Asymmetric information flows (PNF)

Transcription

1 Asymmetric information flows (PNF) Vasiliki Athanasakou London School of Economics Houghton Street, London, WC2A 2AE, UK Norman C. Strong Manchester Business School Booth Street West, Manchester, M15 6PB, UK Martin Walker Manchester Business School Booth Street West, Manchester, M15 6PB, UK We thank seminar participants at International Hellenic University, Ivey Business School (University of Western Ontario), London School of Economics, Nottingham Business School, Stirling Management School, the Research Day in Accounting (IESEG School of Management), the Multinational Finance Society 2015 Annual Conference, the Canadian Academic Accounting Association 2015 Annual Meeting, and the American Accounting Association 2015 Annual Meeting.

2 Asymmetric information flows (PNF) ABSTRACT We develop a measure of asymmetric information flows (PNF) based on negative and positive abnormal stock returns to capture strategic withholding of bad relative to good news and, thereby, an inverse measure of the quality of a firm s corporate information environment. We find that PNF is positively related to factors associated with the scope and managerial incentives for strategically withholding bad news, and negatively associated to withholding constraints. PNF is also related to lower earnings quality and a higher risk of shareholder litigation. These results suggest that PNF captures strategic withholding of bad news. As such, it usefully complements research that directly scores the quantity and quality of corporate disclosures. JEL codes: G20; G30; M40, M41 Keywords: bad news; good news; corporate information environment; withholding; information quality 1

3 1. Introduction Corporate news flows shape a firm s information environment and form a critical part of the price discovery process. Assessing how companies communicate corporate news to capital markets has gained considerable regulatory attention, especially with regard to the properties of corporate disclosures (Hoberg and Lewis 2015). A challenge in assessing firms communication strategies is observing all corporate flows. The information economics literature often uses abnormal stock returns to proxy for firm-specific information flows. Evidence attributing substantial variation in abnormal stock returns to firm disclosures (Beyer et al. 2010, p.300) supports this theoretically driven approximation. Stock returns offer a comprehensive measure of corporate information flows, allowing assessments of flows beyond observable financial narratives. This is an important advantage given that technological change has facilitated the emergence of various forms of continuous communication (Miller and Skinner 2015) and managers have shown an increasing preference for private communication channels (Bushee, Gerakos, and Lee 2014). In this paper we develop a return based measure of corporate disclosure quality that complements existing measures of the quality of narrative disclosures for assessing the firm s information environment. As we develop a return based measure of disclosure quality, we investigate properties of stock returns that reflect corporate disclosure policy, focusing on disclosure policies that trigger adverse consequences, e.g., shareholder litigation. In a typical litigation, a plaintiff s allegation refers to the firm s failure to disclose material adverse information, i.e., withholding bad news (Francis et al. 1994). There is extensive theoretical and empirical literature on managerial incentives to asymmetrically release bad relative to good news. We investigate how stock returns reflect this asymmetry. We start by assuming, similar to prior literature, that managers receive 2

4 information randomly during the accounting period and that the information arrival process is not systematically different between good and bad news (Acharya et al. 2011; Kothari, Shu, and Wysocki, 2009; King and Wallin, 1996). Given this assumption, if managers promptly disclose information, its dissemination should generate symmetrically distributed stock returns, that is, the average magnitude of positive returns to good news should be equal to that of negative returns to bad news (Kothari et al. 2009, p.246). If, however, managers act strategically and accumulate and withhold bad news up to a threshold level while promptly disclosing or even accelerating good news, which we refer to henceforth as strategically withholding bad news, this creates an asymmetry in the distribution of stock returns. This return asymmetry reflects the asymmetric flows of bad versus good news, i.e., lumpier flows of bad news compared to regular flows of good news, and the market reaction to the asymmetric releases. 1 We develop a measure to capture this asymmetry in the distribution of stock returns that captures the flows of negative versus positive returns and their relative magnitude. Our measure, PNF, captures the difference in the flows of positive versus negative returns. Higher values of PNF indicate a higher concentration of negative than positive returns or a higher magnitude of negative than positive returns. We calculate PNF for each firm financial year. We believe that corporate disclosures drive sufficient variation in abnormal stock returns during the financial year (e.g., about 30% of stock return variance occurs on days when there are observable corporate disclosures) for stock returns to reflect systematic patterns associated with firms disclosure policies. In this respect, while we are mindful of several market-based explanations of an asymmetry in the distribution of stock returns (e.g., investor sentiment, investor asymmetric learning, volatility feedback effects), we share the common belief that the asymmetry is, at least partly, due to corporate 1 Kothari et al. (2009) predict that managers asymmetric disclosure behavior leads to an asymmetric price reaction to news releases. The main argument for an asymmetric market response is the market reaction to the withholding of bad news versus the frequent leaking of good news. 3

5 disclosure policies (Kalev, Liu, Pham, and Jarnecic, 2004) and particularly that of disclosing bad news (Hutton, Marcus and Tehranian 2009, Acharya, De Marzo, and Kremer, 2011; Kothari et al. 2009, Shin, 2003, 2006). To explore the validity of PNF in capturing the strategic withholding of bad news and, therefore, in serving as an inverse measure of corporate disclosure quality, we develop hypotheses about its association with known determinants and outcomes of such a corporate disclosure policy. In line with our predictions we document a positive association between PNF and factors associated with the scope for strategic withholding of bad news, i.e., business volatility and growth. We also document a positive association between PNF and managerial incentives to strategically withhold bad news, e.g., poor operating and stock market performance, equity and debt issues, and insider ownership. At the same time we document a negative association between PNF and constraints on withholding bad news such as public visibility and institutional investor trading frequency (investor turnover). Looking at other facets of financial reporting quality, we document a positive association between PNF and common inverse measures of reported earnings quality. Finally in terms of outcomes we find that, as expected, PNF is significantly higher for firms subject to a class action lawsuit in the periods leading up and in the year of the lawsuit. PNF also has incremental power to other known determinants in predicting the likelihood of a lawsuit. In totality, our results provide support for the ability of PNF to capture the strategic withholding of bad news. We complement our primary validity tests with a battery of additional tests to ensure that PNF captures corporate disclosure policy instead of external or structural factors. Competing, market-based explanations of an asymmetry in the distribution of stock returns relate to investor sentiment, learning, overconfidence, the asymmetric transmission speeds of bad versus good 4

6 news, and volatility feedback. A return asymmetry could also reflect an asymmetric market response to financial analysts recommendations or the varying credibility of bad versus good news. We discuss these competing explanations in detail in section 5 and run several additional validity tests. First, we investigate changes in PNF associated with CEO successions and with changes in corporate governance regimes. To the extent that our measure reflects CEOdetermined strategic withholding of bad news, we expect CEO successions to cause a structural shift in the practice. We examine changes in PNF in the four years surrounding CEO successions and document a substantial fall in PNF in the two years following CEO successions, consistent with prior evidence of higher management discipline following CEO turnover. Second, we identify two structural shifts in PNF following the enactment of corporate governance and disclosure regulations during our sample period that arguably constrained firms incentives and opportunity to withhold bad relative to good news. Consistent with this argument and the rationale for our measure, we find that these regulations constrained variation in PNF driven by incentives to withhold bad relative to good news. Third, we test variations of our measure that filter out market and industry-wide variation. We repeat the analysis using industry-adjusted PNF and our results continue to hold. To further mitigate concerns that PNF reflects an asymmetry in the market s transmission speed of bad versus good news or volatility feedback effects, we repeat the analysis focusing on adjusted transmission rates of negative returns using firms with similar transmission rates of positive returns instead of industry peers as a benchmark, and again obtain similar results. These tests justify keeping the construction PNF as simple as possible. Also, our measure does not reflect an asymmetric market response to analyst stock recommendations, as our results are robust to re-calculating PNF after removing days with revisions in analyst earnings forecasts or stock recommendations. 5

7 An alternative explanation for the asymmetric market response to positive versus negative firm news is that managers accelerate bad news disclosures. For example, Skinner (1994) and Soffer, Thiagarajan, and Walther (2000) find that firms pre-empt negative earnings surprises at mandatory announcement dates by pre-releasing the bad news voluntarily. In this sense an asymmetric return distribution may still reflect corporate disclosure policy, but one of releasing bad news to the market more promptly than good news. Such a disclosure policy may parallel conservative accounting practices in financial reports (e.g., Watts 2003). A policy of disclosing bad news promptly but releasing good news gradually leads to bad news releases that are more concentrated than good news releases, similar to a withholding strategy. This allows for alternative interpretations of asymmetric flows; they may reflect higher relative promptness in releasing bad news, or they can reflect accumulated, withheld bad news (Kothari et al. 2009). The market reaction can help to distinguish the two interpretations. In an efficient market, if a firm discloses bad news promptly to the market and good news gradually, there should be a larger stock price reaction per unit of good news than bad news, to reflect investors rationally inferring the undisclosed good news based on the disclosed good news portion. This is the opposite prediction to a withholding story, where the market reaction to bad news exceeds the market reaction to good news, because firms withhold and accumulate bad news while regularly releasing good news to the market. As a final test we examine the extent to which the asymmetric spread and magnitude of positive versus bad news contribute to PNF capturing the withholding of bad news. We find that the spread and magnitude of bad versus good news are incrementally informative in reflecting withholding incentives and outcomes, so their combined inclusion increases the power of PNF in capturing withholding disclosure policy. 6

8 Our paper contributes to the literature in three key ways. First, we construct a return based measure of the quality of the firm s information environment. Our measure is based on theories of voluntary disclosure and can be generated for the universe of listed firms. PNF constitutes a comprehensive measure of corporate information flows, and therefore provides a measure of the quality of a company s communication with financial markets that goes beyond the quality of financial reporting in the audited financial statements and observable financial disclosures (e.g., annual report tone and readability scores). This is the important contribution of our measure. The closest comparable measures of disclosure quality in terms of comprehensiveness are the AIMR-FAF ratings that evaluate the quality of U.S. firms corporate disclosures in annual reports, analyst meetings, and conference calls, but these scores have discontinued since Though stock return provides an endogenous classification of bad and good news, results of our additional analysis mitigate endogeneity concerns and, though not definitive, support the construct validity of PNF in capturing properties of corporate disclosure policy. As an indirect proxy for the quality of the information environment, PNF complements research that seeks to score the quantity and quality of firm disclosures directly. PNF avoids the subjectivity involved in directly scoring corporate disclosures, while capturing all, and not just observable corporate disclosures to equity markets. PNF may also be a useful complement to textual scores that the SEC and other regulatory bodies and institutions use to assess the quality of corporate disclosures (Hoberg and Lewis 2015). 2 Second, our paper systematically links the strategic management of information flows to stock returns. We argue that the strategic management of information flows, i.e., accelerating good news and withholding bad news, leads to an asymmetry in the distribution of stock returns. 2 The SEC is seeking ways of analyzing a firm s 10-K MD&A disclosure to identify firms that have committed fraud and shall therefore be pursued with an AAER (accounting, auditing and enforcement action). 7

9 Our evidence contributes to the literature examining how corporate information flows affect stock return volatility (e.g., Hutton et al. 2009, Acharya, De Marzo, and Kremer, 2011; Kothari et al. 2009, Shin, 2003, 2006; Kalev et al. 2004) by highlighting the implications of accelerating good news and withholding bad news for the properties of stock returns. A key conclusion of our analysis is that transmission rates of positive versus negative abnormal stock returns reflect corporate disclosure policy. As we wish to construct an inverse measure of disclosure quality we construct PNF to capture asymmetry in the distribution of stock returns that reflects the strategic withholding of bad news. We find that PNF is positive for a substantial fraction of firm years, about 30%, and this fraction peaks in periods of economic slowdown. On average, however, PNF is negative, consistent with the strategic withholding of bad news not being widespread in the U.S. This may be due to aspects of SEC regulations that induce more conservative disclosure practices (Trueman 1997). This finding deserves further research to explore the reasons for the variation in PNF across jurisdictions, e.g., the litigious environment in the U.S. or political incentives in China (Piotroski, Wong, and Zhang, 2015). Our analysis of corporate governance regimes suggests that PNF captures the effects of institutional changes. 2. The PNF measure and hypotheses In this section we review the rationale underlying our measure and outline its construction and key hypotheses about its construct validity. 2.1 The strategic withholding of bad news and asymmetric stock returns We begin by postulating how the strategic withholding of bad news is associated with the distribution of stock returns, which forms the basis for constructing our measure. We focus on information dissemination, i.e., managers discretion in timing the flow of information after they 8

10 receive it. We assume that managers receive information randomly during the accounting period and that the arrival of good news does not systematically differ from the arrival of bad news, similar to Acharya, DeMarzo, and Kremer. (2011), Kothari, Shu, and Wysocki (2009), King and Wallin (1996). 3 Given this assumption, if managers disclose information as it arrives, communicating this information should generate symmetrically distributed stock returns. If, however, managers strategically time information flows to capital markets, accumulating and withholding bad news while disclosing good news on time, then the market reaction is unlikely to be symmetric. We next consider the theoretical rationale for such a disclosure policy and the underlying managerial incentives. The theoretical literature on corporate financial disclosure has analyzed the circumstances and incentives behind the timing of information flows to capital markets, focusing on key qualifications of the disclosure principle that introduce disclosure thresholds and justify partial disclosure equilibria. Dye (1985) focuses on investor uncertainty about whether managers have received private information. When managers do not disclose, investors are unsure whether managers have received information but chosen to conceal it, or whether they have not received information. This uncertainty induces a rational expectations equilibrium in which managers withhold news that is below a threshold, while releasing news above the threshold. Thus, managers may withhold bad news, with investors pooling non-disclosers with firms whose managers have genuinely received no news. 4 While investor uncertainty about the information a 3 Dye (2010) is one of the few papers that examines information arrival. Focusing on managers timing of information acquisition, assuming immediate release upon acquisition, Dye shows that it may be optimal for managers to acquire information in bunches. Since Dye (2010) assumes real time reporting, his model carries no direct empirical implications for disclosure bunching in practice. Also, his model does not distinguish between bad and good news as, ex ante, managers cannot choose this; asymmetric timing of bad versus good news flows arises only from the disclosure decision, given the acquisition decision. 4 Verrecchia (1983] focuses on proprietary costs in analyzing discretionary disclosures. Proprietary costs increase the firm s disclosure threshold because when the manager does not disclose, investors are unsure whether information withheld is bad or not sufficiently good to offset the proprietary cost. 9

11 manager possesses enables firms to withhold bad news, the likelihood of continuous withholding decreases over time as investors beliefs that managers are informed increase (King and Wallin 1996). 5 This provides a rationale for managers releasing good news in early announcements and bad news in late announcements in the accounting period. Empirical studies of the strategic timing of information flows have investigated several managerial incentives to delay the release of bad news. These incentives relate to the tendency of firms to delay releasing bad news when managers wish to manage market perceptions, face greater career concerns, or have personal wealth at stake (Dye and Sridhar (1995), Shin (2003), Kothari, Shu, and Wysocki (2009), Edmans et al. (2014)). 6 Recent literature also probes the role of market news in explaining asymmetric corporate information flows. A key proposition in this literature is that negative market news triggers a clustering of negative announcements by firms, even when the arrival of the underlying information is not clustered (Acharya, DeMarzo, and Kremer (2011), Gennotte and Trueman (1996), Dye and Shridhar (1995), Aragon and Nanda (2014), Sletten (2012)). If firm value is correlated with market news, firms have incentives to cluster negative announcements with the release of negative market news as the interpretation of the firm s news is then relatively more favorable. The release of correlated negative market news effectively lowers the equilibrium disclosure threshold, acting as an additional incentive to withhold bad news. 5 King and Wallin (1996) extend Dye s (1985] model to allow investors to revise their beliefs that a manager has received information and show that the threshold level for which the manager withholds bad news is unique and inversely related to the probability that the manager has access to private information. This has an important implication: the equilibrium threshold for disclosing news falls over time as investors beliefs that managers are informed increase. 6 The literature also suggests that firms have incentives to accelerate bad news when they want to preempt shareholder litigation or lower the exercise price of stock options to maximize option gains (Kasznik and Lev (1995], Skinner (1994], (1997], Trueman (1997], Baginski, Hassel, and Kimbrough (2002], Aboody and Kasznik (2000], Donelson, McInnis, Mergenthaler, and Yu (2012]). We incorporate all incentives for strategic timing of information flows suggested by the literature when building a model to assess the construct validity of our measure. 10

12 Having reviewed the rationale and incentives to strategically withhold bad news, we next consider how it induces an asymmetry in the distribution of stock returns. In terms of the distribution of news flows, we postulate that firms that strategically withhold bad news accelerate the release of good news but withhold bad news, releasing it later as the equilibrium threshold for disclosing news falls over time or at the same time as subsequent good news. The latter disclosure policy is consistent with a sanitization strategy, which Shin (2003) describes, whereby managers receiving multiple news items prioritize releasing good news, holding on to pieces of bad news in anticipation of a better signal (Pae 2005, Guttman, Kremer, and Skrzypacz 2014). 7 Withholding bad news and accelerating good news induces an asymmetry in the way information flows, such that bad news flows in lumpier pieces than good news, i.e., bad news flows are more concentrated than good news flows. A result of managers asymmetric disclosure behavior is that the market reaction to news flows is also asymmetric. As Kothari et al. (2009) argue, when managers withhold bad news while releasing good news regularly, the market reaction to bad news is larger than the market reaction to good news with the asymmetry reflecting the withholding and accumulation of bad news versus the regular release of good news. Taken together, strategic withholding of bad news is likely to result in an asymmetric stock return distribution, reflecting the higher concentration of bad compared with good news flows and the higher magnitude of the market reaction to bad compared with good news flows. We construct a measure to capture this asymmetry in spread and magnitude of information flows associated with the strategic withholding of bad news. In this way we operationalize the notion 7 Shin (2003] shows that in equilibrium managers follow a sanitization strategy when receiving multiple signals, i.e., they remove bad news and leave the good news. Also in a multiple-signal setting, Pae (2005] shows that a firm that receives two signals discloses both if they are favorable and confirm each other, discloses the most favorable signal if it is sufficiently favorable relative to the other, and discloses neither when both are unfavorable. This implies that managers may withhold bad news when they simultaneously receive good news. Guttman, Kremer, and Skrzypacz (2014] identify additional conditions that may lead managers to withhold bad news. Adding multi-periods to a multiple-signal setting, they show that the market interprets later disclosures more favorably, and managers may withhold disclosure to take advantage of the option to wait for a better signal. 11

13 of the asymmetric distribution of stock returns reflecting corporate disclosure policy (Acharya, DeMarzo, and Kremer 2011, Gennotte and Trueman (1996, Dye and Shridhar 1995, Aragon and Nanda 2014, Sletten 2012, Hutton et al. 2009) to produce an empirical proxy for the quality of companys communication with capital markets. 2.2 PNF: A measure of asymmetric information flows We capture firm specific information flows using the abnormal returns from a market model. 8, 9 This association is well supported theoretically and empirically (Acharya, DeMarzo, and Kremer (2011), Shin (2003), (2006), Kalev et al. (2004)). Beyer et al. (2010) report evidence supporting the theoretical association between abnormal returns and firm-specific information by reviewing how key sources of accounting disclosures contribute to the information in security prices. Their results indicate that for the average U.S. firm, about 30% of its quarterly stock return variance occurs on days when there are accounting disclosures, with management forecasts contributing the most information, followed by earnings announcements and SEC filings. This evidence attributes substantial variation in stock returns to observable voluntary and mandatory firm disclosures, while at the same time leaving enough variation attributable to additional observable or non-observable channels (e.g., conference calls, road shows, investor days, meetings with investors, meetings with analysts, press releases, and letters to shareholders). An important assumption in using abnormal returns to measure firm-specific good and bad news is that, consistent with market efficiency, the unconditional probabilities of good and bad news at the start of the year are equal. Although the exogenous stochastic process generating 8 We regress daily firm stock returns (Ret) on daily market value-weighted returns and obtain abnormal returns (AR) by adding the intercept to the regression residual. We set AR to zero when Ret equals zero. We obtain similar results recalculating our measure using raw instead of abnormal returns. 9 Since we want a measure of firm specific information flows, i.e., a measure that separates firm-specific from market-wide information flows, we follow the standard approach of using a statistical model of asset returns, in particular the market model (see, e.g., Fama (1998]). 12

14 information may induce an asymmetry in positive relative to negative stock market movements (e.g., due to volatility feedback), we postulate that it also, at least partly, reflects the way that firms manage the flow of information to the capital market (Kalev et al. 2004) and particularly managers policy of withholding bad news (Hutton et al. 2009, Acharya, De Marzo, and Kremer, 2011; Kothari et al. 2009, Shin, 2003, 2006). 10 To capture the asymmetry in the distribution of stock returns that reflects both the asymmetric flows of bad versus good news and the asymmetric magnitude of the market reaction to these flows, we calculate flow measures of positive and negative returns. We label these two measures Good flows and Bad flows and define PNF as the difference between Good flows and Bad flows. PNF = Good flows Bad flows (1) We take this difference because if firms strategically withhold bad news, bad news flows fall short of good news flows. The flow measures reflect both the spread and magnitude of negative versus positive returns, so higher values of PNF indicate a higher concentration of negative than positive returns or a higher magnitude of negative than positive returns. In particular, Good flows and Bad flows are the standardized flows, in number of days, of positive and negative returns, calculated using the sum of abnormal returns transmitted daily weighted within each firm year in order of magnitude (assigning one to the highest). Taking the difference between Good and Bad flows, instead of concentrating on just Bad flows, controls for firm characteristics that arguably affect the speed of transmitting news into the capital market irrespective of type (e.g., communication channels and frequency, growth rate, operating performance). We standardize the flows by total returns in the year, separately for each news type. In detail, 10 Hutton et al. (2009) associate stock price crashes with managerial accounting choices, claiming that the crashes reflect managers withholding of bad news until its accumulation reaches a tipping point sufficient to result in a stock price crash. 13

15 p,, i1,, (2) Good flows AR Tot AR Rank Tot AR it id it id it n,, i1,, (3) Bad flows AR Tot AR Rank Tot AR it id it id it where AR is firm i s abnormal return on day d, if positive, Tot AR id, it, positive abnormal stock returns in year t, Rank is the rank of AR id,, id is the sum of firm i s daily for firm i in year t and p is the number of days of positive abnormal stock returns. Similarly, AR id, return on day d if negative, Tot AR it, is firm i s abnormal stock is the sum of firm i s daily negative abnormal returns in year t, id, Rank is the rank of AR, id for firm i in year t, and n is the number of days of negative abnormal stock returns. To provide intuition for PNF, figure 1 provides two examples. Both examples assume that managers receive the same amount of good and bad news at the same rate over days 1 to 3, but release it in different ways, both consistent with strategic withholding. In the first example managers release good news as it arrives, but withhold bad news to release it all on day 4. PNF for this example is 1.6, capturing the lower flow, in number of days, of bad (1.6 days) versus good news (3.3 days). Note that an opposite disclosure strategy, i.e., the timely disclosure of bad news and withholding of good news to day 4 yields a PNF of 1.6. In the second example managers release good news on time, but withhold bad news to partly release it on day 3 alongside good news, while releasing the remaining bad news on day 4. This scenario is consistent with a sanitization disclosure strategy where firms withhold bad news to release it with subsequent good news to cancel the effect. PNF in this case is positive but slightly lower than in the first example, i.e., 1.25, reflecting the more prompt release of some bad news on day 3 instead of day 4. In the extreme, where managers are completely successful in burying bad with good news, the lower limit of PNF is zero. PNF, however, identifies more realistic cases 14

16 where firms cannot bury all bad news with good news and eventually release some of the accumulated bad news separately, giving rise to bad news with a larger magnitude than the average good news. 2.3 Manifestations of PNF To assess the validity of PNF in capturing the strategic withholding of bad news and therefore serving as an (inverse) proxy for disclosure quality, we examine how it associates with the scope, incentives, and constraints on strategically withholding bad news, with other facets of information quality, e.g. the quality of reported earnings in the financial statements, and with the unfavorable outcome of shareholder lawsuits. Scope relates to investor uncertainty about whether managers have private information that enables withholding of bad news. A volatile business environment increases uncertainty about managers private information, making investors less sure whether the manager has received information but chosen to withhold it or has not received information. Managers of firms with volatile operations therefore have greater flexibility to strategically time bad news releases. Investor uncertainty about whether a manager is informed is higher for newly listed and high growth firms. Managers of newly listed firms face stronger incentives to withhold bad news as they strive to maintain a positive outlook in the market to survive the listing and maintain a record of growth. Newly listed firms also face a higher likelihood of negative earnings news leading to lawsuits (Beneish (1997)). Growth stocks face an asymmetric market response to bad news (Skinner and Sloan (2002)). Investors may raise their expectations of future sales and growth for firms with high past growth, inflating price multiples. Over time, if managers of these firms disclose information suggesting that optimistic expectations in price multiples are not sustainable, inferior stock price performance ensues. Taken together, if PNF captures strategic 15

17 withholding of bad news, we expect it to be positively associated with investor uncertainty that the manager is informed, which is positively associated with business volatility and firm growth. This leads to our first hypothesis. H1: PNF is positively associated with business volatility and firm growth. Managers are likely to strategically withhold bad news releases to financial markets if they perceive net benefits to doing so. Managerial incentives can relate to operating performance, corporate financing events, and other contractual considerations. Managers may have incentives to time the flow of earnings news to the market throughout the accounting period. Bagnoli and Watts (2007) hypothesize that negative earnings news creates incentives for managers to voluntarily disclose private information to mitigate the market s response to negative earnings surprises in the financial report. Miller (2002) observes a rise in disclosure ahead of reporting an earnings decline, with firms shifting to regular disclosures that focus on the positive short-term results and avoiding discussing the impending decline. 11 This is consistent with managers strategically managing the flow of good versus bad news during the year when facing reporting an earnings decline at the end of the year. Poor capital market performance may also affect firm disclosure practices. If firm value is correlated with market news, firms have incentives to cluster negative announcements with the release of negative market news as the interpretation of the firm s news is then relatively more favorable (Aragon and Nanda (2014), Sletten (2012), Acharya, DeMarzo, and Kremer (2011), Gennotte and Trueman (1996), Dye and Shridhar (1995)). So firms may have incentives to accumulate and withhold bad news until the release of bad market news. If PNF captures the strategic withholding of bad news, we expect it to be 11 Such disclosure strategies enable managers to reap the benefits of high disclosure while avoiding the negative effect of unreliable forward-looking statements on their reputations and exposure to litigation risk. 16

18 positively associated with poor reported performance and poor stock market performance. Our second hypothesis is: H2: PNF is positively associated with poor reported and stock market performance. Capital raising activities may also induce managers to strategically withhold bad news. If a firm intends to issue additional equity to finance future operations, managers have incentives to withhold information that might have a negative price impact in order to avoid exacerbating the negative stock market reaction due to information asymmetry. Research suggests that around equity offerings firms make more frequent, detailed, and optimistic disclosures about their performance in anticipation of the offering (Lang and Lundholm (2000), Shroff, Sun, White, and Zhang (2013)). 12 Managers also face strong incentives to withhold bad news around debt issues as debtholders are concerned more about bad than good news. Due to their asymmetric payoff function, debtholders have a limited ability to benefit from increases in firm value. Conversely, bad news implies a higher risk of default that affects their payoff directly. Consistent with this, Easton, Monahan, and Vasvari (2009) find that bond trades increase around earnings announcements, especially when they convey bad news, and bond prices react more strongly to negative unexpected earnings. Similarly, DeFond and Zhang (2011) find that the bond market reacts more strongly to bad than to good news. 13 When it comes to the role of news in the debt market an additional factor to consider is the riskiness of existing debt (Easton, Monahan, and Vasvari (2009)). When a company approaches default, debtholders face a higher risk of economic loss and, therefore, news about firm performance is more relevant. As the relevance of 12 There is arguably a cost to strategically timing information releases around equity offerings. Lee and Masulis (2009] find that lower information quality, proxied by the quality of accruals, is associated with higher underwriting fees and a higher probability of a withdrawn equity offering. For equity offerings to induce withholding of bad news, the net gain that managers make by hiding information must outweigh the implied costs. 13 Similar to equity issues there is a related reputational cost to strategically withholding bad news during debt issues. Ashbaugh, Collins, and LaFond (2006], Bharath, Sunder, and Sunder (2008], and Graham, Li, and Qiu (2008] provide evidence that rating agencies and bondholders assign lower credit ratings and charge higher debt costs to firms with poor information quality. 17

19 bad news for the debt market increases with default risk, so do managers incentives to withhold bad news. Default risk may also lead to biased information flows through their effect on managers career concerns. Kothari, Shu, and Wysocki (2009) argue that when managers approach financial default, their concerns about contract termination increase and they face greater incentives to delay bad news. So taken together, if PNF captures the strategic withholding of bad news we expect it to be positively associated with both equity and debt issues and financial distress. H3: PNF is positively associated with equity and debt issues and financial distress Of the managerial incentives to strategically withhold bad news, we finally consider manager s equity ownership. A manager s equity ownership may affect the strategic timing of information flows, with the effect varying with ownership level. At lower managerial ownership levels evidence suggests that managers engage in stronger selling activity (Core and Larcker (2002), Ofek and Yermack (2000)), which may induce a strategy of withholding bad news. 14 Consistent with this, Noe (1999) and Edmans et al. (2014) provide evidence of a positive association between selling activity (or stock option vesting) and the release of more discretionary good news. At higher levels of ownership, however, managers are less likely to strategically time disclosures. Firms with higher managerial ownership have reduced agency and monitoring costs, so their managers are less likely to be concerned about the stock price reaction to unfavorable information. So if PNF captures the strategic withholding of bad news we expect 14 Core and Larcker (2002] find that firms with low managerial ownership are more likely to grant options to managers to increase their equity exposure, while Ofek and Yermack (2000] show that managers receiving option grants diversify their risk by selling shares. 18

20 it to exhibit a non-linear association with managerial ownership, increasing with lower and decreasing with higher ownership levels. 15 Our fourth hypothesis is: H4: PNF is positively associated with managerial ownership at lower (higher) levels. Among constraints to strategically withholding bad news we first consider public visibility. Larger firms have superior communication channels that increase the market s ability to be informed about the firm. Therefore, managers of larger firms have less flexibility to time information releases. Also, more analysts usually follow larger firms, making more information available to outsiders (Duchin, Matsusaka, and Ozbas (2010)) and constraining managers ability to strategically time information flows to the market. Larger firms also face a higher risk of shareholder litigation, and this may induce managers to disclose bad news on time (Skinner (1994), Kasznik and Lev (1995), Field, Lowry, and Shu (2005)). Therefore if PNF captures withholding of bad news we expect it to be negatively associated with firm size. H5: PNF is negatively associated with firm size. We next consider the role of institutions in constraining the strategic timing of information flows. The ability of institutional investors to acquire private information may restrain managers strategic timing of information flows. Similar to Maffett (2012), we evaluate institutions ability to acquire private information based on their investment horizon. Prior research suggests transient institutions, i.e., those with short investment horizons, are more likely to seek and trade based on private information than non-transient institutions, i.e., those with long-term investment horizons (Yan and Zhang (2009), Bushee and Goodman (2007), Bushee and Noe (2000)). The key factor inducing the information acquisition process is trading frequency, which allows investors to exercise governance through the alternative exit channel of 15 Yeo, Tan, Ho, and Chen (2002] also find a non-linear association between managerial ownership and the informativeness of earnings; at low (high) levels of management ownership the informativeness of earnings increases (decreases) with managerial ownership. 19

21 selling shares (Edmans, Fang, and Zur (2013), Edmans (2009)). If PNF captures the strategic withholding of bad news by managers, we expect it to be negatively associated with investor turnover, i.e. the frequency with which institutional investors rotate positions in their portfolios. H6: PNF is negatively associated with investor turnover. Looking beyond the determinants of the strategic withholding of bad news to other potential manifestations, we investigate the association of PNF with the properties of other aspects of the company s financial reporting. A company s strategic management of bad versus good news flows is likely to be part of a general lower quality reporting and disclosure policy. Prior research shows that there is a positive association between the quality of mandatory reporting and that of voluntary disclosures due to information quality being endogenous (Francis, Nanda, and Olsson (2008)). So if reporting and disclosure choices align, we expect firms that strategically withhold bad news to have lower quality reported earnings. Therefore, if PNF captures strategic withholding of bad news we expect it to be positively associated with poor earnings quality. H7: PNF is positively associated with poor earnings quality. Turning finally to the outcome of strategically withholding bad news, releasing bad news in lumpier pieces due to cumulative withholding may induce large stock price drops upon earnings announcements and trigger class action lawsuits (Francis, Philbrick, and Schipper (1994)). An important implication of strategic withholding of bad news is, therefore, a higher lawsuit probability. As a result, if PNF captures the strategic withholding of bad news, we expect to find a positive association between the possibility of a class action lawsuit and PNF. Our final hypothesis is: H8: The likelihood of a class action lawsuit is positively associated with PNF. 20

22 3. Methods and variable definitions To test our first six hypotheses, we regress PNF on the three vectors of variables capturing the business environment, management incentives, and constraints and a set of controls, PNF Business environment Management incentives it, it, it, Constraints Controls e it, it, it, (4) Below we describe the components of each vector and Figure 2 summarizes the components and variables. We capture business environment volatility, using the firm s idiosyncratic return volatility, σar, and an indicator of membership of a high-tech industry, HighTech. 16 Including σar also allows for higher discretion to time information flows afforded by higher rates of news arrival (Kalev et al. (2004), Berry and Howe (1994)). We capture firm growth using the number of years the company has been listed (YrsListed) and the firm s compound annual sales growth (Salesgrowth). According to H1, we expect PNF to be positively associated with σar, HighTech, YrsListed, and Salesgrowth. We capture poor operating performance using an indicator of reported earnings declines (NegΔEarn) and of losses (NegEarn). We measure poor stock performance using the proportion of days in the fiscal year with negative market returns (NegMRetDays). According to H2, we expect PNF to be positively associated with NegΔEarn, NegEarn, and NegMRetDays. We capture equity offerings using an indicator of issues of common or preferred stock (SEO) and debt issues using an indicator of issues of long-term debt (DebtIssues). We capture financial distress using the company s debt to market value of equity ratio, which we label Distress. According to H3, we expect PNF to be positively associated with SEO, DebtIssues, and Distress. We capture managerial ownership using the percentage of stock held by directors, DirStk%, and include DirStk% and its square, DirStk% 2, to account for the non-linear effect of 16 Appendix A provides exact definitions of all variables. 21

23 insider ownership. As DirStk% 2 captures the effects of lower ownership levels we expect PNF to be positively associated with DirStk% 2 and negatively associated with DirStk% according to H4. We capture firm size using the company s market capitalization, MktCap. According to H5, we expect PNF to be negatively associated with MktCap. We calculate a measure of investor turnover (InvestorTurnover) based on the churn rates of firms institutional holdings, i.e., the average frequency with which institutional investors rotate positions in their portfolios (Gaspar, Massa, and Matos (2005)). 17 Higher rates indicate higher trading frequency and shorter investment horizons. 18 We expect to find a negative association between PNF and InvestorTurnover according to H5. In our set of control variables we include the level of operating performance measured as return on assets (ROA). We then control for any residual effects of stock options not captured by insider ownership. Evidence suggests that when managers receive stock option grants, they have incentives to disclose bad news and withhold good news, to minimize stock price (Aboody and Kasznik 2000). Conversely, when managers consider exercising stock options, they have incentives to delay bad news or accelerate good news. 19 We control for the different effects of option grants and exercisable options by including the number of options granted, #OptionsGrant, and the number of exercisable stock options, #OptionsEx. We control for analyst coverage, as we wish to control for variation in PNF attributable to financial analysts, #Analysts. 17 Calculating investor turnover rates involves two stages. First, we calculate for each institutional investor a measure of portfolio turnover in each quarter (Gaspar, Massa, and Matos (2005]). Second, we calculate the investor turnover ratio at the firm level by calculating the weighted average of the total portfolio churn rates of its institutional investors over the four quarters each year. 18 As an alternative measure of investment turnover we use the classification of Bushee (2001] and Bushee and Noe (2000] of institutional investors into transient and non-transient institutions. Transient institutions have high portfolio turnover. Non-transient institutions include dedicated institutional investors that hold large and stable holdings in a small number of firms and quasi-indexers that hold large diversified portfolios and trade infrequently (Bushee and Noe (2000]). As data on this classification is available up to 2009, we report our main results using InvestorTurnover. 19 A constraint on this disclosure pattern is litigation costs associated with insider sales strategies. Cheng and Lo (2006] find that insider sales do not motivate changes in disclosure. We account separately for the effect of litigation when considering constraints on the asymmetric timing of information flows. 22

24 We finally control for the risk of litigation as fear of litigation may affect the extent of bad news withholding. To account for the endogeneity involved in this variable we follow Field, Lowry, and Shu (2005) and include an indicator variable for firms operating in high legal exposure industries, HighLit in our vector of control variables. To test H7 weuse four earnings quality proxies: accruals quality (AQ); absolute abnormal accruals (AbsAA); earnings variability (σearn); and a combined measure based on the common factor score of these metrics (EQ). Accruals quality, AQ, is based on the Dechow and Dichev (2002) model, extended by McNichols (2002), and measures the extent to which working capital accruals map into cash flows in the current, prior, and future periods and changes in revenues and property, plant, and equipment. The absolute value of abnormal accruals, AbsAA, is based on the modified Jones (1991) model. The standard deviation of earnings, EarnVar, correlates with various earnings quality measures, such as earnings smoothness, earnings predictability, poor matching of revenue and expenses, etc. (e.g., Francis et al. (2004), Dichev and Tang (2009)). Higher values of AQ, AbsAA, and EarnVar indicate noisier reported earnings (e.g., accrual estimation errors, volatility). The combined measure, EQ, is the common factor score from a factor analysis of AQ, AbsAA, and EarnVar. EQ offers a more comprehensive measure of earnings quality. It has the same ordering as the underlying variables, so larger values of EQ indicate lower quality reported earnings. According to H7, we expect to find a positive association between PNF and EQ and the other earnings quality proxies. Finally to test H8, we model the probability of a class action lawsuit (Lawsuit = 1) as a function of PNF and other factors affecting the probability of litigation that Field, Lowry, and Shu (2005) suggest. We include, market capitalization (MtkCap), stock volatility (σret), past returns ( Rett ), stock turnover (Turnover), and indicator variables for high tech and regulated 1 23