Abnormal accruals and external financing

Transcription

1 Abnormal accruals and external financing Theodore H. Goodman Eller College of Management University of Arizona McClelland Hall Tucson, AZ August 2007 ABSTRACT In this paper, I analyze the information conveyed by abnormal accruals by examining the conditions where debt and equity investors place a weight on abnormal accrual when deciding to provide a firm with external financing. I find evidence consistent with the debt market containing sophisticated investors that are able to identify when abnormal accruals predict future cash flow. Upon observing abnormal accruals that anticipate future operating cash flow, debt investors withhold capital from firms where abnormal accruals signal that future cash flow is low and extend capital to firms that where abnormal accruals signal that future cash flow is high. Debt investors also place a positive weight on abnormal accruals that reverse, but only when these reversals are due to increases in cash flow realized in the following period. While debt investors appear to rely on abnormal accruals when these accruals are informative, equity investors appear less sophisticated consistent with the literature on earnings management at equity offerings. This paper is based on my dissertation at the University of Pennsylvania. I would like to thank my committee members: Bob Holthausen, Dave Larcker, Scott Richardson, Cathy Schrand (chair), and Joel Waldfogel. I also thank the following for their helpful comments and suggestions: Dan Bens, Brian Bushee, Mary Ellen Carter, Dan Dhaliwal, Joseph Gerakos, Chris Ittner, Sarah McVay, Jonathan Rogers, Tjomme Rusticus, Bill Schwartz, Mark Trombley, Irem Tuna, Rodrigo Verdi, Sarah Zechman, and workshop participants at the University of Arizona, Baruch College, Michigan State University, Northwestern University, Penn State University, the University of Pennsylvania, Purdue University, and the University of Rochester. I am grateful for financial support from the Deloitte & Touche foundation and the University of Arizona. Any remaining errors are my own.

2 Abnormal accruals and external financing ABSTRACT In this paper, I analyze the information conveyed by abnormal accruals by examining the conditions where debt and equity investors place a weight on abnormal accrual when deciding to provide a firm with external financing. I find evidence consistent with the debt market containing sophisticated investors that are able to identify when abnormal accruals predict future cash flow. Upon observing abnormal accruals that anticipate future operating cash flow, debt investors withhold capital from firms where abnormal accruals signal that future cash flow is low and extend capital to firms that where abnormal accruals signal that future cash flow is high. Debt investors also place a positive weight on abnormal accruals that reverse, but only when these reversals are due to increases in cash flow realized in the following period. While debt investors appear to rely on abnormal accruals when these accruals are informative, equity investors appear less sophisticated consistent with the literature on earnings management at equity offerings.

3 1. Introduction While the use of estimates in accrual accounting affords firms the opportunity to manipulate the resulting accounting figures, the inclusion of estimates also may convey useful information about firm performance. A large literature has examined earnings management initiated with the intent of inflating firm valuations through positive abnormal accruals prior to security offerings (e.g., Teoh, Welch, and Wong, 1998). However, another literature has provided evidence that accruals provide an informative signal for predicting future cash flows (e.g., Dechow, Kothari, and Watts, 1998). In this paper, I examine whether abnormal accruals provide useful information to investors in the context of external financing decisions. I analyze the information conveyed by abnormal accruals by examining the conditions where debt and equity investors place a weight on abnormal accrual when deciding to provide a firm with external financing. This evidence provides information about both the information that may be gleaned from abnormal accruals and the sophistication of these different groups of investors with respect to abnormal accruals. While research on earnings management suggests that positive abnormal accruals reflect discretionary and opportunistic accounting choices, there is still debate over whether discretionary accrual models are well specified (e.g., Guay, Kothari, and Watts, 1996; Ball and Shivakumar, 2006). Specifically, accruals that are above (below) the expected value from a model (e.g., the Jones Model) may indicate future cash flow is expected to increase (decrease). In such a case, abnormal levels of accruals result from accrual accounting conveying useful information about the firm s future operations and an association between abnormal accruals and external financing reflects investors reliance on informative accruals. In contrast to work that examines only the mispricing of abnormal accruals, I provide evidence on mispricing, but also 1

4 cases where investors rely on abnormal accruals and these abnormal accruals correctly indicate when to provide capital and when to withhold capital. I find evidence consistent with the debt market containing sophisticated investors that are able to identify when abnormal accruals predict future cash flow. Upon observing abnormal accruals that accurately predict whether future operating cash flow will increase or decrease, debt investors withhold capital from firms where there is greater risk that their principal will be recovered (i.e., future cash flow is low) and extend capital to firms that will be able to repay funds in the future (e.g., future cash flow is high). These findings suggest that debt investors rely on abnormal accruals to infer the probability that a claim will be repaid. Debt investors also place a positive weight on abnormal accruals that reverse, but only when these reversals are due to increases in cash flow realized in the following period. These results suggest that debt investors understand when abnormal accruals reflect information useful for forecasting future cash flow and when accrual reversals are due to the matching principle (e.g., a decrease in receivables due to a cash collection). The support for this informative role of abnormal accruals raises further questions on the extent to which abnormal accruals effectively measure discretionary accruals. While debt investors appear to rely on abnormal accruals when these accruals are informative, equity investors appear less sophisticated consistent with the literature on earnings management at equity offerings. Equity investors rely on abnormal accruals that appear opportunistic ex-post (i.e., positive abnormal accruals preceding a decrease in operating cash flow). In addition, equity investors tend to place a positive weight on abnormal accruals that will reverse, but not due to an increase in cash flow consistent with an accrual correction of previously inflated abnormal accruals. 2

5 Analysis of both debt and equity financing also provides insight into the conditions that make each of these financing alternatives preferable. When positive abnormal accruals anticipate increases in future cash flow firms appear to prefer issuing debt, despite the fact that equity investors appear to reward positive abnormal accruals that precede decreases in future cash flow. In addition, debt investors provide less capital to firms whose negative abnormal accruals anticipate a decrease in cash flow, but equity investors appear to provide these firms with external financing. This combination of debt and equity results suggests that information about the distribution of cash flow from abnormal accruals may influence the choice between external financing alternatives, where positive abnormal accruals that are informative make debt preferable and negative abnormal accruals that are informative make equity preferable. A large body of research has examined the level of abnormal accruals around external financing events and the reversal of these accruals in future periods (e.g., Rao, Teoh, and Wong, 1998; Teoh, Welch, and Wong, 1998). This paper contributes to this literature by further examining whether abnormal accruals are related to security issuances because these accruals are informative or opportunistic. This paper s motivation is to identify which observations contribute to the overall association between abnormal accruals and external financing. I examine if investors use of abnormal accruals depends upon whether these abnormal accruals are informative or opportunistic ex-post. Comparing the behavior of investors across these ex-post partitions provides insight into the extent to which investors can distinguish between these groups ex-ante. While I do not expect that investors have perfect foresight to discern which cases abnormal accruals are accurate, evidence of differences across these partitions indicates that investors possess the ability to interpret accrual information on-average. The results in this paper 3

6 indicate that the debt investors posses a sophistication that enables these investors to distinguish between these groups. The remainder of this paper is organized as follows. Section 2 contains the hypothesis development, the empirical specification, and the predicted coefficient signs. Section 3 outlines the variable measurement. Section 4 presents the empirical results and section 5 is the conclusion. 2.1 Hypothesis development This paper s research question asks whether accruals convey useful information to investors. To address this question I examine the weight investors place on a firm s abnormal accruals when deciding to provide a firm with financing. Empirical analysis that examines both debt and equity investors allows for the possibility that these different groups of investors may have different levels of sophistication with respect to interpreting abnormal accruals. This paper s central hypothesis is that sophisticated investors extract information about future cash flow from abnormal accruals, while less sophisticated investors rely on accruals that are more opportunistic. Analysis of this hypothesis provides insight into investor sophistication and whether abnormal accruals effectively measure opportunistic accounting or whether these accruals provide useful information about future cash flow in some circumstances. Adverse selection models predict that if a seller has private information, then a rational buyer will price protect against the possibility that the good is over-valued (e.g., Akerlof, 1970). Myers and Majluf (1984) extend these models to the setting of firms issuing securities to fund investing activities, demonstrating that an information asymmetry between the firm and new investors can raise the cost of acquiring external capital relative to internally generated cash. This incremental cost forces a value maximizing firm to not issue securities and under-invest 4

7 when internal resources are low (forgone positive NPV projects). Furthermore, the magnitude of this incremental cost is smaller for securities where the value of the security is less sensitive to the firm s private information, suggesting that to minimize information asymmetry related costs firms prefer to issue safe securities before risky securities (pecking order theory). 1 While rational investors, such as those in the Myers and Majluf (1984) model, would not place a positive weight on a signal that is expected to be uninformative about the value of the firm, empirical research on earnings management suggests that the association between abnormally positive accruals and equity financing indicates that equity investors purchase overvalued securities (e.g., Rao, Teoh, and Wong, 1998; Teoh, Welch, and Wong, 1998; Rangan, 1998; Cassar, 2005). This empirical evidence is consistent with the hypothesis that equity investors are unsophisticated with respect to accruals. However, this interpretation relies on the assumption that abnormal accruals effectively measure discretionary or uninformative accruals. 2 In addition, there is evidence that upon learning of an equity offering, equity investors penalize a firm for recent positive abnormal accruals (Shivakumar, 2000). Prior research has examined whether the relatively more sophisticated parties within the equity market (e.g., insiders) understand the implications of accruals for future earnings and returns (Beneish and Vargus, 2002). The debt market also presents a group of investors that may be able to understand the implications of abnormal accruals. The debt market contains a large proportion of institutional investors, which are expected to have greater skill in interpreting 1 Consistent with theories in which external capital is more costly than internal funds, extant empirical research documents that investing activities are positively associated with cash flow from operations (e.g., Fazzari, Hubbard, and Petersen, 1988). In addition, firms with volatile cash flow invest less on-average due to cash flow shortfalls (Minton and Schrand, 1999). To mitigate these concerns, firms with volatile cash flow tend to hold larger cash reserves (Opler, Pinkowitz, Stulz, and Williamson, 1999) and constrained firms decrease (increase) reserves when cash flow from operations is low (high) (Almeida, Campello, and Weisbach, 2004). However, the benefits of cash reserves must be weighed against the potential agency costs associated with large levels of slack (Harford, 1999). 2 Several papers have debated whether abnormal accrual models are well-specified and if this mis-specification affects the inferences from earnings management research (e.g., Guay, Kothari, and Watts, 1996; Kothari, Leone, and Wasley, 2005; Ball and Shivakumar, 2006). 5

8 financials. In addition, some parties in the debt market (e.g., banks, credit agencies) have access to non-public information which should be useful in evaluating the information presented in accrual. As a result, if debt investors are sophisticated then these investors should respond to abnormal accruals when these abnormal accruals contain useful information about future cash flow. Debt investors also differ from equity investors based on the type of information that is useful for valuation. If positive abnormal accruals signal that a firm will have high future cash flow, then this signal reduces debt investor concerns about the firm s ability to repay debt in the immediate future. As debt claims have a fixed upside (i.e., the principal and coupon payments), a signal that a firm will have high cash flow reduces the sensitivity of the debt to the firm s assets (or private information about those assets). As an extreme example, Myers and Majluf (1984) note the case of risk-free debt, where the purchasers of risk-free debt are not concerned about any information asymmetry regarding the total value of assets in place because there is no risk that this debt is over-valued. It is important to note that for a firm s debt to be risk-free does not require removing all uncertainty related to the firm s assets, only the uncertainty that it will make its debt payments. It is more difficult to construct cases where there is no uncertainty regarding an equity claim as this would require removing all uncertainty about these good outcomes. 3 In addition, the longer horizon of equity (relative to debt) further requires that an abnormal accrual signal be highly persistent for it to reduce uncertainty about cash flows many years in the future. Recent work by Cassar (2005) presents evidence that the results suggesting earnings management around equity financing appear to extend to debt financing: debt financing is 3 For example, if debt investors learn that the value of the firm s assets will be greater than an amount with certainty, then the firm can issue risk-free debt as long as the face value of the debt is less than that amount. In contrast, if equity investors learn that the value of assets is above some amount with certainty, then there still could be uncertainty and possibly asymmetry about the value of the equity claim as long as there is still variation in the value of the firm s assets conditional on information that the value is above a bound (e.g., the distribution is continuous). 6

9 positively related to contemporaneous abnormal accruals and debt financing is negatively related to the association between abnormal accruals and future earnings. 4 I examine when abnormal accruals appear to convey useful information to debt and equity investors. I analyze whether investors appear to distinguish between different types of abnormal accruals. In addition, instead of examining if an increase in financing coincides with more persistent accruals, I examine whether investors appear to obtain information from abnormal accruals that informs their decision to extend or to withhold financing. Specifically, sophisticated investors are expected to provide capital when positive abnormal accruals are accurate ex-post, but also withhold capital when negative abnormal accruals are accurate ex-post. 5 In addition, I provide further details on the interpretation of transitory accruals by distinguishing between different types of reversals based on whether the reversal coincides with an increase in cash flow. 2.2 Empirical models and predictions This paper s investigates whether investors with different degrees of sophistication (debt vs. equity) place different weights on informative and opportunistic abnormal accruals (RES_ACC t-1 ) when determining amount of cash to offer the firm in exchange for securities in year t (DEPVAR t ). I begin my analysis with the following model: DEPVAR t = α 0 + α 1 RES_ACC t-1 + CONTROLS + error (1) 4 While Cassar (2005) documents a positive association between contemporaneous abnormal accruals and debt financing, this result is sensitive to matching choices and the use of control variables. When the association is estimated after matching based on total financing the association becomes negative and significant when control variables are excluded and positive and insignificant when control variables are included. In addition, when examining whether debt financing affects the association between abnormal accruals and future earnings, these results are sensitive to whether debt financing is also interacted with other independent variables (cash flow from operations and normal accruals). 5 In such a case, sophisticated investors rely on abnormal accruals that convey useful information both when providing and withholding capital. As a result, if persistence is used as a measure of when abnormal accruals convey useful information, then I would predict greater persistence both when sophisticated investors provide capital and also when sophisticated investors withhold capital. 7

10 Where: DEPVAR t =Financing cash flow in year t (TOTFIN t, DEBT t, EQUITY t ) TOTFIN t = EQUITY t + DEBT t dataxx Refers to Compustat data item number XX DEBT t =Cash proceeds from the issuance of debt (data 111) plus the change in notes (data 301) less the payments to reduce long term debt (data 114) ) divided by average assets EQUITY t =Cash proceeds from the sale of common/preferred stock (data108) cash dividends (data127) - repurchases (data 115) divided by average assets RES_ACC t-1 =Abnormal accruals in year t-1 CONTROLS =Other determinants of DEPVAR The coefficient on RES_ACC t-1 (α 1 ) in equation 1 provides evidence of whether a group of investors rely on abnormal accruals. Equation 1 is similar to the empirical specification used in prior work to examine the association between abnormal accruals and financing activities. Using a similar model, Dechow et. al (2006) find that after controlling for other changes in the balance sheet contemporaneous with the measurement of total accruals, total accruals are negatively associated with net equity financing activities. In fact, firms with high total accruals tend to distribute cash to equity investors. 6 Equation 1 provides evidence of whether investors rely on abnormal accruals on-average. However, there may be variation in the extent to which abnormal accruals contain useful information. In such a case, further analysis is necessary to infer whether a positive estimate of α 1 reflects reliance on informative or opportunistic abnormal accruals. To identify whether an investor s use of abnormal accruals reflects a reliance on informative or opportunistic abnormal accruals, I examine whether α 1 is significant in different sub-samples and whether the coefficients are different across sub-samples. 6 In Table 9, Dechow et. al (2006) report that accruals in period t are positively correlated with disbursement to equity investors in period t+1 and negatively related to disbursements to debt investors in period t+1. 8

11 To distinguish between informative or opportunistic abnormal accruals, I evaluate whether the firm s realized future performance corresponds to the abnormal accruals. Specifically, I examine whether the weight investors place on abnormal accruals varies depending on whether the sign of these abnormal accruals is consistent with the sign of the expost change in cash flow. I use future cash flow to define positive and negative future outcomes as the incremental ability to accruals to predict future cash flow is often noted as an advantage of accrual accounting relative to cash basis accounting (Dechow, Kothari, and Watts, 1998; Barth, Cram, and Nelson, 2001). 7 Partitioning based on ex-post cash flow data identifies cases where the level of accruals is abnormal relative to a benchmark model (e.g., the Jones model), but this deviation may have been due to an anticipation of future cash flow performance. If investors are able to distinguish between these different types of abnormal accruals ex-ante, then their weight on abnormal accruals should vary across these samples. By examining partitions based on ex-post future cash flow, I assume that firms have private information about the change in cash flow from pre-issuance to post-issuance. This assumption is similar to the Myer and Majluf (1984) assumption that firms have private information about the value of asset in place: firms know the sum of its discounted cash flows from theses assets. By assuming that some of an issuing firm s information advantage is due to knowledge of short-term cash flows, I examine whether abnormal accruals are able to communicate information to sophisticated investors and reduce the costs related to this information asymmetry. The inclusion of ex-post cash flow data into equation 1 is as follows. First, I replace RES_ACC t-1 in equation 1 with a piece-wise specification splitting RES_ACC t-1 into a positive 7 Dechow, Kothari, and Watts (1998) find that forecasts of cash flows based on earnings yield lower mean squared errors than forecasts based on historical cash flow. In addition, accruals have an incrementally positive coefficient in models predicting future cash flow (Barth, Cram, and Nelson, 2001). 9

12 (POS_RES_ACC t-1 ) and negative component (NEG_RES_ACC t-1 ). Next, I interact POS_RES_ACC t-1 and NEG_RES_ACC t-1 with indicators for whether future cash flow has increased (POS_ CF t+1 ) or decreased (NEG_ CF t+1 ) between year t-1 and year t+1. This yields the following model: DEPVAR t = β 0 + β 1 NEG_ CF t+1 *POS_RES_ACC t-1 + β 2 POS_ CF t+1 *POS_RES_ACC t-1 + β 3 NEG_ CF t+1 *NEG_RES_ACC t-1 + β 4 POS_ CF t+1 *NEG_RES_ACC t-1 + CONTROLS + error (1a) Tests of the coefficients in equation 1a are designed to infer whether positive and negative accruals are associated with financing activities and whether this association is stronger when accruals predict future cash flow or do not predict future cash flow. 8 If equity investors are misled by abnormal accruals, as suggested by the research on earnings management, then these investors will place a weight on abnormal accruals when preissuance abnormal accruals are high and realized future cash is low (β 1 >0). In such a case, issuing firms appear to benefit from issuing equity when accruals are opportunistic given the firm s knowledge of future cash flow: abnormal accruals are positive, but future cash flow performance will decrease. If debt investors are able to infer useful information from abnormal accruals then the weight on abnormal accruals will be positive in cases where the sign of abnormal accruals correctly predicts the sign of future cash flow changes (β 2 >0 and β 3 >0). As a result, debt investors withhold (provide) financing when they infer from abnormal accruals that future cash flow will be (high) low. The prediction of a positive coefficient on 8 This regression model is similar to the piece-wise model estimated by Beneish and Vargus (2002) in their analysis of whether positive and negative accruals are more strongly associated with a dependent variable (future earnings and returns) when they coincide with another variable (insider trading indicators). 10

13 NEG_ CF t+1 *NEG_RES_ACC t-1 is distinct from the hypothesis that debt investors prefer firms with less volatile accruals as in this case where accruals are accurate ex-post (i.e., lower volatility), but because the sign of the accrual is negative the positive coefficient implies lower debt financing. Equation 1a also provides insight into the interaction between debt and equity financing. Specifically, it provides the opportunity to examine if equity financing becomes more appealing when negative abnormal accruals make debt financing more costly by signaling. In such a case, debt financing is less attractive as the probability of default is higher due to lower expected future cash flow. However, equity investors may not penalize firms for negative abnormal accruals (DEPVAR= EQUITY t, β 3 <0), if the persistence of this bad news is limited. Alternatively, a negative weight on these negative abnormal accruals by equity investors may reflect that in addition to over-weighting positive accruals that are inaccurate ex-post, equity investors also mistakenly under-weight negative abnormal accruals that are accurate ex-post. In addition a comparison of the use of abnormal accruals by debt and equity investors provides insights into the extent to which firms prefer debt over equity. When positive abnormal accruals anticipate positive future cash flow, debt investors should be willing to provide capital as these investors are able to interpret the abnormal accrual. In this case, equity investors may be willing to provide capital as well if these investors myopically place a positive weight on positive abnormal accruals. If both debt and equity investors are willing to provide the same amount of capital, then a firm s choice between these alternatives provide insight into the relative costs of debt and equity. If firms exhibit a preference for debt over equity (pecking order theory), then these firms may use abnormal accruals to obtain debt financing (DEPVAR= DEBT t, β 2 >0), but not issue equity (DEPVAR= EQUITY t, β 2 =0). Interestingly, 11

14 this would result in equity investors primarily providing capital to positive abnormal firms, where these abnormal accruals are opportunistic. The following table summarizes the predictions for equation 1a: DEPVAR t = DEBT t DEPVAR t = EQUITY t Predicted sign for β 1? + Predicted sign for β Predicted sign for β Predicted sign for β 4?? In addition to testing whether the coefficient on a sub-sample of abnormal accruals is different from zero, these coefficients can be compared with other sub-samples (e.g., β 1 with β 2 or β 3 with β 4 ). These comparisons do not require perfect foresight of future cash flow by investors. Instead these comparison only requires that sophisticated investors are able to glean information about future cash flow from analyzing abnormal accruals and as a result put a more positive weight on abnormal accruals that are accurate ex-post (β 2 >β 1 and β 3 >β 4 ). Data on future cash flow changes also provides the opportunity to further explore the interpretation of accrual reversals around financing events. Research in earnings management often implies that reversals of abnormal accruals following an issuance are evidence of opportunism. These studies suggest that abnormal accruals decrease because firms must recognize losses following the issuance to reverse the fictional pre-issuance gains. However, it is important to emphasize that the losses that drive the reversal are some type of accrual correction, not a cash flow. For example, if inventory had been opportunistically inflated, then inventory will decrease through a write-down, but this write-down does not affect cash flow. If this inventory is sold, then there would be a cash inflow. In such a case, the ex-post cash flow 12

15 obtained from liquidating the inventory provides a signal on the extent to which the inventory was over-valued ex-ante. While accrual reversals could be evidence of opportunism, a decrease in abnormal accruals following an issuance could also reflect the application of the matching principle to account for a transitory cash flow. For example, if a firm builds up its inventory for a sale that is non-recurring, then the firm will experience a cash inflow from this sale and a decrease in inventory. However, once the sale is made the firm will not need to restock its inventory, so inventory accruals will appear to reverse. In such a case, abnormal accruals correctly anticipate a cash inflow, albeit a transitory cash flow. These types of abnormal accruals reversals should be particularly useful for valuing debt, due to its relatively shorter maturity compared to equity. The inclusion of data on ex-post cash flow and reversals in abnormal accruals into equation 1 is as follows. First, I replace RES_ACC t-1 in equation 1 with a piece-wise specification splitting RES_ACC t-1 into a cases where there is a reversal (REV_RES_ACC t-1 ) and cases where there is no reversal (NOREV_RES_ACC t-1 ). Next, I interact REV_RES_ACC t-1 with an indicator for whether future cash flow has increased (POS_ CF t+1 ) or decreased (NEG_ CF t+1 ) to explore the different types of reversals. This yields the following model: DEPVAR t = γ 0 + γ 1 NEG_ CF t+1 *REV_RES_ACC t-1 + γ 2 POS_ CF t+1 *REV_RES_ACC t-1 + γ 3 NOREV_RES_ACC t-1 + CONTROLS + error (1b) Tests of the coefficients in equation 1b are designed to examine whether the weight placed on abnormal accruals depends on the whether these accruals reverse and if this reversal coincides with an increase in cash flow. 13

16 If debt investors are sophisticated and accruals that eventually decrease due to the application of matching (i.e., cash flow increases), then debt investors will place a positive weight on these abnormal accruals (γ 1 >0). The prediction that debt investors value information about transitory short-term cash flow is motivated by the relatively shorter horizon relative of these securities compared to equity. As a result, I expect that empirical support for the predictions of equation 1b predicting debt financing would be strongest when the debt is shortterm. To address this concern, I examine changes in notes payable as a robustness test (section 4.5) If equity investors are unable to anticipate that abnormal accruals will decrease in a manner consistent with opportunistic account (i.e., cash flow from operations does not increase), then these investors will place a positive weight on these abnormal accruals (γ 2 >0). This prediction is motivated by the claims that accrual reversals following equity issuances signal opportunism, by further refining the definition to measure cases where the reversal are no due to normal accrual accounting (e.g., a collection of a receivable). The following table summarizes the predictions for equation 1b: DEPVAR t = DEBT t DEPVAR t = EQUITY t Predicted sign for γ 1? + Predicted sign for γ 2 +? As with the comparisons of the coefficients in equation 1a, a comparison of γ 1 with γ 2 provides the opportunity to examine whether sophisticated investors are able to distinguish between these types of reversals. 3. Variable measurement 3.1 Measurement of cash flow and accruals 14

17 As the distinction between cash flow and accruals is important for my research design, I use cash flow statement data to measure CF and RES_ACC. This measurement of accruals and cash flow is preferable to changes in balance sheet accounts as it is not confounded by non-cash transactions (e.g., mergers and acquisitions, foreign currency translations) (Collins and Hribar, 2002). As in Kothari, Leone, and Wasley (2005), I define RES_ACC t as the residual from the Jones model after removing the level of abnormal accruals for firms with comparable ROA (return on assets) performance. I follow their specification where a constant is included when estimating the cross-sectional Jones Model for each industry-year, where industry is equal to a firm s 2 digit SIC code. To control for abnormal accruals related to ROA performance, I match based on performance by taking the average level of abnormal accruals from a firm s ROA peer group within the same industry-year. Peer group are assigned by dividing all industry-year firms into quintiles based on ROA in year t. The following definitions are used to calculate RES_ACC t : ACC t Jones Model (1/ASSETS t-1 ) SALES t PP&E t ROA RES_ACC t =Net Income (data18) Cash flow from operations (data308) divided by lagged total assets (data6) ACC t = δ 0 + δ 1 (1/ASSETS t-1 ) + δ 2 SALES t + δ 3 PP&E t + ε This model is estimated at the 2 digit SIC code level each year. The Jones model is only estimated using observations where the dependent is less than one in absolute value. The model is only estimated for industryyears with at least 10 observations. =1 divided by lagged total assets =Change in sales (data12) in year t divided by lagged total assets =Gross PP&E (data8) in year t divided by lagged total assets =Net Income (data18) divided by average assets = ε i for firm i - average ε i of peer group ROA peer groups are obtained by dividing the 2 digit SIC code into quintiles based on ROA. When calculating the average ε for firm i s peer group firm i is excluded. 15

18 I define CF as the amount of funds produced by operations that can be used in investing activities. To measure cash flow produced by operations before investing choices I remove R&D from reported operating cash flow. The definition is listed below: CF = Cash flow from operations, defined as reported operating cash flow (data308) before R&D expense (data46) divided by average assets. R&D expense is set equal to zero when missing. While this definition of CF is used in Richardson (2006), it is in contrast to research in finance that has used earnings before interest, taxes, depreciation and amortization (EBITDA) to measure cash flow. For example, Bushman, Smith, and Zhang (2005) note that research in finance uses EBITDA to measure cash flow from operations and that a large portion of the association between EBITDA and investment is attributable to working capital. Because this paper s hypotheses center on predicting cash flow using accrual information, the distinction between cash flow (which can be used in investing activities) and accruals (which can predict future cash flow, but not directly fund investment) is important. 3.2 Measurement of dependent variables The dependent variable in equation 1 is equal to the amount of financing received from different sources (TOTFIN t, DEBT t, and EQUITY t ) calculated using data from the statement of cash flows and balance sheet as in Richardson and Sloan (2003). By examining the proceeds of an issuance, I rely on Myers and Majluf s (1984) result that when information asymmetry costs are too high firms decide to not issue new securities. Theory would also support a prediction related to the incremental cost of external capital. Several papers have examined the association between the cost of external capital and both cash flow volatility (Minton and Schrand, 1999) and the characteristics of earnings (Francis, Lafond, Olsson, and Schipper, 2005; 16

19 Bharath, Sunder, and Sunder, 2006; Jiang, 2006). However, empirical analysis of proceeds requires controlling for the amount of cash needed to avoid under-investment, while analysis of the cost of external capital requires controlling for the internal cost of capital. 9 In my analysis, I include control variables to measure growth opportunities and internal resources which should determine the amount of cash needed from external financing to prevent under-investment. Using a dependent variable that reflects the magnitude of external financing assumes that when the proceeds from an issuance are larger, the incremental cost of external capital is smaller. As a sensitivity analysis, I convert the dependent variables into indicators for whether a firm obtained external financing of a given type and the main results of this study are qualitatively similar. 3.3 Measurement of control variables I include control variables in equation 1 to limit four omitted variable concerns. First, I include cash flow in year t-1 and year t to ensure that RES_ACC t-1 captures information about expected future cash flow, rather than information about historical cash flow or cash flow that is contemporaneous with the financing activity. Second, I include proxies for growth opportunities to limit concerns that RES_ACC t-1 captures the expected benefits from outside financing. Third, I include balance sheet information about a firm s liquidation values to examine the incremental information conveyed by RES_ACC t-1 which is based on income statement data. Fourth, I include variables measuring the firm s leverage and past financing activities which may influence the costs of additional external financing. 9 Examining the cash obtained from investors and the cash used in investing activities in a given firm-year also alleviates concerns inherent in measuring financing constraints indirectly through cash flow-investment sensitivities. For example, Kaplan and Zingales (1997) note that cash flow-investment sensitivities may not be an appropriate measure of firm financing constraints as theory does not necessarily predict a monotonic relation between these sensitivities and the cost of external capital. However, examining financing cash flows in specific firm-years requires identifying a sample that needs external capital to prevent under-investment. 17

20 To ensure that the association between RES_ACC t-1 and financing activity reflects the presence of accounting information that is used to shape investor expectations of future cash flow and not differences in the pattern of cash flows, I include the level of CF in year t-1 and t as control variables. Including CF t-1 controls for other information that is available in year t-1. Including CF t also provides a measure of the need for external capital during year t (Richardson, 2006). When issuing securities a firm balances the cost of external capital with the benefits from initiating new projects. To control for variation in these benefits I include proxies for growth opportunities. Growth opportunities are expected to be negatively related to firm age (AGE) and the firm s book to market ratio (BTM) and positively related to investing expenditures made during year t-1 (ITOTAL). The definitions of the growth opportunity variables are listed below: AGE BTM ITOTAL =Log(years since company first appears in the CRSP monthly stock file) =BV of assets (data6) divided by MV of assets (MV of equity (data25*data199) + BV of liabilities (data181)) =[Capital expenditures (data128) + R&D (data46) + Acquisitions (data129) Sale of PPE (107)] divided by average assets To examine the association between RES_ACC t-1 and external financing also requires controlling for the information contained in the balance sheet. Berger (1999) notes that when assessing the use of income statement data as an indicator of the probability of default, it is important to control for balance sheet data that provides information on the liquidation value of a firm s assets. As an asset s location on the balance sheet (e.g., current, fixed, etc.) is associated with its liquidation value (Berger, Ofek, and Swary, 1996), the composition of a firm s assets is relevant when evaluating their liquidation value of the firm. In addition, recent work by Almeida and Campello (2005) indicates that asset liquidation values are related to cash flow-investment 18

21 sensitivities; firms with higher liquidation values are less likely to be constrained, but among constrained firms the affect of cash flow from operations on investing activity increases in liquidation value for financially constrained firms. The definitions of the asset liquidity variables are listed below (variables measured at the beginning of year t): CASH AR INV =Cash (data1) divided by total assets =Accounts Receivable (data 2) divided by total assets =Inventory (data3) divided by total assets By including the ratio of a firm s primary current assets to total assets (CASH, AR, INV) as control variables, the remaining percent of total assets composed of non-current assets (PP&E and intangibles) is the reference group. I include variables measuring the firm s leverage and past financing activity which may reflect various costs associated with raising new capital. The level of firm leverage may reflect either debt capacity or the presence of other constraints which affect capital structure (Kaplan and Zingales, 1997). Prior cash flow from financing ( DEBT t-1, and EQUITY t-1 ) are included to limit concerns that expectations formed in year t-1 are related to financing activity in year t-1, which prior work suggests will have implications for financing activities in year t-1 (Dechow et. al, 2006). The definitions of the capital structure variables are listed below (all variables are measured in the beginning of year t): LEV =Debt/(Debt + Equity), where debt=(data9 + data34) and equity=data60 4 Empirical Results 4.1 Descriptive statistics 19

22 I begin my sample selection with the universe of non-financial firm-year observations that have non-missing data for CF and RES_ACC for a given year (year t) and the two adjacent years (t-1 and t+1). Next, I require that all sample firms have data for all dependent variables in year t ( DEBT t and EQUITY t ) and all control variables. I also eliminate extremely small firms (sales or average assets under $10 million) and firms where the absolute value of CF or RES_ACC exceeds 1 in year t-1, t, or t+1. Finally, I remove firms listed as ADRs on CRSP. 10 There are 45,409 firm-year observations that meet these criteria from 1988 to In addition, to limit the influence of extreme observations, I winsorize the top and bottom one percent of all variables before estimating equation 1. Panel A of table 1 presents descriptive statistics for both the dependent and independent variables used in this paper s analysis. The medians for DEBT t and EQUITY t are both zero, while the means are both positive. Both the mean and median for RES_ACC t-1 are slightly negative (Mean=-0.008, Median=-0.007). As this residual was estimated using all firms in Compustat with available data and there are other sample selection criteria to reach this study s final sample, this mean may not equal zero for the sample. Panel B of table 1 presents descriptive statistics on the univariate correlations between RES_ACC t-1 and financing activities. Consistent with prior research, there is a positive and significant univariate correlation between RES_ACC t-1 and all measures of external financing ( EQUITY t, DEBT t, and TOTFIN t ) when calculated using either pearson or spearman correlations. While this univariate correlation is consistent with investors relying on accruals in general, it does not distinguish between the hypothesis that this correlation is due to earnings management from the hypothesis that this correlation is due to informative accruals. In addition, 10 ADRs are identified as all firms where the CRSP share code is greater than 29 and less than

23 the correlation matrix documents a negative and significant correlation between EQUITY t and DEBT t, consistent with prior research on external financing (Richardson and Sloan, 2003). Table 2 presents OLS estimates of equation 1, which predicts the amount of cash flow from the different financing activities as a function of abnormal accruals and other control variables. In column 1, where the model is predicting DEBT t, the coefficient on RES_ACC t-1 is positive and significant (t=11.407) consistent with the univariate correlations in panel b of table 1. However, in column 2, where the model is predicting EQUITY t, the coefficient on RES_ACC t-1 is negative and significant (t=-4.178) in contrast to the positive univariate correlation between RES_ACC t-1 and EQUITY t. While this change in sign contrasts with the univariate descriptive statisics it is consistent with empirical results presented in Dechow et. al (2006). Dechow et. al (2006) document a positive association between accruals and future debt financing and a negative association between accruals and future equity financing. 11 However, this analysis does not yet clarify whether the association between external financing and abnormal accruals reflects earnings management or informative accruals. Column 3 of table 2 indicates that there is a positive association between RES_ACC t-1 and TOTFIN t. While the results in columns 1 and 2 were consistent with the work in Dechow et. al (2006) the evidence in column 3 differs with Dechow et al. s finding that total accruals are positively related to total distributions (debt plus equity). This difference with respect to total financing may be due to the measurement of accruals, the presence of additional control variables or a difference in samples. 12 In subsequent tests, I explore the factors that influence the 11 The analysis in Dechow et al. (2006) is based on total accruals (both current and long-term) instead of abnormal accruals. 12 Dechow et. al (2006) only include variables capturing the change in the balance sheet accounts during the prior year as independent variables. The model contains lagged total accruals, the lagged change in cash balance, and lagged financing activity as independent variables. Rather than include the change in cash balance, the specification in this paper includes cash flow from operations (CF t-1 ), cash flow used in investing activities (ITOTAL t-1 ), and financing activities ( EQUITY t-1, DEBT t-1 ) which together approximate the total change in the cash balance. 21

24 weight on abnormal accruals for debt, equity, and total financing, which also clarifies the association between total financing and abnormal accruals. 4.2 Empirical analysis of Equation 1a Panel A of table 3 provides evidence on whether positive or negative abnormal accruals have different effects on external financing. This analysis provides a baseline model examining whether investors weight on abnormal accruals depends on the sign, while the models in panel B provide evidence on whether investors discern between cases where the sign of abnormal accruals matches the sign of changes in future operating cash flow (equation 1a). Column 1 indicates that debt investors place a positive weight on both positive and negative abnormal accruals; the coefficients on both POS_RES_ACC t-1 and NEG_RES_ACC t-1 are positive and significant (t=5.582 and t=8.926, respectively). While the coefficient on NEG_RES_ACC t-1 is larger than the coefficient on POS_RES_ACC t-1 the difference between these coefficients is not significant at the ten percent level (p-value=0.1189, not tabled). Overall, the positive coefficients on both positive and negative abnormal accruals provides further evidence that on-average debt investors rely on accounting information for both good and bad news when deciding to provide the firm with capital. Column 2 suggests that the weight placed on abnormal accruals by equity investors depends on the sign of abnormal accrual. Equity investors provide capital to firms with positive abnormal accruals; the coefficient on POS_RES_ACC t-1 is positive and significant (t=3.841). In addition, equity investors provide financing to firms with negative abnormal accruals; the coefficient on NEG_RES_ACC t-1 is negative and significant (t=-8.644). This asymmetry between positive and negative abnormal accruals is consistent with existing work where positive 22

25 abnormal accruals play a more prominent role in the accrual anomaly (Beneish and Vargus, 2002; Kothari, Loutskina, and Nikolaev, 2006). These equity results in column 2, in combination with the debt results in column 1, also provide evidence of pecking order theory. If firms prefer to issue debt as postulated by pecking order theory, then firms only choose to issue equity when other costs make debt too costly. If debt investors place a greater weight on accounting signals of near-term performance relative to equity investors, then firms with negative abnormal accruals choose obtain less debt financing as it is now more costly (a positive coefficient on NEG_RES_ACC t-1 in column 1) and instead choose to issue equity (a negative coefficient on NEG_RES_ACC t-1 in column 2). Column 3 provides evidence on total financing activities. Firms with positive abnormal accruals tend to obtain more total external financing, a positive and significant coefficient on POS_RES_ACC t-1 (t=7.081). However, the coefficient on NEG_RES_ACC t-1 is insignificant. Given the results in the columns 1 and 2, a negative abnormal accrual may limit debt financing, but this is offset by more equity financing, resulting in no effect on total financing. Furthermore, these results suggest that the association between abnormal accruals and total financing estimated for an entire sample is likely to be influenced by the percent of firms in that sample with positive abnormal accruals in that sample. Panel B of table 3 presents of estimates of equation 1a. Column 1 indicates that the positive coefficients on POS_RES_ACC t-1 and NEG_RES_ACC t-1 documented in column 1 of panel A are due to cases where the sign of the accruals is consistent with the sign of the change in cash flows from year t- to t+1. Specifically, the coefficients on both POS_ CF t+1 *POS_RES_ACC t-1 and NEG_ CF t+1 *NEG_RES_ACC t-1 are positive and significant (t=9.727 and t=7.995, respectively). These positive coefficients indicate that positive 23