Household Investor Expectations of Risk and Return on Stocks: Are Sharpe Ratios Countercyclical? Gene Amromin and Steven Sharpe Chicago Fed and the Federal Reserve Board January 2, 2009 paper & remarks reflect our own views, and not necessarily those of the Board of Governors or the Federal Reserve System 1 Motivation 1: What drives cyclicality of returns Huge literature on predictability of stock returns Grown from findings that macro variables “predict” equity returns/premium Fama and French (1989) – D/P, other “business cycle” factors Lettau and Ludvigson (2001) – CAY, consumption-wealth ratio, evokes cycle story: • “When excess returns are expected to be higher, forward-looking investors will react by… allowing consumption to rise above its common trend w/ wealth” • Rational story, still leaves question, why expected (required) returns vary Lead to… New theories of household risky asset demand (Cochrane 2005) Time-varying risk aversion: Campbell and Cochrane (1999) Time-varying risk: Constantinides and Duffie (1996) 2 Motivation 1 (cont) Which has led to… New studies of household-level behavior Household portfolio dynamics -- Brunnermeier and Nagel (2008) Evidence on habit formation -- Dynan (2000) Ravina (2005) So we step back, consider… What could we learn if we could ask relevant households about their beliefs? How do their expected equity returns vary with perceptions of business cycle? (especially the most influential--the more sophisticated or wealthy) How do their perceptions of risks in equity returns vary with the cycle? 3 Motivation 2: Broader Q: What influences investor beliefs? Controlling for perceptions of economy, how do perceptions of RETURN & RISK vary: Demographic characteristics Education Past experience Measured in cross section, but also potential time series interp. The relevance of survey-reported perceptions relevant: Related to respondent portfolio decisions? Wish granted: In 1999, devised insert to Michigan survey of consumer sentiment 4 Previous studies: Survey beliefs & stock market Individual investor expectations for returns Fisher and Statman, 2002 Vissing-Jorgensen, 2003; • UBS-Gallup Survey: Persistence of past returns; Effect of wealth Dominitz and Manski, 2003, 2005 (Michigan survey) • “Probability typical mutual fund will increase” (related to expected return, also risk) • Document effects of expected business conditions, cross-sectional heterogeneity, extrapolation; gender & education effects CFO expectations: Graham & Harvey (2003); G&H with Ben-David (2007) • Expected Returns & Risk: ST forecasts show persistence, no risk-return relation • Evidence of overconfidence: tighter return distribution --> aggressive corp. policies Studies of Consumer Confidence Index (Michigan) & stock returns Qui and Welch (2006) – “sentiment” & actual returns Lemmon and Portniaguina (2006) – “sentiment” vs. fundamentals 5 Road map & Summary of results Expected returns Measures contradict inferences of predictability studies (D/Y, CAY) • Gallup-UBS survey data Expected returns are procyclical • positively related to expected business conditions • expected by self and by “consensus” (so not expected news) Determinants of perceived risk Uncertainty varies inversely with expected economic conditions; • Given above, implies procyclical Sharpe ratios Individual characteristics, heuristics have strong affect perceived risk Portfolio allocations consistent with beliefs? Reported portfolio equity shares (+) in returns and (−) in risk 6 Data – Michigan Survey special supplement Criterion household needs to pass: Equity ≥ $5000 35%-45% of respondents 150-250 respondents per survey month 22 irregularly spaced surveys, Sept. 2000 – Oct. 2005 Data quality filters Response to all 3 questions on ER Survey-giver’s codes indicating low quality responses Analysis in appendix 7 Expected stock returns, survey means Gallup/UBS 12-month ahead (own) vs. Michigan 3-yr (mkt) 8 Gallup/UBS 12-mo ER vs. CAY (+) coef. in realized return regressions (so L-L are on to something, but their interpretation contradicts that of actual consumers) 9 Gallup/UBS 12-mo. ER vs. log(D/P) Literature: Positive coef. in regression using realized returns, low R-squares 10 Next step: Relating Expected Return (Mich. survey) to expected economic conditions BUS5. Looking ahead, which is more likely -Business continuous good times during the next 5 years, or cycle periods of widespread unemployment or depression, or what? [coded -2,-1,0,1,2] Nearterm “news” BEXP. A year from now, do you expect that in the country as a whole, business conditions will be better, or worse than at present, or about the same? What do you think chances are your family income will increase by more than the rate of inflation in the next five years or so? Own Prospects 11 Expected Return Regressions (3-yr ER) Regressors Good times, next 5 yrs [2, -2] Coefficient (t-stat) 0.28 (5.8) Good times-survey mean 1.52 (8.2) Good times-deviation from mean 0.23 (4.6) Better Conditions-12 mos. + (3.1) + (2.9) Chance own income > inflation + (4.3) + (4.0) Past S&P return (time-series) + (10.2) + (9.3) Gender=male + (2.7) + (2.9) (1) Procyclical ER; Past return (+); gender effect (2) Consensus effect even stronger Findings identical for half of sample w/ largest equity holdings 12 Measuring perceived risk (volatility) The survey asks for confidence interval around ER: “… what is the chance that the average return over the next 10 to 20 years will be within 2 percentage points of your [expected return]…?” Define uncertainty as inverse: 100 – probability in interval with distributional assumption, can map uncertainty σ (std. dev.) 13 Regressors to explain Perceived Risk Measures of expected economic conditions Confidence in own ability to predict (uncertainty) Knowledge: Higher education / years of investment experience Gender: male (Beyer, 1990) Representativeness (Tversky & Kahneman, 1982) Outcomes “representative” of available evidence may seem more likely If event (Return being close to ER) consistent with “salient” available evidence (past recalled S&P returns), put higher probability on event 14 Expected Risk Regressions Dependent Variable: Uncertainty = Prob |R − Re|>2% (2) all obs. (2069) (3) Prob≠50 (1413) Specification (3) excludes 50-50 answers Good times, next 5 yrs -0.70 (2.4) -1.03 (3.0) Better conditions next 12 months -0.14 (0.3) -0.25 (0.4) Better conditions/times reduces perceived risk Chance own income > inflation -0.10 (6.5) -0.12 (5.3) Abs [Expected R – Recalled R] 0.61 (9.5) 0.94 (6.5) Male -4.27 (5.1) -5.69 (4.6) College Degree -6.69 (6.0) -9.06 (4.8) Pseudo R-squared 0.106 0.143 15 Better own prospects reduce market risk Representativeness 10% discrepancy raises uncertainty 6.1% Confidence, knowledge lowers uncertainty Can reported ER and σ explain actual behavior? Are expectations summarized earlier relevant to portfolio decisions? Survey question: Fraction of financial wealth in stocks five discrete buckets {<10, 10-25, 25-50, 50-75, >75} Classic Samuelson portfolio (CRRA preferences) Portfolio fraction = (R i - rf) / γ i σi 2 For regression: log (fraction) = log (R − rf) − log σ2 − log γ 16 Portfolio choice regressions: Samuelson model Dependent Variable: Log portfolio fraction in stocks (Table 7) Log expected excess returns 0.04 (3.4) 0.15 (2.9) Log expected returns Log expected volatility Adjusted R-squared -0.09 (7.0) -0.09 (7.3) 0.042 0.046 Expected return, risk significant; signs consistent with theory Coefficients small compared to theory Not shown: sluggish adjustment (yrs invest experience matters) 17 Conclusions Summary of results Expected returns vary (+) with expected macro conditions (procyclical) Uncertainty (risk) varies (-) with expected macro conditions Uncertainty varies (-) with individual’s knowledge, self-confidence, “Representativeness” of prospective period Investor portfolios reflect these beliefs Results not just for dummies, investors with small portfolios Implications/interpretations ER appears to covary negatively with usual conditioning vars Sharpe ratios are procyclical – HH investors do not appear to expect a premium in bad times, hold less equity Other types of investors need higher returns to “take up slack” 18 Conclusions (cont’d) Implications for equilibrium asset prices? Equity valuations lower during recession – and subsequent returns higher – because HH investors overly pessimistic (extrapolating too much) Individual investors presumably ‘expropriated’ by smart (institutional) investors But presumably rational investors do not entirely offset systematic irrational trading by HH investors • Limits to arbitrage • Active “smart” traders profit by “riding the bubble” – positive feedback trading • Observe countercyclical returns – Given these facts, what is simplest explanation? 19 The End 20