Does the Stock Market See a Zero or Small Positive Earnings Surprise as a Red Flag?

1. Does the Stock Market See a Zero or Small Positive Earnings Surprise as a Red Flag? Zhi-Xing Lin Michael Shih NUS Business School National University of Singapore

2. Earnings management is costly (reduces earnings in the future and heightens earnings expectations in the future) • Manipulation of analyst expectations is costly (lose credibility with analysts) • Thus, firms are likely to manage earnings and/or analyst expectations just enough to meet or narrowly beat analysts’ earnings forecasts.

3. Hypothesis 1: Investors and analysts see a zero or small positive earnings surprise as a red flag on the earnings announcement date in their attempt to identify manipulators of earnings and/or analyst earnings expectations.

4. there is an increasing tendency of firms to manage earnings and/or analyst expectations to avoid a small negative earnings surprise, as our test results show. • Academic research showed in late 1990s (Burghstahler and Dichev 1997; Degeorge et al. 1999) the frequency of firms meeting or narrowly beating earnings benchmarks is higher than expected. • Thus, investors and analysts should be more suspicious of firms that report a zero or small positive earnings surprise.

5. Hypothesis 2: Investors and analysts see a zero or small positive earnings surprise as more of a red flag on the earnings announcement date in late 1990s and early 2000s.

6. CAR = α + β(ES/P) + ּּּּּ + ε where CAR is cumulative abnormal returns attributed to the earnings announcement, ES is the earnings surprise, calculated as actual earnings per share minus the latest consensus analyst forecast, and P is the price per share. The coefficient β in the regression model is the earnings response coefficient (ERC).

7. Divide sample into earnings surprise class/range: • Class R-7: Earnings surprises less than -8¢ per share • Class R-6 comprises ES’s in the [-8¢,-6¢) range • Class R-5 comprises ES’s in the [-6¢, -4¢) range; • Class R-4 comprises ES’s in the [ -4¢,-3¢) range; • Class R-3 comprises ES’s in the [-3¢ ,-2¢) range; • Class R-2 comprises ES’s in the [-2¢ , -1¢) range; • Class R-1 comprises ES’s in the [-1¢,0); • Class R0 comprises ES’s in the [0, 1¢] range; • Class R1 comprises ES’s in the (1¢,2¢] range; • Class R2 comprises ES’s in the (2¢,3¢] range; • Class R3 comprises ES’s in the (3¢,4¢] range; • Class R4 comprises ES’s in the (4¢,6¢] range; • Class R5 comprises ES’s in the (6¢,8¢] range; nd • Class R6 comprises ES’s greater than 8¢.

8. CAR[-1,1]: Cumulative abnormal returns from 1 day before the earnings announcement date to 1 day after; • CAR[-20,-2]: cumulative abnormal returns from 20 days before the earnings announcement date to 2 days before; • ES/P: earnings surprise (ES) scaled by the share price twenty-one days before the earnings announcement date (P); • Dk: earnings surprise dummy, equal to one if the unscaled earnings surprise (ES) falls in Class Rk and zero otherwise.

9. FREV is the analyst forecast revision for next-quarter earnings, calculated as the first consensus analyst next-quarter earnings forecast after the current-quarter earnings announcement minus the latest consensus analyst next-quarter earnings forecast before the current-quarter earnings announcement. • The coefficient βk, k=-7…6, captures analysts’ response to earnings surprises in each range. We refer to this coefficient as the analyst earnings response coefficient (hereafter AERC).

10. Sample: Firm-quarters in 1992-2004 in the intersection of I/B/E/S, CRSP and Compustat. • Sample size: 118,136 firm-quarters for 7,394 firms.

11. ERCs for earnings surprises in various ranges (full sample 1992-2004)

12. Results for 1992-1997 and 1998-2004

14. AERCs for earnings surprises in various ranges (full sample 1992-2004)

15. Results for 1992-1997 and 1998-2004

17. Sensitivity analysis • Control for market to book, size, earnings persistence, earnings volatility, earnings growth, analyst following – the results remain the same • Use actual earnings as scaler for the earnings surprise – the results remain the same • Test the hypotheses using non-linear models – the results remain the same • Control for the sign of estimated discretionary accruals and the trajectory of analyst earnings forecasts (see next slide)

18. where: ES’ = ES/P, or earnings surprise scaled by share price; ES’(≥ 0) = ES’ if ES’ is non-negative (zero or positive), and zero otherwise; ES’(<0) = ES’ if ES’ is negative, and zero otherwise; PDA = positive discretionary accruals dummy, equal to 1 if estimated discretionary accruals are positive, and zero otherwise; TEF = trajectory of analyst earnings forecast dummy, equal to 1 the trajectory is downward before the earnings announcement, and zero otherwise.

22. Cumulative Abnormal Returns in the Event Window [+1, +60] after Earnings Announcement (1992-1997)

23. 0.04 ES<0¢ 0¢<=ES<=1¢ ES>1¢ 0.03 0.02 0.01 0 -0.01 0 10 20 30 40 50 60 Cumulative Abnormal Returns in the Event Window [+1, +60] after Earnings Announcement (1998-204)

24. QRET: size-adjusted abnormal returns after the earnings announcement; SURP: scaled decile rank of quarter t’s earnings surprise divided by share price. MNB: dummy variable, equal to 1 if unscaled earnings surprise of quarter t is in the [0,1¢] range, and zero otherwise; B/M: .book to market ratio at the end of fiscal quarter t-1; SIZE : log market value at the end of fiscal quarter t-1; BETA: market beta, FREV: analyst earnings forecast revisions for quarter t+1 after the earnings announcement for quarter t,