May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

DYNAMIC PORTFOLIO INSURANCE

WITHOUT OPTIONS
9
DOMINIK DERSCH
Rumfordstr. 6, 80469 München, Germany
[email protected]

Dynamic portfolio strategies are an interesting alternative to classical option-based

investment and protection strategies. One of the most prominent techniques is Con-
stant Proportion Portfolio Insurance (CPPI). In this chapter, we provide a review of
various techniques and formulate a general framework for investment and protection
strategies. The common feature of this strategy is that it empowers the investor to repli-
cate various option like pay-off profiles without the usage of options. These strategies
may replicate a simple floor type or advanced path-dependent look-back options that
implement all-time-high strategies with a given participation rate. We illustrate the
different strategies that employ features like various types of lock-in, trailing, lever-
age, and risky portfolio strategies with historical simulations. We include features
that allow the simulation under realistic market conditions taking into account trans-
action costs and the avoidance of excessive rebalancing through transaction filters.
We discuss the use of exchange traded funds (ETF) to invest in broadly diversified
multi-asset portfolios. The goal of this chapter was to illustrate different protection
strategies and to show how a practical implementation of these strategies could look
like. This chapter can serve as a guideline for simple spread-sheet models.

201
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

202 Dersch

9.1. INTRODUCTION

Institutional investors who require portfolio protection or a minimum absolute per-

formance target are facing the problem to trade off risk and return. On the one hand,
exposure to high return risky asset classes is desirable; on the other hand, the downside
potential that accompanies higher risk should be kept at the investor’s level of com-
fort. Holistically, the tasks of asset allocation and risk management are closely linked
and should not be performed in two separate steps, as this leaves the investor most
likely with sub-optimal solutions. However, a fairly large arsenal of risk management
techniques has been developed that work completely independent of the investment
strategy and respective asset classes. The sole requirements are that the asset classes are
investible and allow — at least in theory — a liquid market and continuous trading. The
selection of the underlying depends on the specific requirements of the investor with
respect to holding period, taxation, investment guidelines, etc., and the asset class
itself. Examples of different underlyings are direct investments in stocks or bonds,
futures, and funds. It is important to note that the investment universe and the respec-
tive protection strategy have to be closely aligned. Investments with large bid–offer
spreads would require a strategy with a low reallocation frequency.
For the sake of simplicity, this chapter focuses on this two-step approach described
above: Step 1 asset allocation, Step 2 risk management framework. Throughout this
chapter, the first step of asset allocation is simulated by investing in a set of indices.
The performance of each investment is assumed to follow the performance of the index
time-series. The considered investment universe includes stock, bond, commodity, and
hedge fund indices.
Our choice of the investment universe poses no restriction on the second step of
the investment process — the risk management framework. Within the fairly general
requirements of “investability” and liquidity, any asset classes and (propriety) trading
system may be embedded in this risk management framework.
The recent history of financial markets posed a huge challenge to portfolio man-
agers as reflected in massive declines in asset values, historical highs in volatility, and a
break down in correlations. Dynamic portfolio strategies are an interesting alternative
to classical option-based investment and protection strategies that allow one to cope
with the market turmoil. One of the most prominent techniques is Constant Proportion
Portfolio Insurance (CPPI) [2, 3]. CPPI went out of fashion because of a number of
drawbacks like the fixed time horizon, the inability to take profits and recover from
a major draw-down. One of the major shortcomings — the pro-cyclical behavior —
has been blamed for huge market movements and the stock market crash of 1987.
However, a number of advanced features allow one to overcome the shortcomings
of the first generation model and allow a practical application in asset management.
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

Dynamic Portfolio Insurance Without Options 203

A comparison of option-based strategies — mentioned above — and CPPI is shown

in [1].
The mature market for exchange traded funds (ETF) gives access to a huge universe
of different asset classes that may be traded with high liquidity and low bid–offer
spreads. This enables the investor to apply advanced portfolio insurance strategies as
described above and investment in a broadly diversified universe.
The remainder of this chapter is structured as follows. In the next two sections,
we review simple portfolio strategies and plain vanilla CPPI and illustrate them with
historical simulations. In Sec. 9.4, we formulate a more general framework of CPPI
with various features like different types of lock-in, trailing, leverage, and risky port-
folio rebalancing strategies. We include features that allow simulation under realistic
market conditions taking into account transaction costs and the avoidance of exces-
sive rebalancing through transaction filters. We further show that our framework also
includes an extension of CPPI named TIPP [4]. In Sec. 9.5, the strategies are illustrated
using historical simulations. In particular, we investigate how the different strategies
cope with the historical market evolution. In Sec. 9.6, we discuss the use of exchange-
traded funds (ETF) to implement protection strategies. This chapter concludes with
final remarks on different risk transfer mechanisms of option strategies versus dynamic
portfolio strategies.

9.2. SIMPLE STRATEGIES

9.2.1. Buy-and-Hold
Probably the most simple and most common risk management framework is the buy-
and-hold strategy. Not just among retail investors, either deliberately or not deliberately
buy-and-hold is a widespread approach. Many ambitious strategies will eventually drift
into passive sit-and-wait strategies as the investor sits out long periods of negative
performance or hesitates to take profits on time. Cheekily, buy-and-hold or strategic
trades very often stem from short-term tactical trades turned bad. The only reason why
this strategy remotely qualifies as some kind of protection strategy is that in the absence
of leverage the total loss is limited to the initial investment. However, buy-and-hold
has performed well over long periods and across many asset classes.

9.2.2. Stop-Loss
A stop-loss strategy is the first non-trivial step toward a risk management framework.
Here, we distinguish between an investment target IT on a present value base and a
target on a given time horizon T . In the first case, the position has to be switched into a
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

204 Dersch

risk-free investment matching the investment horizon if the portfolio value falls below
the target value at any time
PVportfolio (t) > IT . (9.1)
In the second case, the portfolio value must not fall below
PVportfolio (t) > e−r(T −t) · IT (9.2)
or equivalently
PVportfolio (t) − e−r(T −t) · IT > 0. (9.3)
Here, T − t is the time until the investment horizon is reached and r is the risk-free
rate of this time span. The second case is slightly more complex. Here, we must
ensure that the investor reaches his/her investment target at maturity. Therefore, the
portfolio value must not fall below a certain floor value given by the right hand side
of Eq. (9.2). In case, the stop-loss level is reached, the portfolio must be liquidated
and invested in the risk-free asset. Investing the floor value with the risk-free rate will
ensure the given target value at maturity. The implementation of a stop-loss strategy
requires monitoring both the portfolio performance and the total return of the risk-
free investment. The risk-free investment is usually implemented with treasury bills or
bonds with a maturity matching the investment horizon. The re-investment of coupons
paid until maturity must also be considered. In case, the stop-loss level is reached,
a single portfolio re-allocation occurs. Strictly speaking, stop-loss is therefore not a
dynamic strategy.
In the above analysis, we assume that the investments and the target level are in
the same consolidation currency. Otherwise, the respective FX spot or forward rates
have to be additionally monitored. Please note that a stop-loss strategy with a target
value of zero is the same as the above buy-and-hold strategy.

9.2.3. The Bond Floor Strategy

The stop-loss strategy bears the risk that the portfolio value may be exposed to large
volatility. In addition, it carries a short fall or gap risk. This is the risk that the stop
level is missed in large market movements and the investor is left with a final portfolio
value below the target. The bond floor strategy takes a more cautious approach. Here,
we only invest in the risky portfolio the amount we are willing to lose in the first place.
Let us take Eq. (9.3) at the beginning of the investment horizon
PV portfolio (t = t0 ) − e−rT · IT = C. (9.4)
C is the amount given by the difference between the initial investment and the present
value of the target amount. Here, we assume an investment in a risk-free bond with a
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

Dynamic Portfolio Insurance Without Options 205

maturity matching the investment horizon. The bond floor strategy is usually imple-
mented with a target value close to 100%. According to the above equation, a given
investment of EUR 100, a target value of EUR 101 in one year, and a risk-free rate of
2.5% would leave us with a risk budget of EUR 1.5 and a bond investment of EUR 98.5.
That means EUR 1.5 is invested with a buy-and-hold strategy in a risky investment.
A total loss of the risky investment would still guarantee the target amount of 101 EUR
in one year. For a target value close to 100%, the bond floor strategy — as compared
to the stop-loss strategy — swings the pendulum in the other direction of extreme risk
aversion with the consequences of little upside potential beyond the target value. There
are two things worth mentioning: The bond floor must not be larger than the amount
that may be earned with the risk-free investment and a bond floor of zero is the same
as the above buy-and-hold strategy.

9.2.4. Plain Vanilla CPPI

CPPI tries to bridge the gap between high risk and high risk aversion of the above strate-
gies. The Constant Proportion Portfolio Insurance technique [2] was first introduced
by Black and Jones in 1987. It may be seen as a further generalization of stop-loss
and bond floor and it is literally a dynamic strategy. Similar to both strategies, a target
level and an investment horizon are given. In contrast to the bond floor strategy, we
assume that a total loss of the risky investment is highly unlikely. We rather accept
that the risky investment may fall by a factor of 1/M within a given time horizon. This
means that if we invest twice (e.g., M = 2) the amount given by the right hand side
of Eq. (9.3) in the risky portfolio, we still meet the investment target if the investment
will lose less than 50% of the initial value. This is the key idea of CPPI. The strategy
may be summarized by the following set of steps:
(1) Calculate the current risk budget C(t) similar to Eq. (9.4) according to
C(t) = PV portfolio (t) − e−r(H−t) · T, (9.5)
with PV portfolio (t = t0 ) = N0 .
(2) Calculate the exposure E(t) invested in the risky portfolio according to
E(t) = M · C(t). (9.6)
(3) Rebalance the portfolio by investing the amount E(t) in the risky portfolio and the
remaining amount in the risk-free investment.
(4) Wait until !t has passed and go back to Step (2).
(5) Repeat the above steps until the end of the investment horizon T is reached. In
case C(t) according to Eq. (9.5) is zero, the portfolio will be allocated in the risk-
free investment until the end of the horizon. In that case, only the target value is
achieved.
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

206 Dersch

Table 9.1 Comparison of Simple Protection Strategies and Their Relation to CPPI.

Strategy Risk appetite Upside potential Short fall risk Relation to CPPI

Buy-and-hold High High No CPPI with zero

target
Stop-loss Medium, depends High, depends on Yes CPPI with very
on stop level stop level large multiplier
and target equal
to stop-loss level
Bond floor Low, depends on Low, depends on No CPPI with target
bond floor level bond floor level level equals bond
floor and
multiplier M = 1
CPPI Low, depends on Medium, depends Yes, but small Yes
target level on target level

Depending on the parameterization, CPPI allows one to implement a given risk

appetite: The larger the multiplier and the lower the target level, the higher the risk
appetite. It is easy to show that CPPI contains the above strategies buy-and-hold,
stop-loss, and bond floor for different settings of target level and multiplier.
The discounted target level is also called the floor
F(t) = E−r(T −t) · IT. (9.7)
The floor is the present value of the target.
Table 9.1 summarizes the risk and reward characteristics for the above strategies
and shows the link to CPPI.

9.3. HISTORICAL SIMULATION I

In this section, we show sample simulations for the above protection strategies. Our
risky portfolio is the Dow Jones Euro Stoxx 50. The simulation covers a period of
close to 10 years. For comparative reasons, we use similar protection levels where
applicable.
• Table 9.2 summarizes the simulation parameter.
• Figure 9.1 shows the buy-and-hold strategy.
• Figure 9.2 shows the stop-loss strategy.
• Figure 9.3 shows the bond floor strategy on the DJ Euro Stoxx 50 index.
• Figure 9.4 shows the CPPI strategy with a target level of 100% and a multiplier of
6. The floor is calculated using a discount rate of 3.25%.
• Table 9.3 summarizes the simulation results.
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

Dynamic Portfolio Insurance Without Options 207

Table 9.2 Simulation Parameters of Simple Strategies.

Simulation period 4 January, 1999 — 12 December, 2008

Investment EUR 100 mn
Risk-free and risky investment EUR Overnight liquidity, DJ Euro Stoxx 50
Target 100% (except buy-and-hold)
Discount rate 3.25% (except buy-and-hold)
Multiplier 6 (CPPI only)

180

160

140

120

100

80

60

40

20

0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 9.1 Buy-and-hold strategy on the DJ Euro Stoxx 50 performance index for an initial
portfolio value of EUR 100 mn. The performance of the portfolio (in EUR mn) is the performance
of the index.

In this market environment, the bond floor strategy seems to perform best over the
complete period both in the absolute return and the size of the worst draw-down. The
three other strategies all reach the same maximal portfolio amount of EUR 156.3 mn
(up 56%), but fail to benefit at maturity. The stop-loss and CPPI strategy show a very
similar picture. They are both stopped out during the sharp market decline in mid-2002
and realize a slight loss as compared to the initial portfolio value. This translates into
a slightly negative annual return of −0.17% (CPPI) and — 0.30% (stop-loss). Buy-
and-hold ranks last with respect to final portfolio amount annual return, and suffers
the worst portfolio draw-down of 65% of the previous all-time high value.
Stop-loss and CPPI both fail the target by a small amount. This may be due to two
reasons. The first is fundamental and is caused by rapid market movements when the
position may only be liquidated below the theoretical stop level. We can reduce this
fundamental risk by monitoring the position intraday. The second reason is caused by
a simplification of our simulation framework. Monitoring the stop-loss or floor level
requires one to monitor the zero coupon bond with a maturity equal to the remaining
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

208 Dersch

180
DJ EUR STOXX 50 risk-free
160

140

120

100

80

60

40

20

0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 9.2 Stop-loss strategy on the DJ Euro Stoxx 50 performance index with a target level
of 100% at maturity. The risk-free investment is EUR overnight liquidity. The y-axis shows the
portfolio value in EUR mn.

160
DJ EUR STOXX 50 risk-free
140

120

100

80

60

40

20

0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 9.3 Bond floor strategy on the DJ Euro Stoxx 50 index with a target level of 100% at
maturity. The risk-free investment is EUR overnight liquidity. The y-axis shows the portfolio
value in EUR mn.

investment horizon. As a simplification, a fixed discount rate is used in our simulation

framework to calculate the stop-loss level and floor. Our simplified risk-free investment
is EUR overnight liquidity rather than the corresponding zero bond with matching
investment horizon. We may fail to reach the target if the overnight investment fails to
earn — over the remaining investment horizon — the return implied by the floor level.
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

Dynamic Portfolio Insurance Without Options 209

180
DJ EUR STOXX 50 risk-free floor
160

140

120

100

80

60

40

20

0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 9.4 CPPI strategy on the DJ Euro Stoxx 50 performance index with a target level of
100% at maturity. The risk-free investment is EUR overnight liquidity. The floor (black line) is
assumed to follow an annual rate of 3.25%. The y-axis shows the portfolio value in EUR mn.

Table 9.3 Historical Simulation of Four Different Protection Strategies Using the
Dow Jones Euro Stoxx 50. We Calculate the Annual Return Over the Simulation
Period and the Worst Draw-Down. A Draw-Down of, e.g., 65% Means That the
Portfolio Lost 65% of Its Previous All-Time-High.

Strategy Buy-and-hold Stop loss Bond floor CPPI

Ann. return −1.69% −0.30% 2.03% −0.17%

Reached protection NA Yes No Yes
Worst draw-down 64.64% 48.47% 27.26% 47.34%

9.4. ADVANCED FEATURES

The above simulations illustrate different protection strategies. However, they are not
suitable for practical use for a number of reasons:

• Transaction costs may have an impact on the real world performance.

• Frequent rebalancing may cause excessive transaction costs and should therefore
be constrained by transaction filters.
• Investors require more sophisticated protection strategies like lock-in of gains.
• Investors may wish a leveraged exposure to the risky portfolio.
• For non-trivial risky portfolios, e.g., more than one risky asset, different rebalancing
strategies for the risky portfolio may be applied.
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

210 Dersch

In our simulation environment, we implemented a fairly general framework that takes

into account the above features.

9.4.1. Transaction Costs

Independent of asset classes and markets, transaction costs is a relevant factor that may
significantly impact the performance of a trading system. Trading systems that look
good on paper, e.g., paper trading, may fail in reality because transaction costs were not
considered. The impact of transaction costs is more dominant for frequent trading and
for long investment periods. The latter is important because of the compounding effect.
Transaction costs that reduce the portfolio amount in an early stage are no longer
available for future investment. The impact is hard to estimate and therefore has to be
simulated.
In our simulation environment, we model transaction costs as a percentage of the
transaction volume. Different transaction costs can be set for the risk-free and the risky
assets and for buying and selling the asset. The bid–offer spreads of ETF, for example,
vary widely from 2 bp to up to 100 bp depending on the asset class and time. This
corresponds to a transaction cost of 0.0001–0.005 times the transaction volume. The
transaction cost is only half because the bid–offer spread is paid on the full round trip
to get in and out of the asset.

9.4.2. Transaction Filter

With the ability to model transaction costs, the impact can be analyzed and optimized.
Here, we have to trade off flexibility versus rigidity — rapid adaptation to changes
in the market environment on the one hand with the downside of a large number of
transactions and high transaction costs. On the other hand, less frequent trading reduces
the transaction costs but poses the risk that the system is not flexible to respond to large
market moves. The introduction of transaction filters ", λ, and !t allows the investor
to trade off these two effects.
Strictly speaking a rebalancing in the plain vanilla CPPI is required whenever
Eq. (9.6) is violated. We calculate the deviation after time t = t " + !t has passed
according the following equation:
E(t)
1 − "buy ≤ target ≤ 1 + "sell . (9.8)
E (t)
A rebalancing is performed only if the target exposure Etarget (t) deviates from the
current exposure E(t) beyond the given boundaries. Typically, the parameter " for buy
and sell are in the range of [0,0.2], with "sell slightly smaller than "buy , in order to react
quicker to a market downturn. Larger " defines a higher threshold for rebalancing,
" = 0 requires instant rebalancing.
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

Dynamic Portfolio Insurance Without Options 211

The volatility filter λ scales the size of the rebalance — either buy or sell — if a
rebalancing is triggered according to Eq. (9.8).
!E = λ · |Eold (t) − Enew (t)|, (9.9)
with λ in the range of [0.7,1.0]. For λ smaller than one, we follow only a fraction of
the rebalancing amount. The idea behind the volatility filter is that the market trades
in a range rather than follows a trend.
The re-observation period !t is also a transaction filter: The larger !t, the more
time elapses between a test of Eq. (9.8). The re-observation period has to be in line
with ". Small " and a large !t or " very close to one and small !t do not fit well
together. Typically, !t is in the range of one day. This implies that we test Eq. (9.8) at
close, but ignore intra-day movements outside the range defined by ". This reflects the
observation that the intra-day volatility is typically larger than the day-to-day volatility.

9.4.3. Lock-in Levels

The plain vanilla CPPI does not protect any gains. To remedy this weakness, different
strategies to lock-in gains by raising the target level have been proposed. The obser-
vation period for lock-in may be given by !t, its multiples (k · !t), or by any other
discrete lock-in dates like every week or month.
To implement lock-in, we have to distinguish between the lock-in trigger and
the lock-in action. The trigger can be a simple trigger in time as described above, or
a trigger caused by a certain portfolio level or given by both. The different lock-in
actions are performed if a lock-in trigger is reached. They are summarized as follows:
• Discrete lock-in steps: X% gain in the portfolio amount is locked in by raising the
floor defined in Eq. (9.7). The discrete lock-in steps may refer to the fraction of the
initial notional at the beginning of the investment period, like every EUR 100,000
or to the notional at the previous lock-in level (compounding lock-in), like 10% of
the portfolio amount at the last lock-in.
• Newly reached all-time-high levels are locked in by raising the floor with respect to
the previous high level.
• Trailing: Y % gain in the portfolio amount is locked in by raising the target level
by the given gain. The discrete lock-in steps may refer to the initial target level
at the beginning of the investment period or refer to the previous lock-in level
(compounding lock-in).
• All-time-high trailing: Newly reached all-time-high levels are locked in by raising
the target level with respect to the previous high level.

The difference between the first two and the last two lock-in actions is that the lock-in
is applied to the floor level rather than the target level.
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

212 Dersch

The advantage of a lock-in is that gains that have been reached in the past are
protected. On the other hand, a lock-in may also cause a reduction in the exposure
due to a rising floor level and therefore limits participation in a future rise in the risky
portfolio.

9.4.4. Leverage and Constrain of Exposure

The lock-in levels discussed above support the requirement to protect gains in the risky
portfolio. On the other hand, investors may request further upside potential through
higher exposure to the risky portfolio. We may introduce leverage by rewriting Eq. (9.6)
E(t) = min{Emax (t), M · C(t)}. (9.10)
In the absence of leverage, we define
Emax (t) = PV portfolio (t). (9.11)
This means that the exposure to the risky portfolio may not be larger than the cur-
rent portfolio value. We cannot invest more than our current notional amount. More
generally, we may write
Emax (t) = K · PV portfolio (t). (9.12)
For K > 1, we allow leverage up to a certain level. For 0 < K < 1, we constrain the
maximum exposure to the risky portfolio. The latter has a similar effect as a lock-in. In
addition, we may scale Emax with respect to the initial investment PV (t0 ) rather than
the current portfolio value.
The leverage and constraint in the above description refer to the risky portfolio.
Similarly, we may define leverage with respect to the risk-free investment defined by
Riskfree(t) = max{Riskfreemin (t), PV portfolio (t) − E(t)}. (9.13)
In the absence of leverage, we find
Riskfreemin (t) = 0. (9.14)
This means that it is not allowed to borrow funds in order to invest in the risky portfolio.
More generally, we may write:
Riskfree(t) = k · PVportfolio (t). (9.15)
For k < 0, we allow borrowing and thus leverage up to a certain level. For 0 < k < 1,
we request a minimum amount to be invested in the risk-free asset, which constrains
the exposure to the risky portfolio. This has a similar effect as a lock-in. As proposed
above, we may also scale the risk-free exposure with respect to the initial investment
PV (t0 ) rather than the current portfolio value.
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

Dynamic Portfolio Insurance Without Options 213

In our framework, we implemented leverage and constraint of the risky portfolio

and the risk-free asset. Both features may be used at the same time. The weaker criterion
determines the overall portfolio leverage or constraint.

9.4.5. Rebalancing Strategies for the Risky Portfolio

Up to now, we have not discussed how a rebalancing of the risky portfolio is propagated
to the portfolio constituents. In principle, we could perform a complete portfolio opti-
mization in each step and adjust the portfolio weights accordingly. This may include
Markowitz, VaR, or CVaR optimization with different types of linear, non-linear con-
straints, or boundary conditions. For the sake of simplicity, we show three types of
simple portfolio rebalancing strategies:

• Balanced: Here, the notional of all N risky portfolio constituencies is rebalanced in

accordance with the initial weights defined at the start of the strategy. This approach
reduces the amount of the above-average-performing risky assets and increases the
amount of the below-average-performing risky assets. The balanced approach is an
anti-cyclical profit-taking strategy that assumes a mean-reverting market within the
universe of the risky assets.
• Proportional: Here, the risky portfolio constituencies are rebalanced proportional to
current weights. This strategy implements a simple trend-following approach with
a soft competition among the risky assets.
• Squared: Here, the risky portfolio constituencies are rebalanced proportional to the
square of the current weight. Similar to the proportional strategy, this approach
implements a trend follower but with fierce competition among the risky assets due
to the squared-weighting factor.

9.4.6. CPPI and Beyond

The described framework allows implementation of a wide range of different strate-
gies customized to the risk appetite and investment guidelines of the investor. These
strategies are independent of the invested asset classes. The advanced features may
also make the simple strategies buy-and-hold, stop-loss, and bond floor more flexible.
When these strategies are extended from a static to a dynamic portfolio approach, the
transaction filters are also very useful.
The advanced features also offer the opportunity to remedy the weaknesses of
the original CPPI, like the lack of protecting gains, the fixed time horizon, and the
limited potential to recover from a large draw-down. The guaranteed target level at a
fixed time horizon is a feature that may not be necessarily required by an investor as
this requirement has also a significant downside: There is little upside potential to take
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

214 Dersch

advantage of a market recovery if the risk budget is strongly depleted after a downturn
in the risky portfolio.
The situation is worse if this happens in the early stage of the strategy. One pos-
sibility to remedy this weakness is the minimum exposure CPPI. Here, the allocation
in the risky portfolio is held at a minimum guaranteed level. This ensures participa-
tion over the complete lifetime of the strategy. But the participation comes at a cost.
An additional option has to be purchased to guarantee the target level at maturity. In
this chapter, we focus on portfolio strategies without options and therefore follow a
different approach.
If we forgo the requirement of a target level at a fixed time horizon, we can
instead attempt to secure — at any time — a fraction, e.g., 80% of a past portfolio
value. In our advanced CPPI framework, this approach can be implemented by setting
the discount rate in Eq. (9.7) to zero. Now target and floor are the same. In order to
provide a risk budget, the initial target value has to be below 100% of the notional
amount invested. Adding the lock-in type all-time-high trailing results in a strategy
that has been described as Time-Invariant Portfolio Protection (TIPP) [4]. The main
characteristics are

• No fixed investment horizon is required.

• Guaranteed instantaneous target level with an initial value below the investment
amount.
• Guarantee to recover from a market rebound because the risk-free portfolio contin-
uously generates a new risk budget.
• Implements a sequence of all-time-high look-back options with a participation rate,
which equals the target level but without the requirement of a fixed expiry date.
• The present value of the portfolio may fall with a growing investment horizon in
falling or sideward-moving markets.

The last point is a drawback of the approach. However, the investor can redeem the
investment at any time.

9.5. HISTORICAL SIMULATION II

In this section, we illustrate the above-advanced features with individual simulations

and analyze the impact on the performance in the current volatile market environment.

9.5.1. Transaction Costs and Transaction Filter

We mentioned above that it is important to include transaction costs in a simulation in
order to get an idea of the performance under realistic conditions.
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

Dynamic Portfolio Insurance Without Options 215

Table 9.4 Simulation Parameters to Study the Impact of Transaction Filters

on Transaction Costs. Four Different Simulations are Carried Out.
Simulation period 4 January, 1999–26 December, 2006
Investment EUR 100 mn
Target 100%
Multiplier 3
Risk-free investment EUR overnight liquidity
Risky investment DJ Euro Stoxx Select Dividend 30
Discount rate 3.25%
Trans. filter (Sim. 1 and 3) "(buy) = 0.12, "(sell) = 0.08, λ = 0.85, !t = 1 day
Trans. cost (Sim. 1 and 2) Risky investment 60 bp, risk-free investment 1 bp

In the following, we carry out four simulations with and without transaction fil-
ters combined with and without transaction costs. The risky asset is the DJ Euro Stoxx
Selected Dividend 30 index. The simulation parameters are summarized in Table 9.4.
To illustrate the impact of transaction costs, we selected a period of a rising market. The
transaction costs are set to 60 bp of the transaction volume for the risky investment and
to 1 bp for the risk-free investment. The transaction costs are deducted from the port-
folio at the time of the transaction. In our simulation, we ignore the effect of slippage
and partial execution. Figure 9.6 shows one out of the four simulations using transac-
tion filters and considering transaction costs. Table 9.5 summarizes a comparison of
different simulations with and without transaction filters and transaction cost.
Without the transaction filter, there is a huge impact of transaction costs. The
difference is EUR 54.45 mn. This means that the naïve approach — Simulation 4 —
would suffer a drop of more than 20% of the final portfolio value in case transac-
tion costs have to be taken into account (Simulation 2). Using the transaction filter
results in much smaller dependency on transaction costs. The difference here is only

Table 9.5 Comparison of Different Historical CPPI Simulations With (Without) Trans-
action Filters, Simulations 1 and 3 (Simulations 2 and 4) and With (Without) Transaction
Costs, Simulations 1 and 2 (Simulations 3 and 4). For the Different Combinations, We
Show the Total Number of Rebalancing Steps, the Portfolio Value at Maturity and the
Total Transaction Costs.
Trans. Trans. # of PV final in Transaction cost
Simulation filter costs transactions EUR (mn) in EUR (mn)

1 Yes Yes 66 230.64 −2.04

2 No Yes 1923 215.40 −8.84
3 Yes No 52 254.53 —
4 No No 1688 269.85 —
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

216 Dersch

EUR 23.90 mn (9.4%). The above simulation clearly reveals the damaging effect of
transaction cost in the absence of the transaction filter. There is a primary effect caused
by pure transaction costs that are deducted from the portfolio. This accounts for EUR
−2.04 mn (Simulation 1, with transaction filter) and EUR −8.84 mn (Simulation 2,
without transaction filter). There is a secondary effect that shows the total effect of
transaction cost. A comparison of the final portfolio values shows that in case of
transaction filters the difference between portfolio values of EUR −23.90 mn contains
transaction costs in the amount of EUR −2.04 mn. In the absence of transaction filters,
the performance gap is EUR −54.45 mn, where EUR 8.84 mn is the pure transaction
cost. In this context, the transaction cost ratio
PV (tc) − PV (ntc)
tcr = (9.16)
TC
is a useful ratio to measure the secondary effect of transaction cost. Here, PV (tc) is the
performance with transaction cost and PV (ntc) without transaction cost and TC the
pure transaction cost incurred over the observation period. A tcr close to one means
that the performance difference with and without transaction cost is mainly caused by
pure transaction cost. A tcr of two means that the transaction cost causes a performance
reduction of twice the amount of the pure transaction cost. In our simulation, the tcr
is equal to 11.71 (with transaction filter) and 6.16 (without transaction filter). The tcr
depends on the compounding effect and the leverage of exposure. In a CPPI simulation
with a multiplier of 3, every EUR of risk capital (cushion) changes the exposure to
the risky portfolio by EUR 3. Transaction costs that reduce the exposure early in
the investment period may have a tremendous impact on the performance later on.
Ironically, in a falling market, transaction costs may even have a positive impact in the
presence of a high multiplier as they may force the early reduction of the exposure and
save the portfolio from otherwise higher losses. We also found that if the floor level
is reached early in a simulation, transaction costs become less significant as it makes
no difference whether a further falling market or transaction costs are the cause for a
decline in portfolio value. In general, we conclude that the effect of transaction costs
is not very easy to estimate. Here, simulations shed light on the effect.

9.5.2. Lock-in Levels

The sample simulation of the simple strategies illustrated a fundamental drawback
of the protection strategies: the inability to take profits. An historical simulation on
the DJ Euro Stoxx 50 index illustrated how the gains of more than 50% that were
accumulated in the year 2000 are wiped out again (please compare Fig. 9.4). In order
to remedy this weakness, we introduced a number of different lock-in strategies that are
designed to protect gains. In this section, we perform historical simulations to evaluate
the performance of different lock-in strategies. In order to get a direct comparison, we
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

Dynamic Portfolio Insurance Without Options 217

Table 9.6 Parameter Settings for Three Different Advanced CPPI Simulations. The
Simulation Parameters Differ only in the Discount Rate for the Floor, the Lock-In Trigger,
and Lock-In Action (the Last Three Rows).

Simulation period 4 January, 1999–12 December, 2008

Investment EUR 100 mn
Target 80%
Multiplier 6
Risk-free EUR overnight liquidity
Risky investment DJ Euro Stoxx 50
Transaction filter "(buy) = 0.12, "(sell) = 0.08, λ = 0.85, !t = 1 day
Transaction cost Risky investment 8 bp, risk-free investment 1 bp
Discount rate 3.25% (Simulations 1 and 2), 0% (Simulation 3)
Lock-in trigger NA (Simulation 1, no lock-in), monthly, all-time-high (Simulations 2
and 3)
Lock-in action NA (Simulation 1, CPPI without lock-in), trail all-time-high by moving
up the floor (Simulations 2 and 3)

perform our simulation on the Dow Jones Euro Stoxx 50 index. Table 9.6 shows the
simulation parameters.
For comparative reasons, we first show the CPPI simulation without lock-in
(Fig. 9.5). As compared to the previous simulation shown in Fig. 9.4, we use a transac-
tion filter, consider transaction costs, and a target of only 80% as compared to 100%.

250
DJ EUR STOXX Select Dividend 30 risk-free floor

200

150

100

50

0
1999 2000 2001 2002 2003 2004 2005 2006

Figure 9.5 Historical CPPI simulation under realistic condition on the DJ Euro Stoxx Selected
Dividend 30 index using transaction filters and transaction costs. For a detailed specification,
please refer to the text above. The y-axis shows the portfolio value in EUR mn.
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

218 Dersch

180
DJ EUR STOXX 50 risk-free floor
160

140

120

100

80

60

40

20

0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 9.6 Historical CPPI simulation on the DJ Euro Stoxx 50. The simulation parameters
are shown in Table 9.6. The areas show the allocation of the two assets. The black line is the
floor level. The y-axis shows the portfolio value in EUR mn.

Qualitatively, we find a similar result in Fig. 9.6 as before. The decline in the stock
market leads to a complete switch into the risk-free asset at the end of the investment
horizon. All previous gains are wiped out again. Due to the lower target of 80%, the
portfolio is still invested in the risky asset after the sharp decline in 2002 as compared
to a target level of 100%.
Figure 9.7 illustrates the impact of the all-time-high trailing. The floor is moved
upwards with a rising market. The initial target level of 80% trails the complete upward
market movement of 56%, resulting in a final portfolio amount of 133.8% of the initial
value. The simulation clearly demonstrates that trailing allows one to protect gains that
have been previously accumulated. The floor value is slightly missed for the reasons
already mentioned above.
As a consequence, the increased target level causes a complete exit from the risky
investment until the end of the investment horizon. The strategy protects gains, but is
unable to recover from the draw-down. This weakness is removed in the next simulation
shown in Fig. 9.8.
Compared to the previous simulation, we now discount the floor level with zero.
As a result, the floor is a horizontal line shifted upwards when new all-time-highs are
reached. The floor level marks the guarantee level. Discounting the floor with a rate of
zero reduces the risk budget because the floor is guaranteed instantaneously and not
at maturity.
The first half of the simulation period shown in Fig. 9.8 is similar to the simulation
in Fig. 9.7. The floor trails the rising index value and the following decline results in a
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

Dynamic Portfolio Insurance Without Options 219

180
DJ EUR STOXX 50 risk-free floor
160

140

120

100

80

60

40

20

0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 9.7 Historical CPPI simulation on the DJ Euro Stoxx 50 using the all-time-high trailing
with monthly lock-in triggers. The further simulation parameters are shown in Table 9.6. The
areas show the allocation of the two assets. The y-axis shows the portfolio value in EUR mn.
The black line is the floor level. The floor value is slightly missed for reasons mentioned above.

200
DJ EUR STOXX 50 risk-free floor
180

160

140

120

100

80

60

40

20

0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 9.8 Historical CPPI simulation on the DJ Euro Stoxx 50. The simulation parameters
are shown in Table 9.6. The areas show the allocation of the two assets. The black line is the
floor level. The y-axis shows the portfolio value in EUR mn.

reallocation to the risk-free investment. Here, this reallocation occurs earlier because
of the reduced risk budget and we therefore do not reach the same all-time-high in
the portfolio value. The second half of the simulation shows a significantly different
picture as compared to Fig. 9.7. The strategy participates in the rising market starting
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

220 Dersch

in 2003. In 2005, the strategy is again allocated in the risky investment to a very high
proportion for a very short period of time. The guarantee level is moved upwards again.
As a result, we reach a final portfolio value of EUR 167.50 nm. For the given example,
the simulation shown in Fig. 9.8 is superior to the other strategies because of two
reasons:

• The ability to recover from the floor.

• The target value is guaranteed instantaneously and not at a pre-set maturity.

The reason for the rising from the dead like behavior of this strategy is the fact that
even on the floor the risk-free asset continuously regenerates a cushion. Supported by
a large multiplier, the exposure is quickly scaled up again. By the end of 2005, a new
all-time-high levels are reached. They are trailed with a participation rate which equals
the target level (80%).
Table 9.7 summarizes the results for the three different simulations. The TIPP
strategy performs best as compared to the two other strategies.

9.5.3. The Use of Leverage

Lock-in is a conservative feature. It allows the protection of past gains. On the other
hand, each lock-in reduces the risk budget and therefore may constrain future expo-
sure to the risky asset. Leverage has an opposite effect. It increases the exposure by
borrowing risk-free and investing it in the risky portfolio. It therefore allows a lever-
aged investment in the risky portfolio. In the following, we illustrate leverage with
one example on the Credit Suisse/Tremont Investable Hedge Fund Index. The index
consists of 60 different hedge funds that represent 10 different strategies. For more
information on the index, please see [6]. There exist institutional and retail products
on this index. Table 9.8 shows the parameter settings for this simulation. Please note
that the rebalancing frequency is monthly and we assume borrowing at the risk-free
overnight rate.

Table 9.7 Comparison of Different Historical CPPI Simulations With and Without
Lock-In (Columns 2 and 3) and with Lock-In and Flat Floor (Column 4). For the Three
Combinations, We Show the Portfolio Return, the Final Value and the Final Floor Value.
The Floor Value for Simulation 2 is Slightly Missed for the Reasons Mentioned Above.

Strategy (1) CPPI (2) CPPI with lock-in (3) TIPP: lock-in and flat floor

Ann. return −2.25% 2.97% 5.33%

End value EUR mn 79.76 133.78 167.53
Floor Final in % 80.00% 136.26% 166.80%
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

Dynamic Portfolio Insurance Without Options 221

Table 9.8 Parameters for an Historical Advanced CPPI Simulation With Leverage.

Simulation period 4 January, 2000–4 December, 2008

Investment USD 1 mn
Target 80%
Multiplier 6
Risk-free USD overnight liquidity
Risky investment Credit Suisse Tremont Investable index
Transaction filter "(buy) = 0.12, "(sell) = 0.08, λ = 0.85, !t = 1 day
Transaction cost Risk-free investment 1 bp, risky investment 50 bp
Discount rate 0%
Lock-in trigger Portfolio PV increased by 10% of previous value tested on a monthly
base
Lock-in action Increase target by 10% of initial portfolio PV
Leverage Up to 60% of the initial investment of USD 1 mn may be borrowed at
any time to increase the exposure to the risky asset

2.5

2.0

1.5

1.0

0.5

0.0

-0.5

-1.0 CS Trem. Inv. risk-free floor portfolio index

2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 9.9 Historical advanced CPPI simulation on the Credit Suisse Tremont Investable
Index. The y-axis shows the portfolio value in EUR mn. The leverage is limited to USD 0.6 mn.
The areas show the allocation of the two assets. Leverage shows up in a negative allocation of
the risk-free asset (light gray area). The gray-and-black-dashed lines are the performance of the
index and portfolio, respectively. The black line is the floor.

Figure 9.9 shows an historical simulation of the above strategy. The light gray
area represents the risk-free investment. Until 2008, we find a leveraged investment.
The allowed leverage of USD 600,000 is utilized to the maximum level at certain
times in 2000, 2005, 2006, 2007, and January 2008. Each increase of the target level
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

222 Dersch

(black line) is followed by a reduction of leverage. In this example, the leveraged

investment results in an outperformance of the hedge fund index (compare dashed
black line versus dashed gray line). That means that the proportion based on leverage
yields an additional positive return after borrowing and transaction costs.
The final portfolio value is USD 1.57 mn. This corresponds to an annual return
of 5.19%. The current floor level is USD 1.52 mn. There has been no short position in
the risk-free investment since October 2008. Currently, the portfolio still holds a 30%
investment in hedge funds. In this example, the lock-in feature mitigates the effect
of leverage. Without lock-in, the leverage budget of USD 600,000 mn would be fully
utilized from 2001 until the end of the simulation period. A higher leverage together
with lock-in would not make a significant difference as the lock-in and multiplier affect
the maximum amount to be borrowed (data for both simulations are not shown). This
emphasizes the requirement that strategy parameters are interdependent and have to
be carefully adjusted.

9.5.4. CPPI on a Multi-Asset Risky Portfolio

In this example, we illustrate a risky portfolio of different asset classes represented
by different performance indices, namely, equity (DAX), fixed income (Rex), and
commodities (Dow Jones AIG Commodity Index) classes. For the three assets, there
exist exchange-traded funds. The three asset classes have been selected based on the
low historical correlation of daily log-returns. In this example, we make use of the
portfolio rebalancing feature proportional. The simulation parameters are shown in
Table 9.9.

Table 9.9 Parameter Settings for an Historical Advanced CPPI Simulation

on a Multi Asset Portfolio Including the DAX, Rex, and Dow Jones AIG.

Simulation period 3 January, 2000–12 December, 2008

Investment EUR 100 mn
Target 80%
Multiplier 6
Risk-free EUR overnight liquidity,
Risky investment DAX (initial weight 30%),
Rex (initial weight 50%),
Dow Jones AIG Commodity (initial weight 20%)
Transaction filter "(buy) = 0.12, "(sell) = 0.08, λ = 0.85, !t = 1 day
Transaction cost Risky investment 20 bp, risk-free investment 1 bp
Discount rate 0%
Lock-in trigger Monthly, all-time-high
Lock-in action Trail all-time-high by moving up the floor
Rebalancing strategy Proportional
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

Dynamic Portfolio Insurance Without Options 223

160
REX DAX risk-free DJ AIG Commodity Index floor
140

120

100

80

60

40

20

0
2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 9.10 Advanced historical CPPI simulation on a risky portfolio of three different low-
correlated asset classes. The four different shaded areas indicate the allocation over time. The
black line is the floor level.

Figure 9.10 illustrates how the allocations in different asset classes evolve over
time. From the end of 2000 until the beginning of 2005, the target level is constant.
There are two periods with a full investment in the risky asset (2000 and the end of 2004
until the end of 2005). The final portfolio value is EUR 131.32 mn. This corresponds to
a return of 3.0% p.a. mainly attributed to the years 2000 and 2005–2007. The present
value time weighted asset allocation is 45% Rex, 21% AIG Commodity Index, 17%
DAX, and 17% risk-free. This contrasts the current allocation of 22% Rex, 7% AIG
Commodity Index, 6% DAX, and 66% risk-free. This is intuitive as we are currently
very close to the floor level of EUR 123.95 mn.

9.6. IMPLEMENT A DYNAMIC PROTECTION

STRATEGY WITH ETF

The strategies shown in this chapter may be implemented by a direct investment in

shares or baskets of shares, bonds, or by an indirect investment via Futures. Exchange
Traded Funds (ETF) are an attractive alternative investment vehicle. ETF offers an
investment in a wide range of different asset classes. All sample investments used in
this chapter — except the Hedge Fund and Commodities index — may be implemented
by investing in a corresponding ETF. The ETF on the Dow Jones Stoxx Selected Divi-
dend 30 closely replicates the corresponding index. The index composition implements
a portfolio strategy in its own sense as the portfolio constituencies are dynamically
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

224 Dersch

Table 9.10 Comparison of Various Characteristics of ETF’s Please See Also [5].

Bid-offer Management
ETF ISIN spread in bp fee in bp pa

eb.rexx Goverm. Germany 5.5–10.5 DE 000 628 949 9 7 15

DJ Euro Stoxx 50 DE 000 593 395 6 8 15
DJ Stoxx 50 DE 000 593 394 9 43 50
DJ Euro Stoxx Select. Dividend 30 DE 000 263 528 1 56 30
DJ Stoxx Selected Dividend 30 DE 000 263 529 9 72 30
Dow Jones-AIG Commodity DE 000 A0H 0728 99 45

adjusted from the Dow Jones Stoxx 600 universe based on their dividend yield and
dividend consistency. For an exact definition of the index, please refer to the cor-
responding description of Dow Jones. Due to the nature of the index creation, the
portfolio follows a more conservative anti-cyclical profit-taking strategy. Generally,
the underlying rational of the index, its dynamic and characteristics have to be taken
into account when setting the parameters of the dynamic protection strategy. Table 9.10
compares typical bid–offer spreads and management fees of various ETF. The shown
bid–offer spreads are snapshots and may vary from day to day. For further information,
please compare [5].
Bid–offer spreads reflect the characteristics of the underlying and are influenced
among other factors by liquidity and taxation issues. Larger bid–offer spreads in the
investment would favor the transaction filter with larger ", !t, and small λ.

9.7. CLOSING REMARKS

We have demonstrated how the main drawback of CPPI, namely, its pro-cyclical behav-
ior, the lack of recovering potential once the floor has been hit, and the fixed investment
horizon can be overcome by introducing a number of advanced features. The above
strategies may be implemented with simple spread sheet models avoiding the usage
of options. However, the investor must be aware of the different risk aspects related to
this approach.
We illustrated how dynamic portfolio strategies empower an investor to replicate
fairly complex option profiles including path dependent look-back options. Under
certain conditions, the investor may save hedging costs, e.g., the option premium. But
the savings come at a price. The investor is left with the risk that the strategy will miss
the investment target. Buying an option allows one to lock in the implied volatility
at the time of purchase. If the realized volatility over the lifetime is higher than the
implied volatility of the option, the option is superior. Buying insurance — in the form
May 12, 2010 17:47 WSPC/SPI-B913 b913-ch09 FA

Dynamic Portfolio Insurance Without Options 225

of options — is therefore fairly expensive in the current market environment; however,

if such a strategy was implemented one and a half years ago, an option-based strategy
back then looks very cheap now in terms of strikes and volatility.

Acknowledgment

The author is grateful to Thorsten Weinelt and UniCredit Research to support this
work, Peter Hieber for layout and formating, and to David Dakshaw for proof reading
the manuscript. Views expressed in this chapter are those of the author and do not
necessarily reflect positions of UniCredit Research.

References

[1] Bertrand, P and JL Prigent (2001). Portfolio insurance strategies: Obpi versus cppi. CERGY
Working Paper, 30.
[2] Black, F and R Jones (1987). Simplifying portfolio insurance. The Journal of Portfolio
Management, 14(1), 48–51.
[3] Black, F and AF Perold (1992). Theory of constant proportion portfolio insurance. Journal
of Economic Dynamics and Control, 16, 403–426.
[4] Estep, T and M Kritzman (1988). Tipp: Insurance without complexity. The Journal of
Portfolio Management, 14(4), 38–42.
[5] iShares (2008).
[6] Credit Suisse Tremont (2008).