Asset Liability Management Techniques in Private Banking

Transcription

1 EDHEC Asset Management Days 2007 Geneva, March 12 th 10:00-11:30 Asset Liability Management Techniques in Private Banking Lionel Martellini Professor of Finance, EDHEC Business School Scientific Director, EDHEC Risk and Asset Management Research Centre

2 Outline Introducing ALM in PWM: Why? Sources of Added-Value in Wealth Management ALM as a Client-Driven Approach to PWM A Typology of Clients Profiles ALM as a Real Client-Driven Approach to PWM Introducing ALM in PWM: How? Cash-Flow Matching & Immunization Methods Surplus Optimization Techniques LDI Solutions A Brief History of ALM Techniques & their Use in PWM Introducing ALM in PWM: For what Benefits? Illustrations of the Illustration # 1: Improving Retirement Benefits Usefulness of an ALM Illustration # 2: Including Bequest Motives Approach to PWM Illustration # 3: Preparing for Real Estate Acquisition

3 Introducing ALM in PWM: Why? Sources of Added-Value in Wealth Management ALM as a Client-Driven Approach to PWM A Typology of Clients Profiles Introducing ALM in PWM: How? Cash-Flow Matching & Immunization Methods Surplus Optimization Techniques LDI Solutions Introducing ALM in PWM: For what Benefits? Illustration # 1: Improving Retirement Benefits Illustration # 2: Including Bequest Motives Illustration # 3: Preparing for Real Estate Acquisition

4 Sources of Added-Value in Wealth Management Most private bankers implicitly promote an ALM approach to wealth management. In particular, they claim to account for the client s goals and constraints. The technical tools involved, however, are often non-existent or ill-adapted. Current practice in addressing clients specific goals and constraints is relatively limited. While the client is routinely asked all kinds of questions regarding current situation, goals, preferences, constraints, etc., the resulting service and product offering mostly boil down to a rather basic classification in terms of risk profiles. Private bankers real added-value should lie in the ability to account for the client s specific goals and constraints. Several situations exist, corresponding to varying levels of sophistication & consideration of clients needs.

5 ALM as a Client-Driven Approach to PWM Level # 0: do not use any portfolio construction tool The solution does not take into account clients objective, risk-aversion, or constraints. This is simply not acceptable. Level # 1: do asset management only The solution consists of optimal design of different portfolios with different risk profiles. Clients time-horizon, goals and constraints are not taken into account. Level # 2: test for ALM impact of asset allocation decisions For example use a model to test probability of a shortfall at horizon. Optimal asset portfolio, however, is designed independently of clients needs. Level # 3: incorporate the client s full profile in portfolio construction Only this can ensure the proper addressing of clients needs. This ALM approach to PWM is the target of the present presentation.

6 A Typology of Clients Profiles There are at least four dimensions in a client profile. Objective profile Time-horizon profile Constraints and risk-aversion profiles Contribution profiles Each of these dimensions is related to the definition of a client s liabilities.

7 A Typology of Clients Profiles Con t Dimension # 1: Objective profile Related to the particular type of liability a client is facing. Examples are pension needs, real estate acquisition, paying for children s education, etc. Dimension # 2: Time-horizon profile It can be shown that the optimal solution depends on the risk-horizon. Often the horizon is long, sometimes with intermediate, short-term, constraints or goals. Dimension # 3: Constraints and risk-aversion profiles Often classification boils down to subjective classification in terms of risk tolerance. A better understanding can be obtained from the perspective in terms of risk constraints. Dimension # 4: Contribution profiles A key distinction exists between fixed and flexible contribution schedules. This is expected to have an impact on portfolio construction.

8 A Typology of Clients Profiles - Example Objective profiles Pension-related objectives. Client is retired and has already contributed. Client is not retired yet and will contribute in the future. Expenditure-related objectives. Accumulate for future expenditure based on a single contribution. Accumulate for future expenditure based on a schedule of contributions. Time-horizon profiles: 0-3 years, 3-5 years, 5-10 years, years, years. Constraints and risk-aversion profiles Short-term constraints: no annual protection, annual capital protection up to a 100%, 95%, 90%, 85% confidence level. Long-term constraints: no protection at horizon, full protection at horizon in real or nominal terms, probability of not reaching the goal kept below 10%, 15%, 20%, 25%, 50%. Contribution profiles Fixed contribution schedule. Uncertain contribution schedule. Presence of a schedule of minimum contributions. Absence of a schedule of minimum contributions.

10 A Brief History of ALM Introducing a real focus on addressing clients needs involves a new approach to private banking, similar to the recent focus on ALM in institutional money management. Sophisticated ALM techniques are naturally suited to design an offering of products and services that really account for clients goals and constraints, not only in institutional money management, but also in private money management. There are three types of techniques used in ALM, each one corresponding to a specific ALM technique.

11 A Brief History of ALM Con t These ALM techniques are the natural counterparts from a relative return perspective to corresponding approaches to AM.

12 A Brief History of ALM Con t Technique # 1 Immunize interest rate risk exposure. Cash-flow or duration matching strategies. Technique # 2 Modeling stochastic scenarios for asset and liability returns. (Surplus) optimization-based strategies. Technique # 3 Separate (liability) risk management from asset management. LDI solutions.

13 A Brief History of ALM Con t Different ALM techniques Cash-flow matching & Immunization: CF matching involves ensuring a perfect match between the cash flows from the portfolio of assets and the commitments in the liabilities; inflation-linked instruments are often used in that perspective. To the extent that perfect matching is not possible, immunization techniques allow the residual interest rate risk created by the imperfect match between the assets and liabilities to be managed in an optimal way. Counterpart in AM: invest in risk-free asset. Surplus optimization: In a concern to improve the profitability of the assets, and therefore to reduce the level of contributions, it is necessary to introduce asset classes (stocks, nominal bonds) that are not perfectly correlated with the liabilities into the strategic allocation. Counterpart in AM: invest in optimal risky portfolio. LDI Solutions Advocate a separation between two conflicting goals: risk management and performance generation. Leverage is used. Counterpart in AM: two-fund separation theorem in the static form, portfolio insurance in the dynamic form.

14 A Brief History of ALM Con t Risk/Return Profile Asset Management (absolute risk) Asset-Liability Management (relative risk) Zero risk no access to risk premia Investment in Risk-free asset Cash-flow matching and/or immunization Optimal risk-return trade-off Optimally diversified portfolio of risky assets Optimisation of the surplus Fund separation theorem Capital market line (static mix of cash and optimal performanceseeking risky portfolio) LDI solution (static mix of cash, liability matching portfolio and optimal risky portfolio) Dynamic and skewed risk management (non linear payoffs) Portfolio insurance (dynamic mix of risk-free asset and optimal risky portfolio) Dynamic LDI (also known as contingent immunization) Overview of ALM techniques and the corresponding techniques in asset management

15 Surplus Optimization Techniques versus LDI Relative MSD Frontier 0.15 Exceess retur Performance-Seeking Portfolio 0.1 Liability Hedging Portfolio 0.05 Liability Matching Portfolio Relative Risk

16 LDI Solutions Dynamic Strategies High risk aversion leads to dominant investment in liability-hedging portfolio, which implies: Low extreme funding risk (zero risk in complete market case) Low performance and hence high necessary contribution Low risk aversion leads to dominant investment in standard AM optimal portfolio, which implies: High funding risk Higher performance, and hence lower contributions Can we get the best of both worlds? Dynamic extensions of LDI strategies exist; they represent the equivalent of portfolio insurance in an ALM context (in particular extended CPPI strategies). They imply a zero extreme funding risk beyond the pre-specified threshold as well as an access to upside potential (limited because of the cost of insurance).

17 Dynamic LDI Strategies Under these strategies, the fraction of wealth A t allocated to the performance portfolio is equal to a constant multiple m of the cushion, i.e., the difference between the asset value and the floor defined as the difference between asset value and minimum liability value: A t -kl t. This is reminiscent of CPPI strategies, which the present setup extends to a relative risk management context; intuitively these strategies are based on the recognition that how much risk you take should depend on margin for error, and not only on risk aversion! While CPPI strategies are designed to prevent final terminal wealth to fall below a specific threshold, extended CPPI strategies (or extended contingent immunization strategies) are designed to protect asset value not to fall below a pre-specified fraction of some benchmark value, here liability value. More general strategies actually exist, which can be regarded as extensions of OBPI and exotic OBPI strategies.

19 Illustrations of the Usefulness of an ALM Approach In what follows, we present a set of examples of the usefulness of asset-liability management techniques in private banking. Our examples are drawn from the typology of client s profiles documented in the first section of the presentation. We use a standard model for generating stochastic scenarios for risk factors affecting asset and liability values. The model is calibrated with respect to long time-series; other choices of parameter can be adopted.

20 Illustration # 1 Client problem We consider a 65-year-old individual who is already retired. His/her goal is to ensure a stream of inflation-protected fixed payments (normalized at 100) for the next 20 years (i.e., from age 65 to age 85). (*) To achieve this goal the individual is prepared to invest a fixed amount of money. Client profile Pension-related objective # 1 Time-horizon: 20 years Constraints: no ST constraints; robustness checks on LT constraints. Contributions: fixed-schedule set at zero additional contributions. (*) We thus assume away the complexity related to mortality risk (can be dealt with through an annuity contract provided for by an insurance company).

21 Illustration # 1 Client s Profile Objective profiles Pension-related objectives. Client is retired and has already contributed. Client is not retired yet and will contribute in the future. Expenditure-related objectives. Accumulate for future expenditure based on a single contribution. Accumulate for future expenditure based on a schedule of contributions. Time-horizon profiles: 0-3 years, 3-10 years, years. Constraints and risk-aversion profiles Short-term constraints: no annual protection, annual capital protection up to a 95% confidence level. Long-term constraints: no protection at horizon, full protection at horizon in real or nominal terms, probability of not reaching the goal kept below 10%. Contribution profiles Fixed contribution schedule.

22 Illustration # 1 Strategies One natural solution consists of buying equal amounts of zerocoupon TIPS with maturities ranging from 1 year to 20 years (liability matching portfolio). The performance, however, is poor. The client needs a very high current contribution to sustain his/her future consumption needs. It is reasonable to add risky asset classes to spice up the return. There is a risk involved, however.

23 Illustration # 1 Experimental Protocol Using the aforementioned stochastic model, we generate random paths for the prices of a stock index, a bond index, and 20 zerocoupon TIPS with parameters consistent long-term estimates. We find the time- and state-dependent value of liabilities by taking the discounted sum of cash-flows; it is identical to the current value L(0) of the portfolio of TIPS. We test three different strategies Cash-flow matching strategies. Surplus optimization strategies. Dynamic LDI strategies.

24 Illustration # 1 Experimental Protocol Con t Cash-flow matching Find present value of liability-matching portfolio L(0); we obtain L(0) = (rather close to 20x100, very expensive!) Distribution of surplus at date 20: trivial (no possible deficit). Surplus optimization Start with same initial amount L(0). Find the best fixed-mix strategy that consists of investment in stocks, bonds and liability-matching portfolio (regarded as a whole) so as to generate an efficient frontier in a surplus space (trade-off between expected surplus and variance of the surplus). Minimum risk portfolio = 100% in liability-matching portfolio. Plot the distribution of surplus at date 20 for a few points on the efficient frontier. Formally, assume that asset portfolio is liquidated each year, liability payment is made, and the remaining wealth invested in optimal portfolio; in scenarios such that remaining wealth is not sufficient for making liability payment, we assume that borrowing at the risk-free rate is performed so as to make up for the difference. Estimate probabilities of not meeting the objectives (probability of a deficit).

25 Illustration # 1 Optimal ALM Portfolios weights expected Surplus volatility of surplus Prob(S<0) expected shortfall necessary nominal contribution relative Contribution Saving p.a. stocks bonds Liab-PF A 0% 0% 100% % 0.00 (0.00%) % B 19% 22% 59% % (11.50%) % C 43% 41% 16% % (24.08%) % D 52% 45% 3% % (28.28%) % A' 10% 90% 0% % (24.33%) % B' 20% 80% 0% % (21.72%) % C' 43% 57% 0% % (25.75%) % D' 51% 49% 0% % (28.09%) % * all values are given as present values at starting date (initial investment as of that is the present value of liabilities) ** losses relative to L(0) in parentheses

26 Illustration # 1 Optimal ALM Portfolios Efficient frontier with a liability-matching portfolio (blue line) and without a liability-matching portfolio (green line)

27 Illustration # 1 Distribution of Deficit/Surplus Distribution of final surplus/deficit

28 Illustration # 1 ALM versus AM The ALM exercise consists of finding the portfolios that are optimal from standpoint of protecting the investors liabilities. A pure asset management exercise, on the other hand, focuses on designing the portfolios with the optimal risk-return trade-off. Nothing guarantees that AM efficient portfolios will be efficient from an ALM perspective (and vice-versa); in particular, the focus is on nominal return from an AM perspective, while it is on real return from an ALM perspective. To test for the ALM performance of AM efficient portfolios, we have conducted the following experiment Find the standard (Markowitz efficient) frontier based on horizon returns, i.e., the portfolios that achieves the lowest level of volatility (across scenarios at horizon) for a given expected return (across scenarios at horizon). Plot these portfolios in the (expected surplus-volatility of the surplus) ALM space.

29 Illustration # 1 ALM versus AM Con t A portfolio efficient in an AM sense is not necessarily efficient in an ALM sense, and vice-versa.

30 Illustration # 1 Dynamic Portfolio Strategies We now turn to dynamic portfolio strategies. We consider 6 different implementations of the extended contingent immunization method. We take the stock-bond portfolio with highest Sharpe ratio as performance portfolio (with our choice of parameter values, and 4% risk-free rate, we obtain the following portfolio: 30.9% in stocks and 69.1% in bonds) Cases 1 to 3: protection level k=90%, and multiplier m=2,3,4 Cases 4 to 6: protection level k=95%, and multiplier m=2,3,4 Results show very significant risk management benefits.

31 Illustration # 1 Static versus Dynamic Strategies dynamic CPPI expected Surplus volatility of surplus Prob(S<0) expected shortfall necessary nominal contribution p.a. relative Contribution Saving p.a. m=2 k= % (3.74%) % m=3 k= % (5.47%) % m=4 k= % (6.70%) % static CPPI expected Surplus volatility of surplus Prob(S<0) expected shortfall necessary nominal contribution p.a. relative Contribution Saving p.a. m=2 k= % (4.55%) % m=3 k= % (6.74%) % m=4 k= % (8.94%) % dynamic CPPI expected Surplus volatility of surplus Prob(S<0) expected shortfall necessary nominal contribution p.a. relative Contribution Saving p.a. m=2 k= % (1.84%) % m=3 k= % (2.71%) % m=4 k= % (3.29%) % static CPPI expected Surplus volatility of surplus Prob(S<0) expected shortfall necessary nominal contribution p.a. relative Contribution Saving p.a. m=2 k= % (2.27%) % m=3 k= % (3.46%) % m=4 k= % (4.55%) %

32 Illustration # 1 ALM Extended CPPI Strategies

33 Illustration # 2 We consider a wealthy 65-year-old individual who is already retired. He/she has a significant wealth (say 100 million euros) and wishes to maintain a standard of living (annual expenses say at 2 million euros + inflation) with an additional bequest motive in 20 years. (*) The analysis aims at finding the optimal policy so as to generate the highest possible bequest level with a given probability denoted by α. Consider two variants in which ½ of the client wealth (100 million) is held as stock of his/her own private company, which will be sold in 5 years from now (in this case, impose a 50% lower constraint on equity allocation). The value of existing property is accounted for (e.g., the client has 10 million worth of property value in addition to the 100 million euros). (*) We thus assume away the complexity related to mortality risk (can be dealt with through an annuity contract provided for by an insurance company).

35 Illustration # 2 Results Discounted final bequest distribution (no constraints) Target Percentile Weights Expected Volatility of Bequest Percentiles Stocks Bonds LMP Bequest Bequest median no constraints alpha= 5% 11% 49% 41% alpha=10% 25% 38% 37% alpha=20% 53% 23% 24%

36 Illustration # 2 Variants Discounted final bequest distribution (with additional 10m initially held in real estate) Discounted final bequest distribution (min 50% in stocks) min 50% in stocks additional 10m in real estate property at T0 Target Percentile Weights Expected Volatility of Bequest Percentiles Bequest Bequest Stocks Bonds LMP median alpha= 5% 50% 40% 10% alpha=10% 50% 15% 35% alpha=20% 53% 23% 24% alpha= 5% 22% 36% 42% alpha=10% 24% 44% 32% alpha=20% 52% 23% 25%

37 Illustration # 3 We consider an investor who wishes to invest fixed annual contributions ( x) for a future expenditure, e.g., buy a house in 5 years, the current value of which is normalized at 100 (or interpret this as the required down payment). For simplicity assume that house prices increase with inflation; more accurately, model the dynamics of real estate prices. In practical terms, the goal is to generate a lump sum payment at horizon date (5 years).

39 Illustration # 3 Experimental Protocol In this case, it is not possible to find a perfect liability-matching portfolio. The existence of a perfect liability-matching portfolio is only ensured at the following two conditions: Investor can borrow against future income and can invest at the initial date the present value of the future contributions. There exists an investment vehicle (e.g., REITS) for which the payoff is directly related to real estate price uncertainty. In what follows, we test two different situations: Opportunity set contains stocks, bonds and TIPS. Opportunity set contains stocks, bonds, TIPS + real estate (modelled as an investment that will pay the compounded return on real estate). To generate comparable portfolios, we have looked at the improvement in surplus volatility for a given level of expected surplus.

40 Illustration # 3 Results Distribution of house prices at final date; mean value = , standard deviation =

41 Illustration # 3 Results (Con t) ALM Efficient Frontiers without Real Estate (A, B, C, D, E, F) and with Real Estate (A, B, C, D, E, F ) much better of course.

42 Illustration # 3 Results (Con t) weights Portfolio Stocks Bonds Tips Real Estate expected Surplus volatility of surplus Prob(S<0) expected shortfall necessary nominal contribution p.a. relative Contribution Saving p.a. A 0% 0% 100% % (12.5%) % B 10% 90% 0% % (11.3%) % C 33% 67% 0% % (12.1%) % D 55% 45% 0% % (14.2%) % E 78% 22% 0% % (17.5%) % F 100% 0% 0% % (21.0%) % A' 0% 0% 100% 0% % (12.5%) % B' 0% 0% 68% 32% % 7.44 (8.2%) % C' 0% 0% 37% 63% % 4.57 (5.0%) % D' 0% 0% 5% 95% % 2.07 (2.3%) % E' 46% 0% 0% 54% % 7.36 (8.1%) % F' 100% 0% 0% 0% % (21.0%) % (*) Results obtained for a 20 annual contribution for 5 years, the present value of which amounts to given our choice of parameter values. Expected shortfall is expressed as a percentage of this value.

43 Conclusions from ALM to AM ALM is an essential step that allows the strategic inter-classes allocation to be determined. The success or failure of the satisfaction of the client s long term objectives is highly dependent on this phase. AM comes as a response to the implementation constraints of the ALM decisions. On the one hand, it allows the risk and return parameters supporting the ALM for each asset class to be delivered/enhanced. On the other hand, it allows short-term constraints to be managed as a complement (e.g. capital preservation at a 85%, 90% or 95% confidence level - VaR) or as full capital protection.