the agreed exchange rate
and the forward exchange rate when the deal is to be settled.
The value of the deal can be seen as the difference in the discounted value of
the forward cash flows, converted to a common currency. RAZOR uses this
approach to value FX Physical trades.

©2009 Razor Risk Technologies Page 105 of 373

Razor Financial Principals

Let
C ccy1 = the cash flow denominated in currency ccy1

C ccy 2 = the cash flow denominated in currency ccy 2

df ccy1 = the ccy1 discount factor from spot to settlement

df ccy 2 = the ccy 2 discount factor from spot to settlement

1 = the ccy1 → val exchange rate

val
S ccy

2 = the ccy 2 → val exchange rate

val
S ccy
Then

V = Cccy1df ccy1S ccy

val
1 + Cccy 2 df ccy 2 S ccy 2
val

9.2 FX Vanilla Options

9.2.1 Description of Instrument
FX Vanilla Options give the purchaser the right to buy (vanilla calls) or sell
(vanilla put) one of the currency in exchange for the other currency at the
strike price. Note that a call on one currency is the same as a put on the other
currency.

9.2.2 XML Representation

FX Vanilla Options are specified in the FPML version 3 standards.

XML Schema

fpmlFXOptionLeg Schema
Name: Type Occurs Size Description
productType
1..1 50 Indicates the type of product
string
buyerPartyReference
1..1
fpmlPartyReference
sellerPartyReference
1..1
fpmlPartyReference
expiryDateTimeExpiryDate
1..1
date
expiryDateTimeExpiryTime
1..1
time
expiryDateTimeCutName
0..1
string
exerciseStyle The manner in which the option can be
1..1
string exercised

©2009 Razor Risk Technologies Page 106 of 373

Razor Financial Principals

fpmlFXOptionLeg Schema
Name: Type Occurs Size Description
fxOptionPremium Premium amount or premium installment
0..n
fpmlFXOptionPremium amount for an option.
valueDate The date on which both currencies traded
1..1
date will settle
This optional element is only used if an
option has been specified at execution
cashSettlementTerms
0..1 time to be settled into a single cash
fpmlFXCashSettlement
payment. This would be used for a non-
deliverable option
putCurrencyAmountCurrency The currency in which an amount is
1..1
string denominated
putCurrencyAmountAmount
1..1 The monetary quantity in currency units
decimal
callCurrencyAmountCurrency The currency in which an amount is
1..1
string denominated
callCurrencyAmountAmount
1..1 The monetary quantity in currency units
decimal
fxStrikePriceRate
1..1
decimal
fxStrikePriceStrikeQuoteBasis The method by which the strike rate is
1..1
string quoted
quotedAs
1..1 Describes how the option was quoted
fpmlQuotedAs

FX European Option Example

<fxSimpleOption>
<productType>Foreign Exchange</productType>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>
<expiryDate>2004-05-31</expiryDate>
<expiryTime>1200</expiryTime>
</expiryDateTime>
<exerciseStyle>European</exerciseStyle>
<fxOptionPremium>
<payerPartyReference href="ITE"/>
<receiverPartyReference href="BAN"/>
<premiumAmount>
<currency>AUD</currency>
<amount>1000.00</amount>
</premiumAmount>
<premiumSettlementDate>2003-06-02</premiumSettlementDate>
</fxOptionPremium>
<valueDate>2004-06-02</valueDate>
<putCurrencyAmount>

©2009 Razor Risk Technologies Page 107 of 373

Razor Financial Principals

<currency>USD</currency>
<amount>550000.000000</amount>
</putCurrencyAmount>
<callCurrencyAmount>
<currency>AUD</currency>
<amount>1000000.000000</amount>
</callCurrencyAmount>
<fxStrikePrice>
<rate>0.550000</rate>
<strikeQuoteBasis>PUTCURRENCYPERCALLCURRENCY</strikeQuoteBasis>
</fxStrikePrice>
</fxSimpleOption>

9.2.3 Pricing
When valuing currency options we will call one currency involved in the option
contract the foreign currency. This currency is the denominator in the quote
convention for the currency pair. The other currency, the numerator in the
currency pair’s quote convention, we will call the domestic currency. Our
valuation will then be pricing in terms of the domestic currency. We are going
to assume that the exchange rate follows a geometric Brownian motion
process.
Let
S = the exchange rate
r = the annualised domestic rate, continuously compounded
r f = the annualised foreign interest rate, continuously compounded
σ = the exchange rate volatility
X = the strike rate
Tx = the time from today to expiry, annualised
Td = the time from spot to delivery, annualised
( r − r f )Td
F0 = S 0 e
N is the cumulative normal distribution function.

 F  σ Tx  F  σ Tx
2 2
ln  0  + ln  0  −
d1 =   d2 =  
X 2 X 2
σ Tx σ Tx
The European call price is given by
c = e − rTd ( F0 N ( d1 ) − XN ( d 2 ))
The European put price is given by
p = e − rTd ( XN (− d 2 ) − F0 N ( − d1 ))

Please refer to section 17.1.4 - generalised Black-Scholes option pricing

formula for further detail and the Greeks.

©2009 Razor Risk Technologies Page 108 of 373

Razor Financial Principals

9.3 FX Single Barrier Options

9.3.1 Description of Instrument
Barrier options are a family of options that cause an underlying product to
come into existence or cease to exist when predetermined trigger levels are
reached. There are two major types of single barrier options - Knock Ins and
Knock Outs. Knock In options cause an underlying vanilla option to be created
when the trigger level is breached and Knock out options cause the underlying
vanilla option to cease to exist upon a breach of the barrier. The barrier can be
any traded variable and may or may not be directly related to the underlying
of the original option.

9.3.2 XML Representation

XML Schema

fpmlFXBarrierOption Schema
Name: Type Occurs Size Description
productType
1..1 50 Indicates the type of product
string
buyerPartyReference
1..1
fpmlPartyReference
sellerPartyReference
1..1
fpmlPartyReference
expiryDateTimeExpiryDate
1..1
date
expiryDateTimeExpiryTime
1..1
time
expiryDateTimeCutName
0..1
string
exerciseStyle The manner in which the option can be
1..1
string exercised
fxOptionPremium Premium amount or premium instalment
0..n
fpmlFXOptionPremium amount for an option
valueDate The date on which both currencies traded
1..1
date will settle
This optional element is only used if an
option has been specified at execution
cashSettlementTerms
0..1 time to be settled into a single cash
fpmlFXCashSettlement
payment. This would be used for a non-
deliverable option.
putCurrencyAmountCurrency The currency in which an amount is
1..1
string denominated
putCurrencyAmountAmount
1..1 The monetary quantity in currency units
decimal
callCurrencyAmountCurrency 1..1 The currency in which an amount is

©2009 Razor Risk Technologies Page 109 of 373

Razor Financial Principals

fpmlFXBarrierOption Schema
Name: Type Occurs Size Description
string denominated
callCurrencyAmountAmount
1..1 The monetary quantity in currency units
decimal
fxStrikePriceRate
1..1
decimal
fxStrikePriceStrikeQuoteBasis The method by which the strike rate is
1..1
string quoted
quotedAs
0..1 Describes how the option was quoted
fpmlQuotedAs
An optional element used for FX forwards
and certain types of FX OTC options. For
deals consummated in the FX Forwards
Market, this represents the current
market rate for a particular currency
spotRate
0..1 pair. For barrier and digital/binary
decimal
options, it can be useful to include the
spot rate at the time the option was
executed to make it easier to know
whether the option needs to move ”up”
or ”down” to be triggered.
Information about a barrier rate in a
fxBarrier
0..n Barrier Option - specifying the exact
fpmlFXBarrier
criteria for a trigger event to occur.
The amount of currency which becomes
triggerPayout
0..1 payable if and when a trigger event
fpmlFXOptionPayout
occurs

fpmlFXBarrier Schema
Name: Type Occurs Size Description
This specifies whether the option
fxBarrierType becomes effective (”knock-in”) or is
0..1
string annulled (”knock- out”) when the
respective trigger event occurs.
quotedCurrencyPairCurrency1 The first currency specified when a pair
1..1
string of currencies is to be evaluated
quotedCurrencyPairCurrency2 The second currency specified when a
1..1
string pair of currencies is to be evaluated
quotedCurrencyPairQuoteBasis The method by which the exchange rate
1..1
string is quoted
The market rate is observed relative to
triggerRate
1..1 the trigger rate, and if it is found to be
decimal
on the predefined side of (above or

©2009 Razor Risk Technologies Page 110 of 373

Razor Financial Principals

fpmlFXBarrier Schema
Name: Type Occurs Size Description
below) the trigger rate, a trigger event
is deemed to have occurred
informationSource
0..n
fpmlInformationSource
The start of the period over which
observationStartDate
0..1 observations are made to determine
date
whether a trigger has occurred
The end of the period over which
observationEndDate
0..1 observations are made to determine
date
whether a trigger event has occurred

FX Single Barrier Option Example

<fxBarrierOption>
<productType>FX Single Barrier</productType>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>
<expiryDate>2004-05-31</expiryDate>
<expiryTime>1200</expiryTime>
</expiryDateTime>
<exerciseStyle>European</exerciseStyle>
<fxOptionPremium>
<payerPartyReference href="ITE"/>
<receiverPartyReference href="BAN"/>
<premiumAmount>
<currency>AUD</currency>
<amount>1000.000000</amount>
</premiumAmount>
<premiumSettlementDate>2003-06-02</premiumSettlementDate>
</fxOptionPremium>
<valueDate>2004-06-02</valueDate>
<putCurrencyAmount>
<currency>USD</currency>
<amount>550000.000000</amount>
</putCurrencyAmount>
<callCurrencyAmount>
<currency>AUD</currency>
<amount>1000000.000000</amount>
</callCurrencyAmount>
<fxStrikePrice>
<rate>0.550000</rate>
<strikeQuoteBasis>PUTCURRENCYPERCALLCURRENCY</strikeQuoteBasis>
</fxStrikePrice>
<spotRate>0.585000</spotRate>
<fxBarrier>
<fxBarrierType>KNOCKIN</fxBarrierType>
<quotedCurrencyPair>
<currency1>AUD</currency1>

©2009 Razor Risk Technologies Page 111 of 373

Razor Financial Principals

<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<triggerRate>0.650000</triggerRate>
<informationSource>
<rateSource>Reuters</rateSource>
<rateSourcePage>AUD=</rateSourcePage>
</informationSource>
</fxBarrier>
</fxBarrierOption>

9.3.3 Pricing
The equations below are derived from Reiner and Rubinstein [1991a] equations
for pricing standard barrier options. The implementation in RAZOR identifies
two different time periods; the time from the spot date to the delivery date of
the underlying product, which is applied to the exchange and strike rate
components in the pricing model, and the time from the pricing date to option
expiry, which is applied to the volatility components.
Let
ln( S / X ) ln( S / H )
x1 = + (1 + µ )σ T x2 = + (1 + µ )σ T
σ T σ T
ln( H 2 / SX ) ln( H / S )
y1 = + (1 + µ )σ T y2 = + (1 + µ )σ T
σ T σ T
ln( H / S ) b −σ 2 / 2 2r
z= + λσ T µ= λ = µ2 +
σ T σ 2
σ2
A = φSe ( b − r )T N (φx1 ) − φXe − rT N (φx1 − φσ T )

B = φSe ( b − r )T N (φx2 ) − φXe − rT N (φx2 − φσ T )

H 2 ( µ +1) H
C = φSe ( b − r )T ( ) N (ηy1 ) − φXe − rT ( ) 2 µ N (ηy1 − ησ T )
S S
H 2 ( µ +1) H
D = φSe (b − r )T ( ) N (ηy2 ) − φXe − rT ( ) 2 µ N (ηy2 − ησ T )
S S
H 2µ
E = Ke − rT ( N (ηx2 − ησ T ) − ( ) N (ηy2 − ησ T ))
S
H µ +λ H
F = K (( ) N (ηζ ) + ( ) µ − λ N (ηζ − 2ηλσ T ))
S S
c di ( X > H ) = C + E ,η = 1, φ = 1

c di ( X < H ) = A − B + D + E ,η = 1, φ = 1
cui ( X > H ) = A + E ,η = −1, φ = 1
cui ( X < H ) = B − C + D + E ,η = −1, φ = 1

p di ( X > H ) = B − C + D + E ,η = 1, φ = −1

©2009 Razor Risk Technologies Page 112 of 373

Razor Financial Principals

p di ( X < H ) = A + E ,η = 1, φ = −1
pui ( X > H ) = A − B + D + E ,η = −1, φ = −1
pui ( X < H ) = C + E ,η = −1, φ = −1
c do ( X > H ) = A − C + F ,η = 1, φ = 1
c do ( X < H ) = B − D + F ,η = 1, φ = 1
cuo ( X > H ) = F ,η = −1, φ = 1
cuo ( X < H ) = A − B + C − D + F ,η = −1, φ = 1

p do ( X > H ) = A − B + C − D + F ,η = 1, φ = −1
p do ( X < H ) = F ,η = 1, φ = −1
p uo ( X > H ) = B − D + F ,η = −1, φ = −1

puo ( X < H ) = A − C + F ,η = −1, φ = −1

Rebates
An optional rebate may be paid out the barrier is hit (for a knock-out
option) or not hitting the barrier (for a knock-in option.
Rebates in Razor are cash rebates factored into the intrinsic value of the
option as either a digital no-touch (knock-in) or a digital (binary) one-
touch (knock-out). Rebates are implemented for knock-ins that do not
knock-in at expiry as a digital no-touch paid in units of CCY2. Rebates for
options that knock out prior to expiry are implemented as a digital one-
touch paid in units of CCY2.

Knock-In Rebate at Expiry

If the option has not been knocked-in during its lifetime a rebate can be
specified which is paid out at expiration. The implementation of the
rebate in this model is equivalent to a single barrier digital no-touch
option.

Rebate of K units of CCY2

 2

( )
H
(
r = Ke− rT  N η x2 − ησ T −   N η y2 − ησ T 
S 
)
 

Rebate of K units of CCY1

2µ
 
( )
H
(
r = K ⋅ Se − rT  N η x1 − ησ T −   N η y1 − ησ T 
S 
)
 

Knock-Out Rebate at Hit

©2009 Razor Risk Technologies Page 113 of 373

Razor Financial Principals

If the option had been knocked-out during it’s lifetime a rebate can be
specified which is paid out at hit. The payout term is equivalent to a
single barrier digital one-touch at hit

Rebate of K units of CCY2

2 µ +λ 2 µ −λ
 H  
r = K  
 S 
H 
N (ηζ ) −  
S 
(
N ηζ − 2ηλσ T 

)

Rebate of K units of CCY1

 H  2 µ + λ 2 µ −λ

r = S  
H
N (ηζ ) −   (
N ηζ − 2ηλσ T  )
 S  S  

Knock-Out Rebate at Expiry:

If the option had been knocked-out during it’s lifetime a rebate can be
specified which is paid out at hit. The payout term is equivalent to a
single barrier digital one-touch at expiry

Rebate of K units of CCY2

2 µ −λ
 
( H
r = K  N −η x2 − ησ T −  
 S 
) N η y2 − ησ T 

( )
Rebate of K units of CCY1
2 µ −λ
 
( H
r = K  N −η x1 − ησ T −  
S 
) N η y1 − ησ T  ( )
 

9.4 FX Double Barrier Option

9.4.1 Description of Instrument
A double barrier option knocks in or out the underlying vanilla option if the
underlying price touches an upper or lower barrier prior to expiration.

9.4.2 XML Representation

XML Schema
This product uses the fpmlFXBarrierOptionSchema.

FX Double Barrier Option Example

The following is an example of a double no touch:
<fxBarrierOption>
<productType>DOUBLEBARRIER</productType>
<buyerPartyReference href="ITE" />
<sellerPartyReference href="BAN" />
<expiryDateTime>

©2009 Razor Risk Technologies Page 114 of 373

Razor Financial Principals

<expiryDate>2005-03-04</expiryDate>
<hourMinuteTime>10:00</hourMinuteTime>
</expiryDateTime>
<exerciseStyle>EUROPEAN</exerciseStyle>
<fxOptionPremium>
<payerPartyReference href="ITE" />
<receiverPartyReference href="BAN" />
<premiumAmount type="Money">
<currency>USD</currency>
<amount>192765.35</amount>
</premiumAmount>
<premiumSettlementDate>2005-01-07</premiumSettlementDate>
</fxOptionPremium>
<valueDate>2002-03-06</valueDate>
<putCurrencyAmount>
<currency>JPY</currency>
<amount>2500000000</amount>
</putCurrencyAmount>
<callCurrencyAmount>
<currency>USD</currency>
<amount>23798191.34</amount>
</callCurrencyAmount>
<fxStrikePrice>
<rate>105.05</rate>
<strikeQuoteBasis>PUTCURRENCYPERCALLCURRENCY</strikeQuoteBasis>
</fxStrikePrice>
<spotRate>106</spotRate>
<fxBarrier>
<fxBarrierType>KNOCKOUT</fxBarrierType>
<quotedCurrencyPair>
<currency1>USD</currency1>
<currency2>JPY</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<triggerRate>102</triggerRate>
<informationSource>
<rateSource>Reuters</rateSource>
<rateSourcePage>JPY=</rateSourcePage>
</informationSource>
</fxBarrier>
<fxBarrier>
<fxBarrierType>KNOCKOUT</fxBarrierType>
<quotedCurrencyPair>
<currency1>USD</currency1>
<currency2>JPY</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<triggerRate>115</triggerRate>
<informationSource>
<rateSource>Reuters</rateSource>
<rateSourcePage>JPY=</rateSourcePage>
</informationSource>
</fxBarrier>

©2009 Razor Risk Technologies Page 115 of 373

Razor Financial Principals

</fxBarrierOption>

9.4.3 Pricing
The pricing model uses Ikeda and Kunitomo (1992) formula. The value of a call
double barrier can be expressed as follows:

ln( SU 2 n / XL2 n ) + (b + σ 2 / 2)T ln( SU 2 n / FL2 n ) + (b + σ 2 / 2)T

d1 = d2 =
σ T σ T
ln( L2n+2
/ XSU 2 n ) + (b + σ 2 / 2)T ln( L2n+2
/ FSU 2 n ) + (b + σ 2 / 2)T
d3 = d4 =
σ T σ T
2(b − δ 2 − n(δ1 − δ 2 )) (δ1 − δ 2 )
µ1 = +1 µ 2 = 2n
σ2
σ 2

2(b − δ 2 + n(δ1 − δ 2 ))
µ3 = +1
σ2

F = Ueδ1T
∞  Un L Ln+1 
c = Se(b − r )T ∑ n=−∞ ( n ) µ1 ( ) µ2 ( N ( d1 ) − N ( d 2 )) − ( n ) µ3 ( N (d3 ) − N ( d 4 ))  −
 L S U S 
n +1
 U µ1 − 2 L µ2
n
L µ −2 
∞ ( n ) ( ) ( N ( d1 − σ T ) − N (d 2 − σ T )) − ( n ) 3 
Xe ∑ n =−∞  L
− rT
S U S 
( N ( d − σ T ) − N (d − σ T )) 
 3 4 
The value of a double barrier put option can be expressed as follows:

ln( SU 2 n / EL2 n ) + (b + σ 2 / 2)T ln( SU 2 n / XL2 n ) + (b + σ 2 / 2)T

y1 = y2 =
σ T σ T
ln( L2n+2
/ ESU 2 n ) + (b + σ 2 / 2)T ln( L2n+2
/ XSU 2 n ) + (b + σ 2 / 2)T
y3 = y4 =
σ T σ T
E = Le δ 2T
 U n µ1 − 2 L µ 2 Ln +1 µ − 2 
( n ) ( ) ( N ( y1 − σ T ) − N ( y2 − σ T )) − ( n ) 3 ( N ( y3 − σ T ) −
 L S U S −
− rT ∞ 
p = Xe ∑n = −∞  N ( y − σ T )) 
4 
∞  Un L Ln +1 
Se(b − r )T ∑n = −∞ ( n ) µ1 ( ) µ 2 ( N ( y1 ) − N ( y2 )) − ( n ) µ3 ( N ( y3 ) − N ( y4 )) 
 L S U S 

Strike Outside the Barriers

The Ikeda and Kunitomo formula only holds when the strike price is inside the
barrier range. When the strike is outside the range, we price the option in the
following ways:
Double knock-Ins must be priced as the Vanilla European Option less the
Double knock-out.

©2009 Razor Risk Technologies Page 116 of 373

Razor Financial Principals

Notation:

In the following, C({strike},{lower barrier},{upper barrier}) refers to the

existing RAZOR model for double barriers. The convention is “C” for call
on the foreign currency, “P” for put on the foreign currency.

Double knock outs:

Case 1:

Where the prices are expressed in units of domestic currency per

foreign currency,
e.g. 1.5554 USD per EUR,

(L − K )
1. C '( K , L,U ) = C ( L, L,U ) + ( C ( L, L,U ) + P(U , L,U ) )
(U − L)

where K < L. If K > U then value is zero.

(K − U )
2. P '( K , L,U ) = P(U , L,U ) + ( C ( L, L,U ) + P(U , L,U ) )
(U − L)
where K > U. If K < L then value is zero.

Case 2:

Where the prices are expressed in units of foreign currency per domestic
currency,
e.g. 106.92 JPY per USD,

C’(K,L,U) = C(U,L,U) + (K-U)/(U-L) * (C(U,L,U) + P(L,L,U)),

Where K > U. If K < L then value is zero.

4. P’(K,L,U) = P(L,L,U) + (L – K)/(U-L)*( C(U,L,U) + P(L,L,U)),

Where K<L. If K> H then value is zero.

9.5 FX Digital Option

9.5.1 Description of Instrument
Digital options (also known as “binary options”) are options that either pay out
a fixed amount if the option expires in the money, or nothing. If the currency
used to value the digital is the same as the payoff currency at expiry, the
binary option will have either an exposure based on the payoff amount, or
zero. If the payoff currency is different to the valuation currency, the digital

©2009 Razor Risk Technologies Page 117 of 373

Razor Financial Principals

option will have an exposure that changes with the exchange rate between the
payoff currency and the valuation currency.

9.5.2 XML Representation

XML Schema

fpmlFXDigitalOption Schema
Name: Type Occurs Size Description
productType
1..1 50 Indicates the type of product
string
buyerPartyReference
1..1
fpmlPartyReference
sellerPartyReference
1..1
fpmlPartyReference
expiryDateTimeExpiryDate
1..1
date
expiryDateTimeExpiryTime
1..1
time
expiryDateTimeCutName
0..1
string
fxOptionPremium Premium amount or premium instalment
0..n
fpmlFXOptionPremium amount for an option
valueDate The date on which both currencies
1..1
date traded will settle
quotedCurrencyPairCurrency1 The first currency specified when a pair
1..1
string of currencies is to be evaluated
quotedCurrencyPairCurrency2 The second currency specified when a
1..1
string pair of currencies is to be evaluated
quotedCurrencyPairQuoteBasis The method by which the exchange rate
1..1
string is quoted
An optional element used for FX
forwards and certain types of FX OTC
options. For deals consummated in the
FX Forwards Market, this represents the
current market rate for a particular
spotRate
0..1 currency pair. For barrier and
decimal
digital/binary options, it can be useful
to include the spot rate at the time the
option was executed to make it easier
to know whether the option needs to
move ”up” or ”down” to be triggered.
A European trigger occurs if the trigger
fxEuropeanTrigger criteria are met, but these are valid
0..n
fpmlFXEuropeanTrigger (and an observation is made) only at the
maturity of the option

©2009 Razor Risk Technologies Page 118 of 373

Razor Financial Principals

fpmlFXDigitalOption Schema
Name: Type Occurs Size Description
An American trigger occurs if the trigger
fxAmericanTrigger
0..n criteria are met at any time from the
fpmlFXAmericanTrigger
initiation to the maturity of the option
The amount of currency which becomes
triggerPayout
1..1 payable if and when a trigger event
fpmlFXOptionPayout
occurs

FX European Binary Option Example

<fxDigitalOption>
<productType>Euro Binary</productType>
<buyerPartyReference href="ITE" />
<sellerPartyReference href="BAN" />
<expiryDateTime>
<expiryDate>2004-01-24</expiryDate>
<expiryTime>1200</expiryTime>
<businessCenter>AUSY</businessCenter>
<cutName>Sydney</cutName>
</expiryDateTime>
<fxOptionPremium>
<payerPartyReference href="ITE" />
<receiverPartyReference href="BAN" />
<premiumAmount type="Money">
<currency>USD</currency>
<amount>369000</amount>
</premiumAmount>
<premiumSettlementDate>2001-09-14</premiumSettlementDate>
</fxOptionPremium>
<valueDate>2004-01-26</valueDate>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<spotRate>0.6500</spotRate>
<fxEuropeanTrigger>
<triggerCondition>Above</triggerCondition>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<triggerRate>0.6000</triggerRate>
<informationSource>
<rateSource>REUTERS</rateSource>
<rateSourcePage>AUD=</rateSourcePage>
</informationSource>
</fxEuropeanTrigger>
<triggerPayout>
<currency>AUD</currency>

©2009 Razor Risk Technologies Page 119 of 373

Razor Financial Principals

<amount>10000</amount>
<payoutStyle>IMMEDIATE</payoutStyle>
</triggerPayout>
</fxDigitalOption>

9.5.3 Pricing
Let
S = the exchange rate
r = the annualised domestic interest rate, continuously compounded
r f = the annualised foreign interest rate, continuously compounded
σ = the exchange rate volatility
X = the strike rate
Tx = the time from today to expiry, annualised

Td = the time from spot to delivery, annualised

( r − r f ) Td
F0 = S0e
N is the cumulative normal distribution function.

 F  σ Tx  F  σ Tx
2 2
ln  0  + ln  0  _
d1 =   d2 =  
X 2 X 2
σ Tx σ Tx
The European digital asset price is given by:
d asset = e − rTd F0 N ( d1 )
The European digital cash price is given by:
d cash = e − rTd XN (d 2 )

9.6 FX Digital Barrier Option

9.6.1 Description of Instrument
Digital barrier options are options containing a barrier that, when breached,
either causes an underlying cash payment to occur, or knocks-out an
underlying cash payment.

9.6.2 XML Representation

This product uses the FPML FXDigitalOption product schema in FPML version 3.

FX One Touch Digital Option Example

<fxDigitalOption>
<productType>One Touch</productType>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>

©2009 Razor Risk Technologies Page 120 of 373

Razor Financial Principals

<expiryDate>2004-01-22</expiryDate>
<expiryTime>1600</expiryTime>
<cutName>AUSY</cutName>
</expiryDateTime>
<valueDate>2004-01-24</valueDate>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<spotRate>0.5500</spotRate>
<fxAmericanTrigger>
<touchCondition>Touch</touchCondition>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<triggerRate>0.7500</triggerRate>
<informationSource>
<rateSource>REUTERS</rateSource>
<rateSourcePage>AUD=</rateSourcePage>
</informationSource>
<observationStartDate>2001-09-14</observationStartDate>
<observationEndDate>2004-01-22</observationEndDate>
</fxAmericanTrigger>
<triggerPayout>
<currency>AUD</currency>
<amount>100000000</amount>
<payoutStyle>DEFERRED</payoutStyle>
</triggerPayout>
</fxDigitalOption>

FX No Touch Digital Option Example

<fxDigitalOption>
<productType>No Touch</productType>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>
<expiryDate>2004-01-22</expiryDate>
<expiryTime>1600</expiryTime>
<cutName>AUSY</cutName>
</expiryDateTime>
<valueDate>2004-01-24</valueDate>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<spotRate>0.5500</spotRate>
<fxAmericanTrigger>
<touchCondition>Notouch</touchCondition>
<quotedCurrencyPair>

©2009 Razor Risk Technologies Page 121 of 373

Razor Financial Principals

<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<triggerRate>0.6000</triggerRate>
<informationSource>
<rateSource>REUTERS</rateSource>
<rateSourcePage>AUD=</rateSourcePage>
</informationSource>
<observationStartDate>2001-09-14</observationStartDate>
<observationEndDate>2004-01-22</observationEndDate>
</fxAmericanTrigger>
<triggerPayout>
<currency>AUD</currency>
<amount>100000000</amount>
<payoutStyle>DEFERRED</payoutStyle>
</triggerPayout>
</fxDigitalOption>

9.6.3 Pricing
Riener and Rubinstein present a set of formulas that can be used to price the
different types of digital barrier options.
ln( S / X ) ln( S / H )
x1 = + ( µ + 1)σ T x2 = + ( µ + 1)σ T
σ T σ T
ln( H 2 / SX ) ln( H / S )
y1 = + ( µ + 1)σ T y2 = + ( µ + 1)σ T
σ T σ T
ln( H / S ) b −σ 2 / 2 2r
z= + λσ T µ= λ = µ2 +
σ T σ2 σ2

A1 = Se (b − r )T N (φx1 )

B1 = Ke − rT N (φx1 − φσ T )

A2 = Se (b − r )T N (φx2 )

B2 = Ke − rT N (φx2 − φσ T )

A3 = Se ( b − r )T ( H / S ) 2 ( µ +1) N (ηy1 )

B3 = Ke − rT ( H / S ) 2 µ N (ηy1 −ησ T )

A4 = Se(b − r )T ( H / S ) 2( µ +1) N (ηy2 )

B4 = K e− rT ( H / S ) 2 µ N (ηy2 − ησ T )

A5 = K (( H / S ) µ +λ N (ηz ) + ( H / S ) µ −λ N (ηz − 2ηλσ T )

By using the formulas A1 to A5 and B1 to B4 in various combinations we can
value the different types of binary barrier options.

©2009 Razor Risk Technologies Page 122 of 373

Razor Financial Principals

# Condition Value η φ # Condition Value η φ

1 S>H A5 1 18 S < H , X > H B1 − B2 + B4 −1 −1
2 S<H A5 −1 X <H B1 −1
3 S > H,K = H A5 1 19 S > H , X > H A2 − A3 + A4 1 −1
4 S < H,K = H A5 −1 X <H A1 −1
5 S>H B2 + B4 1 −1 20 S < H , X > H A1 − A2 + A3 −1 −1
6 S<H B2 + B4 −1 1 X <H A3 −1
7 S>H A2 + A4 1 −1 21 S > H , X > H B1 − B3 1 1
8 S<H A2 + A4 1 −1 X <H B2 − B4 1 1
9 S>H B2 − B4 1 1 22 S < H , X > H 0
10 S<H B2 − B4 −1 −1 X <H B1 − B2 + B3 − B4 −1 1
11 S>H A2 − A4 1 1 23 S > H , X > H A2 − A4 1 1
12 S<H A2 − A4 −1 −1 X <H A2 − A4 −1 −1
13 S > H , X > H B3 1 24 S < H , X > H 0
X <H B1 − B2 + B4 1 1 X <H A1 − A2 + A3 − A4 −1 1
14 S < H , X > H B1 1 25 S > H , X > H B1 − B2 + B3 − B4 1 −1
X <H B2 − B3 + B4 −1 1 X <H 0
15 S > H , X > H A3 1 26 S < H , X > H B2 − B4 −1 −1
X <H A1 − A2 + A4 1 1 X <H B1 − B3 −1 −1
16 S < H , X > H B1 1 27 S > H , X > H A1 − A2 + A3 − A4 1 −1
X <H B2 − B3 + B4 −1 1 X <H 0
17 S > H , X > H B2 − B3 + B4 1 −1 28 S < H , X > H A2 − A4 −1 −1
X <H B1 −1 X <H A1 − A3 −1 −1

9.7 FX Double Digital Option

9.7.1 Description of Instrument
These derivative products have a double barrier, and a digital payoff is created
if either barrier is breached, or if neither barrier is breached. The payoff may
be either upon the breach of the barrier – “at hit” – or at expiry. Comment [M.L.1]: Currently Not
Supported

9.7.2 XML Representation

XML Schema

fpmlFXDigitalOption schema
Name: Type Occurs Size Description
productType
1..1 50 Indicates the type of product
string
buyerPartyReference
1..1
fpmlPartyReference
sellerPartyReference
1..1
fpmlPartyReference

©2009 Razor Risk Technologies Page 123 of 373

Razor Financial Principals

fpmlFXDigitalOption schema
Name: Type Occurs Size Description
expiryDateTimeExpiryDate
1..1
date
expiryDateTimeExpiryTime
1..1
time
expiryDateTimeCutName
0..1
string
fxOptionPremium Premium amount or premium instalment
0..n
fpmlFXOptionPremium amount for an option
valueDate The date on which both currencies
1..1
date traded will settle
quotedCurrencyPairCurrency1 The first currency specified when a pair
1..1
string of currencies is to be evaluated
quotedCurrencyPairCurrency2 The second currency specified when a
1..1
string pair of currencies is to be evaluated
quotedCurrencyPairQuoteBasis The method by which the exchange rate
1..1
string is quoted
An optional element used for FX
forwards and certain types of FX OTC
options. For deals consummated in the
FX Forwards Market, this represents the
current market rate for a particular
spotRate
0..1 currency pair. For barrier and
decimal
digital/binary options, it can be useful
to include the spot rate at the time the
option was executed to make it easier
to know whether the option needs to
move ”up” or ”down” to be triggered.
A European trigger occurs if the trigger
fxEuropeanTrigger criteria are met, but these are valid
0..n
fpmlFXEuropeanTrigger (and an observation is made) only at the
maturity of the option
An American trigger occurs if the trigger
fxAmericanTrigger
0..n criteria are met at any time from the
fpmlFXAmericanTrigger
initiation to the maturity of the option
The amount of currency which becomes
triggerPayout
1..1 payable if and when a trigger event
fpmlFXOptionPayout
occurs

FX Double One Touch Digital Option Example

<fxDigitalOption>
<productType>Double one touch</productType>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>

©2009 Razor Risk Technologies Page 124 of 373

Razor Financial Principals

<expiryDate>2004-11-26</expiryDate>
<hourMinuteTime>1400</hourMinuteTime>
<businessCenter>AUSY</businessCenter>
<cutName>SYDNEY</cutName>
</expiryDateTime>
<fxOptionPremium>
<payerPartyReference href="ITE"/>
<receiverPartyReference href="BAN"/>
<premiumAmount type="Money">
<currency>AUD</currency>
<amount>78000</amount>
</premiumAmount>
<premiumSettlementDate>2001-11-14</premiumSettlementDate>
</fxOptionPremium>
<valueDate>2004-11-28</valueDate>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<spotRate>0.5580</spotRate>
<fxAmericanTrigger>
<touchCondition>Touch</touchCondition>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<triggerRate>0.5800</triggerRate>
<informationSource>
<rateSource>REUTERS</rateSource>
<rateSourcePage>AUD=</rateSourcePage>
</informationSource>
<observationStartDate>2001-11-12</observationStartDate>
<observationEndDate>2004-11-26</observationEndDate>
</fxAmericanTrigger>
<fxAmericanTrigger>
<touchCondition>Touch</touchCondition>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<triggerRate>0.5400</triggerRate>
<informationSource>
<rateSource>REUTERS</rateSource>
<rateSourcePage>AUD=</rateSourcePage>
</informationSource>
<observationStartDate>2001-11-12</observationStartDate>
<observationEndDate>2004-11-26</observationEndDate>
</fxAmericanTrigger>
<triggerPayout>
<currency>AUD</currency>
<amount>2000000</amount>
©2009 Razor Risk Technologies Page 125 of 373
Razor Financial Principals

<payoutStyle>IMMEDIATE</payoutStyle>
</triggerPayout>
</fxDigitalOption>

FX Double No Touch Digital Option Example

<fxDigitalOption>
<productType>Double one touch</productType>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>
<expiryDate>2004-11-26</expiryDate>
<hourMinuteTime>1400</hourMinuteTime>
<businessCenter>AUSY</businessCenter>
<cutName>SYDNEY</cutName>
</expiryDateTime>
<fxOptionPremium>
<payerPartyReference href="ITE"/>
<receiverPartyReference href="BAN"/>
<premiumAmount type="Money">
<currency>AUD</currency>
<amount>78000</amount>
</premiumAmount>
<premiumSettlementDate>2001-11-14</premiumSettlementDate>
</fxOptionPremium>
<valueDate>2004-11-28</valueDate>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<spotRate>0.5580</spotRate>
<fxAmericanTrigger>
<touchCondition>Notouch</touchCondition>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<triggerRate>0.5800</triggerRate>
<informationSource>
<rateSource>REUTERS</rateSource>
<rateSourcePage>AUD=</rateSourcePage>
</informationSource>
<observationStartDate>2001-11-12</observationStartDate>
<observationEndDate>2004-11-26</observationEndDate>
</fxAmericanTrigger>
<fxAmericanTrigger>
<touchCondition>Notouch</touchCondition>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>USD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>

©2009 Razor Risk Technologies Page 126 of 373

Razor Financial Principals

<triggerRate>0.5400</triggerRate>
<informationSource>
<rateSource>REUTERS</rateSource>
<rateSourcePage>AUD=</rateSourcePage>
</informationSource>
<observationStartDate>2001-11-12</observationStartDate>
<observationEndDate>2004-11-26</observationEndDate>
</fxAmericanTrigger>
<triggerPayout>
<currency>AUD</currency>
<amount>2000000</amount>
<payoutStyle>IMMEDIATE</payoutStyle>
</triggerPayout>
</fxDigitalOption>

9.7.3 Pricing
Razor uses the Hui (1996) closed form approximation with 5 terms.

knock-out:

2
H  1  2b  1  2b  r
L = ln  2  α = −  2 − 1 β = −  2 − 1 − 2 2
 H1  2 σ  4 σ  σ

  S α i S 
α

  − ( −1)    1  iπ  
2

2π iK  H   iπ  S   2  L  
∞ −  − β  σ 2T
= ∑ 2  1   H2 
cknock − out  sin  L ln  H   e
 iπ 
2
L  1 
i =1
 α2 +   
 L 

knock-in:

C knock −in = Ke rT − C knock − out

American binary knock-out option

In this option, if S ever reaches one barrier, H2 , then the option is
worthless; thus on the line H2 the option value is zero. If S ever reaches
another barrier, H1 , the payment isR at the time of touching the
payment barrier.

   
2 − 1  iπ  − β σ 2T
2
  
  iπ     S  
β − 2  L     ln   
α
 S  ∞ 2   e   iπ  S   
  L  H1   
cknock − out = R  ∑   sin  ln    + 1 − 
 H1   i =1 iπ   iπ 
2
  L  H1    L 
  −β  
  L    
  

©2009 Razor Risk Technologies Page 127 of 373

Razor Financial Principals

9.8 FX Average Rate Options

9.8.1 Description of Instrument
An average rate option is a derivative security that gives the buyer a payoff at
the maturity of the option based on the difference between the average of
exchange rates sampled at points along the life of the option, and a strike
rate.
An average strike option is similar to an average rate option, but the payoff is
the difference between an average of exchange rates sampled at points along
the life of the option, and the exchange rate upon maturity of the option.

9.8.2 XML Representation

XML Schema

fpmlFXAverageRateOption Schema
Name: Type Occurs Size Description
productType
1..1 50 Indicates the type of product
string
buyerPartyReference
1..1
fpmlPartyReference
sellerPartyReference
1..1
fpmlPartyReference
expiryDateTimeExpiryDate
1..1
date
expiryDateTimeExpiryTime
1..1
time
expiryDateTimeCutName
0..1
string
exerciseStyle The manner in which the
1..1
string option can be exercised
Premium amount or premium
fxOptionPremium
0..n instalment amount for an
fpmlFXOptionPremium
option
valueDate The date on which both
1..1
date currencies traded will settle
putCurrencyAmountCurrency The currency in which an
1..1
string amount is denominated
putCurrencyAmountAmount The monetary quantity in
1..1
decimal currency units
callCurrencyAmountCurrency The currency in which an
1..1
string amount is denominated
callCurrencyAmountAmount The monetary quantity in
1..1
decimal currency units
fxStrikePriceRate 1..1

©2009 Razor Risk Technologies Page 128 of 373

Razor Financial Principals

fpmlFXAverageRateOption Schema
Name: Type Occurs Size Description
decimal
fxStrikePriceStrikeQuoteBasis The method by which the
1..1
string strike rate is quoted
An optional element used for
FX forwards and certain types
of FX OTC options. For deals
consummated in the FX
Forwards Market, this
represents the current market
rate for a particular currency
spotRate
0..1 pair. For barrier and
decimal
digital/binary options, it can
be useful to include the spot
rate at the time the option was
executed to make it easier to
know whether the option needs
to move ”up” or ”down” to be
triggered.
The ISO code of the currency in
payoutCurrency which a payout (if any) is to be
1..1
string made when a trigger is hit on a
digital or barrier option
The method by which the
averageRateQuoteBasis
1..1 average rate that is being
string
observed is quoted
Specifies the rounding
precision in terms of a number
of decimal places. Note how a
percentage rate rounding of 5
roundingPrecision
0..1 decimal places is expressed as
int
a rounding precision of 7 in the
FpML document since the
percentage is expressed as a
decimal, e.g. 9.876543.
The description of the
payoutFormula
0..1 mathematical computation for
string
how the payout is computed
The primary source for where
the rate observation will occur.
primaryRateSource
1..1 Will typically be either a page
fpmlInformationSource
or a reference bank published
rate.
An alternative, or secondary,
secondaryRateSource source for where the rate
0..1
fpmlInformationSource observation will occur. Will
typically be either a page or a

©2009 Razor Risk Technologies Page 129 of 373

Razor Financial Principals

fpmlFXAverageRateOption Schema
Name: Type Occurs Size Description
reference bank published rate.
The time at which the spot
currency exchange rate will be
fixingTime observed. It is specified as a
1..1
fpmlBusinessCenterTime time in a specific business
centre, e.g. 11:00am London
time.
averageRateObservationSchedule Parametric schedule of rate
1..1
fpmlFXAverageRateObservationSchedule observations
averageRateObservationDate One of more specific rate
0..n
fpmlFXAverageRateObservationDate observation dates
Describes prior rate
observations within average
rate options. Periodically, an
average rate option agreement
will be struck whereby some
rates have already been
observedRates
0..n observed in the past but will
fpmlObservedRates
become part of computation of
the average rate of the option.
This structure provides for
these previously observed rates
to be included in the
description of the trade.

FX Average Rate Option Example

<fxAverageRateOption>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>
<expiryDate>2003-11-30</expiryDate>
<hourMinuteTime>12:30</hourMinuteTime>
<businessCenter>AUSY</businessCenter>
</expiryDateTime>
<exerciseStyle>EUROPEAN</exerciseStyle>
<fxOptionPremium>
<payerPartyReference href="ITE"/>
<receiverPartyReference href="BAN"/>
<premiumAmount type="Money">
<currency>USD</currency>
<amount>1750</amount>
</premiumAmount>
<premiumSettlementDate>2003-08-18</premiumSettlementDate>
</fxOptionPremium>
<valueDate>2003-12-04</valueDate>
<putCurrencyAmount>
<currency>AUD</currency>
<amount>10000000</amount>
©2009 Razor Risk Technologies Page 130 of 373
Razor Financial Principals

</putCurrencyAmount>
<callCurrencyAmount>
<currency>USD</currency>
<amount>5855000</amount>
</callCurrencyAmount>
<fxStrikePrice>
<rate>0.5855</rate>
<strikeQuoteBasis>CALLCURRENCYPERPUTCURRENCY</strikeQuoteBasis>
</fxStrikePrice>
<payoutCurrency>USD</payoutCurrency>
<averageRateQuoteBasis>CALLCURRENCYPERPUTCURRENCY</averageRateQuote
Basis>
<primaryRateSource>
<rateSource>AUUS</rateSource>
</primaryRateSource>
<fixingTime>
<hourMinuteTime>18:00</hourMinuteTime>
<businessCenter>AUSY</businessCenter>
</fixingTime>
<averageRateObservationDate>
<observationDate>2003-11-01</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-02</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-05</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-06</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-07</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-08</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-09</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-13</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-14</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
©2009 Razor Risk Technologies Page 131 of 373
Razor Financial Principals

</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-15</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-16</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-19</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-20</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-21</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-23</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-26</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-27</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-28</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-29</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
<averageRateObservationDate>
<observationDate>2003-11-30</observationDate>
<averageRateWeightingFactor>1</averageRateWeightingFactor>
</averageRateObservationDate>
</fxAverageRateOption>

9.8.3 Pricing
RAZOR uses the Turnbull and Wakeman approximation to price arithmetic
average rate options.
Note that the formula doesn’t work when the cost of carry is zero.

©2009 Razor Risk Technologies Page 132 of 373

Razor Financial Principals

ln(M 2 )
σA = − 2bA
T
ln( M 1 )
bA =
T
ln(S / X ) + (bA + σ A2 / 2)T2
d1 =
σ A T2

d 2 = d1 − σ A T2

c ≈ Se( bA − r )T2 N ( d1 ) − Xe − rT2 N (d 2 )

p ≈ Xe − rT2 N ( d 2 ) − Se( bA − r )T2 N ( d1 )
The exact first and second moments of the arithmetic average are:

ebT − ebτ
M1 =
b(T − τ )

2e ( 2b +σ )T 2e ( 2b +σ )τ  1 e b (T −τ ) 
2 2

M2 = + 
2 
− 
(b + σ )(2b + σ )(T − τ ) b(T − τ )  2b + σ
2 2 2 2
b + σ 2 

9.9 FX Fade-In Option

9.9.1 Contract Definition

A fade-in option is a type of barrier option. There is a preset barrier value

for the underlying asset at a predetermined future date. There are four
types of barriers:

Up-and-in
If the underlying asset value is above the barrier value on the
predetermined future date, then the option turns to the plain-vanilla
option. Otherwise, the option expires without value at maturity.
Up-and-Out
If the underlying asset value is above the barrier value on the
predetermined future date, then the option expires immediately without
any value. Otherwise, the option pays off like a plain-vanilla option at
maturity.
Down-and-in
If the underlying asset value is below the barrier value on the
predetermined future date, then the option turns to the plain-vanilla
option. Otherwise, the option expires without value at maturity.
Down-and-out
If the underlying asset value is below the barrier value on the
predetermined future date, then the option expires immediately without

©2009 Razor Risk Technologies Page 133 of 373

Razor Financial Principals

any value. Otherwise, the option pays off like a plain-vanilla option at
maturity.

9.9.2 Model Specification

Define

t= the valuation date.

T1 = barrier observation date.
T2 = option maturity date.
T3 = option settlement date.
K = strike price.
h = barrier value.
S (t ) = the underlying asset value at time t .
r (t , T ) = the domestic risk-free rate applied from time t to T .
q (t , T ) = the foreign risk-free rate applied from time t to T .
σ (t, T ) = the volatility of the underlying asset applied from time t to T
.

α (s, t , DC ) = length of time from s to t in years and DC is the day count

 actual  t − s
basis., e.g. α  s, t , = .
 365  365

Φ(x1 , x 2 , Σ ) = cumulative value of a standard 2-D normal distribution

evaluated at x1 for first variable, x 2 for the second variable and Σ is the
covariance matrix.

Note that r (t , T ) , q (t , T ) and σ (t, T ) can be obtained from the term

structure curve.

Option Payoff

We define:
δ = ±1 for up and down options respectively.
η = ±1 for in and out options respectively.

We also define the following function:

f (S (T1 ); δ ,η , h ) = η if δ ⋅ S (T1 ) ≥ δ ⋅ h
f (S (T1 ); δ ,η , h ) = 1 − η otherwise.

The payoff at maturity of the fade-in option is:

VT = f (S (T1 ); δ ,η , h ) ⋅ [β ⋅ max(S (T2 ) − K ,0 )]

©2009 Razor Risk Technologies Page 134 of 373

Razor Financial Principals

where
β = ±1 for call and put respectively.

9.9.3 Pricing Formulas

The price of a fade-in option at the valuation date t is:

 1 + δ 
Vt = df (t , T3 ) ×  − δη  ⋅ β ⋅ [F2 Φ (α , γ , Σ1 ) − KΦ(α ' , γ ' , Σ1 )] +
 2 
1− δ  
 + δη  ⋅ β ⋅ [F2 Φ (− α , γ , Σ 2 ) − KΦ(− α ' , γ ' , Σ 2 )]
 2  

where
ln h − m1
α= − v1
v1
ln h − m1
α'=
v1

 F  v + v2
ln 2  + 1
γ =β⋅  
K 2
v1 + v 2
 F  v + v2
ln 2  − 1
γ '= β ⋅  
K 2
v1 + v 2
v1
ρ = −β ⋅
v1 + v 2
v1 = σ 2 (t , T1 )T1
1
m1 = ln F1 − v1
2
v 2 = σ (t , T2 )T2 − v1
2

F1 = S (t )e ( r (t ,T1 )− q (t ,T1 ))α (t ,T1 , DC )

F2 = S (t )e (r (t ,T2 )− q (t ,T2 ))α (t ,T2 , DC )

Σ1
=[1 ρ ; ρ 1] which is a 2 × 2 matrix for the standard 2-D normal
distribution
Σ2
=[1 − ρ ; − ρ 1] which is a 2 × 2 matrix for the standard 2-D normal
distribution.

©2009 Razor Risk Technologies Page 135 of 373

Chapter 10
Base Metals

10.1 Metal Forwards

10.1.1 Description of Instrument
Spot and Forward Metal transactions involve the exchange of metal for cash in
a particular currency. Because the underlying exchange is very similar to an FX
spot or forward transaction, with one of the currencies being the metal, this is
how the product is modelled and the exposure calculated in RAZOR.

10.1.2 XML Representation

fpmlFXLeg Schema
Name: Type Occurs Size Description
productType
1..1 50 Indicates the type of product
String
This is the first of the two currency flows
exchangedCurrency1
1..1 that define a single leg of a standard
fpmlCurrencyFlow
foreign ex- change transaction
This is the second of the two currency
exchangedCurrency2
1..1 flows that define a single leg of a
fpmlCurrencyFlow
standard foreign exchange transaction
valueDate The date on which both currencies traded
1..1
Date will settle
exchangeRate The rate of exchange between the two
1..1
fpmlFXRate currencies
Used to describe a particular type of FX
nonDeliverableForward
0..1 forward transaction that is settled in a
fpmlFXCashSettlement
single currency

Metal Forward Example

<fxleg>
<exchangedCurrency1>
<payerPartyReference href = "ITE"/>
<receiverPartyReference href = "BAN"/>
<paymentAmount>
<currency>AUD</currency>
<amount>10000000</amount>
</paymentAmount>
</exchangedCurrency1>
<exchangedCurrency2>
<payerPartyReference href = "BAN"/>
<receiverPartyReference href = "ITE"/>
<paymentAmount>
<currency>XAU</currency>
Razor Financial Principals

<amount>1852</amount>
</paymentAmount>
</exchangedCurrency2>
<valueDate>2003-12-21</valueDate>
<exchangeRate>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>XAU</currency2>
<quoteBasis>CURRENCY1PERCURRENCY2</quoteBasis>
</quotedCurrencyPair>
<rate>539.9568</rate>
</exchangeRate>
</fxleg>

10.1.3 Pricing
Let
C ccy = the cash flow denominated in currency ccy

Cmetal = the amount of the metal metal

df ccy = the ccy discount factor from spot to settlement

df metal = the metal discount factor from spot to settlement

val
S ccy = the ccy → val exchange rate
val
S metal = the metal → val exchange rate
Then

V = Cmetal df metal S metal

val
+ Cccy df ccy S ccy
val

10.2 Metal Vanilla Options

10.2.1 Description of Instrument
Metal vanilla options give the purchaser the right to buy (vanilla calls) or sell
(vanilla put) the metal at the given strike price. Again we can treat these
products in a similar manner to FX vanilla options.

10.2.2 XML Representation

XML Schema
We are using the FPML fpmlFXSimpleOption schema to represent this product.

Metal European Option Example

<fxSimpleOption>
<productType>Metal Call Option</productType>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>
<expiryDate>2004-05-31</expiryDate>

©2009 Razor Risk Technologies Page 137 of 373

Razor Financial Principals

<expiryTime>1200</expiryTime>
</expiryDateTime>
<exerciseStyle>European</exerciseStyle>
<fxOptionPremium>
<payerPartyReference href="ITE"/>
<receiverPartyReference href="BAN"/>
<premiumAmount>
<currency>AUD</currency>
<amount>50000</amount>
</premiumAmount>
<premiumSettlementDate>2003-06-02</premiumSettlementDate>
</fxOptionPremium>
<valueDate>2004-06-02</valueDate>
<putCurrencyAmount>
<currency>AUD</currency>
<amount>1000000</amount>
</putCurrencyAmount>
<callCurrencyAmount>
<currency>XAU</currency>
<amount>1852</amount>
</callCurrencyAmount>
<fxStrikePrice>
<rate>539.9568</rate>
<strikeQuoteBasis>PUTCURRENCYPERCALLCURRENCY</strikeQuoteBasis>
</fxStrikePrice>
</fxSimpleOption>

10.2.3 Pricing
Let
S = the current spot price of the metal
r = the continuously compounded interest rate
r f = the continuously compounded yield on the metal
σ = the price volatility
X = the strike price
Tx = the time from today to expiry, annualised

Td = the time from spot to delivery, annualised

( r − r f )Td
F0 = S 0 e

 F  σ Tx
2
ln 0  +
d1 =  
X 2
σ Tx

 F  σ Tx
2
ln 0  −
d2 =  
X 2
σ Tx

©2009 Razor Risk Technologies Page 138 of 373

Razor Financial Principals

The European call price is given by

c = e − rTd ( F0 N ( d1 ) − XN ( d 2 ))
The European put price is given by
p = e − rTd ( XN (− d 2 ) − F0 N ( − d1 ))

10.3 Metal Single Barrier Options

10.3.1 Description of Instrument
Barrier options on metal products are very similar to barrier options on FX
products.

10.3.2 XML Representation

XML Schema
RAZOR uses the FPML FXBarrierOption schema to represent single barrier
options on metal products.

Metal Single Barrier Option Example

<fxBarrierOption>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>
<expiryDate>2002-02-06</expiryDate>
<hourMinuteTime>10:00</hourMinuteTime>
<businessCenter>AUSY</businessCenter>
</expiryDateTime>
<exerciseStyle>EUROPEAN</exerciseStyle>
<fxOptionPremium>
<payerPartyReference href="ITE"/>
<receiverPartyReference href="BAN"/>
<premiumAmount type="Money">
<currency>AUD</currency>
<amount>50000</amount>
</premiumAmount>
<premiumSettlementDate>2001-11-06</premiumSettlementDate>
</fxOptionPremium>
<valueDate>2002-02-08</valueDate>
<putCurrencyAmount>
<currency>AUD</currency>
<amount>1000000</amount>
</putCurrencyAmount>
<callCurrencyAmount>
<currency>XAU</currency>
<amount>1852</amount>
</callCurrencyAmount>
<fxStrikePrice>
<rate>539.9568</rate>
<strikeQuoteBasis>CALLCURRENCYPERPUTCURRENCY</strikeQuoteBasis>
</fxStrikePrice>

©2009 Razor Risk Technologies Page 139 of 373

Razor Financial Principals

<spotRate>533.6190</spotRate>
<fxBarrier>
<fxBarrierType>KNOCKIN</fxBarrierType>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>XAU</currency2>
<quoteBasis>CURRENCY1PERCURRENCY2</quoteBasis>
</quotedCurrencyPair>
<triggerRate>542.20</triggerRate>
<informationSource>
<rateSource>Reuters</rateSource>
<rateSourcePage>XAU=</rateSourcePage>
</informationSource>
</fxBarrier>
</fxBarrierOption>

10.3.3 Pricing
Let
ln( S / X ) ln( S / H )
x1 = + ( µ + 1)σ T x2 = + ( µ + 1)σ T
σ T σ T
ln( H 2 / SX ) ln( H / S )
y1 = + ( µ + 1)σ T y2 = + ( µ + 1)σ T
σ T σ T
ln( H / S ) b −σ 2 / 2 2r
z= + λσ T µ= λ = µ2 +
σ T σ 2
σ2
A = φSe ( b − r )T N (φx1 ) − φXe − rT N (φx1 − φσ T )

B = φSe ( b − r )T N (φx2 ) − φXe − rT N (φx2 − φσ T )

H 2( µ +1) H
C = φSe(b − r )T ( ) N (ηy1 ) − φXe − rT ( ) 2 µ N (ηy1 − ησ T )
S S
H 2 ( µ +1) H
D = φSe(b − r )T ( ) N (ηy2 ) − φXe − rT ( ) 2 µ N (ηy2 − ησ T )
S S
H 2µ
E = Ke − rT ( N (ηx2 − ησ T ) − ( ) N (ηy2 − ησ T ))
S
H µ +λ H
F = K (( ) N (ηζ ) + ( ) µ − λ N (ηζ − 2ηλσ T ))
S S
c di ( X > H ) = C + E ,η = 1, φ = 1

c di ( X < H ) = A − B + D + E ,η = 1, φ = 1
cui ( X > H ) = A + E ,η = −1, φ = 1

cui ( X < H ) = B − C + D + E ,η = −1, φ = 1

p di ( X > H ) = B − C + D + E ,η = 1, φ = −1
p di ( X < H ) = A + E ,η = 1, φ = −1

©2009 Razor Risk Technologies Page 140 of 373

Razor Financial Principals

pui ( X > H ) = A − B + D + E ,η = −1, φ = −1

pui ( X < H ) = C + E ,η = −1, φ = −1

c do ( X > H ) = A − C + F ,η = 1, φ = 1
c do ( X < H ) = B − D + F ,η = 1, φ = 1

cuo ( X > H ) = F ,η = −1, φ = 1

cuo ( X < H ) = A − B + C − D + F ,η = −1, φ = 1

p do ( X > H ) = A − B + C − D + F ,η = 1, φ = −1

p do ( X < H ) = F ,η = 1, φ = −1
puo ( X > H ) = B − D + F ,η = −1, φ = −1

puo ( X < H ) = A − C + F ,η = −1, φ = −1

10.4 Metal Double Barrier Option

10.4.1 Description of Instrument
Double barrier metal options are similar to FX double barrier options, where
the barrier levels are upper and lower barriers on the metal price.

10.4.2 XML Representation

XML Schema
This product uses the fpmlFXBarrierOptionSchema.

Metal Double Barrier Option Example

<fxBarrierOption>
<productType>DOUBLEBARRIER</productType>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>
<expiryDate>2002-03-04</expiryDate>
<hourMinuteTime>10:00</hourMinuteTime>
<businessCenter>AUSY</businessCenter>
</expiryDateTime>
<exerciseStyle>EUROPEAN</exerciseStyle>
<fxOptionPremium>
<payerPartyReference href="ITE"/>
<receiverPartyReference href="BAN"/>
<premiumAmount type="Money">
<currency>AUD</currency>
<amount>50000</amount>
</premiumAmount>
<premiumSettlementDate>2002-01-07</premiumSettlementDate>
</fxOptionPremium>

©2009 Razor Risk Technologies Page 141 of 373

Razor Financial Principals

<valueDate>2002-03-06</valueDate>
<putCurrencyAmount>
<currency>AUD</currency>
<amount>1000000</amount>
</putCurrencyAmount>
<callCurrencyAmount>
<currency>XAU</currency>
<amount>1852</amount>
</callCurrencyAmount>
<fxStrikePrice>
<rate>539.9568</rate>
<strikeQuoteBasis>PUTCURRENCYPERCALLCURRENCY</strikeQuoteBasis>
</fxStrikePrice>
<spotRate>538.9789</spotRate>
<fxBarrier>
<fxBarrierType>KNOCKOUT</fxBarrierType>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>XAU</currency2>
<quoteBasis>CURRENCY1PERCURRENCY2</quoteBasis>
</quotedCurrencyPair>
<triggerRate>541.354</triggerRate>
<informationSource>
<rateSource>Reuters</rateSource>
<rateSourcePage>XAU=</rateSourcePage>
</informationSource>
</fxBarrier>
<fxBarrier>
<fxBarrierType>KNOCKOUT</fxBarrierType>
<quotedCurrencyPair>
<currency1>AUD</currency1>
<currency2>XAU</currency2>
<quoteBasis>CURRENCY1PERCURRENCY2</quoteBasis>
</quotedCurrencyPair>
<triggerRate>534.55</triggerRate>
<informationSource>
<rateSource>Reuters</rateSource>
<rateSourcePage>XAU=</rateSourcePage>
</informationSource>
</fxBarrier>
</fxBarrierOption>

10.4.3 Pricing
The value of a call double barrier can be expressed as follows:

ln(SU 2 n / XL2 n ) + (b + σ 2 / 2)T

d1 =
σ T
ln( SU 2 n / FL2 n ) + (b + σ 2 / 2)T
d2 =
σ T

©2009 Razor Risk Technologies Page 142 of 373

Razor Financial Principals

ln( L2 n + 2 / XSU 2 n ) + (b + σ 2 / 2)T

d3 =
σ T
ln( L2 n+ 2 / FSU 2 n ) + (b + σ 2 / 2)T
d4 =
σ T
2(b − δ 2 − n(δ1 − δ 2 ))
µ1 = +1
σ2
(δ1 − δ 2 )
µ 2 = 2n
σ2
2(b − δ 2 + n(δ1 − δ 2 ))
µ3 = +1
σ2
F = Ueδ1T
∞  Un L Ln +1 
c = Se (b − r )T ∑n = −∞ ( n ) µ1 ( ) µ 2 ( N ( d1 ) − N ( d 2 )) − ( n ) µ 3 ( N ( d 3 ) − N ( d 4 ))  −
 L S U S 
n +1
 U µ1 − 2 L µ 2
n
L µ −2 
∞ ( n ) ( ) ( N ( d1 − σ T ) − N ( d 2 − σ T )) − ( n ) 3 
Xe ∑n = −∞  L
− rT
S U S 
( N (d − σ T ) − N (d − σ T )) 
 3 4 
The value of a double barrier put option can be expressed as follows:

ln(SU 2 n / EL2 n ) + (b + σ 2 / 2)T

y1 =
σ T
ln(SU 2 n / XL2 n ) + (b + σ 2 / 2)T
y2 =
σ T
ln(L2 n + 2 / ESU 2 n ) + (b + σ 2 / 2)T
y3 =
σ T
ln(L2 n+ 2 / XSU 2 n ) + (b + σ 2 / 2)T
y4 =
σ T
E = Le δ 2T
 U n µ1 − 2 L µ 2 Ln +1 µ − 2 
( n ) ( ) ( N ( y1 − σ T ) − N ( y2 − σ T )) − ( n ) 3 ( N ( y3 − σ T ) − 
 L S U S −
∞
p = Xe − rT ∑ n = −∞  N ( y − σ T )) 
 4 
∞  Un L Ln +1 
Se( b − r )T ∑ n = −∞ ( n ) µ1 ( ) µ 2 ( N ( y1 ) − N ( y2 )) − ( n ) µ 3 ( N ( y3 ) − N ( y4 ))
 L S U S 

©2009 Razor Risk Technologies Page 143 of 373

Chapter 11
Commodities and Energy

11.1 Commodity Forwards

11.1.1 Description of Instrument
Commodity forwards are forward agreements to buy or sell a commodity at a
specific price. These types of agreements are modelled in RAZOR in a similar
manner to FX forwards.
Commodity products such as oil have standard units of measurement. This is
encapsulated in the asset ID.
The settlement price can be a singe-date price or an average over a pre-
specified period. If the contract is to be settled against an average price, the
Accumulated Average needs to be an input from the Source System.

11.1.2 XML Representation

Commodity Forward Example
<fxleg>
<exchangedCurrency1>
<payerPartyReference href = "ITE"/>
<receiverPartyReference href = "BAN"/>
<paymentAmount>
<currency>AUD</currency>
<amount>10000000</amount>
</paymentAmount>
</exchangedCurrency1>
<exchangedCurrency2>
<payerPartyReference href = "BAN"/>
<receiverPartyReference href = "ITE"/>
<paymentAmount>
<currency>OILBRENT</currency>
<amount>50000</amount>
</paymentAmount>
</exchangedCurrency2>
<valueDate>2003-12-21</valueDate>
<exchangeRate>
<quotedCurrencyPair>
<currency1>OILBRENT</currency1>
<currency2>AUD</currency2>
<quoteBasis>CURRENCY1PERCURRENCY2</quoteBasis>
</quotedCurrencyPair>
<rate>20</rate>
</exchangeRate>
</fxleg>
Razor Financial Principals

11.1.3 Pricing
The value of a Commodity Forward is calculated as follows:
V = DT N (F − X )
where:
V - commodity payer swap price
F − the projected forward commodity value derived from the underlying
commodity index forward curve.
X − commodity forward contract price.
DT - discount factor to the payment date of contract

N - contract notional.

11.2 Electricity Futures

11.2.1 Introduction
In Razor, it provides two methods to determine the price of electricity
futures. One method is to obtain the futures price from the forward price
curve while the other method is to use the market quoted price directly.

11.2.2 Definition
ft ( T ) = futures price at time t that matures at time T .
K = contract price.
df ( t1 , t2 ) = discount factor for the period ( t1 , t2 ) .

11.2.3 Valuation
The value of the electricity futures vt at time t is computed as
vt = df ( t , T ) × ( ft (T ) − K ) .

11.3 Commodity Vanilla Options

11.3.1 Description of Instrument
Vanilla commodity option gives its holder the right to buy (Call) or sell (Put) a
commodity at a predetermined price (strike) at a given date (maturity).

©2009 Razor Risk Technologies Page 145 of 373

Razor Financial Principals

Commodity vanilla options are treated in a similar manner to metal vanilla

options.

11.3.2 XML Representation

XML Schema
Commodity vanilla options use the FX vanilla specification of the FPML version
3 standard.

Commodity European Option Example

<fxSimpleOption>
<productType>Call Commodity Option</productType>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>
<expiryDate>2003-06-04</expiryDate>
<hourMinuteTime>1400</hourMinuteTime>
<businessCenter>AUSY</businessCenter>
<cutName>Sydney</cutName>
</expiryDateTime>
<exerciseStyle>European</exerciseStyle>
<fxOptionPremium>
<payerPartyReference href="ITE"/>
<receiverPartyReference href="BAN"/>
<premiumAmount type="Money">
<currency>AUD</currency>
<amount>1000</amount>
</premiumAmount>
<premiumSettlementDate>2002-12-06</premiumSettlementDate>
</fxOptionPremium>
<valueDate>2003-06-06</valueDate>
<putCurrencyAmount>
<currency>AUD</currency>
<amount>1000000</amount>
</putCurrencyAmount>
<callCurrencyAmount>
<currency>OILBRENT</currency>
<amount>50000</amount>
</callCurrencyAmount>
<fxStrikePrice>
<rate>20.00</rate>
<strikeQuoteBasis>PUTCURRENCYPERCALLCURRENCY</strikeQuoteBasis>
</fxStrikePrice>
</fxSimpleOption>

11.3.3 Pricing
Vanilla Put and Call commodity options are priced using the standard Black-
Scholes formula

Let

©2009 Razor Risk Technologies Page 146 of 373

Razor Financial Principals

S = the current spot price of the commodity

r = the continuously compounded interest rate
rf = the continuously compounded yield on the commodity
σ = the price volatility
X = the strike price
Tx = the time from today to expiry, annualised
Td = the time from spot to delivery, annualised
( r − r f )Td
F0 = S 0 e , where ( r − rf ) is the cost of carry

if the future price F0 of the commodity is provided, then the cost of carry is
zero.

 F  σ Tx
2
ln 0  +
d1 =  
X 2
σ Tx

 F  σ Tx
2
ln 0  −
d2 =  
X 2
σ Tx
The European call price is given by
c = e − rTd ( F0 N ( d1 ) − XN ( d 2 ))
The European put price is given by
p = e − rTd ( XN (− d 2 ) − F0 N (− d1 ))

Please refer to section 17.1.4 - generalised Black-Scholes option pricing

formula for further detail and the Greeks.

11.4 Commodity Single Barrier Options

11.4.1 Description of Instrument
Barrier options on commodities and energy are treated in a similar way to
barrier options on FX products.

11.4.2 XML Representation

XML Schema
RAZOR uses the FPML FXBarrierOption schema to represent single barrier
options on commodities.

©2009 Razor Risk Technologies Page 147 of 373

Razor Financial Principals

Commodity Single Barrier Option Example

<fxBarrierOption>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>
<expiryDate>2002-02-06</expiryDate>
<hourMinuteTime>10:00</hourMinuteTime>
</expiryDateTime>
<exerciseStyle>European</exerciseStyle>
<fxOptionPremium>
<payerPartyReference href="ITE"/>
<receiverPartyReference href="BAN"/>
<premiumAmount type="Money">
<currency>AUD</currency>
<amount>10000</amount>
</premiumAmount>
<premiumSettlementDate>2001-11-06</premiumSettlementDate>
</fxOptionPremium>
<valueDate>2002-02-08</valueDate>
<putCurrencyAmount>
<currency>AUD</currency>
<amount>1000000</amount>
</putCurrencyAmount>
<callCurrencyAmount>
<currency>OILBRENT</currency>
<amount>31736</amount>
</callCurrencyAmount>
<fxStrikePrice>
<rate>31.51</rate>
<strikeQuoteBasis>PUTCURRENCYPERCALLCURRENCY</strikeQuoteBasis>
</fxStrikePrice>
<spotRate>30</spotRate>
<fxBarrier>
<fxBarrierType>KNOCKIN</fxBarrierType>
<quotedCurrencyPair>
<currency1>OILBRENT</currency1>
<currency2>AUD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<triggerRate>34</triggerRate>
</fxBarrier>
</fxBarrierOption>

11.4.3 Pricing
Using the Reiner and Rubinstein Single Barrier Model
Let

©2009 Razor Risk Technologies Page 148 of 373

Razor Financial Principals

ln( S / X ) ln( S / H )
x1 = + ( µ + 1)σ T x2 = + ( µ + 1)σ T
σ T σ T
ln( H 2 / SX ) ln( H / S )
y1 = + ( µ + 1)σ T y2 = + ( µ + 1)σ T
σ T σ T
ln( H / S ) b −σ 2 / 2 2r
z= + λσ T µ= λ = µ2 +
σ T σ 2
σ2
A = φSe ( b − r )T N (φx1 ) − φXe − rT N (φx1 − φσ T )

B = φSe ( b − r )T N (φx2 ) − φXe − rT N (φx2 − φσ T )

H 2( µ +1) H
C = φSe( b − r )T ( ) N (ηy1 ) − φXe − rT ( ) 2 µ N (ηy1 − ησ T )
S S
H 2 ( µ +1) H
D = φSe( b − r )T ( ) N (ηy2 ) − φXe − rT ( ) 2 µ N (ηy2 − ησ T )
S S
H 2µ
E = Ke − rT ( N (ηx2 − ησ T ) − ( ) N (ηy2 − ησ T ))
S
H µ +λ H
F = K (( ) N (ηζ ) + ( ) µ − λ N (ηζ − 2ηλσ T ))
S S
c di ( X > H ) = C + E ,η = 1, φ = 1

c di ( X < H ) = A − B + D + E ,η = 1, φ = 1

cui ( X > H ) = A + E ,η = −1, φ = 1

cui ( X < H ) = B − C + D + E ,η = −1, φ = 1

p di ( X > H ) = B − C + D + E ,η = 1, φ = −1

p di ( X < H ) = A + E ,η = 1, φ = −1
pui ( X > H ) = A − B + D + E ,η = −1, φ = −1

pui ( X < H ) = C + E ,η = −1, φ = −1

c do ( X > H ) = A − C + F ,η = 1, φ = 1
c do ( X < H ) = B − D + F ,η = 1, φ = 1

cuo ( X > H ) = F ,η = −1, φ = 1

cuo ( X < H ) = A − B + C − D + F ,η = −1, φ = 1

p do ( X > H ) = A − B + C − D + F ,η = 1, φ = −1

p do ( X < H ) = F ,η = 1, φ = −1

puo ( X > H ) = B − D + F ,η = −1, φ = −1

puo ( X < H ) = A − C + F ,η = −1, φ = −1

©2009 Razor Risk Technologies Page 149 of 373

Razor Financial Principals

11.5 Commodity Double Barrier Option

11.5.1 Description of Instrument
Double barrier options on commodities and energy are treated in a similar way
to double barrier options on FX products.

11.5.2 XML Representation

XML Schema
This product uses the fpmlFXBarrierOptionSchema.

Commodity Double Barrier Option Example

<fxBarrierOption>
<productType>DOUBLEBARRIER</productType>
<buyerPartyReference href="ITE"/>
<sellerPartyReference href="BAN"/>
<expiryDateTime>
<expiryDate>2002-03-04</expiryDate>
<hourMinuteTime>10:00</hourMinuteTime>
</expiryDateTime>
<exerciseStyle>EUROPEAN</exerciseStyle>
<fxOptionPremium>
<payerPartyReference href="ITE"/>
<receiverPartyReference href="BAN"/>
<premiumAmount type = "Money">
<currency>AUD</currency>
<amount>10000</amount>
</premiumAmount>
<premiumSettlementDate>2002-01-07</premiumSettlementDate>
</fxOptionPremium>
<valueDate>2002-03-06</valueDate>
<putCurrencyAmount>
<currency>AUD</currency>
<amount>1000000</amount>
</putCurrencyAmount>
<callCurrencyAmount>
<currency>OILBRENT</currency>
<amount>31736</amount>
</callCurrencyAmount>
<fxStrikePrice>
<rate>31.51</rate>
<strikeQuoteBasis>PUTCURRENCYPERCALLCURRENCY</strikeQuoteBasis>
</fxStrikePrice>
<spotRate>30</spotRate>
<fxBarrier>
<fxBarrierType>KNOCKOUT</fxBarrierType>
<quotedCurrencyPair>
<currency1>OILBRENT</currency1>
<currency2>AUD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
©2009 Razor Risk Technologies Page 150 of 373
Razor Financial Principals

</quotedCurrencyPair>
<triggerRate>34</triggerRate>
</fxBarrier>
<fxBarrier>
<fxBarrierType>KNOCKOUT</fxBarrierType>
<quotedCurrencyPair>
<currency1>OILBRENT</currency1>
<currency2>AUD</currency2>
<quoteBasis>CURRENCY2PERCURRENCY1</quoteBasis>
</quotedCurrencyPair>
<triggerRate>27</triggerRate>
</fxBarrier>
</fxBarrierOption>

11.5.3 Pricing
The value of a call double barrier can be expressed as follows:

ln(SU 2 n / XL2 n ) + (b + σ 2 / 2)T

d1 =
σ T
ln( SU 2 n / FL2 n ) + (b + σ 2 / 2)T
d2 =
σ T
ln( L2 n + 2 / XSU 2 n ) + (b + σ 2 / 2)T
d3 =
σ T
ln( L2 n+ 2 / FSU 2 n ) + (b + σ 2 / 2)T
d4 =
σ T
2(b − δ 2 − n(δ1 − δ 2 ))
µ1 = +1
σ2
(δ1 − δ 2 )
µ 2 = 2n
σ2
2(b − δ 2 + n(δ1 − δ 2 ))
µ3 = +1
σ2
F = Ueδ1T
∞  Un L Ln +1 
c = Se ( b − r )T ∑n = −∞ ( n ) µ1 ( ) µ 2 ( N ( d1 ) − N ( d 2 )) − ( n ) µ 3 ( N ( d 3 ) − N ( d 4 ))  −
 L S U S 
n +1
 U µ1 − 2 L µ 2
n
L µ −2 
∞ ( n ) ( ) ( N ( d1 − σ T ) − N ( d 2 − σ T )) − ( n ) 3 
Xe ∑n = −∞  L
− rT
S U S 
( N (d − σ T ) − N (d − σ T )) 
 3 4 
The value of a double barrier put option can be expressed as follows:

ln(SU 2 n / EL2 n ) + (b + σ 2 / 2)T

y1 =
σ T

©2009 Razor Risk Technologies Page 151 of 373

Razor Financial Principals

ln(SU 2 n / XL2 n ) + (b + σ 2 / 2)T

y2 =
σ T
ln( L2 n + 2 / ESU 2 n ) + (b + σ 2 / 2)T
y3 =
σ T
ln( L2 n+ 2 / XSU 2 n ) + (b + σ 2 / 2)T
y4 =
σ T
E = Le δ 2T
 U n µ1 − 2 L µ 2 Ln +1 µ − 2 
( n ) ( ) ( N ( y1 − σ T ) − N ( y2 − σ T )) − ( n ) 3 ( N ( y3 − σ T ) − 
 L S U S −
∞
p = Xe − rT ∑ n = −∞  N ( y − σ T )) 
 4 
∞  Un L Ln +1 
Se( b − r )T ∑ n = −∞ ( n ) µ1 ( ) µ 2 ( N ( y1 ) − N ( y2 )) − ( n ) µ 3 ( N ( y3 ) − N ( y4 ))
 L S U S 

11.6 Commodity Average Rate Swap

11.6.1 Description of Instrument

Commodity Average Swap is represented as the stream of Commodity

Average Forward transactions on the same commodity with successive
delivery and maturity days that can have different notional amounts on
the floating side of the swap.

11.6.2 XML Representation

<commoditySwap>
<productType>Commodity</productType>
<commodityForward>
<commodityPhysicalSettlement>
<firmness>firm</firmness>
</commodityPhysicalSettlement>
<buyerParty href='EUROPE' />
<sellerParty href='DKW' />
<commodityBuyerPrice>
<formulaPrice>
<fixedPrice>
<precision>0</precision>
<currency>USD</currency>
<amount></amount>
<perUom></perUom>
</fixedPrice>
<indexPrice>
<margin>1.00</margin>
©2009 Razor Risk Technologies Page 152 of 373
Razor Financial Principals

<index>WTI</index>
<indexPct>1.0</indexPct>
<indexTenor>
<period>W</period>
<periodMultiplier>1</periodMultiplier>
</indexTenor>
<fixingDateOffset>
<period>M</period>
<periodMultiplier>-2</periodMultiplier>
<businessDayConvention>FOLLOWING</businessDayConvention>
<dayType>Business</dayType>
</fixingDateOffset>
</indexPrice>
<averagingTerms>
<averagingMethod>Unweighted</averagingMethod>
<averagingStartDate opt='y'>
<anchor>FixingDate</anchor>
<adjustment>optional not used</adjustment>
<offset>
<periodMultiplier>-2</periodMultiplier>
<period>D</period>
<businessDayConvention>FOLLOWING</businessDayConvention>
<dayType>Business</dayType>
</offset>
</averagingStartDate>
<averagingEndDate opt='y'>
<anchor>FixingDate</anchor>
<adjustment>optional not used</adjustment>
<offset>
<periodMultiplier>2</periodMultiplier>
<period>D</period>
<businessDayConvention>FOLLOWING</businessDayConvention>
<dayType>Business</dayType>
</offset>
</averagingEndDate>
</averagingTerms>
</formulaPrice>
</commodityBuyerPrice>
<commodityUnits>
<commodity>WTI</commodity>
<totalVolume>60000</totalVolume>
<totalUom>BBL</totalUom>
<volumeAmount>0</volumeAmount>
<volumeUom>X</volumeUom>
<volumePerFreq>TERM</volumePerFreq>
</commodityUnits>
<commodityDeliveryPeriod>
<startDate>2006-07-01</startDate>
<endDate>2006-08-01</endDate>
</commodityDeliveryPeriod>
©2009 Razor Risk Technologies Page 153 of 373
Razor Financial Principals

</commodityForward>
<commodityForward>
<commodityPhysicalSettlement>
<firmness>firm</firmness>
</commodityPhysicalSettlement>
<buyerParty href='DKW' />
<sellerParty href='EUROPE' />
<commodityBuyerPrice>
<formulaPrice>
<fixedPrice>
<precision>0</precision>
<currency>USD</currency>
<amount></amount>
<perUom></perUom>
</fixedPrice>
<indexPrice>
<margin>1.00</margin>
<index>WTI</index>
<indexPct>1.0</indexPct>
<indexTenor>
<period>M</period>
<periodMultiplier>1</periodMultiplier>
</indexTenor>
<fixingDateOffset>
<period>M</period>
<periodMultiplier>-2</periodMultiplier>
<businessDayConvention>FOLLOWING</businessDayConvention>
<dayType>Business</dayType>
</fixingDateOffset>
</indexPrice>
<averagingTerms>
<averagingMethod>Unweighted</averagingMethod>
<averagingStartDate opt='y'>
<anchor>FixingDate</anchor>
<adjustment>optional not used</adjustment>
<offset>
<periodMultiplier>-2</periodMultiplier>
<period>D</period>
<businessDayConvention>FOLLOWING</businessDayConvention>
<dayType>Business</dayType>
</offset>
</averagingStartDate>
<averagingEndDate opt='y'>
<anchor>FixingDate</anchor>
<adjustment>optional not used</adjustment>
<offset>
<periodMultiplier>2</periodMultiplier>
<period>D</period>
<businessDayConvention>FOLLOWING</businessDayConvention>
<dayType>Business</dayType>
©2009 Razor Risk Technologies Page 154 of 373
Razor Financial Principals

</offset>
</averagingEndDate>
</averagingTerms>
</formulaPrice>
</commodityBuyerPrice>
<commodityUnits>
<commodity>WTI</commodity>
<totalVolume>60000</totalVolume>
<totalUom>BBL</totalUom>
<volumeAmount>0</volumeAmount>
<volumeUom>X</volumeUom>
<volumePerFreq>TERM</volumePerFreq>
</commodityUnits>
<commodityDeliveryPeriod>
<startDate>2006-07-01</startDate>
<endDate>2006-08-01</endDate>
</commodityDeliveryPeriod>
</commodityForward>
</commoditySwap>

11.6.3 Pricing
The value of commodity swap contract can be calculated as the sum of
rolling commodity forward contracts values:

N M
V swap = ∑V
i =1
i = ∑ (F
i =1
i − X i )D T i N i ,

where:
Vswap - commodity payer swap price
M - number of legs in commodity swap
Vi - i’th leg commodity forward contract value
Fi − the projected forward commodity value derived from the underlying
commodity index forward curve.
X i − i’th leg commodity forward contract price. This can be the same for
all N legs of commodity swap.
DTi - discount factor to the payment date of i’th contract
N i - i’th contract notional.
Average price commodity swaps also referred to as Asian type contracts
are settled against the average of prices for an underlying commodity
over a period of time.
The value of the average rate commodity swap contract is represented
by:
N M
V swap = ∑
i =1
Vi = ∑ (A
i =1
i − X i )D T i
Ni

©2009 Razor Risk Technologies Page 155 of 373

Razor Financial Principals

with Ai defined as the floating arithmetic average rate accumulated

through the i’th leg averaging period.
Depending if the considered averaging period is pending or current the
average commodity rate is either derived from the underlying forward
curve adjusted by simulated or observed values.
For pending averaging period starting after the valuation date the
average rate is determined from the interpolated values derived from the
forward curve
K
1 i

Ai =
Ki
∑
j =1
F ti

K i - the number of required observations in i’th leg averaging period,

Fti - interpolated from forward curve projected observation at time t i

within i’th averaging period.

For current averaging period started before the valuation date the last
formula should be adjusted:
Ki
mA + ∑F
j = K i − m +1
ti

Ai =
m + Ki
where A is current accumulated average and m is the number of
past observations.
The XML presentation of commodity swaps and forwards allows definition
of the averaging period schedule of each leg.
Average price options, also referred to as Asian options or average rate
options, are settled against the average of prices for an underlying
commodity over a period of time.
Average price options are financially settled upon expiration and cannot
be exercised into the underlying futures contract.

11.7 Commodity Average Rate Options

11.7.1 Description of Instrument
Average price options, also referred to as Asian options or average rate
options, are settled against the average of prices for an underlying
commodity over a period of time. These instruments have become
popular in the over-the-counter markets during the past decade as a way
of dampening market volatility.

Average price options are financially settled upon expiration and cannot
be exercised into the underlying futures contract.

11.7.2 XML Presentation

©2009 Razor Risk Technologies Page 156 of 373

Razor Financial Principals

<commodityOption>
<commodityUnderlying>
<commodityPhysicalSettlement>
<firmness>firm</firmness>
</commodityPhysicalSettlement>
<buyerParty href='NTHAMERICA' />
<sellerParty href='BARCAP' />
<commodityBuyerPrice>
<formulaPrice>
<indexPrice>
<index>BCO</index>
<indexPct>1</indexPct>
<margin>0</margin>
<indexMethod>Defined by delivery</indexMethod>
<fixingDateOffset>
<dayType>Business</dayType>
<periodMultiplier>-1</periodMultiplier>
<period>M</period>
<businessDayConvention>Business</businessDayConvention>
</fixingDateOffset>
</indexPrice>
<fixedPrice>
<precision>0</precision>
<currency>USD</currency>
<amount>0</amount>
<perUom>BBL</perUom>
</fixedPrice>
<averagingTerms>
<averagingMethod>Arithmetic</averagingMethod>
<averagingStartDate>
<offset>
<dayType>Business</dayType>
<periodMultiplier>-2</periodMultiplier>
<period>D</period>
<businessDayConvention>Business</businessDayConvention>
</offset>
<anchor>FixingDate</anchor>
</averagingStartDate>
<averagingEndDate>
<offset>
<dayType>Business</dayType>
<periodMultiplier>2</periodMultiplier>
<period>D</period>
<businessDayConvention>Business</businessDayConvention>
</offset>
<anchor>FixingDate</anchor>
</averagingEndDate>
</averagingTerms>
</formulaPrice>
</commodityBuyerPrice>
©2009 Razor Risk Technologies Page 157 of 373
Razor Financial Principals

<commodityUnits>
<commodity>BCO</commodity>
<totalVolume>100000</totalVolume>
<totalUom>BBL</totalUom>
<volumeAmount>0</volumeAmount>
<volumeUom>X</volumeUom>
<volumePerFreq>TERM</volumePerFreq>
</commodityUnits>
<commodityDeliveryPeriod>
<startDate>2006-06-11</startDate>
<endDate>2006-07-11</endDate>
</commodityDeliveryPeriod>
</commodityUnderlying>
<productType>Commodity</productType>
<buyerParty href='NTHAMERICA' />
<sellerParty href='BARCAP' />
<payoutFormula>AverageRate</payoutFormula>
<optionType>CALL</optionType>
<strike>
<strikePrice>70</strikePrice>
<buyer>NTHAMERICA</buyer>
<seller>BARCAP</seller>
</strike>
</commodityOption>

11.7.3 Pricing
The Levy Approximation formula is used for pricing arithmetic average-
rate commodity option.
C asian ≈ S A N ( d1 ) − Xe − rT2 N ( d 2 )
where
C asian - average-rate commodity call option
X – is the commodity option strike price,
S A - arithmetic average of the known commodity price fixing and
SA =
S
T *b
( )
e ( b − r )T2 − e − rT2 ,
S − representing commodity spot price, r - risk free interest rate, b –
commodity cost of carry rate, T2 -remaining time to option maturity from
valuation date and T – original time to option maturity.
Other components of Levy approximation formula calculated according
to:
1  ln( D) 
d1 =  − ln( X * )  and d 2 = d 1 − V
V  2 
The last expressions use following notations:
T − T2
D = 2 , V = ln(D) − 2[rT2 + ln(S A )] and X * = X −
M
SA .
T T

©2009 Razor Risk Technologies Page 158 of 373

Razor Financial Principals

2 S 2  e ( 2b +σ )T2 − 1 e bT2 − 1
2

M =  − .
b + σ 2  2b + σ 2 b 

©2009 Razor Risk Technologies Page 159 of 373

Chapter 12
Interest Rate Derivatives

12.1 Forward Rate Agreements

12.1.1 Description of Instrument
A FRA is the right to buy or sell a short-term money market instrument at some
date in the future. It is essentially a contract that locks in a forward interest
rate for a counterparty.

12.1.2 XML Representation

FRAs are represented in FPML version 3.

fpmlFRA Schema
Name: Type Occurs Size Description
adjustedEffectiveDate
1..1 The start date of the FRA
fpmlAdjustedDate
adjustedTerminationDate
0..1 The end date of the FRA
fpmlAdjustedDate
paymentDate The date the payment occurs on the
1..1
fpmlUnadjustedDate FRA
fixingDateOffset The date offset that the floating
1..1
fpmlRelativeDateOffset rate is fixed
dayCountFraction The day count fraction used to
0..1 10
string calculate the payment amount
calculationPeriodNumberOfDays 0..1 The number of days in the
int calculation period
notional
1..1 The notional face value of the FRA
fpmlAmount
fixedRate
1..1 The fixed rate of the FRA
double
floatingRateIndex The floating rate index to use as a
0..1 50
string reference rate
indexTenor
1..1 The tenor of the FRA
fpmlTenor
fraDiscounting Whether discounting is applied to
1..1
boolean the FRA

12.1.3 Pricing
There are essentially two formulas for pricing FRAs depending on the market in
which the FRA is traded — the BBA formula which is typically used for FRAs
traded outside of Australia, and the AFMA formula.
Razor Financial Principals

The AFMA formula is as follows:

Let
F fra = The notional face value of the FRA

r fra = The agreed FRA rate

rm = The market benchmark rate, the forward rate derived from the index
curve
rzero = the zero coupon rate from pricing date to settlement

rmat = the zero coupon rate from pricing date to maturity

d fra = number of days from settlement to maturity

d fwd = number of days from valuation to settlement

D = number of days in the year
V fra = FRA exposure
Then

( 1+frar frafrad frafra/ D ) − ( 1+frarmmd frafra/ D )

F r d /D F r d /D

V fra =
1 + rzero d fwd / D
The BBA formula for pricing FRAs is:
( F fra r fra d fra / D ) − ( F fra rm d fra / D )
V fra =
1 + rmat d fra / D

12.2 Interest Rate and Cross Currency Swaps

12.2.1 Description of Instrument
A swap is an agreement whereby two counterparties exchange periodic interest
payments. In the case of a cross currency swap, these period payments are
denominated in different currencies.

12.2.2 XML Representation

Interest Rate Swaps are represented in FPML version 3.

fpmlSwap Schema

Name: Type Description

©2009 Razor Risk Technologies Page 161 of 373

Razor Financial Principals

fpmlSwap Schema

Name: Type Description

The fpmlInterestRateStream
swapStream
structure that defines the interest
fpmlInterestRateStream
rate flows for the swap
<swap>
<swapStream>
<payerPartyReference href="ITE" />
<receiverPartyReference href="BAN" />
<calculationPeriodDates id="CalcPeriodDates0">
<effectiveDate id="">i
<unadjustedDate>2003-05-29</unadjustedDate>
</effectiveDate>
<terminationDate id="">
<unadjustedDate>2004-05-31</unadjustedDate>
</terminationDate>
<calculationPeriodDatesAdjustments>
<businessDayConvention>MODFOLLOWING</businessDayConvention>
</calculationPeriodDatesAdjustments>
<calculationPeriodFrequency>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
<rollConvention>IMM</rollConvention>
</calculationPeriodFrequency>
</calculationPeriodDates>
<paymentDates>
<calculationPeriodDatesReference href="#CalcPeriodDates0" />
<paymentFrequency>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
</paymentFrequency>
<payRelativeTo id="">CalculationPeriodEndDate</payRelativeTo>
<paymentDatesAdjustments>
<businessDayConvention>MODFOLLOWING</businessDayConvention>
</paymentDatesAdjustments>
</paymentDates>
<resetDates>
<calculationPeriodDatesReference href="#CalcPeriodDates0" />
<resetRelativeTo id="">CalculationPeriodStartDate</resetRelativeTo>
<fixingDates>
<periodMultiplier>-2</periodMultiplier>
<period>D</period>
<businessDayConvention>NONE</businessDayConvention>
<dateRelativeTo id="#resetDates0">ResetDate</dateRelativeTo>
</fixingDates>
<resetFrequency>
©2009 Razor Risk Technologies Page 162 of 373
Razor Financial Principals

<periodMultiplier>6</periodMultiplier>
<period>M</period>
</resetFrequency>
<resetDatesAdjustments>
<businessDayConvention>MODFOLLOWING</businessDayConvention>
</resetDatesAdjustments>
</resetDates>
<calculationPeriodAmount>
<calculation>
<notionalSchedule>
<notionalStepSchedule>
<initialValue>1000000.000000</initialValue>
<currency>AUD</currency>
</notionalStepSchedule>
</notionalSchedule>
<fixedRateSchedule>
<initialValue>0.000500</initialValue>
</fixedRateSchedule>
<dayCountFraction>ACT/365</dayCountFraction>
</calculation>
</calculationPeriodAmount>
<principalExchanges>
<initialExchange>false</initialExchange>
<finalExchange>false</finalExchange>
<intermediateExchange>false</intermediateExchange>
</principalExchanges>
</swapStream>
<swapStream>
<payerPartyReference href="BAN" />
<receiverPartyReference href="ITE" />
<calculationPeriodDates id="CalcPeriodDates0">
<effectiveDate id="">
<unadjustedDate>2003-05-29</unadjustedDate>
</effectiveDate>
<terminationDate id="">
<unadjustedDate>2004-05-31</unadjustedDate>
</terminationDate>
<calculationPeriodDatesAdjustments>
<businessDayConvention>MODFOLLOWING</businessDayConvention>
</calculationPeriodDatesAdjustments>
<calculationPeriodFrequency>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
<rollConvention>IMM</rollConvention>
</calculationPeriodFrequency>
</calculationPeriodDates>
<paymentDates>
<calculationPeriodDatesReference href="#CalcPeriodDates0" />
<paymentFrequency>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
</paymentFrequency>
<payRelativeTo id="">CalculationPeriodEndDate</payRelativeTo>
<paymentDatesAdjustments>
©2009 Razor Risk Technologies Page 163 of 373
Razor Financial Principals

<businessDayConvention>MODFOLLOWING</businessDayConvention>
</paymentDatesAdjustments>
</paymentDates>
<resetDates>
<calculationPeriodDatesReference href="#CalcPeriodDates0" />
<resetRelativeTo id="">CalculationPeriodStartDate</resetRelativeTo>
<fixingDates>
<periodMultiplier>-2</periodMultiplier>
<period>D</period>
<businessDayConvention>NONE</businessDayConvention>
<dateRelativeTo id="#resetDates0">ResetDate</dateRelativeTo>
</fixingDates>
<resetFrequency>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
</resetFrequency>
<resetDatesAdjustments>
<businessDayConvention>MODFOLLOWING</businessDayConvention>
</resetDatesAdjustments>
</resetDates>
<calculationPeriodAmount>
<calculation>
<notionalSchedule>
<notionalStepSchedule>
<initialValue>1000000.000000</initialValue>
<currency>AUD</currency>
</notionalStepSchedule>
</notionalSchedule>
<floatingRateCalculation>
<floatingRate>
<floatingRateIndex>BBSW</floatingRateIndex>
<indexTenor>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
</indexTenor>
<spreadSchedule>
<initialValue>0.000000</initialValue>
</spreadSchedule>
</floatingRate>
<averagingMethod></averagingMethod>
<negativeInterestRateTreatment></negativeInterestRateTreatment>
</floatingRateCalculation>
<dayCountFraction>ACT/365.FI</dayCountFraction>
</calculation>
</calculationPeriodAmount>
<principalExchanges>
<initialExchange>false</initialExchange>
<finalExchange>false</finalExchange>
<intermediateExchange>false</intermediateExchange>
</principalExchanges>
</swapStream>
</swap>

©2009 Razor Risk Technologies Page 164 of 373

Razor Financial Principals

12.2.3 Pricing
Pricing interest rate swaps is a matter of present-valuing both legs of the
swap. The formula presented below gives an example of pricing a vanilla fixed-
floating swap. The pricing model is very general however and will price
floating-floating and cross currency swaps.
Let
F float = the face value of the floating leg of the swap

F fixed = the face value of the fixed leg of the swap

r fixed = the coupon rate of the swap

r j −1→ j = the implied forward rate of the swap

p float = the floating payment periods per year

p fixed = the fixed payment periods per year

Vswap = the exposure of the swap

df j = the discount factor from time 0 → j

Then
n rfixed n rj −1→ j
Vswap = ∑ F fixed dfi + ∑ F float df j
i =1 p fixed j =1 p float

12.2.4 Other Variations of Swaps

Averaging Swaps
These contracts are similar to normal swaps except that the cashflows of the
floating leg are based on the average of some index rates.
Compounding Swaps
These contracts are swaps that the cashflows of the floating leg are based on
some index rates compounded over the accruing period.
Overnight Index Swap (OIS)
In a OIS transaction, the floating rate used to determine the floating leg
cashflows is based on an overnight rate that is reset daily. This typically is the
interbank overnight or call interest rate.

Pricing
The pricing of these variations of swaps are the same as the conventional
interest rate swaps, i.e. the present value of the difference between the
floating leg cashflow value and fixed leg cashflow value. The required unknown
floating rates used to determine the floating leg cashflows can always be
obtained from the forward rate curve as in the conventional swap case.

©2009 Razor Risk Technologies Page 165 of 373

Razor Financial Principals

12.3 Interest Rate Futures

12.3.1 Description of Instrument
There are three types of interest rate futures supported in Razor, namely, 30
day interbank cash rate futures, 90 day bank bill futures and 3 and 10 year
bond futures.
1. 30 Day Interbank Cash Rate Futures
These contracts are based on the interbank overnight cash rate published
by the Reserve Bank of Australia and allow users to hedge against
fluctuations in the overnight cash rate and better manage their daily cash
exposures. These contracts allow the buyer to fix the cash rate in a
specified future date for 30 days.
2. 90 Day Bank Bill Futures
These contracts are the major short term interest rate derivative
products.These contracts allow the holder to enter into a 90 day bank bill
at a predetermined price in a specified future date.
3. 3 and 10 Year Bond Futures
These contracts are the major medium and long term interest rate
derivative products in Australia.These contracts allow the buyer to enter
into a 3 or 10 year bond at a predetermined price in a specified future
date.
Note Razor supports futures contracts in all major economies and different
maturity dates for bill futures and bond futures are also supported.

12.3.2 XML Representation

Bill Futures
<deal>
<trade>
<tradeHeader>
<tradeId>FUT:10909</tradeId>
<tradeDate>2007-09-12</tradeDate>
<tradeType>SFBAB</tradeType>
<dealer>84751</dealer>
<counterparty>83976</counterparty>
<internalUnit>GTYA</internalUnit>
<buySell>BUY</buySell>
<status>OPEN</status>
<location>Bank1</location>
<domicileCountry>154</domicileCountry>
</tradeHeader>
<extensions>
<extension>
<value>ValueByMTMYieldCurve</value>
<name>FuturesValueStyle</name>
</extension>

©2009 Razor Risk Technologies Page 166 of 373

Razor Financial Principals

<extension>
<value>true</value>
<name>DiscountFuturesValueToValueDate</name>
</extension>
</extensions>
<product>
<irFuture>
<productType>BankBillFuture</productType>
<instrumentId>SFBAB</instrumentId>
<settlementDate>
<unadjustedDate>2007-12-13</unadjustedDate>
</settlementDate>
<currency>AUD</currency>
<exchangeCode>SFE</exchangeCode>
<numberContracts>500</numberContracts>
<tradedPrice>93.27</tradedPrice>
<priceQuote>FuturesPrice</priceQuote>
<bond>
<bondStream>
<payerPartyReference href="83976" />
<receiverPartyReference href="GTY" />
<calculationPeriodDates id="CalcPeriodDates0">
<effectiveDate>
<unadjustedDate>2007-12-13</unadjustedDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
</effectiveDate>
<terminationDate>
<unadjustedDate>2008-03-12</unadjustedDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
</terminationDate>
<calculationPeriodDatesAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</calculationPeriodDatesAdjustments>
<firstPeriodStartDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
<unadjustedDate>2007-12-13</unadjustedDate>
</firstPeriodStartDate>
<calculationPeriodFrequency>
<rollConvention>NONE</rollConvention>
<periodMultiplier>90</periodMultiplier>
©2009 Razor Risk Technologies Page 167 of 373
Razor Financial Principals

<period>D</period>
</calculationPeriodFrequency>
</calculationPeriodDates>
<paymentDates>
<calculationPeriodDatesReference href="CalcPeriodDates0" />
<paymentFrequency>
<period>D</period>
<periodMultiplier>90</periodMultiplier>
</paymentFrequency>
<payRelativeTo>CalculationPeriodEndDate</payRelativeTo>
<paymentDatesAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</paymentDatesAdjustments>
</paymentDates>
<calculationPeriodAmount>
<calculation>
<notionalSchedule>
<notionalStepSchedule>
<currency>AUD</currency>
<initialValue>1000000</initialValue>
</notionalStepSchedule>
</notionalSchedule>
<fixedRateSchedule>
<initialValue>0</initialValue>
</fixedRateSchedule>
<dayCountFraction>ACT/365</dayCountFraction>
</calculation>
</calculationPeriodAmount>
<principalExchanges>
<intermediateExchange>false</intermediateExchange>
<initialExchange>false</initialExchange>
<finalExchange>true</finalExchange>
</principalExchanges>
</bondStream>
<issuer href="Bank1" />
<securityId>DEC07</securityId>
<position>1</position>
<paymentAmount>
<paymentType>SETTLEMENT</paymentType>
<receiverPartyReference href="GTY" />
<payerPartyReference href="83976" />
<paymentAmount>
<currency>AUD</currency>
<amount>0</amount>
</paymentAmount>
<adjustedPaymentDate>2007-12-13</adjustedPaymentDate>
©2009 Razor Risk Technologies Page 168 of 373
Razor Financial Principals

</paymentAmount>
<exInterestDays>0</exInterestDays>
</bond>
</irFuture>
</product>
</trade>
<dealHeader>
<dealId>FUT:10909</dealId>
<dealType>SFBAB</dealType>
<dealDate>2007-09-12</dealDate>
<status>OPEN</status>
</dealHeader>
</deal>

Bond Futures
<deal>
<trade>
<tradeHeader>
<tradeId>FUT:10910</tradeId>
<tradeDate>2007-09-12</tradeDate>
<tradeType>SFYTB</tradeType>
<dealer>84751</dealer>
<counterparty>83976</counterparty>
<internalUnit>CFXA</internalUnit>
<buySell>BUY</buySell>
<status>OPEN</status>
<location>Bank1</location>
<domicileCountry>154</domicileCountry>
</tradeHeader>
<extensions>
<extension>
<value>ValueByMTMYieldCurve</value>
<name>FuturesValueStyle</name>
</extension>
<extension>
<value>true</value>
<name>DiscountFuturesValueToValueDate</name>
</extension>
</extensions>
<product>
<irFuture>
<productType>BondFuture</productType>
<instrumentId>SFYTB</instrumentId>
<settlementDate>
<unadjustedDate>2007-09-15</unadjustedDate>
</settlementDate>

©2009 Razor Risk Technologies Page 169 of 373

Razor Financial Principals

<currency>AUD</currency>
<exchangeCode>SFE</exchangeCode>
<numberContracts>500</numberContracts>
<tradedPrice>93.61</tradedPrice>
<priceQuote>FuturesPrice</priceQuote>
<bond>
<bondStream>
<payerPartyReference href="83976" />
<receiverPartyReference href="CFX" />
<calculationPeriodDates id="CalcPeriodDates0">
<effectiveDate>
<unadjustedDate>2007-09-15</unadjustedDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
</effectiveDate>
<terminationDate>
<unadjustedDate>2010-09-15</unadjustedDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
</terminationDate>
<calculationPeriodDatesAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</calculationPeriodDatesAdjustments>
<firstPeriodStartDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
<unadjustedDate>2007-09-15</unadjustedDate>
</firstPeriodStartDate>
<calculationPeriodFrequency>
<rollConvention>NONE</rollConvention>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
</calculationPeriodFrequency>
</calculationPeriodDates>
<paymentDates>
<calculationPeriodDatesReference href="CalcPeriodDates0" />
<paymentFrequency>
<period>M</period>
<periodMultiplier>6</periodMultiplier>
</paymentFrequency>
<payRelativeTo>CalculationPeriodEndDate</payRelativeTo>
<paymentDatesAdjustments>
<businessDayConvention>NONE</businessDayConvention>
©2009 Razor Risk Technologies Page 170 of 373
Razor Financial Principals

</paymentDatesAdjustments>
</paymentDates>
<calculationPeriodAmount>
<calculation>
<notionalSchedule>
<notionalStepSchedule>
<currency>AUD</currency>
<initialValue>100000</initialValue>
</notionalStepSchedule>
</notionalSchedule>
<fixedRateSchedule>
<initialValue>0.06</initialValue>
</fixedRateSchedule>
<dayCountFraction>ACT/360</dayCountFraction>
</calculation>
</calculationPeriodAmount>
<principalExchanges>
<intermediateExchange>false</intermediateExchange>
<initialExchange>false</initialExchange>
<finalExchange>true</finalExchange>
</principalExchanges>
</bondStream>
<issuer href="66096" />
<securityId>FUT-3YR-SEP7</securityId>
<position>1</position>
<paymentAmount>
<paymentType>SETTLEMENT</paymentType>
<receiverPartyReference href="CFX" />
<payerPartyReference href="83976" />
<paymentAmount>
<currency>AUD</currency>
<amount>0</amount>
</paymentAmount>
<adjustedPaymentDate>2007-09-15</adjustedPaymentDate>
</paymentAmount>
<exInterestDays>0</exInterestDays>
</bond>
</irFuture>
</product>
</trade>
<dealHeader>
<dealId>FUT:10910</dealId>
<dealType>SFYTB</dealType>
<dealDate>2007-09-12</dealDate>
<status>OPEN</status>
</dealHeader>
©2009 Razor Risk Technologies Page 171 of 373
Razor Financial Principals

</deal>

Cash Rate Futures

<deal>
<trade>
<tradeHeader>
<tradeId>FUT:10912</tradeId>
<tradeDate>2007-09-12</tradeDate>
<tradeType>SFIBC</tradeType>
<dealer>84751</dealer>
<counterparty>83976</counterparty>
<internalUnit>GTYA</internalUnit>
<buySell>SELL</buySell>
<status>OPEN</status>
<location>Bank1</location>
<domicileCountry>154</domicileCountry>
</tradeHeader>
<extensions>
<extension>
<value>HISTORICAL</value>
<name>HistoricalRateSet</name>
</extension>
<extension>
<value>CASHFUTURE_0M</value>
<name>HistoricalRateClass</name>
</extension>
<extension>
<value>ValueByMTMForwardCurve</value>
<name>FuturesValueStyle</name>
</extension>
<extension>
<value>true</value>
<name>DiscountFuturesValueToValueDate</name>
</extension>
</extensions>
<product>
<irFuture>
<productType>CashRateFuture</productType>
<instrumentId>SFIBC</instrumentId>
<description>30 DAY I/BANK CASH RATE FUTURES</description>
<settlementDate>
<unadjustedDate>2007-10-31</unadjustedDate>
</settlementDate>
<currency>AUD</currency>
<exchangeCode>SFE</exchangeCode>
<numberContracts>400</numberContracts>

©2009 Razor Risk Technologies Page 172 of 373

Razor Financial Principals

<tradedPrice>93.67</tradedPrice>
<priceQuote>FuturesPrice</priceQuote>
<fixedTickValue>24.66</fixedTickValue>
<fixedTickUnit>1</fixedTickUnit>
<bond>
<bondStream>
<payerPartyReference href="83976" />
<receiverPartyReference href="GTY" />
<calculationPeriodDates id="CalcPeriodDates0">
<effectiveDate>
<unadjustedDate>2007-10-01</unadjustedDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
</effectiveDate>
<terminationDate>
<unadjustedDate>2007-10-31</unadjustedDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
</terminationDate>
<calculationPeriodDatesAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</calculationPeriodDatesAdjustments>
<firstPeriodStartDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
<unadjustedDate>2007-10-01</unadjustedDate>
</firstPeriodStartDate>
<calculationPeriodFrequency>
<rollConvention>NONE</rollConvention>
<periodMultiplier>30</periodMultiplier>
<period>D</period>
</calculationPeriodFrequency>
</calculationPeriodDates>
<paymentDates>
<calculationPeriodDatesReference href="CalcPeriodDates0" />
<paymentFrequency>
<period>D</period>
<periodMultiplier>30</periodMultiplier>
</paymentFrequency>
<payRelativeTo>CalculationPeriodEndDate</payRelativeTo>
<paymentDatesAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</paymentDatesAdjustments>
©2009 Razor Risk Technologies Page 173 of 373
Razor Financial Principals

</paymentDates>
<calculationPeriodAmount>
<calculation>
<notionalSchedule>
<notionalStepSchedule>
<currency>AUD</currency>
<initialValue>3000000</initialValue>
</notionalStepSchedule>
</notionalSchedule>
<fixedRateSchedule />
<dayCountFraction>ACT/365</dayCountFraction>
</calculation>
</calculationPeriodAmount>
<principalExchanges>
<intermediateExchange>false</intermediateExchange>
<initialExchange>false</initialExchange>
<finalExchange>true</finalExchange>
</principalExchanges>
</bondStream>
<issuer href="Bank1" />
<securityId>OCT07</securityId>
<position>1</position>
<paymentAmount>
<paymentType>SETTLEMENT</paymentType>
<receiverPartyReference href="GTY" />
<payerPartyReference href="83976" />
<paymentAmount>
<currency>AUD</currency>
<amount>0</amount>
</paymentAmount>
<adjustedPaymentDate>2007-09-15</adjustedPaymentDate>
</paymentAmount>
<exInterestDays>0</exInterestDays>
</bond>
</irFuture>
</product>
</trade>
<dealHeader>
<dealId>FUT:10912</dealId>
<dealType>SFIBC</dealType>
<dealDate>2007-09-12</dealDate>
<status>OPEN</status>
</dealHeader>
</deal>

©2009 Razor Risk Technologies Page 174 of 373

Razor Financial Principals

12.3.3 Pricing
We define some notations before moving to pricing of each type of interest
rate futures:
F = face value.
f0 = futures price at contract initiation.

ft = futures price at valuation date.

TV = tick value.
TS = tick size.
P = price of underlying asset, e.g. bank bill if it is a bank bill future.
N = number of futures contracts.

Note:
1. Futures prices are quoted as 1 − yield % .

TV
2. represents the value of 0.01% of premium.
TS

Pricing of Interest Rate Futures

1. 30 Day Interbank Cash Rate Futures
Depend on the user’s requirement, Razor can return four different types of
values in regard to the values of the cash rate futures.
(a) The Trade Price Amount

 100 − f 0 30 
This is equal to N × F ×  1 + × .
 100 365 
(b) The Futures Price Amount

 100 − ft 30 
This is equal to N × F ×  1 + × 
 100 365 
(c) The Net Settlement Amount

 TV 
This is equal to N × F × ( f t − f 0 ) ×  .
 TS 
(d) The present values of (a), (b) and (c) are also supported.

It should be noted that for 30 day interbank cash rate futures, because the
contracts always have a face value with fixed value F = 3, 000, 000 at
contract initiation rather than expiry with a fixed term to maturity of 30
TV 30
days, the 0.01% premium is always = 3000000 × 0.0001× = 24.66 .
TS 365
2. 90 Day Bank Bill Futures

©2009 Razor Risk Technologies Page 175 of 373

Razor Financial Principals

The price of a bank bill is:

F × 365
P= .
 yield % × DaystoMaturity 
365 +  
 100 

For a 90 day bank bill future, the price is simply:

F × 365
P= .
 (100 − f ) × 90 
365 +  
 100 

Again, the four types of values Razor can return for the bond futures are:
(a) The Trade Price Amount
N × F × 365
This is equal to .
 (100 − f 0 ) × 90 
365 +  
 100 
(b) The Futures Price Amount
N × F × 365
This is equal to .
 (100 − ft ) × 90 
365 +  
 100 
(c) The Net Settlement Amount
This is equal to
 
 
N × F × 365 ×  .
1 1
−
  (100 − f t ) × 90   (100 − f0 ) × 90  
 365 +   365 +   
  100   100 
(d) The present values of (a), (b) and (c) are also supported.

3. 3 and 10 Year Bond Futures

The price of bond is:

 c (1 − v nf ) 
P = N ×F× + v nf  ,
 i f 
where
100 − f 1
if = × .
100 2

v f = (1 + i f )
−1
.

©2009 Razor Risk Technologies Page 176 of 373

Razor Financial Principals

6 if 3 year bond
n= .
 20 if 10 year bond
CouponRate% 1
c= × .
100 2
The values can be returned by Razor are:
(a) The Trade Price Amount

This is equal to N × F × 
(
 c 1 − v nf
0
) +v n

.
f0
 i f0 

(b) The Futures Price Amount

This is equal to N × F × 
(
 c 1 − v nf
t
) +v n

.
ft
 i ft 

(c) The Net Settlement Amount

The net settlement amount of a bond future is:

N × F × 
(
 c 1 − v nf
t
)
+ v nft
  c 1 − v nf
− 0
( )
+ v nf0   .
 i ft   i f0 
   
(d) The present values of (a), (b) and (c) are also supported.

Note that we have semi-annual compounding in the examples above. Razor

is able to set frequencies in other conventional intervals, e.g. quarterly and
yearly.

12.4 Bond Options, Options on Bond Futures, and Options

on Money Market or Bill Futures
12.4.1 Description of Instrument
Options on interest rate products are the right given to the holder of the
options to enter into the interest rate products in the future at a
predetermined rate or price.

12.4.2 XML Representation

Option on Bond Future: - note for Option on Bill Future - just replace bond
section with example bond section from bill future.
<deal>
<trade>
<tradeHeader>
<tradeId>OPT:1289</tradeId>
<tradeDate>2007-09-07</tradeDate>
<tradeType>SFYTO</tradeType>
©2009 Razor Risk Technologies Page 177 of 373
Razor Financial Principals

<dealer>84793</dealer>
<counterparty>78907</counterparty>
<internalUnit>TDXA</internalUnit>
<buySell>BUY</buySell>
<status>OPEN</status>
<location>Bank1</location>
<domicileCountry>154</domicileCountry>
</tradeHeader>
<product>
<bondOption>
<underlyingType>Future</underlyingType>
<optionType>CALL</optionType>
<cashSettlement />
<strike>
<strikePrice>93</strikePrice>
<priceQuote>FuturesPrice</priceQuote>
<buyer>78907</buyer>
<seller>TDX</seller>
</strike>
<numberOfOptions>100</numberOfOptions>
<premium>
<pricePerOption>1629.679</pricePerOption>
<payerPartyReference />
<receiverPartyReference />
<paymentDate>
<unadjustedDate>2007-09-07</unadjustedDate>
</paymentDate>
</premium>
<europeanExercise>
<expirationDate>
<adjustableDate>
<unadjustedDate>2007-12-15</unadjustedDate>
</adjustableDate>
</expirationDate>
</europeanExercise>
<bond>
<bondStream>
<payerPartyReference href="78907" />
<receiverPartyReference href="TDX" />
<calculationPeriodDates id="CalcPeriodDates0">
<effectiveDate>
<unadjustedDate>2007-12-17</unadjustedDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
</effectiveDate>
©2009 Razor Risk Technologies Page 178 of 373
Razor Financial Principals

<terminationDate>
<unadjustedDate>2010-12-17</unadjustedDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
</terminationDate>
<calculationPeriodDatesAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</calculationPeriodDatesAdjustments>
<firstPeriodStartDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
<unadjustedDate>2007-12-15</unadjustedDate>
</firstPeriodStartDate>
<calculationPeriodFrequency>
<rollConvention>NONE</rollConvention>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
</calculationPeriodFrequency>
</calculationPeriodDates>
<paymentDates>
<calculationPeriodDatesReference href="CalcPeriodDates0" />
<paymentFrequency>
<period>M</period>
<periodMultiplier>6</periodMultiplier>
</paymentFrequency>
<payRelativeTo>CalculationPeriodEndDate</payRelativeTo>
<paymentDatesAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</paymentDatesAdjustments>
</paymentDates>
<calculationPeriodAmount>
<calculation>
<notionalSchedule>
<notionalStepSchedule>
<currency>AUD</currency>
<initialValue>100000</initialValue>
</notionalStepSchedule>
</notionalSchedule>
<fixedRateSchedule>
<initialValue>0.06</initialValue>
</fixedRateSchedule>
<dayCountFraction>ACT/ACT</dayCountFraction>
</calculation>
</calculationPeriodAmount>
©2009 Razor Risk Technologies Page 179 of 373
Razor Financial Principals

<principalExchanges>
<intermediateExchange>false</intermediateExchange>
<initialExchange>false</initialExchange>
<finalExchange>true</finalExchange>
</principalExchanges>
</bondStream>
<issuer href="66096" />
<securityId>FUT-3YR-DEC</securityId>
<position>1</position>
<paymentAmount>
<paymentType>SETTLEMENT</paymentType>
<receiverPartyReference href="TDX" />
<payerPartyReference href="78907" />
<paymentAmount>
<currency>AUD</currency>
<amount>0</amount>
</paymentAmount>
<adjustedPaymentDate>2007-09-10</adjustedPaymentDate>
</paymentAmount>
<exInterestDays>0</exInterestDays>
</bond>
<optionWriterReference href="0" />
<optionHolderReference href="78907" />
</bondOption>
</product>
</trade>
<dealHeader>
<dealId>OPT:1289</dealId>
<dealType>SFYTO</dealType>
<dealDate>2007-09-07</dealDate>
<status>OPEN</status>
</dealHeader>
</deal>

12.4.3 European Exercise Pricing

European bond options, options on bond futures, and options on money market
or bill futures are all priced using the Black-76 model.
This is achieved by converting strike quotes to a consistent dollar price quote,
and converting yield volatilities if provided to price volatilities. These
conversions are outlined below the model details.

The Black-76 Model

Let
F = the forward market price

©2009 Razor Risk Technologies Page 180 of 373

Razor Financial Principals

σ = the price volatility

X = the strike price
Tx = the time from today to expiry, annualised
TS = the time from today to settlement, annualised
r = the continuous yield to settlement
N = the cumulative normal distribution function.
σ 2T x
ln( FX ) +
d1 = 2

σ Tx

d 2 = d1 − σ Tx

The European call price is given by

c = e − rTS [FN ( d1 ) − XN ( d 2 )]

The European put price is given by

p = e − rTS [XN ( −d 2 ) − FN ( −d1 )]

Please refer to section 17.1.4 - generalised Black-Scholes option pricing

formula for further detail and the Greeks.

Strike Quote Conversions

Futures Price = 100 – Yield
Dollar Price = ForwardPrice(Yield, T), where T is forward valuation date for the
underlying

Yield to Price Volatility Conversion

σ P = σ Y yD where

σ P = price volatility
σ Y = yield volatility
y = forward yield (continuous)
D = MacCaulay Duration

12.4.4 American Exercise Pricing

American bond options, options on bond futures, and options on money market
or bill futures are all priced using the Binomial CRR model. For details, please

©2009 Razor Risk Technologies Page 181 of 373

Razor Financial Principals

refer to Binomial Model on page 264. Pricing uses the same model as an equity
option, with coupons instead of discrete dividends. Due to the recursive
binomial calls, performance with many coupons can be an issue. This can be
avoided by using no more steps than coupons. Volatility is assumed to be yield
vol, and is converted into price vol by multiplying by MacCaulay duration and
YTM.

12.5 Caps and Floors

12.5.1 Description of Instrument
Caps and floors are akin to a swap with embedded options on the floating rate
payments, which in effect “cap” the floating rate exposure, or limit the
downside liability.

12.5.2 XML Representation

fpmlCapFloor Schema

Name: Type Description

capFloorStream
fpmlInterestRateStream

fpmlInterestRateStream Schema
Name: Type Occurs Size Description
payerPartyReference
1..1 The identifier of the paying party
fpmlPartyReference
receiverPartyReference
1..1 The identifier of the receiving party
fpmlPartyReference
calculationPeriodDates Indicates the schedule that the
1..1
fpmlCalculationPeriodDates floating rate calculations occurs
paymentDates Indicates the schedule that the date
1..1
fpmlPaymentDates payments occur
resetDates This structure indicates when
1..1
fpmlResetDates floating rate resets occur
calculationPeriodAmount This structure indicates how the
1..1
fpmlCalculationPeriodAmount amounts to be paid is determined
This structure indicates how the
stubCalculationPeriodAmount
0..1 stub period amounts to be paid are
fpmlStubCalculationPeriodAmount
determined

©2009 Razor Risk Technologies Page 182 of 373

Razor Financial Principals

fpmlInterestRateStream Schema
Name: Type Occurs Size Description
principalExchanges Determines if and when the
0..1
fpmlPrincipalExchanges exchange
cashflows This type gives us the fixed cash
0..1
fpmlCashflows flows represented by the product

12.5.3 Pricing
Each cap or floor is considered to be made up of “caplets” or “floorlets”, that
limit the liability or exposure for each floating rate payment. The caplets or
floorlets are valued separately as options and then summed to produce the
overall value for the cap or floor.
We can use the Black model to allow us to price caps by inputting rf as the
current forward price, and rx as the strike price. The output of the Black
model is in terms of a yield percentage per anum. To convert this into a dollar
amount, we need to multiply the yield by the interest sensitivity of the option.
This interest sensitivity will be determined by the principal amount of the
option and the term of the underlying interest period.
Let
S = the forward market rate
X = the cap/floor rate
σ = the yield volatility
F = the implied forward rate at each caplet maturity as the underlying asset.
τ = number of days in the forward rate period.
basis = number of days per year used in the market.
T = the time until expiry, annualised
ln( XS ) + σ 2T
2

d1 =
σ T
ln( XS ) − σ 2T
2

d2 =
σ T
τ
Notional ×
V floorlet = basis e− rT ( XN ( −d ) − SN ( −d ) )
τ 2 1
1+ F
basis
n
V floor = ∑ V floorlet
i =1

τ
Notional ×
Vcaplet = basis e− rT ( SN ( d ) − XN ( d ) )
τ 1 2
1+ F
basis

©2009 Razor Risk Technologies Page 183 of 373

Razor Financial Principals

n
Vcap = ∑ Vcaplet
i =1

12.6 Collars
12.6.1 Description of Instrument
A collar is like an agregation of a cap and floor, in that the interest rate
exposure is limited to a certain range, with both a cap and a floor.

12.6.2 XML Representation

The collar product makes use of the FPML cap and floor product specifications.

12.6.3 Pricing
Let
S = the underlying market par rate
X cap = the cap rate

X floor = the floor rate

σ = the yield volatility
T = the time until expiry, annualised
ln( XS ) + σ 2T
2

d1 =
σ T
ln( XS ) − σ 2T
2

d2 =
σ T
The collar premium is given by:
τi
Notional ×
( i 1 cap ,i 2 )
n
Vcollar = ∑ basis e − rTi S N ( d ) − X N ( d )
τi
i =1
1 + Fi
basis
τi
Notional ×
( floor ,i N ( −d2 ) − Si N ( −d1 ) ) .
n
−∑ basis e − rTi X
τi
i =1
1 + Fi
basis
Note that a collar can be structured as buying a cap and selling a floor
simultaneously with the same underlying details and expiry.

©2009 Razor Risk Technologies Page 184 of 373

Razor Financial Principals

12.7 Swaption
12.7.1 Description of Instrument
A swaption is an agreement between two counterparties to enter into a
currency or interest rate swap at an agreed fixed rate at a date in the future.

12.7.2 XML Representation

fpmlSwaption Schema
Name: Type Occurs Size Description
premium
1..1 The premium for the swaption
fpmlPremium
europeanExercise
1..1 The exercisable period
fpmlEuropeanExercise
calculationAgentPartyReference
1..1
fpmlRef
cashSettlement
0..1
fpmlCashSettlement
swaptionStraddle
0..1
boolean
swap
1..1 The underlying swap
ftmlSwap

<swaption>
<europeanExercise>
<commencementDate id="" >
<unadjustedDate>2001-05-31</unadjustedDate>
</commencementDate>
<expirationDate id="">
<unadjustedDate>2003-05-29</unadjustedDate>
</expirationDate>
</europeanExercise>
<swap>
<swapStream>
<payerPartyReference href="ITE" />
<receiverPartyReference href="BAN" />
<calculationPeriodDates id="CalcPeriodDates0">
<effectiveDate id="">
<unadjustedDate>2003-05-29</unadjustedDate>
</effectiveDate>
<terminationDate id="">
<unadjustedDate>2004-05-31</unadjustedDate>
</terminationDate>
<calculationPeriodDatesAdjustments>
<businessDayConvention>MODFOLLOWING</businessDayConvention>
</calculationPeriodDatesAdjustments>
<calculationPeriodFrequency>
<periodMultiplier>6</periodMultiplier>
<period>M</period>

©2009 Razor Risk Technologies Page 185 of 373

Razor Financial Principals

<rollConvention>IMM</rollConvention>
</calculationPeriodFrequency>
</calculationPeriodDates>
<paymentDates>
<calculationPeriodDatesReference href="#CalcPeriodDates0" />
<paymentFrequency>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
</paymentFrequency>
<payRelativeTo id="">CalculationPeriodEndDate</payRelativeTo>
<paymentDatesAdjustments>
<businessDayConvention>MODFOLLOWING</businessDayConvention>
</paymentDatesAdjustments>
</paymentDates>
<resetDates>
<calculationPeriodDatesReference href="#CalcPeriodDates0" />
<resetRelativeTo id="">CalculationPeriodStartDate</resetRelativeTo>
<fixingDates>
<periodMultiplier>-2</periodMultiplier>
<period>D</period>
<businessDayConvention>NONE</businessDayConvention>
<dateRelativeTo id="#resetDates0">ResetDate</dateRelativeTo>
</fixingDates>
<resetFrequency>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
</resetFrequency>
<resetDatesAdjustments>
<businessDayConvention>MODFOLLOWING</businessDayConvention>
</resetDatesAdjustments>
</resetDates>
<calculationPeriodAmount>
<calculation>
<notionalSchedule>
<notionalStepSchedule>
<initialValue>1000000.000000</initialValue>
<currency>AUD</currency>
</notionalStepSchedule>
</notionalSchedule>
<fixedRateSchedule>
<initialValue>0.000500</initialValue>
</fixedRateSchedule>
<dayCountFraction>ACT/365</dayCountFraction>
</calculation>
</calculationPeriodAmount>
<principalExchanges>
<initialExchange>false</initialExchange>
<finalExchange>false</finalExchange>
<intermediateExchange>false</intermediateExchange>
</principalExchanges>
</swapStream>
<swapStream>
<payerPartyReference href="BAN" />
<receiverPartyReference href="ITE" />
©2009 Razor Risk Technologies Page 186 of 373
Razor Financial Principals

<calculationPeriodDates id="CalcPeriodDates0">
<effectiveDate id="">
<unadjustedDate>2003-05-29</unadjustedDate>
</effectiveDate>
<terminationDate id="">
<unadjustedDate>2004-05-31</unadjustedDate>
</terminationDate>
<calculationPeriodDatesAdjustments>
<businessDayConvention>MODFOLLOWING</businessDayConvention>
</calculationPeriodDatesAdjustments>
<calculationPeriodFrequency>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
<rollConvention>IMM</rollConvention>
</calculationPeriodFrequency>
</calculationPeriodDates>
<paymentDates>
<calculationPeriodDatesReference href="#CalcPeriodDates0" />
<paymentFrequency>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
</paymentFrequency>
<payRelativeTo id="">CalculationPeriodEndDate</payRelativeTo>
<paymentDatesAdjustments>
<businessDayConvention>MODFOLLOWING</businessDayConvention>
</paymentDatesAdjustments>
</paymentDates>
<resetDates>
<calculationPeriodDatesReference href="#CalcPeriodDates0" />
<resetRelativeTo id="">CalculationPeriodStartDate</resetRelativeTo>
<fixingDates>
<periodMultiplier>-2</periodMultiplier>
<period>D</period>
<businessDayConvention>NONE</businessDayConvention>
<dateRelativeTo id="#resetDates0">ResetDate</dateRelativeTo>
</fixingDates>
<resetFrequency>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
</resetFrequency>
<resetDatesAdjustments>
<businessDayConvention>MODFOLLOWING</businessDayConvention>
</resetDatesAdjustments>
</resetDates>
<calculationPeriodAmount>
<calculation>
<notionalSchedule>
<notionalStepSchedule>
<initialValue>1000000.000000</initialValue>
<currency>AUD</currency>
</notionalStepSchedule>
</notionalSchedule>
<floatingRateCalculation>
<floatingRate>
©2009 Razor Risk Technologies Page 187 of 373
Razor Financial Principals

<floatingRateIndex>BBSW</floatingRateIndex>
<indexTenor>
<periodMultiplier>6</periodMultiplier>
<period>M</period>
</indexTenor>
<spreadSchedule>
<initialValue>0.000000</initialValue>
</spreadSchedule>
</floatingRate>
<averagingMethod></averagingMethod>
<negativeInterestRateTreatment></negativeInterestRateTreatment>
</floatingRateCalculation>
<dayCountFraction>ACT/365.FI</dayCountFraction>
</calculation>
</calculationPeriodAmount>
<principalExchanges>
<initialExchange>false</initialExchange>
<finalExchange>false</finalExchange>
<intermediateExchange>false</intermediateExchange>
</principalExchanges>
</swapStream>
</swap>
</swaption>

12.7.3 Pricing
European swaptions are priced using the Black-76 model. The Black-76 value is
multiplied by a factor adjusting for the tenor of the swaptions as shown by
Smith (1991). American and Bermudan swaptions can be priced either using the
Black model or using a numerical approximation (such as HGM/J). Most
numerical approximations for Bermudan swaptions will have a large
detrimental impact on the simulation process, so directing these trades to the
PELookup Server is recommended unless a fast analytical solution is used.

The Black-76/Smith Model

Let
t1 = Tenor of swap in years.
F = Forward rate of underlying swap.
X = Strike rate of swaption.
r = Risk-free interest rate.
T = Time to expiration in years.
σ = Volatility of the forward-starting swap rate.
m = Compoundings per year in swap rate.

Price of payer swap is:

©2009 Razor Risk Technologies Page 188 of 373

Razor Financial Principals

 1 
1 − t1 × m 
 1 + F  
  − rT
c= 
m
F  e [FN (d1 ) − XN (d 2 )]
 
 
 
 
Price of receiver swap is:

 1 
1 − t1 × m 
 1 + F  
  − rT
p= 
m
F  e [XN (− d 2 ) − FN (− d1)] ,
 
 
 
 
where

F  σ 
2
ln  +  T
X   2 
d1 =
σ T
d 2 = d1 − σ T .

12.7.4 Calculating Greeks

The Greeks are calculated numerically using central finite difference
approximations for both the first and second derivatives. Central differences
2
are used since the order of accuracy O(h ) (for the first order) is greater than
single (forward or backward) differences which is O(h)
The first and second derivatives in relation to the central finite differences are:
∂ f ( x + h ) − f ( x − h)
f ' ( x) = = lim
∂x h → 0 2h
∂ 2
f ( x + h) + f ( x − h) − 2 f ( x )
f '' ( x) = 2 = lim
∂x h →0 h2
where
∂ f ( x + h) − f ( x − h) 
≈  h ∈ , h = ( 0, ε )
∂x 2h 
The limit can be approximated by selecting a sufficiently small non-zero value
for h. In the case of delta,gamma,vega and rho, h is set to 1 basis point or
0.0001. Theta h is chosen as 1 day in units of years or 1/ 365 ≈ 0.00274 . For
theta we use only the backward finite difference defined as:
∂ f ( x ) − f ( x − h)
f ' ( x) = = lim
∂x h → 0 h

©2009 Razor Risk Technologies Page 189 of 373

Razor Financial Principals

Delta
Delta is defined as the difference in value of the swaption by perturbing
the forward swap rate F by ±1bp where all other parameters remain
constant.

Let
f = a function defined as the value of the payer or receiver
swaption
F = Forward rate of underlying swap
1bp = 1 basis point or 0.0001

∂V f ( F + 1bp) − f ( F − 1bp )
∆= ≈
∂F 2bp

Gamma
Gamma is defined as the difference in delta by perturbing the forward
swap rate F by ±1bp where all other parameters remain constant. In Razor
we approximate gamma by using the second order finite central
difference.

Let
f = a function defined as the price of the payer or receiver
swaption
F = Forward rate of underlying swap
1bp = 1 basis point or 0.0001
∂∆ ∂ 2V f ( F + 1bp) + f ( F − 1bp) − 2 f ( F )
Γ= = 2≈
∂F ∂F 1bp 2

Vega
Vega is defined as the difference in value of the swaption by perturbing
the forward swap volatility σ by ±1% where all other parameters remain
constant.

Let
f = a function defined as the price of the payer or receiver
swaption
σ = Volatility of the forward-starting swap rate.
1% = 1 percent as a decimal or 0.01

∂V f (σ + 1%) − f (σ − 1%)
Λ= ≈
∂σ 2%

Rho
Rho is defined as the change in value of the swaption by perturbing the
risk free rate r by ±1bp where all other parameters remain constant.

©2009 Razor Risk Technologies Page 190 of 373

Razor Financial Principals

In Razor, pricing formulas accept the risk free rate in terms of a

continuously compounded discount factor.

Let
f = a function defined as the price of the payer or receiver
swaption
df = a continuously compounded discount factor of r in terms of T

dfs = discount factor shift amount as dfs = e−0.0001 T

1bp = 1 basis point or 0.0001

Here we use a modified version of the central derivative. Since the

continuously compounded discount factor is logarithmic we substitute in
the following relations:

ln a + ln b = ln(ab)
ln a − ln b = ln(a / b)
∂V f (df × dfs ) − f ( df / dfs )
ρ= ≈
∂r 2bp

Theta
Theta is defined as the difference in value by shifting the value date
forward by one day where all other parameters remain constant. Since
time moves in a forward direction, for theta we use the backward finite
difference which is defined as:

∂ f ( x ) − f ( x − h) 
≈ h > 0
∂x h 

In Razor, pricing formulas accept the risk free rate in terms of a

continuously compounded discount factor. Therefore, an adjustment to
the discount factor is made to account for the decay of 1 day.

Let
f = a function defined as the price of the payer or receiver
swaption
T = Time to expiration in years.
T −1 = Time to expiration minus 1 day in years.
df = a continuously compounded discount factor of r in terms of T

dfT−1 = a continuously compounded discount factor of r in terms of T−1

1dy = 1 day in years or 1/365

Where
©2009 Razor Risk Technologies Page 191 of 373
Razor Financial Principals

T −1
dfT−1 = df T

∂V f (T−1 , dfT−1 ) − f (T )
Θ=− ≈
∂T 1dy

12.8 European Swaption

12.8.1 Contract Definition
The owner of a payer (receiver, respectively) swaption has the right to
enter at option maturity time the underlying forward payer (receiver,
respectively) swap settled in arrears. A forward payer swap pays fixed
rate and receives floating rate while a forward receiver swap pays float
rate and receives fixed rate.

Define
β = 1 for payer swaption; -1 for receiver swaption.
N = notional amount.
t= valuation date.
T = maturity date of the option.
T0 = the underlying swap start date, T0 ≥ t .
Tn = the underlying swap end date.
Ti float
= i = 1,..., n float , the floating side payment dates of the underlying
swap.
Ti fix = i = 1,..., n fix , the fixed side payment dates of the underlying
swap.
Ti = i = 0,..., n float − 1 , the forward period start dates of the swap’s
reference index.
κ = the strike rate.
δ i float = the day count fraction between Ti −float 1 and Ti float .
δ i fix = the day count fraction between Ti −fix1 and Ti fix .
B (t,T )
= the price of zero-coupon bond at time t paying $1 at time T .
δ tenor = the tenor of the reference index of the underlying swap.
L ( t , Ti , Ti + δ tenor )
= the simple-compounded forward interest rate prevailing at time t ,
starting at time Ti and ending at time Ti + δ tenor .
1  df ( t , Ti ) 
=  − 1 + rspread
δ tenor  df ( t , Ti + δ tenor ) 
rspread = the spread added to the reference index of the underlying
swap.
df ( t , t j )

©2009 Razor Risk Technologies Page 192 of 373

Razor Financial Principals

= the discount factor from time t to time t j .

Swaption Payoff
The swaption payoff at maturity time T is:
n 
float
n fix
N ⋅ β  ∑ B (T , Ti float ) ⋅ δ i float ⋅ L (T ; Ti −1 , Ti −1 + δ tenor ) − ∑ B (T , Ti fix ) ⋅ δ i fixκ  .
 i =1 i =1 

12.8.2 Pricing Formulas

Using Black’s model, the price of payer swaption at the valuation date is:
PStblack = N ⋅ G ( t ) {κ ( t , T ) N ( d1 ) − κ N ( d 2 )} .
The price of receiver swaption at the valuation date is:
RStblack = N ⋅ G ( t ) {κ N ( − d 2 ) − κ ( t , T ) N ( − d1 )} .
 κ ( t, T ) 
 + σ T −t ,Tn −T0 ⋅ (T − t )
2
ln 
κ
d1 =   .
σ T −t ,Tn −T0 ⋅ T − t
d 2 = d1 − σ T −t ,Tn −T0 T − t .
σ T −t ,T −T is the implied volatility with option term T − t and swap term
n 0

Tn − T0 .
n fix
G ( t ) = ∑ δ i fix df ( t , Ti fix ) .
i =1

κ ( t , T ) is the forward swap rate and is equal to:

n float

∑ L (t, T i −1 (
, Ti −1 + τ tenor ) df t , Ti float )
κ (t, T ) = i =1
.
G (t )

12.9 Bermudan Swaption

12.9.1 Contract Definition
For the time Tk < Th < Tn , a Bermudan swaption is contract that gives the
holder the right to enter at any time Tl ( Tk ≤ Tl ≤ Tn ) into an interest-rate
swap with first reset in Tl , last reset in Tn and fixed rate K .

Define
{l1 , l 2 ,..., l n−1 , l n } = the vector of time nodes at the swap reset dates.
{c1 , c2 ..., cn−1 , cn } = the vector of time nodes at the exercise dates.
t k = the actual time at time node k .
Bi, j (q ) = zero-coupon bond price at node (i, j ) valued with maturity at
time node q .
©2009 Razor Risk Technologies Page 193 of 373
Razor Financial Principals

ri , j = interest rate at node (i, j ) implied by the interest rate tree.

IRS i , j = the interest rate swap value at node (i, j ) .
CC i , j = the rolling back value at node (i, j ) .
f = the fraction of a year for the swap payment, e.g. if the swap is
1
paying semi- annually then f = .
2

Note that the actual time of time node l1 ( t l1 ) is Tl and the actual time of
time node l n ( t ln ) is Tn since Tl and Tn are the first swap reset time and
last swap reset time respectively by definition. Also the actual time of
time node c1 ( t c1 ) is Tl since the swap payments start at the first exercise
date.

12.9.2 Pricing
Bermudan swaption has an early exercise feature. Unlike American-style
options, Bermudan swaption only allows early exercises at some discrete
point time, i.e. on the swap reset dates. Representing Bermudan
swaption price in closed-form solutions is very difficult. However, it can
be priced numerically using the interest rate tree approach through
backward induction.

Step 1:
Construct an interest-rate tree starting today to the last swap reset date
with interest rate nodes covering all interim swap reset dates.

Step 2:
Introduce a vector of bond prices Bln , j (l n ) = 1 for all the nodes at time
node l n , i.e. at time Tn .

Step 3:
Calculate the vector of bond prices Bln −1 , j (l n ) for all the nodes at time
node l n −1 using backward induction:

Bln −1 , j (l n ) = e
(
− rl n −1 , j tl n − tl n −1 )
(p B
u l n , j +1 (l n ) + p m Bl , j (l n ) + p d Bl , j −1 (l n ))
n n

Introduce a new vector of bond prices Bln −1 , j (l n −1 ) = 1 for all the nodes at
time node l n −1 .

Step 4:
Calculate both bond vector prices Bln − 2 , j (l n ) and Bln − 2 , j (l n −1 ) from the bond
vector prices Bln −1 , j (l n ) and Bln −1 , j (l n −1 ) respectively using the backward
induction similar to the above step.

©2009 Razor Risk Technologies Page 194 of 373

Razor Financial Principals

Introduce a new vector of bond prices Bln − 2 , j (l n − 2 ) = 1 for all the nodes at
time node l n − 2 .

Step 5:
Such backward induction and new vector bond prices introduction will
continue until the first swap reset date at time node l1 , i.e. at time Tl .

Step 6:
During the rolling back process, if the nodes hit the last exercise date at
time node cn , we calculate the interest rate swap (IRS) value:

ln
IRS cn , j = 1 − Bcn , j (l n ) − ∑ fKB (k ) . cn , j
k = cn +1

Note that Bi, j (q ) can be obtained from the rolling back bond vector
prices.
We also define:
CC cn , j = IRS cn , j .

Step 7:
We calculate CC cn −1, j for all nodes at time node c n − 1 using backward
induction:

CC cn −1, j = e
(
− rc n −1, j t c n − t c n −1 )
( p CC
u cn , j +1 )
+ p m CC cn , j + p d CC cn , j −1 .

Step 8:

Such rolling back process of CC continues until it reaches node (0,0) .

However, during any of the time nodes that match the exercise dates, we
need to calculate the IRS i , j values:
ln
IRSi , j = 1 − Bi , j (l n ) − ∑ fKB (k ) . i, j
k = i +1

Note that Bi, j (q ) can be obtained from the rolling back bond vector
prices.

We then compare CC i , j and IRS i , j . If CC i , j ≥ IRS i , j , we keep CC i , j as its

original value. If CC i , j < IRS i , j then we set CC i , j to IRS i , j and all the
rolling back process of CC will use the new CC i , j value instead of the old
one.

Step 9:
The rolling back CC 0 , 0 is the Bermudan swaption price.

©2009 Razor Risk Technologies Page 195 of 373

Razor Financial Principals

Pricing can also be done using an older method called Geske

Approximation (please refer to section 12.19). This model will be used if
pricing parameter or trade extension “ModelGeskeBermudan” is set with
value 1.

12.10 Constant Maturity Swap

12.10.1 Description of Instrument

A Constant Maturity Swap, also known as a CMS, is a swap that allows the
purchaser to fix the duration of received flows on a swap. The floating leg
of an interest rate swap typically resets against a published index. The
floating leg of a constant maturity swap fixes against a point on the swap
curve on a periodic basis.

Constant Maturity Swaps can be either single currency or cross currency

swaps. The prime factor therefore for a constant maturity swap is the
shape of the forward implied yield curves. A single currency constant
maturity swap versus LIBOR is similar to a series of differential interest
rate fix (or “DIRF”) in the same way that an interest rate swap is similar
to a series of forward rate agreements.

12.10.2 XML Representation

The CMS uses the fpmlInterestRateStream with the additional of a few
elements. A constantMaturity element was added as part of the
floatingRateCalculation. This consisted of the term of the swap/treasury rate
and its frequency.

12.10.3 Pricing

Brotherton-Ratcliffe and Iben show that the convexity adjustment that

must be made to the forward rate (f) is:

[ ]
− 0.5 f 2σ 2T B '' ( f ) / B ' ( f )

Where
f = forward CMS/CMT rate
σ = volatility
T = time to maturity of the forward
B( f ) = the price at time t of a security that provides
coupons equal to the
forward CMS/CMT rate over the life of the bond as a function of yield ( Y )
.

©2009 Razor Risk Technologies Page 196 of 373

Razor Financial Principals

B' ( f ) = first derivative of the bond price (B ) with respect to

yield ( Y ) .
B '' ( f ) = second derivative of the bond price (B ) with respect to
yield ( Y ) .

To cater for the convexity effect on the forward rate, the adjusted
forward rate becomes:

[ ]
f − 0.5 f 2σ 2T B '' ( f ) / B ' ( f ) .

This means that to obtain the expected forward rate, the convexity
adjustment should be added to the forward swap rate (for CMS) or bond
yield (for CMT).

Example:

f = 5% pa
Y = 5% pa
σ = 10% pa

Assume that the instrument provides a payoff in 3 years time linked to

the 3 year rate.

The value of the instrument is given by:

B (Y ) = f / (1 + Y ) + f / (1 + Y ) + (1 + f ) / (1 + Y )
2

The first and second derivatives are given by:

B ' ( f ) = − f / (1 + Y ) − 2 f / (1 + Y ) − 3(1 + f ) / (1 + f ) = -2.7232
2 3 4

B '' ( f ) = 2 f / (1 + Y ) + 6 f / (1 + Y ) + 12* (1 + f ) / (1 + Y ) = 10.2056

3 4 5

The convexity adjustment is given by:

(− 0.5)(0.05)2 (0.10 )2 3(11.8469 / − 2.7232 ) = 0.0001631 or 1.631 bps.
To adjust the forward rate with the convexity effect, the adjusted
forward rate is 5.01631% pa instead of 5.00% pa.

The volatility to use for the convexity adjustment for a specific forward
forecast swap rate will be dependent on the swap rate term, and for
which forward start date. Finding a swaption volatility for expiry
matching the forward start date and underlying maturity matching the
swap rate term would be the ideal.

A practical compounding solution in determining B ' ( f ) and B '' ( f ) is

shown below:
©2009 Razor Risk Technologies Page 197 of 373
Razor Financial Principals

Consider one cashflow where:

m = compounding frequency per annum

n = number of years
x = rate per annum
c = cashflow amount (coupon payment equivalent)

Then:

B (x ) =
c
mn
 x
1 + 
 m

The first derivative:

− c.n
B ` (x ) = mn +1
 x
1 + 
 m

And the second derivative:

c.n.( m.n + 1)
B `` (x ) = mn + 2
 x
m.1 + 
 m

Hence the convexity adjustment can be found by the summation of these

cashflows along with the principal.

12.11 Power Reverse Dual Currency Swap

A power-reverse dual-currency swap, or PRDC, pays FX-linked coupons in
exchange for floating-rate payments.

• Leg 1: Stream of coupons linked with FX options.

• Leg 2: Fixed or Floating rates.

12.11.1 Pricing FX-Linked Coupon Leg

Let us define the given inputs:

• A tenor structure 0 < T1 < ... < Tn .

• A tenor period τ i = Ti − Ti −1 .

©2009 Razor Risk Technologies Page 198 of 373

Razor Financial Principals

• St the spot FX rate at time t .

• rd domestic risk-free rate at time t .
• rf foreign risk-free rate at time t .
• σ volatility of foreign exchange rate.
• w1 the value generated by some deterministic function, e.g. deterministic
foreign coupon rate.
• w2 the value generated by some deterministic function, e.g. deterministic
domestic coupon rate.
• bl the floor rate.
• bu the cap rate.
• t the valuation date.

An FX-linked, or PRDC, coupon for the period [Ti −1 , Ti ] , i = 1,.., n , pays the
amount:

(Ti − Ti −1 ) Ci ( ST )
i

at time Ti , where:

( ) ( (
Ci STi = min max w1STi − w2 , bl , bu .) )
In the classical structure, bl = 0 and bu = ∞ (payoff floored at 0 and no
cap) the payoff is simply:

( ) (
Ci STi = h ⋅ max STi − K , 0 )
w2
where the strike price is defined as K = and the payment rate is
w1
h = w1 .

©2009 Razor Risk Technologies Page 199 of 373

Razor Financial Principals

If we assume a deterministic risk-free rate1, w1 and w2 are values

generated by some deterministic functions then a single cash flow of the
option leg paid for tenor time interval (Ti −1 , Ti ) paid on time Ti evaluated
at time t , call it CFti , is:
 B (t )
( )

EQ 
( )
( ) − r (T − t )
h max STi − K , 0 | Ft  = h St e f i N ( d1 ) − Ke d ( i ) N ( d 2 ) ,
− r T −t

 B Ti −1 
where
S   σ2 
ln  t  +  rd − rf +  (Ti − t )
K  2 
d1 = ,
σ Ti − t
d 2 = d1 − σ Ti − t .
Note that all rd , rf and σ can be obtained from the term structure
curve.
The value of the option leg, VLPj , can be priced as:

VLpj = ∑ (Ti f − Ti −f1 ) CFti .

i> j

Important Assumptions
• Risk-free rates are deterministic.
• w1 and w2 are values generated by some deterministic functions.
If any of the above “deterministic” assumptions does not hold, then this
pricing method does not work.

12.11.2 Pricing Floating Leg

The floating leg is:

1
It is not really realistic because the floating interest rate in the floating leg is
stochastic.
©2009 Razor Risk Technologies Page 200 of 373
Razor Financial Principals

 
N (U (Ti −1 , Ti ) + si ) (Ti −1 , Ti ) Ft 
B(t )
VLRJ = E O  ∑
 i > j B(Ti ) 
(
= N ∑ B ( t , Ti )(Ti − Ti −1 ) E O U (Ti −1 , Ti ) Ft  + si
i> j
Ti

)
Note that E Q (U (Ti −1 , Ti ) ) | Ft  can be obtained from section 17.1
Ti

depending on what forward floating rate it is.

12.12 Cancellable Yield Curve Basket Swap with Floor

A yield curve basket swap with floor is a swap contract type with the
following payoff:

• Leg 1: Rate paid based on the basket of constant maturity treasury (CMT)
rate.
• Leg 2: Fix / Floating rate.

12.12.1 Pricing Option Leg

The option leg consists of a basket of rates.

Define:

U k ( t , Ti ) = the k th CMS forward rate in the basket.

K = the floor level.
µ k (t, T ) = drift function of the k th CMS rate at time t .
σ k (t,T ) = volatility function of the k th CMS rate at time t .
FT j = the floor price at date T j evaluated at valuation date t .
fTi = the floorlet price at tenor date Ti evaluated at
valuation date t .
n = number of tenor dates.
wk = weighting for the k th CMS in the basket.
t = valuation date.
ρ ij = correlation between rate i and j .
µ b (t,T ) = basket drift function at time t .
σ (t, T )
b
= basket volatility function of basket at time t .
U ( t , Ti )
b
= the basket CMF forward fate.
QTi T
Wt ,k
= Wiener process Wt Qi for the k th asset.
Ti T
Wt Q ,b
= Wiener process Wt Qi for the basket.

©2009 Razor Risk Technologies Page 201 of 373

Razor Financial Principals

We need to find the distribution of U b ( t , Ti ) to find the basket floor

value.

We know that each individual floating rate under the forward martingale
measure is distributed as:

dU k ( t , Ti )
= µ k ( t , Ti ) dt + σ k ( t , Ti ) dW Q
Ti
,k
.
U k
( t, Ti )
Thus the basket floating rate under QTi is distributed as:

dU b ( t , Ti )
U b ( t , Ti )
∑ w dU ( t , T )
k k
i
= k

∑ w U (t,T )
k
k k
i

∑ w U ( t , T ) µ ( t , T ) ∑ w U ( t , T ) σ ( t , T ) dW
k k
i
k
i
k k
i
k
i t
QTi , k

= k
dt + k

∑ w U (t,T )k
k
∑ w U (t, T )
k
i
k
k k
i

∑ w U (t, T ) µ (t, T )
k k
i
k
i
= k
dt + ∑ λ ( t , T ) σ ( t , T ) dW k k QTi ,k
.
∑ w U (t,T )k
k k
i k
i i t

It is clearly not lognormally distributed. However, to retain the lognormal

property for the basket rates, we need to freeze the drift and volatility.

dU b ( t , Ti )
U b ( t , Ti )
≈ µ b ( t , Ti ) dt + ∑ λ k ( 0, Ti ) σ k ( t , Ti ) dWt Q
Ti
,k

= µ b ( t , Ti ) dt + ∑∑ λ ( 0, T ) λ ( 0, T ) ρ σ ( t , T ) σ ( t , T )dW
∗
k l k l QTi ,b
i i ij i i t
k l

=µ b∗
( t , Ti ) dt + σ ( t , Ti ) dWt Q
Ti
b ,b
.
It is assumed that σ b
(t , T )f
is provided by the user.
Note that µ b ( t , Ti ) represents the basket drift under measure QTi .
∗

µ b ( t , Ti ) is approximated by:
∗

E Q U b (Ti −1 , Ti ) | Ft 
Ti

Ti  
= E Q ∑ wkU k (Ti −1 , Ti ) | Ft 
 k 

©2009 Razor Risk Technologies Page 202 of 373

Razor Financial Principals

= ∑ wk E Q U k (Ti −1 , Ti ) | Ft  .
Ti

Note that the computation of E Q U k (Ti −1 , Ti ) | Ft  is given in section 17.1.

Ti

For example, if the rate is LIBOR rate, then E Q U k (Ti −1 , Ti ) | Ft  = U k ( t , Ti )

Ti

. If the rate is CMS rate, then we need convexity adjustment.

The basket rate can now be treated as a single rate.

An interest rate floor is a series of floorlets:
n
FT j = ∑ fTi .
i> j

The value of a floorlet is:

fTi = (Ti − Ti −1 ) B ( t , Ti ) E Q  max K − U b (Ti −1 , Ti ) | Ft  . ( )

i T

Note that U b ( t , Ti ) is lognormally distributed and we know the drift and

volatility of U b ( t , Ti ) from above. Thus:
E Q  max ( K − U b (Ti −1 , Ti ) ) | Ft 
i T

= KN ( −d 2 ) − E Q U b (Ti −1 , Ti ) | Ft  N ( − d1 ) ,
Ti

where:
b∗
E Q U b (Ti −1 , Ti ) | Ft  = U b ( t , Ti ) e ( t ,Ti )(Ti −t )
Ti
µ
,
n
U b ( t , Ti ) = ∑ wkU k ( t , Ti ) ,
k =1
n
µ b∗
( t , Ti ) = ∑ wk E Q U k (Ti −1 , Ti ) | Ft  .
Ti

k =1

Note that the computation of E Q U k (Ti −1 , Ti ) | Ft  is given in section 17.1.

Ti

For example, if the rate is LIBOR rate, then E Q U k (Ti −1 , Ti ) | Ft  = U k ( t , Ti )

Ti

. If the rate is CMS rate, then we need convexity adjustment.

 E Q i U b (Ti , Ti −1 ) | Ft   σ b ( t , T )2 (T − t )
T

ln   + i i

 K  2
d1 =   ,
σ ( t , Ti ) Ti − t
b

 E Q i U b (Ti −1 , Ti ) | Ft   σ b ( t , T )2 (T − t )
T

ln   − i i

 K  2
d2 =   .
σ ( t , Ti ) Ti − t
b

12.12.2 Pricing Floating Fix Leg

©2009 Razor Risk Technologies Page 203 of 373

Razor Financial Principals

Pricing floating and fix leg can be done using the usual forward
martingale methods.

The floating leg value is:

( )
VLPj = N ⋅ ∑ B ( t , Ti )(Ti − Ti −1 ) E Q (U (Ti −1 , Ti ) ) | Ft  + si .
i> j
Ti

Note that E Q (U (Ti −1 , Ti ) ) | Ft  can be obtained from section 17.1

Ti

depending on what forward floating rate it is.

The fix leg value is:

(
VLpj = N ⋅ ∑ B ( t , Ti )(Ti − Ti −1 ) fixedTi + si .
i> j
)

12.13 Cancellable Swap with Equity Trigger

Cancellable swap with equity trigger is an interest rate swap contract

where the underlying swap can be cancelled by either the booking-party
or counter-party or automatically terminated earlier than maturity
depending on the series of specified equity trigger levels. The underlying
interest rate swap can be receiving floating interest rate payments on
one side and paying fixed interest payments on another side, or receiving
fixed interest payments and paying floating interest payments. Either
party has the right to cancel the swap:

• At the strike price on the exercise date.

• Automatically terminated according to the equity termination rate and
termination date.

The underlying swap is a standard swap (as the equity trigger is only
considered for early termination in simulation, and not for MTM). The two
legs are:

• Leg 1: It is the fixed leg and is based on fixed rate

• Leg 2: It is the floating rate and is based on LIBOR with a spread.

12.13.1 Pricing Fixed Leg

The fixed leg is a stream of fixed rate plus a spread. The value of the
fixed leg is:

(
VLpj = N ⋅ ∑ B ( t , Ti )(Ti − Ti −1 ) fixedTi + si .
i> j
)
©2009 Razor Risk Technologies Page 204 of 373
Razor Financial Principals

12.13.2 Pricing Floating Leg

The floating leg is:

( )
VLPj = N ⋅ ∑ B ( t , Ti )(Ti − Ti −1 ) E Q (U (Ti −1 , Ti ) ) | Ft  + si .
i> j
Ti

Note that E Q (U (Ti −1 , Ti ) ) | Ft  can be obtained from section 17.1

Ti

depending on what forward floating rate it is.

12.13.3 Cancellable Swap

It is the same as early exericisable swaps and can be structured as a
Bermudan swaption and the non-cancellable swap.

12.13.4 Cancellable Swap with Equity Trigger

The early termination for the cancellable swap is subject to an equity
trigger. The evaluation of this callable swap is as for standard callable
swap so without taking into account the equity trigger for the MtM. This
equity trigger only has an impact in the simulation: if the path simulated
implies a termination then the MtM becomes 0 (like for the PRDC with
early terminations).

12.14 Cancellable Reverse Interest Rate Swap

A swap contract type where the rate paid on one side increases as market
floating rates decline and the rate on another side is based on fixed rate.

The two legs are:

• Leg 1:Reverse floating leg.

• Leg 2: Fixed or floating rates.

12.14.1 Pricing Reverse Floating Side

The rate paid on the reverse floating side is set by the fixed rate minus
the multiple of floating reference rate for each payment period.

We write:
fTi = fixed rate at time Ti .
wTi = weight at time Ti .

©2009 Razor Risk Technologies Page 205 of 373

Razor Financial Principals

Hence for each call date T j :

VLPj
 B (t ) 
= E Q ∑ ( (
N fTi − wTi ⋅ U (Ti −1 , Ti ) + sTi )) (T − T ) | F 
 i > j (Ti )
i i −1 t
B 

i> j
(
= N ∑ B ( t , Ti )(Ti − Ti −1 ) fTi − wTi ⋅ E QTi
U (Ti −1 , Ti ) | Ft  − sTi . )
Note that E Q (U (Ti −1 , Ti ) ) | Ft  can be obtained from section 17.1
Ti

depending on what forward floating rate it is.

12.14.2 Reverse Floating Side with Floor

The reverse floating side can also have a floor. If the floor rate is K , the
value of this single cash flow at time ti , call it CFti is:

{ (
CFti = B ( t , Ti ) E Q  max K − fTi + wTi ⋅ U (Ti −1 , Ti ) − sTi , 0 | Ft 
Ti

  ) }
   fT    .
K
= wTi E Q  max U (Ti −1 , Ti ) −  sTi − + i  , 0  | Ft 
Ti

 wTi wTi   
   
The above can be evaluated using Black’s call option pricing formula with
K fT
E Q U (Ti −1 , Ti ) | Ft  to replace the current stock price and sTi − + i to
Ti

w w Ti Ti

replace the strike price. Note that E U (Ti −1 , Ti ) | Ft  can be evaluated as
QTi

in section 17.1 depending on what type of rate it is.

Note that we did not include notional amount to keep the length of the
formula shorter. It can be incorporated by multiplying the notional
amount to the formula as a scalar constant.
The value for floating leg at callable time t j evaluated at time t is:

VLPj = ∑ CFti .
i> j

12.14.3 Pricing Fixed-Floating Leg

The floating leg value is:

(
VLPj = N ⋅ ∑ B ( t , Ti )(Ti − Ti −1 ) E Q (U (Ti −1 , Ti ) ) | Ft  + si .
i> j
Ti

)
Note that E Q (U (Ti −1 , Ti ) ) | Ft  can be obtained from section 17.1
Ti

depending on what forward floating rate it is.

The fix leg value is:

©2009 Razor Risk Technologies Page 206 of 373

Razor Financial Principals

(
VLpj = N ⋅ ∑ B ( t , Ti )(Ti − Ti −1 ) fTi + si .
i> j
)

12.15 Interest Rate Cliquet

An interest rate cliquet consists of a series of forward starting options
where the strike price for the next exercise date is set equal to a positive
constant times the asset price as of the previous exercise date.

12.15.1 Payoff of Interest Rate Cliquet

The payoff of interest rate cliquet for the tenor period (Ti −1 , Ti ) paid on
payment date Ti is:
 B (t ) 
(Ti − Ti −1 )  max (U (Ti −1 , Ti ) − U (Ti− 2 , Ti −1 ) , 0 )
.
 B (Ti ) 
The above is only the payoff for a single period. The whole interest rate
cliquet payoff is the sum of each individual payoff (Ti −1 , Ti ) for i = 1, 2,..., n ,
assuming there are n tenor dates in total.

12.15.2 Pricing
We first focus on the pricing of the claim for the tenor period (Ti −1 , Ti ) .
The price of the claim, fTi evaluated on the valuation date t is:
  B (Ti )  
fTi = E Q0 (Ti − Ti −1 )  max (U (Ti −1 , Ti ) − U (Ti − 2 , Ti −1 ) , 0 )  | Ft 


 B ( t )  

(
= (Ti − Ti −1 ) B ( t , Ti ) E Q  max (U (Ti −1 , Ti ) − U (Ti − 2 , Ti −1 ) , 0 ) | Ft  )
i T

 

For the first payment, U (T0 , T1 ) is observable. Thus it is a standard cap

with strike rate U (T0 , T1 ) . For later payments, by assuming U (Ti −1 , Ti ) and
U (Ti −2 , Ti −1 ) are bivariate-lognormally distributed,

(
E Q  max (U (Ti −1 , Ti ) − U (Ti −2 , Ti −1 ) ,0 ) | Ft  )
iT

 
= B ( t , Ti ) E ( QTi
)
U (Ti −1 , Ti ) | Ft  N ( d1 ) − E Q U (Ti −2 , Ti −1 ) | Ft  N ( d 2 ) ,
Ti

where
E Q U (Ti −1 , Ti ) | Ft  can be obtained from section 17.1,
Ti

E Q0 U (Ti − 2 , Ti −1 ) | Ft  can be obtained from section 17.1 (we can do this
Ti

because we assume the correlation of rates between different tenors is

0),
©2009 Razor Risk Technologies Page 207 of 373
Razor Financial Principals

 E Q i U (Ti −1 , Ti ) | Ft  
T

ln  QTi   
 E U (Ti − 2 , Ti −1 ) | Ft   1
 
d1 =  + σ i2 ⋅ (Ti − t ) − σ i2−1 ⋅ (Ti −1 − t ) ,
σ i ⋅ (Ti − t ) − σ i −1 ⋅ (Ti −1 − t )
2 2 2

d 2 = d1 − σ i2 (Ti − t ) − σ i2−1 (Ti −1 − t ) ,

σ i and σ i −1 can be calibrated from market data and the values of σ i and
σ i −1 provided by the user.

The price of the cliquet at the valuation date t is:

Ct = ∑ fTi .
i

12.16 Range Accrual Swap

For range accrual swap, one leg is the range accrual note and the other
leg is either a floating rate leg or fixed rate leg. Range accrual notes are
similar to fixed rate bonds except that the coupons are only paid when
some reference rate falls within a particular range.

The legs for range accrual swap are:

• Leg 1: Range accrual note.

• Leg 2: Fixed or floating rate.

12.16.1 Pricing Range Accrual Note Leg

Range accrual notes can be treated as structured products that the

payoffs of range accrual notes can be replicated by a series of digital
floorlets.

Define:
aj = the accrual factor for the period t j −1 and t j .
Bmax = the upper bound of reference rate range for coupon
payment.
Bmin = the lower bound of reference rate range for coupon
payment.
D(t , T , K ) =
the price of digital floorlet on the reference rate at time t
maturing at time T with strike rate K .
tj = time for the j th coupon.
t = the valuation date.

©2009 Razor Risk Technologies Page 208 of 373

Razor Financial Principals

Step 1
We first consider the coupon contribution to the j th coupon by a single
day say at time ti where t < ti < t j .

The coupon contribution for the j th coupon by time ti is:

1 U ( ti −1 , ti ) if Bmin ≤ U ( ti −1 , ti ) ≤ Bmax

aj  .
365 0 otherwise

The above simply means that for the j th coupon, an amount of

U ( ti −1 , ti )
aj is contributed by time ti if the reference at time ti −1 falls
365
within the target bound. Note that the accrual factor a j is there to keep
the generosity because we don’t always start the range accrual note on
the coupon payment day.

Step 2:
The payoff of a digital floorlet D ( t , ti , K ) is:

U ( ti −1 , ti ) if U ( ti −1 , ti ) ≤ K
 .
0 otherwise

By longing D ( t , ti , Bmax ) and shorting D ( t , ti , Bmin ) with notional amount

1
aj , this structured payoff at time ti is:
365

1 U ( ti −1 , ti ) if Bmin ≤ U ( ti −1 , ti ) ≤ Bmax

aj  .
365 0 otherwise

This completely replicates the range accrual notes except that the timing
of this payoff for range accrual note is at time t j while the structured
payoff is at time ti . Time correction is required for an exact replication.
We now look at the present values of these two payoffs:

We assume the discount factor is non stochastic, the expected present

value for the range accrual note is:
1 Q  B (t ) 
aj E  U ( ti −1 , ti )1{B ≤U ( t ,t )≤ B } | Ft 
365  B ( t j ) min i −1 i max


= B (t, t j ) a j
1 Qti 
E U ( ti −1 , ti )1{Bmin ≤U ( ti−1 ,ti )≤ Bmax } | Ft  .
365  

©2009 Razor Risk Technologies Page 209 of 373

Razor Financial Principals

The expected present value for the structured digital floorlets is:
1 Q  B (t ) 
aj E  U ( ti −1 , ti )1{B ≤U ( t ,t )≤ B } | Ft 
365  B ( ti ) min i −1 i max


= B (t, t j ) a j
1 Qti 
E U ( ti −1 , ti )1{Bmin ≤U ( ti−1 ,ti )≤ Bmax } | Ft  .
365  

1 B (t, t j )
Thus by holding a notional of aj for the digital floorlets, we
B ( t , ti ) 365
can fix the time mismatch problem and completely replicate the range
accrual note claim.

Step 3
We need to know the price of digital floorlet. The payoff of digital
floorlet is:
U ( ti −1 , ti ) if U ( ti −1 , ti ) ≤ K
 .
0 otherwise
It is known that U ( ti −1 , ti ) is lognormally distributed under Q ti , the price
of the floorlet can be evaluated using the normal asset-or-nothing digital
floorlet, i.e.:
D ( t , ti , K ) = B ( t , ti ) E Q U ( ti −1 , ti ) | Ft  N ( −d ) with
ti

E Q U ( ti −1 , ti ) | Ft  can be computed as in section 17.1,

ti

 E Q i U ( ti −1 , ti ) | Ft   σ 2 ⋅ ( t − t )
t

ln   + e i
 K  2
d=   .
σ e ti − t
Note that σ e here is the volatility of the LIBOR rate and can be obtained
from the term structure curve.

Step 4
The above two steps demonstrated the replication of range accrual note
for a single day contribution by time ti using floorlets maturing at time ti .
To replicate the whole range accrual, we simply use a series of long and
short floorlets maturing for each of the time t ≤ ti ≤ t j , where
j = 1, 2,3,..., n with n denoting the number of coupon payments.

Step 5
We need to evaluate the digital floorlets. If the reference rate is LIBOR
rate, we can construct a measure such that the forward LIBOR rate is a
geometric Brownian distributed martingale. If the reference rate is not
LIBOR rate then one way we can do is to make an assumption that the
reference rate is geometric Brownian distributed. In either way, the
digital floorlets can be evaluated using Black’s formula.

©2009 Razor Risk Technologies Page 210 of 373

Razor Financial Principals

Note that if the coupon payment is based on fixed rate rather than
floating rate, then we have cash-or-nothing digital floorlet formula
instead of asset-or-nothing digital floorlet formula.

12.16.2 Pricing Floating-Fix Leg

Pricing floating and fix leg can be done using the usual forward
martingale methods.

The floating leg value is:

( )
VLPj = N ⋅ ∑ B ( t , Ti )(Ti − Ti −1 ) E Q (U (Ti −1 , Ti ) ) | Ft  + si .
i> j
Ti

The fix leg value is:

VLpj = N ⋅ ∑ B ( t , Ti )(Ti − Ti −1 ) fTi + si .

i> j
( )

The forward LIBOR rates U ( t , Ti ) can be obtained from the LIBOR term
structure curve.

12.17 Constant Maturity Cap and Floor

Constant maturity cap and floor are like normal cap and floor except for
that the underlying rate becomes constant maturity swap rae. The pricing
can be done using Black-Scholes’ formula.

12.17.1 Pricing

The price of constant maturity cap is:

(
B ( t , Ti ) E Ti
PQ
U (Ti −1 , Ti ) | Ft  N ( d1 ) − KN ( d 2 ) . )
The price of constant maturity floor is:
(
B ( t , Ti ) KN ( −d 2 ) − E
PQ
Ti U (Ti −1 , Ti ) | Ft  N ( −d1 ) .)
Note that E Ti U (Ti −1 , Ti ) | Ft  can be calculated as in section 17.1.
PQ

©2009 Razor Risk Technologies Page 211 of 373

Razor Financial Principals

12.18 Early Exercisable Interest Rate Derivatives

Some of the interest rate derivatives from section 12.12 to 12.17 (except
for section 12.16 on interest rate cliquet) can have early exercise
features. This section discusses on how to cater for early exercise
features for those contracts.

Step 1
For any interest rate swap product, there is a payer leg and a receiver
leg. The payer leg generates a series of cash flows and the receiver leg
generates a different set of cash flows. We treat each leg separately and
the value of the swap is simply the difference between the values of the
two legs.

Step 2
We need to determine the value of each leg. To calculate the value of
the payer leg, we find a market proxy such that the proxy generates
similar cash flows as the payer leg. The value of the payer leg is
approximated by the price of the proxy. The same procedure is used to
calculate the value of the receiver leg. The value of the interest rate
swap is difference between the values of the two legs.

Step 3
The cash flows of longing the interest rate swap with early exercise
features can be structured as longing a Bermudan swaption on the
interest rate swap (with pay and receive legs reversed) with the
corresponding early exercise dates and longing the interest rate swap
without the early exercise features. The value on the valuation date for
the interest rate swap with early exercise features is the sum of the value
of the Bermudan swaption and the value of the interest rate swap without
early exercise features on the valuation date. The value of Bermudan
swaption can be approximated using Geske’s approximation and the value
of interest rate swap without early exercise features can be calculated as
described in Step 2.

Note that if the other party also has the right to early exercise, then this
early exercise right can be structured as shorting the Bermudan swaption.

©2009 Razor Risk Technologies Page 212 of 373

Razor Financial Principals

If both parties have callable rights on an underlying swap, then there will
be both a long and a short Bermudan swaption with explicit exercise
dates specific to each party.

12.19 Bermudan Swaption – Geske’s Approximation

In this section, we provide a closed form solution for valuing Bermudan
swaptions. A Bermudan swaption is an option to exercise into an interest rate
swap at a finite number of callable dates T1c , T2c ,..., Tnc . In this document all
Bermudan Swaptions are priced with Geske’s approximation2.

Without loss of generality, we assume the valuation date t = 0 . Let us denote

VLpj and VLRj the value of the time T jc residual legs paid (resp. received) by the
option holder being evaluated on the valuation date. The European option
maturing at time T jc is given by the following formula:

V j = VLRj Φ ( d1, j ) − VLPj Φ ( d 2, j ) ,

and the formula for computing a general Bermudan swaption:

V j Φ ( d 2, j )
( )∑
V = V j max + 1 − Φ ( d 2, j max ) ⋅
j ≠ j max

Φ ( d 2, j )
,
∑ j ≠ j max

where

 VLR  1 2 c
ln  Pj
 VL j  + σ j T j
d1, j =   2 ,
σ j T jc

d 2, j = d1, j − σ j T jc ,

2
Callable feature in Triple-A program, Christophe Michel, Oct. 23, 2007
©2009 Razor Risk Technologies Page 213 of 373
Razor Financial Principals

and Φ ( ⋅) is the cumulative density function of the normal centred distribution.

j max is the index for which V j max is the maximum (is the maximum European
price). The volatility σ , used to compute the probabilities to be ITM and OTM
in d1, j and d 2, j , is assumed ex ante (given as a term structure).

12.20 Numerical Interest Rate Models

Very often when valuing financial instruments, we assume the underlying

interest rate to be deterministic. It is a reasonable assumption if the
major risk components inherited in the instruments are not the interest
rate risk. By assuming deterministic interest rates, we actually assume
the price for the interest rate risk is zero. However, deterministic
interest rate is clearly not valid when valuing interest rate products as it
does not make any sense to price a product with a zero price for the
inherited risk. It is then natural to assume the interest rate to follow
some stochastic processes and it leaves with the problem of choosing a
suitable model to describe the interest rate process. There are many
literatures on modelling interest rate. Some focused on modelling the
short rate (e.g. Hull-White Model), some focused on modelling the
forward rate (e.g. HJM approach) and also some attempted to model in
the context of observable market rates (e.g. BGM Model). In this
document, we will concentrate on the method of modelling short rate
using Hull-White one factor model. We will first state some definitions.

Define:
rt = the short rate at time t .
B(t , T ) = the zero-coupon bond price at time t that matures at time T .
Bt = the savings account at time t .
Wt = the Weiner Process at time t .

We should note that by definition, the following equality should hold:

t

Bt = e ∫u = 0
r j du

B(t , T ) = Bt E P∗ (BT−1 | Ft )
where
P ∗ is a probability measure such that the discounted bond price is a
martingale under this measure (called the spot-martingale measure) and
Ft is the filtration to time t .

©2009 Razor Risk Technologies Page 214 of 373

Razor Financial Principals

12.20.1 Hull-White One Factor Interest Rate Tree Model (Equal Time
Step)

Hull and White (1990) proposed a model that used the following
stochastic differential equation to model the short rate process:

drt = (θ (t ) − art )dt + σdWt

where

θ (t ) = time-dependent deterministic function.

a = constant mean-reversion rate of the short rate.
σ = constant volatility of the short rate.

The Hull-White model has a mean reversion feature and the speed of
mean reversion of the short rate is determined by the parameter a . The
time-dependent deterministic function θ (t ) is a free function allowing us
to fit our interest rate model to the initial term-structure curve (i.e. it is
an arbitrage-free model).

Using the Hull-White one factor model, closed-form solutions can be

derived for some relatively more simple interest rate products, e.g. zero-
coupon bonds and bond options. However, for more exotic and
complicated products (e.g Bermudan swaptions), it is often difficult to
obtain closed-form solutions. Hull and White (1994) proposed a numerical
model for the short rate by discretising the analytical Hull-White one
factor model using interest rate tree.

Hull-White Interest Rate Tree

Hull and White (1994) proposed a trinomial interest tree that is
essentially a discrete-time, discrete-state version of the Hull-White
model. The interest rate can move in three different ways in the next
time period. The interest rate model is very useful when valuing
complicated interest rate products especially those with early exercise
features.

Steps to construct the trinomial tree

Step1:
Recall
drt = (θ (t ) − art )dt + σdWt .
We define another process
drt∗ = − art∗ dt + σdWt .
We need to construct the tree for rt∗ first.
We discretise the interest rate movement and time period so that we
have ∆rt∗ and ∆t instead of drt∗ and dt .

Step2:

©2009 Razor Risk Technologies Page 215 of 373

Razor Financial Principals

We set ∆rt∗ = σ 3∆t . This choice of ∆rt∗ is good for error minimisation.
We define node (i, j ) as the node where t = i∆t and rt∗ = j∆rt∗ ( i implies
the time point and j implies the interest rate state). We can then
construct a tree for the ∆rt∗ process. The following table is a
demonstration:

j=
∗ ∗
3∆rt 3∆rt 3
∗ ∗
2∆rt 2∆rt 2∆rt∗ 2
∆rt∗ ∆rt∗ ∆rt∗ ∆rt∗ 1
0 0 0 0 0 0
− ∆rt∗ − ∆rt∗ − ∆rt∗ − ∆rt∗ -1
− 2∆rt∗ − 2∆rt∗ − 2∆rt∗ -2
∗ ∗
− 3∆rt − 3∆rt -3
i=0 1 2 3 4

Note that the above demonstration is capped and floored at j = ±3 . It is a

good practice that we impose cap and floor for the interest rate tree to
emphasize its mean reversion feature. Hull and White showed that
probabilities are always positive if we set the maximum j value (denoted
j max ) equal to the smallest integer greater than 0.184 / ∆t and the
minimum j value (denoted j min ) equal to − j max .

Step 3:
Define pu , p m and pd be the probabilities that the interest rates move
up one state, remain stable at the current state and drop down one state
respectively. Note that if we are at the upper bound rate, i.e. at j = j max ,
then pu , p m and pd become the probabilities that the interest rates
remain stable at the current state, drop down one state and drop down
two states respectively. Similar adjustments but in different directions
can be followed when we are at the lower bound, i.e. at j = j min .

Hull and White proposed the following probabilities:

At normal state:

1 a 2 j 2 ∆t 2 − aj ∆t 2 1 a 2 j 2 ∆t 2 + aj∆t
pu = + pm = − a 2 j 2 ∆t 2 pd = +
6 2 3 6 2

At upper bound state:

7 a 2 j 2 ∆t 2 − 3aj∆t
pu = +
6 2

©2009 Razor Risk Technologies Page 216 of 373

Razor Financial Principals

1
p m = − − a 2 j 2 ∆t 2 + 2aj∆t
3
1 a 2 j 2 ∆t 2 − aj∆t
pd = +
6 2
At lower bound state:
1 a 2 j 2 ∆t 2 + aj∆t
pu = +
6 2
1
p m = − − a 2 j 2 ∆t 2 − 2aj∆t
3
7 a 2 j 2 ∆t 2 + 3aj∆t
pd = + .
6 2

Step 4:
We now need to choose the free deterministic function θ (t ) in the Hull-
White Model to make sure that our interest rate model is consistent with
the initial term structure curve (i.e. to ensure the model is arbitrage-
free).

Define:
α t = rt − rt∗ .

It can be shown that

dα t = [θ (t ) − aα (t )]dt and then
σ2
α t = F (0, t ) +
2a 2
(1 − e )
− at 2

where
F (0, t ) = the forward rate implied at time t from time 0.

We need to calculate α t in the discrete state and time setting. Once we

can calculate α t , we can then construct the arbitrage-free interest rate
tree for rt from rt∗ . Note that rt = rt∗ + α t .

Step 5:
Hull and White proposed the following formulas to calculate α t . In the
discrete time setting, α t is a step function.

Define nm to be the number of nodes on each side of the central node at

time m∆t for m to be integers ≥ 0 .

nm

∑Q e − j∆rt ∆t − ln B(0, m + 1)
∗
ln m, j
j = − nm
αm =
∆t
Qm +1, j = ∑ Qm, k p (k , j ) exp[− (α m + k∆R )∆t ]
k

©2009 Razor Risk Technologies Page 217 of 373

Razor Financial Principals

where
p(k , j ) is the probability of moving from node (m, k ) to node (m + 1, j ) and
the summation is taken over all values of k for which this is nonzero.

Note that both α 0 and B(0, m + 1) are immediately available from the
initial term structure. α 0 is simply the continuous rate of the zero-
coupon bond B(0, ∆t ) .

12.20.2 Hull-White One Factor Interest Rate Tree Model and Calibration
Method (Unequal Time Step)

Summary
This document specifies the procedures on generalising the equal time
step Hull-White tree to the unequal time step cases. This generalised
version of Hull-White tree is important for pricing financial products with
early exercise features because exercise dates do not usually fall on equal
time step nodes. The tree construction method is based on the book by
Brigo and Mercurio. Calibration of model is also mentioned in this section.

Definitions and notations

We need to redefine the notations slightly for unequal time step cases:
xi , j = the interest rate at node (i, j ) .
∆t i = ti+1 − ti .
∆xi = change in interest rate in vertical direction at time ti .
pu , pm and pd are probabilities of moving up, middle and down as usual.
E {X (ti +1 ) | X (ti ) = xi , j } = M i , j .
Var{X (ti +1 ) | X (ti ) = xi , j } = Vi ,2j .

Assumption
∆xi is assumed to be constant across time ti .

With the above assumption, we can write xi , j = j∆xi .

Methodology
The Hull-White one factor model is by definition:
dr = (θ ( t ) − ar ) dt + σ dWt ∗ .
In the equal time step case, we start by constructing the tree for the
process dx = − axdt + σdWt ∗ and then determine θ (t ) by fitting the initial
term structure curve. For unequal time step case, we use the same
approach. This document only discusses on how to construct

©2009 Razor Risk Technologies Page 218 of 373

Razor Financial Principals

dx = − axdt + σdWt ∗ since the fitting of initial term structure is the same as
equal time step case.

Note that xi +1, j +1 = xi +1, j + ∆xi+1 and xi+1, j −1 = xi+1, j − ∆xi+1 as in the equal time
step case. However, in the equal time step case, the following also hold:
xi+1, j +1 = xi , j + ∆xi +1 .
xi+1, j = xi , j .
xi+1, j −1 = xi , j − ∆xi+1 .

The three equations above are no longer true for the unequal time step
case, thus the formula for pu , pm and pd are also different. By setting
∆xi+1 = Vi 3 (this minimises the variability of the tree, please refer to Hull
and White 1994), we can find the following formula for unequal time step
cases:
1 η η
2

pu = + j ,k2 + j ,k .
6 6Vi 2 3Vi
2 η j ,k
2

pm = − .
3 3Vi 2
1 η j ,k η j ,k
2

pd = + 2
− .
6 6Vi 2 3Vi
η j ,k = M i , j − xi +1,k by definition.

It can be shown for the Hull-White model, the following are true:
M i , j = xi , j − axi , j ∆ti .
Vi , j = σ ∆t i .

Note that in the above formulas, we did not specify what is k . We now
explain the meaning of k and how to determine the value of k .

As we have mentioned before, for equal time step case, xi+1, j = xi , j , i.e., if
we branch to the next time period through the middle path, the vertical
node in the next time period we have branched to is still j and the rate
stays the same as the previous time period. However, it is not true for
the unequal time step case. Although we branch to the next node at time
ti +1 from time ti through the middle path, the vertical node at time ti +1
does not have to be j but can be some other number k . k is chosen in a
way such that the central node we branch to at time ti+1 , i.e., xi+1,k , is
closest to M i , j . It is now obvious that
M 
k = round  i , j  .
 ∆xi+1 
We then have the full specification of the model.

©2009 Razor Risk Technologies Page 219 of 373

Razor Financial Principals

We can now construct the whole tree for xi , j of the process

dx = − axdt + σdWt ∗ . As we mentioned before, the free parameter θ ( t ) can
be fitted using initial term structure curve and the method is similar to
the equal time step case using the following formulas recursively:

Qi +1, j = ∑ Qi , h ph, j exp ( − (α i + h∆xi ) ∆ti ) and

h

∑Q i, j exp ( − j∆xi ∆ti )

1
αi = ln
j
.
∆t i P ( 0, ti +1 )

where
P ( 0, ti +1 )
= the initial bond price at time 0 maturing at time ti +1 and can be
observed from the initial term structure curve.
ph , j = probability moving from state h to j .
The final fitted interest rate tree can be calculated as:
ri , j = xi , j + α i .
Thus we can construct the whole tree for ri , j of the process
dr = (θ ( t ) − ar ) dt + σ dWt ∗ .

Calibration
We have assumed the parameters a and σ are known in the interest tree
model. However, in practice, we do not know these two values. We
should find a way to determine these parameters or in another words, to
calibrate the model. The usual way to calibrate the model is to use the
market instruments to find the market implied values for the parameters.

Step One:
Find a set of ‘calibrating instruments’, i.e. the instruments that are used
to estimate the parameter values. These calibrating instruments must be
instruments that the prices can be calculated from the calibrated model,
i.e. the Hull-White unequal time step tree in this case. For example, we
can choose European swaptions as our calibrating instruments. The
number of calibrating instruments must be no less than the number of
parameters to be estimated. The calibrating instruments chosen should
be as similar as possible to the instrument being valued. For example, if
we want to calculate the value for a Bermudan swaption, we then should
use say European swaptions as the calibrating instruments.

Step Two:
We calculate the calibrating instruments values using the calibrated
model and we denote Vi to be the model value of the i th calibrating
instrument. Note that Vi is a function of a and σ .

©2009 Razor Risk Technologies Page 220 of 373

Razor Financial Principals

Step Three:
We can observe the actual market values for the calibrating instruments
and we denote U i to be the observed market value of the i th calibrating
instrument. Assuming there are n calibrating instruments, we calculate
the sum of least squares between the observed market values and model
values of the calibrating instruments using the formula:

n
SSE = ∑ (U i − Vi ) .
2

i =1

Step Four:
We choose a and σ such that the SSE is minimised. It can be done using
constrained non-linear optimisation algorithm (Box Constrained
Levenberg-Marquardt algorithm).

©2009 Razor Risk Technologies Page 221 of 373

Chapter 13
Commercial Lending

13.1 Collateral
Collateral refers to the practice of providing assets to secure an obligation.
Collateral can take many forms: property, inventory, equipment, receivables,
oil reserves, etc.
Collateralisation agreements are often used to secure repo, securities lending
and derivatives transactions. Under this agreement, a party who owes an
obligation to another party posts collateral, usually cash or securities, to
secure the obligation. In the event that the party defaults on the obligation,
the secured party may seize the collateral.
The arrangement can be unilateral where only one party is obliged to post
collateral, or bilateral where both parties may be obliged to post collateral.
Alternatively, the net obligation may be collateralised, in which case the party
who is the net obligator posts collateral for the value of the net obligation.
Periodically, the secured obligation is revalued and the collateral is adjusted
to reflect changes in value. The securing party adjusts the collateral holdings
depending on the current revaluation of the security.

13.2 Cash Advance Facility

A cash advance facility is where the bank gives a company a facility to draw
money from – essentially a loan with an agreed set of terms and conditions.
Cash Advance Facilities may be committed or uncommitted. Committed
facilities give the company guaranteed access to the full amount of the
facility. Cash Advance Facilities may be revolving or non-revolving. Revolving
facilities are ones where the borrow can repay some or all of the principal and
then draw back down on it again.

13.2.1 Cash Advance Facility XML Representation

Cash Advance Facilities are represented as a “CreditLine” product, which uses
the following XML Schema.

XML Schema

CreditLine Schema
Name: Type Occurs Size Description
currency
1..1 3
string
drawnAmount
1..1
decimal
startDate 1..1
date
Razor Financial Principals

CreditLine Schema
Name: Type Occurs Size Description
periods 1..1
CreditLinePeriods
floatingRateDefinition
0..1
fpmlFloatingRateDefinition
calculationPeriodFrequency 0..1
fpmlCalculationPeriodFrequency
fixedRate
0..1
double

CreditLinePeriod Schema

Name: Type Description

lineAmount
decimal

maturityDate
date

committed
boolean

capitaliseInterest
boolean

Credit Line Example

<creditLine>
<currency>USD</currency>
<drawnAmount>100000000.000000</drawnAmount>
<startDate>2003-05-29</startDate>
<periods>
<period>
<lineAmount>200000000.000000</lineAmount>
<maturityDate>2009-05-29</maturityDate>
<committed>true</committed>
<capitaliseInterest>false</capitaliseInterest>

©2009 Razor Risk Technologies Page 223 of 373

Razor Financial Principals

</period>
</periods>
<calculationPeriodFrequency>
<periodMultiplier>3</periodMultiplier>
<period>M</period>
<rollConvention>MF</rollConvention>
</calculationPeriodFrequency>
<fixedRate>0.050000</fixedRate>
</creditLine>

The CreditLine contains a list of CreditLinePeriod structures that allows the

user to define amortization periods or complex structured loans.

13.2.2 Cash Advance Facility Credit Issues

Committed facilities give the borrower a credit exposure equal to the
maximum amount of the facility. Uncommitted facilities give the borrower a
credit exposure equal to the amount they are currently drawn.

13.2.3 Cash Advance Facility Pricing

The exposure on an uncommitted cash advance facility is considered to be
simply the amount drawn.
The exposure on a committed cash advance facility is considered to be the full
limit of the facility.

13.3 Generic Interest Rate

13.3.1 Description of Instrument
A number of financial instruments can be represented using the
GenericInterestRate product.

13.3.2 XML Representation

GenericInterestRate Schema

Name: Type Description

referenceParty
fpmlPartyReference

interestRateStream
fpmlInterestRateStream

The Generic Interest Rate product encapsulates the FPML InterestRateStream

representation.
©2009 Razor Risk Technologies Page 224 of 373
Chapter 14
Securities

14.1 Promissory Note

14.1.1 Description of Instrument
A Promissory Note or PN is a document issued by a borrower promising to repay
a loan under agreed-upon terms.

14.1.2 XML Representation

This product is represented using the GenericInterestRate schema.
Pricing
Let
F pn = the face value of the promissory note

r pn = the annualised yield to maturity of the promissory note

d pn = the days to maturity of the promissory note

D = the number of days in the year
V pn = the exposure of the promissory note
Then
F pn
V pn =
1 + rpn
d pn
D

14.2 Commercial Bill

14.2.1 Description of Instrument
Commercial Bills are a short term security issued by a company. The holder of
the instrument at maturity receives the face value of the bill from the
acceptor who in turn seeks repayment from the drawer.

14.2.2 XML Representation

This product is represented using the GenericInterestRate schema.
Credit Implications
Should the acceptor default, the holder may seek recourse from previous
endorsers.
Razor Financial Principals

14.2.3 Pricing
Let
F pn = the face value of the bill

r pn = the annualised yield to maturity of the bill

d pn = the days to maturity of the bill

D = the number of days in the year
V pn = the exposure of the bill
Then
F pn
V pn =
1 + rpn
d pn
D

14.3 Certificate of Deposit

14.3.1 Description of Instrument
Certificates of Deposit (or CD) are short term securities issued at a discount to
face value by a bank. The holder of the security receives from the bank the
face value of the CD at maturity.

14.3.2 XML Representation

This product is represented using the GenericInterestRate schema.

14.3.3 Pricing
Let
F pn = the face value of the bill

r pn = the annualised yield to maturity of the bill

d pn = the days to maturity of the bill

D = the number of days in the year
V pn = the exposure of the bill
Then
F pn
V pn =
1 + rpn
d pn
D

14.4 Commercial Paper

14.4.1 Description of Instrument
Commercial Paper is a discount instrument issued by a company to raise funds.
It is unsecured, and thus generally trades at a higher required rate of return.

©2009 Razor Risk Technologies Page 226 of 373

Razor Financial Principals

The company has the obligation to repay the face value of the paper to the
bearer at maturity.

14.4.2 XML Representation

This product is represented using the GenericInterestRate schema.

14.4.3 Credit Implications

As commercial paper is an unsecured investment instrument, the amount at
risk is the discounted face value of the paper held.

14.4.4 Pricing
The value of the Commercial Paper is the present value of the face value of
the paper held.
Let
Pcp = the face value of the commercial paper

df n = the discount factor at time n

Vcp = the value of the commercial paper
Then
Vcp = Pcp df n

14.5 Bank Accepted Bill

14.5.1 Description of Instrument
A bank accepted bill is a discount instrument redeemable for the face value of
the bill upon maturity. A bank accepted bill provides security as the bank has
undertaken to pay the bearer of the bill the face value at maturity.

14.5.2 XML Representation

This product is represented using the GenericInterestRate schema.

14.5.3 Pricing
The value of the bank accepted bill is the present value of the face value of
the bank bill.
Let
Pbab = the face value of the bank accepted bill

df n = the discount factor at time n

Vbab = the value of the bank accepted bill
Then
Vcp = Pbab df n

©2009 Razor Risk Technologies Page 227 of 373

Razor Financial Principals

14.6 Bonds
14.6.1 Description of Instrument
Bonds are secured loans that investors make to corporations and governments.
Corporations and governments issue bonds when they want to raise capital.
Bonds typically pay out a stream of cash-flows to the bearer (the coupon
payments), including repayment of the face value at maturity.
A bond can be described by the following attributes:
The issue date is the day on which the life of a bond starts. The term to
maturity defines the period of time, or the life of the bond. The bond’s
maturity date is the date on which the last payment is due.
The face value (also called par value or principal sum) of a bond represents the
amount that will be repaid to the bondholder at maturity.
The coupon is the nominal annual rate of interest that is paid to the
bondholder on a regular basis. It is usually expressed as a percentage of the
face value (coupon rate). The coupon rate is either fixed or variable. The
coupon rate is given as an annualised percentage of the face value. For
example, a 7% semi-annual bond pays 3.5% of the face value twice a year.
The purchase price is the price the investor pays to buy the bond, i.e., to
receive this series of cash flows (coupon and face value).
The coupon period is not necessarily the same for all bonds. The coupon
payments are made semi-annually, which is common in the USA, or annually,
which is more common in Europe. The coupon payment dates are fixed.
There are many different types of bonds being traded in the market. Bonds can
pay interest that can be fixed (Coupon Bond), floating (Floating Rate Bond) or
payable at maturity (Zero Coupon Bond). Other types of bonds can be
converted into stock at maturity rather than paid interest (Convertible Bond).
Still other bonds can be called back by the issuing company before maturity
(Callable Bond).

14.6.2 XML Representation

As there is no bond schema yet defined in FpML, RAZOR’s bond schema has
been developed by IT&e, but using base components defined by FPML version
3.
The RAZOR bond schema supports the parametric representation of a bond.
The parametric information is designed to capture all the economic
information required to calculate the exposure of the bond the dates, amounts
and rates that imbue the bond with value.
A bond is considered as being a product that contains FPML’s Interest Rate
Stream structure. This provides the parametric representation of the bond.

fpmlBond Schema
Name: Type Occurs Size Description
fpmlPartyReference 1..1
calculationPeriodDates
1..1
fpmlCalculationPeriodDates

©2009 Razor Risk Technologies Page 228 of 373

Razor Financial Principals

fpmlBond Schema
Name: Type Occurs Size Description
paymentDates
1..1
fpmlPaymentDates
calculationPeriodAmount
1..1
fpmlCalculationPeriodAmount
paymentAmount
0..1
fpmlAmount

14.6.3 Credit Implications

There are different contexts against which the credit risk of a bond can be
measured. Issuer risk is the risk the institution has to the issuer of the bond.
This is the risk that the issuer will default on the obligations imbued in the
bond. Counterparty risk is the risk that the counterparty that the bond has
been bought from defaults on the exchange, so that the purchaser doesn’t end
up owning the bond.

14.6.4 Pricing
The value of a bond is equal to the present value of the expected cash flows.
The interest rate or discount rate used to compute the present value depends
on the yield offered on comparable securities in the market. The term
structure used to present value the bond’s cash flows is assumed to be
bootstrapped from comparable securities.
The first step in determining the value of a bond is to work out what it’s cash
flows are. The cash flows of a non-callable fixed rate coupon bond consist of
periodic coupon payments to the maturity date and the payment of the face
value at maturity. Determining the number and timing of cash flows that a
bond has is based on the roll convention and date basis convention of the
bond.
We can now calculate the bond price when we have the cash amounts to be
paid and the dates when they are to be paid. Each cashflow is valued to today
using our bootstrapped yield curve (the yield curves in RAZOR store spot
discount factors to enable faster cash flow discounting).
In RAZOR we take the pessimistic view with callable bonds, assuming that they
will not be called.
Let
Ci = the cashflow at time i
P = the bond’s principal
r = the coupon rate of the bond
d i = the number of days in the coupon period i
D = the number of days in the year
n = the number of coupon payments
Vbond = the value of the bond

©2009 Razor Risk Technologies Page 229 of 373

Razor Financial Principals

di
Ci = Pr
D
Then
n
Vbond = Pdf n + ∑ Ci df i
i=0

14.6.5 Spread Calculation

If a bond is priced against a yield curve, a spread is needed so that the
bond can be properly priced to the market. This is due to the yield curve
being typically composed of interest rate instruments that have a
different credit profile from the bond in question for example: LIBOR
versus the US treasury rate. The spread can be recovered from the bond’s
market price and the yield curve using an iterative solver method.

Define:
Pc = clean price of a bond given explicitly from the market
P ( y ) = market bond price based on yield curve
ai = accrued interest per $1 notional or face value
s = spread
N = accrued interest per $1 notional or face value

Method:
From the yield curve, the market value of the bond price is: NP ( y ) . The
value of the given bond with accrued interest is: N ( P c + ai ) . To calculate
the spread, we solve for s iteratively using a bisection method such that:
NP ( y + s ) = NP ( P c + ai ) .

Unless it is provided the initial guess for spread is fixed at 1%.

14.7 Bonds – Capital Indexed

14.7.1 Description of Instrument
A capital indexed bond is a bond where the PVd cashflows of the bond is
indexed in line with an increase in the Consumer Price Index from the date of
issue of the security to the settlement date. I.e, the yield is expressed in real
terms. Thus, an indexation factor is required.

14.7.2 XML Representation

Using the bond schema for the Vanilla bond, the floating rate multiplier
schema is used to represent the historical inflation rates.
An inflation Adjustment indicator is set to determine which type of inflation
adjusted bond it is.
©2009 Razor Risk Technologies Page 230 of 373
Razor Financial Principals

14.7.3 Credit Implications

The credit implications are the same as a normal bond.

14.7.4 Pricing
The indexation factor is determined by the percentage change of the CPI from
the date of issue of the security to the settlement date.
K t = K t −1 (1 + P ) K is the indexation factor at next interest payment date.
P is the average percentage change in the Consumer Price Index over the two
quarters ending in the quarter which is two quarters prior to that in which the
next interest payment falls (for example, if the next interest payment is in
November, p is based on the average movement in the Consumer Price Index
over the two quarters ending the preceding June quarter)

 CPI t 
 − 1
 CPI t − 2  where CPI is the Consumer Price Index for the second
t
2
quarter of the relevant two quarter period, and CPI t − 2 is the Consumer
Price Index two quarters previously.
If the settlement date occurs between two payment dates then the indexation
factor is discounted back from the next payment date.

i.e. = K t (1 + P )
f
−
where f is number of days from next payment date and d is
d

the number of days between next payment date and previous payment date.

Pricing Formula:
The Market Value of a CIB is calculated according to the formula:
f /d
 
 
MV = 

v 
p 
[ ]
∗ g ( x + a n ) + 100v n ∗
K t FV
∗
100 100
 1 + 
 100 
where

v = 1 / (1 + i )

i = Market Yield / (Frequency*100)

f = Number of days from settlement to next coupon date

d = Number of days between previous and next coupon dates

p = Average percentage change in CPI over the two quarters ending in the
quarter two quarters before that in which the next interest payment occurs =
[CPI t / CPI t −2 − 1] ∗100 / 2 . For example: if next coupon payment is in August
2004, CPI t will be the 2004 March quarter CPI and CPI t − 2 will be the 2003
September quarter CPI .

©2009 Razor Risk Technologies Page 231 of 373

Razor Financial Principals

g = Coupon rate / Frequency

x = 1 if there is an interest payment at the next coupon date; 0 otherwise

n = Number of coupon periods from next coupon date to maturity

an = (1 − v )/ i
n

Kt = Nominal value of principal at next coupon date = K t −1 ∗ [1 + p / 100] .

K t −1 = Nominal value of principal at previous coupon date. If there was no previous

coupon date, K t −1 = $100 . Rounded to 2 decimal places.

FV = Face Value

14.8 Bonds – Indexed Annuity

14.8.1 Description of Instrument
An Indexed Annuity bond is one where the future annuities are multiplied by an
indexation factor, and present valued.

14.8.2 XML Representation

Using the bond schema for the Vanilla bond, the floating rate multiplier
schema is used to represent the historical inflation rates.
An inflation Adjustment indicator is set to determine which type of inflation
adjusted bond it is.

14.8.3 Credit Implications

The credit implications are the same as a normal bond.

14.8.4 Pricing
CPI t
The indexation factor is IF = . If the settlement date occurs
CPI t −1
between two payment dates then the indexation factor is discounted
back from the next payment date.
f
−
 CPI t  d
i.e. =   where f is number of days from next payment date and d is
 CPI t −1 
the number of days between next payment date and previous payment date.

Pricing Formula:
The Market Value of an IAB is calculated according to the formulae:

©2009 Razor Risk Technologies Page 232 of 373

Razor Financial Principals

f /d
v
∗ (Z + a n ) ∗ Bt −1 ∗ q ∗
FV
MV =   if the CPI value for the next coupon
q 100
payment is not yet known.
f /d
v
∗ (Z + a n ) ∗ Bt ∗
FV
MV =   if the CPI value for the next coupon
q 100
payment is known

where

v = 1 / (1 + i )

i = Market Yield / (Frequency*100)

q = Quarterly inflation factor = CPI t / CPI t −1 (with a minimum value of 1)

CPI 0 = Base CPI

f = Number of days from settlement to next coupon date

d = Number of days between previous and next coupon dates

Z = 1 if the price includes the next payment; 0 otherwise

n = Number of coupon periods from next coupon date to maturity

an = (1− v )/ i
n

Bt = Annuity Payment at time t = B0 ∗ CPI t / CPI 0

B0 = Base Annuity Payment

FV = Face Value

14.9 Inflation Linked Bond with Indexation Lag

14.9.1 UK ILG Eight-Month Indexation Lag
Valuation of UK ILG Eight-month Indexation Lag
These are bonds whose coupon and principal payments are linked to the UK RPI index,
assuming an eight-month lag.

Valuation Method

This method relates to UK ILGs issued prior to 1 January 2005.

©2009 Razor Risk Technologies Page 233 of 373

Razor Financial Principals

a) The MTM calculation is similar to a regular coupon paying bond with the
difference being that the coupon and principal repayments are scaled by
an “Indexation Coefficient” based on the UK Retail Price Index (RPI).
b) In order to calculate the Indexation Coefficient, the following are
required:
i. The RPI applicable when the bond was originally issued, the “Base
RPI”. This will be sourced from bond static data.
ii. The RPI applicable for the coupon payment concerned. Note the
8-month lag means that the RPI quote is taken for the month 8-
months prior to the coupon payment. RPI quotes will be sourced
from market data.
c) For the current coupon period the RPI is known. For future coupons and
the principal repayment, the RPI is forecast using an assumed annual
rate of inflation. This is the parameter value to be modelled, for
example, the forecast inflation rate is assumed to be 3% pa.
d) The NPV for an ILG with Eight-Month Indexation Lag is calculated as
follows:
n
Trade NPVILG8 = ∑ I i .Ci .Fi .DFi + P.Fn .DFn
i =1
where,
Fi = Indexation Coefficient for ith coupon (rounded to
5 decimal places)
RI imi − 8
=
RI Base
RI imi −8 = Reference Index Value for ith coupon, being 8
months prior to mi (if quote is not available, then
value is approximated as described below)
mi = month/year in which the ith coupon payment
falls
RIBase = Reference Index Value applicable at bond issue
date
n = number of future coupon payments
Ci = ith coupon payment based on appropriate
accrual basis
Ii = ex-interest indicator for ith coupon payment; Ii=0
for ex-interest, Ii=1 for cum-interest.
P = bond par amount
DFi = bond curve discount factor from valuation date
to ith coupon date

e) For clarification, the Reference Index Value relating to a coupon

payment in month mi, is the index value for month mi -8. For example,
the Reference Index Value applicable to a coupon payment date on any
day in Sep-09 is the index value for Jan-09.
f) For future coupon payments where the relevant Reference Index Value
m −8
has not yet been published, RI i i is approximated using a growth rate
assumption applied to the most recently published index value, as
follows:

©2009 Razor Risk Technologies Page 234 of 373

Razor Financial Principals

mi − 8− M
RI imi −8 = RI M (1 + z ) 12

where,
RIM = most recently published RPI quote
mi = month/year in which the ith coupon payment
falls
M = month/year of the most recently published RPI
z = assumed index annualized growth rate (3%)

g) Accrued Interest for an ILG with Eight-Month Indexation Lag is

calculated as follows:

AccInt ILB 8 = Real Accrued Interest x Index Coeff for next coupon
 No. of actual accrued days up to settlement date 
C1 .  * F1
=  No. of actual days in coupon period 
where,
C1 = next coupon payment (ie. coupon rate / coupon
frequency x notional)
F1 = Indexation Coefficient for the next coupon

h) UK ILGs issued prior to 1 July 2005 (ie. 8 month index lag) trade on an
inflation-adjusted basis. Accordingly, the price of such gilts reflects
inflation since it was originally issued, eg. most of these ILGs trade at a
price well in excess of £100 per £100 nominal.
i) The following examples describe the application of the rules for
determining the appropriate Indexation Coefficient:
i. Given a Valuation Date of 10 April 2009, a Base RPI of 102, the
most recent RPI published for index month Feb-09 of 125, and a
next coupon payment date of 1 September 2009 (or any other day
in Sep-09):
mi = Sep-09
mi-8 = Jan-09
RI imi −8 = RPI for index month Jan-09 (published during Feb-
09), eg. 120
M = Feb-09
Fi = 120 / 102 = 1.176471

ii. Given a Valuation Date of 10 April 2009, a Base RPI of 102, the
most recent RPI published for index month Feb-09 of 125, and an
ith coupon payment date of 1 March 2012 (or any other day in
Mar-12):
mi = Mar-12
mi -8 = Jul-11
M = Feb-09
RIM = 125
z = 0.03
RI imi −8 = 125 * ( 1 + 0.03 ) ^ ( 29 / 12 )
= 134.255877
Fi = 134.255877 / 102 = 1.316234

©2009 Razor Risk Technologies Page 235 of 373

Razor Financial Principals

14.9.2 Inflation-Linked Bonds with Three-Month Indexation Lag

(“Canadian Model”)
Valuation of ILBs with Three-Month Indexation Lag
These are inflation-linked bonds (ILBs) whose coupon and principal payments
follow the standard “Canadian Model” which assumes a three-month indexation
lag.
For example, the following method applies to the following securities:

Country Reference Issuing

Index Authority
UK (issued Aggregate ONS
after 2005) RPI
France CPI ex- INSEE
tobacco
Germany HICP ex- EUROSTAT
tobacco
Italy HICP ex- EUROSTAT
tobacco
Greece HICP ex- EUROSTAT
tobacco

RPI refers to the Retail Price Index in the UK.

CPI refers to the Consumer Price Index in France.
HICP refers to the Harmonized Index of Consumer Prices in the Euro-Zone.

Valuation Method

a) The valuation method for ILBs with three-month indexation follows the
de facto industry standard and is commonly referred to as the “Canadian
Model”.
b) The Indexation Coefficient under the Canadian Model:
(i) is based on index quotes 3-months and 2-months prior to the
coupon payment,
(ii) is calculated using linear interpolation.
c) The NPV for an ILB with Three-Month Indexation Lag is calculated as
follows:
n
Trade NPVILB3 = ∑ I i .C i .Fi .DFi + P.Fn .DFn
i =1
where,
Fi = Indexation Coefficient for ith coupon (rounded to
5 decimal places)

©2009 Razor Risk Technologies Page 236 of 373

Razor Financial Principals

RI iinterp
=
RI Base
RI iinterp = Reference Index Value for ith coupon based on
linear interpolation as described below
RIBase = Reference Index Value applicable at bond issue
date
n = number of future coupon payments
Ci = ith coupon payment based on appropriate
accrual basis
Ii = ex-interest indicator for ith coupon payment; I=0
for ex-interest, I=1 for cum-interest.
P = bond par amount
DFi = bond curve discount factor from valuation date
to ith coupon date

interp
d) The interpolated Reference Index Value for the ith coupon, RI i , is
determined using the following rules:
1. The Reference Index Value applicable to the first day of the month
in which the coupon falls, mi, is the index value for month mi -3.
For example, the Reference Index Value applicable to June 1 is the
index value for March.
2. The Reference Index Value for any other day in month mi is
calculated by linear interpolation between the index values for
month mi -3 and month mi -2, according to the following formula:

 d −1
( )
RI iinterp =  i m  RI imi − 2 − RI imi −3 + RI imi −3
 NDi 
i

where,

mi = month/year in which the ith coupon payment

falls
di = day of month in which the ith coupon payment
falls
ND imi = number of days in month, mi, in which the ith
coupon payment falls
RI imi −3 = Reference Index Value for month mi -3
RI imi − 2 = Reference Index Value for month mi -2

3. When calculating the interpolated Reference Index Value, if any

RI imi −3 or RI imi − 2 have not yet been published, these values are
approximated using a growth rate assumption applied to the most
recently published index value, as follows:
mi − Lag − M
RI imi − Lag = RI M (1 + z ) 12

where,

RIM = Most recently published RPI quote

mi = month/year in which the ith coupon payment
falls
©2009 Razor Risk Technologies Page 237 of 373
Razor Financial Principals

Lag = number of months prior to mi for which the index

value is required, ie. either 2 or 3
M = Month/Year of the most recently published RPI
z = assumed index annualized growth rate (3%)

e) Accrued Interest for an ILB with Three-Month Indexation Lag (Canadian

Model) is calculated as follows:
AccInt ILB 3 = Real Accrued Interest x Current Index Coefficient
 
AccInt ILB 3 = C1 . * No. of actual accrued days up to settlement date  * FRS
 No. of actual days in coupon period 
where,
C1 = next coupon payment (ie. coupon rate / coupon
frequency x notional)
FRS = Indexation Coefficient as at Regular Settlement
date
interp
RI RS
=
RI Base
interp
RI RS = Interpolated Reference Index Value as at Regular
Settlement date based on linear interpolation

= 
ND
(
 d RS − 1  mRS − 2
)m RS − 3
 RI RS − RI RS
m RS 
m RS − 3
+ RI RS
 RS 

f) ILBs based on the Canadian Model trade on a real price basis. That is,
the impact of inflation since the bond was first issued is effectively
stripped out from the price for trading purposes, and hence will
typically have a quoted price around par. However, settlement prices
will be on the basis of the inflation-adjusted price. The relationship
between traded real price and settlement inflation-adjusted prices is as
follows:
Inflation-Adjusted Dirty Price = (Real Clean Price + Real Acc Int) x FRS
where,
FRS = Indexation Coefficient as at Regular Settlement
date, calculated as per (e) above

g) The following examples describe the application of the rules for

determining the appropriate Indexation Coefficient:
i. Given a Valuation Date of 25 March 2009, a Base RPI of 102, the
most recent RPI published for index month Feb-09 of 125, and a
next coupon payment date of 1 April 2009:
mi = Apr-09
mi -3 = Jan-09
RI imi −3 = RPI for index month Jan-09 (published during Feb-
09), eg. 120
RI iinterp = RI imi −3 = 120
Fi = 120 / 102 = 1.176471

©2009 Razor Risk Technologies Page 238 of 373

Razor Financial Principals

ii. Given a Valuation Date of 25 March 2009, a Base RPI of 102, the
most recent RPI published for index month Feb-09 of 125, and a
next coupon payment date of 20 April 2009:
mi = Apr-09
mi -3 = Jan-09
mi -2 = Feb-09
RI imi −3 = RPI for index month Jan-09, eg. 120
RI imi − 2 = RPI for index month Feb-09, eg. 125
di = 20
ND imi = 30
RI iinterp = ( 20 -1 ) / 30 * ( 125 – 120 ) + 120
= 123.166667
Fi = 123.166667 / 102 = 1.207516

iii. Given a Valuation Date of 25 March 2009, a Base RPI of 102, the
most recent RPI published for index month Feb-09 of 125, and a
next coupon payment date of 21 July 2009:
mi = Jul-09
mi -3 = Apr-09
mi -2 = May-09
M = Feb-09
RPIM = 125
z = 0.03
RI imi −3 = 125 * ( 1 + 0.03 ) ^ ( 2 / 12 ) = 125.617328
RI imi − 2 = 125 * ( 1 + 0.03 ) ^ ( 3 / 12 ) = 125.927134
di = 21
ND imi = 31
RI iinterp = ( 21 -1 ) / 31 * (125.927134 – 125.617328 ) +
125.617328
= 125.817203
Fi = 125.817203 / 102 = 1.233502

14.10 Perpetuity Bond

Valuation of Perpetuity
Razor system shall value a bond that pays coupons in perpetuity using the Net
Present Value method (NPV).

Valuation Method
a) Future cashflows up to the last discount curve point will be discounted
directly using curve DFs.
b) Coupons beyond the last curve point will be valued using a yield-to-
perpetuity calculation.
c) The NPV for a perpetuity is calculated as follows:

©2009 Razor Risk Technologies Page 239 of 373

Razor Financial Principals

n
Cn
Trade NPVPerp = ∑ Ci DFi + .DFn
i =1 r
where,
Ci = ith next coupon payment based on appropriate accrual basis.
n = number of coupon payments up to last bond discount curve point.
DFi = bond curve discount factor from valuation date to ith coupon date.
r = DFLast −
1
YFLast . f − 1 (using DF) or
r = (1 + z Last ) f − 1 (using annualized zero rate).
1

YFLast = year fraction between valuation date and last date on bond discount
curve.
DFLast = bond curve discount factor from valuation date to last date on bond
discount curve.
zLast = annual compounding zero rate at last date on bond zero curve.
f = coupon frequency, ie. Number of coupons per year.

14.11 Cash Bond

We first define some notation.

Terminology - Dates
TD = Trade Date
V = Valuation Date
TS = Trade Settlement Date
RS = Regular Settlement Date
CD j = Exdividend or coupon date if exdivided date equals coupon date
Terminology – Accruals

α V , RS = Appropriat e Accrual from V to RS =

(RS − V )
λ
α V ,TS = Appropriat e Accrual from V to TS =
(TS − V )
λ
(TS − CD j )
α CD TS = Appropriate Accrual from CD j to TS =
j
λ
α RS ,TS = Appropriat e Accrual from RS to TS =
(TS − RS )
λ
λ = Appropriat e Currency Accrual Basis (this should be driven by the bond
Trade XML)
The difference between the days is driven by the daycount convention.
Terminology - Prices
PV = Bond Clean Price(s) at the Valuation Date
AI RS = Accrued Interest to RS
Terminology - Rates
LDFt = Libor Discount Function at t
RIFt = Repo Interest Function at t

©2009 Razor Risk Technologies Page 240 of 373

Razor Financial Principals

 LDFV  1 
LiborV ,TS = Libor Curve Derived Rate from V to TS =  − 1  
 LDFTS   α V ,TS 

 LDFV  1 
LiborV , RS = Libor Curve Derived Rate from V to RS =  − 1  
 LDFRS  α V , RS 

 RIFRS  1 
R RS ,TS = Repo Curve Derived ForwardRate from RS to TS =  − 1  
 RIFTS   α RS ,TS 

 RIFCD  1 
RCD ,TS = Repo Curve Derived Forward Rate from CD to TS =  − 1  
 RIFTS   α CD ,TS 
Terminology - Cashflows

C j = Cashflow j

n = Number of Cashflows from V to TS

N = Bond Nominal Amount

DRS = Bond Dirty Value at RS = [N (PV + AI RS )]

B = Forward Value of Bond Dirty Value, at TS = D RS [1 + R RS ,TS α RS ,TS ]

[ ]
n
FC = Sum of Forward Values of Interim Coupons, at TS = ∑ C j 1 + RCD j ,TS α CD j ,TS
j

For clarity purposes: if RS < exDivDate < TS then FC <>0 otherwise FC=0

FCV = Future Cash Value (s ) - this is equivalent to the bond cash

settlement/payment amount.

Cash Bond
This is a separate valuation methodology for forward settling bonds. Cash bonds
settled on or after the ex-dividend, or coupon date if no ex-dividend date exists,
will be deemed to be ex-coupon. Therefore, if the ex-divided/coupon date
occurs between RS and TS , then FC will have some value otherwise FC = 0 . If
both the ex-dividend date and the coupon date occur between RS and TS then
the coupon must obviously only accrue interest from the coupon date to TS
rather than from the ex-dividend date. However if the ex-divided date falls
before TS while the coupon date falls after TS then the coupon will need to be
discounted back to TS .

NPV for the buyer of the cash bond trade is:

©2009 Razor Risk Technologies Page 241 of 373

Razor Financial Principals

 1   1   1 
Trade NPV =  B  −  FC  −  FCV 
 1 + LiborV ,TS αV ,TS   1 + LiborV ,TS αV ,TS   1 + LiborV ,TS αV ,TS 

where:
PV in DRS = PVBid = Bond Clean Bid Price at the Valuation Date
Formatted: Body Text 2, Line
spacing: Multiple 1.1 li
NPV for the seller of the cash bond trade is:

 1   1   1 
Trade NPV =  FCV  −  B  −  FC 
 1 + LiborV ,TS α V ,TS   1 + LiborV ,TS αV ,TS   1 + LiborV ,TS αV ,TS 

where:
PV in DRS = PVAsk = Bond Clean Ask Price at the Valuation Date

14.12 Floating Rate Notes

14.12.1 Description of Instrument
A floating rate note (FRN) is a debt instrument similar to a bond in that it
makes regular interest payments. Unlike a bond, however, the FRN pays an
amount based on a floating benchmark rate and a fixed margin.

14.12.2 XML Representation

The FRN is represented using FPML’s InterestRateStream structure.

14.12.3 Pricing
In the case of a bond, valuation consists of present valuing a series of known
cash flows. With an FRN, the only cash flows we know are the coupon interest
payment at the end of the current interest payment period and the repayment
of the principal at maturity — all of the intermediate coupon payments are
unknown.

There are two methods to evaluate the price of floating rate notes. The first
method is to use the discounting of cash flows similar to the floating side of an
interest rate swap and another method is to use a closed form pricing formula.
Note that MBS has the same payment structure as FRN except that MBS has a
chance of early repayment. Due to the similarity in the payment structure of
these two types of contracts, the pricing methodologies are similar. We shall
discuss the pricing of both FRN and MBS in the following.
1. Discounting Cash Flows

©2009 Razor Risk Technologies Page 242 of 373

Razor Financial Principals

For FRN, the valuation is the same as floating leg of an interest rate swap.
The intermediate coupon payments can be calculated from the forward
curve. For MBS, it is the same as FRN except that we do not know the
length of the contract. We can simply use the weighted average life ( WAL
) as a proxy for the length of MBS. The maturity date used for the MBS is
the WAL if it falls on a coupon date, otherwise it will be set to the coupon
date following the WAL . The valuation is then carried out the same way as
the FRN. The only difference is that we multiply the final discounted value
of the MBS by the bond factor ( BF ) to obtain the final price.
2. Closed Form Pricng Formula
We first define some notations:
FV = Face value.

 IM  d
C = Next coupon amount per $100 FV ,  BBSW + × .
 100  365
IM = Issue margin.
TM = Trading margin.
f = Number of days from settlement to next coupon date.

d = Number of days between previous and next coupon dates.

1 − (1 + i )
−n

a = Annuity factor .
i
TM
s+
i = 100 .
freq ×100
s = Swap rate.
r = Discount rate.
n =
Number of coupon periods between the next coupon date and the WAL date
WAL − NCD
with the appropriate n rounding convention = × Frequency .
365.25

The value of an MBS is:

 IM − TM 
100 + C +  × a
 100 × freq  × FV × BF .
 TM   f  100
1+  r + × 
 100   36500 
The value of a FRN is by substituting time to maturity for WAL and 1 for
BF .

©2009 Razor Risk Technologies Page 243 of 373

Razor Financial Principals

14.13 Repurchase Agreements

14.13.1 Description of Instrument
Repurchase Agreements involve the selling of securities to a counterparty and
simultaneously agreeing to buy the securities back on a predetermined date at
a predetermined price based on the Repo rate. Effectively this results in
borrowing cash for the period and providing the bond as collateral.
A reverse Repo is the transaction of opposite side.
If the repo includes an underlying security, it is priced separately first, then
included in the pricing of the repo. At the moment, we support repo with an
underlying irBond. Users can specify a curve (called the bond curve) to use for
pricing the underlying bond instead of the reval curve that is being used
currently. One additive and one multiplicative factor can be specified to be
applied to the bond curve before pricing takes place.

Razor values the following Eligible Securities for specific or general collateral:
a) Bonds
i. Fixed rate.
ii. Zero Coupon.
iii. Perpetual.
b) Strips.
c) Bills
i. Discount yield-quoted.
ii. Money market yield-quoted.
d) Inflation -Linked Bonds
i. UK ILG Eight-Month Indexation Lag.
ii. UK ILG Three-Month Indexation Lag (Canadian Model).
iii. EU ILB Three-Month Indexation Lag (Canadian Model).

14.13.2 XML Representation

Repurchase agreements are modelled using two trades The purchase/sale of a
bond on the near date and the sale/purchase of the bond on the far date.
Repo is implemented as a structural deal combining the short position in
underlying asset and long cash position reflecting the borrowed amount at
Repo start date and a forward transaction of underlying asset repurchase at
repo termination date. Coupon payments during the life of the Repo
Transaction are optional

14.13.3 Credit Implications

Repurchase agreements have similar credit implications to the underlying
security. In the case where the underlying security is a debt security (bond),
©2009 Razor Risk Technologies Page 244 of 373
Razor Financial Principals

there is counterparty risk with seller of Repo and an underlying security issuer
risk. Under a classical Repo agreement the borrower of the cash (lender of the
underlying asset) may still receive any cash flows generated by the asset prior
to termination of the deal and that passed on by Repo seller.

14.13.4 Pricing

Define
S1 = Cash payment for the spot leg at t1.
S2 = Cash payment for the forward leg at t1.
B1 = Bond price for the spot leg per 100 at t0.
B2 = Bond price for the forward leg per 100 at t0.
RR = Repo rate from repo curve.
RL = Discount rate from RF curve.

The repo valuation formula is

MTM = MTM Leg1 + MTM Leg 2

 
 B1  t1 − t0    1 
MTM Leg1 =  S1 − N 1 + RR  ×  
 100  360   1 + R t1 − t0 

L
360 
 
  t2 − t1  B2  t2 − t0    1 
MTM Leg 2 =  − S 2 1 + R + N  1 + RR   × 
  360  100  t −
360   1 + R 2 0  t

L
360 

Points to note:
Although the above uses ACT/360 the day count convention should follow trade
and market configuration as usual.
The bond price B is obtained by pricing the bond (discounted cashflows) using
the bond curve and value date t0. It is the dirty price at t0.
Given the selling and purchase prices of the bond are not the same due to bid-
ask spread, the above pricing formula has made spread adjustment to cater for
this mismatch.

14.14 General Collateralised Repo

There are two parties to a repo trade: A (the seller) and B (the buyer). On the
trade date, the two parties enter into an agreement whereby on a set date, the
settlement date, A will sell to bank B a nominal amount of securities in
exchange for cash. The agreement also demands that on the termination date B
will sell identical stock back to A at the previously agreed price, and
consequently Bank B will have its cash returned with interest at the agreed repo

©2009 Razor Risk Technologies Page 245 of 373

Razor Financial Principals

rate. If a coupon is paid it will be handed over to the seller on the coupon
payment date.

14.14.1 Sterling GC Trades

Valuation of Sterling GC trades
a) Razor system shall value Sterling GC trades using the NPV method.
b) For the long collateral holder of a Sterling GC trade, the NPV is
calculated follows:

 TS − TS1 TS − max(VD, TS1 ) 

Trade NPVGC = S1  R 2 − rmax(VD ,TS1 ),TS2 2 .DF2
 Y Y
where, VD = valuation date
TS1 = trade opening leg date
TS2 = trade maturity leg date
S1 = GC trade notional, ie. Initial cash payment value
R = trade repo rate (retrieved from the xml trade)
ri,j = market forward repo rate between time i and j
Y = appropriate year accrual basis
DF2 = libor discount factor from VD to TS2

c) For the short collateral holder of a Sterling GC trade, the NPV is

calculated follows:
 TS − TS1 TS − max(VD, TS1 ) 
Trade NPVGC = − S1  R 2 − rmax(VD,TS1 ),TS2 2 .DF2
 Y Y

14.14.2 Euro GC Trades

Valuation of Euro GC trades
a) Razor system shall value Euro GC trades using the NPV method.
b) For the long collateral holder of a Euro GC trade, the NPV is calculated
exactly as per a Sterling GC trade, given as follows:

 TS − TS1 TS − max(VD, TS1 ) 

Trade NPVGC = S1  R 2 − rmax(VD ,TS1 ),TS2 2 .DF2
 Y Y
where, VD = valuation date
TS1 = trade opening leg date
TS2 = trade maturity leg date
S1 = GC trade notional, ie. initial cash payment value
R = trade repo rate (retrieved from the xml trade)
ri,j = market forward repo rate between time i and j
Y = appropriate year accrual basis
DF2 = euribor discount factor from VD to TS2

c) For the short collateral holder of a Euro GC trade, the NPV is calculated
exactly as per a Sterling GC trade, given as follows:

 TS − TS1 TS − max(VD, TS1 ) 

Trade NPVGC = − S1  R 2 − rmax(VD,TS1 ),TS2 2 .DF2
 Y Y

©2009 Razor Risk Technologies Page 246 of 373

Razor Financial Principals

14.15 Bill Index Deposits

Product Decscription

14.15.1 Guaranteed Bill Index Deposits

The Guaranteed Bill Index Deposit (GBID) is a short-term deposit product

offering guaranteed
returns every 90 days determined by reference to an index. If the index
performance is positive
for the period of the deposit, the return will be positive. However, if the
index performance is
negative for the period of the deposit, the return will be negative and an
amount equivalent to the
negative return will become payable to the issuer. The GBID provides a
money market return without
the need to actively manage the cash and bill market investment
alternatives.

14.15.2 Protected Bill Index Deposits

The Protected Bill Index Deposit (GBID) is similar to the Guaranteed Bill
Index Deposit (GBID) but which also protects the investors capital, by
flooring any return to zero and so protecting the investors from paying for
any negative return on the index .

GBID maturity dates will be the quarter end dates of March, June,
September and December. Interest is calculated by reference to the daily
rate of return of the UBS Australian Bank Bill Index. Interest is paid
quarterly in arrears on the maturity date.

Finmark Schema and Example

This finmark schema is supported by the bond schema.

Example
- <deal>
- <trade>
- <tradeHeader>
<tradeDate>2007-09-13</tradeDate>
<tradeType>CDBD</tradeType>
<counterparty>56789</counterparty>
<internalUnit>1234</internalUnit>
<buySell>SELL</buySell>
<status>OPEN</status>
</tradeHeader>
- <extensions>
- <extension>

©2009 Razor Risk Technologies Page 247 of 373

Razor Financial Principals

<value>HISTORICAL</value>
<name>HistoricalRateSet</name>
</extension>
- <extension>
<value>UBSBBI</value>
<name>AUD BBI 1D</name>
</extension>
</extensions>
- <product>
- <bond>
- <bondStream>
<payerPartyReference href="56789" />
<receiverPartyReference href="1234" />
- <calculationPeriodDates id="CalcPeriodDates0">
- <effectiveDate>
<unadjustedDate>2007-08-13</unadjustedDate>
- <dateAdjustments>
<businessDayConvention>MODFOLLOWING</businessDayConvention>
</dateAdjustments>
</effectiveDate>
</calculationPeriodDates>
- <calculationPeriodAmount>
- <calculation>
- <notionalSchedule>
- <notionalStepSchedule>
<currency>AUD</currency>
<initialValue>100000</initialValue>
</notionalStepSchedule>
</notionalSchedule>
- <floatingRateCalculation>
<floatingRateIndex>BillIndex</floatingRateIndex>
</floatingRateCalculation>
<dayCountFraction>ACT/360</dayCountFraction>
</calculation>
</calculationPeriodAmount>
- <principalExchanges>
<intermediateExchange>false</intermediateExchange>
<initialExchange>false</initialExchange>
<finalExchange>false</finalExchange>
</principalExchanges>
</bondStream>
<productType>GBID</productType>
<securityId>GBID</securityId>
</bond>
</product>
</trade>
- <dealHeader>
<dealType>CDBD</dealType>
<dealDate>2009-02-19</dealDate>
<status>OPEN</status>
</dealHeader>
</deal>

©2009 Razor Risk Technologies Page 248 of 373

Razor Financial Principals

Calculation of UBS Warburg Bank Bill Index

The UBS Warburg Bank Bill Index is based on a series of 13 91-day bank
bills that mature at 7 day intervals (every Tuesday). Upon maturity, the
face value of the maturing bill is reinvested in a new 91-day bill at its
discounted price.

The Index is calculated daily by valuing these 13 bank bills. The yield for
each bill is determined by interpolation between the 10:15 am cash rate,
1-month BBSW and 3-month BBSW rates, as shown in the table below. The
Index is calculated by summing the discounted values of the 13 bills and
dividing by 10000.

Days to Maturity Interpolated Rate

0-7 R1
8-14 2/3R1 + 1/3R2
15-21 1/3R1 + 2/3R2
22-28 R2
29-35 8/9R2 + 1/9R3
36-42 7/9R2 + 2/9R3
43-49 6/9R2 + 3/9R3
50-56 5/9R2 + 4/9R3
57-63 4/9R2 + 5/9R3
64-70 3/9R2 + 6/9R3
71-77 2/9R2 + 7/9R3
78-84 1/9R2 + 8/9R3
85-91 R3

R1 = cash rate.
R2 = 1 month BBSW rate.
R3 = 3 month BBSW rate.

Example
Cash Rate = 6.5%.
1 Month BBSW Rate = 7.07%.
3 Month BBSW Rate = 7.0983%.
Current Date = 2007-9-13.
Yield Rate = Simple Interest.
Day Convention = ACT/365.

Maturity No. of Yield Rate Face Value DF PV

Date Days
2007-11-20 68 0.0708886667 4599097 0.9869654866 4539150.01
2007-11-27 75 0.0709201111 4590987 0.9856366856 4525045.21
2007-12-04 82 0.0709515556 4595705 0.9843102894 4523599.72
2007-12-11 89 0.0709830000 4782951 0.9829862928 4701575.27
2007-9-18 5 0.0650000000 4678551 0.9991103812 4674388.87
2007-9-25 12 0.0669000000 4646588 0.9978053749 4636390.48
2007-10-2 19 0.0688000000 4599276 0.9964314106 4582863.07
©2009 Razor Risk Technologies Page 249 of 373
Razor Financial Principals

2007-10-9 26 0.0707000000 4632012 0.9949890715 4608801.32

2007-10-16 33 0.0707314444 4590717 0.9936457371 4561546.38
2007-10-23 40 0.0707628889 4587825 0.9923048376 4552520.94
2007-10-30 47 0.0707943333 4598925 0.9909663678 4557380.00
2007-11-6 54 0.0708257778 4620371 0.9896303227 4572459.24
2007-11-13 61 0.0708572222 4593062 0.9882966973 4539308.01
Sum 59563722.01

Valuation
There are 4 methods (parameter “MTMMethod”) for modeling GBID/PBID
products.
For the get MTM function – the value returned for MTMMETHOD of …
MTM – cashflow based, notional and interest – overnight or at maturity - pv’d to value date
of flows
FACE – cashflow based, notional and interest – overnight or at maturity - sum of face value
of flows
STEP – step schedule – returns current balance (notional outstanding)
BBI – method calculates return on BBI Index

Interest Calculation

The GBID/PBID net interest calculation is the sum of the historic accrued
interest and forecast interest calculation.

The balance at time t will be calculated from the original deposit and any
notional adjustments which have occurred before time t

The daily interest is calculated as follows

 Daily return 
InterestT = BalanceT ×  
 100 
 BBI T 
Daily GBID return =  − 1 × 100, Daily PBID return = max(Daily GBID return, 0 )
 BBI T −1 
Where
BBI T = UBS Australian Bank Index for Day T
BBI T −1 = UBS Australian Bank Index for Day T − 1

The BBI Index is only calculated and provided by TCV for business days.
Interest is calculated daily for each day of the month including non-
business days. For non business days the interest is calculated from the
last known daily return.

Today’s interest amount will be calculated from the BBI observation as

set yesterday (normally 11am).

©2009 Razor Risk Technologies Page 250 of 373

Razor Financial Principals

All previous daily interest amounts are calculated from the historical
observations for the BBI. If the BBI Index History is not provided then
accrued interest will be set to zero.

For historic interest to be calculated for time t – the rate of return currently is BBI(t)/BBI(t-1)
BUT should be BBI(t-1)/BBI(t-2) – so interest is P(t-1)*( BBI(t-1)/BBI(t-2) – 1)
No interest is calculated for start date – ie on day of initial deposit the user does not earn
interest.
On start date the first days interest can only be forecast.
On first days interest day – the interest is calculated from BBIs observed at COB on start date
ie t-1 and previous day ie t-2
Forecast accrued interest is implemented in the setting up of the transition cache – so there
is no impact from forecast accrued interest for VaR or stress/sensitivity – ie its not
dependent on the current scenario and only the base scenario – needs to be fixed – but
lower priority
Forecast and Historic daily interest mostly caters for changing notional correctly – except on
the days of notional change – interest should be calculated on previous notional
Weekend handling for historic interest calculation – if observation for t-1 is not known due
to being a weekend the previous return is applied for interest at time t

The interest accrued today is calculated from yesterdays BBI return

applied to yesterdays outstanding principal (not today’s).

Calculation of Interest

– daily interest calculation for day at time

 – outstanding balance at time

 – deposit start date
 – value date
 – deposit maturity date
 , – annualised rate of return for period from  to

 – discount factor from time
from the BBI curve

 , – forward discount factor from time
to time  from the BBI
curve

NOTE:   ,

NOTE: for PBID the daily return is floored at 1 – ie calculated daily

interest is floored at zero

Historical Accrued: for time     1 to 


      1


Forecast: for time     1 to 

©2009 Razor Risk Technologies Page 251 of 373

Razor Financial Principals


,     1!


Net Interest: is the sum of daily interest already accrued plus sum of the
forecast daily interest
$ '

"  #   # ,
% & %$&

Annex:
In case of interest to anyone, justification for this calculation for forward
forecast daily interest is …

For  , …

 , 

1
 ,  -

(1   ,  ,
365

Hence …
 1
 -

 (1   ,  ,
365

And …
365 
 ,  (  1,

  

The calculated interest for a forward period from time  to

time
is


 ,     , 
365

By substituting for (,
, above, we get


 ,    (  1,


The calculated interest for a forward daily period from time
 1 to time

is then found by setting  
 1

,    (  1,


Example of Interest Calculation

BBI on 24/01/2001 = 5000.

©2009 Razor Risk Technologies Page 252 of 373
Razor Financial Principals

Opening Balance = $10,000,000.

Transaction Details:

Date Transaction Amount Balance

25/01/2001 Deposit 10,000,000 $10,000,000
28/01/2001 Deposit 5,000,000 $5,000,000
31/01/2001 Interest 1,869.81 $1,869.81

Principal = $15,000,000.
Interest = $1,869.81.
Closing Balance = $15,001,869.81.

Calculation of Interest Payable

Date BBI Daily Return Balance Interest

25/01/2001 5000.10 0.002% $10,000,000 $200.00
26/01/2001 5000.15 0.001% $10,000,000 $100.00
27/01/2001 5000.35 0.004% $10,000,000 $399.99
28/01/2001 5000.40 0.001% $15,000,000 $149.99
29/01/2001 5000.80 0.008% $15,000,000 $119.99
30/01/2001 5000.90 0.002% $15,000,000 $299.95
31/01/2001 5001.10 0.004% $15,000,000 $599.89
Total Interest $1869.81

©2009 Razor Risk Technologies Page 253 of 373

Chapter 15
Equities

15.1 Ordinary Shares

15.1.1 Description of Instrument
Ordinary shares give the holder limited liability ownership of the company.

15.1.2 XML Representation

The asset representation of shares allows us to specify either discrete
dividends, or a dividend yield.

Equity Example
<equity>
<instrumentId>ITE</instrumentId>
<description>ITE Ordinary</description>
<effectiveDate>
<unadjustedDate>2003-12-12</unadjustedDate>
</effectiveDate>
<numberShares>100000</numberShares>
<exchangeCode>ASX</exchangeCode>
<currency>AUD</currency>
</equity>

15.1.3 Credit Implications

RAZOR will simulate the price of the stock. Fixed dividend payments may also
be assigned to the stock, or an annualised dividend yield. Because the stock
price is quoted in terms of the currency where the exchange is based that the
stock is traded on, the holding could also incur exchange rate risk.

15.1.4 Pricing
Let
Pccy = price per share in currency ccy

N = number of shares held

Vccy = Exposure in currency ccy
Then
Vccy = NPccy

15.2 Equity Forwards

15.2.1 Basket Equity Forward
A Basket Equity Forward is an agreement to purchase a specified basket
of equity shares for at an agreed future time for an agreed value.
Razor Financial Principals

To support this product, Razor requires updates to the following

components:
Pricing Adapter
XML Schema
Deal Entry Screen

XML Representation
<equity>
<instrumentId>Basket</instrumentId>
<forwardType>Vanilla</forwardType>
<basket>
<basketAsset>
<id>ANZ</id>
<weightingFactor>0.3</weightingFactor>
</basketAsset>
<basketAsset>
<id>TEL</id>
<weightingFactor>0.2</weightingFactor>
</basketAsset>
<basketAsset>
<id>BHP</id>
<weightingFactor>0.5</weightingFactor>
</basketAsset>
</basket>
<numberShares>100000</numberShares>
<exchangeCode>ASX</exchangeCode>
<currency />
<paymentAmount>
<paymentType>SETTLEMENT</paymentType>
<paymentAmount>
<currency>AUD</currency>
<amount>1666000</amount>
</paymentAmount>
<adjustedPaymentDate>2005-12-09</adjustedPaymentDate>
</paymentAmount>
</equity>

Pricing

The Razor pricing adapter for Equity Forwards currently supports the
following methods of pricing for a single share:

Future Equity Price Calculation

Future Equity Price Calculation - Forward Equity Price on Trade
PV(Future Equity Price Calculation - Forward Equity Price on Trade)

Please Note that the Forward Equity Price on the Trade is expected to be
specified as an average equity price across the basket, which is multiplied
by the number of shares to get the total basket value.

©2009 Razor Risk Technologies Page 255 of 373

Razor Financial Principals

Where Future Equity Price Calculation (F(tj)) at the Forward Settlement

Date tj is defined as:

b ( t j )(t j − t )
F (t j ) = S (t )e

Where: S (t ) = Current Market Price of Equity

b(t j ) = rt j − q

And: rt j = zero rate at tj

q = annualized dividend yield

In order to cater for Basket Equity Forwards, the Future Equity Price is
now a Future Equity Basket Price and is required to take all equity shares
that make up the basket into account, hence the Future Equity Basket
Price Calculation becomes:

N
Fb (t j ) = ∑ wi Fi (t j )
i

Where: N = Number of Equities that make up the basket

wi = weight of Equity i (Si) in the basket
bi ( t j )(t j − t )
Fi (t j ) = Si (t )e

And: S i (t ) = Current Market Price of Equity (i)

bi (t j ) = rt j − qi
qi = annualized dividend yield for Equity (i)

XML Schema

The Current Equity Schema (used for Equity Forwards) does not support
basket definitions. An additional tag is required in the Equity schema to
enable the basket definition as shown below, where subsequent
underlying equities and their respective weights toward the basket can be
defined.

Please Note: The weights should be normalized and the number of shares
specified on the trade is required to be equal to the total number of
shares that make up the basket across all underlying equities. Note this is
the same implementation as for Equity Basket Asian Options.

fpmlEquity Schema
Name: Type Occurs Size Description

©2009 Razor Risk Technologies Page 256 of 373

Razor Financial Principals

fpmlEquity Schema
Name: Type Occurs Size Description
instrumentId
1..1 Equity Id or “Basket”
string
description
0..1 The long name of the security
string
basket
0..1 Basket, grouping of assets and weightings
rzmlBasket
forwardType
0..1 The type of the equity forward
string
settlementDays The number of settlement days for this
0..1
positiveInteger equity
numberShares
1..1 Number of shares
double
exchangeCode
1..1 Stock Exchange Code eg:ASX
ExchangeCode
currency
1..1 Currency
fpmlCurrency
paymentAmount
0..1 Payment amount
fpmlPayment

15.2.2 Foreign Equity Forward

A foreign equity forward contract is an agreement to purchase a certain
number of shares of an individual equity at an agreed future time, at an
agreed price, or probably with agreed cash dividends, which are in a
different currency than the equity.

We will consider three types of equity forwards, namely: The quanto

equity forwards, the domestic-valued foreign equity forwards (DVFEF) and
the basket version of foreign equity forwards.

Quanto
In quanto equity forward, the exchange rate is fixed.

XML Representation
<equity>
<instrumentId>Basket</instrumentId>
<forwardType>Quanto</forwardType>
<basket>
<basketAsset>
<id>ANZ</id>
<weightingFactor>0.3</weightingFactor>
<guaranteedFXRate>1.2</guaranteedFXRate>
</basketAsset>

©2009 Razor Risk Technologies Page 257 of 373

Razor Financial Principals

<basketAsset>
<id>TEL</id>
<weightingFactor>0.2</weightingFactor>
<guaranteedFXRate>1.4</guaranteedFXRate>
</basketAsset>
<basketAsset>
<id>BHP</id>
<weightingFactor>0.5</weightingFactor>
<guaranteedFXRate>1</guaranteedFXRate>
</basketAsset>
</basket>
<numberShares>100000</numberShares>
<exchangeCode>ASX</exchangeCode>
<currency/>
<paymentAmount>
<paymentType>SETTLEMENT</paymentType>
<paymentAmount>
<currency>USD</currency>
<amount>1666000</amount>
</paymentAmount>
<adjustedPaymentDate>2005-12-09</adjustedPaymentDate>
</paymentAmount>
<settlementDays>4</settlementDays>
</equity>

Continuous Dividend Yield

Define
t= maturity date of the contract.
ρ = the correlation between the underlying stock and FX rate.
σS = the equity price volatility coming from the volatility matrix
and is a
deal specified quantity.
σX = the volatility of FX rate.
X GER = the guaranteed exchange rate.
Q (t ) = the continuous dividend yield of foreign equity.
zcdf f ( 0, t )
= discount factor on time period ( 0, t ) from the foreign zero curve.
zcdf d ( 0, t )
= discount factor on time period ( 0, t ) from the domestic zero curve
(USD).
MSD = the market settlement date.
zcdf f ( t , t + MSD )

©2009 Razor Risk Technologies Page 258 of 373

Razor Financial Principals

= forward discount factor on time period ( t , t + MSD ) from the foreign zero
curve which is defined as:
zcdf f ( 0, t + MSD )
zcdf f ( t , t + MSD ) = .
zcdf f ( 0, t )
S = spot price denominated in specified foreign currency.

The forward price denominated in USD is:

− Q t + ρσ σ ⋅t  zcdf f ( t , t + MSD ) 
FGER ( t ) = X GER Se ( ( ) S X )  .
 zcdf f ( 0, t + MSD ) 
 

Discrete Dividends

Define
di = represent decalred and projected discrete dividends going ex-
dividend
on or before t .
B (t ) = the basis function to time t .
ρ = the correlation between the foreign equity and the exchange
rate
(domestic/foreign).
σX = the volatility of the exchange rate (domestic/foreign).
FX S / D = the spot FX rate used to convert discounted dividends from
their
currency to that of the underlier.
q ( tN , t )
= forward dividend yield in time period ( t N , t ) which is defined as:

q ( ti −1 , ti ) =
(Q (t ) ⋅ t ) − (Q (t ) ⋅ t ) .
i i i −1 i −1

ti − ti −1
zcdf D ( 0, Di )
= the discount factors on time period ( 0, Ti ) from the zero curve of
the
currency of each discrete cash dividend.
N = index reference for the last discrete dividend which will go
ex-dividend
on or before time t .

The forward price denominated in USD is:

FGER( t )
= X GER ⋅
  e ( ) S X ( N ) ( N ) 
B t ⋅t − ρσ σ ⋅t − q t ,t ⋅ t − t
   N

 S ⋅ zcdf f ( t , t + MSD )  S /D ∑ D (
− FX zcdf 0, PDi) ⋅ d i  ⋅ .
   i =1   zcdf f ( 0, t + MSD ) 
©2009 Razor Risk Technologies Page 259 of 373
Razor Financial Principals

Domestic – Valued Foreign Equity Forward (DVFEF)

In DVFEF, the exchange rate is not fixed.

XML Representation
<equity>
<instrumentId>Basket</instrumentId>
<forwardType>DVFEF</forwardType>
<basket>
<basketAsset>
<id>ANZ</id>
<weightingFactor>0.3</weightingFactor>
</basketAsset>
<basketAsset>
<id>TEL</id>
<weightingFactor>0.2</weightingFactor>
</basketAsset>
<basketAsset>
<id>BHP</id>
<weightingFactor>0.5</weightingFactor>
</basketAsset>
</basket>
<numberShares>100000</numberShares>
<exchangeCode>ASX</exchangeCode>
<currency/>
<paymentAmount>
<paymentType>SETTLEMENT</paymentType>
<paymentAmount>
<currency>USD</currency>
<amount>1666000</amount>
</paymentAmount>
<adjustedPaymentDate>2005-12-09</adjustedPaymentDate>
</paymentAmount>
<settlementDays>4</settlementDays>
</equity>

Define
X = the spot exchange rate (domestic/foreign).
f = foreign currency.
d = domestic currency.
D = dividends.

The DVFEF price is:

 zcdf d ( 0, t + MSD )   1 
FDV ( t ) = XSe − Q( t )⋅t   / zcdf d ( 0, t + MSD ) = XSe − Q( t )⋅t 
 zcdf d ( 0, t )   zcdf ( 0, t ) 
   d 
.

©2009 Razor Risk Technologies Page 260 of 373

Razor Financial Principals

If there is a basis function, then it should be applied as in the quanto

case. Also, if there are discrete dividend payments, then the DVFEF price
is:
FDV ( t )
=X⋅
  zcdf d ( 0, t + MSD )   N
 e ( ) ( N )( N)
B t ⋅t − q t ,t ⋅ t −t

 S ⋅   −  FX S / D ⋅ ∑ zcdf D ( 0, PDi ) ⋅ di   ⋅ .

  zcdf d ( 0, t )  
 i =1   zcdf d ( 0, t + MSD )


Basket Equity Forward

Define
n = number of forward contracts in the basket.

The basket price is simply a linear combination of the individual forward

contract in the basket:
n
FB = ∑ wi Fi .
i =1

15.3 Equity Options

15.3.1 Description of Instrument
Razor supports European and American call options. American call options on
company stock give the holder of the option the right but not the obligation to
purchase the stock at the strike price at any time up until the expiry date of
the option.
European call options on company stock give the holder of the option the right
but not the obligation to purchase (Call) the stock or to sell the stock (Put) at
the predetermined strike price at the date of option maturity.
American call options on company stock give the holder of the option the right
but not the obligation to purchase or sell the stock at the strike price at any
time up until the expiry date of the option.

15.3.2 XML Representation

fpmlEquityOption Schema
Name: Type Occurs Size Description
productType
0..1 Indicates the type of product
fpmlProductType
optionType
1..1 Indicates the type of option
fpmlOptionType
underlying
1..1 The underlying equity
fpmlInstrumentRef
Strike 1..1 The option strike price

©2009 Razor Risk Technologies Page 261 of 373

Razor Financial Principals

fpmlEquityOption Schema
Name: Type Occurs Size Description
fpmlStrike
numberOfOptions
1..1 The number of options traded
Int
optionEntitlement
1..1 The entitlement per option
Double
equityExercise
1..1 The option details
fpmlEquityExercise
equityOptionFeatures
0..1 The index option features
fpmlEquityOptionFeatures
equityPremium
1..1 The option premium
fpmlPremium

15.3.3 Pricing
European Style
Vanilla Call and Put options are priced using the standard Black-Scholes
formula

Vanilla Call price is:

C = Se − qT N (d 1 ) − Xe − rT N ( d 2 ) ,

Vanilla Put price is:

P = Xe − rT N ( −d 2 ) − Se − qT N ( −d1 )

with d1 and d 2 defined as:

S σ2 S σ2
ln( ) + (r − q + )T ln( ) + (r − q − )T
d1 = X 2 d2 = X 2 = d1 − σ T
σ T and σ T .

and N ( ) - cumulative normal distribution function.

Other variables in formulas represent:

S – stock spot price,
X – option strike price,
r - continuously compounded risk free rate until option maturity
σ - volatility of the relative price change of the underlying stock price,

©2009 Razor Risk Technologies Page 262 of 373

Razor Financial Principals

T – time to expiration in years.

q - a known dividend yield that underlying stock pays during option life.

Please refer to section 17.1.4 - generalised Black-Scholes option pricing

formula for further detail and the Greeks.

If a European option is written on stock that is due to pay dividends during

option life that are presented in discrete form (t i , Di ) then the stock price S in
the above formulas should be adjusted:

S=S−
∑D ei
i
− rt i

where
t i < T - is time in years to i’th dividend payout

Di - the known value of i’th dividend.

American Style
American options on stock can be priced in various ways, using either closed-
form approximations or numerical solutions – binomial tree model or finite
differences method. All of which are implemented in Razor.

Bjerksund and Stensland

The Bjerksund and Stensland approximation is an efficient model for pricing
American stock options in which case an annualised dividend yield is applied as
input to the pricing model.

©2009 Razor Risk Technologies Page 263 of 373

Razor Financial Principals

α = (I − X )I −β
2
1 b   b 1 r
β = − 2 
+  2 −  +2 2
2 σ  σ 2 σ
I 2 ln( I / S )
φ ( S , T , λ , γ , H , I ) = e λ S γ ( N (d ) − ( ) k N ( d − ))
S σ T
1
λ = ( −r + γb + γ (γ − 1)σ 2 )T
2
ln(S / H ) + (b + (γ − 12 )σ 2 )T
d =−
σ T
2b
κ = 2 + ( 2γ − 1)
σ
I = B0 + ( B∞ − B0 )(1 − e h (T ) )
 B0 
h(T ) = −(bT + 2σ T ) 
 B∞ − B0 
β
B∞ = X
β −1
r
B0 = max( X , ( )X )
r −b
C = αS β − αφ ( S , T , β , I , I ) + φ ( S , T ,1, I , I ) − φ ( S , T ,1, X , I ) −
Xφ ( S , T ,0, I , I ) + Xφ ( S , T ,0, X , I )
P( S , X , T , r , b,σ ) = C ( X , S , T , r − b,−b,σ )

Black Scholes
The generic Black-Scholes model can be extended to cover options on a cash
SPI. The major difference is that we must now incorporate an asset income in
the form of dividends. We will assume the underlying index is broad-based and
thus the payment of dividends is approximately continuous.

Binomial (Cox-Ross-Rubinstein Binomial Tree)

Razor’s Binomial model is based on the Cox-Ross-Rubinstein Binomial Tree.
This numerical model is most useful for American options where no closed form
solution is possible. The impact of this computational intensive model is a
detrimental effect on performance.

For a background to the binomial model see Options, Futures, and Other
Derivatives (fifth International edition) by John C. Hull pages 392 to 403,
Chapter 12 Paul Wilmott on Quantitative Finance, or alternatively pages 123-
140 of Modelling Derivatives in C++ by Justin London.

©2009 Razor Risk Technologies Page 264 of 373

Razor Financial Principals

Non dividend paying stock:

Let
r = the continuously compounded risk free rate
X= the strike of the option
T = time to maturity
δt = time interval length
n = number of tree steps
R = the tree’s growth factor
u = the tree’s up step size
d = the tree’s down step size
pUp = probability of an up movement
i = the i’th time step
j = the j’th node
S ij = the price at the i’th step and the j’th node

Vij = the price of the option at the i’th step and the j’th node

Then the parameters to construct the tree are:

T
δt =
n
u = eσ δt

d = 1/ u
R = e rδt
R−d
pUp =
u−d
S ij = S 00 u j d i − j

For a call at the terminal nodes i = n :

Vnj = max(0, S 00 u j d i − j − X )

For a call at each node where 0 ≤ i ≤ (n − 1)and 0 ≤ j ≤ i :

Vij = max(S 00 u j d i − j − X , ( pUp * Vi +1 j +1 + (1 − pUp) *Vi +1 j ) / R)

For a put at the terminal nodes i = n :

©2009 Razor Risk Technologies Page 265 of 373

Razor Financial Principals

Vnj = max(0, X − S 00 u j d i − j )

For a put at each node where 0 ≤ i ≤ (n − 1)and 0 ≤ j ≤ i :

Vij = max( X − S 00 u j d i − j , ( pUp * Vi +1 j +1 + (1 − pUp) *Vi +1 j ) / R)

Dividend yield paying stock:

Let
r = the continuously compounded risk free rate
X= the strike of the option
T = time to maturity
δt = time interval length
n = number of tree steps
R = the tree’s growth factor
u = the tree’s up step size
d = the tree’s down step size
pUp = probability of an up movement
i = the i’th time step
j = the j’th node
S ij = the price at the i’th step and the j’th node

Vij = the price of the option at the i’th step and the j’th node
q = the dividend yield

Then the parameters to construct the tree are:

T
δt =
n
σ δt
u = e
d = 1/ u
R = e rδt
R
q δt
−d
pUp = e
u−d
S ij = S 00 u j d i − j

©2009 Razor Risk Technologies Page 266 of 373

Razor Financial Principals

At the terminal nodes i = n :

For a call: Vnj = max(0, S 00 u j d i − j − X )

For a put: Vnj = max(0, X − S 00 u j d i − j )

At intermitted nodes 0 ≤ i ≤ (n − 1)and 0 ≤ j ≤ i :

For a call: Vij = max(S 00 u j d i − j − X , ( pUp * Vi +1 j +1 + (1 − pUp) * Vi +1 j ) / R )

For a put:

Vij = max( X − S 00 u j d i − j , ( pUp * Vi +1 j +1 + (1 − pUp) * Vi +1 j ) / R)

Discrete Dividend paying stock:

Let
r = the continuously compounded risk free free rate
X= the strike of the option
T = time to maturity
δt = time interval length
n = number of tree steps
R = the tree’s growth factor
u = the tree’s up step size
d = the tree’s down step size
pUp = probability of an up movement
i = the i’th time step
j = the j’th node
S ij = the price at the i’th step and the j’th node

Vij = the price of the option at the i’th step and the j’th node

k = the k’th dividend

Dk = the dividend amount at time k

t k = the time to the k’th ex-dividend date

Then the parameters to construct the tree are:

T
δt =
n
σ δt
u = e

©2009 Razor Risk Technologies Page 267 of 373

Razor Financial Principals

d = 1/ u
R = e rδt
R−d
pUp =
u−d

At the terminal nodes i = n :

S nj = S 00 u j d n − j

For a call: Vnj = max(0, S 00 u j d n − j − X )

For a put: Vnj = max(0, X − S 00 u j d n − j )

At other intermitted nodes 0 ≤ i ≤ (n − 1)and 0 ≤ j ≤ i but not where

δt * i ≤ t k ≤ δt * (i + 1)and 0 ≤ j ≤ i :
S ij = S 00 u j d i − j

For a call: Vij = max(S 00 u j d i − j − X , ( pUp * Vi +1 j +1 + (1 − pUp) * Vi +1 j ) / R )

For a put:

Vij = max( X − S 00 u j d i − j , ( pUp * Vi +1 j +1 + (1 − pUp) * Vi +1 j ) / R)

With the discrete dividend model, the binomial tree does not linkup and jumps
are required from the last node prior to an ex-dividend date to the ex-dividend
date. From the ex-dividend date a new tree is created with the following
parameters as before with the exception that now:
T − tk
δt =
n−i
The new tree then continues as per normal until the next ex-dividend date
where a new jump and tree needs to be created.

At a jump point however, where δt * i ≤ t k ≤ δt * (i + 1) and 0 ≤ j ≤ i :

S t j = S ij − Dk
k

Imposing the no-arbitrage constraint that the option value is either the option
value with the dividend, or the payoff:
For a call:

V = max(S 00 u j d i − j − X , ( pUp *Vi +1 j +1 + (1 − pUp) *Vi +1 j ) / R)

t j
k

For a put:

V = max( X − S 00 u j d i − j , ( pUp *Vi +1 j +1 + (1 − pUp) *Vi +1 j ) / R)

t j
k

©2009 Razor Risk Technologies Page 268 of 373

Razor Financial Principals

And just prior to the jump:

For a call:

Vij = max(S 00 u j d i − j − X ,Vtk j − Dk )

For a put:

Vij = max( X − S 00 u j d i − j , Vtk j − Dk )

Since the binomial tree does not recombine, and a new tree has to be started
for each ex-dividend date, recursion is used to simplify the implementation.
Considering the vast number of extra nodes that this non recombining model
generates a significant impact on performance will occur, particularly for
frequent dividends on a long dated expiry option.

For further reading on this discrete dividend payment model see pages 129-131
and 146 to 147 of Paul Wilmott on Quantitative Finance. For an
implementation visit Financial Principle’s Discrete Dividend model by Bernt
Arne Odegaard.

15.3.4 Equity Option Greeks

For American equity options, greek sensitivities are calculated
numerically using a finite difference scheme. European option
sensitivities are provided as analytical solutions to the Black-Scholes
model (refer to section 17.2.4 for details).

Delta
Delta is defined as the difference in value of the option by perturbing the
underlying equity price S by ±1bp where all other parameters remain
constant.

Let
f = a function defined as the price of the option
S = Underlying equity spot price
1bp = 1 basis point or 0.0001

∂P f ( S + 1bp) − f ( S − 1bp )
∆= ≈
∂S 2bp

Gamma
Gamma is defined as the difference in delta by perturbing the underlying
equity spot price S by ±1bp where all other parameters remain constant.
In Razor we approximate gamma by using the second order finite central
difference.

Let

©2009 Razor Risk Technologies Page 269 of 373

Razor Financial Principals

f = a function defined as the price of the option

S = Underlying equity spot price
1bp = 1 basis point or 0.0001
∂∆ ∂ 2 P f ( S + 1bp) + f ( S − 1bp) − 2 f ( S )
Γ= = ≈
∂S ∂S 2 1bp 2

Vega
Vega is defined as the difference in value of the option by perturbing the
equity volatility σ by ±1% where all other parameters remain constant.

Let
f = a function defined as the price of the option
σ = Implied volatility of the underlying equity
1% = 1 percent as a decimal or 0.01

∂P f (σ + 1%) − f (σ − 1%)
Λ= ≈
∂σ 2%

Theta
Theta is defined as the difference in value by shifting the value date
forward by one day where all other parameters remain constant. Since
time moves in a forward direction, for theta we use the backward finite
difference which is defined as:

∂ f ( x ) − f ( x − h) 
≈ h > 0
∂x h 

In Razor, pricing formulas accept the risk free rate in terms of a

continuously compounded discount factor. Therefore, an adjustment to
the discount factor is made to account for the decay of 1 day.

Let
f = a function defined as the price of the option
T = Time to expiration in years.
T−1 = Time to expiration minus 1 day in years.

df = a continuously compounded discount factor of r in terms of T

dfT−1 = a continuously compounded discount factor of r in terms of T−1

1dy = 1 day in years or 1/365

Where
T−1
dfT−1 = df T

©2009 Razor Risk Technologies Page 270 of 373

Razor Financial Principals

∂P f (T−1 , dfT−1 ) − f (T )
Θ=− ≈
∂T 1dy

15.4 Equity Index Options

15.4.1 Description of Instrument
The Equity Index Option is a variant of the Equity Option. The differences
between equity and index options occur primarily in the underlying instrument
and the method of settlement. Equity Index options are predominately traded
via exchanges hence the other common term used are equity ETO’s, Exchange
Traded Options.

15.4.2 XML Representation

The index option is represented similarly to the equity option. It uses the
fpmlEquityOption schema with the optional equityOptionFeatures element
made mandatory as below:
<equityOptionsFeatures>
<etoFeatures>
<indexOptionTickUnit>0.5</indexOptionTickUnit>
</etoFeatures>
</equityOptionsFeatures>

The other slight differences occur in the numberOfOptions and the

optionEntitlement elements. In an index option sense the numberOfOptions
implies the number of contracts, while the optionEntitlement element implies
the Tick Value.

15.4.3 Pricing
The index Option is priced exactly the same as the equity option above. The
only factor that needs to considered is in the amount or position of the trade
where the Tick Value needs to be factored in.
EquityIndexOptionPosition = EquityOption Pr ice * numberOfOptions * optionEntitlement * TickUnit

15.5 Equity Basket Option

15.5.1 Contract Definition

An equity basket option is an option contract where the underlying asset
is a basket of individual equities. There are two valuation methods for
equity basket options.

15.5.2 Valuation 1
The basket option price in Razor is approximated using single-equity
Black-Scholes formula with adjustments to the input variables. The spot
rate, cost of carry and volatility in the single-equity Black-Scholes
©2009 Razor Risk Technologies Page 271 of 373
Razor Financial Principals

formula are replaced by basket spot rate, basket cost of carry and basket
volatility.

Define:
t = the valuation date.
T = the option expiration date.
i
St = the price of asset i in the basket at time t .
bti,T = the cost of carry of asset i in the basket for time T implied
at time t .
σ ti,T = volatility of asset i in the basket for time T implied at time
t.
K = the strike price.
Stb = the basket price at time t .
btb,T = the basket cost of carry for time T implied at time t .
σ b
t ,T = the basket volatility for time T implied at time t .
rt ,T = the risk-free rate for time T implied at time t .
wi = weighting of asset i in the basket.
ρij = correlation coefficient between asset i and asset j .
n = number of assets in the basket.

The basket variables are approximated as:

n
Stb = ∑ wi Sti .
i =1
n
b b
t ,T = ∑ wi bti,T .
i =1
2 i −1

∑ w (σ )
n n
σ tb,T = i
2 i
t ,T + 2∑∑ wi w j ρijσ ti,T σ t j,T .
i =1 i =1 j =1

Note that bti,T , rt ,T and σ ti,T can be obtained from the term-structure curve
for each individual asset i .

Pricing Formulas:

Call option price is given as:

(b ) N d − Ke − r T N d .
( 1) ( 2)
b
− rt ,T T
c = Stb e t ,T t ,T

Put option price is given as:

( bb − r )T
p = Ke t ,T N ( − d 2 ) − Stb e t ,T t ,T N ( − d1 ) .
−r T

©2009 Razor Risk Technologies Page 272 of 373

Razor Financial Principals

 ( ) 
2
 Sb
  b σ tb,T
+
ln  t + T
 b
  
t ,T
K 2
d1 =   .
σ T
d 2 = d1 − σ t ,T T .
b

This valuation method only supports single currency. All assets must be
denominated in the same currency as the strike price. It also supports
continuous dividend yield but not discrete dividend payout.

The Case of Single Asset

If we only have one asset in the basket, Razor calculates the price using
the single asset option pricing formula instead of the basket formula.

15.5.3 Valuation 2
Razor also supports another method of calculating equity basket option.
We can use the Asian quanto basket formula to calculate the equity
basket option price. This valuation method has the advantage that it
supports cross currencies for the asset basket. We first recall the formula
for Asian quanto basket.

Define:
t = the valuation date.
T = the option expiration date.
na = number of underlying assets.
nf = number of fixings.
m = number of observed fixings.
C = payoff currency.
Si ( t ) = i by 1 array of asset prices i at time t .
Ci = currency i for asset i .
σ S (t j )
i
= i by j matrix of volatilities of asset i at time t j .
Xi
= i by 1 array of exchange rate quoted as payoff currency per
currency i ,i.e. C / Ci .
σ X (t j ) i
= i by j matrix of volatilities X i at time t j .

ρ ij = i by i matrix of instantaneous correlations of ln S i and

ln S j .
ρi = i by 1 array of instantaneous correlations of S i and X i .
rCi ( t j ) =
i by j matrix of instantaneous risk-free interest rates in C i market.
bi ( t j ) = i by j matrix of cost of carries for asset i at time t j .
©2009 Razor Risk Technologies Page 273 of 373
Razor Financial Principals

K = strike price is payoff currency.

W = Wiener process.

We further define:

S (t n ) = ∑∑ wij S i (t j ) .
na nf

i =1 j =1

Option Payoff:

Call: (( ) )
max S t n f − K ,0
Put: max(K − S (t ),0) . nf

Pricing Formulas:

We have assets denominated in different currencies. We need to find a

measure say C , such that all assets under this measure are risk-neutral.
It can be shown (Datey, Gauthier and Simonato) that in such risk-neutral
measure, the asset prices are distributed as:

( )
dS i (t ) = bi (t ) − ρ i ⋅ σ Si (t ) ⋅ σ X i (t ) ⋅ S i (t ) ⋅ dt + σ Si (t ) ⋅ S i (t ) ⋅ dWi C (t ) , ∀i = 1,2,..., n a
.

Regardless all the technical terms, the important thing is that the
underlying asset price still has the form of geometric Brownian motion
with different drift and diffusion term. It means that we can use similar
approaches in Asian non-quanto options for Asian quanto options.

S (t n ) = ∑∑ wij S i (t j ) = S (t m ) + ∑ ∑ w S (t ) = S (t ) + M
na n na nf

ij i j m t
i =1 j =1 i =1 j = m +1

where

∑ w S (t ) .
na nf
Mt = ∑ ij i j
i =1 j = m +1

We denote bi∗ (t j ) = bi (t j ) − ρ i ⋅ σ Si (t j ) ⋅ σ X i (t j )
and define:
Fi (t j ) = S i (t )e i j j
b ∗ (t )(t −t )

∑ w F (t ) .
na nf
E ∗ (M t ) = ∑ ij i j
i =1 j = m +1

( ) ∑ ∑ ∑ w w F (t )F (t ) ×
na nf na nf
E ∗ M t2 = ∑ ij kl i j k l
i =1 j = m +1 k =1 l = m +1

©2009 Razor Risk Technologies Page 274 of 373

Razor Financial Principals

( )
exp ρ ik σ Si (min (t j , t l ))σ Sk (min(t j , t l ))min (t j − t , t l − t ) .
m
We can then use the fixed strike formulas with all K − ∑ α i S ti replaced
i =1

by K − S (t m ) . Exercise for certain and out-of-money for certain cases are

m
the same by replacing K − ∑ α i S ti with K − S (t m ) .
i =1

To find the price for equity basket option, we simply assume there is only
one fixing date, i.e. n f = 1 , on the option expiration date and the
correlation between the underlying asset and its denominated currency to
be zero, i.e. ρi = 0 .

15.5.4 Equity Basket Option with Quanto Feature

If one or more assets in the basket are denominated in any currency
different from the paying currency, then we cannot use the method
proposed in Valuation 1. We need to use method proposed in Valuation 2
by taking the correlation between the underlying asset and its
denominated currency into account. It can be achieved using the above
Asian quanto basket formula by assuming there is only one fixing date,
i.e. n f = 1 , on the option expiration date and input the required
correlation coefficients.

This valuation method supports continuous dividend yield but not discrete
dividend payout.

15.5.5 Summary
In summary, we have two different valuation methods to calculate the
equity basket price for non-quanto case and one method (the Asian
quanto basket option formula) for the quanto case. Currently, both
valuation methods support for continuous dividend yield but not discrete
dividend payout.

15.6 Equity Futures

15.6.1 Description of Instrument
A Futures are legally binding contracts to buy or sell a particular asset (or cash
equivalent) on a specified future date. Equity futures are futures traded in the
equity market on organized exchanges.
Equity Futures can consist of contracts of an individual equity, a basket of
equities or an equity index.

15.6.2 XML Representation

An example of the XML product section of an equity index future is as follows:
©2009 Razor Risk Technologies Page 275 of 373
Razor Financial Principals

<equityFuture>
<instrumentId>XJO7U</instrumentId>
<currency>AUD</currency>
<numberShares>1</numberShares>
<exchangeCode>ASX</exchangeCode>
<numberContracts>2</numberContracts>
<futurePrice>58050.00</futurePrice>
<indexTickUnit>10</indexTickUnit>
<settlementDate id="">
<unadjustedDate>2007-09-20</unadjustedDate>
</settlementDate>
</equityFuture>

15.6.3 Pricing
The individual equity, basket, or equity index are all similarly priced. It’s in
the input values in the 3 above fields, numberShares, numberContracts and
indexTickUnit where their values vary from the default of 1.0.

Equity Future:
Let
Vccy = value of equity future in currency ccy

Fccy = the projected forward price derived from the underlying stock/index
forward curve in currency ccy.
X = equity future contract price.
df = discount factor to the payment date of contract
N = number of shares

Then:
Vccy = df * N * ( Fccy − X )

Equity Index Future:

Let
Vccy = value of equity future in currency ccy

Fccy = the projected forward price derived from the underlying stock/index
forward curve in currency ccy.
X = equity future contract price.
df = discount factor to the payment date of contract
C = number of contracts
k = index tick unit

©2009 Razor Risk Technologies Page 276 of 373

Razor Financial Principals

Then:
Vccy = df * C * ( Fccy − X ) / k

15.7 Asian Options: Introduction

Asian Options encompass various types of average options including
Average Rate (Fixed Strike) and Average Strike (Floating Strike) Options.
Asian options are options where the payoff depends on the average price
of the underlying asset during at least some part of the life of the option.

This document specifies the various pricing models used for valuation of a
subset of Asian options. The first section will focus on Asian options with
geometric average features. The second section will concentrate on Asian
options with continuous arithmetic average features. The third section
will also be on Asian options with discrete arithmetic average features
but with equal fixing length and weights and the final section will be on
Asian options with discrete arithmetic average features with unequal
fixing length, weights and time-dependent underlying parameters.

Define:
S t = the current underlying asset price at time t .
S = the average price of the underlying asset during some part of the
life of the
option.
K = the strike price of the option.
T = the original maturity time of the option.
τ= T − t is the remaining time to maturity.
r = the risk-free rate.
σ = the volatility of the underlying asset.
b = the cost of carry
N (⋅) = the cumulative normal distribution function.

Unless otherwise stated, the above definitions will be assumed for all the
formulae.

Example: The four fundamental types of payoff for Asian Options:

Type Payoff Name

(1) max(S − K ,0) Fixed Strike Call
(2) max(K − S ,0 ) Fixed Strike Put
(3) max(S T − S ,0) Floating Strike Call
(4) max(S − S T ,0 ) Floating Strike Put

©2009 Razor Risk Technologies Page 277 of 373

Razor Financial Principals

15.7.1 Types of Averaging and Fixing

We can see from the payoff table that the payoff of Asian options
depends on S , the average price of the underlying asset S . Two
approaches to taking the average of the asset prices are the geometric
average and the arithmetic average.

Geometric Average

1. Assume we have n asset prices S1 , S 2 ,..., S n .

The discrete geometric average of these n asset prices is defined as:
S = (S1 S 2 ...S n ) n .
1

2. The continuous geometric average over the time period, say T , is

defined as:
1 T 
S = exp ∫ log S u du 
T 0
.

Arithmetic Average

1. Again, assume we have n asset prices S1 , S 2 ,..., S n .

The discrete arithmetic average of these n asset prices is defined as:
S + S 2 + ... + S n
S = 1
n .
2. The continuous arithmetic average over the time period, say T , is
defined as:
1 T
S = ∫ S u du
T 0 .

Types of Fixing
Fixings are stock prices used in averaging. In discrete arithmetic average,
S + S 2 + ... + S n
S = 1
fixings can be assumed to have equal weights, i.e. n or
n
S = ∑ wi S i
unequal weight, i.e. i =1 where wi is the normalised weight for
th
the i fixing. Fixings can also be equally spaced (e.g monthly) or they
can be unequally spaced in an average period.

©2009 Razor Risk Technologies Page 278 of 373

Razor Financial Principals

15.7.2 Geometric Average Asian Option Pricing Models

Fixed Strike Continuous Geometric Asian Option

Define:
S t = the current underlying asset price at time t .
S = the continuous geometric average price of the underlying asset
during some part of the life of the option.
K = the strike price of the option.
T = the original maturity time of the option.
τ = T − t is the remaining time to maturity.
r = the risk-free rate.
σ = the volatility of the underlying asset.
b = the cost of carry.
N (⋅) = the cumulative normal distribution function.

Option Payoff:
Call: max(S − K ,0)
Put: max(K − S ,0)

Pricing Model:

Case1: Valuation date on or before average start date

Kemma Vorst (1990) derived the following formulae for pricing fixed
strike Asian options:
c fix (t ) = S 0 e (bA −r )τ N (d1 ) − Ke − rτ N (d 2 )
p fix (t ) = Ke − rτ N (− d 2 ) − S 0 e (bA −r )τ N (− d1 )
where
S   1 
ln 0  +  b A + σ A2 τ
 
d1 =  
K 2
σA τ
d 2 = d1 − σ A τ
σ
σA =
3
1 σ2 
b A =  b − 
2 6 .

Case 2: Valuation date in the averaging

We further define:

©2009 Razor Risk Technologies Page 279 of 373

Razor Financial Principals

Gt = the time t observed geometric mean of the underlying asset.

Pricing Model:
c fix (t ) = S q N (d1 ) − Ke − rτ N (d 2 )
p fix (t ) = Ke − rτ N (− d 2 ) − S q N (− d1 )
where
 Sq   1 
ln  +  r + σ A2 τ
K  2 
d1 =  
σA τ
d 2 = d1 − σ A τ
t
G  T − q (t ,T )τ
S q = S t  t  e
 St 
σ τ
σA = ⋅
3 T
1 1  τ 1 τ2
q (t , T ) = r −  b − σ 2  ⋅ − σ 2 2
2 2  T 6 T .

Reference: Generalisation of Peter Buchen’s method.

Fixed Strike Discrete Geometric Asian Option

Define:
S t = the current underlying asset price at time t .
S = the discrete geometric average price of the underlying asset during
some part of the life of the option.
K = the strike price of the option.
T = the original maturity time of the option.
τ = T − t is the remaining time to maturity.
r = the risk-free rate.
σ = the volatility of the underlying asset.
b = the cost of carry.
t i = time to the i th fixing.
σ i = the implied global volatility for an option expiring at time t i .
vi = the local volatility between each time fixing.
n= total number of fixings.
N (⋅) = the cumulative normal distribution function.

Option Payoff:
Call: max(S − K ,0)

©2009 Razor Risk Technologies Page 280 of 373

Razor Financial Principals

Put: max(K − S ,0)

Pricing Model:

c fix (t ) = S 0 e (bG −r )τ N (d1 ) − Ke − rτ N (d 2 )

p fix (0) = Ke − rτ N (− d 2 ) − S 0 e (bG − r )τ N (− d1 )

where
S   σ2 
ln 0  +  bG + G T
K  2 
d1 =
σG T
d 2 = d1 − σ G T
σ G2 1 n
 σ i2 
bG =
2
+
nT
∑ 

i =1 
b −
2
t i



From Levy (1997), the σ G is calculated using the Levy Approach which
uses global volatility as an input.
n −1
1  n 
σ G2 = 2 ∑ σ i2 t i + 2∑ (n − i )σ i2 t i 
n T  i =1 i =1 

15.7.3 Arithmetic Average Asian Option Pricing Models (Continuous

Averaging)

Fixed Strike Continuous Arithmetic Asian Option

Define:
S t = the current underlying asset price at time t .
S = the continuous arithmetic average price of the underlying asset
during some part of the life of the option.
K = the strike price of the option.
T = the original maturity time of the option.
τ = T − t is the remaining time to maturity.
r = the risk-free rate.
σ = the volatility of the underlying asset.
b = the cost of carry.
N (⋅) = the cumulative normal distribution function.
©2009 Razor Risk Technologies Page 281 of 373
Razor Financial Principals

Option Payoff:
Call: max(S − K ,0)
Put: max(K − S ,0)

Pricing Models:

Turnbull and Wakeman Approximation

Define:
TA = the length of the total average period.
t1 = time to the beginning of the average period from current time t .
τ A = TA − τ .

Pricing Formulas:

Case1 : Valuation date on or before average start date

c fix (t ) ≈ S t e (bA − r )τ N (d1 ) − Ke − rτ N (d 2 )

p fix (t ) ≈ Ke − rτ N (d 2 ) − S t e (bA −r )τ N (d1 )

σ A2
ln(S t / K ) + (b A + )τ
d1 = 2
σA τ
d 2 = d1 − σ A τ

ln(M 2 )
σA = − 2b A
τ
ln(M 1 )
bA =
τ
The exact first and second moments of the arithmetic average are:
e bτ − e bt1
M1 =
b(τ − t1 )
2e ( 2b +σ )τ 2e ( 2b +σ )t1 e b (τ −t1 )
2 2
 1 
M2 = +  − 
(b + σ )(2b + σ )(τ − t1 )
2 2 2
b(τ − t1 ) 2  2b + σ
2
b +σ 2 

When we have zero cost of carry (i.e. b = 0 ):

M1 = 1

M2 =
2 2
[
2eσ τ − 2eσ t1 1 + σ 2 (τ − t1 ) ]
σ (τ − t1 )
4 2
.

Case 2: Valuation date in averaging

©2009 Razor Risk Technologies Page 282 of 373
Razor Financial Principals

It is clear from the above formula that it does not handle the situation
when we need to calculate option price at times when the average period
already began. However, this problem can be solved by a slight
modification of the above formula.
We denote S A to be the average asset price realised or observed. To
adjust the formula for the in progress case (i.e. τ A > 0 ), we simply
replace the strike price K by K̂ and the option value must be multiplied
TA
by τ , where
T τ
Kˆ = A K − A S A
τ τ .
Also we set 1 to be zero for all cases when τ A > 0 .
t

Case 3: Exercise for certain

TA τA
K− S <0
If we are inside the average period (i.e. τ A > 0 ), and τ τ A ,
then a call option is for certain to be in-the-money and the put option is
for certain to be out-of-the-money at the maturity. This leads to the
following results:
([ ]
c fix (t ) = e − rτ E S A∗ − K )
p fix (t ) = 0
where
TA − τ τ
[ ]
E S A∗ = S A
TA
+ SM 1
TA
t1 is zero for this case because τ A > 0 .

Levy Approximation

Define:
S A = the arithmetic average of the known asset price fixings.

Pricing Formulas:

c fix (t ) ≈ S E N (d1 ) − K ∗ e − rτ N (d 2 )
p fix (t ) ≈ c fix − S E + K ∗ e − rτ

1  ln( D ) 
d1 =  − ln( K ∗ ) 
V  2 
d 2 = d1 − V

©2009 Razor Risk Technologies Page 283 of 373

Razor Financial Principals

SE =
St
T *b
(
e (b − r )τ − e − rτ )
T −τ
K* = K − SA
T
V = ln( D) − 2[rτ + ln(S E )]
M
D= 2
T

 e ( 2b +σ )τ − 1 e bτ − 1
2
2 S t2
M =  − 
b +σ 2  2b + σ
2
b 

Floating Strike Continuous Arithmetic Asian Option

Define:
S t = the current underlying asset price at time t .
S = the continuous arithmetic average price of the underlying asset
during some part of the life of the option.
K = the strike price of the option.
T = the original maturity time of the option.
τ= T − t is the remaining time to maturity.
r = the risk-free rate.
σ = the volatility of the underlying asset.
b = the cost of carry.
N (⋅) = the cumulative normal distribution function.

Option Payoff:

Call: max(S − S ,0)

Put: max(S − S ,0)

Pricing Models:

Henderson-Wojakowski Identity

Define:
λ = some constants ∈ ℜ .

Different Asian option prices can be computed as:

c float (S t , λ , r , b, σ , t , T ) = c float = e − r (T −t ) E ∗ (λS T − S )
+

p float (S t , λ , r , b, σ , t , T ) = p float = e − r (T −t ) E ∗ (S − λS T )
+

©2009 Razor Risk Technologies Page 284 of 373

Razor Financial Principals

c fix (S t , K , r , b, σ , t , T ) = c fix = e − r (T −t ) E ∗ (S − K )
+

p fix (K , S t , r , b, σ , t , T ) = p fix = e − r (T −t ) E ∗ (K − S )
+

∗
Note that E is the expectation taken under risk-neutral measure.
c fix p
and fix are the prices of fixed strike Asian call and put options
respectively. When λ = 1 , float and float are the prices of floating strike
c p
Asian call and put options respectively.

Pricing Formula:

Henderson and Wojakowski proved the following identity:

c float (S t , λ , r , b, σ , t , T ) = p fix (λS t , S t , r − b,−b, σ , t , T )
.
 
p float  S t , , r , b, σ , t , T  = c fix (K , S t , r + b,−b, σ , t , T )
K
 S 
t .
We assume λ = 1 , the above identities can be summarised in words as:
The first identity simply says that the price of a floating strike call option
with current stock price S t at time t , risk-free rate r , cost of carry b ,
volatility σ that matures at time T is equal to the price of an at-the-
money fixed strike put option with the risk free rate replaced by r − b ,
cost of carry replaced by − b and rest of variables holding constant.
The second identity says that an at-the-money floating strike put option
with current stock price S t at time t , risk-free rate r , cost of carry b ,
volatility σ that matures at time T is equal to a fixed strike call option
with the risk-free rate replaced by r + b , cost of carry replaced by − b
and rest of the variables holding constant.

With the above two identities, we can compute the floating strike options
once we know the fixed strike options.

The above identities can be extended to forward-starting options at some

time t > 0 for the option prices at the times before the averaging begins.
However, it does not work for in-progress case. It also only works for
continuous arithmetic averages.

Bouaziz, Briys and Crouhy Model

Define:
TA = the total average time period.
τ A = TA − τ .

Pricing Formulas:

©2009 Razor Risk Technologies Page 285 of 373

Razor Financial Principals

Case1: Valuation date on or before average start date

 T  3rˆ T A  −3 rˆ TA 
2

 + σ T A ⋅ e 8σ 2 
2
c float (t ) ≈ S t exp(− rT A )rˆ ⋅ A ⋅ N 
 2  2σ  6π 
  .

p float ≈ 
(
 1 − e − rTA  )
− 1 S t + c float
 rT A 
where
1
rˆ = r − σ 2
2 .
Set TA = T , then we obtain the case for valuation date on the beginning
of average period.

Case 2: Valuation date in averaging

Define:
t
1
Mt =
TA ∫S u du
T −TA and note that M t is known at time t .

c float ( t )
  m   − m2 
≈ S t exp(− rτ ) mN 
v
 + exp 
  v 2π  2v 
where
τ rˆτ 2
m = 1− + rˆτ − − Mt
TA 2T A .
 τ3 τ2
v = τ + 2 − σ 2
 3T A T A  .

Set T = T A leads to the case when averaging was taken from the initiation
of the option contract.

By put-call parity:
1 − e − rτ 
p float (t ) ≈ M t e − rτ + S t  − 1 + c float
 rT A 

Continuous Reciprocal Asian Option

Define:
S t = the current underlying asset price at time t .

©2009 Razor Risk Technologies Page 286 of 373

Razor Financial Principals

S = the continuous arithmetic average price of the underlying asset

during some part of the life of the option.
K = the strike price of the option.
T = the original maturity time of the option.
τ = T − t is the remaining time to maturity.
r = the risk-free rate.
σ = the volatility of the underlying asset.
b = the cost of carry.
N (⋅) = the cumulative normal distribution function.

Option Payoff:
 
 
 1 1 
max T − ,0
1 ST 
 ∫ u S du 
Call:  T 0 
 
 
1 1
max  − T ,0
 ST 1 


∫
T 0
S u du 

Put:

Pricing Model:

Dufresne (2000) proposed a model to evaluate reciprocal Asian options

with continuous arithmetic average using the Laguerre series. Dufresne
model works for the case when average period and the option maturity
coincides exactly, i.e. the payoff for the call and put options are
respectively:
and
.

The generalised Laguerre polynomials are defined as:

Γ ( n + a + 1) ( − x )
k
x−a x d
( )
n
Lan ( x ) = e x n+a − x
e = ∑
k = 0 Γ ( k + a + 1) k !( n − k ) !
n
n ! dx
.

We further define:
Γ(⋅) to be the gamma function.
1
v =b− σ2
2 .
t
At( µ ) = ∫ e2 µ s + 2Ws ds
0 , t ≥ 0 , µ ∈ℜ .
Dufresne proved the following expectation:

©2009 Razor Risk Technologies Page 287 of 373

Razor Financial Principals

( )
−k
E 2 At( µ ) = ∫ φµ ( k , t , y )ψ µ ( −1, y ) dy
0 , k = 1, 2,...
µ t y
2 2
− −
2 2t
ye
φµ (1, t , y ) =
2π t 3 .
1 ∂
φµ ( k , t , y ) = φµ ( k − 1, t , y ) + ( k − µ − 1) φ ( k − 1, t , y )
2 (1 − k ) ∂t , k = 2, 3,...
cosh[(µ − 1) y ]
ψ µ (− 1, y ) =
sinh ( y ) .

The no-arbitrage price for the fixed strike reciprocal Asian call option is
equal to:
σ 2T R
e− rT C ( µ,T , x)
2S0 .
where
∞
C R ( µ , t , x ) = c a +1 x b e − cx ∑ an ( t ) Lan ( cx )
n=0 , 0< x<∞,
( )
−( a −b + k + 2 )
E 2 At( )
µ
n !( −c )
k
n
an ( t ) = ∑
k = 0 Γ ( k + a + 1) k !( n − k ) ! ( a − b + k + 1)( a − b + k + 2 )
, n = 0,1,...
σ T
2
t=
4 .
2v
µ=
σ2 .
2S 0
x=
σ 2 KT .
a , b and c are constants and we can set to zero for computational
convenience.

( )
−k
E 2 At( µ ) = ∫ φµ ( k , t , y )ψ µ ( −1, y ) dy
0 .
can be evaluated using numerical integration, for example, Gauss-
Legendre Quadrature.

1 ∂
φµ ( k , t , y ) = φµ ( k − 1, t , y ) + ( k − µ − 1) φ ( k − 1, t , y )
2 (1 − k ) ∂t
can be calculated
using symbolic packages such as Mathematica or it can be done manually.
Usually we require only 10 terms by truncating at the point k = 10 . Thus
φ (1, t , y ), φ µ (2, t , y )..., φ µ (10, t , y )
we only need µ .
Γ ( z ) = z!
if z is an integer ≥ 0.
Generally, if z is a complex number with positive real part, then
©2009 Razor Risk Technologies Page 288 of 373
Razor Financial Principals

∞
Γ ( z ) = ∫ t z −1e − z dt
. This covers the non-integer cases of z and this integral
0

can be shown to converge definitely. However, we should choose a and

b in a way such that non-integer value of z is not present for easier
computation.

The no-arbitrage price for the fixed strike reciprocal Asian put option is
equal to:

−1
1 1 T  
− rT vt +σ Wt
e  − E  ∫ S0 e dt  
K  T 0  

where
−1
1 T 
E  ∫ S0evt +σ Wt dt 
T 0 
σ T2
 1 
= E 
2 S 0  2 At( µ ) 
− µ 2t
cosh[(µ − 1) y ]
∞ − y2
σ T e
2 2
=
2πt
2S 0 3 ∫ ye 2t
sinh ( y )
dy
0 .
and again
1
v =b− σ2
2 .
σ 2T
t=
4 .
2v
µ=
σ2 .
2S 0
x=
σ 2 KT .
a , b and c are constants and we can set to zero for computational
convenience.

©2009 Razor Risk Technologies Page 289 of 373

Razor Financial Principals

15.7.4 Arithmetic Average Asian Option Pricing Models (Discrete

Averaging, Equal Fixing Weighting and Constant Underlying Variables)

Fixed Strike Discrete Arithmetic Asian Option

Define:
S t = the current underlying asset price at time t .
S = the discrete arithmetic average price of the underlying asset during
some part of the life of the option.
K = the strike price of the option.
T = the original maturity time of the option.
τ= T − t is the remaining time to maturity.
r = the risk-free rate.
σ = the volatility of the underlying asset.
b = the cost of carry.
N (⋅) = the cumulative normal distribution function.

Option Payoff:
Call: max(S − K ,0)
Put: max(K − S ,0)

Pricing Models:

Turnbull-Wakeman Model

©2009 Razor Risk Technologies Page 290 of 373

Razor Financial Principals

Define:
t1 = time to the beginning of average period from current time t .
n = number of average points.

Pricing Formulas:

Case 1: Valuation date on or before average start date

The valuation formula (Levy 1997 and Haug, Haug and Margrabe 2003) is:
c fix (t ) ≈ e − rτ [FA N (d1 ) − KN (d 2 )]
p fix (t ) ≈ e − rτ [KN (− d 2 ) − FA N (− d1 )]

where
F  σ
2
ln A  + A τ
d1 = 
K  2
σA τ
d 2 = d1 − σ A τ

τ − t1
h=
n −1
S t bt1 1 − e bhn
FA = E [Aτ ] = e
n 1 − e bh

σA =
[ ]
ln (E Aτ2 ) − 2 ln (E [ Aτ ])
τ
[ ]
E Aτ 2

S t2 e (2b +σ )t1 1 − e (2b +σ )hn

 1 − e bhn 1 − e (2b +σ )hn 
2 2 2

 2 
=  (2 b +σ 2 )h
+ −
 1 − e 1 − e (b +σ )h
 1 − e bh 1 − e (2b +σ 2 )h 
2 2
n   .
In the case when cost of carry b = 0 ,
E [Aτ ] = S t
2 t1
S 2 eσ 1 − eσ 2hn  σ hn 
[ ] n − 1− e 2
2
2 
E Aτ = t 2
2
 +
σ 2h
1 − eσ h  1 − eσ h 
 1 − e
2
n  

Case 2: Valuation date in averaging

We now look at the case when we want to find a price at a time inside
the average period.

Define:
m = number of observed average points.
S A = the arithmetic average of the known asset price fixings.

If we are inside the average period, i.e. m > 0 , then the strike price is
replaced by:
©2009 Razor Risk Technologies Page 291 of 373
Razor Financial Principals

nK − mS A m
Kˆ = n−
n−m n−m
Also we set t1 = 0 for all cases m > 0 .

Case 3: Exercise for certain

n
SA > K
If m , then the exercise is certain for a call, and put is certain to
be out-of-the-money. This leads to following results:
( )
c fix (t ) = e − rT Sˆ A − K
p fix (t ) = 0
where
n−m
Sˆ A = S A + E [ A]
m
n n .
Note E[ A] can be calculated as E [Aτ ] with t1 = 0 because m > 0 in this
case.

Case 4: One fixing left:

If there is only one more fixing before maturity, the following formula is
used:
[ ]
c fix (t ) = Se (b− r )τ N (d1 ) − K ∗ e − rτ N (d 2 )
1
n
[ ]
p fix (t ) = K ∗ e − rτ N (− d 2 ) − Se (b − r )τ N (− d 1 )
1
n
where
 S   σ2 
ln ∗  +  b + τ
K   2 
d1 =
σ τ
d 2 = d1 − σ τ
K ∗ = nK − (n − 1)S A .

Curran’s Approximation

Define:
t1 = time to the beginning of average period from current time t .
n = number of averaging points.

Pricing Formulas:

Case1: Valuatin date on or before average start date

c fix (t ) ≈ e  ∑ e
− rτ 1 2
N 
( )
 n µi + σ i  µ − ln Kˆ σ xi 
2

+  − KN 
( )
 µ − ln Kˆ
  σ 
 n i =1  σx σx   x 

where

©2009 Razor Risk Technologies Page 292 of 373

Razor Financial Principals

τ − t1
h=
n −1
 σ2 
µ i = ln(S t ) +  b − t i
 2 
σ i = σ 2 [t1 + (i − 1)h ]
  i (i − 1)  
σ xi = σ 2 t1 + h (i − 1) − 
  2n  

µ = ln(S t ) +  b −
σ 2 
 t1 +
(n − 1)h 
 2   2 

 h(n − 1)(2n − 1) 
σ x = σ 2 t1 + 
 6n 
 σ xi2 
 σ 2
− 
σ [ln(K ) − µ ] σ x2 
i
1 n 
Kˆ = 2 K − ∑ expµ i + xi + 
n i =1  σ x2 2 
 

Case 2: Valuation date in averaging

Define:
m = number of observed average points.
S A = the arithmetic average of the known asset price fixings.

If the average already began at the time when we want to find the option
price (i.e. m > 0 ), then the strike price is replaced by:
nK − mS A m
Kˆ = n−
n−m n−m .
Also we set 1t = 0 for all cases m > 0 .

Case 3: Exercise for certain

n
SA > K
If m , then call is for certain going to be in-the-money and put is
for certain to be out-of-money. This leads to the following result:
(
c fix = e − rτ Sˆ A − K )
p fix = 0
where
n−m
Sˆ A = S A + E [A]
m
n n
1− e bhn
E[ Aτ ] = t e bt1
S
n 1 − e bh
.

©2009 Razor Risk Technologies Page 293 of 373

Razor Financial Principals

Note E[ A] can be calculated as E [Aτ ] with t1 = 0 because m > 0 in this

case.

Case 4: One fixing left:

If there is only one more fixing before maturity, the following formula is
used:
[ ]
c fix = S t e (b − r )τ N (d 1 ) − K ∗ e − rτ N (d 2 )
1
n
[ ]
p fix = K ∗ e − rτ N (− d 2 ) − S t e (b − r )τ N (− d 1 )
1
n
where
 S   σ2 
ln ∗  +  b + τ
K   2 
d1 =
σ τ
d 2 = d1 − σ τ
K ∗ = nK − (n − 1)S A .

Levy’s Approximation
We will adapt some slightly different notations and conventions for Levy’s
model.

Define:
S t = the current asset price at time t .

the asset price of the (i + 1) fixing (i.e. t0 is the asset price of

S ti = th
S
the first fixing).
K = the strike price of option.
r = the risk-free rate.
b = the cost of carry.
T = the original time to maturity.
τ = T − t , the remaining time to maturity.
t i = time to the (i + 1)th fixing from current time t (i.e. t 0 is the time to
the first fixing).
σ = volatility of asset.
m = number of observed average points after the first fixing.
n = total number of fixings minus one.
S A = the arithmetic average of the known asset price fixings.
E ∗ = the expectation under risk-neutral measure.

We assume that there are n + 1 number of fixings in total at times

t 0 , t1 ,...t n . Thus t 0 is the time to the first fixing and t n is the time to the
(n + 1)th fixing (the last fixing) from current time t .

We further define:
©2009 Razor Risk Technologies Page 294 of 373
Razor Financial Principals

t n − t0
h= n .
t − tm
ζ = h .
m +1 1 n 1 m m +1 1 n
Mt = S − SA = ∑
n + 1 n + 1 i =0
S ti − ∑
m + 1 i=0
S ti = ∑ St
n + 1 n + 1 i = m +1 i .

S A (m + 1)
K− <0
If n +1 then call will be in-the-money for certain and put
will be out-of-money for certain, thus:

(
c fix (t ) ≈ e − rτ Sˆ − K )
p fix (t ) ≈ 0

where
m +1
Sˆ = S A + E ∗ [M t ]
n +1 .
S t b (1−ζ )h 1 − e b ( n −m )h 
E ∗ [M t ] = e  
n +1  1− e
bh
.

S A (m + 1)
K− ≥0
If n +1 , then the discrete average version of Levy’s
approximation for the case where t ≥ t 0 is:

  S (m + 1) 
c fix (t ) ≈ e − rτ  E ∗ [M t ]N (d1 ) −  K − A  N (d 2 )
  n +1  
  S (m + 1) 
p fix (t ) ≈ e −rτ  E ∗ [M t ] ⋅ [N (d1 ) − 1] −  K − A  ⋅ [N (d 2 ) − 1]
  n +1  

where
S (m + 1) 
1
[ ] 
ln E ∗ M t2 − ln  K − A
n + 1 
d1 =
2 
vt
d 2 = d1 − vt

[ ]
vt = ln E ∗ M t2 − 2 ln E ∗ [M t ]
.
b ( n − m )h
S t b (1−ζ )h 1 − e 
E ∗ [M t ] = e  
n +1  1− e
bh


©2009 Razor Risk Technologies Page 295 of 373

Razor Financial Principals

 1 

[ ] S t2 − 2ζ  b + σ 2  h
E ∗ M t2 = e  2 
( A1 − A2 + A3 − A4 )
(n + 1)2
where
e (2b +σ )h − e (2b +σ )(n − m+1)h
2 2

A1 =
( )(
1 − e bh 1 − e (2b +σ )h
2
)
e [b ( n − m + 2 )+σ ]h − e (2b +σ )(n − m +1)h
2 2

A2 =
(1 − e bh )1 − e (b+σ )h ( 2
)
e (3b +σ )h − e [b (n − m+ 2 )+σ ]h
2 2

A3 =
(1 − e bh )1 − e (b+σ )h ( 2
)
e 2 (2b +σ )h − e (2b +σ )(n − m +1)h
2 2

A4 =
(
1 − e (b +σ )h 1 − e (2b +σ )h
2
)( 2
)
For the case when t < t 0 , we need to modify E [M t ] and E M t .
∗ ∗ 2
[ ]
1 − e b (n +1)h 
E ∗ [M t ] = t e b (t0 −t ) 
S

n +1  1− e
bh


[ ]
E ∗ M t2 =
S t2
e (2b +σ )(t0 −t ) (B1 − B2 + B3 − B4 )
2

(n + 1) 2

1 − e (2b +σ )(n +1)h

2

B1 =
(1 − e )(1 − e (
bh 2 b +σ 2 h ) )
− e (2b +σ )(n +1)h
b ( n +1)h 2
e
B2 =
(1 − e bh )1 − e (b+σ )h ( 2
)
b ( n +1)h
e −e
bh
B3 =
(1 − e )(1 − e (
bh b +σ 2 h ) )
e (2b+σ )h − e (2b +σ )(n +1)h
2 2

B4 =
(
1 − e (b+σ )h 1 − e (2b +σ )h
2
)( 2
).
Levy’s model covers all plain-vanilla, forward start and in progress cases.
It is also not limited by the case where the last fixing date does not
coincide with the option maturity date.

©2009 Razor Risk Technologies Page 296 of 373

Razor Financial Principals

15.7.5 Arithmetic Average Asian Option Pricing Models (Discrete

Averaging, Variable Fixing Weighting and Time-Dependent Underlying
Parameters)

Fixed Strike Asian Option

Define:
S t = the asset price at current time (or valuation date) t .
S ti = th
the asset price of the i fixing.
α i = the normalised weight of the i th fixing.
i

∑α
S ti = u =1 u tu
S
.
n = total number of fixings.
m = number of observed fixings.
K = the strike price.
rti =
the risk-free rate at time t i .
bti =
the cost of carry at time t i .
σ ti = the volatility of the underlying asset at time t i .
T = the original time to maturity.
τ = the remaining time to maturity.
df t ,T =
the discount factor for the period from t to T .
N (⋅) = the cumulative normal distribution function.
E ∗ = the expectation under risk-neutral measure.

Note:
©2009 Razor Risk Technologies Page 297 of 373
Razor Financial Principals

∑α i =1
i =0 .

Similarly to the equal time interval case, we define:

n
M t = S tn − S tm = ∑α S i ti
i = m +1 .
Further, we define:
[ ]
Fti = E ∗ S ti = S t e
bt i (t i −t )

Option Payoff:

Call: max(S − K ,0)

Put: max(K − S ,0)

Derivations:

E ∗ [M t ]
 n 
= E ∗  ∑ α i S ti 
i = m +1 

∑ α E [S ]
n
∗
= i ti
i = m +1
n
= ∑α F
i = m +1
i ti

E ∗ M t2[ ]
 n n 
= E ∗  ∑ α i S ti ∑ α j S t j 
i = m+1 j = m +1 
 n n 
= E ∗  ∑ ∑ α i α j S ti S t j 
i = m +1 j = M =1 
[ ]
n n
= ∑ ∑α α
i = m +1 j = m +1
i j E ∗ S ti S t j

(t i , t j ) min (t i −t ,t j −t )
n n
σ min
2

= ∑ ∑α α
i = m +1 j = m +1
i j Fti Ft j e

c fix ( t )
≈ dft ,T E ∗  max Stn − K , 0 
  ( )
©2009 Razor Risk Technologies Page 298 of 373
Razor Financial Principals

  n 
= dft ,T E ∗  max  ∑ α i Sti − K , 0  
  i =1 
  m n

= dft ,T E ∗  max  ∑ α i Sti + ∑ α i Sti − K , 0  
  i =1 i = m +1 
  m
 
= dft ,T E ∗  max  ∑ α i Sti +M t − K , 0  
  i =1 
   m
 
= dft ,T E ∗  max  M t −  K − ∑ α i Sti  , 0   .
   i =1   

Pricing Formulas:

Case 1: Exercise for certain and out-of-money for certain

m
K − ∑ α i S ti < 0
If i =1 then call will be in-the-money for certain and put will
be out-of-money for certain, thus:

(
c fix (t ) ≈ df t ,T Sˆ − K )
p fix (t ) ≈ 0

where
m
Sˆ = ∑ α i S ti + E ∗ [M t ] = S tm + E ∗ [M t ]
i =1 .
Case 2: Valuation date before last fixing
m
K − ∑ α i S ti ≥ 0
If i =1 then we assume M t is lognormally distributed like the
equal time interval case, the resulted formulas for variable fixing time
intervals are:

  m
 
c fix (t ) ≈ df t ,T  E ∗ [M t ]N (d1 ) −  K − ∑ α i S ti  N (d 2 )
  i =1  
 ∗  m
 
p fix (t ) ≈ df t ,T  E [M t ] ⋅ [N (d1 ) − 1] −  K − ∑ α i S ti  ⋅ [N (d 2 ) − 1]
  i =1  

where
[ ]  
m
1
ln E ∗ M t2 − ln  K − ∑ α i S ti 
d1 =
2  i =1 
vt
d 2 = d1 − vt
[ ]
vt = ln E ∗ M t2 − 2 ln E ∗ [M t ]
.

©2009 Razor Risk Technologies Page 299 of 373

Razor Financial Principals

n
E ∗ [M t ] = ∑α F i ti
i = m +1

[ ] ∑ ∑α α (t i , t j ) min (t i −t ,t j −t )
n n
σ min
2

E ∗ M t2 = i j Fti Ft j e
i = m +1 j = m +1 .

Case 3: Valuation date on or after last fixing

{
c fix (t ) ≈ df t ,T S tm e rt (T −t ) − K }
p fix (t ) ≈ df t ,T {K − S tm e rt (T −t )
}.

Floating Strike Asian Option

Define:
S t = the asset price at current time (or valuation date) t .
S ti = th S
the asset price of the i fixing (i.e. t0 is the asset price of the
first fixing).
α i = the normalised weight of the i th .
i

∑α
S ti = u =1 u u
S
.
n = total number of fixings.
m = number of observed fixings.
K = the strike price.
rti =
the risk-free rate at time t i .
bti =
the cost of carry at time t i .
σ ti =
the volatility of the underlying asset at time t i .
T = the original time to maturity.
τ = the remaining time to maturity.
df t ,T =
the discount factor for the period from t to T .
N (⋅) = the cumulative normal distribution function.
E ∗ = the expectation under risk-neutral measure.

∑α i =1
i =1 .
n
M t = S tn − S tm = ∑α S i ti
i = m +1 .
©2009 Razor Risk Technologies Page 300 of 373
Razor Financial Principals

[ ]
Fti = E ∗ S ti = S t e
bt i (ti −t )
.

Assumptions:
1. t 0 = 0 .
2. t n = T .
3.
[ ( )
ln S tn , ln(S T ) ]
follows a bivariate normal distribution.

Option Payoff:

Call: max(S − K ,0)

Put: max(K − S ,0)

Pricing Formulas:

Case 1: Valuation date before last fixing and last fixing matches the
option expiry date

The pricing formulas are:

{ [ ] }
c float (t ) ≈ df t ,T FT N (d1 ) − E ∗ S tn N (d 2 )
p float (t ) ≈ df t ,T {E [S ]N (− d ) − F N (− d )}
∗
tn 2 T 1

where
 F 
ln ∗ T 
 E St [ ]  1
d1 =  n + σ
σS 2
S

d 2 = d1 − σ S
[ ] [
E ∗ S tn = E ∗ S tm + M t = S tm + E ∗ [M t ]]
n
E ∗ [M t ] = ∑α F i ti
i = m +1

 E [M ] 
 n
( ) 
 ∑ wi Fti exp σ t2i ⋅ (t i − t 0 )
 ×  E [M t ] 
2
∗ ∗
σ S2 = σ 2 (T ) + σ x2  ∗ t  − 2 ln i = m +1
 E St 
 n  [ ] 

E ∗ [M t ]
 
[ ]
  E ∗ St 
n 
 

σ 2 (τ ) = σ T2 (T − t 0 )
σ x2 = v 2 in the fixed strike case.

Case 2: Valuation date after the last fixing

The payoffs for call and put are respectively:

©2009 Razor Risk Technologies Page 301 of 373

Razor Financial Principals

(
max S T − S tn ) and (
max S tn − S T ).
The option prices can be evaluated using standard Black-Scholes model
S
since t n is known and can be treated as a constant strike.

Case 3: Maturity Mistmatch

The Levy model assumes the last fixing time is on the maturity day. In
reality, the last fixing can happen before the maturity day. In such a
case, we can use the double average rate option (DARO) model to
approximate the option price as floating strike Asian option is just a
special case of DARO.
c∗ (t )
To find float , we set β = 1 w1 = 1 , w2 = 1 , and K = 0 in the DARO
(1)
model. We also assume there is only one asset fixing for S and the
fixing point is at option maturity.

p ∗float (t )
To find , we set β = 1 w1 = 1 , w2 = 1 , and K = 0 in the DARO
(2 )
model. We also assume there is only one asset fixing for S and the
fixing point is at option maturity.

For details on DARO model, please refer to the section on DARO.

Asian Basket Options 1

Define:
wis = weight of asset i in the basket
S i (t ) = price of asset i at time t .
bi (t ) = cost of carry of asset i at time t .
σ i (t ) = volatility of asset i at time t .
ρ ij =
correlation between asset i and j
S b (t ) = price of the basket at time t .
σ b (t ) = volatility of the basket at time t .
na = number of assets in the basket.
nf =
number of fixings.

Option Payoff:

max(∑∑ S i (t j ) − K ,0)
na nf

Call: i =1 j =1

 
max K − ∑∑ S i (t j ),0 
na n f

Put:  i =1 j =1 
©2009 Razor Risk Technologies Page 302 of 373
Razor Financial Principals

Pricing Formulas:

na
S b (t ) = ∑ wis S i (t )
i =1 .
na
bb (t ) = ∑ wis bi (t )
i =1 .

∑ (w ) σ (t ) + 2∑∑ w w ρ σ (t )σ (t )
na na i −1
σ b (t ) = i
s 2
i
2
i
s s
j ij i j
i =1 i =1 j =1

Then we can apply the normal fixed strike or floating strike Asian option
formulas to calculate the Asian basket option price.

Asian Quantos Basket Options

Quantos (also called fixed exchange rate foreign equity options) are
option contracts that are denominated in another currency than that of
the underlying equity exposure. Asian Quanto incorporates both the
average rate and cross-currency natures.

Define:
na = number of underlying assets.
nf =
number of fixings.
m= number of observed fixings.
C= payoff currency.
S i (t ) = i by 1 array of asset prices i at time t .
Ci = currency i for asset i .
σS i
(t ) =
j i by j matrix of volatilities of asset i at time t j .
Xi = i by 1 array of exchange rate quoted as payoff currency per
currency i ,i.e. C / Ci .
σ X (t j ) =
i i by j matrix of volatilities X i at time t j .

ρ ij = i by i matrix of instantaneous correlations of ln S i and ln S j

.
ρi = i by 1 array of instantaneous correlations of S i and X i .
rCi (t j ) = i by j matrix of instantaneous risk-free interest rates in C i
market.
bi (t j ) = i by j matrix of cost of carries for asset i at time t j .
©2009 Razor Risk Technologies Page 303 of 373
Razor Financial Principals

K = strike price is payoff currency.

W= Wiener process.

We further define:

S (t n ) = ∑∑ wij S i (t j )
na nf

i =1 j =1 .

Option Payoff:

Call:
(( ) )
max S t n f − K ,0
max(K − S (t ),0)
Put: . nf

Pricing Formulas:

We have assets denominated in different currencies. We need to find a

measure say C , such that all assets under this measure are risk-neutral.
It can be shown (Datey, Gauthier and Simonato) that in such risk-neutral
measure, the asset prices are distributed as:

( )
dS i (t ) = bi (t ) − ρ i ⋅ σ Si (t ) ⋅ σ X i (t ) ⋅ S i (t ) ⋅ dt + σ Si (t ) ⋅ S i (t ) ⋅ dWi C (t )
, ∀i = 1,2,..., n a
.

Regardless all the technical terms, the important thing is that the
underlying asset price still has the form of geometric Brownian motion
with different drift and diffusion term. It means that we can use similar
approaches in Asian non-quanto options for Asian quanto options.

S (t n ) = ∑∑ wij S i (t j ) = S (t m ) + ∑ ∑ w S (t ) = S (t ) + M
na n na nf

ij i j m t
i =1 j =1 i =1 j = m +1

where

∑ w S (t )
na nf
Mt = ∑ ij i j
i =1 j = m +1 .

bi∗ (t j ) = bi (t j ) − ρ i ⋅ σ Si (t j ) ⋅ σ X i (t j )
We denote
and define:
Fi (t j ) = S i (t )e
( )(
bi∗ t j t j −t )

∑ w F (t )
na nf
E ∗ (M t ) = ∑ ij i j
i =1 j = m +1 .

©2009 Razor Risk Technologies Page 304 of 373

Razor Financial Principals

E ∗ (M t2 ) = ∑ ∑ ∑ ∑ w w F (t )F (t ) ×
na nf na nf

ij kl i j k l
i =1 j = m +1 k =1 l = m +1

(
exp ρ ik σ Si (min (t , t ))σ (min (t , t ))min (t − t , t
j l Sk j l j l ).
− t)
m
K − ∑ α i S ti
We can then use the fixed strike formulas with all i =1 replaced
by K − S (t )
m . Exercise for certain and out-of-money for certain cases are
m
K − ∑ α i S ti
the same by replacing i =1 with K − S (t m ) .

Asian Basket Options 2

Another method to price the Asian basket options is similar to Asian
σ (t )
quanto basket. We simply set the the array ρ i to 0, the matrix X i j to
b (t ) b (t )
∗
0 and use i j instead of i j in the Asian quanto basket options
formula.

Double Average Rate Options

Double average rate options are also called Asian spread options. The
payoff is dependent on the differences of two averages of the underlying
asset. The underlying asset can be a single asset or a basket of assets. For
this documentation, we will assume that there is only a single underlying
asset.

Define:
S t(1) = the asset price at time t for the first average.
(2 )
St = the asset price at time t for the second average.
(1)
S ti = th
the asset price of the i fixing for the first average.
S t(i2 ) = th
the asset price of the i fixing for the second average.
α i(1) = th
the normalised weight of the i fixing for the first average.
α i(2 ) = th
the normalised weight of the i fixing for the second
average.
i
(1)
S ti = ∑α ( )S ( ) u
1 1
tu
u =1 .
i

S t(i 2 ) = ∑α ( )S ( ) u
2
tu
2

u =1 .
n1 = total number of fixings in the first average.
n2 = total number of fixings in the second average.
m1 = number of observed fixings in the first average.
©2009 Razor Risk Technologies Page 305 of 373
Razor Financial Principals

m2 = number of observed fixings in the second average.

K = the strike price.
rti =
the risk-free rate at time t i .
bti =
the cost of carry at time t i .
σt = i the volatility of the underlying asset at time t i .
T= the original time to maturity.
τ= the remaining time to maturity.
df t ,T =
the discount factor for the period from t to T .
N (⋅) = the cumulative normal distribution function.
E∗ = the expectation under risk-neutral measure.

(1) (2 )
Note S t = S t for single asset case.

Further, we define:
( )
[ ]
b ( i ) ti( i ) −t
Fti(i ) = E ∗ S t(ii ) = S t(i ) e
ti

Note:
n1

∑α ( ) = 1i
1

i =1 .
n2

∑α ( ) = 1i
2

i =1 .

We further define:
{
Τ1 = t1(1) < t 2( 2 ) < ... < t n(11 ) } and Τ = {t ( ) < t ( ) < ... < t ( ) },
2 1
2
2
2 2
n2 Τ1 ∩ Τ2 is a null set,
where
t i(1) ' s , t i( 2 ) ' s are the time to the average points (i.e. fixing points) in the
first and second average respectively.

Also:
t n(1i ) ≤ Tt (2 ) ≤ T
and n2 so that the last fixings of the two averages can be
either before or on the option maturity date.

Option Payoff:

Call:
[ ]
max w1 S n(11) − w2 S n(22 ) − K ,0
max[K + w S ( ) − w S ( ) ,0]
2 1
Put: 2 n2 1 n1

where w1 , w2 > 0 .

©2009 Razor Risk Technologies Page 306 of 373

Razor Financial Principals

Derivations:

X = w1 S n(11) − w2 S n(22 )
To price the DARO, we assume is normally distributed
with mean µ v
and variance . The probability density function of X
 ( x − µ )2 
f X (x ) =
1
exp − 
2 πv  2v 
becomes .
We need to find the parameters µ and v .

The following equalities can be shown (the derivations are in appendix):

( )
E ∗ S t(i1) = Fti(1)
E (S ( ) ) = F ( )
∗
ti
2
ti
2

( ) ∑ α ( ) E (S ( ) )
n1
E ∗ S t(n11) = S t(m11) + i
1 ∗
ti
1

i = m1 +1

( ) ∑ α ( ) E (S ( ) )
n2
E ∗ S t(m22) = S t(m22) + i
2 ∗
ti
2

i = m2 +1
(1 ) (
(1 ) )
E (S ( ) S ( ) ) = F ( ) F ( ) e
σ2 (1 ) , t (1 )  min ti −t ,t j −t
min  t i
∗ 1 1 1 1  j 

ti tj ti tj
(2) (2) ( )
E (S ( ) S ( ) ) = F ( ) F ( ) e
σ2 ( 2 ) , t ( 2 )  min ti −t ,t j −t
min  t i
∗ 2 2 2 2  j 

ti tj ti tj
(1 ) (
(2 ) )
E (S ( ) S ( ) ) = F ( ) F ( ) e
σ2 (1 ) , t ( 2 )  min ti −t ,t j − t
min  t i
∗ 1 2 1 2  j 

ti tj ti tj
(1) (2 )
Note S t = S t for single asset case.
( ) α i(1) E ∗ (S t(1) ) + (S ( ) S ( ) )
n1 n1 n1
E ∗ S t(n11) = S t(m11) + 2 S t(m11) ∑ ∑ ∑ α ( )α ( ) E ∗
2 2
1 1 1 1
i i j ti tj
i = m1 +1 i = m1 +1 j = m1 +1

( ) α i(2 ) E ∗ (S t(2 ) ) + (S ( ) S ( ) )
n2 n2 n2
E ∗ S t(n22 ) = S t(m22) + 2 S t(m22) ∑ ∑ ∑ α ( )α ( ) E ∗
2 2
2 2 2 2
i i j ti tj
i = m2 +1 i = m2 +1 j = m2 +1

(
E ∗ S t(n1) S t(n2 )
1 2
)
( ) ∑ α i(1) E ∗ (S t(1) ) + (S ( ) S ( ) )
n2 n1 n1 n2
= S t(m11) S t(m22) + S t(m11) ∑ α (j2 ) E ∗ S t(2 ) + S t(2 ) j m2 i ∑ ∑ α ( )α ( ) E i
1
j
2 ∗
ti
1 2
tj
j = m2 +1 i = m1 +1 i = m1 +1 j = m2 +1

We can obtain parameters µ and v :

( )
µ = w1 E ∗ S t(n1) − w2 E ∗ S t(n2 ) 1
( ) 2

v
( ( ) ( ) )+ w (E (S )− E (S ) )
= w12 E ∗ St(n1) − E ∗ St(n1)
2 2 2
2
∗ (2 )
t n2
2
∗ (2 ) 2
t n2

( ( ) ( ) ( ))
1 1

− 2w1 w2 E ∗ S t(n1) S t(n2 ) − E ∗ S t(n1) E ∗ S t(n2 )

1 2 1 2

The price of DARO is:

c DARO (t )
= e − rτ E ∗ (max[β ( X − K ,0 )])

©2009 Razor Risk Technologies Page 307 of 373

Razor Financial Principals

Pricing Formulas:

We assumed X ~ N (µ , v ) , it follows that:

c DARO (t )
 v  1 ( K − µ )2  1 + β  Κ − µ  
≈ df t ,T  exp −  + (µ − K ) − N   
 2π 
 2 v   2  v   .
β is the call-put index (1 for a call, -1 for a put).

Reciprocal Asian Options

Asian options are options where the payoff depends on the average price
of the underlying asset during at least some part of the life of the option.
Reciprocal Asian options are a particular class of Asian options and the
payoff depends on the reciprocal of the average price of the underlying
asset. In this document, we will demonstrate the methods to value fixed
strike European reciprocal Asian call and put options for both discrete and
continuous arithmetic averages.

Define:
S t = the asset price at current time t .
S ti = th
the asset price of the i fixing.
α i = the normalised weight of the i th fixing.
i

∑α
S ti = u =1 u tu
S
.
n = total number of fixings minus one.
m = number of observed fixings minus one.
K = the strike price.
rti =
the risk-free rate at time t i .
bti =
the cost of carry at time t i .
σ ti =
the volatility of the underlying asset at time t i .
T = the original time to maturity.
τ = the remaining time to maturity.
df t ,T =
the discount factor for the period from t to T .
N (⋅) = the cumulative normal distribution function.
E ∗ = the expectation under risk-neutral measure.

Note:
n

∑α i =1
i =1 .

©2009 Razor Risk Technologies Page 308 of 373

Razor Financial Principals

Further, we define:
n
M t = S tn − S tm = ∑α S i ti
i = m +1 .
∗
Fti = E S ti = S t e [ ] bt i (t i −t )

Option Payoff:
 1 1 
N max  − ,0
Call:  S tn S T  .
1 1 
N max  − ,0
Put: 
 S T S tn 
.

N is the notional amount.

n
S t n = ∑ α i S ti
i =1 .

Derivations:

Assume t m ≤ τ ≤ t m+1 , we know all the fixings up to time t m .

m n n
S t n = ∑ α i S ti + ∑α S i ti = S tm + ∑α S i ti = S tm + M t
i =1 i = m +1 i = m +1 .

M t = S tn − S tm = S tn − c
.
c = S tm
We define and it is a known constant.

c fix (t )
  1 1 
= Ndf t ,T E ∗ max − ,0 
  
 S tn K 
  1 1 
= Ndf t ,T E ∗ max − ,0 
  M t + c K 
1 1
> ⇒ Mt < K − c
For the call option to have a non-zero payoff, M t + c K .
We also know that the average of the underlying asset price cannot be
negative. It suggests that the call option has non-zero values when
0 < Mt < K −c.

c fix (t )
  1 1 
≈ Ndf t ,T E ∗ max − ,0 
  M t + c K 

©2009 Razor Risk Technologies Page 309 of 373

Razor Financial Principals

K −c
 1 1
= Ndf t ,T ∫  x + c − K  f (x )dx Mt
x=0 .
f M t (x )
is the probability density function of M t .

Similarly, for put option to have a non-zero payoff,

1 1
> ⇒ Mt > K − c
K Mt + c and we the average of the underlying asset
price cannot be negative.
p fix (t )
 1 1 
≈ Ndf t ,T E ∗ max − ,0 
  K M t + c 
∞
1 1 
= Ndf t ,T ∫(  −  f M ( x )dx
x = max 0, K − c ) 
K x + c t
f M t (x )
is the probability density function of M t . It is sufficient for the
upper limit for the integral to be
E ∗ (M t ) + 10 Var (M t ) = E ∗ (M t ) + 10 E ∗ M t2 − E ∗ (M t ) ( ) 2
.

We assume ln(M t ) is normally distributed with mean α t and standard

deviation vt . α t and vt can be approximated using Levy (1992)
approach.

Pricing Formulas:

K −c
 1 1
c fix (t ) ≈ Ndf t ,T ∫  x + c − K  f (x )dx Mt
x =0 .
( )
E ∗ ( M t )+10 E ∗ M t2 − E ∗ ( M t )2
1 1 
p fix (t ) ≈ Ndf t ,T ∫(  −  f M t (x )dx.
x = max 0 , K − c ) K x+c

 (ln x − α t )2 
f M t (x ) =
1
exp − 
xvt 2π  2vt2  x > 0
, .
α t = 2 ln E [M t ] − ln E M t
∗ 1
2
∗ 2
[ ]
.
[ ]
vt = ln E ∗ M t2 − 2 ln E ∗ [M t ]
.
n
E ∗ [M t ] = ∑α F i ti
i = m +1

[ ] ∑ ∑α α (t i , t j ) min (t i −t ,t j −t )
n n
σ min
2

E ∗ M t2 = i j Fti Ft j e
i = m +1 j = m +1 .
©2009 Razor Risk Technologies Page 310 of 373
Razor Financial Principals

Note the integral can be evaluated numerically using Gauss-Legendre

Quadrature.

Callable Asian Options

Note that support is only required for arithmetic averaging, with fixed
strike.

Let S( t ) be a price process of a given underlying asset, {t 1 < L < t n } be a

set of reset dates and T ≥ t n be a payoff settlement date. A callable
Asian option with the underlying, S, is a derivative security whose
matured payoff at the settlement date is given by

I N ⋅ N + N ⋅ max{0, β ( A − K ) / K } (1)

where
N is the notional principal, I N (= 0 or 1) is the notional payment indicator
function, β is the call-put index, K is the strike, and A is the arithmetic
average of S( t i ), i = 1, L , n , with equal weight.

Equation (1) indicates that the principal is protected if I N = 1 . A payoff

type without notional is also allowed, which removes the first part of
equation (1) at maturity, but the owner of the option can still receive the
call amount if the option is called on the call date.

Call Feature
Let t c be the call date. At time t c , t c < t 1 , if the underlying stock price
S( t c ) is above a barrier, Pc , the deal will be cancelled, either
automatically or optimally by the underwriter, in which case the owner of
the option will receive a call premium, C (automatic cancellation) or the
minimum of C and the value of (1) at time t c (optimal cancellation); if
the deal will not be cancelled, the payoff at the settlement date is (1).
Forward Start Feature
The callable Asian option can also allow a forward start feature. This
t
option specifies that at time f , the strike price of the Asian option, K,
S (t f )
will be set equal to the current spot price, . With the forward start
feature the payoff at the settlement date is
I N ⋅ N + N ⋅ max {0, β ( A − S (t f )) / S (t f )}
(2)
if the deal will not be cancelled. Otherwise, the owner of the option will
receive a call premium, C (automatic cancellation) or the minimum of C
and the value of (2) at time t c (optimal cancellation).
©2009 Razor Risk Technologies Page 311 of 373
Razor Financial Principals

Pricing for the Call Feature Case

Let t be the current valuation date and assume t < t c . Also denote
Vtc (S (t c ))
to be the value of the option at time t c .

Automatic Cancellation

  β ( A − K ) 
Vtc (S (t c )) = I (S (t c ) < Pc ) × df tc ,T × E tc  I N ⋅ N + N × max 0, 
  K 
+ I (S (t c ) ≥ Pc ) × C
  β ( A − K ) 
df tc ,T × Etc  I N ⋅ N + N × max 0, 
where   K  is the expression for the
value of a vanilla Asian option which can be approximated with Michael
df
Curran formula and tc ,T is the discount factor for the period from time
t c to T .

  β ( A − K ) 
MC (S (t c )) = df tc ,T × Etc  I N ⋅ N + N × max 0, 
We denote   K  where,
MC (S (t c )) is the Michael Curran formula including the notional
adjustments. We can then write:

Vtc (S (t c )) = I (S (t c ) < Pc ) × MC (S (t c )) + I (S (t c ) ≥ Pc ) × C
.

Optimal Cancellation
With optimal cancellation, we can write:

Vtc (S (t c )) = I (S (t c ) < Pc ) × MC (S (t c )) + I (S (t c ) ≥ Pc ) × min {C , MC (S (t c ))}

.

For both cancellation conditions, the value of the option at time t , is:

[ ]
Vt (S (t )) = df t ,tc × E t Vtc (S (t c ))

where

∞
[ ]
Et Vtc (S (t c )) = ∫ Vtc (S (t c )) f (S (t c ))dS (t c )
0 ,
f (S (t c )) is the density function of S (t c ) .

Derivation of f (S (t c )) :
Under risk-neutral world,

©2009 Razor Risk Technologies Page 312 of 373

Razor Financial Principals

dS (t ) = btc S (t )dt + σ tc S (t )dWt

.

Taking Ito’s Lemma, and integrates, we then get:

ln S (t c )
 
= ln S (t ) +  btc − σ t2c (t c − t ) + σ t2c Wtc −t
1
 2 
   
~ N  ln S (t ) +  btc − σ t2c (t c − t ), σ t2c (t c − t )
1
  2  ,
i.e. S (t c ) is lognormally distributed with
 (ln S (t c ) − µ )2 
f (S (t c )) =
1
exp − 
S (t c )v 2π  2v 2  S (t c ) > 0
; ,
where

 
µ = ln S (t ) +  bt − σ t2 (t c − t )
1
 2 
c c

v 2 = σ t2c (t c − t )
btc σ t can be obtained from the term structure for cost of carries
and c

and volatilities with maturity t c .

The integral can be evaluated using Gauss-Legendre quadrature with the

care to divide the integral at the non-differentiable points of the
integrand. For automatic cancellation, the integrand is non-differentiable
at Pc . For optimal cancellation, the integrand is non-differentiable at Pc
and/or the point S (t c ) , such that C = MC (S (t c )) .
∗ ∗

Forward cost-of-carries and volatilities

It is important to note that the callable Asian options are evaluated at
time t c first and then brought back to time t . Thus we need the term
structure curve for cost of carries and volatilities starting from time t c
when applying the Michael Curran formula. The formulas for transforming
term structure curves are:
bt t i − btc t c
btc ,ti = i ,
ti − tc
σ t2 t i − σ t2 t c
σ 2
= i c
.
ti − tc
t c ,t i

Quanto case
Callable Asian option with quanto feature is the same except the cost of
carries needs to be adjusted as in the quanto Asian basket options case,

©2009 Razor Risk Technologies Page 313 of 373

Razor Financial Principals

i.e. bi∗ (t j ) = bi (t j ) − ρ i ⋅ σ Si (t j ) ⋅ σ X i (t j ) . Note that cost of carries should be

adjusted for quanto before converting to forward rates.

Pricing for the Forward Start Case

If the option is forward starting, both the future asset price and strike
price are random variables at the valuation time t . We can write the
Michael Curran formula as:

MC (S (t f ), S (t c )) = MC (K = S (t f ), S (t c ))
.

Vtc (S (t f ), S (t c ))
The value the option, , at time t c , is given by:

Automatic Cancellation

Vtc (S (t f ), S (t c )) = I (S (t c ) < Pc ) × MC (S (t f ), S (t c )) + I (S (t c ) ≥ Pc ) × C
.

Optimal Cancellation

Vtc (S (t f ), S (t c )) = I (S (t c ) < Pc ) × MC (S (t f ), S (t c )) +
I (S (t c ) ≥ Pc ) × min {C , MC (S (t f ), S (t c ))}
.

Vtc (S (t f ), S (t c ))
Therefore, given , the value of the option, Vt (S (t )) , at time
t is given by:

[
Vt (S (t )) = df t ,tc × E t Vtc (S (t f ), S (t c )) ]
where

[
Et Vtc (S (t f ), S (t c )) ]
[ [
= Et Et f Vtc (S (t f ), S (t c )) ]]
∞∞
= ∫ ∫ Vtc (S (t f ), S (t c )) ⋅ f Stc |St f (S (t c )) f St f (S (t f ))dS (t c )dS (t f )
0 0 ,

(S (t c )) S (t f )
is the conditional density function of S (t c ) given
f St | St f
c
, and
f St (S (t f )) S (t f ).
f
is the density function of

With similar derivations, we can show:

 (ln S (t c ) − µ )2 
f Stc |St f (S (t c )) =
1
exp − 
S (t c )v 2π  2v 2  S (t c ) > 0
; ,

©2009 Razor Risk Technologies Page 314 of 373

Razor Financial Principals

where

v 2 = σ t2f ,tc (t c − t f )
σ t − σ t2 t f
2

σ
tc c
2
= f

tc − t f
t f ,t c

σt tf σt
can be obtained from the volatility structure with maturity
f
and c

can be obtained from the volatility structure with maturity t c .

 (ln S (t f ) − µ )2 
f (S (t f )) = 1
exp− 
S (t f )v 2π  2v 2  ; S (t f ) > 0 ,
where

µ = ln S (t ) +  bt − σ t2 (t f − t )
 1 
 2 
f f

v = σ t2f (t f − t )
σt tf
f
can be obtained from the volatility structure with maturity .

We can approximate the double integral using two-dimensional Gauss-

Legendre quadrature and tensor product. The non-differentiable points
for the integrand in automatic cancellation is at Pc . In optimal
cancellation, the non-differentiable points for the integrand are at Pc
and/or the point
S ∗ (t c ), S ∗ (t f ) [
, such that
]
C = MC S ∗ (t c ), S ∗ (t f )
.
( )
Tensor Product Rule
Tensor product is a technique used in multi-dimensional numerical
integration. For the following, we will focus on two-dimensional integral
since it is highest dimension we need in the callable model. However, the
same concept can be applied to higher dimensional problems.

Define f ( x1 , x 2 ) be a function of two variables. We want to evaluated the

following two-dimensional integral:

I= ∫ ∫ f (x , x )dx dx
1 2 1 2
x1 x2 .

For each x1 , the inner integral can be approximated as:

I 2 = ∫ f ( x1 , x 2 )dx 2 ≈ ∑ wˆ j f (x1 , xˆ 2, j )
x2 j

ŵ j xˆ 2 , j
where and are Gauss-Legendre weights and points.

©2009 Razor Risk Technologies Page 315 of 373

Razor Financial Principals

The outer integral can be evaluated similarly as:

∑ wˆ i ∑ wˆ j f (xˆ1,i , xˆ 2, j )
i j

where again
ŵi and xˆ1, j are Gauss-Legendre weights and points.

The two-dimensional integral is:

I = ∫ ∫ f ( x1 , x 2 )dx1 dx 2 ≈ ∑∑ wˆ i wˆ j f (xˆ1,i , xˆ 2, j ) = ∑ wˆ l f ( xˆ l )
i j l
x1 x2 .

It is obvious that ŵl contain all the combinations of products of ŵi and
ŵ j (xˆ , xˆ )
and x̂l contain all the combinations of the pair i j .

Non-smooth points
It is important to note that Gauss-Legendre Quadrature can only be
applied to smooth functions. It is clearly that for callable payoff, the
integrands are not smooth functions for both call-feature and forward-
start cases. We need to split the integrand into sections so that the
integrand in each section is smooth. We can then integrate each smooth
integrand using Gauss-Legendre Quadrature and then sum each section
together to obtain the final integral result.

Call feature case

The integrand for call feature case is not smooth at the point when
S (t c ) = Pc with automatic cancellation and MC (S (t c )) = C for the optimal

cancellation case. We need to find those S (t c ) such that S (t c ) = Pc and

∗ ∗

MC (S ∗ (t c )) = C . We then split the integral at those S ∗ (t c ) points so that

the integrand between two S (t c ) points is smooth.

∗

Forward start case

Forward start case
Forward start case is more complicated because of the double integral.
However, the idea is similar. We first find out the Gauss-Legendre points
for both the outer and inner integrals. We then have a set of Gauss points
S (t f )
for the outer integral and S (t c ) for the inner integral. For each of
S (t f )
the point, we fix it in the inner integral and treat it as a single
integral. We can then obtain the non-smooth points and calculate the
integral just as we did in the call feature case. We then repeat for each
S (t f )
of the points. We obtain the final double integral by summing each
S (t f )
of the inner integral each with the corresponding Gauss point .

Upper bound
Theoretically, we need to integrate the integral to infinity for both call
feature and forward start cases. However, it is unnecessary to integrate

©2009 Razor Risk Technologies Page 316 of 373

Razor Financial Principals

over the values above say 15 standard deviations as the probabilities of

reaching those values are very trivial. The following are examples of
truncating the integral at proper points:

1
µ + v2
E ∗ (S ) = e 2
.
Var (S ) = e (2 µ +v ) e v − 1 .
∗ 2
( 2
)
We can set upper limit to be:
upper = E ∗ (S ) + 15 ⋅ Var ∗ (S ) .
We can apply this idea for both call feature and forward start case with
corresponding µ and v for each integral.

Valuation date on or after the forward start date or the call date
If the valuation date is on or after the forward start date but before the
call date then the strike price is known and the option can be treated as
a non-forward start callable Asian option with the strike price to be set to
the observed spot price on the forward start date. If the valuation date is
after the call date then the callable Asian option becomes a standard
Asian option with price equal to

V (t )
 
= df t ,T × E ∗  I N ⋅ N + × max{0, β ( A − K )} | Ft 
N
 K 
= df t ,T × I N ⋅ N + × E ∗ [max {0, β ( A − K )} | Ft ]
N
K
= df t ,T × I N ⋅ N + × MichaelCur ranFormula (t )
N
K

Valuation date on or after the last fixing date

{ }
c(t ) ≈ df t ,T S t e rt (T −t ) − K
p(t ) ≈ df t ,T{K − S e }.
( )
t
rt T −t

Forward Start Asian Options

Forward start Asian option is like a standard Asian option but with the
strike price being determined as a predetermined multiple of the
underlying asset price at some predetermined date (the strike reset
date).

Define:
na = number of underlying assets.
nf =
number of fixings.
m= number of observed fixings.

©2009 Razor Risk Technologies Page 317 of 373

Razor Financial Principals

S i (t ) = i by 1 array of asset prices i at time t .

σ S (t j ) =
i i by j matrix of volatilities of asset i at time t j .
ρ ij = i by i matrix of instantaneous correlations of ln S i and ln S j
.
rCi (t j ) = i by j matrix of instantaneous risk-free interest rates in C i
market.
bi (t j ) = i by j matrix of cost of carries for asset i at time t j .
tf =
time of the strike reset date.
λi = the strike reset ratio of asset i .
K = strike price is payoff currency.
W= Wiener process.

Option Payoff:

  nf f  
  ∑ wi S j (t i )  
 na
  
max β ⋅ ∑ w aj  i =1 − 1,0 
Kj
 j =1
  
 
   .
β = 1 if it is a call and β = −1 if it is a put.

Pricing Formulas:

t0 < t f
Case 1:

Define:
na
W1 = ∑ w aj λ j
j =1 .
wjλ j
wj =
W1 .
S j (t i )
R j (t f , t i ) =
S j (t f )
.
1
K1 =
W1 .

The payoff in this case can be expressed as:

   na n f  
maxW1 ⋅ β ⋅  ∑∑ w j wia R j (t f , t i ) − K 1 ,0 
   .
   j =1 i =1

©2009 Razor Risk Technologies Page 318 of 373

Razor Financial Principals

R j (t f , t i )
can be expressed as:

ti  1  ti
∫t f  b j ( s )− 2σ j ( s )  ds + ∫t f σ j ( s )dW j ( s )
R j (t f , t i ) = e
2

R j (t f , t i )
The relative price can be viewed as the price of an asset with
S (t )
same dynamics as the asset price j but with the price starts from 1 at
tf
time . The above payoff is exactly the same as the Asian basket option
and can be priced using the Asian basket option formula.

t0 ≥ t f
Case 2:

Define:
na
wjλ j
W2 = ∑
j =1 S j (t f ) .
wjλ j
wˆ j =
W2 S j (t f )
1
K2 =
W2 .

The payoff in this case can be expressed as:

   na n f  
maxW2 ⋅  β ⋅  ∑∑ wˆ j wia S j (t i ) − K 2 ,0 
   .
   j =1 i =1

The payoff is the same as the Asian basket case. The price can be
evaluated using the Asian basket case formula.

Note for Asian forward starting option formula can be applied to the case
with quanto features. Same adjustment for the drift term as in the
quanto case is used to modify the formula.

Forward cost-of-carries and volatilities

Since R j (t f , t i ) is from time t f , we need the forward cost of carries and
volatilities. The formulas to obtain these numbers are:
bti t i − bt f t f
bt f ,ti =
ti − t f
σ t2 t i − σ t2 t f
σ 2
= i f

ti − t f
t f ,t i

Quanto case

©2009 Razor Risk Technologies Page 319 of 373

Razor Financial Principals

Callable Asian option with quanto feature is the same except the cost of
carries needs to be adjusted as in the quanto Asian basket options case,
i.e. bi∗ (t j ) = bi (t j ) − ρ i ⋅ σ Si (t j ) ⋅ σ X i (t j ) . Note that cost of carries should be
adjusted for quanto before converting to forward rates.

Foreign Exchange Fixed Strike/Floating Strike/Double Average Rate Asian Options

The pricing formula is exactly the same as equity Asian options with the
r − rf
cost of carry replaced by d (if there is no dividend payout) where rd
r
is the domestic risk-free rate and f is the foreign risk-free rate.

Foreign Exchange Reciprocal Averaging Asian Options

The above adjustment for foreign exchange Asian options cannot be used
for reciprocal averaging Asian options. The reason is because the
numeraire used in foreign exchange reciprocal averaging Asian options is
foreign currency rather than the domestic currency. However, the
adjustments required for reciprocal averaging are still small.

Recall that in equity reciprocal Asian option model,

[ ]
Fti = E ∗ S ti = S t e
.
bt i (t i −t )

The first adjustment is that there is an extra term for futures price, i.e.
[ ]
Fti = E ∗ S ti = S t e
(bti )
+σ t2i (ti 0t )
b = rtid − rtif
, where ti for the foreign exchange
d f
case if there is not dividend payout and r and r are domestic and
foreign risk-free rates respectively. We then use the equity reciprocal
formula to calculate the option price with the changes of the above
adjustments.

The second adjustment required is the resultant price based on equity

reciprocal formula needs to multiply the spot exchange rate on the
valuation date and the foreign discount factor to obtain the final foreign
exchange reciprocal Asian option price. The exchange rate is quoted as
domestic/foreign.

15.7.6 References
Buchen, P., “Stocks, Derivatives and Exotics”, Lecture Notes for the
School of Mathematics, the University of Sydney.

Callable Call Option section is provided by RBC.

Castellacci, G., and Siclari, M., “Asian Basket Spreads and Other Exotic
Averaging Options”, OpenLink.

Curran, M. (1992): “Beyond Average Intelligence”, Risk Magazine, 5(10).

©2009 Razor Risk Technologies Page 320 of 373

Razor Financial Principals

Curran, M. (1994): “Valuing Asian and Portfolio Options by Conditioning

on the Geometric Mean Price”, Management Science, 40(12), 1705-1711.

Daniel Dufresne (2000), “Laguerre Series for Asian and Other Options”,
Mathematical Finance 10(4) 407-428.

Datey, J. Y., Gauthier, G., and Simonato, J.G (2003): “The Performance
of Analytical Approximations for the Computation of Asian Quanto-Basket
Option Prices”, Multinational Finance Journal 7(1&2), 55-82.

Double Average Rate Option is provided by RBC.

Fixed Strike Reciprocal Asian Option Model with Discrete Arithmetic

Average is provided by RBC.

Foreign Exchange Fixed Strike Asian Option Model Specification is

provided by RBC.

Foreign Exchange Floating Strike Asian Option Model Specification is

provided by RBC.

Foreign Exchange Reciprocal Averaging Asian Option Model Specification

is provided by RBC.

Foreign Exchange Double Average Rate Asian Option Model Specification is

provided by RBC.

Forward Start Asian Option Model Specification is provided by RBC.

Haug, E.G.: “The Complete Guide to Option Pricing Formulas”, First

Edition.

Haug, E.G.: “The Complete Guide to Option Pricing Formulas”, Second

Edition.

Henderson, V., and R. Wojakowski (2001): “On the Equivalence of

Floating and Fixed-Strike Asian Options”, Journal of Finance, 52(3), 923-
973.

Kemna, A. and A. Vorst (1990): “A Pricing Method for Options Based on

Average Asset Values”, Journal of Banking and Finance, 14, 113-129.

Levy, E. (1992): “Pricing European Average Rate Currency Options”,

Journal of International Money and Finance, 11, 474-491.

Levy, E., and S. Turnbull (1992): “Average Intelligence”, Risk Magazine,

5(2).

©2009 Razor Risk Technologies Page 321 of 373

Razor Financial Principals

Levy, E. (1997): “Asian Options”, in Exotic Options: The State of the Art,
ed. L. Clewlow and C.Strickland (Washington, DC: International Thomson
Business Press).

Royal Bank of Canada, “Callable Asian Option”, Risk Infrastructure

Development, Market and Trading Credit Risk, Group Risk Management.

Turnbull, S. M., and L. M. Wakeman (1991): ‘A Quick Algorithm for Pricing

European Average Options”, Journal of Financial and Quantitative
Analysis, 26, 377-389.

15.7.7 Appendix (Proofs of the Identities for Double Average Rate

Options):

( ) ∑ α ( ) E (S ( ) )
n1
E ∗ S t(n11) = S t(m11) + i
1 ∗
ti
1

i = m1 +1
1.
E (S ( ) )
∗ 1
t n1

 n1 
= E ∗  ∑ α i S t(i1) 
 i =1 
 m1 n1

= E ∗  ∑ α i S t(i1) + ∑ α i S t(i1) 
 i =1 i = m1 +1 

∑ α E (S ( ) )
n1
= S t(m11) + i
∗
ti
1

i = m1 +1

©2009 Razor Risk Technologies Page 322 of 373

Razor Financial Principals

( ) ∑ α ( ) E (S ( ) ) + ∑ ∑ α ( )α ( ) E (S ( ) S ( ) )
n1 n1 n1
E ∗ S t(n11) = S t(m11) + 2 S t(m11) ∗ ∗
2 2
1 1 1 1 1 1
i ti i j t1 t1
i = m1 +1 i = m1 +1 j = m1 +1
2.
E (S )
∗ (1) 2

t n1

  m1 n1
 
2
 (1) (1)
= E  ∑ α i S ti + ∑ α i S ti 
∗ (1) (1)
 i =1 
 i = m1 +1  
 2

(1) (1) 
n1
= E  S tm1 + ∑ α i S ti  
∗
 (1) 
 
 i = m1 +1  
 2 2

(1) (1)  (1) (1) 
n1 n1
= E S tm1 + 2 S tm1 ∑ α i S ti +  ∑ α i S ti  
 ∗(1) (1)  
 
 i = m1 +1  i = m1 +1  
  n1  
2

( )
n1
= S t(m11) + 2 S t(m11) ∑ α i(1) E ∗ S t(i1) + E ∗   ∑ α i(1) S t(i1)  
2

 i = m +1 
i = m1 +1
 1  
 
∑ α ( ) E (S ( ) ) + E  ∑ ∑ α ( )α ( ) S ( ) S ( ) 
n1 n1 n1
= S t(m11) + 2 S t(m11) ∗ ∗
2
1 1 1 1 1 1
i ti i j ti tj
i = m1 +1  i = m1 +1 j = m1 +1 

∑ α ( ) E (S ( ) ) + ∑ ∑ α ( )α ( ) E (S ( ) S ( ) )
n1 n1 n2
= S t(m11) + 2 S t(m11) ∗ ∗
2
1 1 1 1 1 1
i ti i j ti tj
i = m1 +1 i = m1 +1 j = m1 +1

3.
(
E ∗ S t(n1) S t(n2 )
1 2
)
( ) ∑ α ( ) E (S ( ) ) + ∑ ∑ α ( )α ( ) E (S ( ) S ( ) )
n2 n1 n1 n2
= S t(m11) S t(m22) + S t(m11) ∑ α (j2 ) E ∗ S t(2 ) + S t(2 ) j m2 i
1 ∗
ti
1
i
1
j
2 ∗
ti
1 2
tj
j = m2 +1 i = m1 +1 i = m1 +1 j = m2 +1

(
E ∗ S t(n1) S t(n2 )
1 2
)
  m1 n1
 m2 n2

= E ∗   ∑ α i(1) S t(i1) + ∑ α i(1) S t(i1)  ∑ α i(2 ) S t(i2 ) + ∑ α i(2 ) S t(j2 )  
 i =1 
 i = m1 +1  i =1 i = m2 +1 
 m1 m2 m1 n2 n1 m2
= E ∗  ∑ α i(1) S t(i1) ∑ α i(2 ) S t(i2 ) + ∑ α i(1) S t(i1) ∑ α i(2 ) S t(i2 ) + ∑ α i(1) S t(i1) ∑ α i( 2 ) S t(i2 )
 i =1 i =1 i =1 i = m2 +1 i = m1 +1 i =1

n1 n2

+ ∑ α ( ) S ( ) ∑ α ( ) S ( ) 
i
1
ti
1
i
2
ti
1

i = m1 +1 
i = m2 +1

 n2 n1 n1 n2

= E ∗  S t(m11) S t(m22) + S t(m11) ∑ α i(2 ) S t(i2 ) + S t(m22) ∑ α i(1) S t(i1) + ∑ ∑ α i(1)α i(2 ) S t(i1) S t(i2 ) 
 i = m2 +1 i = m1 +1 i = m1 +1 i = m2 +1 

∑ α ( ) E (S ( ) ) + S ( ) ∑ α ( ) E [S ( ) ] + ∑ ∑ α ( )α ( ) E [S ( )S ( ) ]
n2 n1 n1 n2
= S t(m11) S t(m22) + S t(m11) i
2 ∗
ti
2 2
t m2 i
1 ∗
ti
1
i
1
i
2 ∗
ti
1 2
tj
i = m2 +1 i = m1 +1 i = m1 +1 i = m2 +1

4.
( )
µ = w1 E ∗ S t(1) − w2 E ∗ S t( 2 )n1
( ) n2

µ
= E ∗ (X )
©2009 Razor Risk Technologies Page 323 of 373
Razor Financial Principals

( )
= E ∗ w1 S t(n1) − w2 S t(n2 )
= w E (S ) − w E (S ( ) )
1 2
∗ (1) ∗ 2
1 t n1 2 t n2

5.
v
( ( ) ( ) )+ w (E (S )− E (S ) )
= w12 E ∗ St(n1) − E ∗ St(n1)
2 2 2
2
∗ (2 )
t n2
2
∗ (2 ) 2
t n2

( ( ) ( ) ( ))
1 1

− 2w1 w2 E ∗ S t(n1) S t(n2 ) − E ∗ S t(n1) E ∗ S t(n2 )

1 2 1 2

v
= var( X )
(
= var w1 S t(n1) − w2 S t(n2 ) )
( )( ) ( )
1 2

= w var S tn + w var S t(n2 ) − 2w1 w2 Cov S t(n1) , S t(n2 )

2
1
(1) 2
2

= w (E (S ( ) ) − E (S ( ) ) ) + w (E (S ( ) ) − E (S ( ) ) )
1 2 1 2

2 ∗ 1 2
∗ 1 2 2 ∗ 2 2
∗ 2 2
1 t n1 t n1 2 t n2 tn2

( (
− 2w1 w2 E ∗ S t(n1) S t(n2 ) − E ∗ S t(n1) E ∗ S t(n2 )
1 2
) ( ) ( ))1 2

15.8 Equity Swap

15.8.1 Contract Definition

An equity swap is an agreement between two parties where one party
agrees to make payments based on the return of equities and another
party agrees to make payments based on the return of a fixed/floating
interest rate plus a spread (or less usually based on the return of other
equities). There is no exchange of notional at start or end.

15.8.2 Definition for Notations

t = valuation date.
St = price of the equity at time t .
Tpm = the payment date associated with the mth reset period.
Tsm = the reset date associated with the m th reset period.
rt →m
= risk-free rate applied for time m based on the simulated term structure
at time t .
qt →m
= dividend yield applied for time m based on the term structure at time
t.
dft ,t j = discount factor from time t to time t j .
©2009 Razor Risk Technologies Page 324 of 373
Razor Financial Principals

Ts0 = the last reset date before time t .

S0
= the last reset price of the equity before time t .
f t ,t1 →t2 =
the forward interest rate applied for time t2 based on the term structure
at time t1 .
f fix = fixed interest rate.
r0 = the last reset interest rate before time t .
N = the notional amount.
fxt = spot foreign exchange rate (Domestic/Foreign) at valuation
date t .
n = number of cash flows.
E [ ] =
∗
expectation under risk-neutral measure.
Ef [ ]= expectation under forward martingale measure.
Bt = savings account at time t .

Note that ft ,t1 →t2 = E f  rt1 →t2 | Ft  .

15.8.3 Valuation
Equity swaps consist of two streams of cash flows: cash flows from equity
(equity leg) and cash flows from interest rate (interest rate leg). The
parties under the contract receive one stream of cash flows and pay the
other stream of cash flows. Thus the value of the equity swaps at any
valuation date can be calculated as the difference between the values of
the equity leg and interest rate leg at the valuation date. We
demonstrate on how to calculate the values of the two legs in the
following.

Equity Leg
ST m
The return of the equity leg for the m th reset period for is s
−1 .
ST m−1
s

ST m
The value of s
− 1 at the valuation date t is:
ST m−1
s

 B  ST m  
E∗  t  s − 1 | Ft 
 BTpm  ST m−1 
  s  
 ST m 
= dft ,T m E f  s − 1| Ft 
p
 STsm−1 

p  ((
≈ dft ,T m exp rt →T m − qt →T m
s s
) (T − t ) − ( r
s
m
t →Tsm −1 s
) )
− qt →T m−1 (Tsm −1 − t ) − 1 .


©2009 Razor Risk Technologies Page 325 of 373

Razor Financial Principals

Note that dft ,T m is derived from the zero-coupon bond price.

s

If m = 1 and Ts0 ≤ t < Ts1 then the value for the first reset period of the
equity leg can be derived similarly as:

((
 S exp r 1 − q 1 T 1 − t 
)( ))
df t ,T 1  − 1 .
t t →Ts t →Ts s

p
S0
 

To get the value of the whole equity leg, we simply sum each equity
return value for each reset period. We also need to multiply by the
foreign exchange rate if cross currency is present.

Thus, to summarise, the present value of the equity leg at valuation date
t taking the cross currency into consideration is:

Case 1: Tm0 ≤ t < Tm1

pvequity ( t )
n
= ∑ df t ,T m × N × fxt
p
k =2

 ((
 s s
)(   s
) (
×  exp  rt →T m − qt →T m Tsm − t  −  rt →T m−1 − qt →T m−1 Tsm−1 − t  − 1
s  
)( ) )
 t ((
 S exp r 1 − q 1 T − t
t →Ts t →Ts s
1 
)( ))
+ dft ,T 1 × N ×  − 1 × fxt .
p
 S 0 
 

Case 2: t < Ts0

pvequity ( t )
n
= ∑ df t ,T m × N × fxt
p
m =1

  ((
×  exp  rt →T m − qt →T m
s s
) (T − t ) − ( r
s
m
t →Tsm −1 s
)( )  )
− qt →T m−1 Tsm−1 − t  − 1 .


Interest Rate Leg

The way to calculate the value of interest rate leg is similar to equity leg,
and is the same as the valuation of such a leg as part of an interest rate
swap.

The return of the interest rate leg for the mth reset period for is
(rTsm −1 →Tsm )(
+ spread Tms − Tms −1 . )
( )(
The value of the of rT m−1 →T m + spread Tms − Tms −1 at the valuation date t is:
s s
)

©2009 Razor Risk Technologies Page 326 of 373

Razor Financial Principals

 B 
 BTpm s
( s
)
E ∗  t rT m−1 →T m + spread (Tms − Tms −1 ) | Ft 

 

p  s (
= dft ,T m E  rT m−1 →T m + spread (Ts − Ts ) | Ft 
f 
s
m m −1
 )
(
= dft ,T m f t ,T m−1 →T m + spread Tsm − Tsm−1
p s s
)( )
If m = 1 and Ts0 ≤ t < Ts1 then the value for the first reset period of the
interest rate leg can be modified as:
df t ,T1, p × N × ( r 0 + spread ) × (Ts1 − Ts0 ) × fxt .

To get the value of the whole interest rate leg, we simply sum each
interest rate return value for each reset period. We also need to multiply
by the foreign exchange rate if cross currency is present.

Thus, to summarise, the present value of the interest rate leg at

valuation date t taking the cross currency into consideration is:

Case1: Ts0 ≤ t < Ts1

The present value if it is floating interest rate equals:
pv float ( t )

(( f ) )
+ spread × (Tsm − Tsm −1 ) × fxt
n
= ∑ dft ,T m × N × t ,Tsm−1 →Tsm
p
m=2

+ dft ,T1, p × N × ( r 0 + spread ) × (Ts1 − Ts0 ) × fxt .

Case 2: t < Ts0

pv float ( t )

(( f ) )
+ spread × (Tsm − Tsm −1 ) × fxt .
n
= ∑ df t ,T m × N × t ,Tsm−1 →Tsm
p
m =1

If the interest rate leg pays fixed interest rate, then it is much simpler.

The present value if it is fixed interest rate equals:

pv fix ( t )

(( f ) )
+ spread × (Tsm − Tsm −1 ) × fxt .
n
= ∑ df t ,T m × N × t ,Tsm−1 →Tsm
p
m =1

In general, the present value of the equity swap at time t is:

pv ( t ) = pvreceivingLeg ( t ) − pv payingLeg ( t ) .

For example, if a party enters an equity swap by paying floating interest

rate and receiving equity return, the preset value of the equity swap at
time t is:

©2009 Razor Risk Technologies Page 327 of 373

Razor Financial Principals

pv ( t ) = pvequity ( t ) − pv floating ( t ) .

Within Razor current support for equity swaps, there is consideration for
discrete dividend payments associated with equity leg and similarly no
support for compounding interest rate leg.

©2009 Razor Risk Technologies Page 328 of 373

Razor Financial Principals

Chapter 16
Credit Derivatives

16.1 Credit Default Swaps

16.1.1 Description of Instrument
Credit Default Swaps are products in which the purchaser of protection makes
periodic payments in exchange for the payment of the protection amount
contingent on there being a default event. What constitutes a default event is
defined within the CDS documentation.

16.1.2 XML Representation

Credit Default Swaps became a standard in FPML version 4.
The schema for fpmlCreditDefaultSwap is defined as follows

fpmlCreditDefaultSwap Schema
Name: Type Occurs Size Description
productType
0..1 Indicates the type of product
fpmlProductType
generalTerms Defines dates and reference
1..1
fpmlGeneralTerms obligation of the CDS.
feeLeg
1..1 Premium payment definition.
fpmlFeeLeg
This is where the credit events and
protectionTerms obligations that are applicable to
1..1
fpmlProtectionTerms the credit default swap trade are
specified.
cashSettlementTerms
0..1 Cash settlement details.
fpmlCashSettlementTerms
physicalSettlementTerms
0..1 Physical settlement details.
fpmlPhysicalSettlementTerms

Note the fee leg contains the stub position implicitly (as per fpml spec). There
are 5 stub position types: none, ShortFirst, LongFirst, ShortFinal, LongFinal.
ShortFirst and LongFirst require a firstPaymentDate.
ShortFinal and LongFinal require lastRegularPaymentDate.
A sample representation is as follows:

<creditDefaultSwap>
<generalTerms>
<effectiveDate>

©2009 Razor Risk Technologies Page 329 of 373

Razor Financial Principals

<unadjustedDate>2006-10-18</unadjustedDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
</effectiveDate>
<scheduledTerminationDate>
<adjustableDate>
<unadjustedDate>2007-01-18</unadjustedDate>
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
</adjustableDate>
</scheduledTerminationDate>
<sellerPartyReference href="ECHIDNA" />
<buyerPartyReference href="PLATYPUS" />
<dateAdjustments>
<businessDayConvention>NONE</businessDayConvention>
</dateAdjustments>
<referenceInformation>
<referenceObligation>
<primaryObligor id="KANGAROO" />
</referenceObligation>
</referenceInformation>
</generalTerms>
<feeLeg>
<periodicPayment>
<paymentFrequency>
<period>M</period>
<periodMultiplier>3</periodMultiplier>
</paymentFrequency>
<rollConvention>NONE</rollConvention>
<fixedAmount>
<currency>EUR</currency>
<amount>0</amount>
</fixedAmount>
<fixedAmountCalculation>
<dayCountFraction>ACT/360</dayCountFraction>
<calculationAmount>
<currency>EUR</currency>
<amount>0</amount>
</calculationAmount>
<fixedRate>0.017200</fixedRate>
</fixedAmountCalculation>
</periodicPayment>
</feeLeg>
<protectionTerms>
<calculationAmount>
<currency>EUR</currency>
<amount>10000000.00</amount>
</calculationAmount>
</protectionTerms>
</creditDefaultSwap>

©2009 Razor Risk Technologies Page 330 of 373

Razor Financial Principals

16.1.3 Credit Implications

Pricing Contexts
There are two or more contexts that a credit default swap has when pricing.
There is an exposure to the issuer of the Credit Default Swap contract, and
there are exposures to the counterparties that the swap is providing the cover
against.

Timing of payment on default

Counterparties typically wait up to three months after the default event has
occurred in order to give the price of the reference security time to settle at a
new level.

Physical delivery
Often there will be physical delivery of the reference security upon default,
and there could be liquidity risk that the writer of the credit default swap
incurs upon trying to purchase the reference security in order to satisfy this
physical delivery requirement.

16.1.4 CDS Pricing

A credit default swap (CDS) is a financial contract in which the buyer of credit
protection pays a periodic premium in return for a promise from the credit
protection seller to compensate the default losses on some agreed upon third
party reference credit obligation.
CDS’s are by far the most important instruments enabling credit risk transfer.
Although a CDS is now viewed as a relatively simple instrument, it continues to
play the essential role as the key building block in a wide range of more
complex credit structures.

The cash flow profile of a single name credit defaults swap consists of two
legs:
Premium Leg: the buyer of protection pays periodic payments to the protection
seller until the earlier of a credit event or maturity of the CDS contract.
Protection Leg: the seller of protection pays the difference between par and
the recovery value of the referenced portfolio exposure value should a credit
event occur during the contract.

The value of a CDS protection seller is given as the risk-neutral expected value
of the present value of premium leg less the protection leg.
At the inception of a CDS, the premium is set such that the value of the
Premium Leg must equal the Protection Leg. i.e., on-market CDS must have
zero net present value.

Mathematically, the value of two legs can be expressed as:

The risk-neutral expected value of present value of the Premium Leg:

©2009 Razor Risk Technologies Page 331 of 373

Razor Financial Principals

[{ } ]
M
Premium_Le g = ∑ d Ti p (T0 , Ti ) τ i S{N (Ti −1 )α (Ti −1 , Ti )}
i =1
(1)
Where:
T0 - is the date of CDS valuation,
T0 < T1 < … < TM - in arrears premium payment dates,
τi - ith payment period as a fraction of years,
S - per-annum spread for this CDS,
p(T0,Ti) - is the risk-neutral survival probability of the reference
entity from time T0 to Ti,
by convention p(T0,Ti) = p(Ti) if T0 is the valuation date.
N(Ti) - Notional value of the CDS at time Ti,
α(Ti-1,Ti) - is the accrual factor from Ti-1 to Ti,
dT - risk free discount factor from time T0 to T

The risk-neutral expected value of present value of the Protection Leg:

[ ]
D
Protection_Leg = ∑ d ti {( p (T0 , t i −1 ) − p (T0 , t i )}Loss (VR(t i ))
i =1
(2)
Where:
T0 < t1 < ..... < t D - credit event dates and their modeled periodicity can be
different from the premium payment dates.
Loss (VR(t i )) - the amount of loss occurring on reference credit value at
time t i .

The value of CDS for protection seller is then the difference between a
premium (fee) leg and protection (loss) leg:

ValueCDS (T0 ) = Premium_Leg - Protection_Leg . (3)

The price of a credit default swap will reflect several factors. The key inputs
would include the following:
The risk-neutral probability of default of the reference asset or alternatively
its survival probabilities p(T0,Ti) should be provided as an input term structure.
In practice these probabilities are derived from the CDS premium spread values
S(Ti ) - constant or time dependent and known for existing CDS contracts.
Risk free discount rates determined from the appropriate curve.

©2009 Razor Risk Technologies Page 332 of 373

Razor Financial Principals

Loss (VR(t i )) - loss of an underlying referenced asset defined as some

deterministic function that is usually modeled as:
Loss (VR(t i )) = (1 − R (t i )) N (t i )
where R(ti ) - modeled recovery rate at time t i .
Recovery rate must be assumed or estimated separately as the (ideal) joint
calibration on default rates and recovery. In Razor it is currently assumed to
be constant.

©2009 Razor Risk Technologies Page 333 of 373

Razor Financial Principals

Calibration.

In the pricing of credit derivatives it is assumed that default events follow a

Poisson distribution and hence the survival probability can be expressed as:
T
− ∫t 0 λ ( s ) ds
SP ( t 0 , T ) = e
Where λ (t ) is called the default intensity, or the instantaneous forward
default rate, or hazard rate.
The CDS instrument prices this default risk. Hence, from the observed CDS
spreads in the market we can derive the risk-neutral survival probabilities.
There are two approaches to tackling this problem, the parametric
bootstrapping and piece wise constant bootstrapping. Both methods are
described below. Currently Razor implements the parametric bootstrapping for
CDS curve calibration.

Parametric Bootstrapping

It is assumed there exists a functional form of the survival probability curve.

Inputting observed CDS market prices results in a system of linear equations.
This can be solved to determine the parameters of the original functional form
of the survival probability curve.
This approach has the advantage of controlling the possible resulting curve
shapes combined with its smoothing features.
Razor employs the following functional form:
N αj
S (t ) = ∑ (4)
(t − t j ) 2 + ξ j
2
j =1

where
t1 , t 2 ,..., t N - tenors of observed CDS spreads quotes,
N – number of quotes,
ξ1 ,..., ς N - weights that are used to control a degree of confidence in
the accuracy to target CDS spread observations.

The direct substitution of (4) into equations (1)-(3) leads to a system of linear
algebraic equations that can be easily solved against unknown parameters
α 1 ,..., α N using standard methods of linear algebra.
Such a representation allows for the exact fitting of the term structure to the
underlying CDS spread quotes, delivers a smooth solution even for a limited
number of market observations and allows for the avoidance of kinks especially
at the short and long ends of credit curve where market observations are often
unavailable.

©2009 Razor Risk Technologies Page 334 of 373

Razor Financial Principals

As an example, the following survival probability curve is built from three CDS
spreads of 3, 5 and 7 years maturity:
S 3 = 34.76,
S 5 = 57.27,
S 7 = 82.42

Piece-Wise Constant Hazard Rate Bootstrapping

The hazard rate is assumed piece-wise constant and starting from the shortest
maturity is ‘bootstrapped’ similarly to the bootstrapping procedure for the
interest rate term structure.
The iterative algorithm can be described by the following procedure:
- Assume a constant hazard rates λ (t ) for all time intervals t i − t i −1 with
t i , i = 1,..., M
representing intermediate payment dates and maturity dates of all CDS quotes.
Assume that we successfully determined values of λ1 ,..., λ k , k < M hazard
rates.
This also guarantees the knowledge of all survival probabilities
S (t 0 , t i ), i = 0,1,..., k .
Note that:
Bi +1
S (t0 , ti +1 ) = ,
Ai +1
with
©2009 Razor Risk Technologies Page 335 of 373
Razor Financial Principals

Ak + 1 = S k +1τ k +1 d k +1 p ( t k ) + (1 − R ) d k +1 p ( t k ) ,

k 
Bk+1 =−Sk+1τk+1∑dti p(ti ) + (1− R)  ∑dti ( p(ti−1) − p(ti )) + dtk p( tk )  ,
k

i=1  i=1 
where
τ k +1 - fee payment period of the k+1 CDS spread,

dt - risk free discount factor up to t,

p(t) - survival probabilities up to time t,
S k +1 - k+1 (st) CDS spread value taken either from the available
market quotes or linearly interpolated from them
R - constant recovery rate.

Then λk +1 hazard value can be found as the solution of the following equation:

S ( t 0 , t k +1 )
λ k +1 = − ln( ) / τ k +1 .
S (t0 , tk )

The resulting hazard curve can have the following step ladder form:

©2009 Razor Risk Technologies Page 336 of 373

Razor Financial Principals

16.1.5 Configuration of CDS Spread Curve

The user has the option of specifying a CDS Spread Curve as containing CDS
spreads, or to specify the probabilities directly as either survival or default
rates. In the case of survival or default rates no market calibration is
performed. A CDS curve must contain only one CDS Quote Type within its
structure. A CDS curve has a recovery rate associated with it.
The bootstrapping method for a CDS curve is derived from the QUOTE TYPE of
the “CdsSpread“ asset as follows:

SURVIVAL_RATE - actual survival rates are taken directly

DEFAULT_RATE - survival rate = 1 – default rate
CDS_SPREAD - standard market CDS spread quotes used to compute
survival rates

For example:

<asset id="DefaultProbability" type="cdsSpread">

<cdsSpread>
<cdsQuoteType>DEFAULT_RATE</cdsQuoteType>
<frequency>
<months>3</months>
</frequency>
</cdsSpread>
</asset>

and

<asset id="CdsSpread" type="cdsSpread">

<cdsSpread>
<cdsQuoteType>CDS_SPREAD</cdsQuoteType>
<frequency>
<months>3</months>
</frequency>
</cdsSpread>
</asset>

16.1.6 Pricing Parameters

To price a CDS the following pricing parameters must be set.

Name: Type Description

The family of credit curves to use. Typically it would be
CreditCurveName
the industry of the name. This is optional.
The rating of the name. This effective adds another
CreditRating*
specifier the credit curve name. This is optional.

* If this parameter is not present it will attempt to obtain the credit rating
from the trade value:
©2009 Razor Risk Technologies Page 337 of 373
Razor Financial Principals

creditDefaultSwap/generalTerms/referenceInformation/referenceObligationAt[
0]/seniority
The CDS curve selected from the market can depend on the name and its
rating. The curve name will be built according to the base CDS type,
CreditCurveName, and CreditRating. If either parameter is omitted it does not
impact the requested curve.
"CdsSpread [CreditCurveName] [CreditRating]"
Examples:
"CdsSpread FINANCE A+"
"CdsSpread MINING EUR"
"CdsSpread MANUFACTORING"
"CdsSpread"

16.2 Collateralised Debt Obligation

16.2.1 Description of Instrument
A Collateralized Debt Obligation is a securitization product backed by an
underlying portfolio of credit products called the collateral consisting of
instruments such as loans or CDS. The resulting cashflows from this collateral is
repackaged into a number of tranches providing a variety of risk return profiles
to investors.

16.2.2 XML Representation

Collateralized Debt Obligations are not covered by FPML version 4 and form a
razor extension.
The schema for rzmlCollateralizedDebtObligation is defined as follows:

fpmlCollateralisedDebtObligation Schema
Name: Type Occurs Size Description
productType 0..1 Indicates the type of product
fpmlProductType
tranche 1..1 Tranche being priced
rzmlTranche
paymentFrequency 1..1 Frequency of premium payment
fpmlCalculationPeriodFrequency
effectiveDate 1..1 CDO start date.
xsd:date
maturityDate 1..1 CDO termination date.
xsd:date
referencePool 1..1 Reference pool of collateral
rzmlReferencePool

A sample representation is as follows:

<collateralizedDebtObligation>
<effectiveDate>2006-02-01</effectiveDate>

©2009 Razor Risk Technologies Page 338 of 373

Razor Financial Principals

<maturityDate>2011-02-01</maturityDate>
<tranche>
<attachment>0.05</attachment>
<detachment>0.1</detachment>
<spreadRate>0.005</spreadRate>
<notional>
<amount>1000000.0</amount>
<currency>EUR</currency>
</notional>
</tranche>
<paymentFrequency>
<periodMultiplier>M</periodMultiplier>
<period>3</period>
<rollConvention>NONE</rollConvention>
</paymentFrequency>
<referencePool>
<collateral>
<creditDefaultSwap>
<generalTerms>
<effectiveDate>
<unadjustedDate>2006-10-18</unadjustedDate>
<dateAdjustments>
<businessDayConvention>
NONE
</businessDayConvention>
</dateAdjustments>
</effectiveDate>
<scheduledTerminationDate>
<adjustableDate>
<unadjustedDate>2007-01-18</unadjustedDate>
<dateAdjustments>
<businessDayConvention>
NONE
</businessDayConvention>
</dateAdjustments>
</adjustableDate>
</scheduledTerminationDate>
<sellerPartyReference href="ECHIDNA" />
<buyerPartyReference href="PLATYPUS" />
<dateAdjustments>
<businessDayConvention>
NONE
</businessDayConvention>
</dateAdjustments>
<referenceInformation>
<referenceObligation>
<primaryObligor id="KANGAROO" />
</referenceObligation>
</referenceInformation>
</generalTerms>
<feeLeg>
<periodicPayment>
<paymentFrequency>
<period>M</period>

©2009 Razor Risk Technologies Page 339 of 373

Razor Financial Principals

<periodMultiplier>3</periodMultiplier>
</paymentFrequency>
<rollConvention>NONE</rollConvention>
<fixedAmount>
<currency>EUR</currency>
<amount>0</amount>
</fixedAmount>
<fixedAmountCalculation>
<dayCountFraction>
ACT/360
</dayCountFraction>
<calculationAmount>
<currency>EUR</currency>
<amount>0</amount>
</calculationAmount>
<fixedRate>0.017200</fixedRate>
</fixedAmountCalculation>
</periodicPayment>
</feeLeg>
<protectionTerms>
<calculationAmount>
<currency>EUR</currency>
<amount>10000000.00</amount>
</calculationAmount>
</protectionTerms>
</creditDefaultSwap>
</collateral>
</referencePool>
</collateralizedDebtObligation>

16.2.3 Credit Implications

CDO are structured using a Special Purpose Vehicle (SPV) removing any credit
exposure to the organizing entity. The CDO has a reference pool of assets with
an income stream and exposure to potential losses due to credit events. The
CDO defines tranches with attachment and detachment points [a, b] which
reimburse credit event losses from a% to b% of the notional value of the CDO.
The losses of a tranche of a lower attachment point must be fully absorbed
before losses are allocated to the next tranche.
In return for this protection each tranche receives a premium on the non-
defaulted notional amount it covers. This rate is higher for tranches with a
lower attachment point due to the increased credit risk exposure.
The method for the allocation of premium and losses is fully defined in the
waterfall documentation of the CDO.

©2009 Razor Risk Technologies Page 340 of 373

Razor Financial Principals

SPV

Tranches
Senior
Collateral Payments Subordinated
Reference Mezzanine
Pool ….
Allocation of Equity
Losses defined
via waterfall

Originally CDO's were used to securitize a bank's portfolio of loans removing

them from the balance sheet. All income such as interest payments are
forwarded on to the holders of the tranches in the CDO. Any losses from
defaults are allocated to the tranches defined by the waterfall structure.
The loan backed CDO form is now a small market which is now dominated by
the Synthetic CDO which is a CDO whose collateral pool consists entirely of
CDS. Razor CDO pricing assumes a Synthetic and static collateral pool, where
the reference pool may not vary throughout the life of the CDO.

16.2.4 CDO Pricing

In this pricing framework, it is taken that the risk neutral probabilities can be
bootstrapped from the CDS spreads (see CDS product). However in CDO pricing
we consider the time to default τ and whether it is before a given time.
P(τ≤t) - unconditional risk neutral probability the time to default is ≤
t.

The following definitions are required.

t1, … ,tn = T - premium dates
n - number of premium dates
t0 - valuation date
tn =T - maturity
si - effective premium rate paid over period (ti-1, ti]
d1, … ,dn = T - discount factor corresponding to premium date
K - number of names in collateral pool
(k)
N - recovery adjusted notional for name k (1 ≤ k ≤ K)
J - number of tranches
l - attachment point of tranche being valued

©2009 Razor Risk Technologies Page 341 of 373

Razor Financial Principals

u - detachment point of tranche being valued

S = (u – l) - size of tranche being valued (assuming a notional of 1)
Li - cumulative losses of the tranche being valued up to time ti.
Λi - cumulative losses of the entire reference pool up to time ti.

We can express the losses allocated to a tranche at time ti as:

Li = min(S, max(Λi-l, 0)
Therefore the tranche losses is a function of the pool losses, we define this
function to be Ψ(Λ).
Ψ(x) = min(S, max(x-l, 0)

To price a CDO tranche, the risk neutral expected value of the discounted
premiums, must equal to the risk neutral expected value of the discounted
losses. This can be expressed as:

 n   n 
E Q ∑ si (ti − ti −1 )( S − Li ) d i  = E Q ∑ ( Li − Li −1 )d i 
 i =1   i =1 
Hence during the life of the CDO when the premium has been set the value for
a tranche buyer is:
n n
Vbuy = ∑ si (ti − ti −1 )( S − E Q [ Li ]) d i − ∑ ( E Q [ Li ] − E Q [ Li −1 ])d i
i =1 i =1

It remains to determine the expected accumulated losses E[Li]. The default

process will be specified using a single factor Gaussian copula model to
describe the correlation structure, with the credit events simulated
independently.
Denote unconditional risk neutral probability the time to default of name k:
πˆ i( k ) = P (τ ( k ) ≤ t i )
Then conditional on the single credit driver x, the risk-neutral distribution of
default time is given by:

πˆi( k ) = Φ
( )
 Φ −1 πˆ i( k ) − β ( k ) x 

 σ (k ) 
Where:
∞
∫ πˆi( k ) dΦ ( x) = πˆi( k )
−∞

Φ is the standard normal cumulative distribution function and Φ-1 is its inverse.
Define: π i( k ) ( x ) = 1 − πˆ i( k ) ( x )
Assuming the names are conditionally independent the mean tranche loss is:

E[Li ] = ∫ Ex [Li ]d Φ( x)
∞

−∞

Where Ex is the expectation conditional on X = x.

©2009 Razor Risk Technologies Page 342 of 373
Razor Financial Principals

Now dropping the i notation:

E[L ] = E[ψ (Λ)]
 
E[L ] = ∑ [ y − l ]P(Λ = 1) + s 1 − ∑ P(Λ = 1)
l < y <u  y<u 

Homogenous Model
Assuming a homogenous pool the number of defaults in (t0, ti] is binomially
distributed.
Furthermore, assuming a large pool size, the binomial distribution can be
approximated by a Poisson distribution.
K
Now defining λˆ = ∑ πˆ
k =1
(k )

Leads to the result:

( ˆ
)
E [L ] = S 1 − e − λ − e − λ  S ∑
ˆ λˆm
+ ∑
λˆm
[ 
u − mN (1) ]
 1< m < l / N (1) m! l / N (1) < m < u / N (1) m! 

Heterogenous Model
Extending this to the general case requires the following definitions:

N * = min N ( k )
k

N = max N ( k )
*
k

πˆ ( k )
f (N ) = ∑
=N λ
k:N ( k )
ˆ
Where:

N* ≤ N ≤ N *
f *m is the m-fold convolution of f with itself.
Then we arrive to the final result:

( ˆ ˆ
)
E [L ] = S 1 − e − λ − e − λ  S ∑
λˆm
∑ f *m
( N ) + ∑
λˆm
∑

[u − N ] f * m ( N ) 
 1< m < l / N * m! mN * < N < l 1< m < u / N * m! l < N < u 

References
The pricing model was derived from the following sources:
J Hull and A White, 2003
Valuation of a CDO and an nth to default CDS without Monte Carlo
simulation
Chaplin, 2005
Credit Derivatives

©2009 Razor Risk Technologies Page 343 of 373

Razor Financial Principals

I Iscoe, A Krenin and B De Prisco, 2005

Loss in translation
Risk June 2005

16.2.5 Pricing Parameters

The CDO pricing adapter uses the following pricing parameters.

Name: Type Description

Base correlation at the tranche attachment point.
AttachmentCorrelation
Default 0.
Base correlation at the tranche detachment point.
DetachmentCorrelation
Default 0.
HomogenousPool – use the homogenous model.
CdoModel
Otherwise (default) – use the heterogenous model.

The Homogenous model is faster but is restricted to cases where the reference
pool is homogenous in nature. The heterogenous model whilst being more
general is also slower to process.

©2009 Razor Risk Technologies Page 344 of 373

Razor Financial Principals

Chapter 17
Pricing

17.1 Distribution of Forward Rates

This section is a fundamental building block for pricing various interest
rate derivatives in the following sections. We shall discuss on the
dynamics of two types of forward rates, i.e. LIBOR rate and CMS rate, in
this section and these are important for the pricing of interest rate
derivative contracts in later sections.

17.1.1 Definition
U ( t , Ti )
= any forward floating rate for the tenor period (Ti −1 , Ti ) implied by
the structure at time t .
σ ( t , Ti )
U
= volatility function of the forward floating rate U ( t , Ti ) .
EQ = expectation under risk-neutral measure for currency e .
QTi
E
= expectation under forward martingale measure for currency e with
respect to time Ti .
F= { Fu }0≤t ≤T
= filtration generated by the Wiener process.
B (t ) = savings account at time t .
B (t, T )
= zero-coupon bond price at time t maturing at time T .
σ (t, T ) = zero-coupon bond volatility function.

r (t ) = short rate at time t .

Q
Wt
= Wiener process under risk-neutral measure at time t .
QTi
Wt
= Wiener process for currency e under forward martingale measure
with respect to time T at time t .
N = notional amount.
si = spread of tenor period Ti −1 to Ti .
Ti = i th tenor date.
t = the valuation date.

©2009 Razor Risk Technologies Page 345 of 373

Razor Financial Principals

17.1.2 Pricing
For any interest rate derivative, we need to calculate the present value
of the cashflow U (Ti −1 , Ti )(Ti − Ti −1 ) for the floating rate leg.
The expected present value of the cashflow is:
 B (t ) 
U (Ti −1 , Ti )(Ti − Ti −1 ) | Ft  = (Ti − Ti −1 ) B ( t , Ti ) E Q U (Ti −1 , Ti ) | Ft  .
Ti
E Q0 
 B (Ti ) 
Thus if we know E Q U (Ti −1 , Ti ) | Ft  , we can find out the expected
Ti

present value of the required cash flow for the floating rate leg of the
swap. The computation of the expectation E Q U (Ti −1 , Ti ) | Ft  depends on
Ti

whether the underlying forward rate is a LIBOR or CMS rate.

17.1.3 Forward LIBOR Rate

The distribution of forward LIBOR rate is:
dU ( t , Ti ) = σ U ( t , Ti )U ( t , Ti ) dWt Q .
Ti

Therefore:
E Q0 U (Ti −1 , Ti ) | Ft  = U ( t , Ti ) .
Ti

17.1.4 Forward CMS Rate

For CMS rate, we need convexity adjustment. Thus the expectation of
CMS rate is:
σ U ( t , Ti ) f '' (U ( t , Ti ) ) 2
2

E Q U (Ti −1 , Ti ) | Ft  ≈ U ( t , Ti ) − U ( t , Ti )(Ti − t ) .

Ti

2 ( (
f ' U t,T i ))
Note that f ( x ) is the price-to-yield function and can be defined as:
n
Ti − Ti −1 1
f ( x ) = x0 ∑ + .
(1 + x ) (1 + x )
Ti −T0 Tn −T0
i =1

The above convexity adjustment is exactly the same as the one in section
12.10.3.

17.2 Greeks Computation

17.2.1 Introduction
This section shows the Greeks of a generalised Black-Scholes option
pricing formula.

17.2.2 Definition
c = price of call option.
p = price of put option.
S = underlying asset price.

©2009 Razor Risk Technologies Page 346 of 373

Razor Financial Principals

K = strike price.
b = cost of carry.
r = risk free rate.
rf = foreign risk free rate.
q = continuous dividend yield.
T = time to maturity.
σ = price volatility.
n ( ⋅) = standard normal probability density function.
N ( ⋅) = cumulative standard normal distribution function.

17.2.3 Generalised Black-Scholes Formula

The call option price is:
c = Se( b − r )T N ( d1 ) − Ke − rT N ( d 2 ) .
The put option price is:
p = Ke − rT N ( −d 2 ) − Se(b − r )T N ( − d1 ) .
We define:
S  σ2 
ln   +  b + T
K  2 
d1 = and
σ T

S  σ 
2
ln   +  b − T
K  2 
d2 = .
σ T

The generalised Black-Scholes formula can be summarised as:

b=r gives the Black-Scholes (1973) stock option model.
b = r−q gives the Merton (1973) stock option model with q .
b=0 gives the Black (1976) futures option model.
b = r − rf gives the Garman and Kohlhagen (1983) currency option
model.

17.2.4 Greeks
Delta
∂c
∆ call = = e( b − r )T N ( d1 ) .
∂S
∂p
∆ put = = −e( b − r )T N ( −d1 ) .
∂S
Gamma
∂ 2 c n ( d1 ) e
(b − r )T
Γ call = 2 = .
∂S Sσ T

©2009 Razor Risk Technologies Page 347 of 373

Razor Financial Principals

∂ 2 p n ( d1 ) e
(b − r )T
Γ put = = .
∂S 2 Sσ T
Rho
∂c
ρcall = = TKe − rT N ( d 2 ) .
∂r
∂p
ρ put = = −TKe − rT N ( −d 2 ) .
∂r
For option on foreign exchange options, the ρ f on foreign risk free rate
is:
∂c
= −TSe(b − r )T N ( d1 ) = −TSe f N ( d1 ) .
−r T
ρcall
f
=
∂rf
∂p
= TSe( ) N ( − d1 ) = TSe f N ( −d1 ) .
b−r T −r T
ρ put
f
=
∂rf
For option on futures ( b = 0 ), we have:
∂c
ρcall = = −Tc .
∂r
∂p
ρ put = = −Tp .
∂r
Vega
∂c
vegacall = = Se( b − r )T n ( d1 ) T .
∂σ
∂p
vega put = = Se( b− r )T n ( d1 ) T .
∂σ
Theta
∂c Se(b − r )T n ( d1 ) σ
θ call = − =− − ( b − r ) Se( b − r )T N ( d1 ) − rKe− rT N ( d 2 ) .
∂T 2 T
∂p Se ( b − r )T
n ( d1 ) σ
θ put = − =− + ( b − r ) Se( b − r )T N ( − d1 ) + rKe − rT N ( − d 2 ) .
∂T 2 T

17.3 Interest Rate Risk Analytics

Introduction

The following interest rate risk analytics are defined on any generalized
series of cashflows.

The following definitions and derivations will default to use an annual

compounding yield to maturity rate ra . This is to be consistent with the
Razor approach which uses an annual compounding yield to maturity for
reporting.

Where appropriate the calculations of duration and convexity will use the
compounding frequency and coupon periods per year.

©2009 Razor Risk Technologies Page 348 of 373

Razor Financial Principals

Sometimes these formulae below are expressed from a bond pricing

formulae starting point using k + f rather than nti . In the end the
d
formulae are always consistent and equivalent by substitution, as
1
t= k+ f
n (
d . )
Definitions

ti - time in years for i ’th cashflow, calculated using security days

basis or day count convention

dfi - discount factor at time ti from bootstrapped yield curve

- discount factor at time ti derived from yield to maturity rate r of

df r ,i
specified yield basis

Annual Compounding Rate ra

1
va =
1 + ra , for all i
df a ,i = (1 + ra ) − ti = ( va ) i
t

N-Period Compounding Rate rn

1
vn =
r
1+ n
n , for all i
r
df n ,i = (1 + n ) − nti = ( vn ) i
nt

Simple Annualised Rate rs

1
vs ,i =
1 + rs ti
df s ,i = (1 + rs ti )−1 = vs ,i

Continuous Compounding Rate rc

1
vc = rc
e
df c ,i = e − rc ti = ( vc ) i
t

Annualised Discount Rate rd

vd ,i = (1 − rd ti )

©2009 Razor Risk Technologies Page 349 of 373

Razor Financial Principals

df d ,i = vd ,i

17.3.1 Valuation for Interest Rate Securities

Market Value

P = ∑ dfi Ci ≡ ∑ df r ,i Ci
i i

Accrued Interest
f
AI i = Ci - cum interest
d
f 
AI i = Ci  − 1 - ex interest
d 

where

d = d e − d s - number of days of accrued interest, with respect to days

basis
f = d e − d v - number of days in coupon period, with respect to days basis
ds - date of previous coupon
de - date of next coupon
dv - value date

Clean Price
CP - clean price

P = CP + AI

Per $100 Face Value Equivalents

P100 FV = P *100 / FV
CP100 FV = CP *100 / FV
AI100 FV = AI *100 / FV

Ex Interest Date Calculation

If an interest rate security is defined to support ex-interest days
(calendar or business days), then the calculation to determine if a coupon
is ex-interest is as follows.

Calendar Days
Ex-Interest if (Coupon Date – Calendar Days) >= Value Date, otherwise Cum-Interest

Business Days
Ex-Interest if (Coupon Date – Elapsed Business Days) >= Value Date, otherwise Cum-Interest

©2009 Razor Risk Technologies Page 350 of 373

Razor Financial Principals

17.3.2 Security Interest Rate Risk Analytics

Annual Compounding Rate ra

Price

P = ∑ (1 + ra ) − t i Ci
i

Macaulay Duration

∑ (1 + r ) a
−ti
Ci t i
D= i
P , or

∑ df C t i i i
D= i
P - Razor uses this form

Modified Duration

dP 1
D' = −
dra P

dP
= −∑ (1 + ra ) − ti −1 Ci ti
dra i

= −(1 + ra ) −1 ∑ (1 + ra ) −ti Ci ti
i

= −va ∑ (va )ti Citi

i

= −va ∑ dfi Ci ti
i

D
D' = = va D
(1 + ra ) - Razor uses this form

Convexity

d 2P 1
CX =
dra2 P

©2009 Razor Risk Technologies Page 351 of 373

Razor Financial Principals

d 2P d  
2
=  −∑ (1 + ra ) − ti −1 Ci ti 
dra dra  i 
= ∑ (1 + ra ) − ti − 2 Citi2 + (1 + ra ) − ti − 2 Ci ti 
i

= (1 + ra ) −2 ∑ (1 + ra ) − ti Ci ti2 + (1 + ra ) − ti Ci ti 
i

= va2 ∑  dfi Citi2 + dfi Ci ti 

i

 
= va2 ∑ dfi Ci ti2 + ∑ dfi Ci ti 
 i i 

[
CX = va D '' + D ' , where '' ]
D = va ∑ dfiCiti2
1
i P

N-Period Compounding Rate rn

For example, as for bonds

Price

rn − nt i
P = ∑ (1 + ) Ci
i n

Macaulay Duration

rn
∑ (1 + n ) − nt i
Citi
D= i
P , or
∑ dfiCiti
D= i

P - Razor uses this form

Modified Duration

dP 1
D' = −
drn P

©2009 Razor Risk Technologies Page 352 of 373

Razor Financial Principals

dP r
= −∑ (1 + n ) − nti −1 Citi
drn i n
rn −1 r
= −(1 + ) ∑ (1 + n ) − nti Ci ti
n i n
= −vn ∑ (vn ) nti Ci ti
i

= −vn ∑ dfi Ci ti
i

D
D' = = vn D
r
(1 + n )
n - Razor uses this form

Convexity

d 2P 1
CX =
drn2 P

d 2P d  rn − nti −1 
=
drn2 drn  −∑ (1 + n ) Ci ti 
 i 
 r 1 r 
= ∑  (1 + n ) − nti − 2 Citi2 + (1 + n ) − nti − 2 Ci ti 
i  n n n 
rn −2  r 1 r 
= (1 + ) ∑ (1 + n ) − nti Ci ti2 + (1 + n ) − nti Citi 
n i  n n n 
 1 
= vn2 ∑  dfi Ci ti2 + df iCi ti 
i  n 
 1 
= vn2  ∑ df iCi ti2 + ∑ dfi Ci ti 
 i n i 

 1  1
CX = vn  D '' + D '  D '' = vn ∑ dfi Ci ti2
 n  , where i P

Simple Annualised Rate rs

This is equivalent to converting to an effective annual compounding rate,

and then applying the annual compounding forms of these analytics. For
example as used for pricing a Bank Accepted Bill.

Price

P = ∑ (1 + rs ti ) Ci
−1

©2009 Razor Risk Technologies Page 353 of 373

Razor Financial Principals

Macaulay Duration

∑ (1 + r t )
−1
s i Ci ti
D= i
P , or

∑ df C t i i i
D= i
P - Razor uses this form

For a BAB or similar single cashflow based instrument, Duration simplifies

to term to maturity, ie …

D = ti

Modified Duration

dP 1
D' = −
drs P

dP
= −∑ (1 + rs ti ) −2 Ci ti
drs i

= −∑ (vs ,i ) 2 Ci ti
i

= −∑ ( dfi ) Ci ti
2

∑ ( df )
2
i Ci ti
D' = i
P

For a BAB or similar single cashflow based instrument, Modified Duration

simplifies to …

D' = vs ,i ti = dfi ti = vs ,i D

Convexity

d 2P 1
CX =
drs2 P

©2009 Razor Risk Technologies Page 354 of 373

Razor Financial Principals

d 2P d  
2
=  −∑ (1 + rs ti ) −2 Ci ti 
drs drs  i 
= 2∑ (1 + rs ti ) −3 Ci ti2 
i

= 2∑ ( vs ,i ) Ci ti2
3

2∑ ( vs ,i ) Ci ti2
3

CX = i
P

For a BAB or similar single cashflow based instrument, Convexity

simplifies to …

2 ( vs ,i ) Ci ti2
3

( )
= 2 ( vs ,i ) ti2 = 2 D '
2 2
CX =
(v ) Cs ,i i

Continuous Compounding Rate rc

This is equivalent to effective annual compounding rate.

Annualised Discount Rate rd

Discount rates are not currently supported by Razor.

Price

P = ∑ (1 − rd ti ) Ci
i

Macaulay Duration

∑ (1 − r t ) C t d i i i
D= i

P , or
∑ dfiCiti
= i

For a single cashflow based discount instrument, Duration simplifies to

term to maturity, ie …

D = ti

Modified Duration

©2009 Razor Risk Technologies Page 355 of 373

Razor Financial Principals

dP 1
D' = −
drd P

dP
= −∑ Ci ti
drd i

∑C t i i
D =
' i
P

Convexity

d 2P 1
CX =
drs2 P

d 2P d  
= −∑ Ci ti  = 0
drs2 drs  i 

Duration of Futures, FRA, Forward Bond and Interest Rate Swap

In this section, we discuss on the methods Razor uses to compute the
Macaulay and Modified duration of futures, FRA, forward bond and
interest rate swaps.

Duration of Futures

Cash Futures
The Macaulay duration of cash futures is the length of the time in years
from the valuation date to the futures maturity date using ACT/365 day
count basis. Modified duration is calculated from scaling Macaulay by the
yield curve forward rate.

Tm − t Dmac
Dmac = Dmod =
365 (1 + ytm)

Bond Futures
The Macaulay duration of a bond in future in Razor is the Macaulay
duration of the underlying bond.

Dfi CFi ti Dmac

Dmac = ∑ Dmod =
i P (1 + ytm)

©2009 Razor Risk Technologies Page 356 of 373

Razor Financial Principals

Duration of a FRA
The duration of a FRA is the weighted average of the duration of incoming
leg and duration of the outgoing leg weighted by the present value of
each of the leg. The duration of the incoming and outgoing legs is the
length of time period from the forward starting date to the forward
ending date.

∑ N ⋅ Df ⋅τ i i i
Dmac
Dmac = i
Dmod =
∑ PV i
i (1 + ytm)

Duration of Forward Bond

Define
t0 = valuation date.
Tb = bond starting date.
Te = bond maturity date.
P = price of the bond at bond starting date.

Currently Razor supports two methods to calculate forward bond

durations.
Method 1
The first method is simply to calculate the forward bond duration as for a
normal bond but to assume the valuation date is at Tb .

Method 2
The second method is again the same as the duration for a normal bond
but to assume the cashflow at the bond starting date, i.e. P , is part of
the bond cashflows (with the same + or – sign as other bond cashflows).

Duration of Interest Rate Swap

It is clear that if we hold a pay-fix interest rate swap and a fixed rate
bond with the swap tenor dates matching the bond coupon paying date,
then we obtain cashflows from the portfolio that replicate the cashflows
of a floating rate bond.

We have the relationship

Swap + Fixed Rate Bond = Floating Rate Bond.

Given the cashflows on both side of equations are exactly the same, we
can deduce
Swap Duration + Fixed Rate Bond Duration = Floating Rate Bond Duration.

It should be noted that duration of a floating rate bond is zero thus

Swap Duration = - Fixed Rate Bond Duration.

Duration of fixed rate bond can be computed using usual method.

©2009 Razor Risk Technologies Page 357 of 373
Razor Financial Principals

17.3.3 Portfolio Risk Analytics

The portfolio level approximation for risk analytics such as Macaulay
Duration, Modified Duration and Convexity is defined as the net sum of
trade level analytic weighted by relative trade market value. The weight
by function is configurable.

For any portfolio risk analytic RP , where Rti , and Vti

∑R ti *Vti
RP = i =1
n

∑V i =1
ti

Therefore, specifically

Portfolio Macaulay Duration

n

∑D ti *Vti
DP = i =1
n

∑V i =1
ti

Portfolio Modified Duration

n

∑D '
ti *Vti
DP' = i =1
n

∑V i =1
ti

Portfolio Convexity
n

∑ CX ti *Vti
CX P = i =1
n

∑V i =1
ti

17.4 Yield and Price Volatility

While either yield or price volatility may be used to calculate an interest rate
option price, the inputs to the pricing model will need to be adjusted
depending on whether yield or price volatility is used.
The implied volatility is the standard deviation of the expected forward price
distribution. If the current yield of a forward bond is 5% per annum and has an
implied yield volatility of 10%, then the volatility implies that two-thirds of the
changes in yield over the next year will be within a range of 4.5% amd 5.5% per
annum. If we can calculate the price per US$100 of a bond given the yield,
then we can also imply the range of price changes implied by these yields.

©2009 Razor Risk Technologies Page 358 of 373

Razor Financial Principals

An approximation can be used to calculate price volatility from yield volatility

using modified duration:
V price = V yield * Modified duration * Yield
Where modified duration refers to the underlying forward bond.

17.5 Day Count Conventions

From [Per Annex to the 2000 ISDA Definitions (June 2000 Version), Section
4.16. Day Count Fraction ]
Defines a scheme of values for specifiying how the number of days between
two dates is calculated for purposes of calculation of a fixed or floating
payment amount and the basis for how many days are assumed to be in a year.
The full set of conventions is (Invalid, ACT/360, ACT/365, 30/360, 30E/360,
ACT/ACT, ACT/365.25, ACT/365.FIXED, ACT/ACT.ISDA, ACT/ACT.ISMA,
ACT/ACT.AFB, ACT/365.END, 1/1, ACT/365.ACT).
Currently Invalid defaults to ACT/365.ACT .

30/360
This calculation assumes a year has 360 days comprised of 12 months with 30
days each. In the bond calculation, d is given by the number of months
between payment dates multiplied by 30 days. The coupon paid is always the
annual coupon rate divided by the payment frequency. Determining (d − f ) is
more complex as each part month needs to be treated, as if there are 30 days.
This type of interest convention is common in Europe and is often referred to
as the annual coupon basis.
As part of this calculation if the start day of month is the 31st, then it is reset
to the 30th, and if the end date is also the 31st then it is also reset to the 30th.

30E/360
As above for 30/360, except if either the start or end day of month is the 31st,
then it will be reset to the 30th.

Actual/Actual (ACT/ACT)
The interest accrual is based on the actual number of days elapsed since the
last interest payment date (d − f ) in an interest period divided by the actual
number of days between the last and next interest payment dates (d ) . The
coupon paid is always the annual coupon rate divided by the payment
frequency. This method is commonly used in English-speaking countries and
includes instruments such as US Treasury Bonds, A$ and NZ$ government issues
and UK Gilts.

ACT/ACT.ISMA
As for Actual/Actual.

Actual/360 (ACT/360)
This method is also referred to as the US Money Market basis because interest
is calculated in the same way as for US dollar money market instruments. The
interest accrual is based on the actual number of days elapsed since the last

©2009 Razor Risk Technologies Page 359 of 373

Razor Financial Principals

interest payment date (d − f ) in an interest period divided by an assumed 360

days a year (d ) . In this method, the coupon payment is not always an even
amount.

Actual/365 (ACT/365)
As for Actual/360 except d is 365 days per anum.

Actual/365.FIXED (ACT/365.FIXED)
As for Actual/365.

Actual/365.25 (ACT/365.25)
As for Actual/360 except d is 365.25 days per anum.

ACT/365.END
As for Actual/360 except d is 366 if end date is a leap year, else 365.

ACT/365.ACT
As for Actual/360 except d is 366 if start date is a leap year, else 365.

ACT/ACT.ISDA
If the number of whole years between the start and end dates is greater than
0, then the day count fraction is calculated as (start year fraction + end year
fraction + number of years – 1). The days in year used for each date will be 366
if a leap year, and otherwise 365.
If the number of whole years is 0, then days in year is by default 365.

ACT/ACT.AFB
If number of whole years between start and end dates is greater than 0, then
calculate the days in year according to whether the fractional part of the day
count fraction at start end passes over the 29th February.
If the number of whole years is 0, then days in year is by default 365.

BUS/252
Business days / 252. This method is only used for Brazilian Real (BRL).
Weekends and public holidays are excluded from the calculation. This requires
the use of a calendar. Razor will look for a BUS252 calendar if defined,
otherwise only weekends will be excluded.

1/1
This is currently not supported.

17.6 Roll Conventions

The convention for determining the sequence of calculation period end dates.
It is used in conjunction with a specified frequency and the regular period start
date of a calculation period, e.g. semi-annual IMM roll dates

©2009 Razor Risk Technologies Page 360 of 373

Razor Financial Principals

EOM, FRN, IMM, IMCAD, SFE, NONE, TBILL, [1, 2 … 29, 30], MON, TUE, WED,
THU, FRI, SAT, SUN
Currently FRN is not implemented, so defaults to NONE.

EOM - Rolls on month end dates irrespective of the length of the month and the
previous roll day.
FRN - Roll days are determined according to the FRN Convention or Eurodollar
Convention as described in ISDA 2000 definitions.
IMM - IMM Settlement Dates. The third Wednesday of the (delivery) month.
IMMCAD - The last trading day/expiration day of the Canadian Derivatives
Exchange (Bourse de Montreal Inc) Three-month Canadian Bankers' Acceptance
Futures (Ticker Symbol BAX). The second London banking day prior to the third
Wednesday of the contract month. If the determined day is a Bourse or bank
holiday in Montreal or Toronto, the last trading day shall be the previous bank
business day. Per Canadian Derivatives Exchange BAX contract specification.
SFE - Sydney Futures Exchange 90-Day Bank Accepted Bill Futures Settlement
Dates. The second Friday of the (delivery) month.
NONE - The roll convention is not required. For example, in the case of a daily
calculation frequency.
TBILL - 13-week and 26-week U.S. Treasury Bill Auction Dates. Each Monday
except for U.S. (New York) holidays when it will occur on a Tuesday.
[1, 2 … 29, 30] - Rolls on the specific day of the month.
[MON, TUE, WED, THU, FRI, SAT, SUN] - Rolls weekly on the specific day of the
week.

17.7 Business Day Conventions

FOLLOWING, MODFOLLOWING, PRECEDING, MODPRECEDING, NONE, FRN

FOLLOWING -The non-business date will be adjusted to the first following day
that is a business day.
MODFOLLOWING - The non-business date will be adjusted to the first following
day that is a business day unless that day falls in the next calendar month, in
which case that date will be the first preceding day that is a business day
PRECEDING - The non-business day will be adjusted to the first preceding day
that is a business day
MODPRECEDING - The non-business date will be adjusted to the first preceding
day that is a business day unless that day falls in the previous calendar month,
in which case that date will be the first following day that us a business day
FRN - Per 2000 ISDA Definitions, Section 4.11. FRN Convention; Eurodollar
Convention.
NONE -The date will not be adjusted if it falls on a day that is not a business
day.
©2009 Razor Risk Technologies Page 361 of 373
Razor Financial Principals

NotApplicable - The date adjustments conventions are defined elsewhere, so it

is not required to specify them here.

Currently FRN is not implemented, so defaults to NONE.

This enumeration is also referenced as 'Date Roll Convention'.

17.8 Interpolation Methods

17.8.1 Linear Interpolation
Linear interpolation calculates the value of a point lying on a straight line
between two other points.
The formula for linear interpolation is:
Ti − T1
ri = (r2 − r1 ) + r1
T2 − T1

17.8.2 Log-Linear Interpolation

( )
Ti −T1

 r  T2 −T1
ri =  2  r1
 r1 

17.8.3 Cubic Interpolation

(Ti − T2 )(Ti − T3 )(Ti − T4 )
ri = r1 +
(T1 − T2 )(T1 − T3 )(T1 − T4 )
(Ti − T1 )(Ti − T3 )(Ti − T4 )
r2 +
(T2 − T1 )(T2 − T3 )(T2 − T4 )
(Ti − T1 )(Ti − T2 )(Ti − T4 )
r3 +
(T3 − T1 )(T3 − T2 )(T3 − T4 )
(Ti − T1 )(Ti − T2 )(Ti − T3 )
r4
(T4 − T1 )(T4 − T2 )(T4 − T3 )

©2009 Razor Risk Technologies Page 362 of 373

Razor Financial Principals

17.9 Cumulative Normal Distribution Function

The standard numerical procedure to evaluate the cumulative normal
distribution function adopts a polynomial approximation algorithm3. This
method provides numerical accuracy to six-decimal-place. Unfortunately,
it can be inadequate sometimes. Razor implements a better polynomial
approximation utilizing the Hart Algorithm (1968)4. This method gives
double precision (14-16 decimal places) accuracy. Furthermore, the
complexity of this procedure is low, hence its performance is as good as
that provided by the standard one.

17.10 Cumulative Bivariate Normal Distribution Function

The two dimensional cumulative normal probability density function with
a mean of µ = [0, 0] and standard deviation of σ = [1,1] can be described
as:

1
1 − xT Σ −1 ( x ) Formatted: Lowered by 19 pt
f ( x, y , Σ ) = e 2
2π Σ Formatted: Lowered by 14 pt

 σ2 ρσ xσ y  Formatted: Lowered by 36 pt
Σ= x  Field Code Changed
 ρσ xσ y σ y2 
1 ρ
= , σ xy = (1,1) , Σ ≠0
ρ 1 
1
1 − xT Σ −1 ( x ) Field Code Changed
f ( x1 , x2 , Σ ) = e 2

2π Σ

3
M.Abramowitz and I. Stegun, Handbook of Mathematical Functions. New York, Dover
Publications, 1972.
4
E. G. Haug, The Complete Guide to Option Pricing Formulas. New York, Mcgraw-Hill,
1998.
©2009 Razor Risk Technologies Page 363 of 373
Razor Financial Principals

 σ x21 ρσ x σ x  Field Code Changed

Σ= 1 2

 ρσ x1σ x2 σ x2 2 
1 ρ
= , where σ x1 = σ x2 = 1, Σ ≠ 0
ρ 1 
Formatted: Indent: Left: 0 cm

Here we use the scalar form, in terms of the correlation ρ , by making the
following substitutions:

Σ = (1 − ρ 2 ) Formatted: Lowered by 7 pt

Σ = (1 − ρ 2 ) Field Code Changed

1  1 −ρ   x 
xT Σ −1 ( x) =
(1 − ρ 2 )  − ρ1   y 
for σ xy = (1,1)
x 2 + y 2 − 2 ρ xy
=
(1 − ρ 2 )

E[ XY ]
ρ= = E[ XY ] where E[ XY ] is the covariance of XY
σ xσ y

x 2 + y 2 − 2 ρ xy
1 −
f ( x, y , ρ ) = e 2(1− ρ 2 ) , ρ ∈ [ −1,1]
2π 1 − ρ 2

The cumulative bivariate normal distribution function is defined as:

©2009 Razor Risk Technologies Page 364 of 373

Razor Financial Principals

a b
Φ xy (a, b, ρ ) = ∫∫ f ( x, y, ρ )dxdy
−∞ −∞
a b x 2 + y 2 − 2 ρ xy
1 −
=
2π 1 − ρ 2
∫ ∫e
−∞ −∞
2(1− ρ 2 ) dxdy

This integral is solved numerically in Razor using the Drezner (1978)5

algorithm with the following added special conditions:

17.10.1 Special Conditions

The Drezner approximation does not correctly handle the limiting case of
ρ = ±1 . Also in cases where ρ = 0 and a=b=0, optimizations can be made
to simplify computation. The following additions have been made to the
Drezner algorithm:

Case 1: ρ = 1

Φ xy ( a, b,1) = Φ (min( a, b))

where Φ is the standard cumulative normal distribution function in one

dimension

Case 2: ρ = −1

 0 b ≤ −a
Φ xy ( a, b, −1) = 
Φ (a) + Φ (b) − 1 b > −a

Case 3: ρ <ε , ε = 1×10−8

a b x 2 + y 2 − 2 ρ xy
1 −
Φ xy (a, b, 0) = lim
ρ →0
2π 1 − ρ 2 ∫ ∫e
−∞ −∞
2(1− ρ 2 ) dxdy

a b x2 + y 2
1 −
=
2π ∫ ∫e
−∞ −∞
2 dxdy

1  a − x 2   b − y 2  
=  ∫ e 2 dx  ⋅  ∫ e 2 dy  
2π  −∞   −∞  
= Φ ( a ) ⋅ Φ (b)

5
John C. Hull, Options Futures and Other Derivatives. New York, Mcgraw-Hill, 2002.
©2009 Razor Risk Technologies Page 365 of 373
Razor Financial Principals

Case 4: a , b < ε , ρ ∈ [−1,1], ε = 1×10−8

When a,b are very small, the probability density function then reduces to:

x 2 + y 2 − 2 ρ xy
1 − 1
lim e 2(1− ρ 2 ) =
x , y →0
2π 1 − ρ 2
2π 1 − ρ 2

Therefore, the cumulative bivariate normal distribution function can be

solved analytically as:

1 1
Φ xy ( a, b, ρ ) = ∫ dρ
2π 1− ρ 2
sin −1 ( ρ ) 1
= +
2π 4

17.11 Reiner and Rubinstein Single Barrier Model

For pricing “In” and “Out” barrier options we will use the following
notations:

S S
ln(
) ln(
)
x1 = X + (1 + µ )σ T x2 = H + (1 + µ )σ T
σ T σ T
H2 H
ln( ) ln( )
y1 = SX + (1 + µ )σ T y2 = S + (1 + µ )σ T
σ T σ T
H σ2
ln( ) b−
S + λσ T 2r
z= µ= 2 λ = µ2 +
σ T σ 2
σ2

A = ϕ Se( b− r )T N (ϕ x1 ) − ϕ Xe − rT N (ϕ x1 − ϕσ T )
B = ϕ Se( b− r )T N (ϕ x2 ) − ϕ Xe − rT N (ϕ x2 − ϕσ T )
H 2( µ +1) H
C = ϕ Se( b − r )T ( ) N (η y1 ) − ϕ Xe − rT ( ) 2 µ N (η y1 − ησ T )
S S
H H
D = ϕ Se( b − r )T ( ) 2( µ +1) N (η y2 ) − ϕ Xe − rT ( ) 2 µ N (η y2 − ησ T )
S S
H 2µ
E = Ke [ N (η x2 − ησ T ) − ( ) N (η y2 − ησ T )]
− rT

S
H µ +λ H µ −λ
F = K [( ) N (η z ) + ( ) N (η z − 2ηλσ T )]
S S

©2009 Razor Risk Technologies Page 366 of 373

Razor Financial Principals

with
S representing the price of underlying asset,
H – option barrier value,
T - option time to maturity,
X - option strike price,
K – pre-specified cash rebate that is paid out at option expiration if the
barrier is not hit during option life,
σ - underlying asset volatility of correct maturity
r - continuously compounded risk-free interest rate
b – cost of carry.

“In” Barrier Options.

“In” barrier options come into existence if the underlying asset price
process hits the barrier level H before option expiration.
Call option with “down and in” type barrier pays payoff max( S T -X,0) if
S t ≤ H at any moment t before maturity T and a (possible)
predetermined amount K otherwise.

The price of such option can be expressed as:

Cdi ( X > H ) = C + E η = 1, ϕ = 1
Cdi ( X < H ) = A − B + D + E η = 1, ϕ = 1

Call option with “up and in” type barrier that pays payoff max( S T -X,0) if
S t ≥ H at any moment t before maturity T and a predetermined amount
K otherwise priced according to the following equations:

Cui ( X > H ) = A + E η = −1, ϕ = 1

Cui ( X < H ) = B − C + D + E η = −1, ϕ = 1

For corresponding put options:

Put option with “down and in” barrier, payoff at maturity -
max(X- S T ,0) if S t ≤ H at any moment t before maturity T and amount K
otherwise priced as:

Pdi ( X > H ) = B − C + D + E η = 1, ϕ = −1
Pdi ( X < H ) = A + E η = 1, ϕ = −1

The value of put “up and in” option is:

Pui ( X > H ) = A − B + D + E η = −1, ϕ = −1

Pui ( X < H ) = C + E η = −1, ϕ = −1

“Out” Barrier Options.

These type of call and put options remains vanilla type options unless
they are knocked out and become worthless at the moment when the
©2009 Razor Risk Technologies Page 367 of 373
Razor Financial Principals

underlying asset process hits barrier level. Similar to “in” type options it
is also possible include a predetermined rebate amount K that
compensate the knock out event.
“Down and out” calls have payoff at maturity T as max( S T -X,0) if S t > H
for all t ≤ T or K otherwise and prices defined:
C do ( X > H ) = A − C + F η = 1, φ = 1
C do ( X < H ) = B − D + E η = 1, φ = 1

“Up and out” calls with payoff at maturity T as max( S T -X,0) if S t < H
for all t ≤ T and K compensator priced according to:
C uo ( X > H ) = F η = −1, φ = 1
C uo ( X < H ) = A − B + C − D + F η = −1, φ = 1

Put options values are:

for “Down and out” case -
Pdo ( X > H ) = A − B + C − D + F η = 1, φ = −1
Pdo ( X < H ) = F η = 1, φ = −1

and “Up and out” case:

Puo ( X > H ) = B − D + F η = −1, φ = −1
Puo ( X < H ) = A − C + F η = −1, φ = −1

17.12 Finite Difference Methods

17.12.1 Introduction
Though renown Black-Scholes (1973) and Merton (1973) equation and
formulae can price the majority of vanilla type financial derivatives many
of OTC traded options often include more "exotic" than vanilla features
and thus their pricing frequently rely on numerical techniques.
Finite difference methods is one of the most efficient and simple of such
numerical procedures that allows calculation of option premium by
solving the partial differential equation that represents option price
evolution:
df df 1 2 2 ∂ 2 f
+ rS + σ S = rf
dt dS 2 ∂S

The method requires the construction of a discrete mesh that represents

the continuous domain of the state variables of underlying partial
differential equation. The time space [0,T] (T – time to option maturity)
is discretised into a finite collection of N+1 equally spaced intervals of
T
length ∆t = and underlying asset space (stock prices) is represented by
N
M+1 equally spaced intervals
©2009 Razor Risk Technologies Page 368 of 373
Razor Financial Principals

S
0, 0, ∂S ,2∂S ,3∂S ,..., S max of length ∆S =.
M
The value S max is chosen sufficiently high that, when it is reached, the put
payoff has no value. M is finite enough to get the required accuracy. The
values of continues partial derivatives from the above differential
equation are approximated then on the pre-built mesh nodes with an
explicit and implicit numerical schemes or with their combination.
A numerical scheme commonly used is Crank-Nicholson scheme, which is
unconditionally stable and can be considered as the average of the
implicit and explicit methods:

u mn +1 − u mn 1 u mn ++11 − 2u mn +1 + u mn +−11 1 u mn +1 − 2u mn + u mn −1
= + ,
∆t 2 ∆S 2 2 ∆S 2

where u mn ≈ f (m∆S , n∆t ) - the approximation of option premium f at each

node of the mesh.
Using above approximation the option price partial differential equation
is converted into a series of ordinary difference equations, and the
difference equations are solved iteratively.

Finite difference methods value a derivative by solving the differential

equation that the derivative satisfies. The differential equation is converted
into a series of difference equations, and the difference equations are solved
iteratively.
The differential equation that an American option must satisfy is:

∂f ∂f 1 2 2 ∂ 2 f
+ rS + σ S = rf
∂t ∂S 2 ∂S 2

17.12.2 The Time Axis

Suppose that the life of an option is T . We divide this into N equally spaced
T
intervals of length δt = . A total of N + 1 times are therefore considered.
N
0, δt ,2δt ,..., T

17.12.3 The Price Axis

Suppose the S max is a stock price sufficiently high that, when it is reached, the
S max
put has virtually no value. We define δS = and consider a total of M + 1
M
equally spaced stock prices:
0, δS ,2δS ,..., S max
The level S max is chosen so that one of these is the current stock price.

©2009 Razor Risk Technologies Page 369 of 373

Razor Financial Principals

17.12.4 The Grid

The time points and price points define a grid consisting of a total of
( M + 1)( N + 1) points. The (i , j ) point on the grid is the point that corresponds
to time iδt and price jδS . We will use the variable f i , j to denote the value of
the option at the (i , j ) point.

17.13 Implicit/Explicit Finite Difference Methods

17.13.1 The Forward Difference Approximation
∂f
For an interior point (i , j ) on the grid, can be approximated as:
∂S
∂f f i , j +1 − f i , j
=
∂S δS

17.13.2 The Backward Difference Approximation

∂f
For an interior point (i , j ) on the grid, can also be approximated as:
∂S
∂f f i , j − f i , j −1
=
∂S δS

17.13.3 Averaging the Forward and Backward Approximations

We can use a more symmetrical approximation by averaging the forward and
backward approximations:

∂f f i , j +1 − f i , j −1
=
∂S 2δS
For ∂f / ∂t we will use a forward difference approximation so that the value at
time iδt is related to the value at time (i + 1)δt :

∂f f i +1, j − f i , j
=
∂t δt
The finite difference approximation for ∂ 2 f / ∂S 2 at the (i, j ) point is:
∂2 f f i , j +1 − f i , j f i , j − f i , j −1
∂S 2
=( δS − δS )
δS
or

∂2 f f i , j +1 + f i , j −1 − 2 f i , j
=
∂S 2
δS 2
Substituting into the differential equation and noting that S = jδS gives

f i +1, j + f i , j f i , j +1 + f i , j −1 1 f i , j +1 + f i , j −1 − 2 f i , j
+ rjδS + σ 2 j 2δS 2 = rf i , j
δt 2δS 2 δS 2

©2009 Razor Risk Technologies Page 370 of 373

Razor Financial Principals

for j = 1,2,..., M − 1 and 0,1,..., N − 1. Rearranging terms we obtain:

a j f i , j −1 + b j f i , j + c j f i , j +1 = f i +1, j
Where
1 1
aj = rjδt − σ 2 j 2δt
2 2
b j = 1 + σ j σt + rδt
2 2

1 1
c j = − rjδt − σ 2 j 2δt
2 2

©2009 Razor Risk Technologies Page 371 of 373

Concept Index
Bank Accepted Bill, 227
Bonds, 228, 230, 231
Caps and Floors, 182
Cash Advance Facility, 222
Certificate of Deposit, 226
Collars, 184
Collateral, 222
Commercial Bill, 225
Commercial Paper, 226
Commodity Average Rate Options, 156
Commodity Average Rate Swap, 152
Commodity Double Barrier Option, 150
Commodity Forwards, 144
Commodity Single Barrier Option, 147
Commodity Vanilla Options, 145
Credit Default Swaps, 325, 334
Credit Risk, 14, 15
Equity Futures, 272
Equity Index Options, 267
Equity Options, 259
Floating Rate Notes, 241
Forward Rate Agreements, 160
FX Average Rate Options, 128
FX Digital Barrier Option, 120
FX Digital Option, 117
FX Double Barrier Option, 114
FX Double Digital Option, 123
FX Forwards, 104
FX Single Barrier Options, 109
FX Vanilla Options, 106
Generic Interest Rate, 224
Interest Rate and Cross Currency Swaps, 161
Metal Double Barrier Option, 141
Metal Forwards, 136
Metal Single Barrier Option, 139
Metal Vanilla Options, 137
Monte Carlo Simulation, 17
Monte Carlo Specification, 18
Ordinary Shares, 252
Promissory Note, 225
Razor Financial Principals

Repurchase Agreements, 241

Swaption, 185

©2009 Razor Risk Technologies Page 373 of 373