Finding market timing patterns when they are unlikely to exist

Transcription

1 Finding market timing patterns when they are unlikely to exist Michael Kisser Loreta Rapushi November, 2017 Abstract We show that periods during which firms issue equity and simultaneously retire debt reflect market timing patterns: such leverage decreasing recapitalizations (LDRs) occur during periods of high equity valuation which are followed by a decrease in valuation ratios. Nevertheless, we argue that market timing unlikely explains those issues as the documented pattern also persists among firms that presumably have little leeway to time the market - such as firms with high (excess) leverage, an implied default credit rating or those that violate financial covenants. The underlying intuition is that in these cases creditors are likely to have substantial control rights. Instead, we show that the subsequent decrease in valuation ratios is explained by physical investment, which is consistent with an ex-post exercise of growth options. This interpretation is further supported by asset pricing tests which show that abnormal stock returns of firms undertaking LDRs are indistinguishable from zero. Keywords: equity issue; recapitalization; market timing; covenants; abnormal returns We have benefitted from the comments and suggestions of Leonidas Barbopoulos, Paul Borochin, Ettore Croci, Amil Dasgupta, Ran Duchin, Halit Gonenc, Gerard Hoberg, Tim Jenkinson, Minna Martikainen, Francisco Santos and Kate Suslava. We also thank seminar participants at the 2017 European Financial Management Association (doctoral seminar) and the Norwegian School of Economics. Norwegian School of Economics (michael.kisser@nhh.no) Norwegian School of Economics (loreta.rapushi@nhh.no)

2 1 Introduction It is a well known fact that firms which conduct seasoned equity offerings have high valuations (Asquith and Mullins, 1986; Masulis and Korwar, 1986). Moreover, these firms experience pre-issue stock price run-ups that are large and positive, whereas (abnormal) returns following the SEO are often negative (Loughran and Ritter, 1995; Spiess and Affleck-Graves, 1995). However, there is little agreement as to the underlying explanation for these empirical findings. For example, Baker and Wurgler (2002) suggest that market timing efforts drive equity issues and thereby they also have a long-lasting impact on corporate capital structures. Leary and Roberts (2005), on the other hand, conclude that the high valuations reflect growth opportunities and that the corresponding effect on capital structures can be rationalized with the existence of leverage adjustment costs. Kim and Weisbach (2008) further observe that firms stockpile cash following periods of equity issues and argue that this behavior is consistent with market timing efforts. DeAngelo, DeAngelo, and Stulz (2010) instead argue that the increase in cash reflects asset growth effects and that - without the SEO - firms would have quickly run out of funds. Results are also ambiguous for studies relating asset growth to stock returns following seasoned equity offerings. Butler, Cornaggia, Grullon, and Weston (2011) find that investment based factor models explain the negative stock returns following seasoned equity offerings. On the other hand, Lewis and Tan (2016) find that managers are more likely to issue equity when analysts are optimistic about long-term growth prospects. Further controlling for research and development (R&D) expenses as an additional investment factor, they further find that abnormal equity returns of equity issuers are negative and interpret this as consistent with the managerial attempt to time the market. Finally, Huang and Ritter (2017) show that the frequency and size of equity (and debt) issues is negatively correlated with future abnormal stock returns. In this paper, we take a different route and investigate the potential impact of market timing efforts by focusing on an over-looked subsample of equity issuers: firms that perform a leverage decreasing recapitalization (LDR) by issuing equity and using the proceeds to actively retire debt. The focus on LDRs is interesting for several reasons. First, it attempts to isolate periods of neutral asset growth and could therefore help in identifying the impact of market timing efforts. Second, because shareholders would never find it optimal to recapitalize to a lower leverage outside of bankruptcy or strategic debt 1

3 renegotiation (Fischer et al., 1989; Leland, 1994; Goldstein et al., 2001; Strebulaev, 2007; Admati et al., 2017), their existence likely reflects the exercise of creditor control rights. As a consequence, it becomes less likely that market timing efforts drive these equity issues. Using a large Compustat sample of 14,321 firms (147,256 firm-years) over the period from 1965 to 2015, we find that LDR account for 17 percent of all observed net equity issues. While net equity issues become more frequent during boom periods, this is not the case for LDRs. To the extent that market timing is more prevalent during hot markets, this raises the possibility that LDRs are driven by other factors than the average net equity issue. Finding similar patterns than documented previously in the literature, would therefore raise questions regarding the corresponding interpretation of the data. Turning to firm characteristics, we find that firms performing LDRs have higher leverage and lower cash holdings than the full sample of net equity issuers. Interestingly, both groups of issuers are on average unprofitable, invest substantial amounts into both Capex and R&D and are valued at high valuation ratios (Q). Moreover, valuation ratios peak and decrease after the issue for both samples. However, we also document substantial heterogeneity in firm characteristics among the group of LDR firms. We find that approximately one half of all LDRs occur among firms with an implied default credit rating or high leverage, whereas a non-trivial fraction of LDR firms exhibits low leverage and high investment financing needs. We then perform several tests to investigate whether LDR reflect market timing efforts of management. We begin this task by extending the annual valuation framework of Fama and French (1998) to specifically account for LDRs. This exercise relates a scaled version of the market-to-book ratio (precisely its excess value over one) to various fundamental factors and an indicator variable indicating the presence of a LDR. Our findings suggest that LDRs take place during periods of high valuation which are followed by a decrease in valuation ratios. This result is robust and obtains for both small and large firms. The pattern is similar to the sample of net equity issuers and it is a necessary (yet not sufficient) condition for the presence of successful market timing abilities. Empirically, it is however very difficult to disentangle market timing from rational behavior in the context of growth opportunities (Leary and Roberts, 2005). In simple words, equity values could be high because the company faces substantial (yet unexercised) growth opportunities and the leverage decreasing recapitalization solely occurs in an attempt to mitigate the debt overhang problem that arises when firms exercise growth options (Myers, 1977). Because the market-to-book ratio (or variants of it) identify both 2

4 periods of high equity valuations (market timing) and the presence of growth opportunities, it is hard to distinguish between the two forces - even in the context of a valuation model that controls for research and development expenses (among its fundamental factors) in an attempt to isolate growth opportunities. We address this problem using capital structure theory to guide our test design. First, dynamic tradeoff theory of capital structure implies that LDR should not exist outside of bankruptcty or strategic debt renegotitation (Fischer et al., 1989; Leland, 1994; Goldstein et al., 2001; Strebulaev, 2007; Admati et al., 2017). Our descriptive evidence shows that up to 50% of all LDRs occur among firms with an implied default credit rating or among those with high leverage ratios. We further hypothesize that for those firms, market timing efforts are (even more) unlikley to be the main driver of the recapitalization. The underlying intuition is that once a firm approaches a so-called default boundary, there is presumable little time left to time the market. Moreover, these periods are typically associated with falling equity values for the company (Bhamra et al., 2010b,a). Second, contracting theory rationalizes the existence of debt covenants to mitigate the ex-post suboptimality of LDRs (Smith and Warner, 1979; Aghion and Bolton, 1992; Dewatripont and Tirole, 1994). That is, shareholders and creditors can agree ex-ante on a set of (financial) covenants which the company must meet. If breached, creditors typically gain substantial control rights which - among others - can force the company to issue equity. We hypothesize that for those firms, market timing efforts are unlikley to be the main driver of the recapitalization. However, our findings show that the positive relation between LDR and equity value (as well as the subsequent decrease in valuation ratios) also exists among firms with high leverage, for those that are over-levered, that have an implied default credit rating or among firms that violated financial covenants. We argue that for all these subsamples, market timing considerations are unlikely to be the driving force as - by construction - in the cases creditors have substantial influence over the firm. In other words, we find market timing patterns where they are unlikely to exist. Using a setup where market timing is unlikely to exist, our findings emphasize how easy it is to be misled by equity value dynamics. That is, our findings cast substantial doubt over the use of equity value dynamics (which affect our valuation ratios) in the attempt to identify market timing efforts. Instead, our results raise the possibility that the existence of growth opportunities explains the decision of firms to undertake a leverage decreasing recapiatlization (Myers, 1977). Accordingly, we show that investment rates increase substantially in periods of recapitalizations which is consistent with the subsequent exercise 3

5 of growth options. Moreover, while investment rates increase valuation ratios decrease which could be consistent with a decrease in the firm s risk (Carlson et al., 2004). Consistent with this intuition, we further show that abnormal returns following periods of LDR are statistically indistinguishable from zero in the context of an empirical asset pricing model and further decrease in magnitude when one controls for the existence of growth opportunities. Taken together, our findings are difficult to square with the market timing hypothesis but are consistent with a rational decrease in valuation ratios of LDR firms due to the exercise of growth opportunities. The paper proceeds as follows. Section 2 presents the sample and provides descriptive evidence on LDR. Section 4 estimates the impact of LDR on shareholder value, while Section 5 attempts to provide explanations for the observed valuation effect of LDR. Section 6 concludes the paper. 2 Data and descriptive evidence 2.1 Sample Construction The sample consists of U.S. public industrial corporations listed on Crisp/Compustat (CCM) over the period from 1971 to As usual, we exclude financial firms, utilities and government entities. In addition, we require the availability of one-year lagged information on our main variables (to be introduced below). All other sample selection criteria are standard and are in Table 1. The final sample consists of 14,321 firms and 147,256 firm-years. In this paper, we focus on leverage decreasing recapitalizations (LDR) which we define as periods during which firms issue equity and use the proceeds to retire debt. This definition is consistent with dynamic trade-off of capital structure (Fischer et al., 1989) in which cash holdings are absent and, moreover, any period of capital structure inactivity is driven by security issuance costs. We precisely define this transaction as follows LDR t = 1 if NEI t A t > th and D t A t < th (1) where N EI are common and preferred stock issues net of dividends and share repurchases (obtained from the cash flow statement), A is the book value of assets and D is the change in the book value of long-term debt. The variable th is a size threshold which, in most situations, is set equal to 5%. The 4

6 choice of this cutoff value is motivated by Hovakimian et al. (2001) who show that the 5% classification scheme produces similar results than a direct merge with the SDC Database. A particularly interesting (yet often overlooked) feature of dynamic trade-off models of capital structure is that LDR should not exist outside bankruptcy or strategic debt renegotiation. This is because the recapitalization involves a transfer of wealth from shareholders to bondholders which is why the former find it optimal to postpone the recapitalization for as long as possible (Danis et al., 2014; Eckbo and Kisser, 2017). Table 2 displays annual values for the numbers of U.S. publicly listed firms (column 1) and net equity issues (column 2). Both numbers exhibit substantial variation over time, yet column (2) reveals that the cyclicality is particularly strong for equity issues which peak in the late 1990s. Column (3) displays the fraction of net equity issues that are classified as LDR and shows that, on average, 17 percent of all net equity issues are used to reduce long-term debt. Moreover, the corresponding fraction varies over time and is negatively correlated with the total number of equity issues. To the extent that market timing is more prevalent during hot markets, the descriptive information raises the possibility that LDR are driven by other factors than the average net equity issue. Table 3 reports descriptive information on selected financial characteristics for the full sample of firms, those performing a net equity issue and the subsample of firms which the NEI is classified as a leverage decreasing recapitalization. The table shows that firms performing a N EI end up with a more conservative capital structure than the full sample of firms: market leverage (L) is lower, the fraction of all-equity financed firms (AE) is higher, as is the ratio of cash holdings to assets (CR). The leverage of LDR firms lies between the two extremes (L equals 25% for the full sample of firms vs. 14% for NEI firms and 20% for LDR firms). Furthermore, the table shows that 6% of all firms performing a LDR choose to become all-equity financed and 31% become almost all-equity financed (which we define as those with a market leverage of less than 5% (Strebulaev and Yang, 2013)), suggesting an active decision to abstain from debt financing. Interestingly, the table shows that the cash holdings of LDR performing firms are comparable to the full sample of firms (18% versus 16%). However, firms performing either a N EI or a LDR are substantially less profitable than the average sample firm. The average ratio of operating profits to assets is -9% for LDR firms (compared to -14% for N EI firms and +7% for the average sample firm). Notwithstanding the low current profitability, equity valuations - measured as the market-to-book ratio (Q) - are higher for firms performing either a NEI 5

7 or a LDR. The combination of low profitability, high equity valuations and contemporaneous net equity issues is precisely what makes the focus on LDR performing firms an interesting research topic. In other words, the pattern is similar than for NEI firms, yet - as we argue below - the likelihood that market timing explains that pattern is much lower. Finally, we can see that LDR firms are somewhat smaller and they invest more into intangible capital (R&D). 2.2 Why do firms perform LDRs? Trade-off theory of capital structure implies that firms recapitalize only when close to bankruptcy (Danis et al., 2014; Admati et al., 2017). Note, however, that although an early recapitalization is ex-post suboptimal for shareholders, it may still reflect trade-off behavior. For example, the existence of bond covenants as well as (for researchers unobserved) agreements with private creditors might induce shareholders to commit to (and execute) pro-active leverage reductions (Nini et al., 2009). As a first approximation, we infer a firm s credit risk by mapping its interest coverage ratio (ICR) onto credit ratings. The ICR is a measure of the firm s ability to make payments to creditors and it compares a firm s earnings before interest and taxes (EBIT) to the level of interest payments. Using Damodaran s mapping table (displayed in Appendix Table 2), we then assign a synthetic credit rating for each sample firm. The notation is based on Standard and Poor s (S&P): AAA ratings denote the safest assets, ratings below BBB are referred to as non-investment grade and D identifies bankruptcy. Table 6 displays synthetic credit ratings for the full sample of firms, net equity issuers and those performing LDRs. For LDRs, half of the observed recapitalizations occur among firms with an implied default rating. These cases are broadly consistent with trade-off theory of capital structure which implies that leverage decreasing recapitalizations should occur close to (or at) bankruptcy (Fischer et al., 1989; Danis et al., 2014; Admati et al., 2017). 1 Reassuringly, the corresponding fractions are lower for all net equity issuers or the full sample of firms (46% and 23%, respectively). While the large fraction of LDR firms with an implied default rating suggests these might be driven by trade-off considerations, Table 6 nevertheless suggests that alternative forces might be at play. For example, 14% of all LDR firms receive at least a AA rating and 27% are rated investment grade. These findings are interesting and suggest the possible co-existence of two alternative types of firms performing 1 We use the word broadly to emphasize that the synthetic credit rating is based on a univariate mapping using the interest rate coverage ratio only. 6

8 LDRs: those that adjust leverage downwards towards a (possibly) optimal capital structure and another type of financially sound firms. To better understand, Table 5 describes the evolution of leverage, Q and the financing deficit over a five year event window surrounding the year of the LDR (Panel A) and the net equity issue (Panel B). Panel A shows that leverage peaks (and cash holdings hit a bottom) in the year before listing (L is 31%, BL 40%, CR is 15%). Turning to the decomposition of the financing deficit, we can see that LDR firms continue to invest and incur regular operating losses. Note that cash holdings would just suffice to finance one single average financing deficit and LDR firms thus (have to) issue equity in each event year (on average). Notwithstanding the peristent losses and high leverage, Q is high, peaks in the year of the LDR and decreases subsequently. These dynamics are consistent with previous findings on equity return surrounding stock issues (Asquith and Mullins, 1986; Masulis and Korwar, 1986; Loughran and Ritter, 1995; Spiess and Affleck-Graves, 1995) and have been interpreted as a necessary, yet not sufficient condition for the existence of market timing abilities. Panel B complements the picture by displaying corresponding information for all net equity issuers. Contrary to above, leverage is low and cash holdings are abundant. At the same time, the pattern for Q is more pronounced, as is the magnitude of operating losses. As a result, the financing deficit is large (31% of assets in event year 0) and firms in turn issue substantial amounts of net equity each year as existing cash holdings could only finance a single deficit (DeAngelo et al., 2010). Taken together, Tables 6 and 5 provide some evidence consistent with trade-off theory while also suggesting that financing deficit considerations are important. To further disentangle these two effects, Table 6 categorizes firms based on their market leverage ratio prior to the issue. To be precise, we first compute the distribution of lagged market leverage for the full sample of firms and then use the corresponding decile cutoff values to categorize LDR firms (Panel A) as well as the full sample of net equity issuers (Panel B) into ten different groups (ranging from low to high leverage). Panel A confirms that LDR firms on average have relatively high leverage. While only 9% of the LDR firms are in the lowest three leverage deciles (and thus among the third of all firms with the lowest market leverage), 38% of are in the three highest (leverage deciles). Conditioning on lagged market leverage further reveals systematic differences in cash policy, profitability, market valuation and investment: LDR firms with low initial leverage have large cash holdings, are very unprofitable but are valued at high multiples (Q). Moreover, these firms invest heavily into R&D and capital expenditures. While those 7

9 firms meet the technical condition of a LDR (requiring that net equity issues and debt retirements exceed 5% of book assets), columns 13 and 14 show that the net equity issue substantially exceeds the net debt retirement. On the other hand, LDR firms with high initial leverage are profitable, have low Q, are large and invest significantly in Capex but little into R&D. Also, the size of net equity issue and net debt retirements are very similar. Panel B shows that the underlying leverage distribution among the full sample of net equity issuers is quite different. Nearly one half (46%) of all net equity issuers are concentrated among the three lowest leverage deciles, while only 17% are in the three highest (leverage deciles). Conditioning on past leverage, the pattern is similar as for LDRs in Panel A. Low leverage firms have high cash holdings, are very unprofitabled, invest heavily and are valued at high multiples. High leverage firms, on the other hand, are more profitable, they invest less in R&D and they are valued at low multples. Taken together, the descriptive findings in this section suggest two potential explanations for why firms perform a leverage decreasing recapitalization. First, the LDR may be driven by the main objective to change the firm s capital structure. This explanation is consistent with trade-off behavior of capital structure and the existence of financial constrains and is likely among firms with relatively high market leverage. Alternatively, the LDR could be described as more of a by-product of a very significant equity issue where the proceeds are used not only to retire net debt but also to increase cash holdings and physical investment. 3 Valuation framework and hypothesis development Our univariate findings above suggest that the dynamics in valuation ratios of firms performing LDRs are similar to the full sample of net equity issuers: market-to-book ratios peak in the year of the issue and decrease subsequently. These patterns are consistent with previously documented evidence on stock returns surrounding stock issues (Asquith and Mullins, 1986; Masulis and Korwar, 1986; Loughran and Ritter, 1995; Spiess and Affleck-Graves, 1995) and have been interpreted as indicative of market timing. However, as discussed in the introduction, there is an ongoing debate whether those patterns constitute a sufficient (and not only necessary) condition. 8

10 3.1 Valuation framework The approach below is based on an extension to Fama and French (1998) who estimate the value impact of debt and dividend payments. 2 To arrive at our regression specification, we start from the well-known fact that levered firm value (V L ) can be decomposed into the value of the firm s unlevered assets (V U ) and the tax shield associated with debt financing (τd): 3 V L = V U + τd Assuming that unlevered firm value consists of both assets in place (V A ) and growth options (V G ), we can further write that V L = V A + V G + τd Using the book value of assets (A) as an approximation for the value of assets in place, leads to the following regression specification V L A = α + βv G + γd + ɛ In order to estimate the valuation model, one needs to control for the value of growth opportunities. Therefore (levels and changes of) operating profits (prof), R&D expenses (rd) and capital expenditures (capex) are included as additional control variables. Standardizing all variable by assets to both deal with heteroskedasticity and to also make sure that the largest firms do not drive results, implies that Q E t = α + β 1 P rof t + β 2 RD t + β 3 Capex t + η L dx t A t + η F dx t+v A t + γbl t + ɛ t where Q E t is (Vt L A t )/A t and can be interpreted as the the excess of Q over one. The variables P rof, RD, Capex denote the ratios of prof, rd and capex to assets and the compact notation dx t (dx t+v ) denotes the lagged one year (future v-year) change in the variable of interest (prof, rd or capex). 4 Using 2 The Fama-French valuation framework has been used extensively in the empirical cash literature which attempts to estimate the shadow value of cash holdings (Pinkowitz and Williamson, 2004; Pinkowitz et al., 2006; Kisser, 2013). 3 If the financial markets are competitive and corporations are taxed then, ceteris paribus, the value of the levered firm equals that of the unlevered firm plus the value of the debt tax shield, i.e., V L = V U + τd, where the L and U denote levered and unlevered, respectively, and (τ D) denotes the value of the debt tax shield (Modigliani and Miller, 1958). 4 Specifically, dx t = (X t X t v)/a t and dx t+v = (X t+v X t)/a t. 9

11 a two-year future change is in line with evidence that two years is as far ahead as the market can predict (Fama, 1990; Fama and French, 1998). Finally, BL is the book leverage ratio. 3.2 Hypotheses development The focus on LDR firms in this paper is motivated by the assumption that LDRs are likely driven by a financial motivation to decrease leverage. We therefore identify recapitalizations by of focusing on periods during which a firm issues substantial amounts of equity (in excess of dividends and share repurchases) and simultaneously retires a substantial amount of net debt (Leary and Roberts, 2005; Danis et al., 2014). The valuation framework of the previous section allows us to investigate whether LDRs systematically reflect market timing patterns. We first test whether LDR firms have a systematically higher valuation ratio and estimate Q E t = α + β 1 P rof t + β 2 RD t + β 3 Capex t + η L dx t A t + η F dx t+v A t + γbl t 1 + δ 1 I t + ɛ t (2) Note that the regression employs the lagged book leverage ratio in order to disentangle the effect of the leverage decreasing recapitalization (captured by the indicator variable It ) from the level of leverage. A positive coefficient estimate of δ 1 is consistent with (though not proving) the effect of market timing. Second, we investigate whether the period of the LDR is followed by a decrease in valuation ratios Q E t = α + β 1 P rof t + β 2 RD t + β 3 Capex t + η dx t A t + φ dx t+v A t + γbl t 1 + δ 2 I t + ɛ t (3) where Q E t = Q E t Q E t 1 and a negative value of δ2 would again be consistent with (though again not proving) the effect of market timing. Taken together, the analysis of the coefficient estimates δ and φ allows us to specify the following hypothesis for LDR firms (H1) LDR firms are unlikely to be driven by market timing and therefore should not exhibit dynamics in valuation ratios that are consistent with a market timing interpretation (δ 1 > 0 and δ 2 < 0) The underlying intuition of H1 is that valuation dynamics which are present for the full sample of net equity issuers (and which in the past have been interpreted as evidence consistent with market timing) 10

12 should not exist for LDR firms if the latter are primary driven by a motivation to decrease leverage. In the Appendix, we therefore provide complementary estimates of equations 2 and 3 for the full sample of net equity issuers. The evidence in Section 2 has revealed significant heterogeneity in terms of firm characteristics among the sample of LDR firms. That is, while a significant part of all LDR firms has indeed high leverage or an implied default credit rating, a non-trivial number of LDR firms exhibits low leverage and a strong need to finance a deficit that reflects large investment outlays. To improve our test, we therefore propose the following modified hypothesis: (H2) LDR firms which are in financial distress are very unlikely to be driven by market timing and therefore should not exhibit dynamics in valuation ratios that are consistent with a market timing interpretation (δ 1 > 0 and δ 2 < 0) The test of H2 assumes that market timing considerations should be even more unlikely in case firms are close to default as such cases are typically associated with falling (instead of rising) equity values (Bhamra et al., 2010b,a). Below, we identify the proximity to a default boundary using different measures including absolute leverage, its excess over a leverage target or the existence of bond covenant violations. 4 Do LDRs exhibit market timing patterns? 4.1 H1: LDRs and market timing patterns Table 7 displays the correlation between LDRs and shareholder value. Specifically, columns (1) to (3) test whether LDRs occur during periods of high valuations and present estimates of equation 2. Next, columns (4) to (6) investigate whether valuations decrease following the LDR and correspond to equation 3. To maximize sample size, we focus on one year future changes in the control variables (v = 1). 5 Results are provided using OLS regression (columns 1 and 4), accounting for firm-fixed effects (columns 2 and 5) as well as cross-sectional regressions in columns 3 and 6 (Fama and MacBeth, 1973). Fama-MacBeth regressions have the advantage that they identify the average cross-sectional effect, but come with the drawback of relatively little test power when applied to yearly data. 5 Alternative results when using two year future changes in the control variables (v = 2) are quantitatively similar and are displayed in the Appendix. 11

13 Focusing on the coefficient of the LDR indicator variable in column 1, we can see that the existence of a LDR increases excess Q by 0.55 units. In other words, this suggests that firms undertaking a leverage decreasing recapitalization have a market-to-book ratio that is approximately 0.6 units higher than for the average sample firm. Moreover, the coefficient is highly statistically significant and robust to alternative estimation methods including the presence of firm fixed effects (column 2) or FMB regressions (column 3). Investigating the period after the LDR, columns 4 to 6 provide strong evidence that the LDR is followed by a decrease in valuation ratios. Independent of the estimation method (OLS, FE, FMBA) we find that that excess Q decreases by approximately 0.15 units. Appendix Tables 3 and 4 show that the the pattern is similar when investigating the valuation effects of all net equity issues (as opposed to only LDRs) or when investigating the subsequent two-year (instead of one year) period. Also interesting, the coefficient estimate of operating profitability (P rof) shows that more profitable firms have lower excess market-to-book ratios. In other words, the negative correlation implies that low profitability firms on average have higher valuations, which is consistent with characteristics of high market-to-book firms (Fama and French, 1992; Novy-Marx, 2011). In addition, the correlation with lagged leverage is negative. Both coefficient estimates reflect the descriptive evidence above (Table 6) showing that LDR firms with low initial leverage are unprofitable but valued highly (whereas those with high leverage are more profitable and have lower market-to-book ratios). Table 7 suggests that firm size to a large part explains the negative impact of profitability as for large firms (with FE estimation or FMB) the correlation with excess Q becomes positive. Irrespective of size, LDRs are associated with high valuation periods (δ 1 > 0 in 6 out of 6 regressions) and decreasing subsequent valuations (δ 2 < 0 in 5 out of 6 regressions). Taken together, these patterns reject H1 as valuations peak in the year of the LDR and decrease subsequently. The documented pattern for LDRs is thus consistent with previously documented evidence on stock returns surrounding stock issues (Asquith and Mullins, 1986; Masulis and Korwar, 1986; Loughran and Ritter, 1995; Spiess and Affleck-Graves, 1995), which opens up for different interpretations. First, it is possible that - against all odds - leverage decreasing recapitalizations are executed to time the market. Second, it is possible that the heterogeneity among LDR firms (which was documented above in Section 2) introduces too much noise into our identification strategy. Third, the patterns reflect the existence and exercise of growth options. We turn to these issues next. 12

14 4.2 H2: Financially distressed LDRs and market timing patterns We now turn to a test of H2 which states that LDRs undertaken by financially distressed firms are very unlikely to reflect market timing patterns. In Table 9, we approximate financial distress using different measures including a firm s absolute leverage, its excess over a leverage target or an implied junk credit rating. Panel A investigates the correlation between the LDR indicator variable and excess Q and robustly confirms the previously found positive relation. Columns (1) to (3) contain results for firms for which the lagged market leverage is in the upper tercile of the distribution for the entire sample. This condition is satisfied for 42% (1,183 out of a total 2,755) LDRs. Irrespective of whether OLS, FE or FMB regressions are used, we find that a LDR raises Q E by 0.2 to 0.3 relative to the average Q E of all firms that are in the upper tercile of the leverage distribution. Even though these firms have substantial leverage (L = 52%) at the beginning of the year, it is possible that the absolute leverage ratio does not successfully sort on a firm s distance to default. Therefore, columns (4) to (6) present results based on whether firms have excess leverage. That is, we first estimate the lagged leverage target of each firm and then categorize each firm as overed-levered in case the lagged market leverage ratios exceeds the estimated target. 6 Results show that the correlation between LDR and Q E is again positive with a larger economic effect (γ 1 ranges between 0.22 and 0.48). Finally, columns (7) to (9) display results for firms with an implied junk credit rating. The correlation remains positive with similar economic magnitudes than in columns 4 to 6. Panel B tests whether LDRs of financially distressed firms are followed by a decrease in valuation ratios. Using the same classifications (absolute leverage, excess leverage, junk rating) as in Panel A, we find that all coefficient estimates γ 2 of the LDR indicator variable are negative, with seven out of 9 further being statistically different from zero. The table further shows that Q E decreases by 0.07 and 0.10 relative to the year of the net equity issue. Taken together, the results presented in Table 9 suggest that financially distressed LDR firms exhibit dynamics in valuation ratios that are consistent with a theory of market timing. This, however, is surprising given that our focus on distressed firms intends the reduce the ex-ante likelihood that firms engage in market timing. In other words, H2 hypothesizes that once a firm approaches a so-called default 6 To avoid mechanical mean reversion in leverage ratios, we estimate the target on a rolling basis. The set of control variables is standard and includes lagged values of size, profitability, market-to-book ratio, cash ratio, asset tangibility, R&D expenses, capital expenditures and the median industry market leverage ratio. In addition, we account for firm-fixed and time-fixed effects. 13

15 boundary, there is presumable little time left to time the market. Moreover, these periods are typically associated with falling equity values for the company. Before discussing the meaning of our findings, we present one more test that should further reduce the likelihood of finding market timing patterns in the data. Contracting theory rationalizes the existence of debt covenants to mitigate the ex-post suboptimality of LDRs (Smith and Warner, 1979; Aghion and Bolton, 1992; Dewatripont and Tirole, 1994). That is, shareholders and creditors can agree ex-ante on a set of (financial) covenants which the company must meet. If breached, creditors typically gain substantial control rights which - among others - can force the company to issue equity. We hypothesize that for those firms, market timing efforts are unlikley to be the main driver of the recapitalization. The underlying assumption of this exercise is that the covenant violation occurs for reasons that are unrelated to the high equity valuation of the company. We argue that this assumption is plausible as covenant violations are more likely during periods of falling equity values (which increase leverage) or during periods of deteriorating profitability. Practically, the identification of such forced recapitalizations is done using data from Nini, Smith, and Sufi (2009). This dataset is based on quarterly SEC filings for public U.S. corporations over the period from 1996 to For those firms, the authors identify whether a (at least one) financial covenant was violated or not. We then merge this dataset with our full annual Compustat sample, which results in the successful merge of 40,503 firm-years. In 12% (or 4,944 cases) financial covenants are violated for our sample firms. Moreover, we find that the frequency of covenant violations is relatively larger during periods of strict LDRs (17%, or 166 cases). We then re-visit the correlation between the value over cost spread and the LDR dummy variable by focusing only on firms that violated financial covenants. Tables 10 displays corresponding results. Columns (1) to (3) employ Q E as the dependent variable and show that the cross-sectional correlation between LDRs and valuation also persists among firms that violated financial covenants. These findings are robust to using OLS, FE or FMB regressions. Columns (4) to (6) investigate the impact of the LDR on the subsequent change in excess Q. All three coefficient estimates are negative, yet only significant at the 10 percent level. This, of course, reflects the relatively little power this test as there are only 166 observations for LDRs violating financial covenants. Taken together, the findings in this paper reject hypotheses H1 and H2. In other words, we find that LDR firms exhibit valuation dynamics that are consistent with a market timing interpretation. 14

16 Valuation ratios peak in the year of the net equity issue financing the debt retirement and valuation ratios subsequently decrease. These findings also obtain for LDR firms with high absolute leverage, positive excess leverage, an implied junk rating and when violating financial covenants. Because the latter firms are likely to be in a position to engage in market timing, the findings likely imply that dynamics in valuation ratios are not a good variable to test market timing theories. In other words, variation valuation is more likely to be driven by variation in growth opportunities - a point previously made by (Leary and Roberts, 2005) (among others). 5 LDRs, the exercise of growth options and abnormal returns The findings above reveal a robust positive correlation between the presence of a leverage decreasing recapitalization and excess Q. Moreover, it was shown that this correlation also obtains among subsamples of firms which are relatively close to a default boundary or where creditors have substantial influence over the firm due to a breach of financial covenants. Given the low likelihood that those distressed firms are in a position to time the market, the more plausible explanation for the positive relation between LDR and Q (and its subsequent decrease) relates to the existence of growth opportunities. That is, even though the valuation framework above controls for R&D, P rof and Capex, it is possible that some incremental information on growth options is captured by the decision to do a leverage decreasing recapitalization. To investigate this possibility, we perform two main tests. The first one explores whether the presence of LDRs leads to an (increased) exercise of growth options. The second one investigates the value and asset pricing implications of LDRs. Specifically, we hypothesize that while the exercise of growth options leads to a decrease in subsequent valuations ratios, abnormal stock returns of LDR firms should indistinguishable zero. 5.1 The exercise of growth options following periods of LDRs It is possible that the positive relation between LDRs and excess Q is explained by the presence of substantial growth opportunities among LDR firms. If this is indeed the case, we would expect that - ceteris paribus - the LDR allows firms to invest in a less constrained way. Put differently, the LDR should be followed an increase in the physical investment activity of firms which is conceptually equivalent to 15

17 the exercise of growth options. Figure 1 illustrates the investment dynamics in event time for firms undertaking a LDR in event year 0. The figure tracks, over the next three years, the evolution of three different measures of investment into fixed assets (all of which are scaled by the lagged value of book assets in order to accurately measure the resulting asset growth). The solid blue line shows capital expenditures (I CX ) which equal 9% of lagged book assets in the year of the LDR. The red dashed line also includes cash outlays for patent purchases and acquisitions, as well as net reductions resulting from asset sales, decreasing the total cash investment (I CF ) to 7% in year 0. Comparing the two measures suggests that LDR firms were selling assets in order to help finance capital expenditures. Finally, the long-dashed green line is based on the broadest measure of fixed asset investment (I F A ) and is computed from yearly changes in fixed assets in the firm s balance sheet (Lewellen and Lewellen, 2016). 7 What happens following the LDR is interesting. All three measures of investment increase substantially. Investment into capital expenditures and total cash financed investment are now similar and equal to 10% of assets. This suggests that the sale of assets in year 0 was temporary. Finally, we can see that total fixed asset investment (as measured by I F A ) nearly doubles from 14% to 23% of lagged book assets. To investigate whether the illustrated effects also holds in a multivariate context, we further estimate dinv t+v A t = α + β 1 P rof t + β 2 RD t + η dx t A t + φ dx t+v A t + γbl t 1 + δldr t + ɛ t (4) where dinv t+v = Inv t+v Inv t and dx t+v now only includes subsequent changes in profits (prof) and R&D outlays (rd). Table 11 again employs three different measures of fixed asset investment: capital expenditures, total cash financed investment and total fixed assets investment. Note, however, that the regression scales by current assets (and not lagged assets as in the Figure). For each investment measure, results are shown using OLS, FE and FMB regressions. The table uses a one year horizon suggests that capital expenditures increase by 1.8 percentage points (pp), total cash financed investment by pp. and total investment into fixed assets by pp. These effects are not only statistically significant but are also economically large (as visualised by Figure 1). 7 The yearly changes in fixed assets are adjusted for non-cash charges that affect fixed assets such as depreciation and write-downs. 16

18 5.2 Value and return implications of LDRs In this section, we further investigate the stock performance of firms performing LDRs. To this end, we merge our annual sample of Compustat data with monthly stock returns using the CRSP database. Following Huang and Ritter (2017), the merge imposes a four month lag between the date of the fiscal year-end (reported in Compustat) and the first monthly stock return (used from Crsp). This choice is somewhat more restrictive than market timing papers relying on security issue announcement returns and it is driven by the fact that we use the broader Compustat database (instead of SDC) to identify net equity issues and leverage decreasing recapitalizations. 8 Finally, monthly market returns, risk-free rates and returns of the book-to-market, size and momentum factors are obtained from Ken French s data library. 9 The merge with CRSP reduces the overall sample somewhat to 13,624 firms and 1,546,301 firm-months. Table 12 displays descriptive information on buy-and-hold stock returns for firms performing net equity issues (Panel A) and those undertaking LDRs (Panel B). Returns are computed over different buy-and-hold periods (ranging from T = 12 to 36 months) using different measures: the excess of the firm s gross return over the risk-free rate (Ret r f ) and an abnormal return relative to the market portfolio (Ret R m ). Panel A confirms the well established fact that net equity issuers have low subsequent stock returns. For example, the average abnormal return (Ret R m ) is -10.3% over a 24-months period following the portfolio formation, with half of all firms dropping by more than 35%. Panel B shows that - consistent with above evidence - the stock returns of LDR firms exhibit similar dynamics. Returns in excess of the risk free rate (Ret r f ) are positive, yet the distribution is skewed to the left as is indicated by the negative median values. Moreover, abnormal stock returns (Ret R m ) are negative over both a one and two year period following the LDR. While the stock return performance is less extreme than for NEIs, they are consistent with the above findings that valuation ratios drop after the leverage decreasing recapitalization. However, if LDRs also decrease the risk of a firm s equity then the negative stock returns relative to the market portfolio are not necessarily surprising. To disentangle the effect of market timing from a reduction in systematic risk, we turn to an estimation of abnormal returns implied by different empirical 8 Note that, while conservative, the imposition of a fourth month lag guarantees that the underlying net equity issue activity is public information. 9 Consistent with Fama and French (1993), we compute the book-to-market ratio using the seven month lagged value of market equity and we drop negative book-equity firms from the sample. 17

19 asset pricing models (Sharpe, 1964; Fama and French, 1992, 2014). Table 13 presents monthly valueweighted excess returns of a trading strategy investing in firms that performed a LDR (Panel A) or a net equity issue in general (Panel B). Excess returns are computed relative to three competing asset pricing models: the market-model, the Fama and French three-factor model and the five factor model which also accounts for profitability and investment. Panel A shows that abnormal returns when investing into firms that undertook a net equity issue are negative and economically large. Abnormal returns equal -48 basis points per month (p.m.) under the CAPM (t-statistic of -3.1) and increase to -24 basis points under the three-factor model (t-statistic of -1.83). However, accounting also for investment and profitability as additional risk factors completely kills any remaining abnormal return (α is 1 basis point with t-statistics of 0.07). The magnitude of these estimates and the documented effect of investment and profitability is consistent with other findings (Huang and Ritter, 2017). Turning to an analysis of firms performing a LDR (Panel B), we find that estimating the capital asset pricing model (CAPM) is already sufficient to generate a statistically insignificant estimate of the abnormal return. These findings are difficult to square with successful market timing efforts of management. Moreover, accounting for size and value as additional risk factors further increases abnormal returns (from -15 basis points p.m. to -4 basis points, with a corresponding increase in t-statistics from to -0.19), which is consistent with the argument that variations in growth opportunities explain the decrease in valuation ratios of LDR firms. As above, the inclusion of investment and profitability removes any abnormal return effect (α is 2 basis point with t-statistics of 0.10). Similar findings also obtain when we focus on the three previously used subsamples of financially distressed LDR firms. That is, when we invest into LDR firms with high leverage, those that are overlevered or have an implied junk credit rating, we continue to find that excess returns are statistically indifferent from zero. 6 Conclusion This paper performs an analysis of leverage decreasing recapitalizations and shows that half of them occur for firms with an implied default credit rating. We then show that LDRs reflect many valuation patterns that are frequently interpreted as being consistent with market timing efforts: LDRs occur during periods 18

20 of high equity valuations, and are followed by subsequent decreases in valuation ratios. These findings are particularly surprising as they prevail even among a subset of LDR firms that is very unlikely to be driven by market timing. For example However, our analysis further suggests that the patterns are more likely driven by the existence (and subsequent exercise) of growth options. Firms increase investment rates in periods following the LDR, valuation ratios decrease but abnormal stock returns are statistically indifferent in the context of single and multi-factor asset pricing models. 19