Do Institutional Investors Demand Public Disclosure?

Transcription

1 Do Institutional Investors Demand Public Disclosure? Andrew Bird and Stephen A. Karolyi July 1, 2015 Abstract We implement a regression discontinuity design to examine the effect of institutional ownership on corporate disclosure policy. With a novel, hand-collected data set encompassing every 8-K filing between 1996 and 2006, we find that positive shocks to institutional ownership around Russell index reconstitutions increase disclosure quantity, form, and quality. Firms just-added to the Russell 2000 index increase institutional ownership by 9.8%, and disclose 4.7% longer 8-K filings with 21.3% more embedded graphics. The increased disclosure is reflected in greater information content of 8-K filings as measured by analyst reactions, trading volume, and absolute announcement returns, which increase by 45.9%. We thank Alan Crane, Andy Koch, Sebastien Michenaud, Marina Niessner, and Tom Ruchti for valuable comments and discussion. All remaining errors are ours. Tepper School of Business, Carnegie Mellon University. Address: Posner Hall, 5000 Forbes Ave, Pittsburgh, PA, USA,

2 1. Introduction Do institutional investors demand corporate disclosure? A central question in finance and accounting is whether corporate transparency benefits or hurts investors. This question is further complicated by the fact that information provision could impact groups of investors differentially. Theoretically, public information may crowd out the private information advantage of some institutional investors (Gao and Liang [2013], Armstrong et al. [2012]). But these same institutional investors may benefit from an endogenous increase in broader market participation, which arises as a result of an increase in monitoring by other stakeholders (Bens [2002]). This tradeoff is reflected in mixed empirical evidence on the relationship between institutional ownership and corporate disclosure (Ajinkya et al. [2005], Bushee et al. [2003], Bushee and Noe [2000], Healy et al. [1999], O Brien and Bhushan [1990]). To address this tradeoff faced by institutional investors, we analyze the revealed preference for corporate disclosure by institutional investors and the associated impact on the information content of corporate disclosure. Empirically, identifying a causal effect of institutional ownership on corporate disclosure policy is difficult because experimental settings and direct measures of corporate disclosure quantity and characteristics are scarce. We propose a two part solution to these problems. First, we utilize exogenous changes in institutional ownership around Russell 2000 index reconstitutions in a regression discontinuity design to identify the effect of institutional ownership on corporate disclosure policy. Second, we directly measure the characteristics of corporate disclosure using a hand-collected data set comprised of all 8-K filings between 1996 and This experimental setting and novel data provide a platform to contribute to our knowledge about institutional investor demand for corporate disclosure along the dimensions of quantity, form, and quality. Our goal is to estimate the effect of an exogenous shock to institutional ownership 1

3 on corporate disclosure policy. We find that Russell 2000 inclusion has a positive impact on the quantity of disclosure. Relative to firms just-excluded from the Russell 2000 index, firms justincluded disclose 1.7% more 8-K filings and 2.0% more items per 8-K filing. Is this increase in disclosure the result of changes in other corporate activities, superficial window-dressing, or index reconstitution-induced attention? Whereas other papers have shown that corporate operating activities change in response to Russell index inclusion (Crane et al. [2014], Mullins [2014], Appel et al. [2015]), the difference in the disclosure cannot be fully explained by increases in corporate operating activity that requires mandatory disclosure for two reasons. First, 8-K filings of firms just-included to the Russell 2000 index include 21.3% more embedded graphics, are 4.7% longer, which we measure by the number of characters in the filing, and include 1.4% more exhibits compared to firms just-excluded. Second, we find that Russell 2000 inclusion has a positive impact on the quality of disclosure. The information content of 8-K filing announcements, as measured by absolute cumulative abnormal returns (CAR), increases by a factor of 1.46, on average. These results suggest that institutional investors demand more corporate disclosure and that these incremental disclosures contain valuable information. Furthermore, this average increase in stock return reaction is directly related to the increase in length, embedded graphics, and exhibits per 8-K filing. We first examine the relationship between 8-K characteristics and absolute CARs to document the importance of these characteristics. We find that the length and the number of items, graphics, and exhibits are all economically and statistically significant in explaining equity market reactions to 8-K filings. This result complements Lerman and Livnat [2010], which documents the information content of distinct items in 8-K disclosures. To control for the information content of different 8-K items, we include indicator variables for each item type and find that the information content of the length, graphics, and exhibits increases even within 8-K item types. 2

4 After controlling for these 8-K characteristics and the interaction of these 8-K characteristics with Russell 2000 index inclusion, the impact of Russell 2000 inclusion on absolute CARs is not as economically or, in some specifications, statistically significant. However, the interaction terms have statistically and economically significant positive effects, suggesting that the increase in the information content of 8-K filings is explained by the change in disclosure characteristics. For example, each additional embedded graphic corresponds to a 10.5 basis point increase in absolute CARs, on average. These results suggests that the increase in the information content of 8-K announcements is a result of changes in disclosure quantity and form, rather than index-membership-induced changes in investor attention. To abstract from the potential market microstructure effects of Russell 2000 index inclusions, we focus on two alternative measures to test for changes in the information content of 8-K filings. These measures are based on equity analyst forecast accuracy and forecast revisions, which provide real outcomes that are directly related to information revelation. Without new valuable information, analysts have no reason to revise their forecasts. Similarly, incremental disclosure of valuable information should allow analysts to make more accurate forecasts. Our analyst results are in line with our results on equity market reactions to 8-K filings. On average, the consensus forecast deviation from actual earnings-per-share (EPS) of firms just-included in the Russell 2000 index decreases by 16.2% relative to those just-excluded. Similarly, the number of forecast revisions per broker in the five days following an 8-K filing increases by 59.1% for firms just-included in the Russell 2000 index relative to those just-excluded. Using the same interaction methodology as in the absolute CARs tests, we find that the increases in analyst forecast accuracy and forecast revisions are directly related to the increase in length, embedded graphics, and exhibits per 8-K filing. These results suggest that the equity analysts react to the change in disclosure by making more forecasts and also more accurate forecasts. 3

5 In the equity market and analyst tests, when we allow the slope effect with respect to disclosure characteristics to differ, our estimates of the level effect of Russell 2000 index inclusion on information content become somewhat smaller. If viewed in isolation, these coefficient estimates might suggest that the increase in disclosure may crowd out the information content effect of Russell 2000 index inclusion. However, we only find that result when we allow the incremental information content to vary with the length, number of items, graphics, or exhibits. In these tests, for an average 8-K filing, the net effect of Russell 2000 index inclusion on the information content of 8-K filings remains statistically and economically positive, which suggests that the incremental length, items, graphics, and exhibits that 8-K filings of just-added Russell 2000 index firms compared to just-excluded firms contain incremental information content. We rely on the recent literature on Russell index membership that provides evidence of a discontinuity in institutional ownership around the Russell 1000/2000 index membership thresholds (Chang et al. [2015], Crane et al. [2014], Mullins [2014], Appel et al. [2015]). Russell index membership satisfies the key aspects of a regression discontinuity design because membership is based on the May 31 closing-price implied market capitalization rank. While the market capitalization rank discontinuities are public knowledge and consistent through time, the underlying market capitalization thresholds are time-varying and depend on the cross-sectional distribution of market capitalization at the end of the May 31 trading day. Firms in the top 1000 ranked market capitalization on that day become members of the Russell 1000 and the subsequent 2000, those ranked between 1001 and 3000, comprise the Russell Therefore, especially close to these market capitalization rank thresholds, Russell index reconstitutions are quasi-random with respect to corporate policies. Russell index membership is closely followed by institutional investors, particularly those who benchmark to one of the Russell indexes. Using the same index reconstitutions, Appel et al. [2015] finds that the increase in quasi-index investors improves governance, suggesting that even 4

6 institutional investors that are constrained to using governance through voice and not exit can have an impact on corporate governance. In particular, consistent with the stated governance objectives of large passive institutional investors 1, Appel et al. [2015] provides evidence that quasi-indexers increase the number of independent directors, remove poison pills and restrictions on shareholders ability to call special meetings, reduce dual-class share structures, and are less likely to support management proposals. Crane et al. [2014] use Russell index reconstitution-induced variation in total institutional ownership to identify the effect of institutional ownership on corporate payout policy. Moreover, Crane et al. [2014] also document a discontinuity in voting participation at the Russell 1000/2000 index threshold, supporting the notion that these institutional investors are not simply passive stakeholders from a governance perspective. Mullins [2014] provides further evidence that other blockholders, including insiders (i.e. an officer or director) and outsiders, are not displaced by these new institutional owners, which implies that retail investors, the group least likely to exert monitoring effort in governance, are the displaced shareholders. Our results suggest that the average increase in institutional ownership of approximately 10% causes a 2.0% increase in 8-K filings, a 2.1% increase in items filed, a 21.3% increase in embedded graphics, 4.7% increase in length, and 1.4% increase in the number of exhibits. Chang et al. [2015], Crane et al. [2014], Mullins [2014], and Appel et al. [2015] find that the discontinuity in institutional ownership around the Russell 1000/2000 index threshold is related to improvements in corporate governance along many margins. This paper departs from those in two important ways. First, we provide direct evidence that these changes in corporate policies, including corporate disclosure policy, which we study, are a result of a reduction in shareholders marginal cost of monitoring. This follows from our evidence that corporate disclosure increases in response to positive shocks to institutional ownership. Second, we ask whether this increase 1 Vanguard, 2014, Our views on corporate governance, 5

7 in institutional ownership improves one facet of governance-corporate disclosure policy-where it is not clear whether institutional investors would prefer stronger or weaker governance. Because disclosure improves corporate transparency to shareholders, increases in disclosure may improve market participation at the cost of crowding out the potential private information advantage of institutional investors. Consistent with institutional investors improving corporate governance, we provide evidence that institutional investors demand disclosure policies characterized by a larger quantity of disclosure with more content, graphical and otherwise, per disclosure. A contemporaneous paper, Boone and White [2015], investigates the relationship between institutional ownership and managerial guidance, a measure of voluntary disclosure, focusing on the effects of managerial guidance on market microstructure. Relative to that paper, we show not only that positive shocks to institutional ownership increase disclosure along the intensive margin of mandatory disclosure through 8-K filing frequency and characteristics, but also that the information content and consequences of these disclosures increase. In particular, we are focused on the production of valuable information whereas Boone and White [2015] speak more to symmetric information diffusion through trading environment outcomes such as quarterly average PIN and bid-ask spreads. In fact, because managerial guidance is likely to interact with mandatory corporate disclosure, the positive relationship between institutional ownership and managerial guidance found in Boone and White [2015] is likely to dampen our results on mandatory disclosure in 8-K filings. Managerial guidance, an aggregation of many pieces of information into a single dimensional forecast, is likely to contain some of the information that could otherwise be produced in 8-K filings. Furthermore, our analysis of 8-K filings as information events is crucial because it distinguishes the information produced as valuable. If the institutional ownership induced increase in 8-K filing frequency and characteristics had no incremental impact on the information content of disclosure, the change in 8-K filings could have been a result of managerial or institutional investor 6

8 obfuscation (Bushee et al. [2015]). The rest of the paper is organized as follows. Section 2 describes the data and sample selection. Section 3 discusses measurement and identification using Russell index reconstitutions. Section 4 analyzes the real effects of disclosure using equity market and analyst reactions to 8-K filings and Section 5 concludes. Supplementary results are presented in the Appendix. 2. Data Our empirical methodology depends upon changes in Russell 2000 index membership over time. Two approaches have been used in the literature to identify these index membership lists. First, because the Russell 1000 and 2000 indexes are explicitly determined by market capitalization rank as of the last trading day in May each year, the index memberships can be constructed using CRSP data. Second, Russell provides data for academic use. Because Crane et al. [2014] argue that predicting index inclusion for a fixed sample size can induce a bias from index misclassifications, we use index membership lists for the subset of reconstitutions covered by the Russell-provided data in robustness checks of our main results. An important caveat to this identification approach is that Russell changed their indexing methodology in 2007 by introducing banding. The banding policy smoothes over changes in index membership because it allows firms that should be excluded from the indexes to remain members if they are close to the May 31 market capitalization threshold. Because of this index methodology change and the introduction of consistent and recoverable 8-K filings, we restrict our sample period to begin in 1996 and end in To address the question of whether institutional investors demand corporate disclosure, we must construct proxies for the quantity and characteristics of corporate disclosure. We do this in two ways. First, we rely on prior literature to construct measures of financial reporting quality (Beatty 7

9 et al. [2010]), auditor choice (Behn et al. [2008], Khurana and Raman [2004]), adverse auditor opinions (Li et al. [2010], Ashbaugh-Skaife et al. [2009]), and accruals management (Demerjian et al. [2012], Yu [2008], Bergstresser and Philippon [2006]). These variables are readily constructed from Compustat data. Table 1 presents summary statistics of these measures for firms in the Russell 1000 and Russell 2000 indexes along with accompanying tests of difference in means between these two groups. These univariate tests reveal that firms in the Russell 2000 are less likely to use a Big Four auditor, less likely to receive an adverse opinion from their auditor, have lower absolute accruals as a fraction of assets, and have higher accounting quality as measured by Beatty et al. [2010]. At the firm-year level, 79.06% of Russell 1000 members use Big Four auditors compared to only 61.47% of Russell 2000 members. Second, we hand-collect a dataset of all 8-K filings between 1996 and 2006 that includes not only the number and timing of 8-K filings, but also their characteristics. The SEC requires publicly traded companies to make 8-K filings to report material events in a timely manner. Material events include entry into bankruptcy or receivership, the acquisition or disposition of assets, a change of auditor, and the departure or appointment of directors and executives, among others. Because the objective of the SEC is to promote the full public disclosure of relevant company information, and the 8-K, or the current report, has been the tool for firms to disclose timely information on important operational or financial changes between periodic financial reports, we believe 8-K filings are an ideal and direct measure of corporate disclosure. To reflect the quantity of corporate disclosure, we construct variables based on the number of 8-K filings, the length of each 8-K filing, the number and type of items reported in each 8-K filing, the number of exhibits included in each 8-K filing, and the number of graphics used in each 8-K filing. We posit that these variables reflect the quantity of corporate events disclosed as well as disclosure methods. Table 2 presents summary statistics of these measures for firms in the Russell 8

10 1000 and Russell 2000 indexes along with accompanying tests of difference in means between these two groups. These univariate tests reveal that firms in the Russell 1000 provide more content both in the quantity of 8-K filings and items that they disclose, but also in the form of this incremental disclosure. At the firm-month and filing levels, Russell 1000 members disclose longer 8-K filings and embed more graphics than Russell 2000 members. Per month, Russell 1000 members disclose K filings compared to 1.56 for Russell 2000 members. Similarly, Russell 1000 members embed 1.99 graphics per filing, about 0.55 graphics per filing more than Russell 2000 members. These differences in reporting quality and 8-K filing disclosure between Russell 1000 and Russell 2000 index members likely reflect differences in observable firm characteristics, but they highlight the importance of utilizing an experimental setting like ours to identify the effect of institutional ownership on disclosure. This is because the Russell 2000 index is more widely followed than the Russell 1000 index, so the cross-sectional relationship between institutional ownership and disclosure is opposite to the one we identify by focusing on changes in institutional ownership around the index threshold discontinuity. Finally, to determine whether changes in the quantity and characteristics of corporate disclosure have consequences for the information content of disclosure, we rely on data from CRSP on stock returns and trading volume and data from I\B\E\S on analyst forecast accuracy and revisions. 3. Index Reconstitutions, Institutional Ownership, and Corporate Disclosure 3.1 Identification Our goal is to estimate the effect of an exogenous shock to institutional ownership on corporate disclosure policy and the information content of disclosures. We observe several proxies for accounting 9

11 quality that have been highlighted in the previous literature and construct measures based on the quantity and style of 8-K disclosures using a novel hand-collected data set that includes information about every 8-K filed with the SEC between 1996 and We provide evidence, consistent with the literature on Russell 1000 and Russell 2000 membership, of a discontinuity in institutional ownership around the index membership thresholds. Russell index membership satisfies the key aspects of a regression discontinuity design because membership is based on the May 31 closingprice implied market capitalization rank. While the market capitalization rank discontinuities are public knowledge and consistent through time, the underlying market capitalization thresholds are time-varying and depend on the cross-sectional distribution of market capitalization at the end of the May 31 trading day. Firms in the top 1000 ranked market capitalization on that day become members of the Russell 1000 and the subsequent 2000, those ranked between 1001 and 3000, comprise the Russell Therefore, especially close to these market capitalization rank thresholds, Russell index reconstitutions are quasi-random with respect to corporate disclosure policy. We estimate the following model: IO i,t = α + τd i,t + f(r i,t ) + πx i,t + v t + ξ i,t (1) Y i,t = β 0 + β 1ÎO i,t + g(r i,t ) + β 2 X i,t + ω t + ε i,t (2) where v t represents year-month fixed effects, D i,t is an indicator variable that equals one if firm i is a Russell 2000 index member in year t and zero otherwise, R i,t represents the market capitalization rank of firm i in year t minus 1,000, IO i,t represents the fraction of firm i s shares outstanding owned by institutions in year t, and Y i,t represents measures of financial reporting quality from the literature on corporate disclosure and based on our database of 8-K filings. X i,t includes a set of time-varying firm characteristics as controls. We include year-month fixed effects to remove the possibility that the results are being driven by secular changes in corporate disclosure or 10

12 disclosure regulation. 2 The functions f and g are parametrized as k-order polynomials in the market capitalization ranking, R i,t, on either side of the Russell 1000/2000 threshold. The parameter k is chosen to maximize the Bayesian information criterion (BIC) as in Lee and Lemieux [2010]. It is important that we acknowledge the methodological differences among the other papers that exploit the Russell index reconstitution setting for identification. In the results presented here we follow Crane et al. [2014] and and impose a k-order polynomial control function on the market capitalization ranks provided and used by Russell, but, in untabulated results we confirm that our main findings hold if we alternatively impose a k-order polynomial control function on the market capitalization ranks that we calculate using May 31 closing prices from CRSP as in Appel et al. [2015]. 3 Intuitively, this system of equations ensures that the variation in institutional ownership that we use to identify our coefficient of interest, β 1, comes from Russell index reconstitutions. Because the effect of index reconstitutions on disclosure policy operates through the institutional ownership channel, we utilize a fuzzy regression discontinuity design. Technically, the change in disclosure policy happens stochastically with respect to the threshold; an observed change in disclosure policy around the rank threshold does not happen with certainty, which is the differentiating requirement that a sharp regression discontinuity design would require. A fuzzy regression discontinuity design requires a two stage least squares design as in equations (1) and (2), which means that we can reinterpret the Russell 2000 dummy as an instrumental variable. Thus, the conditions for instrument validity must be satisfied. We observe relevance in Figure 1, which plots the discontinuity in institutional ownership around the rank threshold of All of the results that follow are robust to inclusion of firm fixed effects to mitigate selection on time-invariant, unobservable firm characteristics. This further means that effects on disclosure are similar even when restricting the identifying variation to come only from firms that switched indexes over the sample. 3 This is, however, a slightly different empirical specification than that of Boone and White [2015] in that the k-order polynomial control functions are based on market capitalization ranks rather than rank bins. 11

13 Firms just to the left of the rank threshold are included in the Russell 1000 and firms just to the right of the threshold are included in the Russell Beyond the statistical relevance of the the rank threshold, relevance is satisfied because of two institutional features of the Russell indexes. First, approximately twice as much institutional investor money is invested in the Russell 2000 compared to the Russell 1000 (Chang et al. [2015]). Second, Russell indexes are value-weighted, meaning that a firm just-included in the Russell 1000 index will have a much lower index weight than a firm just-excluded from the Russell 1000 index. These two features suggest that the discontinuity in institutional ownership around the May 31 market capitalization rank threshold of 1000 is a result of the Russell index methodology. Valid instruments must also satisfy the exclusion criterion. In our case, exclusion is satisfied because other observables are locally continuous at the rank threshold. That is, inclusion in the Russell 1000 or Russell 2000 should not impact disclosure policy directly. Rather, its effect on disclosure policy exists only because of its effect on institutional ownership. Because the fuzzy regression discontinuity design is appropriate to our setting, the smooth local polynomial plots we present in Figures 2-7 are reduced-form in the sense that they abstract away from the change in institutional ownership at the rank threshold. In these figures, we can only interpret the jump in disclosure around the rank threshold as a result of institutional ownership because we also observe the discontinuity in institutional ownership at the rank threshold, which is shown in Figure 1, and because of the aforementioned features of the Russell indexes. These two features suggest that the discontinuity in institutional ownership around the May 31 market capitalization rank threshold of 1000 is a result of Russell index methodology. Note that the two stage empirical strategy employed in our formal tests directly connects the variation in institutional ownership due to index reconstitutions with changes in disclosure policy at the same firm. Table 3 presents estimates of τ from equation (1) in which k = 0, 1, and 2 for parsimony. 4 We 4 We investigate up to k = 10 and find similar results. 12

14 rely on Crane et al. [2014] for evidence of a discontinuity in institutional ownership across institution types. Our evidence is consistent with theirs in that we find strong evidence of a discontinuity in institutional ownership around the Russell 1000/2000 threshold. Regardless of k, our estimates of the Russell 2000 inclusion effect on institutional ownership at the Russell 1000/2000 threshold are statistically significant at least at the 5% level. For k = 2, the polynomial form chosen by the BIC, we find that firms just-included in the Russell 2000 index have a 9.8% increase in institutional ownership relative to those just-included in the Russell 1000 index. Russell index membership is closely followed by institutional investors, particularly those who benchmark to one of the Russell indexes. Among firms just-excluded from the Russell 1000 index, total institutional ownership increases by 10%. This is not particularly surprising given the prevalence of benchmarking investment performance against indexes, and, in particular, the greater institutional following of the Russell 2000 index relative to the Russell 1000 index, as documented by Chang et al. [2015]. Furthermore, because the Russell indexes are value-weighted, the index weight applied to firms at the top of the Russell 2000 index have a much larger weight in that index than firms at the bottom of the Russell 1000 index (Chang et al. [2015]). Mullins [2014] provides further evidence that other blockholders, including insiders and outsiders, are not displaced by these new institutional owners, which implies that retail investors, the group least likely to exert monitoring effort in governance, are the displaced shareholders. Moreover, Crane et al. [2014] also document a discontinuity in voting participation at the Russell 1000/2000 index threshold, suggesting that these institutional investors are not simply passive stakeholders from a governance perspective. These results provide causal evidence that Russell 2000 index membership has an economically large impact on institutional ownership and activity. This evidence complements arguments from Azar et al. [2014] and McCahery et al. [2014], which suggest that even small, apparently passive institutions engage in corporate governance. This 13

15 means that even changes in dedicated indexer ownership are likely to impact corporate policies. 3.2 Reporting Quality Measures from the Literature Among the three groups of disclosure proxies, the first group includes variables that are consistent with the literature on financial reporting quality and can be constructed using Compustat data from annual 10-K filings. These variables are Big4 i,t, an indicator that equals one if firm i has a Big Four auditor in year t and zero otherwise 5, AudSwitch i,t, an indicator that equals one if firm i switches from any auditor to another in year t and zero otherwise, AdvOpinion i,t, an indicator that equals one if firm i s auditor issues an Adverse opinion on the effectiveness of the firm s internal control over financial reporting and zero otherwise, AbsAcc i,t is firm i s absolute value of accruals scaled by total assets in year t, and AccQuality i,t is the measure of accounting quality proposed by Beatty et al. [2010] for firm i in year t. Table 4 presents evidence from proxies for corporate disclosure policy from the prior literature. The estimation results suggest that the average increase in institutional ownership associated with Russell 2000 inclusion causes a 3.3% increase in the probability of choosing a Big 4 auditor, a 2.4% marginal increase in the probability of switching auditors, a 1.3% decrease in the probability of having an adverse opinion, and 2.8% lower absolute abnormal accruals as a percentage of total assets. Additionally, we find a statistically insignificant positive relationship between Russell 2000 index inclusion and accounting quality, as defined in Beatty et al. [2010]. Together, this evidence suggests that, relative to firms just excluded, firms just added to the Russell 2000 index seek out new and better auditors, are less likely to have adverse auditor opinions, and decrease their use of earnings management. 5 We exclude Arthur Andersen from our Big4 i,t measure because our sample period spans 1996 to 2006, including the post-scandal period.. 14

16 3.2 Measuring Corporate Disclosure with 8-K Filings The second group of measures are based on our hand-collected data set of 8-K filings. For a panel of firm-year-month observations, we track the number of 8-K filings in a given year-month, or Num8kF ilings i,t, the cumulative length of these filings, NumCharacters i,t, the number of separate items disclosed in these filings, NumItems i,t, the number of graphics provided in these filings, NumGraphics i,t, and the number of exhibits in these filings, NumExhibits i,t. With these variables, our intention is to capture both the quantity and the richness of each firms reporting environment, so that we can assess not only how much but also what kinds of disclosures institutional investors demand from corporations. These estimation results in Panel A of Table 5 suggest that a 10% increase in institutional ownership increases the number of 8-K filings by 0.37 per year, the length of filings by 60,609 characters per year, the number of embedded graphics by 3.63 per year, and the number of items and exhibits increase by 0.67 and 0.42 per year, respectively. 6 These estimates suggest that the exogenous increase in institutional ownership at the Russell 2000 inclusion threshold is associated with a higher quantity of corporate disclosure, as measured by the number of 8-K filings and the content within these 8-K filings. The third group of measures also use our hand-collected data set of 8-K filings, but instead of focusing on absolute changes in the characteristics of all filings within a given month, we focus on the changes in the average filing. This analysis provides evidence as to whether firms are changing the quantity of disclosure on the extensive margin or the intensive margin. The evidence in Panel B of Table 5 suggests that not only does the quantity of disclosure increase, but also the amount of information contained in each filing increases for firms just-added 6 We arrive at these estimates by multiplying the coefficients from Panel A of Table 5 by 10% to yield the per-month change in 8-K filing characteristics, and then multiplying them by 12 to annualize the estimates. For example, the coefficient on institutional ownership for the number of graphics regression is Thus, the per-month increase in graphics is , which, when annualized, yields an increase of graphics per year. 15

17 to the Russell 2000 index relative to firms just excluded. The length of each filing increases by 4,181 characters, the number of items increases by 0.04, the number of graphics by 0.21, and the number of exhibits by These estimates are economically significant because they are (i) close in magnitude to the monthly results, and (ii) represent marginal increases on the order of 2% to 15%. In particular, the increase in the length and number of embedded graphics per filing of approximately 12% and 15%, respectively, suggests that the increase in corporate disclosure cannot be fully explained by increases in the number of corporate events that involve mandatory disclosure that may also be associated with increases in institutional ownership. Table A2 of the Appendix investigates the time series properties of this change in 8-K filing frequencies and amounts. We explore the potential for heterogeneous effects before and after the passage of Regulation Fair Disclosure (Reg. FD). The results are uniformly larger economically, but statistically different post-reg. FD only for length and the number of graphics. If Reg. FD decreased the prevalence of private information, these results are consistent with institutional investors incrementally demanding public disclosure to offset this lost private information channel. We also test for differences before and after 2004 when the SEC changed the structure of mandatory 8-K filings to make the items more granular. We find no economic or statistical differences across these two subperiods. The results in the previous three tables are consistent with four inferences about institutional investor demand for corporate disclosure policies. First, as a shock to institutional investor holdings, index membership increases the quantity of corporate disclosures. Second, to the extent that Big Four auditors have a better auditing technology, this shift in disclosure quantity is accompanied by higher internal reporting scrutiny, evidenced by the switch to Big Four auditors and drop in adverse opinions. Third, the change in corporate disclosure policy is not restricted to the total quantity of information, but also the quantity of information per 8-K filing and variety of information 16

18 transmission tools utilized. This third inference is supported mainly by the results presented in the Panel B of Table 5, but also to a lesser extent by the traditional measures of reporting quality in Table 4. Fourth, given the observation that the control group in the Russell 2000 index tests includes Russell 1000 index members, the magnitude of the economic and statistical effects for Russell 2000 index membership represent a marginal effect on corporate disclosure policy relative to firms already included in the Russell 1000 index. 4. Real Effects of Disclosure 4.1 Equity Market Reactions to Corporate Disclosure In this section, we examine the real effects of the shift in corporate disclosure policy due to institutional investor demand. For this analysis, we focus on two dimensions: 8-K announcement returns and trading volume. Because we are concerned with the effect of changes in corporate disclosure on 8-K announcement returns, we focus on the magnitude of the 8-K announcement returns rather than the direction. Thus, our first dependent variable of interest is abs(car i,t ), the three day, market-adjusted cumulative abnormal returns around firm i s 8-K announcement on date t. 7 Our second dependent variable of interest has to do with trading volume around the 8-K announcement date. We construct ln AbnormalV olume i,t, the natural log of the three day, cumulative abnormal trading volume of firm i s 8-K announcement on date t. 8 Do equityholders care about the information contained in 8-K filings? We first examine the relationship between abs(car i,t ), AbnormalV olume i,t, and our measures of 8-K filing character- 7 In untabulated results using five day cumulative abnormal announcement returns, we find consistent results, which suggests that the information is not simply being incorporated into prices more quickly due to differential investor attention. 8 Abnormal trading volume is defined as the residual from a regression of daily trading volume on firm-by-month fixed effects. These residuals can be interpreted as the abnormal daily trading volume that cannot be attributed to firm-specific factors that vary at the year-month frequency. 17

19 istics. In the full sample of 1,373,778 8-K filings between 1996 and 2006, we estimate the following regression: Y i,t = α + βx i,t + u i + v t + ɛ i,t (3) where u i and v t represent firm and year-month fixed effects, Y i,t represents abs(car i,t ) or ln AbnormalV olume i,t, and X i,t represents a vector of 8-K filing characteristics that includes NumCharacters i,t, NumItems i,t, NumGraphics i,t, and NumExhibits i,t. The results presented in Table 6 suggest that the characteristics of corporate disclosure have an incremental relationship with the information content of 8-K filings. The length and the number of items, graphics, and exhibits in 8-K filings are positively correlated with the magnitude of 8- K announcement returns and abnormal trading volume, on average. That the return response is increasing in the number of items and exhibits is not surprising because each of these correspond to unique information events. However, the positive relationship between the number of graphics, length, and absolute CARs suggests that the amount of information about each event and the graphical transmission of information increases the information content of 8-K filings. 9 We next focus on the difference in stock market reactions to 8-K announcements between firms just included and excluded from the Russell 2000 index. By imposing our discontinuity restriction and controlling for the absolute rank distance from the index inclusion threshold, we preserve our regression discontinuity design in the context of announcement returns and trading volume. We estimate the following regression on the sample of 1,373,778 8-K filings between 1996 and 2006: Y i,t = α + β IOÎO i,t + β X X i,t + u i + v t + ɛ i,t (4) where u i and v t represent firm and year-month fixed effects, respectively, ÎO i,t is the instrumented institutional ownership of firm i in year-month t, and Y i,t represents abs(car i,t ) or 9 Appendix Table A1 replicates the results in Table 6 using day-of-week fixed effects. This alleviates the concern that the information content of 8-K filing characteristics depends on strategic disclosure timing (Niessner [2015]). 18

20 ln AbnormalV olume i,t. X i,t includes 8-K filing and firm characteristics as controls. The previous model identifies the average effect of exogenous changes in institutional ownership associated with Russell 2000 index inclusion on announcement returns and abnormal trading volume. In the following model, we allow this average effect to vary with the change in corporate disclosure policy found in Section 3. In the same sample of 8-K filings, we estimate: Y i,t = α + β IOÎO i,t + β Z Z i,t + β Interaction IO i,t Z i,t + β X X i,t + u i + v t + ɛ i,t (5) where Z i,t represents four measures that proxy for the breadth and depth of disclosure, ln NumCharacters i,t, NumItems i,t, NumGraphics i,t, and NumExhibits i,t. In Panel A of Table 7, we present abs(car) i,t coefficient estimates for β IO, β Z, and β Interaction for each of the four proxies for breadth and depth of disclosure. We instrument for the endogenous interaction term, IO Z i,t, by interacting Z i,t with D i,t, R j i,t, and D i,t R j i,t for each j chosen by the BIC in the first stage. On average, increases in institutional ownership associated with Russell 2000 index inclusion increase the information content of 8-K filings. Announcement returns increase in magnitude by 6.6 bps, on average. Furthermore, the estimation results reflect the two channels by which corporate disclosure policy impacts the information content of 8-K filings. First, part of this increase in 8-K announcement returns is due to changes in the amount of content included in each filing. The interaction results suggest that the per-filing return reaction is smaller, on average, for firms justincluded in the Russell 2000 index relative to those just-excluded, but that the return reaction is incrementally larger for filings with more information events. Second, the other part of the magnitude increase in announcement returns is that the filings contain more content per information event and more graphical content, reflected by the length and number of graphics, respectively. For example, for an average increase in institutional ownership of 9.8%, each additional graphic is associated with an additional 10.5 bps of return reaction. 19

21 We next focus on the incremental effect of Russell 2000 inclusion on abnormal trading volume around 8-K filings. In Panel B of Table 7, we present ln AbnormalV olume i,t coefficient estimates for equations 4 and 5. These estimation results are consistent with the hypothesis that, on average, inclusion in the Russell 2000 index increases the information content of 8-K filings. Abnormal trading volume increases by 2.9%, on average. As in the return reaction test above, the results reflect the two channels by which corporate disclosure policy impacts the information content of 8-K filings. Part of this increase in abnormal trading volume around 8-K announcements is related to an increase in the number of information events contained in each filing and part can be attributed to the increase in content per information event and graphical content. Table A3 of the Appendix investigates the time series properties of this effect. We explore the potential for heterogeneous effects before and after the passage of Regulation Fair Disclosure (Reg. FD). The results are uniformly larger economically, but statistically different post-reg. FD only for selective measures. If Reg. FD decreased the prevalence of private information, these results are consistent with private information crowding out or otherwise discouraging public disclosure. We also test for differences before and after 2004 when the SEC changed the structure of mandatory 8-K filings to make the items more granular. We find no economic or statistical differences across these two subperiods. The results in this section provide causal support for the effect of two corporate disclosure channels on equity market reactions to corporate disclosure. Combined with the results from Section 3, these results suggest that exogenous increases in institutional ownership are related to (i) more corporate disclosure, (ii) higher quality corporate disclosure, (iii) larger equity market reactions to corporate disclosures, and (iv) these equity market reactions are due to changes in the number of corporate events disclosed and the tools utilized to disclose information about corporate events. From these results, we can infer that institutional owners demand greater corporate disclosure, 20

22 which in turn impacts equity market returns and trading volume around corporate announcements. 4.2 Analyst Forecasts and Revisions Prior literature has argued that increases in corporate disclosure are reflected in analyst forecast accuracy and revisions. Given our regression discontinuity setting, we have the opportunity to investigate the causal effect of changes in corporate disclosure on analyst forecasts and revisions. Furthermore, because the stock market results in the previous section are subject to market microstructure concerns related to index inclusion, analyst forecast accuracy and revisions provide real outcomes that are directly related to information revelation. Without new valuable information, analysts have no incentive to revise their forecasts. Similarly, incremental disclosure of valuable information should allow analysts to make more accurate forecasts. In this section, we examine the effects of the shift in corporate disclosure policy due to institutional investor demand on equity analysts. For this analysis, we focus on analyst forecast accuracy and the number of forecast revisisons per broker in response to 8-K filings. As in the previous section, it is important to first establish that equity analysts condition on the information content of 8-K filings, so we start by examining the relationship between Deviation i,t, NumRevisions i,t, and our measures of 8-K filing characteristics. In the full sample of 1,373,778 8-K filings between 1996 and 2006, we estimate equation (3) again, where Y i,t represents Deviation i,t or NumRevisions i,t and report extimation results in columns (5)-(8) of Table 6. These results suggest that the characteristics of corporate disclosure have an incremental relationship with the information content of 8-K filings. The length, number of items, graphics, and exhibits in 8-K filings are positively correlated with the number of forecast revisions per broker in the five days after 8-K filings and negatively correlated with the absolute deviation of analyst EPS forecasts to actual EPS realizations. On average, an additional graphic embedded in 8-K filings is 21

23 associated with 1.6 cents smaller forecast deviation from actual EPS realizations and more forecast revisions per broker in the five days following 8-K filings. Having established that the characteristics of 8-K filings have information content for equity analysts, we turn to the question of whether the index inclusion shock to institutional ownership increases or decreases the incremental information content of 8-K filings and their characteristics for equity analysts. We estimate equation (4) in the full sample of 59,247 firm fiscal quarters between 1996 and 2006 with analyst forecast accuracy and revisions as the dependent variables of interest. Table 8 presents the estimation results. The analyst response to 8-K filings is consistent with higher information content of these filings for firms just-added to the Russell 2000 index relative to those just-excluded. On average, 8-K filings of firms just-added to the Russell 2000 index are associated with a 1.4 cent smaller deviation from actual EPS realizations compared to firms justexcluded. Because the average deviation from actual EPS realizations is approximately 5.7 cents in our sample period, the average decrease of 1.4 cents corresponds to a 24.6% decrease. Similarly, on average, 8-K filings of firms just-added to the Russell 2000 index are associated with a 0.32 more revisions per broker in the five days following the filing relative to firms just-excluded. Given that the average number of forecast revisions per broker in the five days following 8-K filings is 0.72 during our sample period, the average increase of 0.32 forecast revisions per broker corresponds to a 44.4% increase. That the average equity analyst response to 8-K filings is larger for firms just added to the Russell 2000 index than those just excluded is consistent with institutional owners demanding more informative corporate disclosures. But is this increase in the information content of corporate disclosures a result of the changes in disclosure characteristics that we observe? Consistent with the equity market results in the previous section, we find that equity analysts react to the incremental content per filing. We re-estimate equation (5) with analyst forecast accuracy and revisions as the 22

24 dependent variables of interest and show the estimation results in Table 8. These results reveal that part of the average increase in analyst reactions to 8-K filing is due to changes in the amount of content disclosed in each filing. The interaction results suggest that the per-filing analyst reaction is larger, on average, for firms just-included in the Russell 2000 index relative to those just-excluded, but also that the return reaction is incrementally larger for filings with more information events. Moreover, analyst reactions also increase because filings contain more content per information event and more graphical content, reflected by the length and number of graphics, respectively. While analyst forecasts provide an environment to examine the index-inclusion-induced variation in the information content of 8-K filings which is less susceptible market microstructure changes, it does not come without potential confounds. In particular, analysts present potential alternative explanations of our results. The common theme of these alternatives is that they depend upon equity analysts having a preference for public information. For example, because analysts and institutional investors interact, the possibility exists that analysts demand disclosure from firms as a mechanism to improve their service to institutional investor clients. This explanation is consistent with the institutional investor monitoring mechanism we describe, but in this explanation, it is proactive analysts that interact with corporations as agents of disclosure for their institutional clients. We believe that these alternatives are unlikely given the growing set of evidence that analysts do not prefer more public disclosure to less. Armstrong et al. [2012] show that analyst following decreases in response to plausibly exogenous increases in financial statement informativeness. In another setting, Xie [2013] shows that though analysts issue public forecasts and recommendations, they provide additional private signals about these recommendations to their best institutional clients, which suggests that analysts extract private information rents from institutional investors. Similarly, Barth et al. [2001] show that analyst following is higher for firms with more intangible assets, which provide more profitable opportunities for private information collection. 23