Empirical Financial Economics The Efficient Markets Hypothesis Stephen J. Brown NYU Stern School of Business 2009 Merton H. Miller Doctoral Seminar Major developments over last 35 years Portfolio theory Major developments over last 35 years Portfolio theory Asset pricing theory Major developments over last 35 years Portfolio theory Asset pricing theory Efficient Markets Hypothesis Major developments over last 35 years Portfolio theory Asset pricing theory Efficient Markets Hypothesis Corporate finance Major developments over last 35 years Portfolio theory Asset pricing theory Efficient Markets Hypothesis Corporate finance Derivative Securities, Fixed Income Analysis Major developments over last 35 years Portfolio theory Asset pricing theory Efficient Markets Hypothesis Corporate finance Derivative Securities, Fixed Income Analysis Market Microstructure Major developments over last 35 years Portfolio theory Asset pricing theory Efficient Markets Hypothesis Corporate finance Derivative Securities, Fixed Income Analysis Market Microstructure Behavioral Finance Efficient Markets Hypothesis ln pt E[ln pt | it ] E[ln pt | t ] which implies the testable hypothesis ... E [ln pt E (ln pt | t )] zt 0 t where zist part of the agent’s information set it In returns: E [rt E ( rt | t )] zt 0 wher rt ln pt ln pt t e Efficient Markets Hypothesis E rt E rt | t zt 0 t Tests of Efficient Markets Hypothesis What is information? Does the market efficiently process information? Estimation of parameters What determines the cross section of expected returns? Does the market efficiently price risk? Tests of Efficient Markets Hypothesis E rt E rt | t t 0 t Weak form tests of Efficient Markets Hypothesis 1 sell Example: trading rule tests Semi-strong form tests of EMH 1 Example: Event studies t 0 1 t 0 1 bad news no news good news Strong form tests of EMH Example: Insider trading studies (careful about conditioning!) hold buy Random Walk Hypothesis E rt E rt | t zt 0 t Random Walk Hypothesis E rt E rt | t zt 0 t zt rt t Random Walk Hypothesis E rt E rt rt 0 Serial covariance tests Random Walk Hypothesis E rt E rt rt 0 Serial covariance tests: E rt E rt E rt 0 Random Walk Hypothesis E rt E rt rt 0 Serial covariance tests E rt E rt rt E rt 0 Random Walk Hypothesis E rt E rt rt 0 Serial covariance tests Variance Ratio tests 1 Var( rt ) k 1 2 (1 ) ( k )2( 1)] 1 Var( rt ) k 1 Random Walk Hypothesis E rt E rt rt 0 Serial covariance tests Variance Ratio tests Momentum literature E rj ,t E rj ,t rj ,t 0 Random Walk Hypothesis E rt E rt rt 0 Serial covariance tests Variance Ratio tests Momentum literature E rj ,t E rj ,t rj ,t rt 0 Random Walk Hypothesis E rt E rt rt 0 Serial covariance tests Variance Ratio tests Momentum literature E rj ,t E rj ,t rj ,t rt 0 Zero investment portfolio Random Walk Hypothesis E rt E rt rt 0 Serial covariance tests Variance Ratio tests Momentum literature Assumes stationarity Random Walk Hypothesis E rt E rt rt 0 Serial covariance tests Variance Ratio tests Momentum literature Assumes stationarity t Random Walk Hypothesis E rt E rt rt 0 Serial covariance tests Variance Ratio tests Momentum literature Assumes stationarity Neither necessary nor sufficient for EMH Trading rule tests of EMH E rt E rt | t t 0 t 1 t 0 1 sell hold buy Trading rule tests of EMH E rt E rt | t t 0 t Timmerman (2007) survey 1 t 0 1 Naïve models using past sample means hard to beat Recent financial data is most relevant Short lived episodes of limited predictability sell hold buy Trading rule tests of EMH E rt E rt | t t 0 t Timmerman (2007) survey 1 t 0 1 sell hold buy Naïve models using past sample means hard to beat Recent financial data is most relevant Short lived episodes of limited predictability Predictability is not profitability Necessity: Do not consider all possible patterns of returns Sufficiency: Cannot profit if all markets rise and fall together Trading rule tests of EMH E rt E rt | t t 0 t Timmerman (2007) survey 1 t 0 1 sell hold buy Naïve models using past sample means hard to beat Recent financial data is most relevant Short lived episodes of limited predictability Predictability is not profitability Necessity: Do not consider all possible patterns of returns Sufficiency: Cannot profit if all markets rise and fall together An important seminal reference … Trading Rules: Cowles 1933 Cowles, A., 1933 Can stock market forecasters forecast? Econometrica 1 309-325 William Peter Hamilton’s Track Record 1902-1929 Classify editorials as Sell, Hold or Buy Eˆ [rt E rt | t ] t 3.5% Return on DJI Novel bootstrap in strategy space 1 t 0 1 41 sell 74 hold 140 buy Trading rule predicting sign of excess return January 1970 - December 2005 Trading rule value S&P500 value Factor-augmented AR logit based on prior 120 month rolling window Cowles Bootstrap Jan 1970-Dec 2005 Annualized excess fund return Sharpe ratio of fund Sharpe ratio of S&P500 Peseran & Timmermann (1992) p-value Cowles bootstrap p-value 2.203% 0.063 0.049 4.83% 6.32% Standard Event Study approach EVEN T rt1 u01 u11u21 … EVEN T rt2 u02 u12u22 … EVEN T rt3 u03 u13u23 … 0 EVEN T EVEN T u04 u14u24 … u05 u15u25 … 5 10 15 20 25 30 rt4 t Orthogonality condition Event studies measure the orthogonality condition E[rt E[rt | t )] zt 0 i I all using the average value of the residual across events u i [ri ,ti E ( ri ,ti | ti , rM ,ti )] zti wherezt 1 is good news zand t 1 is bad news If the residuals are uncorrelated, then the average residual will be asymptotically Normal with expected value equal to the orthogonality condition, provided that the event zt has no market wide impact Fama Fisher Jensen and Roll Cumulat ive average residual - Um Cumulat ive residuals around st ock split 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 -30 -20 -10 0 10 20 Mont h relat ive t o split - m 30 FFJR Redux Cumulat ive average residual - Um Cumulat ive residuals around st ock split 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 -30 -20 -10 0 10 20 Mont h relat ive t o split - m 30 Original FFJR results Cumulat ive average residual - Um Cumulat ive residuals around st ock split 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 -30 -20 -10 0 10 20 Mont h relat ive t o split - m 30 Asset pricing models: GMM paradigm E rt E rt | t zt 0 t Match moment conditions with sample moments Test model by examining extent to which data matches moments Estimate parameters Example: Time varying risk premia Time varying risk premia t 0 X t 1 imply a predictable component of excess returns rt rf 0 X t 1 f t B t where the asset pricing model imposes B constraint Estimating asset pricing models: GMM Define residuals t rt (rf 0 X t 1 ft B) Residuals should not be predictable using instruments zt-1 that include the predetermined variables1Xt-1 t zt 1 E{[rt E (rt | , X t 1 ) ]zt 1} 0 T t Choose parameters to minimize residual 1 predictability z 0 T t t t 1 Estimating asset pricing models: Maximum likelihood t Define residuals rt (rf 0 X t 1 ft B) 1 2 Choose parameters to minimize t T t Relationship to GMM: when instruments zt include the predetermined variables Xt-1 1 FOC : t zt 1 0 T t Conclusion Efficient Market Hypothesis is alive and well EMH central to recent developments in empirical Finance EMH highlights importance of appropriate conditioning in empirical financial research