The Information Content of Fiscal-Year-End Earnings

The Information Content of Fiscal-Year-End Earnings Linda H. Chen, George J. Jiang, and Kevin X. Zhu January, 2018 Linda Chen is from the Department of Accounting, College of Business and Economics, University of Idaho, Moscow, Idaho 83844-3161; email lindachen@uidaho.edu; tel. (208) 885-7153. George Jiang is the Gary P. Brinson Chair of Investment Management in the Department of Finance and Management Science, Carson College of Business, Washington State University, Pullman, WA 99164; email george.jiang@wsu.edu; tel. (509) 335-4474; fax: (509) 335-3857.. Kevin X. Zhu is from the International Business School Suzhou, Xi'an Jiaotong-Liverpool University, Suzhou, China; e-mail Kevin.Zhu@xjtlu.edu.cn. We wish to thank seminar participants at Hong Kong Polytechnic University for helpful comments and suggestions. This research was supported by Hong Kong GRF B-Q50J. The usual disclaimer applies.

The Information Content of Fiscal-Year-End Earnings Abstract We compare the information content of fiscal-year-end (FYE) quarter earnings and non- FYE quarter earnings to shed light on the effect of earnings characteristics on market response to earnings announcements. We also use this setting to examine the effect of regulations related to accounting practice or financial information disclosure on market response to earnings announcements. Our results show that there is significantly less information content in FYE quarter earnings than non-fye quarter earnings. Relative to non-fye quarter earnings, market response to FYE quarter earnings is significantly weaker not only during the announcement window, but also during both the pre- and post-announcement windows. We show that the results are not driven by stale analyst forecasts and are robust to controlling for various earnings characteristics, such as earnings persistence, earnings quality, and reporting lag, as well as firm characteristics, such as size, BM ratio, etc. We show evidence that extraordinary items contribute partly but not fully to the lower information content in FYE quarter earnings. However, FASB Statement No. 145, which significantly shortens the list of extraordinary items, fails to improve the information content of FYE quarter earnings. Nevertheless, we show evidence that Regulation Fair-Disclosure, which restricts corporate disclosure to selected market participants, effectively reduces the pre-earnings announcement drift and the difference in pre-announcement market responses to FYE quarter earnings versus non-fye quarter earnings. Keywords: Fiscal-year-end and non-fiscal-year-end earnings announcements; Market reactions; Earnings response coefficient; Regulation Fair Disclosure; FASB Statement No. 145 1

I. Introduction Market reaction to earnings announcements, measured by earnings response coefficients (ERCs), has received great attention in the literature as it sheds light on the information content of earnings and the efficiency of the financial market. Much of the literature focuses on the cross-sectional variation of ERCs among different firms, namely, whether the estimates of earnings response coefficients are correlated with certain properties of the firms earnings across firms. For instance, Kormendi and Lipe (1987) and Easton and Zmijewski (1989) show that the more persistent the time-series of a firm s earnings, the larger the price change or the earnings response coefficient is. Collins and Kothari (1989) predict a positive marginal effect of a firm s growth opportunities on the earnings response coefficient. In this paper, we examine variations in market responses to fiscal-year-end (FYE) quarter earnings versus non-fye quarter earnings. Complementary to existing studies that have mostly focused on the determinants of ERCs across firms, we focus on market responses to different earnings announcements by the same firm. The contrast between FYE quarter earnings and non- FYE quarter earnings presents a unique setting to examine the effect of earnings characteristics rather than firm characteristics on earnings response coefficients. In addition, we also use this setting to examine the effect of regulations related to accounting practice or financial information disclosure on market responses to earnings announcements. The literature presents mixed arguments about whether FYE earnings is more or less informative than non-fye earnings. One view is that information in FYE earnings is less uncertain and is more creditable to investors. This is because management has less discretion in annual statements and FYE quarterly earnings (Mendenhall and Nichols, 1988). This is because the information in FYE earnings is subject to audit before 10-Ks are filed. In addition, FYE 2

results are also likely subject to more scrutiny by market participants and regulators. Cornell and Landsman (1989) suggest that the fourth quarter announcement appears to provide more information to analysts and investors than interim announcements. When announcing FYE earnings, management not only have to explain the performance over the past fiscal year but very often have to provide guidance about the firm s business prospects. On the other hand, there is also reason to believe that information in FYE quarterly earnings can be noisy since the reporting of FYE earnings is subject to more managerial discretion. Palepu (1988) points out that FYE quarter earnings are noisier than the first three quarters' earnings. One likely reason is that nonrecurring or extraordinary expenses/revenues are more frequently recognized in the FYE quarter than in the other three quarters (Palepu, 1988). In addition, there is evidence that management seems to have more incentives to manage FYE earnings via accruals or discretional accruals. The main data used in our empirical analysis includes the IBES database for earnings announcement dates and analyst forecasts, the CRSP database for stock returns, Compustat for other firm characteristics, and the Thomson-Reuters 13F database for institutional holdings. The stock sample includes common stocks traded on the NYSE, AMEX, or NASDAQ in the CRSP stock database. The sample period is from January 1984 to December 2015. Our main results show that there is a significantly weaker market reaction to FYE earnings surprises than to non-fye earnings surprises. This is inconsistent with the view that FYE earnings are subject to less information uncertainty and are more credible to investors. For example, the return spread between the top and bottom deciles of earnings surprises during the two-day earnings announcement window is 4.656% for FYE quarter earnings announcements, whereas the corresponding number is 6.007% for non-fye quarter earnings announcements. The 3

difference is highly significant with a t-statistics of 6.43. The results also show that market responses to earnings surprises are also significantly weaker for FYE announcements than for non-fye earnings surprises during both the pre- and post-announcement windows. The return spreads between the top and bottom deciles of earnings surprises during the five-day preearnings announcement window and two-week post-earnings announcement window are, respectively, 1.522% and 0.939% for FYE quarter earnings announcements, whereas these corresponding numbers are 2.412% and 1.877% for non-fye quarter earnings announcements. Both differences are highly significant. This is evidence that the weaker market reaction to FYE quarter earnings announcements can not be attributed to stronger pre-announcement market reactions or delayed market reactions, but less information content. Moreover, our results show that the weaker market reaction is more pronounced for negative earnings surprises, i.e., earnings response coefficients are particularly lower when firms announce negative earnings news in FYE quarters. We perform a number of robustness checks for our main findings. First, one concern is that the weaker market reaction may be due to the fact that there are more stale analyst forecasts for FYE quarter. If analysts do not update earnings forecasts timely, earnings surprises can be less informative. We exclude stale analyst forecasts in our analysis and show that the results are consistent. Second, we perform multivariate tests and show that the results are robust to controlling for various firm characteristics, such as size and BM ratio, earnings characteristics, such as earnings quality and reporting lag, as well as past stock returns. In particular, our results show that FYE quarter earnings are less persistent than non-fye quarter earnings. Nevertheless, market reaction to FYE earnings remains significantly weaker even after controlling for earnings persistence. 4

As pointed out in Palepu (1988), nonrecurring or extraordinary expenses/revenues are more frequently recognized in the FYE quarter than in the other three quarters. These extraordinary items may be reasons that FYE earnings are less persistent and less informative to investors. As such, immediate market reactions to FYE earnings are weaker. Our results show that indeed a higher percentage of FYE income statements contain extraordinary items (PEOI) than non-fye income statements. As expected, among stocks in different FE deciles, stocks in the bottom two FE deciles (with negative earnings surprises) have a significantly higher percentage with extraordinary items in FYE income statements. We further examine whether extraordinary items contribute to weaker market responses to FYE earnings. The results confirm that extraordinary items in FYE income statements have a significantly negative effect on earnings response coefficients. Nevertheless, even after controlling for the presence of extraordinary items, there is still a weaker market response to FYE earnings. That is, while extraordinary items are potential explanation of the weaker information content of FYE earnings, they do not fully explain why the market responds significantly weaker to FYE earnings. Moreover, we note that in 2003, FASB introduces FASB Statement No. 145 and significantly shortens the list of extraordinary items a firm can include in income statements. We examine market reaction to FYE earnings during the subsample periods prior to and post FASB Statement No. 145. Our results show no evidence that FASB Statement No. 145 helps improve the information content of FYE earnings. Finally, we are interested in what may contribute to the weaker market reaction to FYE quarter earnings during the pre-announcement period. Regulation Fair Disclosure, adopted on August 10, 2000 by the Securities and Exchange Commission, aims to prevent selective disclosure by public companies to a group of market professionals and certain shareholders. Reg 5

FD has a clear effect on investor access to information during pre-announcement period. The effect may be stronger for FYE quarter earnings news since the management is more likely to pre-announce or disclose information related to firm performance over the entire fiscal year. We examine whether Reg FD has any effect on market responses to earnings announcements during the pre-announcement window. Our results show that compared to pre-reg FD era, there is a significantly lower ERCs over pre-earnings announcement window during post-reg FD era. As a result, the difference in pre-announcement market responses between non-fye and FYE quarter earnings is no longer significant during the post-reg FD period. The rest of the paper is structured as follows. Section II describes the data used in our analysis. Section III presents the main empirical findings, with further analysis in Section IV. Section V concludes. II. Data and Summary Statistics The main data used in our analysis includes the IBES database for earnings announcement dates and analyst forecasts, the CRSP database for stock returns, Compustat for other firm characteristics, and the Thomson-Reuters 13F database for institutional holdings. CRSP stock returns are adjusted for delistings to avoid survivorship bias, following Shumway (1997). The stock sample includes common stocks traded on the NYSE, AMEX, or NASDAQ with analyst forecasts available in the IBES database. The sample period is from January 1984 to December 2015. The key variable in our analysis is analyst forecast error (FE). We follow Livnat and Mendenhall (2006) and compute analyst forecast error as follows: FE ~ ( X X ) i, t i, t i, t (1) Pi, t 6

where Xi,t is primary earnings per share before extraordinary items for firm i in quarter t and Pi,t is the price per share for firm i at the end of quarter t from Compustat. Both Xi,t and Pi,t are unadjusted for stock splits and ~ is the median of forecasts reported to IBES in the 90 days X i, t prior to the earnings announcement. Based on the IBES data, the total number of firm-quarter observations of FE is 313,823. We then merge the above sample to CRSP database and restrict to common stocks traded on NYSE/AMEX/NASDAQ. This reduces the total number of firmquarter observations to 306,100. We follow convention and limit our analysis to common stocks by excluding American Depositary Receipts (ADRs), Shares of Beneficial Interest (SBIs), certificates, units, Real Estate Investment Trusts (REITs), closed-end funds, and companies incorporated outside the U.S. from the sample. In addition, we restrict to announcements with non-missing return and price observations from the CRSP daily stock file. The total number of firm-quarter observations remaining is 301,465. Furthermore, since we compute abnormal returns using the size decile portfolio formed at the beginning of the year as benchmark, we require a firm to have valid observation of size at the end of previous year from the CRSP database, the total number of firm-quarter observations becomes 296,073. Finally, we exclude earnings announcements made more than one quarter after the reporting period, our final sample includes 293,599 total number of firm-quarter observations. Over our sample period, the average number of firms per year is 3,015, ranging from 994 in year 1984 to 4,467 in year 1998. Firm characteristics in our analysis include size (SIZE), the book-to-market ratio (BM), momentum (MOM), the Amihud illiquidity measure (ILLIQ) (Amihud, 2002), idiosyncratic volatility (IVOL), leverage (LEV), and analyst coverage (COV). They are constructed following the convention of the literature (e.g., Fama and French, 2008), as described in the following. SIZE: the natural log of market capitalization at the end of June of a year. 7

BM: the natural log of the book-to-market ratio. The book value of equity is stockholders equity plus balance sheet deferred taxes and investment tax credit (TXDITC, from Compustat), if available, minus the preferred stock liquidating value (PSTKL), if available, or the redemption value (PSTKRV), if available, or the carrying value (PSTK). Depending on availability, stockholders equity is Compustat variable SEQ, or CEQ + PSTK, or AT - LT, in that order. All Compustat items are measured for the fiscal year ending in calendar year t - 1. The market value of equity is computed at the end of December of year t - 1, from CRSP. We exclude firms with a negative book value. MOM: skip one-month lagged 11-month buy-and-hold returns. ILLIQ: Amihud s (2002) illiquidity measure is calculated as the ratio of the absolute daily stock return to the daily dollar trading volume, averaged over a given period. Since trading volume is defined differently for NASDAQ stocks and NYSE AMEX stocks, the trading volumes of NASDAQ stocks are adjusted by a factor of 0.7 (Boehmer, 2005). IVOL: the standard deviation of the residuals in the Fama French three-factor model estimated based on daily returns over a given time period. LEV: the ratio of the natural log of book assets to market equity. Book assets are total assets (from Compustat) at the end of the month of the fiscal year ending in the previous calendar year. COV: the number of analysts following a firm in a quarter. Table I reports the cross-sectional statistics of firm characteristics for selected years in our sample period. The sample size is much smaller in year 1984 due to a small number of firms with analyst coverage. The average of analyst forecast errors (FE) are mostly negative, consistent with the literature that analyst forecasts are mostly optimistic and have positive biases. Average 8

firm size increases over time. The negative book-to-market ratio indicates that the book value is on average lower than the market value. The level of ILLIQ suggests lower market liquidity in early years than in later years. Finally, the analyst coverage has increased steadily over our sample period. Table II reports summary statistics of the rank of analyst forecast errors (FE) separately for all quarters, FYE quarters and non-fye quarters. The literature suggests that firms may have a tendency of reporting bad news at the fiscal year end. Thus, FYE quarter earnings news may be generally worse than non-fye quarters. In addition, as pointed out in Palepu (1988), FYE quarter earnings may contain more nonrecurring or extraordinary expenses/revenues. These items are by nature less persistent and may reduce the predictability of FYE earnings. For a firm in a given fiscal year, we assign a value of 1 to the quarter with the highest FE, a value of -1 to the quarter with the lowest FE, and 0 to other quarters. For each firm, we then calculate the average rank of FE for FYE and non-fye quarters, separately. We also compute the first-order autocorrelation of same-fiscal-quarter earnings as a measure of persistence. For each firm, the first-order autocorrelations of same-fiscal-quarter earnings are calculated for all quarters, FYE and non-fye quarters, respectively. The table reports mean and median of these statistics. The differences in median and mean between FYE and non-fye quarters as well as their t-statistics are also reported. The results show that the differences of the mean and median rank of analyst forecast errors between non-fye and FYE quarters are statistically insignificant. This suggests that there is no significant difference in earnings news, whether it is good or bad, between non- FYE and FYE quarters. However, non-fye quarterly earnings are more persistent. Relative to FYE quarterly earnings, current non-fye quarterly earnings contains more information about future earnings of the same fiscal quarter. 9

Figure 1 shows the total number of FYE and non-fye quarter earnings announcements during each calendar quarter. The plot shows that in the first calendar quarter, about 72% of earnings announcements are for FYE quarters and about 28% for non-fye quarters. This is due to the fact that a majority of firms have fiscal year end in December. For other calendar quarters, only about 7.5% of all earnings announcements are for FYE quarters and about 92.5% for non- FYE quarters. III. Main Empirical Analysis A. Market Responses to Earnings Announcements: FYE vs. non-fye Quarters In this section, we examine market responses to earnings prior to, during and post the announcement. Following existing literature, each quarter we assign stocks to deciles based on analyst forecast error (FE) breakpoints of the previous quarter. The FEs are estimated following the procedure described in Section II. As showed in Figure 1, the numbers of FYE quarter and non-fye quarter earnings announcements are uneven among different calendar quarters. To mitigate this problem and ensure sufficient number of FYE quarter and non-fye quarter earnings announcements in each subsample, we form deciles portfolios over a four-quarter cycle from the second calendar quarter of year t to the first calendar quarter of year t+1. Over this cycle, we have both FYE quarter and non-fye quarter earnings announcements of the same fiscal year for a majority of firms. It thus allows us to examine the differences in market responses to FYE quarter and non-fye quarter earnings announcements. Panel A of Table III reports the average FE, average buy-and-hold abnormal stock returns (BHARs) (in percentage term) for each FE decile portfolio as well as spreads between the top and bottom deciles (D10-D1) over the two-day announcement window [0, 1] as well as the one- 10

week pre-announcement window [-5, -1] and two-week post-announcement window [2, 16]. D1 includes firms with the lowest FE rank and D10 includes firms with the highest FE rank. The buy-and-hold abnormal stock returns is calculated as the difference between the buy-and-hold return of the stock and that of the corresponding size decile portfolio formed at the beginning of each calendar year. Specifically, for each stock, we compute the BHARs as the difference between the buy-and-hold return of the stock and that of the corresponding size decile portfolio over the same horizon: t T t T BHAR [ t, T ] iq (1 rik ) (1 rpk ), (2) k t where rik is the return of stock i on day k and rpk is the return of the corresponding size decile portfolio on day k. t and T are the beginning date and the ending date of a holding period. The post-announcement window covers about two weeks following earnings announcement. The reason of focusing on a shorter post-announcement window is to avoid or minimize the confounding effect of SEC filing of 10-K/10-Q, which typically occurs two weeks following earnings announcements. The table also reports t-statistics of the return spreads based on Newey-West (Newey and West, 1987) standard errors that are adjusted for both heteroskedasticity and serial correlation in returns. The results in Panel A of Table III show that there is a significant immediate market reaction to earnings surprises. The abnormal returns of stocks in the top FE decile (D10) are significantly higher than those in the bottom FE decile (D1) during the two-day announcement window. The return spreads between the top and bottom deciles are also positive and highly significant over both pre- and post-announcement windows. The spreads are 2.189%, 7.447%, 5.694%, and 1.615% over [-5, -1], [0, 16], [0, 1], and [2, 16] horizons, respectively. The results over the pre- and post-announcement windows are consistent with findings in existing studies on k t 11

the pre- and post-earnings announcement drifts (e.g., Ball and Brown, 1968; Foster, Olsen, and Shevlin, 1984; and Bernard and Thomas, 1989, 1990). In particular, the post-earnings announcement drift or the PEAD suggests that firms reporting positive unexpected earnings on average outperform those reporting negative unexpected earnings after the earnings announcement. The pattern is interpreted as evidence of investor underreaction to earnings information. To compare market responses to FYE earnings surprises and non-fye earnings surprises, we divide earnings announcements in each FE decile into two subsamples: FYE quarterly announcements and non- FYE quarterly announcements. Since the deciles are formed based on FE breakpoints of the previous quarter, this is equivalent to forming FE decile portfolios separately within each subsample of earnings announcements. We are mainly interested in the differences in market responses between FYE earnings announcements and non- FYE earnings announcements. If FYE quarterly earnings are subject to less information uncertainty and are more creditable to investors than non-fye quarterly earnings, we should expect stronger immediate market reactions and subsequent weaker underreaction to FYE quarterly earnings announcements. On the other hand, if FYE quarterly earnings are subject to more information uncertainty and are less creditable to investors than non-fye quarterly earnings, we should observe the opposite pattern in market reactions. For example, firms tend to include extraordinary items in the fiscal year end earnings, adding complexity to the FYE income statements. In this case, we may observe a weaker immediate market reactions to FYE quarterly earnings surprises, followed by a stronger drift after earnings announcements. Panel B of Table III reports analyst forecast error (FE) and cumulative abnormal returns (BHARs in percentage term) of the top and bottom FE deciles and the return differentials 12

between the top and bottom deciles as well as their t-statistics for the subsamples of FYE earnings announcements and non- FYE earnings announcements. The t-statistics are based on Newey-West standard errors. At the bottom of the table, we also report the differences in FE spreads between the two subsamples, differences in return spreads over announcement window and the pre- and post-earnings announcement windows. The results in Panel B of Table III show that differences in FE spreads between these two earnings announcement subsamples are insignificant, suggesting that any differences in stock returns are likely driven by other effects rather than earnings surprises. The results show that there is a significantly weaker immediate market reaction to FYE earnings information than to non-fye earnings information. The difference in return spreads (Δ(D10 D1)) over the two-day announcement window [0, 1] between FYE and non-fye quarters is -1.350 with a t-statistics of 6.43. In addition, market reactions to FYE earnings information are also weaker than to non-fye earnings information during both the pre- and post-announcement windows. Specifically, during the preannouncement period [-5, -1], the difference in return spreads (Δ(D10 D1)) between FYE and non-fye quarters is -0.890 with a t-statistics of 4.53. The difference in return spreads (Δ(D10 D1)) during the two-week earnings announcement window [2, 16] between FYE and non-fye quarters is -0.939 with a t-statistics of 5.03. This is evidence that the significantly weaker immediate market response to FYE earnings than to non-fye earnings cannot be attributed to either stronger pre-announcement market reactions or delayed reactions. Instead, the results are evidence that FYE quarter earnings contain less information content than non-fye quarter earnings. The results in Panel B of Table III also show that the weaker market response to FYE quarter earnings is more pronounced for stocks in the lowest FE decile (D1), i.e., when FYE quarter earnings contain negative surprises. 13

B. Robustness Check: Excluding Stale Analyst Forecasts One concern of the analysis in the previous section is that the results may be driven by stale analyst forecasts. If analysts may fail to update their revisions in a timely manner, the noise in forecast error may lead to lower ERCs to FYE earnings announcements. As shown in Panel B of Table III, market response to FYE earnings announcements and particularly for FYE earnings announcements with negative surprises. One plausible story of such finding is that firms disclose bad news to the market way before FYE quarterly announcements. This leads to weaker market reactions right before earnings announcements. Moreover, if analysts fail to adjust their earnings forecasts downgrade on a timely manner, it will also lead to a lower market responses or ERCs at the earnings announcement. To address this concern, we replicate our analysis in the previous subsection by excluding stale analyst forecasts. We calculate the number of days between the most recent date an analyst issues earnings forecast and the date of earnings announcement, referred to as the number of inactive days. Stale analyst forecasts are classified as those in the top 5% with the highest number of inactive days each quarter. Table IV reports the results after excluding stale analyst forecasts. It reports average abnormal returns of the top and bottom FE deciles over different horizons and the average return differentials (in percentage term) between the top and bottom deciles as well as their Newey- West t-statistics for subsamples of FYE earnings announcements and non-fye earnings announcements. The differences in return spreads between two subsamples are also reported. The results are very similar to those in Panel B of Table III. That is, the findings of our main analysis are unlikely driven by stale analyst forecasts. 14

C. Robustness Check: Controlling for Other Firm and Earnings Characteristics As a further robustness check, we perform multivariate tests on market responses to FYE quarterly earnings by controlling for other earnings and firm characteristics. Specifically, we perform the following event-based Fama-MacBeth (1973) regression of buy-and-hold abnormal returns (BHARs) over different horizons following earnings announcements on the FE rank and its interaction with a FYE dummy as well as other control variables: BHAR a D [t-5,t-1] FYE i, [ t 1, t h] 0 ( ) i, t 1 ( ) i, t LRET β 4 10 FE [t-11,t-6 ] d β LRET IVOL β COV β 11 5 D 12 FE PERS β PRRET LRET β SIZE β BM β LEV β ILLIQ 6 13 2 7 [ 5, 1] RLAG Other Controls 8 3 [t-1,t] 9 i, t (3) where BHARi, [ t 1, t h] denotes BHARs over the horizon [t + 1, t + h] for firm i with an earnings announcement on day t, D(FE) is an indicator of FE decile rank, and d FYE is a dummy variable that is set equal to 1 if the earnings announcement is for FYE quarter and 0 otherwise. The main difference between an event-based Fama-MacBeth regression and a conventional Fama-MacBeth regression is that, in this setting, stock returns and lagged variables are defined on event dates instead of calendar dates. We include lagged stock returns and pre-announcement returns as control variables. As shown in Aboody, Lehavy, and Trueman (2010), stocks with the highest prior 12-month returns experience significantly negative market-adjusted returns immediately following earnings announcements. Following the literature (e.g., Grinblatt and Moskowitz, 2004), we include past returns over different horizons as control variables. LRET denotes lagged cumulative stock returns over various horizons. For example, LRET [ t 5, t 1] is the lagged cumulative stock return over the past five months. PRRET[-5, -1] is pre-earning-announcement return over the window [-5, -1]. Other control variables include firm characteristics, namely the natural log of market capitalization (SIZE), the natural log of the book-to-market ratio (BM), the 15

natural log of book assets to market equity ratio (LEV), the Amihud (2002) illiquidity ratio (ILLIQ), idiosyncratic volatility (IVOL), and the natural log of 1 plus analyst coverage (COV), as well as their interactions with the FE rank. All firm characteristics are lagged by at least one quarter. For details on the definitions of these variables, please refer to Section II. Earnings characteristics include reporting lag (RLAG), earnings persistence (PERS), and FYE dummy d FYE. Existing literature documents that firms tend to announce good news earlier than bad news. As shown in Table II, there is a higher persistence in non-fye earnings than in FYE earnings. We thus include earnings persistence (PERS), calculated as the first-order autocorrelation of same-fiscal-year quarterly earnings per share over the past eight years with a minimum four observations. Each year we perform the cross-sectional regressions in Eq. (3). Since BM is included as a control variable in Eq. (3), we exclude financial firms in the regressions. Table V reports the average coefficient estimates of the cross-sectional regressions with t-statistics based on Newey- West standard errors. The results show that, consistent with Aboody, Lehavy, and Trueman (2010), stock returns following earnings announcements generally have a negative relation with lagged returns. As expected, market reactions are negatively related to reporting lag. Consistent with the sorting results in Table III, there is a significant immediate market reaction to earnings surprises and the FYE quarter does significantly affect the market reaction over the earnings announcement windows [0, 1]. More importantly, the results show that the coefficient estimates of the interaction term between the FE rank and the FYE quarter dummy are negative and highly significant. The weaker ERC for FYE earnings is consistent with lower earnings persistence of FYE earnings, as reported in Table II. Also consistent with the sorting results in Table III, the coefficient estimates of the interaction term between the FE rank and the FYE quarter dummy 16

are negative and highly significant for both pre- and post-announcement windows. The results confirm that the empirical findings in Panel B of Table III are robust to controlling for other firm and earnings characteristics. IV. Further Analysis A. The Effect of Extraordinary Items: FYE versus non-fye Quarters As pointed out in Palepu (1988), management tends to use more discretions in managing FYE quarter earnings. For instance, nonrecurring or extraordinary expenses/revenues are more frequently recognized in the FYE quarter than in the other three quarters. In this section, we examine the extent to which extraordinary items in income statements contribute to weaker market reactions to FYE earnings. Table VI reports average FE and the percentage of earnings income statements with extraordinary items (PEOI) for stocks in each FE decile. D1 includes firms with the lowest FE rank and D10 includes firms with the highest FE rank. The results are reported separately for FYE and non-fye quarterly earnings announcements. The table also reports the differences in PEOI between FYE and non-fye quarterly earnings announcements for the whole stock sample and each FE decile as well as their Newey-West t-statistics. The results in Table VI show that indeed the percentage of FYE income statements with extraordinary items (PEOI) is higher than the percentage of non-fye income statements. Among stocks in different FE deciles, stocks in the bottom two FE deciles, with negative earnings surprises, have a significantly higher percentage with extraordinary items in FYE income statements. This is consistent with the finding in Panel B of Table III that the weaker market response to FYE quarter earnings is more 17

pronounced for stocks in the lowest FE decile (D1), i.e., when FYE quarter earnings have negative surprises. We perform cross-sectional regressions to test whether extraordinary items contribute to weaker market responses to FYE earnings. The main premise is that if the difference in information content between FYE and non-fye EA is mainly driven by extraordinary items in the income statements, we should see a significantly weaker ERC of FYE quarter earnings for firms with extraordinary items in the FYE income statements. We define a dummy d PEOI =1 if there is at least one extraordinary item in FYE quarter and 0 otherwise. If the weaker information content of FYE quarter earnings is mainly driven by extraordinary items, we expect these firms have even weaker information content in FYE quarter earnings than other firms. More importantly, after controlling for the effect of extraordinary items, we should see the difference in information content between FYE and non-fye EA significantly weaker or even insignificant. Specifically, we perform the following regressions: BHAR a D [t-5,t-1] FYE FYE PEOI i, [ t 1, t h] 0 ( ) i, t 1 ( ) i, t 2 ( ) i, t LRET β 4 10 FE [t,t] ILLIQ β d LRET 11 5 D [t-11,t-6 ] IVOL β 12 FE β LRET 6 COV β d 13 d PERS β D FE PRRET β SIZE β BM β 14 7 8 3 LEV RLAG Other Controls 9 [ 5, 1] i, t (4) where BHARi, [ t 1, t h] denotes BHARs over the horizon [t + 1, t + h] for firm i with an earnings announcement on day t. The control variables in the regression are the same as those in Eq. (3). Each year we perform the cross-sectional regressions in Eq. (4). We require that there are enough observations of extraordinary items over a year, i.e., at least 10% of firms report extraordinary items. Table VII reports the average coefficient estimates of the cross-sectional regressions with t-statistics based on Newey-West standard errors. The coefficient estimates of all control variables are generally consistent with those in Table V. The results show that the coefficient estimates of the interaction term between the FE rank, the FYE quarter dummy, and 18

extraordinary item dummy are negative during the pre-announcement period, significantly negative during the post-announcement period, but insignificant during the announcement window. That is, extraordinary items in FYE income statements do seem to have a negative effect on the information content of FYE earnings. Nevertheless, even after controlling for the effect of extraordinary items, there is still a weaker market reaction to FYE earnings. B. The Effect of FASB Statement No. 145 We note that on April 30, 2003, FASB introduces FASB Statement No. 145 which significantly shortens the list of extraordinary items by repealing the requirement that all early extinguishment of debt be treated as extraordinary. This eliminated most gains and losses previously treated as extraordinary items. Given the evidence in the previous subsection that extraordinary items in FYE income statements do seem to have a negative effect on the information content of FYE earnings, we are interested in whether the introduction of FASB statement No. 145 has a significant effect on market reactions to FYE earnings. We perform cross-sectional regressions in Eq. (3) over two sub-periods: pre- FASB Statement No. 145 period and post- FASB Statement No. 145 period. The two sub-periods are, respectively, April 1984 to March 2003 and April 2003 to March 2015. We focus on market reactions over the announcement window [0, 1] and post-announcement window [2, 16]. Table VIII reports the average of the coefficient estimates of annual regressions as well as the absolute values of their Newey-West t-statistics. The results show that, for regressions over both windows [0, 1] and [2, 16], the coefficient estimates of the interaction term between the FE rank and the FYE quarter dummy are negative and highly significant in the both periods: pre- FASB Statement No. 145 period and post- FASB Statement No. 145 period. In fact, the coefficient 19

estimate of the interaction term between the FE rank and the FYE quarter dummy actually has a higher magnitude during the post- FASB Statement No. 145 period. That is, there is no evidence that FASB Statement No. 145 helps improve the information content of FYE earnings. C. The Effect of Regulation Fair Disclosure Trading activities prior to earnings announcements can be driven by private information possessed by investors or information voluntarily disclosed by the management. In both cases, we may observe the pre-earnings announcement drift. On August 10, 2000, the Securities and Exchange Commission adopted Regulation Fair Disclosure which aims to prevent selective disclosure by public companies to a group of market professionals and certain shareholders. Reg FD has a clear effect on the informational environment and market responses to earnings during the pre-announcement period. In the section, we are interested in the effect of Reg FD, if there is any, on market reaction to FYE quarter earnings during the pre-announcement window. To examine the effect of Regulation Fair Disclosure, we perform analysis during the preand post Reg FD era. We allow two quarters for Reg FD to take effect in our analysis so the post Reg FD era only starts in the second quarter of 2002. Specifically, we perform cross-sectional regressions in Eq. (3) over two sub-periods: April 1984 to March 2001 and April 2002 to March 2015. The purpose is to see, with the introduction of Reg FD, whether (a) pre-announcement response to earnings surprises is generally weaker; and, more importantly, (b) the difference in pre-announcement response between FYE earnings announcement and non-fye earnings announcement becomes less significant. Table IX reports the average of the coefficient estimates of annual regressions as well as the absolute values of their Newey-West t-statistics. The results are reported for pre- 20

announcement window [-5, -1] and, for comparison purpose, the announcement and postannouncement window [0, 16]. The results show that compared to the pre-reg FD period, there is clearly a lower ERCs over the pre-earnings announcement window during the post-reg FD period. This is evidence that Reg FD has a significant effect in reducing corporate disclosure to selected market participants prior to earnings announcements. In addition, for regressions over the pre-reg FD era, the coefficient estimate of the interaction term between the FE rank and the FYE quarter dummy is negative and highly significant over the pre-announcement window [-5, - 1]. However, the coefficient estimate of the interaction term between the FE rank and the FYE quarter dummy is insignificant over the pre-announcement window [-5, -1] over the post-reg FD era. The difference in market responses between non-fye and FYE earnings is no longer significant during post-reg FD period. V. Conclusion The literature presents mixed arguments about whether FYE earnings is more or less informative than non-fye earnings. Our results show that there is significantly less information content in FYE quarter earnings than non-fye quarter earnings. Relative to non-fye earnings, market response to FYE earnings is significantly weaker not only during the announcement window, but also during both the pre- and post-announcement windows. This is inconsistent with the view that information in FYE earnings is less uncertain and is more creditable to investors. Instead, the results are consistent with the view that information in FYE earnings can be noisy since the reporting of FYE earnings is subject to more managerial discretion. We show that nonrecurring or extraordinary expenses/revenues are more frequently recognized in the FYE quarter income statements than in the other three quarters. We show evidence that extraordinary 21

items indeed contribute but only partly to the lower information content of FYE quarter earnings. However, FASB Statement No. 145, which significantly shortens the list of extraordinary items, fails to improve the information content of FYE quarter earnings. We also examine the effect of Regulation Fair-Disclosure on market reaction to earnings news over pre-announcement window and show that Reg FD has a significant effect in reducing pre-earnings announcement drift. As a result, it reduces the difference in pre-announcement market responses to FYE earnings versus non-fye earnings in the post-reg FD era. 22

References Aboody, D., Lehavy, R., & Trueman, B. (2010). Limited attention and the earnings announcement returns of past stock market winners. Review of Accounting Studies, 15(2), 317 344. Alford, A. W., & Berger, P. G. (1999). A simultaneous equations analysis of forecast accuracy, analyst following, and trading volume. Journal of Accounting, Auditing & Finance, 14(3), 219-240. Amihud, Y. (2002). Illiquidity and stock returns: Cross-section and time-series effects. Journal of Financial Markets, 5(1), 31 56. Ayers, B. C., & Freeman, R. N. (2003). Evidence that analyst following and institutional ownership accelerate the pricing of future earnings. Review of Accounting Studies, 8(1), 47-67. Ball, R., & Brown, P. (1968). An empirical evaluation of accounting income numbers. Journal of Accounting Research, 159 178. Bamber, L. S., Barron, O. E., & Stevens, D. E. (2011). Trading volume around earnings announcements and other financial reports: Theory, research design, empirical evidence, and directions for future research. Contemporary Accounting Research, 28(2), 431-471. Bernard, V. L., & Thomas, J. K. (1989). Post-earnings-announcement drift: Delayed price response or risk premium. Journal of Accounting Research, 27(3), 1 36. Bernard, V. L., & Thomas, J. K. (1990). Evidence that stock prices do not fully reflect the implications of current earnings for future earnings. Journal of Accounting and Economics, 13(4), 305 340. Boehmer, E. (2005). Dimensions of execution quality: Recent evidence for US equity markets. Journal of Financial Economics, 78(3), 553 582. Chan, K., Chan, L. K., Jegadeesh, N., & Lakonishok, J. (2001). Earnings quality and stock returns (No. w8308). National bureau of economic research. 23

Collins, D.W. and Kothari, S.P., 1989. An analysis of intertemporal and cross-sectional determinants of earnings response coefficients. Journal of accounting and economics, 11(2-3), pp.143-181. Cornell, B. and Landsman, W.R., 1989. Security price response to quarterly earnings announcements and analysts' forecast revisions. Accounting Review, pp.680-692. Dechow, P. M., & Dichev, I. D. (2002). The quality of accruals and earnings: The role of accrual estimation errors. The accounting review, 77(s-1), 35-59. Easton, P.D. and Zmijewski, M.E., 1989. Cross-sectional variation in the stock market response to accounting earnings announcements. Journal of Accounting and economics, 11(2-3), pp.117-141. Fama, E. F., & French, K. R. (2008). Dissecting anomalies. Journal of Finance, 63(4), 1653 1678. Fama, E. F., & MacBeth, J. D. (1973). Risk, return, and equilibrium: Empirical tests. Journal of Political Economy, 607 636. Foster, G., Olsen, C., & Shevlin, T. (1984). Earnings releases, anomalies, and the behavior of security returns. The Accounting Review, 59(4), 574 603. Francis, J., Lafond, R., Olsson, P., & Schipper, K. (2007). Information uncertainty and postearnings announcement drift. Journal of Business Finance & Accounting, 34(3 4), 403 433. Grinblatt, M., & Moskowitz, T. J. (2004). Predicting stock price movements from past returns: The role of consistency and tax-loss selling. Journal of Financial Economics, 71(3), 541 579. Hirshleifer, D., Lim, S. S., & Teoh, S. H. (2009). Driven to distraction: Extraneous events and underreaction to earnings news. Journal of Finance, 64(5), 2289 2325. Hirshleifer, D., Lim, S. S., & Teoh, S. H. (2011). Limited investor attention and stock market misreactions to accounting information. Review of Asset Pricing Studies, 1(1), 35 73. Hou, K., Xiong, W., & Peng, L. (2009). A tale of two anomalies: The implications of investor attention for price and earnings momentum. Available at SSRN 976394. 24

Jiang, G., Lee, C. M., & Zhang, Y. (2005). Information uncertainty and expected returns. Review of Accounting Studies, 10(2 3), 185 221. Kormendi, R. and Lipe, R., 1987. Earnings innovations, earnings persistence, and stock returns. Journal of business, pp.323-345. Kothari, S. P., Leone, A. J., & Wasley, C. E. (2005). Performance matched discretionary accrual measures. Journal of accounting and economics, 39(1), 163-197. Kumar, A. (2009). Hard-to-value stocks, behavioral biases, and informed trading. Journal of Financial and Quantitative Analysis, 44(06), 1375 1401. Liang, L. (2003). Post-earnings announcement drift and market participants' information processing biases. Review of Accounting Studies, 8(2-3), 321-345. Livnat, J., & Mendenhall, R. R. (2006). Comparing the post-earnings announcement drift for surprises calculated from analyst and time series forecasts. Journal of Accounting Research, 44(1), 177 205. Loh, R. K. (2010). Investor inattention and the underreaction to stock recommendations. Financial Management, 39(3), 1223-1252. Mendenhall, R. R., & Nichols, W. D. (1988). Bad news and differential market reactions to announcements of earlier-quarters versus fourth-quarter earnings. Journal of Accounting Research, 63-86. Newey, W. K., & West, K. D. (1987). A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica, 55(3), 703 708. Palepu, K.G., 1988, Discussion of Bad News and Differential Market Reactions to Announcements of Earlier-Quarters Versus Fourth-Quarter Earnings, Journal of Accounting Research. Shumway, T. (1997). The delisting bias in CRSP data. Journal of Finance, 52(1), 327 340. Zhang, X. (2006). Information uncertainty and stock returns. Journal of Finance, 61(1), 105 137. 25

Dhaliwal, D.S., C.A. Gleason, & L.F. Mills (2004). Last-chance earnings management: Using the tax expense to meet analysts forecasts. Contemporary Accounting Research, 21(2), 431-459. McVay, S.E. (2006). Earnings management using classification shifting: An examination of core earnings and special items. The Accounting Review, 81(3), 501-531. Barua, A., S. Lin, & A. M. Sbaraglia (2010). Earnings management using discontinued operations. The Accounting Review, 85(5), 1485-1509. Lee, L.F. (2012). Incentives to inflate reported cash from operations using classification and timing. The Accounting Review, 87(1), 1-33. Roychowdhury, S. (2006). Earnings management through real activities manipulation. Journal of Accounting and Economics, 42(3), 335-370. Zang, A.Y. (2012). Evidence on the trade-off between real activities manipulation and accrualbased earnings management. The Accounting Review, 87(2), 675-703. Gunny, K.A. (2010). The relation between earnings management using real activities manipulation and future performance: Evidence from meeting earnings benchmarks. Contemporary Accounting Research, 27(3), 855-888. 26

Table I. Summary Statistics of Analyst Forecast Error and Firm Characteristics This table reports summary statistics of analyst forecast error (FE) and firm characteristics for selected years in our sample period from 1984 to 2015. Firm characteristics include the natural log of market capitalization (SIZE), the natural log of book-to-market ratio (BM), momentum (MOM), the natural log of the ratio of book assets to market equity (LEV), the Amihud (2002) illiquidity ratio, pre-multiplied by 1,000,000, (ILLIQ), idiosyncratic volatility in percentage term (IVOL), analyst coverage (COV), and discretionary accruals (DAC). N denotes the number of stocks. The sample includes common stocks traded on NYSE, AMEX, and NASDAQ with earnings announcement dates and analyst forecasts available in the IBES database. Year Variable N 5% Mean Median 95% StDev 1984 FE 828-0.039 0.009 0.000 0.047 0.259 SIZE 826 3.431 5.926 5.979 8.426 1.552 BM 706-1.639-0.469-0.381 0.483 0.708 MOM 786-0.500-0.049-0.037 0.331 0.259 LEV 706-1.215 0.189 0.241 1.426 0.837 ILLIQ 826 0.001 0.251 0.042 1.246 0.634 IVOL 826 0.963 1.946 1.757 3.625 0.832 COV 828 1.000 2.723 2.000 7.000 1.959 1994 FE 2827-0.014-0.002 0.000 0.008 0.029 SIZE 2803 3.242 5.741 5.587 8.673 1.682 BM 1976-2.135-0.870-0.815 0.181 0.757 MOM 2431-0.485-0.011-0.048 0.590 0.360 LEV 1976-1.577-0.141-0.117 1.196 0.845 ILLIQ 2820 0.001 0.884 0.056 4.382 2.988 IVOL 2820 1.084 2.733 2.410 5.470 1.450 COV 2827 1.000 4.493 3.000 14.000 4.406 2004 FE 2680-0.008-0.002 0.000 0.009 0.075 SIZE 2644 4.507 6.877 6.721 9.810 1.640 BM 2059-2.389-0.989-0.922 0.194 0.810 MOM 2548-0.451 0.149 0.112 0.918 0.433 LEV 2059-1.707-0.288-0.308 1.197 0.888 ILLIQ 2656 0.000 0.116 0.005 0.317 0.906 IVOL 2656 0.891 2.261 2.002 4.440 1.183 COV 2680 1.000 5.754 4.000 17.000 5.548 2014 FE 2561-0.012 0.000 0.000 0.013 0.026 SIZE 2505 4.173 7.241 7.217 10.432 1.879 BM 1859-2.616-1.035-0.948 0.150 0.864 MOM 2317-0.509 0.032 0.024 0.532 0.358 LEV 1859-1.820-0.287-0.253 1.083 0.865 ILLIQ 2520 0.000 0.208 0.002 0.198 2.612 IVOL 2520 0.774 2.075 1.680 4.510 1.421 COV 2561 1.000 6.822 5.000 20.000 6.459 27