DRAFT

WORK IN PROGRESS

Technical specifications assessing the

development of the Zero rate curve for
Caribbean jurisdictions

June 2017

This guidance is a non-mandatory resource for members and it imposes no

obligation upon members.
DRAFT 13 November 2013
WORK IN PROGRESS
Contents

Contents
1. Executive Summary i
2. Summary of methodology for assessment 1
3. Introduction 2
3.1. Purpose 2
3.2. Supporting documents 3
4. Scope 4
4.1. Valuation Rates: Definitions 4
4.2. Yield curve basics 5
4.3. Discount bonds and spot rates 5
5. Draft Recommendations for the development of the Risk Free Spot
Zero rate curve 6
5.1. Overview 6
5.2. Market Data Sourcing: Choice of reference instruments 8
5.3. CAA Principle Recommendation on which instruments 8
5.4. CAA Principle Recommendation on the relevant subset of Government bonds 9
5.5. Determination of the observable set of market data points and the last liquidity
point (LLP) 9
5.6. CAA Principle Recommendation on the use of illiquid price points 9
5.7. CAA Principle recommendation on the use of historic price points 10
5.8. Technical specification for initial assessment for the longest reliable horizon (LRH) 11
5.9. Technical specification for assessment – methodology to incorporate past prices 11
5.10. Recommendation for interpolation and extrapolation 12
5.11. A consensus has emerged for a “macroeconomic”/phased approach 14
5.12. Overview of The Smith Wilson – macro economic approach 14
5.13. Considerations for specifying the ultimate forward rate 15
5.14. Considerations for specifying alpha 16
5.15. CAA Recommendation for interpolation and extrapolation method 16
5.16. Technical specification of the Smith Wilson Approach for initial assessment 17
Appendix A. Implications of incomplete data/missing observations 18
Appendix B. Overview of the common methods for the extrapolation for
the “unobservable” part of the curve 20
Appendix C. EIOPA’s Deep Liquid and Transparent (DLT) assessment for
reference instruments and the general bond market 22
Appendix D. Market Data Sourcing 24
Appendix E. A Technical Note on the Smith-Wilson Method 26
Appendix F. Valuation Rates – Understanding the Basic definitions
and relationships 27
Appendix G. The Sovereign rate adjustment to the Extrapolated
Zero Curve 34
DRAFT 13 November 2013
WORK IN PROGRESS
Contents

Appendix H. Overview of the rational for the matching rate

adjustment/Illiquidity premium to the Extrapolated Zero Curve 36
Appendix I. The estimation of the interest rate term structure in markets
with infrequent trading 38
DRAFT
WORK IN PROGRESS
Executive Summary

1. Executive Summary
1.1. The purpose of this document is to describe a recommended method for the
initial assessment of the development of the interest rate term structure (aka
valuation rates).

1.2. The recommendation of this approach is for testing should not be understood
as pre-empting or in any way restricting the final agreement on the
methodology for the final Guidelines. Furthermore, a number of technical
assumptions contained in this document have been made for pragmatic
reasons1 and for the purpose of the assessment only. The majority of areas
where pragmatic short-cuts have been taken are marked with a disclaimer, but
potentially not all of them.

1 For example recommended by other standard setting bodies not customized for our markets pe se as we still need to collect
more historic market price data for each jurisdiction

i
DRAFT
WORK IN PROGRESS
Summary of methodology for assessment

2. Summary of methodology for assessment

Topic CAA Principle recommendation Technical specification for testing
Market data • Reference financial instruments will be
sourcing Government bonds
• Both on the run and off the run issues
Determining • Use of historic price data to define the • Maximum of 2 years of historic price data
observable set of observable market data price set to be considered
market price points • Function of trading frequency at
each maturity
• Use of moving average yields at each
observable maturity
• Expert judgement to remove or adjust for
market prices pre-sovereign rating
downgrades or restructuring, or known
fire sale trades
Methodology for • Macro-economic method used for both • Use of Smith Wilson method
interpolation or interpolation and extrapolation
extrapolation
• Ultimate forward rate (UFR) • TBD by jurisdiction
parameterisation = historic average short
term interest rates plus benchmark for
target inflation
• Alpha – short period to converge to UFR for • 10 years for testing
all jurisdiction as the LLP is relatively short
~10 to 15 years
Sovereign risk • TBD • Use sovereign risk premium as a
adjustment conservative measure proxied by
differential in local yields and double A US
yield adjusted for inflation
Matching rate • Should be reflected for eligible liabilities • None
adjustment (MA)- • Only based on assets currently held as such • Company Specific – not tested at
Company Specific is Company specific and should be this point
determined by each Company
• Recommend general approach in Appendix
H
• A general approach for the calculation of
the MA determined by considering which
assets are not exposed to losses on
forced sales
• Application ratio is company-specific • None
to eligible liabilities be determined by • Company specific – not tested at
each Company this point
• Where the MA is applied in full (application
ratio of 100%)
• Where the MA is not applied at all
(application ratio of 0%),

Page | 1
DRAFT
WORK IN PROGRESS
Introduction

3. Introduction
3.1. Purpose
3.1.1. The valuation of insurance/actuarial liabilities) requires financial parameter inputs or
valuation rates such as discount factors or a term structure of interest rates derived from
available capital market prices.

3.1.2. The CAA’s position is that for each member currency, a relevant risk-free interest rate term
structure should be defined following a consistent methodology. This interest rate term
structure should be used to measure the time value of cash-flows payable in the member
currency. For a given currency and valuation date, each insurance and reinsurance
undertaking should use the same relevant risk-free interest rate term structure.

3.1.3. The purpose of this document is to describe a recommended method for the initial
assessment of the development of the interest rate term structure (aka valuation rates).

3.1.4. The recommendation of this approach is for testing should not be understood as pre-empting
or in any way restricting the final agreement on the methodology for the final Guidelines.

3.1.5. The purpose of testing is to collect data and provide a reliable basis for an informed decision
on the final methodology. Furthermore, a number of technical assumptions contained in this
document have been made for pragmatic reasons and for the purpose of the assessment
only. The majority of areas where pragmatic short-cuts have been taken are marked with a
disclaimer, but potentially not all of them.

3.1.6. The methodology in this document is designed to reflect the high level principles of the IFRS
Phase 2 Insurance contract standards. Most notably the June 2013 exposure draft of the
standard states that:

“Discount rates that reflect the characteristics of the cash flows of an insurance
contract may not be directly observable in the market. An entity shall maximise the
use of current observable market prices of instruments with similar cash flows, but
shall adjust those prices to reflect the differences between those cash flows and the
cash flows of the insurance contract in terms of timing, currency and liquidity.
However, the draft IFRS Standard does not prescribe the method for making
those adjustments.”

Page | 2
DRAFT
WORK IN PROGRESS
Introduction

3.1.7. Currently, the Caribbean markets, similar to other developing economies, are generally
characterised by illiquid bond markets with low trading volume, (sparse market price data)
and with a non-negligible amount of sovereign risk. To the extent that the characteristics of
the cashflows of the relevant insurance contracts require under the principles IFRS Phase 2
the use of “risk free”, liquid interest rates for the purpose of discounting, the methodology
developed needs to reflect the constraints of the available market bond price data in this
region and make appropriate adjustments where possible.

3.1.8. In summary, the approach will need to

• Take into account the issues of illiquidity, incompleteness, sovereign risk in the Caribbean
region, and
• Be consistent with the principles of the IFRS Phase 2 insurance contract standard where
deemed appropriate given the characteristics of the underlying insurance cashflows

3.1.9. There is currently no specific benchmark term structure model of interest rates issued by
independent agencies for the member currencies which sufficiently addresses these issues. As
such, in order to ensure more consistency of data sources, calculation methods and control
processes in the regional insurance industry, the CAA will initially publish valuation rate data
and the underlying referenced bond yields on a quarterly basis for key member currencies,
including Trinidad, Jamaica, Barbados and EC.

3.1.10. It is not compulsory for members to comply with this material.

3.2. Supporting documents

3.2.1. In addition to the IFRS Phase II exposure draft, this document relies heavily on the previous
research work conducted on estimating the term structure of interest rates. The information
in this document is supported by or references a number of documents including:

• Barrie and Hibbert – Yield curve extrapolation – A work in progress May 2013
• CEA/Insurance Europe, Examples of the Macroeconomic Extrapolation of the benchmark
risk-free yield curve, January 2010
• EIOPA – Technical Specifications part II on the Long-Term Guarantee Assessment –
January 2013
• CRO forum Best Practice Paper – Extrapolation of Market Data August 2010
• FSA Norway – A technical specification of the Smith Wilson Method
• Term structure estimation in markets with infrequent trading , International Journal of
Economics and Finance – 2007, Cortozar, Schwartz, Naranjo
• IAA Educational Monograph_20120810 FINAL

Page | 3
DRAFT
WORK IN PROGRESS
Scope

4. Scope
4.1. Valuation Rates: Definitions
4.1.1. The “risk-free” interest rate term structure of the member currencies should be provided at
least on a quarterly basis. Where market conditions are volatile, the term structures may be
provided more frequently.

4.1.2. The CAA will provide quarterly files of valuation rate data such as zero rates or discount
curves, for each major currency. The valuation rate data files should be based on well-defined
transparent sources and created in a controlled and documented environment.

4.1.3. They will then be distributed via the CAA website. The quarterly files will be named by
currency and by date. There will be two versions of a curve for each currency, one reflecting
sovereign risk the other excluding sovereign risk.

The table below provides a brief description of each valuation rate data to be provided.

Valuation rate Description

Spot zero curves • A spot interest rate is the rate at which an individual cashflow is discounted. If spot
rates for payments at all dates in the future are known, then the price of the
financial asset/liability can be equated to the present value of future cashflows
• The spot zero curve provided is a set of zero coupon interest rates valid on the date
of publication for annual terms to maturity up to 120 years
• The spot zero-coupon interest rate, denoted rt, with term to maturity t, is the
continuously compounded rate of interest on a zero-coupon bond with term
to maturity t
Spot discount Curve • The current discount curve provided is the set of discount factors for all annual
maturities up to 120 years at the current date. Discount factor, Dt, is the value today
of a dollar payable at time t
Forward rate matrix • The forward rate matrix is the set of forward zero rates at all annual maturities for
each future year. Alternatively described, it is the series of future zero curves
implied by the current spot zero rates

Page | 4
DRAFT
WORK IN PROGRESS
Scope

4.1.4. Since the discount function, par yield curve, zero coupon yield curve and implied forward
yield curve are all algebraically related, knowing one of these four means that we can readily
compute the other three. This paper focuses on the derivation of the spot zero rate curves
also known as the term structure of interest rates.

4.2. Yield curve basics

4.2.1. The terms “yield curve” and “the term structure of interest rates” are often used quite loosely.
Here is a very basic guide to the terminology.

4.3. Discount bonds and spot rates

4.3.1. At the most basic level we are concerned with the price of financial contracts which pay €1 at
future dates in exchange for today’s prices. A “discount bond” or “zero-coupon bond” is the
simplest type of bond available and the prices of these bonds can also be expressed in terms
of a “spot” interest rate for a given maturity. For example, if the price of a 10-year discount
bond is $0.7441, this is equivalent to:

• An annually-compounded interest rate of 3.00% (0.7441 * 1.03 10 = 1.000)

• A semi-annually-compounded interest rate of 2.98% (0.7441 * 1.0149 20 = 1.000)
• A continually-compounded (log) interest rate of 2.96% (0.7441 * exp (0.0296 * 10)
= 1.000)
4.3.2. Because these compounding conventions can cause confusion, we will always publish yield
curve data in the form of discount bond prices which are unambiguous.

See Appendix F for a fuller review of valuation rates and their relationships.

Page | 5
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

5. Draft Recommendations for the development of the Risk Free

Spot Zero rate curve
5.1. Overview
5.1.1. The appropriate risk-free interest rate term structure is necessarily constructed from a finite
number of data points of sufficient liquidity. Therefore, both interpolation between these
data points and extrapolation beyond the last available data point of sufficient liquidity is
required. This zero curve derivation process can be broken down into 3 key steps:

A. Market Data Sourcing

B. Interpolation
C. Extrapolation

5.1.2. In most capital markets, there will be bond market price data available at certain maturities.
For example, in a case where there is adequate bond market price data with maturities
between 3 and 10 years, the observable part of the yield curve will span 3 years to 10 years.
However, there will likely not be a price point at each maturity between 3 to 10 years. As such,
a methodology is required to interpolate between the available price points to fit the
“observable” part of the yield curve. That is, the objective of interpolation is to fit the
observable or “liquid” part of the yield curve. In addition, in this example, the last observable
point or “last liquidity point” (LLP) is the 10 year maturity. As such, an extrapolation
methodology is also required to extrapolate or complete the yield curve beyond the 10 years.

5.1.3. Thus for both markets, developed and emerging, the common questions which need to be
answered include:

1. What are the appropriate sources of market price data which can be used to develop the
zero curve. That is, what type of financial instruments should be referenced?
2. Which set of price points can be incorporated to fit the observable part of the zero
curve? What method will be used to define the last observable price point or the last
liquidity point (LLP) which is appropriate from which the zero curve needs to be
extrapolated
3. What is the most appropriate method for interpolation
4. What is the most appropriate method for extrapolation

Page | 6
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

5.1.4. For emerging or developing markets these 3 steps are not as evident as the commonly used
methods assume the availability of relatively complete, liquid risk free market price data
points across the range of the term structure.

5.1.5. The methodology developed for this region also needs to reflect the issues specific to the
Caribbean, including:

1. Incomplete or sparse market price data

2. Sovereign risk
 Depending on the use and the nature of the liability cashflows, the rates may need to
be adjusted for sovereign risk. Appendix G provides a simple methodology to
estimate the market based sovereign risk premium
5.1.6. The CAA will provide reference curves for the major currencies with two versions, one
reflecting sovereign risk the other excluding sovereign risk. Both will exclude any matching
adjustment as this should be a Company specific determination.

5.1.7. According to the proposed IFRS Phase 2 exposure draft for the purpose of the valuation of
illiquid insurance cashflows, there is an allowance for a matching adjustment or an illiquidity
premium for certain life insurance liability cashflows. Appendix H to this document provides
the rationale and an outline of a methodology for the matching adjustment which can be
applied to reference discount curves.

5.1.8. The approaches and assumptions described in this document (e.g. Ultimate Forward Rate,
Last Liquid Point (LLP), and sovereign risk adjustment) often reflect the fact that the
respective technical Standards under IFRS Phase II are still currently under development.
Therefore none of these assumptions should be seen as an indication for the final
implementation, but rather as a pragmatic approach chosen for this assessment only.

5.1.9. This document provides the recommendations on market data sourcing, derivation
observable market price data points, the last liquidity point, interpolation and extrapolation
method for the purposes of the technical assessment.

5.1.10. The Appendices describe further the insights and the considerations and rational for the
selection of the recommended methodologies in more detail.

Page | 7
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

5.2. Market Data Sourcing: Choice of reference instruments

5.2.1. To determine the basic risk-free term structure, considerations have to be made in respect of
the availability and the relevance of data. In addition, those data have to be adjusted for the
inherent credit risk or take into account the peg to another lead currency, if applicable.

5.2.2. In most member jurisdictions, Government bonds are the closest, most available financial
instruments to use as a starting point to build an appropriate discount curve for long term
actuarial liabilities (See Appendix D –market data sourcing).

5.3. CAA Principle Recommendation on which instruments

5.3.1. The CAA recommends the use of regional Government bonds to determine the liquid part of
the “risk free” term structure in each jurisdiction2.

5.3.2. Deciding to use government debt as a reference for building a risk-free yield curve is often the
easy part of the process. But the analysis grows more difficult when the actuary starts to
consider the appropriate subset of all possible government securities that should be included
in the analysis the various subsets include:

• On-the-run securities (the most recently issued security of a particular maturity)

• Off-the-run securities (securities other than the most recently issued)
• A combination of on-the-run and off-the-run securities

2 Insurance company survey for 4 jurisdictions demonstrated that Government bonds comprise over 50% of the asset portfolio
of most life insurers.

Page | 8
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

5.4. CAA Principle Recommendation on the relevant subset of Government bonds

5.4.1. In the Caribbean region, where Government are thinly traded we would recommend the use
of both on-the-run and off-the-run securities.

5.5. Determination of the observable set of market data points and the last
liquidity point (LLP)

5.5.1. Background

5.5.1.1.First relevant question is: What instruments or price points should be included? If traded assets do not
offer some minimum level of liquidity, should they be ignored?

5.5.1.2.There are strongly conflicting views on what market data should be “respected” in fair valuation work
and what might be ignored. One point of view argues that only prices from highly-liquid markets should
be used. Solvency II regulations demand that input prices (for estimating the traded part of the curve)
must meet the accountants’ active market definition and additionally satisfy a “Deep, Liquid and
Transparent” (DLT) requirement3.

5.5.1.3.Crudely, another camp – led by the accounting standard-setting bodies – seeks to maximize the
amount of information used from active markets i.e. where “quoted prices are readily and regularly
available … and those prices represent actual and regularly occurring market transactions on an arm’s
length basis. Further where markets are deemed to be inactive the analyst/accountant “cannot ignore
relevant market data (including observable transaction prices) when it is clear that market participants
would use that data in determining the price at which they would be willing to enter into a transaction
for the financial asset”.

5.5.1.4.There is a move from the strict definition of liquidity to reflect the fact that the impact of the
instrument’s liquidity should be separated from the horizon at which grading towards the ultimate
forward rate (UFR) begins. A good method should focus less on the last liquidity point but more on the
longest reliable horizon for forward rates4,5.

5.6. CAA Principle Recommendation on the use of illiquid price points

5.6.1. The bond markets in the Caribbean will not meet most of the DLT criteria. However it is our
view that the limited price points which are available are still very relevant as

3 See EIOP Technical specification –DLT assessment for reference instruments and the general bond market for definition of DLT.
4 Ref – CRO Forum extrapolation of Market Data August 2010. Strict liquidity criteria require that that market participants can
rapidly execute large-volume transactions with little impact on the prices of the financial instruments used in the replication;
or – based on expert judgment – allow to include market data that does not meet the strict liquidity requirement, but is still
considered to be a good reflection of the cost of the replicating portfolio
5 Ref – Yield Curve Extrapolation – Work in Progress – May 2013, Barrie & Hibbert

Page | 9
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

• They will be considered by market participants when pricing

• They are held by most insurers and as such those trade points are a better reflection of
the cost of the replicating portfolio than an extrapolation derived by excluding them

The second relevant question is: Should historic market price data points be ignored?

5.6.2. There are other features besides goodness of fit to observed current prices that are desirable
in a term structure model, such as the time-series stability of the term-structure curves
obtained. It might well be the case that the model fits very well the existing bond prices (or
yields), but it implies large daily movements of yields for maturities that are not traded. This is
not an issue for liquid markets, but can be a serious problem for thin markets.

5.6.3. Of particular note, the research paper in the International Journal of Economics and Finance
entitled “Term Structure Estimation in markets with infrequent trading” provides a
methodology for the development of a term structure using historic prices. The paper
demonstrates it is preferable to incorporate historic price data vs. calibrating models based
only on using current but thin market price information.

5.7. CAA Principle recommendation on the use of historic price points

5.7.1. In order to minimize model-induced volatility which is not reflective of actual volatility in the
region’s markets, we recommend that past prices should be incorporated where needed
to both:

1. Define what is the “liquid” or observable part of the curve, or the longest reliable horizon
or last “liquidity” point
2. Calibrate the parameters of the chosen model
5.7.2. The historic data price points incorporated need to reflect the trading frequency and must
also be adjusted for any sovereign rating downgrades or debt restructuring, fire sales (if
known) over the period chosen based on expert judgment6.

5.7.3. In order to ensure transparency, we recommend that the “observable” market bond price
data set used be published quarterly as well along with the zero rate curve or discount curve.

• Next step: In order to design and parameterize the final approach:

 The historic market price information for each currency needs to be provided,
including trading frequency historic information, average yields and standard

6 A lot more work needs to be done here to develop objective criteria, in order to avoid that expert judgment is needed to
determine the observable price data set

Page | 10
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

deviation of past Government yields to confirm and parameterize this approach or

any other approach
 We will require testing and back testing

5.8. Technical specification for initial assessment for the longest reliable horizon
(LRH)

5.8.1. For the purpose of the assessment to define the set of observable data points and the last
liquidity point before extrapolation is required, we propose that we test the following to the
define the longest reliable horizon.

• The need for at least one observable price point at the longest maturity within the last
two years to define the longest reliable horizon. That is all observable trades within the
last two years will be used to parameterize the model (subject to any considerations of
sovereign rating downgrades, restructuring, known firesales)
• The bond with the longest maturity during the two year history will define the longest
reliable horizon or last liquidity point

5.9. Technical specification for assessment – methodology to incorporate

past prices

5.9.1. Cortozar’s 2007 research paper in the International Journal of economics and finance titled
“Term structure estimation in markets with infrequent trading” proposes a stochastic process
to incorporate three years of historic prices using a Kalman filter to parameterise the curve
fitting models. This approach is used in Chilean capital markets. However, it is not clear that
there are adequate historic data points in this region to parameterize and replicate this
process. This is still being investigated. In the interim, we recommend for the assessment:

• Use of moving average yield for each maturity over the maximum of two years of historic
price data
• If there is more than one price point at a specific maturity within the last two years, two
years of historic price point data may not be required. For example, if there were over 3
bonds traded or issued at the 5 year maturity point over the most recent quarter, there
will not be a need to use 2 years of historic information
• Ideally, to reflect this, there should be a transparent algorithm to determine the number
of data points at each maturity to be incorporated

Page | 11
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

• We propose for the assessment to define the use of historic price period to be reflected
based on the trading frequency7, TF, criteria as follows:
 TF >20% => maximum of 3months of historic price data to be use. That is the 3 month
moving average yield of the last three months on traded government bonds will be
the observable price point
 10% <TF<20% – maximum of 6months of historic price data to be used. That is the 6
month moving average yield on traded government bonds will be the observable
price point
 5% <TF<10% – maximum of one year of data. That is the 12 month moving average
yield will be the observable price point
 TF<5% – maximum of two years of data. That is the 24 month moving average yield
will be the observable price point

5.10. Recommendation for interpolation and extrapolation

5.10.1. Background

5.10.1.1. The appropriate interest rate term structure will in practice be constructed from a finite number
of liquid market data points as defined in the section above. Therefore, both interpolation between
these data points and extrapolation beyond the last liquid point (LLP) are required.

Interpolation:

The objective of interpolation is to fit the observable “liquid” part of the yield curve term
structure for which bond market price data is available.

Extrapolation

The unavailability of quoted bonds or other market instruments which have cashflows
beyond a certain date is called an “incomplete market”. Hence models are also needed to
extrapolate the prices or zero rates for these longer dated cashflows which are common in
life insurance. Yield curve extrapolation is concerned with the construction of yields for
maturities that fall beyond the traded market.

7 For example a trading frequency of 20% means that at least one bond with that maturity was traded an average of 50 days per
year

Page | 12
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

5.10.1.2. Ideally, a consistent approach for both the interpolation and extrapolation would be preferred.
Different approaches can lead to inconsistencies between the interpolated and extrapolated part of the
same curve and also to inconsistencies over time for each part of the curve. However, common
methods for fitting the market don’t easily extend to non-traded maturities. There is a large body of
technical literature concerned with fitting yield curves to traded instrument prices. The practical
application of this work has been led by the central banks – such as the US Federal Reserve and UK
Bank of England – who routinely publish yield curves (as opposed to simply reporting raw bond prices
and individual bond yields). A good survey of central bank methods can be found in in the BIS 2005
survey8. The natural focus of central banks is the traded market and the information that can be
inferred from market prices about both the cost of funding government borrowing and – for the
purposes of policy formation – expectations, risk premia and so on. These curves are not constructed
for the purpose of extrapolation and, it turns out that they are an inadequate starting point
for extrapolation9.

5.10.1.3. Given that changes in long term discount rates can cause substantial changes in the value of long
term liabilities and thereby lead to procyclical effects, the focus, for our purposes, should be primarily
on the choice of the extrapolation technique.

5.10.1.4. The CRO forum identified four key principles for extrapolation. Most of these high level principles
should also be applied to the non-extrapolated part of the curve. They include:

1. Use all relevant observed market data where available

2. Extrapolated market data should be arbitrage-free
3. The extrapolation method should be theoretically sound
 That is, it should be consistent with generally accepted economic theory or clearly
observed historical patterns and take into account the effect on financial stability. In
particular, the method should not result in spurious volatility which is not consistent
with past historic patterns. This is particularly important for currencies in thin
markets where liquid reference rates are only available for short term maturities and
simple extrapolation of these short term interest rates may cause excessive volatility
4. The extrapolation should follow a smooth path from the entry point to the unconditional
ultimate long-term level

5.10.1.5. Notable contributions to the design of methods suitable for long-term institutions which focus
explicitly on the behaviour of the extrapolation include:

• Papers by Smith & Wilson (2000) and Thomas & Maré (2007)
• A series of technical papers and an Exposure Draft 1 setting out practice which was
reviewed and endorsed by the B&H

8
The most prominent examples used by the ECB and other central banks are the Svensson method and the Nelson-
Siegel method
9 Barrie and Hibbert – Yield curve extrapolation – A work in progress May 2013

Page | 13
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

• Technical Advisory panel in 2008/9

• Various papers published by CEIOPS (now EIOPA) for the field study QIS5 in 2010
• A CRO Forum paper setting out the large European insurance firms’ views on
best practice
• Various IFRS staff papers

5.11. A consensus has emerged for a “macroeconomic”/phased approach

5.11.1. Any method will – either explicitly or implicitly – say something about the forces that drive
the fair value of ultra-long term cash flows. Within these papers there is a degree of
consensus on the basic framework which is the macreconomic method. This approach is
referred to as “macroeconomic” because long-term economic assumptions drive the choice of
limiting forward rate.

5.11.2. So-called “macroeconomic” methods all aim to address three parts of the extrapolation
problem – a phased approach to building a full curve:

1. Providing a fit to some set of observable bond prices or interest rate quotations
2. Setting an assumption for a limiting/unconditional/ultimate interest rate – most usually a forward
interest rate
3. Combining the observed curve and ultimate assumption. In particular, defining the speed of
adjustment from an observed forward rate towards the chosen ultimate rate

Of note, the method chosen by the EIOPA Solvency II is the Smith-Wilson macroeconomic approach

5.12. Overview of The Smith Wilson – macro economic approach

5.12.1. The Smith-Wilson technique is a macroeconomic approach: a spot (i.e. zero coupon) rate
curve is fitted to observed prices of financial instruments, with the macroeconomic ultimate
long term forward rate as input parameter 1.

5.12.2. In the Smith-Wilson method the pricing function P(t), for all t>0, is set up as the sum of a term
e-UFR·t for the asymptotical long term behavior of the discount factor and a linear
combination of N kernel functions5 Ki(t), i=1,2,…,N (the number N of kernel functions being
equal to the number of input instruments).

5.12.3. The kernel functions are appropriately defined functions of the input market data and two
input parameters:

• The ultimate forward rate (UFR) and

• A parameter (alpha) that determines how fast the estimated forward rates converge
to UFR

Page | 14
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

5.12.4. If N input instruments are given, we know N market prices and can thus set up N linear
equations. In most of the cases the resulting system of linear equations (SLE) can be solved
automatically, i.e. without interfering from the outside. By plugging the solution of the SLE
(the solution assessed for the maturities of the N input instruments) into the Smith-Wilson
pricing function at any given time t we receive the discount function for maturity t. With the
discount function, the spot rate curve is known10. See Appendix E for more details.

5.13. Considerations for specifying the ultimate forward rate

5.13.1. Financial economists identify a number of factors that determine the level of forward
interest rates:

• Expectations for future short-term interest rates. Analysts often think about these
expected short rates in terms of an inflation component and an expectation for the real
rate of interest
• A “convexity” element which means that, where there is interest rate uncertainty, the
forward rate must incorporate a downward adjustment to ensure bonds of different
maturities all offer consistent expected (i.e. average) returns because of the non-linear
relationship between prices and yields
• A risk premium or discount to reflect investors’ attitude to risk and the balance of supply
and demand for risk-free cash flows at the specified maturity
• An adjustment for trading liquidity and liquidity risk. Assets which offer higher trading
liquidity – especially during stress periods – sell at (relatively) high prices and lower yields

10 Financial Supervisory Authority of Norway- A technical note on the Smith Wilson Method

Page | 15
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

5.13.2. Note that all of these elements might be expected to vary over time. However, the question
of how variations in risk premia and liquidity adjustments should be applied to maturities
beyond the traded curve turns out to be very difficult to answer.

5.14. Considerations for specifying alpha

5.14.1. The alpha parameter in the Smith-Wilson method determines both the speed of convergence
to the UFR in the extrapolated part, and the smoothness of the curve in the interpolated part.
Larger values of alpha give greater weight to the UFR, while smaller values of alpha give more
weight to the liquid market data.

5.15. CAA Recommendation for interpolation and extrapolation method

5.15.1. We recommend the Smith Wilson (S-W) Approach for both interpolation and extrapolation,
primarily because

• The method emphasizes smoothing as there is a mechanism in the method that provides
stability to the outcomes of the extrapolation, avoiding spurious movements in the long
part of the curve
• There is growing industry consensus toward this particular macroenomic method
• S-W is a method in the open domain. Both the formulae and a computing tool are to be
published on CEIOPS homepage. Thus, the method is wholly transparent and full
accessible to all companies
• S-W provides a perfect fit of the estimated term structure to the liquid market data. In
many other methods the term structure is assessed as a smoothed curve. In S-W all
relevant data from the liquid market is taken as input, no smoothing is performed
• S-W can be applied directly to the raw data from financial markets. No bootstrapping or
other methods are needed to transform market par swap rates into zero coupon bond
rates, as is the case for example in the linear extrapolation method, where the input has
to be first converted into zero coupon bond rates
• S-W is a uniform approach, both interpolation between the liquid market data points and
extrapolation beyond the last data point are performed
• In S-W the ultimate forward rate will be reached asymptotically. How fast the
extrapolated forward rates converge to the UFR will depend on how the rates in the liquid
part of the term structure behave and on an exogenous parameter alpha. For higher alpha
the extrapolated forward rates converge faster to the UFR, i.e. the market data from the
liquid part of the curve are of less impact for the extrapolated rates

Page | 16
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

5.15.2. The key disadvantages of the Smith-Wilson approach is that the parameter alpha has to be
chosen outside the model. Thus, in general, expert judgment would be needed to assess this
input parameter for each currency and each point in time separately.

5.15.2.1. Parameterisation

5.15.2.2. The ultimate forward rate (UFR) is the percentage rate that the forward curve converges to at
the pre-specified maturity. For this region, the UFR will be defined as a function of long-term
expectations to the inflation rate, and to the long-term average of the short-term real rate. The other
potential factors are difficult to estimate in this region

5.15.2.3. As this value is assessed in line with long-term economic expectations it is expected to be stable
over time and only change due to changes in long-term expectations.

5.15.2.4. In order to have a harmonized approach over all the major currencies we will for all currencies
use the same alpha.

5.16. Technical specification of the Smith Wilson Approach for initial assessment
5.16.1. Ultimate forward rate (TBD)

5.16.2. For the purpose of this assessment we propose to define the UFR as the sum of the

• Expected future short term interest rate for each jurisdiction using historic 15 year moving
average of the short term real rate
• Target inflation rate for each currency stated by the region’s home Central Bank
• These rates are to be provided shortly for December 2013

5.16.3. Parameter alpha the speed of convergence

5.16.3.1. In order to have a harmonized approach over all the major currencies we will for all currencies
use the same alpha. For the purpose of the assessment, the alpha parameter is calibrated so that the
extrapolated part of the forward curve converges to within 3 bps from the UFR within 10 years from
the LLP.

5.16.3.2. If this alpha is not appropriate for the currency it is applied to, we will increase it iteratively, until
it is deemed – based on given criteria – to be appropriate. A lot more work needs to be done here to
develop objective criteria for setting the alpha, in order to avoid that expert judgment is needed in all
these cases.

Page | 17
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

Appendix A. Implications of incomplete

data/missing observations11
A.1.1. Empirical evidence shows that in well-developed markets, where numerous bonds are traded
every day for different maturities, static curve fitting methods generate yield curves that
accurately fit current bond transactions (Bliss, 1996). As such most static curve fitting
methods propose fitting curves to current bond prices without regard to past prices12.
However there are problems when these methods are applied to markets with sparse risk
free bond prices and infrequent trading, which are common characteristics of many emerging
or developing markets, including the Caribbean13.

A.1.2. There are other features besides goodness of fit to observed current prices that are desirable
in a term structure model, such as the time-series stability of the term-structure curves
obtained. It might well be the case that the model fits very well the existing bond prices (or
yields), but it implies large daily movements of yields for maturities that are not traded. This is
not an issue for liquid markets, but can be a serious problem for thin markets

A.1.3. A second problem of these static curve-fitting methods when used in markets with infrequent
trading occurs when the prices for short or for long-term bonds are not available, even if the
number of observed prices is sufficient for the estimation. Curve-fitting methods provide
reasonable estimates within the time range spanned by the available prices, but provide
much less reliable estimates for extrapolations outside this range.

A.1.4. In many emerging markets it is common that for some dates long-term bonds are not traded;
but the need for complete term-structure estimation for valuation of long term actuarial
liability cashflows still remains. It is well-known that the term structure of volatilities is
downward sloping due to mean reversion in interest rates. This means that the volatility of
long rates obtained from any curve fitting model should be lower than the volatility of
short rates.

11 International journal of finance and economics: term-structure estimation in markets with infrequent tradinga91 - IJFE 07
Cortazar-Schwartz-Naranjo (4)
12
McCulloch (1971,1975), Vasicek and Fong (1982) and Fisher et al. (1994), among others, use spline curve-fitting methods to
estimate the current term structure. Nelson and Siegel (1987) and Svensson (1994) use parsimonious representations of the
forward curve limiting the number of parameters and giving more stability to the term structure
13 For example, traditional curve-fitting methods render unreliable estimates of the current term structure without a sufficient
number of observations or without short or long term bond prices

Page | 18
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

A.1.5. Cortazar’s paper provides a case example which illustrate that curve fitting methods such as
the Svensson model on the Chile market provide unstable estimates of long rates when no
long-term bonds are traded. The paper provides a plot of the volatility of interest rates
calculated from daily estimations of the term-structure in Chile between 1997 and 2001 using
the Svensson (1994) method. It can be seen that this term-structure of volatilities is not
consistent with mean reversion in interest rates: it implies very high volatilities for long rates.
Moreover, the Svensson volatility estimates are much higher than the empirical estimates
obtained directly from bond prices, suggesting that missing observations induce unreliable
rate estimates. Similar results are obtained when using other curve-fitting methods like
Nelson and Siegel (1987). Cortazar proposes to solve the problems of term-structure
estimation in markets with infrequent trading by using also past price information to infer the
current term structure. This requires a dynamic model of the stochastic behaviour of interest
rates to be able to mix current and past prices in a meaningful way.

A.1.6. Given the considerations above, for interpolation, we recommend methods which
incorporate the use of historic price data where appropriate to minimize model induced
volatility and to determine a more reliable estimate of the “liquid/observable” part of
the curve.

Page | 19
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

Appendix B. Overview of the common methods for the

extrapolation for the “unobservable” part of the curve
B.1.1. The choice of the extrapolation method is very important to determination of the insurance
liabilities. There are many techniques available for extrapolating the interest rate curve.

B.1.2. The CRO Forum has identified 4 principles for the extrapolation of market data for market
consistent valuation purposes:

1. Use all relevant observed market data where available

 As described in the previous subsection on determining the last liquidity point – in less
developed markets were longer term tenor bonds are not frequently traded consideration
needs to be given to any single trade or issues over the recent past as it is still relevant to
understanding the cost of the replicating portfolio
2. Extrapolated market data should be arbitrage-free
 Arbitrage opportunities should be avoided to ensure that the best-estimate is appropriate. For
example, when the interest rate curve is extrapolated based on the last observed liquid spot
rate this implies in most cases that you observe an abrupt change in slope of the curve, which
implies a jump in the instantaneous forward interest rate. Any arbitrage free extrapolation
therefore should have a continuous curve plus a continuous forward curve. Extrapolating based
on forward rate achieves this by construction and is therefore a good starting point for any
extrapolation method
3. The extrapolation method should be theoretically sound
 Extrapolation methods that are in violation of generally accepted economic theory or clearly
observed historical patterns should be avoided. The extrapolation method should not result in
spurious volatility. Volatility arises from genuine changes in expected levels of long-term best-
estimates and changes in risk premia
4. The extrapolation should follow a smooth path from the entry point to the unconditional ultimate
long-term level
 The view on the long-term level should be forward looking, reflecting market analysis, future
expectations, economic theory and historic experience. This view should be documented
appropriately and regularly reviewed
 There is also wide support to allow for some mechanism in the method that may provide
stability to the outcomes of the extrapolation, avoiding spurious movements in the long part of
the curve. This is particularly important for currencies where liquid reference rates are only
available for short term maturities and simple extrapolation of these short term interest rates
may cause excessive volatility

Page | 20
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

Method Description Comment

Simple extrapolation Simple extrapolation techniques require no deeper This approach does not meet the
analysis of the fundamentals or shape of the curve. arbitrage free criteria and could
In its purest form, the simple extrapolation generate very volatile curves in
technique assumes that the final liquid data point is markets with very short maturity
also the long term interest rate level. From the final observable rates (under 10 years).
liquid point onwards, the curve is therefore a Could result in extrapolation to
horizontal line. extraordinary levels as yield curves
tend to be steep at the shorter end
of the curve.
Macroeconomic Macroeconomic extrapolation techniques involve This is arbitrage free, simple to
extrapolation identifying a long term equilibrium interest rate, apply, suitable in markets where
techniques usually but not necessarily through economic there are only observable rates for
analysis, and interpolating between the available shorter term maturities,
data points and this additional long term compatible with most
equilibrium point. interpolation methods using
forward rates.
Parameterisation Parameterisation techniques emphasize smoothing This is arbitrage free, but not
techniques and provide an objective construction of the term suitable in markets where there
structure if the parameterization technique is fixed. are only observable rates for
They can be (but it is not necessarily the case) based shorter term maturities.
on economic assumptions. This category of
extrapolation techniques is the one currently most
used in market practice. There are many types of
parameterisation techniques: for example, constant
forward rates, the Svensson method, the one-factor
or Vasicek class of models, to mention only a few.
Constant or variable Constant or variable spread methods are alternative A priori we don’t think this
spread methods methods US currencies: first an appropriate approach or relationship is
extrapolation technique for the US is defined, then consistent with historic data for all
the rates for the other currencies are extrapolated jurisdictions, we would like to
by using the US curve plus In the case of a constant apply the same method to all
spread method, the constant spread between the US jurisdictions.
and the relevant currency for the last available liquid
data point of the relevant currency. In the case of a
variable spread, the spreads might be derived by
fitting a curve to the spreads observed in the non-
extrapolated part of the curve. This technique is
based on one unique extrapolation technique for
all currencies.

B.1.3. It should be understood that it is not possible to identify a single method performing the best
extrapolation for all currencies. However, it is our view important for the method to be
consistent for all territories. For all member jurisdictions we recommend the use of the
macroeconomic extrapolation technique as it is arbitrage free and suited for markets with
only observable rates for shorter term maturities.

Page | 21
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

Appendix C. EIOPA’s Deep Liquid and Transparent (DLT)

assessment for reference instruments and the general
bond market
C.1.1. The general bond market refers to all bonds in a currency, i.e. sovereigns plus corporate
(including financial) bonds. When assessing whether a market is transparent it needs to be
assessed for at least:

• Appropriateness of data provider choice

• Frequency of data update
• Simple availability checks, and
• Plausibility checks and monitoring

C.1.2. In addition, an assessment of depth and liquidity has to be performed both for the reference
instruments as well as for the general bond market.

C.1.3. A market is assumed deep if transactions involving a large quantity of financial instruments
used in the replications can take place without significantly affecting the price of the
instruments. Conversely, a market is liquid if financial instruments can readily be converted
through an act of buying or selling without causing a significant movement in the price.

C.1.4. There are a number of methods to measure whether a market fulfils the aforementioned
definitions. A non-exhaustive list of indicators for the assessment of depth and liquidity in a
market is described below:

• Bid-ask spread: the price difference between the highest price a buyer would pay and the lowest
price for which a seller would settle
• Trade frequency: number of trades that take place within a defined period of time
• Trade volume
• Trader quotes/dealer surveys (incl. dispersion of answers)
• Quote counts (1): number of dealer quotes within a few day window
• Quote counts (2): number of dealers quoting
• Number of pricing sources
• Assessment of large trades and movement of prices (depth)
• Residual volume approach (for bonds only)

Page | 22
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

C.1.5. For the bond market, the assessment includes an analysis of the ability of insurers to match
their insurance liabilities with bonds. Where it would no longer be possible for insurers to
match insurance liabilities with bonds of the same currency, this is reflected in the last liquid
point. The assessment of the reference instrument(s) and the general bond market are
performed independently of each other.

C.1.6. The consideration of a number of different measures/indicators takes account of a number of

issues. First, liquidity (including depth) is an abstract concept that essentially measures to
what extent supply meets demand. There are essentially three principal dimensions to
liquidity: depth (the amount of trade volume that can be executed without impacting price),
tightness (or breadth; the ease of purchase/sale or the ratio of suppliers/demanders, typically
measured by the bid ask spread) and resilience (the amount of time before prices return to
pre-large trade levels. None of the proxies is able to cover perfectly all dimensions of liquidity,
notwithstanding the fact that the indicators may also be “diluted” by factors unrelated
to liquidity.

C.1.7. Second, markets can considerably differ, e.g. in size, and one all-encompassing methodology
may not appropriately capture this difference. Furthermore, a single indicator may not
capture well enough new market developments. For the aforementioned reasons, common
thresholds for all currencies are also not appropriate and may be inconsistent with some of
the general requirements for the risk free rate, such as robustness, practicability, or incentive
effects. The analysis shall have regard to the specificities of the market and apply expert
judgement where appropriate.

Page | 23
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

Appendix D. Market Data Sourcing

To determine the basic risk-free term structure, considerations have to be made in respect of the
availability and the relevance of data. In addition, those data have to be adjusted for the inherent credit
risk or take into account the peg to another lead currency, if applicable .

Under proposed IFRS 4 Phase 2 “risk free interest rates” can be derived either using the “bottom up
approach” using relatively risk free Government bond data or the “top down approach” using other risky
financial instruments adjusted for risk not relevant to the insurance liability cashflows.

The actuary would consider excluding any bonds with characteristics that render the bond inappropriate
for purposes of matching the timing and amount of expected liability payments. For example, the
actuary would consider excluding bonds with one or more of the following features: callable (unless the
call option includes a make-whole provision or the actuary is comfortable that the call option does not
have a material effect on the bond price), putable, convertible, sinkable, extendable, perpetual, variable
coupon, and inflation linked.

Ideally for the development of a “Risk free” zero rate curve, the instrument on which the relevant risk-
free interest rate term structure is based should also have the following qualities 14:

• No credit risk
 Relatively immaterial credit risk or sovereign risk
• Realism
 It should be possible for the institutions/funds to earn the specified risk-free rate in a risk free
manner. Otherwise, liabilities will incorporate hidden losses which would materialize during the
run-off period of the liabilities
• Reliability
 Structure should be robust. It should result in a reliable estimate. This criterion should in
particular apply in time of market crisis or turbulent
• Highly liquid for all maturities
• The rates should be based on financial instruments for which a reliable market value is observable
from deep, liquid and transparent markets

Deep liquid and transparent markets can be described as markets whereby

• Market participants can rapidly execute large-volume transactions with little impact on the prices of
the financial instruments used in replications

14 CEIOPS L2 Final Advice on TP Risk Free Rate -2010 3.1

Page | 24
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

• Current trade and quote information of those prices is readily available to the public

Below we recommend the use of data in the follow order of priority15.

First option:

If government bonds are available that meet the criteria (a) to (e) defined in section 3.1.2 (“risk-free rate
criteria”), then government bonds should be used to determine the relevant risk-free rate.

Second option:

If government bonds are available, but they do not meet the risk-free rate criteria, then they should be
adjusted for their deficiencies relating to these criteria. The adjusted rates should approximate
government bond rates which meet the risk-free criteria. The adjusted rates should be used to
determine the relevant risk-free rates.

Third option:

If government bonds are not available or if government bond rates cannot be adjusted to meet the risk-
free rate criteria for practical or theoretical reasons, other financial instruments should be used to
derive the risk-free interest rates to the extent they exist or are available. These instruments should be
as similar to government bonds as possible. Their rates should be adjusted for credit risk and any other
deviations from the criteria with the objective to approximate government bond rates which meet the
risk-free criteria.

The second option is the only available option for the Caribbean region. For most of the member regions
in the Caribbean the preferred source for the interest rate will be the Government bonds/securities in
each currency. While the Government bonds in most Caribbean jurisdictions would not pass both the
“no credit risk” “highly liquid criteria”, there are no other easily accessible/permissible asset types in the
region which are more liquid or representative of the insurance contract cashflows 16. As such they are
the most common assets supporting actuarial liabilities. For life insurers, Government securities
comprise over 70% of most of the asset portfolios.

Recommendation: The use of Government bonds (domestic or foreign) adjusted for sovereign risk,
if applicable.

15 CRO Forum
16 Corporate bond/Private Placements – less liquid with lower volumes, low availability Mortgages – callable, higher credit risks

Page | 25
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

Appendix E. A Technical Note on the Smith-Wilson Method

The following link provides resource material on this topic.

http://www.ressources-actuarielles.net/EXT/ISFA/fp-
isfa.nsf/2b0481298458b3d1c1256f8a0024c478/bd689cce9bb2aeb5c1257998001ede2b/$FILE/A_Techni
cal_Note_on_the_Smith-Wilson_Method_100701.pdf

Page | 26
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

Appendix F. Valuation Rates – Understanding the Basic definitions

and relationships
F.1.1. Zero rates and discount factors

F.1.2. The n-year zero rate, rn, is the market rate of interest earned on an investment that starts
today and lasts for n-years where the entire interest and principal amount is realised at the
end of the of n-years. The zero rates provided are continuously compounded rates. From a
future value perspective, a continuously compounded n-year zero rate, rn , is defined such that,
in return for an investment of P dollars today the investor receives, P*e rn*n , in n years. This is
illustrated in the figure below.

R
rn *n
P *e P
Time

P
0 n

Principal Interest

Page | 27
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

F.1.3. The zero rate curve is constructed by interpolating between annual zero rates, as is illustrated
in the graph below.

US$ Swap Zero Curve January 2005

6%
Rate

4%

2%
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

Maturity

Spot Swap Curve

F.1.4. Used for pricing, zero coupon rates are translated into a present value perspective through
the Discount Factor, Define the n-year discount factor D n as the value today (present value) of
a dollar which is payable at the end of year n, assuming continuous compounding. Under this
definition, we can write the discount factor as a function of the spot zero coupon rate, e.g.

Dn  e rn *n

In the past, some insurers/pension actuaries have used the same interest rate to value a series of future
cashflows due at different times in the future; using continuously compounded rates, this practice

would be equivalent to
Dn  e  r*n

The practice is to use the appropriate zero coupon rate, and therefore discount factor, for each cash

flow maturity, e.g.

Dn  e  rn *n , as is illustrated below:

Page | 28
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

Example 1: Discounting using Zero Curve

Timing of Expected Present Value Appropriate Zero Rate% Discount Factor, Dn

Cashflow Future of Future rn, (continuously
D n  e  rn *n
(n years) CashFlow Cashflow compounded)
0.5 10 9.876 r0.5= 2.50% 0.98758
1.0 20 19.390 r1.0 =3.100% 0.96948
2.5 30 27.428 r2.5= 3.586% 0.91425
3.0 40 35.800 r3.0= 3.698% 0.89500
Total 92.494

To calculate the present value of the first cashflow of 10 we discount it at 2.5% for 6 months; to
calculate the present value of the second cashflow of 20 we discount it at 3.1% for 1 year and so on.
Therefore, the value today, V0, of these cashflows is

 0 .0 2 5*0 .5  0 .0 3 1*1 .0  0 .0 3 59*2 .5  0 .0 37 0*3

V
0
 10 * e  20 * e  30 * e  40 * e = 92 .4 9

Another alternative to calculate the present value of these cashflows is to first derive the corresponding
discount factors for each maturity from the zero curve provided (see column 4 of table above) and then
multiply each of the individual liability cashflows in each period by the appropriate discount factor, Dn.
That is:

V0 = 10 * D0.5 + 20 * D1.0 + 30* D2.5 + 40* D3.0 = 92.49

Page | 29
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

F.1.5. Forward Interest Rates and Discount factors

F.1.5.1. Forward interest rates are the future rates of interest implied by the current zero rates which can
be locked in today. They can be used to calculate the value at a future time point of a series of
cashflows, this future value is also referred to as the forward value.

F.1.5.2. In this section we will first describe forward interest rates in more detail and then explain how they
can be used to calculate the forward value of a series of future cashflows.

F.1.6. What are Forward Interest Rates

F.1.6.1. The table below describes the notation for forward interest rates at annual maturities. The current
zero rates are in the first row and the forward rates begin in the second row of the table. The
forward rates constitute a matrix of n columns and m rows.

F.1.6.2. The n*m forward interest rate is defined as the interest rate which can be locked in today and is
payable on a zero-coupon bond that commences n years from the current date and matures in
m years.

Maturity (year) 1 2 3 4 5 …..M

Time 0 (Current) Zero r1 r2 r3 r4 r5 rm
Curve
(n* m) forward interest rates
Time 1 Zero Curve f11 f12 f13 f14 f15 f1M
Time 2 Zero Curve f21 f22 f23 f24 f25 f2M
Time 3 Zero Curve f31 f32 f33 f34 f35 f3M

: : : : : : :
Time N Zero Curve fN1 fN2 fN3 fN4 fN5 fNM

Page | 30
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

F.1.6.3. In the matrix above of forward rates, the rows represent the implied zero curves in future years.
For example, the Time 2 zero curve represents the implied zero curve in 2 years. Hence, f 24
represents the implied 4 year zero rate in 2 years. The columns are referred to as the n-year
forward curves. Hence, the first column is the 1-year forward curve which represents all future 1-
year zero rates (fn,1 for the next N years). For example, f31 represents the implied 1-year zero rate
in three years.

F.1.7. How to calculate forward values

F.1.7.1. A forward valuation is very similar to a time 0 valuation whereby each individual future cashflow is
multiplied by its appropriate discount factor and these values are summed over all cashflows.
However, in a forward valuation, the appropriate discount factors are the forward discount factors,
Dn,m, where Dn,m is the forward value at time n of one dollar payable at time n+m. This is illustrated
in the following exhibit.

R
f m , n *m
P *e P
Time

P
0 n n+ m

Principal Interest

F.1.7.2. There are three different, but equivalent, ways of calculating the forward discount factor. The first
defines the forward discount factors, Dn,m using the spot discount factors such that

Dnm
Equation (1) D n ,m 
Dn

Page | 31
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

F.1.7.3. However, many existing models use forward rates to derive the forward discount factors. There
are two additional approaches to do this calculation and both are equivalent to the equation 1
above. The second approach derives the appropriate discount factor from the Time n zero rates,
such that,

 f n,m*m
Equation (2) Dn,m  e

F.1.7.4. The final approach derives the discount factors from the 1-year forward rates, such that

n m  f *1
Equation (3) Dn,m   e i,1
i n

F.1.7.5. All three equations will generate the same forward discount factors.

The current zero rates and the entire matrix of forward rates at annual maturities will be provided.
Hence, the user will be able to use any of the above 3 equations to calculate forward values.

Example 1: Calculation of Forward Rates

Maturity (year) 1 2 3 4 5 6 7
Time 0 Zero Curve 3.100% 3.475% 3.698% 3.891% 4.066% 4.210% 4.355%
Time 1 Zero Curve 3.849% 3.996% 4.155% 4.307% 4.432% 4.565% 4.663%
Time 2 Zero Curve 4.143% 4.308% 4.459% 4.578% 4.708% 4.799%
Time 3 Zero Curve 4.472% 4.617% 4.723% 4.849% 4.930%
Time 4 Zero Curve 4.763% 4.849% 4.974% 5.045%
Time 5 Zero Curve 4.935% 5.080% 5.138%

The forward value in 2 years of a dollar payable 5 years from today is represented as D 2,3. Using the
valuation rate table above this value is calculated under all 3 equations as follows:

D5 e4.066%*5 0.81605
Equation (1) D2,3     0.8748
D2 e 3.475%*2 0.9328

 f 2 ,3 *3
Equation (2) D2,3  e  e 4.459%*3  0.8748

 f 2 ,1 *1  f 3,1 *1  f 4 ,1 *1  f 5,1 *1
D2,3  e *e *e *e
Equation (3)   e 4.143% * e 4.472% * e 4.763% * e 4.935%
  0.8748

Page | 32
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

Page | 33
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

Appendix G. The Sovereign rate adjustment to the Extrapolated

Zero Curve
G.1.1. Sovereign risk represents the country risk and the credit risk of the country. There are
different views on the CAA Committee related the reflection of sovereign risk . There is no
emerging consensus thus far.

G.1.2. One view – Sovereign risk is not characteristic of the insurance liability cashflows and should
be excluded

G.1.3. Alternate view – needs to be included as not an insurable risk by private sector insurance
companies in emerging markets where investment Government bonds reflect over 50% of the
invested asset portfolio primarily due to regulations

G.1.4. It is also not clear where IFRS4 Phase 2 will land on this issue at this point in time. At a very
high level, to the extent that the characteristics of the cashflows of the relevant insurance
contracts require the use of “risk free” rates for the purpose of discounting, the methodology
developed will need to remove the embedded sovereign risk premium.

G.1.5. Using this principle, arguments may also be made that sovereign risk is characteristic of the
cashflows of certain insurance cashflows and the sovereign risk premium should be reflected.
Further research and discussion is required to establish criteria and methodology for the
final standard.

G.1.6. In the interim, for the purpose of the technical assessment we propose that two curves be
created and tested for each jurisdiction, one curve reflecting the sovereign risk premium, the
other curve deducting the sovereign risk premium.

G.1.7. Sovereign risk represents the country risk and the credit risk of the country. Simply put, the
sovereign risk premium is the difference between the yield of the risk-free triple-A rated
government bond (or double A US bond yields) and a bond issued by the local government
(with the sovereign risk embedded in it) minus the inflation differential of the two currencies
involved. If local bonds are issued in US$, the inflation differential should not be deducted.

Double A risk free Inflation

Sovereign risk Local Govt
= + US bond yield … differential
premium bond yield
(US$) (US$/Local)

Page | 34
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

G.1.8. It is likely that the estimated sovereign risk premium exceeds the actual average expected
losses from sovereign default and that increases in the premium in stress conditions are
driven by increased illiquidity, rather than default, expectations which may not be
appropriate for the purposes of the development risk free curve.

G.1.9. Recommendation – Explore alternative methods to estimate the risk sovereign default which
do not reflect illiquidity premium.

Page | 35
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

Appendix H. Overview of the rational for the matching rate

adjustment/Illiquidity premium to the Extrapolated
Zero Curve17
H.1.1. The insurance industry has for many years highlighted the need for the discount curve to take
into account the long-term nature of insurance business.

H.1.2. The method for developing the curve needs to fully recognise the very significant difference
between the credit risk faced by a company which trades bond-like assets and one which,
because of its long-term liabilities and long-term investment strategy, holds those assets
to maturity.

H.1.3. A company trading the assets is exposed to the full market value movements, while one
holding to maturity is only exposed to the defaults This matters because the risk of losses due
to defaults varies significantly less than the movement of credit risk reflected in market values
(spread movements), as we have seen very clearly in the recent crisis.

H.1.4. Failure to address this risks driving the industry away from offering long-term guarantees and
long-term investment strategies, with major unintended consequences for policyholders,
markets and the industry.

H.1.5. The matching adjustment (MA) is a mechanism that prevents changes in the value of assets,
caused by spread movements, from flowing through to companies’ balance sheets for
portfolios where companies have fully or partially mitigated the impact of these movements
This prevents non-economic (artificial) volatility and pro-cyclicality in Own Funds The MA
adjusts the best estimate liabilities to ensure that where insurers may need to sell such assets
to meet their unpredictable liabilities, they are exposed to these short-term asset value
fluctuations; but not where they hold the assets to maturity In order to assess this exposure,
the MA considers the extent to which the company is exposed to “losses on forced sales”.

H.1.6. Losses on forced sales describes the situation where a company may be forced to sell assets
that it had intended to hold to maturity in order to cover unexpected liability payments (such
as lapses), and hence realises any losses in the value of those assets since they
were purchased.

17 The Matching Adjustment Theory and practice –Insurance Europe

Page | 36
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

H.1.7. Without the MA, where the company is not exposed to losses on forced sales, asset
movements cause unrealised losses or gains to impact the balance sheet. This would create
artificial volatility Insurers are not exposed at all to losses on forced sales in the following
situations: When insurers are not exposed at all to forced sales because liabilities are
predictable and the timing of asset cash flows enables the timely payment of liability cash
flows, e.g. for annuities When insurers are exposed to forced sales but can pass on the
potential losses to policyholders, e.g. via market value adjustment (MVA) clauses Additionally,
there are situations where insurers are only partially exposed to losses on forced sales.

H.1.8. For some products, such as annuities, the exposure to spread movements can be completely
eliminated where there are no options for policyholders to lapse and where deaths do not
trigger a liability cash flow Where the company uses an investment strategy under which it
has the intention and ability to hold the assets to maturity, there is no risk that the company
will need to sell assets early and potentially incur a loss from the forced sale In this situation,
the company is only exposed to losses on assets due to defaults in respect of these products
For products such as this, the full MA in respect of the backing assets can be applied.

H.1.9. There is strong evidence that credit spreads exceed the actual average losses from default
and that increases in these spreads in stress conditions are driven by increased illiquidity,
rather than default, expectations Many studies of historical data and market-consistent
modelling support this. None that we are aware of provide evidence against it.

H.1.10. A general approach for the calculation of the MA can be determined by considering which
assets are not exposed to losses on forced sales. This approach has been determined by an
industry working group, based on high-level principles agreed by Insurance Europe, the CFO
Forum and the CRO Forum. See attached presentation.

Page | 37
DRAFT
WORK IN PROGRESS
Draft Recommendations for the development of the Risk Free Spot Zero rate curve

Appendix I. The estimation of the interest rate term structure in

markets with infrequent trading
The following link provides resource material on this topic.

https://www.researchgate.net/publication/5115185_Term-
structure_Estimation_in_Markets_with_Infrequent_Trading

Page | 38