Advertising Investments, Information Asymmetry, and Insider Gains

Transcription

1 Accepted Manuscript Advertising Investments, Information Asymmetry, and Insider Gains Kissan Joseph, M. Babajide Wintoki PII: S (13)00011-X DOI: doi: /j.jempfin Reference: EMPFIN 650 To appear in: Journal of Empirical Finance Received date: 30 May 2012 Revised date: 26 November 2012 Accepted date: 12 February 2013 Please cite this article as: Joseph, Kissan, Wintoki, M. Babajide, Advertising Investments, Information Asymmetry, and Insider Gains, Journal of Empirical Finance (2013), doi: /j.jempfin 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 proof before it is published in its final 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.

2 AdvertisingInvestments,InformationAsymmetry,andInsiderGains KissanJoseph a,m.babajidewintoki a, Abstract a SchoolofBusiness,UniversityofKansas,Lawrence,KS66045,USA Extant research has documented various sources of informational advantages enjoyed by company insiders including firm size, analyst following, dividend payout policy, book-to-market ratio, and the presence or absence of R&D investments. Surprisingly, despite this large body of work, virtually no research has investigated the contribution of advertising investments to information asymmetry. This omission is particularly glaring since:(a) advertising investments constitute a significant fraction of many firms ongoing expenditures, and(b) the received literature provides strong theoretical arguments relating advertising investments and information asymmetry. Accordingly, the primary objective in this study is to empirically address this gap. Usingadvertisingandinsidertransactiondataatover12,000firmsfrom ,wefindthat insider gains are significantly greater at firms characterized by advertising investments. Specifically, a zerocost portfolio that is long on firms with net insider purchases and advertising investments, and short on firms with net insider purchases and devoid of advertising investments, garners annual abnormal returns of 5.5%. In addition, we find that investors reaction to news of insider purchasing is significantly more pronounced at firms characterized by advertising investments investors rationally recognize the greater information content associated with insider purchases at these firms. JEL Classification: G1; G3 Keywords: Advertising, Insider trades, Information asymmetry, Insider gains WewouldliketothankPaulKoch,seminarparticipantsatUniversityofKansas,andthe2010meetingoftheEasternFinance Association, for helpful comments on earlier drafts. Any errors of analysis and interpretation are our own. Correspondingauthor:M.BabajideWintoki;Tel.:+1-(785) ;Fax:+1-(785) addresses: kjoseph@ku.edu(kissan Joseph), jwintoki@ku.edu(m. Babajide Wintoki) Preprint submitted to Elsevier November 26, 2012

3 1. Introduction Corporate insiders generate signals when they trade in the equities of their own firms. Lakonishok and Lee(2001) examine companies that trade on the NYSE, AMEX, and Nasdaq markets during the period markets and observe that insiders generate plenty of signals, about 50,000 trades per year (p. 80). These signals are a major part of how market participants form their opinions of firm value; indeed, insider trades are a key channel by which information is transmitted from the black box of the firm to outside investors whose trading activity determines the price of shares of the firm. A large body of research (e.g., Lorie and Niederhoffer, 1968; Jaffe, 1974; Finnerty, 1976; Seyhun, 1986; Bettis, Vickrey, and Vickery, 1997; Jeng, Metrick, and Zeckhauser, 2003) has documented that insiders gain from trading in their firms shares because of informational advantages. To date, the specific sources of information asymmetry that have been explored include firm size (Seyhun, 1986; Lakonishok and Lee, 2001), analyst following (Frankel and Li, 2004), dividend payout policy (Khang and King, 2006), book-to-market ratio (Huddart and Ke, 2007), and the presence or absence of R&D expenditures (Aboody and Lev, 2000; Huddart and Ke, 2007). Surprisingly, despite this large body of work, virtually no research has examined the contribution of advertising investments to information asymmetry. This omission is particularly glaring since advertising investments constitute a significant fraction of many firms ongoing expenditures. Specifically, between 2004 and 2008, aggregate advertising expenditure for all COMPUSTAT firms was twice as large as aggregate R&D expenditure, which is a previously documented source of information asymmetry between insiders and outsiders. 1 The omission of advertising investments is also glaring because received literature (Chauvin and Hirschey, 1993; Hirschey, 1982; Lodish, Abraham, Kalmenson, Livelsberger, Lubetkin, Richardson, and Stevens, 1995a; Lodish, Abraham, Livelsberger, Lubetkin, Richardson, and Stevens, 1995b) reveals that advertising investments can, in fact, endow insiders with 1 Between 2004 and 2008, the ratio of advertising to sales for all COMPUSTAT firms was 1.2% while the ratio of R&D to sales was 0.6% (Hirschey, Skiba, and Wintoki, 2012). 1

4 substantial informational advantages. Accordingly, our primary objective in this research is to empirically address this gap. Our empirical findings strongly suggest that advertising expenditure is indeed an important source of information asymmetry between insiders and outside investors. Specifically, we find that a zero-cost portfolio that is long on firms with net insider purchases and advertising investments, and short on firms with net insider purchases and devoid of advertising investments, garners annual abnormal returns of approximately 5.5%. In addition, abnormal returns following disclosures of stock purchases by insiders are significantly larger in firms that advertise than in firms that do not advertise. Notably, this incremental return enjoyed by insiders at advertising firms relative to their counterparts at non-advertising firms holds even over longer windows such as six and twelve months. Our empirical strategy in this paper is three-fold. First, we compare the magnitude of insider gains at firms with positive reported advertising investments (Advertising firms) to the magnitude of insider gains at firms with no reported advertising investments (No- Advertising firms). In our initial analysis, we track the performance of all individual insider purchases, both within and across firms. We find that both the mean as well as the median return following insider purchases, adjusted for market returns, is significantly higher at Advertising firms as compared to No-Advertising firms. Moreover, these findings extend over three different time intervals: from transaction date to SEC filing date, from transaction date to six months following the transaction date, and from transaction date to twelve months following the transaction date. Next, we augment these initial findings by aggregating insider purchases within a firm. Here, we compose portfolios comprising of firms where insiders were net purchasers during the period (i.e., shares purchased exceeded shares sold). We then employ a strategy of going long on insider purchases in Advertising firms and short on insider purchases at No-Advertising firms. Doing so, we again find that the mean and median returns at Advertising firms are significantly greater than the mean and median returns at No-Advertising firms. Moreover, the level of analysis here also allows us to demonstrate that these results are robust to the inclusion of other risk factors. Specifically, we find that the difference in returns persists 2

5 even when we include the well-known risk-factors encompassed in the three-factor model of Fama and French (1993). The differences in returns are also robust to the inclusion of the momentum risk-factor suggested in Carhart (1997). These results thus corroborate our initial findings by revealing that the informational advantages possessed by insiders at Advertising firms are greater than those possessed by insiders at No-Advertising firms even after controlling for well-known risk-factors. That is, advertising investments are not driving increased returns because of their possible association with other key characteristics such as firm size, growth opportunities, and momentum. The third component of our empirical strategy is to compare the short-term investor reaction to announcements of insider purchasing at firms characterized by advertising investments to the short-term investor reaction to announcements of insider purchasing at firms characterized by no advertising investments. This analysis is designed to demonstrate that outsiders correctly anticipate insider purchases to be more informative at firms characterized by advertising investments; consequently, they respond more strongly to insider purchases at such firms. Accordingly, we examine three time intervals around the date of the filing of insider purchases at the SEC: the day of the filing, a two-day return, and a threeday return. We find that the mean (market-adjusted) returns are significantly higher at Advertising firms as compared to No-Advertising firms for all three time intervals. In effect, external investors rationally recognize the greater information content of insider purchases at firms characterized by advertising investments and consequently respond more strongly to disclosure at insider purchases at such firms. We believe our findings offer implications for both practice and scholarship. With regard to practice, it establishes the value of utilizing investments in advertising investments as an additional screen when following a strategy of mimicking insiders. Specifically, the incremental long-term returns portended by insider purchases at advertising firms relative to their counterparts at firms devoid of advertising can potentially guide investment practice. The identification of such a guideline is clearly one important contribution of our work. In terms of scholarship, our work sheds important new light on the nature and sources of information 3

6 asymmetry between insiders and outsiders. Specifically, we empirically verify the theoretical argument that advertising investments can generate information asymmetries. The rest of the paper is organized in the following manner. In the next section, we review the related literature and formally state our research hypothesis. Then we discuss our data, variables, and our empirical findings. We then demonstrate that our results are robust to the consideration of other investments such as R&D that may generate informational asymmetry. We also show that our results are robust to including other sources of information asymmetry in a cross-sectional analysis. For completeness, we also analyze insider sales and confirm that these are not as informative as insider purchases. Finally, we provide a summary, discuss alternative explanations, and outline the implications of our research endeavor. 2. Literature review and hypotheses development In this section, we provide a brief review of the literature and formally state our hypotheses. Our brief review of the literature is divided into two parts. We first review the extant literature on the insider gains. We then summarize research which examines the information asymmetries that might drive such gains. We begin by reviewing the evidence that documents the gains from insider trading Gains from insider trading Researchers in the finance literature have documented that corporate insiders, defined by the 1934 Securities and Exchange Act as corporate officers, directors, and owners of 10% or more of any equity class of securities, gain from trading in the securities of their firms. However, estimates of the gains from insider trading vary widely. Early studies (Lorie and Niederhoffer, 1968; Jaffe, 1974; Finnerty, 1976) report abnormal gains ranging from 3% to 30% for holding periods up to three years. Seyhun (1986) finds more modest gains to insiders; over 300 days subsequent to insider trading, the average risk-adjusted gains were 4.3% for stock purchasers and 2.3% for sellers. Finally, Jeng et al. (2003) find insider gains on purchases of 6% annually and insignificant abnormal returns for insider sales. Note that across all the studies discussed, there seems to be a more pronounced effect 4

7 for insider purchases of stock compared to insider sales of stock. This seems reasonable in light of the following observation: insider purchases are an unambiguous signal of intent to purchase an asset that is believed to be appreciating in value. In contrast, insider sales are at best an imprecise signal of intent to dispose an asset that is believed to be depreciating in value insider sales could also reflect a need for liquidity or diversification. For example, while Chowdhury, Howe, and Lin (1993) argue that aggregate insider trades have a small economically significant relation to subsequent stock returns, they show that this effect is entirely driven by insider purchases of stock rather than insider sales. Indeed, Lakonishok and Lee (2001), among others, formalize this observation when they write: Only insider purchases appear useful, while sales are not associated with low returns (p. 109). Given this understanding, the focus in our subsequent empirical analysis centers mostly on examining returns associated with insider purchases Information asymmetry Aboody and Lev (2000) note that all corporate investments create information asymmetries. This is because insiders can continually observe changes in investment productivity on an individual asset basis whereas investors can only view highly aggregated information on investment productivity at discrete points in time. However, the extent of information asymmetry is likely to vary across various investments. For example, the extent of information asymmetry associated with R&D is likely to be higher than tangible investments in property, plant, and equipment (PP&E) because of the relative uniqueness (idiosyncrasy) of R&D. For instance, efficacy trials of a pharmaceutical agent or alpha testing of software are difficult to evaluate for profit impact because the import of such developments is hard to gauge for outsiders. Moreover, the absence of organized markets for R&D contributes further to information asymmetry; whereas investors can derive estimates of firm PP&E assets from prices of comparably traded assets in organized markets, there is no direct price-based information on firm-specific changes in the value and productivity of R&D investments. Finally, 2 In additional analysis in Section 3.7, we examine and confirm that are no cross-sectional differences related to advertising in returns following insider sales. 5

8 the practice of expensing R&D further obscures the current value and changes in current value of R&D investments. Aboody and Lev (2000) use these observations to motivate their investigation of insider gains in a specific context, namely, between firms that do invest in R&D and those that do not invest in R&D. They find that insider gains at R&D firms are substantially larger than insider gains at firms without R&D. They thus conclude that R&D is a major contributor to information asymmetry between insiders and external investors. A careful review of the marketing literature suggests that similar arguments can be developed for advertising investments. Like R&D, the response to advertising is unique to a particular firm and industry (idiosyncratic). Indeed, the meta-analysis of Lodish et al. (1995a), conducted across many industries and many time periods, reveals that: (i) there is no simple correspondence between increased advertising expenditures and increased sales, regardless of whether the increased spending is compared to competition or not, (ii) advertising is more likely to work when there are changes in brand/copy/media strategy, and (iii) standard measures of recall and persuasion are unlikely to impact the sales effectiveness of the copy. Lodish et al. (1995a) conclude that there are many aspects of advertising sales effectiveness that seem to be unique to a particular brand, competitive situation, copy strategy and execution, and media strategy and tactics (p. 138). These observations all make a good case for the theoretical prediction that internal officers of the firm can better assess the productivity of ongoing advertising investments. Logically, this is because insiders obtain more information pertaining to the idiosyncratic impact of advertising investments via initial sales reports, future customer orders, and the like. Additionally, advertising investments are often characterized by long-term payoffs. Lodish et al. (1995b), for example, report substantial long-term sales effects associated with advertising. Specifically, the long-term effect of advertising, when successful, is on average approximately double the initial impact although a given campaign may exhibit higher or lower impact. 3 Since insiders are privy to continuous information on the sales impact of advertising whereas outsiders obtain such information only at discrete points in time (earn- 3 The data interval associated with this conclusion is a year. 6

9 ings announcements, investor conferences, and the like), the long-term payoffs associated with advertising further exacerbates the difficulty faced by external investors in assessing the productivity of advertising. Finally, like R&D, there is no direct price-based information on firm-specific changes in the value and productivity of advertising investments. As noted by Hirschey (1982) and Chauvin and Hirschey (1993), advertising leads to the creation of intangible capital that has valuation implications but is difficult to value. Specifically, there are no price-based markets that assist in discerning the goodwill associated with advertising. In addition, akin to R&D, the practice of expensing advertising further obscures the current value, and changes in current value, of advertising investments. Of course, there is a rich literature on other informational aspects of advertising. Grullon, Kanatas, and Weston (2004), for example, report that firms that advertise have shares that are more liquid and have smaller bid-ask spreads. They attribute this to the fact that advertising attracts more local small-scale investors to the firm. We argue that this finding speaks to the fact that advertising reduces informational asymmetries across external investors, rather than between insiders and outside investors. Thus, in summary, the idiosyncratic, long-term, and non-traded aspects of advertising imply that substantial information asymmetries can be generated on account of advertising investments between insiders and external investors Research hypotheses Following our conceptualization that advertising endows insiders with additional informational advantages and the well-established finding that informational advantages provide insiders with superior returns, we are now in a position to formally state our research hypotheses. We partition the impact of informational advantages on returns into two temporal windows. Here, we rely on Kyle s (1985) theoretical model to posit that trades by the informed trader do get incorporated into prices gradually even though not all information is incorporated into prices by the end of trading. Accordingly, we first examine the impact of insider purchases from the time of purchase to just before its formal announcement to the 7

10 market. Next, we delineate the reaction of the market following the announcement of insider purchases. For either window, we expect a more pronounced effect for purchases at firms with advertising investments relative to those without advertising advertising investments because of the heightened information asymmetry at such firms. Following these arguments, we formally state our hypotheses as follows: H1: Market reaction following insider purchases but prior to its formal announcement will be more pronounced at firms that advertise than at firms that do not advertise, and H2: Market reaction to insider purchases after its formal announcement will be more pronounced at firms that advertise than at firms that do not advertise. 3. Data, variables and empirical findings 3.1. Data and variables The insider trading data for this study is taken from the Thomson Reuters Insiders database. It covers all insider transactions reported to the SEC on Forms 3, 4, 5 and 144 between January 1, 1986 and December 31, We focus on 409,734 insider purchase transactions over this period made by inside officers of firms. 4 We exclude all transactions where the total number of shares involved exceed 20% of the shares outstanding as these may include change of corporate control events. We eliminate 31,129 duplicate transactions and 103,697 transactions for which we were unable to find any returns data on CRSP or firm data on COMPUSTAT. This leaves us with a sample consisting of 284,908 insider purchases in 12,822 firms. 4 These were actual open-market purchases and we eliminate transactions that involved conversion of a derivative security (e.g. stock options) into a non-derivative security, or exercises of stock options. The definitions of insiders (and their Thomson Reuters codes) included in our sample are as follows: Chairman of the Board (CB), Chief Executive Officer (CEO), Chief Operating Officer (CO), General Counsel (GC), President (P), Chief Financial Officer (CFO), Chief Investment Officer (CI), Chief Technology Officer (CT), Officer, Director and Beneficial Owner (H), Managing Director (MD), Officer (O), Officer and Beneficial Owner of more than 10% of a Class of Security (OB), Officer and Director (OD), Officer of Parent Company (OP), Officer of Subsidiary Company (OS), Officer and Treasurer (OT), Divisional Officer (OX), Treasurer (TR), Vice Chairman (VC), Senior Vice President (SVP), Vice President (VP). 8

11 Our empirical analysis and presentation closely matches that of Aboody and Lev s (2000) study of insider gains in the context of R&D expenditures. In Table 1, we present the distribution of insider purchases. We classify firms as Advertising or No-Advertising based on whether they reported advertising expenditures during the period January 1986 December 2011; the latter are those for which COMPUSTAT does not report any advertising expenditures. A firm is classified as an Advertising firm in a particular fiscal year if it reported advertising expenditure in a that fiscal year and it is classified as a No-Advertising firm if it does not. 5 This binary classification scheme is consistent with that employed by Aboody and Lev (2000) in the context of R&D. Table 1 reveals that the total number of insider transactions, total number of shares, and total value of transactions are all higher at No-Advertising firms. This is simply a reflection of the greater number of No-Advertising firms in our sample; as such, no inference is associated with these aggregate numbers Insider gains and advertising Table 2 reports mean and median market-adjusted returns (raw minus the return on a value-weighted NYSE/AMEX/Nasdaq index) for three data intervals subsequent to insider transactions: from transaction date to 1 day before the SEC filing (an average of 16 trading days in our sample); and 6 and 12 months following the transaction. 6 For all three data intervals, the mean as well as the median returns are higher for Advertising firms compared to 5 Disclosure of advertising expenditure is guided by the AICPA SOP 93-7 Reporting on Advertising Costs. SOP 93-7 requires that generally, the costs of all advertising should be expensed either in the periods in which those costs are incurred or the first time the advertising takes place. However, it allows firms to capitalize advertising expenses under the stringent exception that the advertising is direct-response advertising (a) whose primary purpose is to elicit sales to customers who could be shown to have responded specifically to the advertising, and (b) that results in probable future economic benefits (future benefits). This means that there may be some firms which advertise but do not expense or report advertising expenditures. Thus, our Advertising sample is a far more conservative representation of firms that actually advertise than is our No-Advertising sample is of firms that do not advertise. We believe this actually biases us against finding any significant information asymmetry differences between our Advertising and No-Advertising samples. 6 Since the time interval between the date of insider purchase and the date on which the purchase is reported to the SEC varies across purchases, we find the average daily return for each purchase, over the period between the transaction date and report date interval, and multiply this average daily return by 16 days, which is the average interval between the transaction date and report date in our sample. 9

12 No-Advertising firms; moreover, these differences are all statistically significant. In addition, the difference in mean market-adjusted returns for the three periods are also economically significant they amount to 1.07% over 16 days, 5.90% over 6 months, and 9.56% over 12 months ( , , and , respectively). We find similar returns when raw returns are adjusted for the returns of a typical firm in the same size, book-tomarket and momentum quintile, where the benchmarks for the quintiles are as described in Daniel, Grinblatt, Titman, and Wermers (1997) and Wermers (2004). It is worth noting that the excess returns, when properly accounting for the length of time, are highest for the shortest time horizon, a period when the information asymmetry is presumably the largest. It is also worth noting that the excess returns are skewed in that the mean returns are always higher than the median returns at both Advertising and No-Advertising firms. In fact, median returns are sometimes zero or even negative. This skewness suggests that advertising investments do not perfectly identify opportunities for insider gains and there are a cluster of purchases at which the returns are substantially positive. Such skewness is also reported by Aboody and Lev (2000) in the context of R&D investments. Consequently, it appears to be endemic to the nature of gains associated with insider purchases. Skewness notwithstanding, the returns data in Table 2 clearly reveal that insider gains in Advertising firms are higher than insider gains in No-Advertising firms. However, the underlying transaction data are not independent (there are multiple transactions per firm), and the returns over the six and twelve month periods are overlapping. Moreover, firm attributes (risk, size) related to advertising investments may also influence the documented returns. Therefore, the returns data in Table 1 should be viewed as descriptive. In further analysis, we aggregate insider transactions by firms and control for known risk factors. Accordingly, in Table 3, we focus on the returns in the period between the time the insider makes the purchase transaction (transaction date) and the time the company files with the SEC and makes the transaction known to the public (filing date) while controlling for systematic risk factors that explain expected returns. We form portfolios as follows: for each trading day between January 1, 1986 and December 31, 2011, we calculate the mean 10

13 returns of a portfolio of all firms where insiders are net purchasers of shares in the firm. Each firm is kept in the portfolio from the transaction date until the trading day before the filing date provided the insiders of the firm remain net purchasers up to the day before SEC filing. Multiple transactions per firm reported to the SEC on the same date are considered here as one transaction and the firm is added to the portfolio on the day of the first insider purchase for that period. 7 Thus we have equally-weighted portfolios, which are re-balanced daily when a new insider purchase enters the sample. We calculate the mean returns separately for firms with Advertising (ADV) and without Advertising (N OADV). To calculate the difference in the returns of the two portfolios we also calculate the daily mean returns of a portfolio that is long the shares of firms with Advertising and those without Advertising (ADV NOADV). To account for systematic risk factors, we report (in Panel A) the results from the regression of the daily time series of returns on three factors from Fama and French (1993): the excess return on the market (R m R f ); the return difference between a portfolio of small and big stocks (SMB), and the return difference between a portfolio of high and low book-to-market stocks (HM L). These factors have been found to explain cross-sectional differences in stock returns (see for example, Fama and French, 1993; Kothari and Warner, 2008). Thus, our results in Panel A are obtained by carrying out the following regression: R p,t R f,t = α+β m (R m,t R f,t )+β s SMB t +β h HML t +ǫ it (1) where R p,t R f,t is the return of the portfolio on day t, adjusted for the risk-free rate. In Panel B, we report the results from the regression of the daily time series of returns on R m R f,smb and HML, augmented with a momentum factor from Carhart (1997), which is the return difference between a portfolio of stocks with high returns in the past year and a portfolio of stocks with low returns in the past year (UMD). In both Panels A and B, α is the daily abnormal return (in percentage terms). Thus, our results in Panel B are 7 This means that if there are any insider sellers after the first insider purchase but the number of shares sold remains less than the number of shares bought, the firm remains in the portfolio. 11

14 obtained by carrying out the following regression: R p,t R f,t = α+β m (R m,t R f,t )+β s SMB t +β h HML t +β u UMD t +ǫ it (2) The implied 16-day return is calculated as (1+α) 16 1, where 16 days is the average number of trading days between the transaction date and the SEC filing date. Consistent with our findings from Table 2, the abnormal daily returns from portfolios of insider purchases are both positive and significantly different from zero in all specifications. So, for example, in the four-factor model of Panel B in Table 3, this is true both for Advertising firms (α = , t = 14.02), as well as No-Advertising firms (α = , t = 16.05). This abnormal daily returns translates to 16-day returns of 1.81% and 1.46% for the Advertising and No-Advertising firms, respectively. (Very similar insights emerge from perusing the three-factor model reported in Panel A; consequently, we do not elaborate on them). To examine the extent to which insiders are able to exploit the higher level of information asymmetry in firms that employ advertising, we examine the abnormal returns from a longshort portfolio that buys the firms that advertise and shorts the firms that do not advertise (ADV NOADV). A key benefit of the long-short portfolio is that it is a portfolio with zero cost the long purchase is offset by the short sale. The results in Table 3 show clearly that the abnormal return here is positive and significant. So, for example, in the four-factor model of Panel B, the daily abnormal return, α is (t = 2.50), which corresponds to an implied 16-day return of 0.34%. To further put this in an economic context, the % excess daily return that insider purchases in Advertising garner relative to insider purchases in No-advertising firms translates to an annual return of 5.5% a year or 280% return over the entire sample period of 1986 to (As before, very similar insights emerge from perusing the three-factor model reported in Panel A of Table 3; consequently, we do not elaborate on them). 12

15 3.3. Investor reaction to the disclosure of insider purchases Next, we examine investors reactions to the disclosure of insider purchases. If advertising enhances information asymmetry and rational investors are aware of this asymmetry, we expect the market to respond more positively to announcement of insider purchases in firms with advertising than those without. Towards this end, Table 4 reports the market reaction tothedisclosureofinsiderpurchases. Day0isthedatewhenthefilingoftheinsiderpurchase was received by the SEC; this is the date when information about the insider purchase is fully revealed to the public. Days (0, +1) refers to the returns on day 0 and the following trading day. Days (0, +1, +2) is the three-day return from the day the filing was received by the SEC. There are 36,749 insider purchases for Advertising firms and 82,492 transactions for No Advertising firms. The total number of insider purchases is smaller in this Table compared to corresponding numbers in Table 1 (85,113 and 199,795 respectively) because multiple transactions per firm reported to the SEC on the same date are considered here as one transaction. This means that even if an insider makes purchases on numerous days within a particular period, it is considered as one transaction if they are all reported on the same day and in the same SEC filing. Again, the smaller number associated with Advertising firms is accountable to their smaller numbers in our sample; as such, no inference is associated with it. Perusing Table 4, we find that the market reaction to the disclosure of investor purchases is much more pronounced at Advertising firms as compared to No-Advertising firms. Specifically, across all three time horizons, namely, Day 0, Day (0, +1), and Day (0, +1, +2), the mean market reaction is significantly higher at Advertising Firms. Across the three time intervals, the differences are 0.07%, 0.23%, and 0.30% (corresponding to , , and , respectively). A similar picture arises when comparing the difference in median returns, although the differences are significant only in the Day (0, +1) and Day (0, +1, +2) time horizons. As a point of comparison, the aforementioned difference in the mean returns are comparable in magnitude to those reported by Aboody and Lev (2000) in the context of R&D the corresponding numbers there are 0.16%, 0.16%, and 0.19%. The 13

16 market reaction to insider purchases (in the ten day period around the announcement) is illustrated in Figure 1. The figure shows that while there is in general a positive announcement reaction to all insider purchases, the reaction is significantly higher in firms that advertise compared to those that do not. Overall, the evidence reveals that investors react more strongly to public disclosures of insider trades at Advertising firms than No-Advertising firms, consistent with the notion that advertising investments contribute to information asymmetry Robustness tests One possible criticism of our analysis is that our results are driven not by differences in information asymmetry arising from the presence(or absence) of advertising but rather by the presence of other sources of information asymmetry associated with internal firm investments, whose presence is correlated with the presence of advertising. Indeed, careful analysis of our sample indicates that 63% of the insider transactions involving firms with advertising also involve firms that have R&D expenditure. As we have noted earlier, R&D expenditure itself has been shown to significantly contribute to information asymmetry between insiders and outside investors (Aboody and Lev, 2000). To ascertain that our results are not confounded by the information asymmetry arising from R&D expenditures, we carry out our tests of insider gains and investors reactions to the disclosure of insider purchases in a sample of firms that have been purged of any firms with R&D expenditure. Table 5 shows the difference in insider gains in Advertising and No-Advertising firms (purged of R&D firms) over the period between the transaction and its public revelation via an SEC filing. The results are similar to those obtained in Table 4. While the absolute 16-day abnormal returns from the Advertising and No-Advertising samples are comparable to those in Table 4, the relative difference between the samples is actually bigger. So, for example, using the four-factor model (Panel B of Table 4), a portfolio that s long the shares of Advertising firms, and short the shares of No-Advertising firms yields a daily abnormal return of α = (t = 3.29), which translates into an 16-day return of 0.63%. This is somewhat higher than the 16-day return of 0.34% obtained in the sample of firms characterized by both R&D 14

17 and Advertising investment. This empirical finding confirms that while R&D may be a major source of information asymmetry between insiders and outside investors, advertising investments make a contribution to information asymmetry that is distinct from that arising from R&D. Table 6 shows the investors reaction to the announcement of insider purchases in the Advertising and No-Advertising sample purged of R&D firms. Again, the results are similar to those obtained in Table 4 with the broader sample that includes firms with both R&D and Advertising investments. Investors react more strongly to disclosures of insider purchases in the Advertising firms than they do to disclosures of insider purchases in the No-Advertising firms. The results strongly suggest that outside investors recognize the nature of information asymmetry inherent in advertising expenditures beyond whatever information asymmetry is due to R&D expenditures. We offer another perspective on our empirical findings by pictorially highlighting the distribution of insider purchases and firm-level abnormal returns across firms in four distinct states: (1)FirmsthatinvestonlyinAdvertising,(2)FirmsthatinvestonlyinR&D,(3)Firms that invest in both Advertising and R&D, and (4) Firms that invest in neither Advertising nor R&D (Please see Figure 2). Specifically, Figure 2 reports the number of insider purchases at firms in each one of these four states as well as the abnormal firm-level returns (α s) from the four-factor regression. Figure 2 also reports the 16-day abnormal returns for firms in each one of these four states. So, for example, R 1 is the abnormal return from holding a portfolio of firms that invested in advertising and characterized by a net of insider purchasing. The findings displayed in Figure 2 reveal that firms characterized by insider purchases all yield positive abnormal returns. In addition, the long-short portfolio in Table 3 reports the difference between the weighted average of R 1 and R 3 (Advertising) and the weighted average of R 2 and R 4 (No-Advertising). The long-short portfolio in Table 5 (the robustness check) reports the difference between R 1 (Advertising and No-R&D) and R 4 (No-Advertising and No-R&D). 15

18 3.5. Cross-sectional analysis We complement our portfolio analysis from the previous sections with cross-sectional analysis of the effect of advertising investments on insider gains. The cross-sectional analysis enables us to directly control for those firm characteristics, or characteristics of the firm s environment, that may be correlated with advertising but have been suggested in prior literature as being sources of information asymmetry between insiders and outsiders. These factors include firm size (Seyhun, 1986; Lakonishok and Lee, 2001), analyst following (Frankel and Li, 2004), dividend yield (Khang and King, 2006), market-to-book ratio (Huddart and Ke, 2007), and the presence or absence of R&D expenditures (Aboody and Lev, 2000; Huddart and Ke, 2007). We include two additional variables in our cross-sectional regressions. First we include industry concentration. There is empirical evidence that industry concentration affects affects advertising expenditure; Willis and Rogers (1998) find that advertising-to-sales ratios are highest in industries with the highest concentration. However, industry concentration may itself affect the firm s information environment through its effect on a firm s disclosure policy. While disclosure may improve a firm s information environment, such disclosure may put the firm at a competitive disadvantage (see, for example, Darrough and Stoughton (1990)). Along these lines, Ali, Klasa, and Yeung (2012) report empirical evidence that firms in more concentrated industries disclose less information. Thus, ignoring industry concentration in any regression of insider returns (or returns around disclosure of insider trades) on advertising may lead to an omitted variable bias. Finally, in order to assess the incremental explanatory power of advertising in explaining insider returns, we include a measure of the market reaction to previous earnings announcements; Huddart and Ke (2007) suggests that this is a good proxy for other unobservable aspects of information asymmetry. Our empirical proxies, all of which are measured at the fiscal year-end prior to the SEC filing date of the insider transaction, are as follows: Size is the log of the market value of equity; analyst following is the number of analysts who provided a fiscal year-end earnings estimate (from the IBES database); dividend yield is the firm s total cash dividend divided 16

19 by market value of equity; market-to-book is the sum of the market value of equity and the book value of assets minus the book value of equity, all divided by the book value of assets; a dummy variable (R&D DUM) that equals one for firms that report R&D expenditure on COMPUSTAT, and zero otherwise; industry concentration which is measured by a Herfindahl index constructed from the distribution of COMPUSTAT sales in a firm s two-digit SIC code industry; and LN(Median AR) which is the median magnitude of the cumulative abnormal returns from two days before to the day of the quarterly earnings announcement date for the prior five fiscal years. As before, our key variable of interest, ADV DUM, equals one for firms that report advertising expenditure on COMPUSTAT, and zero otherwise. We examine the effect of this variable on two dependent variables: (i) insider gain which is measured from the time of the transaction and (ii) the cumulative return around the SEC filing data, or public announcement, of the insider purchases. In all our cross-sectional regressions, we include two-digit SIC industry dummies. It is also possible that the residuals for a given firm can be correlated across years (serial correlation), or that the residuals of a given year may be correlated across different firms (cross-sectional dependence); both of these could bias the standard errors downwards and inflate the t-statistics (Petersen, 2009). Thus, to avoid biased inferences, we report robust t-statistics that are clustered by firm and year in all our regression results. Table 7 shows the results from OLS regressions of insider returns on advertising and other potential sources of information asymmetry. In column (1), the dependent variable is the market adjusted return from the date of the insider transaction to the date when the information about the insider purchase is made publicly available via an SEC filing. The results show that over this period, stocks in firms with insider purchases return an average of 1.08% (t = 2.02) more in firms that advertise than in those that do not advertise even after controlling for other potential sources of information asymmetry that may be correlated with advertising. This difference is similar in magnitude to what we obtained from the univariate analysis of Table 2 and the portfolio analysis of Table 3. We see a similar result in column (2) of Table 6 in which the dependent variable is the market adjusted return in the six-month 17

20 period following the insider purchase: insider purchases at firms that advertise are followed by returns that are 1.54% (t = 2.11) more than the returns to insider purchases at firms that do not advertise. Table 8 shows the results from OLS regressions of investor reaction (CAR around SEC filing dates) on advertising and other potential sources of information asymmetry. In column (1), the dependent variable is the two day (0, +1) CAR around the SEC filing date. Here again the results are in line with those we obtained from the portfolio analysis: investors react more strongly to the disclosure of insider purchases at firms that advertise than they do to those from firms that do not advertise. The two-day CAR around SEC filing is, on average, 0.09% (t = 1.72) higher in firms that advertise than in those that do not. Similarly, in column (2), we show that the three-day CAR around SEC filing is, on average, 0.14% (t = 3.28) higher in firms that advertise than in those that do not. Thus, we find that investors respond more strongly to announcements of insider purchases at advertising firms even after controlling for other potential sources of information asymmetry that may be correlated with advertising Does the level of advertising matter? Here, we investigate whether the level of advertising impacts the extent of information asymmetry. Accordingly, in Table 9, we first sort firms into quintiles based on the ratio of advertising-to-sales, with the lowest quintile (Q1) having the lowest ratio and the highest quintile(q5)havingthehighestratio. Afirmisincludedintheportfolioonaccountofinsider purchasing and it is kept in the portfolio from the transaction date until the trading day before the SEC announcement. We then calculate abnormal returns for firms by quintilies. We also regress insider gains on various sources of information asymmetry as well as the ratio of advertising-to-sales. These results are reported in Panels A and Panel B, respectively. The results in Panel A of Table 9 reveal that the the daily abnormal return, α, does not vary by the level of the advertising-to-sales ratio. Moreover, the abnormal return associated with Q5 Q1 is statistically insignificant. In addition, the findings in Panel B of Table 9 reveal that the ratio of advertising-to-sales does not contribute to explaining the abnormal 18

21 return associated with insider purchases. These findings suggest that the amount of information asymmetry generated by advertising is not driven by the size of the investment; rather, as suggested Lodish et al. (1995a), it is the idiosyncratic features such as competitive situation, copy strategy and execution, and media strategy and tactics that determine advertising effectiveness, and consequently, generate information asymmetry The effect of advertising on returns following insider sales Thus far, we have focused our analysis on returns following insider purchases. As we noted in Section 2.1, insider sales are, at best, an imprecise signal of the insider s belief of the value of the firm; selling may simply arise for liquidity or diversification. Nevertheless, it is possible that there may be cross-sectional differences related to advertising in returns following insider sales. We examine this possibility in Table 10. Panel A of Table 8 presents the four factor daily abnormal returns for the period between transaction date and the (day before) filing date for firms where insiders were net sellers. The results show that the abnormal returns are insignificantly different from zero for both the Advertising and No-Advertising portfolios. The results also show that the difference between the Advertising and No-Advertising portfolios is not significant. Panel B of Table 8 shows the two-day and three-day market reaction to the disclosure of insider sales via SEC filing. While the market reaction is significantly negative for both Advertising and No-Advertising firms, here again we find no difference between the two sets of firms. 4. Summary, discussion, and future research 4.1. Summary In this investigation, we find that insider gains are significantly greater for insider purchases at firms that invest in advertising relative to insider purchases at firms devoid of advertising, even after controlling for other factors that may be associated with advertising 19

22 such as firm size, growth opportunities, and momentum. Specifically, a zero-cost portfolio that is long on firms with net insider purchases and advertising investments and short on firms with net insider purchases and devoid of advertising investments garners annual abnormal returns of 5.5%. Clearly, this is an important and economically significant difference. We also find that investors reaction to news of insider trading is significantly more pronounced at Advertising firms external investors rationally recognize the greater information content associated with these purchases. We also examine if the excess returns are simply an artifact of the fact that firms that invest in advertising are also those that invest in R&D. Accordingly, we redo our analysis on a reduced sample purged of all firms that invest in R&D. Our conclusions survive this robustness check. In additional cross-sectional analysis, we find that advertising is significantly related to insider gains and investor reaction to insider purchases, even after controlling for other sources of information asymmetry that may be related to advertising. These include size, market-to-book ratio, R&D, analyst following and dividend yield. Our conclusions also survive the assumption of homoscedasticity assumed in estimating our returns equation. Finally, consistent with previous research, we find that insider selling is not informative in that there are no differences between advertising and non-advertising firms when we examine returns following insider sales Discussion Here, we elaborate on another explanation that may be driving our results. We conceptualize that advertising creates information asymmetry. An alternative conceptualization is that advertising increases investor attention, which manifests itself in a stronger market reaction to signals emanating from the firm (see, for example, Barber and Odean, 2008). In this connection, recall that our results span three time periods. In the first row of Table 2, and via the portfolio tests in Table 3, we display findings from the time of purchase to the day of the filing. Clearly, the higher returns associated with insider purchases at firms characterized by advertising investments cannot be attributed to greater attention because the insider activity is yet to be reported. Rather, the effect is consistent with Kyle s (1985) notion that informed purchasing transfers information from the firm to market participants. 20