Strategic Asset Allocation session 1 Andrei Simonov Strategic Asset Allocation 1 4/9/2015 Agenda Introduction, Course Outline, Requirements, Resources Reminder: SAA Definitions Historical Records of Returns on different securities Crisis in Investment Industry Strategic Asset Allocation 2 4/9/2015 Introduction The field of Finance and Investments – Individual agents making decisions to supply capital to the markets – Firms getting capital from the financial markets (when, where, how?) – Capital Markets acting as market clearing device. Goal of the course: – To familiarize you with ”real world” of investments. – To give broad overview of modern investment issues. By June one should know what does that mean to be investment professional. Strategic Asset Allocation 3 4/9/2015 Resources and requirements: Course web page andreisimonov.com/NES Articles (package+web site) Provide deeper insight, latest developments No econometrics, just general idea Access to Internet, some Excel experience, basic knowledge of econometrics It is assumed that basic courses are still remembered by you. Groups of 2-3 (pls let TA know by the end of the week) Strategic Asset Allocation 4 4/9/2015 Cases What case report is NOT: – Not copy of textbook or article. – Not exercise in history of economics or finance. I do not care (at least, in that class) who got Nobel Prize for what... Ideal case report is similar to consulting report: – – – – – – Analysis of data that is in the case (preferrably statistical analysis) Covering all relevant issues (pros and cons) Take the position and defend it! Case report is not War and Peace. Be brief! Please understand what you are writing about. Cases are due before the discussion session. Do not spend more than 2 days on ANY case! Class discussion is part of the case work. Strategic Asset Allocation 5 4/9/2015 My assumptions about you: You know and understand basic regression analysis (what is R2, statistical significance, etc.) You remember conditions of optimality from Econ 101 You remember basics from Finance I You are willing to learn... Strategic Asset Allocation 6 4/9/2015 Agenda Individual’s preferences, utility function Measurement of risk by variance Diversification – – – – A bit of math Industry diversification International diversification Latest evidence Shortcut to math: Excel! Risk accounting Strategic Asset Allocation 7 4/9/2015 First Approximation Model of Investors’ Behavior: Assumptions: Single holding period Investors are risk-averse Investors are ”small” The information about asset payoffs is common knowledge Assets are in unlimited supply Assets are perfectly divisible No transaction cost Wealth W is invested in assets Strategic Asset Allocation 8 4/9/2015 Investors´ preferences Attitude to risk Time horizon (do not confuse with holding period) Non-traded risks (liabilities, labor income, human capital) Constraints Strategic Asset Allocation Advisor & Investor Type: Fidelity Conservative Moderate Aggressive Merrill Lynch Conservative Moderate Aggressive New York Times Conservative Moderate Aggressive Portfolio Cash Bond Stock Bonds/Stock 50 20 5 30 40 30 20 40 65 1.500 1.000 0.462 20 5 5 35 40 20 45 55 75 0.778 0.727 0.267 20 10 0 40 30 20 40 60 80 1.000 0.500 0.250 9 4/9/2015 Investor’s preferences:Meanvariance framework Representation by utility function of wealth W – u’(W)>0, u’’(W)<0 Taylor Expansion: 1 ~ ~ ~ ~ ~ ~ ~ ~ u (W ) u ( E (W )) u ' ( E (W ))(W E (W )) u ' ' ( E (W ))(W E (W )) 2 2 Applying Expectations operator: 1 ~ ~ ~ E u (W ) u ( E (W )) u ' ' ( E (W ))s 2 2 Simplest utility function is quadratic:u=W-0.5bW2 b ~ ~ b ~ ~ b ~ ~ E u (W ) E (W ) E (W 2 ) E (W ) E (W ) 2 s 2 f ( E (W )) s 2 2 2 2 Problem: satiation Arbitrary preferences: Asset returns are distributed as multivariate normal A dominates B if E(rA) (>) E(rB) and sA <() sB Strategic Asset Allocation 10 4/9/2015 Indifference curves All portfolios on a given indifference curve are equally desirable Any portfolio that is lying on indifference curve that is ”further North-west” is more desirable than any portfolio that is lying on indifference curve that is ”less Northwest” Different investors (e.g., in risk aversion) have different indifference curves Strategic Asset Allocation Solid line, b=1, dashed line, b=2 1.8 1.7 1.6 Exp. return 1.5 1.4 1.3 1.2 1.1 1 0 1 2 3 Std. Dev. 4 5 6 11 4/9/2015 Measuring risk by variance Variance – definition: probability weighted squared deviations from the expected value – based on probability distribution Any drawbacks of this measure? – People do not behave that way (read Odean): Overconfidence (“wrong” probability distribution) Regret (distinguish “gains” from “losses”) Should we use semi-variance? – Particularly in case of delegated portfolio management? Strategic Asset Allocation 12 4/9/2015 How to live with risk? Know and classify risks into asset classes. On what basis? Price risk (Country (incl. Political risk), $17 Industry,statistical categories) $15 Credit risk, counterparty risk $13 Tail risk or risk of ruin Most important classification concept: statistical correlation pitfalls of correlations quasi-arbitrage opportunities (“convergence trades”): LTCM and limits of arbitrage (Shleifer &Visny) Strategic Asset Allocation Large vs. Small Cap: Total Return (01.1954=$1) $11 Large Cap Small Cap Large Cap Small Cap $9 $7 HIGH CORRELATION r =0.99 $5 LOW CORRELATION r =0.5 $3 $1 1954 1959 1964 1969 13 4/9/2015 The same story: Nasdaq vs. S&P 500 Strategic Asset Allocation 14 4/9/2015 Strategic Asset Allocation 15 4/9/2015 Henry Lowenfeld, 1909 “It is significant to see how entirely all the rest of the Geographically Distributed stocks differ in their price movements from the British stock. It is this individuality of movement on the part of each security, included in a well-distributed Investment List, which ensures the first great essential of successful investment, namely, Capital Stability.” From: Investment and Exact Science, 1909. Strategic Asset Allocation 16 4/9/2015 Globalization and Financial Linkages Common wisdom is that globalization and integration of markets accentuates financial linkages (correlations) – – – – – Business cycle synchronization Policy coordination Coordination of institutions Decrease in “home bias” of investors Globalization of firms Globalization and integration also allows country specialization Strategic Asset Allocation 17 4/9/2015 What is the overall effect? Decrease in expected returns Higher correlation between asset markets More markets for investment Increase in the types of marketed securities Potential synchronization of business cycles Increased policy coordination Net effect? Strategic Asset Allocation 18 4/9/2015 International Diversification 2: Time-Varying Correlations 1. Correlations between countries are highly timevarying. 2. Result of Solnik can be due to segmentation period used. 3. There is striking similarities between end of XIX and XX centuries. (Based on Goetzmann et. al. NBER W8612) Strategic Asset Allocation 1872-2000 UK US France US France Germany 0.265 0.351 0.143 0.163 0.083 0.189 Average correlation =0.199 Integration, 18721914, 1972-2000 US France Germany UK 0.345 0.467 0.369 US 0.301 0.284 France 0.520 Average correlation =0.381 Segmentation, 1914-1971 US France Germany UK 0.193 0.311 0.097 US 0.101 0.041 France 0.135 Average correlation =0.146 20 4/9/2015 The Role of Emerging Markets Expand the investment opportunity set Are imperfectly correlated with existing markets What is the relative contribution of changing correlations and evolution in the investment opportunity set for diversification benefits? Strategic Asset Allocation 21 4/9/2015 Strategic Asset Allocation 23 4/9/2015 Decomposing the Benefits of International Diversification equally-weighted portfolio variance / average market variance Ratio portfolio volatility / average market volatility 1.0 Core Countries (limited diversification) 0.8 Average four countries All Countries (unlimited diversification) 0.6 0.4 0.2 0.0 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 Bottom Line: International Diversification Does Not Work as it Used to... •Trade barriers disappear (NAFTA, EU, ASEAN, etc.) •Globalization of Business Enterprises, •Wave of intra-industry M&A (incl. cross-border M&A) “…active portfolio managers will have increasing difficulty adding value by using a top-down strategy through European country allocation.” (Freiman, 1998) New Holy Graal: Industry Diversification Strategic Asset Allocation 25 4/9/2015 Industry vs. International Diversification APT-style estimation: Ri=ai(t)+SdijbijNatlMarketIndexj+ Sd1ijgijGlobalIndustryIndex+ i where dij (d1)=1 if firm i belongs to country (industry) j. This can be further simlified as Ri=ai(t)+Sdijbij(t)+ Sd1ikgik (t) + i 2-stage estimation as in Fama-McBeth procedure (time-series + cross-section) gives us time-series of prices of national and industry risk. One can interpret ai(t)+bij (t) is return on geographically diversified industry portfolio. ai(t)+gij(t) is return on industry-diversified national portfolio. Small Print: (a) We miss all “other” firm characteristics-size, b/m, dividend payout ratio, leverage, etc. (b)We also assume that securities in country i have same exposure to domestic and foreign factors. (c) We do not address Ericsson problem. (d) Cavaglia et. al. (2001) consider 35 industries in 21 countries. Strategic Asset Allocation 26 4/9/2015 Random diversification: international vs. industrial Strategic Asset Allocation 27 4/9/2015 Eiling, Gerard, Hillion & de Roon (2009) Adding currency deposits to industry portfolios is a winning recepie. Our conditional results show that international equity returns are primarily driven by industry and currency risk factors…The dominance of global industry factors over country factors is in line with the seven developed countries in our sample being among the most integrated equity markets in the world. Finally, we find that currency returns significantly improve the meanvariance efficiency of country, industry and world market portfolio returns. Strategic Asset Allocation 28 4/9/2015 K. Lewis: Breaks are country breaks Strategic Asset Allocation 29 4/9/2015 K. Lewis 2 Firms correlations are determined by markets’ integration, not vv. Strategic Asset Allocation 30 4/9/2015 How non-diversifiable risk changes with time (Campbell et al, 2000)? It increases... When before you were OK with 10 stocks, now you have to use 50. Why? – Younger companies are on the market – Internal capital markets are gone – Competition – Institutions Strategic Asset Allocation 31 4/9/2015 Do you really have to go abroad to achieve international diversification? (based on Diermeier-Solnik 2001) No, It is enough to invest into companies that do business abroad. Ri=ai+biLocInd+SgijIndj+ SdijCurrencyj+ i gij is “exposure” to foreign market risk, dij is “exposure” to foreign currency risk. Exposure is explained well by % of foreign sales, gij =li+mijForSalesj Adjusted R2 Strategic Asset Allocation International Market Exposure Country France 0.13 Germany 0.31 Italy 0.40 Japan 0.22 Netherlands 0.49 Switzerland 0.32 UK 0.22 US 0.02 Foreign Currency Exposure 0.06 0.09 0.00 0.24 0.19 0 0.17 0.02 32 4/9/2015 Word of caution: “Trust companies…have reckoned that by a wide spreading of their investment risk, a stable revenue position could be maintained, as it was not to be expected that all the world would go wrong at the same time. But the unexpected has happened, and every part of the civilized world is in trouble…” Chairman of Alliance Trust Company (1929) Strategic Asset Allocation 33 4/9/2015 Strategic Asset Allocation 34 4/9/2015 Non traded risks Human capital and death insurance Investment in residence Other consumption needs: saving for retirement and life insurance Liabilities: B/S optimization You must consider that these are part and parcel of your portfolio, but with immutable weights Strategic Asset Allocation 35 4/9/2015 Human Capital Most of the ”normal” individual wealth is in the form of HUMAN CAPITAL. Assume that human capital supply (willingness to work) is flexible and tradeable. Value of future cash flow decreases with time. Share of stocks will go down with time The higher is the riskiness of human capital, the less is the willingness to invest in stock Strong effect on portfolio decisions. Real estate can amplify riskiness of human capital Strategic Asset Allocation 36 4/9/2015 Normative multi-period AA: theory One risk-free asset (return r) and n risky assets with e=E[R] and var-covar matrix V. Investor’s consumption-investment problem: T t max EU Ct , Ct 1 ,...,CT max E d U C Ct , wt Ct , wt t ~ ' ~ s.t. Wt 1 Wt Ct 1 w t 1r w t R t 1 Constant relative risk aversion (CRRA) utility: C1g , g 0, g 1 U (C ) 1 g ln C , g 1 Strategic Asset Allocation 37 4/9/2015 Optimal dynamic portfolios: C U x H w AM BH M WU C W C W W C CC M V 1 E R r1, H V 1 ν, ν ' (s 1x ,...) M is mean-variance portfolio H is hedge portfolio against changes in variable x. H does not matter for non-stochastic opportunity set or log –utility function. Strategic Asset Allocation 38 4/9/2015 Constraints Liquidity Regulations: public or self imposed SEC Pension funds: Employee Retirement Income Security Act (ERISA); European directives no more than 5% in any publicly traded company Mostly domestic assets Mutual funds: No borrowing. Association for Investment Management and Research (AIMR) Taxes Unique needs: internal restrictions Strategic Asset Allocation 39 4/9/2015 Frontier with constraints S&P500 (51.8%) IA Small Stocks (7.5%) MSCI Europe (19.7%) S&P500 (20.0%) MSCI Pacific TR (7.5%) IA 5YR Gvt (4.5%) IA Small Stocks (2.6%) MSCI Europe (11.9%) Real Estate (20.0%) IA 1 Year Gvt (20.0%) Real Estate (23.5%) MSCI Pacific TR (10.2%) Expected Return 19.0 18.0 IA Small Stocks 17.0 16.0 14.0 Source:Ibbotson Assoc. Portfolios with s=20% No short sales B: 20% max MSCI Pacific TR CRSP MidCap 15.0 S&P500 MSCI Europe 13.0 12.0 Russell 2000 11.0 10.0 9.0 8.0 Real Estate 7.0 6.0 IA Corporate IA 1 Year Gvt IA 5YR Gvt 5.0 4.0 IA 20Yr Gvt Non-US LT Gvt 30 Day TBILL 3.0 Gold 2.0 Strategic Asset Allocation 0.0 3.0 6.0 9.0 12.0 15.0 18.0 21.0 24.0 Standard Deviation (Risk) 27.0 30.0 33.0 36.0 40 4/9/2015 39.0 42.0 Time ”Diversification” Can you reduce risk by holding assets longer? – Uncertainty in annual rate of return goes down – BUT!!! Uncertainty of total returns goes up Position 56 Return Percentiles Position 56 Compound Annual Return 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% -10.0% -20.0% Nov 2001 Source: Ibbotson Assoc. R=15%, s=20% S&P500 (37.6%) IA Small Stocks (23.2%) 95th Percentile Dec 2005 Dec 2010 Time Expected Value Dec 2015 Nov 2020 5th Percentile Wealth Percentiles Position 56 Wealth (USD) 50 10 Real Estate (8.1%) 1 1 MSCI Pacific TR (12.5%) MSCI Europe (18.7%) Nov 2000 95th Percentile Strategic Asset Allocation Dec 2005 Expected Value Dec 2010 Time 5th Percentile Dec 2015 Nov 2020 41 4/9/2015 Practicality: Estimation Risk Óptimization results are usually suffering from: – Huge short positions in many assets in no-constraint case. – “Corner” solutions with zero positions in may assets if constraints are imposed. – Huge positions in obscure markets with small cap – Large shifts in positions when exp. returns or covariances changes just a bit… All of those are coming from one common cause: difficulties in estimation of expected returns. Strategic Asset Allocation 42 4/9/2015 Another example (GS 2003): Here are forecasts given by “Wall Street Protagonist” (Columns 1 and 2) and set of portfolio weights that are generated by it. RETURN STD DEV Japanese Gov Bonds 4.7% 4.2% EU Gov Bonds 5.1% 3.6% US Gov Bonds 5.2% 4.6% US Equities 5.4% 15.5% Global Fixed income 6.0% 3.6% EU Equities 6.1% 16.6% US Inv Grade Corp Bonds 6.3% 5.4% EAFE 8.0% 15.3% Hedge Fund Portfolio 8.0% 5.2% US High Yield 8.9% 7.3% Private Equity 9.0% 28.9% Emerging Debt 9.0% 17.6% REITs 9.0% 13.0% Japanese Equity 9.5% 19.6% Emerging Equities 11.8% 23.4% Portfolio ER Portfolio Volatility Asset Pricing Models Unconstrained No Short weights Sales -2.02 0.00 -3.21 0.00 -4.84 0.00 -0.11 0.00 14.93 0.00 -2.58 0.00 -3.86 0.00 3.14 0.00 0.58 0.55 -0.10 0.36 0.01 0.00 -0.29 0.00 0.04 0.08 -0.72 0.01 0.03 0.00 4.90% 18.20% 5.10% 8.40% 43 4/9/2015 Equilibrium and individual asset’ expected return One can expect that ERP is between 4 and 6% and is fairly stable with time (see later in the course) One can make forecast for individual assets that are different from long term premium. But by forecasting one asset class, we are implicitly making forecast for other asset classes as well. Asset Pricing Models 44 4/9/2015 Practicality: How to express views? Method is due to Black & Litterman (Goldman Sachs). The core themes: equilibrium returns and views. Investors normally have views/preferences. They are NOT incorporated into optimization process. Views=mathematically expressed preferences of individual investors. Step Action 1 2 3 4 5 6 Calculate equilibrium returns Define weight for news Set target tracking error Set target market exposure Get portfolio weight Examine risk distribution Asset Pricing Models Purpose Set neutral reference point Dampen impact of aggressive news Control risk wrt benchmark control directional effect Find allocation that maximize performance Is risk diversifies? 45 4/9/2015 Equilibrium optimal portfolio Imagine that the investor thin that US is still in recession. Thus, stocks will perform badly, and bonds will perform OK. Mathematically, it is equivalent to assuming that bonds will go up 0.8%, and stocks will drop 2.5% Result: see Table 8: Asset Pricing Models 46 4/9/2015 Updating of discrete probabilities 1. We have a probability estimate for event H: prior probability P(H) 2. New information D is gained 3. Update the estimate using Bayes’ theorem: posterior probability P(H|D) Asset Pricing Models 47 4/9/2015 The Bayes’ theorem The updating is done using the Bayes’ theorem: P( D | H ) P( H ) P( H | D) P( D) Asset Pricing Models 48 4/9/2015 Example: Using Bayes’ theorem 1,5 % of the population suffer from schizophrenia P(S) = 0.015 (prior probability) Brain atrophy is found in – 30 % of the schizophrenic P(A|S) = 0.3 – 2 % of normal people P(A|S) = 0.02 If a person has brain atrophy, the probability that he is schizophrenic (posterior probability) is: P( A | S ) P( S ) P( S | A) P( A | S ) P( S ) P( A | S ) P( S ) 0.3 0.015 0.186 0.3 0.015 0.02 0.985 Picture: Clemen s. 250 Figure: Posterior probability with different prior probabilities. Asset Pricing Models 49 4/9/2015 Updating of continuous distributions Choose a theoretical distribution, P(X=x|), for the physical process of interest. Assess uncertainty about parameter : prior distribution, f() Note: Uncertainty about X has two parts: Observe data x1 Update using Bayes’ theorem: posterior distribution of , f(|x1) Asset Pricing Models 1. Due to the process itself, P(X=x|). 2. Uncertainty about , f(), later updated to f(|x1). 50 4/9/2015 Updating of continuous distributions Bayes’ theorem for continuous : f ( | x1 ) f ( x1 | ) f ( ) f ( x1 | ) f ( )d f(x1|) is called the likelihood function of with a given observed data x1. In most cases the posterior distribution can not be calculated analytically, but must be solved numerically. Asset Pricing Models 51 4/9/2015 Normal distribution 1. The physical process of interest is normal distributed: X ~ N(m, s2) (s is assumed to be known) 2. Prior distribution for m: m ~ N(m0, s20) (notation: s20 = s2 / n0) 3. Observe a sample of the physical process: – sample size: n1 – sample mean: x1 4. The posterior distribution, calculated using the Bayes’ theorem, gets reduced to: m0s 2 n1 x1s 02 n0 m0 n1 x1 m* 2 2 m ~ N(m*, s2*), where s n1 s 0 n0 n1 s 02s 2 n1 s2 s * 2 2 s n1 s 0 n0 n1 2 Asset Pricing Models 52 4/9/2015 Expressing Views Thus, one need to specify set of “views” and “precisions” of views for each asset f(x|m). “No views” is equivalent to having sx For this case posterior=prior. Model will deviate further for assets where views are stronger. All assets are affected: Asset Pricing Models 53 4/9/2015 Further risk control & results: Minimize tracking error Mkt. Exposure of the portfolio (b) (neutral should be 1) Look at diversification (are all eggs in one basket?) Results (according to GS, 95-97): (+103 bp, +83 bp, -26bp) Asset Pricing Models 54 4/9/2015 Other uses: Black-Litterman model is essentially Tactical Asset Allocation model (provided that algorithm of selecting “views” is specified). But it can be used effectively in updating priors on the distribution of the signals. It can be used to bring in new asset classes for which the recorded history is short or unreliable (venture capital funds, hedge funds, emerging markets, etc.) Asset Pricing Models 55 4/9/2015 What is expected risk premium? Expected risk premium= Var(rM)/E[W/] Plays central role in any discussion about the market What is that? How to measure it? What will it tell us about mankind and economy (or Asset Pricing Model)? – Historical perspective – Equilibrium perspective Asset Pricing Models 56 4/9/2015 Asset Classes Returns: US History Asset Pricing Models 57 4/9/2015 Asset Classes Returns: Swedish History 1 SEK invested in 1900 1000000 100000 10000 DMS Global Sweden Bill TR DMS Global Sweden Inflation DMS Global Sweden Equity TR DMS Global Sweden Bond TR 1000 100 10 1 1900 0.1 Asset Pricing Models 1920 1940 1960 1980 2000 58 4/9/2015 Goetzmann&Jorion: International Evidence Asset Pricing Models 59 4/9/2015 Goetzmann&Jorion: International Evidence Median Market RR 0.75% GDP-weighted RR 4% Asset Pricing Models 60 4/9/2015 Estimates from historical data Ibbotson & Sinquefield (76): Real ERP= 5% Ibbotson & Chen (2000) 4% Fama & French (2002) longer period 4.4% Jagannathan, McGrattan,Scherbina (2000) – 1926-70 ERP=7% – 1971-99 ERP=0.7% Asset Pricing Models 61 4/9/2015 What is going on 2day? Strategic Asset Allocation 62 4/9/2015 Equilibrium Approach: C-CAPM Individuals have preferences over consumption C described by CRRA u=-C1-g Certainty case: marginal utility of consumption today =discounted marginal utility of consumption tomorrow times teturn of asset ri: C-gt[(1+ri)/(1+r)] C-gt+1 In case of uncertainty C-gtE[(1+ri)/(1+r)] C-gt+1 Introducing consumption growth g=C(t+1)/C(t)-1 Asset Pricing Models 63 4/9/2015 C-CAPM(2) E 1 r 1 g 1 r g i T aylorexpansion: 1 ri 1 g 1 ri gg ggri g g (1 g ) 2 g2 ApplyingExpectations operator: r Eri gEg g E ri Eg covri , g g (1 g ) Eg Varg 2 If t is small, thenquadratic termscan be disregarded : g (1 g ) r Eri gEg g covri , g Varg 2 T hiseq. should be true for both risky and riskless assets: g (1 g ) r rf gE g Varg Eri rf g covri , g 2 Asset Pricing Models 2 64 4/9/2015 C-CAPM: Equity premium puzzle Mehra&Prescott(85); Mankiv &Zeldes(91): – – – – 1890-1979[1948-88]: Risk premium=0.06 [.08] Std of consumption growth =0.036 [0.014] Std of market returns=0.167 [0.14] Correlation between consumption growth and market returns = 0.40 [0.45] – 0.06=g*0.40*0.167*.036 => g=25 [90] Asset Pricing Models 65 4/9/2015 Equity premium puzzle Gamble: take 20% paycut if state of the world is ”bad” (prob=1/2) and stay at your current salary in good state, or agree on permanent cut of X%: 0.5*(0.81g+1)=x1g Asset Pricing Models 66 4/9/2015 Equity premium puzzle Gamble: take 20% paycut if state of the world is ”bad” (prob=1/2) and stay at your current salary in good state, or agree on permanent cut of X%: 0.5*(0.81g+1)=x1g If g=25 then x=17.7% Realistic estimate for gamma=3 Asset Pricing Models 67 4/9/2015 How can we solve it? Habit formation u=-(Ct-lCt-1)1-g – Increases demand for bonds, lower Rf – “Keeping up with the Joneses”: instead C(t-1) there is AVERAGE consumption in the reference group. Idiosynchratic labor risk Disaster states and survivorship bias. Liquidity premium Limited Participation Asset Pricing Models 68 4/9/2015