Financial Constraints and Future Stock Price Crash Risk

Transcription

1 Financial Constraints and Future Stock Price Crash Risk Guanming He Warwick Business School, University of Warwick Tel: Helen Ren Warwick Business School, University of Warwick Phone: Coventry, United Kingdom CV4 7AL I

2 Financial Constraints and Future Stock Price Crash Risk Abstract: This study investigates the association between financial constraints and future stock price crash risk. Using the crash risk measure in Hutton et al. (2009) and financial constraint proxy developed by Hadlock and Pierce (2010), we find that financial constraints increase future crash risk. This finding can be explained via both managerial bad news hoarding and default risk mechanisms. The desire to secure external financing of a financially constrained firm induces manager to withhold bad news, the accumulation of which would result in a price crash once the upper limit is exceeded. Also, financially constrained firms are subject to higher default risk and firms with higher probability to fail are more likely to experience a price crash at default. Our results are robust to the use of a difference-indifferences method and longer forecast windows of crash risk to control for potential endogeneity; and to the test of default risk mechanism using the proxy of credit rating. We further conduct cross-sectional analyses and find that the positive association between financial constraints and future crash risk is more prominent for firms with high accruals but less pronounced for firms that commit tax avoidance. Keywords: financial constraints; crash risk; accruals; tax avoidance; default risk JEL classification: G19; G31; H26; M41 II

3 1. INTRODUCTION The recent financial crisis of and corporate scandals (e.g., Enron, Worldcom, Fannie Mae) have triggered the growth of new stream of research in probability of price crashes, namely, stock price crash risk, which is normally observed in the far-left tail of firm-specific return distributions (e.g., Hutton et al., 2009; Zhu, 2016). The motivation to study the risk of extreme negative residual returns involves its importance in determining the expected stock returns (Conrad et al., 2013), return volatility and option pricing (Merton, 1976). The objective of this study is to explore whether and how a firm s financing constraint status affects future stock price crash risk. Lamont et al. (2001) define financial constraints as frictions that prevent firms from funding their desired investments. Previous literature (e.g. Fazzari et al., 1988; Lamont et al., 2001; Denis and Sibilkov, 2010) typically examines the association of financial constraints with investment, firm value, risk and expected returns, but none has evaluated the information management of financially constrained firms and its impact on stock price crash risk. Our study attempts to fill this gap in the literature. The difficulties in raising external funds induce managers in financially constrained firms to withhold bad news, the accumulated bad news and overvalued stock prices would increase risk of future stock price crashes. Moreover, financially constrained firms are subject to higher probability of corporate failure and are more likely to experience price crashes at the point of default. However, if investors are able to decipher these implications and discount the financially constrained stocks promptly, stock prices will reduce gradually over time without triggering a sudden price crash, lowering the future crash risk. Therefore, the relation between financial constraints and future stock price crash risk remains ambiguous, which constitutes another 1

4 motivation for our study. Among the mechanisms that could engender stock price crash risk, two explanations are particularly relevant to our study. Firstly, the accumulation of bad news is regarded as the fundamental cause of stock price crashes (Chen et al., 2001; Hutton et al., 2009; Kim et al. 2011a; Chang et al., 2016). Previous studies (Bolton et al., 2006; Benmelech et al., 2010) suggest that managerial equity incentives induce managers to engage in short-termist behavior such as bad news hoarding to inflate current stock prices. Because negative earnings surprise can result in costly equity and debt, managers in financially constrained firms are prone to hide bad news for an extended period in order to secure external funds. According to Jin and Myers (2006) s model, the amount of downside risk and losses related to bad performance of the firm that managers are willing or able to absorb is limited. Once a critical level is crossed, managers would abandon their positions and the exposure of accumulated bad news in one fell swoop would result in a sudden, dramatic price drop, that is, a stock price crash (Hutton et al., 2009). Given a greater incentive to secure external finance, firms in financial constraints have higher future crash risk compared to the unconstrained firms. An alternative explanation of stock price crash is the default risk 1, which is rarely examined. A price crash resulted from corporate failure is more likely to happen for a firm with high default risk, that is, an elevated probability of failure to meet its debt obligations. Prior research (e.g., Fazzari et al., 1988; Almeida et al., 2004; Acharya et al., 2007; Campello et al., 2010) shows that investment spending of financially constrained firms is more sensitive to internal cash flows than that of unconstrained firms. Under capital market imperfections, financially constrained firms tend to hold more cash, which plays an important role in 1 For default risk, we mean the probabilities of default, financial distress, economics distress, or bankruptcy that are usually used interchangeably in the literature (Campbell et al., 2008). 2

5 avoiding corporate default. By contrast, when the external financing is costly, any shortage in internal funds can easily push the firm into default. Thus, a firm confronted with financial constraints is more susceptible to internal funds availability and is subject to higher default risk. Campello et al. (2010) find that firms facing financing frictions have difficulties in pursuing desired investments and are forced to sell off assets during financial crisis. The higher cost of capital and suspension of profitable projects encountered by financially constrained firms would increase their probabilities to fail, resulting in a higher stock price crash risk. Although both bad news hoarding and default risk mechanisms predict that firms in financial constraints have higher future crash risk, tension exists when taking consideration of investor s ability in deciphering the implications of financial constraints. Prior studies (e.g. Farre-Mensa and Ljungqvist, 2016) point out that financial constraints faced by a firm are not directly observable. Empirical literature applies various indirect proxies to infer financial constraints from firm characteristics and investment behavior. Yet considerable debates persist on measures of financial constraints. At the market level, investors may observe cross-sectional differences in firm characteristics but no evidence suggests that they can discern the financing constraint status of a firm. Thus, we conjecture that investors cannot identify the presence of financial constraints in absence of appropriate expertise. In addition, previous evidence (Lamont et al., 2001; Whited and Wu, 2006; Livdan et al., 2009) shows that financially constrained firms do not earn significantly higher equity returns, indicating irrationality of market participants. If investors can discern and decipher the valuation impact of financial constraints, a return premium would be required for financially constrained stock to compensate for higher risk taken. The evidence implies that constrained 3

6 stocks are generally overpriced and hence are more likely to undergo future price crashes. Zhu (2016) argues that market inefficiency could in itself reinforce the magnitude of future stock price crash risk. Also, even if markets are efficient in valuing financial frictions-related information that is conveyed in management s selected disclosures, the amount of bad news hoarded by managers and the level of default probability can hardly be appraised by outsiders (Dye, 1985; Jung and Kwon, 1988; Dichev, 1998; Griffin and Lemmon, 2002; Campbell et al., 2008). Overall, it follows that investors are unable to observe and comprehend financial constraint s implications in bad news hoarding and default risk for future crash risk. Therefore, we predict that the association between financial constraints and future crash risk is positive. We further explore whether the positive relation between financial constraints and future stock price crash risk is mediated by earnings management and corporate tax avoidance. Earnings management, or abnormal accruals, could facilitate bad news hoarding behavior (Hutton et al., 2009; Zhu, 2016; Kim et al., 2011a). Zhu (2016) argues that managers seeking to withhold bad news are inclined to make aggressive income-increasing accrual estimation. The financial opacity resulting from earnings management makes it more difficult for outside investors to discern any hidden bad news, providing stronger incentives for managers to conceal adverse performance using positive accruals. Therefore, we expect that earnings management, as a main approach facilitating bad news hoarding, has an incrementally positive impact on the predicted positive relation between financial constraints and future crash risk. Next, recent studies (Desai and Dharmapala, 2006, 2009; Kim et al., 2011a) find consistent evidence with agency perspectives on corporate tax avoidance. Complex tax 4

7 avoidance transactions obfuscate financial reporting and facilitate earnings management that conceals bad news and managerial rent diversion (He et al., 2016). Despite the fact that tax avoidance could incentivize bad news hoarding, the primary purpose for managers in financially constrained firms to pursue tax planning activities is to generate cash flows and mitigate default risk. The influence of financial constraints on cash flows has been largely documented in prior research (e.g. Almeida et al., 2004; Faulkender and Wang, 2006; Pinkowitz and Williamson, 2006; Denis and Sibilkov, 2010). The internal funds will be more valuable when firms face with costly external financing. Among the strategies to obtain extra cash flows, tax avoidance has relatively less adverse impact on firm operation (Edward et al., 2016), the increased cash resulting from tax savings helps lower the default risk and mitigate the occurrence of future crashes. Therefore, we expect that the relationship between financial constraints and future crash risk would be less pronounced for firms that involve in aggressive tax planning activities. Following Hutton et al. (2009), we use an indicator variable crashrisk as our primary measure of crash risk in empirical analyses, which captures the likelihood of extreme negative stock returns. We use three other proxies for crash risk in robustness checks: (i) the number of crash weeks (count) with negative extreme weekly returns; (ii) down-to-up volatility of firm-specific weekly returns (duvol) and (iii) the minimum value of firm-specific weekly returns (minreturn). The degree of financial constraints is gauged using the SA index developed by Hadlock and Pierce (2010) who find that firm size and age are particularly strong predictors of constraints. Using a sample of U.S. firms for the period , we find that the association between financial constraints and future (i.e., one-year-ahead) stock price crash risk is significantly positive. The effect is economically as 5

8 well as statistically significant, confirming that investors are incapable of fully appreciating the prospects of firms in financial constraints. Additional analyses reveal that this positive relationship is more pronounced for firms with high abnormal accruals but less pronounced for firms that commit tax avoidance. We also perform several tests to address the endogeneity concerns. Firstly, following Almeida et al. (2013), we employ a difference-in-differences analysis by exploiting the collapse of junk bond market in 1989 as an exogenous shock to financial constraints. A series of events in bond market unexpectedly reduced the supply of credit (especially the below-investment-grade credit), which tightened up the financial constraints of junk bondissuing or speculative-grade companies (Lemmon and Roberts, 2010). We find that the increase in future crash risk for the treatment group (junk-bond issuers) is significantly higher than that for the control group (unrated firms) during the post-collapse period. This test mitigates concerns about potential measurement errors in financial constraint variable and provides corroborating evidence on the positive causal impact of financial constraints on future crash risk. Second, we test the predictability of financial constraints for future crash risk in longer-run windows to deal with potential simultaneity problems. Results show that financial constraints remain positively correlated with crash risk in two and three years. Third, prior research provides ample evidence on bad news hoarding mechanism, we carry out an additional test to support the default risk mechanism for our hypotheses. By using credit rating as a proxy for corporate default probability, we construct high and low default risk subsamples consisting of speculative- and investment-grade firms respectively, based on S&P s benchmark. Results confirm that the positive association between financial constraints and future crash risk is only observed in speculative-grade firms that have relatively higher 6

9 default risk. Our paper contributes to the literature in several ways. Prior research largely focuses on the impact of financial constraints on investment decisions, cost of capital and firm value. We add to the financial constraint literature by examining the role of financial constraints in information management and providing the first piece of firm-level evidence on a significant positive association between financial constraints and future stock price crash risk. We extend the related literature on managerial incentives by revealing that the desire to secure or lower the cost of external financing constitutes an important aspect of managers concerns in hiding negative firm-specific information. On that basis, we complement the emerging literature on determinants of future crash risk. Previous studies examine the influences of executives option incentives (Kim et al., 2011b), corporate tax avoidance (Kim et al., 2011a), institutional investor instability (Callen and Fang, 2013), accrual manipulation (Zhu, 2016), stock liquidity (Change et al., 2016), corporate governance (Andreou et al., 2016) and so on. We show that firm s financing constraint status has an impact on its future crash risk via both bad news hoarding and default risk mechanisms. It confirms that opportunistic managerial behavior arises in environments that accentuate agency risk. This study provides further insights into the occurrence of future extreme outcomes (i.e., higher-moment of stock return distribution), which, albeit rare, would have material impact on investor s welfare. Our findings provide some important implications. It should be of particular interest to investors making portfolio investment decisions. In understanding the factors that drive the cross-sectional variations in tail risk, investors could price the constrained stocks in a timely manner and better predict and eschew future stock price crashes. This is also relevant to 7

10 creditors, suppliers, customers and other stakeholders who have interests in the firm s creditworthiness and ability in meeting its financial obligations. The remainder of the paper is structured as follows. Section 2 reviews the related literature and develops the hypotheses. Section 3 describes the sample, measurement of key variables and research design. Section 4 and Section 5 present the empirical results and further robustness checks. Section 6 concludes. 2. LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT 2.1 The Association between Financial Constraints and Stock Price Crash Risk Bad news hoarding Prior literature has proposed a number of explanations of firm-level price crashes. Most studies view managers bad news hoarding as the major determinant in the formation of a stock price crash (e.g. Jin and Myers, 2006; Bleck and Liu, 2007; Hutton et al., 2009; Benmelech et al., 2010; Kim et al. 2011a, 2011b; Chang et al., 2016). Stock price crashes are typically triggered by the arrival of unexpected bad news, among the various types of bad news events, Ak et al. (2015) find that earnings announcements are the most common cause of stock price crashes. While a firm is likely to be hit by one piece of bad news, the occurrence of a price crash is actually attributed to the sudden overrun of the bad news hoarding limit, a threshold point at which the managers can no longer withhold any unfavorable information. The bad news hoarding explanation is derived from some theoretical models (Jin and Myers, 2006; Bleck and Liu, 2007; Benmelech et al., 2010) linking managerial accumulation 8

11 of bad news to crash risk. According to Jin and Myers (2006), managers are willing to personally absorb some costs of bad performance in order to maintain their job positions. However, as the bad news accumulates, the costs become unaffordable such that managers are likely to abandon their positions, leaving all the hidden news coming out at once and resulting in a sudden price plunge. In reality, the maximum amount of bad news that managers can withhold varies unforeseeably with the firm s changing environment, making it difficult for themselves to anticipate when such a threshold point would be crossed. Therefore, managers usually cannot prevent the price crashes from happening (He, 2015). Under the first explanation, the managerial tendency to conceal negative information from outside investors would engender the risk of future crashes and hence crash risk is largely determined by managerial incentives in bad news disclosures. The greater incentives managers have to withhold the firm s unfavorable information, the higher the future crash risk. Managers asymmetric information disclosure behavior has been widely documented in prior literature (e.g., Healy and Palepu, 2001; Verrecchia, 2001; Kothari et al., 2009; Ali et al., 2015). It suggests that career and compensation concerns (Graham et al., 2005; Ball, 2009; Kothari et al., 2009; Jiang et al., 2013; Ali et al., 2015) incentivize managers to withhold bad news. Specifically, the managerial incentives include those monetary motives such as intention to influence stock prices, value of option grants, bonus plans and promotion opportunities, as well as nonfinancial motives such as empire building and maintaining the esteem of people. Gibbons and Murphy (1992) argue that career concerns arise when promotions largely depend on employer s assessment of managers ability, which either stimulates good performance or induces concealment of bad news. Moreover, Graham et al. (2005) find that managers tend to withhold temporary bad performance, in hope of a 9

12 turnaround from subsequent corporate projects that would allow them to bury current unfavorable news. Since negative earnings surprise can result in costly equity and debt, managers are prone to hide bad news for an extended period, especially in financially constrained firms. Generally, the demand for external financing depends upon both how much cash and stock of funds available in the firm and the level of desired investments (Dechow et al., 1996). Given the existing difficulties in funding desired projects, financially constrained firms are more eager to obtain external funds at a low cost and secure future investments. Thus, managers in constrained firms have stronger incentives to withhold bad news relative to those in unconstrained firms, leading to a higher future stock price crash risk in financially constrained firms Default risk Meanwhile, the corresponding higher default risk of financially constrained firms provides an alternative explanation for their future crash risk. Default risk, or risk of financial distress, generally refers to the probability that firms fail to meet their financial obligations (Vassalou and Xing, 2004; Campbell et al., 2008; Garlappi et al., 2008). It is argued that firms with high default risk are more likely to fail and experience price crashes at the point of default (Zhu, 2016). The default risk explanation is most evident during the financial crisis period, yet little research attention has been paid to this mechanism. Previous literature emphasizes the role of internal funds in impelling constrained firms to default, Fazzari et al. (1988), Almeida et al. (2004) and Acharya et al. (2007) argue that the 10

13 investment spending by financially constrained firms is more sensitive to cash flows than that by unconstrained firms. Firms with costly external finance tend to save more cash. When cash plays an important role in helping constrained firms avoid default, cash shortages are more likely to precipitate corporate default for firms with restrictions in accessing external financing. By contrast, this impact is insignificant when a firm is financially unconstrained (Davydenko, 2007). These arguments imply that a financially constrained firm is more susceptible to internal funds availability and more prone to default than an unconstrained firm. Prior literature (e.g., Lamont et al., 2001; Whited and Wu, 2006; Gomes et al., 2006; and Livdan et al., 2009) examines the effects of financial constraints on equity returns, with mixed evidence documented. Others (e.g., Dichev, 1998; Griffin and Lemmon, 2002; Vassalou and Xing, 2004; Garlappi et al., 2008) have provided evidence on relationship between corporate default probability and stock returns. Although a firm s inability to fund investments is undoubtedly closely related to its default probability, few studies have investigated the relation between financial constraints and default risk. Kaplan and Zingales (2000, pp.710) argue that financially constrained firms share similar characteristics with financially distressed firms and point out that financial distress is a form of being financially constrained. It follows that corporate constraint status is an important aspect in determining firm s default risk. In additional, empirical evidence from financial crisis reveals that firms in financial constraints are subject to higher default risk. Ivashina and Scharfstein (2010) report that banks dramatically reduced the lending to corporate sector during the crisis period , the exogenous contraction in supply of credit resulted in a substantial increase in cost of 11

14 external finance. Yang (2011) finds that the impact of credit crisis on default risk is stronger for financially constrained firms. According to Campello et al. (2010) s survey, majority of the financially constrained firms encountered restrictions in pursuit of profitable projects and were forced to cancel valuable investments during the crisis. Drawing upon findings of prior research, we predict that higher cost of capital and suspension of profitable projects due to restricted funds increase the default probabilities of financially constrained firms and hence raise their future crash risk Financial constraints and market reaction Previous theories and studies suggest that stock price crashes can be triggered by a sudden release of accumulated bad news or result from corporate failure. At the firm-level, bad news hoarding and default risk mechanisms provide consistent prediction of a positive association between financial constraints and future crash risk. However, if, at the market level, investors are able to discover financial constraint and perceive its implications in managerial bad news hoarding and default probability, then stock price would be discounted promptly and decline gradually towards its fundamentals over time, instead of engendering a sudden price crash. We consider the necessary conditions for the proposition that financial constraints lower future crash risk to become true, and according to our following discussions, these conditions are unlikely to be fully satisfied. Firstly, there is no indicative evidence showing that investors are able to observe financial constraint status of a firm. Farre-Mensa and Ljungqvist (2016) point out that financial constraints are not directly observable. Most studies (e.g., 12

15 Hadlock and Pierce, 2010; Farre-Mensa and Ljungqvist, 2016) suggest that more constrained firms tend to be smaller, younger, less leveraged, less profitable and have more intangible assets. Nevertheless, no consensus has been achieved on measurements of financial constraint. The general definition suggests that financial constraints arise due to frictions in supply of capital, which largely result from information asymmetries between the firm and outside investors (Tirole, 2006). Almeida and Campello (2002) observe that constrained firms are normally featured with inelastic supply of capital curve. It follows that firm s financial constraints depend on both the level of internal funds and the extent of capital market imperfections (Povel and Raith, 2002). These features, however, are neither measurable nor observable to investors who do not have necessary expertise. Second, if investors can decipher the valuation impact of financial constraints, a risk premium would be required to compensate for the risk they bear. The risk-return tradeoff theory suggests a positive relationship between financial constraints and equity returns. However, empirical evidence (Lamont et al., 2001; Whited and Wu, 2006; Livdan et al., 2009) shows that financially constrained stocks do not earn significantly higher returns than unconstrained stocks. For example, Lamont et al. (2001) find that financially constrained firms earn lower average returns than unconstrained firms, which implies the irrationality of market participants. Whited and Wu (2006) and Livdan et al. (2009) document higher average returns earned by financially constrained firms than that by unconstrained ones but the difference is not significant. These findings indicate that financially constrained stocks are generally overpriced in the markets, thereby generating higher future crash risk. On the other side, even the market is efficient in pricing constrained stocks, it is still unlikely for investors to decipher the implications of financial constraints for future crash risk. 13

16 Without the access to private information, outside investors are not able to appraise the amount of hidden bad news and adjust stock prices for bad news hoarding. Also, prior evidence (Garlappi et al., 2008; Livdan et al., 2009) suggests that investors are not capable of evaluating the potential default probability of a firm, with negative distress-return relation reliably documented. Therefore, by refuting the opposite proposition, we posit that financially constrained firms are more prone to future price crashes through either bad news hoarding or default risk mechanism. Our first hypothesis is stated as follows: H1: Future stock price crash risk is positively associated with financial constraints. 2.2 Cross-Sectional Analyses of Association between Financial Constraints and Crash Risk Earnings management Recent studies (Hutton et al., 2009; Zhu, 2016) find that earnings (or accruals) management is positively associated with future stock price crash risk. Under an accrual accounting system, the estimate of firm performance depends on both accruals (e.g., accounts receivables and inventories) and actual cash flows. The estimated accruals are expected future net cash flows that associated with economic transactions during a given reporting period (Hutton et al., 2009). Accordingly, the subjectivity in accrual estimation provides managers with greater discretion in hiding bad news. Zhu (2016) argues that managers who seek to withhold bad news tend to make more aggressive income-increasing accruals estimates and such aggressive use of abnormal accruals would be elevated when managers have stronger incentives to conceal adverse performance. To provide some intuition, managers could withhold bad news such as negative product 14

17 market shocks by delaying inventory write-offs. Specifically, if inventories become obsolete or decline in market value, the loss should be reported as write-offs to shareholders, either in the cost of goods sold account or in a separated expense account. However, by delaying this recognition, managers can prevent investors from timely observing the negative shocks in inventory values. Other examples include delayed recognition of receivables revaluations and provisions for operating costs arising from adverse events (Ball and Shivakumar, 2006). When a specific customer invoice is identified as uncollectable after the record of original sales, the amount of payment should be written-off by removing from the accounts receivables. In such a case, managers could conceal the customer-related bad news and inflate the earnings by delaying the bad debt expense recognition. Therefore, we expect that the increased financial opacity resulted from aggressive use of inaccurate accruals will make it more difficult for investors to discern any accumulated bad news in a financially constrained firm. Moreover, Bleck and Liu (2007) s model predicts that financial opacity hampers shareholders ability to discriminate good projects from bad ones at an early stage and hence allows the bad projects to continue. Because outside investors cannot effectively abandon the bad projects, the poor performance would accumulate over time and eventually lead to a crash. Given the main intention of earnings management to facilitate bad news hoarding, it is less likely that abnormal accruals would impose an incrementally negative effect on crash risk by mitigating the default risk. Therefore, we establish the second hypothesis based on the former argument as: H2: The positive association between financial constraint and future stock price crash risk is more pronounced for firms that have high abnormal accruals. 15

18 2.2.2 Corporate tax avoidance Corporate tax avoidance generally refers to any tax planning activities that reduce explicit taxes (Hanlon and Heitzman, 2010). The traditional theory suggests that tax avoidance effectively transfers state wealth to a firm, resulting in increased cash flows. Recent studies (Desai and Dharmapala, 2006, 2009; Kim et al., 2011a) find consistent evidence with agency perspectives on corporate tax avoidance. With the separation of ownership and control, complex tax avoidance transactions obfuscate financial reporting and facilitate managerial opportunism such as bad news hoarding and rent diversion (Kim et al., 2011a, b; He et al., 2016). Despite the fact that tax avoidance increases financial opacity and provides managers with masks and tools to manipulate earnings and conceal adverse performance, the main intention of avoiding taxes for a financially constrained firm is indeed to obtain additional internal funds and mitigate default risk. In an imperfect capital market, external finance is not perfect substitute for internal capital. Since debt and equity financing are particularly costly and difficult to access for financially constrained firms, firms with high costs of external finance tend to rely more on cash holdings and cash flows (Fazzari et al., 1988; Almeida et al., 2004; Acharya et al., 2007; Denis and Sibilkov, 2010). According to Campello et al. (2010) s survey, evidence from financial crisis shows that financially constrained firms made significant changes in corporate policies and planned deeper cuts in investment, technology, employment and capital expenditures. Edwards et al. (2016) find that an increase in financial constraints incentivizes firms to increase tax planning in order to generate additional internal funds (i.e. cash tax savings). They argue that reducing tax payment has less impact on firm s operations than other cost-cutting strategies in building cash reserves. Thus, tax avoidance helps raise internal 16

19 funds and strengthens the firm s ability to repay financial obligations (i.e., lower the default risk), thereby reducing the future crash risk. The last hypothesis can be established as follows: H3: The positive association between financial constraint and future stock price crash risk is less pronounced for firms that commit tax avoidance. 3. DATA AND RESEARCH METHODOLOGY 3.1 Sample selection The initial sample consists of firm-year observations obtained from three major sources, Center for Research in Security Prices (CRSP), Compustat, and Factset. The crash risk measures are constructed using weekly stock price and return data from the CRSP database. Specifically, we follow Kim et al. (2011a) to assign weekly stock returns to the 12-month period ending 3 months after the fiscal year-end of a firm for each firm-year observation. Firm s financial and stock information is mainly collected from the merged Compustat/ CRSP database. Institutional ownership data are obtained from Factset. The overall sample period is from 1995 to 2015 to examine the lead-lag relationship between financial constraints and future crash risk. During the data filtering process, we require that firms have all necessary data to construct variables of interest for our empirical analyses. The final sample ends up with 26,501 firm-year observations corresponding with 6,308 unique firms. Table 1 reports the descriptive statistics of all the key variables used. 3.2 Crash risk measures 17

20 Following prior literature (Chen et al., 2001; Hutton et al., 2009; Kim et al., 2011a, 2011b; Callen and Fang, 2013), we employ four firm-specific measures of stock price crash risk: (1) an indicator variable (crashrisk) capturing the likelihood of negative extreme firm-specific weekly returns; (2) the number of crash weeks (count) with negative extreme firm-specific weekly returns; (3) the down-to-up volatility (standard deviation) of firm-specific weekly returns (duvol); and (4) the negative of minimum weekly returns (minreturn) over a fiscal year. We use the crashrisk measure for our main tests and the other three variables in robustness check. As per Hutton et al. (2009) and Kim et al. (2011a, b), price crashes can be defined as firm experiencing the firm-specific weekly returns falling 3.2 standard deviations 2 below the mean firm-specific weekly returns for a fiscal year. Crash risk is measured based on the number of price crashes that firm experiences in a given fiscal year, crashrisk equals to 1 for a firm-year if one or more crashes are experienced, and zero otherwise. Accordingly, the second measure count shows the exact number of crash weeks in which firm experiences negative extreme weekly returns over a fiscal year. The third measure duvol stands for down-to-up volatility. Following Chen et al. (2001), it is calculated as the standard deviation of down-week firm-specific weekly returns divided by the standard deviation of up-week firm-specific weekly returns. The down-week and up-week are defined as weeks with returns below and above the period mean over a fiscal year, respectively. We have the expression: 2 2 duvolit ( nu 1) R æ ö = - it ç ( nd -1) Rit DOWN è UP ø å å (1) where the firm-specific weekly returns are scaled by the number of down- and up-weeks (n " 2 Following Kim et al. (2011a), 3.2 is chosen to generate a frequency of 0.1% in the normal distribution. 18

21 and n # ) minus one. The last variable measures the minimum weekly returns, minreturn is computed as the negative of minimum value of firm-specific weekly returns, subtracting the mean of weekly returns, then divided by the standard deviation of the firm-specific weekly returns over the fiscal year. Although the indicator variable crashrisk has the limitation that it cannot reflect the relative magnitude of the price crash, duvol and minreturn involve potential misclassifications of gradually decreased stock price, featuring high down-to-up variance or minimum weekly returns as a crash. Thus, our empirical analyses are mainly based on the crashrisk measure. 3.3 Financial constraint index For our primary measure of financial constraints, we use the SA index constructed by Hadlock and Pierce (2010) as: 2 SA index size size age =- + - (2) where size is the natural logarithm of the book value of total assets and age is the number of years since the firm s incorporation or founding. By construction, more financially constrained firms have higher SA indices. Hadlock and Pierce (2010) argue that firm size and age are particularly useful in predicting firm s financial constraint status. Despite the considerable dispute in measurement of financial constraints, we use the SA index as main approach to gauge a firm s constraint status. Since financial constraint is not directly observable, extant empirical studies identify the 19

22 constrained firms by relying on indirect proxies and popular measures such as the KZ (Kaplan and Zingales, 1997), WW (Whited and Wu, 2006) and SA (Hadlock and Pierce, 2010) indices, which are constructed based on observable firm characteristics such as firm size, age, cash flow, leverage, sales growth, etc. Among the three indices, Kim and Park (2015) argue that KZ index better captures the financial distress characteristics of a financially constrained firm than the WW and SA indices. They find that constrained firms identified by KZ index display a greater preference of debt repayment to investment. Although it is difficult to distinguish financial distress from constraint (Kaplan and Zingales, 2000; Whited and Wu, 2006), we do not intend to use financial constraint to mean financial distress (as discussed in Section 2.1.2). As noted, one of the reasons that we choose Hadlock and Pierce s SA index is that firm size and age are relatively exogenous to firm s financing choices compared to other factors. Second, as Hadlock and Pierce construct the index by updating Kaplan and Zingales (1997) s approach in searching the annual reports and/or 10-Ks of selected firms over the period , their sample period is the closest to ours among the others. We employ the Hadlock and Pierce (2010) s coefficients directly in our sample to construct the financial constraint (SA) index. 3.4 Research specification We estimate the following pooled regression equation to test our main hypothesis H1: å Crashrisk = a + + a SA + a Controls k it, it, k it, k + ( YearDummies) + ( IndustryDummies) + e it, (3) Logistic regressions are used to estimate one-year-ahead crash risk measured by the dummy 20

23 variable crashrisk t+1. In order to support H1, the coefficient on SA should be significantly positive. Following prior literature (e.g. Chen et al., 2001; Jin and Myers, 2006; Hutton et al., 2009; Kim et al., 2011a, b; Callen and Fang, 2013), we include a set of control variables to ensure that the association between financial constraints and future stock price crash risk is not driven by other factors. In testing H2 and H3, we construct two moderator variables of abnormal accruals and corporate tax avoidance. Firstly, we employ the common balance sheet approach (Dechow et al., 1995; Sloan, 1996) to measure the abnormal, or discretionary accruals (da). It is computed as the discretionary portion of the total accruals using the information from financial statements. The detailed definition is presented in the appendix. As discussed in Section 2.2.1, abnormal accruals allow greater subjectivity in recognizing earnings, the resulting financial opacity helps conceal adverse firm performance. Zhu (2016) provides strong evidence on the positive association between accruals and subsequent price crashes. Second, corporate tax avoidance facilitates earnings management and increases financial opacity, providing the opportunities for managers to withhold bad news (Desai and Dharmapala, 2006). Kim et al. (2011a) find that tax avoidance is positively related to future crash risk. We control for firm s residual domestic book-tax difference (ddmpbtd) as a proxy for tax avoidance, which is developed by Desai and Dharmapala (2006) based on Manzon and Plesko (2002) book-tax difference (mpbtd). Following Cheng et al. (2012), we exclude the foreign tax expense from the total tax expenses and regress mpbtd on the total accruals (ta) as follows: mpbtd = b ta + µ + e (4) it, 1 it, it, it, in order to get the residual ε %,' or ddmpbtd, which is expected to be largely free from 21

24 accruals management. We include a sufficient set of control variables in our regression model (3). A number of studies (e.g. Chen et al., 2001; Hutton et al., 2009; Kim et al., 2011a) document the positive relation between firm size and crash risk. Chen et al. (2001) show that negative skewness is more likely in stocks with large market capitalization. We include size (size) as the natural logarithm of market value of a firm at the end of fiscal year t. We control for book-to-market ratio (btm) as a proxy for growth opportunity following Harvey and Siddique (2000) and Chen et al. (2001) who find that growth stocks with low BTMs are more prone to future crashes. Callen and Fang (2013) find consistent evidence with the monitoring view of institutional investors that greater institutional ownership curbs the bad news hoarding behavior and hence reduces the future crash risk. Therefore, we include percentages of institutional investors stock holding (insti) as a proxy for external monitoring. We control for analyst coverage (lanacov) to capture the earnings pressures from financial analysts. Previous studies document a positive relationship between analyst coverage and future crash risk (Chen et al., 2001; Callen and Fang, 2013, 2015). We also include leverage (debt) and return on assets (roa) to control for firm characteristics. Hutton et al. (2009) show that financial leverage and firm s operating performance are both negatively associated with crash risk. In addition, they construct a measure of financial reporting opaqueness and show that opaque firms are more likely to experience future crashes. We compute opacity as the three-year moving sum of the absolute value of annual discretionary accruals (Hutton et al., 2009). Chen et al. (2001) find that stocks with high past returns as well as more volatile stocks are more likely to crash, we control for standard deviations of weekly returns (stdret) in the regression in order to capture the return volatility. Also, Chen et al. (2001) point out that trading volume 22

25 proxies for intensity of disagreement among investors and find that an increase in the differences of opinion would lead to greater negative skewness. It is expected that trading volume (tradevol) is positively correlated with crash risk. Lastly, we control for year-fixed effects and industry-fixed effects in future crash risk by including year and industry dummies in our regression models. All the variables are defined in detail in the appendix. As indicated by Ai and Norton (2003), the coefficients of the interaction terms estimated in a logit model could be biased when using the methods applied in linear models. Therefore, to evaluate the cross-section variations in association between future crash risk and financial constraints, we undertake subsample analyses to test H2 and H3. In testing the incremental effects of abnormal accruals and tax avoidance, we separate our sample into two sub-groups based on the sample median of discretionary accruals (da) and residual book-tax difference (ddmpbtd), respectively. On that basis, we construct indicator variables dda and dddmpbtd to classify corresponding high- and low-accruals subsamples and high- and low-tax avoidance subsamples 3. Therefore, we estimate the model (3) separately for the two subsamples in each test. If the coefficient on SA is significantly more positive for the high-accruals firms than the low-accruals firms, then H2 holds. Similarly, if the coefficient on SA is incrementally less positive for firms with higher level of tax avoidance than lower level of tax avoidance, H3 holds, suggesting the less pronounced association between financial constraints and future crash risk in presence of corporate tax avoidance. 4. EMPIRICAL RESULTS 3 Indicator variable dda (dddmpbtd) equals to 1 for high-accruals (tax avoidance) subsample and 0 for low-accruals (tax avoidance) subsample. 23

26 Table 2 presents the main results to the question whether financial constraints predict future crash risk of the firm. Column (1) reports the results from the logistic regression model using crashrisk as the dependent variable. The coefficient for SA is significantly positive at 0.1% level. This result supports H1, suggesting that crash risk in year t+1 is positively associated with financial constraints in year t. Also, we find that the marginal effect of the financial constraint index on crash risk is 3.34%. It suggests that an increase of one standard deviation in SA leads to an increase in the probability of a crash (crashrisk) by 3.34%, which accounts for 14.16% of the mean value (23.61%) of crashrisk in our sample and is economically significant. The predictive power of financial constraints in crash risk is consistent with both managerial bad news hoarding behavior and higher default risk of the corresponding constrained stocks. Column (2) reports the ordered logistic regression results for crash risk measure count, Column (3) and (4) report the OLS regression results using alternative proxies duvol and minreturn for crash risk, as the robustness checks. The coefficients for SA remain statistically positive at 1% level across all columns, corroborating our inferences. Not surprisingly, size is highly significant (0.1%) with a positive coefficient in all three models, consistent with the findings in Chen et al. (2001) and Hutton et al. (2009). Also, tradevol is positively associated with crash risk at least at the conventional level across all the columns. As expected, institutional ownership (insti) is negatively related to crashrisk as shown in Column (1) to (3), which, however, is not evident with minreturn. The insignificant results may be attributed to the endogeneity of institutional holdings (Callen and Fang, 2013). Table 3 reports the logistic regressions results with crashrisk as the dependent variable for test of H2. As discussed previously, model (3) is estimated separately in two subsamples: 24

27 low-accruals (dda=0) and high-accruals (dda=1) firms based on the sample median of discretionary accruals (da). The coefficient for SA is statistically significant (p=0.001) and positive in high-total accruals firms, which is in line with H1. By contrast, the coefficient for SA in the low-abnormal accruals subsample, albeit positive, is not statistically significant. These results imply that the positive association between future crash risk and financial constraint is only evident for firms with high abnormal accruals, supporting H2. It is consistent with our prediction that the probability of price crash is elevated via the bad news hoarding channel which is facilitated by earnings management. Table 4 shows the logistic regression results for test of H3. As shown in Column (1), the coefficient for SA is significantly positive in low-tax avoidance subsample (dddmpbtd=1) at 5% level. However, the association between financial constraints and one-year-ahead crash risk is not evident in high-tax avoidance subsample (Column (2) with dddmpbtd=0). These results are consistent with H3, suggesting that the positive relation between financial constraints and future crash risk is less pronounced for firms that conduct aggressive tax planning activities. It supports that the purpose of avoiding tax is mainly to generate internal funds, whereby mitigating the financing difficulties of the constrained firms and reducing the crash risk through default risk channel. 5. ROBUSTNESS CHECK AND ADDITIONAL TESTS 5.1 Control for endogeneity Collapse of junk bond market (1989) and future crash risk Further to our control for omitted firm characteristics, we recognize the potential existence of other unobserved cross-sectional heterogeneity that affects both future crash risk 25

28 and financial constraints. To address the endogeneity concerns, we follow Almeida et al. (2013) s approach exploiting the collapse of bond market in 1989 as the exogenous shock to the firm s financial constraints in a difference-in-differences analysis. Lemmon and Roberts (2010) argue that a series unexpected events in the bond market in 1989 leads to a substantial decline in the supply of credit, especially the issues of below-investment-grade credit. As a result, the financial constraints of the firms that rely on the junk bonds prior to the shock tightened in the post-collapse period. Therefore, if there is a casual relationship between financial constraints and crash risk, we would expect that an increase in financial constraints of junk bond-issuing (or speculative-grade) firms leads to significantly higher future crash risk in post-collapse period relative to firms that do not rely on bond financing. Following prior studies, we conduct the difference-in-differences (DID) test over an event window from 1987 to 1992 and assign the three-year periods and as the pre- and post-collapse periods, respectively. We classify firms based on the Standard &Poor s long-term domestic issuer credit rating. According to S&P s ratings, firms rated BBB- and higher are investment-grade (IG). Since we only focus on below-investment grade (or speculative-grade) and unrated bonds, treatment firms are defined as junk bond issuers that are rated BB+ or lower whereas control firms are defined as those without S&P credit rating. We use the following regression model for the DID analysis: Crashrisk = a + a Post + a Junk + + a Post * Junk it, it, 2 it, 3 it, it, å + a Controls + ( YearDummies) + ( IndustryDummies) + e k k k i, t i, t (5) where Post t is the time-indicator variable that equals to 1 if a firm is in the post-collapse period and 0 otherwise; the group-indicator variable Junk t equals to 1 if a firm is junk-bond 26