Informed trading before earnings shock

Transcription

1 Informed trading before earnings shock Tae-Jun Park Sungkyunkwan University and KCMI Youngjoo Lee Sogang University Kyojik Roy Song * Sungkyunkwan University roysong@skku.edu November 2013 Acknowledgement: This work was supported by the National Research Foundation of Korea Grant funded by the Korean Government (NRF-2012S1A5A2A ). We thank seminar participants at Sungkyunkwan University and 2012 Allied Korea Finance Associations Meetings for their helpful comments. All remaining errors are our own. * Corresponding author: Business school, SKKU 53 Myeongnyun-dong 3-ga, Jongno-gu, Seoul , Korea Phone:

2 Informed trading before earnings shock Abstract This paper investigates whether institutional investors trade profitably around positive or negative earnings surprises. Using Korean data over the period of , we find that information asymmetry is larger before negative earnings surprises among investors and that the trading volume decreases only before the negative events due to the severe information asymmetry. We also find that institutions sell their stocks prior to negative earnings surprises whereas individual and foreign investors do not anticipate bad news. Finally we find that institutional trade imbalance is positively related to the announcement abnormal returns of the negative events. This study complements and extends prior literature on informed trading around earnings announcements by documenting evidence that domestic institutions exploit their superior information around only the negative earnings surprises. JEL classification: G10; G30 Keywords: Information Asymmetry; Earnings surprises; Trading volume; Trade imbalance; Institutional investors - 1 -

3 Informed trading before earnings shock I. Introduction Informed trading around earnings announcements has been one of the main topics in finance and accounting literature. When assessing the value of a firm, investors should collect and interpret information regarding the future earnings or cash flows of the firm. Since earnings announcements convey new information to the market, investors would actively seek information in the pre-disclosure period. However investors can differ in their information acquisition abilities or resources, as well as in their information processing skills. Consequently, information asymmetry among investors may increase around earnings announcements. A stream of research establishes that information asymmetry increases around earnings announcements. Lee, Mucklow and Ready (1993) find that bid-ask spreads increase and quoted depths decrease before earnings announcements. Krinsky and Lee (1996) report that adverse selection component of spreads increases around earnings announcements. Chae (2005) finds that the cumulative trading volume decreases before earnings announcements. In general, prior research suggests that investors widen their bid-ask spreads, decrease order size, and/or postpone trading around earnings announcements. We point out that prior research on information asymmetry around earnings announcements does not distinguish the content of earnings announcements. That is, positive earnings surprises are treated as a similar event to negative surprises. As a result, higher information asymmetry around earnings announcements can be ascribed to both positive surprises and negative surprises. However, the magnitude of information asymmetry may differ with the different direction of surprises. Firms can voluntarily communicate with investors - 2 -

4 through investor relations, media and other means, prior to earnings announcements. Verrecchia (1983) argues that firms actively disclose only positive news whose informational value exceeds the costs of disclosure, and withhold other information. Firms can avoid the negative impact of withholding negative news on their stock prices because investors are unable to distinguish whether the firms withhold negative news or positive news whose informational value is just below a certain threshold. Supporting Verrecchia (1993), previous literature finds empirical evidence that managers voluntarily disclose good earnings news and delay the disclosure of bad news (Lev and Penman (1990); Kothari, et al. (2009)). In the survey literature of Graham, Harvey and Rajgopal (2005), some of the interviewed executives claim that they delay in releasing bad news in order to further study and interpret the information. Withholding the bad news may also be related to managers concern over their careers. Moreover, underperforming firms are more likely to abandon a project, sell some of their assets, change business strategy, or even replace their management. These actions would generate more uncertainty and therefore increase the likelihood of information asymmetry in trading stocks of bad-news firms. The findings of prior studies imply that investors may face greater information asymmetry prior to negative earnings surprises than positive earnings surprises. In this study, we explore evidence that information asymmetry is higher before negative earnings surprises than positive surprises. We also investigate whether sophisticated investors aggressively exploit their informational advantage when information asymmetry is higher. We argue that the trading patterns of informed and uninformed investors are more likely to be distinctive when information asymmetry is higher since informed investors attempt to exploit their informational advantage and uninformed investors try to avoid trades initiated by informed investors. To test this conjecture, we examine the trading pattern of each type of investors around - 3 -

5 positive versus negative earnings surprises. In our study, we employ Korean data since it has a couple of advantages. First, the Korea Exchange provides daily trading volume by investor types for all stocks traded on the Korean Stock Exchange (KSE) and on the Korea Securities Dealers Automated Quotation (KOSDAQ). Also, the Korean stock market is one of the emerging markets in which financial institutions from developed markets have actively invested. Foreign investors are the largest investor group in the Korean stock market. Therefore, the Korean data is well suited for investigating foreign investors informational advantage, relative to domestic investors. In Figure 1, we exhibit the fraction of market capitalization owned by domestic institutional investors, individual investors, and foreign investors. Foreign ownership has dramatically increased in Korea after the ownership limit on foreign investors was lifted in Foreign investors owned 18% of the capitalization of Korean stocks at the end of 1999; their ownership increased to 40% in 2004 and then gradually decreased to 32% in Meanwhile domestic institutional investors ownership marginally increased from 14% in 1999 and 16% in Individual investors are the second largest group in the Korean market. Their ownership decreased from 32% in 1999 and to 24% in We first test whether firms with negative earnings surprises are more likely to have higher information asymmetry than those with positive earnings surprises, using the proxies such as size, analyst following, idiosyncratic risk, and Amihud s (2002) illiquidity ratio. We find that firms with negative earnings surprise are smaller in size, and have smaller analyst following, higher idiosyncratic volatility, and higher illiquidity ratio. These findings indicate that firms announcing negative earnings surprise are more likely to have higher information asymmetry

6 Next, we examine the stock price returns and trading volume around earnings announcement in order to test whether positive news is more likely to be leaked before the event. We find that stock prices tend to increase starting one month before the announcements of positive earnings surprises and do not decline before negative earnings surprises, indicating that investors reflect on the upcoming positive news of stock prices during the pre-disclosure period. These results suggest that, in general, investors anticipate positive earnings surprises, but do not have much information regarding upcoming negative news. More evidence comes from trading volume prior to earnings announcements. We find that abnormal trading volume is significantly negative before the earnings announcements, which is consistent with Chae (2005). More importantly, we document that abnormal trading volume is not statistically different from zero before positive news, but it is significantly negative before negative news. This indicates that Chae s (2005) finding may be driven by abnormal trading volume prior to negative earnings surprises. Chae (2005) argues that decrease in trading volume prior to scheduled events is positively related to information asymmetry. Therefore, our finding supports the conjecture that information asymmetry among investors is greater before negative earnings surprises than positive surprises. In order to further explore the link between trading volume and information asymmetry, we investigate the trading patterns of each type of investors around earnings announcement. We classify investors into three groups domestic institutional investors, individual investors, and foreign investors, and assume that institutional investors are informed investors. Unlike individual investors, institutional investors frequently communicate with publicly traded firms as well as with brokerage firms through their investment banking, lending and asset management arms, in order to acquire information. Prior literature establishes that institutional investors are - 5 -

7 informed traders. Hendershott, Livdan and Schürhoff (2012) provide evidence that institutional investors order flow predicts the occurrence of news announcements, the sentiment of the news and the stock market reaction on news announcement days; it also forecasts earnings announcement surprises. Ke and Petroni (2004) argue that institutional investors obtain information from private communication with management, which is used to predict negative earnings. If firms tend to release positive news earlier and avoid leaking the negative news to the markets, institutional investors would have more incentives to collect private information for firms that withhold negative news. Therefore, institutions may have better private information on the upcoming negative news, compared to other types of investors, and they can make substantial short-term trading profits by moving ahead of the uninformed investors. In contrast, uninformed investors such as individual investors who are at disadvantage in information gathering and processing skills are more likely to postpone the trading of the firms stocks prior to negative earnings surprises than positive ones. Supporting this conjecture, we find that abnormal turnover by individual investors is significantly negative before the negative event, but is not significantly different from zero. We also find that institutional investors abnormal trading volume is not significantly different from zero before either positive or negative surprises. Finally, we test whether informed investors successfully exploit their superior information. If institutional investors have informational advantage prior to negative shocks, their buying and selling activities around the time of the negative news would be more distinctive than those of other investors. For the test, we estimate the standardized trade imbalance (our proxy for buying pressure) by investor type before earnings announcements. Over the trading days from -5 to -1 before negative earnings surprises, the trade imbalance by domestic institutions is significantly negative, which suggests that domestic institutional - 6 -

8 investors sell their stocks with the upcoming negative news. The trade imbalance by individuals is significantly positive, while that of foreigners is not significantly different from zero. The result indicates that individual investors trades are opposite to what the upcoming news implies; moreover, it reveals that foreigners trades are not related to the upcoming earnings news. The trade imbalance by any investor type is significantly different from those by other investor types. Previous literature provides inconclusive evidence that foreigners perform better than domestic investors in trading stocks. For instance, Grinblatt and Keloharju (2000) find that foreign investors act like momentum investors, with high levels of sophistication, and that they outperform local individual investors in the Finnish stock market. In contrast, Choe, Kho and Stulz (2005) do not find evidence that foreign investors outperform domestic institutions in the Korean market. Ekholm (2006) finds foreign investors react to earnings information in a very similar way as do the majority of investors in the market. Our result is more consistent with Choe et al. (2005) and Ekholm (2006), implying that foreign investors do not own informational advantage regarding earnings releases. To strengthen our argument that institutional investors exploit informational advantage prior to negative earnings shocks, we test whether the institutions trade imbalance before earnings announcements predicts stock returns after announcements. We find that the trade imbalance of institutions is positively related to stock returns after negative earnings surprises, but is not related to stock returns after positive events. We also find that the trade imbalances of individual and foreign investors are not related to stock returns after positive or negative earnings surprises. The result suggests that the selling pressures by institutional investors predict stock returns upon the announcements of negative earnings surprises

9 This paper adds new evidence to the extant literature on the relation between information asymmetry and earnings announcements. While the prior literature provides evidence of higher information asymmetry before earnings announcement, we differentiate the level of information asymmetry before positive vs. negative surprises, and find that the information asymmetry is higher before negative surprises. This result is consistent with the theory and empirical findings that firms voluntarily disclose positive news, but hide negative news. Our paper also complements prior research by providing evidence that institutions exploit their short-term informational advantage over other investors, especially before negative earnings surprises. We also contribute to the growing literature on foreign investors trading in the emerging markets. We find that foreign institutions trading is not systematic around earnings announcements and does not predict stock returns. This shows that foreign institutions do not possess informational advantage around earnings announcements, compared to local institutions. The rest of the paper is organized as follows. Section II reviews the related literature and develops the hypotheses. Section III describes the data and Section IV explains the empirical findings. Section V concludes the paper. II. Related literature and hypothesis development Previous literature has extensively examined information asymmetry around the earnings announcements. For instance, Kim and Verrecchia (1991) show that investors acquire private information for the opportunity to trade before and at the earnings announcements. McNichols and Trueman (1994) also argue that information asymmetry should increase before earnings announcement as there is a risk that trades are initiated by informed investors. Lee, Mucklow and Ready (1993) find that bid-ask spreads increase and quoted depths decrease before - 8 -

10 earnings announcements. Krinsky and Lee (1996) report that adverse selection component of spreads increases around earnings announcements. Chae (2005) documents that prior to earnings announcements, the cumulative trading volume decreases; which is correlated with the extent of information asymmetry. He finds that decrease in trading volume prior to earnings announcement is more prominent for small firms and firms with small analyst following and interprets the results as evidence that decreasing trading volume is associated with information asymmetry. Atiase (1985) provides empirical evidence that firm size (capitalization) is positively related to the amount of pre-disclosure information dissemination. Several studies document the positive relation between analyst following and market liquidity (Brennan and Subrahmanyam (1995); Roulstone (2003)). The results imply that analysts reduce information asymmetry by providing public information. Lang and Lundholm (1993) find that analyst following increases with more informative disclosure. Roll (1988) argues that idiosyncratic price changes can be attributable to informed trading rather than public information. Based on this notion, Ferreira and Laux (2007) use idiosyncratic risk as the proxy for information asymmetry and find that antitakeover provisions are negatively related to idiosyncratic risk. They reason that investors may have greater incentives to collect private information on firms with small anti-takeover provisions because those firms are more likely to be targeted by other firms or investors. Amihud (2002) argues that his measure defined as the ratio of absolute stock returns to trading volume over long periods reflects the price impact of informed trading. Following this literature, we use size, analyst following, idiosyncratic risk, and Amihud s illiquidity ratio as the proxy for information asymmetry

11 In this study, we divide earnings announcements into positive and negative events. Lang and Lundholm (1993) demonstrate that successful firms provide more information than unsuccessful firms. Their analysis also indicates that successful firms have less information asymmetry and less uncertainty than bad-news firms. Lev and Penman (1990) find that on average, firms voluntarily disclose good earnings news to differentiate themselves from firms with bad news and they do not find the evidence that the market punishes non-disclosing firms. Kothari, Shu, and Wysocki (2009) document empirical evidence that managers delay the disclosure of bad news, relative to good news. Specifically they find that the magnitude of the negative stock price reaction to bad news disclosures is greater than the magnitude of the positive stock price reaction to good news disclosure. If firms actively leak information before positive news, the stock prices of firms would begin to reflect on the upcoming positive news in the pre-disclosure period. In contrast, the stock prices would not reflect on the upcoming negative earnings surprises in the pre-disclosure period because bad news is not leaked to the markets. Therefore, we first test the following hypothesis: H1: Stock prices increase prior to positive earnings surprises, but do not decrease prior to negative earnings surprises. If information leakage is more prevalent before positive earnings releases, there is little incentive for investors to acquire private information in anticipation of positive announcements. In contrast, investors have incentives to acquire private information prior to negative earnings shocks. Thus, information asymmetry across investors becomes larger prior to negative earnings surprises than positive earnings surprises. Foster and Viswanathan (1995) and Chae (2005) find that the trading volume decreases before scheduled announcements, such as earnings

12 announcements, since uninformed investors try to avoid trades initiated by informed investors. We conjecture that decrease in trading volume is more prominent prior to negative earnings surprises than positive earnings surprises since information asymmetry is higher prior to negative earnings surprise than positive earnings surprise. We test the following hypothesis: H2: The decrease in trading volume prior to negative earnings surprises is larger than that prior to positive earnings surprises. A stream of research investigates informed trading of institutional investors since institutions are generally viewed as better-informed, rational and sophisticated investors who have better access to corporate information and better ability to analyze it. For instance, Yan and Zhang (2009) find that stocks for which the quarterly short-term institutional ownership has increased tend to have positive earnings surprises. To overcome the problem of low-frequency data, some recent papers have used proprietary data with a small sample size or a short time period. Griffin, Harris and Topaloglu (2003) use the trade data on Nasdaq100 stocks from Nasdaq s transaction confirmation service, and find that institutions purchase stocks that have recently risen; however, they do not predict future daily returns. Froot and Teo (2008) use custodial data from State Street Corporation, and find that institutions style investing is related to future stock returns. Baker, Litov, Wachter and Wurgler (2010) find that the average mutual fund displays the stock-picking skill in that subsequent earnings announcement returns on its weight-increasing stocks are significantly higher than those on its weight-decreasing stocks. Recently, Griffin, Shu and Topaloglu (2011) find that most institutional and individual investor groups are uninformed prior to takeovers and earnings announcements; yet, a group of large hedge fund traders sells prior to negative earnings announcements. These studies, which are

13 based on proprietary data, have disadvantages since their samples are restricted in their coverage of institutional investors, and the time span they investigate is relatively short. Other papers use the TAQ database in order to identify trading by institutional and individual investors based on trade size. Lee and Radhakrishna (2000) evaluate the performance of several alternative cutoff rules and find that a $20,000 cutoff most effectively classifies institutional trades in small stocks. However, Campbell, Ramadorai and Schwartz (2009) find that Lee and Radhakrishna s approach performs poorly when benchmarked against the quarterly institutional holdings in 13F filings. In this study, we test whether domestic or foreign institutions can take advantage of their private information before the earnings announcements using the Korean data that includes the daily net buys grouped by individuals, institutions and foreigners. Many previous studies have examined the behavior of institutions and individuals around earnings announcements. It has been argued that institutional trading is driven by their superior information gathering and processing skills. Hendershott, Livdan and Schürhoff (2012) present that the institutional order flow (buy volume minus sell volume) increases during a period of more than five days prior to the announcement of good news and decreases prior to bad news announcement. We argue that institutions have more incentive to acquire private information before negative earnings surprises than before positive earnings surprises since more private information can be acquired prior to the negative event. Accordingly, informed investors, institutions, may trade stocks with anticipated negative earnings surprises more actively in order to exploit their private information. To test this argument, we measure a standardized trade imbalance by investor types as a proxy for buying pressure before earnings announcements. We then test the following hypotheses:

14 H3: The trade imbalance by informed investors, institutions, is negative before negative earnings surprises. H3-1: The trade imbalance by informed or uninformed investors should be close to zero before positive earnings surprises since information asymmetry is small before the positive event. The extant literature provides evidence that foreign investors follow momentum trading strategies unlike local individual investors. For instance, Choe, Kho and Stulz (1999) find such evidence in the Korean market, and Kalev, Nguyen and Oh (2008) also find similar evidence for stocks traded on the Helsinki stock exchange. Brennan and Cao (1997) argue that foreign investors momentum strategies reflect their informational disadvantage in the local markets. In this paper, we investigate whether foreign investors have informational advantage on the upcoming earnings announcements, relative to local institutional or individual investors. We expect that the trade imbalance by foreign institutions would not be directional if foreigners do not have informational advantage relative to local investors. We test the following hypothesis: H3-2: The trade imbalance by foreign institutions should be close to zero before negative or positive earnings surprises. If domestic institutions trade profitably using their superior information over the short term, their trade imbalance before earnings announcement should be positively related to stock returns after the announcement. This relation should hold for negative earnings surprises since we expect that domestic institutions have informational advantage only on the upcoming negative news. We also expect that trade imbalances by domestic individuals or foreigners are not related to the stock returns. This conjecture is not consistent with Kaniel et al. s (2012)

15 finding, in which the individuals trade profitably ahead of earnings announcements. We test the following hypotheses: H4: The trade imbalance by domestic institutions before negative earnings surprises is positively related to stock returns after the events. surprises. H4-1: The trade imbalance by domestic institutions does not forecast positive earnings H4-2: The trade imbalance by individuals or foreign institutions is not related to stock returns after negative or positive earnings surprises. One might think that institutional investors can use options to take advantage of their superior information. In such a case, institutional investors trading activities may not be incorporated into the stock market. However, the Korea Exchange provides a limited number of options to investors. Currently only 33 individual stock options are traded in the derivative market, while the total number of listed companies in the Korea Exchange is 1,789, as of the end of Therefore, institutional investors are not able to take full advantage of their superior information through option trading. III. Data To test our hypotheses, we first obtain the analysts reports on South Korean companies from a database, FnConsensus of FnGuide. 1 The South Korean financial data provider, FnGuide, 1 We also collect analysts forecast data from IBES. The dates of earnings announcements from IBES tend to be later than the actual announcement dates. Therefore, we use the data from the local data provider, FnGuide

16 has collected data on analysts reports since Due to this time limitation, our sample includes analysts forecast data for South Korean companies over the fiscal years of We exclude earnings forecast data for the upcoming quarters and only include forecast data for the fiscal year-end EPS. We collect each analyst s last earnings forecast prior to earnings announcements as well as the actual earnings of each company from the FnConsensus. We then combine accounting data, trading volume by investor types, and the stock return data with analysts forecast data. We obtain the annual accounting data over the sample period of from Total Solution 2000 (TS 2000), a database compiled by the Korean Listed Companies Association. We obtain the daily trading volume of each stock by domestic individuals, domestic institutions and foreigners from the Korea Exchange. We also obtain stock return data and the Korean market index from a database (KIS-value) of Korean Information Service (KIS). The KIS is affiliated with Moody s and is a leading provider of credit-related information and services in South Korea. We exclude financial companies since such companies are heavily regulated and are less subject to information asymmetry compared to industrial companies. Our final sample consists of 1,978 firm-years (or earnings announcements) representing 590 distinctive firms, which have been traded for some period over the years Table 1 reports the number of firms over our sample period and the descriptive statistics of firm characteristics. Since we assume that information asymmetry is not the same before positive vs. negative earnings surprises, we divide our sample into firms announcing positive earnings surprises and firms announcing negative earnings surprises based on the following analyst earnings forecast error (AEFE)

17 AEFE i,t Actuali,t - Forecast i,t (1) Actual i,t where AEFE i,t denotes the analyst earnings forecast error for firm i in fiscal year t, Actual i,t is the reported EPS (earnings per share) of firm i for fiscal year t and Forecast i,t is the consensus EPS forecasts of firm i for the same year. The consensus EPS forecast for a firm is the average of the last forecasts of analysts covering the firm for the fiscal year-end EPS. If AEFE i,t is positive (negative), the firm announces a positive (negative) earnings surprise. Panel A of Table 1 exhibits the number of firms announcing positive and negative earnings surprises by year. For the year 2001, our sample includes only 39 firms since a small number of firms were covered by analysts in the early 2000s. The number of firms has increased to 275 in 2005, and following 2005, remains at a figure greater than 200. Out of 1,978 firm-years in the total sample, 590 firmyears (29.8%) announced positive earnings surprises while 1,388 firm-years (70.2%) announced negative earnings surprises. [Insert Table I about here] Panel B of Table 1 summarizes the variables that represent the characteristics of the sample firms. The mean total assets and market capitalization are approximately US$1.7 billion and US$1.3 billion, respectively at the end of the fiscal year. The mean actual earnings per share (EPS) is about US$3.4, with a standard deviation of US$7.9. The mean profitability (EBIT/total assets) is 8.7% and the mean leverage (total debt/total assets) is 40.5%. Also, the mean dividend yield is 2.1% and the mean market to book ratio of equity is 1.7. We then divide the stocks with negative earnings surprises into three portfolios (P1, P2, and P3) with equal number of stocks, and also divide stocks with positive earnings surprises into

18 three portfolios (P4, P5, and P6) with equal number of stocks based on analysts EPS forecast errors (AEFE). On average, P1 has the smallest AEFE and P6 has the largest AEFE. To test our hypotheses from H1 to H4, we use the sample firms in P1, P2, P5 and P6 since firms in P3 and P4 announce actual earnings close to market expectation. The number of firm-year observations in P1, P2, P5, and P6 is 1,319 (394 firm-year observations with positive surprises and 925 firmyear observations with negative surprises). Our main empirical tests are based on the sub-sample of 1,319 firm-years, hereafter. IV. Empirical findings 4.1. Information asymmetry prior to earnings announcements In this section, we first examine whether firms with negative earnings surprises are more likely to have higher information asymmetry, using the proxies such as firm size, analyst following, idiosyncratic risk and Amihud s illiquidity ratio. Similar to Ferreira and Laux (2007), we first estimate the market model using daily returns over the two periods, from -51 to -26 and from -25 to -1, respectively, based on Equation 2, and measure the idiosyncratic risk for each stock based on Equation 3. r i,t = α i + β i r m,t + e i,t (2) 2 = σ 2 i σ i,m 2 σ2 (3) m σ i,e where r i,t is the excess return of stock i on day t, r m,t is the value-weighted excess 2 market index return on day t, σ i is the variance of stock i, σ i,m is the covariance between r i,t 2 and r m,t, and σ m is the variance of r m,t. We also measure illiquidity of stocks following Amihud (2002) and use the Amihud s illiquidity ratio as our proxy for information asymmetry

19 since this measure reflects the price impact of informed trading. Amihud s illiquidity ratio is defined as a ratio of absolute stock returns to trading volume. Using daily stock returns and trading volume, we measure Amihud s illiquidity ratio over the two periods, from -50 to -26 and from -25 to -1, respectively, based on Equation 4. Amihud i,[t1,t 2 ] = t R i,d d=1vol i,d D i,[t 1,t2] 10 9 (4) where R i,d is the return of stock i on day d, Vol i,d is the trading volume of stock i on day d in Korean won, and D i,[t1,t 2 ] is the number of days over the event period from t 1 to t 2. In Table 2, we report that firms with negative earnings surprise are smaller in size and have smaller analyst following, higher idiosyncratic risk and higher illiquidity ratio than firms with positive earnings surprises. The result on firm size suggests that small firms may announce negative earnings surprises more frequently. The finding that analyst following is larger for positive earnings surprises indicates that public information is more prevalent before positive earnings surprises than negative earnings surprises. The result on idiosyncratic risk implies that negative-surprise firms induce more informed trading based on private information than do positive-surprise firms. Higher illiquidity ratio for negative-surprise firms implies that the price impact of informed trading is higher for negative-surprise firms than positive-surprise firms in the pre-disclosure period. In general, our findings indicate that firms announcing negative earnings surprise are more likely to have higher information asymmetry Stock return movement around earnings announcements Next we investigate how stock prices move prior to earnings announcements. We conjecture that stock prices increase prior to positive earnings surprises due to information

20 leakage but do not decrease prior to negative earnings surprises (H1). To test this hypothesis, we measure the abnormal return (AR) and the cumulative abnormal return (CAR) around earnings announcements using a standard event study. 2 We measure the abnormal return based on the market model. We use value-weighted market index of all listed companies as a proxy for the market portfolio and one-year period prior to each event (trading days of -275 to -26) as an estimation period. Then we calculate CAR over the event period from day t 1 to t 2 as below: CAR t2 AR (5) (t 1,t 2 ) t t=t1 Table III reports the mean CARs over several event periods around positive and negative earnings surprises. The results show that stock prices begin increasing before the announcements of positive earnings surprises. The mean CAR of the pre-disclosure period, trading days from -25 to -1, is 1.19% and is significantly different from zero at the 5% confidence level. The results also show that the stock prices continue to increase after the announcement of the positive event. However, stock prices do not decline before the announcements of negative earnings surprises. The mean CAR over the period of days from 0 to 5 is -1.51% and statistically significant at the 1% confidence level, which means that stock prices reflect the negative information only after the announcement. The results of the mean difference tests show that CARs are significantly different over all event periods around positive vs. negative earnings surprises. [Insert Table III about here] As mentioned in the previous section, we focus on the abnormal returns of the stocks in P1, P2, P5 and P6 since firms in P3 and P4 announce actual earnings that are close to market 2 The returns are calculated using a logarithmic return

21 expectation. We report the stock returns of the firms in P1, P2, P5 and P6 in Table IV. We find that stock prices of the firms in P5 and P6 start to rise before the announcements and continue to rise after positive news releases over some period, which is consistent with the post-earningsannouncement drift (PEAD). 3 In contrast, stock prices of the firms in P1 and P2 do not decline before the announcements and start to decline only after the announcements of the negative news releases. The results in Tables III and IV are consistent with our first hypothesis (H1), that information on upcoming positive news is more likely to be leaked to the market before the event than information on upcoming negative news. [Insert Table IV about here] 4.3. Abnormal turnover around earnings announcements We investigate the trading volume around earnings announcements following Foster and Viswanathan (1995) and Chae (2005). They find that the trading volume decreases before the scheduled announcements since uninformed investors try to avoid trades initiated by informed investors. We calculate a turnover (TO) of firm i on a specific day t as follows: Trading Volume, i,t TOi,t log( ) Outstanding i,t (6) 3 Since Ball and Brown (1968) reported the phenomenon of PEAD, the subsequent literature has extensively investigated the phenomenon

22 where Trading Volume i,t means trading volume of firm i on day t and Outstanding i,t means the number of the firm s outstanding stock on the same day. 4 We then calculate the abnormal trading volume (ATO) of firm i on a specific day t as follows: ATO i,t TO i,t - TOi, (7) where the average turnover in the pre-event period (days -55 to -26) is calculated as TO i -55 t=-26 TO 30 i,t. (8) Table V reports the cumulative abnormal turnovers (ATO) over different event periods around the announcements of earnings surprises. Panel A reports the ATO by aggregate investors. We find in the total sample that the ATO is significantly negative before earnings announcements, but is significantly positive over the event period of days 0 to 5. The result indicates that investors tend to decrease trading activities due to information asymmetry before the event, but increase trading activities after the event, which is consistent with Chae s (2005) result. Moreover, we discover that ATO is not statistically different from zero before the announcements of the positive earnings surprises, whereas ATO is negative and is statistically significant at the 1% confidence level before the announcements of the negative events. This finding is consistent with our conjecture that information asymmetry is more severe before negative events. The finding also indicates that Chae s result might be driven by information asymmetry prior to the negative event, even though our sample is different from Chae s. [Insert Table V about here] 4 We add a small constant of in Equation 4 in order to handle the problem of log transformation of a zero trading volume. The value of a constant is chosen following Llorente, Michaely, Gideon, and Wang (2002)

23 Panel B of Table V reports abnormal turnover by institutions around earnings announcements. We find that ATO is not statistically different from zero around positive or negative earnings surprises. Panel C shows that abnormal turnovers by individual investors decrease around earnings announcements. We specifically find that the ATO over the event period of days -5 to -1 is not statistically significant before positive earnings surprises, while the ATO over the same period is significantly negative at the 5% level prior to negative earnings surprises. Panel D reports ATO by foreigners and shows that the pattern of their trading activities is similar to that of individual investors trading activities. These results suggest that domestic institutions are more informed on upcoming earnings news, compared to individuals and foreigners; and that the information asymmetry problem is more severe before negative earnings surprises Trade imbalance around earnings announcements The trading volume does not show who buys (or sells) the stock with upcoming positive or negative earnings surprises. Following Malmendier and Shanthikumar (2006) and Lai and Teo (2008), we utilize a standardized trading imbalance in order to measure the direction of trading by each type of investor before the earnings news. We first calculate the trade imbalance (TI) on firm i by investor type x on day t as follows: TI i,x,t Buy Volume i,x,t - Sell Volume Buy Volume Sell Volume i,x,t i,x,t i,x,t (9) Using the standard deviation of the TI over the year, we then normalize TI in order to attain the standardized trade imbalance (STI) as follows:

24 STI i,x,t TI i,x,t - TIi,x,year(t) (10) std(ti ) i,x,year(t) Panel A of Table VI shows the cumulative STI by individuals, institutions and foreigners around the announcements of positive earnings surprises. We find that the STIs by any type of investors are not different from zero over the event period of days from -5 to -1, from 0 to 1 or from 0 to 5 around the positive earnings surprises. The ANOVA test also shows that no STI is different from those by other investor types. The result indicates that no investors, regardless of individuals, institutions, or foreigners, directionally trade over short-term periods around the announcements of positive earnings surprises. Panel B shows the STI by investor type around the announcements of negative earnings surprises. We find that the STI by individuals is positive and statistically significant before and after the negative news, suggesting that individual investors trade stocks in the opposite direction to the news. The STI by institutions over the period of days from -5 to -1 is negative and statistically significant at the 1% level, implying that domestic institutions sell stocks anticipating negative earnings surprises. We also find that trading by foreign investors is not directional around the negative news. The ANOVA test shows that each STI is significantly different from those by other investor types. Our finding that individual investors are net buyers around negative earnings surprises, and foreign investors do not show systematic response to earnings surprises, is consistent with the findings of Ekholm (2006). He finds that Finnish households are more likely to increase shareholdings at negative earnings news and that the trading activities of foreign investors are not significantly related to earnings surprises. As pointed out in Ekholm (2006), overconfidence of less sophisticated investors can be a possible explanation for individual investors trading behavior that is opposite to earnings surprises. Our result on foreign investors supports the prior literature on the Korean stock market

25 (e.g. Choe, Kho, and Stulz 1999, 2005), which does not find informational advantage of foreign investors over domestic investors. [Insert Table VI about here] In Table VII, we report the STI by investor type around the earnings news for each portfolio formed by analysts EPS forecast errors. In P1, the STI by individual investors over the period of days from -5 to -1 is 0.36 and significantly different from zero; while that by institutional investors is and significantly different from zero. The result indicates that domestic institutions sell stocks anticipating large negative surprises, whereas individual investors buy those stocks. In P1, the STI by any investor type is not different from zero after the earnings news for the stocks. Institutional investors also sell the stocks in P2 before the negative news. [Insert Table VII about here] The result in Table VII also shows that the STI by any investor type is not statistically different from zero for the stocks in P5 and P6 before and after positive news. To summarize: 1) domestic institutions make informed trading before negative earnings surprises, 2) individual investors trade in the opposite direction to the negative news, 3) foreign investors do not directionally trade around negative earnings surprises, and 4) investors do not show any directional trading behaviors around positive earnings surprises. These results are generally consistent with our hypotheses, H3, H3-1 and H Trade imbalance and after-announcement stock returns To further investigate institutions informed trading, we investigate whether their trading predicts the abnormal return in multivariate regressions. Specifically, we attempt to find the

26 relation between cumulative abnormal return (CAR) over the period of days from 0 to 5 and the STI by each investor type over the period of days from -5 to -1. We primarily estimate the following equation. CAR[0, 5] it = β 0 + β 1 STI[ 5, 1] it + β 2 ES it + β 3 CAR[ 5, 1] it + β 4 Firm Size it + β 5 MTB it + β 6 NumAnal it + ε it (11) where CAR[0,5] is the cumulative abnormal returns over the event period of days from 0 to 5; STI[-5,-1] is the standardized trade imbalance over the event period of days from -5 to -1; ES is the earnings surprise dummy, which is one if analysts EPS forecast errors are larger than the median and zero otherwise; CAR[-5,-1] is the cumulative abnormal returns over the event period of days from -5 to -1; Firm Size is the natural log of market capitalization; MTB is the market to book ratio; and NumAnal is the number of analysts covering each firm for the fiscal year. We estimate the equation using OLS with clustered standard errors and OLS with time fixed effects; the results of OLS with clustered standard errors (OLS with time fixed effects) are reported in Panel A (Panel B) of Table VII. 5 Models 1 to 3 (Models 4 to 6) show the regression results for stocks with positive (negative) earnings surprises. The results in Panel A and those in Panel B are qualitatively similar. In Models 1 to 3, the coefficients on the STIs are not statistically significant. This result suggests that investors either do not trade stocks anticipating positive earnings surprises, or they cannot exploit other types of investors since there is little information asymmetry due to information leakage prior to the announcements. In Models 4 to 6, the coefficient on STI[-5,-1] by institutional investors is significantly positive at the 5% confidence level; the coefficients on STI[-5,-1] by individuals and foreigners are not statistically 5 Refer to Petersen (2009) for an explanation on OLS with clustered standard errors and OLS with time fixed effects

27 significant. The result implies that institutions sell their stocks anticipating negative returns after announcements. The results in Table VIII support H4 and H4-1 which state that trade imbalance by institutions is positively related to stock returns only after the announcements of negative earnings surprises. They also support H4-2 that trade imbalances by individual or foreign investors do not predict stock returns after the announcements of earnings news. [Insert Table VIII about here] In summary, we find that stock prices begin to rise before the announcements of positive earnings surprises due to information leakage. We then find that the trading volume decreases only before negative earnings surprises; however, the decrease is driven by individuals and foreign investors. We also find that institutional investors directionally trade only before the negative news since they have informational advantage on the upcoming announcements of negative earnings surprises. Finally, we document that foreign investors do not have superior information on the most important corporate event, earnings announcements Robustness check In the previous sections, we use the analysts earnings forecasts error (AEFE) to classify earnings news into positive and negative earnings surprises. This approach assumes that analysts forecasts reflect the true market expectations about upcoming earnings news. However, analysts forecasts can be a noisy measure for market expectation if analysts failed to reflect all available information into their forecasts and/or they were biased in some ways. Therefore, in this subsection, we use market reactions to earning announcements as an alternative measure for market expectation, assuming that positive (negative) market reaction after announcements is the reflection of positive (negative) market shocks. Specifically, we first measure the cumulative

28 abnormal returns over the event period of 0 to 5. Abnormal return is estimated based on the market model. We define the firms with a positive (negative) CAR as the firms with a positive (negative) earnings surprise. Then we investigate investors trading activities prior to positive vs. negative earnings surprise and report the results in Table IX. Panel A shows that individual investors and foreigners decrease their trading volume prior to negative earnings surprises while they do not decrease it prior to positive surprises. There are no significant changes in institutional investors trading volume prior to both positive and negative surprise. Panel B shows that only institutions sell stocks with negative surprises in the pre-disclosure period. The result of Panel A (Panel B) is qualitatively similar to the results of Table V (Table VI) in which earnings news are classified based on the analysts earnings forecasts error. In general, the results of Table IX are consistent with our conjecture that information asymmetry is higher prior to negative earnings surprises and institutional investors exploit their superior information in the pre-disclosure period of negative news. [Insert Table IX about here] V. Conclusion In this study, we investigate informed trading around earnings announcements. Specifically, we examine 1) whether information asymmetry prior to earnings announcement is related to the content of earnings information (positive vs. negative surprise to the market), and 2) whether informed traders exploit their informational advantage over the short-term period around earnings announcements. We conjecture that information asymmetry is higher before negative earnings surprises, compared to positive surprises, because firms are more likely to leak positive news and delay the disclosure of negative news. In such a case, investors would have greater