Behavioral Biases of Informed Traders: Evidence from Insider Trading on the 52-Week High

Transcription

1 Behavioral Biases of Informed Traders: Evidence from Insider Trading on the 52-Week High Eunju Lee and Natalia Piqueira ** January 2016 ABSTRACT We provide evidence on behavioral biases in insider trading activities. Using a stock s 52-week high, we find that insiders are subject to anchoring biases. They are reluctant to purchase stocks when the stock price is near its 52-week high and to sell stocks when the stock price is far from the 52-week high. Furthermore, we find evidence of insiders disposition effects, their tendency to sell (buy) stocks when the stock price is near (far from) the 52-week high. Despite the evidence of insiders biased trading, the return predictability of insider trading still appears when the price is near and far from the 52-week high. This suggests that insiders are heterogeneous in terms of their information advantage. Overall, our findings verify that insiders possession of private information cannot prevent them from trading with behavioral biases. JEL classification: G12, G14 Keywords: Insider trading, 52-week high, anchoring bias, disposition effect Corresponding author. Manning School of Business, One University Avenue, University of Massachusetts Lowell, Lowell, MA 01854, Tel: , eunju_lee@uml.edu ** C. T. Bauer College of Business, 334 Melcher Hall, University of Houston, Houston, TX, 77204, Tel: , npiqueira@bauer.uh.edu.

2 Behavioral Biases of Informed Traders: Evidence from Insider Trading on the 52-Week High ABSTRACT We provide evidence on behavioral biases in insider trading activities. Using a stock s 52-week high, we find that insiders are subject to anchoring biases. They are reluctant to purchase stocks when the stock price is near its 52-week high and to sell stocks when the stock price is far from the 52-week high. Furthermore, we find evidence of insiders disposition effects, their tendency to sell (buy) stocks when the stock price is near (far from) the 52-week high. Despite the evidence of insiders biased trading, the return predictability of insider trading still appears when the price is near and far from the 52-week high. This suggests that insiders are heterogeneous in terms of their information advantage. Overall, our findings verify that insiders possession of private information cannot prevent them from trading with behavioral biases.

3 1. Introduction Insiders trade for several reasons. The most well-known motive lies in their opportunistic profitseeking behavior based on private information, since they have access to confidential information about the prospects of the firm and its fundamental value. An extensive literature supports this argument by finding subsequent abnormal returns following insider transactions (Jaffe 1974; Seyhun 1986, 1992; Rozeff and Zaman 1988; Lin and Howe 1990; Lakonishok and Lee 2001; Jeng et al. 2003; Ke et al. 2003; Huddart et al. 2007; Agrawal and Cooper 2015). There could be some other non-profit-seeking motives for insiders to trade, such as portfolio diversification and liquidity needs. They may sell their stocks to diversify risk associated with their wealth or in need of money for their personal consumption (Ke et al. 2003; Huddart and Ke 2007; Huddart et al. 2007; Kallunki et al. 2009). In this paper, we offer evidence of another motive for insider trading: behavioral biases. Some studies have found evidence of several behavioral biases of general investors or managers, 1 but to the best of our knowledge, there are very few studies documenting insiders behavioral biases in their trading activity. 2 The fact that corporate insiders may have private information or better ability to predict the prospects of the firm makes it hard to suspect that their trading is driven by behavioral biases. However, all insiders may not be informed, and even if they are all informed, they may not have perfect information about the firm s fundamental value. Some recent studies support this argument by finding behavioral biases in informed trading. Campbell and Sharpe (2009) document that experts' consensus forecasts of macroeconomic data are biased towards the values of the previous month. Feng and Seasholes (2005) show that 1 For example, Malmendier and Tate (2008) and Jin and Kothari (2008) document that managers tend to be overconfident, and Heath et al. (1999) find executives disposition effects in exercising stock options. Meanwhile, an extensive literature has documented general investors behavioral biases, such as disposition effects in their trading (Shefrin and Statman 1985; Odean 1998; Grinblatt and Keloharju 2001; Frazzini 2006). 2 Using Swedish market data, Kallunki et al. (2009) find that 1) male insiders trade more aggressively than women (overconfidence), and 2) insiders tend to hold on to losing their stocks (disposition effect). 1

4 sophistication and trading experience cannot get rid of the disposition effect. Yet, there is no such evidence found in insider trading in the U.S. market. In this sense, our study contributes to the prior literature on behavioral finance and informed trading. To examine the behavioral biases of corporate insiders, we employ a stock s 52-week high price as a proxy that triggers investors behavioral biases. A stock s highest price within the past 52 weeks is one of the market indicators publicly available through the financial media such as the Wall Street Journal. Since it simply indicates a peak price reached within a year, estimating the distance of the current stock price to this highest price should not provide important information about the firm s fundamental value. Regardless, the 52-week high is often used among market participants as a price resistance level or a leading indicator for a price rally. Financial economists provide supportive evidence of it, claiming that the 52-week high price plays a role as an anchor when investors evaluate information (George and Hwang 2004; Huddart et al. 2009; Drissen et al. 2012; Hong et al. 2015). They show that such anchoring behavior of investors leads to underreaction to news, followed by return momentum. When the stock price is near its 52-week high, investors are reluctant to buy the stock at a high price that fully reflects the good news. The price goes up after the information prevails. When the stock price is far from its 52-week high, investors are reluctant to sell the stock at a low price implied by bad news. The price falls after the bad information prevails. In this study, we investigate how corporate insiders trade on the nearness to the 52-week high. Given the prior findings of the informativeness of insider trading, they are highly likely to possess private information about the firm s fundamental value, and thus their trading will not rely on the distance of the current stock price to its 52-week high. Possibly, they may be sophisticated enough to earn profits by exploiting other investors underreaction driven by 2

5 anchoring biases associated with the 52-week high. On the other hand, insiders may be susceptible to anchoring biases like other investors. This is a valid prediction, because we posit that insiders information advantage and behavioral biases are not mutually exclusive. We find evidence of the anchoring biases of insiders in this study. Corporate insiders tend to decrease their stock purchases when the stock price is near its 52-week high and reduce their selling when the price is far from the 52-week high. These results suggest insiders underreaction associated with the distance to the 52-week high, demonstrating that informed traders can be subject to behavioral biases. In addition, we find evidence of insiders disposition effects, their willingness to sell (buy) stocks when the price is near (far from) the 52-week high. Such biased behavior of insiders differs from their contrarian trading documented by Lakonishok and Lee (2001), and the results are robust to various firm characteristics and insider trading measures. We also examine the return predictability of insider trading when the stock price is near and far from its 52-week high. Our results show that high levels of insider trading predict future returns when the price is near and far from the 52-week high. High levels of insider purchases are followed by higher returns than low levels of insider purchases when the stock price is close to the 52-week high. Also, high levels of insider sales turn out to predict lower returns than low levels of insider sales when the stock price is far from the 52-week high. These results suggest that insiders are heterogeneous with respect to their information advantage. Some of them are subject to anchoring biases when the stock price is close to and far from its 52-week high, while some of them still tend to trade based on their private information. Overall, the results show that the anchoring bias is an important determinant of insiders trading behavior. When the stock price is close to its 52-week high, insiders are reluctant to buy stocks with concerns about price declines in the future. When the price is far from the 52-week 3

6 high, they are reluctant to sell stocks. Nonetheless, future return patterns following insider trading show that the informativeness of insider trading is not entirely dominated by their anchoring behavior. To the best of our knowledge, this study is the first to investigate behavioral biases in insiders trading activity in the U.S. market. Although recent studies document biases in trading behavior of individual investors (George and Hwang 2004; Du 2008; Huddart et al. 2009; Hong et al. 2015) or managers corporate decisions (Baker et al. 2012), there is no study that examines the effect of insiders behavioral biases on their trading. On the one hand, given that insiders have information advantage, the results of this study complement research work on the biases of informed traders (Feng and Seasholes 2005; Campbell and Sharpe 2009; Watson and Funck 2012; Beschwitz and Massa 2013; Lee and Piqueira 2015). On the other hand, given mixed evidence on the behavioral biases of informed traders, such as analysts (Campbell and Sharpe 2009; Cen et al. 2013), short sellers (Watson and Funck 2012; Beschwitz and Massa 2013; Lee and Piqueira 2015), and managers (Degeorge et al. 1999; Heath et al. 1999; Baker et al. 2012), our findings imply that whether informed traders are subject to behavioral biases varies depending on different groups of informed traders. Furthermore, our findings of different degrees of anchoring biases among different groups of insiders, such as management and large shareholders, indicate that there exists heterogeneity in terms of the quality (or quantity) of insiders private information, which has not been documented in the prior literature. Our examination on anchoring biases in stock buying and selling also provide important insights into behavioral biases in different types of trading. The disposition effect indicates investors reluctance to sell losers and willingness to sell winners (Shefrin and Statman 1985; Grinblatt and Han 2005), describing asymmetric behavioral biases depending on future capital 4

7 gains and losses earned by investors. However, to our knowledge, there is no study investigating behavioral biases in buying and selling separately, mostly due to the unavailability of data. In this respect, our study suggests interesting avenues for future research on asymmetric heuristics on buying and selling activities. The remainder of this paper is organized as follows. Section 2 discusses our empirical prediction on insider trading based on the nearness to the 52-week high, and Section 3 describes data and methodology. Section 4 explains return patterns associated with the nearness to the 52- week high, and Section 5 discusses main results on insider s reaction to the distance of the stock price to the 52-week high. In Section 6, we examine the return predictability of insider trading when the price is near and far from the 52-week high. Section 7 performs additional tests using trades by other types of insiders and the 52-week low. Section 8 finally concludes. 2. Anchoring Biases in Insider Trading This study focuses on insiders reaction to the nearness of the current stock price to its 52-week high. George and Hwang (2004) find that investors are reluctant to purchase stocks at high prices implied by information when good news pushes stock prices toward the 52-week highs. By the same logic, investors are reluctant to sell stocks at low prices implied by information when bad news pushes the prices far from the 52-week highs. Such underreaction is followed by positive and negative return momentum when the price is near and far from the 52- week high, respectively. They also show that trading strategies exploiting this anchoring behavior dominate the momentum strategies of Jegadeesh and Titman (1993) and Moskowitz and Grinblatt (1999). 5

8 We expect that insiders react to the nearness to the 52-week high, but, as in Lee and Piqueira (2015), the direction of our prediction relies on the relationship between the informativeness of insider trading and behavioral biases. The prevailing view in the literature is that information advantage and heuristics are mutually exclusive, meaning that informed traders are not subject to behavioral biases. This suggests that insider trading will not be affected by anchoring biases associated with the 52-week high, since insiders have access to private information. Moreover, they may exploit other investors anchoring behavior by buying the stock when the price is near the 52-week high and selling the stock when the price is far from the 52- week high. If there exists positive (negative) return momentum for stocks whose prices are near (far from) the 52-week highs, as documented by George and Hwang (2004), insiders can make profits through these trading strategies. If this is the case, our study will provide additional insights into the information advantage of insiders: they not only have private information but also possess better ability to time their trades using public information. In contrast, informed traders can be susceptible to anchoring biases. Although they have information advantage, their information set could be imperfect. In the case of corporate insiders, they may not have perfect information about a firm s prospects or superior skills to analyze the degree of market misvaluation. Given this, insiders may underreact to news when the stock price is near and far from the 52-week high. They may be reluctant to buy stocks when the stock price is close to its 52-week high and to sell stocks when the price is far from the 52-week high. In this case, such biased trading may hurt the return predictability of insider trading. As mentioned in Section 1, many existing studies have shown that insider trading predicts subsequent abnormal returns. If insiders are subject to anchoring biases, however, their trading 6

9 activities for stocks whose prices are near and far from the 52-week highs may not be followed by abnormal returns. 3. Data and Methodology 3.1. Data and Measures We obtain insider trading data from the Thomson Reuters database. Section 16(a) of the Securities and Exchange Act requires insiders to report their trades within 10 days following the end of the trading month. 3 The reports are filed on Form 4 and include information about insiders transaction data, along with their identity and relationship with the firm. Our sample includes all common stocks traded on the NYSE, AMEX, and NASDAQ exchanges from January 1986 to December Following Alldredge and Cicero (2015), we include transactions that are verified in terms of their accuracy with a cleanse code of R, H, L, C, or Y. We exclude transactions related to options exercises and private transactions. We focus on open market purchases and sales of common stocks with trading prices between $5 and $999 and positive trading volume. Following Lakonishok and Lee (2001) and Marin and Olivier (2008), we classify insiders into three groups: Management, large shareholders, and others. Management includes directors, committees, officers, presidents, vice presidents, and affiliates. Large shareholders include direct and indirect shareholders and beneficial owners who own more than 10% of shares but are not involved in management. 4 The rest of insiders, such as controllers, founders, former officers, etc., 3 Due to the enactment of the Sarbanes-Oxley Act of 2002, insiders are required to report their trades to SEC within two business days following their transaction date. Cohen et al. (2012) show that the median delay between trading and report dates during their sample period is three days. 4 Marin and Olivier (2008) define large shareholders as equity holders who own more than 20% of shares, but we follow Lakonishok and Lee (2001) because the Thomson Reuters Insiders Data defines beneficial owners as those who own more than 10% of shares. 7

10 are grouped as others. Our main analysis focuses on trading activities of the management group, because the main goal of this study is to investigate if insiders who have private information are subject to anchoring biases and, according to prior studies, trades by large shareholders who are not in management do not convey information compared to trades by managers (Seyhun 1986; Kahle 2000). 5,6 We aggregate these insider transaction data on a monthly basis and merge them with monthly stock data from CRSP and quarterly financial statement data from COMPUSTAT. We also collect daily stock market data from CRSP to estimate Amihud s (2002) illiquidity measure which is the monthly average of daily illiquidity ratios. Using these datasets, we calculate a stock s 52-week high, which is the highest price within the past 52 weeks. Following George and Hwang (2004), we use a 52-week high ratio as a proxy for the nearness of the stock price to its 52-week high. The 52-week high ratio (52wkHigh i,t ) is computed as follows: 52wkHigh i,t = P i,t P 52wkHi,t (1) where P i,t is price of stock i at the end of month t, and P 52wkHi,t is the highest price of stock i over the 52 weeks ending on the last day of month t. A high level of the 52-week high ratio suggests that the stock price is close to its 52-week high, while a low level of the 52-week high ratio suggests that the price is far from the 52-week high. We employ several insider trading measures following existing studies. Our first measure is turnover ratio, which is calculated as the number of shares traded by insiders scaled by the 5 Lakonishok and Lee (2001) also mention that the definition of large shareholders is not clear and their trades can be motivated by various considerations. 6 We separately analyze trading behavior of large shareholders and discuss the results in Section

11 number of shares outstanding. We also estimate a value turnover ratio, which is calculated as dollar volume, trade price times the number of shares traded, scaled by market capitalization. We calculate these ratios for insider purchases and sales separately so that we can perform independent analyses on buying and selling. Our second measure is the proportion of insider purchases or sales to total transactions, which are the sum of purchases and sales. 7 We calculate these ratios based on the number of shares traded, dollar trading volume, and the number of transactions made by insiders and for insider purchases and sales separately. For instance, we can calculate the proportion of the number of shares purchased by insiders to the total number of shares traded by insiders ( ainsi_p ), and this ratio can be calculated based on insider sales ( dinsi_p ). All these variables are described in more detail in Appendix Summary Statistics Table 1 reports the time-series statistics of the cross-sectional means for key variables in the entire sample. Panel A reports the mean and median values of insider trading variables used in the analysis. These values are presented for each insider group management, large shareholders, and others, and for each trading type purchases and sales, respectively. Insider sales are in general larger than insider purchases for all insider groups. For instance, the average number of shares purchased by management is 28,200 shares, while the number of shares sold by management is 80,910 shares. When it comes to the turnover ratio, which is calculated as the number of shares traded by insiders divided by the number of shares outstanding, the average turnover ratio for management sales (0.239%) is higher than the ratio for management purchases 7 Kahle (2000) addresses that this measure does not include extreme outliers and is less sensitive to changes in trading activity over time than levels of sales or purchases. 9

12 (0.116%). These results are robust to insider trading measures and insider groups, except the proportions of the number of shares, dollar volume, and the number of trades made by large shareholders. When we compare purchases and sales by insider group, we find that large shareholders trade a higher amount of shares compared to management and others. Panel B reports the descriptive statistics of firm characteristic variables, which are described in Appendix 1. The average 52-week high ratio is %, suggesting that the stock price on average is close to its 52-week high. The average trading volume and monthly return are 18.8 million shares and 2.82%, respectively. Table 2 reports time-series averages of cross-sectional correlations among the variables used in the analysis. The relationships between the 52-week high ratio and insider trading variables are of our interest, because the results may hint how insider purchases and sales are associated with the nearness to the 52-week high. When we use turnover ratios for management purchases and sales as insider trading measures, we find that the 52-week high ratio is significantly negatively correlated with insider purchases (-0.06), while it is significantly positively correlated with insider sales (0.014). 8 There results show that, when the stock price becomes close to the 52-week high, insider purchases tend to decline but insider sales tend to increase. Given that anchoring biases explain investors reluctance to buy (sell) stocks when the price is near (far from) the 52-week high, our correlation results suggest that insiders may anchor on the distance to the 52-week high and underreact to news like other investors in the market. Since these results are based on pairwise correlations between the 52-week high ratio and insider trading, we analyze the relation between the nearness to the 52-week high and insider trading in more detail using different methodologies in Section 5. Before implementing the main analysis, 8 We also calculate correlations using other insider trading variables introduced in Table 1. The results are in general unchanged, although the magnitudes are different. The results are available upon request. 10

13 we examine return patterns associated with the 52-week high ratio to verify the existence of other investors anchoring behavior in the next section. 4. Return Patterns Associated with the 52-Week High This study is motivated by prior findings that investors are reluctant to buy stocks when the stock price is near the 52-week high and to sell stocks when the price is far from the 52-week high. As such, it is important to verify that such anchoring behavior exists in our sample during the sample period. This can be confirmed by subsequent return momentum when the price is near and far from the 52-week high. That is, when the price is close to the 52-week high, we expect to find positive subsequent returns or upward return momentum. On the other hand, negative subsequent returns or downward momentum will follow when the price is far from the 52-week high. For this analysis, we run a fixed-effects regression of future return on the nearness to the 52-week high with firm- and month-fixed effects: ret i,t+1 = α + β 1 52wkHigh i,t 1 + β 2 ret i,t + β 3 mom i,t + β 4 lnmcap i,t 1 + β 5 bmratio i,t 1 + β 6 illiq i,t 1 + ε i,t+1 (2) where the dependent variable (ret i,t+1 ) is the return of stock i in month t+1, and the independent variables are the 52-week high ratio in month t-1 (52wkHigh i,t 1 ) and other control variables described in Appendix 1. We also use subsequent returns from month t+1 to months t+3 and t+6 as dependent variables to see if the return predictability of the 52-week high ratio appears over the long horizon. To avoid the bid-ask bounce effect, we skip one month between the estimation of the 52-week high ratio and future return. 11

14 Secondly, we estimate Fama-MacBeth (1973) regressions to see if the results from the fixed-effects regressions are robust. For the estimation, we run the cross-sectional regressions of Equation (2) every month and calculate the time-series averages of the coefficient estimates. Table 3 reports the estimation results. Panel A presents the fixed-effects regression results. When we regress one-month raw return on the 52-week high ratio in column (1), the coefficient on the 52-week high ratio (1.802) is significantly positive. This suggests that a positive (negative) return follows when the stock price is near (far from) its 52-week high. The results are robust when we extend holding periods for returns to 3 and 6 months in columns (2) and (3), but the magnitudes of the coefficients get smaller with the longer return horizons. Panel B presents the results of Fama-MacBeth regressions. Positive coefficients on the 52-week high ratio in columns (4)-(6) are consistent with the results in Panel A. Although the coefficients are of smaller magnitudes and less significant compared to those in Panel A, the overall results are significant and robust. The results in Table 3 confirms the findings of George and Hwang (2004) that the nearness to the 52-week high leads to investors' underreaction. Given that the nearness to the 52- week high is associated with sporadic news, the results are also consistent with Griffin and Tversky (1992), who argue that individuals underreact to intermittent news. Overall, our results indicate that investors' anchoring behavior based on the nearness to the 52-week high leads to return momentum. The results therefore verify the validity of the 52-week high as an anchor when we investigate insiders trading activities in the following section. 5. Insider Trading and the Nearness to the 52-week high 12

15 We have shown that the nearness to the 52-week high is positively associated with future returns, suggesting investors underrection when the stock price is near and far from the 52-week high. Next we examine if insiders trade based on the nearness to the 52-week high. As mentioned in Section 2, how insiders react to the nearness to the 52-week high depends on the quality or quantity of private information possessed by them. They may exploit other investors underreaction by increasing their purchases and sales when the price is near and far from the 52- week high, respectively. If they are not informed enough to exploit others anchoring biases, they themselves may be susceptible to anchoring biases associated with the 52-week high. We first examine this by soring stocks into quintiles based on the nearness to the 52-week high and looking into insider purchases and sales for each portfolio. Table 4 reports insider trading for portfolios sorted on the nearness to the 52-week high. Q1 in the first row indicates the portfolio of stocks for which prices are farthest from the 52-week highs, and Q5 indicates the portfolio of stocks for which prices are nearest the 52-week highs. With respect to insider purchases, we find that insiders tend to decrease their buying as the price is close to the 52-week high. The turnover ratio for insider purchases (ainsi) is 0.137% for Q1, and it decreases to 0.105% for Q5. The difference in insider purchases between Q1 and Q5 (Q5-Q1) is negative (-0.032%), even though it is marginally significant. The results are more significant when we use the proportions of insider purchases to total transactions. The difference in the proportions of the number of shares purchased by insiders between Q1 and Q5 is significantly negative (-0.117%), suggesting that insiders are reluctant to buy the stock when the price is close to the 52-week high. Therefore, our results support insiders anchoring behavior on the buy side. Results on the sell side also suggest insiders biased trading associated with the distance to the 52-week high. When we use the turnover ratio for analysis, insider sales for the portfolio 13

16 farthest from the 52-week high (Q1) (0.225%) are significantly smaller than those for the portfolio nearest the 52-week high (Q5) (0.267%). The difference between Q1 and Q5 (0.042%) is significantly positive, indicating that insiders tend to sell less stocks when the price is far from the 52-week high. These results are robust to different insider trading measures. Additionally, high levels of insider sales (purchases) when the stock is near (far from) the 52-week high can be interpreted as insiders disposition effects, their tendency to realize gains when the price reaches its past high price. Overall, the results in Table 4 provide evidence that insiders tend to anchor on the nearness to the 52-week high. They are reluctant to buy the stock when the price is close to the 52-week high and to sell the stock when the price is far from the 52-week high. Moreover, their willingness to sell (purchase) stocks with high (low) 52-week high ratios is consistent with Grinblatt and Keloharju (2001) and Barberis and Xiong (2009), who support investors disposition effects. We confirm the results by estimating fixed-effects regressions of insider purchases and sales on the nearness to the 52-week high and other control variables used in existing studies: insi i,t = α + β 1 52wkHigh i,t 1 + β 2 ret i,t + β 3 mom i,t + β 4 lnmcap i,t 1 + β 5 bmratio i,t 1 + ε t (3) The dependent variable, insi i,t, is insider purchases or sales for stock i in month t, and the key explanatory variable is the 52-week ratio (52wkHigh i,t-1 ). If insiders exploit these price anomalies after observing them in the previous month, they will increase buying and selling with high and low 52-week high ratios, respectively. In contrast, decreasing insider purchases and 14

17 sales with high and low 52-week high ratios will indicate that insiders are subject to anchoring biases. Table 5 reports the regression results. Panel A summarizes the estimation results for insider purchases. When we use a turnover ratio for insider buying in column (1), the coefficient on the 52-week high ratio (-0.028) is negative, but it is not significant. When we use the proportion of the number of shares purchased by insiders to total transactions, the coefficient on the 52-week high ratio (-0.155) is significantly negative, and this result is robust in columns (3)- (5). These results suggest that insiders decrease their purchases when the price is close to the 52- week high. The results are consistent with our findings in Table 4, suggesting that insiders are subject to anchoring biases when they purchase stocks. Panel B reports the results for insider sales. It turns out that the coefficient on the 52-week high ratio is significantly positive no matter what insider trading measure we use for the analysis. When we use turnover (dinsi) and value turnover ratios (to_dvalue), the coefficients on the 52-week high ratio are and 0.134, which are significantly positive. The results confirm insiders biased sales, demonstrating that they decrease their sales when the stock price is far from the 52-week high. Furthermore, this positive coefficient on the 52-week high ratio provides additional evidence that insiders are subject to disposition effects, because they are willing to sell stocks for which prices are close to the 52- week highs. We also find that such anchoring behavior of insiders is differentiated from their strategies based on return momentum documented by Jegadeesh and Titman (1993). The coefficient on the past six-month return is in general negative with the dependent variables of insider purchases in Panel A, but it is not statistically significant. In Panel B, the coefficient on return momentum is significantly positive with the dependent variables of insider sales in 15

18 columns (1) and (2), indicating that insider sales tend to increase with high return momentum. These results are consistent with Rozeff and Zaman (1998) and Lakonishok and Lee (2001), who find that insiders are contrarians. 9 After controlling for this momentum effect, the coefficient on the 52-week high ratio is still significantly negative for insider buying and positive for insider selling, confirming that the effect of the 52-week high on insider trading is not the reflection of insiders behavior of exploiting other price anomalies such as momentum. A significantly negative coefficient on market capitalization in Panel A suggests that insiders prefer to buy small-cap stocks, but the coefficients in Panel B are mixed. Meanwhile, the coefficient on book-to-market is significantly positive for insider purchases and negative for insider sales, suggesting that insiders tend to buy high book-to-market (value) stocks and sell low book-to-market (glamour) stocks. These results are in line with Lakonishok and Lee (2001), who find that insiders prefer to buy small and value stocks and sell glamour stocks. Our findings suggest that the answer to the question of whether informed traders are subject to behavioral biases may vary depending on the type of informed traders. Lee and Piqueira (2015) find that short sellers, considered informed traders by many empirical studies, are able to exploit other investors anchoring biases associated with the 52-week and historical highs. Our results provide evidence contradicting their findings, implying that different types of informed traders have different degrees of sensitivity to heuristics in their trading behavior. In addition, the results provide important implications on the heterogeneity of informed traders in terms of their information advantage. Other types of informed traders, such as short sellers, could be better at evaluating a firm s fundamental value and market information compared to corporate 9 Given the previous finding of mean reversions in stock returns (Fama and French 1988; Poterba and Summers 1988), it is hard to conclude that these contrarian patterns are simply based on contrarian strategies or insiders ability to time the market, as documented in Seyhun (1986, 1992, 1998). We discuss the return predictability of insider trading in Section 6, but we do not further investigate insiders contrarian behavior in this study. 16

19 insiders. In this sense, our results are consistent with recent work by Khan and Lu (2013), which finds increased short selling prior to large insider sales. Overall, the results in Table 5 provide evidence supporting insiders anchoring biases. Corporate insiders underreact to good news when the stock price is close to the 52-week high and underreact to bad news when the stock price is far from the 52-week high. These findings raise the question of how such anchoring behavior affects the return predictability of insider trading documented by the prior literature. This will be further discussed in Section The Effect of Insiders Biased Trading on Return Predictability As mentioned above, our evidence of insiders anchoring biases seems to contradict existing studies, since the informativeness of insider trading and its return predictability are extensively documented in the literature. In this section, we revisit the return predictability of insider trading when the stock price is near and far from the 52-week high. Our examination on this issue will not only provide evidence on the information content of insider trading around the 52-week high but also offer insights into the heterogeneity of insiders in terms of the degree of their information advantage. According to prior studies, insiders are able to predict future abnormal returns. If the stock price is close to and far from the 52-week high, the return predictability of insider trading could be affected by insiders anchoring behavior. More specifically, the return predictability could be weakened due to insiders anchoring or even disappear. To examine this, we sort stocks into quintiles based on the nearness to the 52-week high in month t-1 and then, within each quintile, sort the stocks into five portfolios based on turnover ratios for insider purchases and sales in month t. For each of the 25 portfolios, we calculate 17

20 average raw and abnormal returns over the subsequent six months. The abnormal return is calculated as the difference between stock return and an equal-weighted market index return. 10 Table 6 reports average raw and abnormal returns for stocks for which prices are nearest their 52-week highs (Q5) and for which prices are farthest from the 52-week highs (Q1). We focus our analysis on these two portfolios, because insiders anchoring behavior is clearly shown in these two extreme portfolios. Panel A reports raw and abnormal returns for stocks for which prices are nearest the 52-week highs. The subsequent raw returns are in general positive, verifying that there is upward return momentum when the stock price is near its 52-week high. This also shows that insiders are able to predict future returns. On the other hand, average abnormal returns show less significance for most of the periods. When we compare subsequent returns for lowest and highest portfolios sorted on insider purchases (IP), it turns out that the highest IP portfolio has higher returns compared to the lowest IP portfolio across the subsequent two months. The average raw return for the lowest IP portfolio is 0.01 for month t+1, which is lower than the return for the highest IP portfolio (0.027) for the same month. The return difference between these two portfolios (0.017) is statistically significant for month t+1, and it becomes insignificant from month t+3 to month t+6. The patterns are similar in abnormal returns. The results show that high levels of insider purchases predict future returns better than low levels of insider purchases for the following month. Given our main finding of insiders anchoring behavior, we can conclude that insiders tend to be subject to anchoring biases when the price is close to the 52-week high, but insiders who are actively involved in buying are still able to predict returns. 10 We also run the tests using abnormal returns calculated as the difference between stock returns and valueweighted market index returns, but the results are unchanged. 18

21 Panel B reports subsequent returns for stocks for which prices are farthest from the 52- week highs (Q1). Subsequent returns for the highest insider sales (IS) portfolio are lower than those for the lowest IS portfolio in month t+1, and the returns are reversed in month t+3. The abnormal return for the lowest IS portfolio (0.02) is higher than the return for the highest IS portfolio (-0.005) in month t+1, and the return difference between the highest and lowest IS portfolios (Q5-Q1) (-0.025) is significantly negative. These results confirm the return predictability of insider sales, showing that the highest IS portfolio is followed by lower subsequent returns than the lowest IS portfolio. Overall, the results verify the return predictability of insider trading when the stock price is near and far from the 52-week high. Although insiders are subject to anchoring biases, high levels of insider purchases and sales are still able to predict subsequent returns. The results suggest that insiders are heterogeneous. While some of them anchor on the nearness to the 52- week high, some of them trade based on their information advantage and predict future returns. These results also imply that the degree of their information advantage varies among insiders, leading to different trading behavior based on the nearness of the price to the 52-week high. 7. Additional Tests 7.1. Are Other Types of Insiders Subject to Anchoring Biases? So far we have focused on insider trading made by management including directors, committees, officers, etc. The reason behind this is that other types of insiders, such as large shareholders or controlling person, might not be relatively informed compared to management. If this is the case, they could be more subject to anchoring biases associated with the nearness to the 52-week high compared to management. On the other hand, given the prior finding that their 19

22 trades could be initiated for some other motives, we may not be able to find behavioral biases in their trading. Motivated by this, we repeat the analysis in Section 5 with a focus on large shareholders. They are classified as direct and indirect shareholders and beneficial owners who own more than 10% of shares. However, we should note that, as Lakonishok and Lee (2001) pointed out, the definition of large shareholders is ambiguous in terms of information advantage. Therefore, it is hard to predict that they are more or less biased on the nearness to the 52-week high. Table 7 reports regression results of large shareholders buying and selling on the 52- week high ratio and other control variables. When we regress insider purchases in Panel A, the coefficients on the 52-week high ratio are mixed depending on the choice of a dependent variable, and they are not statistically significant. When we investigate large shareholders sales in Panel B, the coefficient on the 52-week high ratio is positive across columns (1)-(5), but it is statistically significant only when we use the value turnover ratio as a dependent variable in column (2). The results suggest that large shareholders tend to decrease their selling when the stock price is far from the 52-week high, but the evidence is weak compared to our findings in Section 5. We also perform the same analysis on transactions made by others, such as controllers, founders, former officers, etc. Unreported results do not show that these insiders are subject to anchoring biases associated with the nearness to the 52-week high. Overall, we do not find evidence of anchoring behavior of other types of insiders, such as large shareholders and controllers. Nonetheless, given the previous finding that they are less informed than management, our evidence confirms that informativeness and behavioral biases are not mutually exclusive. 20

23 7.2. Does the 52-Week Low Lead to Behavioral Biases? While the effect of the 52-week high is well documented, the effect of the 52-week low on investors trading behavior is mixed in the literature. George and Hwang (2004) fail to find evidence on anchoring biases associated with the 52-week low, while Huddart et al. (2009) show that trading volume is high when the stock price is below the 52-week low. Following these studies, we examine if the 52-week low has the same impact as the 52-week high on insiders trading behavior. If insiders are subject to anchoring biases based on the nearness to the 52-week low, they are reluctant to sell (buy) stocks when the stock price is near (far from) the 52-week low. Our examination on this issue will verify that the evidence of insiders biased trading is robust to other price extremes. For this analysis, we calculate the 52-week low ratio as the lowest price achieved within the past 52 weeks divided by the stock price. 11 Thus, a high level of the 52-week low ratio suggests that the stock price is close to its 52-week low, while a low level of the 52-week low ratio suggests that the price is far from the 52-week low. We replace the 52-week high ratio in Equation (3) with this 52-week low ratio and re-estimate the regression model. If insiders are subject to anchoring biases, we can expect to find a negative (positive) association between insider sales (purchases) and the 52-week low ratio. Table 8 reports the estimation results. When we regress insider purchases on the 52-week low ratio in Panel A, we find that the coefficient on the 52-week low ratio is positive and is statistically significant in columns (3)-(5). These results suggest that insider tend to decrease their purchases when the price is far from the 52-week low. On the other hand, when we use insider sales as dependent variables in Panel B, the coefficient on the 52-week low ratio is 11 This is the reciprocal of the 52-week high ratio which is calculated as monthly stock price divided by the highest price achieved within the past 52 weeks. 21

24 significantly negative across all the columns, which suggests that insiders tend to decrease their sales when the price is near the 52-week low. The overall results support that insiders anchor on the nearness to the 52-week low. Moreover, their willingness to buy stocks when the price is close to the 52-week low and to sell stocks when the price is far from the 52-week low suggests the disposition effects of insiders, which are consistent with what we found in Section 5. Linking these results to our main findings of insiders biased trading based on the distance to the 52-week high, we can conclude that insiders are susceptible to behavioral biases associated with the past price extremes, not only the 52-week high but also the 52-week low. To sum up, we find that insiders tend to anchor on the nearness to the 52-week low. Also, their tendency to buy (sell) stocks when the price is near (far from) the 52-week low indicates that they are subject to disposition effects. 8. Conclusions We investigate how the nearness of the stock price to its 52-week high affects insiders trading activities. We first confirm through return analysis that investors underreaction exists when the price is near and far from the 52-week high. Investors are reluctant to buy stocks when the stock price is close to the 52-week high and to sell stocks when the price is far from the 52- week high. Insiders, as informed traders, may make profits by exploiting these anchoring biases, or they may anchor on the nearness to the 52-week high like other investors. We find that insiders are subject to anchoring biases based on the nearness to the 52-week high. They tend to decrease their purchases and sales when the stock price is near and far from the 52-week high, respectively. In addition, we show the evidence of insiders disposition effect by finding that 22

25 they are willing to buy stocks when the price is far from the 52-week high and to sell stocks when the price is close to the 52-week high. These results provide evidence against the prevailing view that informed traders are not subject to behavioral biases. However, our findings of the return predictability of insider trading around the 52-week high indicate that insiders are heterogeneous. While some of them are susceptible to behavioral biases associated with the 52-week high, some of them tend to trade based on their information advantage and are still able to predict future returns. Since our main analysis focuses on trades made by one particular group of insiders, management, we perform robustness tests using purchases and sales by other types of insiders such as large shareholders but do not find evidence of their biased behavior. When we examine insiders behavioral biases using the 52-week low, we find evidence supporting anchoring behavior and disposition effects associated with the nearness to the 52-week low. All these results are robust to various insider trading measures and firm-characteristic variables such as return momentum, market capitalization, and book-to-market. Overall, our results show that informed traders can be susceptible to behavioral biases like other investors. This implies that informed traders are heterogeneous in terms of their information advantage and that the information content of informed trading may also vary. Our findings make important contributions to the existing literature of behavioral finance and informed trading, by showing that behavioral biases are compatible with sophistication. 23

26 References Agrawal, A., Cooper, T., Insider trading before accounting scandals. Journal of Corporate Finance 34, Alldredge, D., Cicero, D.C., Attentive insider trading. Journal of Financial Economics 115, Amihud, Y., Illiquidity and stock returns: Cross-section and time-series effects. Journal of Financial Markets 5, Baker, M., Pan, X., Wurgler, J., The effect of reference point prices on mergers and acquisitions. Journal of Financial Economics 106, Barberis, N., Xiong, W., What drives the disposition effect? An analysis of a long-standing preference-based explanation. Journal of Finance 64, Beschwitz, B., Massa, M., Biased shorts: stock market implications of short sellers' disposition effect. Working paper, INSEAD. Campbell, S., Sharpe, S., Anchoring bias in consensus forecasts and its effect on market prices. Journal of Financial and Quantitative Analysis 44, Cen, L., Gilles H., Wei, K., The role of anchoring bias in the equity market: Evidence from analysts' earnings forecasts and stock returns. Journal of Financial and Quantitative Analysis 48, Cohen, L., Malloy, C., Pomorski, L., Decoding inside information. Journal of Finance 67, Degeorge, F., Patel, J., Zeckhauser, R., Earnings management to exceed thresholds. Journal of Business 72, Driessen, J., Lin, T., Hemert, O.V., How the 52-week high and low affect option-implied volatilities and stock return moments. Review of Finance 17, Du, D., The 52-week high and momentum investing in international stock indexes. Quarterly Review of Economics and Finance 48, Fama, E.F., French, K.R., Permanent and temporary components of stock prices. Journal of Political Economy 96, Fama, E., MacBeth, J., Risk, return and equilibrium: Empirical tests. Journal of Political Economy 81,