4. P ROYECCIÓN DIDÁCTICA
4.2. C ONTEXTUALIZACIÓN
Information hierarchy. Several papers in the insider trading literature investigate the
“in-formation hierarchy hypothesis” (Seyhun, 1986), which holds that trades by those insiders who have more information have a higher price impact.139 We should therefore expect that insiders who are more informed and who have more information to hide will engage more in stealth trading because they face more adverse selection problems.
Hypothesis 3 (Information hierarchy): Stealth trading is more likely for insiders who are more informed.
Based on our data we can distinguish between the CEO, officers and directors other than the CEO, directors who are not officers, the chairman of the board, and other insiders who hold none of these roles (mostly large shareholders).
Insert Table 5.8 here
Before we interpret the multivariate regression results from Table 5.7 in relation to the infor-mation hierarchy hypothesis, we discuss the univariate results presented in Table 5.8. The univariate results show that CEOs, chairmen of the board, and other insiders use stealth trad-ing most frequently, whereas other officers and directors use stealth tradtrad-ing less than all other categories of insiders. The findings regarding CEOs and chairmen of the board support the information hierarchy hypothesis, whereas the findings regarding other insiders contradict it.
Other insiders are mostly large shareholders, who own more than 10% of the shares of the company but do not have a formal role in the firm. These insiders trade relatively large vol-umes (mean volume per transaction is $416,100) compared to CEOs ($294,000), officers ($262,100) and directors ($279,300). Only chairmen trade on average larger volumes
139 In the literature on insider trading, the information hierarchy hypothesis holds that trades by insiders who are closer to the firm have a larger information content. The evidence on this hypothesis is mixed. Seyhun (1986) shows that the directors and officers trade on more valuable information than other insiders. Lin and Howe (1990) show that trades by the CEO and the officers and directors of the firm have a higher informa-tion content than those of unaffiliated shareholders. Fidrmuc, Goergen, and Renneboog (2006) find no evi-dence for the information hierarchy hypothesis.
($420,500). Measuring transaction size by Stake reveals a similar picture. In this case, other insiders actually execute the largest transactions on average. Given that other insiders trade relatively large volumes, possibly because they hold large stakes in the firm, they have strong incentives to split trades. The final verdict on the information hierarchy hypothesis should therefore rely on multivariate regressions that also control for transaction size.
The regression analysis in Table 5.7 includes dummy variables for all categories of insiders except directors, so the coefficients for the four remaining insider groups have to be inter-preted relative to this group. The multivariate results are consistent with the univariate analy-sis. The coefficients for OtherInsider are positive and significantly larger than the coefficients of all other insider groups across all specifications, even though we control for StakeDecile.
This result is inconsistent with the information hierarchy hypothesis unless we assume that officers and directors are less informed than other insiders, which we find implausible. Across all specifications, chairmen and CEOs do more stealth trading than other officers and direc-tors. However, this difference is only significant for the large sample (model (3) and (4)). The finding that CEOs and chairmen are more inclined to stealth trading than officers and direc-tors are consistent with the information hierarchy hypothesis. Nonetheless, the overall support for this hypothesis is only mixed. Interestingly, the results for chairmen and for other insiders are stable across specifications and therefore independent of how we control for liquidity. By contrast, the coefficient for CEO is much larger and that for officers is much smaller in the larger sample compared to the NYSE-subsample.
Direction of trade. The insider trading literature has shown that purchases have a larger in-formation content compared to sales, probably because sales are more likely to be motivated by liquidity considerations, whereas purchases are more likely to be motivated by information
advantages.140 Our hypothesis is that stealth trading is a strategy to hide trades with a larger information content, so that stealth trading should be related to the direction of trades.
Hypothesis 4 (Direction of trade): Stealth trading is more frequent for purchases than for sales.
The univariate results in Table 5.5 suggest exactly the opposite of Hypothesis 4, namely that stealth trading is more likely for sales than for purchases. Only 25.1% of all aggregated stealth trades are purchases, whereas 36.9% of all non-stealth trades are purchases. However, this could be explained by the fact that sales are on average much larger (across all insider groups) than purchases and therefore offer more scope for stealth trading. As in the case of the infor-mation hierarchy hypothesis, we have to control for trade size to test Hypothesis 4.
Table 5.7 shows that the coefficient of PurchaseDummy has the predicted sign in all specifi-cations but is statistically significant only in models (3) and (4), where we can use the large sample. In the smaller sample for which also the spread variables are available, the effect is statistically insignificant. Hence, when we control for trade, firm and insider characteristics, stealth trading is more frequent for purchases than for sales. The fact that PurchaseDummy is insignificant in all models that use the small sample suggests that the sample size is the reason for the insignificant coefficient and not the fact that we control for the spread in models (1) and (2).
Asymmetric information. Stealth trading should be more attractive if the general scope for
informed trading is larger. This will be the case if there is more asymmetric information, for example, in companies that are more opaque, in companies with more firm-specific risk, and
140 The first to make this observation was Rogoff (1964). See Lakonishok and Lee (2001), Jeng, Metrick, and Zeckhauser (2003) or Fidrmuc, Goergen, and Renneboog (2006) for more recent analyses.
at times before an earnings announcement, whereas it will be less likely immediately after an earnings announcement.141
Hypothesis 5 (Asymmetric information): Stealth trading is: (1) more likely if there is more
asymmetric information and if the company is more opaque; (2) more likely in stocks with more firm-specific risk; (3) more likely before and less likely after earnings announcements.
We investigate part (1) of Hypothesis 5 by looking at reporting periodicity and research and development expenditures as measures of firms’ transparency or opacity. We identify firms with higher quality accounts with QuarterlyReport, which is one for firms that file quarterly reports and zero otherwise. R&D is defined as research and development expenditures scaled by total assets. R&D is set to zero for those firms, where Compustat does not report any re-search and development expenditures. We analyze part (2) by using Volatility, defined as the annualized standard deviation of daily stock returns over the calendar month preceding the transaction. We use this as a measure of firm-specific risk.142 Part (3) of Hypothesis 5 is ana-lyzed by looking at earnings announcements reported by Compustat. We define two dummy variables BeforeEarnAnnounce and AfterEarnAnnounce, which equal one for a period of two weeks (14 days) before, respectively after an earnings announcement.
The impact of QuarterlyReport is as predicted and always significant at the 5% level, al-though the vast majority of our sample firms report earnings quarterly and therefore the em-pirical relevance is only limited (the mean of QuarterlyReport is 0.999, see Table 5.3). The results for R&D are in line with Hypothesis 5. The coefficient for R&D is positive and sig-nificant at the 0.1% level across all models, which implies that insiders from firms with higher R&D expenditures use more stealth trading. A one standard deviation increase of R&D
141 Aboody and Lev (2000) show that insider gains are larger for R&D-intensive firms and interpret R&D as a proxy for asymmetric information. Fidrmuc, Goergen, and Renneboog (2006) and Betzer and Theissen (2009) investigate the impact of news announcements on insider trading.
142 Results do not change materially if we use the standard deviation of daily excess returns from a market model as a proxy for firm specific risk, where we use the CRSP-value weighted index over the preceding calendar year as a measure of market risk.