Stock Returns, Aggregate Earnings Surprises, & Behavioral Finance Kothari, Lewellen & Warner, JFE, 2006 FIN532 : Discussion Plan 1. Introduction 2. Sample Selection & Data Description 3. Part 1: Relation between Earnings and Returns Finding Robustness Check 4. Part 2: Relation among Earnings, Business Condition, and Discount Rates Finding Robustness Check 5. Conclusion 1
Importance of this paper Relates 3 areas of Earnings & Returns Research: 1. Empirical research on the stock market s reaction to earnings announcement 2. Behavioral Asset Pricing Literature 3. Research on the correlation among 1. Stock prices 2. Business conditions, and 3. Discount rates. 2
What questions are we pursuing? How does the market reacts to aggregate earnings surprises? Sign of Reaction? Magnitude of Reaction? Are price reactions at aggregate levels similar to that at firm level suggested by Behavioral Models. Do we see Post Earnings Announcement Drift in Aggregate Returns? Over Reaction or Under Reaction? If so, how much and how long? Can changes in Discount Rate explain Contemporaneous Return-Earnings Association? What is the marginal impact of an earnings surprise after controlling for discount rate? Post Earnings Announcement Drift exists at firm level Bernard & Thomas 90 & 89, Foster, Olsen & Shevlin 84, Watts 78 78, Ball & Brown 68 68. Firm s stock prices react immediately to Earnings Report, thereafter drift in the same direction for 3 quarters, and finally partially reverse in the 4 th quarter. Are Returns at firm level predictable after earnings announcement? Chan, Jegadeesh, and Lakonishok 96: Post Earnings Announcement Drift Price Momentum Do returns behave this way at Aggregate level? 3
Behavioral Finance Models Behavioral finance models suggest pricing anomaly exists at firm level: Prices react slowly to public news Do Behavioral models predict Post Earnings Announcement Drift exists? What about the reaction at the aggregate level? Behavioral Finance Models Behavioral finance models suggest pricing anomaly exists at firm level: Prices react slowly to public news. Bernard & Thomas 90: Naïve Investor Story Investors ignore autocorrelation pattern in stock return at firm level, surprised p by ypredictable ΔE. Earning Persistence, Post Earnings Announcement Drift 4
Behavioral Models 2: Barberis, Shleifer, and Vishny 98 Assume that earnings follow a random-walk, but Investors believe that earnings alternate between two regimes: 1. Mean reversion, 2. Trending Prices under-react in the short run (a single report), but over-react to a string of news. Daniel, Hirshleifer, and Subrahmanyam 98 Prices : over-react to private signals but under-react to public news. This model predicts short-run continuations after earnings announcements followed by Long Term Reversals at firm level. Are these Models consistent with Aggregate Price Behavior? Firm & Aggregate Price Behavior: Why could they differ? - 1 Earnings Predictability Transitory component in firm level earnings. Diversified away at the market level. If investors understand that aggregate earnings are a more reliable signal of value, they will under-react less to aggregate earnings news. Limits to arbitrage - Trading Costs, Sys. Risk, etc. ts to a b t age ad g Costs, Sys. s, etc. Difference in trading costs at aggregate and firm level. Difference in systematic risk in trading strategies based on aggregate earnings vs. firm-level earnings. 5
Firm & Aggregate Price Behavior: Why could they differ? - 2 Shocks to Discount Rates Efficient market: a change in discount rate corresponds to a change in expected returns. Discount Rates strongly correlated across stocks Driven by Business Conditions. Should have market-wide impact. Cash flows likely to have large idiosyncratic component. Vuolteenaho 02: accounts for the bulk of firm level returns. Campbel 91: accounts for less than half of aggregate returns. Should get diversified away at aggregate level. Corr(Earnings, Discount Rate)>0 => Decrease the contemporaneous return-earnings relation but enhance any lead-lag effects; In the absence of under-reaction, earnings would be positively related to future returns. Framework Unexpected Return Campbell (1991): Firm i s, Actual Return in period t can be decomposed into 3 components: R t = r t + η d,t η r,t r t = Expected Return = E t-1 (R t ) η d,t = Shock to Expected Dividend (Yield) / Cashflow η r,t = Shock to Expected Return / Capital Gains r P (P P -1 ) P -1 =R Negative Sign Unexpected Return = R t r t = η d,t η r,t = UR t Price impact of earnings determined by Cov(dE t, R t ) Cov (de t, r t + η d,t η r,t ) = Cov(dE t, η d,t ) Cov(dE t, η r,t ) Negative Price Impact of earnings if Cov(dE t, η r,t ) > Cov(dE t, η d,t ) > 0 Cov(dE t, r t ) = 0 if we have a good proxy for de t 6
Framework continued A negative reaction to aggregate earnings is consistent with a positive reaction to firm s earnings. UR i = (de i + de M ) dr M UR i = unexpected return, firm i de i = earning surprise, firm i specific, idiosyncratic de M = earnings surprise, systematic / market-wide de i + de M = Total Earning Surprise for firm i. dr M = impact of discount rate news, market-wide Discount rate goes up dr M > 0 Assumed to be entirely systematic Corr(dr M, de i ) = 0, Corr(dE M, de i ) = 0 UR M = de M dr M Framework continued Covariance of Firms Returns & Earnings Surprise Cov(UR i, de i + de M ) = Cov(UR i, de i ) + Cov(UR i, de M ) = Var(dE i ) + Cov(UR M, de M ) Since, 2Cov(dE i, de M ) = 0, Cov(dE i, UR M ) = 0. => Firm s Cov(UR i, de i ) can be positive, when dominated by idiosyncratic cash-flow shocks, even if Cov(UR M, de M ) is negative. 7
Framework continued Finance Theory suggests: Corr(dE t, η rt r,t )< 0 Movement in Equity Risk premium ~ counter-cyclical. Investors try to smooth consumption (Lucas 1978) or Aggregate risk aversion varies over business cycle (Campbell & Cochrane 99, Chan & Kogan 02) Lettau & Ludvigan 2004: r t Discount Rate, when investors are rational. Corr(dE t, η r,t ) > 0 Discount Rate Shock η r,t when E t. How? Higher earnings Higher Interest Rates Higher Risk Premium Higher Required Rate of Return = Higher Discount Rate Lower Price. Negative Price Impact of higher earnings. 8
Data used: Table 1,2, 3 Earnings Sample (Quarterly) All NYSE, Amex, and NASDAQ stocks On the Compustat quarterly file From 1970 2000 Excluding Firms with fiscal year not ending in March, June, September and December Top and Bottom 0.5% of firms remove outliers ranked by de/p each quarter Penny stocks CRSP Value Weighted Index Return for Aggregate R t Figure 1 Panel C: #of Firms in the Sample Average: 3,288 stocks per quarter Std Dev: 1,505 stocks per quarter Loss making Firms: <10% --> ~40% 9
10K, 10Q Filings: http://www.sec.gov/answers/form10k.htm Category of Filer Large Accelerated Filer ( Public Float of $700 Million or more ) Deadlines For Filing Periodic Reports at various times Form 10-K ANNUAL Since December 15, 2006: 60 days from fiscal years ending date. Since September 2002: 75 days Earlier: 90 days Accelerated Filer Since September 2002: 75 days ( Float: $75 to 700 Million ) Earlier: 90 days Form 10-Q QUARTERLY Since September 2002: 40 days Earlier: 45 days Non-accelerated Filer (Float: below $75 Million ) 90 days 45 days December 2005, Release No. 33-8644 (Revisions to Accelerated Filer Definition and Accelerated Deadlines for Filing Periodic Reports). Earnings Explained - 1 E = Quarterly Earnings. Net Income bf before Extra-ordinary items and Discontinued i operations. So Earnings exclude below the line items. de = E t E t-4 : Seasonally differenced quarterly earnings Earnings in the Current Quarter Less Four Quarter Prior de/s = de scaled by four quarter lagged S de/s t = (de/s) t = (E t E t-4 ) / S t-4 S = Earnings (E), Book equity (B), or Market Capitalization (P) E, B & P are lagged by four quarters => E t-4, B t-4 and P t-4 are used for de/s t 10
Earnings Explained - 2 de/b agg = Aggregate [ (Σ de i ) (ΣB i ) ] for all firms, i = 1, 2,, N de/p ew = Equally Weighted [Σ(dE/P) i N] for all firms Small firms have greater influence on de/p-ew. de/p vw = Value Weighted [Σ {(de/p) i x w i }] for all firms, i = 1, 2,, N where w i = P i (Σ P i ) Large firms have greater influence on de/p-vw. de/p vw de/p agg. correlation is 0.992 Figure 1 Panel A: Earnings Time Series - 1 E/P-agg E/P-ew Average 2.27 1.34 Std. Dev. 0.89 1.57 11
Figure 1 Panel B: Earnings Time Series - 1 E/B-agg E/B-ew Average 3.58 1.90 Std. Dev. 0.66 1.29 Figure 2 Panel A: Earnings Time Series - 2 Any quarter with Negative Earning? de/e-agg Average 7.84 Std. Dev. 17.77 12
Figure 2 Panel B: Earnings Time Series - 2 de/p-vw de/p-ew Average 0.10 0.25 Std. Dev. 0.34 0.54 Table 1: Summary Statistics: Quarterly Returns & Earnings: 1970-2000 13
Table 1: Summary Statistics: Quarterly Returns & Earnings: 1970-2000 Table 2: Correlation among several measures of aggregate earnings changes Earnings Measures used in the paper: de/b-agg, de/p-vw, and de/p-ew 14
What question are we pursuing? How does the market reacts to aggregate earnings surprises? Sign of Reaction? Magnitude of Reaction? Are price reactions at aggregate levels similar to that at firm level suggested by Behavioral Models. Do we see Post Earnings Announcement Drift in Aggregate Returns? Over Reaction or Under Reaction? If so, how long? 15
Tables # 3 and 4 Table 3: Auto Correlation of de/s AR(1) Model AR(5) Model Fama-MacBeth simple time- series regression 16
Table 4: Finding: Quarterly Return & Earning k=0, measurement quarter, k=1, announcement quarter Actual de/s Forecast error: AR(1) Forecast error: lagged de/s & Annual Return Tables # 5, 6, and 7 17
Sample for Robustness Check: Table 5,6,7 Annual Returns & Earnings: Table 5 Firms in Compustat t with December fiscal year end E, B, P, #of Shares Outstanding information present Size Portfolios: Table 7 Similar to sample for Earnings Sample (Quarterly) S&P 500: Table 6 Quarterly returns over different sample periods 1936 2000 = overall 1957 2000 = why is this included? 1970 2000 = sample period 1936 1969 = pre-sample period Table 5: Annual Returns & Earnings 18
Table 6: S&P 500: Quarterly Table 7: Size Portfolio 19
What s the Take Away! Aggregate g Earnings and Stock Returns are contemporaneously negatively related Corr (R t, de/s t ) < 0 Earning Surprises contain little information about future returns Corr (R t+k, de/s t ) ~ 0 No Post Earnings Announcement Drift! 20
What questions are we pursuing? Can changes in Discount Rate explain Contemporaneous Return-Earnings Association? What is the marginal impact of an earnings surprise after controlling for discount rate? Data used: Table 8, 9, 10 Discount Rate Proxies ΔTBILL = change in One Year T-Bill Yield ΔTERM = change in Term spread R b10 R b01 TERM is one of the Lead Indicators of Business Cycle ΔDEF = change in Default spread R AAA R BBB Measure of Economic Activity GDP = per capita growth in GDP CONS = per capita growth in Consumer Spending IPROD = growth in Industrial Production IIP is one of the Coincident Indicators of Business Cycle Consumer Sentiments ΔSENT = change in consumer sentiment from University of Michigan Survey Research Center 1979-2000 21
Tables # 8 and 9 Table 8: Earnings & Macro economy 22
Table 9: Quarterly Returns & Earnings, Controlling for discount rates Table # 10 23
Table 10: Annual Returns & Earnings, Controlling for discount rates Conclusion of this Paper Evidence that Returns and past earnings at aggregate level are not correlated. So, prices neither over-react nor under-react to earnings news => No Drift at the aggregate level. This is quite different from evidence at firm level as found in past research & behavioral models. Aggregate Earnings and discount rates are correlated. Provides new evidence that discount rate shocks explain a significant fraction off aggregate returns. 24
What is this discussion s value addition? Empirical illustrations: Fama MacBeth Regression, AR(1), AR(5), and Autocorrelations 2SLS fitted values and residuals in OLS regression, Detailed Robustness Checks Result at Aggregate Level versus Firm Level: Counter Intuitive Corr(dE t, R t ) and Corr(dE t, ke t ) 3 Behavioral Models inconsistent prediction. Finance & Accounting Keywords / Concepts: Seasonally differenced Earnings Contemporaneous Returns & Earnings Aggregate Earning Surprises Discount Rates, Expected Returns & Business Cycle Behavioral Models, Cognitive and Psychological Biases Price Momentum, Post Earnings Announcement Drift Shock to Dividend, Shock to Expected Returns C est tout? Thank You for your participation 25