Week 10 GD31403 Multifactor Models of Risk and Return Arbitrage Pricing Theory • CAPM is criticized because of: – The many unrealistic assumptions – The difficulties in selecting a proxy for the market portfolio as a benchmark – The growing empirical evidence suggesting the need to consider additional risk variables such as size, P/E, BV/MV. 10-2 • An alternative pricing theory with fewer assumptions was developed by Stephen Ross (Ross, 1976, 1977): Arbitrage Pricing Theory (APT) Arbitrage Pricing Theory • Major assumptions of APT: – Capital markets are perfectly competitive – Investors always prefer more wealth to less wealth with certainty – The stochastic process generating asset returns can be expressed as a linear function of a set of K factors • In contrast to CAPM, APT doesn’t assume: – Normally distributed security returns – Quadratic utility function – A mean-variance efficient market portfolio 10-3 Arbitrage Pricing Theory The APT Model E Ri 0 1bi1 2bi 2 ... K biK where: λ0 = the expected return on an asset with zero systematic risk λj= the risk premium related to the j-th common risk factor bij= the responsiveness of asset i’s returns to movements in the j-th common risk factor 10-4 Arbitrage Pricing Theory CAPM versus APT 10-5 CAPM E ( Ri ) = RFR + bi éë E ( RM ) - RFR ùû APT E Ri 0 1bi1 2bi 2 ... K biK CAPM APT Form of Equation Linear Linear Number of Risk Factors Factor Risk Premium Factor Risk Sensitivity “Zero-Beta” Return 1 [E(RM) – RFR] βi RFR K (≥ 1) { λj } { bij } λ0 Arbitrage Pricing Theory • Unlike CAPM that is a one-factor model, APT is a multifactor pricing model • However, unlike CAPM that identifies the market portfolio return as the factor, APT model does not specifically identify these risk factors in terms of quantity (how many there are) or their identity (what they are) • These multiple factors include: 10-6 − − − − Inflation Growth in GNP Major political upheavals Changes in interest rates Arbitrage Pricing Theory • Instead of SML, APT implies a security market plane- K risk factors and one dimension for the security ’ s expected return (see Exhibit 9.2 of Textbook, pp.240) • Similar to CAPM, the APT assumes that the unsystematic risk will be diversified away in a large portfolio. • In equilibrium when the unsystematic risk are fully diversified, the APT requires that the return on a zero-systematic-risk portfolio is zero (λ0 =0). 10-7 APT: Illustration 1 APT E Ri 0 1bi1 2bi 2 ... K biK (1) Selecting Risk Factors (How many and what are these factors?) • As discussed earlier, the primary challenge with using the APT is identifying the risk factors • For this illustration, assume that there are 2 common factors: − Unanticipated changes in the rate of inflation 10-8 − Unexpected changes in the growth rate of real GDP APT: Illustration 1 (2) Determining the Risk Premium 10-9 • λ1: The risk premium related to the first risk factor is 2 percent for every 1 percent change in the rate (λ1=0.02) • λ2: The average risk premium related to the second risk factor is 3 percent for every 1 percent change in the rate of growth (λ2=0.03) • λ0: The rate of return on a zero-systematic risk asset (i.e., zero beta) is 4 percent (λ0=0.04) APT: Illustration 1 (3) Determining the Sensitivities for Asset X and Asset Y bX1 = The response of asset X to changes in the inflation factor is 0.50 (bX1=0.50) bX2 = The response of asset X to changes in the GDP factor is 1.50 (bX2=1.50) bY1 = The response of asset Y to changes in the inflation factor is 2.00 (bY1=2.00) bY2 = The response of asset Y to changes in the GDP factor is 1.75 (bY2=1.75) 10-10 APT: Illustration 1 (4) Estimating the Expected Return, E(Ri) APT Asset X E Ri 0 1bi1 2bi 2 = 0.04 + 0.02 bi1 0.03 bi 2 E RX 0.04 + 0.02 0.50 0.031.50 = 0.095 @ 9.5% Asset Y E RY 0.04 + 0.02 2.00 0.031.75 = 0.1352 @ 13.25% 10-11 Is it possible to overvalued assets? identify undervalued or APT: Illustration 2 • Suppose that 3 stocks (A, B, C) and 2 common systematic risk factors have the following relationship (assume λ0=0 ): E(RA) = (0.8) λ1 + (0.9) λ2 E(RB)=(-0.2) λ1 + (1.3) λ2 E(RC)=(1.8) λ1 + (0.5) λ2 APT 10-12 E Ri 1bi1 2bi 2 APT: Illustration 2 • Further assume that the risk premium for the 2 common factors are: λ1= 4% λ2= 5% E(RA) = (0.8)(0.04) + (0.9)(0.05) = 0.077 @ 7.7% E(RB) = (-0.2)(0.04) + (1.3)(0.05) = 0.057 @ 5.7% E(RC) = (1.8)(0.04) + (0.5)(0.05) = 0.097 @ 9.7% 10-13 APT: Illustration 2 • Now, assume that all three stocks are currently priced at $35 and do not pay a dividend • With the expected returns of E(RA)=7.7%, E(RB)=5.7% and E(RC)=9.7%, we can estimate the stock price (consistent with investor return expectations) one year later: E(PA) = $35(1+7.7%) = $37.70 E(PB) = $35(1+5.7%) = $37.00 10-14 APT: Illustration 2 • If one “ knows ” actual future prices (from fundamental or technical analysis) for these stocks are different from those previously estimated, then these stocks are either undervalued or overvalued • Assume are : Stock A Stock B Stock C 10-15 that the forecasted future prices $37.20 $37.80 $38.50 (Est. Return = 6.29%) (Est. Return = 8.00%) (Est. Return = 10.00%) APT: Illustration 2 Expected Price A $37.70 B $37.00 C $38.40 Forecasted Price Evaluation Recommend $37.20 Overvalued SELL $37.80 Undervalued BUY $38.50 Undervalued BUY You will get the same conclusion if using returns! 10-16 APT: Illustration 2 Computed from CAPM Required Return A 7.7% B 5.7% C 9.7% Derived from fundamental or technical analysis Estimated Return Evaluation Recommend 6.29% Overvalued SELL 8% Undervalued BUY 10% Undervalued BUY Undervalued: Estimated Return > Required Return Overvalued: 10-17 Estimated Return < Required Return Empirical Tests of the APT Roll-Ross Study (1980) • The methodology used in their study is as follows: 10-18 - Estimate the expected returns and the factor coefficients from time-series data on individual asset returns - Use these estimates to test the basic crosssectional pricing conclusion implied by the APT • The authors concluded that the evidence generally supported the APT, but acknowledged that their tests were not conclusive Empirical Tests of the APT Extensions of the Roll-Ross Study 10-19 • Cho, Elton, and Gruber (1984) examined the number of factors in the return-generating process that were priced • Dhrymes, Friend, and Gultekin (1984) reexamined techniques and their limitations and found the number of factors varies with the size of the portfolio • Connor and Korajczyk (1993) developed a test that identifies the number of factors in a model that does allow the unsystematic components of risk to be correlated across assets Empirical Tests of the APT Shanken’s Challenge to Testability of the APT 10-20 • APT has no advantage because the factors need not be observable, so equivalent sets may conform to different factor structures • Empirical formulation of the APT may yield different implications regarding the expected returns for a given set of securities • Thus, the theory cannot explain differential returns between securities because it cannot identify the relevant factor structure that explains the differential returns [similar to the Roll (1977) critique of CAPM] Multifactor Models • The empirical challenge with CAPM is to accurately estimate the market portfolio. • The main practical problem with the APT is that neither the identity nor the exact number of the risk factors are developed by theory. 10-21 • Multifactor models of risk and return attempt to bridge this gap by specifying the exact number and identity of risk factors (macroeconomic variables or microeconomic variables). Multifactor Models In Practice 1) Macroeconomic-Based Risk Factor Models: Risk factors are viewed as macroeconomic in nature 1) Microeconomic-Based Risk Factor Models: Risk factors are viewed at a microeconomic level by focusing on relevant characteristics of the securities themselves 1) Extensions of Characteristic-Based Risk Factor Models 10-22 Macroeconomic-Based Risk Factor Models Security return are governed by a set of broad economic influences in the following fashion by Chen, Roll, and Ross in 1986: Rit ai [bi1 Rmt bi 2 MPt bi 3 DEIt bi 4UI t bi 5UPRt bi 6UTSt ] eit where: Rm= the return on a value weighted index of NYSE-listed stocks MP=the monthly growth rate in US industrial production DEI=the change in inflation, measured by the US consumer price index UI=the difference between actual and expected levels of inflation UPR=the unanticipated change in the bond credit spread UTS= the unanticipated term structure shift (long term less short term RFR) 10-23 Macroeconomic-Based Risk Factor Models 10-24 • The economic significance of the risk factors changed dramatically over time • The coefficient for stock market proxy is never significant Macroeconomic-Based Risk Factor Models Burmeister, Roll, and Ross (1994) analyzed the predictive ability of a model based on the following set of macroeconomic factors: • • • • • 10-25 Confidence risk Time horizon risk Inflation risk Business cycle risk Market timing risk Microeconomic-Based Risk Factor Models Fama and French (1993) developed a threefactor model specifying the risk factors in microeconomic terms using the characteristics of the underlying securities: ( Rit - RFRt ) ai bi1 ( Rmt - RFRt ) bi 2 SMBt bi 3 HMLt eit – SMB (i.e. small minus big) is the return to a portfolio of small capitalization stocks less the return to a portfolio of large capitalization stocks – HML (i.e. high minus low) is the return to a portfolio of stocks with high ratios of book-to-market values less the return to a portfolio of low book-to-market value stocks 10-26 Microeconomic-Based Risk Factor Models Carhart (1997), based on the Fama-French three factor model, developed a four-factor model by including a risk factor that accounts for the tendency for firms with positive past return to produce positive future return ( Rit - RFRt ) ai bi1 ( Rmt - RFRt ) bi 2 SMBt bi 3 HMLt bi 4 MOM t eit where, MOMt = the price momentum factor, the return difference between stocks with positive and negative excess returns in the recent past 10-27 Extensions of Characteristic-Based Risk Factor Models One type of security characteristic-based method for defining systematic risk exposures involves the use of index portfolios (e.g. S&P 500, Wilshire 5000) as common risk factors such as the one by Elton, Gruber, and Blake (1996), who rely on four indexes: 10-28 • The S&P 500 • The Lehman Brothers aggregate bond index • The Prudential Bache index of the difference between large- and small-cap stocks • The Prudential Bache index of the difference between value and growth stocks Extensions of Characteristic-Based Risk Factor Models The BARRA Model: Develop a model using the following Characteristic-based the risk factors 10-29 • Volatility (VOL) • • • • • • • • • • • • Momentum (MOM) Size (SIZ) Size Nonlinearity (SNL) Trading Activity (TRA) Growth (GRO) Earnings Yield (EYL) Value (VAL) Earnings Variability (EVR) Leverage (LEV) Currency Sensitivity (CUR) Dividend Yield (YLD) Nonestimation Indicator (NEU) Conclusion • The CAPM and APT will continue to be used to value assets. • Future work on this area will continue to seek to identify the set of factors that best captures the relevant dimension of investment risk. • Another area is to explore the intertemporal dynamics of the models (e.g., factor betas and risk premia that change over time). 10-30