Annals of the University of North Carolina Wilmington International Masters of Business Administration.

Transcription

1 Annals of the University of North Carolina Wilmington International Masters of Business Administration

2 THE EFFECT OF MACROECONOMIC FACTORS ON CAPITAL STRUCTURE DECISIONS Sergey A. Chekanskiy A Thesis Submitted to the University of North Carolina Wilmington in Partial Fulfillment of the Requirements for the Degree of Master of Business Administration Cameron School of Business University of North Carolina Wilmington 2009 Approved by Advisory Committee Peter Schuhmann Ravija Badarinathi Nivine Richie Chair Accepted by Dean, Graduate School

3 TABLE OF CONTENTS ABSTRACT... v LIST OF TABLES... vi 1. INTRODUCTION LITERATURE REVIEW Tradeoff Theory Pecking Order Theory Market Timing Theory Macroeconomic influence OBJECTIVES, RESEARCH QUESTIONS AND HYPOTHESIS METHODOLOGY Target leverage estimation Firm-specific target leverage variables Macroeconomic target variables Debt-Equity choice regressions RESULTS AND DISCUSSION Target leverage estimation Macroeconomic factors Lag vs. Lead Financing Choice Regressions Pure issue Pure repurchase iii

4 5.2.3 Mixed transactions Debt maturity choice CONCLUSION AND RECOMENDATIONS REFERENCES APPENDICES iv

5 ABSTRACT I investigate the effect of macroeconomic factors on financing decisions. I support the notion that companies have target debt ratios and rebalance their capital structure accordingly. I estimate the target leverage using a set of firm-specific and macroeconomic variables. I find that macroeconomic forecast plays a very important role in the target leverage decisions. I show that such decisions are based on the medium-term economic forecast to the greater extent than on the retrospective economic information. I introduce mixed issue/repurchase choice regressions and illustrate the importance of macroeconomic factors. I find that macroeconomic factors gain exceptional significance for good and bad decisions prior a change in the state of economy and prove that winners had a better economic forecast. I highlight how the agency problem affects the capital structure of a company and find further support for the market timing theory. Finally, I estimate the maturity of debt regression and find how macroeconomic and firm-specific factors affect the maturity of debt decision. My findings support the importance of the balance between short and long term debt. I find that companies try to keep that balance and try to avoid short-term debt during downturns. v

6 LIST OF TABLES Table Page 1. Summary Statistics Determinants of target leverage Descriptive statistics for debt-equity choice Pure issue choice regression results Pure repurchase choice regression results Mixed transaction regression results Debt maturity choice regression results vi

7 1. INTRODUCTION Capital structure decisions are among the most important in financial management. As shown in studies proving the relevance of capital structure (Ibrahimo and Barros, 2006). An important question is which factors determine capital structure and what decisions are related to it. Theoretical works explore the effect of managers' preferences of internal sources of financing to external ones, the effect of the tax shield, and the costs of financial distress. Empirical studies examine endogenous factors, such as firm size and asset tangibility. Despite the fact that the importance of macroeconomic factors is recognized (e.g. Hackbarth, Miao, and Morellec, 2006), few empirical studies test for the effect of exogenous factors, particularly macroeconomic factors. Recently in The Wall Street Journal, Michael Milken points out that capital structure significantly affects both value and risk (April 21, 2009 p. A21) and defines six factors to consider when making a financing decision. These factors include the state of capital markets and the economy. Where a change in environment should signal a change in the optimal capital structure. Mr. Milken argues that during the last forty years many companies suffered because of the wrong capital structure. For example, looking back we can say that firms who repurchased their stock in 2007 instead of decreasing leverage made a terrible mistake. They entered a period of increasing credit constrains with low liquidity and extensive debt burden. Moreover, their shares dropped by more than fifty percent a year later. This leads to a conclusion that their current financial problems are most likely self-imposed. In this study I look at the decisions made prior to the huge fall of the market in the end of This paper aims to answer the following questions: First, do firms take into account macroeconomic factors when they make capital structure decisions? Second, are right and wrong (historically speaking) decisions differently affected by macroeconomic factors? To answer these questions, I first define right and wrong decisions, divide the

8 data sample accordingly and use a Logit regression model to identify determinants of that choice following Hovakimian et al., (2001). 2. LITERATURE REVIEW For the last fifty years, since the proposition of the irrelevance of capital structure (Modigliani and Miller, 1958) many studies focused on financial policy. Three main theories were developed concerning capital structure decisions: the trade-off theory, the pecking order theory and the market-timing hypothesis. There is also an agency cost theory, but its' concepts are very close to the trade-off theory and thus I do not look at it separately in this study Tradeoff Theory The trade-off theory focuses on the balance between the tax benefit of debt and costs of financial distress. Evidence supporting the trade-off theory is mixed. The trade-off theory suggests that large and profitable firms should issue more debt to decrease their tax burden. However, many studies find the opposite that higher profitability leads to lower target leverage (see Fama and French, 2002). Graham (2000) estimates the cost and benefits of debt and finds that large and profitable firms with low cost of financial distress use debt with cautious. A classic example is Microsoft, which while being very profitable, maintained zerodebt policy for years. Moreover, some evidence suggests that the target debt to equity ratio, if it exists, is not important. A survey of 392 CFOs by Graham and Harvey (2002) found that approximately half of them have a flexible leverage target or have none at all. Fama and French (2002) show that the speed of adjustment toward target leverage is slow. Nevertheless, some studies support the idea of the trade-off theory. Hovakimian, Opler, and Titman (2001), Korajczyk and Levy (2003), Hovakimian (2004), and Hovakimian, 2

9 Hovakimian, and Tehranian (2004) find evidence supporting the role of the target capital structure in security issuance and repurchasing Pecking Order Theory The pecking order theory is proposed by Myers (1984) and Myers and Majluf (1984). In their theoretical framework, there is no optimal capital structure. And even if there is an optimum, the costs of deviating from it are insignificant in comparison with costs of raising external funds. Investors are willing to buy risky securities only at a discount because of information asymmetry between managers and outside investors. To avoid that problem, managers prefer internal financing. When internal funds are exhausted, managers prefer straight debt, then convertible debt, and finally equity as a measure of last resort. However, information asymmetry is not the only possible reason for a pecking order. In 1961, Donaldson talks about transaction costs. Another reason might be the managerial optimist concerning company s stock price (Heaton, 2002). Optimistic managers always think that their stock is undervalued and thus are reluctant to issue equity. Empirical tests of the pecking order theory have mixed results. Shyam-Sunder and Myers (1999) test pecking order theory against the trade-off theory. Using a sample of 157 firms from 1971 to 1989, they find that the pecking order theory has much more explanatory power. However several researchers have questioned their sample, arguing that tests may provide misleading results when evaluating plausible patterns of external financing. Fama and French (2002) find that profitability is negatively related to leverage, consistent with the pecking order model. Seifert and Gonenc (2008) test how well pecking order behavior applies to US, UK, German and Japanese firms, using a sample of firms from 1980 to Their results are incon sistent with the pecking order model, with the exception of Japanese firms in the period from 1980 to However, later the support for the pecking order hypothesis has diminished. 3

10 Korajczyk and Levy (2003) find that firms are more likely to issue equity when the announcement effects are less negative Market Timing Theory The main difference between the trade-off theory, pecking order theory and the market timing theory is that first two theories assume semi-strong market efficiency. When market timing theory does not require market to be efficient at all. However, market timing hypothesis doesn t say that market is inefficient. Windows of opportunities exist when relative cost of equity varies over time. Market-timing theory says that managers try to time the market, which is the critical assumption for this model. In practice, it seems that CFOs are actively engaged in market timing in their financing decisions. In the survey by Graham and Harvey (2001), managers admit to trying to time the market. Two-thirds of those that considered issuing common stock agree that how much their stock is undervalued or overvalued was an important consideration. Baker and Wurgler (2002) define market-timing theory as that capital structure evolves as the cumulative outcome of past attempts to time the equity market. (p. 23) They find evidence that external finance-weighted average of historical market-to-book ratios is negatively related to current market leverage, which is interpreted as a support of market timing theory. In other words low-levered firms tend to be those that raised funds then their valuations were high. High-levered firms raised capital when their valuations were low. Moreover, the effect of fluctuations in market valuations on capital structure persists for at least a decade. Kayhan and Titman (2007) confirm that firm histories strongly influence their capital structure, though they argue with Baker and Wurgler on the persistence of the effect of market timing on capital structure over long horizons. They find evidence that over time firms tend to balance capital structures towards target debt ratios. That is consistent with the tradeoff theory of capital structure. 4

11 2.4. Macroeconomic influence The importance of macroeconomic risk is widely recognized. It is well known that when a firm s operating cash flow depends on economic conditions it should adjust its leverage in accordance with the economy s business circle phase. Hackbarth, Miao, and Morellec (2006) develop an approach to analyze the impact of macroeconomic factors on the level of credit risk and dynamic capital structure choice. Some of their findings have yet to be shown in an empirical study. Previously, exogenous factors were rarely included in empirical studies of capital structure choice. With the exception of Marsh (1982) who includes a forecast of aggregate debt and equity issue as a measure of market conditions in estimating issue choice. Bayless and Chaplinsky (1991) include a measure of equity market performance and the change in T- bill interest rate in estimating issue choice. Recently, researchers have examined the effect of macro-factors on capital structure. Korajczyk and Levy (2003) examine domestic non-financial corporate profit growth, twoyear equity market returns, and the spread between three month commercial paper and T- bills. Huang and Ritter (2004) find that real GDP growth increases the likelihood of debt issuance. However, its' relation to the likelihood of equity issuance is not clear. Drobetz and Wanzering (2006) test of macroeconomic factors on the pace of capital structure changes on the sample of 91 Swiss firms find that in good conditions (term spread is higher, economic prospects are good) the adjustment speed is higher. Haas and Peeters (2006) also find that higher GDP growth increases the adjustment speed [to target leverage] in Estonia, Lithuania and Bulgaria. The most recent study by Tang and Cook (2009) includes term spread, default spread, GDP growth, and dividend yield and looks precisely at the determinants of the adjustment speed of the capital structure towards its target. 5

12 3. OBJECTIVES, RESEARCH QUESTIONS AND HYPOTHESIS This research breaks into two main parts. The first part is dedicated to capturing the effect of different variables on leverage, selecting variables to be used in the second part, and estimating the target leverage equation. The second part is where I answer main questions of this paper. How different capital structure decisions (right and wrong) were affected by the macroeconomic factors. Consequently, were the right decisions simple luck or were they based on a good macroeconomic forecast. Previous work on the determinants of the capital structure choices and financing decisions start by analyzing drivers behind a single issue/repurchase choice, i.e. debt issue versus equity issue or debt retirement versus equity repurchase (see for example, Hovakimian et al., 2001). Who look at the influence of firmspecific and market return variables on single financing decisions. Korajczyk, Levy (2003) extend the research by adding macroeconomic variables into the model. A year later Hovakimian et al. look at determinants of dual debt-equity issues. Gaud et al. (2007) research the drivers of different financing decisions on the sample of European firms, but ignore the mixed issue/repurchase choices. The primary concern of this paper is exactly mixed issue/repurchase decisions. One of which is equity issue and debt reduction versus debt issue and equity repurchase. This transaction type has potentially the most dramatic effect on the leverage ratio and theoretically should reflect managers concerns about the optimal capital structure better than single issue/repurchase decisions. When the economy is in a good shape costs of financial distress are lower and the adjustment speed towards an optimal leverage can be done faster (Cook and Tang, 2009). However, during the recession distress costs skyrocket, there are fewer resources available, default spread is higher and the adjustment is more difficult to make. Add to that the dependency of the cash flow on the economic circle and the increased relevance on the capital structure during a recession is unquestionable. Furthermore, when the inevitability of the upcoming recession is getting clearer, it is better be 6

13 prepared before, than deal later with a whole lot of self-caused problems. Thus, adjustments made to the capital structure prior to the burst of the bubble should have, theoretically, been influenced by the macroeconomic forecast to the greater extent than those, made while the worse is over and there are no signs of a storm on the horizon. It is obvious that managers who decided to repurchase stocks, while its price was at its peak made worse decision than those, who used the moment to raise cash through equity issue and used it to prepare the firm for the recession by reducing its debt burden. The main hypothesis is that those who won made a better job with an economic forecast than those who lost, and it was not a matter of luck or a simple coincidence. Similar to debt-equity choice analysis I test how macroeconomic factors affect the debt maturity choice. This is also important, because the optimal capital structure is dependent not only on the D/E ratio, but also on the ratio between short and long term debt and the maturity of latter (Philosophov L.V., Philosophov V.L., 2005). Moreover, higher portion of a short-term debt in the capital structure increase the probability of bankruptcy, especially during a recession. 4. METHODOLOGY I use quarterly firm-specific data from 2009 Standard and Poor s Compustat database and macroeconomic data are taken from web sites of the U.S. Treasury 1 and Department of Commerce 2. I exclude financial firms (SIC between ), because their capital structures are likely to be very different from those of non-financial firms. Then I require a firm to have 3 preceding quarters of data. This measure is aimed to filter out young firms, since both firm-specific and macroeconomic factors have less predictive power concerning their capital structure decisions. Furthermore I exclude small firms, those who have less than $7 million in assets. The data sample covers 11 years from 1998 to 2008, including the effect

14 of both the dot com bubble and the beginning of the recent crisis. After trimming the data set from the presence of significant outliers by excluding the top and the bottom 0.5% of the firm-specific variables (Total Assets, Book Value, Current Liabilities, Total Debt, Long- Term-Debt, Stock Return, Sales, Tax Expense, Market-to-Book, Return on Assets) the data sample comprises 130,098 firm-quarter observations for 8,578 firms. I require SGA expenses to be positive, since they are used as normalizing variables. Summary statistics are given in the Table 1. Issue/repurchase events are defined using according data points from Compustat database and are required to be bigger than 5% of Assets to be defined as an event. Table 1 Summary Statistics Variable N Mean Std Dev Minimum Maximum Leverage SIZE TANG ROA droa CASHr dpedil dpbdil MTB RET RDr RDD SE Risk E Ind_Lev Tspread DefaultS Div_Yield Target leverage estimation The target leverage is the debt ratio that firms would choose in the absence of informational asymmetries between managers and shareholders, transaction costs, or other adjustment costs. Even though existing theories explaining firms financing decisions (pecking order, trade off, market timing) do not unanimously support the idea that firms operate around target leverage, there is evidence that target leverage do exist (Hovakimian et 8

15 al. (2001), Hovakimian et al. (2004) 3, Graham and Harvey (2001) 4. Thus, I include it into financing choice regressions. Korajczyk et al. (2003) assume that firm s actual leverage equals its target leverage plus measurement error that is orthogonal to the explanatory variables. Cook and Tang (2009) argue that there are methodological problems when using linear models for fractional data, thus, use quasi-maximum-likelihood (QMLE) estimation model to compute the fitted value of target leverage equation. They specify the target leverage as a function of prior period macro variables and firm-specific variables. Gaud et al. (2007), consistent with Hovakimian et al. (2001) use Tobit regression to determine the target leverage ratio. Following them, I use the Tobit regression model with double censoring at 0 and 1, since the leverage ratio is naturally bounded between zero and one. The following equation describes the model: where,,,, (1),,, The Target Leverage for the firm i in the year t Macroeconomic explanatory variables Firm-specific explanatory variables Vector of time dummy variables Stochastic error term This will also allow me to select significant determinants of leverage for the financing decisions Logit regressions. In particular, tested variables are as follows: 3 They found evidence that firms tend to operate in line with the static trade-off theory, offsetting previous earnings-driven decisions towards the target capital structure. And that the target D/E ratio has different importance for different financing decisions. 4 Roundabout 80% of questioned managers in their sample admit having a strict target or a set range for the D/E ratio. 9

16 4.1.1 Firm-specific target leverage variables SIZE - as a proxy of size I am using the natural logarithm of sales. This measure was previously used by Booth et al. (2001) and Gaud et al. (2007). However, many other researchers prefer using the natural log of total assets as a proxy for size. In this particular case log of total assets would create bias, because it would be highly correlated with the next coefficient. TANG - to hold for the effect of collateral I use the ratio of Tangible Assets / Total Assets, where tangible assets are defined as Net Property Plant and Equipment, following Gaud et al. (2007). ROA is defined as EBITDA / Total Assets and it serves as a measurement of profitability and to some extent is a proxy for firm s internal financing capacity (Miguel & Pindado, 2001, Gaud et al. 2007). CASH is intended to hold for the effect of accumulated financial slack. Cash ratio is defined as Cash and equivalents / Total Assets (Gaud et al., 2007). MTB - Market-to-book ratio is used as a common measure of growth opportunities. (Booth et al., 2001, Gaud et al., 2007). It is defined as a quotient of the sum of total assets and the market value of equity minus book value of equity, divided by total assets. [(Total Assets + Price * Shares Out. Book Value of Equity) / Total Assets] RET is defined as the ratio of the quarterly change in market value of equity to the market value of equity in the previous quarter. This variable is aimed to control for the stock price effects. ATA - is the ratio of Depreciation and Amortization in Total Assets, used as an explanatory variable of non-debt tax shield. Risk according to the trade-off theory higher cost of financial distress should lead to lower target leverage. Higher earnings volatility increases the probability of bankruptcy, thus, increasing the distress cost. So to control for the effect of risk I use the standard deviation of the annual difference of EBIT / Total Assets (Delcoure, 2007) over the preceding 5 years. (σ( (EBIT / Total Assets))). Cook and Tang (2009) control for firm uniqueness by introducing three variables. RD, which is R&D Expense normalized (divided) by Book Assets. RDD, dummy variable, equals 1 if a firm 10

17 reports R&D expense and 0 otherwise. SE, which equals Sales Expenses divided by Total Sales. Firms with high R&D and high sales expense are more likely to have unique assets and consequently higher costs of financial distress (Hovakimian et al., 2004). For that reason such firms might want to protect themselves with lower leverage ratios Macroeconomic target variables Term spread defined as a difference between 20 year Gov bond and three-month-tbill. Although some researchers use 3 month T-bill rate as a macroeconomic variable (Drobez and Wanzenried, 2006), some argue (Estrella, Hardouvelis, 1991) that the slope of the yield curve has more predictive power. Cook and Tang, (2009) lag this variable by one year, because it has been known as a strong predictor of a good economy (Estrella, Mishkin, 1998). Default spread Following Cook and Tang (2009) and Korajczyk and Levy (2003), and Fama and French (1989). Default Spread is defined as the difference between the average yield of Baa rated and Aaa rated corporate bond. Each rated by Moody s and with maturity of years. Fama and French (1989) show that this factor is higher during recessions and lower during expansions. In addition to that I test another variation of the Default spread, defined as the difference between Baa rated corporate bond and 20 year T-bill. This interpretation should be less biased towards the political influence on the highest rating. Because it is the fact that many Aaa rated bonds didn t actually deserve this rating, what was proven right during the GDP growth Annual percent change in GDP in constant 2000 prices. Div Yield Consistent with Drobez and Wanzenried, 2006 I take the total dividend paid on value-weighted NYSE/AMEX/NASDAQ portfolio over a year t-1 divided by the current value of the portfolio (time t). As Drobez and Wanzenried, 2006 indicate because dividends tend to be sticky, high dividend yield means portfolio value is low and thus it is a downturn. 11

18 4.2 Debt-Equity choice regressions In the second stage, to determine the drivers of the particular financing choice, I use the Logit regression model:, 1,,,,,,,,, (2), 1,,,,, The probability of a firm i, operating in time t, choosing one financing option rather than another The deviation from the target leverage Macroeconomic explanatory variables Firm-specific explanatory variables Vector of time dummy variables Stochastic error term Following Korajczyk, Levy (2003) I define a firm as issuing (repurchasing) equity (debt) when net equity (debt) issued (repurchased) for cash in the particular quarter divided by the book value of assets in the previous quarter exceed 5%. Apart from some variables discussed earlier I include specific potential determinants of the financing choice. Hovakimian et al. (2001) argue that managers are involved into the calculation of accounting numbers and are affected by them. For example, managers may be evaluated partly based on accounting numbers. Thus, accounting figures may play significant role in debt-equity decisions. Because, if a firm has low stock price relative to its earnings (P/E) or book value (P/B), issuance of equity will further decrease those ratios. To account for the effect mentioned above, I, following Hovakimian et al. (2001), include two dummy variables. dpedil is a dummy variable equals 1 when after-tax cost of debt exceeds firms E/P ratio, 0 otherwise. [E/P > rd(1 - Tc), E/P = Net Profit / Market Value of Equity] This variable 12

19 reflects if the equity issue will dilute firms earnings per share more than the debt issue. The second dummy variable dpbdil equals 1 when MTB < 1, and zero otherwise. It, similarly to the previous variable shows if the equity issue will dilute firms book value per share. droa equals 1 when ROA < 0, and zero otherwise, controlling for losses, since assets are required to be positive. AdjRET is a spread between firm year return and country return (market). TLevDif is the difference between the target leverage, estimated in the first part and the actual leverage at time t. IndLev is the mean leverage in the industry. The industry is defined as the first for numbers in the SIC code. ObsOPsize is the ratio of absolute value of net amount of the transaction to TA at the beginning of the year 5. bias. 5 Though, it is excluded for regressions related to External versus Internal financing choice, due to possible 13

20 5. RESULTS AND DISCUSSION 5.1. Target leverage estimation In order to estimate the target leverage a company would chose in absence of stockholders influence I use Tobit regression model. I regress Leverage ratio, a company has at the quarter t, on a set of firm-specific and macroeconomic variables. Hovakimian et al. (2004) use one-year-lagged firm-specific variables to determine firms target leverage, but I have found that there is virtually no difference in coefficients between regressions with lagged and regressions with actual values. Nevertheless, the overall fit of the model drastically reduces when lagged figures are used. (See Appendices: Table, Table ). ATA variable, the ratio of depreciation and amortization in total assets, has been excluded from the regression because it is not significant, not even at 10% level. Gaud et al. (2007) studying Europe got the same result for ATA variable across almost all countries. Results for the Tobit leverage estimator are presented in the Table 2. Table 2 Determinants of target leverage Parameter Estimate Parameter Estimate SIZE *** RDD *** TANG *** SE *** ROA *** GDP Growth *** CASHr *** Term Spread *** MTB *** Default Spread *** RET *** Dividend Yield *** Risk * _Sigma *** RDr *** Log Likelihood ***, **, * indicates significance at 1%, 5% and 10% level accordingly. SIZE variable enters regression with a positive sign, which supports the hypothesis that larger companies have higher debt ratios. That is because they have more stable cash flows, which reduces costs of financial distress. Moreover, bigger companies have a higher chance of exhausting the debt tax shield. This is consistent with a trade-off theory and prior studies i.e. Gaud et al., (2007), Hovakimian et al. (2004). 14

21 TANG also enters regression with a positive sign, in line with a hypothesis that tangible assets acts as collateral. Potentially, in case of default, tangible assets have higher residual values than other assets. The more there tangible assets the higher is the firm s debt capacity and debt ratios. Because debt holders have a right to request the selling of assets they will most likely prefer tangible assets as collateral. ROA and CASHr both have a negative sign in the regression. On the one hand, it is not consistent with a trade-off theory, because higher operating margins suggests more stable cash flows, increasing the chances to fully exhaust tax shield and decreases costs of financial distress. Cash, under the trade-off theory, is viewed as a negative debt, which together with higher operation margins should increase debt ratios. On the other hand, signs on those variables are consistent with pecking-order theory. Higher profitability and financial slack imply bigger capacity of internal financing, which is the number one choice under the pecking-order hypothesis. Thus, negative signs on CASHr and ROA support the peckingorder hypothesis. Moreover, Graham and Campbell, 2002 in their survey found that ~58% from 392 questioned CFOs find insufficient internal funds as a main factor to issue more debt. The negative sign on Market to Book ratio suggests that higher MTB value diminishes the residual value of assets, acting as collateral, thus increasing the costs of financial distress. The higher is MTB ratio the lower the leverage should be. This is in line with the trade-off theory. Another market-performance variable is RET. It has the same, negative, sign as a Market to Book ratio. The possible explanation lies in the Market Timing theory, which implies that managers try to time markets, issuing equity when they think their stock is overvalued (Baker and Wurgler, 2002). This is consistent with the pecking-order theory, which implies that the variation in the level of asymmetry of information between managers 15

22 and stakeholders lead to negative underinvestment cots. Thus, if managers actively try to time the market the stock price should negatively impact leverage ratios. Risk variable enter the regression with a positive sign, which is counterintuitive. According to trade-off theory higher earnings volatility increases the possibility of a default and consequently increases distress costs. This in turn should reduce leverage ratios. Compared to other variables in the model significance of the Risk variable is low, it is significant only at 10% level. This variable is excluded from the financing choice regressions. RDr, RDD and SE variables proxy for the uniqueness of company s assets. RDD is a dummy variable equals one if a company reports R&D expenses and zero otherwise. RDr and SE are ratios of R&D and SG&A expenses accordingly in total assets. RDD and SE variables both enter the regression with a negative sign, which is consistent with Hovakimian et al., (2004), who finds that firms reporting R&D expense and having high SG&A expenses are more likely to have unique assets, which are harder to sell in case of default. This increases the cost of distress and in turn according to the trade-off model reduces the target leverage ratios. On the other hand RDr enter the regression with the negative sign. This is counterintuitive according to the trade-off model, because the higher is R&D expense the higher is the distress cost. On the other hand high R&D costs mean that a firm is more likely to have unstable cash flows and requires external financing to fund its costly projects. The second reason outweighs the potential downside of debt financing Macroeconomic factors There are four macroeconomic factors included in the Tobit target leverage estimation model. In the current section I estimate the target leverage a company would have in perfect world in absence of any information asymmetries and stakeholders influence. For that reason macroeconomic variables in target leverage regressions are taken at the time t. More detailed discussion concerning lagging macro variables follows in the next section. 16

23 Term Spread is calculated as a difference between long and short term government bond yields. It enters the regression with a negative sign, and taking into account that high Term Spread is known to be strong predictor of a good economy (Estrella & Mishkin, 1998) I can conclude that it is another consequence of market timing. When prospects of economy are good, stock prices are generally raising and manage are more likely to issue equity. On the other hand when prospects of economy become gloomy it could be reasonable to issue debt, if needed, at a still lower rate than is possible during a recession. But those assumptions are yet to be tested in Logit financing choice regressions. Default Spread is the proxy for the business cycle; it is higher during recessions and lower during expansions. Default Spread is calculated as Baa Moody s rated corporate longterm bond s yield minus government twenty year bond s yield. This measure of Default Spread proved to be more significant than the difference between Baa and Aaa rated bond s yields. Moreover, the argument for using it is that Aaa rated corporate bonds in most cases were overrated and hardly represented the most financially stable companies. It enters regression with a positive sign, which can be explained with the Market Timing theory. During recessions the stock value is low and managers are reluctant to issue equity, because they think that their stock is undervalued. The opposite is true for expansions. Dividend Yield enter regression with a negative sign. It is another indicator of the state of economy. Dividend Yield is calculated as dividends paid on S&P 500 portfolio over the time t-1 divided by its current value at time t. Due to the fact that dividends tend to be sticky, higher Dividend Yield variable indicates a recession and decreasing portfolio value. The sign on this variable is consistent with the Market Timing model. GDP Growth enters regression with a positive sign. Growing economy creates conditions for more stable cash flows and lower distress costs, greater growth opportunities and higher investment needs. The sign on this variable is consistent with the Trade-off theory. 17

24 5.1.2 Lag vs. Lead Since the main purpose of this study is to capture the effect of macroeconomic variables on capital structure decisions, special attention is paid to them. Previous studies in this area mainly used lagged macroeconomic variables to describe the target leverage. The lag varied from 3 quarters (Korajczyk & Levy, 2003) to one year (Cook & Tang, 2009). Because in this paper I use quarterly data, I tested how well macro variables, lagged one, two and three quarters fit into the model. Assumption behind lagging macro variables is that macroeconomic data is not reported immediately and that could create a lagging effect on financing decisions. On the other hand, macroeconomic statistics doesn t predict the future, instead it simply describes the current situation in the economy and probably sets a short term trend. One cannot simply take a historic dataset and predict the future. There are so many different factors affecting the real economy that we can assume at least a semi-strong efficiency. This leads to an idea that it could be more beneficial to use leading macro variables, not lagged. The assumption behind this idea is that leading t+3 macroeconomic factors would be a perfect prognosis at time t. Budgeting decisions, by their nature, should be made taking in consideration economic forecasts, because their effect lasts for years. Moreover, Cook & Tang (2009) show that the rebalancing speed reduces during recessions and increases during booming economic conditions. That is why I test both three quarters lagged and three quarters leading macroeconomic variables. Results are given in Appendices. Table 4 Macro Variable 3 Quarters Lagged" presents results for models with lagged macro variables. What is notable about those regressions in that significance of Default Spread diminishes when lagging it farther back. The return variable becomes insignificant is macro variables are lagged more than one quarter back. The situation with leading variables is different. Table 3 Macro Variable 3 Quarters Forward" presents results for models with leading macro variables. Here all variables remain significant at least at 5% level no matter 18

25 whether it is t+1 or t+3 variables. Though, the significance of Default Spread increases form 5% level at t+1 to 1% level at t+2 and t+3. Significance of Term Spread decreases when leading farther in the future. This is happening because Term Spread, according to Estrella & Mishkin, 1998 already has some predictive power. Overall fit of models including leading variables is higher than including lagged variables. And the overall fit of leading variable models increases when going farther in the future. This proves the assumption that macroeconomic forecast has a significant effect on target leverage ratio. And that such forecast is more likely to be made for a medium-term period than a short-term one. 5.2 Financing Choice Regressions There are eight financing transaction I look into. Three pure financing transactions: debt issue, equity issue and issue of both, three pure payout transactions: debt retirement, share repurchase and both at the same time. Table 3 Descriptive statistics for debt-equity choice Transaction Type / Variable Debt Issue Equity Issue Debt and Equity Issue Debt Reduction Share Repurchase Debt Reduction and Equity Repurchase Debt Issue and Equity Repurchase Equity Issue and Debt Reduction No Transaction SIZE TANG ROA CASHr MTB RET RDr SE TLevDif RDD 7% 30% 12% 9% 24% 8% 10% 17% 10% dpedil 67% 79% 73% 69% 90% 81% 91% 72% 74% dpbdil 26% 7% 11% 32% 12% 31% 15% 10% 23% droa 48% 34% 33% 45% 17% 26% 14% 28% 41% T. Spread Default Spread Div. Yield GDP Growth

26 N This table represents the mean values of variables used in debt-equity choice Logit regressions. The data are from Compustat database and the sample contains all firms operated during the period from 1998 to SIZE is the natural logarithm of sales. TANG is the ration of Property Plant & Equipment in total assets.roa is NI/Assets. CASHr is the ratio of cash and equivalents in total assets. MTB is (TA + (Shares*Price) (TA TL)) / TA. RET is company s stock return adjusted for stock splits. RDr is the ratio of R&D expenses in total assets. SE is the ration of SG&A expenses I total assets. TLevDif is the distance of the actual leverage from the target leverage ratio calculated above. RDD is a dummy variable equals one of a firm reports R&D expense and zero otherwise. droa is a dummy variable equals one if ROA < 1 and zero otherwise. ppedil is a dummy variable equals one if equity issue will dilute EPS more than debt issue and zero otherwise. dpbdil is a dummy variable equals one if MTB < 1 and zero otherwise. And two mixed financing-payout transaction types: debt issue and equity repurchase, equity issue and debt retirement. I do not include dividend payout because it is not the main purpose of this study. Table 3 presents descriptive statistics for financing choices mentioned above. Bigger companies tend to be those who are issuing debt and simultaneously repurchasing equity. On average, companies with dual transactions, such as issuing both debt and equity or repurchasing debt and issuing equity, tend to be much bigger than companies only debt or equity and companies with no transactions. Debt issuers have much higher tangible assets ratio than those issuing equity. This is perfectly in line with the trade-off theory. Moreover, those who decide to retire debt have much more tangible assets than who repurchase equity. This supports the assumption that bigger companies and companies with more tangible assets ratio have more debt capacity, due to more stable cash flows and, thus, lower distress costs. Companies repurchasing equity tend to be more profitable than companies with no transactions or issuing debt or equity. Pure debt issuers tend to be the less profitable. 48% of debt issuers have losses versus around 33% for equity issuers and double issuers. There is a high percentage of troubled companies among equity issuers too, but few of them have MTB less than 1. Firms that reduce debt have the lowest MTB, relatively low RET and the second highest proportion of unprofitable firms 45%. This statistics is similar to those reported by Gaud et al. (2007) for companies in European Union. Those, who repurchase equity, tend to be more profitable than those, who retire debt. Moreover, those who issue or repurchase 20

27 equity tend to have more financial slack, than those who issue or repurchase debt or do nothing. Overall, companies issuing or repurchasing equity tend to have higher MTB ratios than companies issuing or repurchasing debt. Stock returns of equity issuers are significantly higher than of debt issuers, who tend to have negative stock returns. Companies involved in all observed payout transactions tend to have low or negative stock return. That strongly supports the Market Timing theory. Companies repurchasing equity are more likely to be concerned with EPS dilution. Companies repurchasing equity and issuing debt or just repurchasing equity have a 90% chance to have equity dilution dummy equals 1. Equity issuers tend to be under levered and debt issuers tend to be over levered, which doesn t support the hypothesis of rebalancing capital structure towards its target. However, with payout transactions the situation is the opposite. Over levered firms tend to retire debt, whether under levered firms tend to repurchase equity. Companies repurchasing both debt and equity tend to be under levered, whether those with mixed transactions and no financing activity at all have their leverage close to its target. Companies reporting R&D expenses are more likely to issue equity instead of debt and have higher R&D expense than those issuing debt. Notably, many of them decide to repurchase their stock. Same is true for SG&A expense. Though, the mean SG&A expense is almost even among companies who made different financing decisions. Companies tend to issue equity and repurchase debt in better economic conditions and do nothing in worse economic situation. This alone is intuitive. Further investigation into this relationship is conducted in the next section. 21

28 5.2.1 Pure issue ObtSize variable, which holds for the transaction size, is excluded from regressions with passive strategy. Otherwise it would create serious bias. Table 4 Pure issue choice regression results Ind. Lev. TLevDif ROA CASHr MTB AdjRET Debt issue vs. Equity issue ** *** 6.77*** 7.683*** 0.434*** Debt issue vs. Debt and Equity issue Equity issue vs. Debt and Equity issue Debt issue vs. No transaction Equity issue vs. No transaction Debt issue vs. Equity issue Debt issue vs. Debt and Equity issue Equity issue vs. Debt and Equity issue Debt issue vs. No transaction Equity issue vs. No transaction Debt issue vs. Equity issue Debt issue vs. Debt and Equity issue Equity issue vs. Debt and Equity issue *** 7.755*** 1.162*** 0.251*** 0.526*** *** 2.328*** *** *** *** 0.294*** *** 3.912*** 5.445*** 0.049* *** *** 1.553*** *** *** *** *** RDr RDD SE dpedil dpbdil ObtSize *** 0.778*** 6.713*** 0.501*** *** *** ** 0.5*** 2.404*** 0.548*** *** 0.363*** ** *** *** *** *** *** 0.226*** *** ** *** 1.002*** Term Spread Default Spread Div. Yield GDP Growth Log Likelihood 0.051*** *** 188.8*** *** ** 116*** * *** *** *** Debt issue vs. No transaction Equity issue vs. No transaction *** *** *** *** *** 0.087*** *** *** ***, **. * indicates significance at 1%, 5% and 10% respectively. Standard Errors are given in italic. 22

29 Industry leverage and the distance from the target leverage in all issue choice regressions have signs suggesting that target leverage matters, and firms actually try to adjust to it, issuing more equity if a firm is over levered or issuing more debt if it is under levered. Though, industry leverage has been found insignificant for dual issue regression and debt issue versus no transaction regression. Difference from the target leverage is highly significant in all issue choice regressions. These findings are in line with what Gaud et al. (2007) found across the EU. The operating performance variables both tell the same story. ROA and CASH enter debt vs. equity regression with positive sign, which is I line with findings of Gaud et al. (2007), but contradicts findings of Hovakimian et al. (2004), who do not include Cash and report a negative effect of operating performance on the probability of the issuance of debt instead of equity. Nonetheless, even excluding Cash from my regressions I find the same sign on operating performance across all issue choice regressions. Notably, those variables in the target leverage regression have an opposite sign. It is consistent with findings of Korajczyk et al. (2003). This effect has at least two explanations. Firstly, for highly profitable firms debt acts as a disciplinary device. And issuing more debt firms limit their future financial slack, since it is a source of conflict between managers and shareholders. Moreover, it is consistent with a short run pecking order model where internal funds are preferred over external financing. The negative sign on CASH variable in equity vs. no transaction regression supports that idea. Secondly, this effect is consistent with the long run trade-off theory, where highly profitable firms, accessing public markets tend to issue debt. Market performance variables AdjRET and MTB show different signs compared to what Gaud et al. (2007) found in EU. Though, my results are consistent across different regressions and do not change if different measures of return and/or market-to-book ratio are introduced. Observed results suggest that high market performance increases the likelihood of 23

30 debt issuance and poor market performance increases the likelihood of equity issuance. This doesn t support the agency theory. On the other hand it supports the long run trade-off theory, that profitable firms, and you can expect such firms to have their shares doing well, are more likely to issue debt instead of more equity. AdjRET itself shows support for both market timing and pecking order theory. It enters debt issue vs. debt and equity issue regression with a positive sign and equity issue vs. no transaction with a negative sign, which clearly supports the pecking order theory. On the other hand, it enters equity vs. debt and equity issue regression with a positive sign, which supports the idea of market timing. When there is a choice between mixed issue and a pure equity issue high share price increases the likelihood of going with a pure equity issue. The EPS dilution variable dpedil is significant across all issue regressions and has signs consistent with those found in previous studies. In all cases the sign on dilution dummy suggests that managers try to avoid it and supports the idea that EPS one of primary concerns for CFOs. This found support in the survey by Graham and Harvey (2001) when approximately 70% of surveyed CFOs admitted that EPS dilution is the key factor in capital budgeting decisions. The issue size variable suggests that firms tend to stick with one financing instrument in case of big financing needs. This is opposite to what Gaud et al., 2007 found in EU. This can be explained with differences in accessibility of borrowed capital between EU and US and lower borrowing costs in the latter. However, choosing from debt issue and equity issue bigger financing requirements favor equity issue. This result is consistent with findings in previous studies including those done in EU. Three proxies for the uniqueness of assets complement one another. On the one hand, a firm reporting R&D expenses have costly projects and require external financing, thus is likely to issue debt or both debt and equity if possible. On the other hand the higher are R&D 24