How Do Financial Constraints Relate to Financial Reporting Quality? Evidence from Seasoned Equity Offerings

Transcription

1 How Do Financial Constraints Relate to Financial Reporting Quality? Evidence from Seasoned Equity Offerings Ahmet C. Kurt Katz Graduate School of Business University of Pittsburgh This draft: July 2011 Abstract: This paper examines the relation between firms financial constraints and their financial reporting during periods when they attempt to raise equity capital. Specifically, I investigate whether financially constrained firms tend to manipulate their earnings more aggressively around seasoned equity offerings (SEOs) as compared to financially unconstrained firms. I begin by presenting a simple rational expectations framework which links issuers pre-seo financing constraints to their earnings management strategies around the offering. By using different measures of financial constraints, I document that constrained issuers, which cannot credibly signal the absence of aggressive earnings management, report higher income-increasing accruals than unconstrained issuers in the year of the offering. Accordingly, investors discount the stock prices of constrained issuers more than those of unconstrained issuers at the announcements of SEOs. The evidence suggests that the aggressive earnings management by constrained issuers is not simply the result of managerial opportunism but rather a rational response to anticipated market behavior at offering announcements. Keywords: Financial Constraints; Financial Reporting; Earnings Management; Seasoned equity offerings JEL classification: G32, M41 * I appreciate helpful comments from participants at the 2011 Eastern Finance Association meetings and Midwest Finance Association meetings.

2 1. Introduction A significant body of research has examined the relation between financial constraints and firms real decisions, particularly investment policy (e.g., Fazzari, Hubbard, and Petersen, 1988; Kaplan and Zingales, 1997; Cleary, 1999; Campello, Graham, and Harvey, 2010). However, the link between financial constraints and firms non-operational decisions, namely corporate financial reporting, has not been previously explored. This study attempts to fill this gap. Previous studies examining managers financial reporting behavior hypothesize that the trade-off between the costs and benefits of manipulating earnings changes during periods when a firm raises capital (for a review, see Dechow, Ge, and Schrand, 2010). In particular, higher utility associated with the availability or pricing of external capital may increase the benefits of opportunistic reporting. This effect may be more pronounced for financially constrained firms, which incur higher transaction and agency costs when they attempt to raise external capital. Accordingly, I compare the earnings management strategies of financially constrained and unconstrained firms around seasoned equity offerings (SEOs) to test the connection between financial constraints and financial reporting. 1 Specifically, I develop and test a simple rational expectations framework which predicts that the quality of financial reporting around SEOs is lower for constrained firms as compared to unconstrained firms (i.e., constrained issuers report higher income-increasing accruals). The distinctive aspect of the framework is that it suggests that the aggressive earnings management by constrained firms is a rational response to anticipated market behavior at offering announcements (i.e., higher discounting by investors) rather than simply being the result of managerial opportunism. 1 SEO firms are not shut out of the financial markets because they have access to equity markets. Following Korajczyk and Levy (2003), I use the terms constrained and unconstrained to denote a relative relation. However, it is worth noting that most issuers operate on a tight financial leash prior to the SEO (DeAngelo, DeAngelo, and Stulz, 2010). In fact, DeAngelo et al. (2010, p.276) point out: most issuers would have run out of cash by the year after the SEO had they not received the offer proceeds. 1

3 A number of studies have documented that SEO firms, on average, report higher earnings by altering discretionary accruals around the offering (e.g., Teoh, Welch, and Wong, 1998a; Rangan, 1998; Shivakumar, 2000). However, as Dechow et al. (2010, p.384) point out: while the studies provide fairly consistent evidence of accruals management when firms raise capital, they do not expand the analysis to examine cross-sectional variation in accruals management. Notably, why do some issuers manage their earnings aggressively, rendering themselves vulnerable to litigation (DuCharme, Malatesta, and Sefcik, 2004), whereas others choose to manage their earnings more conservatively? I propose that issuers financial constraints frictions preventing firms from funding all desired investments 2 play an important role in shaping their financial reporting behavior around SEOs. Korajczyk and Levy (2003, p.82) write: Theoretically, we define a firm as financially constrained if it does not have sufficient cash to undertake investment opportunities and if it faces severe agency costs when accessing financial markets. In other words, everything else equal, financially constrained firms are more likely to forego projects with positive net present value due to lack of funding available to these firms. Therefore, financially constrained (C) firms are expected to have greater incentive than financially unconstrained (UC) firms to use earnings management as a means of reducing their financing needs and enhance their capacity to finance profitable projects. Thus, SEOs provide an interesting setting to examine the interplay between firms financing constraints and financial reporting behavior. The main hypothesis of this paper is that C firms use income-increasing accruals more aggressively than UC firms around equity offerings. The logic underlying this argument is based on a revised version of the rational expectations model of Shivakumar (2000). That is, investors are rational and anticipate that issuers 2 Lamont et al. (2001, p. 529) note: This inability to fund investments might be due to credit constraints, or inability to borrow, inability to issue equity, dependence on bank loans, or illiquidity of assets. They also point out that they do not use financial constraints to mean financial distress, which is associated with financial constraints. 2

4 facing a high level of financial constraints have a greater tendency to manage earnings upwards than unconstrained issuers. Relatedly, they discount C issuers stock prices more than UC issuers stock prices at offering announcements. Therefore, C issuers rationally report higher positive discretionary accruals than UC issuers. There are several reasons that may lead investors to anticipate that earnings reported by C issuers are managed more aggressively than those of UC issuers. First, the need to raise external financing at favorable prices is an important reason why managers window-dress the financial reports (e.g., Dechow, Sloan, and Sweeney, 1996; Firth, Rui, and Wu, 2011). Given that the value of an additional dollar raised through the offering is higher for C firms than UC firms (Faulkender and Wang, 2006), the utility associated with aggressive earnings management is greater for C issuers. Second, higher offering price ensures that the issuing firm raises a targeted amount of capital with selling less number of shares, preventing the issuer from diluting higher level of ownership to outside investors (Kim and Park, 2005). As the amount of capital raised through the SEO increases, the marginal benefit of managing earnings upward (i.e., additional decrease in dilution of ownership due to increased stock price) increases. Since, everything else equal, C issuers tend to raise a larger amount of capital than UC issuers, the marginal benefit of earnings management is higher for C issuers. Third, C firms are subject to higher information asymmetry (Almeida and Campello 2010), decreasing the credibility of their financial statements. As opposed to the short-term benefits of earnings management, there are certain costs associated with managing earnings aggressively such as higher likelihood of being subject to investigations of the Securities and Exchange Commission (SEC) and deterioration in future operating performance (e.g., Teoh, Wong, and Rao, 1998; DuCharme et al., 2004; Karpoff, Lee, and Martin, 2008). In my framework, I assume that the market participants believe that the 3

5 difference between potential costs and benefits of the aggressive use of income-increasing accruals around the offering is high enough to prevent UC issuers from managing their earnings aggressively. I also contend that UC firms can reveal their type credibly by underfinancing (i.e., raising a smaller amount of equity than the amount required to undertake the upcoming project and financing the remaining amount with internal capital or debt). However, due to the reasons listed above, C issuers cannot credibly signal the absence of aggressive earnings management. Consistent with my conjecture, by using different measures of financial constraints (i.e., firm size, payout ratio, the Whited and Wu (2006) index, and the Size-Age index of Hadlock and Pierce (2010)), I find that the difference in median performance-adjusted discretionary current accruals (as a percentage of total assets) between the constrained and unconstrained groups is significantly positive in the SEO year (e.g., 4.51% vs. 0.53% under the size classification). This result is robust to controlling for several variables that may affect the level of discretionary current accruals such as growth opportunities, operational volatility, analyst following, and CEO equity holdings, as well as using the instrumental variable approach. Furthermore, I document that C firms, but not UC firms, consistently manage their earnings upward during the three-year period prior to the offering. However, the difference in discretionary current accruals between the two groups dissipates after the first year following the offering. As predicted, I also find that the SEO announcement returns of C firms are significantly more negative than those of UC firms (e.g., median seven-day CAR: -4.30% vs % under the size classification), suggesting that investors rationally discount C firms stock prices to a greater extent at offering announcements. The findings of this paper contribute to the literature on financial constraints, as well as earnings management around corporate events. The results suggest that firms financial reporting quality varies with the level of their financing constrains during periods when they raise equity 4

6 capital. To the best of my knowledge, this is the first study to document a connection between financial constraints and firms non-operational decisions. Particularly, this study demonstrates an alternative avenue, i.e., financial reporting, through which financial constraints may be linked to firm value and thus, complements and extends previous research documenting the valuation impact of financial constraints by examining firms real decisions such as investment and cash policy (e.g., Denis and Sibilkov, 2010; Faulkender and Wang, 2006). Furthermore, this paper extends our current understanding of earnings management around corporate events in two important ways. First, as pointed out by Dechow et al. (2010), previous research has paid scant attention to the determinants of cross-sectional variation in discretionary accruals reported by firms issuing SEOs. I document that issuers pre-offering financial constraints are an important factor in explaining the level of discretionary current accruals reported in the year of the SEO. In fact, the results suggest that positive earnings management around SEOs documented in the previous literature seems to be driven by the aggressive earnings management of firms that face relatively higher financial constraints. Second, this study contributes to the nascent empirical literature that proposes rational explanations for the observed changes in managers reporting behavior around corporate events (Erickson and Wang, 1999; Shivakumar, 2000). Given that the extant empirical literature on earnings management is dominated by studies suggesting that the use of discretionary accruals around corporate events is the result of opportunistic behavior of managers 3, who are assumed to use their financial reporting discretion to mislead market participants, this research can be seen as a part of a challenging task aimed at reconciling the evidences of earnings management with the theory of efficient capital markets. 3 There is also a stream of literature suggesting that managers report positive discretionary accruals to signal their private information to the market participants (e.g., Louis and Robinson, 2005; Louis and White, 2007). For instance, Louis and Robinson (2005) argue that managers report income-increasing accruals prior to stock splits to communicate their favorable beliefs to the market. They document a positive relation between the income-increasing accruals and stock split announcement returns, suggesting that investors do perceive positive discretionary accruals reported prior to stock splits as the result of managerial optimism (rather than managerial opportunism). 5

7 The rest of the paper is organized as follows. Section 2 reviews the related literature. Section 3 develops a simple rational expectations framework which links issuers earnings management strategies to their pre-seo financing constraints and presents testable hypotheses. Section 4 discusses the measures of financial constraints and earnings management. Section 5 describes the sample. Section 6 presents the empirical models used to test the hypotheses and the results of the tests. Section 7 concludes the paper by discussing the importance of the findings. 2. Related Literature Previous research suggests that incentives to influence equity market valuation affect firms financial reporting behavior. In particular, several studies have documented that firms report income-increasing accruals around initial public offerings (IPOs) (e.g., Friedlan, 1994; Teoh, Welch, and Wong, 1998b; Morsfield and Tan, 2006), seasoned equity offerings (SEOs) (e.g., Teoh et al., 1998a; Rangan, 1998; Shivakumar, 2000; Cohen and Zarowin, 2010), and stock-for-stock acquisitions (e.g., Erickson and Wang, 1999; Louis, 2004). Teoh et al. (1998b) maintain that upward earnings management enables IPO firms to raise higher capital as buyers, who are guided by earnings but are unaware of earnings inflation, pay higher prices for the equity offered than the level justified by unmanaged earnings. The authors write (p. 1941): Our hypothesis is that the marginal investor does not rationally discount for earnings management in forming expectations about future cash flows. In parallel, Teoh et al. (1998a) and Rangan (1998) argue that firms issuing SEOs use income-increasing accruals to temporarily manipulate their stock prices and that investors overvalue issuers at the time of the offering and are subsequently disappointed by declines in the post-seo earnings performance stemming from the reversal of discretionary accruals. In other words, the motivation of managers 6

8 reporting positive discretionary accruals prior to SEOs is to mislead investors who naïvely extrapolate increases in pre-seo earnings. In contrast, Shivakumar (2000) asserts that pre-seo earnings management cannot be designed to mislead investors. Alternatively, he proposes the managerial response hypothesis which suggests that investors anticipate earnings management before the offering and discount issuers stock prices at the announcement and that earnings management is the rational response of issuers to anticipated market behavior at offering announcements. 4 In that sense, the observed financial reporting behavior of issuers can be viewed as the Nash equilibrium outcome of a non-cooperative game between issuers and the stock market. Erickson and Wang (1999) propose a similar explanation for the reporting behavior of managers of acquiring firms around stock-for-stock mergers. They argue that the target firm would rationally anticipate that the acquirer would try to boost its pre-merger stock price by reporting income-increasing accruals. Accordingly, the target discounts the value of the acquirer s stock. Thus, if the acquirer did not manage its earnings upward as expected by the target, it would end up offering higher number of shares for the target, increasing the cost of acquisition. As a result, it is only rational for the acquirers to overstate earnings prior to stock-for-stock acquisitions. Modeling earnings management prior to equity offerings as the outcome of rational expectations model is consistent with Stein s (1989) and Narayanan s (1985) models of myopic corporate behavior. In these models, managers attempt to boost short-term performance (by making decisions that provide immediate cash flows but hurt long-term performance) even though they do not gain anything in the equilibrium as rational investors anticipate their actions and discount reported performance. Despite the fact that managers do not gain from their myopic behaviors, they cannot afford to deviate from non-cooperative equilibrium, since it would only 4 Although Shivakumar s model explains the presence of earnings management around SEOs, the model is silent about what determines the magnitude of earnings management undertaken by issuers. 7

9 make them worse off. As Stein (1989, p.668) succinctly puts it, The Nash approach clearly exposes the fallacy inherent in a statement such as since managers can t systematically fool the market, they won t bother trying. In the next section, I develop a simple rational expectations framework that links firms financial constraints to their financial reporting strategies during the period when they attempt to raise equity. By allowing unconstrained firms (i.e., good types) to credibly signal their type to the market participants, the framework attempts to shed some light on the cross-sectional variation in the magnitude of earnings management activity observed around SEOs, as well as the SEO announcement returns. 3. Hypothesis Development 3.1. A simple rational expectations framework We consider two firms: firm 1 is financially constrained (C), whereas firm 2 is financially unconstrained (UC). Both firms have the opportunity to invest in a positive NPV project. The cost of the project is I, which is common knowledge and needs to be paid before the firms undertake the project. Managers and investors observe the same signal about the payoff of the project, though their interpretation of this signal may be different (Dittmar and Thakor, 2007). 5 While the C firm has to raise the entire capital required to fund the project through equity financing (i.e., EF C = I), the UC firm has the ability to finance some proportion of the cost of the project (but not the entire amount) through internal capital or debt (i.e., 0 < EF UC < I). 6 5 For ease of exposition, I assume that there is only one type of project available to both firms and it pays $M > I for sure (i.e., the mundane project as characterized by Dittmar and Thakor (2007)). I operationalize this assumption in my tests by controlling for firms market-to-book ratio, which is a proxy for investment opportunities. The information asymmetry between the managers and investors arises mainly due to uncertainty regarding the value of assets-in-place, V. When firms reveal their types at the offering announcement, investors revise their valuations to adjust for the impact of aggressive and conservative earnings management on V. 6 I assume that the UC firm does not finance the project fully with internal funds or debt and thus, issue some equity to 8

10 Both firms manage their earnings upward prior to the announcement of the equity offering. This assumption is plausible since Shivakumar (2000) shows that it is only rational for issuing firms to engage in positive earnings management. Therefore, the magnitude of earnings management is always greater than zero but managers can choose to manage earnings aggressively or conservatively. However, investors cannot perfectly discern the magnitude of the firms earnings management activity. 7 They infer the amount of earnings management undertaken by the firms based on their types. There is no information asymmetry between managers and investors regarding the firm s financial constraint status at the time of the equity announcement. Firms reveal their types (i.e., constrained vs. unconstrained) by announcing how much equity (i.e., EF C or EF UC ) they want to raise. In this setting, underfinancing (i.e., raising EF UC ) works as a credible signal since it is costly for the C firm to commit such an action: the C firm cannot undertake the positive NPV project (i.e., forego future profits) unless it chooses to raise EF C. On the other hand, even if the UC firm chooses to raise EF UC, it can still undertake the project by financing the remaining amount (i.e., I - EF UC ) through internal funds or debt. 8 Accordingly, I model the earnings management before the equity offering as the outcome of the rational expectations model: Investors expect the C (UC) issuer announcing equity offering to manage earnings aggressively (conservatively) and discount its stock price at a higher (lower) rate, and consistent with this undertake the positive NPV project (i.e., 0 < EF UC ). This may be due to several reasons such as maintaining a target level of cash (e.g., Duchin and Dittmar, 2010) or leverage (e.g., Hovakimian, Hovakimian, and Tehranian, 2004; Flannery and Rangan, 2006), and avoiding the underinvestment problem (Myers, 1977; Stulz, 1990). Furthermore, although EF UC may be equal to I, in an economy with rational managers and investors, EF UC is always less than I (and EF C ), which is a necessary condition for separation to occur in my framework. 7 As pointed out by Louis and White (2007, p.214), managers have so much discretion in reporting their firms financial performance that it is improbable for an outsider to know exactly by how much they have manipulated their reports. 8 This is consistent with Hennessy, Livdan, and Miranda (2010) who show that firms signal their types to outside investors by choosing to what extent they finance their projects with equity. They find that bad types do not use debt, hoard cash, and issue only equity. On the other hand, good types do not use full equity financing and signal their type through choosing a mix of equity and debt financing. In addition, cash flows allow good types to substitute internal funds for debt. One could argue that if the UC firm chooses to issue debt simultaneously with equity, it will be subject to tight scrutiny of debt financiers, limiting its ability to manage earnings. In the empirical section, I exclude simultaneous offerings from the sample to perform a cleaner test of the proposed hypotheses. 9

11 behavior, the C firm overstates reported earnings more than the UC firm prior to the offering. To formalize this argument, consider the games depicted in Figures 1 and 2. Figure 1: Payoffs from earnings management game between the constrained (C) firm and market participants At offering announcement Investors believe that the C firm manages earnings conservatively. Investors believe that the C firm manages earnings aggressively. Before offering announcement C firm manages earnings conservatively C firm manages earnings aggressively (0, 0) (H-K, -H) (-H, H) (-E, -E) In this game, investors have two strategies. They either assume that the C firm manages earnings conservatively or believe that the C firm manages earnings aggressively. If market participants believe that a firm manages earnings aggressively, they discount the firm s stock price at a higher rate. The first entry in each box is the payoff to the C issuer, while the second entry is the payoff to market participants. E represents a positive short-term cost of managing earnings aggressively (e.g., cost of diverted resources). On the other hand, without any loss of generality, I assume that the cost of managing earnings conservatively is zero. H represents a net positive short-term payoff (after accounting for E) of earnings management (e.g., % decrease in diluted ownership due to increased share price; in cell 3 (i.e., -H, H) new shareholders get this payoff, since the firm dilutes larger ownership to raise the target amount of capital). K stands for the expected post-seo cost of aggressive earnings management to the extent not anticipated by the investors (e.g., litigation costs). 9 I assume that K < H. That is, the benefit of aggressive earnings management exceeds its expected cost to the C firm, which relies heavily on offering proceeds to undertake the upcoming positive NPV project. This assumption is consistent with previous studies (e.g., Dechow et al., 1996; Dechow et al. 2011) suggesting that the benefit of earnings 9 DuCharme et al. (2004) document that the probability of post-seo lawsuits filed by shareholders against the issuers and settlement amounts increase with pre-seo discretionary accruals. 10

12 misstatement seems to exceed the expected cost of misstatement for the firms seeking to secure external financing, leading these firms to manipulate their earnings but in turn, face enforcement actions by the SEC. If investors can directly observe earnings management, the co-operative outcome is given by the first cell. However, since they cannot perfectly discern the amount of earnings management, the C firm has an incentive to deviate from the co-operative equilibrium. That is, by managing earnings aggressively, the manager of the C firm might try to mislead investors and sell newly issued shares at an artificially high price. Given this incentive, investors assume that the C firm manages reported earnings aggressively and discount its stock price at a higher rate. Thus, in this case, even if the C firm does not manage earnings aggressively, it suffers from a decrease in its stock price. Therefore, it is only rational for the C firm to manage earnings aggressively prior to the announcement, at least to the extent expected by investors. This results in cell 4 (i.e., -E, -E) as equilibrium outcome, which is a unique Nash equilibrium in pure strategies. However, this outcome is Pareto inefficient since E > 0. Next we turn into earnings management game between the UC firm and market participants. Figure 2: Payoffs from earnings management game between the unconstrained (UC) firm and market participants At offering announcement Investors believe that UC firm manages earnings conservatively. Investors believe that UC firm manages earnings aggressively. Before offering announcement UC firm manages earnings conservatively UC firm manages earnings aggressively (0, 0) (L-K, -L) (-L, L) (-E, -E) The structure of the game between the UC firm and investors is similar to the first one. E stands for the short-term cost of earnings management, L stands for a short-term net positive payoff, and K represents the expected post-seo cost of aggressive earnings management. 11

13 However, in this case, I assume that L K. That is, the benefit of aggressive earnings management does not exceed the expected cost of such activity to the UC firms, which relies less on external equity as well as faces lower cost of financing (though it may rise sharply if the firm is revealed to be an aggressive earnings manipulator by the SEC; Dechow et al., 1996). Further, L < H since the marginal benefit of managing earnings upward and raising the stock price is higher for the C firm than for the UC firm since such an action results in a greater decrease in the percentage of ownership diluted to outside investors in the case of the C firm given that it raises a larger amount of equity than the UC firm. Therefore, for any K that lies between L and H, there is no incentive for the UC firm to deviate from the co-operative outcome (whereas the opposite is true for the C firm). Knowing that the UC firm does not have an incentive to manage earnings aggressively, investors assume that the UC firm manages its earnings conservatively and discount the price of newly issued shares at a lower rate. Based on this game structure, I propose the following hypotheses: H1: Ceteris paribus, financially constrained firms manage their earnings more aggressively around SEOs than financially unconstrained firms. H2: Ceteris paribus, financially unconstrained firms raise a smaller amount of capital through the offering (i.e., underfinance) as compared to financially constrained firms. H3: Ceteris paribus, SEO announcement returns are lower for financially constrained firms as compared to financially unconstrained firms Alternative hypotheses In the Myers-Majluf (1984) model, managers, who are better informed than investors about the value of the firm s assets-in-place (as well as its growth opportunities), prefer to issue equity when their private valuation is lower than that of the market participants. However, investors, who rationally infer this managerial preference, reduce the value of offering firms at the 12

14 announcement. Accordingly, the model predicts that firms with ample financial slack and debt capacity choose to finance their projects first with internal funds, then with debt, and finally with equity. This suggests that an equity offering announcement made by the unconstrained firm would generate a more negative response among investors as compared to that made by the constrained firm (since the equity offered by the unconstrained firm is believed to be more overvalued, everything else equal). If this is the case, one would expect that unconstrained firms report higher income-increasing accruals prior to the offering as a response to anticipated investor behavior at offering announcements. Also, since the revelation of being an unconstrained firm is associated with higher cost (i.e., lower announcement returns), unconstrained firms are expected to try to conceal their type rather than signal it through underfinancing. Thus, overall, the alternative hypotheses are the exact opposite of H1, H2, and H3. 4. Variable Measurement 4.1. Financial constraints criteria A number of alternative measures of financial constraints have been proposed in the literature. However, there is no agreement on which measure is the best proxy for financial constraints. Therefore, previous studies (e.g., Almeida, Campello, and Weisbach, 2004; Denis and Sibilkov, 2010; Duchin, 2010) use several alternative measures instead of a single measure to avoid potential problems stemming from the misclassification of firms. Following previous studies, I use four alternative approaches to sort firms into the financially constrained and unconstrained groups. 1. Firm size: Following previous research (e.g., Almeida et al., 2004; Denis and Sibilkov, 2010), I rank the sample firms based on their inflation-adjusted book value of assets (Compustat item 13

15 AT) at the end of the year preceding the offering year and assign those firms in the bottom (top) tercile of the distribution to the financially constrained (unconstrained) group. 10 The reasoning that lies behind this approach is that small firms are younger and less well-known and thus, more vulnerable to capital market imperfections stemming from information asymmetries and collateral constraints. Hennessy and Whited (2007) estimate that financing costs are almost twice as large for small firms as for large firms. 2. Payout Ratio: Following prior studies (e.g., Almeida and Campello, 2007, 2010; Fazzari, Hubbard, and Petersen, 1988) suggesting that unconstrained firms are more likely to have higher payout ratios, I rank the sample firms based on their payout ratio at the end of the year preceding the offering year and assign those firms in the bottom (top) tercile of the distribution to the financially constrained (unconstrained) group. Payout ratio is defined as the ratio of dividends (Compustat item DVC plus Compustat item DVP) and share repurchases (Compustat item PRSTKC) to operating income (Compustat item OIBDP). Payout ratio is set equal to 1 if a firm has negative operating income and positive payout (e.g., Li and Zhang, 2010; Hadlock and Pierce, 2010). 3. The Whited and Wu (2006) index: Employing a structural investment model, Whited and Wu (2006) develop a structural index of firms external finance constraints by using all Compustat firms for the period of The index is a combination of six factors: cash flow, dividend dummy, leverage, firm size, industry sales growth, and firm sales growth. The index increases with the firm s financial constraints and is calculated as follows: WW it = *CF it *DIVPOS it *TLTD it *LNTA it 10 I do not discard those observations in the middle tercile (according to a financial constraints criterion) and keep such opaque issuers in the sample to be used in the regression analysis. Thus, I create a dummy variable (labeled Medium-FC ) that is equal to 1 if an issuer is in the middle tercile and 0 otherwise. Although I expect that these firms also manage their earnings more aggressively as compared to financially unconstrained firms, the significance of the difference in discretionary current accruals between the two groups is ultimately an empirical question. 14

16 *ISG it *SG it (1) where, CF is cash flow (Compustat item IB plus Compustat item DP) divided by lagged total assets; DIVPOS is a dummy variable equal to 1 if the firm pays dividends and 0 otherwise; TLTD is long-term debt (Compustat item DLTT) divided by total assets; LNTA is the natural log of inflation-adjusted total assets; ISG is the firm s three-digit SIC industry annual sales growth; SG is the firm s annual sales growth. I rank the sample firms based on their WW index at the end of the year preceding the offering year and assign those firms in the top (bottom) tercile of the distribution to the financially constrained (unconstrained) group. This approach is similar to those in Franzoni (2009), Fahlenbrach and Stulz (2009) and Duchin (2010). 4. The Size-Age (SA) Index: Recently, Hadlock and Pierce (2010) develop a financial constraints index using two factors: firm size and age. They obtain the index loadings via order logit regression where the dependent variable ranges from 1 (least constrained) to 5 (most constrained) and is coded based on a firm s level of constraints identified through manual inspection of the firm s annual reports and 10-K filings. The index is calculated as follows: SA it = *SIZE it *(SIZE it ) *AGE it (2) where, SIZE is the natural log of inflation-adjusted total assets and AGE is the number of years the firm is listed with a non-missing stock price on Compustat. As suggested by the authors, in calculating the index, SIZE is winsorized at the natural log of $4.5 billion and AGE is winsorized at thirty-seven years. Since the SA index is higher for financially constrained firms, I assign those firms in the top (bottom) tercile of the distribution of the index value at the end of the year preceding the offering year to the financially constrained (unconstrained) group Earnings management Following Teoh et al. (1998a, 1998b) and Louis (2004), I employ discretionary current 15

17 accruals as a measure of earnings management and use the cross-sectional adoption of the modified-jones model (Dechow, Sloan, and Sweeney, 1995) to obtain the discretionary current accruals for a given year. Specifically, for each industry (two-digit SIC code) and each year, I estimated equation (3) using all firms that have necessary accounting data on Compustat and with assets greater than $1M. CA it = β 0 + β 1 (1 / ASSET it-1 ) + β 2 ΔSALE it + ε i (3) where, CA is current accruals divided by total assets at the end of year t-1; ASSET is total assets; ΔSALE is the change in sales (Compustat item SALE) from the prior year divided by total assets at the end of year t-1; ε is the regression residual. 11 Following previous studies (e.g., Teoh et al., 1998a; Fischer and Louis, 2008; Hovakimian and Hutton, 2010), current accruals is defined as change in non-cash current assets (Compustat item ACT minus Compustat item CHE) minus change in current liabilities (Compustat item LCT) plus change in the current portion of long-term debt (Compustat item DD1). Equation (4) estimates expected current accruals using the coefficient estimates obtained from equation (3) with an adjustment for the change in accounts receivables: ECA it = β 0 + β 1 (1 / ASSET it-1 ) + β 2(ΔSALE it - ΔAR it ) (4) where, β 0, β 1, and β 2 are estimated coefficients from equation (3); ECA is the expected current accruals; ΔAR is the change in accounts receivable (Compustat item RECT) from the prior year divided by total assets at the end of year t-1. Discretionary current accruals (DCA) is the difference between current accruals and expected current accruals. Finally, I adjust the DCA obtained from the modified-jones model for performance since Kothari, Leone, and Wasley (2005) show that performance-adjusted accruals are well-specified 11 To mitigate the effects of outliers, I winsorize CA, 1/ ASSET and ΔSALE at the 1% and 99% levels. I also require there be at least 20 observations in each industry and year to accurately estimate the regression coefficients. 16

18 and yield powerful tests. Specifically, following previous studies (e.g., Louis, 2004; Louis and Robinson, 2005; Gong, Louis, and Sun, 2008), for each year and each industry, I build five portfolios by sorting the data into quintiles based on return-on-assets (i.e., income before extraordinary items (Compustat item IB) divided by lagged total assets) from year -2 relative to the issuance year. 12 The performance-adjusted DCA for a sample firm is calculated as the DCA of the firm minus the median DCA for its respective industry-performance-matched portfolio. As suggested by Gong et al. (2008), the portfolio benchmarking approach allows researchers to control not only for performance but also for random effects stemming from other events which may impact accruals or other managerial incentives to engage in earnings management. 5. Sample Formation and Descriptive Statistics The initial sample of SEOs is obtained from the Security Data Corporation s (SDC) New Issues Database for the period over 1983 to Following previous studies (e.g., Lee and Masulis, 2009; Cohen and Zarowin, 2010), I exclude the following: (1) SEOs lacking Compustat annual financial statement data for the four years prior to the SEO filing date 13, (2) close-end funds, unit investment trusts, REITs, and limited partnerships, (3) spin-offs, (4) reverse LBOs, (5) rights issues, (6) pure secondary offerings, (7) simultaneous or combined offers of several classes of securities such as unit offers and warrants, (8) non-domestic and simultaneous domestic-international offers, (9) offering of securities with CRSP share codes other than 10 or 11, (10) SEOs with offer prices less than $5. Furthermore, financial firms (SIC codes ) and 12 Teoh et al. (1998a) document that issuing firms report positive discretionary current accruals in year -1. Therefore, performance-matching based on the ROA reported in year -1 could lead to biased results. However, I reran all the analyses by creating matching portfolios based on the ROA reported in year -1, as well as year 0 and obtained similar results. 13 This restriction is likely to induce survivorship bias, resulting in the inclusion of larger and more successful firms. However, I expect that this will reduce the variation in the earnings management and financial constraints measures and thus, result in a more conservative test of my hypotheses. 17

19 utilities (SIC codes ) are excluded from the sample due to the greater regulation for these firms, limiting their ability to engage in earnings management. The final sample includes 1,645 SEOs for which the accruals data are available in the year of the offering. Summary statistics are reported in Table 1. Panel A reports the size characteristics as of the end of year -1 (where year 0 is the offering year). Panel B and C present the year and industry distribution, respectively. The results suggest that the sample is not dominated by a particular year, though 1983 stands out as a very active year for SEOs as it contains 10.88% of the sample. Furthermore, examination of the industry distribution reveals that the computer equipment/services companies and electronic equipment companies constitute 15.02% and 11.19% of the sample, respectively. Table 2 presents the median firm characteristics (as of the end of year -1) by financial constraints categories under each classification. The results consistently suggest that financially constrained firms tend to have higher market-to-book ratios and experience higher volatility of cash flows, revenue, and sales growth. Under each category, the median financially constrained firm does not pay any dividends or repurchase any shares in the pre-seo year and is followed by only one analyst (two analysts under the payout ratio classification). On the other hand, the median unconstrained firm has positive payout activity in the year prior to the offering and is followed by higher number of analysts. Although the size and SA index measures suggest that unconstrained firms report lower industry-adjusted ROA as compared to constrained firms, the payout ratio measure does not reveal any significant difference in ROA between the two groups. On the other hand, according to the WW index classification, the unconstrained group reports higher industry-adjusted ROA than the constrained group. Furthermore, the results reveal that the offer size (scaled by market value or sales) is significantly larger for the constrained group versus unconstrained group under all classifications. 18

20 6. Empirical Analysis and Results 6.1. Univariate results Table 3 (and Figure 3) presents time-series of performance-adjusted discretionary current accruals (DCA), in percent, for financially constrained and unconstrained firms under different classifications from year -3 to +3 relative to the offering (year 0). Following previous studies (e.g., Cohen and Zarowin, 2010; Shivakumar, 2000), I report medians as they are less likely than averages to be influenced by extreme observations. The results show that under all financial constraint classifications, the median DCA reported by the constrained group in the year of the offering is significantly higher than that of the unconstrained group. For instance, under the firm size classification, the median DCA of the constrained group is 4.51% of total assets, whereas it is 0.53% of total assets for the unconstrained group and the difference between the two groups is statistically significant at 1% level. The median DCA of the constrained (unconstrained) firms under the payout ratio, WW index and SA index classifications is 2.96% (1.15%), 3.90% (0.85%), and 4.12% (0.85%), respectively, and all the differences are significant at 1% level. These results suggest that constrained issuers manage their earnings more aggressively as compared to unconstrained issuers during the year of the SEO, providing initial support for H1. Furthermore, the comparison of the two groups earnings management activity in the years prior to the offering reveals that constrained, but not unconstrained, issuers manage their earnings upward in the pre-seo period as well, albeit the magnitude of income-increasing accruals is smaller as compared to those in the offering year. For instance, under the size classification, the median DCA for the constrained group is 0.96% (p <.01) in the year preceding the announcement, whereas it is only 0.29% (p >.10) for the unconstrained group. 19

21 Finally, the examination of the post-offering discretionary accruals suggests that there is no significant difference in the financial reporting behavior of the two groups beyond the first year following the offering. That is, although constrained issuers continue to manage their earnings in the period immediately following the offering, the difference in median DCA between the two groups dissipates in years +2 and +3. Specifically, both groups report negative but insignificant DCA in year +3. This finding suggests that the documented divergence in the earnings management strategies of constrained and unconstrained firms does not manifest itself during the post-offering period Median regressions of discretionary current accruals in the SEO year To perform a test of H1 in a multivariate setting, I estimate the following model (separately for each financial constraint criterion) using median regression with bootstrapped standard errors (with 200 replications). Since previous studies (Teoh et al., 1998a; Rangan, 1998) show that earnings management activity is at its peak during the offering year, I use performance-adjusted DCA calculated for the year of the offering as a measure of earnings management undertaken by the issuers around the SEO. The model includes several control variables suggested in the literature (e.g., Hribar and Nichols, 2007; Cohen and Zarowin, 2010). PADJDCA = α 0 + α 1 HIGH-FC+ α 2 MEDIUM-FC+ α 3 MB+ α 4 ADJROA+ α 5 CFVOL + α 6 REVVOL + α 7 SGVOL + α 8 LITIGATION + Year fixed effects + ν (5) where, PADJDCA is the performance-adjusted discretionary current accruals (as a percentage of total assets) in the year of the offering; HIGH-FC is a dummy variable equal to 1 if a firm is categorized as financially constrained under a sorting criterion and 0 otherwise; MEDIUM-FC is a dummy variable equal to 1 if a firm is ranked in the middle tercile according to a financial constraint sorting criterion and 0 otherwise; MB is the natural log of the market-to-book ratio in the 20

22 year prior to the SEO year; ADJROA is the industry-median adjusted ROA (i.e., income before extraordinary items divided by lagged total assets) in the year prior to the SEO year; CFVOL is the standard deviation of cash flow deflated by total assets over the five year period (with a minimum of three years) prior to the SEO year; REVVOL is the standard deviation of sales deflated by total assets over the five year period (with a minimum of three years) prior to the SEO year; SGVOL is the standard deviation of annual sales growth over the five year period (with a minimum of three years) prior to the SEO year; LITIGATION is a dummy variable equal to 1 if a firm s SIC code is , , , , and 0 otherwise (Barton and Simko, 2002). The high litigation risk industries include pharmaceuticals/biotechnology and computers/electronics. 14 The model includes market-to-book ratio and industry-adjusted ROA to control for variations in issuers growth opportunities and profitability, respectively. I conjecture that issuers with higher growth opportunities and recent profitability tend to report greater income-increasing accruals. Further, I also control for cash flow volatility, revenue volatility and sales growth volatility to ensure that the results are not driven by the more volatile operating environments of financially constrained firms. Finally, since firms operating in high litigation risk industries use discretionary accruals less aggressively (Cohen and Zarowin, 2010), the model includes an industry-based litigation dummy variable. I expect a negative sign on the litigation variable. Table 4 (columns 1, 3, 5, and 7) presents the results of the median regressions. The results for the firm size measure are reported under column 1. As predicted, the coefficient on financially constrained dummy is positive (4.438) and significant at 1% level, suggesting that the constrained 14 Although it is not stated in equation (5), the natural log of the pre-offer market value is added to the model as an additional control variable when the payout ratio measure is used to determine the financial constraint groups. The reason why I do not control for the market value when other financial constraint measures are employed is that the market value is a proxy for the firm size, which is already captured in the other constraint measures. The Spearman rank (Pearson) correlation of ln(market value) with ln(assets), payout ratio, the WW index, and the SA index is, respectively, (0.805), (0.091), (-0.699), and (-0.654). 21

23 issuers report higher performance-adjusted DCA in the year of SEO than unconstrained issuers. Similar results are obtained under other sorting criteria. Specifically, using the payout ratio measure, WW index, and SA index, respectively, reveals (p <.05), (p <.01), and (p <.01) percentage point difference in the median DCA between the constrained and unconstrained groups, after controlling for other factors. Overall, the results obtained using four different measures support the prediction that constrained issuers manage their earnings more aggressively than unconstrained issuers around the offering. 15 Furthermore, the estimated coefficients on the control variables suggest that firms with higher industry-adjusted ROA and book-to-market ratio in the pre-seo year tend to report larger income-increasing accruals in the year of the offering. However, operating in a high litigation risk industry is negatively associated with upward earnings management. As an additional test, I augment the regression equation (5) with the following variables: the natural log of the number of analysts following the company, big-n auditor dummy, low/high governance index dummies. However, the data availability limits the sample to 1,220 firms for the period 1990 and Analyst data is obtained from the First Call database and calculated as the maximum number of analysts following the firm in the year prior to the offering. Since previous research (e.g., Yu, 2008) documents that firms followed by more analysts manage their earnings less, I anticipate a negative relation between the number of analysts and performance-adjusted DCA. Further, Big-N auditor dummy equals 1 if a firm s financial statements for the SEO year were audited by one of the Big-8, Big-6, Big-5 or Big-4 auditors (depending on the sample period) and 0 otherwise. 16 The issuers whose financial statements are audited by Big-N companies are expected to report less income-increasing accruals. Finally, G-index (Gompers, Ishii, and Metrick, 15 Estimating equation (5) with OLS and clustering standard errors at the firm level lead to the same conclusion with quantitatively similar results % of the sample had their financial statements audited by Big-N auditors. 22

24 2003) is obtained from the RiskMetrics database. Low (High) governance index dummy equals 1 if a firm s G-index is less than (greater than or equal to) nine, which is the sample median, and 0 otherwise. The base category for governance dummy variables includes the firms with missing value for the G-index (as in Bergstresser and Philippon (2006)). 17 Previous research suggests that the use of discretionary accruals is less pronounced among firms with lower G-index. The results of the updated median regression model are presented in columns 2, 4, 6, and 8 of Table 4. The estimated coefficient on the financially constrained firm dummy is still positive and statistically significant under all sorting criteria. The difference in median performance-adjusted DCA between the two groups is (p <.01), (p <.05), (p <.01), and (p <.10) percentage points under the size, payout ratio, WW index, and SA index, respectively. Furthermore, the results suggest that offering firms with higher analyst following report less income-increasing accruals in the SEO year. Although the estimated coefficient on the Big-N auditor dummy is negative, it is not statistically significant. Also, neither low nor high governance index dummy is significantly associated with performance-adjusted DCA Additional test: financial constraints, CEO equity compensation, and earnings management One might argue that financially constrained (vs. unconstrained) firms may compensate their CEOs more with options and stocks rather than cash payments (i.e., salary and bonus). Given that the use of discretionary accruals to manipulate earnings is more pronounced at firms where the CEO s potential total compensation is more closely tied to the value of option and stock holdings (Cheng and Warfield, 2005; Bergstresser and Philippon, 2006), my results can be confounded by the difference in the compensation structure of the two groups. To test the validity of this argument, I hand-collect the data on CEO s option and stock holdings from the issuer s last proxy statements filed prior to the offering date. I limit the sample period to for this analysis. The reason 17 G-index is available for only 23.44% of the firms included in the sample. 23

25 why I hand-collect this data rather than use the data on Execucomp is that the sample size drops significantly when I merge the SEO and Execucomp datasets (small issuers are not covered by Execucomp). The final sample with CEO option and ownership data consists of 660 SEOs. Since the composition of the sample is different than that of the initial sample, I re-categorize firms into the financial constraint groups based on the updated terciles of each measure. Consistent with the argument that constrained firms tend to compensate their managers with more options, the median number of CEO s exercisable (unexercisable) options as a percentage of shares outstanding prior to the offering is 1.019% (0.810%) for the constrained group, as compared to 0.653% (0.442%) for the unconstrained group according to the firm size criterion. 18 The difference between the two groups is statistically significant at 5% (1%) level. The results also suggest that the median CEO stock ownership in constrained firms, 2.474%, is significantly larger as compared to that in unconstrained firms, 0.735% (p <.01). 19 This finding is consistent with Fahlenbrach and Stulz (2009) who document that managerial ownership increases when a firm becomes financially constrained, since managers become more willing to accept shares instead of cash to prevent the firm becoming more constrained. Finally, constrained firms have younger CEOs as compared to unconstrained firms, though the difference in median age between the groups is only 2 years (p <.01), 1 year (p >.40), 2.5 years (p <.01), and 4 years (p <.01) under the size, payout ratio, WW index, and SA index criteria, respectively. To examine whether the positive relation between financial constraints and earnings management still holds after controlling for CEO option holdings and ownership, I augment the regression model (equation (5)) with the following variables: number of exercisable options (as a 18 The results are similar under the other financial constraints criteria, though the differences between the two groups do not reach statistical significance when the payout ratio measure is used. 19 The SA (WW) index produces quantitatively (qualitatively) similar results. However, the payout ratio measure suggests that there is no difference in the CEO ownership between the constrained and unconstrained groups. 24

26 percentage of the number of shares outstanding), number of unexercisable options (as a percentage of the number of shares outstanding), number of shares held by the CEO (as a percentage of the number of shares outstanding). I also control for the CEO s age and duality measured by the CEO/Chairman dummy that is equal to 1 if the CEO is also the chairman of the company and 0 otherwise. The results are presented in Table 5. The results of the median regressions suggest that after controlling for CEO option and stock holdings, there is still a positive relation between issuers financial constraint status and performance-adjusted DCA in the year of the offering. Specifically, the difference in median DCA between the constrained and unconstrained groups is (p <.01), (p <.05), (p <.01), and (p <.05) percentage points under the size, payout ratio, WW index and SA index criteria, respectively. Although the coefficients on exercisable and unexercisable options are not significant in any of the models, the results reveal a positive and significant association between CEO ownership and earnings management. More important, these results suggest that the link between financial constraints and aggressive earnings management around SEOs is not confounded by the CEO s equity related incentives Financial constraints and offering size To test the hypothesis that, everything else equal, unconstrained firms raise smaller amount of equity (i.e., underfinance) than constrained firms, I estimate equation (6) using OLS with firm level clustered standard errors. Ideally, one would like to use the offer amount filed by the issuer as the dependent variable for this test while controlling for the market value of the issuer on the right-hand side of the equation. However, this would give rise to multicollinearity concerns as there is a significant correlation between market value and financial constraints measures (except the payout ratio). One alternative is to use scaled offer size as the dependent variable. Since, in my 25

27 framework, underfinancing is defined relative to the cost of the upcoming project, I choose to scale the offer size by sales, which would be a reasonable proxy for the size of projects undertaken by issuers. 20 Thus, I specify the following regression: ln(offer Size) = α 0 + α 1 LOW-FC+ α 2 MEDIUM-FC+ α 3 MB+ α 4 ADJROA + α 5 CFVOL + α 6 REVVOL + α 7 SGVOL + α 8 CAPEX + Year and industry fixed effects + ζ (6) where, ln(offer Size) is the natural log of the ratio of offer amount to sales in the year prior to the offering. LOW-FC is a dummy variable equal to 1 if a firm is categorized as financially unconstrained under a sorting criterion and 0 otherwise; MEDIUM-FC is a dummy variable equal to 1 if a firm is ranked in the middle tercile according to a financial constraint sorting criterion and 0 otherwise; CAPEX is the capital expenditures plus R&D expenditures divided by sales in the year prior to the offering. Other variables are as defined in equation (5). 21 To mitigate the impact of outliers, I winsorized all the variables at the 1% and 99% levels. Furthermore, since the sample contains multiple offerings by an issuer, White heteroskedasticity robust standard errors are clustered at the issuer level. Table 6 reports the regression results. As predicted, there is a negative and significant relation between offer size and being financially unconstrained. The estimated coefficient on the low-fc dummy created based on the size, payout ratio, WW index, and SA index measures is , , , and , respectively (p s <.01). This suggests that under the firm size classification, the average offer size of the unconstrained group is 74.44% lower (i.e., 100 x (exp(-1.364) 1)) than that of the constrained group. These results provide strong support for the 20 The results remain unchanged if I scale the offer size by cost of goods sold (Compustat item COGS) or market value. 21 Similar to equation (5), the natural log of the pre-offer market value is added to the model as an additional control variable when the payout ratio measure is used to determine the financial constraint groups. 26

28 underfinancing hypothesis (H2). Furthermore, the results also reveal that firms with higher market-to-book ratios and capital expenditures in the pre-seo year tend to raise larger amount of equity relative to their sales. The offer size is also positively associated with cash flow and sales growth volatility Financial constraints and SEO announcement returns To test the prediction that investors discount the stock prices of constrained firms at a higher rate as compared to unconstrained firms at offering announcements, I compare the announcement returns of the two groups. The univariate results are presented in Table 7. I calculate seven-day cumulative abnormal returns using the market-model in which the CRSP value-weighted index is the measure of the market return. 22 The estimation period starts at 300 trading days before the announcement date and ends at 61 trading days prior to the announcement date (event day 0). 23 I use the SEO filing date reported on SDC as the announcement date. Consistent with my conjecture, under all classifications, the difference in seven-day announcement returns between constrained and unconstrained issuers is positive and significant at 1% level. The median announcement returns for the unconstrained (constrained) group is -1.59% (-4.30%), -2.28% (-3.77%), -1.88% (-4.21%), -2.12% (-4.51%) under the size, payout ratio, WW index, and SA index criteria, respectively. The tests for the differences in announcement returns between the two groups are all significant at 1% level. These results provide initial support for H3. Furthermore, I estimate the following model using OLS to test H3 in a multivariate setting: CAR = λ 0 + λ 1 HIGH-FC+ λ 2 MEDIUM-FC + λ 3 MB+ λ 4 RUNUP+ λ 5 VOLATILITY + Year and industry fixed effects + η (7) 22 I obtain similar results if I use three-day or five-day cumulative abnormal returns. 23 The results are robust to the use of alternative estimation windows (e.g., [-220, -21]) and methods (e.g., market-adjusted). 27

29 where, CAR is the seven-day abnormal return around the SEO announcement date; HIGH-FC is a dummy variable equal to 1 if a firm is categorized as financially constrained under a sorting criterion and 0 otherwise; MEDIUM-FC is a dummy variable equal to 1 if the firm is ranked in the middle tercile under a financial constraint sorting criterion and 0 otherwise; MB is the natural log of the market-to-book ratio in the year prior to the SEO year; RUNUP is the cumulative abnormal returns over the period starting 60 trading days prior to the announcement and ending 6 trading days prior to the announcement; VOLATILITY is the standard deviation of the market-model residuals over the period from 300 trading days to 61 days preceding the offering announcement. Under the payout ratio classification, the model also controls for the natural log of the issuer s market value prior to the offering. To mitigate the impact of outliers, all the variables are winsorized at the 1% and 99% levels. Further, the standard errors are clustered at the issuer level. The results are reported in Table 8. The OLS regression results reveal that the average seven-day abnormal announcement return is significantly lower for the constrained group as compared to the unconstrained group under each financial constraint classification. Specifically, the estimated coefficient on the financially constrained dummy is , , , and (p s < 0.05) under the size, payout ratio, WW index, and SA index measures, respectively. This finding suggests that investors discount the stock prices of constrained issuers more than those of unconstrained issuers at offering announcements. 24 Furthermore, there is a negative and significant relation between announcement returns and volatility, indicating that firms with more volatile pre-seo stock returns experience more negative abnormal returns when they announce an equity issuance Extension: Post-SEO stock return performance of constrained and unconstrained issuers 24 This finding goes against the alternative hypothesis suggesting that investors would react more negatively to an equity offering announcement by the unconstrained firm than the constrained firm since they would believe that the equity offered by the unconstrained issuer is more overvalued than that offered by the constrained issuer. 28

30 In this section, I examine post-seo stock return performance of constrained and unconstrained firms during the one- and two-year period following the offering year. If investors correctly price, on average, the offerings of constrained and unconstrained issuers, we should not observe a significant price correction (measured by abnormal stock price performance) after the filing of financial statements for the SEO year. To test this prediction, I run calendar-time portfolio regressions separately for the constrained and unconstrained groups. Following previous research (e.g., Brav, Geczy, and Gompers, 2000; Hertzel and Li, 2011), I use both the Fama-French (1993) three-factor model and Carhart (1997) four-factor model, and measure abnormal portfolio performance using the intercept from the factor regressions. The results are presented in Table 9. The dependent variable in Panel A (B) is the equally- and value-weighted monthly portfolio returns of SEO firms in excess of risk-free rate one-month T-bill rate for the period between April 1984 (1985) and March 2008 (2009). Following Teoh et al. (1998b), to allow for a reporting lag (i.e., time till the filing of annual financial statements after the end of a fiscal year), the event firms are included in the portfolio starting three months after the SEO fiscal year-end. 25 That is, in Panel A (B), SEO firm returns are included in portfolio returns for the period from four to fifteen (twenty seven) months after the SEO fiscal year-end. These portfolios are rebalanced each month to drop all firms that reach the end of their respective post-seo periods and add all firms that have announced an SEO. To obtain reliable estimates, I exclude the calendar months where there are less than five observations. Furthermore, since the number of observations changes across calendar months, I use weighted-least square regression, where the weighting vector is the number of event firms having non-missing returns in a relevant calendar month (e.g., 25 For instance, if a firm with December fiscal year-end announces an SEO in October 1997, the first month in which it appears in the portfolio is April The results are similar if I do not impose any time lags and define the beginning of the post-seo return calculation period as the month immediately following the SEO announcement. 29

31 Lyandres, Sun, and Zhang, 2008). Further, I calculate p-values based on heteroskedasticity-robust standard errors. The full sample results reported in Table 9 are consistent with previous studies (e.g., Loughran and Ritter, 1995; Teoh et al., 1998a). Specifically, the results obtained from the Fama-French (1993) three-factor model suggest that the equally-weighted (EW) portfolio of SEO firms exhibit significant underperformance both in the one-year (Panel A) and two-year (Panel B) period following the offering year (α 1YR = -0.45%, p <.01; α 2YR = -0.42%, p <.01). However, similar to Mittchell and Stafford (2000), I find that using value-weighted (VW) portfolios eliminates underperformance of issuers in both periods. Further, after controlling for the momentum factor as suggested by Brav et al. (2000), the EW intercept becomes small and statistically insignificant in the two-year period. More importantly, both the EW and VW intercepts obtained from the Carhart (1997) four-factor model reveal that neither constrained nor unconstrained issuers (under all classifications) exhibit significant abnormal stock price performance during the year following the offering year. This result continues to hold when the examination window is extended to two-year period. However, the results suggest that even after controlling for the momentum factor, the EW portfolio of opaque (middle tercile) issuers according to the WW index and SA index classifications experience negative abnormal stock price performance in the year after the offering year (WW index: α 1YR = -0.44%, p <.05, SA index: α 1YR = -0.45%, p <.05), though value weighting renders these intercepts insignificant. Further, the VW portfolio of opaque issuers under the size classification has a negative and marginally significant intercept (α 1YR = -0.66%, p <. 10). One possibility that may give rise to these findings is that the incentives and costs of accrual 30

32 manipulation to opaque issuers are not as apparent as they are to constrained and unconstrained issuers, preventing investors from correctly pricing their equity around the offering. Overall, the EW and VW intercepts obtained from the Carhart (1997) four-factor model regressions reveal no significant price correction for both constrained and unconstrained issuers during the period following the SEO year, implying that investors rationally discount the stock prices of each group at appropriate rates around the offering Robustness tests Endogeneity Although I find that higher financial constraints are negatively related to financial reporting quality around SEOs, the specified regression models may not fully account for potential endogeneity in the sample. Particularly, modeling the relation between financial constraints and discretionary current accruals may be problematic due to either an endogenous feedback from earnings quality to financial constraints or an omitted variable driving both financial constraints and discretionary current accruals. Although the reverse causality is a relevant concern when the financial constraint categories are created based on payout ratio and the WW index endogenous measures of financial constraints it is not a very valid concern when one uses firm size and the SA index, which are fairly exogenous measures of financial constraints. Besides, the rational expectations framework presented in the paper is based on a detailed reasoning suggesting that the causality between the two variables is more likely to be influenced by the firm s financial constraints. Nevertheless, I employ the instrumental variable (IV) approach to bolster the integrity of the analysis, as well as to address the omitted variable criticism. Specifically, I use the number of employees (Compustat item EMP) and market share (based on two-digit SIC codes) of issuers in the year prior to the offering as instruments for their financial constraints and re-estimate 31

33 equation (5) with 2SLS using continuous versions of each financial constraint measure (e.g., ln(assets)) instead of the tercile dummy variables used in the previous analyses. 26 However, I acknowledge the fact that number of employees and market share are better suited to be an instrument for the size, WW index, and SA index measures rather than the payout ratio measure. For instance, large and established firms with low leverage and high cash flows are expected to employ higher number of employees, whereas the relation between the payout ratio and the number of employees is less clear. Consistent with my conjecture, the Spearman rank (Pearson) correlation of ln(1+employees) with ln(assets), WW index, and SA index is (0.809), (-0.753), (-0.719), respectively. On the other hand, the correlation between ln(1+employees) and payout ratio is only (0.137). 27 Thus, to avoid problems associated with using a weak instrument, I do not include the payout ratio measure in IV estimations. Also, as a note, the Spearman rank correlations between ln(1+employees) and performance-adjusted DCA and between ln(1+market share) and performance-adjusted DCA are low, and , respectively. 28 The 2SLS regression results confirm the previous results. The estimated coefficient on ln(assets) is (p <.01), suggesting that one standard deviation increase in ln(assets) results in a percentage point decrease in performance-adjusted DCA reported in the SEO year. The results also reveal positive and significant coefficients on the WW index (28.956, p <.01) and the SA index (4.264, p <.01), indicating that one standard deviation decrease in each measure leads to, respectively, and percentage point decline in performance-adjusted DCA. Overall, 26 To mitigate the impact of outliers, all the variables used in 2SLS regressions are winsorized at 1% and 99% levels. 27 Similarly, the Spearman rank (Pearson) correlation of ln(1+market share) with ln(assets), WW index, SA index, and payout ratio is, respectively, (0.453), (-0.427), (-0.369), and (0.104). 28 The IV diagnostics also suggest that my results do not significantly suffer from the weak instrument problem. The F-statistic of the first-stage IVs of ln(assets), the WW index, and the SA index is , , and , respectively (p s <.01). In addition, the p-values of the Hansen-J statistic for ln(assets), the WW index, and the SA index is 0.555, 0.544, and 0.172, respectively. This suggests that the over-identification restriction of both IVs hold under each financial constraints measure. 32

34 these results cast doubt on the reverse causality and omitted variable criticisms and boost confidence in the mechanism proposed by the rational expectations framework Using the cash flow approach to measure current accruals Using the balance sheet approach to calculate accruals may induce measurement error in accrual estimates (Hribar and Collins 2002). Measuring accruals directly from the statement of cash flows can correct for potential estimation errors but the cash flow statement data are not widely available before Thus, using the balance sheet approach helps preserve a larger sample covering a longer period. Nonetheless, I repeat all the tests with current accruals data obtained from the statement of cash flows. Following previous research (e.g., Beatty, Liao, and Weber, 2010), I define current accruals as total accruals (i.e., Compustat item IBC minus OANCF) plus depreciation (Compustat item DP). Although using this approach reduces the sample size to 1,301, it does not alter any of my conclusions. In particular, re-estimations of the regression models in columns 1, 3, 5, and 7 of Table 4 reveal that the difference in median performance-adjusted DCA between the constrained and unconstrained groups is (p <.01), (p =.10), (p <.01), (p <.05) percentage points under the size, payout ratio, WW index, and SA index classifications, respectively Small denominator problem in the estimation of discretionary accruals Ball and Shivakumar (2008) point out that the modified-jones model may suffer from small denominator problem (due to using lagged total assets to scale the variables included in the model) during periods when firms undergo large transactions (e.g., IPOs). Although this concern has not been directed to SEOs, I rerun my analyses using the residuals obtained from the corrected modified-jones model as suggested by Armstrong, Foster, and Taylor (2010) to ensure that my results are robust to small denominator problem. In particular, I use average total assets rather than 33

35 lagged total assets as deflator in the modified-jones model. Using this alternative estimation model does not change my findings. For instance, re-estimating the regression models in columns 1, 3, 5, and 7 of Table 4 under this approach reveals that the difference in median performance-adjusted discretionary accruals between the constrained and unconstrained groups is (p <. 01), (p <.05), (p <.01), and (p <.01) percentage points under the size, payout ratio, WW index, and SA index measures, respectively. 7. Conclusion This paper examines the link between firms financial constraints and their financial reporting behavior around SEOs. Within a simple rational expectations framework, I propose and test that when raising equity, financially constrained, but not unconstrained, firms engage in aggressive earnings management as a response to anticipated severe discounting by investors at offering announcements. On the other hand, unconstrained firms, which can credibly signal the absence of aggressive earnings management (through underfinancing), are expected to manage their earnings conservatively and thus, experience less negative investor response at offerings announcements. By using four different measures of financial constraints (i.e., firm size, payout ratio, the Whited and Wu (2006) index, the Size-Age index of Hadlock and Pierce (2010)), I find that constrained issuers report higher income-increasing accruals around the offering than unconstrained issuers. The difference in performance-adjusted discretionary current accruals between the two groups is most pronounced at the year of the offering and then dissipates after the first year following the offering. 34

36 Furthermore, I document that unconstrained firms raise smaller amount of equity (relative to their sales) than unconstrained firms. This finding is consistent with the underfinancing behavior predicted by the rational expectations framework and enables unconstrained issuers to credibly reveal their types. Accordingly, I find that SEO announcement returns are significantly lower for constrained issuers as compared to unconstrained issuers, suggesting that investors rationally discount the stock prices of constrained firms to a greater extent at offering announcements. Overall, the evidence suggest that the aggressive earnings management by constrained issuers is not simply the result of opportunistic behavior of managers but rather a preemptive strategy aimed at mitigating the impact of large negative announcement returns by boosting stock price via inflated earnings performance. 35

37 References Almeida, H., Campello, M., Financial constraints, asset tangibility, and corporate investment. Review of Financial Studies 20, Almeida, H., Campello, M., Financing frictions and the substitution between internal and external funds. Journal of Financial and Quantitative Analysis 45, Almeida, H., Campello, M., Weisbach, M.S., The cash flow sensitivity of cash. Journal of Finance 59, Armstrong, C.S., Foster, G., Taylor, D.J., Earnings management around initial public offerings: a re-examination. Working paper, University of Pennsylvania and Stanford University. Ball, R., Shivakumar, L., Earnings quality at initial public offerings. Journal of Accounting and Economics 45, Barton, J., Simko, P.J., The balance sheet as an earnings management constraint. The Accounting Review 77, Beatty, A., Liao, S., Weber, J., Financial reporting quality, private information, monitoring, and the lease-versus-buy decision. The Accounting Review 85, Bergstresser, D., Philippon, T., CEO incentives and earnings management. Journal of Financial Economics 80, Brav, A., Geczy, C., Gompers, P.A., Is the abnormal return following equity issuances anomalous? Journal of Financial Economics 56, Campello, M., Graham, J.R., Campbell, R.H., The real effects of financial constraints: evidence from a financial crisis. Journal of Financial Economics 97, Carhart, M.M., On the persistence in mutual fund performance. Journal of Finance 52, Cheng, Q., Warfield, T.D., Equity incentives and earnings management. The Accounting Review 80, Cleary, S., The relationship between firm investment and financial status. Journal of Finance 54, Cohen, D.A., Zarowin, P., Accrual-based and real earnings management activities around seasoned equity offerings. Journal of Accounting and Economics 50,

38 Dechow, P.M., Sloan, R.G., Sweeney, A.P., Detecting earnings management. The Accounting Review 70, Dechow, P.M., Sloan, R.G., Sweeney, A.P., Causes and consequences of earnings manipulation: an analysis of firms subject to enforcement actions by the SEC. Contemporary Accounting Research 13, Dechow, P.M., Ge, W., Schrand, C., Understanding earnings quality: a review of the proxies, their determinants and their consequences. Journal of Accounting and Economics 50, Dechow, P.M., Ge, W., Larson, C.R., Sloan, R.G., Predicting material accounting misstatements. Contemporary Accounting Research 28, DeAngelo, H., DeAngelo, L., Stulz, R.M., Seasoned equity offerings, market timing, and the corporate lifecycle. Journal of Financial Economics 95, Denis, D., Sibilkov, V., Financial constraints, investment, and the value of cash holdings. Review of Financial Studies 23, Dittmar, A., Thakor, A., Why do firms issue equity? Journal of Finance 62, DuCharme, L.L., Malatesta, P.H., Sefcik, S.H., Earnings management, stock issues, and shareholder lawsuits. Journal of Financial Economics 71, Duchin, R., Cash holdings and corporate diversification. Journal of Finance 65, Duchin, R., Dittmar, A., The dynamics of cash. Working paper, University of Michigan. Erickson, M., Wang, S., Earnings management by acquiring firms in stock for stock mergers. Journal of Accounting and Economics 27, Fahlenbrach, R., Stulz, R.M., Managerial ownership dynamics and firm value. Journal of Financial Economics 92, Fama, E., French, K., Common risk factors in the returns of stocks and bonds. Journal of Financial Economics 33, Faulkender, M., Wang, R., Corporate financial policy and the value of cash. Journal of Finance 61, Fazzari, S.R., Hubbard, R.G., Petersen, B., Financing constraints and corporate investment. Brooking Papers on Economic Activity 1, Firth, M., Rui, O.M., Wu, W., Cooking the books: recipes and costs of falsified financial statements in China. Journal of Corporate Finance, forthcoming. 37

39 Flannery, M.J., Rangan, K.P., Partial adjustment toward target capital structures. Journal of Financial Economics 79, Franzoni, F., Underinvestment vs. overinvestment: evidence from price reactions to pension contributions. Journal of Financial Economics 92, Friedlan, J.M., Accounting choices of issuers of initial public offerings. Contemporary Accounting Research 11, Gompers, P., Ishii, J., Metrick, A., Corporate governance and equity prices. The Quarterly Journal of Economics 118, Gong, G., Louis, H., Sun, A.X., Earnings management, law suits, and stock-for-stock acquirers market performance. Journal of Accounting and Economics 46, Hadlock, C.J., Pierce, J.R., New evidence on measuring financial constraints: Moving beyond the KZ index. Review of Financial Studies 23, Hennessy, C.A., Livdan, D., Miranda, B., Repeated signaling and firm dynamics. Review of Financial Studies 23, Hennessy, C.A., Whited, T.M., How costly is external financing? Evidence from a structural estimation. Journal of Finance 62, Hertzel, M.G., Li, Z., Behavioral and rational explanations of stock price performance around SEOs: evidence from a decomposition of market-to-book ratios. Journal of Financial and Quantitative Analysis 45, Hribar, P., Collins, D.W., Errors in estimating accruals: implications for empirical research. Journal of Accounting Research 40, Hribar, P., Nichols, C., The use of unsigned earnings quality measures in tests of earnings management. Journal of Accounting Research 45, Hovakimian, A., Hovakimian, G., Tehranian, H., Determinants of target capital structure: the case of dual debt and equity issues. Journal of Financial Economics 71, Hovakimian, A., Hutton, I., Market feedback and equity issuance: evidence from repeat equity issuance. Journal of Financial and Quantitative Analysis 45, Jones, J., Earnings management during import relief investigations. Journal of Accounting Research 29, Kaplan, S.N., Zingales, L., Do investment-cash flow sensitivities provide useful measures of financing constraints? The Quarterly Journal of Economics 112,

40 Karpoff, J.M., Lee, D.S., Martin, G.S., The cost to firms of cooking the books. Journal of Financial and Quantitative Analysis 43, Kim, Y., Park, M.S., Pricing of seasoned equity offers and earnings management. Journal of Financial and Quantitative Analysis 40, Korajczyk, R. A., Levy, A., Capital structure choice: macroeconomic conditions and financial constraints. Journal of Financial Economics 68, Kothari, S.P., Leone, A., Wasley, C., Performance matched discretionary accrual measures. Journal of Accounting and Economics 39, Lamont, O.A., Polk, C., Saa-Requejo, J., Financial constraints and stock returns. Review of Financial Studies 14, Lee, G., Masulis, R.W., Seasoned equity offerings: quality of accounting information and expected flotation costs. Journal of Financial Economics 92, Li, D., Zhang, L., Does Q-theory with investment frictions explain anomalies in the cross-section of returns? Journal of Financial Economics 98, Loughran, T., Ritter J.R., The new issues puzzle. Journal of Finance 50, Louis, H., Earnings management and the market performance of acquiring firms. Journal of Financial Economics 74, Louis, H., Robinson, D., Do managers credibly use accruals to signal private information? Evidence from the pricing of discretionary accruals around stock splits. Journal of Accounting and Economics 39, Louis, H., White, H., Do managers intentionally use repurchase tender offers to signal private information? Evidence from firm financial reporting behavior. Journal of Financial Economics 85, Lyandres, E., Sun, L., Zhang, L., The new issues puzzle: testing the investment-based explanation. Review of Financial Studies 21, Mitchell, M.L., Stafford, E., Managerial decisions and long-term stock price performance. Journal of Business 73, Myers, S.C., Determinants of corporate borrowing. Journal of Financial Economics 5, Myers, S.C., Majluf, N.S., Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics 13,

41 Morsfield, S.G., Tan, C.E.L., Do venture capitalists influence the decision to manage earnings in initial public offerings? The Accounting Review 81, Narayanan, M.P., Managerial incentives for short-term results. Journal of Finance 40, Rangan, S., Earnings management and the performance of seasoned equity offerings. Journal of Financial Economics 50, Shivakumar, L., Do firms mislead investors by overstating earnings before seasoned equity offerings? Journal of Accounting and Economics 29, Stein, J.C., Efficient capital markets, inefficient firms, a model of myopic corporate behavior. The Quarterly Journal of Economics 104, Stulz, R.M., Managerial discretion and optimal financing policies. Journal of Financial Economics 26, Teoh, S.H., Welch, I., Wong, T.J., 1998a. Earnings management and the underperformance of seasoned equity offerings. Journal of Financial Economics 50, Teoh, S.H., Welch, I., Wong, T.J., 1998b. Earnings management and the long-run market performance of initial public offerings. Journal of Finance 53, Teoh, S.H., Wong, T.J., Rao, G.R., Are accruals during initial public offerings opportunistic. Review of Accounting Studies 3, Whited, T.M., Wu, G., Financial constraints risk. Review of Financial Studies 19, Yu, F., Analyst coverage and earnings management. Journal of Financial Economics 88,

42 Figure 3a: Earnings management around SEOs by financial constraint status (firm size classification) Percentage of Total Assets Median Performance-Adjusted DCA around the SEO year Constrained Unconstrained Figure 3b: Earnings management around SEOs by financial constraint status (payout ratio classification) Percentage of Total Assets Median Performance-Adjusted DCA around the SEO year Constrained Unconstrained 41

43 Figure 3c: Earnings management around SEOs by financial constraint status (the Whited-Wu (2006) index classification) Percentage of Total Assets Median Performance-Adjusted DCA around the SEO year Constrained Unconstrained Figure 3d: Earnings management around SEOs by financial constraint status (the Size-Age index classification) Percentage of Total Assets Median Performance-Adjusted DCA around the SEO year Constrained Unconstrained 42

44 Table 1: Summary statistics The sample includes 1,645 seasoned equity offerings (SEOs) over the period 1983 to Panel A reports pre-offering size characteristics. All monetary variables are adjusted for inflation and represent millions of 2006 constant dollars. Total assets are end of period book assets (Compustat item AT) in the year prior to the offering. Market value is calculated as the closing price at the fiscal year-end (Compustat item PRCC_F) times the number of shares outstanding (Compustat item CSHO) in the year prior to the offering. Market-to-book ratio is calculated as the market value of equity divided by the book value of common equity (Compustat item CEQ). Offer amount is the dollar amount of the SEO filed by the issuer. Panel B reports the yearly distribution. Panel C presents the industry distribution. Panel A: Size characteristics Total assets ($M) Market value ($M) M/B ratio Offer amount ($M) Mean Median Std. dev. 2, , Panel B: Time distribution Year Freq % Cum Freq % , , , , , , , , , ,

45 Table 1 (cont d): Summary statistics Panel C: Industry (two-digit SIC code) distribution Industry Codes Freq % Mining, oil, and gas 10, % Food products % Paper and paper products 24, 25, 26, % Chemical products % Manufacturing 30, 31, 32, 33, % Computer equipment and services 35, % Electronic equipment % Transportation 37, 39, 40, 42, 44, % Scientific instruments % Communications % Durable goods % Retail 53, 54, 56, 57, % Eating and drinking establishments % Entertainment services 70, 78, % Health % All others 16, 22, 23, 29, 47, 51, 52, 55, 82, 87, % 44

46 Table 2: Median firm characteristics by financial constraint status Size Payout Ratio The Whited-Wu Index The Size-Age Index C Mid UC C Mid UC C Mid UC C Mid UC Firm Characteristics Assets ($M) a a a a Market value ($M) a a a a Market-to-book ratio a a a a Cash flow volatility a a a a Revenue volatility a a a a Sales growth volatility a a a a Industry-adjusted ROA (%) a a c Payout ratio a a a a Number of analysts a a a a Offering Characteristics Offer amount / Market value a a a a Offer amount / Sales a a a a CEO Characteristics Stock ownership (%) a a Exercisable options (%) b a Unexercisable options (%) a a a Age a a a a, b, and c denote the statistical significance of the difference in median firm characteristics between the unconstrained (UC) and constrained (C) firms. 45

47 Table 2 (cont d): Median firm characteristics by financial constraint status This table reports the median firm characteristics (in the year prior to the offering) by financial constraint categories. Financial constraint status is determined based on four different measures: (1) firm size measured by total assets, (2) payout ratio, (3) the Whited and Wu (2006) index, (4) the Size-Age index of Hadlock and Pierce (2010). The procedure of categorizing firms into the financial constraint groups is described in section 4.1. The columns labeled as C, Mid, and UC represent the categories of financially constrained, moderately financially constrained, and financially unconstrained firms, respectively. Assets, market value, and M/B ratio are as defined in Table 1. Cash flow volatility is the standard deviation of cash flow (Compustat item IB plus Compustat item DP) deflated by total assets over the five year period (with a minimum of three years) prior to the SEO year. Revenue volatility is the standard deviation of sales (Compustat item SALE) deflated by total assets over the five year period (with a minimum of three years) prior to the SEO year. Sales growth volatility is the standard deviation of annual sales growth over the five year period (with a minimum of three years) prior to the SEO year. ROA is the income before extra ordinary items (Compustat item IB) as a percentage of lagged total assets. Industry-adjusted ROA is the firm s ROA minus the industry-median ROA. Payout ratio is defined as the ratio of dividends (Compustat item DVC plus Compustat item DVP) plus share repurchases (Compustat item PRSTKC) to operating income (Compustat item OIBDP). Payout ratio is set equal to 1 if a firm has negative operating income and positive payout. Number of analysts is the maximum number of analysts following the firm in the year prior to the offering and obtained from the First Call database for the period over The data on CEO s option holdings and stock ownership are hand-collected from the last proxy statement filed by the issuers prior to the offering. This sample consists of 660 offerings (over the period 1997 to 2006) with available proxy statements on the SEC s website. Ownership is calculated as the number of shares held by the CEO divided by the number of shares outstanding prior to the offering. Exercisable (unexercisable) represents the number of exercisable (unexercisable) options held by the CEO divided by the number of shares outstanding prior to offering. Age is the CEO s age. The differences in firm characteristics between the constrained and unconstrained groups are tested using Wilcoxon signed rank test. a, b, and c denote statistical significance at 1%, 5%, and 10% level, respectively. 46

48 Table 3: Median performance-adjusted discretionary current accruals around SEOs by financial constraint status This table reports time-series of median performance-adjusted discretionary current accruals (DCA), as a percentage of total assets, from year -3 to year +3 relative to the seasoned equity offering (i.e., year 0). Financial constraint status is determined based on four different measures (i.e., firm size, payout ratio, the Whited-Wu index, and the Size-Age index) in the year prior to the offering. See section 4.1 for details of the procedure used to categorize firms into the financial constraint groups. Significance of tests that median DCA are different from zero is assessed using Wilcoxon signed rank test. a, b, and c denote statistical significance at 1%, 5%, and 10% level, respectively. Panel A: Firm size Year Performance-adjusted discretionary current accruals Constrained (C) Median 0.88 a 1.03 a 0.96 a 4.51 a 1.94 a 0.73 b N Unconstrained (UC) Median a 0.52 b b Panel B: Payout ratio N p-value (C UC) < Constrained (C) Median 0.60 b 0.95 a 0.62 a 2.96 a 1.25 a 0.49 c N Unconstrained (UC) Median 0.36 c a N p-value (C UC) Panel C: The Whited and Wu (2006) index Constrained (C) Median 0.88 a 1.38 a 0.60 a 3.90 a 1.32 a N Unconstrained (UC) Median a 0.42 b N p-value (C UC) < Panel D: The Size-Age index of Hadlock and Pierce (2010) Constrained (C) Median 0.91 a 1.02 a 0.82 a 4.12 a 1.29 a 0.78 c N Unconstrained (UC) Median a 0.85 a 0.47 b N p-value (C UC) <

49 Table 4: Median regression of performance-adjusted discretionary current accruals in the year of the offering This table presents the results of the median regression of performance-adjusted discretionary current accruals, as a percentage of total assets, in the SEO year. The sample includes 1,645 offerings over the period 1983 to Each column is labeled with the financial constraint measure (i.e., firm size, payout ratio, the Whited-Wu (WW) index, and the Size-Age (SA) index) used to categorize the sample firms as constrained and unconstrained in the year prior to the offering (year -1). High-FC is a dummy variable equal to 1 if a firm is categorized as financially constrained under a sorting criterion and 0 otherwise. Medium-FC is a dummy variable equal to 1 if a firm is ranked in the middle tercile according to a financial constraint criterion and 0 otherwise. Litigation is a dummy variable equal to 1 if a firm s SIC code is , , , , and 0 otherwise. Big-N auditor is a dummy variable equal to 1 if the firm s annual report is audited by one of the Big-8, Big-6, Big-5 or Big-4 auditors depending on the year of the observation and 0 otherwise. Low (High) Governance Index is a dummy variable equal to 1 if the firm s G-index is less than (greater than or equal to) nine and 0 otherwise. Firms with missing G-index constitute the base category for the G-index dummy variables. Other variables are as defined in Tables 1 and 2. p-values are reported in parentheses and based on bootstrapped standard errors with 200 replications. a, b, and c denote statistical significance at 1%, 5%, and 10% level, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Size Size Payout Payout WW index WW index SA index SA index Sample Period High-FC a a b b a a a c (0.000) (0.002) (0.014) (0.029) (0.000) (0.008) (0.000) (0.051) Medium-FC b c (0.031) (0.600) (0.485) (0.644) (0.098) (0.208) (0.138) (0.285) Ln(Market-to-book) b a a c b b (0.185) (0.047) (0.000) (0.000) (0.086) (0.040) (0.308) (0.024) Industry-adjusted ROA a a a a a a a a (0.001) (0.001) (0.000) (0.000) (0.000) (0.001) (0.000) (0.001) Cash flow volatility (0.714) (0.987) (0.754) (0.664) (0.898) (0.677) (0.895) (0.616) Revenue volatility (0.850) (0.756) (0.700) (0.327) (0.884) (0.591) (0.967) (0.381) Sales growth volatility c b b b (0.158) (0.084) (0.048) (0.139) (0.328) (0.034) (0.155) (0.017) Litigation c b b b c c (0.123) (0.056) (0.019) (0.011) (0.043) (0.079) (0.155) (0.060) 48

50 Ln(1 + Analyst) b a a (0.028) (0.137) (0.008) (0.001) Big-N auditor (0.178) (0.271) (0.414) (0.279) Low governance index (0.621) (0.190) (0.696) (0.547) High governance index (0.313) (0.438) (0.427) (0.343) Ln(Market Value) a a (0.000) (0.002) Constant b a a c b (0.760) (0.025) (0.000) (0.002) (0.314) (0.051) (0.167) (0.035) Year dummies? Yes Yes Yes Yes Yes Yes Yes Yes Observations 1,645 1,220 1,645 1,220 1,645 1,220 1,645 1,220 Pseudo R-squared

51 Table 5: Median regression of performance-adjusted discretionary current accruals in the year of the offering (controlling for CEO equity incentives). This table presents the results of the median regression of performance-adjusted discretionary current accruals, as a percentage of total assets, in the SEO year. The data on CEO s stock and option holdings are hand-collected from the last proxy statement filed by the issuers prior to the offering. The sample consists of 660 offerings (over the period 1997 to 2006) with available proxy statements on the SEC s website. CEO/Chairman is a dummy variable equal to 1 if the CEO is also the chairman of the company and 0 otherwise. Other variables are as defined in Tables 1, 2, and 4. p-values are reported in parentheses and based on bootstrapped standard errors with 200 replications. a, b, and c denote statistical significance at 1%, 5%, and 10% level, respectively. (1) (2) (3) (4) Size Payout WW index SA index High-FC a b a b (0.004) (0.048) (0.008) (0.020) Medium-FC (0.185) (0.805) (0.656) (0.174) Ownership b c a b (0.012) (0.056) (0.003) (0.014) Exercisable (0.247) (0.245) (0.162) (0.404) Unexercisable (0.771) (0.728) (0.513) (0.382) CEO/Chairman (0.764) (0.688) (0.762) (0.921) Ln(Age) (0.223) (0.461) (0.323) (0.163) Ln(Market-to-book) b a b c (0.038) (0.001) (0.032) (0.052) Industry-adjusted ROA b a a b (0.012) (0.010) (0.009) (0.012) Cash flow volatility (0.996) (0.743) (0.990) (0.798) Revenue volatility (0.583) (0.931) (0.397) (0.543) Sales growth volatility (0.271) (0.183) (0.483) (0.182) Litigation b b b b (0.011) (0.038) (0.017) (0.040) Ln(Market value) b (0.017) Constant (0.229) (0.797) (0.296) (0.152) Year dummies? Yes Yes Yes Yes Observations Pseudo R-squared

52 Table 6: OLS regression of offer size. This table presents the results of the OLS regression of the natural log of offer amount (deflated by sales) filed by the issuers. Capex is defined as the ratio of capital expenditures (Compustat item CAPX) plus R&D (Compustat item XRD) to sales. Other variables are as defined in Tables 1, 2, and 4. To mitigate the impact of outliers, all the variables are winsorized at the 1% and 99% levels. p-values are reported in parentheses and based on White standard errors clustered by firm. a, b, and c denote statistical significance at 1%, 5%, and 10% level, respectively. (1) (2) (3) (4) Size Payout WW index SA index Low-FC a a a a (0.000) (0.002) (0.000) (0.000) Medium-FC a c a a (0.000) (0.088) (0.000) (0.000) Ln(Market-to-book) a a a a (0.000) (0.000) (0.000) (0.000) Industry-adjusted ROA (0.209) (0.743) (0.235) (0.231) Cash flow volatility b b a b (0.042) (0.041) (0.008) (0.015) Revenue volatility c c (0.099) (0.116) (0.318) (0.051) Sales growth volatility a a a a (0.000) (0.000) (0.000) (0.000) Capex ratio a a a a (0.000) (0.000) (0.000) (0.000) Ln(Market value) a (0.000) Constant a a a c (0.007) (0.000) (0.000) (0.061) Year Dummies? Yes Yes Yes Yes Industry Dummies? Yes Yes Yes Yes Observations 1,633 1,633 1,633 1,633 R-squared

53 Table 7: SEO announcement returns by financial constraint status This table presents seven-day abnormal returns (in percentages) around the SEO announcement date. Abnormal returns are calculated using the market model, in which the CRSP value-weighted index is the measure of market return. The market model parameters are estimated by OLS using a firm s daily returns over trading days -300 to -61 prior to its SEO announcement date. Panel A, B, C, and D report the results according to the firm size, payout ratio, Whited-Wu index, Size-Age index classifications, respectively. Significance of tests that the difference in mean (median) cumulative abnormal returns between the constrained and unconstrained groups is different from zero is assessed using t-test (Wilcoxon signed rank test). a, b, and c denote statistical significance at 1%, 5%, and 10% level, respectively. CAR[-3, +3] Panel A: Firm size Mean Median N Unconstrained Constrained Difference a a Panel B: Payout ratio Mean Median N Unconstrained Constrained Difference a a Panel C: The Whited-Wu index Mean Median N Unconstrained Constrained Difference a a Panel D: The Size-Age index Mean Median N Unconstrained Constrained Difference a a 52

54 Table 8: OLS regression of SEO announcement returns This table presents the results of the OLS regression of seven-day abnormal returns around the SEO announcement date. Abnormal returns are calculated as defined in Table 7. Runup is the cumulative abnormal return over the period starting 60 trading days prior to the announcement and ending 6 trading days prior to the announcement. Volatility is the standard deviation of the market-model residuals over the period from 300 trading days to 61 trading days prior to the offering announcement. Other variables are as defined in Table 4. To mitigate the impact of outliers, all the variables are winsorized at the 1% and 99% levels. p-values are reported in parentheses and based on White standard errors clustered by firm. a, b, and c denote statistical significance at 1%, 5%, and 10% level, respectively. Dependent Variable = CAR [-3, +3] (1) (2) (3) (4) Size Payout WW index SA index High-FC a b b b (0.004) (0.046) (0.022) (0.014) Medium-FC b c (0.020) (0.829) (0.096) (0.126) Ln(Market-to-book) (0.886) (0.362) (0.710) (0.900) Runup (0.194) (0.170) (0.204) (0.219) Volatility b c b b (0.048) (0.086) (0.030) (0.031) Ln(Market value) b (0.034) Constant (0.864) (0.228) (0.818) (0.948) Year dummies? Yes Yes Yes Yes Industry dummies? Yes Yes Yes Yes Observations 1,640 1,640 1,640 1,640 R-squared

55 Table 9: Calendar-time portfolio regressions of post-seo stock returns of issuers by financial constraint status This table reports the intercepts obtained from calendar-time portfolio regressions of issuers stock returns during the one-year (Panel A) and two-year (Panel B) period following the offering year. Each panel presents both equally- and value-weighted intercepts calculated using the Fama-French (1993) three-factor model and Carhart (1997) four-factor model. The intercepts are reported for the full sample and each financial constraint category determined based on the firm size, payout ratio, Whited-Wu index, and Size-Age index classifications. The dependent variable in Panel A (B) is the monthly portfolio returns of SEO firms in excess of one-month T-bill rate for the period between April 1984 (1985) and March 2008 (2009). Post-SEO returns are calculated beginning three months after the SEO fiscal year-end to allow for a reporting lag. Accordingly, in Panel A (B), SEO firm returns are included in portfolio returns for the period from four to fifteen (twenty seven) months after the SEO fiscal year-end. The portfolios are rebalanced each month to drop all the firms that reach the end of their respective post-seo periods and add all the firms that have announced an SEO. The models are estimated with weighted-least square regression, where the weighting vector is the number of firms with non-missing returns in a given calendar month. The calendar months where there are less than five observations are excluded to obtain reliable estimates. a, b, and c denote statistical significance at 1%, 5%, and 10% level, respectively. Panel A: One-year post-seo period 3-Factor / EW 3-Factor / VW 4-Factor / EW 4-Factor / VW Full sample a c Firm size Constrained c Middle b b c Unconstrained Payout ratio Constrained b Middle c Unconstrained c The Whited-Wu index Constrained Middle a b Unconstrained c The Size-Age index Constrained Middle a b Unconstrained c

56 Panel B: Two-year post-seo period 3-Factor / EW 3-Factor / VW 4-Factor / EW 4-Factor / VW Full sample a Firm size Constrained b Middle a b Unconstrained Payout ratio Constrained b Middle b c Unconstrained b The Whited-Wu index Constrained Middle a Unconstrained b The Size-Age index Constrained b Middle b c Unconstrained