Proactive Leverage Increases around the World

Transcription

1 Proactive Leverage Increases around the World By Aaron Brauner Latest Version: November 2017 Abstract Large debt issuances that increase firm leverage beyond estimated long-run targets are pervasive in both middle and high income countries. The evolution of debt ratios after a proactive leverage increase is determined by cash flows rather than attempts to rebalance toward a target. In middle income countries, firms utilize this method of external finance more frequently, and the link between debt ratio dynamics and cash flows is stronger. These countries have weaker institutions and capital markets that lead to higher transaction and agency costs of equity relative to debt. This suggests that debt capacity is more important for financial flexibility and capital structure decisions in countries where equity is a costlier form of external finance. Contact information : Aaron Brauner- abrauner@albany.edu 1

2 I. Introduction Financial flexibility in the form of unused debt capacity is a major determinant of capital structure. Market frictions combined with volatility in either the investment opportunity set or cash flows makes flexibility valuable to firms as a means of avoiding costly underinvestment. Stockpiling cash or issuing equity to finance investment projects can be costly due to information asymmetries that create adverse selection or moral hazard problems. In countries with capital markets and institutions that fail to resolve these information asymmetries, debt is often the only source of financial flexibility when external capital is needed. We examine large, often investment-driven, debt issuances that increase firm leverage beyond estimated long-run targets in 30 high and middle income countries. We find that firms in middle income countries (MICs) conduct proactive leverage increases more frequently, operate with higher leverage pre- and post-jump, and that the post-jump evolution of their debt ratios is more strongly tied to cash flows than firms in high income countries (HICs). The majority of firms could not have implemented their chosen strategies with cash on hand; they required external capital. The results suggest that debt capacity plays a larger role in the capital structures of firms in countries with weaker institutions and less developed capital markets. In these environments information asymmetries likely lead to higher agency costs (frictions) of cash and equity, leaving debt as the primary alternative. Financial flexibility has received increased attention since Graham and Harvey (2001) reported that CFOs consider it to be the most important determinant of capital structure. It can be obtained through a combination of cash reserves, payout policy, and capital structure policy. Higher retained cash flows, resulting in lower equity payouts to shareholders and higher cash balances, and high debt capacity allow firms to respond in a value maximizing manner to shocks to their investment opportunity set and cash flows. A key feature of financial flexibility is that firms must have a volatile investment opportunity set and/or cash flows, and there must be market frictions that make raising external capital costly. If the firm had perfect foresight, capital markets were frictionless, and there was complete information, there would be no need to maintain financial flexibility. These 2

3 frictions are represented by costs associated with raising internal capital (retained cash) or external capital (equity and debt issuance) and vary based on the institutional environment and capital market structure. This paper addresses how these country factors affect firms use of debt as a source of financial flexibility. Debt capacity as a source of financial flexibility is based on its prominence in the competing theories of capital structure. In the pecking order theory, firms prefer debt over equity when external capital is needed due to costs imposed by asymmetric information problems (Myers & Majluf, 1984). Although firms ultimately prefer internal cash to external sources, excess cash can have agency costs, such as managerial overinvestment, that make debt preferable (Jensen, 1986). Further, optimal contracting models endogenize the financial contract by starting from an agency problem, like managerial overinvestment, and conclude debt to be to the optimal response. 1 In trade-off models, firms target an optimal debt ratio by weighing the costs and benefits of additional debt: tax benefits, financial distress costs, and costs due to agency or asymmetric information problems. DeAngelo, DeAngelo, & Whited (2011) combine elements of both the pecking order and trade-off theories in a dynamic model and conclude transitory debt issues are an important source of financial flexibility. There is also substantial empirical evidence that financial flexibility in the form of unused debt capacity plays an important role in capital structure choices. Sufi (2009) reports that revolving credit lines comprise a large proportion of firm debt obligations. Similarly, Kahl, Shivdasani, & Wang (2015) show the commercial paper market provides firms with financial flexibility needed when raising capital for a new investment. Lins, Servaes, & Tufano (2010) survey CFOs from 29 countries who confirm that lines of credit are their primary source of liquidity. Denis & McKeon (2012), hereafter DM, examine large debt issuances in the USA that drive the firm leverage beyond estimated long-run targets. The evidence from these proactive leverage increases (PLIs) suggests that substantial debt issues are done primarily in response to operating and investment needs. They also find that the evolution of the debt ratio is driven by the realization of a financial surplus or deficit rather 1 See (Stein, 2003, pp ) for an overview of optimal contracting literature. 3

4 than attempts to rebalance toward a target. In fact, subsequent deficits are covered with more debt even though firms exhibit leverage ratios well above estimated targets. Overall, this evidence supports the notion that debt capacity and transitory debt issues are important sources of financial flexibility around the world. The empirical results of Denis & McKeon (2012) closely fit the predictions of a model developed by DeAngelo, DeAngelo, & Whited (2011) which are summarized by Denis (2011, p. 672): In their model, firms face volatility in both cash flows and the investment opportunity set, thereby creating the need for financial flexibility in order to avoid costly underinvestment. However, because stockpiling cash is itself costly due to tax and agency costs, the optimal financial policy consists of low, long-run leverage targets that preserve debt capacity. Subsequent debt issues (and repurchases) then represent pro-active responses to shocks to the firm's investment opportunity set and its cash flows. In other words, transitory debt issues are an important source of financial flexibility. Their model combines the agency costs of cash and equity caused by information asymmetries in the pecking order theory, with the leverage targets featured in trade-off theory. The innovation is that the costs of debt include the opportunity cost of the inability to borrow in the future. Thus, leverage targets conservatively embed the option to borrow later at favorable terms. We provide further support for the model presented by DeAngelo, DeAngelo, & Whited (2011), hereafter DDW, by examining proacive leverage increases in an international setting. By examining PLIs across a range of institutional environments and capital markets we can evaluate the validity of the DDW model. Namely that higher agency costs make equity more expensive relative to debt as a source of external finance. This makes transitory debt issues such as PLIs an even more essential source of financial flexibility. Countries with weaker institutions and less developed capital markets are less efficient at solving the information asymmetries that lead to the costly agency problems associated with equity issues. Therefore, in these countries we expect firms to rely more heavily on PLIs. Unused debt capacity is also a more valuable lifeline to these firms. Thus, they will use surpluses to pay down existing debt to preserve their debt capacity after a PLI. These predictions are consistent with our results from MICs and are developed more formally in the next section. 4

5 It is important to examine PLIs in an international setting since it has been shown that access to capital markets varies across countries. This heterogeneous access to capital markets has consequences for firm investment and performance. Firms are often constrained by the capital market conditions of the country in which they operate; poorly functioning capital markets can prevent firms from obtaining funding needed to pursue growth opportunities. Heterogeneous institutional environments that cause differences in financial development and access to capital across countries also affect a range of firm financial choices and outcomes. Property rights, accounting standards, creditor rights, and protections against self-dealing have been shown to influence the issuance of stock and debt, ownership dispersion, voting premiums, and payout policies of the firm as well as the development of financial markets (Foley & Manova, 2015). A study of debt issues across countries is also timely given present economic trends. In response to accommodative monetary policy in developed countries over the past decade there has been a documented increase in emerging market leverage. Specifically, the debt of nonfinancial emerging market firms more than quadrupled from US$ 4 trillion to US$ 18 trillion between 2004 and 2014 (Alter & Elekdag, 2016). With this large increase in emerging market debt it is natural to ask if PLIs of the type documented by DM occur internationally, and how these patterns vary across heterogeneous institutional and economic environments. This paper is our attempt to address these questions as well as test the DDW model in an international setting. In this paper, we examine 7,639 instances from 30 countries between 1981 and 2015 in which firms aggressively increase their leverage ratio through a substantial debt issuance (as opposed to a stock price decline). To be identified as a PLI the increase in leverage must be at least 0.10 and the resulting post-jump leverage must be at least 0.10 above the estimated long-run target. On average, firms increase leverage by 0.24, resulting in a leverage ratio that is 0.26 above the estimated target. While MIC firms operate at much higher levels of leverage, the changes in leverage are very similar between the two income groups. 5

6 We then track the use of proceeds using the DM methodology to shed light on the motivations for the debt issuance. Our analysis indicates that the funds are used primarily for long-term investment (55%), followed by additions to working capital (26%), and then to cover operational shortfalls (14%). Very few PLIs were used to make a payout to equity holders (2%). Interestingly, this breakdown of uses of proceeds is similar in both HIC and MIC groups and is consistent over time. PLIs are conducted for similar purposes across varying economic and institutional environments. Approximately 85% of the sample firms would have been unable to pursue their chosen operating policy using internally generated cash had they not raised external capital. After a PLI firms do not behave as if rebalancing toward a target leverage ratio is a first order concern. Ex-post debt ratio dynamics are instead explained by the evolution of cash flows, with surpluses being associated with a decrease in debt levels and deficits being associated with an increases; in some cases, firms running deficits actually continue to decrease leverage. This link is strongest in MICs and our coefficient on surpluses dwarfs the other determinants of leverage ratios in regression analysis. When we restrict our sample to only surplus years the significance of the other determinants (including target leverage ratios) drops out completely, meaning that firms running a surplus after a PLI behave as if they are indifferent to being above or below their target leverage. In cross-sectional analysis we provide evidence indicating that firms in less financially developed countries, with weaker property rights, less product market competition, and more corruption conduct relatively more PLIs than firms in more advanced countries with stronger institutions. Economies with weaker institutions and less financial development have smaller and less liquid stock markets. In the absence of deep, liquid equity markets, debt issues are a more frequent method for raising external finance. This implies that debt capacity is a more important component of financial flexibility, and capital structure in general, in less developed financial markets. Our contribution is to show that firms in MICs behave as if debt is their primary source of financial flexibility, while firms in HICs behave as if flexibility can be achieved through a combination of debt, equity, and cash. We extend the methodology of Denis & 6

7 McKeon (2012) for identifying PLIs to an international setting and find results consistent with theirs. We then use it to test several predictions that come from applying the dynamic capital structure model of DeAngelo, DeAngelo, & Whited (2011) to various institutional environments with variation in the agency costs of cash and equity. We find that where agency costs are likely to be higher PLIs are relatively more frequent. That is, when equity and cash are costlier firms rely on transitory debt issuances for investment. We have contributed to the ongoing discussion of financial flexibility (Denis, 2011), and debt as a major source of that flexibility [ (Sufi, 2009); (Lins, Servaes, & Tufano, 2010); (Kahl, Shivdasani, & Wang, 2015) ]. We also contribute to the discussion of other financial policies that provide flexibility such as payout policies [ (Jagannathan, Stephens, & Weisbach, 2000); (Bonaime, Hankins, & Harford, 2014)]. This paper also contributes to the literature on country factors that influence corporate financial policy and financial development. From La Porta et al. (2013, p. 438): One of the foundational assumptions of law in finance is that the central agency problem of the firm is the expropriation of investors by corporate insiders, whether controlling shareholders or managers. This agency problem is at the heart of the DDW model that we test using our sample of PLIs. We test it by exploiting cross-sectional differences in property rights and capital markets that change the costs of equity and debt contracts. In the contractual view of the firm, the protection of property rights from corporate insiders is essential to assuring the flow of capital to firms and makes investors willing to provide capital at lower cost. 2 The assumptions that these agency costs disproportionately affect equity securities in the DDW model are consistent with our result that firms in MICs with weaker property rights conduct many more PLIs than their counterparts in HICs with stronger property rights. The rest of this paper is arranged as follows. Section 2 provides additional background and develops our hypotheses. Section 3 describes our data and identfication of PLIs. Section 4 describes debt ratio dynamics before and after the jump, as well as the use of 2 See (La Porta, Lopez-de-Silanes, & Shleifer, 2013) for an extensive discussion of Law and Finance. 7

8 proceeds. Section 5 presents a cross-sectional analysis of the frequency of PLIs across countries and the factors that explain this variation in PLI frequency. Section 6 concludes. II. Background and Hypothesis Development Until recently, the empirical corporate finance literature has consisted of running a horse race between the pecking order (Myers & Majluf, 1984) and tradeoff theories. In the pecking order model, costs due to information asymmetries result in a preference for internal funds or debt over equity for external finance. In tradeoff models, firms identify their optimal/target leverage by weighing the costs and benefits of an additional dollar of debt. Both have significant empirical shortcomings: the pecking order theory fails to explain frequent equity issuances by firms; the trade-off theory fails to explain lower than expected debt ratios and slow rebalancing to estimated targets (Fama & French, 2002). Recently however, elements from both models have been combined with financial flexibility by assuming the opportunity cost of borrowing today is the potential inability to borrow in the future. In other words, uncertainty regarding future earnings, investment opportunities, and security prices incentivizes managers to pursue financial policies that maximize flexibility to respond to unanticipated future shocks to these factors (DeAngelo & DeAngelo, 2007). Volatility in both cash flows and the investment opportunity set, and costly stockpiling of cash are woven together by DDW (DeAngelo, DeAngelo, & Whited, 2011) in a model where leverage dynamics are driven primarily by flexibility considerations, and debt capacity is the primary source of that flexibility. Firms deliberately but temporarily deviate from permanent leverage targets via transitory debt issues to fund investment. Leverage targets conservatively embed the option to issue debt in the future, and rebalancing is driven by investment needs, cash flows, and market conditions in future states of the world. The emphasis on financial flexibility in recent theoretical work is consistent with survey evidence by Graham & Harvey (2001), in which CFOs identify financial flexibility as the most important determinant of capital structure. Another implication of the theoretical literature is that debt capacity is a vital source of financial flexibility. This is consistent with Sufi (2009) who finds that firms maintain unused lines of credit twice as large as the utilized credit capacity. Similarly, Lins, Servaes, & Tufano (2010) survey CFOs from 29 countries 8

9 who site lines of credit as their dominant source liquidity. It is also consistent with studies that show leverage ratio dynamics are driven by debt issuances (Welch, 2004), and that debt issuances increase the deviation of a firm s debt ratio from target (Hovakimian, 2004). Lastly, it is consistent with observed investment behavior such as investment spikes after periods of conservative leverage (Marchica & Mura, 2010), and proactive leverage increases with slow rebalancing in response to operating needs (Denis & McKeon, 2012). Denis (2011, p. 667) defines financial flexibility as the ability of a firm to respond in a timely and value-maximizing manner to unexpected changes in the firm s cash flows or investment opportunity set. In the frictionless environment of Modigliani & Miller (1958) perfect capital markets yield firms complete financial flexibility in the form of costless capital structure adjustments. Thus, it is only in the presence of financing frictions that financial flexibility becomes an interesting concept. Such frictions are captured through several assumptions regarding the costs of raising internal and external capital in the DDW model. First, internal cash is costly because of agency costs, as in Jensen (1986), that allow managers with an informational advantage to over-retain resources for self-interested projects. Second, external equity is costly because of the adverse selection problem, as in Myers and Majluf (1984), in which informationally advantaged managers can sell securities above their intrinsic value and refuse to sell them below their intrinsic value. Lastly, debt capacity is assummed to be finite making the option to issue debt in future periods valuable and whose optimal usage depends on current and future shocks to the investment opportunity set. 3 The result of this model is that information asymmetries and moral hazard make it costly for managers to raise capital from both outside and inside the firm. These costs are prime examples of the market frictions that make financial flexibility a critical determinant of capital structure. It is noteworthy that the adverse selection argument used for costly equity issues in this model is not applied to debt issues. The adverse selection argument is that managers with private information of a likely default will borrow more. Alternatively, from a moral hazard view, managers who borrow more have an increased incentive to take risks that lead to default, since they get all of the upside. It has been shown by Jaffee & Russel 3 Additional costs for cash are taxes and foregone interest payments on other liquid assets (i.e. treasury bills); security flotation costs for equity issues; interest payments for debt (DeAngelo, DeAngelo, & Whited, 2011). 9

10 (1976), Stiglitz & Weiss (1981), and others that these types of considerations can lead to credit rationing. The rationale for omitting these types of costs in debt markets comes from Myers (1984), and Myers & Majluf (1984), who argue that adverse selection problems are more severe in the equity market because equity values are more sensitive than debt values to managers private information (Stein, 2003). Put simply, debt is easier to value than equity in markets with agency conflicts and incomplete information. 4 This framework, in which information asymmetries make it costly to raise capital and debt capacity is the primary source of financial flexibility, is useful for thinking about how financial flexibility is affected by institutional environment. The central message of law and finance research is that legal protections for investors have substantial implications for the organization and development of capital markets (La Porta, Lopez-de-Silanes, & Shleifer, 2013). While firms can certainly affect their financial flexibility through cash management, payout, and capital structure policies, the relative costs and benefits governing this choice are determined by the structure of the capital markets and the underlying institutional environment. According to the DDW model, firms in countries whose institutions do not adequately resolve information asymmetries will have higher agency costs of cash and equity and will be forced to rely on debt for financial flexibility more than firms in countries with strong institutions. The fact that debt is not subject to severe agency costs that can be mitigated with strong institutions, while equity and cash are, allows us to exploit crosscountry differences to test the model empirically. The central idea of this paper is that debt capacity becomes the dominant source of financial flexibility when weak institutions and underdeveloped financial markets lead to high costs of using equity as a source of external finance; whether due to illiquid equity markets or high agency costs. This conjecture yields several testable hypotheses when we compare MICs that are likely to have costly equity, to HICs who have been shown to have stronger institutions and more developed capital markets. 4 Additional rationale comes from literature that seeks to endogenize the financial contract by starting from a specified agency problem and asking what claim represents the optimal response to the problem. Often the contract that emerges is a standard debt contract with no outside equity financing (Stein, 2003). 10

11 Hypothesis 1: Firms in MICs will be more levered prior to (and after) a PLI. If the capital markets and institutions in a given country lead to higher agency costs of equity and cash, then firms will be forced to use more debt in their capital structures ceteris paribus. Thus, regardless of whether a firm conducted a PLI, we expect firms to be more highly levered in middle income countries. Hypothesis 2: Firms in MICs will exhibit debt ratio dynamics driven primarily by cash flows and debt capacity, as opposed to a permanent long run target. In countries where agency costs of equity are high, the option to issue debt in the future at desirable terms is valuable and will be preserved. Thus, firms in MICs running surpluses will aggressively pay down their debt to regain this valuable option and preserve their debt capacity. This hypothesis leads to 2 possible outcomes for firms running deficits. Firms running deficits but with high debt capacity will rely on this debt capacity to cover operational losses; they may continue to take on additional debt despite being well above their target leverage. However, firms running deficits with lower debt capacity may aggressively pay down their debt, possibly going below their targets, in order to regain their scarce debt capacity. Hypothesis 3: Firms in MICs will conduct relatively more PLIs than their counterparts in HICs. Further, cross-sectional variation in the frequency of PLIs will be driven by proxies for legal protections (i.e. property rights) and capital market development. Conditional on firms raising external capital, we would expect to see relatively more PLIs in MICs than in HICs. The agency costs of equity are determined by the capital markets and legal institutions ability to solve the information asymettry problem between investors and corporate insiders. When they fail to adequately do this, firms must use debt more aggressively since equity is so costly. Thus PLI frequency should be associated with proxies that capture higher agency costs of equity (i.e. less developed capital markets and weaker property rights). 11

12 III. Data and Sample Selection We use the Worldscope database for firm-level accounting data for companies across the globe. Worldscope extracts data from filings with financial regulatory institutions and adjusts them for differences in accounting practices to facilitate comparisons between companies and industries both within and across national boundaries. The items reported are in local currency by country of domicile. The universe of companies available represents approximately 95% of the global market capitalization and contains both active and inactive companies. The earliest annual information available is for 1980 for large companies in developed markets, with smaller companies and emerging markets becoming available in the 1990s and early 2000s. The starting sample consists of all firms in the Worldscope database with assets greater than $1M and non-missing data for industry (SIC code) and country of a domicile. We exclude financial firms (SIC codes ), regulated utilities (SIC codes ), and government entities (SIC codes ). We define market leverage as Market Leverage it = DEBT it DEBT it + MCAP it, (1) where DEBT is the sum of long and short term debt and represents all interest bearing and capitalized lease obligations. MCAP is market capitalization and represents the total market equity value of the company; it equals market price per share of common equity at the fiscal period end multiplied by common shares outstanding. Market Leverage is hereby abbreviated as MLEV and the denominator DEBT+MCAP is abbreviated as MASSETS (market assets). To proxy for the long-run target leverage ratio we use the following double-sided Tobit model censored at 0 and 1 and estimated separately within each year and country: 12

13 MLEV it = α + β 1 (Ind Med MLEV i,t 1 ) + β 2 ( M B ) i,t 1 + β 3 ( PPE ASSETS ) + β 4 ( EBITDA i,t 1 ASSETS ) i,t 1 (2) + β 5 (ln(assets$)) i,t 1. The target leverage ratio for firm i is estimated by the fitted value from this model MLEV it. Differences between a firms realized leverage and its estimated target (MLEV it MLEV it ) measure the deviation in leverage from that of firms in similar industries (Ind Med MLEV), with similar characteristics such as market-to-book ratio ( M B ), tangibility (PPE ASSETS ), profitability ( EBITDA ASSETS ), and size (ln(assets$)). These factors come from (Frank & Goyal, 2009) who find them to be the most consistent predictors of leverage decisions in publicly traded firms. Estimating separate regressions within each country-year pair allows us to exclude expected inflation from the model since it is constant across all firms within each country-group. Industry median leverage is defined as the median leverage in each country-industryyear group. We require that there be at least ten observations in the group defined by the four-digit SIC code. If there are not ten observations, we expand it to three or two-digit SIC codes. Market-to-book ratio is defined as M B = ASSETS it EQUITY it DEFTAX it + MCAP it + PRFSTK it, (3) it ASSETS it where ASSETS is total assets, EQUITY is book equity, DEFTAX is deferred taxes, and PRFSTK is the liquidation value of preferred stock. Asset tangibility is defined as the ratio of net property, plant, and equipment (PPE) to total assets. Profitability is defined as the ratio of earnings before interest, taxes, and depreciation (EBITDA) to total assets. When EBITDA is missing we use EBIT. Size is defined as the log of total assets expressed in US dollars. Since ln(assets$) is the only independent variable that is not a ratio we use the Worldscope data item that provides assets converted to US$. The rest of the independent variables use assets denominated in local currency resulting in a unit-less ratio. 13

14 To identify firms that proactively increase leverage through large debt issuances we follow DM and define a variable $ΔMLEV which captures the value of additional debt not arising from asset growth with a constant leverage ratio. It is defined as $ MLEV it = DEBT it DEBT i,t 1 ( MASSETS it MASSETS i,t 1 ) (4) This allows us to isolate firms that increase leverage primarily through the result of a debt increase as opposed to a decrease in equity. Leverage increases of this nature would otherwise be difficult to identify empirically because large buybacks or stock price declines could reduce equity causing an increase in leverage. DM provide an intuitive example for how their measure captures the value of additional debt represented by the change in leverage normalized by change in market assets: a firm starts with 100 in assets and 20 in debt in year t 1, so the leverage ratio is During year t, the firm issues 30 in debt and its assets grow to130. If the firm kept its leverage ratio constant, given a 30% increase in assets (130/100), debt would increase to 26 ( ). However, its debt rose to 50, so the additional 24 is the increase in debt not arising from larger assets: $ MLEV it = (50 20 ( ) = 24). We identify a proactive leverage increase (PLI) using the following criteria: 1. The change in leverage must be at least 0.10: MLEV it = MLEV it MLEV i,t The post jump leverage must be at least 0.10 above target: MLEV it MLEV it The change in total debt must be at least 90% of $ MLEV it : DEBT it DEBT i,t 1 $ MLEV it The firm must not have committed a PLI in the previous 7 years. Clustered PLI s within the 7 year tracking period are treated as the same event. 5. Between 80% and 200% of the increase in debt must be identifiable via cash flows. Criterion 5 regarding identification of cash flows is discussed in detail in the next section on uses of the proceeds; it allows for an analysis of the use of debt proceeds by tracking the cash. Criteria 5 and 6 are relaxed in the final section where we develop a measure of the frequency 14

15 of PLIs at the country level for a cross-sectional analysis. We relax them because we want to estimate how frequently firms engage in PLIs not necessarily what they are using the proceeds for. Table 1 presents the sample of 7,639 PLIs that meet our selection criteria in the Worldscope database. We omit countries where we observe less than 20 PLIs. The unbalanced panel structure of the database is evident as we can track PLIs in the USA going back to the first possible year in 1981, but HICs do not enter the sample in any significant way until 1989, while MICs come in two waves in 1995 and This is due to countries populating Worldscope unevenly, with much of the data from emerging markets only becoming available in the late 1990s or 2000s. About 50% of the PLIs observed in the sample are from HICs outside of the USA, while 20% come from MICs, and the remaining 30% come from the USA. The high incidence of observations from the USA is why we separate it from the other HICs in the rest of this paper- the USA is not included in the High Income Country (HIC) group. We can see from Table 1 that PLIs do not seem to be explained by differences in national income. A substantial number of proactive leverage increases occur in both HICs and MICs. We will see later that MIC firms actually engage in relatively more PLIs than HIC firms. The final two columns in Table 1 show that the PLIs are spread out amongst the countries in our sample with no country, besides the USA, having more than 9% of the observations. Panel A of Figure 1 complements this by showing the total number of PLIs of each country and the income classification. The countries with the most and the least PLIs are groups dominated by HICs, with MICS clustered in the center of the distribution. Panel B of Figure 1 shows that once the emerging markets come online in the late 1990s and early 2000s the frequencies from year-to-year closely track those of developed markets. The correlation between the HIC and MIC counts is Taken together this implies that proactive leverage increases are pervasive across time and countries, rather than limited to developed countries. 15

16 IV. Leverage Ratio Dynamics and Use of Proceeds Table 2 reports statistics describing the distribution of leverage ratios before and after the PLIs in HIC and MIC groups (the USA has been excluded from HIC group). The increases in leverage are dramatic for both country groups. The median HIC-PLI results in a 25-point jump in the firm s market leverage ratio going from a pre-jump leverage of to a post-jump leverage of The median MIC-PLI results in a 23-point jump from to The magnitudes of these increases are consistent across the most and least levered firms measured by pre-jump leverage. Firms in HICs increase leverage by 28 points and 22 points at the lower and upper quartile of pre-jump leverage respectively. Similarly, firms in MICs increase leverage by 25 and 20 points at the lower and upper quartiles of pre-jump leverage. This consistency implies that debt capacity is at least as important a driver of financial policy in MICs as it is in HICs. Indeed, debt capacity looks to be even more important for financial policy in MICs than it is in HICs. Consistent with hypothesis 1, across every quartile the pre- and post-jump leverage are consistently points higher for MIC firms. When we look at a firm s leverage in excess of their peers, we see that pre- and post-jump excess leverage ratios follow a nearly identical trend in the two country groups. This result suggests that the mechanics behind PLIs are the same in both income groups but that firms in MICs consistently operate with higher leverage ratios. In an environment such as this, a firm s debt capacity should play a dominant role in financial decision-making. In order to track the use of funds from a PLI we follow DM and calculate the total amount of cash used for each of 4 categories. INVESTMENT: All funds used for investment activities such as capital expenditures, acquisitions or other investment activities INVESTMENT = CAPX + ACQUISITIONS + INCINV SPPE DECINV OTHERINV (5) 16

17 Where CAPX represents funds used to acquire fixed assets not associated with acquisitions; ACQUISITIONS represents assets acquired through pooling of interests or mergers; INCINV represents investments bought during the fiscal period; SPPE represents cash the company received from the sale of plant, property, and equipment; DECINV represents investments sold during the fiscal year; and OTHER represents other miscellaneous investment activities. 5 WORKING CAPITAL: Positive figures for changes in working capital (ΔW) are considered uses of funds to fund working capital needs. ΔW = CASH RECEIVABLES INVENTORY PAYABLE TAXES OTHERATSLBS INVESTMENTS (6) OTHERUSES Where ΔCASH represents the increase in cash and short term investments; ΔRECEIVABLES represents the decrease in net receivables; ΔINVENTORY represents the decrease in inventories reported on the cash flow statement; ΔPAYABLE represents the increase in accounts payable; ΔOTHERATSLBS represents the decrease/increase in other assets/liabilities in balance sheet items related to operations besides the other items mentioned; ΔINVESTMENTS represents short term investments change as the decrease in investments minus the increase in investments during the fiscal period; OTHERUSES represents all other uses of funds besides payouts to equity holders, debt reduction, investments/acquisitions or additions to fixed assets. OPERATIONS: When operating cash flow (OCF) results in a negative number this is considered a use of cash for the purpose of covering an operational shortfall. OCF = OPCASH RECEIVABLES INVENTORY PAYABLE (7) TAXES OTHERATSLBS EXCHANGE Where OPCASH represents the net cash from operations of the company; EXCHANGE represents the effect of exchange rates on the cash flow of the company when translating from one currency to another. 5 Detailed descriptions of the component variables are available in the appendix. 17

18 PAYOUT: The deviation from the previous year s payout. PAYOUT t = (DIVIDENDS t + REPURCHASES t ) DIVIDENDS t 1 (8) Where the expected payout at year t is equal to the dividend from the previous year. This measures the payout in the year of the PLI as a deviation from the pre-jump dividend. Once we have calculated the 4 uses of cash in equations (5)-(8), we divide each item by the change in total debt to identify the specific use of funds as a percentage of the new debt the firm has taken on. We sum the percentages from each of the 4 categories to determine the proportion of the debt issuance that is captured; it is this number that must be between 80% and 200% as described in criterion 5 for identifying a PLI. Sums greater than 100% are common since firms almost always have additional sources of funds besides the debt issuance (equity issuances, retained earnings, etc.). We divide each category s proportion of debt issuance by the total percentage identified and if a category comprises greater than 50% of the total percentage captured (not the proportion of new debt) then that category is labeled as the primary use of funds for the PLI. If no category meets this condition, then the PLI is labeled as MULTIPLE. As an example, in 2008 the medical device manufacturer Dynotronics increased debt by $5.39M. The proceeds were used as follows: they had negative operating cash flow of $1.43M (OPERATIONS 26.5%), repurchased $0.28M worth of stock without paying a dividend (PAYOUT 5.2%), added $0.69M to working capital (WORKING CAPITAL 12.9%) and made long-term investments of $3.19M (INVESTMENT 59.2%). The four categories captured 103.8% of the debt issuance. INVESTMENT captures 59.2/103.8 = 57.0% of the total percent captures thus INVESTMENT is flagged as the primary use of funds. If the four categories capture below 80% or above 200% of the debt issuance we consider the use of proceeds too poorly identified to flag as a PLI. Additionally, if the firm has conducted a PLI during the previous 7 years we do not flag the debt increase as a PLI. In the final cross-sectional analysis, we relax these conditions to determine the frequency with which firms in a given country utilize PLIs for external finance. 6 6 The cross-sectional results are robust to using the stricter definition of a PLI as well. 18

19 Table 3 presents the percentages of total PLIs for each use of proceeds category by income groups. The USA has been separated from the HICs. Globally, the most common use of proceeds from a PLI are investments; this accounts for over 50% of the total in each group. The next most common use of PLIs, across all three groups, is additions to working capital. As a proportion of the total PLIs, firms in MICs use PLIs for working capital twice as frequently as American firms and more than 1/3 as frequently as firms in other HICs. Investments and working capital account for 80%, 87%, and 76% of all PLIs in HICs, MICs, and the USA respectively. Covering operational shortfalls is the third most common usage category and is significantly lower in MICs than in HICs or the USA. PLIs are rarely used for payouts to equity holders in any group. Figure 2 indicates that these proportions are stable for both income classifications over the period The cutoff of 2000 is chosen due to the small sample of MICs available before this, however, the proportions in HICs are stable back to the early 1990s. The stability in the use of proceeds is noteworthy in light of the volatility in total PLIs over the sample period. The dashed line shows a large increase in number of PLIs for both income groups in the period leading up to the financial crisis in 2008 where it rapidly declined, before increasing again after The preliminary evidence supports the notion that debt capacity is just as important for making investments and managing working capital in MICs as it is in HICs; it may even be more important. In addition to tracking the use of proceeds from the PLI, we also analyze whether the firm could have implemented their chosen operating policy absent the debt issuance. This allows us to distinguish whether external finance was required, as opposed using discretionary cash to cover the observed use of proceeds. Discretionary cash (DC) is defined as the sum of cash on hand at the end of the previous year and operating cash flow less the cash used for the three categories other than the primary use of funds. Excess cash (EC) is defined as discretionary cash (DC) less the cash used for the primary use of funds for the PLI. Specifically, they are defined for each category as: INVESTMENT: 19

20 DC it = CASH i,t 1 + OCF it PAYOUT it W it (9) EC it = DC it INVESTMENT it (10) WORKING CAPITAL: DC it = CASH i,t 1 + OCF it PAYOUT it INVESTMENT it (11) EC it = DC it W it (12) OPERATIONS: DC it = CASH i,t 1 PAYOUT it INVESTMENT it W it (13) EC it = DC it + OCF it (14) PAYOUT: DC it = CASH i,t 1 + OCF it INVESTMENT it W it (15) EC it = DC it PAYOUT it (16) Note that operating cash flow for the OPERATIONS category is added to discretionary cash as this is a cash inflow, while the other categories are cash outflows. Once discretionary and excess cash have been calculated they are both scaled by pro forma total assets. Pro forma total assets are defined as total assets in year t less the net debt issuance from the PLI. Thus, a negative value for EC/Assets indicates the firm would have faced a cash shortfall had they engaged in their chosen operating policy but not conducted the debt issuances. Table 4 presents the cash coverage analysis results. Nearly all of the firms conducting PLIs would have exhausted their cash and been unable to implement their primary use of proceeds without altering other financial or operating policies (i.e. issuing equity or debt, investing less, lowering payouts, or lowering changes to working capital). Approximately 9 out of every 10 firms lacked the cash to carry out their chosen policies in MICs. The figure is closer to 8 out of 10 in HICs. Excluding payouts, the median excess cash shortfall ranges from 11.7% to as high as 39.0% of pro forma assets. The deficits in excess cash are slightly higher in MICs than in HICs. However, the deficits in the USA are often more than double the deficits for both HICs or MICs. These results imply there is slightly greater reliance on PLIs to fund 20

21 operating and financial policies in MICs than in HICs, but firms in the USA are able to fund much larger deficits than firms in countries with less robust credit markets. Table 5 presents the year-by-year mean leverage and mean excess leverage (firm leverage less target leverage) for our sample in the years around the PLI event. The data indicate that leverage ratios decline in the years following the initial jump. In HICs the average firm leverage ratio doubles from 0.23 to 0.46 in the year of the PLI before declining over the 7 year tracking period to The average firm recovers about 10 points of the initial 23-point jump over the tracking period. Consistent with hypothesis 1, firms operate at higher leverage and are even slower to rebalance in MICs; the average firm increases from 0.37 to 0.59 in the year of the PLI before declining over the 7 year tracking period to Here, the average firm only recovers 8 points of the initial 22-point jump over the tracking period. The slow reduction of leverage is evident in their mean excess leverage ratios. Before the PLI, the average firm in both groups is no more levered than similar firms operating in their country and industry. However, after the jump they are both approximately 25 points above comparable firms; and after 7 years they are still 12 and 14 points more levered than similar firms in HICs and MICs respectively. Similar results are found in unreported analysis when we examine subgroups by use of proceeds. These results confirm that leverage decreases in the years after a PLI but that the reduction is neither rapid, nor does it return the firm to pre-jump leverage or target leverage. As many as 7 years after the initial jump, firms are still dramatically over-levered relative to peers. The lack of any urgency to return to target is striking in MICs where they are already operating with high leverage ex ante. After the initial jump to in year 0, the average MIC firm actually increases leverage further in years 1 and 2 reaching a peak of This surprising result implies that the target is at best a second order concern in regards to the firm s financial policy and capital structure decisions. To investigate whether firms rebalancing efforts are affected by deficits and surpluses we define financial surplus as FS it = OCF it DIV i,t 1 INVESTMENT it W it + CASH it (17) 21

22 where OCF is operating cash flow, DIVt-1 is the previous year s dividend payment, INVESTMENT is the net investment, and ΔW is the change in working capital, and ΔCASH is the change in cash and short-term investments. This formulation takes operating cash flow, working capital changes (other than changes in cash) and net investment as given and identifies the cash that a firm can pay out the equity holders, use to reduce debt, or stock pile as cash reserves as the financial surplus. The formulation uses the prior year s dividend instead of the current year because, consistent with our earlier definition of PAYOUT, we view a dividend increase as a discretionary use of the surplus. 22 To allow for a direct comparison to leverage, we scale financial surplus (FS) by the market value of assets to create the scaled financial surplus (SFS). SFS it = FS it MASSETS it (18) Table 6 presents evidence comparing the relative importance of cash flow versus rebalancing toward a stationary target as determinants of capital structure decisions. The table presents mean and median changes in market leverage (ΔMLEV) and scaled financial surplus (SFS) for event years +1 to +7. We partition the sample based on surplus and deficit firm-years, and according to whether the firm is above or below their estimated long run target leverage. Because the tracking period is immediately after a PLI event that, by construction, drove the firm a minimum of 10 points above their target, the majority of observations are above target. The evidence indicates there is a dichotomous approach to managing capital structure in the years after a PLI. Across all country groups, the median firm realizing a deficit while above target increases leverage by in the HIC group, in the MIC group, and in the USA. The median firm realizing a deficit while below target actually decreases leverage by in the HIC group, in the MIC group, and in the USA. Thus, firms running deficits after a PLI event manage their capital structures in one of two distinct ways: cover deficits with additional debt despite already being above target; or aggressively reduce leverage even once they re below target, despite running deficits. The majority of observations in each country group fall into the former category, indicating that already over-levered firms are willing take on additional debt and remain above target for

23 long periods. Consistent with hypothesis 2, the distinction between these two approaches is sharpest in the MIC group where firms below target exhibit much larger leverage reductions than the other country groups. Firms generating financial surpluses tend to decrease leverage. However, the magnitude of the decrease is not symmetric between firms above target and below target. The median firm generating a surplus while above target reduces leverage by just in the HIC group, in the MIC group, and in the USA. On the other hand, the median firm generating a surplus while below target reduces leverage by a much larger in the HIC group, in the MIC group, and in the USA. Thus, firms generating surpluses after a PLI event also manage their capital structures in one of two distinct ways: using surpluses to make small, incremental reductions in leverage despite being above target; or aggressively reducing leverage even once they re below target. Once again, the majority of observations fall into the former category, indicating that already over-levered firms are willing to remain above target for long periods and would prefer to use their surplus funds for purposes besides debt reduction to a stationary target. The distinction between these two approaches is, again, sharpest in the MIC group where the median below target firm reduces leverage by nearly 11 times as much as the median above target firm. The firms aggressively reducing leverage with their surpluses in MICs, even while below target, are consistent with hypothesis 2. However, large fractions of firms in both country groups continue to be above target and only make minor reductions in leverage. This is neither consistent with them trying to regain their debt capacity, nor get back to a stationary target. The qualitative results from this analysis, in which below target firms reduce leverage more aggressively than above target firms when they realize a deficit or a surplus, are supported by two additional points. First, the differences are even greater if you consider the mean leverage changes instead of the median. The mean increases (decreases) in leverage for firms running a deficit while above (below) target are even greater. In unreported analysis we also looked at quartiles and found qualitatively identical results; this asymmetric effect is seen in the entire distribution rather than just means or medians. Second, these effects are not driven by differences in the magnitude of the scaled financial surplus (SFS). The surpluses between the above and below target groups are comparable at 23

24 both the mean and median and would not be enough to explain the large differences in capital structure adjustments. The only firms that move toward their target in the 7 years after a PLI event are the firms generating a surplus while above target; and they make small, incremental reductions. This group s slow rebalancing, and the further deviation from target in the other groups is contrary to a policy of active rebalancing toward a stationary target as a first-order determinant of capital structure decisions. It is more likely that the realization of a surplus or deficit, combined with a firm s debt capacity are more important considerations for managing capital structure. This would be consistent with the dichotomous behavior observed in Table 6. Firms with a high debt capacity will only slowly reduce leverage or may even increase it further in a ratchet effect of leverage. Firms with a lower debt capacity will need to pay down debt rapidly and even get below target so they do not face financial distress costs and will be able to raise funds in the future if needed. Debt capacity driving capital structure decisions would also be consistent with the observation that this pattern was strongest in the MICs. Firms in this group operate in less developed financial markets with less opportunity to raise external funds, particularly equity. Decreasing leverage through asset sales may also be difficult if asset markets are illiquid. Under these conditions, firms with high debt capacity will maintain high leverage and possibly increase it further if selling assets/equity is costly or unfeasible. Firms with low debt capacity will place an emphasis on paying down debt, particularly after a PLI, in case they need to raise external funds in the future. In other words, the observed patterns should be magnified in MICs where deep, liquid financial markets are not typical. In Table 7 we evaluate the effects of financial surpluses on change in debt scaled by the market value of assets controlling for other determinants of capital structure using OLS regressions. The observations are restricted to only the firm-years where the firm generated a financial surplus (where SFS is positive). We partition the sample by country groups as well as event years and non-event years among firms that have conducted a PLI. We also examine a third subsample of firms who have never conducted a PLI. ABOVE is and indicator variable taking a value of 1 when the firm is above target leverage and 0 when it is below. 24

25 The coefficient on SFS is negative and significant across all specifications and larger in MICs than in HICs for each comparable subsample. It is also much larger in magnitude than any of the other common determinants of capital structure. This suggests the realization of financial surpluses plays a critical role the decision to retire debt. It is interesting to note the coefficient is largest in magnitude in both MICs and HICs in the years after a PLI and significantly smaller in years in which a PLI hasn t been conducted recently. This implies that firms place greater emphasis on using surpluses to retire debt in the years after a PLI than they do in other years. Further, firms that never conducted a PLI have an even smaller, but still significant, coefficient on SFS in both MICs and HICs. This is consistent with the notion that firms that do not conduct PLIs do not use surpluses to pay down debt as aggressively as firms that do conduct PLIs, and in particular firms that have recently conducted a PLI. In other words, preserving or regaining unused debt capacity is a larger concern for firms that frequently use that capacity by conducting PLIs. The effect of SFS on the change in total debt is more than twice as strong in event years following a PLI than it is in non-event years in MICs; it is slightly less than double in HICs. The link is also the strongest in event years in MICs where the coefficient on SFS is a whopping The adjusted R-squared is much higher in the MIC event years than the other specifications. To get a sense of the economic magnitude of these results we can use the average SFS for an above target firm running surpluses in each country group reported in Table 6 to get the predicted decrease in leverage for each country group based on the coefficients on SFS, ABOVE, and the interaction term in Table 7. Performing this calculation, we find that an above target firm generating an average surplus that committed a PLI in the last 7 years will be expected to reduce leverage by 0.05 in the HIC group and by 0.08 in the MIC group. If they have not committed a PLI in the last 7 years, the numbers are lower but the order remains the same. This confirms the result that the link between financial surpluses and the decision to retire debt is strongest in the years following a dramatic debt increase, and that debt capacity and surplus realization play a magnified role in capital structure decisions in MICs with less developed financial markets. The coefficient on ABOVE is insignificant in both MICs and HICs in the 7 years after a PLI. It is only significant in HICs in non-event years and in MICs for the subsample of firms 25

26 that have never conducted a PLI. In both cases, it is economically small and positive, indicating that once surplus cash flow is accounted for, being above target leverage is actually associated with deviating further away from the target rather than rebalancing. This is the opposite of what we would expect if firms were rebalancing toward a target. The interaction term is also insignificant in all specifications except for the subsample of firms that have never conducted PLIs in MICs. This indicates that, in this subsample, the link between surplus cash and debt retirement is stronger for firms that are above target. Overall, this evidence is very consistent with hypothesis 2 that cash flows and debt capacity are driving capital structure decisions rather than deviations from target. In Table 8, we rerun the analysis but restrict our attention to event years where the firm runs a financial deficit. The results are strikingly different from surplus years and across the various subsamples. Since a deficit year results in negative SFS, a positive coefficient on SFS or the interaction term indicates larger deficits are associated with debt reduction, while a negative coefficient indicates larger deficits are associated with increases in total debt. In both HICs and MICs we see further evidence for the dichotomous approach to capital structure amongst firms running deficits after a PLI event. The coefficient on SFS positive and economically large in both the MIC and HIC event-year subsamples; although it is only statistically significant in the HIC group. This is consistent with firms reducing their debt despite running deficits. The interaction term of SFS*ABOVE is negative and much larger than the coefficient on SFS for both groups. The interpretation is that firms running deficits and already above their target leverage are covering their deficits with additional debt. Thus some firms, likely with low debt capacity, aggressively pay down their debt after a PLI despite running deficits. While other firms, likely with high debt capacity, continue to take on additional debt. The wedge between these two approaches is greatest in MICs where the coefficient on SFS goes from to an enormous and negative on the interaction term. The adjusted R-squared is, once again, much higher in this subsample than the others. We can also determine economic magnitude of these results using the average SFS for an above and below target firm running deficits in each country group reported in Table 6 to get the predicted change in leverage for each country group based on the coefficients on SFS, ABOVE, and the interaction term in Table 8. For an above target MIC firm running an average 26

27 deficit the expected change in debt is 0.065; they are expected to increase leverage by more than 6 points despite already being over-levered. On the other hand, a below target MIC firm running an average deficit is expected to decrease leverage by despite already being below target and running deficits. These are drastically different responses and cause firms to deviate further from predicted targets in the years after a PLI. Similarly, an above target HIC firm running an average deficit will increase leverage by while a below target firm will decrease leverage by In both HIC cases firms are deviating further away from target in the years after a PLI. They either take on additional debt to cover deficits or aggressively pay down debt beyond their estimated target. There are also interesting results in the other subsamples in Table 8. The coefficient on SFS declines in magnitude and significance in non-event years indicating the link between cash flows and changes in debt is strongest in the years after a PLI. The interaction term reduces in magnitude but remains negative and significant in all specifications. This is consistent with above target firms taking on debt to cover deficits rather than rebalancing to target. It is interesting that it is negative and significant for both firms that conduct PLIs as well as those that do not. It is also larger in magnitude in each of the MIC subsamples than it is in the HIC subsamples, which would be consistent with debt capacity playing a larger role in capital structure decisions in MICs. Lastly, the coefficients on SFS actually reverse and become negative and significant for both MIC and HIC subsamples of firms that never conduct a PLI. This demonstrates how financial policy is different in the years after a PLI than it is in non-event years or for firms that never conduct a PLI. After a PLI a lot of the firm s debt capacity has been used and many firms aggressively pay down the debt regardless of surpluses or deficits. However, for firms that do not conduct PLIs they, presumably, still have ample debt capacity and so have no problem covering deficits with additional debt. This is why we see negative and significant coefficients on SFS for the No PLI sample in both HICs and MICs. The coefficient is nearly double the size in MICs consistent with the notion debt capacity is more important to these firms. Overall, the results in Tables 7 and 8 present strong evidence in support of hypothesis 2: that cash flows and debt capacity dominate target leverage in capital structure decisions. Once cash flows have been accounted for being above or below target appears to have no 27

28 impact on the changes to debt levels for firms running surpluses. Firms running deficits appear to either take on additional debt to cover them, despite being well above target already, or they aggressively pay down their debt, despite running deficits, to regain unused debt capacity. Consistent with the idea that debt capacity is most important for firms in countries with high equity costs, our results are strongest in MICs. In the next section we will examine the frequency with which firms conduct PLIs to compare their relative importance as a method of external finance across the countries in our sample. V. Relative Frequency of PLIs around the World Our findings thus far indicate that after a PLI firms remain highly levered and may even increase leverage further in subsequent years. There is also a smaller portion of firms that rapidly decrease leverage to levels well below estimated targets. Both scenarios run contrary to models of capital structure that assume firms actively rebalance toward stationary long-run targets. They are consistent with the interpretation that debt capacity and cash flows are more relevant determinants of capital structure, and that the resulting evolution of leverage can be explained by the realization of financial surpluses rather than rebalancing. We have also shown the link is strongest in middle income countries with underdeveloped financial markets which make leverage reducing equity issuances and asset sales difficult. In light of these facts, we expect that firms in countries with under developed financial markets will conduct relatively more PLIs than firms in countries with more abundant sources of external finance. To explore this, we construct a variable representing the relative frequency with which firms conduct PLIs in a given country. We relax criteria 4 and 5 regarding clustering of PLIs and identifying the use of funds on the statement of cash flows. 7 We do this because we no longer wish to identify the primary motivation of a PLI but rather the frequency with which they are conducted. We sum the number of PLIs conducted by firms in a given country and divide this total by the total number of firm-years in that country, excluding firms that never conducted a single PLI. This number represents the frequency with which firms 7 The cross-sectional results are qualitatively identical if we maintain these restrictions. 28

29 conduct PLIs in a country, conditional on them having done at least 1. Put differently, we are measuring how frequently firms are returning to drink from the well of large debt issuances to quench their thirst for external finance within each country. Figure 3 presents our measure of the relative frequency of PLIs plotted against country level factors that proxy for financial market development, the quality of legal/political institutions, and competitive product markets. To proxy for financial market development we use the natural logarithm of the ratio of private credit markets to GDP in 2000 provided by the World Bank. There is a strong negative association between PLI frequency and credit market development. On the far left, the 4 countries whose firms conduct the most PLIs are also the 4 with the smallest credit markets relative to their overall economies: Russia, Vietnam, Poland, and Turkey. Nearly one in every four firm-years in Russia and Vietnam can be classified as a PLI (conditional on firms conducting at least one PLI). On the far right, Japan, Switzerland, and the United States have the largest credit markets relative to the size of their overall economies and firms conducting PLIs here do so relatively less than the other countries in our sample. Similar results are presented in the next panel displaying property rights on the x- axis. The intuition for the negative association between property rights and PLI frequency is that property rights are necessary for the development of financial markets. If the equity holders or debt holders do not have sufficient rights then financial markets will be illiquid and underdeveloped due to lack of investor demand for corporate debt/equity. In these markets, PLIs will be relatively more common due to debt (most likely private credit) being the only option for external finance of large projects. Similarly, if enforcement of these rights is weak, ineffective, or plagued by corrupt institutions then equity will also be off the table as a source of external finance. This is exactly what we see in the third panel where more corrupt institutions are associated with more frequent PLIs. Finally, the last panel shows that product market competition is associated with less frequent PLIs. The two countries on either end of the spectrum really tell the story: Russiaan economy characterized by quasi-state owned monopolies that stifle competition and discourage innovation rely on PLIs more frequently than any other country except Vietnam; 29

30 at the other end is the United States- a country founded on competitive free markets where firms rely relatively less on PLIs. Firms in economies with little competition may conduct more PLIs simply because they are the only game in town for lenders, so credit is cheap for them. The implicit backing of the government in quasi-state owned corporations lowers any potential costs of financial distress and lowers the cost of debt even further. Table 9 presents OLS regressions using the relative frequency of PLIs as the dependent variable and the country factors as the explanatory variables. The regression results confirm the graphical analysis from Figure 3. All four factors have coefficients that are negative and significant. The adjusted R-squared values are also fairly high with the highest being for Log(Debt/GDP). Column (5) presents a horse race between all four factors: Log(Debt/GDP) and Property Rights contribute the most to explaining the variation in relative frequency of PLIs. Much of this variation is explained by these four factors as the adjusted R-squared for this specification Overall, these results provide strong evidence in support of hypothesis 3: that firms rely on PLIs more in countries with underdeveloped financial markets and weaker legal institutions (and to a lesser extent weak product market competition and corrupt political institutions). VI. Conclusion In this paper, we study large debt-driven increases in leverage across 30 high income and middle income countries. These proactive leverage increases represent large deviations from estimated long-run target leverage ratios for the purposes of long term investments or additions to working capital. We find they are a common method of raising external finance in both high income and middle income countries, and that firms in both groups do not behave as if rebalancing toward a target is a first-order concern. Instead, the evolution of leverage is determined by cash flow realizations: surpluses are used to retire debt while deficits are covered with additional debt, regardless if the firm is above or below target. In the deficit group there is a subsample of firms that aggressively pay down debt even once they are below target to regain unused debt capacity. These patterns are strongest in middle income countries with less developed financial markets, weaker property rights, and weaker institutions. We argue that these the institutional and capital market conditions of these 30

31 countries are likely to result in high agency costs of equity and so debt and unused debt capacity is more important and frequent source of financial flexibility. Our analysis is also a test of a dynamic model of capital structure presented by DeAngelo, DeAngelo, & Whited (2011). In their model, volatility in cash flows and investment opportunities creates the need for financial flexibility to avoid underinvestment. Equity as a source of financial flexibility is costly because of an adverse selection problem in which informationally advantaged managers can sell equity securities above their intrinsic value. Cash as a source of financial flexibility is costly because of agency costs that allow informationally advantaged managers to over-retain resources for self-interested projects. The optimal contract that solves these problems is a debt contract and the cost of using debt is the possibility of not being able to issue debt on favorable terms in the future. In the context of our study of PLIs, debt capacity is a vital source of financial flexibility in the DDW model and will be more aggresively used and preserved in countries that have higher agency costs of equity and cash. The initial evidence from the debt ratio dynamics after a PLI support this hypothesis; firms in MICs place a greater emphasis on using surpluses to pay down debt after a PLI; while those running deficits after a PLI either aggressively pay down their debt going below their long-run targets, or they continue to take on additional debt to cover the deficits. This is consistent with debt capacity driving capital structure decisions and it being a more important determinant in countries where equity is costlier. Further, we conduct a study of the frequency with which firms conduct PLIs, conditional on them having conducted at least 1. We find that firms in MICs conduct relatively more PLIs than firms in HICs. We also find that the cross-sectional variation in PLI frequency at a country-level is highly negatively correlated with capital market development and property rights; and to a lesser extent product market competition and corruption. Firms in MICs with less developed capital markets and weaker property rights rely on PLIs and debt capacity as a source of financial flexibility even more than firms in HICs where the agency costs of cash and equity are not as high. Our results highlight the importance of country factors in determining the costs and benefits of different sources of finance, in determining financing frictions, and ultimately setting corporate financial policy. 31

32 32

33 REFERENCES Alter, A., & Elekdag, S. (2016). Emerging market corporate leverage and global financial conditions. IMF Working Paper. Bonaime, A. A., Hankins, K. W., & Harford, J. (2014). Financial flexibility, risk management, and payout choice. The Review of Financial Studies, 27(4), DeAngelo, H., & DeAngelo, L. (2007). Capital structure, payout policy, and financial flexibility. Working Paper at the Marshall School of Business of the University of Southern California. DeAngelo, H., DeAngelo, L., & Whited, T. M. (2011). Capital structure dynamics and transitory debt. Journal of Financial Economics, 99, Denis, D. J. (2011). Financial flexibility and corporate liquidity. Journal of Corporate Finance, 17(3), Denis, D. J., & McKeon, S. B. (2012). Debt financing and financial flexibility evidence from proactive leverage increases. The Review of Financial Studies, 25(6), Djankov, S., La Porta, R., Lopz-de-Silanes, F., & Shleifer, A. (2002). The regulation of entry. The Quarterly Journal of Economics, 117(1), Fama, E. F., & French, K. R. (2002). Testing trade-off and pecking order predictions about dividends and debt. The Review of Financial Studies, 15(1), Foley, C. F., & Manova, K. (2015). International trade, multinational activity, and corporate finance. Annual Review of Economics, 7, Frank, M. Z., & Goyal, V. K. (2009). Capital structure decisions: Which factors are reliably important? Financial Management, 38, Graham, J. R., & Harvey, C. R. (2001). The theory and practice of corporate finance: Evidence from the field. Journal of Financial Economics, 60(2), Hovakimian, A. (2004). The role of target leverage in security issues and repurchases. The Journal of Business, 77(4), Institute for International management Development. (2001). World competitiveness Report. Lausanne, Switzerland: IMD. Jaffee, D. M., & Russel, T. (1976). Imperfect information, uncertainty, and credit rationing. The Quarterly Journal of Economics, 90(4),

34 Jagannathan, M., Stephens, C. P., & Weisbach, M. S. (2000). Financial flexibility and the choice between dividends and stock repurchases. Journal of Financial Economics, 57, Jensen, M. C. (1986). Agency costs of free cash flow, corporate finance, and takeovers. The American Economic Review, 76(2), Kahl, M., Shivdasani, A., & Wang, Y. (2015). Short-term debt as bridge financing: Evidence from the commercial paper market. The Journal of Finance, 70(1), La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (2008). The economic consequences of legal origins. Journal of Economic Literature, 46(2), La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (2013). Law and Finance After a Decade of Research. In G. M. Constantinides, M. Harris, & R. M. Stulz (Eds.), Handbook of the Economics of Finance (Vols. Volume 2, Part A, pp ). Elsevier. Lins, K. V., Servaes, H., & Tufano, P. (2010). What drives corporate liquidity? An international survey of cash holdings and lines of credit. Journal of Financial Economics, 98, Marchica, M.-T., & Mura, R. (2010). Financial flexibility, investment ability, and firm value: Evidence from firms with spare debt capacity. Financial Management, 39(4), Modigliani, F., & Miller, M. H. (1958). The cost of capital, corporation finance and the theory of investment. The American Economic Review, 48(3), Myers, S. C. (1984). The capital structure puzzle. The Journal of Finance, 39(3), Myers, S. C., & Majluf, N. S. (1984). Corpoate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics, 13, Stein, J. C. (2003). Agency, information and corporate investment. In G. M. Constantinides, M. Harris, & R. M. Stulz (Eds.), Handbook of the Economics of Finance (pp ). Stiglitz, J. E., & Weiss, A. (1981). Credit rationing in markets with imperfect information. The American Economic Review, 71(3), Sufi, A. (2009). Bank Lines of Credit in Corporate Finance: An Empirical Analysis. The Review of Financial Studies, 22(3), Welch, I. (2004). Capital Structure and Stock Returns. Journal of Political Economy, 112(1),

35 Table 1 PLIs by Year and Country The sample includes 7,639 firm-year observations between 1981 and 2015 from the WorldScope Database. A proactive leverage increase (PLI) is identified as a firm year in which the firm increases its leverage by at least 0.10 to a level at that is at least 0.1 above its estimated long-run target and for which 90% of the leverage shift is driven by a change in debt as opposed to a decline in equity value. Market Leverage is defined as total debt divided by market value of assets. We must also be able to identify at least 80% of the increase in debt on the statement of cash flows so we may track the use of funds. Subsequent PLI s in years 1-7 following an initial PLI are not counted as an additional PLI thus we are underestimating the frequency with which this activity occurs by counting clustered events as a single event. Countries for which we observe less than 20 PLI are excluded. Year High Middle United Income Total Income Income States Group Country #PLI H AUS M BRA H CAN H CHE M CHN H DEU H FRA H GBR H GRC H HKG M IDN M IND H ISR H ITA H JPN H KOR M LKA M MYS H NOR M PAK H POL M RUS H SGP H SWE M THA M TUR H TWN USA 2, M VNM M ZAF Total 3,622 1,609 2,408 7,639 Total 7,639 35

36 Table 2 Description of Leverage Changes The sample includes 7,639 firm-year observations with 3,622 observations from High Income countries, 1,609 observations from Middle Income countries, and 2,408 observations from the United States between 1981 and 2015 in the Worldscope Database. A proactive leverage increase (PLI) is identified as a firm year in which the firm increases its leverage by at least 0.10 to a level at that is at least 0.1 above its estimated long-run target. Market Leverage is defined as total debt divided by market value of assets. We compare market leverage for the fiscal year ending just prior to the jump year (pre-jump leverage) with that for the fiscal year of the jump (post-jump leverage). Excess leverage is defined as the difference between the observed leverage and the firm s estimated target leverage. Countries for which we observe less than 20 PLI are excluded. Lower Quartile Median Mean Upper Quartile High Income Pre-jump Leverage Post-jump Leverage Pre-jump Excess Leverage Post-Jump Excess Leverage Middle Income Pre-jump Leverage Post-jump Leverage Pre-jump Excess Leverage Post-Jump Excess Leverage United States Pre-jump Leverage Post-jump Leverage Pre-jump Excess Leverage Post-Jump Excess Leverage

37 Table 3 Use of Proceeds The sample includes 7,639 firm-year observations between 1981 and 2015 from the WorldScope Database. Income classifications come from the World Bank. A firm is categorized into a primary motivation if the majority of the cash proceeds from the debt issuance can be traced to a particular use. INVESTMENT captures capital expenditures and cash acquisitions, the WORKING CAPITAL motivation captures increases in working capital the PAYOUT motivation captures dividend increases and repurchases, and the OPERATIONS motivation captures operational cash shortfalls resulting primarily from earnings shocks. The MULTIPLE category includes those firms for which no single motivation dominates. Countries for which we observe less than 20 PLI are excluded. TOTAL INVESTMENT WORKING CAPITAL OPERATIONS PAYOUT MULTIPLE High Income 3, % 26.8% 14.7% 1.1% 3.8% Middle Income 1, % 35.8% 9.4% 0.1% 3.0% United States 2, % 17.3% 15.4% 4.8% 3.9% 37

38 Table 4 Cash Coverage Ratios by Use of Proceeds The sample includes 5,029 firm-year observations between 1981 and 2015 from the WorldScope Database. A firm is categorized into a primary motivation if the majority of the cash proceeds from the debt issuance can be traced to a particular use. Discretionary cash (DC) is defined as cash plus operating cash flow less uses of cash outside the primary motivation. It measures the amount of cash available to the firm to execute a change in their respective motivation category. Excess cash (EC) is defined as DC less the amount of cash used for their primary motivation. Both figures are scaled by pro forma total assets, defined as total assets less net debt issuance. Countries for which we observe less than 20 PLI are excluded. Firm years for which we are unable to calculate excess or discretionary cash because of missing data are excluded. # PLI Median DC/Assets Median EC/Assets # PLI w/ EC < 0 % EC < 0 High Income INVESTMENT % WORKING CAPITAL % OPERATIONS % PAYOUT % Middle Income INVESTMENT % WORKING CAPITAL % OPERATIONS % PAYOUT % United States INVESTMENT % WORKING CAPITAL % OPERATIONS % PAYOUT % 38

39 Table 5 Evolution of Leverage The sample includes 7,639 firm-year observations between 1981 and 2015 in which the firm increases its market leverage by at least 0.1 to a level that is at least 0.1 above its estimated long-run target. A firm is categorized into a primary motivation if the majority of the cash proceeds from the debt issuance can be traced to a particular use. Market leverage is defined as total debt over the sum of total debt and market value of equity. Excess leverage is defined as the difference between the observed leverage and the firm s estimated target leverage. The table reports leverage ratios for up to 4 years before the jump and 7 years after the jump. Countries for which we observe less than 20 PLI are excluded. Event Time Mean Leverage High Income Middle Income United States Mean Excess Leverage Obs. Mean Leverage Mean Excess Leverage Obs. Mean Leverage Mean Excess Leverage , , , , , , , , , , , , , , , , , , , , , , , , , , , , , Obs. 39

40 Table 6 Leverage Changes and Realization of a Financial Surplus or Deficit The sample includes 21,471 firm-year observations between 1981 and Market leverage is defined as total debt over the sum of total debt and market value of equity. Change in market leverage (ΔMLEV) is the difference between leverage in the observation fiscal year and leverage in the year immediately preceding it. Excess leverage is defined as the difference between the observed leverage and the firm s estimated target leverage. When excess leverage takes a positive (negative) value, the firm is labeled ABOVE (BELOW) TARGET. Financial surplus (FS) is defined as operating cash flow less prior-year dividend less net investment less non-cash changes in working capital. Positive (negative) values of FS are labeled SURPLUS (DEFICIT). Scaled financial surplus (SFS) is defined as FS divided by the sum of total debt and market equity. SFS is winsorized at the 1st and 99th percentiles Event Years 1-7 High Income Middle Income United States Deficit SFS<0 Surplus SFS>0 Deficit SFS<0 Surplus SFS>0 Deficit SFS<0 Surplus SFS>0 Mean ΔMLEV ABOVE TARGET Median ΔMLEV Mean SFS Median SFS Observations 4,753 4,192 2,803 1,648 1,787 1,571 Mean ΔMLEV BELOW TARGET Median ΔMLEV Mean SFS Median SFS Observations 1,151 1,

41 Table 7 Debt Changes in Surplus Years The table presents different sub-samples for firm-years where SFS is positive; the firm is generating a financial surplus. Financial surplus (FS) is defined as operating cash flow less prior-year dividend less net investment less non-cash changes in working capital. Scaled financial surplus (SFS) is defined as FS divided by the sum of total debt and market equity. The dependent variable is scaled change in total debt (SΔTD) and is defined as the change in total debt at t-1 divided by market assets at t. When excess leverage (the difference between observed and estimated target leverage) takes a positive value, the dummy variable ABOVE takes a value of 1. Event years are the 7 years after, but not including, the year of a PLI. Non-event years are other firm-years but exclude firms who never conducted a PLI. No PLI is the subsample of firms that never conducted a PLI. All ratios are winsorized at the 1 st and 99 th percentiles. Standard errors are corrected for heteroscedasticity and T-statistics are in parentheses. Coefficients are significant at 10%, 5%, and 1% when marked with*, **, and *** respectively. Event Years 1-7 High Income Non-Event Years No PLI Event Years 1-7 Middle Income Non-Event Years No PLI SFS *** *** *** *** *** *** (-10.52) (-11.92) (-16.54) (-9.44) (-7.33) (-11.29) ABOVE *** ** (1.15) (4.48) (1.17) (-0.39) (1.15) (2.04) SFS*ABOVE * (1.08) (1.20) (-1.31) (1.57) (1.64) (-1.87) EBITDA/ASSETS ** * (-0.17) (-0.35) (1.58) (-0.21) (-2.37) (-1.78) Log(RETURN) *** *** * ** (-1.06) (-3.45) (-6.30) (-1.96) (0.12) (-2.17) Log(ASSETS $U.S.) 0.004*** 0.005*** 0.001*** 0.014*** 0.009*** 0.007*** (3.28) (7.04) (3.87) (5.33) (5.63) (13.47) MARKET-TO-BOOK *** 0.002*** *** 0.001** (1.03) (3.42) (6.65) (0.42) (2.58) (1.97) PPE/ASSETS * *** *** *** *** (-1.95) (-8.04) (-16.10) (-1.01) (-6.27) (-11.05) CONSTANT *** *** *** *** *** (-2.96) (-5.18) (-0.55) (-4.51) (-4.67) (-10.48) Observations 3,247 10,571 29,169 1,199 2,532 9,851 Adjusted R

42 Table 8 Debt Changes in Deficit Years The table presents different sub-samples for firm-years where SFS is negative; the firm is running a financial deficit. Financial surplus (FS) is defined as operating cash flow less prior-year dividend less net investment less non-cash changes in working capital. Scaled financial surplus (SFS) is defined as FS divided by the sum of total debt and market equity. The dependent variable is scaled change in total debt (SΔTD) and is defined as the change in total debt at t-1 divided by market assets at t. When excess leverage (the difference between observed and estimated target leverage) takes a positive value, the dummy variable ABOVE takes a value of 1. Event years are the 7 years after, but not including, the year of a PLI. Non-event years are other firm-years but exclude firms who never conducted a PLI. No PLI is the subsample of firms that never conducted a PLI. All ratios are winsorized at the 1 st and 99 th percentiles. Standard errors are corrected for heteroscedasticity and T-statistics are in parentheses. Coefficients are significant at 10%, 5%, and 1% when marked with*, **, and *** respectively. Event Years 1-7 High Income Non-Event Years NO PLI Event Years 1-7 Middle Income Non-Event Years NO PLI SFS 0.140** *** *** (2.23) (1.53) (-3.38) (1.46) (-1.18) (-3.69) ABOVE 0.045*** 0.032*** 0.024*** * 0.028*** 0.009** (5.00) (8.11) (12.17) (-1.66) (4.10) (2.57) SFS*ABOVE *** *** *** *** *** *** (-4.50) (-4.92) (-3.20) (-4.59) (-4.92) (-8.32) EBITDA/ASSETS 0.012*** (6.75) (0.80) (1.62) (1.44) (0.82) (-0.47) Log(RETURN) * *** (1.61) (0.41) (1.84) (1.29) (0.62) (3.36) Log(ASSETS $U.S.) 0.013*** 0.006*** 0.006*** 0.005*** 0.003** 0.005*** (8.08) (8.24) (17.89) (3.32) (2.24) (10.30) MARKET-TO-BOOK ** *** ** ** *** (-1.19) (-2.24) (-4.12) (-2.23) (-2.00) (-9.53) PPE/ASSETS *** 0.018*** ** 0.024*** (1.05) (4.59) (7.72) (-0.49) (2.55) (5.32) CONSTANT *** *** *** * *** (-7.77) (-7.02) (-15.90) (-1.62) (-1.83) (-8.54) Observations 2,662 8,137 20,521 1,625 2,865 9,441 Adj. R

43 Table 9 Relative Frequency of PLIs vs Country Factors The dependent variable is the relative frequency of proactive leverage increases (PLIs) in each of the 30 countries in our sample. It is defined as the number of firm-years flagged as a PLI divided by the total number of firm years for that country conditional on the firm having conducted at least one PLI. The table presents OLS estimated coefficients that are significant at 10% when marked with * at 5% when marked with ** and at 1% when marked with ***. T-statistics are in parentheses. (1) (2) (3) (4) (5) (6) (7) Log(Debt/GDP) *** ** ** ** (-3.78) (-2.49) (-2.14) (-2.32) Property Rights *** * Index (-2.93) (-1.91) (-1.07) (-0.86) Product Market *** Competition (-2.83) (-0.91) (-0.80) Corruption Index ** (-2.34) (1.59) Constant 0.247*** 0.191*** 0.340*** 0.148*** 0.381*** 0.313*** 0.245*** (8.36) (9.87) (4.73) (19.06) (5.06) (4.89) (8.24) Observations Adj. R

44 Figure 1 Number of Leverage Changes by Year and Country # PLI # PLI LKA ITA CHE GRC POL NOR BRA Proactive Leverage Increases by Country ZAF RUS TUR PAK SWE ISR IDN THA VNM FRA SGP DEU MYS HKG TWN 2408 CHN AUS KOR CAN JPN IND GBR USA* High Income Middle Income United States Proactive Leverage Increases by Year High Income Middle Income United States 44

45 Figure 2 Use of Proceeds Over Time High Income Countries Proportion Payout/Multiple Operations Working Capital Investment Total Year Middle Income Countries Proportion Operations Working Capital Investment Total Year 45

46 Figure 3 Relative Frequency of PLIs vs Country Factors 46