UFDC Home  myUFDC Home  Help 



Full Text  
PAGE 1 1 CREDIT LINES, CASH HOLDINGS, AND CAPITAL STRUCTURE By G. BRANDON LOCKHART A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009 PAGE 2 2 2009 G. Brandon Lockhart PAGE 3 3 To my wife, children parents, and family for their unconditional love support and patience PAGE 4 4 ACKNOWLEDGMENTS I gratefully acknowled ge my dissertation committee Mark Flannery (chair) David Brown, Joel Houston, and Sarah Hamersma, for many helpful discussions, suggestions, and encouragement. I also thank Harry DeAngelo, Mike Faulkender, Jason Karceski, Ed Nelling, Jay Ritter, Michael Roberts, Mike Ryngaert, Sherrill Shaffer, Irina Stefanescu, Amir Sufi, and seminar participants at University of Florida, University of Mississippi, Lehigh University, University of Wyoming, University of Nebraska Lincoln, Drexel University, and Auburn Uni versity for their comments. Shelby Cohen, Valentina Gonzalez, and Kristy Lockhart provided excellent research assistance. PAGE 5 5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................................... 4 LIST OF TABLES ................................................................................................................................ 7 LIST OF FIGURES .............................................................................................................................. 9 ABSTRACT ........................................................................................................................................ 10 CHAPTER 1 ADJUSTING TO T ARGET CAPITAL STRUCTURE: THE EFFECT OF CREDIT LINES .......................................................................................................................................... 11 Introduction ................................................................................................................................. 11 Empirical Methodology .............................................................................................................. 16 Partial Adjustment Model ................................................................................................... 16 A Two Step Partial Adjustment Model .............................................................................. 17 Determinants of Adjustment Speed in Stage Tw o of the Estimation ............................... 18 Predicted Relationship between Credit Lines and the Speed of Adjustment ................... 20 Firms with mechanical movement to ta rget (subsets 3 6) ........................................ 20 Firms with mechanical movement away from target (subsets 1, 2, 7, and 8) .......... 21 Summary of Line Effects Predictions ................................................................................. 21 Data and Summary Statistics ...................................................................................................... 22 Empirical Results ........................................................................................................................ 26 Target Equation .................................................................................................................... 26 Summary Statistics Post Target Estimation ....................................................................... 28 Adjustment Speeds by Subset ............................................................................................. 30 Line Effects on Adjustment Speeds .................................................................................... 30 Line effects for firms with mechanical drift to target ................................................ 30 Line effects for firms with mechani cal drift away from target ................................. 32 Adjustment speed analysis summary .......................................................................... 33 Leverage Effects on Adjustment Speeds ............................................................................ 33 Cash Flow Effects on Adjustment Speeds ......................................................................... 34 How Are Firms Adjusting? ................................................................................................. 35 Robustness of Results .......................................................................................................... 38 Endogeneity of the existence of a credit line .............................................................. 38 Financial constraint ...................................................................................................... 38 Conc lusion ................................................................................................................................... 40 2 CREDIT LINES AND THE SUBSTITUTABILITY OF CASH AND DEBT ...................... 61 Introduction ................................................................................................................................. 61 Empirical Methodology .............................................................................................................. 66 Data and Summary Statistics ...................................................................................................... 69 PAGE 6 6 CRSP/Compustat and Credit Line Data ............................................................................. 69 Summary Statistics .............................................................................................................. 71 Constrained and Unconstrained Firm Identification Criteria ............................................ 72 Cash Balances b y Line Existence and Financial Constraint ............................................. 73 Incidence of Financial Covenant Violations by Financial Constraint Status .................. 74 Empirical Re sults ........................................................................................................................ 76 Financially Constrained versus Financially Unconstrained ............................................. 76 Existence of a Credit Line and Financial Constraint ......................................................... 77 Line Availability and Financial Constraint ........................................................................ 79 The Substitutability of Cash and Line of Credit Debt ....................................................... 80 Conclusion ................................................................................................................................... 82 APPENDIX A VARIABLE DEFINITIONS FOR CHAPTER 1 ...................................................................... 95 B VARIABLE DEFINITIONS FOR CHAPTER 2 ...................................................................... 97 C DATA COLLECTION ............................................................................................................. 100 Construction of the Full Sample ............................................................................................... 100 Construction of the Random Sample ....................................................................................... 101 D ECONOMETRIC ISSUES ....................................................................................................... 103 Short Panel Bias ........................................................................................................................ 103 Generated Regressors ................................................................................................................ 104 E REPLICATION OF FAULKENDER AND WANG (2006) ................................................. 105 LIST OF REFERENCES ................................................................................................................. 109 BIOGRAPHICAL SKETCH ........................................................................................................... 115 PAGE 7 7 LIST OF TABLES Table page 1 1 Summary statistics ................................................................................................................. 45 1 2 Su mmary statistics by HASLINE status ............................................................................... 46 1 3 Results from estimating the partial adjustment model ......................................................... 47 1 4 Summary of results from estimati on of target leverage ....................................................... 49 1 5 Transition matrix for leverage, cash flow, and line of credit subsets ................................. 50 1 6 Adjustment speed regression of Equation 1 6 ...................................................................... 51 1 7 Leverage effects on adjustment speeds ................................................................................. 53 1 8 Cash flow effects on adjustment speeds ............................................................................... 54 1 9 Net line effect predictions and empirical results .................................................................. 55 1 10 Balance sheet changes by subset ........................................................................................... 56 1 11 Cash flow statement analysis by subset over levered firms with negative cash flow ...... 57 1 12 Cash flow statement analysis by subset over levered firms with positive cash flow ....... 58 1 13 Cash flow statement analysis by subset under levered firms with negative cash flow .... 59 1 14 Cash flow statement analysis by subset under leve red firms with positive cash flow. .... 60 2 1 Summary statistics, full sample 1996 2006 ......................................................................... 84 2 2 Summary statistics, random sample 1996 2006 .................................................................. 85 2 3 Cash holdings and credit line characteristics by financial constraint, full sample ............. 86 2 4 Cash holdings and credit line charac teristics by financial constraint, random sample ...... 87 2 5 Covenant violations by financial constraint status ............................................................... 88 2 6 Regression results f or full sample by financial constraint status ........................................ 89 2 7 Marginal effects after regressions on full sample by financial constraint status ............... 91 2 8 Marginal effects after regressions on fandom sample by financial constraint status ......... 92 2 9 Substitutability of cash and debt regressions on random sample ........................................ 93 PAGE 8 8 E 1 Replication of Faulkender and Wang (2006), summary statistics .................................... 106 E 2 Replication of Faulkender and Wang (2006), regression results ...................................... 107 E 3 Regressions over alternative time periods. ......................................................................... 108 PAGE 9 9 LIST OF FIGURES Figure page 1 1 Description of mutually exclusive subsets of the data ........................................................ 43 1 2 Depiction of leverage and liquidity effects of existence of a revolving line of credit on a firms adjustment speed to target leverage ................................................................... 44 PAGE 10 10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CREDIT LINES, CASH HOLDINGS, AND CAPITAL STRUCTURE By G. Brand on Lockhart August 2009 Chair: Mark J. Flannery Major: Business Administration How important are transactions costs for firm financial policies? Transactions costs play an important role in theoretical and empirical explanations of firm cash holdings an d capital structure, but the extent of their economic importance is unknown. I use unique data on corporate credit lines to investigate the importance of transactions costs in the firms leverage rebalancing and cash holding decisions. Credit lines are important for this research because they provide access to debt and liquidity with minimal ex post fixed transactions costs. I extend the capital structure literatures partial adjustment model by estimating cross sectional adjustment speeds to target leverag e based on current leverage relative to target, firm demand for liquidity, and access to a credit line. I show that the access to liquidity provided by a credit line provides economic benefit to shareholders. Specifically, I find that shareholders of finan cially unconstrained firms value credit line availability and cash holdings similarly. Financially constrained firms can increase firm value by increasing cash holdings and credit line debt by the same amount. The results provide strong evidence that trans actions costs shape firm financial policy. PAGE 11 11 CHAPTER 1 ADJUSTING TO TARGET CAPITAL STRUCTURE: T HE EFFECT OF CREDIT LINES Introduction How important are transactions costs for firm financial policy? Transactions costs play a primary role in many theoretical and empirical explanations of corporate capital structure perhaps most prominently in their role in the firms leverage rebalancing decision. But the extent to which these transactions costs are economically important for leverage rebalancing is unknown I provide novel empirical evidence on the role of transactions costs in shaping financial policy by investigating how access to credit lines impacts firms leverage rebalancing decisions. Credit lines are relevant for this question because, once a line i s in place, the marginal cost of a line transaction is essentially zero, and thus, their existence potentially decreases the effect of transactions costs in the firms management of leverage. Indeed, I find that credit lines are associated with 50% to 77% faster speeds of adjustment to target leverage when the firm is under levered relative to target large effects given that under levered firms without credit lines close an average of 19% of the distance from target per year. Over levered firms close over 40% of the distance from target annually, but access to a credit line does not hasten these firms rebalancing efforts. Instead, these over levered firms appear to be more concerned with access to liquidity than with a marginal adjustment to target levera ge. The dynamic tradeoff theory of capital structure characterizes the firms leverage rebalancing activities as a pursuit to an optimal leverage ratio. The tax benefits of leverage are partially offset by expected distress and net agency costs resulting i n an optimal leverage ratio for the firm. However, transactions costs impede full adjustment to the target, and thus, firms allow leverage to vary over a range of ratios, and they choose to rebalance when the benefits exceed the costs of adjustment. Levera ge adjustments should result in debt ratios where the PAGE 12 12 marginal benefit of an incremental change equals the marginal cost. Thus, as discussed extensively in Leary and Roberts (2005), the characteristics of transactions costs influence optimal management tow ard the target. For example, in a world where transactions costs have only a fixed component, the optimal leverage range will increase with the magnitude of the fixed cost (Fischer, Heinkel, and Zechner (1989)). Once leverage crosses the optimal range boundary, the firm will adjust completely to target there is no incremental adjustment cost to keep the firm from adjusting to target completely. The result is a pattern of lumpy adjustments long excursions from target and large complete adjustments. Alt ernatively, in a world with only incremental transactions costs, the firm will make leverage adjustments to just within the optimal leverage range boundary, resulting in small leverage adjustments clustered in their timing. And when there are both fixed an d incremental costs of adjustment, the leverage ratio at which the marginal benefit of adjustment equals the marginal cost will be somewhere between the target and the range boundary. Plainly stated, the firm does not adjust leverage all the way to target because of convex incremental costs (Altinkili and Hansen (2000)), but it puts enough distance between the ex post leverage ratio and the optimal range boundary to minimize the likelihood of incurring the fixed cost again next period. Flannery and Rangan (2006) estimate a dynamic partial adjustment model to analyze the leverage rebalancing decision: leverage next period is highly dependent on leverage this period and the target, and the coefficient on the firms distance from the estimated target measures the speed of adjustment. A nontrivial estimate of adjustment speed is consistent with active leverage rebalancing; the alternative empirical outcome has been argued as evidence supporting PAGE 13 13 alternative capital structure theories.1 The first papers employing a partial adjustment model assumed constant adjustment speed for all firms.2 Among these first studies, constant adjustment speeds are estimated on the order of 7% to 35% of the distance from target per year depending on the empirical methodology and the definition of leverage. At this point the literature has concluded that the method of estimating targets is very important to conclusions about speeds of adjustment (e.g., see Table 8 of Huang and Ritter (2008)), with most of the focus placed on the model ing of fixed effects and the importance of leverage dynamics. An important extension to this research is documentation of an association between adjustment speeds and transactions costs in the cross section. Instead of assuming all firms adjust to target w ith the same speed, an association between adjustment speeds and transaction costs across firms is further evidence of active leverage rebalancing.3 My paper follows this intuition by estimating cross sectional variation in adjustment speeds based on firm characteristics related to transactions costs, such as access to a credit line, the current leverage ratio relative to target, and the firms demand for liquidity. Firms with credit lines in place should have lower costs of adjusting leverage. Specificall y, ex post the credit line should lower the firms fixed cost, and weaken the convexity of variable costs, of adjustment. As a result, we should observe more complete adjustment to target for firms with access to a line thus, greater estimated speeds of adjustment if the lower transactions costs are economically important. Draws and repayments on the line are low cost and convenient 1 See Myers and Majluf (1984) for the pecking order theory, Baker and Wurlger (2002) for the market timing theory, and Welch (2004) for the managerial inerti a theory. 2 For example, see Fama and French (2002), Flannery and Rangan (2006), Lemmon, Roberts, and Zender (2008), Antoniou, Funey, and Paudyal (2008), and Huang and Ritter (2008). 3 Cross section variation in adjustment speeds are explored in Byoun (2008), Dudley (2008), and Faulkender, Flannery, Hankins, and Smith (2008) with regard to the estimated leverage residual, large investments, and cash flow shocks, respectively. PAGE 14 14 to the firm by construction; a phone call to the banker gets the funds transferred provided compliance with loan covenant s. Repayments of credit line debt are relatively costless as well; there are no coordination problems often attendant the public debt markets, and lines are often senior debt and sometimes secured, thus, a credit line balance is largely immune to incentive distortions caused by debt overhang. In short, the line offers the firm a convenient tool with which to make leverage adjustments. But, firms maintain credit lines for important reasons other than to have access to a convenient leverage rebalancing tool they are also in place to fund investment and manage corporate liquidity (Sufi (2007), Lins, Servaes, and Tufano (2007), and Yun (2008)). Thus, the credit lines usefulness for leverage adjustments will depend on current leverage relative to target and th e firms demand for liquidity. The lines dual role as leverage tool and liquidity tool makes these contracts especially interesting for study of the firms tradeoff of liquidity and leverage (Gamba and Triantis ( 2008)). The partial adjustment model impli citly assumes that cash and debt are always perfect substitutes: if the firm is at target leverage then no leverage adjustment is made regardless of the firms short term cash position. Also, if a firm is outside of the target leverage range, the firms de mand for liquidity is exogenous to the predicted leverage adjustment. But Acharya, Almeida, and Campello (2007) and Gamba and Triantis (2008) show that there are times when cash is not a perfect substitute for debt and this friction yields important implic ations for firm investment and financial policies. The substitutability of cash and debt is relevant for this study because the firm can use the line for access to cash and for leverage adjustments, and with regard to credit lines, perfect substitutability of cash and debt can fail due to their contingent nature. Noncompliance with the lines financial covenants makes access to the line tenuous. A firm with no expectation of a covenant violation might be indifferent between cash balances and line PAGE 15 15 availabil ity (ignoring interest and tax differentials), however, there can be other cases in which this contract contingency is important, and this friction may overshadow the lines provided benefit of relative lower transactions costs for leverage rebalancing. In summary, credit lines provide an opportunity for a novel test of the economic importance of transactions costs for leverage rebalancing, but the relative importance of these lower transaction costs will depend on the relative importance of other functions served by the line. This chapter contributes to the empirical capital structure literature in a few ways. First, it extends the empirical work of Leary and Roberts (2005), Faulkender et al. (2008), and Dudley (2008) empirical papers which analyze the ch aracteristics of transactions costs and implications for management of capital structure.4 I find that the lower transactions costs associated with credit lines are economically important for leverage rebalancing, but that the relative importance of lower transactions costs varies in the crosssection. Second, this paper contributes to the work on external capital market frictions and resulting effects on the firms real and financial decisions. In this regard the paper is motivated by the work on the effe cts of the lack of substitutability of cash and debt on firm financing and investment decisions,5 and the recent work on the effects of credit control rights on management of capital structure and firm investment policies.6 I find results consistent with the contingent nature of credit lines leading to 4 Leary and Roberts (2005), Faulkender et al (2008), and Dudley (2008) are motivated by the theories of Fis c her et al (1989), Goldstein, Ju, and Leland (2001), Stebulaev (2007) and Dudley (2008) which model leverage rebalancing activity with adjustment cos ts featuring fixed and variable components. 5 Almeida, Campello, and Weisbach ( 2005) finds that firms add to cash stocks from cash flows in anticipation of future financing frictions. Acharya et al (2007) finds that firms with low correlation between investment opportunities and cash flows will adjust cash and debt policies to trans fer debt capacity from high cash flow states to states with high investment opportunities. Gamba and Triantis (2008) show theoretically that a firm might prefer cash holdings to debt capacity if transactions costs of access to liquidity are sufficiently im portant. 6 Sufi (2007) finds that negative cash flows is the most important predictor of a covenant violation for firms in his sample of credit lines, and a covenant violation is associated with a 15% to 25% drop in the total and available portions of the credit line. Roberts and Sufi (2008) find that net debt issuance decreases following a financial PAGE 16 16 preference for cash vis vis leverage adjustments at times when the credit lines contingency is likely most important for the firm and lender. Third, this paper adds to the young but growing literature on the role of credit lines in corporate financial management. Sufi (2007) documents the widespread use of corporate credit lines, analyzes the lines role in corporate liquidity management, and provides evidence that these contingent contracts are sometimes very different from the committed credit lines described in the theoretical literature.7 Lins et al. (2007) analyze results from an international corporate survey on credit line use, and Yun (2007) analyzes the interplay of corporate line use, cash polici es, and corporate governance. To start, I estimate leverage targets by estimating a dynamic partial adjustment model for leverage ratios. Next, I calculate the firms distance from target and note whether the firm is over or under levered relative to targ et. Then, I note whether the firm has excess or deficit free cash flows to categorize the firms demand for liquidity. These three characteristics: 1) line or no line, 2) over or under levered, and 3) cash surplus or deficit, yield eight mutually exclusiv e subsets of the data. I then estimate leverage adjustment speeds for the eight subsets, and test hypotheses on the importance of the three characteristics and their interactions. In my last set of results, I study the firms cash flow statement to observe the effect of credit lines on differential responses to a one dollar cash flow (outflow) across the eight subsets of firms. Empirical Methodology Partial A djustment M odel Flannery and Rangan (2006) estimate a partial adjustment model for market leverage on a large panel of U.S. firms over 1965 2001, covenant violation, and that over one quarter of firms violate a financial covenant over their sample period. Nini, Smith, and Sufi (2008) show that firm inves tment policy reacts sharply to a financial covenant violation 7 See Boot, Thakor, and Udell (1987) and Holstrom and Tirole (1998) for theoretical work on credit lines. PAGE 17 17 + 1 = + 1 + + 1 (1 1 ) where + 1 is the firms observed market leverage ratio at the end of year t+1 Their regression implicitly assumes firms have target debt ratios ( + 1 ), they trade off costs and benefits of adjusting to their (unobserved and dynamic) targets, and all firms adjust at the same constant speed ( ) where estimates the portion of the distanc e from target leverage that the firm closes per year Because the target is unobserved, to move forward emp irically we must first rewrite E quation 1 1 to include an vector of variables which determine the firms leverage ratio target, + 1 = ( 1 ) + ( + ) + + 1 (1 2) where + = + 1 Estimating Equation 1 2 returns estimates of ( 1 ) and the firm fixed effects We then divide and by and calculate the estimated target leverage ratio, + 1 A nontrivial estimate for is consistent with active management to target capital structures. A less than one indicates that the firm does not completely realize the target within one year. A T wo S tep P artial A djustment M odel One contribution of this chapter is the relaxation of the requirement for adjustment speeds to be constant across all firms. Estimating variable adjustment speeds requires a two step procedu re, and the first step is to estimate a target. I follow the literature by estimating Equation 1 2, and then I calculate + 1 = + so that I can re write Equation 1 2 in terms of the deviation from target, PAGE 18 18 + 1 = + ( 1 3) where + 1 + 1 , + 1 is the leverage residual and is a function of a vector of firm characteristics affecting adjustment speed, = ( 1 4) Substituting Equation 1 4 into Equation 1 3 yields + 1 = + (1 5 ) I can then test hypotheses on the effect of firm characteristics, including the presence of a line of credit, on estimated adjustment speeds There are two econometric issues important to t his section. First, the Nickell (1981) bias results if Equation 1 2 is estimated via least squares dummy variable regression (LSDV) due to the presence of the lagged dependent variable and firm fixed effects on the right hand side of the equation. I follow Lemmon et al (2008) and Faulkender et al. (2008) and estimate Equation 1 2 with system GMM (Blundell and Bond, 1998) to address the Nickell (1981) bias. I can estimate Equation 1 5 with OLS as the fixed effects are included in the targets calculation. S econd, inferences on estimates from Equation 1 5 must contemplate the presence of generated regressors (Pagan (1984)). I use a bootstrap program to address this issue. As the former issue has been discussed at length in Lemmon et al (2008), Huang and Ritt er (2008), and Faulkender et al. (2008), I leave discussion of both econometric issues to Appendix D Determinants of A djustment S peed in S tage T wo of the Estimation Because my innovation lies in the estimation of cross sectional adjustment speeds via Equa tion 1 5 I focus attention now on adjustment speed determinants and return to target determinants in a later section In order to allow adjustment speeds to depend on a firms 1) PAGE 19 19 current leverage residual, 2) demand for liquidity, and 3) existence of a li ne of credit, I create eight mutually exclusive subsets of the data by interacting dummy variables for the three dimensions. Figure 1 1 details the arrangement of the eight subsets. Firms in Subsets 1 through 4 (5 through 8) are over levered (under levered ) relative to estimated target subsets in the top half of Figure 1 1 are for firms overlevered and subsets in the bottom half of the figure are for under levered firms. The firms in Subsets 1, 2, 5, and 6 (3, 4, 7, and 8) have negative (positive) free c ash flow outcomes the subsets in the top halves of over and under levered subsets are for firms with negative free cash flows. For line effects analysis, peer groups are by leverage residual and cash flow outcome differing by line of credit status. Thus for firms over levered, Subsets 1 and 2 contain firms with negative cash flow, and Subsets 3 and 4 contain firms with positive cash flow. In the bottom half of the figure, Subsets 5 and 6 contain firms under levered with negative cash flows, and Subsets 7 and 8 contain firms under levered with positive cash flows. Finally, firms in the even (odd ) numbered subsets have (do not have) a committed revolving line of credit. Expositing an analysis among eight subsets is a challenge, and to keep it as simple a s possible I refer to subsets according to this numbering scheme and encourage the reader to refer to the figure as needed. Estimating separate adjustment speeds across the eight subsets of the data allows the line of credit effects to vary according to le verage characteristics and cash flow outcomes. I model the firm specific adjustment speeds in Equation 1 5 as a function of the leverage residual, the cash flow outcome, and the existence of a line, + 1 = 8 = 1 + (1 6 ) where 1 is the estimated adjustment speed for firms belonging to Subset 1: a firm year observation with a negative cash flow, no line of credit, and a negative estimated leverage PAGE 20 20 res idual The dummy variables indicate the Subset to which the firm belongs at time t With estimates of the eight we can test hypotheses on the effects of credit lines on adjustment speeds. Predicted R elationship between C redi t L ines and the S peed of A djustment The credit lines usefulness for leverage adjustments toward target depends on the firms current distance from target (leverage residual) and the firms demand for liquidity; Figure 1 1 identifies four relevant outcom es based on combinations of 1) a positive or negative leverage residual, and 2) positive or negative free cash flow. Firms over levered (under levered) relative to target with positive (negative) cash flows will mechanically move closer to target if cash f lows and leverage are left unmanaged (Subsets 3 6). Likewise, negative (positive) cash flows will push over levered (under levered) firms away from target (Subsets 1,2,7, and 8). Firms with m echanical m ovement to target ( s ubsets 3 6) In Subsets 3 6, adjust ment speed to target is increased if the line of credit is used to manage the excess or deficit liquidity. For example, the over levered firm with excess cash flows (Subset 4) can increase adjustment speed to target (relative to firms in Subset 3) by using some of the cash to repay some of a credit line balance. And under levered firms with negative cash flows (Subset 6) can get closer to target (relative to firms in Subset 5) by drawing funds from the credit line to meet cash demands. Thus we should observe faster speeds of adjustment to target if the credit lines lower transactions costs are economically important for leverage rebalancing. Alternatively, the friction derived from the credit lines contingency might overwhelm any importance of the lower tr ansactions cost for Subset 4. For example, if firms are over levered due to past negative performance or if over levered firms are more likely to violate a financial covenant, then Subset 4 firms might value realized cash stocks more than they value line of PAGE 21 21 credit availability (i.e., cash is not equal to negative debt), and in this case the adjustment speeds associated with credit lines will not necessarily be greater than those estimated for firms in Subset 3. Under levered firms might face more monitoring from the lender given its negative cash flows (Subset 6), decreasing the lines effectiveness of lowering transactions costs. On the other hand, negative cash flows and/or increased monitoring might encourage Subset 6 firms to prefer cash over line of cre dit availability, and resulting draws on the line will increase adjustment speeds to target. Firms with m echanical m ovement a way from target ( s ubsets 1, 2, 7, and 8) Over levered (Under levered) firms with negative (positive) excess cash flows will mechan ically move away from target if cash flows and leverage are left unmanaged. Over levered firms with negative cash flows will be faced with the trade off of liquidity and leverage adjustment needs. Is it more important to manage the negative cash flows or t o adjust back to target? A credit line balance will be a convenient debt source to repay to lower leverage, but a repayment of debt decreases cash availability further. If managing the cash flows is more important for these over levered firms, then adjustm ents to debt will not rebalance leverage to target raising cash via the debt markets will attenuate other rebalancing activities. To adjust back to target these firms need to issue equity or shrink assets. Under levered firms with excess cash flows will also move away from target leverage if cash flows and leverage are not managed. All else equal, access to the line will make the firm more comfortable with its liquidity position and perhaps more likely to use cash or line availability for investment or for equity repurchases. Summary of L ine Effects P redictions Figure 1 2 summarizes the predictions for line effects on adjustment speeds to target leverage. The figure decomposes the net line effects on adjustment speeds into two components: the leverage eff ect and liquidity management effect. Leverage effects refer to the partial PAGE 22 22 effect of the credit line on adjustment speeds if the line is used to manage the free cash flow outcome. Liquidity management effects are zero if cash and debt are perfect subst itutes. If cash and debt are not perfect substitutes then the firm prefers realized liquidity over the optional liquidity provided by the credit line. As a result, the firm is more inclined to draw down availability off of the line or less inclined to use cash to restore line availability. The net line effect is negative for firms over levered with negative cash flows. Draws on the credit line to manage the negative cash flows will attenuate other rebalancing activities. If cash and debt are perfect substi tutes then the net line effect is positive for firms over levered with positive free cash flows. Cash can be used to repay credit line debt, and the lower transactions costs will lead to faster adjustment for firms with such access. However, if cash is not equal to negative debt then the firm will be less inclined to use cash to restore credit line availability, and if this liquidity management effect overwhelms the benefit of the credit lines lower transactions costs then the net line effect will be negat ive. The net line effects for under levered firms with negative cash flows are positive because draws off the line increase leverage to target, and any liquidity management effects will exacerbate the incentives to increase leverage. The net line effect fo r under levered firms with positive free cash flows is hard to predict ex ante. Firms can use excess cash flows to repay a line balance (if one exists), but doing so moves the firm further from target such an action is consistent with having little use f or cash stocks. If the line of credit availability and positive free cash flows make the firm more comfortable with its liquidity position, then the firm will be more likely to repurchase equity, which increases leverage. Data and Summary Statistics Much l ike the Sufi (2007) study of lines of credit and corporate liquidity management policies, I need data on the existence of lines of credit for a panel of firms through time. I require PAGE 23 23 firms to have at least four years of continuous data available in the CRS P / Compustat merged database during the sample period of 1996 2006. I follow Sufi (2007) for data screens. I require four consecutive years of positive assets and book leverage ratios between zero and one (inclusive). Non missing data for at least four cons ecutive years is required for total liabilities (data181), sales (data12), operating income before depreciation expense (data13), common shares outstanding (data25), fiscal year end stock price (data25), preferred stock (data10), deferred tax assets (data3 5), and convertible debt (data79). These screens leave me with 5,476 firms afte r omitting financials (SIC 6000 6999), utilities (SIC 4900 4999) and firms not in the Fama French 49 industry groups. For the resulting firms, I program a text search of every 1 0 K form available on the SECs Edgar server.8 From the search program results, I form a dummy variable, HASLINE equal to one if the firm has a committed line of credit available at year end, and equal to zero otherwise. I am able to compile this variabl e for 5,299 firms and 41,696 firm year observations over 1996 2006. After data loss due to missing data for regressors and requiring one year asset growth to be less than 100%, the actual usable size of the dataset for adjustment speed regressions consists of 5,171 firms and 31,513 observations.9 I use a free cash flow variable, FCF, to measure the amount of cash flow over which the firm likely has discretion. For its construction I follow Acharya et al (2007) by taking operating income before depreciation expense in year t+1 (data13) less the sum of the following items for year t : depreciation and amortization expense (data14), interest expense (data15), taxes (data16), preferred dividends (data19), and common dividends (data20), all scaled by year t asset s (data6). 8 D etails of the search prog ram and data collection are available in Appendix C. 9 I ha ve only 31,513 usable observations because my dependent variable is in year t+1 and thus my last usable year for regressors is 2005. I lose my first observation for each firm when I first difference Equation 12 for system GMM. PAGE 24 24 It is important to use a variable such as FCF as the proxy for the firms demand for liquidity as opposed to a measure combining cash flows and existing cash stocks. Combining cash flows and stocks implicitly assumes the firm addresses any demand for liquidity completely before considering leverage adjustments. But when are cash stock adjustments more important than leverage adjustments? We can learn more about the firms trade off of cash stock and leverage adjustments by analyzing the eight mut ually exclusive subsets. Definitions of all variables used herein are listed in Appendix A, and variables used in regressions are winsorized at the 1% and 99% levels. Table 1 1 reports a few summary statistics on the sample. Leverage at time t has a mean o f 0.237 of market assets and the inter quartile range spans 0.021 to 0.379. The median market leverage is much lower at 0.159, as 13.8% of the firm year observations have zero debt outstanding.10 While the mean EBITTA is negative at 3.5% of book assets, the median firm has operating earnings of 6.0% of assets. The mean (median) market to book ratio is 1.85 (1.19) and the mean LNTA is 19.426, therefore the average firm has book assets of $273 million in 1983 U.S. dollars. Nonzero research and development ex pense is reported for 48.4% of the observations, and for these firms the average expense is 11.6% of assets (0.116 = 0.056/0.484). The median firm holds 7.7% of assets in cash at year t In Table 1 2 the summary statistics are delineated by whether the fir m has a line of credit at year end t Tests for equal means and medians are reported by t and z statistics, respectively. The existence of a line of credit sorts firms in potentially important ways. Market leverage is much lower for firms without lines of credit the mean market leverage for firms without (with) lines is 0.119 (0.269), and the median market leverage for firms without lines is a mere 1% of assets. But the standard deviation (not tabulated) of leverage for firms without lines is 1.76 times 10 Yet 44.1% of these zerodebt observations have a line of credit. PAGE 25 25 the mean and the inter quartile range covers zero leverage firms to those with leverage ratios of 0.147. Notice the differences in cash holdings among the firms with and without lines of credit. Firms without (with) lines of credit hold an average 39.1% (11.7%) of book assets as cash, and the medians for CASHTA tell the same story. This result highlights the potential importance of credit lines data for studies on corporate liquidity management. Firms with (without) a credit line have book assets of about $ 389 ($75) million in 1983 book assets at the mean. Firms with lines have more tangible assets, as the median FATA is 0.231 versus 0.115 for firms without. Firms without lines have more research and development expense, but also have higher market to book r atios. Median free cash flows also show that firms without lines of credit have negative cash flows of almost 10% of assets, while firms with lines of credit have positive free cash flows of 3.2% of assets. In untabulated results, almost 21.5% of firms wit hout lines of credit will have one in place at year end t+1 Of the total 5,171 firms there are 505 (9.8%) which never have a line of credit. Out of these 505 firms 2,369 observations, 21.7% (115 firms) also never have any outstanding debt at year end. In summary, Table 1 1 and Table 1 2 show that the existence of a credit line sorts firms in other potentially important ways, and it is important to be mindful of these data features as we investigate the role of lines of credit in the leverage adjustment de cision. The firms differ by size, earnings, cash flows, and perhaps overall cash management policies. Firms without lines have negative free cash flows and fewer tangible assets with which to offer as collateral, perhaps making the acquisition and/or maint enance of a line of credit more difficult. Higher market to book ratios and greater research and development activities for firms without lines, coupled with the fact that 21.5% of firms without lines at year t+1 had one at year t suggests that these firm s indeed have a need for such a contract but are being rationed. PAGE 26 26 Empirical R esults Target Equation The first step to relaxing the assumption of constant adjustment speeds is to estimate a target leverage ratio. In my first stage estimation of leverage targ ets (Equation 1 2), I follow recent literature, and pay particular attention to Flannery and Rangan (2006) and Lemmon et al (2008). Flannery and Rangans partial adjustment model was the first to explore the importance of both the firm fixed effects and t ime varying components for the firms target leverage ratio. But their focus was also on estimating constant adjustment speeds among the dynamic targets, and thus time varying firm characteristics such as size, profitability, and investment opportunities w ere modeled along with firm fixed effects to proxy for the leverage target. Lemmon et al. (2008) use both U.S. Compustat data and a UK dataset containing pre IPO capital structures to show that a majority of a firms target leverage ratio is determined by a time invariant firm effect.11 As discussed in Appendix D there are important econometric details to consider when estimating Equation 1 2 due to the presence of fixed effects and the lagged dependent variable on the right hand side of the equation. Huang and Ritter (2008) document the importance of the methodology chosen to estimate targets and adjustment speed s. I report various alternative sp ecifications mainly to show comparability to previous work, which is important here because the sample period is much shorter relative to those studied elsewhere. My OLS, least squares dummy variable (LSDV), and Arellano Bond results are comparable to other papers reporting similar regressions. Like Faulkender et al (2008) and Lemmon et al (2008), I use Blundell and 11 DeAngelo, DeAngelo, and Whited (2008) model the role of transitory debt in the firms capital structure and show that heterogeneity in investment shocks and the use of transitory debt can also generate the empirica l patterns of leverage documented in Lemmon et al (2008). PAGE 27 27 Bonds (1998) system GMM estimator to estimate Equation 1 2. The adjustment speed estimate reported in Column 4 of Table 1 3 is one minus the reported coefficient ( = 1 0 775 = 0 225 ) and results from the assumption of constant adjustment speed much like that which is reported in the existing literature. The constant adjustment speed estimate in this first stage is not my focus here it is merely a byproduct of the estimation of target leverage, where target leverage is the sum of the firm effect, year ef fects, and the linear combination of My choices for the components of the determinants of the leverage ratio target, are the literatures standard time varying firm characteristics: profitability, investment opportunities, demand for tax shields, size, asset tangibi lity, and industry leverage. EBITDA controls for firm demand for tax shields, or alternatively, for firm demand for external finance. An increase in investment opportunities has a negative effect on target leverage because the firm desires to maintain debt capacity, or alternatively, because the intermediary requires more monitoring activity.12 Firms with higher levels of DEPTA have less need, ceteris paribus for interest tax shields, and firms with large amounts of FATA have more tangible collateral with which to secure loans (DeAngelo and Masulis (1980)). Larger firms tend to have fewer information asymmetries important to external financiers, and size should also be related to scale economies associated with the cost of accessing external markets (Faulke nder and Petersen (2006)). Research and development expense is another control for investment opportunities and hard to value assets (Hovakimian, Opler, and Titman, 2001). Signs on GMM BB point estimates of in Column 4 for the time varying firm characteristics are also consistent with expectations. On the margin, firms with more profitability will have less need for external financing, and I find that the point estimate for EBITTA is 0.023. 12 See Johnson (1997) and Billett, King, and Mauer (2007) for studies on loan contract features and the attenuation of the negative effect of investment opportunities on leverage. PAGE 28 28 Investment o pportunities are modeled with the market to book ratio and the research and development expense variables. The estimate for MB is zero, but the point estimates for RDD and RDTA indicate that future growth opportunities and hard to value assets tend to lowe r target leverage. Firms with high depreciation expense will have less need for interest tax shields, ceteris paribus and I find a negative point estimate for DEPTA 0.191. Theoretically, firm size and the presence of more tangible assets tend to increas e target leverage. Here size returns a positive point estimate, and the point estimate for FATA has the correct sign but it is not statistically different from zero. In summary, the GMM BB constant adjustment speed estimate is consistent with other studies and point estimates on target determinants agree with capital structure theory.13, 14 Next I use the GMM BB results to calculate the estimated targets and determine the leverage residual for each observation.15 Interacting dummy variables for the sign of th e leverage residual, the sign of FCF, and HASLINE determines the observations Subset assignment for the remaining tests. Summary S tatistics P ost Target Estimation Table 1 4 reports summary results from the target calculation, + 1 = + for all observations by leverage, free cash flow, and line of credit status Subsets. The mean target leverage is 0.263 of market assets, and the average firm is under levered by 2.6 percentage points. Estimated targets vary significantly over the various subsets; firms over levered with positive free cash flow and no line of credit have an average target of 9.9% of market assets, but 13 Frank and Goyal (2008) show that a negative coefficient on profitability is not necessarily in discord with the trade off theory if leverage is modeled as a ratio of debt over assets. 14 Using too many instruments can over fit and thus fail to remove endogeneities. Accordi ng to Roodman (2006), a useful heuristic is to become concerned with the toomany instruments problem if the number of instruments approaches the number of firms or if results are not robust to reasonable alternative instrument sets. Here the number of fir ms is nearly ten times the number instruments, and the results are robust to various instrument specification suggestions discussed in Roodman (2006). 15 The papers results are robust to the definition of the leverage target. PAGE 29 29 firms under levered with negative free cash flows have an average target of 45.6% of assets. The average d eviation from target conditional on being over levered (under levered) ranges from 11.9 to 17.1% (8.9 to 20.4%) of assets depending on the Subset. Firms with a negative cash flows and a line of credit show the highest absolute deviation from target. And fi rms with positive cash flows and no line of credit are estimated as closest to target. The results in Table 1 2 showed that access to a credit line sorts firms in other potentially important ways, which could be problematic for later tests if firms are sta tic along this dimension. But the last column of Table 1 4 gives a preliminary look at the within firm variation in HASLINE Over 21% of firms over levered without lines of credit at time t+1 (Subset 1) had a line of credit in place at time t and almost o ne quarter of firms in Subset 3 had a line of credit the previous fiscal year end. So there appears to be some nontrivial migration across subsets. Table 1 5 describes the migration directly by reporting the probability of belonging to Subset q at time t+ 1 conditional on belonging to Subset p at time t For example, conditional on belonging to Subset 1 at time t the firm is 45% likely to remain in Subset 1 at time t+1 but also 25% likely to be in Subset 5 at time t+1 The three right most columns of the t able show the probability of maintaining leverage residual, free cash flow, and line of credit statuses, respectively, at year t+1 Firms in Subset 1 are 35.9% (1 0.641) likely to be under levered at year t+1 The diagonal of the transition matrix alwa ys contains the largest probability for assignment at t+1 but significant migration occurs in one period. In untabulated results, I analyze the extent of firm movement among the subsets. On average, firms migrate to a new subset every other year. Thus, a firm with eleven observations will migrate across five to six of the total eight Subsets, and a firm with six observations will at some point belong to three different Subsets. PAGE 30 30 Adjustment S peeds by Subset The results of the variable adjustment speed estima tion (Equation 1 6 ) are reported in Table 1 6 The partial adjustment model assumes no leverage change if the leverage residual is zero, thus no constant is modeled in the regression of Equation 1 6 all subsets are modeled, and estimates are interpreted a gainst the null hypothesis of zero. OLS results show wide variation in estimates Estimates range from as little as 0.135 to as great as 0.520, with an (untabulated) average of 0.337. Over (Under ) levered firms adjust to target with an average speed of 0.425 (0.267). Firms over levered with a negative cash flow and no line of credit (Subset 1) exhibit the greatest speed of adjustment, and firms under levered with a positive cash flow and no line of credit (Subset 7) have smallest adjustment spee d estimates.16 These results are new to the literature releasing the constraint of constant adjustment speed seems to be important in the cross section. In the next subsections I analyze line effects, leverage effects, and cash flow effects to test for s tatistical significance of various differences between the Table 1 6 adjustment speed estimates. Line Effects on A djustment S peeds Line e ffects for firms with m echanical d rift to t arget Now that we have estimated variable adjustment speeds, we can analyze the effects of the existence of a line of credit. The line of credit is perhaps most useful for adjustments to target leverage when the leverage residual and cash flow outcomes are complementary to desired leverage and cash stock adjustments. For example, when the firm has excess cash and is over levered, or needs cash and is under levered, the line of credit can potentially address the firms leverage adjustment and liquidity needs, simultaneously. Estimates for Subsets 3 and 4 show the 16 Excluding firms which never have a line and never have debt makes no difference to the results. PAGE 31 31 effect of a line of credit on adjustment speeds for firms over levered with a positive free cash flow. Here the firm chooses between adding to cash stocks and reducing leverage. Likewise, estimates for Subsets 5 and 6 show the effect of a line of credit on adjustment speeds for firms under levered with a negative cash flow. Simply funding the negative cash flows with new debt issues addresses both sides of the tradeoff. The Line Effects column of Table 1 6 reports these line effects. Firms in Subset 3 have essentially the same estimated adjustment speed as firms in Subset 4. Here the line of credit does not appear to affect the firms adjustment speed. This zero line effect is consistent with firms choosing not to use cash flows to repay debt on the line of credit different ially the liquidity management effect offsets the leverage effect of the line of credit. Perhaps the firm discounts the probability of having the line next period and values cash today greater than line availability. Alternatively, firms in Subset 4 coul d be using the cash flow to invest in projects instead of adjusting to target. But the bottom line seems to be that these firms are treating their line much like Subset 3 firms treat term debt. Whatever benefits the line offers in decreased transactions co sts appears to be dominated by the benefits derived from other uses of the cash. Put another way, the benefits derived from the lines lower transactions costs are dominated by the benefit derived from the lines intrinsic flexibility. I revisit this issue in a later section. Firms in Subset 6, however, adjust to target 12. 6 percentage points faster than firms in Subset 5, a very large effect given the Subset 5 adjustment speed of 0.241; here the line appears to allow the under levered firms with negative c ash flows to adjust faster than similarly situated firms without lines. If there are lower transactions costs associated with draws on the line then we expect to see this result. In addition, if the Subset 6 firms have reason to doubt the availability of t he line next period, then they will prefer to draw funds today leading to faster adjustment. PAGE 32 32 Line e ffects for firms with m echanical d rift a way from t arget For Subsets 1 and 2 (over levered firms with negative cash flows) recall our discussion of Figure 1 2 these firms likely need to raise funds for operations and can do so by issuing debt and/or equity. The choice of security depends on relative issuance costs, and if a line of credit is available for use then it will act to attenuate any adjustment speed generated by rebalancing activity (e.g., equity issuance). Further, if the firm doubts future line availability, perhaps due to lender action, then line availability tomorrow is discounted and a line draw today is more attractive. The estimated line effec t for Subset 2 is consistent with this story the Subset 2 firms have estimated adjustment speeds 15.4 percentage points lower than the estimates for firms in Subset 1. The extremely fast adjustment speed for Subset 1 firms is striking. These firms are pu shed from target by the negative cash flows but they close of over half the leverage gap. This action indicates the potential importance of equity issues for Subset 1 firms. Indeed, if there is a high fixed cost of accessing the equity market, then consist ent with the importance of a fixed component of transactions costs (Fischer et al. (1989) ), once the firm can justify these transactions costs (a decision aided by being over levered with negative cash flows), then the adjustment should be large. Subset 7 and 8 firms are also mechanically pushed away from target if cash flows and leverage are left unmanaged. Firms in Subset 8 adjust to target leverage 10.4 percentage points faster than firms in Subset 7 a very large effect given that firms in Subset 7 adj ust at an annual rate of only 0.135. If firms with lines of credit are more certain about future financial flexibility then they will be more inclined to distribute excess cash to shareholders, and this action increases leverage toward target. Alternativel y, if firms with lines are more certain about future financial flexibility following positive cash flows, then they are more inclined to use todays resources (cash and debt capacity or line availability) for investment. PAGE 33 33 Adjustment s peed a nalysis s ummary T able 1 9 restates Figure 1 2 with a summary of the expectations and results for the n et line effects estimated with Equation 1 6 Estimates of line effects are consistent with those expected in all cases. In the case for over levered firms with a positive cash flow, we could not sign the net line effect ex ante as its sign depends on the relative importance of the leverage and liquidity management effects. The statistical (and economic) result for this net line effect in the context of the theoretical expectation indicates that the leverage and liquidity effects on adjustment speed are equally important for firms over levered with positive cash flows. The other net line effects range from 0.104 to 0.154 in magnitude results economically meaningful given that the unconditional estimated adjustment speed is 0.337. In three of the four cases, line effects are consistent with the importance of the substitutability of cash and debt for leverage adjustment speeds. In a later s ection I will analyze the firm action s behind these adjustment speed differences, but first I want to briefly discuss a few of other results on adjustment speeds and how they support the results in Byoun (2008) and Faulkender et al. (2008). Leverage Effects on A djustment S peeds Byoun (2008) f inds an asymmetry in adjustment speeds based on the sign of the leverage residual. We can also analyze this question and observe the credit lines effect on the asymmetric adjustment speeds for over and under levered firms. Table 1 7 reports these leverage effects as if the firm moves from over to under target. For each of the four leverage effects I hold the free cash flow sign and the line of credit status constant, thus the first reported leverage effect is equal to 5 1 : the adjustmen t speed difference for over levered firms and under levered firms conditional on negative free cash flows and no line of credit. Generally, over levered firms adjust to target faster than respective under levered firms; this holds for all but one case. For example, conditional on negative cash flows and no line of credit, firms in Subset 1 (over  PAGE 34 34 levered) adjust 27.4 percentage points faster than firms in Subset 5 (under levered). The results are consistent with Byoun (2008), and are consistent with two (ind istinguishable) stories: different cost structures for equity and debt transactions, and firm preference for flexibility, if flexibility is greater when under levered. If Subset 1 firms are rebalancing primarily by issuing equity, and Subset 5 firms are ad justing to target primarily by borrowing, then some of the adjustment speed difference in Table 1 7 will reflect the different characteristics of transactions costs in those markets (e.g., fixed and variable cost components). If proportional transactions c osts of debt issuances are more convex compared to that of equity issuances, then equity issuances will close more of the leverage residual conditional on a transaction. More complete adjustment to target for over levered firms is also consistent with the management to flexibility. If the desire for financial flexibility is a first order effect for leverage adjustments (Graham and Harvey (2001)), and flexibility is greater when under levered, then firms adjust relatively aggressively if over levered and les s aggressively if under levered. Cash F low E ffects on A djustment S peeds Faulkender et al. (2008) focuses on the effects of large free cash flow outcomes on adjustment speeds. With my eight subsets I can also analyze this question, and the results are repor ted in Table 1 8 Cash flow effects are statistically significant at the 1% level in three out of the four cases. And in the other case the effect of moving from a negative free cash flow to a positive free cash flow if over levered without a line of cre dit (moving from Subset 1 to Subset 3) the bootstrapped z statistic maps to a p value of 0.126 (assuming normality), but it is statistically significant at the 10% level if we analyze only constrained firms (unreported). The signs of the statistically significant cash flow effects are consistent with the importance of holding cash and/or avoiding transactions in the equity market. In order to match the adjustment PAGE 35 35 speeds of negative cash flow firms, firms with positive cash flows must burn liquidity today or transact in the equity market today. For example, firms in Subset 2 must repay more debt or issue more equity to match the adjustment speed of firms in Subset 4. And firms in Subset 5 (Subset 6) must borrow less or issue more equity to match adjustment speeds of firms in Subset 7 (Subset 8). Choosing to hold more liquidity today highlights the importance of the firms trade off of the deviation from target leverage and its demand for liquidity. How A re F irms A djusting? The results thus far are consistent with the importance of transactions costs and demand for liquidity/flexibility for the firms management of capital structure. The results are intuitive, but we can learn more about the cash and leverage trade off by analyzing how the firms are making adj ustments. It is important to remember that a change in the line of credit balance is not a necessary condition for the existence of a line effect on adjustments to the leverage ratio. A necessary condition for such an effect is a change in firm behavior du e to the presence of a line. Generally, any action not in the firms choice set because of the absence of a line of credit can constitute a line effect (as defined here) if as a result the firms capital structure is affected. Observing these actions can t each us more about how the firm internalizes its management of leverage and liquidity. Table 1 10 reports means on the change in several firm characteristics by Subset (all scaled by year t book assets): net equity issues in the year t+1 cash flow statemen t, change in debt is the change in the balance sheets total debt from year t to year t+1 change in assets is the change in book assets, and change in cash is the change in balance sheet cash and equivalents.17 Direct your focus in Table 1 10 to compariso ns across Subset pairs (1, 2), (3, 4), (5, 6), and (7, 8) as the 17 For reference to variable definitions in Appendix A: net equity issues ( NEITA ), change in debt ( CHGDEBT ), change in assets ( CHGASSETS ), and change in cash ( CHGCASH ). PAGE 36 36 adjustment speed differences across these Subset pairs constitute the net line effects in Table 1 6 The most striking result is the net equity issuance differences across the Subset pairs (1, 2) and (5, 6). Firms with negative cash flows and no line of credit issue a great deal of equity Subset 1 (Subset 5) firms issue 21.1% (22.2%) of lagged assets. This result is interesting because the Subset 1 firms were estimated to close over one half the leverage gap, and they had the largest adjustment speed estimate among all Subsets. The equity issuance by firms in Subset 5 is particularly striking in that these firms are pushed away from target by such action. Subsets 1 and 5 are issuing nearly four times the amount of equity issued by firms in peer Subsets 2 and 6. Debt policies appear to be different across peer Subsets for Subset pairs (5, 6) and (7, 8). Subset 6 firms issue more debt than is issued by their peer firms in Subset 5 a difference of 7.6% of time t book assets. As a percentage of assets, Subset 8 firms issue three times the amount of debt issued by Subset 7 firms, and Subset 7 firms increase cash stocks by almost four times the cash increase for Subset 8 firms. These balance sheet changes are consistent with the line effect differences reported in Table 1 6 The results in Table 1 10 are consistent with the story described herein, but the large differences in equity issuance across Subsets in pairs (1, 2) and (5, 6) warrant an even closer look. It could be that firms in Subsets 1 and 5 simply have larger negative cash flows than firms in Subset 2 and 6, and if so, it could be that the equity issuance is merely a response to the level of the cash flow shock and in no way related to th e absence of a line. Alternatively, it could be that the presence of the line of credit in Subsets 2 and 6 enables these firms to avoid issuing as much equity following the negative cash flows. T he next few tables report a decompos ition of the year t+1 cas h flow statement scaled by year t+1 cash flows. Thus, the statistics here can be interpreted as the response to a one dollar cash flow. Mean differences are reported in the PAGE 37 37 column labeled diff: p q and t statistics and z statistics report statistical tes ts for differences in means and medians, respectively. A summary of firm actions for peer Subset pairs is given for statistics if they are statistically different at the mean or median, and these summaries should be interpreted as Subset p s differential r esponse to one dollar of positive or negative cash flow (depending on the Subset in question) relative to Subset q s response. For example, Subsets 1, 2, 5, and 6 (3, 4, 7, and 8) are responding to negative (positive) FCF, therefore the action summary in T able 1 11 (Subsets 1 and 2) is in response to a one dollar negative cash flow. Table 1 13 (Subsets 3 and 4) lists an action summary in response to a one dollar positive cash flow. Tables 1 11 and 1 13 show that the equity issuance results in Table 1 10 for Subsets 1 and 5 are not just a result of a difference in the level of negative cash flows within peer Subsets. Firms in Subset 1 (Subset 5) issue 20.4 (19.0) cents more equity per one dollar of negative cash flow compared to firms in Subset 2 (Subset 6). These results are very large, statistically significant at the 1% level, they are consistent with the signs of the line effects in Table 1 6 and they are consistent with the story that firms with lines of credit are able to avoid issuing equity following negative cash flows. Another interesting result from analysis in Table 1 6 is the lack of a line effect for firms over levered with positive cash flows (Subsets 3 and 4); a result which could be due to firms in Subset 4 saving cash out of cash flows instea d of repaying the line of credit. Here in Table 1 12 we see that Subset 4 firms invest 21 cents more per dollar of positive cash flow compared to Subset 3 firms. Instead of retiring debt, Subset 4 firms invest 77 cents and retain 15 cents of each dollar of cash flow. In Table 1 6 we found that firms under levered with positive cash flows adjust back to target 13.3 percentage points faster if there is a line of credit. Table 1 14 shows that these Subset PAGE 38 38 8 firms borrow (invest) 32 (41) cents more per dollar of positive cash flow compared to firms in Subset 7. Thus Subset 8 firms combine the free cash flows with new debt issues to make new investment whereas firms without lines (Subset 7) save 22 cents more per dollar of cash flow relative to Subset 8 firms. Ro bustness of R esults Endogeneity of the e xistence of a c redit l ine The choice of a line of credit is perhaps endogenous to the leverage change, and thus I use a Roy Model (Roy (1951)) to control for the potential self selection bias.18 The estimates on both inverse Mills ratios (Mills (1926)) are small in magnitude and not statistically different from zero, and the adjustment speed estimates are largely unchanged from the OLS model.19 The switching model equation for firms with no line of credit returns adju stment speeds slightly smaller in magnitude compared to those reported in Table 1 6 but qualitatively nothing changes; the line effects are robust to use of the selection model. Financial constraint Much recent capital structure work has emphasized the im portance of whether the firm is financially constrained. The distinction might matter for this study if unconstrained firms have lower costs of adjustment or if unconstrained firms have indeterminate cash policies (Acharya et al (2007)). The literature ha s used the firms payout ratio, size, and existence of bond and commercial paper ratings to define whether the firm is constrained. I follow Acharya et al 18 In an effort to sa ve space I do not tabulate the results the selection model. For the first stage probit regression to predicting the existence of a line of credit, the dependent variable is HASLINE at year end t+1 and the regressors are the Xt target determinants from (1 2). Instruments for the unobserved selection determinants are industry cash flow volatility, SIGCF (defined in Appendix A), the twodigit SIC average of HASLINE at year end t ( HLAVG2 ), and RATED a dummy variable equal to one if the firm has a debt ratin g in Compustat at year end t 19 This result on the selection model is one in which the target truncation at [0,1] matters in preliminary tests (see footnote 11). If I omit the [0,1] truncation the bootstrapped z statistics for the inverse Mills ratios are 2.45 ( 1.60) for the HASLINE =0 ( HASLINE =1) equations. Adjustment speeds are lower in magnitude than those reported here, but line effects are qualitatively the same. PAGE 39 39 (2007) and generate five definitions of financial constraint. If the firm is in the bottom or top (b ottom) three deciles by year according to payout ratio then the observation is noted as constrained (unconstrained), respectively. The same method is used for firm size. For bond and commercial paper ratings, the observation is noted as constrained if ther e is never a bond or commercial paper rating for the firm, respectively, in the Compustat data and the firm has debt outstanding at year end. The firm is noted as unconstrained if there is a respective rating at any point in the data. The fifth definition of constraint is an attempt to summarize the previous four where observations are deemed to be constrained or unconstrained if at least three of the respective Acharya et al. definitions based on payout, size, bond ratings, and commercial paper ratings ind icate as such. The adjustment speeds for constrained firms (41% of the sample) exhibit qualitatively the same results as in the full sample OLS. The most noteworthy exceptions are the estimates for Subsets 4 and 7. The adjustment speed estimate for over le vered firms with positive cash flows and a line of credit (Subset 4) increases from 0.433 (reported in Table 1 6 ) to 0.488. This result appeals to the story that relative to unconstrained firms, constrained firms get greater benefit from being closer to le verage target, and conditional on having positive cash flows with which to adjust, the firm will close more of the leverage gap. The estimate for under levered firms with positive cash flows and no line of credit (Subset 7) decreases from 0.135 (reported i n Table V ) to 0.086. Again, a story can be told to explain in the context of financial constraint. The credit markets are more likely to ration constrained firms (relative to unconstrained firms) in Subset 7 and thus these firms are less likely to distribu te positive free cash flows shareholders. Results for unconstrained firms suggest slower overall adjustment speeds if over levered, relative to constrained firms, and faster adjustment speeds if under levered. But conditional on PAGE 40 40 being over or under levered the pattern of adjustment speeds across Subsets for unconstrained firms are qualitatively similar to those estimated for constrained firms. Slower average adjustment speed for over levered unconstrained firms is consistent with the deviation from target leverage being less disruptive for firm operating policies. If by maintaining a small leverage residual the firm protects financial flexibility, then intuitively we should see faster adjustment for constrained firms, ceteris paribus because by definition the unconstrained firms have more flexibility orthogonal to leverage effects. In plain words, the unconstrained firms are not as concerned about being over levered (relative to constrained firms). Conclusion The purpose of this study is not to simply deter mine whether lines of credit affect adjustment speeds to target leverage ratios. The intent is to determine whether the firms adjustment decisions are affected significantly by transactions costs in predictable ways. Lines of credit data offer an interest ing framework for analyzing this question: they have widespread use, they are flexible by design, and they directly touch the firms liquid assets and debt by construction. But perhaps the most important feature of credit lines for this study is that the y link two parties borrower and lender who have well known and well researched frictions. The intrinsic flexibility offered to the borrower by the credit line does not appeal to the lenders concern for moral hazard and adverse selection. Conditional o n establishing the credit line, the firm agrees to a number of contingencies built into the contract to protect the lenders position and to commit the firm to compatible incentives ex post But contracting is not perfect, and the contracts contingencies lead to the lack of substitutability of cash and debt in some cases. Thus, PAGE 41 41 given the importance of the credit line in firms management of flexibility, leverage, and liquidity, we have a unique opportunity to learn more about the firms behavior.20 I use da ta on the existence of a corporate line of credit to show that transactions costs and the firms demand for liquidity are very important for the firms management of capital structure in the cross section. I extend the literatures constant speed dynamic p artial adjustment model to a two stage model for estimation of cross sectional variation in adjustment speeds, and I use a bootstrap program to address the presence of generated regressors in the second stage. My model for variable adjustment speeds allows for inferences on how the lack of substitutability of cash and debt influences the speed with which the firm adjusts leverage to target. I analyze the cash flow statement to see exactly how some of the differential adjustment speeds are attained. Results support the notion that transactions costs play a nontrivial role in the management of capital structure. Specifically, for under levered firms, I find that the existence of a corporate credit line is associated with adjustment speeds 50 77% greater than s peeds estimated for under levered firms without a line. These line effects are large given that under levered firms without access to a credit line close 0.135 (0.247) of the distance from target when realizing positive (negative) free cash flows. The bene fit of lower transactions costs associated with a credit line is offset by the firms demand for liquidity if the firm is over levered; credit lines are associated with either no adjustment speed difference or slower adjustment speeds, depending on the fir ms demand for cash. For over levered firms in situations when the firms demand for cash is likely elevated, the access to a credit line attenuates the speed of adjustment but enables the firm to avoid raising cash by issuing equity a more costly transa ction than raising cash with a credit line draw. Over 20 A working paper by Kahl, Shivdasani, and Wang (2008) analyzes whether firms use the comm ercial paper market to enhance flexibility. PAGE 42 42 levered firms with excess free cash flows choose realized liquidity over a larger adjustment to target leverage, and thus the line is not associated with faster adjustment to target leverage. The result s highlight the importance of considering the firms demand for flexibility and liquidity when studying the capital structure decisions of the firm. The literatures partial adjustment equation model ignores the firms demand for liquidity. The equation ma kes no prediction for a leverage change if the firm is at its target but has a cash stock deficit or surplus. The implicit assumption is that cash is always equal to negative debt. The results here indicate that this assumption can be important; when the s ubstitutability of cash and negative debt fails to hold what follows has important implications for the firms management of capital structure. More work needs to be done to understand the role of credit lines in the interaction of firm financial and inves tment policies. DeAngelo, DeAngelo, and Whited (2008) model leverage rebalancing activity with endogenous investment and the use of transitory debt to fund investment shocks. They show theoretically that cross sectional heterogeneity in investment shocks a nd the use of transitory debt can generate leverage dynamics consistent with the empirical results documented in Lemmon, Roberts, and Zender (2008). This paper takes a first step toward understanding the role of credit lines in the management of capital st ructure. An interesting next step is to analyze the impact of investment shocks on the credit lines role in managing liquidity and leverage. PAGE 43 43 Subset Adjustment speed estimate Leverage residual Cash flow outcome Line of credit status 1 1 Over levered < 0 FCF < 0 HASLINE = 0 2 2 Over levered < 0 FCF < 0 HASLINE = 1 3 3 Over levered < 0 FCF HASLINE = 0 4 4 Over levered < 0 FCF HAS LINE = 1 5 5 Under levered 0 FCF < 0 HASLINE = 0 6 6 Under levered 0 FCF < 0 HASLINE = 1 7 7 Under levered 0 FCF HASLINE = 0 8 8 Under levered 0 FCF HASLINE = 1 Figure 1 1. Descript ion of m utually e xclusive s ubsets of the d ata. Observations are divided into eight mutually exclusive subsets according to the estimated leverage residual ( R ), the firms free cash flow outcome at time t+1 ( FCF ), and whether the firm has a line of credit in place at time t+1 Subsets 1 through 4 (5 through 8) are for over levered (under levered) firms. Subsets 1, 2, 5, and 6 (3, 4, 7, and 8) are for those with negative (positive) FCF. Odd (Even ) numbered subset s are for firms without (with) lines of credit, HASLINE = 0 ( HASLINE = 1). The s are to be estimates of adjustment speed and the subscripts identify the Subset of the data associated with the estimate. PAGE 44 44 Leverage residual FCF outcome Leverage effect Net line effect Cash Negative Negative Cash = Debt Zero Negative Cash Negative Any Cash = Debt Zero Positive Cash Positive Positive Cash = Debt Zero Positive Cash Positive Any Cash = Debt Zero Negative Underlevered Overlevered Liquidity management effect FCF < 0 FCF > 0 FCF < 0 FCF > 0 Negative Positive Positive Negative Figure 1 2. Depiction of leverage and liquidity effects of existence of a revolving line of credit on a firms adjustment speed to target leverage. The Figure illustrates some of the capital structure adjustment issues facing over (under ) levered firms based on the estimated leverage residual, = + 1 at time t + 1 is the firms estimated target leverage ratio for time t+1 is the firms observed leverage ratio at time t FCF is a measure of the free cash flow over which the firm has discretion. Negative cash flows will mechanically i ncrease leverage if left unmanaged, thus the upper portion of the Figure (within over and under levered rows) is for FCF < 0. The lower portion of the Figure (within over and under levered rows) is for FCF > 0, which tends to decrease leverage if left un managed. The column for Leverage effect shows the credit lines effect on the adjustment of leverage to target if used in response to FCF without regard to liquidity effects. For example, following a negative cash flow a draw on the line will increase le verage decreasing (increasing) adjustment speed if over (under ) levered. The column for Liquidity management effect shows the effect on adjustment to target leverage due to the use of the line of credit for cash management purposes only. This effect is theoretically zero if cash and negative debt are perfect substitutes. If cash and negative debt are not perfect substitutes then liquidity management effects through the line of credit act to increase leverage, which decreases (increases) adjustment speed if over (under ) levered. The last column summarizes the expected sign of the total net effect of having a line of credit on adjustment speed to target leverage; the combination of the leverage effects and the liquidity management effects manifested thro ugh credit line existence. PAGE 45 45 Table 1 1. Summary s tatistics. This table reports summary statistics on all firms used in the estimation of adjustment speeds via Equation 1 3. LEVt is the firms market leverage ratio at fiscal year end t MB is the firms mar ket value of assets scaled by book value of assets. LNTA is the natural log of book assets expressed in 1983 U.S. dollars. RDD is a dummy variable equal to one if the firm reports non zero research and development expense and is zero otherwise or if resear ch and development expense is missing. LEV_MED is the median market leverage by Fama French industry groupings by year. The following variables are scaled by book assets at time t EBITTA is earnings before interest and taxes expense, DEPTA is depreciation and amortization expense, FATA is net fixed assets, RDTA is research and development expense or zero if such expense is missing, FCF is free cash flow in year t+1 CASHTA is cash and equivalent assets. N Mean Standard deviation 25th pctile Median 75th pctile LEVt31,513 0.237 0.247 0.021 0.159 0.379 EBITTA 31,513 0.035 0.349 0.049 0.060 0.118 MB 31,513 1.859 2.243 0.810 1.190 1.989 DEPTA 31,513 0.051 0.042 0.026 0.042 0.062 LNTA 31,513 19.427 2.225 17.817 19.371 20.952 FATA 31,513 0.274 0.229 0.093 0.207 0.394 RDD 31,513 0.484 0.500 0.000 0.000 1.000 RDTA 31,513 0.056 0.121 0.000 0.000 0.059 LEV_MED 31,513 0.182 0.130 0.052 0.178 0.268 FCFTA 31,513 0.055 0.305 0.074 0.021 0.078 HASLINE 31,513 0.785 0.411 1.000 1.000 1.000 CASHTA 31,513 0.175 0.221 0.020 0.077 0.246 PAGE 46 46 Table 1 2. Summary s tati stics by HASLINE s tatus. This table reports summary statistics on all firms used in the estimation of adjustment speeds via Equation 1 3 according to whether the firm has a line of credit in place at year end t+1 The t (z ) statistics are tests for equal means (medians) across HASLINE status. LEVt is the firms market leverage ratio at fiscal year end t MB is the firms market value of assets scaled by book value of assets. LNTA is the natural log of book assets expressed in 1983 U.S. dollars. RDD is a d ummy variable equal to one if the firm reports non zero research and development expense and is zero otherwise or if research and development expense is missing. LEV_MED is the median market leverage by Fama French industry groupings by year. The following variables are scaled by book assets at time t EBITTA is earnings before interest and taxes expense, DEPTA is depreciation and amortization expense, FATA is net fixed assets, RDTA is research and development expense or zero if such expense is missing, FCF is free cash flow in year t+1 CASHTA is cash and equivalent assets. N Mean Median N Mean Median t statistic z statistic LEVt6,739 0.119 0.010 24,774 0.269 0.204 49.8 60.0 EBITTA 6,739 0.239 0.071 24,774 0.021 0.071 38.9 48.7 MB 6,739 2.935 1.761 24,774 1.566 1.109 31.6 38.2 DEPTA 6,739 0.052 0.037 24,774 0.051 0.043 2.2 13.1 LNTA 6,739 18.137 17.964 24,774 19.778 19.735 55.5 53.3 FATA 6,739 0.203 0.115 24,774 0.293 0.231 29.6 38.3 RDD 6,739 0.667 1.000 24,774 0.434 0.000 35.6 34.0 RDTA 6,739 0.132 0.064 24,774 0.036 0.000 41.6 50.4 LEV_MED 6,739 0.120 0.056 24,774 0.199 0.200 47.1 47.0 FCFTA 6,739 0.250 0.098 24,774 0.003 0.032 41.2 51.4 CASHTA 6,739 0.391 0.353 24,774 0.117 0.052 76.8 77.9 HASLINE = 0 HASLINE = 1 PAGE 47 47 Table 1 3. Results from estimating the partial adjustment model, Equation 1 2, via pooledOLS, LSDV, Arellano and Bonds (1991) difference GMM estimator, and Blundell and Bonds (1998) system GMM estimator (GMM BB). The estimates on target determinants via GMM BB are used to calculate LEV *t+1 used in the second stage estimation of Equation 1 6 The dependent variable is LEVt+1, the observed market leverage ra tio at time t+1 LEVt is the firms market leverage ratio at fiscal year end t MB is the firms market value of assets scaled by book value of assets. LNTA is the natural log of book assets expressed in 1983 U.S. dollars. RDD is a dummy variable equal to one if the firm reports nonzero research and development expense and is zero otherwise or if research and development expense is missing. LEV_MED is the median market leverage by Fama French industry groupings by year. The following variables are scaled b y book assets at time t EBITTA is earnings before interest and taxes expense, DEPTA is depreciation and amortization expense, FATA is net fixed assets, and RDTA is research and development expense or zero if such expense is missing. All four specification s include year effects. p values for tests for first and secondorder autocorrelation of first differenced residuals are reported. Standard errors are robust to heteroskedasticity, intra firm autocorrelation, and are adjusted for finite sample bias accordi ng to Windmeijer (2005). ***, **, indicate statistical significance at the 1, 5, and 10% levels, respectively (twotailed) PAGE 48 48 Constant adjustment speed (1): LEVt0.831***0.448***0.534***0.775*** (0.00) (0.01) (0.03) (0.02) Implied annual adjustment speed ( )0.169 0.552 0.466 0.225 Target determinants (year and fixed effects not reported):EBITTA 0.024***0.040***0.023 0.023** (0.00) (0.01) (0.02) (0.01)MB 0.001***0.000 0.000 0.000(0.00) (0.00) (0.00) (0.00)DEPTA 0.138***0.099**0.301***0.191*** (0.02) (0.04) (0.10) (0.07)LNTA 0.002***0.031***0.052***0.005** (0.00) (0.00) (0.01) (0.00)FATA 0.019***0.047***0.113**0.001(0.00) (0.01) (0.05) (0.02)RDD 0.012***0.005 0.042 0.017* (0.00) (0.01) (0.03) (0.01)RDTA 0.048***0.036**0.037 0.073*** (0.01) (0.01) (0.05) (0.03)LEV_MED 0.065***0.151***0.158***0.069*** (0.01) (0.02) (0.04) (0.03)Observations 33,140 33,140 33,140 33,140 Firms 5,181 5,181 5,181 5,181 Adjusted R20.855 0.282 OLS LSDV GMM ArellanoBond GMM BlundellBond PAGE 49 49 Table 1 4. Summary of results from estimation of target leverage This table reports results from esti mating Equation 1 2 and calculating the firms target ratio, + 1 Subsets 1 through 8 are mutuallyexclusive and describe the firms combination of leverage residual ( = + 1 ), free cash flow ( FCF ), and line of credit status (HASLINE ). + 1 is the calculated target ratio at time t for leverage at time t+1 according to + from Equation 1 2. is the firms leverage residual at time t based on + 1 HASLIN E is a dummy variable equal to one if the firm has a line of credit in place at time t+1 and equal to zero otherwise. Only firms with one year book asset growth of less than 100% are reported. Subset DescriptionSubset N LEV* R HASLINEt Overlevered / FCF < 0 / HASLINEt+1 = 0 1 1,474 0.120 0.153 0.216 Overlevered / FCF < 0 / HASLINEt+1 = 1 2 3,806 0.234 0.171 0.946 Overlevered / FCF 0 / HASLINEt+1 = 0 3 666 0.099 0.119 0.243 Overlevered / FCF 0 / HASLINEt+1 = 1 4 8,058 0.161 0.134 0.976 Underlevered / FCF < 0 / HASLINEt+1 = 0 5 3,158 0.219 0.153 0.144 Underlevered / FCF < 0 / HASLINEt+1 = 1 6 4,671 0.456 0.204 0.901 Underlevered / FCF 0 / HASLINEt+1 = 0 7 1,472 0.142 0.089 0.151 Underlevered / FCF t+1 = 1 8 8,208 0.345 0.159 0.967 All 31,513 0.263 0.026 0.786 PAGE 50 50 Table 1 5. Transition matrix for leverage, cash flow, and line of credit subsets This table is a transition matrix reporting probabilities of belonging to Subset q at time t+1 conditional on belonging to Subset p at time t Here Subsets 1 through 8 are mutuallyexclusive and describe the firms combination of leverage residual ( = ), free cash flow ( FCF), and line of credit status ( HASLINE ) at time t +1 Rows (Columns) indicate Subset assignment at time t ( t+1 ). The three right most columns summarize the probability of maintaining status across all three dimensions, but separately, thus 1 pr (maintaining status) is the probability of migrating Subsets due to a change in the respective leverage residual, free cash flow, or line of credit status. Only firms w ith one year asset growth of less than 100% are reported. Subset at t N N/Total 1 2 3 4 5 67 8 Leverage residual Cash flow Line status 1 1,059 0.042 0.454 0.078 0.076 0.033 0.257 0.047 0.048 0.007 0.641 0.837 0.835 2 3,014 0.119 0.048 0.413 0.009 0.273 0.023 0.144 0.004 0.087 0.743 0.627 0.917 3 513 0.020 0.097 0.043 0.341 0.111 0.082 0.018 0.242 0.066 0.593 0.760 0.762 4 6,516 0.258 0.002 0.095 0.010 0.582 0.001 0.041 0.006 0.263 0.689 0.861 0.982 5 2,359 0.093 0.145 0.020 0.017 0.008 0.613 0.093 0.091 0.013 0.810 0.871 0.866 6 3,696 0.146 0.025 0.202 0.005 0.097 0.056 0.453 0.008 0.153 0.670 0.736 0.906 7 1,150 0.046 0.034 0.007 0.082 0.027 0.160 0.042 0.546 0.103 0.850 0.757 0.822 8 6,948 0.275 0.002 0.060 0.003 0.225 0.009 0.144 0.013 0.544 0.710 0.785 0.973 Total 25,255 1.000 FCF 0 Overlevered Underlevered Probability of Subset at t+1 Probability of maintaining FCF < 0 FCF 0 FCF < 0 PAGE 51 51 Table 1 6. Adjustmen t speed regression of Equation 1 6 This table reports regression results of adjustment speeds to target leverage ratios via Equation 1 6 and implie d line effects. Each is estimated on an interaction of three dummy variables and the estimated leverage residual ( = + 1 ). The three dummy variables constituting the eight mutually exclusive subsets are the sign of the leverage residual (whether the firm is over or under levered relative to target), the sign of free cash flow ( FCF), and the absence or presence of the line of credit ( HASLINE ) at time t+1 Target determinants are: MB is the market to b ook ratio, LNTA is the natural log of book assets expressed in 1983 U.S. dollars, RDD is a dummy variable equal to one if the firm has nonzero research and development expense at time t and equal to zero otherwise or if missing, LEV_MED is the industry med ian LEVt by Fama French 49 industry groupings. The following are scaled by book assets at time t : earnings before interest and tax expense ( EBITTA ), depreciation and amortization expense ( DEPTA ), net fixed assets ( FATA), and research and development expens e ( RDTA ). Standard errors for coefficients and linear combinations of coefficients (e.g., line effects) are bootstrapped for 499 replications. This bootstrap includes both stages of the estimation: the target Equation 1 2 and the adjustment speed E quation 1 5 for each iteration. Firm year observations realizing one year asset growth of over 100% are excluded from these regressions. ***, **, indicate statistical significance at the 1, 5, and 10% levels, respectively (two tailed) PAGE 52 52 Overlevered relative to target Negative free cash flow Subset 1: Firms with no credit line 10.520*** (0.03)Subset 2: Firms with a credit line 20.367***210.154*** (0.02) (0.03) Positive free cash flow Subset 3: Firms with no credit line 30.444*** (0.04)Subset 4: Firms with a credit line 40.433***430.011 (0.02) (0.04) Underlevered relative to target Negative free cash flow Subset 5: Firms with no credit line 50.247*** (0.02)Subset 6: Firms with a credit line 60.372***650.126*** (0.02) (0.02) Positive free cash flow Subset 7: Firms with no credit line 70.135*** (0.03)Subset 8: Firms with a credit line 80.239***870.104*** (0.01) (0.03)N 31,513 Firms 5,171 AdjustedR20.341 Bootstrap Reps 499 Adjustment Speeds Line Effects PAGE 53 53 Table 1 7 Leverage effects on adjustment speeds: Moving from over levered to under levered This table reports estimated leverage effects on adjustment speeds. Standard errors for coefficients and linear combinations of coefficients (e.g., line effects) are bootstrapped for 499 replications. This bootstrap includes both stages of the estimation: the target Equation 1 2 and the adjustment speed E quation 1 5 for each iteration. ***, **, indicate statistical significance at the 1, 5, and 10% levels, respec tively (two tailed) Negative free cash flow Firms with no credit line 510.274*** (0.03)Firms with a credit line 620.006 (0.02) Positive free cash flow Firms with no credit line 730.309*** (0.05)Firms with a credit line 840.194*** (0.01) PAGE 54 54 Table 1 8 Cash flow effects on adjustment speeds: Moving from negative to positive cash flow: This table reports estimated cash flow effects on adjustment speeds. Standard errors for coefficients and linear combinations of coefficients (e.g., line effects) are bootstrapped for 499 replications. This bootstrap includes both stages of the estimation: the target Equation 1 2 and the adjustment speed E quation 1 5 for each iteration. ***, **, indicate statistic al significance at the 1, 5, and 10% levels, respectively (two tailed) Overlevered relative to target Firms with no credit line 310.076 (0.05)Firms with a credit line 420.067*** (0.02) Underlevered relative to target Firms with no credit line 750.112*** (0.03)Firms with a credit line 860.133*** (0.01) PAGE 55 55 Table 1 9. Net line effect predictions and empirical results The table is a reproduction of Figure 3 with the last column tabulating the estimates of the net line effects from Equation 1 6 reported in Table V, Panel A. The leverage residual is = + 1 at time t + 1 is the firms estimated target leverage ratio for time t+1 is the firms observed leverage ratio at time t FCF is a measure of the free cash flow over which the fir m has discretion. Negative cash flows will mechanically increase leverage if left unmanaged, thus the upper portion of the Figure (within over and under levered rows) is for FCF < 0. The lower portion of the Figure (within over and under levered rows) is for FCF > 0, which tends to decrease leverage if left unmanaged. The column for Leverage effect shows the credit lines effect on the adjustment of leverage to target if used in response to FCF without regard to liquidity effects. For example, following a negative cash flow a draw on the line will increase leverage decreasing (increasing) adjustment speed if over (under ) levered. The column for Liquidity management effect shows the effect on adjustment to target leverage due to the use of the line of credit for cash management purposes only. This effect is theoretically zero if cash and negative debt are perfect substitutes. If cash and negative debt are not perfect substitutes then liquidity management effects through the line of credit act to increa se leverage, which decreases (increases) adjustment speed if over (under ) levered. The last column summarizes the theoretical net line effect (combination of the leverage effects and the liquidity management effects manifested through credit line existen ce), and reports the estimated net line effects from Equation 1 6 Expected Empirical Cash Negative Negative Cash = Debt Zero Negative Cash Negative Any Cash = Debt Zero Positive Cash Positive Positive Cash = Debt Zero Positive Cash Positive Any Cash = Debt Zero Negative Leverage residual FCF outcome Leverage effect Liquidity management effect Underlevered FCF < 0 Positive FCF > 0 Negative Overlevered FCF < 0 Negative FCF > 0 Positive 0.154*** 0.011 0.126*** 0.104*** Net line effect PAGE 56 56 T able 1 10. Balance sheet changes year t to year t+1 scaled by book assets at year t by s ubset This table reports selected means by Subset. Subsets 1 through 8 are mutuallyexclusive and describe the firms combination of leverage residual ( = + 1 ), free cash flow ( FCF ), and line of credit status ( HASLINE ). Subsets 1 4 (5 8) are for firms with < 0 ( 0 ). Subsets 1, 2, 5, and 6 (3, 4, 7, and 8) are for firms with FCF < 0 ( FCF t+1 Odd (Even ) numbered Subsets are for firms with HASLINE = 0 ( HASLINE = 1) at time t+1 Net equity issues is NEITA in Appendix A variable definitions: net equity issues from the cash flow statement in year t+1 scaled by book assets at time t Change in debt is CHGDEBT : short term and long term debt at time t+1 minus short term and long term debt at time t scaled by book assets at time t Change in assets is CHGASSETS : book ass ets at time t+1 minus book assets at time t scaled by book assets at time t Change in cash is CHGCASH : cash stocks at time t+1 minus cash stocks at time t scaled by book assets at time t Overlevered relative to target Negative free cash flow Subset 1: Firms with no credit line 1,474 0.211 0.066 0.087 0.000 Subset 2: Firms with a credit line 3,806 0.054 0.040 0.070 0.003 Positive free cash flow Subset 3: Firms with no credit line 666 0.016 0.029 0.133 0.056 Subset 4: Firms with a credit line 8,058 0.007 0.010 0.101 0.022 Underlevered relative to target Negative free cash flow Subset 5: Firms with no credit line 3,158 0.222 0.047 0.095 0.060 Subset 6: Firms with a credit line 4,671 0.061 0.123 0.049 0.025 Positive free cash flow Subset 7: Firms with no credit line 1,472 0.006 0.019 0.173 0.066 Subset 8: Firms with a credit line 8,208 0.004 0.057 0.168 0.015 Net equity issues Change in debt Change in assets Change in cash N PAGE 57 57 Table 1 11. Cash flow statement analys is by s ubset over levered firms with negative cash flows This table reports a summary of the cash flow statement for over levered firms at year t+1 with negative FCF Cash flow variables are defined in Appendix A. Net equity issues is NEITCF net debt i ssuance is NDITCF net investment is INVTCF other cash flows is OTHCF and CHGCASTF is the change in cash. CHGCASHTCF is scaled by SCFCF NEITCF NDITCF INVTCF and OTHTCF are scaled by SCFCF SCFCF is free cash flow from the firms cash flow statement at year t+1 Thus CHGCAS H TCF + NEIT CF + NDITCF + INVTCF + OTHCF = 1, and means reported can be interpreted as the firms response to a one dollar cash flow Note that these firms have FCF < 0. F irm actions in response to the one dollar cash flow are summar ized in the Actions column according to the sign of FCF The diff( p q ) column is the result of subtracting statistic for Subset q from that for Subset p For example, the column labeled diff(1,2) is the difference in means for a statistic in Subsets 1 and 2. The column labeled t (diff) is a t statistic reporting results of a test with the null hypothesis diff( p q ) is zero. The column labeled z (p q ) reports the results of a test for equal medians. The Actions column reports the Subsets response t o the one dollar of negative cash flow, and the p relative to q sub column summarizes Subset p s action relative to Subset q s action. ***, **, indicate statistical significance at the 1, 5, and 10% levels, respectively (two tailed) Overlevered firms with negative free cash flow N 1,474 3,806 Net equity issues 0.309 0.105 0.2044.82***12.56 ***0.20 moreNet debt issues 0.054 0.105 0.0500.55 4.69 ***0.05 lessNet investment 0.174 0.218 0.0430.41 3.09 ***0.04 lessOther cash flows 0.481 0.461 0.0200.22 0.70Change in cash 0.040 0.064 0.0240.34 0.43 Use cash stocks 1 Subsets 1 (no credit line) & 2 (credit line) t (diff) z (1,2) 2 1 2 diff: 12 Action Issue equity Issue debt Disinvest Shrink balance sheet 1 relative to 2 PAGE 58 58 Table 1 1 2 Cash flow statement analysis by s ubset over levered firms with positive cash flow. This table reports a summary of the cash flow statement for over levered firms at year t+1 with positive cash flows. Cash flow variables are define d in Appendix A. Net equity issues is NEITCF net debt issuance is NDITCF net investment is INVTCF other cash flows is OTHCF and CHGCASTF is the change in cash. CHGCASHTCF is scaled by SCFCF NEITCF NDITCF INVTCF and OTHTCF are scaled by SCFCF SCFC F is free cash flow from the firms cash flow statement at year t+1 Thus CHGCAS H TCF + NEIT CF + NDITCF + INVTCF + OTHCF = 1, and means reported can be interpreted as the firms response to a one dollar cash flow Note that these firms have FCF > 0. Firm actions in response to the one dollar cash flow are summarized in the Actions column according to the sign of FCF The diff( p q ) column is the result of subtracting statistic for Subset q from that for Subset p For example, the column labeled diff(3,4) is the difference in means for a statistic in Subsets 3 and 4. The column labeled t (diff) is a t statistic reporting results of a test with the null hypothesis diff( p q ) is zero. The column labeled z (p q ) reports the results of a test for equal media ns. The Actions column reports the Subsets response to the one dollar of positive cash flow, and the p relative to q sub column summarizes Subset p s action relative to Subset q s action. ***, **, indicate statistical significance at the 1, 5, and 1 0% levels, respectively (two tailed) Overlevered firms with positive free cash flow N 666 8,058 Net equity issues 0.064 0.066 0.0020.07 1.29Net debt issues 0.210 0.144 0.0661.07 0.59Net investment 0.563 0.768 0.2053.11***3.89 ***0.21 lessOther cash flows 0.070 0.025 0.0450.80 1.26Change in cash 0.237 0.145 0.0912.20**3.66 ***0.09 more Subsets 3 (no credit line) & 4 (credit line) t (diff) 3 4 diff: 34 z (3,4) Invest Grow balance sheet Save cash Action Issue equity Retire debt 3 4 3 relative to 4 PAGE 59 59 Table 1 1 3 Cash flow statement analysis by s ubset under levered firms with negative cash flows. This table reports a summary of the cash flow statement for under levered firms at year t +1 with negative FCF. Cash flow variables are defined in Appendix A. Net equity issues is NEITCF net debt issuance is NDITCF net investment is INVTCF other cash flows is OTHCF and CHGCASTF is the change in cash. CHGCASHTCF is scaled by SCFCF NEITCF NDITCF INVTCF and OTHTCF are scaled by SCFCF SCFCF is free cash flow from the firms cash flow statement at year t+1 Thus CHGCAS H TCF + NEIT CF + NDITCF + INVTCF + OTHCF = 1, and means reported can be interpreted as the firms response to a one dollar ca sh flow Note that these firms have FCF < 0. Firm actions in response to the one dollar cash flow are summarized in the Actions column according to the sign of FCF The diff( p q ) column is the result of subtracting statistic for Subset q from that for Subset p For example, the column labeled diff( 5,6 ) is the difference in means for a statistic in Subsets 5 and 6. The column labeled t (diff) is a t statistic reporting results of a test with the null hypothesis diff( p q ) is zero. The column labeled z (p q ) reports the results of a test for equal medians. The Actions column reports the Subsets response to the one dollar of negative cash flow, and the p relative to q sub column summarizes Subset p s action relative to Subset q s action. ***, **, indicate statistical significance at the 1, 5, and 10% levels, respectively (two tailed) Underlevered firms with negative free cash flow N 3,158 4,671 Net equity issues 0.364 0.174 0.1906.74***13.97 ***0.19 moreNet debt issues 0.087 0.014 0.0721.25 2.96 ***0.07 moreNet investment 0.058 0.216 0.1581.84*2.84 ***0.16 lessOther cash flows 0.295 0.500 0.2063.35***7.14 ***0.21 lessChange in cash 0.196 0.090 0.1061.95*6.82 ***0.11 more Shrink balance sheet Use cash stocks Subsets 5 (no credit line) & 6 (credit line) t (diff) z (5,6) Issue debt diff: 56 5 6 Disinvest 5 6 5 relative to 6 Action Issue equity PAGE 60 60 Table 1 1 4. Cash flow statement analysis by s ubset under levered firms with positive cash flow. This table reports a summary of the ca sh flow statement for under levered firms at year t+1 with positive cash flows. Cash flow variables are defined in Appendix A. Net equity issues is NEITCF net debt issuance is NDITCF net investment is INVTCF other cash flows is OTHCF and CHGCASTF is th e change in cash. CHGCASHTCF is scaled by SCFCF NEITCF NDITCF INVTCF and OTHTCF are scaled by SCFCF SCFCF is free cash flow from the firms cash flow statement at year t+1 Thus CHGCAS H TCF + NEIT CF + NDITCF + INVTCF + OTHCF = 1, and means reported ca n be interpreted as the firms response to a one dollar cash flow Note that these firms have FCF > 0. Firm actions in response to the one dollar cash flow are summarized in the Actions column according to the sign of FCF The diff( p q ) column is the r esult of subtracting statistic for Subset q from that for Subset p For example, the column labeled diff( 7,8 ) is the difference in means for a statistic in Subsets 7 and 8. The column labeled t (diff) is a t statistic reporting results of a test with th e null hypothesis diff( p q ) is zero. The column labeled z (p q ) reports the results of a test for equal medians. The Actions column reports the Subsets response to the one dollar of positive cash flow, and the p relative to q sub column summarizes Su bset p s action relative to Subset q s action. ***, **, indicate statistical significance at the 1, 5, and 10% levels, respectively (two tailed) Underlevered firms with positive free cash flow N 1,472 8,208 Net equity issues 0.020 0.031 0.0110.45 4.23 ***0.01 moreNet debt issues 0.101 0.422 0.3216.65***7.55 ***0.32 lessNet investment 0.705 1.116 0.4117.02***9.68 ***0.41 lessOther cash flows 0.083 0.229 0.1463.45***4.51 ***0.15 lessChange in cash 0.270 0.053 0.2176.53***8.46 ***0.22 more Invest Grow balance sheet Save cash Repurchase equity Issue debt Subsets 7 (no credit line) & 8 (credit line) t (diff) z (7,8) Action 7 8 7 relative to 8 7 8 diff: 78 PAGE 61 61 CHAPTER 2 CREDIT LINES AND THE SUBSTITUTABILITY OF CASH AND DEBT Introduction An important open question in corporate finance research is whether transactions costs are economically important for corporate financial and investment policies. Direct transactions costs, such as flotation costs, and indirect transactions costs, such as those stemming from information asymmetries, play a prominent role in both theoretical and empirical corporate finance research. But the extent of the economic importance of transactions costs for financial and investment policy is unknown. I provide evi dence of the importance of these costs for the firms cash policies by analyzing the effects of the existence of, and liquidity available from, a corporate credit line on excess returns to equityholders. Credit lines are relevant for this research because the ex post fixed costs of accessing liquidity via a credit line are minimal. Consistent with the economic importance of transactions costs for firm financial policy, I find that credit lines are an important tool for the management of firm liquidity, that the firm contemplates future credit line access when setting cash policies, and that the resulting firm actions with regard to the credit line have a large impact on shareholder value. By construction, the credit line provides convenient access to liquidi ty for wide ranging purposes including the financing of inventory, equipment acquisitions, and general corporate purposes, among others. Credit lines are most often senior debt and sometimes secured, thus repayment of a line balance is also relatively imm une to distortions due to conflicts of interest between shareholders and holders of risky debt. Terms of credit line use are negotiated ex ante with (most often) a private lender so that ex post transactions costs stemming from information asymmetries are also limited. A theoretical committed and state independent credit line PAGE 62 62 (Holmstrom and Tirole, 1998) links the left and right hand side of the borrowers balance sheet without friction, and allows firms to get closer to first best cash policy. Opler Pinko witz, Stulz, and Williamson (1999) find empirical support for Keynes (1936) precautionary motive for holding cash. Under this explanation, uncertainty over access to funds next period can lead to excess cash holdings today benefits of less uncertainty over availability of future funds outweigh the associated agency costs and tax disadvantages of holding excess cash.1 Direct transactions costs, such as flotation costs, with an important fixed component can also induce the firm to hold excess cash balanc es Large fixed relative to incremental costs of access lead s to economies of scale for capital acquisition and firms might raise excess funds today to fund future growth options that are currently out of the money. The firms optimal liquidity is thus de pendent upon characteristics of investment options, taxes agency costs, and direct and indirect transactions costs. Indeed, Gamba and Triantis (2007) model the firms liquidity decision as such, and find that multiple combinations of cash and debt resulti ng in the same net debt position yield different firm values.2 An implication is that cash and debt are not always perfect substitutes: sometimes cash is preferred to debt capacity, or in this case, line availability. In the words of Opler et al. (1999) fo r this case, cash is not equal to negative debt. A firm with a committed credit line might view line availability similarly to cash on the balance sheet such that cash policy will be dependent upon line characteristics. But even 1 See Bates, Kahle, and Stulz (2008) for a recent review of the literatures theories for holding cash: 1) the transactions motive (Baumol (1952), Miller and Orr (1966), and Mulligan (1997)), 2) the precautionary motive (Opler et al (1999), Almeida, Campello, and Weisbach (2005), and Acharya, Almedia, and Campello (2008)), 3) the tax motive (Foley, Hartzell, Titman, and Twite (2007)), and 4) the agency motive (Jensen (1986), Dittmar, Mahrt Smith, and Servaes (2003), Dittmar and Mahrt Smith (2006), Pinkowitz, Stulz, and Williamson (2006), and Harford, Mansi, and Maxwell (2006), among others). 2 An exception is that agency costs are not in the Gamba and Triantis (2007) model. PAGE 63 63 committed credit lines are contingent contracts. Attached covenants and a material adverse condition clause, combined with the lenders monitoring rights, can lead to limited access to the credit line. Indeed, Sufi (2007) and Lins, Servaes, and Tufano (2007) show that access to a cr edit line might not be available just when it is most needed by the firm. Sufi (2007) reports that 72% of credit lines to public corporations in Loan Pricing Corporations Dealscan database have information listed on financial covenants. Of these credit li nes, only 25% do not have a financial covenant based on cash flow coverage. Sufis analysis suggests that low cash flows best explain the incidence of a covenant violation, and of his random sample firms with a credit line, 35% violate a financial covenant at some point during the panel. A firm using a credit line to manage cash policy will consider the probability of access next period. The more tenuous is next periods access, the more likely line availability is not considered a good substitute for cash balances. A sufficient probability of limited access to availability combined with sufficient cash needs will lead to the firm preferring cash to line availability, and could result in draws on the credit line to fund precautionary cash.3, 4 3 There are numerous press reports of fi rms accessing credit lines in the wake of the economic crisis beginning in mid to late 2008. General Motors in September 2008; GM treasurer Walter Borst: Accessing the (credit line) funds available to us is a prudent liquidity measureDrawing on the revol ver now improves our liquidity position at a time when the capital markets have become more challenging. (CFO.com (09/22/2008)) Ford Motor Company in January 2009; Ford CFO Lewis Booth: We are not drawing on the revolver to fund operations, and not to shore up cash levels, but because we think it is the prudent thing to do. (Financial Times (01/29/2009)) Computer Sciences borrowed its entire $1.5 billion credit facility in late 2008. (Investors Business Daily (12/16/2008)) Goodyear Tire borrowed $600 mi llion in late 2008 to fund working capital and to enhance its liquidity. (Investors Business Daily (12/16/2008)) Monster Worldwide in October 2008; CFO Timothy Yates: We have always viewed our revolving credit as an insurance policy, and given events i n the market, we felt that it was appropriate to access that insurance. (Investors Business Daily (12/16/2008)) Station Casinos in December 2008; CreditSights analyst Christopher Snow: One outcome is that by drawing down on the revolver they still have the potential to negotiate a waiver and pay it back, so it could be part of that negotiation. (Rueters News (12/22/2008)) Tribune Co. in October 2008; company reports: Tribune is borrowing under the revolving credit facility to increase its cash position in todays uncertain credit market. (Reuters News (10/17/2008)) PAGE 64 64 This intuition is related to the work of Acharya et al. (2007), which shows that financially constrained firms are more likely to save cash out of cashflows. Weak correlation between investment opportunities and cash flows, and strong correlation between cash flows and debt capacity, can provide the firm with incentives to resist using excess cash to repay debt in high cash flow states. Instead, excess cash is transferred from high cash flow states to low cash flow states during which investment opportunities are more at tractive. In the present context for this case, cash is not negative debt, and cash balances are preferred to line availability. Following the above discussion, a credit line can affect firm value in at least a couple of ways. First, a credit line can be a useful tool to optimize cash policy. Firms with access to a credit line can limit the carry of excess cash by maintaining line availability. Such policy will limit taxes and agency costs of excess cash balances, benefitting firm value. Second, a credit li ne can lower the fixed cost of accessing funds for general corporate purposes including acquisition of inventory, equipment and other property, and acquisitions. The result is that firms are less likely to forego in the money growth options and less likely to early exercise other growth options currently out of the money. As such, credit line characteristics should be priced by shareholders.5 Faulkender and Wang (2006) show that the market value of an additional dollar of cash inside the firm is more valua ble to shareholders of firms that are financially constrained. They reason that constrained firms are more likely to have attractive projects to fund with additional Fairpoint Communications in September 2008; CFP Al Giammarino: The revolver draw was more of a precautionary measure. (Credit Investment News ( 09/19/2008) ) 4 An increase in credit spreads can also increase incentives for firms to access their credit lines. From an analyst report at Unicredit: Funding has gotten increasingly difficult for banks and short term refinancing rates are very expensivethe refinancing spread is currently higher than the credit spread of their committed credit facilities and liquidity is scarce. Therefore, banks have an incentive to withdraw their credit facilities, especially at companies with lower credit quality. (Dow Jones International News ( 10/14/2008) ) 5 See James (1987) for a study of announcement effects of bank loan agreements PAGE 65 65 cash, but for unconstrained firms, the marginal dollar is more likely to be distributed to shareholders or instead will support the value of debt. Denis and Sibilkov (2007 ) replicate the Faulkender and Wang (2006) result and find that the source of greater value of cash changes for financially constrained firms is due to the relative profitability of marginal investments for financially constrained and unconstrained firms. They do not find evidence consistent with an alternative story of the market compensating financiallyconstrained firms for passing up unprofitable projects (Mikkelson and Par tch (2003)). I analyze the effect of credit line characteristics on the value to shareholders of a marginal dollar of cash inside the firm. Using a unique dataset on the existence of, and availability on, a corporate credit line, I extend the Faulkender an d Wang methodology to analyze the effect of a credit line on the sensitivity of shareholder value to changes in cash holdings If cash and line availability are good substitutes then the cash inside the firm should be valued less by shareholders when a line is available. However, if cash is not negative debt, then the lines existence or availability should not reduce the value of cash inside the firm, ceteris paribus In cases where the firm is financially constrained and/or the lines future access is les s certain, the precautionary motive for holding cash might induce the firm to draw funds off of the line regardless of whether the funds are deployed immediately. As Gamba and Triantis show, these firms can potentially increase value by drawing funds off t he line and increasing cash balances; the firms net debt position is unchanged, but both cash and credit line balances are increased. I find results consistent with the hypothes is above. I start by replicating the Faulkender and Wang (2006) result that th e value of a marginal change in cash is greater for constrained firms. I then show that this effect is weaker for unconstrained firms with a credit line using a dummy variable indicating the existence of a credit line at fiscal year end. I then analyze a r andom subset PAGE 66 66 of the sample firms for which I have collected detail on the credit line characteristics to show that the effect holds after controlling for line availability. Next, I show that the value of a marginal change in line availability is more like the value of a marginal change in cash for firms for which cash and line availability are more likely to be good substitutes. In my last set of results, I estimate the increase in shareholder value associated with a draw of $1 off the line of credit st ored as cash. The increase in shareholder value is economically and statistically significant for firms deemed to be financially constrained. This research contributes to the corporate finance literature in several ways. First, it provides novel empirical support for the economic importance of transactions costs for firm cash policies, particularly via its analysis of the substitution of cash and debt in the spirit of Gamba and Triantis (2007). Second, this research complements recent work on financial cont racting (such as Roberts and Sufi (2008)) by providing empirical support for the importance of the contingent nature of private debt contracts and the resulting effects on firm financial policies. Third, this research adds to the growing body of literature on the importance of firm financial flexibility for firm value.6 Finally, this paper adds to the recent research on corporate credit lines, which has gained momentum of late following the work of Sufi (2007).7 Empirical M ethodology I analyze shareholder v alue effects of credit line characteristics on the marginal dollar of cash held inside the firm. If firms are unable to exercise in the money growth options due to transactions costs of accessing funds, then a dollar of cash inside the firm can increase va lue to shareholders. This precautionary motive for holding cash has offsetting costs due to taxes and 6 In addition to Gamba and Triantis (2007), see Gilson and Warner (1998), Jagannathan, Stephens, and Weisbach (2000), Billett and Garfinkel (2004), DeAngelo and DeAngelo (2007), and DeAngelo, DeAngelo, and Whited (2008). 7 In addition to Sufi (2007) see recent work by Lins, et al (2007), Riddiough and Wu (2009), and Yun (2007). PAGE 67 67 agency costs, but for sufficiently attractive marginal investments, the benefits can outweigh these costs. Credit line access can, in some cases, weaken f irm incentives to hold precautionary cash. In the Lins, et al. (2007) survey of international credit line use, the authors find that 41% of their sample firms respond that they consider their credit line as a substitute liquidity instrument, and among th ese firms, those with large credit lines have less precautionary, or strategic, cash holdings. Further, survey respondents claim that 40% of cash balances are held for strategic purposes. Lins et al.s regression analysis on survey results indicates th at among firms with credit lines, precautionary cash motives are associated with larger credit lines. Faulkender and Wang (2006) analyze the increase in shareholder value associated with one additional dollar held inside the firm. For financially constrain ed firms, they find that the marginal dollar of cash inside the firm indeed raises firm value by more than one dollar to shareholders. They describe their study as much like a longterm event study focused on liquidity policy. The firms excess return is c alculated and regressed on changes in cash and other balance sheet and income statement changes to control for other changes in financing and investment policies affecting shareholder value and correlated with changes in cash. = 0 + 1 1 + 2 1 + 3 1 + 4 1 + 5 1 + 6 1 + 7 1 1 + 8 + 9 1 + 10 1 1 1 + 11 1 + (2 1) PAGE 68 68 where the s indicate a change in the regressor over t 1 to t To the exten t that the regressor changes are unexpected, the regression will identify the associated excess returns.8 The firms excess return is defined relative to its benchmark return, which is based on the Fama and French (1993) size and book to market portfolio r eturns. Since essentially all variables are scaled by the lagged market value of equity, coefficient estimates can be interpreted as the additional value to shareholders resulting from a $1 increase in the regressor.9 The coefficient estimate 1 identifies the increase in value to shareholders associated with a one dollar change in cash inside the firm ( Ci,t), if leverage and lagged cash are both zero. Faulkender and Wang (2006) include other regressors to control for other financing and in vestment changes that affect value and are potentially correlated with changes in cash. Regressors controlling for changes in financial policy are levels of leverage ( Li,t) and net financing ( NFi,t), and changes in interest expense ( i,t) and dividends ( Di,t). Regressors controlling for changes in investment policy are changes in net assets (total assets minus cash and marketable securities, i,t) and R&D expense ( i,t). Changes in firm profitability are controlled for with changes in the firms earni ngs ( i,t). Faulkender and Wang also show that the effect of an additional dollar of cash inside the firm depends on the existing level of cash and leverage, and the interaction terms in Equation 2 1 control for these effects. An increase of $1 at already high cash levels is less important than is the same change at low levels of lagged cash, 8 First differencing assumes lagged values are the expected values for year t realizations. Faulkender an d Wang (2006) check this assumption with three alternative measures of expected cash holdings, and find the results to be robust. 9 Pinkowitz and Williamson (2004) estimate the value of cash holdings related to the firms investment opportunities using the method of Fama and French (1998), which analyzes the level of the firms market to book ratio. Faulkender and Wang (2006) that the method used here is an improvement over the Fama and French (1998) and the Pinkowitz and Williamson (2004) method because 1) regressing market excess returns on firm characteristics contemplates time variation in the markets compensation for risk factors, and 2) part of the variation in firms market to book ratios will be due to variation across firms in accounting methods for book values relative to true replacement costs, and 3) excess returns are easier interpret. Denis and Sibilkov (2007) reproduce Faulkender and Wangs (2006) primary results by regressing Tobins Q on firm characteristics (see Table 4 of Denis and Sibilkov). PAGE 69 69 ceteris paribus and an increase in cash at high leverage levels will be more likely to support risky debt values than to increase shareholder value.10 I modify Equat ion 2 1 with credit line characteristics to analyze the effect of lower transactions costs of access to liquidity on shareholder value. Data and S ummary S tatistics CRSP/Compustat and C redit Line D ata The dataset is comprised of firms in the CRSP/Compustat merged database over 1996 to 2006. There are several screens on the data that follow Sufi (2007). Firms must have at least four years of continuous data available, four years of positive assets and book leverage ratios between zero and one (inclusive). Nonmissing data for at least four consecutive years is required for total liabilities (data181), sales (data12), operating income before depreciation expense (data13), common shares outstanding (data25), fiscal year end stock price (data25), preferred stock (data10), deferred tax assets (data35), and convertible debt (data79). A text search program is used to read 10 K forms downloaded from the SECs Edgar server.11 A dummy variable, HASLINEt, indicating the existence of a committed credit line available at fi scal year end is created from the text search results.12 As described in Appendix C I have collected data on the existence of a credit line at fiscal year end for a panel of firms over 19962006 following the method used in Sufi (2007). For a random subset of 362 firms I have detail on the line availability, use, and whether the firm violated a covenant during the fiscal year. Using the larger sample I modify Equation 2 1 to estimate coefficients for firms with and 10 This papers results are robust to excluding the interaction terms. 11 Text search details are provided in Appendix C. 12 A committed credit line is available to the firm without need for further credit approval upon the request for a draw. The actua l funding of the requested draw is subject to the firms abidance pursuant to the credit lines loan agreement, which can include financial covenants, a material adverse condition clause, and security agreements. PAGE 70 70 without a credit line at fiscal year end, which enables me to test hypotheses about the credit lines effect on the value of an additional dollar of cash held inside the firm. With the smaller random sample, I use the amount drawn on the credit line and the credit line total to test hypotheses about the substitutability of cash and debt. I follow Sufi (2007) and read the 10K forms for credit line detail on a random subset of the dummy variable sample. Sufi (2007) collects detail on credit lines for the whole panel of 300 random firms over 19962003; his 300 firms represents a 6.5% sampling rate of firms surviving his screens. I save Sufis random sample firms and then re sample the remaining firms surviving my application of Sufis screens over a slightly longer time period.13 For these random firms I collect the detail of the credit line : the total commitment available, the total drawn, and whether the firm violated a covenant during the fiscal year. Details of the random sample construction are in Appendix C All ratio variables are W insorized at the 1 and 99 percentiles and I eliminate observations if net assets, the market value of eq uity, or dividends are negative. My full sample includes the dummy variable for credit line existence on 4,640 firms and 26,215 observations. The random sample includes detailed data on credit line characteristics, and is available for 323 firms and 1,836 observations. I follow Faulkender and Wang (2006) and Fama and French (1993) and note the Fama French portfolio to which the firm belongs each month. The Fama F rench size and book to market portfolios are formed at the end of June in year t The size sort uses the firms market value of equity as of the end of June in year t The book to market sort uses the ratio of the 13 Amir Sufi has generously provided his dat aset to the public via his faculty website at the University of Chicago, and this dissertation has benefitted as a result. I replicate Sufis (2007) data to ensure my extension (years 20042006) is collected in a manner consistent with the 19962003 data. PAGE 71 71 firms book value of equity as of the fisc al year end in calendar year t 1 and the market value of equity at the end of December in calendar year t 1 .14 During one fiscal year, a firm can belong to two different portfolios if their fiscal year does not close at the end of June. Monthly returns of t he appropriate portfolios are used to calculate the annual benchmark return.15 Summary S tatistics Table 2 1 (Table 2 2) reports summary statistics for the full ( random ) samples. A credit line exists at fiscal year end for 82.3% of Full Sample observations. The mean firm begins the fiscal year with cash balances equal to 14.9% of the market value of equity, which translates to approximately 18% of assets. Net financing activity ( NFt), which is equal to cash flow statement debt issues minus debt retirements plus the net of equity issues and repurchases, is 2.7% of the lagged market equity at the mean. The mean excess return ( ERt) is slightly negative at 0.009, but is right skewed as the median is 0.106.16 Summary statistics on the random sample are nearly id entical to the full sample for the common variables. Table 2 2 results show that the firms credit line is indeed an important tool for firm financial policy. The statistics on credit line detail ( e.g., t, t, LAt, and LTt) are for the whole sample su ch that to get the mean values conditional on having a credit line, the reported credit line values must be divided by the HASLINEt mean of 0.806. Thus, the mean 14 To be clear, a firm with FYE in January 2000 (which is FYE 1999 for the firm) might belong to two different Fama French portfolios, and therefore needs data for the market value of equity in at monthend June 1998 and 1999 for the size sorts, data for th e market value of equity at monthend December 1997 and 1998 for the bookto market ratios, and data for the book value of equity from the fiscal years ending in months January 1997 and 1998 (which are FYEs 1996 and 1997 for the firm). The firms first month of FYE 1999 will be February 1999, a month for which benchmark returns will be based off of Fama French portfolios formed at the end of June 1998. The Fama French portfolios formed at the end of June 1998 will use the firms market values at monthend J une 1998 and December 1997, and book value of equity at the end of January 1997. 15 I use Fama French portfolio returns including dividends as the stocks return is the firms total return over the fiscal year. 16 These summary statistics match closely to th ose in Faulkender and Wang (2006), despite the difference in sample periods (Faulkender and Wang covers 19722001). Appendix E details a replication of Faulkender and Wang (2006). PAGE 72 72 credit line has a total committed amount equal to 37.8% (0.305/0.806 = 0.378) of the lagged ma rket value of equity. The average firm has 16.0% of lagged equity in cash balances. Conditional on having a credit line, there is an additional 22.6% (0.182/0.806 = 0.226) of lagged equity available. These results suggest the average firm with access to a credit line would be forced to make significant changes to financial policy if such access were limited in the future Constrained and U nconstrained F irm I dentification C riteria Much of the recent corporate finance research on firm financial policy has ana lyzed the effects of firm financial constraints, and there are several methods put forth for identifying financially constrained and unconstrained firms.17 I follow the literature and identify firms as financially constrained and unconstrained according to payout ratio, size, and the existence of a bond rating.18 For a fourth definition of constraint, I mark the observation as constrained (unconstrained) if at least two of the payout, size, or bond rating conditions are met as such. For payout ratio, the obse rvation is considered constrained (unconstrained) if the payout ratio (common dividends plus repurchases divided by earnings) is in the lower (upper) 30th percentile by year. For size, an observation is considered constrained (unconstrained) if total asset s is in the lower (upper) 30th percentile by year.19 Finally, i f a firm has a bond rating at a year end in which it reports positive debt, then the observation is deemed unconstrained, but constrained if there is positive debt and no bond rating. 17 See Whited (1992), Kashyap, Lamont, and Stein (1994), Gilchrist and Himmel berg (1995), Pinkowitz and Williamson (2004), Almeida et al (2004), Denis and Sibilkov (2007), Sufi (2007), and Acharya et al (2007). 18 Firms with high payout ratios are more likely to have sufficient internal funds for investment and financing needs, th us, cash holdings should be less important for shareholder value compared to firms with low payout ratios. Size is likely correlated with access to capital markets, and cash holdings for firms with relatively easier access to funds should be less important than for firms without such access. The existence of a bond rating indicates the public capital markets have experience with the firm, and these firms should have a relatively easier time accessing funds than otherwise, which should decrease the value of marginal cash holdings. 19 These definitions of constraint follow the literature, but most notably Faulkender and Wang (2006). PAGE 73 73 Cash Balan ces b y Line Existence and F inancial C onstraint Table 2 3 reports cash holdings of firms by credit line status and financial constraint. T he mean firm in the full sample without a credit line keeps 28.0% of the lagged market value of equity in cash balances This is in stark contrast to the 12.8% for the mean firm with a credit line in place. Delineating firms by financial constraint status shows that constrained firms without a credit line (C,NL) indeed do hold the greatest cash balances; firms considered c onstrained according to at least two of the three definitions hold cash balances equal to 29.5% of the lagged market value of equity. As a percentage of lagged market equity, this is nearly three times the amount of cash held by firms considered unconstrai ned with a credit line available (U,HL). Sufi (2007) provides evidence that data on the existence of the credit line improves identification of firms considered financially constrained or unconstrained. His result is seen here too: firms identified as unco nstrained, but without a credit line, have cash balances more like constrained firms than like unconstrained firms with a credit line. These unconstrained firms without a credit line (U,NL) on average hold more cash than do firms considered constrained w ith a credit line available (C,HL) and these differences are statistically significant at the 1% level no matter the definition of financial constraint. See row 1 of Table 2 3 for cash mean cash holdings for firms without a credit line by financial constraint status and definition; per constraint definition, left (right) columns are conditional means for firms constrained (unconstrained). Thus, the unconstrained firms without a credit line (U,NL) hold cash balances from 19.1% to 23.3% of lagged market equity, depending on the definition of constraint. Constrained firms with a credit line (C,HL) hold 12.8% to 15.5% of lagged equity in cash balances (see row 2 for cash balances for firms with a credit line available). The bottom panel of Table 2 3 reports te sts for statistical significance of the conditional mean differences. These PAGE 74 74 results suggest that we can learn more about Faulkender and Wangs (2006) results on the value of cash inside the firm by studying the firms access to, and use of, a credit line. Table 2 4 reports the detailed credit line data from my hand collected Random Sample, and take s a closer look at the net cash and credit line balance positions of firms according to financial constraint status Cash holdings and credit line availability ar e reported by constraint status. The firms net line debt ( NLDt, cash holdings minus credit line balance) is reported in the bottom row of the top panel of Table 2 4 Conditional on having a credit line, the average firms drawn line balance exceeds its balance of cash by 2.1% of the lagged market equity. In their study for why firm cash holdings have increased dramatically over time, Bates, et al. (2008) note that the average firm in 2004 had enough cash balances to completely repay all outstanding debt. But the random sample results here indicate that the Bates, et al. result is likely driven by large cash balances held by firms without a credit line in place. The remaining columns of Table 2 4 show that with respect to only cash and credit line balance, constrained firms have a greater net debt position compared to that of unconstrained firms. Among credit line firms, only firms considered unconstrained according to bond rating have a slight net cash position of 0.8% of lagged market equity at the mean. Fi rms with a credit line and considered constrained according to at least two of the three definitions have a greater mean credit line balance than mean cash balance to the tune of 5.8% of lagged market equity. This evidence is consistent with the notion tha t financially constrained firms have greater reliance on their credit lines for their management of liquidity and cash policies. Incidence of F inancial C ovenant V iolations by F inancial C onstraint S tatus Credit lines are contingent contracts subject to fina ncial covenants and/or the lack of a material adverse event for the firm. This contingent nature affects the borrowers willingness or ability to rely upon the line as a cash substitute. Financial covenants attached to the credit line PAGE 75 75 provide an important mechanism for exercising creditor rights, and the credit lines revolving nature and usually general purpose for use (often the credit lines purpose is stated as for general corporate purposes) makes these contracts especially fit for the private lender s monitoring expertise. The low ex post transactions costs associated with the credit line are made possible because of the frequent information flow from the firm to the lender (Aghion and Bolton (1992), Dewatripont and Tirole (1994), and Roberts and Suf i (2008)), but upon a material adverse event or a covenant violation, the lender can impose a wide range of actions including the acceleration of the lines balance. Oftentimes a covenant violation is accompanied by a reduction in the line availability and/or a covenant waiver fee.20, 21 In his Figure 2, Sufi (2007) shows the reaction of the line characteristics to a covenant violation. The line total, availability, and amount drawn are all noticeably decreased following such a violation. Roberts and Sufi (2008) analyze firm quarterly and annual filings with the SEC over 19952005 and find that one third of covenant violator firms disclose that their lenders responded by decreasing the loan 20 As anecdotal evidence, in the first quarter of 2001, AGCO Corporation paid a $2.5 million waiver fee in connection with a covena nt violation and disclosed the following in the 10K: We subsequently received sufficient waivers from the holders of the notes for any violation of the covenant that might have resulted from the dividend payments. In connection with the receipt of waivers, we paid a waiver fee of approximately $2.5 million, which will be expensed in the first quarter of 2001. Currently, we are prohibited from paying dividends until such time as the interest coverage ratio in the indenture is met. We currently are in the pro cess of modifying our capital structure in order to most efficiently meet our funding needs and replace our existing revolving credit facility, which expires in January 2002. In addition, although we currently are in compliance with the financial covenants under all of our indebtedness, the financial covenants in our existing revolving credit facility become more stringent at the end of the second quarter of 2001. As a result, we currently do not anticipate being able to fulfill two of the financial covenan ts contained in the facility, a limitation on the ratio of funded debt to EBITDA and a minimum fixed charge coverage ratio. 21 On December 27, 2002, Avado Brands, Inc. incurred an $8.5 million waiver fee in connection to a failure to meet an EBITDA coverage ratio target. The credit facilitys interest rate was increased and $6.5 million of the waiver fee was designated as refundable provided all indebtedness was repaid by April 27, 2003. Yet, $7.8 million of additional borrowing was available at fiscal year end 12/28/02. PAGE 76 76 facility amount, increasing the interest rate, or requiring more collat eral, and that the firms net debt issuance scaled by lagged assets falls by 418 basis points within six months of the violation. In Table 2 5 I report the incidence of covenant violations during the fiscal year for firms in my Random Sample that are consi dered financially constrained and unconstrained. The unconditional incidence rate is 0.089. Regardless of how constrained firms are defined, they have a greater incidence rate of covenant violations compared to unconstrained firms and all differences in year t are statistically significant at the 1% level. Gamba and Triantis (2007) model the firms management of liquidity in a dynamic framework such that firms make use of the debt for liquid assets today depending on the probability of access to funds in future periods. Thus, I also report the covenant violation rates for years t+1 and t+2 The covenant violation incidence rates are all statistically different across financially constrained and unconstrained firms with the exception of one case (size defi nition and CVt+2). These results suggest that the financial constraint definitions are not only capturing the ability to access needed funds, but also effects related to the likelihood of continued access to existing credit lines going forward, which might be important for the firms reliance on the credit line for liquidity management. In the next section I analyze the effect of a credit line on the value of a marginal dollar held inside the firm. Empirical Results Financially C onstrained versus F inanciall y U nconstrained To begin, I replicate Faulkender and Wangs (2006) primary result for my sample. Via OLS, I estimate a version of (1) with separate estimates for constrained and unconstrained firms. = 0 + 1 + + + (2 2) PAGE 77 77 where FCi,t ( FUi,t) are dummy variables indicating that the firm is financially constrained (unconstrained) in year t and Xi,t is a vector of the regressors included in Equation 2 1. Superscripts C and U on the parameter vectors indicate estimate vectors for financially constrained and unconstrained firms, respectively. Since t is interacted with the lagged level of cash and levera ge (see Equation 2 1), the marginal effects for a one dollar change in cash vary with values of lagged cash and leverage At the mean level of these covariates, we have: Marginal effect of = + 1 1 + ( 2 3) I find that a one dollar increase in cash is valued greater than one dollar for both constrained and unconstrained firms (ME( t)  C is the estimate d value of an additional dollar inside a constrained firm).22 The last row of Table 2 6 reports the difference between the marginal effects of a change in cash for constrained and unconstrained firms ( labeled diff(ME( t)), which equals ME( t)C ME( t)U). A one dollar increase in cash is valued $0.370 to $0.497 greater by shareholders o f constrained firms relative to unconstrained firms depending on the definition of constraint, and the differences are statistically significant at the 1% level. These results are consistent with Faulkender and Wang (2006), Denis and Sibilkov (2007), and w ith the notion that financially constrained firms are unable to execute profitable growth options at the margin. Existence of a C redit Line and F inancial C onstraint Equation 2 2 can be modified to measure how a credit line affects the marginal value to equ ityholders of an additional dollar of cash inside the firm. 22 My estimated marginal effects of a change in cash are greater in magnitude that those found in Faulkender and Wang (2006), which I attribute to the different sample periods. Appendix E provides evidence that the marginal effect of a change in cash has increased over time. PAGE 78 78 = 0 + 1 + 2 + 3 (2 4) + + + + + where is a dummy variable indicating that the firm is financially constrained without (with) a credit line at year end t and Xi, t is a vector of the regressors in ( 2 1) Superscripts CNL, CHL, UNL, and UHL on the parameter vectors are for combinations of financial constraint and HASLINEi,t statuses. To save space, only the marginal effects of an additional dollar of cash are report ed in Table 2 7 The value of an additional dollar of cash inside the firm is greater for firms financially constrained and unconstrained without a credit line (first three rows of MEs: CNL, CHL, and UNL) compared to that for firms unconstrained with a cr edit line (fourth row: UHL). A test for equality of all four marginal effects is rejected no matter the definition of constraint; however, the test for equality of the first three effects cannot be rejected (see p value tests). Thus, the lower marginal eff ect of a change in cash found for unconstrained firms in Table 2 6 is driven by unconstrained firms with a credit line available at fiscal year end (UHL). This result is consistent with Sufis (2007) finding that the existence of a credit line adds informa tion about the firms financial constraint beyond what is captured by the extant literatures sorts. It is also consistent with results reported in Table 2 4 that the unconstrained firms without credit lines (UNL) seem to have cash policies similar to thos e of constrained firms. Shareholders of financially constrained firms with a credit line available (CHL) value an additional dollar of cash inside the firm $0.431 to $0.528 more than do shareholders of financially unconstrained firms also with a credit li ne (UHL). This result (see diff(ME( t))HL) PAGE 79 79 is consistent with the notion that cash and line availability are not good substitutes for constrained firms, and is consistent with the result reported in Table 2 5 that constrained firms have significantly gre ater incidence of covenant violations. Line A vailability and F inancial C onstraint The results presented thus far suggest that with regard to cash policies and the value of cash inside the firm, respectively, financially constrained firms with a credit lin e (CHL) are more like firms without credit lines than they are like unconstrained firm with a credit line (UHL). One explanation for results on the value of cash held inside the firm is that the constrained firms credit line has no remaining availability. Another explanation is that these shareholders discount the value of the credit lines availability due to concern over future access. In this case, cash is not a good substitute for line availability, and the shareholders prefer cash on the balance sheet Table 2 3 restates Equation 2 2 with the addition of two variables in the Xi,t vector representing credit line characteristics from the random sample to estimate the marginal value of a change in cash, controlling for changes in line availability. t is the change in line availability scaled by the lagged market value of equity. But line availability can change due to repayment or draw down of debt on the line, or due to a change in the total commitment. t controls for changes in the amount dra wn so that the effect identified by t is due to changes in the total commitment. The constrained unconstrained result from Table 2 6 holds after controlling for changes in line availability: shareholders of financially constrained firms place a greater value on changes in cash regardless of changes in line characteristics. The constrained unconstrained difference in the marginal effect of a change in cash is estimated to be at least $0.807 per dollar change. Changes in line availability are also associat ed with changes in shareholder value: $0.233 to $0.415 ($0.448 to $0.500) per dollar change for financiallyconstrained (unconstrained) firms. PAGE 80 80 The difference between the marginal effects of a change in cash and a change in line availability is statisticall y significant for financially constrained firms, but not so for financiallyunconstrained firms. This evidence is consistent with the notion that cash and line availability are better substitutes for unconstrained firms. For shareholders of financiallycon strained firms, a change in cash is estimated to be worth $1.460 to $1.805 more than is a change in line availability. Such a result is expected if cash and line availability are not good substitutes. However, cash and line availability seem to be better s ubstitutes for financially unconstrained firms. The difference between the marginal effects of a change in cash and line availability for financially unconstrained firms ranges $0.244 to +$0.683, but it is never significantly different from zero. I explic itly analyze the substitutability of cash and debt in the next subsection. The S ubstitutability of C ash and Line of C redit D ebt The analysis above controlled for changes in financing policy that might be related to changes in shareholder value, but also correlated with changes in cash holdings, via the regressor NFt, which is a combination of the debt and equity issues data from the cash flow statement. This model construction allowed us to examine the value of an additional dollar of cash inside the firm n o matter its source (debt or equity issues or changes in other liabilities). In this subsection, I want to explicitly examine the effect of drawing an additional dollar of cash off of the credit line to hold as cash. This test is in the spirit of Gamba and Triantis (2007), who show that when transactions costs are important for financial policies, multiple combinations of cash and debt resulting in the same net debt position can yield different firm values, resulting in an optimal liquidity for shareholder value. Since credit lines have minimal transactions costs ex post we have a unique opportunity to test this important theoretical result. To execute the tests, I must now control for the source of the change in cash. I drop the net financing regressor NFt from the Xi,t vector in ( 2 4) and replace it with several regressors to PAGE 81 81 control for changes in the right hand side of the firms balance sheet. In place of NFt, I use the following: t: change in the amount drawn on the line of credit from year t 1 to t t: change in other debt, debt excluding the credit line debt, from year t 1 to t t: change in other liabilities, liabilities not debt, from year t 1 to t and t: change in retained earnings from year t 1 to t Thus, the omitted balance sheet item is the change in equity less retained earnings EQt. As such, the interpretation of the marginal effect of a change in cash is now the value of a change in cash by one dollar due to the issuance of one dollar of equity. Since I have omitted only one balance sheet category, the coefficients on LDt and ODt represent the value of increasing (the respective) debt by one dollar and retiring one dollar of equity. I also include L Tt to control for increases in the credit line commitment, which are off bal ance sheet.23 I simply add the estimate on LDt to the marginal effect of a change in cash to analyze the value added to shareholders by drawing one dollar of cash off the credit line and then increasing the cash account by one dollar. The results are repor ted in Table 2 9 One dollar drawn off the credit line used to increase cash by one dollar increases the market value of equity by $1.124 to $1.597 if the firm is financiallyconstrained, a result statistically significant at the 1% level. This action does not increase value for shareholders of unconstrained firms, regardless of constraint definition. These results are consistent with the Gamba and Triantis (2007) theory in that cash is not a perfect substitute for debt when the firm faces significant trans actions costs to raising funds. The firm can increase value, at the margin, by increasing cash balances from the credit line today to ensure availability for use tomorrow. 23 Since we have t and Tt in the regression, the coefficient on Tt tells us the estimate of value added to shareholders due an increase in line availability (since the amount drawn on the credit line is being held constant) Tt is in the regression to ensure that the estimate on Dt identifies to value loss due to utilization of finite debt capacity as in DeAngelo, DeAngelo, and Whited (2008). PAGE 82 82 The precautionary motive for holding cash appears to be valued by shareholders for f inancially constrained firms. Conclusion I use data on corporate credit lines to analyze the effect of having this low transactions cost access to liquidity on the value of a marginal change of cash holdings to the firms shareholders. If transactions cost s are important for firm financial policies, and the firms cash policies have implications for firm value, then credit lines should have an impact on the management of liquidity and the resulting effects on shareholder value. I find that firms without a c redit line hold over two times the level of cash (as a percentage of the lagged market value of equity) of firms with access to a credit line. In addition, I find that among firms with access to a credit line, those firms considered financially constrained hold more cash than do those considered unconstrained. But, firms considered financiallyunconstrained and therefore supposedly with easier access to external capital but without access to a credit line, hold significantly more cash do firms with acce ss to a credit line. I extend the Faulkender and Wang (2006) model by incorporating credit line characteristics to analyze the associated effects of cash policies and credit line use on shareholder value. I find that Faulkender and Wangs result that share holders of financiallyunconstrained firms place less value on a marginal increase in cash holdings is driven by unconstrained firms with access to a credit line; shareholders of financially unconstrained firms without a credit line value marginal changes to cash holdings much like the shareholders of constrained firms value such changes. I analyze the role of the credit line on the substitutability of cash and debt in the spirit of Acharya et al. (2007) and Gamba and Triantis (2007). Consistent with the no tion that line availability is more like cash for financially unconstrained firms, for these firms, I find that the effects on shareholder value associated with changes in the amount available on the credit line PAGE 83 83 (not related to changes in the line draw bal ance) are more like the effects of changes in cash holdings. Consistent with the notion that cash and credit line availability are not good substitutes for financially constrained firms, in my last set of results, I find that shareholder value increases si gnificantly for these firms when both the cash and line draw balances are increased by the same amount. PAGE 84 84 Table 2 1. Summary statistics, f ull s ample 1996 2006. This table reports s ummary statistics for the full sample. ERt is the firms excess return at fi scal year end relative to a benchmark portfolio based on a Fama French 5x5 size and book to market sorts. t is the firms change in cash holdings from year t 1 to year t Ct 1 is lagged cash. Et is the change in earnings. NAt is the change in net assets (total assets net of cash). RDt is the change in research and development expense. It is the change in interest expense. Dt is the change in dividends. Lt is market leverage. NFt is net stock and debt issues. LAt is the change in credit line availability. LTt is the change in the credit line commitment. HASLINEt is a dummy variable equal to one if the f irm has a credit line in place at fiscal year end. All variables except ERt, Lt, and HASLINEt are scaled by the lagged market value of equity. ERt0.009 0.387 0.106 0.212 0.626 t0.005 0.027 0.001 0.033 0.116 Ct10.149 0.025 0.076 0.189 0.205 t0.015 0.030 0.006 0.041 0.193 t0.026 0.061 0.016 0.112 0.337 t0.001 0.000 0.000 0.001 0.021 t0.001 0.002 0.000 0.003 0.021 t0.000 0.000 0.000 0.000 0.006 Lt0.212 0.018 0.144 0.338 0.221 NFt0.027 0.034 0.001 0.053 0.191 HASLINEt0.823 1.000 1.000 1.000 0.382 SD Observations = 26,215 Mean 1st Quartile Median 3rd Quartile PAGE 85 85 Table 2 2. Summary statistics, random s ample 1996 2006. This table reports s ummary statistics for the random sample. ERt is the firms excess return at fiscal year end relative to a benchmark portfolio based on a Fama French 5x5 size and book to market sorts. t is the firms change in cash holdings from year t 1 to year t Ct 1 is lagged cash. Et is the change in earnings. NAt is the change in net assets (total assets net of cash). RDt is the change in research and development expense. It is the change in interest expense. Dt is the change in dividends. Lt is market leverage. NFt is net stock and debt issues. LAt is the change in credit line availability. LTt is the change in the credit line commitment. HASLINEt is a dummy variable equal to one if the f irm has a credit line in place at fiscal year end. LTt is the credit line commitment and LAt is the credit line availability. All variables except ERt, Lt, and HASLINEt are scaled by the lagged market value of equity. ERt0.000 0.393 0.107 0.233 0.634 t0.002 0.032 0.001 0.033 0.119 Ct10.160 0.028 0.080 0.205 0.219 t0.012 0.031 0.007 0.039 0.186 t0.018 0.070 0.017 0.107 0.323 t0.001 0.000 0.000 0.000 0.020 t0.001 0.002 0.000 0.003 0.021 t0.000 0.000 0.000 0.000 0.006 Lt0.218 0.012 0.146 0.357 0.229 NFt0.027 0.031 0.001 0.052 0.182 t0.005 0.015 0.000 0.027 0.166 t0.004 0.013 0.000 0.016 0.194 HASLINEt0.806 1.000 1.000 1.000 0.396 LTt0.305 0.032 0.150 0.370 0.523 LAt0.182 0.015 0.096 0.231 0.297 SD Observations = 1,836 Mean 1st Quartile Median 3rd Quartile PAGE 86 86 Table 2 3. Cash holdings and cred it line characteristics by financial constraint status f ull s ample. This table reports conditional mean cash holdings for the Full Sample. HL (NL) indicates HASLINEt = 1 ( HASLINEt = 0). C (U) indicates the firm is financially constrained (unconstr ained) in year t Financial constraint statuses are determined according to payout ratio, size, and bond rating. Payout ratio is dividends plus repurchases divided by earnings, and a firm is considered constrained (unconstrained) according to payout ratio if the payout ratio is in the lower (upper) 30th percentile by year. Size is total assets, and a firm is considered constrained (unconstrained) according to size if the firm is in the lower (upper) 30th percentile by year. A firm is considered constrained (unconstrained) according to Bondrating if the firm has positive debt and no (has a) bond rating at the fiscal year end. According to 2+, a firm is considered constrained (unconstrained) if at least two of the payout, size, and bondrating definitions indicates as such for year t Ct is cash and marketable securities divided by the lagged market value of equity. C U C U C U C U Ct  No line (NL) 0.280 0.297 0.233 0.285 0.206 0.262 0.219 0.295 0.191 Ct  Has line (HL) 0.128 0.155 0.102 0.155 0.100 0.128 0.104 0.149 0.096 ttests Ct  NL,C Ct  NL,U 0.065 *** 0.079 *** 0.043 ** 0.104 *** Ct  HL,C Ct  HL,U 0.053 *** 0.056 *** 0.024 *** 0.053 *** Ct  NL,C Ct  HL,C 0.143 *** 0.130 *** 0.135 *** 0.146 *** Ct  NL,U Ct  HL,C 0.078 *** 0.051 *** 0.092 *** 0.042 *** Ct  NL,U Ct  HL,U 0.131 *** 0.107 *** 0.116 *** 0.094 *** Conditional Mean Payout Size Bond Rating 2+ Definitions Full Sample Payout Size Bond Rating 2+ Definitions PAGE 87 87 Table 2 4 Cash holdings and credit line characteristics by financial constraint status random s ample.This table reports conditional mean ca sh holdings for the Random Sample. HL (NL) indicates HASLINEt = 1 ( HASLINEt = 0). C (U) indicates the firm is financially constrained (unconstrained) in year t Financial constraint statuses are determined according to payout ratio, size, and bond rating. Payout ratio is dividends plus repurchases divided by earnings, and a firm is considered constrained (unconstrained) according to payout ratio if the payout ratio is in the lower (upper) 30th percentile by year. Size is total assets, and a firm is considered constrained (unconstrained) according to size if the firm is in the lower (upper) 30th percentile by year. A firm is considered constrained (unconstrained) according to Bondrating if the firm has positive debt and no (has a) bond rating at the f iscal year end. According to 2+, a firm is considered constrained (unconstrained) if at least two of the payout, size, and bondrating definitions indicates as such for year t Ct is cash and marketable securities divided by the lagged market value of equ ity. LAt is the credit line availability divided by the lagged market value of equity. LTt is the total credit line commitment divided by the lagged market value of equity. Net line debt ( NLDt) is the firms cash holdings minus the credit line drawn balanc e at fiscal year end t C U C U C U C U Ct  No line 0.311 0.325 0.282 0.373 0.172 0.273 0.231 0.346 0.194 Ct  Has line 0.128 0.158 0.096 0.147 0.097 0.132 0.101 0.154 0.094 LAt  Has line 0.226 0.233 0.200 0.219 0.250 0.213 0.284 0.225 0.261 LTt  Has line 0.378 0.408 0.302 0.414 0.370 0.407 0.385 0.436 0.363 Net line debt 0.021 0.017 0.001 0.048 0.016 0.061 0.008 0.058 0.000 ttests Ct  NL,C Ct  NL,U 0.043 0.200 *** 0.042 0.152 *** Ct  HL,C Ct  HL,U 0.063 *** 0.051 *** 0.031 *** 0.060 *** Ct  NL,C Ct  HL,C 0.167 *** 0.225 *** 0.141 *** 0.192 *** Ct  NL,U Ct  HL,C 0.124 *** 0.025 0.100 ** 0.040 Ct  NL,U Ct  HL,U 0.187 *** 0.076 *** 0.131 *** 0.100 *** LAt  C LAt  U 0.033 0.031 0.071 *** 0.036 LTt  C LTt  U 0.106 *** 0.044 0.022 0.073 ** NLDt  C NLDt  U 0.017 0.032 0.069 *** 0.058 ** 2+ definitions Conditional Mean Payout Size Bond rating 2+ definitions Random sample Payout Size Bond rating PAGE 88 88 Table 2 5. Covenant violations by financial constraint status This table reports i ncidence of covenant violations by financial constraint status at fiscal year end t CVt is a dummy variable equal to one if the firm violated a financial covenant during the fiscal year t C (U) indicates the firm is financially constrained (unconstrained). C U is for a t test on the statistical difference between CV for constrained and unconstrained firms. Payout ratio is dividends plus r epurchases divided by earnings, and a firm is considered constrained (unconstrained) according to payout ratio if the payout ratio is in the lower (upper) 30th percentile by year. Size is total assets, and a firm is considered constrained (unconstrained) a ccording to size if the firm is in the lower (upper) 30th percentile by year. A firm is considered constrained (unconstrained) according to Bondrating if the firm has positive debt and no (has a) bond rating at the fiscal year end. According to +, a fir m is considered constrained (unconstrained) if at least two of the payout, size, and bondrating definitions indicates as such for year t. ***, **, indicate statistical significance at the 1, 5, and 10% levels, respectively. C U CU C U CU C U CU C U CU CVt0.089 0.119 0.047 *** 0.129 0.060 *** 0.116 0.070 *** 0.129 0.060 *** CVt+10.100 0.118 0.061 *** 0.133 0.070 *** 0.128 0.077 *** 0.133 0.069 *** CVt+20.099 0.104 0.060 *** 0.105 0.079 0.124 0.071 *** 0.114 0.066 ** 2+ definitions Random sample Payout Size Bond rating PAGE 89 89 Table 2 6. Regression re sults for f ull s ample by financial constraint status This table reports Full Sample results of regressing the firms excess stock return, ERi,t, on firm characteristics interacted with dummy variables indicating whether the firm is financially constrained or uncontrained. t is the firms change in cash holdings from year t 1 to year t Ct 1 is lagged cash. Et is the change in earnings. NAt is the change in net assets (total assets net of cash). RDt is the change in research and development expense. It is the change in interest expense. Dt is the change in dividends. Lt is market leverage. NFt is net stock and debt issues. FCi,t, is a dummy variable equal to one if the firm is financially constrained at fiscal year end. All regressors except Lt and FCt are scaled by th e lagged market value of equity. ME( t)  C (ME( t)  U) is the marginal effect of a one dollar change in cash, calculated at the mean levels of lagged cash ( Ct 1) and leverage ( Lt). diff(ME( t)) is the difference between the marginal effects of a chang e in cash for constrained and unconstrained firms, equal to ME( t)  C ME( t)  U. Financial constraint status is determined according to payout ratio, size, and bondrating. Payout ratio is dividends plus repurchases divided by earnings, and a firm is considered constrained (unconstrained) according to payout ratio if the payout ratio is in the lower (upper) 30th percentile by year. Size is total assets, and a firm is considered constrained (unconstrained) according to size if the firm is in the lower ( upper) 30th percentile by year. A firm is considered constrained (unconstrained) according to Bondrating if the firm has positive debt and no (has a) bond rating at the fiscal year end. According to 2+, a firm is considered constrained (unconstrained) if at least two of the payout, size, and bondrating definitions indicates as such for year t. ***, **, indicate statistical significance at the 1, 5, and 10% levels, respectively. PAGE 90 90 Constrained t2.307 *** 2.298 *** 2.225 *** 2.309 *** t11.077 *** 1.198 *** 1.075 *** 1.125 *** t1.991 *** 2.136 *** 2.067 *** 2.092 *** t0.340 *** 0.475 *** 0.359 *** 0.383 *** NFt0.103 ** 0.004 0.196 *** 0.139 ** t0.541 *** 0.559 *** 0.557 *** 0.540 *** t0.677 ** 0.651 0.456 0.602 ** t1.928 *** 2.183 *** 1.645 *** 1.816 *** t0.432 6.388 *** 2.868 *** 3.174 ** Ct10.431 *** 0.441 *** 0.410 *** 0.407 *** Lt0.592 *** 0.612 *** 0.734 *** 0.693 *** Unconstrained t1.661 *** 1.927 *** 1.768 *** 1.771 *** t10.884 *** 0.815 ** 0.155 0.144 t1.192 *** 1.986 *** 1.861 *** 2.078 *** t0.282 *** 0.150 *** 0.165 *** 0.140 *** NFt0.302 *** 0.010 0.046 0.018 t0.681 *** 0.599 *** 0.564 *** 0.603 *** t0.130 0.448 0.001 0.223 t2.430 *** 2.249 *** 2.420 *** 2.476 *** t3.769 *** 0.509 0.116 0.911 Ct10.249 *** 0.403 *** 0.529 *** 0.505 *** Lt0.451 *** 0.557 *** 0.620 *** 0.534 *** FCt0.002 0.113 *** 0.059 *** 0.046 *** Intercept 0.009 0.083 *** 0.114 *** 0.073 *** Adj R20.203 0.205 0.219 0.214 Observations 22,165 14,612 22,257 18,360 Firms 4,545 3,099 4,215 4,061 ME( t)  C 1.701 *** 1.731 *** 1.621 *** 1.691 *** ME( t)  U 1.331 *** 1.331 *** 1.170 *** 1.194 *** diff (ME( t)) 0.370 *** 0.400 *** 0.451 *** 0.497 *** Bond rating 2+ definitions Payout Size PAGE 91 91 Table 2 7. Marginal effects after regressions on f ull s ample by financi al constraint status Table 5 reports marginal effects of a change in cash on the firms excess return over the fiscal year, ERi,t. Equation 2 4 regression results are not reported to save space. t is the firms change in cash holdings from fiscal year end t 1 to t scaled by the market value of equity at year end t 1 NL (HL) indicates HASLINEt = 0 ( HASLINEt = 1), where HASLINEt is a dummy variable indicating the existence of a credit line at fiscal year end t C (U) indicates the firm is financially constrained (unconstrained) at fiscal year end t ME( t)  NL,C is the marginal effect of a change in cash holdings on the value to shareholders if the firm has credit line and is financially constrained at fiscal year end. diff(ME( t))  HL is the difference in the marginal effects of a change in cash for firms that have a credit line, and is equal to ME( t)  HL, C ME( t)  HL, U. diff(ME( t)) is the difference in the marginal effects of a change in cash for firms financially unconstrained, and is equal to ME( t)  NL, U ME( t)  HL, U. p values for All ME( t) equal report the results of the test for whether all marginal effects of a change in cash are equal: ME( t)  NL, C = ME ( t)  HL, C = ME( t)  NL, U = ME( t)  HL, U. The last p value reports results of the test for equality of all marginal effects of a change in cash excluding ME( t)  HL, U. Financial constraint status is determined according to payout ratio, size, a nd bondrating. Payout ratio is dividends plus repurchases divided by earnings, and a firm is considered constrained (unconstrained) according to payout ratio if the payout ratio is in the lower (upper) 30th percentile by year. Size is total assets, and a f irm is considered constrained (unconstrained) according to size if the firm is in the lower (upper) 30th percentile by year. A firm is considered constrained (unconstrained) according to Bondrating if the firm has positive debt and no (has a) bond rating a t the fiscal year end. According to +, a firm is considered constrained (unconstrained) if at least two of the payout, size, and bondrating definitions indicates as such for year t. ***, **, indicate statistical significance at the 1, 5, and 10% level s, respectively. Adj R20.213 0.214 0.227 0.224 Observations 22,165 14,612 22,257 18,360 Firms 4,545 3,099 4,215 4,061 ME( t)  NL, C 1.665 *** 1.774 *** 1.741 *** 1.690 *** ME( t)  HL, C 1.667 *** 1.687 *** 1.573 *** 1.656 *** ME( t)  NL, U 1.904 *** 1.708 *** 1.824 *** 1.859 *** ME( t)  HL, U 1.235 *** 1.256 *** 1.089 *** 1.128 *** diff (ME( t))  HL 0.432 *** 0.431 ** 0.485 *** 0.528 *** diff (ME( t))  U 0.670 ** 0.453 0.735 ** 0.731 p values: All ME( t) equal 0.002 *** 0.025 ** 0.000 *** 0.000 *** ME( t)  NL,C = ME( t)  HL,C = ME( t)  NL,U 0.735 0.915 0.492 0.861 Bond rating 2+ defs Payout Size PAGE 92 92 Table 2 8. Marginal effects after regressions on f andom s ample by financial constraint status This table reports marginal effects of a change in cash on the firms excess return over the fiscal year, ERi,t. Regression results are not reported to save space. t is the firms change in cash holdings from fiscal year end t 1 to t scaled by the market value of equity at year end t 1 LAt is the firms change in credit line availability from fiscal year end t 1 to t scaled by the lagged m arket value of equity at year end t 1 C (U) indicates the firm is financially constrained (unconstrained) at fiscal year end t ME( t)  C is the marginal effect of a change in cash holdings on the value to shareholders if the firm is financially con strained at fiscal year end. diff(ME( t)) is the difference in the conditional marginal effects of a change in cash for firms, and is equal to ME( t)  C ME( t)  U. diff (ME( LAt)) is the difference in the conditional marginal effects of a change in c redit line availability, and is equal to ME( LAt)  C ME( LAt)  U. ME( t) ME( LAt)  C is the difference between the marginal effects of a change in cash and credit line availability for firms financially constrained, and is equal to ME( t)  C ME ( LAt)  C. diff(ME(t) + ME( LAt)) is the difference between the sums of the marginal effects of a change and cash and credit line availability for firms constrained and unconstrained, and is equal to [ME( t)  C + ME( LAt)  C] [ME( t)  U + ME( LAt)  U]. Financial constraint status is determined according to payout ratio, size, and bondrating. Payout ratio is dividends plus repurchases divided by earnings, and a firm is considered constrained (unconstrained) according to payout ratio if the payout ratio is in the lower (upper) 30th percentile by year. Size is total assets, and a firm is considered constrained (unconstrained) according to size if the firm is in the lower (upper) 30th percentile by year. A firm is considered constrained (unconstrained ) according to Bondrating if the firm has positive debt and no (has a) bond rating at the fiscal year end. According to 2+, a firm is considered constrained (unconstrained) if at least two of the payout, size, and bondrating definitions indicates as such for year t ***, **, indicate statistical significance at the 1, 5, and 10% levels, respectively. Adj R20.275 0.308 0.317 0.325 Observations 1,233 810 1,393 1,049 Firms 276 174 261 243 ME( t)  C 1.952 *** 1.875 *** 1.952 *** 2.157 *** ME( t)  U 1.144 *** 0.751 0.256 0.270 diff (ME( t)) 0.807 1.124 ** 1.696 *** 1.887 *** ME( LAt)  C 0.233 ** 0.415 *** 0.324 *** 0.352 *** ME( LAt)  U 0.462 ** 0.448 0.500 0.472 diff (ME( LAt)) 0.229 0.033 0.176 0.120 ME( t) ME(LAt)  C 1.719 *** 1.460 *** 1.628 *** 1.805 *** ME( t) ME(LAt)  U 0.683 0.303 0.244 0.202 diff (ME( t) + ME(LAt)) 0.578 1.092 1.520 *** 1.767 *** Bond rating 2+ defs Payout Size PAGE 93 93 Table 2 9. Substitutability of cash and debt regressions on random s ample This table reports Random Sample results of regressing the firms excess sto ck return, ERi,t, on firm characteristics interacted with dummy variables indicating whether the firm is financially constrained or uncontrained. t is the firms change in cash holdings from year t 1 to year t NAt is the change in net assets (total ass ets net of cash). LDt is the change in credit line draw OLt is the change in other liabilities (liabilities not debt) ODt is the change in other debt (debt not a credit line draw) REt is the change in retained earnings Et is the change in earnings RDt is the change in research and development expense. It is the change in interest expense. Dt is the change in dividends. LTt is the change in credit line draw Ct 1 is lagged cash. Lt is market leverage. FCi,t, is a dummy variable equal to one if the firm is financially constrained at fiscal year end. All regressors with the exception of Lt and FCt are scaled by the lagged market value of equity. Here ME( t)  C (ME( t)  U) is the marginal effect of a one dollar change in cash raised by the issu ance of equity, calculated at the mean levels of lagged cash ( Ct 1) and leverage ( Lt). ME( t) + ME( LAt)  C (ME( t) + ME( LAt)  U) is the marginal effect on value to shareholders of increasing cash holdings and the credit line drawn amount by one dolla r for constrained (unconstrained) firms. Financial constraint status is determined according to payout ratio, size, and bondrating. Payout ratio is dividends plus repurchases divided by earnings, and a firm is considered constrained (unconstrained) accordi ng to payout ratio if the payout ratio is in the lower (upper) 30th percentile by year. Size is total assets, and a firm is considered constrained (unconstrained) according to size if the firm is in the lower (upper) 30th percentile by year. A firm is cons idered constrained (unconstrained) according to Bondrating if the firm has positive debt and no (has a) bond rating at the fiscal year end. According to 2+, a firm is considered constrained (unconstrained) if at least two of the payout, size, and bondrat ing definitions indicates as such for year t. ***, **, indicate statistical significance at the 1, 5, and 10% levels, respectively. PAGE 94 94 Constrained t3.085 *** 2.550 *** 2.929 *** 3.153 *** t10.909 0.329 1.123 0.972 t3.520 *** 3.298 ** 3.417 *** 4.029 *** t0.566 ** 0.533 0.380 ** 0.396 t0.977 *** 0.607 0.493 ** 0.495 t0.158 0.047 0.094 0.110 t0.320 0.245 0.279 0.256 t0.261 0.025 0.109 0.023 t0.641 *** 1.018 *** 0.673 *** 0.665 *** t1.001 1.924 2.179 ** 1.774 t0.321 0.729 0.012 0.816 t6.189 8.132 *** 2.421 2.831 t0.220 ** 0.400 *** 0.310 *** 0.331 *** Ct10.596 *** 0.800 ** 0.594 *** 0.648 *** Lt0.712 *** 0.731 *** 0.761 *** 0.858 *** Unconstrained t1.681 *** 1.651 ** 0.620 0.628 t10.574 0.985 ** 0.541 0.165 t1.797 1.291 0.308 0.655 t0.443 ** 0.411 ** 0.307 0.299 t0.874 *** 0.875 *** 0.987 *** 0.923 *** t0.141 0.493 0.452 0.758 ** t0.637 *** 0.600 ** 0.253 0.552 *** t0.153 0.665 ** 0.302 0.294 t0.683 *** 0.673 ** 0.661 *** 0.832 ** t1.369 2.333 2.600 1.753 t4.964 *** 4.395 *** 4.728 *** 4.931 *** t2.984 4.285 6.904 5.536 t0.289 0.256 0.336 0.230 Ct10.295 0.438 *** 0.624 *** 0.465 *** Lt0.456 *** 0.518 *** 0.718 *** 0.454 *** FCt0.054 0.071 0.151 *** 0.006 Intercept 0.011 0.086 *** 0.178 *** 0.061 Adj R20.276 0.317 0.318 0.332 Observations 1,233 810 1,393 1,049 Firms 276 174 261 243 ME( t)  C 2.101 *** 1.905 *** 2.020 *** 2.091 *** ME( t)  U 1.217 *** 1.150 ** 0.452 0.422 ME( t) + ME(t)  C 1.124 *** 1.298 *** 1.527 *** 1.597 *** ME( t) + ME(t)  U 0.343 0.275 0.535 0.501 Bond rating 2+ defs Payout Size PAGE 95 95 APPENDIX A VARIABLE DEFINITIONS FOR CHAPTER 1 CASHTA Cash and liquid securities to book assets at time t = (data1(t) / data6(t)) CHGASSETS Change in assets = (data6(t+1) data6(t)) / data6(t) CHGCASH Change in cash scaled by book assets = (data1(t+1) data1(t)) / data6(t) CHGCASHTCF Change in cash according to year t+1 cash flow statement scaled by SCFCF = (data274(t+1)) / SCFCF CHGDEBT Change in total debt scaled by book assets = (data9(t+1) + data34(t+1) data9(t) data34(t)) / data6(t) DEPTA D epreciation expense to book assets at time t = (data14(t) / data6(t)) EBITTA E arnings before interest and taxes scaled by book assets at time t = (data18(t) + data15(t) + data16(t)) / data6(t) FATA F ixed assets to book assets at time t = (data8(t)/data6(t)) FCF F ree cash flow = (data13(t+1) (data14(t) + data15(t) data16(t) data19(t) data21(t)) FCFTA FCF scaled by year t assets = FCF / data6(t) HASLINE Dummy variable equal to one if firm i had a committed credit line in place at fiscal year end t +1 HLAVG2 T he two digit SIC mean of HASLINE at time t INVTCF Cash flow from investing activities in year t+1 scaled by SCFCF = ( data113(t) + data109(t) + data309(t) data128(t) + data107(t) data129(t) + data310(t)) / SCFCF LEV*t+1 E stimated target leverage for year end t+1 according to Equation (2) LEV_MED Fama French 49industry median market leverage, LEVt. http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html LE Vt M arket debt ratio at time t = book value of short term and longterm debt scaled by market as sets = (data9(t) + data34(t)) / (data9(t) + data34(t) + data199(t) data25(t)) LNTA N atural log of real assets at time t 1983 US dollars = data6 1,000,000 deflated by the Consumer Price Index PAGE 96 96 MB M arket to book ratio, assets = book liabilities plus mar ket equity scaled by book assets at time t = (data9(t) + data34(t) + data10(t) + data199(t) data25(t)) / data6(t) NDITCF N et debt issues in year t+1 scaled by SCFCF = (data111(t+1) data114(t+1) + data301(t+1)) / SCFCF NEITA N et equity issues in yea r t+1 scaled by year t book assets = (data108(t+1) data115(t+1)) / data6(t) NEITCF N et equity issues in year t+1 scaled by SCFCF = (data108(t+1) data115(t+1)) / SCFCF. OTHTCF O ther cash flow in year t+1 scaled by SCFCF = (data124(t+1) + data126(t+1) + data106(t+1) + data213(t+1) + data217(t+1) + data302(t+1) + data303(t+1) + data304(t+1) + data305(t+1) + data307(t+1) + data312(t+1) + data314(t+1)) / SCFCF R D eviation from target leverage at time t : LEV*t+1 LEVt. RATED D ummy variable equal to one if the firm has a Compustat debt rating at year end t RDD Research and development expense dummy, equal to one if R&D expense data46 is nonzero at time t equal to zero if data46 is missing or zero RDTA Research and development expense to book assets = (data46(t) / data6(t)) if data46 is nonzero at time t equal to zero if data46 is missing or zero Sj E ight dummy variables indicating an observation's assignment to Subset j SCFCF Cash flows from the firm's time t+1 cash flow statement = (data123(t+1) + data125(t+1) data127(t+1)) SIGCF S ector cash flow volatility = mean of two digit SIC firm standard deviation of three to ten lags of cash flow to total assets: (data13(t) data16(t) data21(t)) / data6(t) Change in market leverage = LEVt+1 LEVt. PAGE 97 97 APPENDIX B VARIABLE DEFINITIONS FOR CHAPTER 2 BENCHMARKi,t Annualized monthly returns of Fama French 5x5 Size and Book to market portfolios to which firm i belongs over fiscal year t BONDRATINGi,t Dummy variable equal to one if firm i has a longter m bond rating in Compustat for fiscal year t Ci,t1 Cash and marketable securities (data1) for firm i at fiscal year end t 1 CVt Dummy variable equal to one if firm i violated a financial covenant during the fiscal year t Ei,t Earnings for firm i during fiscal year t : (data18 + data15 + data50 + data51). EQi,t Book equity excluding retained earnings for firm i at fiscal year end t : (data216 data36). ERi,t Excess stock return ( RETURN BENCHMARK ) for firm i over fiscal year t FCi,t Dummy variable equal to one if firm i is financially constrained at fiscal year end t FUi,t Dummy variable equal to one if firm i is financially unconstrained at fiscal year end t HASLINEi,t Dummy variable equal to one if firm i had a committed credit line in place at fisca l year end t LDi,t Credit line balance for firm i at fiscal year end t Li,t Market leverage for firm i at fiscal year end t : (data9 + data34) / (data9 + data34 + data54 data199). MEi,t1 Market equity for firm i for fiscal year t 1 : (data54 data199). NAi,t Non cash assets for firm i at fiscal year end t : (data6 data1). NFi,t Net financing for firm i during fiscal year t scaled by MEi,t1: (data108 data115 + data111 data114) / MEi,t1. NLDi,t Net line debt for firm i at fiscal year end t scaled b y MEi,t1: (data1 line draw balance) / MEi,t1. ODi,t Debt not funded by a credit line drawn for firm i at fiscal year end t : (data9 + data34 LDi,t). PAGE 98 98 OLi,t Total liabilities excluding debt for firm i at fiscal year end t : (data181 data9 data34). PA YOUTi,t Payout ratio for firm i during fiscal year t : (data115 + data127) / data13. Rei,t Proportion of distributions in the form of repurchases for firm i at fiscal year end t : data115 / (data115 + data21), or equal to zero if data115 + data21 is zero. RE TURNi,t Total stock return for firm i over fiscal year t from CRSP monthly: (PRCCFYEMt PRCCFYEMt 1 + SUM(DIVFYt)) / PRCCFYEMt 1. SIZEi,t Total assets for firm i at fiscal year end t Sj E ight dummy variables indicating an observation's assignment to Subset j i,t Change in cash and marketable securities for firm i from fiscal year end t 1 to t scaled by MEi,t 1: (data1(t) data1(t 1)) / MEi,t 1. i,t Change in common dividends for firm i from fiscal year end t 1 to t scaled by MEi,t1: (data21(t) data21(t 1)) / MEi,t1. i,t Change in earnings for firm i from fiscal year end t 1 to t scaled by MEi,t 1: (Et Et 1) / MEi,t1. i,t Change in book equity for firm i from fiscal year end t 1 to t scaled by MEi,t1: ( EQt EQt 1) / MEi,t1. i,t Change in in terest expense for firm i from fiscal year end t 1 to t scaled by MEi,t1: (data15(t) data15(t 1)) / MEi,t1. i,t Change in credit line availability for firm i from fiscal year end t 1 to t scaled by MEi,t1. i,t Change in the credit line balance f or firm i from fiscal year end t 1 to t scaled by MEi,t1. i,t Change in the credit line total commitment (balance drawn plus availability) for firm i from fiscal year end t 1 to t scaled by MEi,t1. i,t Change in non cash assets for firm i from fisc al year end t 1 to t scaled by MEi,t1: ( NAt NAt 1) / MEi,t1. i,t Change in other debt for firm i from fiscal year end t 1 to t scaled by MEi,t1: ( ODt ODt 1) / MEi,t1. i,t Change in other liabilities for firm i from fiscal year end t 1 to t sc aled by MEi,t1: ( OLt OLt 1) / MEi,t1. PAGE 99 99 i,t Change in research and development expense for firm i from fiscal year end t 1 to t scaled by MEi,t1: (data46(t) data46(t 1)) / MEi,t1, where data46 is set to zero if missing in Compustat. i,t Change in retained earnings for firm i from fiscal year end t 1 to t scaled by MEi,t1: (data36(t) data36(t 1)) / MEi,t1. PAGE 100 100 APPENDIX C DATA COLLECTION Construction of the Full Sample The creation of the variable HASLINE requires a search of 10 K forms. The 1 0 K forms are electronically available from the SECs Edgar database for some firms prior to 1996, but are widely available for U.S. public firms beginning in fiscal year 1996. I follow Sufi (2007) by coding a text search program to assist me with the data collection. For each firm surviving my initial screens of the CRSP Compustat data, I download all available 10 K forms from the SEC server to my local computer.1 I then code a SAS program to read the 10 K forms for the following keywords (following Sufi (2007)): credit line, credit facility, revolving credit, working capital line, working capital facility, lines of credit, line of credit, and revolving line For each 10 K form for each firm year observation (often firms file amendments to their first filed 10 K), t he computer program counts the number of occurrences of each keyword ( positive hits ), creates a text file containing the 21 lines of text surrounding the positive hit and searches for the following keywords in the three lines s urrounding the positive hit: did not, not have, retired, terminated, equity, expired, not renew, and cancel.2 These negative hits are also counted. Each firm year observation with a 10K search resulting in fewer than five positive hits or at least one negative hit (15,732 observations) is manually read for the existence of a line of credit type facility This procedure is an effort to limit Type I errors an indication of the existence of a line when in fact there is not one available to the firm Of the 15,732 observations 1 10K forms downloaded are: 10K, 10 K/A, 10 K405, 10K405/A, 10KT, 10 KT/A, 10KSB, 10KSB/A, 10KSB40, 10KSB40/A, 10KT405, 10KT405/A NT10 K, NT10 K/A, and NTN10K. 2 The search program ignores spaces and tabs, thus searching for cancel also returns search hits for canceled and cancellation. Line number n was always searched concatenated onto line number n1 to ensure keywords containing spaces returned search hits if stretched across two lines. PAGE 101 101 checked manually, 11,533 (73.3%) were deemed to have a committed line of credit at year end There were 27,416 observations with at least five positive hits and no negative hits, and these observations were coded automatically with HASLINE =1 There were 8,900 observations with neither positive hits nor negative hits and these observations were coded automatically with HASLINE =0 The manual check of the 15,732 observations sometimes revealed a rare reporting st yle which would lead to a likely incorrect automatic inference made by the search program. Of the 27,416 (8,900) observations set automatically as HASLINE =1 ( HASLINE =0), 89 (27) were changed manually after reading the firms 10 Ks for other years. The 10 K search resulted in a dataset of HASLINE for 52,048 observations. I lose 4,340 observations due to the fiscal year being outside my sample range of 19962006, and another 6,051 due to missing CRSP Compustat data for other required variables. The final usable dataset is on 5,396 firms with 41,657 observations. Construction of the Random Sample The data on credit line characteristics (random sample) follows the data collection in Sufi (2007). Sufi has graciously provided his data to the public on his facult y website at the University of Chicago. As detailed on his website and in Sufi (2007), the author began data collection with several screens on the Compustat universe, excluding firms in SIC codes 6000 6999. The screens applied were the following: US base d firms only at least f our consecutive years of positive data on total assets (data6) at least four consecutive years of book leverage between zero and one (data9 + data34)/data6 PAGE 102 102 at least four consecutive years of nonmissing data on: total liabilities (d ata181) total sales (data12) EBITDA (data13) share price (data199) shares outstanding (data25) preferred stock (data10) deferred taxes (data35) convertible debt (data79) Sufis application of these screens on the Compustat 19962003 data yields data for 4,604 firms. From these 4,604 firms, he randomly selects 300 for the detailed data collection for a sampling rate of approximately 300/4,604 = 0.065. Extending the dataset requires a bit of work to avoid survivorship bias: merely gathering data on Sufis 300 random firms over 20042006 will ignore firms entering Compustat in 2001 and later. Before making his data publicly available, Sufi limits the 4,604 firms to 4,503 due to missing data on other variables included in his analyses, but not included in data screens. I apply Sufis screens to the 19962003 Compustat data and am left with 5,492 firms, 888 more than does Sufi likely primarily due to data backfill in Compustat. The intersection of Sufis 4,503 firms and my 5,492 contains 4,426 firms. Thus, I have 5,492 4,426 = 1,066 firms available not verifiably available to Sufi. Next, I apply Sufis screens to the 1996 2006 data, and there are 406 firms surviving these screens that do not survive the 19962003 screens. Of these 406, 23 are in Sufis random sa mple, thus, I have 1,066 + (406 23) = 1,449 firms from which I randomly sample at Sufis 0.065 sampling rate to extend his data without survivorship bias resulting from the extension of Sufis sample. After I compile my extension of the Sufi dataset, I r equire firms to be in the CRSP/Compustat merged database, and I exclude firms in industries with SIC numbers 4900 4999 (regulated) in addition to the 60006999 (financial) restricted by Sufi. PAGE 103 103 APPENDIX D ECONOMETRIC ISSUES Short P anel Bias The firm fixed e ffects in + 1 complicate the estimation of the dynamic equation ( 1 2) since they are negatively correlated with the lagged dependent variable, resulting in a positive bias on and a negative bias on + 1 my estima te of interest from the first stage This short panel bias or Nickell Bias (Nickell (1981)) decreases in T, but Judson and Owen (1999) show that the bias can be nontrivial in panels of length 30. The Blundell and Bond (1998) GMM estimator (GMM BB) wa s derived to address th e short panel bias, and the estimator is incorporated into the Stata commands xtdpdsys and xtabond2. GMM BB is an extension of the Arellano and Bond (1991) model which estimates equations such as ( 1 2) in first differences. Differenc ing removes the fixed effect, but if regressors are not strictly exogenous then their differences are correlated with the transformed error, and Arellano and Bond show that lagged levels of these regressors can be used as instruments. If regressors are highly persistent, as in corporate finance, then first differences will (in the limit) resemble a random walk, and thus limit the ability to find any instrument. T he GMM BB estimator also known as system GMM, adds the levels equation to the Arellano Bond model and uses lagged differences to instrument for regressors in levels Blundell, Bond, and Windmeijer (2000) show that the extra moment conditions of system GMM improve on difference GMM by increasing efficiency and reducing finite sample bias in panels wi th short T and persistent regressors (see Baltagi (2005) Chapter 8 .5). PAGE 104 104 Generated R egressors I estimate ( 1 2) via GMM BB to generate + 1 and then move to the second stage and estimate ( 1 6 ). I can estimate ( 1 6 ) via pooled OLS since the firm effects are included in the calculation of the target estimate. But, the OLS standard errors suffer from the generated reg ressor pr oblem (Pagan (1984)) because most of my statistical tests do not compare an estimated coefficient against zero. The standard fix for th is standard error problem is a Murphy and Topel (1985 ) adjustment to the second stage variance covariance matrix.1 S uch adjustments are often an option, or automatically computed, in statistical packages for estimating two stage models However, the generated regressor here is unlike that in a standard two step model as i t is not simply the predicted value of the first stag es dependent variable. Rather, my generated regressor is an estimate based on a strict subset of the regressors in ( 1 2), and I use transformations of this estimate in my estimation of ( 1 6 ). Thus a clean Murphy Topel adjustment is not available to me I estimate my two step partial adjustment model with a bootstrap program whereby both stages are estimated as a system within each bootstrap iteration (see Guan (2003)). This technique ensures that sampling variation is introduced into the leverage target thereby addressing the generated regressor problem ( see Chapter 6.6 of Cameron and Trivedi (2005), and Horowitz (2001)). In the results reported in Table 1 6 the bootstrapped standard errors are an average 2.50 times larger, and never smaller than 1.68 ti mes larger, than the unadjusted OLS standard errors. 1 For a capital structure application of the Murphy Topel adjustment see Hovakimian, Opler, and Titman (2001). PAGE 105 105 APPENDIX E REPLICATION OF FAULK ENDER AND WANG (2006 ) In this paper I use the Faulkender and Wang (2006) model to estimate the value to shareholders of a marginal change in cash balances. Faulkender an d Wangs data covers 1972 2001. Since my data is limited to 19962006 due to the electronic availability of 10K forms, it is important for me to show that I can first replicate the Faulkender and Wang results for their time period. Before merging my line o f credit data with the CRSP/Compustat data, I build a dataset following Faulkender and Wang for the 1972 2006 time period. Below is a replication of the Faulkender and Wang Tables I and II. The results are indeed consistent with those in Faulkender and Wang (2006). However, there is evidence that the value to shareholders of a marginal change in cash has increased dramatically over time. In Table E 3 I estimate the Faulkender and Wang model over my 1996 2006 time period and by decade over 19722006. The mar ginal effect of a change in cash increases from 0.562 during the 1970s to 1.572 during the 2000 2006 time period. Thus, my estimates for the marginal change in cash over 1996 2006 will be greater in magnitude than those in Faulkender and Wang. The evidence that this value of a cash change has increased over time is interesting in light of the findings of Bates, et al. (2007) that firm cash holdings have dramatically increased over time. Dittmar and Mahrt Smith (2007) also use the Faulkender and Wang (2006) methodology to estimate the importance of firm governance on the value of changes in cash. Their estimates over 19902003 are also greater in magnitude compared to Faulkender and Wang, but the authors do not explore how changing governance over time affec ts the value of cash changes. This research question is left for future research. PAGE 106 106 Table E 1. Replication of Faulkender and Wang (2006), s ummary s tatistics This table reports a replication of Table I in Faulkender and Wang (2006). Summary statistics for fi rms in CRSP / Compustat over 1972 2001. ERt is the firms excess return at fiscal year end relative to a benchmark portfolio based on a Fama French 5x5 size and bookto market sorts. t is the firms change in cash holdings from year t 1 to year t Ct 1 is lagged cash. Et is the change in earnings. NAt is the change in net assets (total assets net of cash). RDt is the change in research and development expense. It is the change in interest expense. Dt is the change in dividends. Lt is market leverage. NFt is net stock and debt issues. All variables except ERt, and Lt are scaled by the lagged market value of equity. ERt0.001 0.342 0.087 0.200 0.586 t0.006 0.036 0.000 0.036 0.151 Ct10.174 0.034 0.093 0.213 0.237 t0.012 0.039 0.006 0.048 0.224 t0.007 0.104 0.024 0.154 0.541 t0.000 0.000 0.000 0.001 0.020 t0.001 0.004 0.000 0.007 0.035 t0.000 0.000 0.000 0.000 0.009 Lt0.288 0.073 0.240 0.458 0.243 NFt0.045 0.033 0.001 0.082 0.268 Mean 1st Quartile Median 3rd Quartile SD PAGE 107 107 Table E 2. Replication of Faulkender and Wang (2006), regress ion results Replication of Table II in Faulkender and Wang (2006). Regression results for excess stock returns over 19722001. ERt is the firms excess return at fiscal year end relative to a benchmark portfolio based on a Fama French 5x5 size and book to market sorts. t is the firms change in cash holdings from year t 1 to year t Et is the change in earnings. NAt is the change in net assets (total assets net of cash). RDt is the change in research and development expense. It is the change in interest expense. Dt is the change in dividends. Ct 1 is lagged cash. Lt is market leverage. NFt is net stock and debt issues. Ret is the proportion of distributions made being repurchases of stock. All variables except ERt, Lt, and Ret are scaled by the lagged market value of equity. ***, **, indicate statistical significance at the 1, 5, and 10% levels, respectively. t0.778 *** 1.555 *** 1.000 *** (0.02) (0.04) (0.05) t0.527 *** 0.521 *** 0.873 *** (0.01) (0.01) (0.03) t0.160 *** 0.169 *** 0.168 *** (0.01) (0.01) (0.01) t1.444 *** 1.364 *** 1.393 *** (0.14) (0.13) (0.24) t1.715 *** 1.636 *** 2.073 *** (0.09) (0.09) (0.12) t3.172 *** 3.109 *** 3.955 *** (0.20) (0.20) (0.25) Ct10.370 *** 0.297 *** 0.221 *** (0.01) (0.01) (0.02) Lt0.511 *** 0.508 *** 0.399 *** (0.01) (0.01) (0.01) NFt0.095 *** 0.069 *** 0.002 (0.01) (0.01) (0.02) t10.717 *** 0.435 *** (0.05) (0.07) t1.593 *** 0.993 *** (0.08) (0.11) Ret0.006 (0.01) Ret 0.160 ** (0.07) Intercept 0.068 *** 0.069 *** 0.061 *** (0.00) (0.00) (0.00) Observations 88,903 88,903 48,799 Firms 11,128 11,128 6,706 Adj R20.197 0.210 0.201 Marginal effect of t0.778 *** 0.971 *** 0.706 *** (0.02) (0.02) (0.03) I II III PAGE 108 108 Table E 3. Regressions over alternative time periods Regression results for various time period subsamples to document the change over time in the margin al effect of a change in cash holdings on the market value of equity. ERt is the firms excess return at fiscal year end relative to a benchmark portfolio based on a Fama French 5x5 size and book to market sorts. t is the firms change in cash holdings f rom year t 1 to year t Et is the change in earnings. NAt is the change in net assets (total assets net of cash). RDt is the change in research and development expense. It is the change in interest expense. Dt is the change in dividends. Ct 1 is lagge d cash. Lt is market leverage. NFt is net stock and debt issues. Ret is the proportion of distributions made being repurchases of stock. All variables except ERt, Lt, and Ret are scaled by the lagged market value of equity. ME( t) is the marginal effect o f a change in cash holdings on the value to shareholders if the firm is financially constrained at fiscal year end, calculated at the mean values of Lt and Ct 1. ***, **, indicate statistical significance at the 1, 5, and 10% levels, respectively. t2.201 *** 2.398 *** 0.971 *** 1.167 *** 1.885 *** 2.172 *** (0.07) (0.23) (0.08) (0.06) (0.07) (0.08) t0.522 *** 0.689 *** 0.578 *** 0.473 *** 0.547 *** 0.511 *** (0.03) (0.11) (0.03) (0.02) (0.02) (0.03) t0.220 *** 0.250 *** 0.103 *** 0.180 *** 0.221 *** 0.205 *** (0.01) (0.06) (0.01) (0.01) (0.01) (0.01) t0.909 *** 1.865 ** 1.296 *** 1.703 *** 1.024 *** 0.657 ** (0.24) (0.85) (0.28) (0.23) (0.20) (0.29) t2.159 *** 3.333 *** 1.044 *** 1.391 *** 2.539 *** 2.181 *** (0.23) (0.80) (0.14) (0.13) (0.18) (0.25) t2.745 *** 0.694 4.776 *** 3.169 *** 1.761 *** 2.995 *** (0.58) (2.11) (0.31) (0.35) (0.39) (0.50) Ct10.412 *** 0.469 *** 0.232 *** 0.236 *** 0.338 *** 0.357 *** (0.03) (0.10) (0.02) (0.02) (0.03) (0.03) Lt0.545 *** 0.546 *** 0.412 *** 0.468 *** 0.628 *** 0.453 *** (0.02) (0.05) (0.02) (0.01) (0.01) (0.02) NFt0.016 0.037 0.025 0.072 *** 0.104 *** 0.101 *** (0.03) (0.11) (0.02) (0.02) (0.02) (0.03) t11.059 *** 0.929 *** 0.484 *** 0.493 *** 0.992 *** 1.045 *** (0.11) (0.34) (0.08) (0.09) (0.10) (0.11) t2.127 *** 2.734 *** 0.805 *** 1.088 *** 1.953 *** 1.855 *** (0.16) (0.47) (0.14) (0.13) (0.14) (0.18)Intercept0.027 *** 0.034 0.091 *** 0.076 *** 0.060 *** 0.022 *** (0.01) (0.02) (0.01) (0.01) (0.01) (0.01)Observations30,185 2,101 18,438 27,408 35,970 22,446Firms4,784 336 3,415 5,691 7,144 5,204 Adj R20.220 0.282 0.228 0.208 0.226 0.221 1.566 *** 1.635 *** 0.562 *** 0.766 *** 1.256 *** 1.572 *** (0.05) (0.17) (0.03) (0.03) (0.04) (0.05) Marginal effect of t 20002006 Full Sample: 19962006 Random Sample: 19962006 19721979 19801989 19901999 PAGE 109 109 LIS T OF REFERENCES Acharya, V., Almeida, H., and M. Campello, 2007, Is cash negative debt? A hedging perspective on corporate financial policies, Journal of Financial Intermediation 16, 515554. Aghion, P., and P. Bolton, 1992, An incomplete contracts approa ch to financial contracting, Review of Economic Studies 59, 473494. Almeida, H., Campello, M., and M. Weisbach, 2004, Th e cash flow sensitivity of cash, Journal of Finance 59, 17771804. Alster, N., 2008, Businesses tapping credit lines, but new ba nk loans drying up, Investors Business Daily December 16, 2008. Altinkili, O., and R. Hansen, 2000, Are there economies of scale in underwriter spreads? Evidence of rising external financing costs, Review of Financial Studies 13, 191218. Antoniou, A., Gun ey, Y., and K. Paudyal, 2008, The determinants of capital structure: Capital market oriented versus bank oriented institutions Journal of Financial and Quantitative Analysis 43, 5992. Arellano, M., and S. Bond, 1991, Some tests of specification for pane l data: Monte Carlo evidence and an application to employment equations Review of Economic Studies 58, 277297. Baker, M., and J. Wurgler, 2002, Market timing and capital structure Journal of Finance 57, 1 30. Baltagi, B., 2005, Econometric Analysis of Panel Data, 3rd edn. (John Wiley, New York ) Bates, T., Kahle, K., and R. Stulz, 2007, Why do U.S. firms hold so much more cash than they used to? Journal of Finance forthcoming. Baumol, W., 1952, The transactions demand for cash: an inventory theoret ic approach Quarterly Journal of Economics 66, 545556. Billett, M., and J. Garfinkel, 2004, Financial flexibility and the cost of external finance for U.S. banks Journal of Money, Credit, and Banking 36, 827852. Billett, M., King, T., and D. Mauer, 2 007, Growth opportunities and the choice of leverage, debt maturity, and covenants Journal of Finance 62, 697730. Blundell, R., and S. Bond, 1998, Initial conditions and moment restrictions in dynamic panel data models Journal of Econometrics 87, 1151 43. PAGE 110 110 Blundell, R., Bond, S., and F. Windmeijer, 2000, Estimation in dynamic panel data models: improving on the performance of the standard GMM estimator Advances in Econometrics 15, 5391. Boot, A., Thakor, A., and G. Udell, 1991, Competition, risk neut rality, and loan commitments Journal of Banking and Finance 11, 449471. Brettell, K., 2008, Station Casinos draws remaining cash on bank line Reuters News December 22, 2008. Byoun, S., 2008, How and when do firms adjust their capital structures towar d targets? Journal of Finance forthcoming. Cameron, A., and P. Trivedi, 2005, Microeconometrics: Methods and Applications (Cambridge University Press, New York ). DeAngelo, H., and L. DeAngelo, 2007, Capital structure, payout policy, and financial flexibility, University of Southern California Working paper. DeAngelo, H., DeAngelo, L., and T. Whited, 2008, Capital structure dynamics and transitory debt University of Southern California Working paper. DeAngelo, H., and R. Masulis, 1980, Optimal capital structure under corporate and personal taxation Journal of Financial Economics 8, 3 29. Denis, D., and V. Sibilkov, 2007, Financial constraints, investment, and th e value of cash holdings Purdue University Working paper. Dewatripont, M., and J. Tirole, 1994, A theory of debt and equity: diversity of securities and manager shareholder congruence Quarterly Journal of Economics 109, 10271054. Dittmar, A., and J. M ahrt Smith, 2007, Corporate governance and the value of cash holdings, Journal of Financial Economics 83, 599634. Dudley, E., 2008, Capital structure and large investment projects University of Florida Working paper. Dittmar, A., Mahrt Smith, J., and H Servaes, 2003, International corporate governance and corporate cash holdings, Journal of Financial and Quantitative Analysis 38, 111133. Fama, E., and K. French, 1993, Common risk factors in the returns on stocks and bonds Journal of Financial Econom ics 33, 356. Fama, E., and K. French, 1998, Taxes, financing decisions, and firm value Journal of Finance 53, 819843. PAGE 111 111 Fama, E., and K. French, 2002, Testing trade off and pecking order predictions about dividends and debt Review of Financial Studies 15, 133. Faulkender, M., Flannery, M., Hankins, K., and J. Smith, 2008, Do adjustment costs impede the realization of target capital structure? University of Kentucky Working paper. Faulkender, M., and M. Petersen, 2006, Does the source of debt affect capital structure? Review of Financial Studies 19, 4579. Faulkender, M., and R. Wang, 2006, Corporate financial policy and the value of cash Journal of Finance 61, 19571990. Fischer, E., Heinkel, R., and J. Zechner, 1989, Dynamic capital structure choice: theory and tests Journal of Finance 44, 1940. Flannery, M., and K. Rangan, 2006 Partial adjustment toward target capital structures Journal of Financial Economics 79, 469506. Foley, C., Hartzell, J., Titman, S., and G. Twite, 2007, Why do firm s hold so much cash? A tax based explanation Journal of Financial Economics 86, 579607. Frank, M., and V. Goyal, 2008, Profits and capital structure University of Minnesota Working paper. Gamba, A., and A. Triantis, 2008, The value of financial flexib ility Journal of Finance forthcoming. Gilchrest, S., Himmelberg, C., 1995, Evidence on the role of cash flow for investment Journal of Monetary Economics 36, 541572. Gilson, S., and J. Warner, 1998, Private versus public debt: evidence from firms tha t replace bank loans with junk bonds Harvard University Working paper. Goldstein, R., Ju, N., and H. Leland, 2001, An EBIT based model of dynamic capital structure Journal of Business 74, 483512. Graham, J., and C. Harvey, 2002, How do CFOs make capit al budgeting and capital structure decisions? Journal of Applied Corporate Finance 15, 823. Guan, W., 2003, From the help desk: bootstrapped standard errors The Stata Journal 3, 7180. Harford, J., Mansi, S., and W. Maxwell, 2006, Corporate governance and a firms cash holdings Journal of Financial Economics 87, 535555. Holmstrom, B., and J. Tirole, 1998, Private and public supply of liquidity, Journal of Political Economy 106, 140. PAGE 112 112 Horowitz, J., 2001, The bootstrap, i n Heckman, J., and E. Leamer, e ds. : Handbook of Econometrics vol 5, ch 52, pp 31593228, (North Holland, Amsterdam ) Hovakimian, A., Opler, T., and S. Titman, 2001, The debt equity choice Journal of Financial and Quantitative Analysis 36, 124. Huang, R., and J. Ritter, 2008, Te sting theories of capital structure and estimating the speed of adjustment Journal of Financial and Quantitative Analysis forthcoming. Jagannathan, M., Stephens, C., and M. Weisbach, 2000, Financial flexibility and the choice between dividends and stoc k repurchases Journal of Financial Economics 57, 355384. James, C., 1987 Some evidence on the uniqueness of bank loans Journal of Financial Economics 19, 217235. Jensen, M., 1986, Agency costs of free cash flow, corporate finance and takeovers Amer ican Economic Review 76, 323329. Johnson, S., 1997, An empirical analysis of the determinants of corporate debt ownership structure Journal of Financial and Quantitative Analysis 32, 4769. Johnson, S., 2008, GM drains credit line to meet challenge CFO.com September 22, 2008. Judson, R., and A. Owen, 1999, Estimating dynamic panel models: a guide for macroeconomists Economics Letters 65, 915. Kahl, M., Shivdasani, A., and Y. Wang, 2008, Do firms use commercial paper to enhance financial flexibil ity? University of North Carolina Chapel Hill Working paper. Kashyap, A., Lamont, O., and J. Stein, 1994, Credit conditions and the cyclical behavior of inventories Quarterly Journal of Economics 109, 565592. Keynes, J., 1936, The general theory of em ployment. In: Interest and Money, Harcourt Brace, London. Leary, M., and M. Roberts, 2005, Do firms rebalance their capital structures? Journal of Finance 60, 25752619. Lemmon, M., Roberts, M., and J. Zender, 2008, Back to the beginning: persistence an d the cross section of corporate capital structure, Journal of Finance 63, 15751608. Lins, K., Tufano, P., and H. Servaes, 2007, What drives corporate liquidity? An international survey of strategic cash and lines of credit University of Utah Working pa per. PAGE 113 113 MacMillan, R., 2008, Tribune draws $250 million to increase cash position Reuters News October 17, 2008. Mikkelson, W., and M. Partch, 2003, Do persistent large cash reserves hinder performance? Journal of Financial and Quantitative Analysis 38, 275294. Miller, M., and D. Orr, 1966, A model of the demand for money by firms Quarterly Journal of Economics 80, 413435. Mills, J., 1926, Table of the ratio: area to bounding ordinate, for any portion of normal curve Biometrika 18, 395400. Mulliga n, C., 1997, Scale economies, the value of time, and the demand for money: longitudinal evidence from firms, Journal of Political Economy 105, 10611079. Murphy, K., and R. Topel, 1985, Estimation and inference in two step econometric models Journal of B usiness & Economic Statistics 3, 370379. Myers, S., and N. Majluf, 1984, Corporate financing and investment decisions when firms have information investors do not have Journal of Financial Economics 13, 187221. Nickell, S., 1981, Biases in dynamic mod els with fixed effects Econometrica 49, 14171426. Nini, G., Smith, D., and A. Sufi, 2008, Creditor control rights and firm investment policy, Journal of Financial Economics forthcoming. Opler, T., Pinkowitz, L., Stulz, R., and R. Williamson, 1999 The determinants and implications of cash holdings, Journal of Financial Economics 52, 346. Pagan, B., 1984, Econometric issues in the analysis of regressions with generated regressors International Economic Review 25, 221247. Pinkowitz, L., Stulz, R., a nd R. Williamson, 2006, Does the contribution of corporate cash holdings and dividends to firm value depend on governance? A cross country analysis Journal of Finance 61, 27252751. Pinkowitz, L., and R. Williamson, 2004, What is a dollar worth? The mar ket value of cash holdings, Georgetown University Working paper. Riddiough, T., and Z. Wu, 2008, Financial constraints, liquidity management, and investment, Real Estate Economics forthcoming. Reed, J., 2009, Ford to draw bank credit of $10.1 billion F inancial Times January 29, 2009. Roberts, M., and A. Sufi, 2008, Control rights and capital structure: an empirical investigation Journal of Finance forthcoming. PAGE 114 114 Roodman, D., 2006, How to do xtabond2: and introduction to 'difference' and 'system' GMM in Stata Center for Global Development Working Paper No. 103. Roy, A., 1951, Some thoughts on the distribution of earnings Oxford Economic Papers 3, 135146. Staff Reports, 2008, Some drew down Lehman loans ahead of filing, Credit Investment News Sept ember 19, 2008. Strebulaev, I., 2007, Do tests of capital structure theory mean what they say? Journal of Finance 62, 17471787. Sufi, A., 2007, Bank lines of credit in corporate finance: an empirical analysis Review of Financial Studies forthcoming. Welch, I., 2004, Capital structure and stock returns Journal of Political Economy 112, 106131. Whited, T., 1992, Debt, liquidity constraints, and corporate investment: evidence from panel data Journal of Finance 47, 14251460. Wilcoxon, F., 1945, Individual comparison by ranking methods Biometrics 1, 8083. Wilson, M., and A. Thomson, 2008, FOCUS: Bank woes rise as companies draw on loans Dow Jones International News October 14, 2008. Windmeijer, F., 2005, A finite sample correction of linear eff icient two step GMM estimators Journal of Econometrics 126, 2551. Yun, H., 2008, The choice of corporate liquidity and corporate governance Review of Financial Studies forthcoming. PAGE 115 115 BIOGRAPHICAL SKETCH Brandon recei ved his Bachelor of Science in m ana gement with certificates in e conomics and finance from the Georgia Institute of Technology in 1998; after which he worked for five years as a commercial banker at SunTrust Banks, Inc. and Security Bank Corporation in Georgia. In 2004, Brandon left industr y as an assistant vice president of commercial lending to begin his Ph.D. studies at the University of Florida. After graduating with the Ph.D. in finance in August 2009, Brandon began work a s an assistant professor of f inance at the University of Nebraska Lincoln. Chapter 1 of this dissertation won the Western Finance Associations 2009 Trefftz Award. 