Attentive insider trading

Attentive insider trading

Author's Accepted Manuscript Attentive insider trading Dallin M. Alldredge, David C. Cicero PII: DOI: Reference: www.elsevier.com/locate/jfec S0304...

542KB Sizes 16 Downloads 120 Views

Author's Accepted Manuscript

Attentive insider trading Dallin M. Alldredge, David C. Cicero

PII: DOI: Reference:

www.elsevier.com/locate/jfec S0304-405X(14)00189-5 http://dx.doi.org/10.1016/j.jfineco.2014.09.005 FINEC2469

To appear in:

Journal of Financial Economics

Cite this article as: Dallin M. Alldredge, David C. Cicero, Attentive insider trading, Journal of Financial Economics, http://dx.doi.org/10.1016/j.jfineco.2014.09.005 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Attentive insider trading

Dallin M. Alldredge University of Tennessee Phone: (865) 974-8253 Email: [email protected] David C. Cicero† University of Alabama Phone: (205)348-9791 Email: [email protected]

Abstract

We provide evidence that some profitable insider stock selling is motivated by public information. At firms that disclose having concentrated sales relationships, insiders appear to sell their own stock profitably based on public information about their principal customers. Supplier insiders also sell more stock when public information about their customers’ recent returns and earnings surprises suggests they will earn larger profits. These results are stronger when outside investor attention could be lower. Outside of this setting, insiders engage in a higher proportion of routine sales and their sales are less profitable. We do not find similar patterns for insider purchases. JEL classification: K22, G14, G30 Keywords: Insider trading, Investor attention, Supply chain †

We thank a very helpful anonymous referee, Anup Agrawal, Lee Biggerstaff, Douglas Cook, Phillip Daves, Jesse Ellis, Joan Heminway, Ashrafee Hossain (Financial Management Association 2013 discussant), Andy Puckett, Bill Schwert (the editor) and seminar participants at Georgia State University and the University of Tennessee for helpful comments.

1

1. Introduction Corporate insiders’ trades predict future abnormal returns.1 If stock prices reflect all publicly available information, this suggests that insiders generally do not respect the legal prohibitions on using inside information to make trading decisions.2 We evaluate whether an alternative explanation can account for the abnormal returns earned by insiders. We pose a hypothesis of attentive insider trading informed by public information. That is, we hypothesize that corporate insiders pay close attention to public information that is relevant to their firms and earn profits by trading when outside investors are relatively inattentive. We explore insider trading in a setting where attentive trading could be distinguished from illegal trading. It is where firms have disclosed that other individual public companies account for a large fraction of their sales (i.e., where suppliers are economically linked to their principal customers). Supplier insiders’ opportunities to trade on public information should be enhanced in this setting. Cohen and Frazzini (2008) and Menzly and Ozbas (2010) argue that outside investors are limited in their ability to understand the full impact of public information across firms and industries, and they show that this leads to return predictability across economically linked firms.3 Cohen and Frazzini (2008), in particular, show that investor inattention allows lagged abnormal returns to principal customers to predict their suppliers’ 1

Examples of work in this area include Lorie and Niederhoffer (1968); Jaffe (1974); Finnerty (1976); Seyhun (1986, 1992, 1998); Bettis, Vickrey, and Vickery (1997); Lakonishok and Lee (2001); Jeng, Metrick, and Zeckhauser (2003); Agrawal and Cooper (2014); and Agrawal and Nasser (2012). Historically, most of the robust evidence of returns following insider transactions was based on stock purchases. More recently, Cohen, Malloy and Pomorski (2012) and Cicero and Wintoki (2014) show that applying simple and intuitive screens on the trading data results in strong evidence of informed stock sales, too. 2 In the United States, as in many other countries, it is illegal to trade securities based on private information. Securities and Exchange Act of 1934, Sections 16(b) and 10(b), and the related Security and Exchange Commission rules and case law. 3 Numerous researchers argue for both the existence and rationality of limited investor attention to information relevant to asset prices (Hong, Torous, and Valkanov, 2007; Bacchetta and van Wincoop, 2010; and Peng and Xiong, 2006). Huang and Liu (2007), in particular, show that investor inattention can lead to cross-sectional return predictability. For the social psychology foundations of limited attention theories, see Kahneman (1973) and Fiske and Taylor (1991).

2

returns. In this paper, we test whether customers’ lagged returns and their information disclosures to the market (which are public information at the time of trading) explain supplier insiders’ trading decisions and the abnormal returns that they earn. Although this is not the only setting in which insiders could trade profitably on public information, it is one in which opportunities should be particularly acute. Therefore, the evidence of attentive trading could contrast more starkly compared with trading outside of this context. Our hypothesis is grounded in an expectation that corporate insiders are among the most attentive traders of their own stocks. They have undiversified economic stakes in their firms, including both their current securities holdings and their future income, giving them high incentives to monitor developments that affect their firms’ prospects. In addition, the nature of their jobs is to be informed about market developments that affect their firms as they work to maximize firm value. Corporate insiders at economically linked suppliers could, therefore, quickly recognize profitable trading opportunities when they observe public information about their large customers such as recent stock returns, earnings announcements, and corporate press releases.4 As a result, insiders at these firms could have increased opportunities to trade profitably on public information beyond that which is available to insiders at other firms. Taken as a whole, our results suggest that some profitable insider stock sales are motivated by public information. We first show that insiders’ sales are more profitable at economically linked suppliers, suggesting that these insiders possess some informational advantage. Interestingly, though, their purchases, although profitable in this setting, are followed by similar return patterns when strong supply chain links are not reported. The asymmetry of this 4

Throughout the paper we refer to stock sales as profitable if they precede negative abnormal returns, as the insider avoids losses by selling.

3

finding supports a conclusion that insiders’ sales are at times motivated by public information. To see why, it is useful to consider the different incentives insiders face when they sell stock versus when they purchase it. As argued by prior researchers, there is significantly more litigation risk is associated with selling stock. If an insider withholds negative information when he trades, other investors would clearly be harmed if they purchase at an inflated price. Selling by insiders when their stock is overvalued could also be used as evidence in a suit claiming fraudulent financial reporting. However, if an insider withholds positive information and purchases stock, the only harm is to other investors who could have sold, and the insider would have missed out on potential gains if the good news had already been released. But as others point out, it is less likely that shareholders will bring a successful derivative lawsuit against insiders when their only losses are best described as opportunity costs.5 Given the asymmetric litigation risk, a reasonable interpretation of our results is that the higher levels of profitable insider selling in this context is driven by the greater abundance of opportunities to trade profitably on public information. The comparably strong profitability of insider purchases across settings suggests that any increased opportunities to trade on public information in the supply chain context is not incrementally useful when insiders are less deterred from exploiting private information.6 To conduct our analysis, we collect a sample of 1,858 firms (6,939 firm-years) that report the existence of large principal customers during the period 1986–2010 (economically linked

5

See, for instance, Skinner (1994), Brochet (2010), and Chen, Martin and Wang (2012). Other research suggests that attentive institutional investors exploit information spillovers when investing in related firms (see, for example, Cohen and Frazzini, 2008; Menzly and Ozbas, 2010; and Huang and Kale, 2012). However, because the mutual fund holdings data are reported at quarterly frequency, they are generally neither able to evaluate the profitability of specific investments nor establish the direction of causation between trading and returns.

6

4

suppliers). We compare insider trading activity at these firms across years when they do and do not report these strong economic links. We also contrast the returns to insider trading at the economically linked firms to those at 5,118 other firms that never report these relations (neverlinked firms). Our focus is on trades that are nonroutine, because Cohen, Malloy, and Pomorski (2012) show that these trades are most likely to be motivated by an informational advantage. We start by evaluating the profitability of insiders’ trades. Abnormal returns following nonroutine insider sale months are significantly larger at economically linked suppliers. For example, NYSE size decile–adjusted one-month cumulative abnormal returns (CARs) following insider sales at economically linked suppliers are -0.67% compared with -0.16% at linked suppliers during non-linked years and -0.32% at never-linked firms. The significance of this result is confirmed in multivariate tests controlling for market returns, firm size, book-to-market equity value, and stock return momentum. In addition, a larger proportion of trade months are profitable when strong economic links exist. In contrast, insiders’ stock purchases are profitable, on average, regardless of the existence of a strong supplier-customer relationship. For example, insiders’ purchases are followed by monthly abnormal returns of 0.92% when an economic link is present and 0.80% when it is not, and these two values are not significantly different. We provide a variety of additional tests to help further identify whether insiders’ trades in this setting are motivated by private or public information. For one, we show that supplier insiders do not sell stock profitably in the first year of a concentrated supplier-customer relationship, which could precede their disclosure of the relationship to the market. If supplier insiders traded on private information about their customers, similar or even larger abnormal returns would be expected during these years because outside investors are less likely to understand the impact of information across economic links before they are disclosed. The 5

subsequent analyses focus on trades by supplier insiders once the relations with principal customers have been disclosed. We find that the abnormal returns following these trades are a function of the lagged returns to the large principal customers. Supplier insiders’ sales earn significant abnormal profits on average only when the lagged abnormal returns of their principal customers were negative. In addition, the abnormal profits earned on sales are a function of the magnitude of the customers’ lagged returns. Supplier insiders also earn greater profits when the market could be less attentive to the economic link between suppliers and principal customers. For example, the abnormal returns following their sales are more pronounced when fewer analysts cover the suppliers. Their sales also become less profitable as the duration of the supplier-customer relationship grows. This is to be expected if insiders are trading on public information because the market should learn about the cross-firm return predictability over time and the faster recognition by outside investors should limit insiders’ profitable trading opportunities.7 The next issue we focus on is whether supplier insiders trade more when they have opportunities to profit on public information. We find that the overall level of nonroutine insider trading is higher at economically linked suppliers than at other firms, suggesting that more opportunities exist to trade profitably in this setting. It is reasonable to expect attentive insiders to notice market signals about their customers’ prospects and to react when they think the market does not fully appreciate the impact of these developments on their own future performance. We find that insiders at economically linked suppliers are more likely to trade when publicly available information about their customers indicates their trades will be profitable. In the first

7

None of the results discussed in this paragraph holds for purchases, with the exception of the fact that supplier insider purchases are profitable only when lagged customer returns are positive.

6

set of tests along these lines, we show that supplier insiders are more likely to sell (purchase) stock when the lagged returns to their principal customers are lower (higher), a situation that predicts they will earn abnormal returns (Cohen and Frazzini, 2008). Moreover, a strong relation exists between supplier insiders’ stock sales and their customers’ specific communications to the market. Supplier insiders trade their own stock around the time their principal customers announce earnings, and their trades are in the direction one would expect, given the impact of the unexpected earnings component on their own firm value. The level of trading is also greater when other investors could be less attentive, in particular when the earnings are released on Fridays. Other investors could be less attentive at the end of the week and, therefore, slower to incorporate new information into the suppliers’ stock prices (DellaVigna and Pollet, 2009). To get a sense of the economic significance of these results, we find increases in the odds of a supplier insider sale of 425% (a 6.50% increase relative to the unconditional odds of a sale of 1.53%) following a large negative earnings surprise announced by a principal customer on a Friday. In addition, given a sale, supplier insiders sell an additional 19,190 shares under these circumstances. We conclude with robustness tests designed to evaluate alternative explanations for our findings, including the possibility that our results are driven by trading on private information. We find that the strong relation between supplier insider selling and the information content of their customers’ earnings announcements does not hold in the two days immediately preceding the announcements. However, for purchases, we find similar trading patterns in this period as we find for the days following customer earnings announcements. A final alternative we consider is that insiders use public announcements to camouflage trades based on other, unrelated, private information. We analyze the trades of a subset of supplier insiders who were already trading 7

prior to their customers’ earnings announcements and in the direction that would be expected given the content of the pending release. For the case of stock sales, we continue to find that trading immediately following the customers’ announcements is related to its content. This would not be expected if they are trading on this particular information ahead of time. Also, given that they are already trading, it seems unlikely that they are waiting for a public news release to camouflage trading motivated by other private information.8 This relation does not hold for purchases. As a whole, our analysis is consistent with a conclusion that some profitable insider selling in this setting is motivated by public information. This conclusion is supported in the case of supplier insider sales because of the robust evidence that public information explains the returns that are earned, that selling patterns following customer announcements are a function of the new information conveyed, and that the level of nonroutine trading and returns earned by supplier insiders are not matched outside of this setting. However, the evidence does not support the same conclusion for stock purchases. Less evidence exists regarding the relation between public information and the returns following purchases, insider selling patterns are similar immediately before and after customer earnings announcements, and insiders earn similar returns outside of this setting. It is beyond the scope of this paper to estimate the relative levels of illegal trading on private information versus attentive trading on public information in our markets. However, we

8

We cannot be certain whether insiders are ever motivated solely by public information when they trade, or if their actions reflect consideration of a mosaic of public and private signals. Assuming they have access to both sets of signals, it seems unlikely that insiders would completely ignore one set when they trade. Even if one has the best of intentions, certainly insiders’ expectations of the impact of new public information on their firms’ values would be colored by the private information they possess. Regardless, our analysis suggests that it is the addition of public information to an insiders’ information set that often leads them to trade.

8

offer some thoughts on this question in Section 4 with a focus on insights from prior research that could be useful for calibrating expectations about the prevalence of each. The remainder of this paper proceeds as follows. In Section 2 we discuss the data and methodologies used in this study. Section 3 contains the empirical results. Finally, in Section 4 we conclude with a summary of the results and discuss the possible extent of attentive insider trading in our markets.

2. Data and methodology The data for our analyses are obtained from several sources. We first identify firms that sell a large fraction of their output to individual principal customers. Under US securities laws, public companies are required to report annually any customer that accounts for more than 10% of their annual sales, and these disclosures are included in the Compustat Customer Segments database. Extracting the supplier-customer relationships from this database required a significant amount of hand-matching. The reporting companies often abbreviate the names of their large customers, so each entry had to be compared with the full company name listed in Compustat. We matched the customer names conservatively, and we were careful to check company websites and the Business Week company profiles to exclude questionable matches. This analysis also requires insider trading data. These data are retrieved from Table 1 of the Thomson Reuters insider database, which reports stock transactions by insiders at publicly traded US firms.9 We conduct our main tests on trades by the broad cross section of insiders that

9

The Thomson Reuters Financial Network Insider Filing Data provides information on executives’ stock and derivative transactions. These transactions must be reported to the Securities and Exchange Commission on Forms 3, 4, 5, and 144. Insiders are required to file Form 3 to report initial beneficial ownership of shares, Form 4 to report changes in beneficial holdings, Form 5 to report annual changes in beneficial ownership, and Form 144 to declare

9

must report their transactions, and we show selected results for subsets of top officers or directors. Transactions are included in our analysis if Thomson Reuters has verified their accuracy and given them a cleanse code of R, H, L, C, or Y. We exclude trades that can be characterized as routine because these trades are unlikely to be motivated by information. Cohen, Malloy, and Pomorski (2012) characterize insiders as routine traders if they trade in the same calendar month in three consecutive years, and they show that trades by these insiders do not predict abnormal returns. We follow their definition for routine traders and exclude all trades made by these individuals. The remaining trades covered in our analysis are, therefore, reasonably characterized as nonroutine. We also exclude insider trades related to options transactions, which leaves us with a sample of nonroutine insider open market sales and purchases. For most of our tests, the trading observation is measured at the monthly level. A sale month (purchase month) is defined as a calendar month in which an insider reported trading at least once and where the trades resulted in a net increase (decrease) in ownership. Evaluating abnormal returns requires stock market returns and firm characteristics. Balance sheet and income statement data are collected from Compustat, and stock returns are collected from the Center for Research in Security Prices (CRSP). We focus our study on common stocks only (CRSP share codes 10 and 11). We eliminate illiquid stocks by dropping those with prices below $5 or market capitalizations below $100 million. Further, to eliminate the effect of outliers, we winsorize all variables at the 99% and 1% levels.

their intention to sell restricted shares. Stock transactions are reported in Table 1 of these forms. Officers, directors, and beneficial owners of more than 10% of a class of a company's equity securities are required to report their transactions.

10

Combining the economically linked suppliers, the insider trading data, stock returns, and firm characteristics resulted in a final sample of 1,858 linked firms and 17,180 unique trade months at those firms (11,805 net sale months and 5,375 net purchase months). Our control observations consist of all other firms for which we have the available CRSP, Compustat, and insider trading data but that do not report the existence of principal customers. The control sample includes 6,330 never-linked firms with 128,942 unique trade months (82,905 net sale months and 46,037 net purchase months). Our main empirical strategy is to test for differences in abnormal stock returns following trades at economically linked suppliers versus those at other firms. To establish the robustness of our results, we calculate abnormal returns in a couple of ways. The first and simplest method consists of comparing stock returns over the calendar month subsequent to trading months with the calendar month returns on the appropriate size decile portfolio of firms based on NYSE size breakpoints (size decile portfolio–adjusted CAR). This method is useful because it accounts for market-related risk factors that affect firms of similar size during the same one-month period. We also determine abnormal returns in a regression framework. We regress excess one month returns following trade months onto the equal-weighted market return and other variables that account for additional risk factors, including the firm’s stock market value, book-to-market value of equity, and prior supplier stock returns.10 This method mimics that of Cohen, Malloy, and Pomorski (2012). We also introduce a variety of explanatory variables derived from the lagged three-month abnormal returns to the suppliers’ principal customers. In the case that a supplier has more than one principal customer, we use the equal weighted average of the 10

Excess returns are defined as the monthly return to a firm’s stock minus the one-month risk-free rate as reported on Ken French’s website: http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html. We confirm that all of our results are robust to using total return and size decile–adjusted CARs as the dependent variable.

11

principal customers’ returns in our regression analysis. We are careful to control for the lagged abnormal return to the supplier firm over the same horizon to account for any industry-wide effect that could codetermine these returns. The second part of our analysis focuses on whether public information about principal customers predicts the trading activity of supplier insiders. We conduct limited dependent variable regressions predicting trading at the economically linked suppliers as a function of their principal customers’ past abnormal returns and unexpected earnings. The analyst forecasts and earnings announcements for this part of our analysis are obtained from the Institutional Brokers’ Estimate System (I/B/E/S) database. We use a normalized measure of quarterly earnings surprise from DellaVigna and Pollet (2009), constructed as

SUEt ,k

actual earningst ,k  expected earningst ,k

, (1) pricet ,k where SUEt,k is the standardized unexpected earnings (SUE) announced by company k for quarter t, actual earningst,k is the actual earnings per share for company k in quarter t, and expected earningst,k is the corresponding median of all analyst earnings per share forecasts issued closest in time to the earnings announcement date but not more than 90 days prior to the fiscal period end. The earnings surprise is scaled by Pricet,k, which is the stock price of company k five days prior to the earnings announcement date in quarter t. For our analysis, we calculate earnings surprises for our firms’ principal customers, as well as for the other firms in the same industry [Standard industrial classification (SIC) 4] as those customers. In unusual cases in which suppliers report more than one principal customer and multiple customers report their earnings on the same day (3.9% of observations), we take the average of the firms’ SUE.

12

Table 1 contains summary statistics for our sample. Economically linked suppliers in the sample are slightly smaller than non-linked firms. This makes sense because it is easier for smaller firms to have one customer account for 10% or more of sales. The mean market capitalization of linked firms with insider sales is $4.4 billion, and the mean market capitalization of non-linked firms with insider sales is $4.8 billion. The book-to-market equity value of linked suppliers is marginally higher than for the control firms. These differences highlight the importance of controlling for these characteristics in a regression format. Linked suppliers with insider sales have slightly higher nonroutine traders and trade days per firmmonth, which suggests more profitable trading opportunities could be found at these firms. In contrast, this relation is reversed for insider purchases. Insert Table 1 near here Fig. 1 shows the duration of the supplier-customer relationships in our sample. The plot illustrates a general decrease in the number of links as the duration increases. The maximum link duration is 25 years, and the average duration is 2.66 years. For most of our tests, we include insider trading months in the analysis only if they occur after the supplier-customer link has been disclosed, such that the link can be characterized as public information. For comparison purposes, we also present one regression of returns following trades that occur during the first year of the relationship, which is before the public can be expected to know of the link, to help distinguish trading based on private or public information. Insert Fig. 1 near here The distribution of linked and non-linked firms across the Fama and French 17 industry classifications is reported in Table 2. The economically linked suppliers are disproportionately 13

concentrated in the machinery and business equipment industry and the drugs, soap, perfumes, and tobacco industry. The machinery and business equipment suppliers are typically producers of parts for machinery and computers (e.g., SanDisk Corp and Integrated Device Technology), and the drugs, soap, perfumes, and tobacco suppliers are predominately pharmaceutical companies (e.g., Alnylam Pharmaceuticals Inc. and Alexza Pharmaceuticals Inc.). As would be expected, relatively few supplier firms are retail stores and financial institutions. The main results of the paper are robust to the exclusion of all financial firms, but for completeness we include them in the analysis. Insert Table 2 near here

3. Empirical results We analyze insider trading in a setting where an increased potential exists for insiders to earn returns by trading on public information. The first part of our analysis focuses on the determinants of returns following insider transactions at firms that have large principal customers. The second part of this analysis focuses on the conditions that lead insiders at these economically linked suppliers to trade.

3.1. Returns following insider trades at economically linked suppliers We first examine the relation between strong supplier-customer relationships and the returns following insider trading months. We compare insider trading returns at economically linked suppliers with those at other non-linked firms. We also make comparisons with insider returns at two mutually exclusive groups within the non-linked firms: the non-linked years for firms that report economic links at some point (linked suppliers in non-linked years), and other firms that never report links (never-linked firms).

14

Table 3 presents NYSE size decile-adjusted one-month CARs following insider trades. The size-adjusted CARs following insider sales are reported in Panel A. Overall the results indicate that the returns earned by insiders on their stock sales are significant when economic links are present. The CARs following insider sales at economically linked suppliers on average are -0.671% per month. In contrast, evidence of profitable insider sales is not very strong outside of this context. On average, insiders at these firms earn only 29 basis points (-0.29% CAR) of abnormal return over the month following their nonroutine sales. Breaking the non-linked firms down further, we find abnormal returns of -0.158% per month at linked firms during non-linked years and -0.322% per month at never-linked firms. The differences from the average returns earned by linked suppliers of 0.513% per month (T-statistic = 3.74) and 0.349% per month (Tstatistic = 3.42), respectively, are highly significant. Insert Table 3 near here The contrast in returns is starkest with top-level officers (chief executive officers, chief financial officers, and chief operating officers), who are most likely to be focused on information affecting their firms’ future prospects. Top-level officers are also under the spotlight of regulators and the press, which could motivate them to avoid trading on private information. Size-adjusted CARs for top level officers, on average, are -1.230% per month at linked firms versus only -0.277% at non-linked firms altogether. Breaking these control observations down further, the average returns are -0.355% per month at never-linked firms and 0.359% at linked firms during non-linked years. The differences in abnormal returns at linked suppliers compared with these two groups are 0.875% (T-statistic = 3.28) per month and 1.589% (T-statistic = 3.67) per month, respectively. Similar patterns are also apparent for outside directors. These results

15

suggest that top-level officers and directors are the most skilled at interpreting and profiting from information about their customers. The size-adjusted CARs following insider purchases are reported in Panel B of Table 3. Consistent with prior research, average CARs following insider purchases are larger than those following insider sales. However, we do not find that the abnormal returns following purchases are greater at economically linked firms. The size-adjusted CARs for insider purchases on average are 0.922% per month at linked firms and 0.709% per month at never-linked firms, a difference of -0.213% (T-statistic = 1.40) per month. Purchases are more profitable, on average, at economically linked firms during non-linked years which are followed by average CARs of 1.398% (5% significance on the difference). This pattern holds generally across the different groups of insiders, but the significance of the differences is lost in the smaller samples. Our next step is to analyze the returns following insiders’ trades in a multivariate regression framework, in which we can control for other risk factors and include explanatory variables to evaluate whether the returns are associated with public information. This analysis builds on the empirical setup of Cohen, Malloy, and Pomorski (2012) and is presented in Table 4 for insiders’ sales (Panel A) and insiders’ purchases (Panel B). We follow the framework illustrated in Fig. 2 for this analysis. Supplier insider trade months are classified as sale (purchase) months in month t if there is at least one net seller (purchaser) of shares during that month and no net purchasers (sellers). Supplier excess returns in month t+1 are regressed onto explanatory variables. The baseline specification (Column 1) includes those risk factors discussed previously and a dummy indicating that a firm reported an economic link to a principal customer in that year (Link). The coefficient on Link in the regression for sales of -0.0037 (Tstatistic = 3.31) in Panel A confirms the incremental profitability of these trades at economically 16

linked suppliers. The similarity of returns following insiders’ purchases across the linked and non-linked settings is also confirmed by the much smaller and insignificant coefficient on Link in Panel B, Column 1. Insert Table 4 and Fig. 2 near here The subsequent regressions in this table explore the extent to which public information explains insiders’ returns. We first conduct regressions using only those trades at economically linked suppliers that are executed in the first year of the reported relationship to a principal customer. During this year, the relationship might not have yet been disclosed in public filings with the Securities and Exchange Commission (SEC). In Panel A, we do not find evidence of abnormal returns following supplier insiders’ sales during this year (insignificant coefficient of 0.0014 on the intercept). This could indicate insiders’ reluctance to sell stock on information before the public could be expected to know the importance of the economic link. In the regression evaluating purchases, in Panel B, we find evidence of larger positive abnormal returns during these years (coefficient of 0.0261 on the intercept), but the result is not statistically significant. Although not significant, the point estimate would suggest that insiders take advantage of the market’s lack of knowledge to make greater profits on purchases. The lack of significance could reflect the much smaller sample size (N = 1,203). Next we consider whether insiders’ profits can be explained by public information. We examine the relation between supplier insiders’ trading profits and public information about their customers’ prior stock returns. Cohen and Frazinni (2008) show that past customer returns predict future abnormal returns in the same direction for their suppliers. The regressions in Columns 3 and 4 of both panels include trades only following positive lagged three month CARs

17

at the principal customers. Similarly, Columns 5 and 6 present regressions in which trades follow negative customer CARs.11 These regressions show that both sales and purchases are only profitable on average (with statistical significance) when the lagged customer returns predict they will be. In particular, sales are only followed by negative abnormal returns when the lagged customer returns are also negative (coefficient on the intercept in Panel A Column 5 of -0.0279, significant at the 1% level); and purchases are only followed by positive abnormal returns when lagged customer returns are positive (coefficient on the intercept in Panel B Column 3 of 0.0326, significant at the 5% level). These abnormal returns are large in economic terms.12 The magnitude of the returns earned by supplier insiders on their sales is also a function of the magnitude of the lagged customer CARs. The coefficient on Customer CARt-3,t-1 is insignificant when supplier insiders sell stock following periods of positive customer abnormal returns (Column 4 in Panel A). However, their profits from sales following negative customer returns are highly sensitive to the magnitude of the customers’ returns (Column 6 in Panel A). In contrast, we do not find the same relation for supplier insider purchases. These results show that insiders’ trading profits are a function of the strength of the public signal when they sell stock, which would not be expected if they were trading solely on private information. In the final two specifications in this table, we account for the level of outside investor attention to the supplier-customer relationship. In Column 7 of each panel we include Log(Link Duration), which indicates the length of the supplier-customer relationship in years. If insiders’

11

In the case that a supplier firm has multiple principal customers, the average customer CAR is used in our analysis. 12 To determine whether the abnormal return patterns we find are driven by risk associated with the existence of a principal supplier-customer relationship, we conduct similar regressions but with abnormal returns that are measured following supplier insider trade months relative to the returns at other supplier return months. The results are qualitatively unchanged.

18

superior returns reflect the sharing of private information from customers to suppliers, we would expect their profits to increase over time as relationships strengthen. Alternatively, if market inattention to public information generates these trading opportunities, we expect the abnormal returns to decline over time as the market learns about return predictability across firms.13 We also include year fixed effects in this specification to alleviate concerns that we could pick up a time trend in the returns to insider trading. We find that the length of the relationship is positively related to the abnormal returns following supplier insider sales, indicating that the returns are more negative earlier in the relationship, which lends additional support to the hypothesis that the insiders’ sales are often motivated by public information. In contrast, we do not find this relation for supplier insiders’ purchases. In Column 8 we include Log(Analysts) and Log(Sales) to account for how attentive outside investors could be to the supplier-customer relationship. Log(Analysts) captures the number of equity analysts covering the supplier. We expect the market to be faster to incorporate information about principal customers into the suppliers’ stock price when more analysts cover the stock, thus reducing the profits earned by supplier insiders if they trade on public information. We find that this is the case for supplier insiders’ sales in Panel A, where the coefficient on Log(Analysts) is positive and significant, indicating less negative returns following sales when there is more attention to the stock. Log(Sales) indicates the percentage of the suppliers’ sales that go to the principal customer. When the economic link is stronger, the market could pay more attention. However, we do not find that the strength of this relation impacts the

13

If suppliers have more than one principal customer, we use in our analysis the average number of months separating the trade month from the first reported relationship with each customer as our measure of the duration of the supplier-customer relationship.

19

returns to either insiders’ purchases or sales. Perhaps this non-result reflects the endogenous choice whether to trade based on the opportunity for profits. 3.2. What causes insiders at economically linked suppliers to trade? The second part of our analysis focuses on the factors that lead insiders to trade. We begin by investigating whether the frequency of trading differs depending on whether firms report having principal customers. In Table 5, Panel A, we compare the number of trade months per year across linked suppliers and non-linked firms.14 We find significantly higher levels of both routine and nonroutine stock sales at economically linked firms. However, the difference in nonroutine trading across these groups of firms is much larger. This leads to a significantly higher difference-in-differences (the difference in nonroutine stock sale months per year minus the difference in routine stock sale months per year) of 0.127 trade months per year for the economically linked suppliers versus other firms. This pattern does not hold for stock purchases. Insiders at economically linked suppliers engage in higher levels of both routine and nonroutine purchases, but the difference-in-differences is essentially zero. Insert Table 5 near here In Table 5, Panel B, we compare trading patterns at just the economically linked firms across years with and without reported links. For both sales and purchases, there is a lower level of routine trading and a higher level of nonroutine trading in years when the economic links are reported. The results in this panel suggest that for these firms, in aggregate, more profitable trading opportunities arise when these links exist. Because the results are similar across sales and purchases, they do not provide a clear indication of the type of information that is available. 14

Because the ratio of nonroutine trade months per year to routine trade months per year is undefined for years containing zero routine trade months, we cannot evaluate a simple ratio of trade types.

20

Among the possibilities, it could be that supplier insiders have both more public and private information predicting returns when these links exist. The increased selling could reflect trading on public information, but the increased purchasing could reflect additional trading on private information, too. The combination of larger abnormal returns following insider sales and more non-routine trading suggests a higher prevalence of profitable insider selling overall. To examine this further, we compare the percentages of trade months that are profitable across settings. We consider a sale month to be profitable if it is followed by a negative size-adjusted CAR in the following month. Insiders at economically linked suppliers execute 17.6% more profitable sales than unprofitable sales overall and 19.4% more profitable nonroutine sales. By comparison, the same figures for linked firms in non-linked years are 6.8% and 12.4%, respectively; and for never linked firms, they are 10.3% and 9.5%. Chi-square tests show that the higher proportions of profitable trade months at linked firms are significantly different from those at the other two groups at the 1% level. The final issue we examine is whether public information about principal customers drives trading by supplier insiders. Supplier insiders could understand the return predictability across economically linked firms and trade when they observe abnormal returns to their customers’ stock. Table 6 presents limited dependent variable regressions predicting trading by supplier insiders. The independent control variables are the suppliers’ market value of equity [Log(Size)], book-to-market equity [Log(B/M)], the contemporaneous market return (Markett), and the three-month lagged abnormal returns to the supplier (SupplierCARt-3,t-1). The variable of interest is either the lagged customer abnormal return (CustomerCARt-3,t-1) or a dummy indicating that the lagged customer return was negative or positive (CustomerCARt-3,t-1Negative 21

for predicting sales or CustomerCARt-3,t-1Positive for predicting purchases). If the coefficients on these variables are significant, then that would suggest that insiders trade when public information predicts they will earn profits. We present both probit regressions that predict whether a supplier insider traded and Tobit regressions that predict the number of shares traded. The regressions predicting supplier insider sales are presented in Columns 1 through 4, and those predicting purchases are reported in Columns 5 through 8. The results clearly indicate that supplier insiders trade more often, and they trade more shares, when their customers’ lagged returns predict that their trades will be profitable. For example, the marginal effects associated with the coefficients on CustomerCARt-3,t-1Negative in the probit regression in Column 2 indicate that supplier insiders are 8.07% more likely to sell shares when their lagged customer returns were negative. The same coefficient from the Tobit regression in Column 4 indicates that, when they do sell stock, they sell 27,290 more shares when their customers’ lagged returns were negative. Similarly, the marginal effects indicate that when lagged customer returns are positive, they are 3.67% more likely purchase shares (Column 6); and they purchase 8,308 more shares under those circumstances (Column 8).15 Insert Table 6 near here These results are consistent with supplier insiders trading based on their knowledge of the return relation between their firm and their customers. Attentive supplier insiders also could trade around the specific news releases of their customers. The next part of the analysis, therefore, offers tests focused on trading immediately around principal customers’ earnings releases. In Table 7 we present limited dependent variable regressions predicting supplier insider 15

The marginal effect presented here is E(shares_sold|shares_sold>0). This is accomplished in Stata with the command mfx, pred((e,0).

22

trading during the two-day window (0,1) around their customers’ earnings releases.16 We remove observations with confounding supplier earnings announcement during an approximately threeweek window [(-8,8) trading days]. We are interested in whether supplier insiders are more likely to trade when customers announce unexpected earnings, which could predict the suppliers’ own future sales and, therefore, profits. To control for the fact that insiders’ trading decisions could be driven by general concerns about the future of the market that are reflected in the earnings of similar firms, we include as control observations the earnings announcements of all firms in the same SIC4 industry as the principal customers (Cust Industry Firm SUE contains the SUE of both customers and their peers). The variable of interest is the standardized unexpected earnings reported by the customer (Customer*Cust Industry Firm SUE), where Customer is a dummy variable indicating earnings announcements made by the actual large customers. Insert Table 7 near here These results are presented in Table 7 for supplier insider sales (Panel A) and purchases (Panel B). The regressions in Columns 1 through 3 of each panel present probit regressions and Columns 4 through 6 present Tobit regressions with the same specifications. The results in Panel A provide strong evidence that supplier insiders’ decisions to sell their stock is impacted by the information in their customers’ earnings releases. The highly significant negative coefficient in Column 1 on the interaction term Customer*Cust Industry Firm SUE demonstrates a negative relation between the customers’ SUE and supplier insider sales. In Column 2, the SUE is replaced by a dummy indicating a negative SUE and the coefficient on the interaction term Customer*Cust Industry Firm SUE Negative remains highly significant. In Column 3, the SUE is 16

Because there are so many non-trading observations, here we winsorize the upper tail of the distribution for the variables Shares Sold and Shares Purchased at the 99% level of the nonzero values.

23

replaced with a dummy indicating a large shortfall (below -0.3% of the stock price, which represents approximately the 10th percentile of earnings surprises). Here the magnitude of the coefficient on the interaction term Customer*Cust Industry Firm SUE<-0.3% increases, indicating that supplier insiders sell more shares in this two-day window when their customers report large earnings shortfalls. In terms of marginal significance, the result in Column 2 indicates a 41.60% increase in the odds that a supplier insider sells shares when its principal customer announces a negative earnings surprise, and the result in Column 3 indicates that when the surprise is large there is a 58.82% increase in the odds of a supplier insider sale. In comparison with the unconditional odds of a supplier insider sale during one of these two-day windows, these results are economically significant. These results are also evident in the Tobit regressions. From the regressions in Columns 5 and 6, we are able to estimate that, conditional on a sale, supplier insiders sell 3,353 more shares when their customers announce disappointing earnings and 4,793 more shares when the earnings shortfall was large. We obtain similar results for supplier insider sales in Panel B when using dummy variables to control for positive customer earnings surprises (Columns 2, 3, 5, and 6), although the result is not evident when using a continuous customer SUE variable (Columns 1 and 4). To conclude this analysis, we present a number of robustness tests in Table 8. These additional tests focus on two issues: (1) whether the decision to trade around customer earnings announcements is a function of how quickly the market incorporates information into prices, and (2) to what extent we can be confident that the supplier insiders are trading on public information. To address the first issue, we evaluate the same regressions reported in Table 7 but limit the sample to earnings announcements made on Fridays. DellaVigna and Pollet (2009) show that investors are generally less attentive to corporate announcements made at the end of 24

the week, by showing that post-earnings announcement drift (PEAD) is extended when earnings surprises are announced on Fridays.17 These regressions demonstrate a stronger relation between the customer earnings surprises announced on Fridays and supplier insiders’ stock trades. The probit regressions predicting supplier insider stock sales in Columns 1 and 2 of Panel A estimate marginal increases in the probability of a stock sale of 213% and 425% when a customer announces earnings shortfalls and large earnings shortfalls, respectively, on Fridays.18 Similar unreported Tobit regressions estimate marginal increases in the number of shares sold by supplier insiders of 11,740 and 19,190 when customers announce earnings shortfalls and large shortfalls on Fridays, respectively, conditional on there being a sale. The results are also more pronounced for supplier insider purchases, reported in Panel B. Here, the probit regressions indicate marginal increases in the odds of a supplier insider purchase of 130% or 328% for positive earnings surprises and large positive earnings surprises, respectively.19 Here, Tobit regressions estimate marginal increases in the number of shares purchased of 1,962 and 3,837, respectively. Insert Table 8 near here The results thus far provide evidence that supplier insider trading is a function of the information content in customers’ public announcements. We conclude our analysis with a number of robustness tests designed to evaluate, as best as possible, whether trading in this

17

A recent working paper by Michaely, Rubin, and Vedrashko (2013) argues that the extended PEAD associated with Friday earnings announcements is not due to investor inattention at the end of the week, but rather the characteristics of firms that make Friday announcements (small market cap, low analyst following, low institutional shareholdings) or the fact that these announcements are more likely to be made after markets close for the day, or both. Regardless of the factors driving the extended PEAD under these circumstances, our test should still be valid to the extent all of these factors are associated with investor inattention. 18 The odds of a stock sale are higher by 3.26% and 6.50%, respectively, following missed earnings and large misses, compared with the unconditional odds of a sale of 1.53%. 19 The odds of a stock purchase are higher by 0.658% and 1.66%, respectively, for positive earnings surprises and large surprises, compared with the unconditional odds of a purchase of 0.506%.

25

context is genuinely motivated by information learned from public sources. The first issue addressed is whether we can be confident that supplier insiders’ trades around the time of their customers’ earnings announcements follow the public release of the information. Over the time period studied we do not have time-stamp data capturing when earnings were released, so we cannot be certain that trading on the day of the release (day 0) follows the announcement. In Columns 3 and 4 of Table 8 we provide regressions that analyze trading in the two-day window (1,2) immediately following the day of the earnings release. For both sales (Panel A) and purchases (Panel B), the results show that the fundamental relation between supplier insider trading and customer earnings surprises remains intact when we focus on this window that follows the earnings release. Again, unreported Tobit regressions that analyze the number of shares traded give consistent results. The next alternative we consider is whether supplier insiders are privy to the information content of their customers’ earnings releases ahead of time and actively trade on this information while it is still expected to be private. Principal customers could convey information to their suppliers in the normal course of business (through sales orders, cancellations, or indications of commitments), or they could intentionally tip suppliers as to the content of upcoming news releases. In Columns 5 and 6 of each panel we analyze trading in the two-day period immediately preceding customer earnings releases. The results are interesting. We find only weak evidence that supplier insiders sell stock during this period when the pending customer announcement predicts they should. But we find that the relation between supplier insider purchases and the content of the pending release is similar to what was found following the release. Furthermore, for the most part, the insiders that trade in the expected direction ahead of customer earnings releases are not the same ones who trade following the announcements. We

26

find that only 11% of insiders who sell in the week following negative customer earnings surprises also sell in the two days leading up to the announcement and only 6% of insiders who purchase following a positive surprise also purchased before the announcement. At the firm level, in only 15% of cases in which an insider sold following a negative customer earnings surprise did another insider at that firm sell in the days before the announcement. The similar measure for purchases is just 10%. We offer a couple of comments on these findings. First, it is reasonable to expect customers to be more careful not to convey negative news before they release it to the public, either due to similar concerns over legal liability or because of a desire to hide negative information but convey positive information quickly.20 From this perspective, selling by supplier insiders following the poor earnings releases looks more like attentive trading on newly public information, whereas purchasing before the earnings releases seems more likely to reflect a willingness to buy stock on private information, if available. The second alternative is that supplier insiders have similar access to positive and negative information from their customers before it is released, but they refrain from selling, in particular, until bad news is released to the public. If the informed insiders wait to trade, then this is analogous to an insider trading following their own public announcements (that they would have had prior knowledge of), with the additional benefit that they are confident they can profit because the market will be slow to respond to the information. Overall, this piece of our analysis confirms the existence of a set of supplier insiders who trade attentively on public information released by their customers, while

20

See Graham, Harvey, and Rajgopal (2005), Anilowski, Feng, and Skinner (2006), and Kothari, Shu, and Wysocki (2009) for evidence that managers prefer to release or leak good news quickly but withhold bad news from the market as long as possible. Similar reasoning could drive firms’ communications with their key business partners.

27

others could exploit their informational advantage prior to the announcements, especially when the news is good. The final analysis that we offer helps determine whether supplier insiders genuinely learn from their customers’ public disclosures. To get at this question, we evaluate the trading decisions of supplier insiders around customer earnings announcements, conditioned on the fact that they are already trading in the predicted direction during the 20 trading days prior to the release. Because they are already trading, it is likely that they either already know the content of their customer’s announcement and are taking advantage of this before it was made public, are already trading on other private signals, or are trading for unrelated reasons in the lead-up period. The results are reported in Columns 7 and 8 of the panels. For the case of stock sales we continue to find some evidence that these insiders’ trading behavior immediately following their customers’ announcements is related to the information content of the customers’ earnings, even in this limited sample. This would not be expected if they are trading on this particular information ahead of time. Also, given that they are already trading, it does not appear that they are waiting for a public news release to camouflage trading motivated by other, unrelated, private information. It is therefore reasonable to expect that the earnings releases, in fact, convey new information to these insiders, motivating their post-announcement stock sales. The exception to this interpretation would be if these insiders begin a sequence of trading on the information about their customer while it is still private that extends into the post-announcement period. 4. Conclusion A body of literature demonstrates that corporate insiders’ trades predict abnormal returns. Because insiders are privy to information about their firms before it is disclosed to the public, it is natural to suspect that these returns reflect trading decisions based on private information.

28

However, there are also significant costs to trading on private information as doing so is in violation of securities laws and invites scrutiny from regulators and outside investors. Another body of literature shows that investors are limited in their ability to understand the impact of relevant public information on stock prices. This can result in opportunities for particularly attentive investors to trade profitably on public information. Corporate insiders likely represent such a group of investors. We analyze corporate insider trading in a setting that allows us to possibly distinguish between illegal trading on private information and attentive trading on public information. The setting is trading at firms that sell a large fraction of their products to principal customers. We find that insiders at these firms execute more profitable stock sales than those at other firms, and we provide evidence that at times they generate these enhanced profits by trading on public information. They sell more stock when public information about their principal customers predicts that their trades will be profitable and their customers’ past returns explain the abnormal returns following their sales. We do not find the same contrasting patterns for insider purchases, suggesting that insiders are more likely to refrain from trading on private information when the litigation risk associated with being detected is high, which is the case with insider sales more so than with purchases. More work would have to be done to estimates the actual levels of attentive insider stock selling. Other researchers have uncovered complimentary patterns of trading and stock return predictability that could point toward the extent of this behavior. For example, Ali, Wei, and Zhou (2011) show that insiders trade in the opposite direction of high volume price-moving trading by mutual funds, which could represent a similar form of insider trading motivated by mispricing relative to public information. Jenter (2005) shows that insiders tend to be contrarian, 29

which could indicate they trade when behavioral biases in other investors cause their stock price to move away from its fundamental value. In terms of market inefficiencies that insiders could take advantage of, Lo and MacKinley (1990) show that small stocks tend to be less efficiently priced than large stocks, leading to return predictability across the two groups. Further, Sadka (2006) argues that stock return momentum and PEAD returns are associated with increased insider trading. The perspective supported by this paper is also related to the one proposed by papers showing a relation between insiders’ trading profits and information asymmetry (Aboody and Lev, 2000; Frankel and Li, 2004; and Huddert and Ke, 2007). That body of work argues that insiders make greater profits on private information when firms are more opaque and, therefore, more private information is available on which to trade. However, the results we present could call for a reinterpretation of those results if proxies for information asymmetry are also related to the time it takes the market to digest public information. Also, given their knowledge and experience with their own firms, insiders likely are attuned to a variety of other forms of public information that bear on their firms’ values but to which the market does not quickly adjust. We caution, though, against overgeneralization of our results. For one, we do not find evidence of attentive insider stock purchasing on public information. In addition, our results for stock sales could characterize insider trading only in the specific situation in which firms maintain strong economic links to other public companies, and they could characterize a subset of the trades by insiders only at these firms. Our evidence should be balanced against prior studies that provide evidence of trading on private information in certain settings. Examples include insider trading during private merger negotiations (Heitzman and Klasa, 2013) or before the revelation of accounting fraud (Agrawal and Cooper, 2014). Jagolinzer (2009) also shows that insiders earn significantly higher profits when they are afforded safe harbors from legal 30

liability, such as when their trades follow pre-determined plans as allowed under SEC Rule 10b5-1. His results could have captured a subset of illegal trading on private information under special circumstances when negative repercussions could be avoided. But, one should also be careful not to overgeneralize these results either. In the case of private merger negotiations, the clear prediction is that target insiders should purchase undervalued shares, so this activity is unlikely to speak to the prevalence of illegal insider selling. And illegal selling during accounting fraud could reflect the character of a limited group of unethical executives, such that generalization to other insiders could be inappropriate. Finally, we submit that our analysis could be useful for other researchers by laying the groundwork for identifying insiders who are more attentive to market developments that will impact their firms in the future. This information could be useful as a measure of executive or director quality or effort. Such a measure could be a valuable indicator of insiders’ ability to adapt to future business conditions, thus impacting future firm values, security returns, and economic growth, while minimizing wasteful overinvestment.

31

REFERENCES Aboody, D., Lev, B., 2000. Information asymmetry, R&D, and insider gains. Journal of Finance 55, 2747-2766. Agrawal, A., Cooper, T., 2014. Insider trading before accounting scandals. Unpublished working paper. Available at SSRN: http://ssrn.com/abstract=929413. Agrawal, A., Nasser, T., 2012. Insider trading in takeover targets. Journal of Corporate Finance 18(3), 598-625. Ali, A., Wei, K. D., Zhou, Y., 2011. Insider trading and option grant timing in response to fire sales (and purchases) of stocks by mutual funds. Journal of Accounting Research 49, 595-632. Anilowski, C., Feng, M., Skinner, D., 2006. Does earnings guidance affect market returns? The nature and information content of aggregate earnings guidance. Journal of Accounting and Economics 44, 36-63. Bacchetta, P., Van Wincoop, E., 2010. Infrequent portfolio decisions: a solution to the forward discount puzzle. American Economic Review 100 (3), 870-904. Bettis, C., Vickrey D., Vickery, D. W., 1997. Mimickers of corporate insiders who make large-volume trades. Financial Analysts Journal 53, 57-66. Brealey, R. A., Myers, S. C., Allen, F., 2011. Principles of Corporate Finance. McGraw-Hill, New York, NY. Brochet, F., 2010. Information content of insider trades before and after the Sarbanes-Oxley Act. The Accounting Review 85 (2), 419-446. Chen, C., Martin, X., Wang, X., 2012. Insider trading, litigation concerns, and auditor going-concern opinions. The Accounting Review 88 (2), 365-393. Cicero, D. C., Wintoki, M. B., 2014. Insider trading patterns. Unpublished working paper. Available at SSRN: http://ssrn.com/abstract=2128127. Cohen, L., Frazzini, A., 2008. Economic links and predictable returns. Journal of Finance 63, 1977-2011. Cohen, L., Malloy, C., Pomorski, L., 2012. Decoding inside information. Journal of Finance 67, 10091043. DellaVigna, S., Pollet, J., 2009. Investor inattention and Friday earnings announcements. Journal of Finance 64, 709-749. Finnerty, J., 1976. Insiders and market efficiency. Journal of Finance 31, 1141-1148. Fiske, S., Taylor, S., 1991. Social Cognition, second ed. McGraw-Hill, New York, NY. Frankel, R. M., Li, X., 2004. Characteristics of a firm's information environment and the information asymmetry between insiders and outsiders. Journal of Accounting and Economics 37, 229-259.

32

Graham, J., Harvey, C., Rajgopal, S., 2005. The economic implications of corporate financial reporting. Journal of Accounting and Economics 40, 3-37. Heitzman, S., Klasa, S., 2013. Private information arrival, trading activity, and price formation: evidence from nonpublic merger negotiations. Unpublished working paper. Available at SSRN: http://ssrn.com/abstract=1938436. Hong, H., Torous, W., Valkanov, R., 2007. Do industries lead stock markets? Journal of Financial Economics 83, 367-396. Huang, L., Kale, J., 2012. Product market linkages, manager quality, and mutual fund performance. Review of Finance 17 (6), 1895-1946. Huang, L., Liu, H., 2007. Rational inattention and portfolio selection. Journal of Finance 62, 1999-2040. Huddart, S. J., Ke, B., 2007. Information Asymmetry and Cross-sectional Variation in Insider Trading. Contemporary Accounting Research 24, 195-232. Jaffe, J., 1974. Special information and insider trading. Journal of Business 47, 410-428. Jagolinzer, A. D., 2009. SEC Rule 10b5-1 and insiders' strategic trade. Management Science, 55 (2), 224239. Jeng, L., Metrick, A., Zeckhauser, R., 2003. Estimating the returns to insider trading: a performance evaluation perspective. Review of Economics and Statistics 85, 453-471. Jenter, D., 2005. Market timing and managerial portfolio decisions. Journal of Finance 60 (4), 1903-1949. Kahneman, D., 1973. Attention and effort. Prenctice Hall, Englewood Cliffs, N.J. Kothari, S. P., Shu, S., Wysocki, P. D., 2009. Do managers withhold bad news? Journal of Accounting Research 47, 241–276. Lakonishok, J., Lee, I., 2001. Are insider trades informative? Review of Financial Studies 14, 79-111. Lo, A. W., MacKinlay, A. C., 1990. When are contrarian profits due to stock market overreaction? Review of Financial studies 3 (2), 175-205. Lorie, J., Niederhoffer, V., 1968. Predictive and statistical properties of insider trading. Journal of Law and Economics 11, 35-53. Manne, H. G., 1966. Insider Trading and the Stock Market. Free Press, New York, N.Y. Menzly, L., Ozbas, O., 2010. Market segmentation and cross-predictability of returns. Journal of Finance 65, 1555-1580. Michaely, R., Rubin, A., Vedrashko, A., 2013. Firm heterogeneity and investor inattention to Friday earnings announcements. Unpublished working paper. Available at SSRN: http://ssrn.com/abstract=2164789

33

Peng, L., Xiong, W., 2006. Investor attention, overconfidence and category learning. Journal of Financial Economics 80, 563–602. Sadka, R., 2006. Momentum and post-earnings announcement drift anomalies: The role of liquidity risk. Journal of Financial Economics 80, 309-349. Seyhun, N., 1986. Insiders' profits, costs of trading, and market efficiency. Journal of Financial Economics 16, 189-212. Seyhun, N., 1992. Why does aggregate insider trading predict future stock returns? Quarterly Journal of Economics 107, 1303-1331. Seyhun, N., 1998. Investment Intelligence from Insider Trading. MIT Press, Cambridge, MA. Skinner, D., 1994. Why firms voluntarily disclose bad news? Journal of Accounting Research 32, 38-60.

34

Number of links

0

1,000

2,000

3,000

4,000

5,000

6,000

7,000

1

(6,229)

2

(3,168)

4

(944)

6

(455)

7

(323)

Link duration (years)

5

(614)

35

8

(198)

Fig. 1 Durations of the supplier-customer links

3

(1,612)

9

(165)

10

(116)

11+

(312)

36

Fig. 2 Framework for calculating abnormal returns as a function of returns to principal customers

Table 1 Summary statistics This table reports summary statistics for the sample firm-months of nonroutine insider trades from January 1986 to December 2010. The sample is divided into linked firm insider sale (purchase) months in Panel A (Panel C) and non-linked firm insider sale (purchase) months in Panel B (Panel D). The variable B/M is the previous year-end book-to-market equity ratio. The variable Size is the previous year-end market value (in millions). Trade days per firm-month, traders per firm-month, and firms per month are also reported. If a firm-month contains both an insider net sale and an insider net purchase, then the observation is removed from the sample. Mean

Median

Standard deviation

5%

95%

Panel A: Linked firm sales (1,375 firms; 11,805 firm-month observations) B/M

0.421

0.367

0.275

0.086

0.923

Size

4,445

676

17,914

123

16,876

Trade days per firm-month

2.61

1.00

3.80

1.00

8.00

Traders per firm-month

1.57

1.00

1.66

1.00

4.00

Firms per month

37.0

34.0

18.0

13.0

70.0

Panel B: Non-linked Firm Sales (5,729 firms; 82,905 firm-month observations) B/M

0.455

0.402

0.301

0.096

0.977

Size

4,825

821

18,626

131

18,713

Trade days per firm-month

2.47

1.00

3.23

1.00

7.00

Traders per firm-month

1.55

1.00

1.14

1.00

4.00

Firms per month

277.3

272.0

100.7

122.0

534.0

Panel C: Linked firm purchases (1,088 firms; 5,375 firm-month observations) B/M

0.510

0.452

0.346

0.107

1.106

Size

2,974

529

12,820

121

11,221

Trade days per firm-month

1.89

1.00

2.05

1.00

5.00

Traders per firm-month

1.40

1.00

0.96

1.00

3.00

Firms per month

17.3

15.0

10.3

4.0

39.0

Panel D: Non-linked firm purchases (5,203 firms; 46,037 firm-month observations) B/M

0.561

0.511

0.360

0.130

1.146

Size

3,509

525

17,058

117

11,695

Trade days per firm-month

1.97

1.00

2.58

1.00

6.00

Traders per firm-month

1.49

1.00

1.21

1.00

4.00

Firms per month

154.6

131.0

81.6

52.0

306.0

37

Table 2 Industry classifications of suppliers and principal customers This table reports the percentage of economically linked suppliers in our sample that are associated with each of the Fama and French 17 industry classifications. The table also shows the percentage difference between the distribution of non-linked firms and linked firms. Finally, it reports the distribution of the principal customers across the Fama and French 17 industry classifications. Fama and French 17 industry classification Food

Non-linked firms 3%

Linked firms 4%

-1%

Principal customers 2%

Mining and minerals

1%

1%

0%

0%

Oil and petro products

3%

5%

-2%

4%

Textiles, apparel and footwear

1%

4%

-3%

0%

Consumer durables

2%

4%

-2%

2%

Chemicals

2%

2%

0%

1%

Drugs, soap, perfumes, tobacco

5%

11%

-6%

12%

Construction

3%

3%

0%

3%

Steel

1%

2%

-1%

1%

Fabricated products

1%

1%

0%

0%

Machinery and business equipment.

12%

22%

-10%

18%

Automobiles

1%

4%

-3%

10%

Transportation

3%

5%

-2%

5%

Utilities

3%

2%

1%

3%

Retail stores

8%

1%

7%

17%

Financial institutions

20%

2%

18%

3%

Other

31%

29%

2%

20%

38

Difference

Table 3 Market-adjusted abnormal returns following supplier insider trades This table reports the one-month NYSE size decile–portfolio adjusted cumulative abnormal returns (CARs) following months with insider trading. The CARs for trade months by insiders at economically linked firms are compared with those at nonlinked firms and with those at two subsets of this control group: linked firms during non-linked years and never-linked firms. Panel A contains insider sales, and Panel B contains insider purchases. Column 1 reports results for all insiders. Column 2 reports results for the top-level insiders, who include chief executive officers, chief financial officers, and chief operating officers. Column 3 reports results for only the directors. Results for all other insiders are reported in Column 4. Standard errors are in parentheses and significance at the 1%, 5%, and 10% level are indicated by ***, **, and *, respectively.

All insiders (1)

Top-level officers (2)

Only directors (3)

Other insiders (4)

Linked suppliers Size-Adj. CAR (percent) Standard errors Number of observations

-0.671*** (0.103) 11,805

-1.230*** (0.262) 2,321

-0.809*** (0.193) 3,499

-0.473*** (0.118) 8,891

Non-linked firms Size-Adj. CAR (percent) Standard errors Number of observations

-0.292*** (0.036) 82,905

-0.277*** (0.101) 13,888

-0.231*** (0.063) 27,612

-0.313*** (0.040) 62,418

0.359 (0.354) 1,510

-0.220*** (0.193) 3,741

-0.218** (0.102) 11,349

-0.355*** (0.105) 12,378

-0.233*** (0.066) 23,871

-0.334*** (0.044) 51,069

Abnormal Returns Panel A: Sales

Linked suppliers in non-linked years Size-Adj. CAR (percent) -0.158* Standard Errors (0.091) Number of observations 15,290 Never-linked firms Size-Adj. CAR (percent) Standard Errors Number of observations

-0.322*** (0.039) 67,615

39

1.586*** (0.146) 7,522

Never-linked firms Size-Adj. CAR (percent) Standard Errors Number of observations

40

2.237*** (0.601) 485

Linked suppliers in non-linked years Size-Adj. CAR (percent) 1.398*** Standard Errors (0.145) Number of observations 5,782

0.709*** (0.052) 40,255

1.626*** (0.142) 8,007

0.796*** (0.049) 46,037

Non-linked firms Size-Adj. CAR (percent) Standard Errors Number of observations

1.630*** (0.405) 1,054

0.922*** (0.153) 5,375

Linked suppliers Size-Adj. CAR (percent) Standard Errors Number of observations

Panel B: Purchases

0.562*** (0.059) 27,990

1.101*** (0.180) 3,297

0.619*** (0.056) 31,287

0.763*** (0.183) 3,403

0.985*** (0.080) 18,670

1.430*** (0.247) 2,033

1.030*** (0.076) 20,703

1.330*** (0.241) 2,415

Number of

CARt-12,t-1

CARt-3,t-1

Markett+1

Log(B/M)

Log(Size)

Log(Sales)

Log(Analysts)

Log(Link Duration)

Link

Customer CARt-3,t-1

Constant

Excess Rett+1

Panel A: Insider sales

94,710

-0.00007 (0.000217) 0.00906*** (0.00208) 1.040*** (0.0111) 0.00388 (0.00271) 0.00824*** (0.00117) 2,924

-0.00344* (0.00178) 0.0190 (0.0153) 1.240*** (0.0613) 0.0306** (0.0150) -0.00468 (0.00523)

(2) 0.00140 (0.0148)

(1) -0.0119*** (0.00205)

-0.00371*** (0.00112)

Linked firms (first year of link)

Linked and non-linked firms

5,754

0.000174 (0.00102) 0.00192 (0.00875) 1.084*** (0.0394) -0.0128 (0.0106) 0.0109** (0.00440)

Linked firms (positive lagged customer CAR) (3) -0.0114 (0.00882)

41

5,754

0.000223 (0.00102) 0.00254 (0.00879) 1.084*** (0.0394) -0.0144 (0.0106) 0.0108** (0.00439)

Linked firms (positive lagged customer CAR) (4) -0.0146 (0.00937) 0.0294 (0.0254)

6,051

0.00101 (0.000960) 0.0150* (0.00829) 1.201*** (0.0415) 0.00751 (0.0103) -0.00420 (0.00432)

Linked firms (negative lagged customer CAR) (5) -0.0279*** (0.00840)

6,051

0.000938 (0.000957) 0.0148* (0.00824) 1.204*** (0.0415) 0.00430 (0.0104) -0.00368 (0.00431)

Linked firms (negative lagged customer CAR) (6) -0.0199** (0.00869) 0.0824*** (0.0249)

94,710

-0.000137 (0.000218) 0.0103*** (0.00211) 1.065*** (0.0115) 0.00287 (0.00272) 0.00849*** (0.00116)

-0.0160*** (0.00606) 0.00330** (0.00157)

(7) -0.00805** (0.00316)

Linked and nonlinked firms

8,445

0.00644*** (0.00206) 0.00532 (0.0122) -0.00198* (0.00113) 0.00783 (0.00771) 1.127*** (0.0372) -0.00569 (0.00869) 0.00272 (0.00353)

(8) -0.0148* (0.00854)

Linked firms

This table compares cumulative abnormal returns (CARs) following insider sales at economically linked suppliers with those at other firms in a multivariate regression framework. It also presents regressions of monthly returns following insider sales at economically linked suppliers on control variables and lagged returns to their principal customers. The dependent variable Excess Rett+1 is the one-month excess return following trade month t. Link is an indicator equal to one if the firm has a principal customer during the given year. Log(Link Duration) is the log of the total number of months separating the trade month and the first reported relationship with the principal customer. Log(Analyst) is the log of number of analysts covering the supplier, and Log(Sales) is the log of the percentage of firm sales associated with principal customer. Customer CARt-3,t-1 is the customer three-month market adjusted return from months t-3 to t-1. Log(Size) is the log of the previous year-end market value. Log(B/M) is the log of the previous year-end book-to-market equity ratio. CARt-3,t-1 is the supplier’s three-month market-adjusted return from months t-3 to t-1, and CARt-12,t-1 is the supplier’s one-year market-adjusted return from months t-12 to t-1. Markett+1 is the equal-weighted market return. Panel A contains insider sales, and Panel B contains insider purchases. Column 2 contains observations from insider trade months in the first year of the supplier-customer relation, and all other regressions in this table omit observations from insider trades during the first year of the supplier-customer relationship. Columns 1 and 7 contain results for insider trade months at linked and non-linked firms. Columns 3 and 4 (Columns 5 and 6) contain regression results for insider trade months at linked firms following three-month positive (negative) lagged customer CARs. Column 8 contains only the regression results for insider trade months at linked firms. Clustered standard errors at the firm level are in parentheses, and significance at the 1%, 5%, and 10% level are indicated by ***, **, and *, respectively.

Table 4 An analysis of supplier insider trading returns compared with insider trading at other firms

0.000171

(0.0126)

0.000197

(0.0369)

(0.0128) 0.0276

(0.00472)

(7) 0.00468

Linked and nonlinked firms

0.153

Number of observations R2

51,412 0.173

1,203 0.144

(0.0252) 0.0142 (0.0123)

(0.00379) 0.00395** (0.00193)

CARt-12,t-1

(0.0828) -0.0201

(0.0135) -0.0232***

CARt-3,t-1

2,575 0.160

(0.0166) -0.00462 (0.00838)

(0.0528) 0.0117

42

2,575 0.160

(0.0165) -0.00459 (0.00838)

(0.0527) 0.0121

(0.0138) 1.029***

2,800 0.199

(0.0165) 0.0116 (0.00738)

(0.0561) -0.0374**

(0.0120) 1.242***

2,800 0.200

(0.0168) 0.0117 (0.00737)

(0.0562) -0.0396**

(0.0120) 1.241***

51,412 0.176

(0.00380) 0.00280 (0.00203)

(0.0139) -0.0227***

(0.00296) 1.062***

3,581 0.179

(0.0141) 0.00255 (0.00708)

(0.0519) -0.0104

(0.0119) 1.181***

(0.00180) 0.0222*

Markett+1

(0.0139) 1.028***

(0.000326) -0.000215

(0.0196) 1.173***

(0.00139) 0.0325***

(0.00288) 1.050***

(0.00139) 0.0322***

Log(B/M)

(0.00154) 0.00354

(0.00266) 0.0280

(0.000324) -0.000125

Log(Size) (0.00154) 0.00386

(0.0236) -0.00110

-0.00442*

-0.00147***

Log(Sales) -0.00164***

-0.00236

(0.0142)

(8) 0.0122

Linked firms

0.152

(0.00381) -0.0251

(0.00228)

(0.00898) -0.00207

-0.00407***

(0.0359)

(0.0132) -0.0197

Linked firms (negative lagged customer CAR) (6) -0.0119

0.164

(0.00159)

-0.00408***

(0.0131)

Linked firms (negative lagged customer CAR) (5) -0.0140

0.162

0.00840

(0.0215)

(0.00293)

Linked firms (positive lagged customer CAR) (4) 0.0345***

0.154

0.000996

(2) 0.0261

(1) 0.0132***

Linked firms (positive lagged customer CAR) (3) 0.0326**

0.154

Log(Analysts)

Log(Link Duration)

Link

Customer CARt-3,t-1

Constant

Excess Rett+1

0.130

Linked firms (first year of link)

0.149

Linked and non-linked firms

Panel B: Insider purchases

observations R2

Table 5 The frequency of insider trading at economically linked suppliers This table presents a comparison of the mean number of trade-months per year across groups of firms. The comparison in Panel A is between linked firms and non-linked firms, and the comparison in Panel B is between linked firms during linked years and linked firms during non-linked years. We analyze insider sale months and insider purchase months separately. The table also includes a comparison of routine and nonroutine trades. Standard errors are in parentheses, and significance at the 1%, 5%, and 10% level are indicated by ***, **, and *, respectively. Trade months Panel A: All firms

Routine

Sale months Non-linked firms

0.416

Linked firms

Nonroutine

Difference-in-differences

1.284

N

81,961

0.457

1.452

-0.041***

-0.168***

-0.127***

(0.014)

(0.020)

(0.023)

Purchase months Non-linked firms

0.346

0.921

81,961

Linked firms

0.229

0.811

9,616

0.117***

0.110***

-0.007

(0.013)

(0.016)

(0.019)

Difference Standard errors

Difference Standard errors

9,616

 Panel B: Economically linked suppliers only Sale months Non-linked years

0.489

1.403

18,097

Linked years

0.457

1.452

9,616

Difference

0.032*

-0.049**

-0.081***

Standard errors

(0.017)

(0.023)

(0.028)

Purchase months Non-linked years

0.250

0.757

18,097

Linked years

0.229

0.811

9,616

Difference

0.021*

-0.054***

-0.075***

Standard errors

(0.012)

(0.016)

(0.019)

43

Number of observations Pseudo R2

Constant

Log(B/M)

Log(Size)

Markett

Supplier CARt-3,t-1

Customer CARt-3,t-1 Positive

Customer CARt-3,t-1 Negative

Customer CARt-3,t-1

97,247 0.0167

0.673*** (0.0256) 1.395*** (0.124) 0.0363*** (0.00370) -0.431*** (0.0291) -1.270*** (0.0296)

(0.0383)

97,247 0.0166

0.662*** (0.0254) 1.398*** (0.124) 0.0364*** (0.00370) -0.429*** (0.0291) -1.284*** (0.0300)

0.0249** (0.0105)

Probit regressions, Sale Indicatort (1) (2) -0.141***

97,247 0.00333

44

3,838*** (170.1) 7,428*** (819.0) 222.6*** (24.32) -2,530*** (193.1) -9,145*** (203.7)

(252.5)

97,247 0.00331

3,782*** (168.3) 7,449*** (818.7) 223.1*** (24.32) -2,522*** (193.1) -9,236*** (206.4)

165.2** (69.34)

Tobit regressions, Shares Soldt (3) (4) -828.5***

97,247 0.0103

-0.583*** (0.0335) -1.530*** (0.144) 0.00145 (0.00467) 0.244*** (0.0343) -1.636*** (0.0370)

(0.0470)

97,247 0.0102

0.0382*** (0.0129) -0.574*** (0.0332) -1.537*** (0.144) 0.00128 (0.00467) 0.242*** (0.0342) -1.652*** (0.0378)

Probit regressions, Purchase Indicatort (5) (6) 0.166***

97,247 0.00372

-966.1*** (58.26) -2,444*** (244.1) -0.669 (7.956) 426.0*** (58.08) -2,831*** (68.92)

(79.84)

97,247 0.00372

62.27*** (21.91) -955.5*** (57.68) -2,454*** (244.1) -0.800 (7.955) 423.8*** (58.05) -2,859*** (70.30)

Tobit regressions, Shares Purchasedt (7) (8) 238.9***

This table presents Tobit regressions analyzing the number of shares sold by supplier insiders as a function of lagged customer returns. In Columns 1 and 2 the dependent variable is Sale Indicatort, an indicator equal to one if the firm-month was a net insider sale month. In Columns 3 and 4 the dependent variable is Shares Soldt, the number of shares sold by all insiders (in thousands of shares) for each firm-month observation. In Columns 5 and 6 the dependent variable is Purchase Indicatort, an indicator equal to one if the firm-month was a net insider purchase month. In Columns 7 and 8 the dependent variable is Shares Purchasedt, the number of shares purchased by all insiders (in thousands of shares) for each firm-month observation. The independent variables are an indicator for negative lagged three-month cumulative abnormal customer returns (Customer CARt-3,t-1 Negative), an indicator for positive lagged three-month cumulative abnormal customer returns (Customer CARt-3,t-1 Positive), a continuous variable for lagged three-month cumulative abnormal customer return (Customer CARt-3,t-1), the three-month cumulative abnormal supplier return (Supplier CARt-3,t-1), the equalweighted stock market return (Markett), the log of market capitalization for the prior year [Log(Size)], and the log of the book-to-market equity ratio for the prior year [Log(B/M)]. Standard errors are in parentheses, and significance at the 1%, 5%, and 10% level are indicated by ***, **, and *, respectively.

Table 6 Predicting supplier insider trading with lagged customer returns

0.000356 (0.00472) -0.329*** (0.0352)

Log(Size)

Log(B/M)

-0.180 (0.268) -4.857*** (1.874)

Event window (0,1) probit, Sale Indicator (1)

Non-Customer* Cust Industry Firm SUE Customer* Cust Industry Firm SUE Non-Customer* Cust Industry Firm SUE Negative Customer* Cust Industry Firm SUE Negative Non-Customer* Cust Industry Firm SUE<-0.3% Customer* Cust Industry Firm SUE<-0.3%

Panel A: Insider sales

0.000331 (0.00472) -0.332*** (0.0352)

0.00118 (0.0134) 0.151*** (0.0370)

Event window (0,1) probit, Sale Indicator (2)

45

0.0105 (0.0154) 0.202*** (0.0556) 0.000512 (0.00472) -0.332*** (0.0352)

Event window (0,1) probit, Sale Indicator (3)

1.334 (1.160) -78.98*** (8.768)

-58.51 (65.83) -1,135** (463.0)

Event window (0,1) Tobit, Shares Sold (4)

1.331 (1.161) -79.45*** (8.773)

1.081 (3.309) 31.15*** (9.261)

Event window (0,1) Tobit, Shares Sold (5)

3.213 (3.806) 43.89*** (13.91) 1.370 (1.161) -79.61*** (8.774)

Event window (0,1) Tobit, Shares Sold (6)

This table presents probit and Tobit regressions predicting insider trading at economically linked suppliers as a function of their customers’ earnings announcements. The dependent variables in Panel A are Shares Indicator and Shares Sold. In Columns 1, 2, and 3 of Panel A, Sale Indicator is equal to one if there is at least one supplier insider sale during the window (0,1) days around an earnings announcement by a principal customer or another firm in the same industry. In Columns 4, 5, and 6 of Panel A, Shares Sold is the total volume of shares sold by supplier insiders (in thousands of shares) during the window (0,1) days around an earnings announcement by a principal customer or another firm in the same industry. The dependent variables in Panel B are Purchase Indicator and Shares Purchased, which are constructed analogously to the variables in Panel A. Five earnings surprise measures are used as independent variables: the earnings surprise for the customer and each firm in the customer’s industry (Cust Industry Firm SUE), an indicator equal to one if the earnings surprise is negative (Cust Industry Firm SUE Negative), an indicator equal to one if the earnings surprise is positive (Cust Industry Firm SUE Positive), an indicator variable equal to one if the earnings surprise is less than -0.3% of the stock price (Cust Industry Firm SUE <-0.3%), and an indicator variable equal to one if the earnings surprise is greater than 0.3% of the stock price (Cust Industry Firm SUE >0.3%). Customer is an indicator equal to one if the earnings surprise is for a customer. Non-Customer is an indicator equal to one if the earnings surprise is for a customer peer. Control variables include the past 30-day cumulative abnormal supplier return (Past Ret), the equal-weighted market return on the earnings announcement day (Market), the log of market capitalization for the prior year [Log(Size)], and the log of the book-to-market ratio for the prior year [Log(B/M)]. Standard errors are in parentheses, and significance at the 1%, 5%, and 10% level are indicated by ***, **, and *, respectively.

Table 7: Predicting supplier insider trading with customer earnings surprises

Market

Past Ret

Log(B/M)

Log(Size)

Non-Customer* Cust Industry Firm SUE Customer* Cust Industry Firm SUE Non-Customer* Cust Industry Firm SUE Positive Customer* Cust Industry Firm SUE Positive Non-Customer* Cust Industry Firm SUE>0.3% Customer* Cust Industry Firm SUE>0.3%

Panel B: Insider purchases

Number of observations Pseudo R2

Constant

Market

Past Ret

-0.0267*** (0.00798) 0.169*** (0.0529) -0.730*** (0.0501) -2.729** (1.249)

0.181 (0.435) -0.609 (2.990)

Event window (0,1) Probit, Purchase Indicator (1)

247,416 0.0150

0.554*** (0.0261) 6.528*** (0.782) -2.036*** (0.0356)

-0.0276*** (0.00799) 0.160*** (0.0531) -0.738*** (0.0503) -2.762** (1.248)

0.0122 (0.0210) 0.206*** (0.0385)

Event window (0,1) Probit, Purchase Indicator (2)

247,416 0.0152

0.555*** (0.0261) 6.517*** (0.782) -2.039*** (0.0360)

46

0.0289 (0.0248) 0.301*** (0.0635) -0.0264*** (0.00798) 0.160*** (0.0530) -0.733*** (0.0502) -2.805** (1.249)

Event window (0,1) Probit, Purchase Indicator (3)

247,416 0.0151

0.555*** (0.0261) 6.504*** (0.782) -2.040*** (0.0358)

-3.388*** (1.003) 15.43** (6.689) -89.17*** (6.591) -285.8* (157.1)

16.95 (54.43) -76.15 (374.6)

Event window (0,1) Tobit, Shares Purchased (4)

247,416 0.00677

131.8*** (6.653) 1,455*** (193.8) -512.8*** (10.87)

-3.487*** (1.005) 14.34** (6.706) -90.13*** (6.621) -290.4* (157.0)

0.772 (2.636) 22.97*** (4.918)

Event window (0,1) Tobit, Shares Purchased (5)

247,416 0.00682

132.0*** (6.659) 1,453*** (193.9) -513.7*** (10.96)

3.465 (3.113) 35.85*** (8.071) -3.348*** (1.004) 14.30** (6.700) -89.56*** (6.603) -295.1* (157.0)

Event window (0,1) Tobit, Shares Purchased (6)

247,416 0.00681

132.0*** (6.657) 1,450*** (193.9) -513.8*** (10.92)

247,416 0.0171

Number of observations Pseudo R2

247,416 0.0189

-2.532*** (0.0605) 247,416 0.0184

-2.532*** (0.0599) 247,416 0.00938

-312.7*** (10.60) 247,416 0.0102

-313.9*** (10.70) 247,416 0.0101

-314.0*** (10.64)

Constant

Market

Past Ret

Log(B/M)

Non-Customer* Cust Industry Firm SUE Negative Customer* Cust Industry Firm SUE Negative Non-Customer* Cust Industry Firm SUE<-0.3% Customer* Cust Industry Firm SUE<-0.3% Log(Size)

Panel A: Insider sales

-0.00835 (0.0152) -0.386*** (0.110) 0.629*** (0.0827) 4.647* (2.763) -2.037*** (0.112)

0.0474 (0.0402) 0.526*** (0.0981)

(1)

0.0495 (0.0428) 0.794*** (0.116) -0.00783 (0.0153) -0.402*** (0.110) 0.629*** (0.0826) 4.605* (2.760) -2.028*** (0.111)

(2)

-0.00586 (0.00482) -0.387*** (0.0361) 0.528*** (0.0266) 5.908*** (0.815) -2.002*** (0.0367)

0.00589 (0.0136) 0.182*** (0.0367)

(3)

47

0.0213 (0.0156) 0.200*** (0.0566) -0.00560 (0.00482) -0.387*** (0.0361) 0.528*** (0.0266) 5.884*** (0.815) -2.003*** (0.0365)

(4)

0.00278 (0.00465) -0.424*** (0.0353) 0.565*** (0.0258) 6.730*** (0.762) -2.004*** (0.0356)

-0.0354*** (0.0134) 0.100*** (0.0378)

(5)

-0.0257* (0.0155) 0.0716 (0.0612) 0.00290 (0.00465) -0.422*** (0.0353) 0.564*** (0.0258) 6.738*** (0.763) -2.009*** (0.0354)

(6)

-0.0574** (0.0259) -0.0400 (0.198) -0.0252 (0.147) 12.89*** (4.543) -1.454*** (0.208)

0.140* (0.0795)

(7)

0.155 (0.112) -0.0587** (0.0259) -0.0559 (0.199) -0.0324 (0.147) 12.56*** (4.551) -1.420*** (0.206)

(8)

This table presents robustness tests using probit regressions predicting insider trading at economically linked supplier firms as a function of their customers’ earnings announcements. The regressions in Panel A (Panel B) predict insider sales (purchases). In Columns 1 and 2 of Panel A, Sale Indicator is equal to one if there is at least one supplier insider sale during the window (0,1) days around an earnings announcement by a principal customer or other firm in the same industry Standard Industrial Classification (SIC) 4. In Columns 3 and 4 of Panel A, Sale Indicator is equal to one if there is at least one supplier insider sale during the window (1,2) days immediately following the day a principal customers or another firm in the same industry announced earnings. In Columns 5 and 6 of Panel A, Sale Indicator is equal to one if there is at least one supplier insider sale during the window (-2,-1) days preceding an earnings announcement by a principal customer or another firm in the same industry. Lastly, in Columns 7 and 8 of Panel A, Sale Indicator is equal to one if there is at least one supplier insider sale during the window (0,1) days around an earnings announcement by a principal customer. The explanatory variables of interest in Panel B are the insider purchase analog to those in Panel A. The observations in Columns 1 and 2 are limited to earnings announcements that occur on Fridays. The observations in Columns 7 and 8 of Panel A (Panel B) are limited to those where the insider made at least one sale (purchase) within the 20 trading days leading up to the earnings announcement. See Table 7 for earnings surprise measures and control variable definitions. Standard errors are in parentheses, and significance at the 1%, 5%, and 10% level are indicated by ***, **, and *, respectively.

Table 8 Robustness tests of supplier insider trading around customer earnings announcements

-2.520*** (0.0597)

Constant

(3.806) -3.078*** (0.171)

(0.158) -4.522 (3.785) -3.069*** (0.172)

Past Ret

Market

Constant

Friday release Days (0,1) 27,294 0.0292

(0.159) -4.467

(0.156) -0.814***

Log(B/M)

Friday release Days (0,1) 27,294 0.0232

(0.157) -0.812***

(0.0220) 0.255

Announcements sample Event window Number of observations Pseudo R2

(0.0221) 0.234

0.0467**

Friday release Days (0,1) 27,294 0.0252

0.174** (0.0679) 0.599*** (0.160) 0.0472**

0.0730 (0.0616) 0.334** (0.141)

Friday release Days (0,1) 27,294 0.0221

Non-Customer* Cust Industry Firm SUE Positive Customer* Cust Industry Firm SUE Positive Non-Customer* Cust Industry Firm SUE>0.3% Customer* Cust Industry Firm SUE>0.3% Log(Size)

Panel B: Insider Purchases

Announcements sample Event window Number of observations Pseudo R2

All Days (1,2) 247,416 0.0193

(0.0622)

(1.238) -2.586***

(0.0515) -2.226*

(0.0542) -0.735***

(0.00821) 0.200***

-0.0257***

0.0160 (0.0216) 0.223*** (0.0390)

All Days (1,2) 247,416 0.0147

48

All Days (1,2) 247,416 0.0187

(0.0616)

(1.239) -2.583***

(0.0514) -2.351*

(0.0541) -0.728***

(0.00821) 0.201***

0.0265 (0.0255) 0.317*** (0.0638) -0.0244***

All Days (1,2) 247,416 0.0145

All Days (-2,-1) 247,416 0.0165

(0.0585)

(1.172) -2.490***

(0.0481) -3.832***

(0.0519) -0.657***

(0.00773) 0.108**

-0.0257***

-0.00464 (0.0205) 0.224*** (0.0366)

All Days (-2,-1) 247,416 0.0173

All Days (-2,-1) 247,416 0.0152

(0.0578)

(1.173) -2.489***

(0.0479) -3.807***

(0.0518) -0.650***

(0.00772) 0.111**

-0.00651 (0.0247) 0.276*** (0.0627) -0.0245***

All Days (-2,-1) 247,416 0.0171

(0.488) Those with prerelease trading Days (0,1) 1,949 0.0337

Those with prerelease trading Days (0,1) 1,949 0.0280

(7.777) -1.662***

(0.386) -15.51**

(0.383) -0.521

(0.0647) 0.0954

0.271 (0.177) -0.0972

Those with prerelease trading Days (0,1) 4,633 0.0111

(0.491)

(7.716) -1.610***

(0.384) -14.63*

(0.383) -0.520

(0.0644) 0.123

-0.106*

0.0782 (0.145)

Those with prerelease trading Days (0,1) 4,633 0.0120