Does Greater Return Synchronicity Mean Less Firm-Specific Information? Evidence from Insider Trading

Transcription

1 Does Greater Return Synchronicity Mean Less Firm-Specific Information? Evidence from Insider Trading Claire Y.C. Liang *, Zhenyang (David) Tang, and Xiaowei Xu This version: Oct 29, 2017 Abstract Morck, Yeung and Yu (2000) suggest that greater stock return synchronicity (or R 2 ) reflects less firm-specific information in stock prices. However, a number of recent studies counter that greater synchronicity actually represents more informative stock pricing. To resolve this debate, we investigate corporate insiders a group of investors who are known to possess superior firmspecific information, and study how their trading profitability is affected by stock return synchronicity. Based on a sample of US insider transactions from 1990 to 2013, we find significantly higher insider profitability in firms with greater return synchronicity. Further, the effect is more pronounced for firms with more opaque earnings, less institutional investors and lower short interest, and for trades by officers and directors who are more likely to possess valuable firm-specific information. Overall, our results suggest that greater return synchronicity means less firm-specific information in stock prices. Keywords: return synchronicity; firm-specific return variation; insider trading; information asymmetry; market efficiency. JEL Classification: G14. * Southern Illinois University. claire.yc.liang@business.siu.edu; Address: 1025 Lincoln Dr, Rehn 124A, MC 4626, Carbondale, IL Clark University. ZTang@ClarkU.edu; Address: 950 Main Street, Worcester, MA University of Rhode Island. Xiaowei@URI.edu; Address: Ballentine Hall, 7 Lippitt Road, Kingston, RI

2 Does Greater Return Synchronicity Mean Less Firm-Specific Information? Evidence from Insider Trading Introduction A stock s return variation can come from three sources: market, industry, and firm. In an attempt to quantify the relative importance of these three components, Roll (1988) first documents that stock return synchronicity (or R 2 ) is surprisingly low in the US, suggesting that most of the stock return variation is firm-specific and cannot be explained by market-wide or industry factors. The finding started an ongoing debate on the nature of return synchronicity. Using an international sample, Morck, Yeung and Yu (2000) show that return synchronicity is higher in emerging markets and in markets with poorer property rights. The authors posit that greater return synchronicity reflects less informative stock pricing: when prices incorporate less firm-level information relative to other sources, firm-specific return variation would be lower, resulting in greater return synchronicity. Consistent with this view, Durnev, Morck, Yeung and Zarowin (2003) find that stock returns better predict future earnings when return synchronicity is lower. Durnev, Morck and Yeung (2004) document higher investment efficiency in firms with lower return synchronicity. Wurgler (2000) and Chen, Goldstein and Jiang (2007) also associate lower return synchronicity with higher sensitivity of investment to firm value. However, a growing number of recent studies suggest otherwise. Dasgupta, Gan and Gao (2010) argue that when future information is quickly incorporated into stock prices, there should be less firm-specific surprise to investors in the future; as a result, stock return synchronicity should be higher. Hou, Peng and Xiong (2013) posit that when stock price fluctuations are caused by investor sentiment, greater return synchronicity should mean more, rather than less, informative 2

3 stock pricing. Empirically, more informative stock pricing implies lower information asymmetry between insiders and outside investors. Consistent with this view, Chan and Chan (2014) document smaller SEO discounts in firms with greater return synchronicity. Teoh, Yang and Zhang (2009), Kelly (2014), and Li, Rajgopal and Venkatachalam (2014) also find positive relations between various information efficiency measures and return synchronicity. We posit that the discrepancy between these two strands of research can be primarily attributed to the selection of information measures. By definition, return synchronicity is negatively associated with firm-specific information and positively associated with market or industry information. In other words, it measures the relative amount of firm-specific information to that of market and industry, rather than the quality of a firm s overall information environment. Information measures used in prior studies, including firm size and age, number of analysts following, liquidity, earnings quality 1, SEO discounts, return anomalies and cross-listings 2, are heavily influenced by market-wide or industry factors. Therefore, the positive associations between return synchronicity and the aforementioned information measures may simply reflect the positive correlations between the industry/market information embedded in both stock prices and these information measures, such as size, liquidity, and analyst coverage etc. Based on the discussions above, we hypothesize that greater return synchronicity should mean less firm-specific information relative to market-wide or industry information, ceteris 1 Earnings numbers include significant industry-specific and market-wide components. For example, Ball and Brown (1967) document a strong association between a firm s earnings with the earnings of other firms in the industry and in the economy; Ayers and Freeman (1997) show that the industry component of earnings is incorporated into stock prices earlier than the firm-specific component. Furthermore, earnings quality is greatly affected by financial analysts following a firm, but analysts have been found to produce primarily industry-specific information rather than firmspecific information (Piotroski and Roulstone 2004; Chan and Hameed 2006). 2 SEO discounts and cross-listings are known to be affected by market timing. Furthermore, they are affected by noninformation factors such as underwriter reputation (Beatty and Ritter 1986; Cline, Fu, Tang and Wiley 2012), liquidity and risk (Karolyi 1998) 3

4 paribus. It is important to note that a spurious association between return synchronicity and quality of the overall information environment may arise through channels other than information. For instance, small firms tend to focus on niche markets in order to differentiate themselves and compete with large, bellwether firms in the industry; therefore, small firms stock prices should be more sensitive to firm-specific information rather than to market-wide or industry information. Smaller firms, at the same time, tend to have greater uncertainty and lower quality of overall information environments. It is thus critical to use a measure that reflects primarily firm-specific information, rather than the overall information environment, together with appropriate controls for the overall information environment such as size, liquidity, and analyst coverage. In this study, we investigate the relation between return synchronicity and firm-specific information by examining a unique group of investors who possess superior firm-specific information corporate insiders. Prior literature suggests that transactions by corporate insiders contain valuable information (Seyhun 1986); furthermore, the information conveyed by corporate insider trading is primarily firm-specific (Piotroski and Roulstone 2004; Jiang and Zaman 2010). These studies imply that the profitability of insider trades should positively correlate with the amount of firm-specific information unavailable to outsiders. Thus, analyzing the association between insider trading profitability and return synchronicity allows us to answer whether greater return synchronicity reflects more or less firm-specific information in stock prices: If greater return synchronicity reflects less firm-specific information, one should expect greater insider trading profitability in firms exhibiting greater synchronicity. Conversely, if greater return synchronicity reflects more informative stock prices, the association between return synchronicity and insider trading profitability should be negative. 4

5 We argue that our setting of insider trading has the following advantages compared to previous studies. First, our insider trading profitability measure reflects mainly firm-specific, rather than market- or industry-level information, which is crucial in our research question. Second, insider trades are not corporate actions and their profitability is less susceptible to complications from factors other than information. This is distinct from information measures related to SEOs and cross-listings (Dasgupta, Gan and Gao 2010; Chan and Chan 2014), which are affected by non-information factors such as underwriter reputation (Beatty and Ritter 1986; Cline, Fu, Tang and Wiley 2012), liquidity, and risk (Karolyi 1998). Third, our data allows us to exploit variations of different insiders information sets within a firm, which helps with identification. For example, some key insiders (such as directors and managers) have better access to firm-specific information compared to others (such as beneficial owners who are not directors or managers); this setting enables us to further examine the information story by testing whether our results are more pronounced for these key insider groups. We study a sample of US firms with insider transactions between 1990 and (2013), Overall, we find significantly higher insider trading profitability in firms with greater return synchronicity. When insiders buy shares in the top quintile of firms ranked by return synchronicity, the 3-day cumulative abnormal return (CAR[0,3]) is 0.195% 3 greater than that in the bottom quintile. The positive association between insider trading profitability and return synchronicity remains statistically and economically significant in different models after controlling for various explanatory variables. In addition, we examine if the positive association differs across firms with different information environments and for insiders with different information sets. We find 3 The 0.195% difference is estimated using the Fama-French 3-factor model; when the market model is used, the difference in 3-day CAR is up to 0.414%. 5

6 stronger results for firms with weaker information environments, e.g. firms with more opaque earnings, firms with less institutional holdings, and firms with lower short interests. We also document stronger results for trades by key insiders such as top managers and officer-directors, who are more likely to have access to high-quality, firm-specific information. Our results hold when firm fixed effects are included, suggesting that our findings are not driven by different amount of firm-specific information across firms. Overall, our results support the view that greater return synchronicity means less firm-specific information. This paper contributes to the literature by employing an information measure directly linked to firm-specific information. Given the wide application of return non-synchronicity (or firm-specific return variation) as a proxy for firm-specific information incorporated in stock prices, whether return synchronicity means more or less firm-specific information can lead to different interpretations of empirical findings. Our findings help resolve the debate over what return synchronicity truly reflects, and corroborate the usage of return non-synchronicity as a measure of firm-specific information in empirical research. This paper also adds to the literature on insider trading by presenting empirical evidence on how insider trading profitability interacts with return synchronicity and other information providers. Our paper is related to a recent study by Gider and Westheide (2016) who investigate insider trading and relative idiosyncratic volatility. Although idiosyncratic volatility and return synchronicity are closely related, the research questions we examine are quite different from theirs. While we study the effect of long-term, persistent return synchronicity on insider trading profitability across different firms, Gider and Westheide (2016) study how insiders time their trades based on short-term relative changes in idiosyncratic volatility within a firm. In another related study, Gangopadhyay, Yook and Shin (2014) show that future stock returns following 6

7 insider purchases are significantly higher for high-ivol stocks. If higher idiosyncratic volatility indicates lower stock return synchronicity, our findings are opposite to those reported by Gangopadhyay, Yook and Shin (2014); the difference supports the view of Li, Rajgopal and Venkatachalam (2014) that stock return synchronicity and idiosyncratic volatility are not the same. The rest of the paper is organized as follows. Section I reviews the literature and develops hypotheses. Section II describes the data and defines important variables used in this study. Section III presents the results. Section IV concludes the paper. I. Literature Review and Hypotheses Development I.A. Return Synchronicity Roll (1988) finds that the majority of stock return variation could not be explained by market-wide or industry factors. This raises an important question: why is the majority of return variation firm-specific? One view is that the firm-specific return variation reflects firm-specific information on fundamentals. Morck, Yeung and Yu (2000) compare stock return synchronicity in different countries and posit that when the stock market is efficient, firm-specific information is quickly incorporated into stock prices, and thus reduces stock price synchronicity and R 2. Several empirical findings support this view: firms with lower return synchronicity are found to have higher association between current returns and future earnings (Durnev, Morck, Yeung and Zarowin 2003), more transparency and lower crash risk (Jin and Myers 2006), more efficient capital allocations (Wurgler 2000; Durnev, Morck and Yeung 2004; Chen, Goldstein and Jiang 2007), more open capital markets (Li, Morck, Yang and Yeung 2004), more innovation outcomes (Mathers, Wang and Wang 2017), closer ties with local institutional investors (Bae, Kim and Ni 2013), and less excessive control from controlling shareholders (Boubaker, Mansali and Rjiba 7

8 2014). In addition, Bris, Goetzmann and Zhu (2007) find that when short-selling is not allowed, the downside return synchronicity (R 2 estimated using negative market returns only) increases because negative firm-specific information is not incorporated into prices. Morck, Yeung and Yu (2013) provide a review of related studies. The other view argues that firm-specific return variation arises from noise. In Dasgupta, Gan and Gao (2010) model, return synchronicity increases when stock prices better reflect future information; they also present empirical evidence showing that return synchronicity is higher for older firms, and after information events such as SEOs and cross-listings. Consistent with the view, firms with greater return synchronicity are documented to have higher analyst coverage (Chan and Hameed 2006), better earnings quality (Teoh, Wang and Zhang 2009), less return momentum and reversal (Hou, Peng and Xiong 2013), and lower SEO discounts (Chan and Chan 2014). Li, Rajgopal and Venkatachalam (2014) conclude that firm-specific return variation reflects noise and cannot be used interchangeably as idiosyncratic risk. An interesting observations is that most studies supporting Morck et al (2000) are at the country-level or industry-level, while studies suggesting greater return synchronicity means a better information environment are primarily firm-level studies. The difference could arise from the fact that low-synchronicity firms tend to be small and less liquid. Without controlling for the overall information environment, the latter may simply capture the positive association between return synchronicity and market/industry information embedded in stock prices. I.B. Insider Trading Profitability Previous studies find that insider trades, especially insider purchases, contain valuable information, despite the fact that the SEC forbids corporate insiders to trade on non-public, 8

9 material information. Seyhun (1986) finds that insider purchases usually predict significantly positive abnormal returns. This is later confirmed by Jeng, Metrick and Zeckhauser (2003) who document an average 6% annual abnormal return following insider purchases and insignificant abnormal returns following insider sales. Subsequent studies investigating the information content of insider transactions suggest that it primarily reflects insiders knowledge of firm-specific information about future earnings/cash flows or changes in firm fundamentals not fully conveyed by public announcements and financial statements (Piotroski and Roulstone 2004; Jiang and Zaman 2010; Cziraki, Lyandres and Michaely 2017). Because insider profitability comes from knowledge of firm-specific information, it should decrease with the amount of firm-specific information available to outside investors. Many studies exemplify this by documenting negative associations between insider trading profitability and information environment measures, such as R&D expenses (Aboody and Lev 2000), size (Lakonishok and Lee 2001), number of analysts (Frankel and Li 2004), discretionary accruals (Aboody, Hughes, and Liu 2005), and earnings response coefficient (Huddart and Ke 2007). We base our first hypothesis, H1, on the information content of insider trades: if greater return synchronicity means less firm-specific information, it should have a positive impact on insider trading profitability, everything else being equal. If the opposite is true, it would suggest that return non-synchronicity reflects noise rather than firm-specific information. H1: Insider trading profitability is significantly higher in firms with greater return synchronicity. I.C. Insider Groups 9

10 For a corporate insider to profit from firm-specific information, the insider must have access to high-quality, firm-specific information. This implies that insider trading profitability may vary with insider roles in the company. For example, officers and directors are considered betterinformed insiders compared to others (Lakonishok and Lee 2001; Frankel and Li 2004). Seyhun (1986) and Agrawal and Jaffe (1995) show that insiders who are both firm officers and directors possess more information about firm decisions and make more profitable trades. Cline, Gokkaya and Liu (2017) find that trades by managers better predict abnormal performance compared to trades by large shareholders or unaffiliated insiders. Agrawal and Nasser (2012) show that trades by top managers, namely Chairman, CEO, COO and President, contain more firm-specific information. Consistent with the view, Ravina and Sapienza (2010) also show that independent directors make less profitable trades compared to executives, although the gap narrows when corporate governance improves. To identify insiders different information sets, we consider their roles in a firm, such as top manager or officer-director roles. We next hypothesize that that the association between insider trading profitability and return synchronicity would be more pronounced for trades by top managers and officer-directors, and less pronounced for trades by non-officer or non-director insiders, such as beneficial owners and other affiliated persons. H2: The positive effect of return synchronicity on insider trading profitability is more pronounced for trades by key insiders. We use two ways to define key insiders. The Top Managers group consists of all insiders with titles such as Chairman, CEO, COO and President, following Agrawal and Nasser (2012). The Officer-Directors group consists of corporate directors who are also officers, following Agrawal and Jaffe (1995). Note that these two groups are not mutually exclusive. We also report the profitability of trades by all officers (the All Officers group consists of insiders who are flagged 10

11 as officers in the database), by all non-officers (the Non-Officers group consists of insiders who are not flagged as officers), by beneficial owners, and by all others who are not directors, officers or beneficial owners. I.D. Earnings Quality We also consider how a firm s information environment could affect the association between return synchronicity and insider trading profitability. Specifically, the association should be significantly stronger if the firm s earnings information is more opaque. Opaque earnings information increases information asymmetry between management and shareholders (Francis, LaFond, Olsson and Schipper 2005; Francis, Nanda and Olsson 2008). Consequently, Aboody, Hughes and Liu (2005) suggest that insiders make more profitable trades in firms with more opaque earnings. If greater return synchronicity reflects less firm-specific information, it would be easier for insiders to profit from their trades if earnings quality is weaker, e.g. more opaque earnings. H3: The positive effect of return synchronicity on insider trading profitability is more pronounced in firms with more opaque earnings. I.E. Alternative Information Providers Finally, we consider alternative information providers. Other informed investors, such as short sellers and institutional investors, may gain access to the private, firm-specific information possessed by insiders. These informed outside investors, subject to less trading restrictions than corporate insiders 4, can introduce competition in private information trading (Massa et al, 2015). 4 Corporate insiders are subject to various restrictions imposed by the SEC or by the firm, such as the short-swing rule and blackout windows. They also face more stringent disclosure requirements. 11

12 We next hypothesize that if a firm attracts more alternative information providers, it would be harder for insiders to profit from their trades. H4: The positive effect of return synchronicity on insider trading profitability is more pronounced in firms with less short interest or with lower institutional ownership. We use short interest and institutional ownership to measure the degree of competition from alternative information providers. Christophe, Ferri and Angel (2004) and Nagel (2005) show that short sellers possess firm-specific information. Drake, Myers, Myers and Stuart (2015) document that short selling strengthens the relation between current returns and future earnings. Khan and Lu (2013) find short sellers may front-run trades by corporate insiders, suggesting that short sellers are able to weaken corporate insiders informational advantage over outside investors. Similarly, institutional investors are also considered informed because they appear to profit from trading based on future earnings and returns (Ali, Durtschi, Lev and Trombley 2004; Ke and Petroni 2004; Ke, Ramalingegowda and Yu 2006). Institutional investors may affect a firm s information environment by direct trading (Boehmer and Kelley 2009) or by encouraging short selling (Nagel 2005). Some institutional investors also appoint directors and thus obtain access to private firm information. Brockman and Yan (2009) show that trades by blockholders, many of which are institutional investors, reveal firm-specific information. In a firm with greater short interest or institutional ownership, insiders would find it more difficult to profit from private information. It is worth noting that many earlier papers consider financial analysts as alternative information providers. However, recent studies suggest the information provided by analysts is primarily at industry and market levels (Piotroski and Roulstone 2004; Chan and Hameed 2006; Crawford, Roulstone and So 2012). Given that analysts have a different information set compared 12

13 to corporate insiders, we do not consider analysts as alternative information providers that may compete with insiders on private information trading in this paper 5. II. Data II.A. Sample Overview In this study, we use insider transactions in the US between 1990 and 2013 sourced from Thomson Reuters Insiders Data. The Thomson data begins in 1986, but we start our analysis from 1990 because it is the first year for which we are able to estimate all key variables 6. We keep openmarket transactions only and exclude records that are coded otherwise, including stock grants, stock gifts and options-related transactions. If an insider has multiple transactions in a day, we combine the transactions into one and replace the trade volume with the total signed volume 7. Because insider sales are documented to be driven by diversification or liquidity needs, rather than profit-seeking considerations (see Lakonishok and Lee 2001; Jeng, Metrick and Zeckhauser 2003; Gider and Westheide 2016), we focus our analysis on insider purchases only 8. This results in an initial sample of 823,606 insider purchases. After merging the sample with CRSP and Compustat, we exclude firms that cannot be matched to CRSP or Compustat, financial and utility firms, firms with negative book values, or firms with less than $10 million total assets from our sample. Applying these filters reduces our sample by 255,792, 216,095, 19,490 and 17,703, respectively. 5 In fact, our empirical results are consistent with this view. We do not find a stronger association between insider trading profitability and return synchronicity for firms with more analysts following. 6 Specifically, in our empirical results we interact return synchronicity with accounting opacity, which is estimated using the Statement of Cash Flow method following Hutton, Marcus and Tehranian (2009) is the first year for which we can estimate accounting opacity using three annual lags of discretionary accruals. 7 This is because most same-day trades by the same insiders appear to be split orders, and many insiders report them separately. 8 In untabulated results, we also analyze the association between return synchronicity and the negative CAR following insider sales. The results are largely insignificant, consistent with the previous findings that insider sales do not contain as much information as purchases do. 13

14 We further delete 100,309 trades that are smaller than $1,000 in trade value. The final sample consists of 214,217 trades by 47,257 corporate insiders. The number of unique firm-year observations is 43,042. We use Eventus to estimate abnormal returns around insider trades. Analyst coverage data is from I/B/E/S. Institutional ownership data is from Thomson Reuters. Market capitalization data comes from CRSP. Short selling data and other financial data come from Compustat. II.B. Return synchronicity We define return synchronicity (SYNCH) in a way similar to Morck, Yeung and Yu (2000) and Piotroski and Roulstone (2004). Specifically, we first regress daily returns on equal-weighted market and industry returns for every firm-year: r i,j,t = α i + β i,m r M,t + β i,j r j,t + ε i,t -- (a) where r i,j,t is the daily return of Stock i in industry j on day t, r M,t is the equal-weighted market return on day t, and r j,t is the equal-weighted Fama-French 48 industry return on day t. R 2 of the regressions are then log-transformed so that it is not bounded between 0 and 1: Ψ i,y = log ( R i,y 2 1 R 2 ) -- (b) i,y 2 where R i,y is the coefficient of determination (R 2 ) of (a) for Stock i in year y. Ψ i,y is our one-year measure of return synchronicity. Using Ψ i,y directly in our study may potentially lead to an endogeneity problem. Insiders are more likely to trade in periods of high volatility (Gider and Westheide 2016). This means a 14

15 positive association between Ψ and insider trading profitability may merely reflect insiders timing the market. As a remedy, we use the lagged three-year average synchronicity instead of Ψ i,y : SYNCH i,y = (Ψ i,y 1 + Ψ i,y 2 + Ψ i,y 3 )/3 SYNCHi,y is our main measure of return synchronicity for firm i in year y. This is similar to the idea of Hutton, Marcus and Tehranian (2009) who use a three-year average of earnings opacity to reduce noise in the measure. In robustness tests, we replicate all results using the one-year synchronicity measure of Ψ i,y, and the results are largely the same. II.C. Insider Trading Profitability We measure insider trading profitability using cumulative abnormal returns (CAR). Specifically, for every insider purchase we use the Fama-French 3-factor model to estimate coefficients in the [-300, -46] window, and calculate daily abnormal returns starting from insider trading dates. In our reported results, we use equal-weighted market returns, but using valueweighted market returns yields similar results. One empirical question with using cumulative abnormal returns is how long the event window should be. Many of the previous studies use a 6-month window because the short swing rule limits the profitability of insider purchases for the following six months (Huddart and Ke 2007). However, long-window CARs may be less accurate because of contaminating events following insider purchases. Some insider trades precede major corporate events, such as earnings announcements (Seyhun 1992; Huddart and Ke 2007), dividend announcements (John and Lang 1991), bankruptcy (Seyhun and Bradley 1997), and to some extent, mergers and takeovers (Arshadi and Eyssell 1991; Harlow and Howe 1993; Agrawal and Jaffe 1995; Agrawal and Nasser 2012). With a large sample like ours, it is difficult to exclude transactions prior to these events. In 15

16 this study, we perform most of our tests using shorter event windows: [0, 3], [0,5] and [0,10]. Adopting short event windows also alleviate the overlapping problem because many insiders make multiple purchases in a single year. Most of our results survive when we use longer windows, and some of our results actually become stronger in longer windows such as one month or three months; however, we do find noisier results when windows of six months or longer are used. II.D. Other Variables Prior literature finds that the following variables affect insider trading profitability. We include them as controls in our analysis. SIZE: log of market capitalization at the beginning of year. Smaller firms tend to have greater return synchronicity (Roll 1988); furthermore, size correlates with many measures of a firm s overall information environment. We expect insider profitability to have a positive loading on SIZE because firm-specific information should be more valuable in small firms. BTM: log of book-to-market ratio at the beginning of year. Similar to SIZE, book-to-market ratio is also shown to affect stock return (Fama and French 1992) and should carry a positive coefficient in our regressions. R&D: research and development expense divided by total assets at the beginning of year. The intensity of R&D serves as an additional control of a firm s overall information transparency, and can positively affect the profitability of insider trading (Aboody and Lev 2000; Huddart and Ke 2007). ILLIQ: illiquidity measure defined following Amihud (2002). Specifically, ILLIQ is the average ratio of the daily absolute return to the dollar trading volume on that day. Illiquid stocks 16

17 tend to move less with the market (Chan, Hameed and Kang 2013; Kelly 2014); Gassen, Skaife and Veenman (2016) show that failing to control for illiquidity can lead to biased conclusions. TRDSIZE: the size of insider purchases as a percentage of the total number of outstanding shares at the beginning of year. It reflects the price pressure of insider trading and should have a positive coefficient. ANALYSTS: log of (the number of analysts following + 1). The number of analysts following is defined as the number of unique analysts issuing 1-year earnings forecasts in I/B/E/S. Analyst coverage is an important proxy for a firm s overall information quality and thus should have a negative coefficient in regressions. In addition, we also identify three variables that may affect the association between insider trading profitability and return synchronicity, as elaborated in Section 2. OPAQUE and HIGH_OPAQUE: continuous and dummy measures of earnings opacity as defined in Hutton, Marcus and Tehranian (2009). Specifically, OPAQUE is the three-year moving average of the absolute value of annual discretionary accruals estimated from the modified Jones model (Dechow, Sloan and Sweeney 1995). HIGH_OPAQUE is a dummy variable which equals one (zero otherwise) if OPAQUE is greater than the sample median. SHORT and LOW_SHORT: continuous and dummy measures of short interest. Specifically, SHORT is the amount of short interest divided by the total number of outstanding shares at the beginning of year. LOW_SHORT is a dummy variable that equals one if SHORT is below the sample median and zero otherwise. These variables reflect trading by informed investors (short sellers) other than corporate insiders. 17

18 INST and LOW_INST: continuous and dummy measures of institutional ownership. Specifically, INST is the number of shares owned by institutional investors divided by the total number of outstanding shares at the beginning of year. LOW_INST is a dummy variable that equals one if INST is below the sample median and zero otherwise. II.E. Descriptive Statistics Table 1 provides an overview of the sample. Panel A reports the descriptive statistics of key variables used in this study, with all continuous variables winsored at the 1% level. The size of insider transactions is skewed: an average insider purchase has a size of 0.144% of total number of outstanding shares, but the median is only 0.023% of outstanding shares. When we analyze trades by insider role in untabulated results, we notice that officers, directors and beneficial owners make most of the insider transactions. [Insert Table 1 here] The three CAR measures are all positive and significant at the 1% level; furthermore, the magnitude (1.207% %) is economically significant. In Panel B, we continue to examine the abnormal returns around insider purchase dates. The patterns of daily abnormal returns are consistent with the view that insider purchases convey valuable information: Stock prices react immediately (0.262%) to insider purchases on Day 0, and the average abnormal return peaks (0.425%) on Day 1. Afterwards it gradually falls to 0.153% on Day 5, and eventually becomes insignificant after Day 10 (untabulated). Daily returns from Day 0 to Day 5 are all statistically significant at the 1% level. The patterns of daily abnormal returns support our decision to use window lengths between 3 to 10 days. 18

19 In untabulated results, we divide insider trades by insider role and examine whether trades by key insiders move prices more. Indeed, we find significantly higher cumulative abnormal returns for key insiders, such as top managers and officer-directors, suggesting that key insiders possess better-quality firm-specific information than other insiders. Panel C of Table 1 reports correlation coefficients among key variables. As expected, we observe positive correlations between SIZE, SYNCH, and ANALYSTS, while ILLIQ has negative correlations with these three variables. Before moving on to formal analysis, we examine the changes in insider trading and return synchronicity over time. In Figure 1, we plot the average 3-day CAR following insider purchases, and average return synchronicity for each year from 1990 to (2013), We first confirm the observation made in prior studies that the average R 2 decreases before 2000, but rises after 2000 in the U.S. (Campbell et al 2001; Morck et al 2000; 2013). Similarly, the average 3-day CAR following insider purchases also exhibits a positive time trend in our sample period, indicating that firm-specific information becomes steadily more valuable over time. 9 This is in clear contrast with the more stringent insider trading regulations over time. Combined with the steady growth in return synchronicity, Figure 1 suggests that the amount of firm-specific information reflected in stock prices may have deteriorated in recent years. [Insert Figure 1 here] III. Empirical Results III.A. Baseline results 9 These two series have a significantly positive correlation of (p-value: 0.000). 19

20 In this section, we present empirical test results on the association between insider trading profitability and return synchronicity. We first sort our sample into quintiles by return synchronicity, and report average CARs following insider purchases in Table 2. We find that in firms with greater return synchronicity, insider purchases are significantly more profitable. The 3- day CAR after insider purchases is 1.350% in the top quintile versus 1.155% in the bottom quintile. The 0.195% difference is statistically significant at the 1% level 10. The results are similar for longer event windows: The differences between top and bottom quintiles are 0.249% and 0.203% for event windows of 5 days and 10 days, respectively. [Insert Table 2 here] In Table 3, we formally test the association using multivariate regressions. We regress cumulative abnormal returns on return synchronicity, together with the control variables listed in Section 3.2. We also include industry and year fixed effects, and cluster standard errors at the firm level. The coefficient estimates with CAR[0,3], CAR[0,5] and CAR[0,10] as the dependent variables are reported in columns 1-3, respectively. Consistent with H1, the coefficient estimates of our return synchronicity measure, SYNCH, are positive and statistically significant at the 1% level in all three columns. A one-standard-deviation increase in SYNCH corresponds to increases of approximately 0.37% in CAR[0,3], 0.45% in CAR[0,5], and 0.67% in CAR[0,10], suggesting the results are economically significant. [Insert Table 3 here] Other control variables have signs similar to those reported in previous studies in most cases. We find that insider trading profitability is significantly higher for insider purchases of 10 Standard errors are clustered at the firm level. 20

21 larger sizes, or in firms that are smaller, with higher book to market ratios, or spend more on R&D. However, two variables (ILLIQ and ANALYSTS) have unexpected signs, likely due to multicollinearity. When we regress CAR[0,3] on ILLIQ only, we find a positive coefficient for ILLIQ. Similarly, if we drop SIZE and BTM from the regressions, the sign of ANALYSTS reverts back to negative. Our main results are unaffected if one of the correlated control variables is dropped. III.B. Insider Roles Overall, results in Table 3 are consistent with H1, suggesting that greater return synchronicity reflects less firm-specific information in stock prices. However, other channels may also lead to similar results. For example, one may argue that our results are driven by alternative trading restrictions. Many firms have self-imposed insider trading restrictions (Bettis, Coles and Lemmon 2000). If firms with lower return synchronicity have more stringent firm-level trading restrictions (although there s no evidence from the prior literature that it is the case), we may observe a positive association between insider trading profitability and return synchronicity. In untabulated results, we conduct two tests to rule out this alternative explanation. First, the most widely used insider trading restriction is the blackout period around earnings announcements. We exclude insider transactions 10-days before or after earnings announcements (practically imposing the same blackout window on all firms), and obtain results similar to those in Table 3. Second, insider trading restrictions usually apply to transactions by officers and directors only, but we are able to get qualitatively similar results after removing trades by officers and directors. One may still wonder whether other unobserved firm-level characteristics may affect both insider trading profitability and return synchronicity through a non-information channel. Fortunately, our setting of insider trading allows us to examine transactions by insiders of different roles. Since insider trading profits come from insiders firm-specific information, we anticipate 21

22 our baseline results to be more pronounced for trades executed by key insiders. If any firm characteristic drives our results through a channel other than information, there is no obvious reason for our results to be stronger for key insiders only. Table 4 reports how the association between insider trading profitability and return synchronicity varies across insider groups. We report regression results with CAR[0,3] as the dependent variable, but the results are similar with [0,5] and [0,10] windows. In Panel A, we regress our insider trading profitability measure CAR[0,3] on return synchronicity, in sub-samples divided by insider role. As described in Section 2, we define key insiders in two ways: top managers (including Chairman, CEO, COO and President) in column (1), following Agrawal and Nasser (2012), and officer-directors (including corporate directors who are also officers) in column (2), following Agrawal and Jaffe (1995). Compared to other insiders such as 10% beneficial owners, these key insiders are involved in making important decisions, and are more likely to have access to firm-specific information that is not available to outside investors. For comparison purposes, we also define a group all officers which includes all firm officers, in column (3), as opposed to non-officers which includes all non-officer insiders in column (4). We finally report two groups of insiders who are neither directors nor officers: beneficial owners of at least 10% shares who are not directors in column (5), and all other insiders who are not officers, directors or beneficial owners in column (6). We find that that the coefficient estimates of SYNCH are greater for top managers and officer-directors; specifically, the coefficient estimate of SYNCH in column (2) is more than twice as large as those in columns (5) and (6). The coefficient estimate for all officers is also about 50% greater compared to the coefficient estimate for non-officers. [Insert Table 4 here] 22

23 Table 4 Panel A shows that return synchronicity s effects on insider trading profitability are stronger with trades conducted by key insiders. In Panel B, we test whether the difference is statistically significant. We employ two dummy variables: TOP_MANAGERS and OFFICER_DIRECTORS, whose respective values would equal 1 (and 0 otherwise) if a trade is conducted by the named group, and interact them with our variable of interest, SYNCH, one at a time. The coefficient estimates of the two interaction terms are both statistically significant (at the 5% and 1% levels) and economically significant. Overall, the results in Table 4 are consistent with H2. These results suggest that our baseline results unlikely come from non-information channels. III.C. Return Synchronicity and Earnings Quality We next investigate whether the effects of return synchronicity on insider trading profitability vary across firms with different earnings quality. In Table 5, we regress post-insiderpurchase CARs on return synchronicity and its interaction with a dummy variable HIGH_OPAQUE, which equals 1 (and 0 otherwise) if a firm s earnings opacity (defined as the absolute value of discretional accrual averaged from year t-3 to t-1, following Hutton, Marcus and Tehranian (2009) is greater than the sample median. Interacting SYNCH with a continuous variable of earnings opacity gives us similar results, but it is easier to interpret the results using a dummy variable. We document significantly positive coefficients for the interaction term in all three columns. The coefficients are economically significant as well: in firms with more opaque earnings information, the effects of return synchronicity on insider trading profitability are about 40% greater, compared to in firms with more transparent earnings information. Besides, the coefficient of earnings opacity is positive and significant, consistent with previous findings by Aboody, Hughes and Liu (2005). The coefficient of SYNCH remains significant. The results are consistent 23

24 with H3: the positive association between insider trading profitability and return synchronicity should be stronger in firms with greater earnings opacity. [Insert Table 5 here] III.D. Return Synchronicity and Other Informed Investors In the last set of tests, we examine the possible effects from other informed investors. If our results so far are caused by the information story suggested by Morck et al (2000), the positive association between insider trading profitability and return synchronicity should be more pronounced when firms are less monitored by other informed investors, such as short sellers and institutional investors. These informed investors are not officially affiliated with the firms; however, they may obtain firm-specific information from an inside connection (see Cohen, Frazzini and Malloy 2008), and use the information to compete with insiders in trading that ultimately affects insider trading profitability (see Massa et al, 2015). If a firm has more institutional investors and short sellers, firm-specific information is more likely to be incorporated into stock prices before insiders can trade on it. We thus hypothesize in H4 that our baseline results would be more pronounced for firms with lower short interest and less institutional ownership. In Table 6, we regress our three CAR measures on return synchronicity and its interaction with a dummy variable LOW_SHORT which equals 1 (and 0 otherwise) if a firm s short interest is lower than the sample median. Similar to Table 5, we report three different windows in columns 1-3. We find the coefficient estimates of the interaction term significantly positive in all three columns. However, when the interaction term is included, the coefficient of SYNCH is no longer significant. The finding suggests that short sellers may indeed compete with corporate insiders in information trading, as argued in Massa et al (2015). Our baseline results seem to primarily come 24

25 from firms with low short interest. Insiders may fail to make significant abnormal returns when their firms are closely watched by informed investors such as short sellers. [Insert Table 6 here] Another group of informed investors are institutional investors. The pattern we observe in Table 6 likely holds for institutional investors as well. We then regress our three CAR measures on return synchronicity and its interaction with a dummy variable LOW_INST which equals 1 (and 0 otherwise) if a firm s institutional ownership is lower than the sample median, and report the results in Table 7. The coefficient estimates of the interaction term are positive and statistically significant across all three columns. However, unlike in Table 6, the coefficient of SYNCH remains statistically significant. The contrast between Tables 6 and 7 indicates that institutional investors are less effective in competing with corporate insiders in informed trading, possibly due to the various trading and disclosure restrictions they face. [Insert Table 7 here] Overall, our results in Tables 2-7 are consistent with the view that stock return synchronicity reflects firm-specific information in stock prices, as suggested by Morck et al. (2000). More synchronous stock returns indicate less firm-specific information in prices, enabling corporate insiders to achieve greater profits from their trades. The association becomes stronger when the trades are conducted by key insiders, when firm earnings are more opaque, or when other informed investors are less active, consistent with the information view of Morck et al. (2000). III.E. Robustness In our robustness checks, we consider alternative regression specification, as well as alternative definitions of the dependent variables and return synchronicity. We first consider a firm 25

26 fixed effects model 11. One can always argue that there are omitted variables that cause our results; in addition, our research design may be flawed if the amount of firm-specific information is positively correlated with return synchronicity (i.e. firms with more firm-specific information tend to have more market or industry level information), though we are not aware of any theory or evidence suggesting such a correlation. To address this concern, we report our baseline results including firm fixed effects in Table 8. The results are similar: SYNCH remains statistically significant at the 1% level. [Insert Table 8 here] In untabulated results, we conduct the tests in Tables 4 7 with firm fixed effects added. The results are qualitatively similar results for most of the tables, although the results in Table 6 and Table 7 become slightly weaker. The interactions remain significant when Fama-MacBeth regressions are used instead. In Table 9, we further vary how the abnormal returns and return synchronicity are measured. In Panel A, a market model, instead of Fama-French three-factor, is used to estimate abnormal returns, and in Panel B, we measure return synchronicity using a lagged one-year value rather than the three-year average. The results are unchanged. [Insert Table 9 here] Finally, in untabulated results, we find that using value-weighted returns to estimate CAR or synchronicity has little impact on the results. Our results also hold when different industry 11 The random effects model is rejected in the Hausman test. 26

27 definitions are used, or when routine insider trades are deleted following Cohen, Malloy and Pomorski (2012). IV. Concluding Remarks In this study, we document a positive association between insider trading profitability and stock return synchronicity. Corporate insiders have been documented to base their transactions, particularly purchases, on firm-specific information. Our results thus suggest that greater return synchronicity reflects less, not more, firm-specific information in stock prices, consistent with the information view suggested by Morck et al. (2000). Further analysis indicates that the positive association between insider trading profitability and return synchronicity is stronger in trades conducted by key insiders, in firms followed by fewer other informed investors, such as institutional investors and short sellers, and in firms with more opaque earnings. These results corroborate the information view and help rule out alternative explanations. This study is closely related to empirical works on the nature of firm-specific return variation or return synchronicity. In contrast to our results, some recent studies document a poorer information environment in firms with lower return synchronicity. Our study makes two improvements over these prior studies. First, our insider trading measure reflects primarily firmspecific information, while in contrast, a number of previous studies use information measures that primarily capture market- or industry-level information, such as analyst coverage. Second, not controlling for liquidity can lead to the opposition conclusion because illiquid firms tend to have both low return synchronicity and an overall weaker information environment. Our results provide an interesting comparison with prior studies that focus on a firm s overall information environment, instead of firm-specific information. We argue that by definition, 27