Fear and Greed: a Returns-Based Trading Strategy Around Earnings Announcements

Fear and Greed: a Returns-Based Trading Strategy around Earnings Announcements

Ivo Ph. Jansen Andrei L. Nikiforov

Associate Professor of Accounting Assistant Professor of Finance Rutgers University – Camden Rutgers University – Camden School of Business School of Business 227 Penn Street 227 Penn Street Camden, NJ 08102 Camden, NJ 08102 (856) 225-6696 (856) 225-6594 [email protected] [email protected]

Journal of Portfolio Management

Summer, 2016

Abstract: This study documents that earnings announcements serve as a reality check on short- term, fear and greed driven price development: stocks with extreme abnormal returns in the week before an earnings announcement experience strong price reversal around the announcement. A trading strategy that exploits this reversal is profitable in 40 of the last 42 years and earns abnormal returns in excess of 1.3% over a two day-window.

______We thank an anonymous referee, Brandon Cline, Lee Sanning, and Uzi Yaari for helpful comments and suggestions.

“Be fearful when others are greedy and greedy when others are fearful” — Warren Buffett

Fear and greed are deeply ingrained in life, and are fundamental attributes to the survival

of man. Without the right dose of fear, we would expose ourselves to unreasonable threats, and

without the right dose of greed, we would forego opportunities to secure the resources that we

need to live. However, too much of either fear or greed is often harmful. An individual overcome

by fear, for example, will not be able to pursue opportunities that will help him secure the

resources he needs; and an individual overcome by greed will not recognize and avoid the threats

to his existence. In other words, for optimum survival, fear and greed need to be “balanced.”

In financial markets, where prices are set by the interactions of thousands of individuals,

“greed” for profits and “fear” of losses drive investors to make value assessments as carefully as

possible, thus contributing to market efficiency. But, in this setting also, excessive or unbalanced

fear or greed can be harmful. Warren Buffett’s quote above succinctly speaks to the dual nature

of fear and greed in investing. In a market with too many greedy or fearful investors, aggressive

buying or selling may cause an overreaction, at which point it is profitable to take a contrarian

position. Many investors understand this, and the history of markets has countless examples of

asset prices becoming irrationally high or low because of greed and fear. The challenge in trying

to profit from this mispricing, of course, lies in timing. That is, it is unclear when the market will

“come to its senses” and prices will revert to levels justified by fundamentals. It is difficult,

therefore, to profitably implement Warren Buffett’s advice, especially in the short-term.1

1 Buffett’s advice is clearly intended as an investment strategy for the long-term; not the short-term. However, the reality is that most investors face powerful short-term performance pressures. For example, Stan Druckenmiller, the lead money manager at George Soros’ Quantum Fund, closed all of his long positions in dot-com stocks in February of 2000 based on the (correct) belief that dot-com prices reflected a bubble. Weeks later, however, after prices had continued their run-up and his performance was significantly lagging that of his colleagues, Druckenmiller reclaimed those long positions. A few weeks later, the bubble burst.

In this study, we develop a trading strategy around earnings announcements that seeks to

profit from predictable reversals of fear and greed driven price development in individual stocks.

We argue that earnings announcements are logical events around which to center such a trading

strategy, because they convey fundamental information about asset prices and thus have the

potential to “break” irrational price development. Moreover, because of heightened information

asymmetry in the period just before an earnings announcement, price development is probably

particularly susceptible to excessive fear or greed. That is, if uninformed investors observe sharp

price changes just before an earnings announcement, they may attribute these to the informed

trading of insiders; start to excessively trade in the same direction themselves; and thus cause an

overreaction. We therefore predict—in the spirit of Warren Buffett’s advice—that stocks that

experience sharp price changes just before an earnings announcement will experience price

reversal at the time of the announcement itself. We test this prediction with a trading strategy

that on the earnings announcement date takes (1) a long position in stocks that experienced

extreme negative abnormal returns in the week prior, and (2) a short position in stocks that

experienced extreme positive abnormal returns in the week prior.

We find that, over the two day window of the earnings announcement date and the day

following, both positions are highly profitable. On average, the long position earns abnormal returns of 1.49%, and the short position earns 1.20%. We furthermore show that these return reversals are about 60% larger than around non-earnings announcement dates, and thus are significantly more pronounced than short-term return reversals documented in the prior literature

(e.g., Lehmann [1990] and Figelman [2007]). We also show that our strategy (1) is profitable in

40 of the 42 years in our sample; (2) is similarly profitable in “bear” and “bull” markets; (3) and is significantly profitable for both large firms and high volume stocks. Since the year 2000—

using a conservative transactions costs estimate of 70 basis points for a round trip trade—we find

that our strategy generates abnormal returns of 0.76% after transaction costs, or 95% on an

annualized basis. We conclude, therefore, that prices are subject to sentiment-driven price

development in the period of elevated information asymmetry just before earnings

announcements, and that the announcements themselves serve as a reality check on that price development.

THEORY AND BACKGROUND

The history of financial markets has countless examples of asset prices rising or falling rapidly, and then suddenly reversing. This has been true for individual securities, certain industries, and entire asset classes. Unbalanced fear or greed in financial markets most

commonly manifests itself as price momentum; which according to Fama [1998] is one of the

two most robust and persistent anomalies posing a challenge to the efficient market paradigm.2

While price momentum is usually argued to have its source in underreaction (Jegadeesh and

Titman [1993]), it often ultimately produces an overreaction (Lehmann [1990], Hong and Stein

[1999], Hirschey [2003], and Figelman [2007]). This could be an overreaction to underlying

fundamentals—for example, Zarowin [1989] documents that firms with a string of good (bad)

earnings news tend to become overpriced (underpriced); or it could be an overreaction to

price/investor behavior—for example, the indiscriminate selling of stocks in the wake of the

financial crisis in 2008. Also, the price momentum can occur over a period of months (e.g.,

Jegadeesh and Titman [1993]), or within a single day (e.g. Fabozzi, Ma, Chittenden, and Pace

[1995], Schulmeister [2009]). The defining feature of unbalanced fear and greed driven price

2 The other one is post-earnings announcement drift: the tendency of stock prices to drift upward (downward) after surprisingly good (bad) earnings news.

momentum, however, is the reversal that occurs afterwards. This price reversal thus offers an

investment opportunity for contrarian investors, as captured in Warren Buffett’s advice to “be

fearful when others are greedy and greedy when others are fearful.” However, timing a price reversal is very difficult, and potentially very costly, even when an investor is correct in his or her assessment of mispricing. Theoretical papers indeed demonstrate that, even when rational investors are aware of mispricing in the market and take positions to exploit it, the mispricing can persist and may, in fact, get worse (De Long, Shleifer, Summers, and Waldmann [1990], and

Shleifer and Vishny [1997]).3

In this study, we argue that earnings announcements are natural candidates for the

implementation of a contrarian trading strategy that seeks to exploit the reversal of fear/greed

driven price momentum. First, the increase in information asymmetry preceding the earnings

announcement makes it more likely that excessive sentiment affects pricing during that time.

Second, the earnings announcement, by conveying fundamental information, serves as a reality

check on that sentiment, making it more likely that price reversal will occur (if there is indeed

mispricing).4 Third, the earnings announcement date is known ex ante, which makes it possible

to anticipate the price reversal and thus implement a trading strategy. These last two arguments are obvious. We elaborate on the first argument below.

The anticipation of earnings announcements makes investors aware of the possibility of being taken advantage of by informed traders. Market makers, for example, protect themselves by increasing bid-ask spreads and reducing depths in the period before the earnings

3 An instructive example is the case of Long-Term Capital Management’s bet on the convergence of Royal Dutch Petroleum and Shell Transport. These two companies co-own Royal Dutch/Shell and receive their income from the same sources. There is no theoretical justification, in other words, for their prices to diverge. LTCM took a position to exploit this mispricing when the divergence was 8%, but was later forced to liquidate the position when divergence had increased to 22% (Lamont and Thaler 2003). 4 Many studies in accounting research indeed document that, at the time of an earnings announcement, stock prices move in the same direction as the earnings surprise (e.g., Wilson [1987], Collins and Kothari [1989]).

announcement (Lee, Mucklow and Ready [1993]). Because there is usually very little public

information about a stock in the week before an earnings announcement, large price movements

are more likely attributable to the trading activities of investors with private information.

Individual investors—who tend to watch their portfolios very closely and “over” trade in them

(Barber and Odean [2001], [2002a], and [2002b])—are probably particularly sensitive to the increase in information asymmetry in the pre-announcement period. We propose that they, in particular, may respond to large price movements at this time by trading in the same direction, thus contributing to price momentum. Specifically, we argue that when individual investors observe a large price increase, they may infer that insiders know that the upcoming earnings announcement will reveal unexpectedly good news. Greedy to profit from this inference, the individual investor will “jump on board” and push the price up even further. Similarly, when individual investors observe a large price decrease, they may infer that insiders know that the upcoming announcement will reveal unexpectedly bad news. Fearful to suffer the corresponding losses, the individual investor will “jump ship” and push the price down even further. We argue that the resulting sentiment-driven price momentum will result in an overreaction; thus setting the stage for our contrarian trading strategy.

TRADING STRATEGY AND METHODOLOGY

We implement our strategy as follows. We first select stocks with extreme abnormal

returns in the week before the earnings announcement (from day -5 to -1, relative to the announcement date reported on Compustat). We use three different benchmarks for extreme abnormal returns during this window: 5%, 10% and 15%. Note that we do not argue that all stocks in our trading strategy have experienced sentiment driven price momentum in the pre- announcement window. We simply rely on the argument that, among the stocks with large price

increases, there are some that experienced greed driven price momentum such that, on average,

stocks with large price increases are overpriced; and vice versa. Then, on the day of the earnings announcement, we take a long position in stocks that experienced extreme negative abnormal returns, and a short position in stocks that experienced extreme positive abnormal returns. We continue to hold both positions until the next day, to capture the abnormal returns for those stocks that announce their earnings after trading hours.

We estimate buy-and-hold abnormal returns (BHARs) using, as our benchmark, the contemporaneous return from the portfolio of firms in the same size-decile. We use size-adjusted abnormal returns because firm size is often found to be a significant determinant of the profitability of trading strategies and the magnitude of anomalies. In sensitivity analyses, we also examined abnormal returns estimated relative to the market model, and to the three-factor Fama-

French [1993] model. The results from those analyses are qualitatively similar to those reported in this study.

DATA AND RESULTS

We obtain quarterly earnings announcement dates and earnings information from

Compustat, starting in the second half of 1971 and ending in 2012. We obtain volume, price and firm size information from CRSP. We focus only on common shares and eliminate all observations with missing data, as well as those securities that are traded on exchanges other than the NYSE, AMEX or NASDAQ. Because of restrictions that brokers impose on trading low-priced stocks (e.g., inability to sell them short or buy on margin) we eliminate all stocks with prices less than $2 per share. Our total sample consists of 643,669 observations.

Trading Strategy Abnormal Returns

Exhibit 1 reports the abnormal returns from our trading strategy. The results are reported

in three panels, corresponding to the three abnormal returns benchmarks we use during our

screen window: 5%, 10% and 15%. The results in all three panels confirm the economic and

statistical significance (p-value<0.0001 in all cases) of the profitability of our trading strategy.

For brevity, we focus our discussion on the results in panel B, where we screen for firms with

abnormal returns in excess of ±10% in the week before the earnings announcement.

There are 25,226 observations (about 3.9% of the total) that experience abnormal returns

below negative 10%, and 37,568 observations (about 5.8% of the total) that experience abnormal

returns above positive 10%. During the (-5,-1) screen window, these stocks have average BHAR

of –15.01% and 18.08%, respectively. Consistent with our argument that this reflects, on

average, at least some overreaction, the stocks indeed experience very significant return reversals

In the subsequent (0,1) holding window. And so the long position in our trading strategy generates a profit of 1.49%, while the short position generates a profit of 1.20%. A weighted average between the two positions equals 1.32% which, on an annualized basis, translates into abnormal returns in excess of 160% before transaction costs.5

Panels A and C further show that the profitability of our trading strategy is increasing in

the magnitude of the abnormal returns benchmark during the screen window. The abnormal

returns for the long positions increase from 0.80% in Panel A, to 1.49% in Panel B, reaching

2.26% in Panel C. The abnormal returns for the short positions similarly increase across the three

panels from -0.68%, to -1.20% and, finally, to -1.74%.

5 We annualize abnormal returns by multiplying the two-day holding window return by 126, because a typical year has 126 such windows (i.e., 252 trading days).

Trading Strategy Abnormal Returns over Time

To investigate whether our trading strategy profits are driven by just a few stellar years,

we calculate the average holding window abnormal returns for each year in our sample for the

stocks that experienced at least ±10% abnormal returns during the screen window (cf., Panel B in

Exhibit 1). The results are reported in Exhibit 2. They show that the long and short positions are

profitable in 40 and 38 years, respectively, of the 42 years in our sample period. Moreover, the

weighted average of the two legs of the strategy is profitable in 40 of the 42 years.

While the preceding results suggest that the trading strategy is profitable in both bull and bear markets, we explicitly investigate its profitability as a function of market sentiment in panel

A of Exhibit 3.6 This investigation is relevant not only because many anomalies show different

returns depending on market sentiment, but our study is a specific investigation of sentiment-

driven price development. We find that the long position generates slightly higher profits in bull

markets (1.63% versus 1.12%), but that the short position generates slightly higher profits in bear

markets (1.63% versus 1.08%). A weighted average abnormal return between the long and short

positions (not reported) shows that the trading strategy’s profitability is actually a little bit higher

in bear than in bull markets (1.40% versus 1.29%). This finding compares very favorably with

the profitability of most well-known trading strategies, which earn positive returns in bull

markets, but which tend to earn negative returns in bear markets.

Trading Strategy Abnormal Returns, Firm Size, and Liquidity

To address the concern that our strategy’s returns might be driven by small and/or illiquid

firms, we investigate the profitability of our trading strategy after partitioning the sample into

6 We define bull and bear markets consistent with J.P. Morgan Asset Management. https://www.jpmorgan.com/cm/BlobServer/Guide_to_the_Markets_4Q_2011.pdf

size terciles and dollar-volume terciles. We do so, once again, after limiting our investigation to those stocks that experienced at least ±10% abnormal returns during the screen window.

Panel B of Exhibit 3 presents the results of our trading strategy for size terciles based on

NYSE breakpoints. Small firms indeed generate the highest holding period returns, but the returns for medium and large firms are also statistically and economically significant. For example, the long position for large firms generates a profit of 1.40%, while the short position earns a profit of -0.48% over two days. Thus, Panel B shows that the trading strategy profits are not simply driven by small firms.

Panel C of Exhibit 3 presents the results of our trading strategy across dollar-volume terciles. To create these terciles, we first divided our original sample into three equal sized dollar-volume portfolios based on total dollar volume during the holding window, after which we identified stocks which experienced at least ±10% abnormal returns during the screen window.

Not surprisingly, the least liquid stocks experience the greatest return reversals during the holding window; but, once again, even for the most liquid stocks, our trading strategy is highly profitable, generating a 1.31% profit in the long, and a 0.4% profit in the short position. Thus, liquidity concerns do not prevent a profitable implementation of the trading strategy.

Trading Strategy Profitability and Transaction Costs

The consistent profitability of our trading strategy over the last four decades clearly suggests that fear and greed affects pricing in the week preceding earnings announcements. To assess implementability, however, we need to consider transaction costs.7 We do so by focusing on the strategy’s profitability since 2000. Estimates of transaction costs vary across sources. For

7 Recall that we eliminate all stocks with prices less than $2 from our analyses. Transaction costs (as well as returns) are often particularly high for small priced stocks. Those stocks, however, are not included in this study.

example, Stoll [2006] estimates transaction costs in 2001 at 28 basis points for a roundtrip trade.

Elkins McSherry, a global expert in trade cost analysis, estimates the average costs in 2007 at 66

basis points and in 2009-2011 at 53 basis points (Pollin and Heintz [2011]). Finally, ITG Global

Cost Review estimates transaction costs to range from 43 to 55 basis points between 2009 and

2013. Compared to these numbers, we use a relatively conservative estimate of transaction costs

of 70 basis points for a roundtrip trade to assess the implementability of our trading strategy.

From 2000 through 2012, the net profit of our strategy—i.e., after subtracting transaction costs of 0.7%—is 0.49% for the long position and 0.96% for the short position. These abnormal returns are statistically significant at the 0.0001 level. Moreover, the annualized abnormal returns would range from 60% for the long to more than 120% for the short position. Based on a weighted average between the long and short positions, the strategy would (1) generate a net profit of 0.76% from 2000 through 2012, for an annualized return of 95%, and (2) be profitable on a net basis in every single year.

Additional Robustness Tests

We discuss the results of several additional robustness tests in this section. They are not

tabulated here, but are available online. First, we investigate whether our results are different

from the generic, weekly return reversal patterns documented in Lehman (1990). To this end, we

estimate return reversals from implementing our trading strategy around five, equally spaced,

non-earnings announcement dates in the same fiscal quarter: 2, 4, 6, 8 and 10 weeks before the

actual earnings announcement. Our results show that the return reversal around the actual

earnings announcement date is about 60% higher than around non-earnings announcement dates.

These differences are highly statistically, and economically significant.

Second, we document the robustness of our trading strategy returns to alternative screen

and holding windows. We find that the trading strategy becomes slightly more profitable when

extending the holding window up to day +5. However, we also find that further extending the

holding window to day +10, decreases profitability. When we extend the screen window to start

on day -7, -10 or -20, instead of on day -5, we find that the trading strategy becomes less

profitable the further we extend the screen window. Finally, we shift both the screen and holding

windows, by “moving” day -1 from the former to the latter. We do so to allow for the possibility

that inside information leakage on the day immediately before the earnings announcement

already causes return reversal at that time. The results indicate that this is the case. Using a

(-5,-2) screen, and a (-1,1) holding window, significantly increases the strategy’s profitability:

from 1.49% to 2.36% for the long, and from 1.20% to 1.59% for the short position.8 Moreover, based on these windows, both legs of the trading strategy are profitable in each of the last 42 years. This is a remarkable finding that first of all speaks to the statistical significance of the trading strategy’s profitability, but especially speaks to its economic significance. To our knowledge, there is no anomaly or other trading strategy—including post-earnings announcement drift—that has been consistently profitable for such a long time.

CONCLUSION

This study documents that a contrarian trading strategy around earnings announcements, in stocks that experienced extreme abnormal returns in the week preceding the announcements, is highly profitable. We attribute the significant price reversal to a sentiment-driven overreaction in stock prices in the pre-announcement period. Specifically, we argue that due to heightened information asymmetry just before earnings are announced, individual investors in particular

8 A further shift to include day -2 in the holding window as well, leads to a significant decline in the strategy’s profitability.

trade in the direction of observed price changes, contributing to price-momentum that ultimately

results in overreaction. Then, when the earnings are announced and information asymmetry is

reduced, the overreaction leads to price reversal. We find that our trading strategy generates

abnormal returns of 1.3% over a two-day window, which is more than 160% on an annualized

basis. After transaction costs, and since the year 2000, the annualized abnormal returns equal

95%. We also document that our trading strategy is similarly profitable in bull and bear markets,

and is robust to firm size and liquidity.

REFERENCES Barber, Brad, and Terrance Odean. “Trading is Hazardous to Your Wealth: The Common Stock Investment Performance of Individual Investors.” Journal of Finance, vol. 55, no. 2, (2002a), pp. 773-806. ______. “Boys will be Boys: Gender, Overconfidence, and Common Stock Investment.” Quarterly Journal of Economics, vol. 116, No. 1 (2001), pp. 261-292. ______. “Online Investors: Do the Slow Die First?” Review of Financial Studies, vol. 15, no. 2 (2002b), pp. 455-487. Collins, Daniel W., and S. P. Kothari. “An Analysis of Intertemporal and Cross-sectional Determinants of Earnings Response Coefficients.” Journal of Accounting and Economics, vol. 11, no. 2-3 (1989), pp. 143-81.

De Long, J. Bradford, Andrei Shleifer, Lawrence Summers, and Robert Waldmann. “Noise Trader Risk in Financial Markets.” Journal of Political Economy, vol. 98 (1990), pp. 703– 738. Fabozzi, Frank J., Ma, Christopher K., Chittenden, William T., and Daniel Price. “Predicting intraday price reversals.” The Journal of Portfolio Management, vol. 21, no 2 (1995), pp. 42- 53. Fama, Eugene F. “Market Efficiency, Long-Term Returns, and Behavioral Finance.” Journal of Financial Economics, vol. 49 (1998), pp. 283–306 Fama, Eugene F., and Kenneth R. French. “Common Risk Factors in the Returns on Bonds and Stocks.” Journal of Financial Economics, vol. 33 (1993), pp. 3–53. Figelman, Ilya. “Stock return momentum and reversal.” The Journal of Portfolio Management, vol. 34, no.1 (2007), pp. 51-67.

Hirschey, Mark. “Extreme return reversal in the stock market.” The Journal of Portfolio Management, vol. 29, no. 3 (2003), pp. 78-90. Hong, Harrison, and Jeremy Stein.“A Unified Theory of Underreaction, Momentum Trading, and Overreaction in Asset Markets.” Journal of Finance, vol. 54 (1999), pp. 2143–2184. Investment Technology Group. 2013. “Global Cost Review Q1/2013.” ITG Peer Analysis (http://www.itg.com/marketing/ITG_GlobalCostReview_Q12013_20130725.pdf) Jegadeesh, Narasimham, and Sheridan Titman. “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency.” Journal of Finance, vol. 48 (1993), pp. 65–91. Jones, Charles M. “A Century of Stock Market Liquidity and Trading Costs.” Unpublished Manuscript (2002) (Available at SSRN: http://ssrn.com/abstract=313681) Lamont, Owen, and Richard Thaler. “Can the Market Add and Subtract? Mispricing in Tech Stock Carve-Outs.” Journal of Political Economy, vol. 111 (2003), pp. 227–268. Lee, Charles, Belinda Mucklow, and Mark Ready. “Spreads, Depths, and the Impact of Earnings Information: an Intraday Analysis.” Review of Financial Studies, vol. 6 (1993), pp. 345–374. Lehmann, Bruce. "Fads, martingales, and market efficiency." The Quarterly Journal of Economics, vol. 105, no. 1 (1990), pp. 1-28. Pollin, R., and J. Heintz. “Transaction Costs, Trading Elasticities and the Revenue Potential of Financial Transaction Taxes for the United States.” PERI Research Brief (2011) (http://www.peri.umass.edu/fileadmin/pdf/research_brief/PERI_FTT_Research_Brief.pdf) Schulmeister, Stephan. “Profitability of Technical Stock Trading: Has It Moved From Daily to Intraday Data?” Review of Financial Economics, vol. 18, no. 4 (2009), pp. 190-201. Shleifer, Andrei and R.W. Vishny. “The Limits of Arbitrage.” Journal of Finance, vol. 52 (1997), pp. 35–55. Stoll, Hans R. “Electronic Trading in Stock Markets.” Journal of Economic Perspectives, vol. 20, no. 1 (2006), pp. 153-174. Wilson, G. Peter. “The Relative Information Content of the Accrual and Funds Component of Earnings after Controlling for Earnings.” The Accounting Review, vol. 62 (1987), pp. 93-322. Zarowin, Paul.. “Does the Stock Market Overreact to Corporate Earnings Information?” Journal of Finance, vol. 44 (1989), pp. 1385–1399. Zhou, Haigang and John Qi Zhu. “Jump on the Post-Earnings Announcement Drift.” Financial Analysts Journal, vol. 68, no. 3 (2012), pp. 63-80.

EXHIBIT 1. Trading Strategy Abnormal Returns. This exhibit reports buy-and-hold, size-adjusted abnormal returns. The screen window runs from day -5 through day -1, relative to the earnings announcement. The holding window runs from day 0 through day +1. Stocks with prices less than $2. All reported returns in this exhibit are significant at the 0.0001 level.

Panel A: Screen Window Abnormal Returns <-5% and >5%

Position N Screen Window Holding Window Long 85,538 -9.33% 0.80% Short 95,473 11.38% -0.68%

Panel B: Screen Window Abnormal Returns <-10% and >10%

Position N Screen Window Holding Window Long 25,226 -15.01% 1.49% Short 37,568 18.08% -1.20%

Panel C: Screen Window Abnormal Returns <-15% and >15%

Position N Screen Window Holding Window Long 8,690 -20.75% 2.26%% Short 17,674 24.81% -1.74%

EXHIBIT 2. Trading Strategy Abnormal Returns over Time This exhibit reports buy-and-hold, size-adjusted abnormal returns by year. The stocks in this figure experienced screen window abnormal returns in excess of 10% positive/negative (cf., Panel B in Exhibit 1). Stocks with prices less than $2 are deleted.

4.0%

2.0% Short 0.0% Long

-2.0%

-4.0% 1971 1976 1981 1986 1991 1996 2001 2006 2011

EXHIBIT 3. Robustness of the Trading Strategy Abnormal Returns This exhibit reports buy-and-hold, size-adjusted abnormal returns over a (0,+1) holding window relative to the earnings announcement. The stocks in this exhibit experienced screen window abnormal returns in excess of 10% positive/negative (cf., Panel B in Exhibit 1). In Panel A, we assess the robustness of the trading strategy abnormal returns to periods of bull versus bear markets. We define bear markets consistent with J.P. Morgan Asset Management. They identify the following bear markets during our sample period: January 5, 1973 – October 3, 1974; November 28, 1980 – August 12, 1982; August 25, 1987 – December 4, 1987; March 24, 2000 – October 9, 2002; and October 9, 2007 – March 9, 2009. All other dates are bull markets. In Panel B, we assess the robustness of the trading strategy abnormal returns to firm size. Stocks are assigned, by year, to three size portfolios, where the portfolio cutoffs correspond to the 33rd percentile and 67th percentile of NYSE market-caps. Finally, in Panel C, we assess the robustness of the trading strategy abnormal returns to liquidity, as measured by $-trading volume. Stocks are assigned, by year, to three equal-sized $-Volume portfolios, based on total dollar volume during the holding window. Stocks with prices less than $2 are deleted. All reported returns in this exhibit are significant at the 0.0001 level, except for the -0.48% in Panel B, which is significant at the 0.003 level.

Panel A: Robustness of Trading Strategy Abnormal Returns to Bull and Bear Markets

Bull Market Bear Market Position N Abnormal Returns N Abnormal Returns Long 18,281 1.63% 6,945 1.12% Short 29,227 -1.08% 8,341 -1.63%

Panel B: Robustness of Trading Strategy Abnormal Returns to Firm Size

Small Medium Large Position N Abnormal Returns N Abnormal Returns N Abnormal Returns Long 18,450 1.66% 4,942 0.87% 1,834 1.40% Short 26,558 -1.29% 7,772 -1.18% 3,238 -0.48%

Panel C: Robustness of Trading Strategy Abnormal Returns to Liquidity

Low $-Volume Medium $-Volume High $-Volume Position N Abnormal Returns N Abnormal Returns N Abnormal Returns Long 7,986 1.90% 9,232 1.29% 7,774 1.31% Short 9,164 -2.22% 14,442 -1.35% 13,593 -0.40%