Corporate financing and target behavior: New tests and evidence

Journal of Corporate Finance 48 (2018) 840–856

Contents lists available at ScienceDirect

Journal of Corporate Finance journal homepage: www.elsevier.com/locate/jcorpfin

Corporate financing and target behavior: New tests and evidence Gaurav Singh Chauhan a,⁎, Fariz Huseynov b a b

Indian Institute of Management Indore, Department of Finance and Accounting, Indore, Madhya Pradesh 453556, India North Dakota State University, College of Business, P.O. Box 6050, Dept. 2410, R. H. Barry Hall 202, Fargo, ND 58108, United States

a r t i c l e

i n f o

Article history: Received 22 June 2016 Received in revised form 18 October 2016 Accepted 20 October 2016 Available online 21 October 2016 JEL classification: G1 G30 G32 Keywords: Capital structure Trade-off theory Target behavior Financing choices Mean reversion

a b s t r a c t This study addresses the recent concerns in the capital structure literature about the reliability of tests of the target-following behavior. Using a novel testing strategy, we examine whether and to what extent firms' financing choices–rather than the movement of their debt ratios per se – concur with the target-following behavior. We find that firms' financing decisions are not generally consistent with systematic target-following. Our results remain similar when we examine an extended period of time and also when we consider that firms may have a range of target debt ratios rather than a unique target or varying financial constraints. Our results are also robust to different target specifications and our methodology can reliably distinguish the target behavior from random financing. Further tests also confirm our results by suggesting that the firms' financing decisions are not primarily driven by deviations from the firms' target debt ratios. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Do firms follow a target capital structure, and if so, how do they respond to deviations from their target debt ratios? Recent studies that estimate the speed of adjustment to examine the mean reversion of firms' debt ratios to a target debt ratio mostly suffer from several econometric challenges.1 Additionally, the speed of adjustment may not be an economically meaningful measure of the firms' target-following behavior (Hovakimian and Li, 2012). Their findings may also be affected by the mechanical mean reversion of debt ratios due to factors such as inter-temporal patterns for external financing needs (Shyam-Sunder and Myers, 1999) and debt ratios that are bounded between zero and unity (Chen and Zhao, 2007). Chang and Dasgupta (2009) show that the results of partial-adjustment, dynamic trade-off models can be obtained even if firms follow a random financing behavior that is independent of a target debt ratio. They argue that using leverage ratios in these tests may be misleading and that future tests should focus on the firms' issuance activities in order to reject alternative, non-target behavior successfully. The authors also suggest that a firm's financing choice to use debt versus equity should be conditioned on whether the firm faces a positive or negative financing deficit.

⁎ Corresponding author. E-mail addresses: [email protected] (G.S. Chauhan), [email protected] (F. Huseynov). 1 See, for example, Fischer et al. (1989), Hovakimian et al. (2001), Hennessy and Whited (2005), Flannery and Rangan (2006), Strebulaev (2007), and Huang and Ritter (2009) for a discussion of these models as well as Flannery and Hankins (2013), Hovakimian and Li (2011), and Welch (2007) for a discussion about the econometric challenges.

http://dx.doi.org/10.1016/j.jcorpfin.2016.10.013 0929-1199/© 2016 Elsevier B.V. All rights reserved.

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

841

Our paper develops a new testing strategy to examine the firms' target behavior such that intentional target behavior can be successfully differentiated from the random financing behavior. Unlike previous studies, where the target behavior was inferred from the movement in the debt ratios per se, we focus on the firms' financing choices to draw such inferences. Because the movement in debt ratios could be misleading, we consider a firm's financing choice as a conscious attempt to move its debt ratio toward the intended target level. To measure the intensity of target-following, we examine whether a firm's financing choices are consistent with the reversion to a target. The intensity of target-following is strong (weak) when a firm issues or retires the intended or the “right” (the unintended or the “wrong”) security to a greater extent. Although we utilize different methods to estimate the target debt ratios, we merely use these ratios to determine the intended direction of the movement in actual debt ratios rather than to identify target behavior, per se. Because our empirical analysis is less sensitive to the functional forms of target debt ratios, we are able to use simple empirical models to test the target behavior in a robust and intuitive manner. We apply our testing methodology by using financial-statement data for U.S. firms during the period between 1988 and 2013. We focus on firms with moderate debt ratios because Chen and Zhao (2007) argue that firms with debt ratios that are too low or too high may not be very concerned about financing choices because their debt ratios tend to mechanically mean revert unless they make extraordinary financing decisions.2 To begin, we separate firms into four categories based on whether they are under-levered or over-levered with respect to their target debt ratios and based on whether they subsequently issue or retire capital (debt and/or equity). Therefore, we are able to identify, in each category, the specific financing choices that are consistent with the target behavior. We expect that under-levered firms, which are seeking the net issuance (retirement) of capital in subsequent periods, issue (retire) more debt (equity) than equity (debt). Likewise, we expect that over-levered firms, which are seeking the subsequent net issuance (retirement) of capital, issue (retire) more equity (debt) than debt (equity). We find that the financing choices for the firms in our sample are not generally consistent with the systematic target behavior. Contrary to the prediction about issuing or retiring the specific “right” security for the firm's respective category, a greater proportion of firms tend to issue or retire more debt than equity across all categories. Furthermore, the proportion of firms that do not conform to the target behavior, issue or retire the “wrong” security quite intensely when they are not predicted to do so. This behavior primarily suggests that adjustment-cost concerns are not primarily driving the issuance or retirement of these “wrong” securities. For example, while the majority of the over-levered firms that resort to net issuances prefer debt to equity, they also issue debt quite intensely or use equity to a minimum extent in the financing mix, contradicting the predictions of the target-following behavior. Although our results suggest that a greater proportion of firms predominantly issue or retire debt, the findings may not suggest a pecking order proposed by Myers (1984). Consistent with Frank and Goyal (2003), we also find that a significant proportion of firms predominantly issue or retire equity.3 Following Chang and Dasgupta's (2009) suggestions, we compare our findings to the results obtained by using simulated datasets for an intentional target behavior with varying intensity, as well as simulated datasets for random financing. We show that our testing methodology successfully differentiates between the target and random financing behavior. We also confirm that our results remain similar even after considering the effect of the adjustment costs and frictions in the financial markets that may prevent immediate reversion.4 The firms' aggregate financing behavior remains unchanged and inconsistent with the target-following for several subsequent financing rounds after experiencing a deviation from the target debt ratios. Furthermore, a significant proportion of firms that exhibit the target behavior for a given period do not seem to pursue it further in the following periods. Our results remain similar even when firms do not have a strictly defined target, but instead follow a range around their target debt ratios. These findings suggest that adjustment costs may not be the prime deterrent to the firms' targetfollowing. Our results are robust to different measures of the target debt ratio. For example, along with the target debt ratios that are estimated through the linear regression models, we estimate the debt ratios using non-linear regression models, such as the two-part fractional logistic model suggested in Ramalho and Vidigal da Silva (2013), to address the bounded nature of debt ratios. Moreover, aggregate financing behavior remains the same even when the targets are replaced by the respective contemporary median debt ratios for a particular industry and also when we use seemingly unrelated target debt ratios. Although the aggregate financing behavior is inconsistent with the systematic target-following, as a primary condition for target-following, firms' financing choices should be largely driven by the extent of deviation between the actual and the target debt ratios. Larger deviations should motivate firms to issue or retire a significant amount of the predicted security in each category in order to effectively close the gap between the actual and target debt ratios. We formally investigate the extent to which deviations from target debt ratios, among other determinants, drive the firms' predicted financing choices for each category. We find that deviations from the target debt ratios do not significantly influence the firms' financing decisions, an important finding because the primary reason to follow the systematic target behavior is missing. However, other firm-specific characteristics seem to significantly influence the firms' debt-equity choice. Factors, such as size, profitability, and growth opportunities, have a significantly higher impact on firms' financing choices when compared to the deviation from target debt ratios. Specifically, larger, profitable, and high-growth firms tend to significantly retire more equity than debt when facing a financing surplus. For firms that have a financing deficit, high-growth firms issue more equity while firms with very low ex-ante distress costs tend to issue

2

Later, as part of the robustness tests, we discuss our results and include firms with extremely low or high debt ratios, too. Further, merely finding a higher proportion of firms that use debt may not suggest the pecking order proposed by Myers (1984) unless we carefully examine several postulates of such a pecking order, as shown in Leary and Roberts (2010). 4 Leary and Roberts (2005), among others, argue that adjustment costs could prevent firms from rebalancing small deviations from their specified target debt ratios. 3

842

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

more debt. Interestingly, these factors systematically influence the debt-equity choice based on whether firms issue or retire capital rather than whether firms are under- or over-levered. While these results reinforce our main findings about the absence of systematic target-following behavior, they also suggest that financing choices are far from being randomly determined. Based on specific attributes, firms tend to carefully consider the choice of debt or equity, depending on whether they are issuing or retiring capital. These results suggest that future capital-structure research should focus more on identifying theories based on the firms' motives to choose debt or equity when they issue or retire capital rather than when they deviate from a plausible target debt ratio. Our paper mostly relates to Byoun (2008) and De Jong et al. (2011). Similar to these studies, we analyze under- and over-levered firms separately. Our results extend Byoun's (2008) findings as we focus on firms' financing choices, which are consistent with target-following, rather than reversion of the debt ratios per se. Therefore, when Byoun (2008) suggests a weak reversion for the over-levered (under-levered) firms with a financial deficit (surplus), we find an intense, non-target-following behavior for the majority of them. We also extend De Jong et al.'s (2011) analysis by estimating the intensity of the debt/equity issuances or retirements in addition to reporting the fraction of firms that are issuing or retiring debt or equity. Although we find an intense issuance or retirement of one security for a fraction of firms in each category, we also see an intense issuance of the opposite security for the remaining fraction of the firms. Importantly, both fractions tend to be quite significant for each of the four categories. Our paper contributes to the literature in several ways. First, we examine the target-following behavior by studying the firms' financing choices rather than the movement for the debt ratios per se. Second, because the target debt ratios are unobservable, we design our testing strategy such that precise estimation of the targets is not critical for our analysis. The targets are merely used to indicate the direction of deviation which can easily be inferred even when the targets are estimated with some noise. Third, our results are reliable, when compared to the past literature, to the extent that our testing strategy could successfully differentiate the target from random financing behavior. Fourth, by way of focusing on the firms' financing choices, our findings add further insight to those of Byoun (2008) and De Jong et al. (2011). Finally, our results highlight the importance of other systematic determinants of the firms' financing choices. The remainder of the paper is organized as follows. Section 2 discusses the major drawbacks of using movements in debt ratios to make inferences about the target behavior and explains the new testing strategy for the target-following behavior. Section 3 presents the data and empirical results. This section validates our testing strategy by using the dataset for simulated intentional target-following and random financing behavior. Section 4 explores the impact of the deviation from the target debt ratios and other factors on the firms' debt-equity choices. Section 5 provides the concluding remarks. We conduct several robustness tests to validate our main results. These tests and their findings are discussed in a separate supplementary data section, Appendix B (Supplementary Appendix XA to XE). 2. Misleading movements of debt ratios and a new testing strategy for the target behavior 2.1. Misleading movements of debt ratios and the mechanical mean reversion Prior literature raises several concerns about the inherent tendency of debt ratios to mean-revert due to mechanical reasons. (See, for example, Shyam-Sunder and Myers, 1999; Chen and Zhao, 2007; and Chang and Dasgupta, 2009.) We add analytical insight to these studies and argue that the debt ratios' movements often cannot capture the firms' conscious attempts to drive their debt ratios in the intended direction. A firm that attempts to increase (decrease) its book debt ratio,5 after realizing that it is under (over) levered with respect to its target, will either issue more debt (equity) or retire more equity (debt). In this case, the retained earnings' effect on the debt ratio remains exogenous unless the firm can fully revert back to its target debt ratio in subsequent time periods. In other words, irrespective of the retained earnings' magnitude and sign in the subsequent periods, issuing/retiring securities in this manner is the best the firm could do to revert back to its target. We show that the movement of debt ratios is misleading and often inconsistent with the net issuance of debt or equity required for the reversion. Let ADRi,t and ADRi,t+1 represent the debt ratio for firm i at time t and t + 1 and be measured as follows: ADRi;t ¼

Di;t Di;t þEi;t

ADRi;tþ1 ¼

Di;t þ di;tþ1 ; Di;t þ di;tþ1 þ Ei;t þ ei;tþ1 þ REi;tþ1

ð1Þ

ð2Þ

where Di,t is the total book value of interest-bearing debt (sum of the current portion of the long-term debt, the short-term debt, and the long-term debt), Ei,t is the total book value of equity (total interest of the firm's common shareholders) for firm i at time t; di,t+1 is the net debt issued (measured as the book-value change for the total debt over the previous year), 5 Following the arguments in Chang and Dasgupta (2009), we primarily use book-value debt ratios for studying mean reversion in this paper, however, later in the paper, we also verify our main results by using the market values of equity to estimate the target debt ratios.

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

843

e i,t+1 is the net equity issued (measured as the change in the difference between the total book value of the equity and retained earnings over the previous year),6 and RE i,t+1 is the net addition in retained earnings during year t + 1. From here on, we drop the subscript i for easier exposition. Let a firm plan to increase its debt ratio such that ADRtþ1 NADRt or Dt þ dtþ1 Dt N Dt þ dtþ1 þ Et þ etþ1 þ REtþ1 Dt þ Et

ð3Þ

Ignoring the effect of retained earnings, to achieve this goal, the firm needs to issue net debt such that dtþ1 Dt N ; or dtþ1 þ etþ1 Dt þ Et

ð4Þ

the firm needs to retire net equity such that etþ1 Et N dtþ1 þ etþ1 Dt þ Et

ð5Þ

Thus, the necessary and sufficient condition for the firm to increase its debt ratio is that the proportion of the net debt issued (equity retired) in the total net external financing for a given period should be greater than the previous period's debt (equity) ratio. Following Eq. (4), even if firms issue more debt than equity to deliberately increase their debt ratios, the firms would not be able to do so if the previous period's debt ratios are higher than the proportion of debt used with the external financing. For example, if a firm has financed 80% of its financing needs through debt, the next period's debt ratio could still decline if the previous period's debt ratio is greater than 0.80. Similarly, if a firm has funded 80% of its financing needs by using equity to not increase its debt ratio, the next period's debt ratio could still increase if the previous period's debt ratio is less than 0.20. Assuming exogenously determined external financing needs, issuing or retiring securities in this manner is the best that a firm could have done to drift its debt ratio in the intended direction. Another important concern is the differential sensitivity of changes in debt ratios due to net debt or equity issuances. This difference can be seen by estimating the percentage change in ADRs solely by net debt or equity issuances. Let X be the net debt or equity issued by the firm to make a transition from its ADR at time t. The next-period ADRs for the issuance of net debt and equity, respectively, can then be expressed as follows: ADRtþ1 ¼

Dt þ X Dt þ Et þ X

ð6Þ

ADRtþ1 ¼

Dt Dt þ Et þ X

ð7Þ

or

The percentage changes in the ADR are then, respectively, given by %ΔADR ¼

X:Et Dt  ðDt þ Et þ XÞ

ð8Þ

%ΔADR ¼

−X ðDt þ Et þ XÞ

ð9Þ

or

Eqs. (8) and (9) suggest that percentage changes in the debt ratios differ depending upon choosing debt or equity to make an intended movement. While percentage changes in ADRs that are caused solely by net-debt issuances are influenced by the initial debt-equity ratios, these changes are independent of the initial debt-equity ratios in the case of net-equity issuances. This result shows that the intensity of the movement of debt ratio toward a target depends on the current level of ADR and the type of security chosen to finance the deficit in the next period. For example, very high debt-ratio firms may find it extremely difficult to de-lever by retiring debt; issuing equity in this situation would make the transition easier. Similarly, very low debt-ratio firms 6 As a robustness check, we test this paper's core results by using the net debt and equity issuances or retirements captured through the firms' cash-flow statements as in Shyam-Sunder and Myers (1999). The results are qualitatively similar to those when using the measures described here for such issuances and retirements. However, having more available data points, we only report the results and carry out extended robustness checks using the measures described here.

844

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

may find it easier to lever themselves by issuing debt rather than retiring equity. Firms can make such choices if the external financing needs are endogenous. If the external financing needs are exogenous, even if firms issue or retire the required securities, the firms may not experience the desired drift intensity for their debt ratios. Therefore, corporate-financing transactions may produce debt ratios that significantly differ from their estimated targets (Hovakimian and Li, 2012). Nevertheless, irrespective of the external financing needs being endogenous or exogenous, and irrespective of the consequent movement direction for the firms' debt ratios, by making a conscious choice between debt and equity, the firms might be doing the best they can do to revert back to their targets. Using debt ratios to make inferences about any target behavior could also be misleading due to the role of retained earnings in the denominator of Eq. (3). The retained earnings' effect might inordinately augment or diminish the firms' conscious attempts to move their debt ratios in the intended direction. When including the effect of retained earnings, Eq. (3) suggests that positive (negative) retained earnings in the next period make it difficult for the firms to increase (decrease) their debt ratios. Depending upon the relative magnitude of the retained earnings and the external financing needs, the offsetting effect might be strong enough even when firms choose to satisfy their external financing needs solely with the debt or equity required for the intended movement in the debt ratios. Importantly, if the target behavior is only inferred from the firms' financing choices to make the intended drift of their debt ratios, the retained earnings effect may not distort the conclusion. 2.2. New testing strategy for the target behavior In order to accommodate adjustment costs, the prior literature has used partial-adjustment, dynamic panel models to test for target behavior. These models use debt ratios at time t (ADRt) to predict the future debt ratios at time t + k (ADRt+k) and to identify the speed of adjustment (SOA) toward a target by estimating the coefficient on ADRt.7 Although such models consider the firms' expectations and adjustment costs, the models prove to be less effective for testing the target behavior. Several studies have recently raised concerns due to the econometric bias generated by using a lagged, dependent variable as an explanatory variable. Although, some of these studies suggest alternative models to generate more reliable coefficients,8 the target behavior in these studies' is inherently inferred by the movements in ADRs only. Hovakimian and Li (2012) find that, while significant deviation from the target behavior assumed by partial-adjustment models is quite common, the SOAs identified in these models may only indicate the frequency of full rebalancing in the sample. Their findings, therefore, question the use of SOAs as an economically meaningful measure of the target adjustments. In a similar vein and suggesting further potential pitfalls with SOAs, Chang and Dasgupta (2009) show that SOAs obtained using a simulated dataset for random financing are very similar to SOAs estimated in the previous studies using partial-adjustment models. We propose a testing strategy that examines the target behavior only in terms of the net issuances of the firm's specified securities to satisfy the external financing needs following a deviation from the target debt ratios. A negative deviation indicates an under-levered firm while a positive deviation indicates an over-levered firm. Following the prior literature, we use the predicted values of debt ratios, which are estimated from several possible determinants of leverage, as the target debt ratios. Later, we show that our testing strategy is not sensitive to the different measures of target debt ratio. Next, similar to the classifications in Byoun (2008) and De Jong et al. (2011), we group firms into four categories based on whether a firm is under- or over-levered at t = 0 and whether the firm seeks the net issuance or retirement of capital at time t in the future: Category Category Category Category

1 2 3 4

(C1): (C2): (C3): (C4):

under-levered firms seeking the net issuance of capital; under-levered firms seeking the net retirement of capital; over-levered firms seeking the net issuance of capital; over-levered firms seeking the net retirement of capital.

The net issuance or retirement of capital, in other words total external financing, is the sum of the debt and equity that is issued and/or retired in a particular period, t. To analyze the target behavior of the firms in our dataset, we focus on conscious reversion of the firms' ADRs to their target debt ratios through the issuance or retirement of the specified security in each category. For example, we consider issuing debt more than equity as a conscious attempt at reversion for the target debt ratios by the firms in category 1 (C1). Following this idea, we specifically define the target behavior for each category at time t as follows: C1: firms issue debt (“right” security) more than equity (“wrong” security); C2: firms retire equity (“right” security) more than debt (“wrong” security); C3: firms issue equity (“right” security) more than debt (“wrong” security); 7 See, for example, Faulkender et al. (2012), Huang and Ritter (2009), Lemmon et al. (2008), Kayhan and Titman (2007), Flannery and Rangan (2006), and Hovakimian et al. (2001), among others, for models that estimate the SOAs using different econometric techniques. 8 See Flannery and Hankins (2013) for a comparison of the dynamic panel models in terms of their abilities to generate reliable coefficient estimates. Iliev and Welch (2010) suggest a non-parametric technique to evolve debt ratios randomly and then test for the target behavior under a null hypothesis of managerial neglect. Hovakimian and Li (2011) suggest utilizing the historical fixed-effect proxies for the target debt ratios and the exclusion of extreme debt ratios from the study sample. The authors further suggest that the partial-adjustment model should keep target and lagged debt ratios separate while the target behavior is studied with the coefficient of target debt ratio. DeAngelo and Roll (2015) attempt to infer the target behavior by simulating the instability of the cross-sectional debt ratios over a long period of time by using different model specifications and actual debt ratios. The authors conclude that firms may either have widely fluctuating time-varying targets or may follow wide target zones for rebalancing.

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

845

C4: firms retire debt (“right” security) more than equity (“wrong” security). Finally, we observe the proportionate net usage of the specified “right” security in the total external financing at time t to estimate the target-following's intensity. Note that, irrespective of whether the financing deficit is endogenous or exogenous and whether the retained earnings are positive or negative, the firms' financing choices, as specified above, are the best steps that the firms can take to make the intended reversion. In contrast, as discussed in the previous section, the intensity of target behavior stipulated through actual ADR movements, would depend on the endogeneity of the financing deficits and the effect of retained earnings. Therefore, our testing strategy, which focuses on the firms' financing choices, clearly has an advantage over assessing the target behavior that is stipulated through actual movement in debt ratios. We use our testing strategy to address concerns from the past literature in the following ways. First, we identify the firms' endogenous efforts to change the leverage toward the intended direction. We only focus on the firms' financing choices and not on the movement of ADRs, per se, to separate the target behavior from the mechanical mean reversion.9 Second, the firms' financing choices over a period of time should alleviate concerns about adjustment costs that may prevent firms from following the target behavior immediately. We, therefore, examine the firms' financing behavior over a period of time after they deviate from their targets at t = 0. Finally, we test the target behavior using simulated datasets with intentional target and random financing behavior. We show that our testing strategy can distinguish the target behavior from random financing. 3. Empirical analysis 3.1. Data For the empirical analysis, we use a dataset consisting of 17,848 U.S. non-financial, public firms (excluding utilities) for which annual financial-statement data are available in Compustat files for a period of 25 years between 1988 and 2013.10 When estimating the target debt ratios, we exclude firm-year observations with debt ratios (debt to total book capitalization) of less than zero or greater than one, and with missing values for any variable of interest. Further, we winsorize the independent, firm-specific variables at the 1st and the 99th percentile to remove the outliers' effect. After applying these filters, the final dataset consists of 82,363 firm-year observations for the variables used to estimate the target debt ratios. To study the financing choices, we scale the net debt and equity issuances as well as their sum (total external financing, ExFIN) in a given year with the total assets from the previous year; we exclude firm-year observations that are greater than or equal to unity in magnitude for these scaled variables to remove the outliers' effect. We also exclude firm-year observations with zero external financing.11 Because the first set of estimated target debt ratios is for 1989, our dataset to test the subsequent financing behavior starts in 1990. For example, to be included in the analysis, a firm might be under- or over-levered with respect to its target in 1994 (t = 0) and could have issued and/or retired net debt and equity in 1995 (t = 1). The final dataset consists of 76,194 firm-year observations in 4 categories for net debt and equity issuances. The descriptive statistics for all the variables are shown in Appendix A. 3.2. Estimating the target debt ratio Following prior studies that use partial-adjustment models, we define a firm's target debt ratio as a function of the firm-specific determinants of leverage that were identified in the literature.12 Specifically, a firm's target debt ratio for a given period, t, is estimated using the predicted values from Eq. (10)13: ADRi;t ¼ α þ β1  ADRi;t−1 þ

X

β j  Xi;t−1 þ ηi þ εi;t ;

ð10Þ

where Xi,t − 1 represents the lagged values of the following significant determinants of cross-sectional leverage identified by Frank and Goyal (2009) for firm i at time t − 1: SIZEi,t − 1 is the firm's size, measured as logarithm of the total assets deflated by the consumer price index; PRFi,t − 1 is the profitability, measured as the ratio of the operating income (earnings before interest, tax, and depreciation) to total assets; ATNi,t − 1 is the asset tangibility, measured as the ratio of the net fixed assets to the total assets; GRWTi,t − 1 is growth opportunities, measured as the ratio of market to book value of equity; MEDi,t − 1 is the book value of the median industry leverage, where industries are identified based on the four-digit Standard Industrial Classification (SIC) codes and where the median debt ratios are estimated for each industry every year; INFt − 1 is the inflation rate, measured as the annual percentage change in the consumer price index. 9 This method differentiates our study from Faulkender et al. (2012) who used a partial-adjustment model involving debt ratios but attempted to segregate the mechanical and active reversion by subtracting the passive-leverage components from the ADRs. 10 In order to avoid any potential effect on the firms' leverage due to the Tax Reform Act of 1986, we collect our data after the act's passage. 11 As a robustness check, we also exclude firm-year observations that involve large asset sales, mergers, and divestitures using Compustat identification for the same in the footnotes. Change in sample composition due to this has no effect on our results. 12 See Frank and Goyal (2009) for a review of studies that suggest the predicted debt ratios as a function of firm-specific and other institutional factors. 13 Although, we demonstrate mean reversion in this section using the functional form of Eq. (10) to estimate the target debt ratios; later, we show that our findings remain qualitatively similar when using alternative estimations for the target debt ratio.

846

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

In addition to these variables, Xi,t − 1 also includes the modified Altman Z score proposed by MacKie-Mason (1990) as a proxy for the distress costs or debt capacity, and research and development expenditures as a proxy for the future investments (Fama and French, 2002) and innovative firms (Titman and Wessels, 1988). ALTZi,t − 1 is the modified Altman Z score which is measured as 3.3 times operating income (EBIT) plus sales plus 1.4 times the retained earnings plus 1.2 times (current assets minus current liabilities) divided by the total assets; RNDi,t − 1 is the ratio of the research and development expenses to the total assets; D_RND is a dummy variable that is set equal to one for firms with missing values of RND i,t − 1 and zero otherwise; ηi represents the time-invariant, firm-specific attributes; and εi,t is the usual stochastic error term. For the rest of the paper, we drop the subscript i from all variables of interest. Following Hovakimian and Li (2011), we use the fixed-effects regression model for historical panel data to obtain the coefficient estimates in Eq. (10). Specifically, for any given year t, Eq. (10) is estimated using ADRt as the dependent variable observed from year 2 to year t and the lagged explanatory variables observed from year 1 to year t − 1. The target debt ratio for any firm in year t, then, is the predicted values of ADRt that are estimated using the identified coefficients and the values of explanatory variables at time t. The out-of-sample estimation is, thus, assumed to control for possible look-ahead bias that could surface if the full sample is used. With this procedure, we estimate the target debt ratios for each year from 1989 to 2013. Because the financing choices for firms with very low or very high debt ratios may be guided by concerns other than the target debt ratio,14 we exclude firms with extreme debt ratios, i.e., firm-year observations with ADRs less than 0.20 and greater than 0.80. We also conduct our analysis by using the full sample including these observations, finding that, although, financing choices of under-levered firms are influenced by these observations, the broad conclusion regarding the firms' target-following behavior remains the same. These results are reported in the Supplementary Appendix XA. We define time t = 0 as a transition point where a typical firm experiences a deviation of the actual debt ratio from its estimated target. We further define the deviation from the target as follows: ΔADRt ¼ ADRt −TARt ;

ð11Þ

where TARt is the target debt ratio that is estimated by using Eq. (10). Because target debt ratios are unobservable, identifying a precise deviation from the target is a challenge. An important utility of our testing strategy is that our analysis does not require knowledge about a precise target debt ratio. The target debt ratios are merely used to indicate the sign of the deviation (positive and negative) from the actual debt ratios. The target behavior is identified by observing the firms' financing choices that move their debt ratios in the intended direction, consistent with the target behavior. Thus, a noisy estimate of the target debt ratio can serve the purpose of testing the target behavior in this paper. To check the robustness of our findings and to address concerns related to unobservable targets, we use different proxies for the target debt ratios. The prior literature has used several proxies for target debt ratios, such as the average for the previous period's debt ratios (e.g., Jalilvand and Harris, 1984; Shyam-Sunder and Myers, 1999) and the median industry leverage (e.g., Leary and Roberts, 2010). Further, to address the non-linearity of the debt ratios, non-linear regression models are used to predict the target debt ratios.15 We conduct several robustness checks using these measures and find that our results are qualitatively similar. We also take a consensus estimate of all these proxies to test the robustness of our results. In addition, DeAngelo and Roll (2015) as well as Leary and Roberts (2005) suggest that firms may operate with a band around the target debt ratios and revert only if the band is breached. Therefore, even if a precise target debt ratio is estimated, it is unknown whether the deviation from the target is significant enough to call for a reversion. We undertake the testing for cases where the actual debt ratios differ from their targets by a significant amount to accommodate a range around the target debt ratios. These results are reported in the Supplementary Appendix XB. Because the target debt ratios are often estimated as debt ratios that are determined by firm-specific attributes (e.g., Flannery and Rangan, 2006; Kayhan and Titman, 2007; Hovakimian and Li, 2011), for our primary results in this section, we use the target debt ratios that are estimated using Eq. (10). 3.3. Intensity of the target behavior Using the framework presented in Section 2.2, we empirically examine the intensity of the firms' target-following in our dataset. Specifically, we estimate the fraction of the “right” security used for the total external financing by the firms in a given period. For an intense target behavior, we expect this fraction to be significantly high. We expect a certain pecking order for the financing that is led by significant net issuance or retirement of the “right” security for each category. Following the idea of an inherent pecking order for each category, we can use the basic Shyam-Sunder and Myers (1999) or SSM model16 to test the extent of the “right” security in the financing deficit for a given period as follows: Si;t ¼ α þ β  ExFINi;t þ εi;t ;

ð12Þ

14 Alti (2006), Chen and Zhao (2007), and Hovakimian and Li (2011), among others, find that the target-following behavior can be significantly influenced by the observations with extreme debt ratios. 15 Welch (2007) suggests that debt ratios are inherently non-linear and that the use of linear models to predict them may not be valid. Non-linear models, such as Tobit and Fractional regression models, are suggested for such predictions to overcome the inherent non-linearity in the debt ratios. 16 The basic SSM model is used to test the proportion of debt in the total external financing mix. A slope coefficient close to 1 (0) signifies the dominance of debt (equity) in the financing mix, as predicted by the pecking-order theory proposed by Myers (1984).

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

847

where Si,t refers to the net issuance of the “right” security that is specified for a given category of firm i in period t and ExFINi,t is the firm's total external financing for the same period. However, Chirinko and Singha (2000) raise serious concerns regarding the stability of Eq. (12)'s slope coefficient. Specifically, the coefficients may not correctly capture the proportion of financing through the “right” security versus the “wrong” security for a sample that includes significant proportions of both types. To address these concerns, we modify the basic testing strategy for the SSM model and separate firm-year observations into two groups: target-following and non-target-following. In each of the previously defined categories, a target-following group consists of firm-year observations with more “right” security issuances or retirements than “wrong” security issuances or retirements, whereas a non-target-following group consists of the remaining firm-year observations. For example, the target- (non-target) following group in category 1 (C1) is the set of firm-year observations with more debt (equity) issuances than equity (debt); for this target- (non-target) following group, we use the net debt (equity) issuances as the dependent variable in Eq. (12) to test the target behavior. Similarly, the target- (non-target) following group in C2 is the set of firm-year observations with more equity (debt) retirements than debt (equity) retirements. Accordingly, we use net equity (debt) retirements–the “right” (“wrong”) security–as the dependent variable in Eq. (12) for the target- (non-target) following groups in this category and so forth. We then run separate pooled regressions using Eq. (12) for both groups in each category. The slope coefficients for both groups provide the intensity of the “right” and “wrong” security issuances, respectively. We use two criteria to identify and confirm the target behavior: (i) the proportion of firm-year observations in the target group should outweigh the proportion in the non-target group; i.e., most firms conform to the target behavior; and (ii) the slope coefficients that reflect the intensity of the security issuances associated with the target (non-target) group should be very high (low); i.e., the intensity of the target (non-target) following should be high (low). Because we isolate the two groups based on the predominance of the “right” or “wrong” security issuances in each category, each group's slope coefficient would be greater than or equal to 0.50.17 Because we anticipate that firms will follow their target debt ratios and issue/retire securities accordingly, we expect firms in the target-following groups to extensively issue the “right” security while using the “wrong” security occasionally and to the minimum possible extent. This prediction would make the slope coefficients closer to one for target-following groups. On the other hand, even if firms for the non-target groups use the “wrong” security more than the “right” security, these firms may be constrained to do so because of some adjustment costs and frictions. In such cases, we can expect the firms to use a mix of the “right” and the “wrong” security, which could lead them to deploy the “right” security to the maximum extent possible. Therefore, we expect the slope coefficients for a non-target-following group to be higher than, but close to, 0.50. A slope coefficient that is closer to one for the non-target-following group indicates an intense issuance of the “wrong” securities against the spirit of the target behavior.

3.3.1. Single-period target-following In this section, we focus on the firms' financing choices during the period immediately following the deviation. We present the results from the modified SSM tests for the two groups (target-following and non-target-following firms) in each category separately. For each category, we also report the proportion of the firm-year observations that exhibit the target behavior as a fraction of the total number of observations. The results reported in Table 1 suggest that debt is the preferred means of raising or retiring capital for a majority of the firms, irrespective of whether they are under- or over-levered. Firms tend to depart from the target behavior more often when equity is the “right” security to issue or retire. Thus, financing choices for a majority of the C1 and C4 firms conform to the target-following behavior while the choices for C2 and C3 do not. Some of our results can be interpreted as supporting Myers' (1984) pecking-order hypothesis. However, contrary to this hypothesis and consistent with Frank and Goyal (2003), we find that a substantial number of firms also issue or retire equity. Furthermore, the intensity of the equity issuances and retirements, as reflected by the coefficient estimates, is quite high and most comparable with the debt issuances. Also given Leary and Roberts' (2010) findings, a relatively higher proportion of debt issued by a majority of the firms in a category is not a sufficient indicator of the pecking-order behavior. Even when firms follow a financing hierarchy, it may not be primarily due to the information asymmetry sought in Myers and Majluf's (1984) model of the pecking-order behavior (Graham and Harvey, 2001; Brounen et al., 2006). We also find high R-squareds for both regressions using the two groups in each of the four categories, indicating a higher confidence in the estimated slope coefficients. Importantly, we find that the proportion of firms that do not conform to the targetbehavior issue or that retire the “wrong” security quite intensely. Thus, for example, even when majority of category 1 (C1) firms issue debt or the “right” security, large and significant coefficient estimates for the remaining firms suggest intense equity -or the “wrong” security- issuance for them. This is not expected if adjustment costs are the primary concern for these firms to issue or retire these “wrong” securities. Accordingly, we may not conclude that C1 firms follow a systematic target behavior. In fact, the results suggest that, at the least, categories for which the predicted “right” security is equity (i.e. C2 and C3), firms collectively follow an intense, non-target behavior. For these categories we find higher proportions and coefficient estimates close to unity for the non-target-following groups, i.e. firms issuing or retiring debt. To validate Chirinko and Singha's (2000) concerns and to highlight the merits of our testing methodology, we estimate the coefficients using the basic SSM model described in Eq. (12) without isolating the firm-year observations into target and non-target-following groups for each category. Table 2 reports the results. Although our dependent variable of interest is the net issuances of the “right” security, we also report the results for the “wrong” security as the dependent variable. Because ExFIN sums 17

However, this scenario is strictly applicable only for a line passing through the origin, i.e., when the intercept in Eq. (12) is zero.

848

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

Table 1 Results from modified Shyam-Sunder and Myers (1999) model in Eq. (12) for target-following and non-target-following groups in each category. Dependent variable is the net issuance/retirement of the “right” (“wrong”) security for target (non-target) group in each category. Our sample includes firms with ADRs between 0.20 and 0.80. Section 2.2 presents details for each category. *** indicates significance of the coefficient at 1% level. ExFIN is the total external financing of a firm in a given year. Category Leverage w.r.t target

1

2

3

4

Under-levered

Under-levered

Over-levered

Over-levered

Net issuance

Net retirement

Net issuance

Net retirement

Issue debt

Retire equity

Issue equity

Retire debt

Net external financing Expected target behavior Group Dependent variable

Target Debt

Non-target Equity

Target Equity

Non-target Debt

Target Equity

Non-target Debt

Target Debt

Non-target Equity

Intercept ExFIN Adj. R2 N Fraction (target)

0.016*** 0.832*** 0.899 7297 63.85%

0.029*** 0.866*** 0.826 4132

−0.023*** 0.834*** 0.672 2091 23.84%

−0.017*** 0.877*** 0.781 6680

0.063*** 0.847*** 0.701 5410 39.04%

0.013*** 0.831*** 0.909 8448

−0.021*** 0.961*** 0.812 10,431 89.37%

−0.032*** 0.823*** 0.649 1241

Table 2 Target-following using basic Shyam-Sunder and Myers (1999) model in Eq. (12) for all the firms in each category without splitting the data into target and non-target groups. Dependent variable is the net issuance/retirement of the “right” and “wrong” security for each category. Section 2.2 presents details for each category. *** indicates significance of the coefficient at 1% level. ExFIN is the total external financing of a firm in a given year. Category Leverage w.r.t target Net external financing Expected target behavior

1

2

3

4

Under-levered

Under-levered

Over-levered

Over-levered

Net issuance

Net retirement

Net issuance

Net retirement

Issue debt

Retire equity

Issue equity

Retire debt

Security issued/retired Dependent variable

Right Debt

Wrong Equity

Right Equity

Wrong Debt

Right Equity

Wrong Debt

Right Debt

Wrong Equity

Intercept ExFIN Adj. R2 N

0.010*** 0.511*** 0.397 11,429

−0.010*** 0.488*** 0.376 11,429

0.016*** 0.406*** 0.208 8771

−0.016*** 0.593*** 0.36 8771

0.003*** 0.510*** 0.328 13,858

−0.003*** 0.489*** 0.31 13,858

−0.019*** 0.845*** 0.619 11,672

0.019*** 0.154*** 0.051 11,672

the issuance or the retirement of the “right” and the “wrong” security, the ExFIN coefficients for the “wrong” security as the dependent variable will simply be one minus the ExFIN coefficient for the “right” security and vice versa.18 However, the adjusted Rsquareds for the “right” and “wrong” security show the model's relative fit or the explanatory power of the estimated coefficients. The results from the basic model show stark differences compared to the results from our proposed testing strategy in Table 1. For example, the results for C3 in Table 2 show that net equity and debt issuances constitute an almost equal proportion (51% and 48.9%) of the total external financing. An almost equal R-squared of 32.8% and 31% further reinforces this finding. However, our previous results for C3 (Table 1) indicate that only 39% of the observations actually issue equity more than debt. Further, C3's coefficient estimates and R-squareds in Table 1 indicate a high intensity for debt issuances, suggesting an intense, non-target behavior. Similarly, although the C1 results in Table 2 suggest that firms tend to use equity and debt equally, our findings from Table 1 indicate that more than 63% of the firms tend to use debt predominantly over equity. Moreover, the results in Table 1 indicate that the remaining firms issue equity intensely, suggesting that these firms may have used equity solely to meet their external financing needs and not as a mix of debt and equity. These deeper insights into the firms' financing behavior is an important utility for our testing strategy. Interestingly, when the proportion of the “wrong” security in the sample is minimal, results for the basic model and our testing strategy converge. This finding is consistent with Chirinko and Singha's (2000) arguments. For firms in C4, the coefficients in Table 2 are consistent with the target behavior, suggesting that a significant majority of these firm-year observations (89.37%) retire debt more than equity. Consistent with Chirinko and Singha (2000), a significant proportion of observations with “wrong” security issuances influences the results of the basic SSM model in Table 2. The stark differences between the results for Tables 1 and 2 indicate that the basic SSM model may not properly capture the target-following behavior when target- and non-target-following firms are pooled together. Our testing strategy successfully addresses these challenges and produces reliable results. The findings obtained

18

However, this finding is exactly true only when the intercept for both regressions is zero.

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

849

Table 3 Financing decisions in multi-period financing. Dependent variable is the net issuance/retirement of the “right” (“wrong”) security for target (non-target) group. Section 2.2 presents details for each category. *** indicates significance of the coefficient at 1% level. ExFIN is the total external financing of a firm in a given year. Panel A to D presents results for subsequent round of financings at t = 2, 3, 4 and 5 respectively. Category Leverage w.r.t target

1

2

3

4

Under-levered

Under-levered

Over-levered

Over-levered

Net issuance

Net retirement

Net issuance

Net retirement

Issue debt

Retire equity

Issue equity

Retire debt

Net external financing Expected target behavior Group Dependent variable

Target Debt

Non-target Equity

Target Equity

Intercept ExFIN Adj. R2 N Fraction (target)

0.016*** 0.839*** 0.897 5260 66.00%

0.025*** 0.843*** 0.812 2710

Intercept ExFIN Adj. R2 N Fraction (target)

0.015*** 0.844*** 0.895 4531 66.92%

Intercept ExFIN Adj. R2 N Fraction (target)

Intercept ExFIN Adj. R2 N Fraction (target)

Non-target Debt

Target Equity

Non-target Debt

Target Debt

Non-target Equity

Panel A: t = 2 -0.024*** -0.018*** 0.829*** 0.854*** 0.605 0.776 1692 4816 26.00%

0.052*** 0.844*** 0.715 3502 35.19%

0.012*** 0.843*** 0.914 6450

-0.018*** 0.965*** 0.844 8246 89.07%

-0.032*** 0.798*** 0.617 1012

0.027*** 0.83*** 0.804 2240

Panel B: t = 3 -0.023*** -0.018*** 0.841*** 0.853*** 0.61 0.773 1515 4058 27.18%

0.052*** 0.857*** 0.703 2680 33.64%

0.011*** 0.853*** 0.913 5286

-0.018*** 0.961*** 0.851 6725 88.87%

-0.029*** 0.831*** 0.598 842

0.016*** 0.842*** 0.892 4010 67.75%

0.025*** 0.837*** 0.818 1909

Panel C: t = 4 -0.023*** -0.018*** 0.842*** 0.838*** 0.605 0.775 1414 3542 28.53%

0.048*** 0.869*** 0.714 2150 32.78%

0.011*** 0.859*** 0.918 4409

-0.017*** 0.965*** 0.846 5552 88.37%

-0.034*** 0.801*** 0.606 731

0.015*** 0.846*** 0.9 3630 68.72%

0.026*** 0.82*** 0.813 1652

Panel D: t = 5 -0.023*** -0.018*** 0.834*** 0.847*** 0.598 0.785 1291 3190 28.81%

0.048*** 0.857*** 0.704 1755 31.98%

0.011*** 0.858*** 0.919 3732

-0.017*** 0.957*** 0.85 4643 88.24%

-0.032*** 0.823*** 0.594 619

using our proposed strategy are reassuring because, once the influence of the “unwanted” observations is eliminated, we obtain sharper coefficients, signaling the relative strength of the issuance/retirement of the “right” and the “wrong” security. Furthermore, knowing the proportion of firms that follow the target behavior complements the results using the coefficients. Finally, it is important to note that the reversion in our results reflects the firms' voluntary attempts to issue or retire the security of their choice; such reversion is independent of any mechanical drift in the debt ratios.

3.3.2. Multi-period target-following Prior studies suggest that deviating from the targets can persist when firms are faced with adjustment costs. For example, Leary and Roberts (2005) show that firms tend to make capital-structure adjustments infrequently but respond to deviations caused by certain corporate actions, such as equity issuances, in order to rebalance their debt ratios within two-to-four years of such perturbations. We, therefore, conduct multi-period tests to incorporate adjustment costs into our methodology. Specifically, we observe a firm's financing behavior for the next five years after it deviates from its target debt ratios at t = 0. To accomplish this task, we re-categorize these firms at t = 2, 3, 4, and 5 by using their relative leverage with respect to the targets which are estimated using Eq. (10) at t = 1, 2, 3, and 4, respectively, and their latest external financing pursuits. We, then, test the target behavior for each year separately.19 The results presented in Panels A-D of Table 3 are qualitatively similar to the core results in Table 1. Contrary to Leary and Roberts' (2005) findings, the firms' financing behavior remains inconsistent with a possible target chase during a relatively long span of five consecutive years. We also examine the intensity of the target behavior for firms that remain under- or over-levered for a sustained time period of five years and then find that results are unchanged. Further, we examine the financing behavior of the firms within each category once they switch from one category (for example, C1 firms at t = 1 are re-categorized as C2, C3, or C4 at t = 2) to another in subsequent periods, finding similar, aggregate financing behavior even after such transitions. Finally, we check the impact of adjustment costs in the next round of financing by separating the firms that exhibited the target behavior during the previous period from those that exhibited the non-target behavior, finding that subsequent aggregate financing is unaffected by the 19

The results in Table 1 pertain to the choice of debt and equity at time t = 1.

850

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

Table 4 Financing behavior with a range of target debt ratios. Dependent variable is the net issuance/retirement of the “right” (“wrong”) security for target (non-target) group. Section 2.2 presents details for each category. *** indicates significance of the coefficient at 1% level. ExFIN is the total external financing of a firm in a given year. Panel A to C presents results for the dataset including firmyear observations only with magnitude of deviation more than 0.025, 0.05 and 0.10 respectively. Category Leverage w.r.t target

1

2

3

4

Under-levered

Under-levered

Over-levered

Over-levered

Net issuance

Net retirement

Net issuance

Net retirement

Issue debt

Retire equity

Issue equity

Retire debt

Net external financing Expected target behavior Group Dependent variable

Target Debt

Non-target Equity

Intercept ExFIN Adj. R2 N Fraction (target)

0.016*** 0.83*** 0.899 5469 64.91%

0.027*** 0.87*** 0.831 2956

Intercept ExFIN Adj. R2 N Fraction (target)

0.017*** 0.826*** 0.896 3692 65.17%

Intercept ExFIN Adj. R2 N Fraction (target)

0.018*** 0.81*** 0.893 1439 66.01%

Target Equity

Non-target Debt

Target Equity

Non-target Debt

Target Debt

Non-target Equity

Panel A: Deviations greater than 0.025 −0.023*** −0.018*** 0.068*** 0.835*** 0.853*** 0.835*** 0.665 0.76 0.686 1716 4806 4421 26.31% 39.45%

0.013*** 0.83*** 0.908 6785

−0.022*** 0.964*** 0.809 8633 90.41%

−0.035*** 0.831*** 0.643 916

0.025*** 0.862*** 0.845 1973

Panel B: Deviations greater than 0.05 −0.025*** −0.016*** 0.073*** 0.818*** 0.841*** 0.831*** 0.65 0.765 0.668 1242 3063 3625 28.85% 39.75%

0.014*** 0.828*** 0.909 5495

−0.022*** 0.971*** 0.814 7100 91.20%

−0.034*** 0.847*** 0.665 685

0.018*** 0.838*** 0.871 741

Panel C: Deviations greater than 0.10 −0.021*** −0.013*** 0.08*** 0.814*** 0.815*** 0.815*** 0.799 0.78 0.637 489 1012 2279 32.58% 40.05%

0.011*** 0.841*** 0.918 3411

−0.024*** 0.971*** 0.808 4587 92.31%

−0.034*** 0.849*** 0.683 382

previous target or non-target behavior and is largely inconsistent with the target-following. These investigations and their results are discussed in Supplementary Appendix XC. In summary, our results suggest that the firms' financing choices are not consistent with any systematic target-following behavior, at least when the “right” security to issue or retire is equity. Rather, a majority of the firms issue or retire debt irrespective of the prediction of target-following behavior to issue or retire a specific security. Multiple-period financing choices further corroborate this finding. The firms may systematically choose debt over equity for several reasons, such as the high transaction costs of issuing equity (Eckbo and Kisser, 2013), adverse signaling due to equity issuances (Masulis and Korwar, 1986), stricter regulations while issuing and retiring equity as compared to debt (Jenkinson, 1990), incentive conflicts (Leary and Roberts, 2010), or concerns for dilution (Young et al., 2008). However, over time, such frictions should fade, allowing firms to issue or retire the required securities consistent with the target behavior. Our results suggest that non-conformance with target-following behavior does not appear to be due to frictions or adjustment costs that prevent firms from issuing or retiring a security of their choice. This conclusion is consistent with the findings in Brounen et al. (2006) that transaction costs may not serve as a key driver of the corporate debt policy. Even after five consequent years of experiencing a deviation from the target debt ratio, firms' aggregate financing behavior does not seem to change much. This finding suggests that deviations from the target debt ratio may not be primarily driving the firms' financing choices. Rather, there may be other factors that influence the choices. We will formally test this possibility in Section 4.

3.3.3. Other adjustment-cost concerns Graham and Harvey (2001) as well as De Jong and Verwijmeren (2010) suggest that (i) managers are heterogeneous in their decision rules for their choice of a target debt ratio and (ii) managers who have a target are more likely to have a target range rather than a specific target ratio. Leary and Roberts (2005) show that firms actively rebalance their debt ratios to stay within an optimal range and suggest that firms may not rebalance unless they breach this optimal band around their target debt ratio. Operating within a range of target debt ratios is consistent with the notion of maintaining a credit rating along with the concern about a downgrade for breaching the firms' target ceilings (Graham and Harvey, 2001; Kisgen, 2006, 2009). Following these arguments, we test the firms' target behavior when the target band might have been breached. Specifically, we only include those firm-year observations where the difference between the actual and target debt ratios exceeds 0.025, 0.05, and 0.10 in magnitude. Because such large differences represent the maximum deviation for a significant number of firms in our sample, we believe that these deviations should be sufficient to warrant an adjustment toward the target debt ratios.20 20 Specifically, the minimum magnitude of 0.025 of the difference is close to the maximum deviation associated with the lowest quartile (0.028) of the firm-year observations when sorted according to the deviations' ascending order.

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

851

Table 5 Financing behavior including adjustment costs due to varying ex-ante probability of distress. Dependent variable is the net issuance/retirement of the “right” (“wrong”) security for target (non-target) group. Section 2.2 presents details for each category. *** indicates significance of the coefficient at 1% level. ExFIN is the total external financing of a firm in a given year. Panel A to D presents results for firms with negative, low, moderate and high Altman Z scores. Category Leverage w.r.t target

1

2

3

4

Under-levered

Under-levered

Over-levered

Over-levered

Net issuance

Net retirement

Net issuance

Net retirement

Issue debt

Retire equity

Issue equity

Retire debt

Net external financing Expected target behavior Group Dependent variable

Target Debt

Non-target Equity

Target Equity

Intercept ExFIN Adj. R2 N Fraction (target)

0.012*** 0.784*** 0.879 283 29.36%

0.069*** 0.855*** 0.806 681

Intercept ExFIN Adj. R2 N Fraction (target)

0.022*** 0.745*** 0.89 689 54.12%

Intercept ExFIN Adj. R2 N Fraction (target)

Intercept ExFIN Adj. R2 N Fraction (target)

Non-target Debt

Target Equity

Non-target Debt

Target Debt

Non-target Equity

Panel A: Negative Z scores −0.003 −0.052*** 0.915*** 0.841*** 0.919 0.65 53 303 14.89%

0.105*** 0.805*** 0.651 1497 69.37%

0.014*** 0.775*** 0.867 661

−0.064*** 0.909*** 0.723 933 91.65%

−0.055** 0.765*** 0.666 85

0.04*** 0.798*** 0.801 584

Panel B: Low Z scores −0.021*** −0.03*** 0.834*** 0.765*** 0.841 0.771 139 663 17.33%

0.068*** 0.803*** 0.654 897 44.27%

0.013*** 0.812*** 0.89 1129

−0.028*** 0.943*** 0.814 1389 90.49%

−0.029*** 0.746*** 0.605 146

0.014*** 0.849*** 0.904 4895 68.04%

0.024*** 0.848*** 0.791 2299

Panel C: Moderate Z scores −0.026*** −0.014*** 0.781*** 0.898*** 0.566 0.797 1421 4407 24.38%

0.047*** 0.827*** 0.706 2356 32.44%

0.012*** 0.847*** 0.921 4906

−0.017*** 0.956*** 0.824 6348 89.60%

−0.032*** 0.792*** 0.53 737

0.013*** 0.88*** 0.901 1136 73.01%

0.025*** 0.842*** 0.859 420

Panel D: High Z scores −0.024*** −0.014*** 0.878*** 0.877*** 0.561 0.817 409 1067 27.71%

0.051*** 0.894*** 0.705 378 25.61%

0.014*** 0.862*** 0.914 1098

−0.014*** 0.961*** 0.872 1234 86.11%

−0.036*** 0.936*** 0.711 199

The results shown in Panels A-C of Table 4 are very similar to our findings in Table 1. While the firms' financing behavior is not consistent with a target chase, their behavior may not be significantly affected by the magnitude of the debt ratios' deviation from their targets unless the firms follow a very wide range of target debt ratios where a specific leverage ratio is only of second-order importance (DeAngelo and Roll, 2015). We also test whether the target behavior is influenced by the ex-ante probability of the firms' distress because firms with a higher ex-ante probability of distress or with a lower debt capacity may have restrictions when making proper adjustments to their capital structure. We sort firms according to their modified Altman's Z-score, as proposed by MacKie-Mason (1990), and separate them into four groups: “negative,” which includes firms with negative Z-score values; “low,” which includes firms with Zscore between zero and one; “moderate,” which includes firms with a Z-score between one and three; and “high,” which includes firms with Z-scores that are greater than three. We then test each group's target behavior separately. The results shown in Panels A-D of Table 5 indicate that the intensity of target behavior varies for all four groups suggesting that the ex-ante distress cost may influence the firms' financing choices. However, more than three-fourths of the observations in our sample have a moderate or high Z-score, for which the results are very similar to our findings in Table 1. Consistent with our core findings, these firms with a low distress cost (moderate and high Z-scores in Table 5) seem to issue or retire debt irrespective of the target-following behavior's prediction. Although we find that over-levered firms with a very high distress cost (negative Zscores) follow the target behavior and de-lever themselves by using the “right” security as expected, this group includes fewer observations compared to firms with relatively high Z-scores. The results in Table 5 suggest that, as the distress costs decrease, debt issuances increase steadily and significantly for firms with positive financing deficits (C1 and C3). Although equity repurchases increase with Z scores for firms with financing surpluses (C2 and C4), the increase is much less significant than the increased debt issuances for firms with positive financing deficits (C1 and C3). Thus, more under-levered and fewer over-levered firms exhibit the target behavior with the decreased ex-ante probability of distress. Assuming that the ex-ante probability of distress correlates with the financial constraints and the debt capacity, we compare our results (Table 5) to De Jong et al.'s (2011) findings. Although the proportion of firms issuing debt is similar to the proportion reported in De Jong et al. (2011), different from their study, we find significantly more firms retiring their debt when they have a low ex-ante probability of distress or a high debt capacity. This finding could be due to using a smaller sample of firms with available senior debt investment grade ratings in De Jong et al. (2011). In addition, our results are more informative about the target

852

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

Table 6 Simulated datasets for intentional target and random financing behavior. Dependent variable is the net issuance/retirement of the “right” (“wrong”) security for target (non-target) group in each category. Section 2.2 presents details for each category. *** indicates significance of the coefficient at 1% level. ExFIN is the total external financing of a firm in a given year. Category Leverage w.r.t target

1

2

3

4

Under-levered

Under-levered

Over-levered

Over-levered

Net issuance

Net retirement

Net issuance

Net retirement

Issue debt

Retire equity

Issue equity

Retire debt

Net external financing Expected target behavior Group Dependent variable

Target Debt

Non-target Equity

Intercept ExFIN Adj. R2 N Fraction (target)

0.000 0.847*** 0.99 13,567 76.43%

0.000 0.645*** 0.99 4183

Intercept ExFIN Adj. R2 N Fraction (target)

0.000 0.746*** 0.94 9014 50.78%

0.000 0.754*** 0.94 8738

Target Equity

Non-target Debt

Target Equity

Non-target Debt

Target Debt

Non-target Equity

Panel A: Intentional target behavior 0.000 0.000 0.847*** 0.641*** 0.99 0.99 8021 2465 76.49%

0.000 0.849*** 0.99 7170 76.35%

0.000 0.653*** 0.99 2221

0.000 0.847*** 0.99 6199 76.50%

0.000 0.647*** 0.99 1904

Panel B: Random financing behavior −0.002 0.001 0.711*** 0.772*** 0.94 0.95 5172 5313 49.33%

0.001 0.738*** 0.93 4777 50.87%

0.000 0.737*** 0.94 4613

0.000 0.738*** 0.94 3994 49.29%

−0.000 0.745*** 0.94 4109

behavior. We find not only the proportion of firms that issue/retire debt or equity consistent with the target behavior, but also the respective intensity of issuances or retirements. Therefore, we find intense, non-target behavior for relatively unconstrained firms that are supposed to use equity as a rebalancing vehicle to follow the target behavior.21 3.3.4. Simulated target and random financing behavior Chang and Dasgupta (2009) argue that a successful testing strategy for the target behavior should reject possible alternatives. Following their study, we test the target behavior using simulated datasets for the intentional target behavior and random financing. The coefficient of ExFIN in these tests is expected to be close to 1 (0.50) for the “right” (“wrong”) security for the intentional target behavior and closer to 0.75 for both security types in the case of perfectly symmetrical, random financing choices for debt and equity. Consistent results suggest that, by using simulated datasets, we can calibrate the model to check the relative importance of the “right” versus the “wrong” security while inferring target behavior in the real dataset. To simulate the dataset for target behavior, we define an aggressive target behavior as follows. We assign a probability of 0.75 for a firm-year observation to be financed to a greater extent by the “right” security, where the net issuance of the “right” security is normally distributed with a mean of 85% and a standard deviation of 5% of the actual external financing, and the remaining external-financing requirement (with a consequent mean of 15%) is financed by the net issuance of the “wrong” security.22 We then assign a probability of 0.25 for a firm-year observation to be financed, to a greater extent, by the “wrong” security where the net issuance of the “right” security is normally distributed with a mean of 35% and a standard deviation of 5% of the total external financing; the remaining external-financing requirement (with a consequent mean of 65%) is financed by the net issuance of the “wrong” security. Thus, the simulated datasets for all four firm categories represent (i) the extensive use of the “right” security in proportions (75% in this case) and (ii) a higher intensity (fraction of the total external financing closer to unity) for the “right” security issuances and a lower intensity (fraction of the total external financing closer to 0.50) for the “wrong” security issuances. This is a close approximation to a target-following behavior when firms voluntarily choose to issue or retire the “right” security usually while choosing the “wrong” security only occasionally due to some adjustment costs and frictions. The results from the simulated dataset are reported in Panel A of Table 6. The proportion of firms that exhibit the target behavior and the coefficients of the financing deficit for the “right” and the “wrong” security in the target and non-target groups, respectively, is consistent with the target behavior. For example, 76.4% of the C1 firms exhibit the target behavior such that they issue more debt than equity. Also, for firms in the target (non-target) following group, debt (equity) issuances finance 84.7% (64.5%) of the total external financing, as reflected by the slope coefficients. Further, a high R-squared suggests a closer fit of the regression line with the data. In effect, the results exactly replicate the design of simulated datasets for the intentional target behavior. Next, we test whether our model can effectively respond to the firms' random financing choices. To simulate the dataset for random financing choices, we uniformly assume that the “right” security issuance for any firm-year observation is a fraction of 21 We also examine whether the size of the firms' financing decisions affects our results. We remove firm-year observations with total external financing that is less than 5% or more than 75% of the total assets for the previous period and test for target behavior. We find that the size of external financing does not materially influence our results. These results are discussed in Supplementary Appendix XD. 22 We run several tests with different probabilities and proportions to confirm our findings. The results are very similar; therefore, we only report one scenario here. The additional results are available, upon request, from the authors.

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

853

Table 7 Relative significance of deviation from target debt ratios to determine debt-equity choices. Dependent variable is a dummy variable that equals one if absolute magnitude of debt is greater than equity and zero otherwise. Section 2.2 presents details for each category. ΔADR is the deviation of actual debt ratio from their targets; ADR is the actual debt ratio; SIZE is the size of the firm, measured as logarithm of total assets deflated by consumer price index; PRF is the profitability, measured as the ratio of operating income (earnings before interest, tax, and depreciation) to total assets; ATN is the asset tangibility, measured as the ratio of net fixed assets to total assets; GRWT is growth opportunities, measured as the ratio of market to book value of equity; MED is the book value of median industry leverage, where industries are identified based on Standard Industrial Classification (SIC) codes and median debt ratios are estimated for each industry every year; INF is the inflation rate, measured as annual percentage change in the consumer price index; ALTZ is the modified Altman Z score measured as 3.3 times EBIT plus sales plus 1.4 times retained earnings plus 1.2 times (current assets minus current liabilities) divided by total assets; ExFIN is the total external financing of a firm in a given year; RND is the ratio of research and development expenses to total assets; D_RND is a dummy variable set equal to 1 for firms with missing values of RND and zero otherwise. *** indicates significance of the coefficient at 1% level. Dependent variable DEit (0,1) Category 1 (C1)

ΔADR ADR SIZE PRF ATN GRWT MED INF ALTZ ExFIN RND D_RND Constant N Pseudo R2

Category 2 (C2)

Category 3 (C3)

Category 4 (C4)

Coeff.

z

Economic impact

Coeff.

z

Economic impact

Coeff.

z

Economic impact

Coeff.

z

Economic impact

−0.739 0.832 0.032 0.381 0.121 −0.050 0.022 8.724 0.110 0.488 −1.793 0.018 −0.584 10,987 0.0728

−1.590 3.440 2.970 2.170 1.870 −6.470 0.110 4.970 7.180 6.590 −6.680 0.550 −6.820

−0.034 0.076 0.072 0.058 0.030 −0.108 0.002 0.086 0.215 0.096 −0.191

−2.073 3.184 −0.191 −1.509 −0.263 −0.099 −0.291 2.409 0.022 1.100 0.186 −0.027 1.267 8462 0.1113

−3.130 9.190 −12.170 −5.080 −3.020 −9.030 −1.110 1.050 1.200 5.660 0.390 −0.640 11.400

−0.093 0.302 −0.421 −0.184 −0.059 −0.180 −0.027 0.024 0.034 0.096 0.015

−0.327 0.448 0.021 0.572 0.222 −0.037 0.860 −4.078 0.074 0.204 −0.006 −0.213 −0.079 12,922 0.0812

−1.150 2.600 2.480 4.790 3.870 −9.570 5.740 −2.290 8.900 3.540 −0.170 −8.300 −1.030

−0.027 0.061 0.045 0.096 0.052 −0.174 0.083 −0.033 0.200 0.045 −0.002

0.072 1.590 −0.136 −0.827 −0.050 −0.061 0.617 9.878 0.029 0.099 1.038 0.062 0.829 11,071 0.1059

0.160 6.110 −10.760 −3.810 −0.570 −13.270 2.700 3.510 2.410 0.600 2.780 1.500 7.110

0.006 0.214 −0.294 −0.109 −0.011 −0.220 0.060 0.085 0.057 0.012 0.094

the external financing that is uniformly distributed between zero and one; the difference for the total external financing and the “right” security issuance is then financed with the “wrong” security issuance. Panel B in Table 6 presents the results from this simulated dataset. The firms are equally distributed in each group, and the slopes are moderate with values of around 0.75 (neither close to unity nor close to 0.50), as expected for each category. Overall, our findings from the simulated datasets indicate that the model has sufficient power to detect the alternative financing behavior compared to the firms' intended target behavior. 4. Determinants of the debt-equity choice Our findings for the firms' unaltered aggregate financing behavior are largely inconsistent with the target-following for an extended period of time and also for varying specifications of the target debt ratios. These findings suggest that the firms' financing choices may not be significantly related to deviation from the targets. We, therefore, investigate whether deviations from the target debt ratios significantly influence the firms' finance choices. If that is the case, the firms' marginal capital-structure decisions would be consistent with the target behavior, even when aggregate financing behavior does not reflect target-following. Specifically, larger magnitude of deviations should motivate firms to issue or retire significant amount of the predicted security in each category so as to effectively close the gap between actual and target debt ratios. We study the marginal impact of deviation from the targets, among several other possible factors, on the firms' debt-equity choices by using the following probit model: DEi;t ¼ α þ βa  ΔADRi;t−1 þ βb  ADRi;t−1 þ βc  ExFINi;t þ

X

β j  Xi;t−1 þ εi;t ;

ð13Þ

where DEi,t represents the dummy variable for firm i's debt-equity choice in period t which is equal to one if absolute magnitude of debt issued or retired is greater than the absolute magnitude of equity issued or retired and zero otherwise; ΔADR is our variable of interest representing deviation of debt ratios from their targets, ADR is the level of debt ratios, ExFIN is the total external financing and, Xi,t − 1 represents the lagged values for the significant determinants of the cross-sectional leverage that were identified in the literature and used for the estimation of target debt ratios in Section 3: i.e., size, profitability, asset tangibility, growth option, median industry leverage, Altman Z score, research and development expenses, a dummy variable for missing research and development (R&D) expenses, and the inflation rate. Table 7 presents the results for all four categories using our main dataset which includes the relative economic significance for each variable of interest.23 The results suggest that, contrary to the target behavior, deviation in debt ratios is statistically

23 The relative economic impact is estimated as the change in the function describing probability of issuing or retiring debt over equity in response to a one-standarddeviation change for the respective variable.

854

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

insignificant in explaining debt-equity choices, except for firms in the C2 category. Even though a large proportion of the C1 and C4 firms seem to issue/retire securities consistent with the target-following (Table 1), debt ratios that deviate from their targets do not significantly influence the debt-equity choice for firms in these categories. Further, even in C2 where the deviation seems to be statistically significant, a negative coefficient signifies that, for a larger magnitude of negative deviations, firms tend to retire more debt than equity which is inconsistent with the target-following.24 Consequently, deviation of debt ratios from their targets does not generally appear to be the first-order economic determinant of the debt-equity choice for any category. For example, a one-standard-deviation change in the deviation of debt ratios from their targets only alters the function describing probability of debt issuance for firms in the C1 category by 3.43%, which is economically less significant than most other variables. Rather, other firm-specific characteristics seem to be much more influential when determining the firms' debt-equity choice. We also verify if deviation from the target remain a significant determinant of the firms' debt-equity choices with significantly larger magnitudes of deviation. If firms only rebalance their debt ratios when the deviation is significant, we can expect the financing choices to be influenced by these deviations. However, when we only consider firms with significantly larger-magnitude deviations, the results remain qualitatively similar to the ones reported in Table 7.25 These results are reported in Supplementary Appendix XE. These results are consistent with the findings of DeAngelo and Roll (2015, pp. 410) that “leverage varies so widely at so many firms that it becomes hard to believe in large benefits from a particular level. It seems more plausible that, over a fairly wide range, leverage per se is of second-order importance for firm valuation, so the main determinants of leverage are factors other than the benefits of adhering closely to a particular debt-to-equity mix.” The results in Table 7 suggest that, along with the size, profitability, and growth opportunities, the level of debt ratios has a greater influence on the firms' financing choices when they are seeking to retire capital (C2 and C4). Larger, profitable, highgrowth and less levered firms tend to retire more equity than debt. For firms that are seeking to issue capital (C1 and C3), growth opportunities and the ex-ante cost of distress, as measured with the Altman Z score, have the greatest influence on the firms' capital structure decisions. While high-growth firms issue more equity, firms with very low ex-ante distress costs (high Z scores) tend to issue more debt. Under-levered firms with higher R&D expenses tend to issue more equity. Consistent with Frank and Goyal (2015), we find that more profitable firms issue debt and retire equity; however, we cannot conclude that they do so to offset deviations from the target debt ratios. Hovakimian et al. (2001) find that deviation from debt ratios are more important for repurchase decisions than for issuance decisions. Our results are partially consistent with Hovakimian et al. (2001) because we do not find deviations from the debt ratios to be an important determinant for the over-levered (C4) firms seeking to retire capital. Interestingly, the marginal impact of firm-specific characteristics is quite different for all four categories. For example, while the asset size influences the debt-equity choice when firms face a financing surplus, the asset size appears as a much less economically important determinant for firms facing a financing deficit. The impact of most factors is such that they systematically influence the firms' financing choices based on whether the firms issue or retire capital rather than based on whether the firms are relatively under- or over-levered. Consequently, the economic impact of these variables is very similar when the financing deficit is either positive or negative, irrespective of whether the firms are relatively under- or over-levered. While these results support our findings about the absence of systematic target-following, they also rule out any possibility that financing choices are randomly determined. Based on specific attributes, firms tend to carefully consider the choice of debt or equity, depending on whether they are issuing or retiring capital. These results, therefore, suggest that future capital structure research should focus more on identifying theories based on the firms' motives to make specific financing choices when they issue or retire capital rather than when they deviate from a plausible target debt ratio. 5. Conclusions The capital-structure literature has yet to explain whether firms follow a mean-reversion process toward specified, target debt ratios when making financing choices. Recent studies have raised serious concerns about the partial-adjustment dynamic panel models regarding the econometric biases introduced by lagged debt ratios when testing the target-following behavior. Furthermore, the mechanical mean reversion induced by the bounded nature of actual debt ratios (ADRs) could impair the ability of these tests to distinguish between the firms' target-following behavior and their random financing choices. ADR movements can also be misleading when interpreting a firm's conscious attempts to chase an intended, target debt ratio, as revealed by its financing choices. We address these concerns by developing a new testing strategy to study the target-following behavior. Specifically, we examine the firms' conscious attempts to issue or retire the stipulated security during a given period to steer their debt ratios toward the specified targets. We use the relative proportions of the desired security in the total external financing to determine whether

24 A negative coefficient for the under-levered firms (negative deviation) in C2 means that equity is preferred over debt when the magnitude of deviation is decreasing. 25 Specifically, we remove firm-year observations in the lowest quartile when the data are arranged in ascending order by the magnitude of the deviations and rerun our test using Eq. (13).

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

855

a firm follows the target behavior. This way, our methodology avoids to make any direct inference about the target behavior based on actual movement in debt ratios, and separates the effect of mechanical mean reversion from the firms' endogenous attempts to mean-revert. An important utility of our testing strategy is that precise knowledge about the target debt ratios is not important; the target behavior can be inferred if we know whether a firm is generally under- or over-levered without knowing the precise extent of that under- or over-leverage. Our strategy can be used to test the target behavior even if the firms follow a range of targets and not a specific, target debt ratio. We find that firms' financing choices are not generally consistent with a systematic target behavior. Firms tend to issue or retire more debt than equity in a given period, irrespective of the prediction for issuing or retiring a specified security consistent with the target-following. We also find that the aggregate financing choices remain unaltered for several subsequent rounds of financing, which is not consistent with the notion that adjustment costs may prevent an immediate reversion and, therefore, firms would gradually move toward their targets. Although our findings may seem to lend some support to Myers' (1984) pecking-order hypothesis, consistent with Frank and Goyal (2003), we find that both the proportion of equity-issuing firms and the intensity with which they issue equity are quite substantial. Irrespective of any target-following concerns, results which are similar to those found with our sample of firms that have moderate debt ratios might be obtained if these firms face harder and more persistent constraints to issue or retire equity as compared to debt. Whether this is the case and why it should be are open questions for future research. We show that our testing strategy has sufficient power to reject alternative-financing behavior that is inconsistent with targetfollowing by using simulated datasets of intentional target behavior and random financing choices. Our findings remain similar when we check their robustness with different specifications for the target debt ratios; adjustment-cost concerns, including accommodation for a range instead of specific target debt ratios and the ex-ante probability of distress or debt capacity; and the undue impact of very small and large external-financing magnitudes. While our results indicate that debt is a preferred mode of financing for a majority of the firms, the firms' relatively unchanged behavior in several subsequent rounds of financing also suggests that financing choices may not be driven by deviations from the target debt ratios. We formally test whether deviations from target debt ratios have significant impact on the firm's debt-equity choices after controlling for other firm-specific determinants. Reinforcing our earlier results that there is no systematic target-following, we find that firms' financing choices are not primarily driven by concerns for bridging the deviation from the target debt ratios. Other factors, such as size, profitability, and the firms' growth potential, have greater influence on the debt-equity choice. Moreover, the marginal impact of these variables differs across firms, depending on whether the firms face negative or positive deficits in the next financing round rather than whether they are under- or over-levered. This result implies that, while the financing behavior is far from random, future research should focus on the firms' financing choice based on whether they issue or retire capital rather than based on whether they deviate from a specified target debt ratio.

Funding sources This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Acknowledgements For helpful comments we are grateful to Ashok Korwar, Sudipto Dasgupta and Jeffry Netter, the editor, and especially an anonymous JCF referee. We also thank Bonnie Cooper and Heidi Mann for editorial assistance.

Appendix A. Descriptive statistics

Variable

Notation

N

Mean

Std. deviation

Q1

Median

Q3

Size Profitability Asset tangibility Growth opportunities Median industry leverage Actual debt ratio Actual debt ratio (market value) Inflation R&D expense Altman Z Net debt issuances Net equity issuances Total external financing

SIZE PRF ATN GRWT MED ADR MADR INF RND ALTZ dt et ExFIN

82,363 82,363 82,363 82,363 82,363 82,363 77,646 82,363 82,363 82,363 76,194 76,194 76,194

4.911 0.021 0.268 3.544 0.209 0.276 0.272 0.03 0.059 1.357 0.018 0.058 0.076

2.231 0.188 0.229 6.985 0.092 0.253 0.247 0.007 0.113 2.526 0.172 0.169 0.232

3.278 −0.013 0.089 1.120 0.147 0.023 0.068 0.025 0.000 0.874 −0.049 0.000 −0.041

4.804 0.063 0.199 1.925 0.209 0.238 0.203 0.028 0.001 1.843 −0.002 0.009 0.021

6.422 0.117 0.385 3.384 0.276 0.453 0.418 0.033 0.067 2.650 0.062 0.054 0.139

856

G.S. Chauhan, F. Huseynov / Journal of Corporate Finance 48 (2018) 840–856

Appendix B. Supplementary data Supplementary appendices mentioned in this article can be found online at doi:10.1016/j.jcorpfin.2016.10.013. References Alti, A., 2006. How persistent is the impact of market timing on capital structure? J. Financ. 61, 1681–1710. Brounen, D., De Jong, A., Koedijk, K., 2006. Capital structure policies in Europe: survey evidence. J. Bank. Financ. 30, 1409–1442. Byoun, S., 2008. How and when do firms adjust their capital structures toward targets? J. Financ. 63, 3069–3096. Chang, Dasgupta, S., 2009. Target behavior and financing: how conclusive is the evidence? J. Financ. 64, 1767–1796. Chen, L., Zhao, X., 2007. Mechanical mean reversion of leverage ratios. Econ. Lett. 95, 223–229. Chirinko, R.S., Singha, A.R., 2000. Testing static tradeoff against pecking order models of capital structure: a critical comment. J. Financ. Econ. 58, 417–425. DeAngelo, H., Roll, R., 2015. How stable are corporate capital structures? J. Financ. 70, 373–418. De Jong, A., Verbeek, M., Verwijmeren, P., 2011. Firms' debt–equity decisions when the static tradeoff theory and the pecking order theory disagree. J. Bank. Financ. 35, 1303–1314. De Jong, A., Verwijmeren, P., 2010. To have a target debt ratio or not: what difference does it make? Appl. Financ. Econ. 20, 219–226. Eckbo, B.E., Kisser, M., 2013. Corporate Funding: Who Finances Externally? Unpublished working paper. Tuck School of Business at Darmouth College Fama, E.F., French, K.R., 2002. Testing trade-off and pecking order predictions about dividends and debt. Rev. Financ. Stud. 15, 1–33. Faulkender, M., Flannery, M.J., Hankins, K.W., Smith, J.M., 2012. Cash flows and leverage adjustments. J. Financ. Econ. 103, 632–646. Fischer, E.O., Robert Heinkel, R., Zechner, J., 1989. Dynamic capital structure choice: theory and tests. J. Financ. 44, 19–40. Flannery, M.J., Hankins, J.W., 2013. Estimating dynamic panel models in corporate finance. J. Corp. Finan. 19, 1–19. Flannery, M.J., Rangan, K.P., 2006. Partial adjustment toward target capital structures. J. Financ. Econ. 79, 469–506. Frank, M.Z., Goyal, V.K., 2003. Testing the pecking order theory of capital structure. J. Financ. Econ. 67, 217–248. Frank, M.Z., Goyal, V.K., 2009. Capital structure decisions: which factors are reliably important. Financ. Manag. 38, 1–37. Frank, M.Z., Goyal, V.K., 2015. The profits–leverage puzzle revisited. Eur. Finan. Rev. 19, 1415–1493. Graham, J.R., Harvey, C.R., 2001. The theory and practice of corporate finance: evidence from the field. J. Financ. Econ. 60, 187–243. Hennessy, C.A., Whited, T.M., 2005. Debt dynamics. J. Financ. 60, 1129–1165. Hovakimian, A., Li, G., 2012. Is the partial adjustment model a useful tool for capital structure research? Eur. Finan. Rev. 16, 733–754. Hovakimian, A., Li, G., 2011. In search of conclusive evidence: how to test for adjustment to target capital structure. J. Corp. Finan. 17, 33–44. Hovakimian, A., Opler, T., Titman, S., 2001. The debt-equity choice. J. Financ. Quant. Anal. 36, 1–24. Huang, R., Ritter, J.R., 2009. Testing theories of capital structure and estimating the speed of adjustment. J. Financ. Quant. Anal. 44, 237–271. Iliev, P., Welch, I., 2010. Reconciling Estimates of the Speed of Adjustment of Leverage Ratios. Unpublished working paper. Pennsylvania State University and Brown University Available at SSRN 1542691. Jalilvand, A., Harris, R.S., 1984. Corporate behavior in adjusting to capital structure and dividend targets: an econometric study. J. Financ. 39, 127–145. Jenkinson, T.J., 1990. Initial public offerings in the United Kingdom, the United States, and Japan. J. Jpn. Int. Econ. 4, 428–449. Kayhan, A., Titman, S., 2007. Firms' histories and their capital structures. J. Financ. Econ. 83, 1–32. Kisgen, D., 2006. Credit ratings and capital structure. J. Financ. 61, 1035–1072. Kisgen, D., 2009. Do firms target credit ratings or leverage levels? J. Financ. Quant. Anal. 44, 1323–1344. Leary, M.T., Roberts, M.R., 2005. Do firms rebalance their capital structures? J. Financ. 60, 2575–2619. Leary, M.T., Roberts, M.R., 2010. The pecking order, debt capacity, and information asymmetry. J. Financ. Econ. 95, 332–355. Lemmon, M.L., Roberts, M.R., Zender, J.F., 2008. Back to the beginning: persistence and the cross-section of corporate capital structure. J. Financ. 63, 1575–1608. MacKie‐Mason, J.K., 1990. Do taxes affect corporate financing decisions? J. Financ. 45 (5), 1471–1493. Masulis, R.W., Korwar, A.N., 1986. Seasoned equity offerings: an empirical investigation. J. Financ. Econ. 15, 91–118. Myers, S.C., 1984. The Capital Structure Puzzle. J. Financ. 39, 574–592. Myers, S.C., Majluf, N.S., 1984. Corporate financing and investment decisions when firms have information that investors do not have. J. Financ. Econ. 13, 187–221. Ramalho, J.S., Vidigal da Silva, J., 2013. Functional form issues in the regression analysis of financial leverage ratios. Empir. Econ. 44, 799–831. Shyam-Sunder, L., Myers, S.C., 1999. Testing static trade-off against pecking order models of capital structure. J. Financ. Econ. 51, 219–244. Strebulaev, I.A., 2007. Do tests of capital structure theory mean what they say? J. Financ. 62, 1747–1787. Titman, S., Wessels, R., 1988. The determinants of capital structure choice. J. Financ. 43, 1–19. Welch, I., 2007. Common flaws in empirical capital structure research. AFA 2008 New Orleans Meetings Paper, pp. 1–33 Available at SSRN 931675. Young, M.N., Mike, W., Peng, M.W., Ahlstrom, D., Bruton, G.D., Jiang, Y., 2008. Corporate governance in emerging economies: a review of the principal–principal perspective. J. Manag. Stud. 45, 196–220.