TWO ESSAYS ON CORPORATE FINANCE By KRISTINE W. HANKINS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006
To my husband and my parents
iii ACKNOWLEDGMENTS I wish to thank all the faculty members in the Department of Finance at the University of Florida for creating a challengi ng and nurturing research environment. In particular, I wish to express my gratitude to Mahendrarajah Nimalendran for many years of helpful feedback and encouragement, and to Steven Slutsky and Sarah Hamersma of the Department of Economics, who both provide d invaluable assistan ce and friendship. I also am fortunate to have Jim Seale from the Department of Food and Resource Economics and Mike Ryngaert as supervisor y committee members who were never too busy to review my last-minute drafts. Above all, I am indebted to Mark Flannery (my supervisory committee chair) for keeping the ba r high and always having an open door.
iv TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iii LIST OF TABLES.............................................................................................................vi ABSTRACT.....................................................................................................................viii CHAPTER 1 INTRODUCTION........................................................................................................1 2 ARE ACQUISTIONS AN OPERATIONAL HEDGE? THE INTERACTION OF FINANCIAL AND OPERAT IONAL HEDGING.......................................................3 Introduction...................................................................................................................3 Financial versus Operational Hedging..........................................................................6 Model of Hedging Alternatives....................................................................................9 Data........................................................................................................................... ..12 Measures of Interest Ra te Exposure and Hedging..............................................14 Measures of Volatility.........................................................................................16 Methodology and Results...........................................................................................17 Do Acquisitions Affect Income Volatility?.........................................................18 Do Managers Use Acquisitions to Manage Risk?...............................................19 Is Derivatives Use Responsiv e to Operational Hedging?....................................22 Conclusion..................................................................................................................24 3 A THEORY OF CAPITAL STRUCTURE ADJUSTMENT SPEED.......................34 A Theory of Capital Structure Adjustment Speed......................................................37 Adjustment Speed Factors: Adjustment Costs....................................................38 Adjustment Speed Factors: Adjustment Benefits................................................39 The Econometrics of Capital Structure Adjustment...................................................40 Partial Adjustment and Capital Structure............................................................41 Dynamic Panel Estimation..................................................................................43 Data......................................................................................................................45 Panel Length Bias................................................................................................46 Bias-Corrected Least Squares Dummy Variable.................................................47 Heterogeneous Adjustments Costs and Benefits.................................................48 Empirical Evidence on the Costs and Be nefits of Rebalancing Leverage..................51
v Adjustment Costs and Financial Constraint Proxies...........................................52 Adjustment Costs and External Financing Proxies.............................................54 Adjustment Costs and Stock Price Movements...................................................57 Adjustment Benefits............................................................................................58 Dominant Adjustment Factors.............................................................................60 Conclusion..................................................................................................................61 4 CONCLUSION...........................................................................................................78 APPENDIX: VARIABLE DEFINITIONS........................................................................79 LIST OF REFERENCES...................................................................................................80 BIOGRAPHICAL SKETCH.............................................................................................86
vi LIST OF TABLES Table page 2-1 All Acquirers: Summary of Derivatives Use............................................................ 26 2-2 Acquirers with Positive Hedging: Summary of Derivatives Use.............................. 26 2-3 All Targets: Summar y of Derivatives Use................................................................ 27 2-4 Targets with Positive Hedging: Summary of Derivatives Use................................. 27 2-5 Pre-Acquisition Intere st Rate Sensitivity.................................................................. 28 2-6 Acquisitionsâ€™ Impact on Interest Rate Sensitivity..................................................... 28 2-7 Average Impact of Acquisitions on Volatility.......................................................... 28 2-8 Percent of Acquisitions Decreasing Volatility.......................................................... 28 2-9 Risk Management and the Propensity to Merge....................................................... 29 2-10 Propensity to Merge an d Acquisition Preference...................................................... 30 2-11 Controlling for Selection: Derivatives Use after Acquisitions.................................. 31 2-12 Controlling for Panel Attributes: Derivatives Use after Acquisitions.......................32 2-13 Volatility and the Change in Financial Hedging....................................................... 33 3-1 Summary Statistics.................................................................................................... 63 3-2 Panel Length Sensitivity............................................................................................ 64 3-3 Baseline Adjustment Speed Estimation.................................................................... 65 3-4 Two-Stage Estimates of the Speed of Adjustment.................................................... 66 3-5 Adjustment Speed Factor Proxies............................................................................. 67 3-6 LSDVC: Adjustment Costs and Financial Constraints............................................. 68 3-7 GMMBB: Adjustment Costs and Financial Constraints............................................. 69
vii 3-8 LSDVC: Adjustment Cost s and External Financing................................................. 70 3-9 GMMBB: Adjustment Costs and External Financing................................................ 71 3-10 LSDVC: Adjustment Co sts and Price Movements...................................................72 3-11 GMMBB: Adjustment Costs and Price Movements................................................... 73 3-12 LSDVC: Adjustment Benefits................................................................................... 74 3-13 GMMBB: Adjustment Benefits.................................................................................. 75 3-14 Dominant Factor Analysis......................................................................................... 76
viii Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy TWO ESSAYS ON CORPORATE FINANCE By Kristine W. Hankins August 2006 Chair: Mark J. Flannery Major Department: Finance, Insurance, and Real Estate Firms have many risk management tools at their disposal. How a firm uses these choices alone and as part of a choice set is less well unde rstood. I examine two major risk management decisions in the corporate fi nance arena. First, I address the use of operational hedging (corporate fina nce activity that reduces firm risk). I document that acquisitions are operational hedges and that firms substitute operational and financial hedging. Next, I explore the speed of capital structure adjustment. Capital structure decisions are an important part of risk ma nagement and I document that the costs and benefits of adjustment are significant factor s in determining leverage. Collectively, my research presents new information on how fi rms use two major risk management tools: operational hedges and capital structure adjustment.
1 CHAPTER 1 INTRODUCTION Firms can earn high returns by undertaking more risk. Modern finance, however, recognizes that this positive relationship be tween risk and return is limited. Both excessive and idiosyncratic risk s create real costs for firm s. Acknowledging these costs, a firm can manage risk to improve firm valu e. My dissertation examines two major risk management decisions in the corporate fina nce arena: the use of operational hedging and the leverage adjustment process. I first document the substitution of operati onal and financial hedging in Chapter 2. Both financial hedging derivatives and gene ral corporate finance decisions are risk management tools. However, it is not unde rstood how managers us e these financial and operational hedges together to manage risk. This chapter document s the substitution of these choices. Acquisitions that reduce firm risk provide an operational hedge. Firms reduce their financial hedging with derivatives when opera tional hedging increases. I also examine the capital structure adjustment process (Chapter 3). Optimal capital structure balances the risks and benef its of debt. While debt provides tax and governance benefits, it also increases the poten tial for financial distress. My study makes three key contributions. Using a recently developed dynamic panel model estimation technique, I provide evidence of rapid rebalanc ing. On average, firms reach their target capital structure in fewer than 5 years. I pres ent a theory of capital structure adjustment. Rebalancing costs and benefits vary among firm s. These factors affect the benefits of
2 achieving the optimal leverage and, therefore, a firmâ€™s adjustment speed. I show that these factors are significant determinants of a firmâ€™s capital structure adjustment speed.
3 CHAPTER 2 ARE ACQUISTIONS AN OPERATIONAL HEDGE? THE INTERACTION OF FINANCIAL AND OPER ATIONAL HEDGING Introduction In the world of Modigliani and Miller ( 1958), risk management is not a tool for value maximization. However, capital market imperfections such as financial distress, tax convexity, and external financing create a cost to cash flow volatility (Smith and Stulz (1984), Tufano (1996)). Empirical ev idence confirms the importance of these costs: 71% of the large1 bank holding companies (BHC) use derivative hedges against interest rate, foreign exchange , equity, or commodity risks. Yet income volatility also can be reduced with operational hedging, such as adjusting operati ng leverage, sourcing parts in different countrie s, or diversifying cash flow s through project choice or acquisitions. Using a large sample of BHCs, I find that acquisitions can reduce income volatility and that managers substitute this operational hedging for financial hedging. Despite this, acquisitions and other firm organization decisions are frequently considered independently of derivatives use. If operational hedges affect volatility and therefore ri sk management, appreciating their relationship to financial hedging is essential to the capital budgeting, hedging, and diversification literatures. My study investigates this rela tionship by introducing a simple model of optimal hedging when multiple avenues for risk management exist and ____________ 1 Large BHCs are defined as those with at least $1 billion in total assets.
4 empirically analyzing the rela tionship between financial a nd operational hedging. In the model, management aims to maximize firm value by limiting costly volatility while taking into account the expe nse of hedging. Optimal risk management balances the relative costs of each hedging c hoice with its contribution to reducing volatility. Any increased use of one hedge should result in an offsetting decline in the alternative hedging tool. That is, if vol atility matters, then operational decisions that reduce idiosyncratic risk will impact th e use of derivatives for hedging. Dramatic changes in operational hedging should provide the eas iest observation of risk management tradeoffs. Amihud and Le v (1981) and Aggarwal and Samwick (2003) argue that diversifying idiosyncratic risk is a key motiv ation for mergers and acquisitions (M&A). Thus, I focus on how this firm organization decision affects hedging with derivatives using a dataset of bank holding comp anies. I concentrat e on interest rate hedging because this exposure comprises the overwhelming majority of BHC derivatives hedging. Consistent with my hypothesis, I document a substitution between hedging with derivatives and acquisition activ ity. Not only is an acquisi tion more likely if risk exposures are not actively hedged with derivati ves, but financial he dging decreases after operational hedging increases. It should be noted that ac quisitions could influence fi nancial hedging even in the absence of operational hedging. If the acquire râ€™s intrinsic risk exposure shifts with an acquisition, derivatives use shoul d adjust. This provides so me information about risk management (e.g., that managers respond to ch anging exposures), but it does not address whether some acquisitions serve as an ope rational hedge and how that influences derivatives use. As aforementioned, th e reliance on financial hedging decreases
5 following acquisitions. Evidence of this subs titution exists even after controlling for changing risk exposure. For example, assume a firm hedges a certain percentage of its risk exposure. If the firm hedges a signifi cantly smaller percent after an acquisition, controlling for the change in exposure due to the acquisition, operational hedging may have increased. I also test whether an ac quisitionâ€™s impact on financial hedging varies with its contribution to operational hedgi ng. Acquisitions that increase the scale of the firm, but do not affect cash flow volatili ty in any material manner, offer little in the way of operational hedging. Conversely, ac quisitions that alter volatility should be factored into the total risk management. The empirical eviden ce presented in my study suggests that those acquisitions offering the most operational hedging lead to the la rgest reductions in hedging with derivatives. If acquisition activity provide s operational hedging, then my research pertains to not only the hedging literature but also basic theory of the firm issues. Coase (1937) established the discussion on what determines the boundaries of a fi rm and whether those boundaries affect resource allocat ion. More recently, Berger et al. (2005) provided empirical support for the notion that organi zational form influences the choice of business activities. This implies that if ri sk management impacts firm organization, the core business activities of a firm may cha nge. Hedging choices could influence firm value not just by minimizing the costs of vola tility but also by cha nging project selection or business practices. My research also may contribute to u nderstanding some of the unexplained crosssectional variation in divers ification discount literature (Stein 2003). Benston, Hunter,
6 and Wall (1995) conclude that banks value more highly those acquisitions that diversify earnings. This could be explained by the re duction in volatility and, possibly, financial hedging costs. Financial versus Operational Hedging No comprehensive survey exists of th e multiple avenues for corporate hedging. Prior research focuses primarily on the relations hip of derivatives to either firm value or leverage but overlooks the risk management asp ects of corporate decisi ons such as capital budgeting and diversification. He dging with derivatives infl uences debt ratios (Graham and Rogers 2002), but it is less clear whether hedging affects firm value. Allayannis and Weston (2001) document a positive correlation be tween firm value and foreign exchange hedging while Jin and Jorion (2005) find no such relationship with commodity hedging in the oil and gas industry. The industry eff ect on hedging decisions also is ambiguous. Nain (2005) finds within-industr y practices are important influences on risk management decisions while Haushalter (2000) documents great variation in w ithin-industry hedging practices. The relationship between firm size and hedging is equally murky. Most small firms do not employ financial hedging â€“ an obser vation frequently attributed to the fixed costs of establishing a he dging program. However, Haushalter (2000) finds some evidence that fraction of production hedged is negatively associated with size amongst actively hedging oil and gas producers. This is somewhat surprising given the assumption that hedging with derivatives has a low marginal cost. My research offers a potential explanation of this phenomenon. It is feasible that firm size is positively correlated with operational hedging if firms increase their total assets through acquisitions, geographic expansion, or any othe r project selection that diversifies cash
7 flow. Larger firms may substitute this in creased operational hedging for alternative risk management, such as derivatives use, l eading to the observed relationship between financial hedging and firm size. The banking literature provides additional evidence that corporate decisions are related to risk management. Both Diamond (1984) and Brewer et al . (2000) find that bank lending is related to hedging. Hughes et al . (1999) argue that bank expansion that diversifies risk will reduce risk manageme nt costs. And Cebenoyan and Strahan (2004) note that active credit risk hedgers hold less capital. Numerous papers consider the relative importance of fi nancial versus operational hedging. Guay and Kothari (2003) contend that derivatives appear to cover only a small part of a firmâ€™s risk profile. They conclude most risk stems from sources that cannot be financially hedged. This finding, coupled with the Froot and Stein (1998) conclusion that unhedgable risks will alter both capital struct ure and investment policy, highlights the potential importance of opera tional hedging. The prior ev idence on whether operational and financial hedging are substitutes or co mplements is ambiguous. Nance, Smith, and Smithson (1993) briefly note that other financial policies, such as adjustments to leverage or dividends, may substitute for derivatives , while operational and financial hedging are found to be complements in the empirical study of exchange rate exposures by Allayannis, Ihrig, and Weston (2001). Gecz y, Minton and Schrand (1999) examine risk management choices in the natural gas i ndustry and find mixed evidence on whether hedging alternatives are complements or subs titutes. In contrast, my investigation documents evidence of operational hedgi ng (through acquisitions) substituting for derivatives use.
8 My research is not the first to suggest that acquisitions may provide an operational hedge. The Wall Street Journal often highlights an acquisitionâ€™s effect on risk exposures and volatility (Editors 2004, Samor 2004). Moreover, the academic literature has recognized the potential risk management benefits of M&A activity since Lewellen (1971). Stulz (1990) asserts that costless acq uisitions that reduced cash flow volatility would benefit shareholders and Santomero (1997) notes that credit risk is diversifiable through acquisitions. Amihud and Lev (1981) conc lude that managerial risk aversion is a significant determinant of acqui sition activity; alt hough my paperâ€™s empirical results are not consistent with this agen cy motivation. I document a d ecline in derivatives use after an increase in operational hedging. Risk aver sion would lead the manager to seek an overall decrease in volatility not maintain th e current level by subs tituting operational for financial hedging. Also, Esty, Narasimhan, and Tufano (1999) examine how the interest rate environment affects bank acquisitions. They find the competitive dynamics of bank mergers change with interest rates movement s and acquisition prices are a function of the current interest rate. Unlike Esty et al . , I examine acquisitions as a risk management tool, rather than a byproduct of risk exposure. Most relevant to this i nquiry is the finding of Benston, Hunter, and Wall (1995) that banks bid more for targets that diversify earnings. This lends indirect support to my hypothesi s that acquisitions can provide operational hedging. Ceteris paribus , acquisitions that reduce volatility should command a premium. Still, there is little consen sus in the literature on acqui sitions and firm value. Recent work on the diversification discount in dicates that the discount may disappear or possibly become a premium after accounting for selection bias (Graham, Lemmon, and Wolf 2002, Villalonga 2004). While my paper does not address the value of hedging, to
9 the extent that acquisitions vary in their contribution to risk management (some reduce volatility greatly while others have no effect ), differences in operational hedging benefits may explain some of the cross-sectional variat ion in the value of diversification. I now present a brief model of risk management choices. This model demonstrates how increased operational hedging will reduce the expenditure on financial hedging which, in turn, could affect firm value. Model of Hedging Alternatives Cash flow volatility is costly for a firm due to capital market imperfections such as tax convexity and costly external financing. Froot, Scharfstein, and Stein (1993) model optimal hedging in response to these costs. I use their model as a foundation but focus on optimal hedging when multiple risk management choices are available. Furthermore, I incorporate the cost of hedging that must (log ically) affect a managerâ€™s risk management decisions. For simplicity, assume there are two risk management choices, which may be thought of as financial hedging and ope rational hedging. Each hedging choice (h1, h2) has a positive cost and I assume hedging in creases firm value only by means of reducing costly volatility. The manager wishes to select the optimal he dging portfolio that maximizes firm value. ihMax) ( ) ( )) , ( (2 2 1 1 2 1h C h C h h VRM (1) where VRM is the value of risk management, is a parameter representi ng the cost of volatility, is firm volatility, a func tion of the hedging choices, hi is the level of hedging choice i, Ci is the cost of hedging choice i.
10 Existing evidence on the costs of hedging suggests that initia ting a derivatives program has a high fixed cost (Mian 1996). Ac quisitions also have a high fixed cost. Therefore, I assume a basic li near cost function for both h1 and h2. i i i i ih c F h C ) ( (2) where Fi is the fixed cost of hedging choice i, ci is the marginal cost of hedging choice i. In making hedging choices, managers are c onstrained in their hedging expenditure. This constraint is assumed to be a function of how costly volatility is for the firm. I therefore maximize Equation 1 such that: ) (2 2 2 1 1 1 K F h c F h c (3) where K is the hedging budget that is a function of . The first order conditions of this co nstrained maximization problem are: 0 ) ( ) (1 1 1 1 c c h h VRM (4a) 0 ) ( ) (2 2 2 2 c c h h VRM (4b) 0 ) (2 2 2 1 1 1 K F h c F h c VRM (5) Solving Equations 4a and 4b for and equating the first order conditions with respect to h1 and h2: 2 2 1 1c h c h (6)
11 Optimal risk management balances each hedging choiceâ€™s cont ribution to reducing volatility against its marginal cost. This holds even without the budget constraint. Next, I solve Equation 5 for h1 and substitute Equation 6, to find the impact of the budget constraint: 2 1 2 1 2 1 1) ( * h h h c F F K h (7) That is, the optimal choice of h1 is a function of the maximum amount of h1 (the amount available to spend on hedgi ng divided by the cost of h1) minus the relative effectiveness of the other hedging choice (h2). Equation 7 illustrates that hedging decisions are a f unction of the other risk management tools available and that the relative costs affect the optimal hedging strategy. If M&A activity increases operational hedging (h2) and there is a budget constraint, then it is optimal for management to reduce their use of financial hedging (h1). As financial hedging becomes more expensiv e (or less useful), operational hedging may be substituted. This model also indicates that hedging is increasing with the cost of volatility ( ). This basic model motivates three empiri cally testable hypotheses concerning the risk management tradeoffs between operational and financial hedging. Acquisitions can be an operational hedge. Whether an acquisition is an operational hedge depends on its potential to re duce volatility. To test this, I estimate how an acquisition will impact an acquirerâ€™s income volatility. For acquisitions where income data is available for the twelve quarters preceding the acquisition for both the target and acquirer, I compare the income vol atility of acquirer alone with that of the
12 target and acquirerâ€™s income if it were co mbined on a quarterly basis over the same twelve quarters. Financial hedging is related to acquisition activity. If BHCs utilize acquisitions to manage risk, acquisitions and derivatives will not be independent. I investigate whether current risk management predicts future acquisition activity by estimating a probit model of the propensity to acquire. In addition, if acquisitions are an alternative tool to reduce volatility, a trad e-off should exist and derivatives usage should decline. I test this by measuring the change in deri vatives following an ac quisition using both a Heckman selection mode l and panel analysis. Operational hedging and financial hedging are substitutes. The degree to that an acquisition reduces volatility determines the change in financial hedging. To test this, I model the post-acquisition change in derivatives use as a function of the acquisitionâ€™s impact on volatility. Data Measuring the hedging activity for most t ypes of firms requires laborious data collection from 10-K filings. However, BHCs report their derivatives use in the quarterly Federal Reserve Y-9C filings. Beginning in 1995, derivatives used for trading purposes and non-trading purposes were reported separa tely. Therefore, my study will use data from BHCs to examine how firms adjust financial hedging followi ng acquisitions. The dataset constructed from 1995 to 2003 Federa l Reserve quarterly filings includes the entire universe of bank holding companies wit h total consolidated assets of $150 million or more. Only top-tier BHCs are examined since risk may be managed across subsidiaries. The Y-9C fili ngs categorize the derivatives into interest rate, foreign exchange, equity derivative, and commodity/other contract s, and identify non-trading
13 (hedging) versus trading positions. The empirical analysis of my research is limited to hedging with interest rate derivatives as su ch contracts comprise 97% of BHC hedging. Detailed deal information for BHCs involve d in business combinations valued at $50 million or more is obtained from the SDC Platinum Mergers database.2 From the SDC Platinum database, there are 487 M&A deals identified involving a bank holding company. This deal informa tion is combined with the panel of BHC quarterly filings. To be included in the sample, both parties must be bank holding companies. This excludes the acquisitions of non-banks or partial acquisitions (such as the acquisition of bank branches or business segments) and ensures that risk exposure, derivatives use, and firm char acteristics are known for both th e target and acquirer. Of the 487 deals, BHC information was availabl e and matched for 448 acquirers. Quarterly bank information, including de rivatives usage, is matched to acquirers. All of these variables are winsorized at the 1st and 99th percentiles to remove potential outliers. Historically, bank regulation has varied by state. Restric tions on bank merger activity were no exception. Some states be gan to permit M&A before 1970 while others resisted deregulation until the early 1990s (S trahan 2003). To control for differences in state legislation that might affect acquisition activity, the time since deregulation (Strahan 2003) is matched to each BHC by state. In addition, I control for the compos ition of the balance sheet as business composition may shape hedging decisions. BHC control variables are generated by dividing the BHC asset categorie s by the total assets (Sched ule HC of the FR-Y9C). ____________ 2 A minimum deal value of $50 million limits possible data errors (such as deal values of zero) and inconsequential acquisitions. At the time of an acquisition, the median total assets for a BHC are $5,309,524,000. The conclusions are robust to a minimum deal value of $20 million.
14 However, Allen and Saunders (1992) show that these quarter end numbers are susceptible to â€˜window-dressingâ€™ adjustments. They note that the most active window-dressing on the asset side is in securities, Federal funds, and loans. Therefore, the quarterly average is substituted for each of thes e three asset groups as well as the total assets throughout the dataset (Schedule HC-K of the FR-Y9C). Measures of Interest Rate Exposure and Hedging Interest rate exposure is exp ected to influence the level of interest rate hedging. Following the methodology of Flannery and Ja mes (1984), a measure of interest rate sensitivity â€“ the one year maturity gap â€“ is constructed by subtracting the reported liability exposure subject to repricing within a year from the asset exposures subject to the same repricing time period (S chedule HC-H of the FR-Y9C). This net sensitivity is measured relative to the average quarterly tota l assets. Flannery and James note that this metric assumes that unexpected changes to the interest rate sensitivity affect the bank in a manner proportional to the short term net sensit ivity. Similar one year gap measures of the mismatch between the asset and liabilit ies exposures are used by Brewer, Jackson, and Moser (2001). t t t tTA s Liabilitie ST Assets ST y Sensitivit IR (8) where ST Assets are those assets that mature or reprice within one year, ST Liabilities are those liabilities that mature or reprice within one year, TA is the quarterly average of consolidated assets. The measure of financial hedging is th e BHCâ€™s end-of-quarter gross notional amount of interest rate derivatives used for he dging divided by total a ssets. To detect the substitution of operational hedging for financ ial hedging, I measure the changing use of derivatives for hedging purposes over one and two year horizons:
15 t t t or t t or tTA IRG TA IRG Hedging IR ) 8 ( 4 ) 8 ( 4 (9) where IRG is the gross notional amount of derivatives used to hedge interest rate risk. The gross notional amount of derivatives does not capture the true hedging position if some of the contracts offset one anothe r (Graham and Rogers 2002). This introduces an upward bias into this measurement. Wh ile the net derivatives would be preferable, empirical examinations indicate that the di fference between net and gross positions is minor.3 Furthermore, gross notional amounts bias against finding any decline in financial hedging. Controlling for the ch ange in interest rate sensitivity, a BHCâ€™s gross notional volume of derivatives would be expected to increase or remain constant following an acquisition. First, acquiring a target without a derivatives program provides economies of scale with respect to th e fixed costs of a hedging progr am. The target could hedge without incurring the in itial fixed costs of establishing its own program. Therefore, derivatives use would increase for the combined firm. Sec ond, combining two firms with derivatives programs would result in a constant use of derivatives. And, lastly, derivative contracts are not normally cancelle d; new ones are just written.4 Therefore, the reorganization of any existi ng contracts with the combin ation of two firms would increase derivatives use. All of these issues bias the empirical analysis against finding a decrease in financial hedging. ____________ 3 Graham and Rogers (2002) state, â€œW e conclude, however, that using net, as opposed to total, positions is only marginally important in helping identify factors that affect corporate hedging decisions.â€ â€œOur important findings with respect to the tax incentives to hedge are unchanged [between gross notional and net positions.]â€ 4 Stulz (2004) discusses that closing derivatives pos itions often involves purchasing an offsetting contract.
16 Given the measure of financial hedging is re lative to the quarterly average assets, changes to the asset size could impact the empirical findings. If the gross notional amount of hedging is constant in the year following the acquisition (IRGt+4 = IRGt), the IR Hedging could decline simply given the acquisitionâ€™s impact on the size of assets. To control for this, a new dependent variable ( IR Hedging_Size) is generated. t t t or tTA IRG IRG Size Hedging IR ) 8 ( 4_ (10) This variable removes the potential size e ffect. The empirical analysis is conducted using both measures for the change in hedgi ng with qualitatively similar results. Measures of Volatility To address the mechanism by which acquisitions provide operational hedging, this paper evaluates the targetâ€™s impact on the acq uirerâ€™s volatility. Volatility calculations based on the BHC net income would include the effect of current hedging. Therefore, a new variable OI is created: t t tDeriv NI OI (11) where OI is operational income, NI is net income, Deriv is the impact on income of derivatives held for hedging. The net change in interest income and expense due to hedging is provided on Schedule HI of the FR-Y9C and is subtracted from the net income on a quarterly basis. From OI , volatility is calculated without the influence of derivatives. Next, I measure the level of operationa l hedging introduced by an acquisition by examining the volatility of the acquirer and targ et had they been a combined entity for the three years preceding the acquisition. This captures how management expected the target to impact the acquirerâ€™s operational income vol atility. The volatility of these twelve
17 combined quarterly observations is compared to the volatility of the twelve quarterly observations of the acquirer alone. OVAcquirer = St. Dev. ( A t A t A t A tTA OI TA OI, 1 , 1 , 12 , 12,... ) (12a) OVCombined = St. Dev. ( T t A t T t A t T t A t T t A tTA TA OI OI TA TA OI OI, 1 , 1 , 1 , 1 , 12 , 12 , 12 , 12,... )) (12b) Impact% Acquirer Combined AcquirerOV OV OV (13) where Impact% is the percentage change in operational volatility due to the acquisition, OIt,A is the operational income of the acquirer at time t, OIt,T is the operational income of the target at time t. Methodology and Results Managing interest rate risk is a priority for BHCsâ€™ risk management. Tables 2-1 and 2-2 show that interest rate derivativ es are employed more than other derivative contracts. The sample is divided into tw o groups; observations wh ere an acquisition is made and observations where no acquisition is made. The median and mean derivatives levels relative to the quarterly average of total assets are presented for both sub-samples. While interest rate hedging a nd trading dominate derivatives use, other derivatives use informs the likelihood of an interior solution. Table 2-1 indicates th at BHCs, on average, exhibit a higher level of hedging, as well as trading, when an acquisition is made. However, Table 2-2 reveals that the reverse holds when the sample is limited to BHCs that use each derivative contract. For active hedgers, th e mean amount of interest rate hedging at the time of an acquisition is 3.9% of average quarterly total assets versus 7.8% when no acquisition is made. Ta ble 2-3 shows that target BH Cs exhibit a similar pattern
18 but (perhaps caused by the small sample si ze) the difference is not statistically significant. While the statistics documented in Tables 2-1 and 2-2 support the hypothesis that acquirers have different risk management practices, these BHCs simply may have a lower level of interest rate exposure â€“ leading to a lower need for hedging. Therefore, Table 2-5 presents the average IR Sensitivity (Equation 8) by target and acquirer status. Acquisitions and targets both ha ve significantly more intere st rate exposure than nonmerging institutions,5 but the larger exposure does not e xplain the difference in financial hedging. Merging BHCs have more exposure to interest rate m ovements but hedge less than other institutions. Yet, acquisitions do not si gnificantly change the average BHCâ€™s interest rate sensitivity (Table 2-6). There is no statis tically significant change in interest rate exposure between the year before and the year after the acquisition. This implies that while acquirers appear to manage risk diffe rently, acquisitions arenâ€™t being used to directly reduce intere st rate exposure. Do Acquisitions Affect Income Volatility? To investigate why acquirers hedge less of their interest rate exposure than other BHCs, I explore whether acquisitions can provide operational hedging. My sample contains 208 matched pairs of acquirers and targets where both are bank holding companies and have at least three years of data before the acquisition. A dealâ€™s Impact% is, relative to the acquirerâ€™s volatility, the difference between the acquirerâ€™s volatility for ____________ 5 IR Sensitivity is significantly higher for acquirers and targets than BHCs not involved in M&A activity. This is true whether it is measured at the time of M&A or one year prior to the event.
19 the three years preceding the deal and the volat ility of the acquirer and the target if they were combined during that period. Volatility is measured as the standard deviation of the quarterly operational income divided by total assets. If the combined net income volatility is smaller than that of the acquirer alone, Impact% is positive. A positive impact implies the acquisition would reduce income volatility ceteris paribus . Reductions in income volatility indicate the target has potential operational hedging benefits or potential savings associated w ith lower costs of convex taxation, potential financial distress, and external capital. Table 2-7 shows that, on average, BHC acquisitions increase operational hedging by reducing income volatility. The Impact% coefficient indicates that, on average, volatility decreases 5.9%. Furt hermore, this average increase is not driven by outliers as 86% of acquisitions create ope rational hedging (Table 2-8). Do Managers Use Acquisitions to Manage Risk? A probit analysis is used to examine th e likelihood of making an acquisition or being a target given current exposures and ri sk management over the short-term horizon. Using the methodology of Billett (1996), the data set is split into four sub-samples: 19961997, 1998-1999, 2000-2001, and 2002-2003. A two year horizon period balances the predictive power of current risk management w ith the need for an adequate sample size. The likelihood of an acquisition during each peri od is predicted using the prior year's first quarter BHC information. For example, the bi nary dependent variable equals unity if an acquisition is made during 1996 or 1997 and is modeled as a function of 1995 Q1 data. This also is done using 1997 Q1 with 1998-1999 da ta, etc. (Using the first quarter of the prior year reduces correlati on with adjacent quarters and minimizes the impact of any adjustments due to the expected merger.) The four non-overlapping sub samples are
20 pooled and a probit model, with year dumm ies and clustering at the individual BHC level, is estimated.6 M&A_dum = IR Sensitivityt + IR Hedgingt + IR Tradingt + Private t (14) + Deregt + BHC Controlst + t M&A_dum = High Sensitivityt + Less Hedgingt + (High Sensitivityt* Less Hedgingt) + Private t + Dereg t + BHC Controls t + t (15) where M&A_dum is a binary variable equaling unity if the observation is an acquirer or a target. IR Sensitivity is the net interest rate exposure over the next year. IR Hedging is the gross notional amount of inte rest rate derivatives used for hedging divided by the quarterl y average of total assets. IR Trading is the gross notional amount of in terest rate derivatives used for trading divided by the quarterl y average of total assets. Private is a binary variable equaling unity if the BHC is not SEC registered. Dereg is the time since the BHCâ€™s state deregulated interstate M&A activity. BHC Controls are the log of the quarterly aver age of total assets and the BHC asset categories divided by the qua rterly average of total assets High Sensitivity is a binary variab le equaling unity if IR Sensitivity is above the median. Less Hedging is a binary variable equaling unity if interest rate derivatives for hedging are not used or are below the IR Hedging median of BHCs that hedge. Table 2-9 presents the results of this anal ysis. The first and third columns indicate that the IR Sensitivity and IR Hedging do not appear to predict merger activity. Yet, the second column shows that BHCs with higher sensitivity to interest rate movements ( High Sensitivity ) but who are less active hedgers ( Less Hedging ) have a higher propensity to merge. That is, the absolute level of sensit ivity to short-term interest rate movements does not appear to be important for future acquisitions. What is important is whether a BHC with a large risk exposur e is actively hedgin g with derivatives. This finding is ____________ 6 This analysis also was conducted using seven overlapping periods (1996-1997, 1997-1998, 1998-1999, 1999-2000, 2000-2001, 2001-2002, 2002-2003) and the results are qualitatively similar and retain similar levels of significance.
21 consistent with the hypothesis that both derivatives and M&A are hedging choices and supports Proposition #2 that financial hedging is related to acquisition activity. Private firms also are less likely to be involved with M&A. Given the unconditional probability of a BH C making an acquisition in any quarter is 1.6%, this is a substantial decrease in the likelihood of M&A. Private firms may find idiosyncratic risk less costly, as argued by Xu and Malkiel ( 2003), or external capital for financing acquisitions may be more difficult to access. Private firms also may be less visible targets or may reap fewer benefits from joining an established hedging program. The increased propensity to acquire documented in Table 2-9 should not immediately be attributed to risk manageme nt preferences. It is conceivable that High Sensitivity, Less Hedging firms â€“ due to their lower us e of financial derivatives â€“ experience more volatile cash fl ows. If these firms survive, they may have more cash relative to active hedgers and be more able to make acquisitions. I attempt to distinguish between the risk management prefer ences and the cash-on-hand hypotheses by investigating the type of acquisitions made.7 If acquisitions were driven by risk management preferences, and not si mply cash-on-hand, I would expect High Sensitivity, Less Hedging firms to prefer volatility-reducing ac quisitions. In Table 2-10, I estimate whether this preference exists using a probit model. OpHedge_dum t = (High Sensitivity t x Less Hedging t) + BHC Controls t + t (16) where OpHedge_dum is a binary variable equaling unity if the acquisition reduces volatility. In the first column, any acquisiti on that reduces volatility counts as an operational hedge. In the second column, only acquisitions reducing volatility at least 2% are counted as an operational hedge. ____________ 7 I wish to thank Bob Jennings for this suggestion.
22 There is modest support for the risk ma nagement hypothesis. Firms not actively hedging a large exposure to in terest rate movements exhibit some preference for acquisitions that provide operational hedging. These firms are more likely to acquire targets that reduce their volatility. The weak empirical results indicate that while risk management is not the only motivation for acquisitions, there is some evidence of endogeneity. Is Derivatives Use Responsive to Operational Hedging? The results presented in Tables 2-9 a nd 2-10 indicate acquisi tion activity may be correlated with risk exposures and risk mana gement. If managers actually recognize the potential hedging benefits of acquisitions and believe they can substitute for financial hedging, they should adjust the use of other ri sk management activities. To test this, I estimate the treatment effect (the acquisiti onâ€™s impact on derivatives use) using the following regressions: IR Hedging t, t+4 (t, t+8) = Acquirer t + IR Sensitivity t, t+4 (t, t+8) + BHC Controls t + BHC Controls t ,t+4 (t ,t+8) + t (17) IR Hedging, Size t, t+4 (t, t+8) = Acquirer t + IR Sensitivity t, t+4 (t, t+8) (18) + BHC Controls t + BHC Controls t, t+4 (t, t+8) + t where IR Hedging is the change in hedging relative to total assets over the next year (or two). IR Hedging, Size is the change in hedging over the ne xt year (or two), controlling for size. Acquirer is a binary variable equaling unity if an acquisition is made that quarter. BHC Controls are the change in the quarterly assets and BHC categories over the same period as the dependent variable. Obviously, addressing the endogenous relationship be tween acquisitions and derivatives â€“ as both appear to be risk ma nagement choice variables â€“ is important for unbiased and consistent estimates of how fi rms manage risk. The average impact of making an acquisition on financial hedging is estimated using the Heckman two-stage
23 selection model, which attempts to minimize the selection bias as an explanation for the treatment outcomes. The selection lambda, co mmonly referred to as the inverse Millâ€™s ratio, is included in the sec ond stage estimation to correct fo r the potential selection bias. This approach estimates the acquisition decisi on with a probit mode l based on the results of Table 2-9. The selection criteria are Total Assets, Private, and IR Sensitivity .8 Table 2-11 presents the Heckman two-stag e coefficient estimates. Regardless of the time horizon and dependent variable specif ication, and controlli ng for interest rate sensitivity changes, financial hedging declin es following an acquisition. The selection lambda is positive and, for the two year horizon measures, significant. Since the acquisition decision is modeled as a function of risk expos ure, BHC size, and private status, acquisitions considered unexpected by this model may not be motivated by risk management. The positive lambda implies th at such unexpected acquisitions decrease their financial hedging less following an acquisition as might be expected with acquisitions driven by concerns other than risk management. A shortcoming of the Heckman approach is that it neglects the dataâ€™s panel attributes. Therefore, in Table 2-12, th e change in financial hedging following an acquisition is examined using both random a nd fixed effect models (again, estimating Equations 17 and 18). Once again, derivatives hedging de creases significantly over both the one and two year horizons even after cont rolling for the change in interest rate exposure and the composition of the BHC. ____________ 8 Note that the specific Heckman selection criteria choice does not materially affect the coefficient estimates presented in this paper.
24 If the post-acquisition decline in derivatives is due to the increase in operational hedging, then acquisitions that create the most operational hedging should lead to the largest declines in financial he dging. Therefore, I regress Impact% against the change in derivatives use. IR Hedging t,t+4 (t,t+8) = Impact%t + BHC Controls t + BHC Controls t,t+4 (t,t+8) + t (19) Table 2-13 indicates that acquisitions that reduce volatility are followed by reduced financial hedging. Since Impact% , the measure of operational hedging, is positive when the acquisition contributes to operational hedgi ng, negative coefficients indicate the postacquisition hedging is negatively related to the volatility im pact. That is, the more operational hedging created, the more financ ial hedging will decline. This supports Proposition #3 that operational and financial hedging are substitutes. Conclusion My research provides empirical evidence on risk management tradeoffs between M&A activity and derivatives use for bank hol ding companies. After providing a simple model of optimal risk management, I presen t three main findings. First, acquisitions reduce income volatility. Second, managers recognize the risk management potential of acquisition activity. And, lastly, operational he dging is substituted for derivatives use. The results imply that risk management is not exogenous to firm or ganization. This has vast implications for the an alysis, specifically the econom etric specification, of hedging and firm value. Furthermore, the implications of my st udy extend beyond the hedging literature. Variations in the diversification discount ma y relate to an acquisi tionâ€™s contribution to hedging. Also, the documented trade off between financial and operational hedging
25 implies that managerial risk aversion may not be a primary motivation for M&A activity. And, most significantly, risk management may affect firm value not only by minimizing the cost of volatility, but also by influencing firm organization. It must be restated that this dataset only examines bank holding companies. Whether non-financial firms recognize acquisiti ons as an operational hedge is unknown. Clearly, there is much more work to be done at this intersection of risk management and corporate finance. That being said, my resear ch highlights some of the possible issues for future researchers to consider.
26 Table 2-1. All Acquirers: Summary of Derivatives Use. An Acquirer in Qtr t Not an Acquirer in Qtr t Difference between Means # Obs Median Mean St.Dev. # Ob s Median Mean St.Dev. Diff. Sigf. Hedging IR 448 0.000 0.016 0.034 54165 0.000 0.007 0.050 0.009 *** FX 448 0.000 0.001 0.002 54109 0.000 0.000 0.003 0.001 *** Equity 448 0.000 0.000 0.000 54097 0.000 0.000 0.000 0.000 *** Commodity 448 0.000 0.000 0.000 54093 0.000 0.000 0.000 0.000 Trading IR 448 0.000 0.046 0.122 54102 0.000 0.007 0.088 0.039 *** FX 448 0.000 0.012 0.039 54093 0.000 0.002 0.022 0.010 *** Equity 448 0.000 0.000 0.000 54078 0.000 0.000 0.001 0.000 *** Commodity 448 0.000 0.000 0.000 54079 0.000 0.000 0.000 0.000 *** This table summarizes the level of derivativ es use for hedging and trading purposes over the four derivatives categor ies of interest rate (IR), foreign exchange (FX), equity, and commodity. Derivatives us e is measured as the gross notional amount relative to total assets. Table 2-2. Acquirers with Positive Hedging: Summary of Derivatives Use. An Acquirer in Qtr t Not an Acquirer in Qtr t Difference between Means # Obs Median Mean St.Dev. # Ob s Median Mean St.Dev. Diff. Sigf. Hedging IR 186 0.020 0.039 0.044 4665 0.036 0.078 0.154 -0.040 *** FX 101 0.003 0.003 0.004 1111 0.004 0.008 0.019 -0.004 *** Equity 9 0.000 0.000 0.000 527 0.001 0.002 0.003 -0.002 *** Commodity 0 0 Trading IR 158 0.025 0.131 0.176 2001 0.068 0.197 0.415 -0.066 *** FX 109 0.030 0.050 0.066 1666 0.025 0.066 0.106 -0.016 *** Equity 18 0.002 0.002 0.001 539 0.002 0.003 0.004 -0.001 *** Commodity 31 0.000 0.000 0.000 363 0.000 0.000 0.000 0.000 *** This table summarizes the level of derivativ es use for hedging and trading purposes over the four derivatives categor ies of interest rate (IR), foreign exchange (FX), equity, and commodity. Derivatives us e is measured as the gross notional amount relative to total assets.
27 Table 2-3. All Targets: Summ ary of Derivatives Use. A Target in Qtr t Not a Target in Qtr t Difference between Means # Obs Median Mean St.Dev. # Ob s Median Mean St.Dev. Diff. Sigf. Hedging IR 448 0.000 0.009 0.061 54165 0.000 0.007 0.050 0.002 FX 448 0.000 0.000 0.001 54109 0.000 0.000 0.003 0.000 *** Equity 448 0.000 0.000 0.000 54097 0.000 0.000 0.000 0.000 Commodity 448 0.000 0.000 0.000 54093 0.000 0.000 0.000 0.000 Trading IR 448 0.000 0.012 0.066 54102 0.000 0.008 0.089 0.004 FX 448 0.000 0.004 0.026 54093 0.000 0.002 0.022 0.002 * Equity 448 0.000 0.000 0.000 54078 0.000 0.000 0.001 0.000 Commodity 448 0.000 0.000 0.000 54079 0.000 0.000 0.000 0.000 This table summarizes the level of derivativ es use for hedging and trading purposes over the four derivatives categor ies of interest rate (IR), foreign exchange (FX), equity, and commodity. Derivatives us e is measured as the gross notional amount relative to total assets. Table 2-4. Targets with Positive Hedging: Summary of Derivatives Use. A Target in Qtr t Not a Target in Qtr t Difference between Means # Obs Median Mean St.Dev. # Ob s Median Mean St.Dev. Diff. Sigf. Hedging IR 64 0.027 0.068 0.158 4787 0.035 0.077 0.152 -0.009 FX 15 0.004 0.003 0.002 1197 0.004 0.007 0.019 -0.004 *** Equity 4 0.001 0.002 0.002 532 0.001 0.002 0.003 0.000 Commodity 0 0 Trading IR 38 0.034 0.148 0.193 2121 0.065 0.193 0.405 -0.044 FX 33 0.022 0.066 0.079 1742 0.026 0.065 0.104 0.001 Equity 11 0.002 0.002 0.001 546 0.002 0.003 0.004 -0.001 *** Commodity 10 0.000 0.000 0.000 384 0.000 0.000 0.000 0.000 This table summarizes the level of derivativ es use for hedging and trading purposes over the four derivatives categor ies of interest rate (IR), foreign exchange (FX), equity, and commodity. Derivatives us e is measured as the gross notional amount relative to total assets.
28 Table 2-5. Pre-Acquisition In terest Rate Sensitivity. An Acquirer in Qtr t Not an Acquirer in Qtr t Difference between Means # Obs Mean St Dev # Obs Mean St Dev Diff. Sigf. IR Sensitivityt-4 446 0.146 0.137 49680 0.068 0.191 0.078 *** A Target in Qtr t Not a Target in Qtr t Difference # Obs Mean St Dev # Obs Mean St Dev Diff. Sigf. IR Sensitivityt-4 447 0.118 0.193 49679 0.068 0.191 0.050 *** This table presents the average IR Sensitivity from one year before the observation. This measure is the difference between the s hort term asset and liability exposure to interest rate movements relative to the quarterly average of total assets. Table 2-6. Acquisitionsâ€™ Impact on Interest Rate Sensitivity. # Obs Mean St Dev Significance IR Sensitivity 439 -0.010 0.128 This table summarizes the acquisitionsâ€™ average impact on interest rate sensitivity for the acquirer. There are 439 acquisitions where the interest sensitivity can be calculated both one year before a nd one year after the acquisition. IR Sensitivity is the difference in between IR Sensitivity at the two times. Table 2-7. Average Impact of Acquisitions on Volatility. # Obs Mean St Dev Significance Impact % 208 0.059 0.116 *** Impact% is the mean percent change in volatil ity (the difference between the prior 12 quarters income volatility of the acquirer alone ( Vol_Acquirer) versus the prior 12 quarters if the target and acqui rer were a combined entity over that period ( Vol_Combined ) relative to the acquirer alone). A positive mean indicates that volatility decreased wi th the acquisition and that operational hedging increased. Table 2-8. Percent of Acquisi tions Decreasing Volatility # Obs Mean St Dev Significance # Obs. Percentage Decrease Volatility 179 86% Increase Volatility 29 14% 208 This reports the percentage of the 208 acqui sitions that decreased volatility for the acquirer.
29 Table 2-9. Risk Management and the Propensity to Merge. Marginal Effects Acquirer Target IR Sensitivity -0.001 0.016 (0.918) (0.372) IR Hedging -0.025 -0.019 (0.610) (0.848) IR Trading 0.019 0.129 (0.356) (0.002) Private -0.023 -0.021 -0.037 -0.033 (0.000) (0.000) (0.000) (0.000) Dereg 0.001 0.001 0.001 0.001 (0.003) (0.002) (0.166) (0.068) High Sensitivity -0.017 -0.036 (0.070) (0.088) Less Hedging -0.009 -0.010 (0.300) (0.487) High Sensitivity & Less Hedging 0.021 0.032 (0.036) (0.137) BHC Controls Yes Yes Yes Yes Year Dummies Yes Yes Yes Yes # Obs 5403 5403 5403 5403 The likelihood of being an acquirer or targ et during each of four sub-samples (19961997, 1998-1999, 2000-2001, 2002-2003) is pred icted using the prior year's data. The four sub-samples are pooled and the probit, with year dummies, is conducted. Errors are clustered at the BHC level. This table presents the marginal effects. High Sensitivity is a binary variab le equaling unity if IR Sensitivity is above the median. Less Hedging is a binary variable equaling unity if the BHC has less th an the median amount of IR Hedging of those with hedging programs or the BHC does not he dge. BHC Controls are included. P values are in parentheses. M&A_dum = IR Sensitivity t + IR Hedging t + IR Trading t + Private t + Dereg t + Total Assets t + BHC Controls t + t M&A_dum = High Sensitivity t + Less Hedging t + (High Sensitivity t x Less Hedging t) + Private t + Dereg t + BHC Controls t + t.
30 Table 2-10. Propensity to Merge and Acquisition Preference. Marginal Effects Acquisitions that Reduce Volatility Any > 2% > 4% > 6% > 8% High Sensitivity & Less Hedging Dummy 0.101 0.160 0.150 0.187 0.199 (0.109) (0.050) (0.101) (0.085) (0.100) BHC Controls Yes Yes Yes Yes Yes # Operational Hedges 179 120 94 72 59 # Obs 208 161 136 115 102 This table shows the preference for acquisiti ons that provide opera tional hedging using a probit analysis, presenting the marginal effects. The dependent variable ( OpHedge_dum ) is an indicator equaling un ity if the acquisition reduced volatility. In the first column, all acq uisitions that reduce volatility are considered operational hedges. By the last column, the acquisition must reduce volatility by at least 8% to be an operational hedge. BHC Controls are included. Robust p values are in parentheses. OpHedge_dum t = (High Sensitivity t x Less Hedging t) + BHC Controls t + t
31 Table 2-11. Controlling for Selection: Derivatives Use af ter Acquisitions. 1 Year Horizon 2 Year Horizon IR Hedging IR Hedging, Size IR Hedging IR Hedging, Size Acquirer t -0.003 -0.027 -0.031 -0.072 (0.799) (0.091) (0.047) (0.005) IR Sensitivity 0.007 0.008 0.010 0.018 (0.002) (0.006) (0.000) (0.000) Constant -0.010 -0.017 -0.010 -0.007 (0.187) (0.086) (0.340) (0.706) BHC Controls Yes Yes Yes Yes BHC Controls Yes Yes Yes Yes hazard: lambda 0.000 0.009 0. 012 0.026 (0.980) (0.189) (0.078) (0.021) # Obs 26206 24909 21107 19968 A Heckman selection model is used with the selection criteria of quarterly assets, interest rate sensitivity, and a private firm i ndicator. The dependent variable is the change in the ratio of interest rate hedging to total assets ( IR Hedging ) in the year (or two years) following th e observation. BHC Controls and BHC Controls also are included. P values are in parentheses. IR Hedging t, t+4 (or t, t+8) = Acquirer t + IR Sensitivity t, t+4 (or t, t+8) + BHC Controls t + BHC Controls t, t+4 (t,t+8) + t IR Hedging, Size t, t+4 (or t, t+8) = Acquirer t + IR Sensitivity t, t+4 (or t, t+8) + BHC Controls t + BHC Controls t, t+4 (t, t+8) + t
32 Table 2-12. Controlling for Panel Attribut es: Derivatives Use after Acquisitions. IR Hedging, Size 1 Year Horizon 2 Year Horizon Random Effects Fixed Effects Random Effects Fixed Effects Acquirer -0.007 -0.006 -0.011 -0.010 (0.035) (0.073) (0.012) (0.032) IR Sensitivity 0.009 0.010 0.013 0.012 (0.002) (0.002) (0.001) (0.004) Constant -0.006 -0.005 -0.021 -0.057 (0.513) (0.818) (0.164) (0.123) BHC Controls Yes Yes Yes Yes BHC Controls Yes Yes Yes Yes # Obs 27054 27054 21791 21791 # Groups 1871 1871 1644 1644 R2 0.033 0.052 This table examines the change in financ ial hedging following an acquisition using random and fixed effects models. The depe ndent variable is the size controlled change ( IR Hedging, Size ). BHC Controls and BHC Controls also are included. P values are in parentheses. IR Hedging, Size t, t+4 (or t, t+8) = Acquirer t + IR Sensitivity t, t+4 (or t, t+8) + BHC Controls t + BHC Controls t, t+4 (t, t+8) + t
33 Table 2-13. Volatility and the Change in Financial Hedging. IR Hedging, Size 1 Year Horizon 2 Year Horizon Impact% -0.022 -0.050 (0.072) (0.000) Constant -0.191 -0.393 (0.000) (0.000) BHC Controls Yes Yes BHC Controls Yes Yes # Obs 189 171 R2 0.111 0.348 This table shows the change in financial hedgi ng using an OLS regression clustered at the BHC level. The dependent variable is the one year change in interest rate hedging following the acquisition adjusted for size ( IR Hedging, Size ) over both a one and two year horizon. BHC Controls and BHC Controls also are included. P values are in parentheses. IR Hedging, Size t, t+4 (t, t+8) = Impact% t + BHC Controls t + BHC Controls t, t+4 (t, t+8) + t
34 CHAPTER 3 A THEORY OF CAPITAL STRUCTURE ADJUSTMENT SPEED Ever since the Miller and Modigliani â€œirrelevanceâ€ decl aration in 1958, academics have debated the determinants of capital structure. Classic arguments focused on the tradeoff between the costs and benefits of debt with seminal papers concerning asset substitution (Jensen and Meckling 1976) and und erinvestment (Myers 1984). The pecking order theory documented by Donaldson (1961) and revived by Myers and Majluf (1984) highlighted the importance of information as ymmetry in capital structure. More recent additions to the litera ture include the behavi oral hypotheses of market timing (Baker and Wurgler 2004) and inertia (Welch 2004). Yet after all this, current capital struct ure research largely advocates a refined version of the trade-off theory (Frank a nd Goyal 2006). The revised trade-off theory recognizes that capital market im perfections create a cost to rebalancing leverage and this tempers the speed of adjustment. Numerous recent papers conclude that dynamic tradeoff models dominate alternative hypotheses (Hovakimian, Opler, and Titman 2001, Mehotra, Mikkelsen, and Partch 2003, Flannery and Ra ngan 2006). This litera ture concludes that firms actively pursue target debt ratios even t hough market frictions lead to an incomplete adjustment in any one period. Understanding the capital structure adjustment process is integral to the trade-off literature. Yet, there is little consensus on even the basic elements of adjustment. Adjustment speed estimates vary enormously. Fama and French (2002) estimate that firms
35 adjust between 7-18% each year, while R oberts (2002) documents adjustment speeds approaching 100% for some industries. Part of the dissension stems from econometric issues. Unobserved firm level heterogeneity now is a recognized component of capital structure decisions (Flannery and Rangan 2006, Lemmon, Roberts, and Zender 2006) but dynamic panel estimation is challenging. Th e unbalanced dynamic panel bias-corrected least squares dummy variable approach deve loped by Bruno (2005) offe rs a new approach for estimating the speed of adjustment. This technique controls for panel length bias of fixed effects estimation (Nickell 1981) wit hout requiring the omissi on of younger or shortlived firms from empirical analysis. Using this, I document an average adjustment of ~22% each year. This indicates that firms actively pursue their optimal debt ratio. While capital market imperfections hinder immediat e rebalancing, this evidence supports the rapid adjustment speeds documented by Flannery and Rangan (2006). Further, little is known about the origin and magnitude of the capital structure adjustment costs even though costs are re cognized as a nontrivial impediment to rebalancing (Fischer et al . 1989, Leary and Roberts 2005). Existing research models an individual firmâ€™s target debt ra tio as a function of certain firm-l evel characteristics, such as the costs of distress. I h ypothesize that if the optimal level of debt varies by firm, so might the cost of adjusting the debt ratio and the value of maintaining that target. Leverage decisions are guided by more than an unrelen ting pursuit of the optimal debt ratio; the costs and benefits of reaching that target are significant determinants of the adjustment process. The cost of changing the debt ratio must be weighed against the costliness of prolonged deviation from the optimum. To the extent that adjustment costs and benefits
36 matter for firmsâ€™ capital structure decisions, vari ations in these factor s will affect the speed of adjustment. Consistent with this hypothesis, severa l recent papers diverge from the earlier assumption of adjustment speed uniformity across firms and provide evidence of great heterogeneity. Hovakimian, Opler, and Titm an (2001) document faster adjustments with debt reductions and with ove rleveraged firms, and Robert s (2002) shows that slower rebalancing is associated with certain industry char acteristics. Leary and Roberts (2005) note that adjustment costs differ with various rebalancing options. For example, debt issuance is less expensive so lower adjustme nt benefits are needed to exceed the rebalancing cost. MacKay and Phillips ( 2005) demonstrate that each firmâ€™s competitive landscape is a factor in adjustment process. In my analysis, I present evidence that the speed of adjustment is not equal across all firm s and, in fact, it varies systematically with the costs and benefits of adjustment. Given that the premise of costly adjustment is vital to the revised trade-off theory, it is surprising that little theoretical or empirical work exists on this topic. With an in-depth investigation of adjustment speeds, my pape r contributes in three ways to the capital structure literature. First, I present a new th eory of capital structure adjustment and discuss expected firm behavior. Capita l structure reflects the costs a nd benefits of rebalancing in addition to the target debt level. Second, as econometric concerns ar e a serious issue with this topic and this form of data, I employ a new econometric application â€“ the corrected least squares dummy variables estimation (LSDVC) technique â€“ in my analysis. Previously unavailable for unbalanced dynamic pa nel data, this is the first paper to my knowledge that estimates capital structure issues (or any corporate finance topic) with this
37 approach. I review the adva ntages and limitations of this technique and compare it to existing methodologies. Lastly, using a dyna mic panel dataset, I document the crosssectional and intertemporal vari ation in capital structure re balancing. This empirical evidence provides substantial support for the theory. A Theory of Capital Structure Adjustment Speed In trade-off theory, the optimal debt ratio ba lances the costs and benefits of leverage. I hypothesize that a similar trad e-off exists for adjustment sp eed decisions. Just as the target debt level fluctuates among firms, so does the importance of ma intaining the target and the cost of rebalancing. While the op timal leverage balances the advantages and disadvantages of debt financing, the speed of adjustment weighs re balancing costs against the costliness of deviating from the target. For example, firms near distress have a different incentive to adjust their debt ratio than more stable firms. As the costs of suboptimal leverage and capital structure adju stment vary, so will the adjustment speed. Slower adjustment speeds are predicted in the presence of higher adjustment costs and faster adjustment speeds are predicted when deviations are more costly. Adjustment costs and benefits reflect many el ements of corporate finance. Financial constraints, external financi ng costs, and stock price moveme nts all contribute to the costs of adjustment. The benefits, meanwhile, de pend on the potential costs of distress, the value of tax shields, and managerial perquisite s. This section introduces these adjustment speed factors and their expected impact on reba lancing. The factor proxies and empirical results are presented later in the paper.
38 Adjustment Speed Factors: Adjustment Costs A firm has four rebalancing options. It can retire debt or issue equity when overleveraged and it can repurchase shares or issue debt when underleveraged. As capital market frictions exist, these actions require either financial flexibility or external financing. To the extent a firm is less financially flexib le or has limited access to the capital markets, changing the debt ratio becomes burdensome. Stock price movements also contribute to rebalancing. Financial constraints. Just as internal cash availa bility may color investment decisions (Myers and Majluf 1984), internal fina ncing constraints may impact the ability to retire debt or repurchase shares and, in turn , play an important role in the adjustment process. A firmâ€™s internal financial fl exibility is comprised of cash inflows and constraints. I examine both facets. For inst ance, higher profitability provides funds for share repurchases as well as the financial stabil ity to issue securities at an attractive rate. Conversely, cash outflows such as invest ments and dividends reduce the amount of available cash and restrict managementâ€™s ability to easily alter leverage. I expect the speed of adjustment to decrease with financial constraints. External financing costs. When leverage rebalancing re quires security issuance, the costs of external financing aff ect the adjustment decision. Mo re costly financing creates a hurdle that may slow the speed of rebalancing. I study three asp ects of external financing. First, Korajczyk and Levy (2003) state that in tertemporal variation in the cost of capital may influence capital structure decisions. In this vein, I investigate the relationship between interest rate movements and the sp eed of adjustment. Second, information asymmetry creates frictions th at increase the difficulty of issuing securities (Myers and
39 Majluf 1984). Faulkender and Petersen (2006) fi nd that this affects th e source and cost of financing so I analyze a number of proxies for information asymmetry. Lastly, I examine access to the capital markets since Myers (1984) notes it is an important determinant of adjustment costs. Stock price movements. Stock price movements can move a firm toward or away from its optimal debt ratio. Welch (2004) contributed the concept of â€œinertiaâ€ whereby price movements shift the firmâ€™s leverage to ward the target without active management involvement. When such effortless price shifts bring a firm closer to its target debt ratio, they provide â€œcostlessâ€ adjustment and shoul d result in faster rebalancing. Price movements, however, can work against achieving the target. This w ould slow adjustment as would general price volatility, which proxi es for information asymmetry. Also, Graham and Harvey (2001) believe that price increases reduce the cost of external financing. For all of these reasons, my paper ex plores stock price movements. Adjustment Speed Factors: Adjustment Benefits Proximity to the optimal debt ratio is bene ficial when suboptimal leverage increases potential distress costs, forfeits potential tax shelters, or affects perceived manager quality.9 Further, loan and debt covenants also may increase target benefits by penalizing deviations from the target. I hypothesize that these adjustment benefits vary among firms. Differences affect the value of rebalancing and the speed at which a firm pursues its target leverage. ____________ 9 Clearly, these trade-offs occur in se parate regimes. The costs of distre ss are most serious for overleveraged firms and the tax story dominates when the firm is underleveraged. Future exploration of these asymmetries is expected to be informative.
40 Costs of (near) distress. Firms that are overleveraged relative to their optimal debt ratio clearly can reduce the probability of incurri ng the costs of financial distress (or near distress) by adjusting their leve rage. However, some firms fa ce a higher potential cost of distress such as those with growth opportuni ties (Opler and Titman 1994). As adjustment benefits increase with potential distress costs, I predict capital stru cture rebalancing will reflect the variation in distress costs. Value of tax shields. The tax benefits of debt increase the value of maintaining the target debt ratio, but like the costs of distress, the benefits are not uniform. First, only firms that exceed their non-debt tax shields can extract value from increased tax shield utilization. Second, the benef its increase with the tax rate. Faster adjustment speeds are predicted for firms with higher tax obligations. Manager benefits. Management may find deviati ons from the optimal capital structure costly in the pres ence of both potential distress and good corporate governance. Compensation or job security may increase if maintaining the debt ratio within certain bounds is associated with manager qualit y (Berger, Ofek, and Yermack 1997). I hypothesize that deviations are more costly for a manager in the presence of takeover threats as the potential for job loss is higher. Such incentives should increase the benefits of achieving the optimal leverage. The Econometrics of Capital Structure Adjustment Underpinning the debate among trade-off, p ecking order, market timing, and inertia theories of capital structure is the question of whether firms conti nually rebalance their leverage in pursuit of a target debt ratio. Trade-off theory is the only one that maintains
41 that each firm has an optimal leverage, but th e large cross-sectional variation in targeting behavior has provided fodder for skeptics of this view. Econometric issues, that make it difficult to estimate the speed of adjustment, onl y exacerbate the issue. In this section, I discuss these impediments and present an alternative estimation technique. Partial Adjustment and Capital Structure To evaluate capital structure adjustment , I start with the market debt ratio10, t i t i t i t iE D D MDR, , , , (1) where Di,t = Total Long-Term Debt i,t + Debt in Current Liabilities i,t Ei,t = Common Shares Outstanding i,t * Share Price i,t Each firmâ€™s target debt ratio, MDR* , must be estimated. Numerous existing papers determine this as a function of accepted leverage factors, X . As these factors and their expected relationship with the target debt ratio are discusse d in detail in Hovakimian, Opler, and Titman (2001), Frank and Goyal (2006), and Flannery and Rangan (2006), I review them only briefly. ] _ , _ & , _ & , _ , , _ , , _ [ Median Ind TA D R Dum D R TA F A L nTA TA D E P MB TA E BI T X (2) where EBIT_TA = (Income before extraordinary items + interest expense + total income taxes) / total assets MB = (Book liabilities plus market value of equity) / total assets DEP_TA = Depreciation and amortization / total assets LnTA = ln (total assets) deflated by th e consumer price index to 1983 dollars FA_TA = Net PPE / total assets R&D_Dum = 1 if Research and devel opment expense > 0, else zero R&D_TA = Research and developmen t expense / total assets Ind_Median = Median debt ratio for the firmâ€™s Fama and French (1997) industry. ____________ 10 The Appendix provides the Compustat data references . The forward-looking market debt ratio is selected over the book leverage following Frank and Goyal (2005) and Flannery and Rangan (2006). Like Flannery and Rangan (2006), our initial estimati ons are not sensitive to the use of bo ok debt ratios in place of market values.
42 Severe unobserved firm heterogeneity is documented by Flannery and Rangan (2006) and Lemmon, Roberts, and Zender (2006). Since these papers highlight the importance of including firm dummies, F, in optimal leverage estimation, each firmâ€™s target debt ratio, MDR*, is modeled as a function of both the observed, X, and unobserved firm attributes, F. i t i t iF X MDR , 1 ,* (3) In Equation 3, the coefficient vector beta and the unobserved firm effects are unknown. Following Flannery and Rangan (2006), I use a partial adjustment model to estimate the capital structure adjustment process give n the potential impact of capital market frictions. 1 , , 1 , , 1 ,) * ( t i t i t i t i t iMDR MDR MDR MDR (4) The adjustment measure () captures the actual change in leverage (MDRi,t+1 MDRi,t) relative to the firmâ€™s distance from its target (MDR* MDRi,t). I then substitute the target from Equation 3 for MDR*. 1 , , , , 1 ,) ( t i t i i t i t i t iMDR F X MDR MDR (5) Rearranging this, it is clear that regressing MDRi,t+1 on X and MDRi,t provides the basis to calculate the adjustment speed, , even with unknown and Fi. Equation 6 presents the model with the coefficients to be estimated in parentheses.11 1 , , , 1 ,) 1 ( ) ( t i i t i t i t iF MDR X MDR (6) Subtracting the coefficient on th e lagged dependent variable (MDRi,t) from one (that is, 1(1)) provides the adjustment speed, . ____________ 11 The firm fixed effect coefficients are not reported.
43 Dynamic Panel Estimation Capital structure adjustment speed estimati on requires careful c onsideration of the econometric issues at hand. A partial adjustme nt model is used given firmsâ€™ incomplete adjustment toward their optimal leverage. Th is requires dynamic panel data. In addition, the unobserved heterogeneity must be consid ered. Ignoring the unobserved firm-level heterogeneity imposes the incorrect assumpti on of zero correlation between the observed variables and the unobserved effect. This l eads to biased and inconsistent estimates (Wooldridge 2002). Including fixed effects in a dynamic panel, however, is problematic. Applying the within transformation removes the time-inva riant fixed effect but it also creates a correlation between the transformed lagged depe ndent variable and the transformed error term (Wooldridge 2002, Baltagi 2005). This in troduces a bias that is substantial with shorter panels (i.e., T is small and N is large) . The degree of inconsistency is a function of the panel length (Nickell 1981) on th e order of (1/T). This bias can be quite large even for panels with 30 observations (Judson and Owen 1999). Capital structure data, like most corporate finance data, is relatively short (small T) and therefore susceptible to this short panel bias. This section will show that this issue is critical to accurately estimating the speed of capital structure adjustment. To address this short panel bias, there ar e a number of choices. If an appropriate instrumental variable (IV) is available, it can be used to instrument for the lagged dependent variable. As I will discuss, tradit ional IV or a generalized method of moments (GMM) instrument are possible choices. Alte rnatively, the bias can be estimated and corrected directly.
44 The first option is to use a traditional inst rumental variables appr oach. Flannery and Rangan (2006) instrument using alternativ e firm characteristics and document an adjustment to the target debt ratio of over 30% a year. However, further analysis will show that those estimates still are sensitive to panel length. Alternatively, the Arellano Bond ge neralized method of moments (GMMAB, or difference GMM) is frequently used (Wooldridge 2002). GMMAB instruments for the first differences of endogenous (or predetermined) va riables with lags of their own levels but these may be weak instruments for the transf ormed lagged dependent variable especially with random-walk movement (Blundell and Bond 1998). The Blundell Bond GMM (GMMBB, or system GMM) employs additi onal moment conditions based on the differences (in addition to the levels) to increase the efficiency of the estimation.12 However, in the presence of second order au tocorrelation, GMM estimation of either form is not well-specified (Baltagi 2005). Second order au tocorrelation invalidat es the choice of lagged levels or differences as the instru ments are not exogenous. GMM specification tests indicate that second or der autocorrelation exists within the capital structure data, suggesting GMM is inappropriate. Recent advances in the econom etrics literature introduced a correction for the fixed effects bias found in short dynamic panels, th e corrected least squares dummy variable (LSDVC) approach for balanced panels (K iviet 1995). Traditional instruments and GMM try to remove the correlation between the tr ansformed lagged dependent variable and the transformed error term by using an instrument that is correlated with transformed lagged ____________ 12 While this could possibly create a â€œmany instrumentsâ€ problem, Hayakawa (2005) and Blundell and Bond (1998) find that GMMBB is less biased than GMMAB.
45 but not the error. LSDVC differs in that it first generates a consistent estimate13 of the short panel bias and then subt racts this bias from the LSDV estimate. Using Monte Carlo simulations, Judson and Owen (1999) documents that the LSDVC estimator dominates the various GMM alternatives for all panel lengths. However, until Bruno (2005), no correction existed for unbalan ced dynamic panel analysis. His contribution enables fixed effects estimation without the co ncern of panel length bias or the exclusion of unbalanced panel data. With this approach, I can test whether the rapid re balancing documented by Flannery and Rangan (2006) is driven by the pane l length bias or refl ects active leverage targeting. Data The sample consists of annual CRSP/Com pustat data from the years 1968 through 2004. Quarterly data would minimize the short panel bias but is inappropriate for studying capital structure. Leary a nd Roberts (2005) find that firm s, on average, adjust their leverage once a year, not on a quarterly basi s. Financial firms (SIC codes 6000-6999) and regulated utilities (SIC code s 4900-4999) are excluded from th e sample as regulation may color their choice of cap ital structure. I omit firm-years with missing data for long-term debt, debt in current liabilities, or any of the leverage factors, X , and require a non-negative book value of equity. Firm-years with data form at codes 4, 5, and 6 also are dropped as format codes 4 and 6 are not identified by Comp ustat and code 5 firms are Canadian. The log of total assets is the one variable that is not a ratio a nd it is deflated to 1983 dollars ____________ 13 To estimate the bias, LSDVC requires an initial matrix be specified. For the starting matrix, I use either the GMMBB initial estimator or a matrix based on earlier coef ficient estimates. Sensitivity analysis shows that the LSDVC estimation is robust to the initial matrix sel ection. LSDVC minimizes the panel length bias by correcting up to order NT2.
46 with the consumer price index from the Bur eau of Labor Statistics. To minimize the potential impact of outliers, all va riables are winsorized at the 1st and 99th percentiles. There are 111,371 firm years in my sample a nd the average panel length is 10.62 years. Summary statistics are presented in Table 3-1. Panel Length Bias To investigate the potential panel length bias, I estimate the speed of adjustment using just those firms with 30 or more years of continuous14 Compustat data. My only interest in this section is to investigate the st ability of the coefficient estimates. Therefore, I ignore the role of adjustment speed factors and interpreta tion bias created by limiting the sample to long-lived firms. Table 3-2 presents panel lengt h sensitivity analysis of th e three potential solutions for a dynamic panel with unobserved hetero geneity. I compare the fixed effects instrumental variable mode l (FE IV) presented in Fla nnery and Rangan (2006), both Arellano-Bond and Blundell-Bond generalized method of moments estimations (GMMAB, GMMBB), and Brunoâ€™s bias-corrected least squa res dummy variable approach (LSDVC) for unbalanced panels. Panel A presents the resu lts for the firms with 30 years of data. In Panels B and C, each firm has been subdivide d into multiple firm identifiers representing non-overlapping 10or 5-year panels for each or iginal firm. For example, Panel A labels the first 30 years of firm X data as one firm, Panel B creates three firms (years 1-10 of firm X, years 11-20 of firm X, etc.) and Panel C creates six firms (years 1-5 of firm X, etc.). The data is the same in each panel. The only difference is the firm fixed effect label. ____________ 14 I thank Christopher Baum of Boston College for this suggestion.
47 This allows us to determine whether shorter panels (the same data estimated over shorter horizons) affect the various estimation methodologies for th e partial adjustment model presented in Equation 6. Adjustment speed estimates ( ) are calculated from the coefficient estimate (1-) for MDRi,t and are presented in the la st row of the table. Pane l A shows a fair amount of variation for the different ec onometric techniques with adjustment speed estimates ranging from 6% to 22%. Panels B and C demons trate the magnitude of the potential bias associated with the panel length. With 10a nd 5-year panel lengths, the range of estimates increases dramatically. For the 10-year subgr oups in Panel B, the estimates range from 1544%. And, in Panel C, the adjustment speeds for the shortest panel length vary from 12% for GMMBB to 65% for the instrumental va riable fixed effects model. Bias-Corrected Least Squares Dummy Variable The GMMBB and LSDVC methodologies appear to be the least sensitive to panel length as they exhibit the least fluctuation am ong Panels A, B, and C. Either could be useful for estimating the speed of capital structure adjustment. Unfortunately, tests for the presence of second order au tocorrelation fa il for both GMMAB and GMMBB. This suggests that both are misspecified. GMMBB also requires continuous pa nels (no gaps within each group), that reduces the availabl e sample slightly. This may influence the results as firms near distress are more likely to be excluded. While GMMBB is potentially problematic, there is no perfect econometric technique with this data. LSDVC assumes the regressors are exogenous but the X variables are predetermined. Sensitivity testing reveals that defining X as exogenous versus predetermined has li ttle impact with GMM estimation so it appears that this issue is not influential.
48 The LSDVC approach has intuitive appeal. It approximates the small sample bias of the LSDV estimator and subtracts this estimat ed bias from the original LSDV estimator (Kiviet 1995). With GMM, one must have fa ith that the instrument s are appropriate and then an adjustment s till must be made for the panel le ngth bias (Windmeijer 2005). For comparison, I present coefficients estimated using both LSDVC and GMMBB. The adjustment speed cannot be estimated fo r the entire sample at once given certain computational constraints of these techni ques. For example, the unbalanced dynamic panel LSDVC estimator (explained in detail in Bruno 2005) involves matrix multiplication with an (NT x NT) matrix to demean the obs ervations. With 11,300 firms in the sample, this exceeds the bounds of practicality. Ther efore, I generate coefficient estimates and standard errors based on 100 replicat ions on a 5% random sample of firms.15 These estimates are presented in Ta ble 3-3 for both LSDVC and GMMBB. The average speed of adjustment is approximately 22% for LSDVC and 17% for GMMBB. This implies a relatively rapid adjustment toward the target even after correcting for the panel length bias associated with the inclusion of fixed effects. The slightly smaller GMMBB estimate is not surprising given the potential bias introduced by dropping non-continuou s panel data. As noted earlier, this omission may exclude firms near distress and this study shows that not all firms adjust uniformly. Heterogeneous Adjustments Costs and Benefits Next, I apply this methodology to the que stion of non-uniform adjustment. I hypothesize that differences in adjustment cost s and benefits generate cross-sectional and ____________ 15 As I am interested in the speed of adjustment, I draw 5% of firms, not 5% of the sample at random. This ensures that the sampling includes the full time-series for each firm selected.
49 intertemporal variation in levera ge rebalancing. The prior se ction introduced this theory and the adjustment speed factors. Z represents the factor proxies that are discussed in more detail in the next section. The baseline mode l presented in Equation 4 can be modified to allow the adjustment speed to vary with Z. The adjustment speed coefficient, , is replaced with a multi-factor coefficient that incorporates a base adjustment speed estimate, 0, and the adjustment speed factor estimate(s), j. j j NewZ 0 (7) With this revised adjustment speed coefficien t, the modified partial adjustment model is: 1 , , 1 , 0 , 1 ,) * )( ( t i t i t i j j t i t iMDR MDR Z MDR MDR (8) Estimating Equation 8 in one-stage requires a nonlinear estimation to constrain the 0, j, and (within the MDR*) estimates across the dependent variables. As there is no means of simultaneously constraining the coe fficients and correcting for the panel length bias, one-stage estimation exhibits severe pa nel length bias with the modified partial adjustment model. Therefore, I estimate Equation 8 with a two-stage process that is described below. The first stage predicts the target debt ratio (MDR*), while the second stage tests whether the adjustment speed factor s are significant leverage determinants. A two-stage methodology is common in the capital structure lit erature and authors frequently estimate the target, MDR*, based on the leverage factors, X, (Hovakimian, Opler, and Titman 2001, Fama and French 2002). However, this omits two relevant facts. First, it ignores the unobserved firm effects documented by that may influence the target debt ratio. Second, it assumes that firms ar e in equilibrium. The partial adjustment doesnâ€™t require this assumpti on and, thus, provides a more realistic framework in the presence of incomplete adjustment. Theref ore, I deviate from tradition and estimate the target debt ratio based on the partial adjustment framework.
50 The target is a function of observed and unobserved firm attributes as shown in Equation 3. With the assumption of unifo rm adjustment, the Flannery and Rangan estimation presented in E quation 6 doesnâ€™t require or to be specified. However, as the one-stage estimation of the modified partial ad justment model suffers from panel bias, the two-stage approach requires MDR* be calculated for each firm-year. This can be accomplished using Equation 6. First, rearra nging the equation shows that the target leverage can be isolated. t i t i t i i t iMDR MDR F X, , 1 , ,) ) 1 ( )( 1 ( (9) The predicted target can be calculated based on the fitted values for the dependent variable ( 1 i,tMDR), the lagged dependent variable (MDRi,t), and the estimated adjustment speed,. ) ) 1 ( )( 1 ( ) (, 1 , ,t i t i i t i iMDR MDR F X (10) I estimate Equation 6 using LSDVC and find the fitted values for the dependent variable ( 1 i,tMDR) and the estimated adjustment speed,. These values are substituted into Equation 10 and the pred icted target leverage is calculated. This target then is used in the second st age to estimate the impact of the adjustment speed factors, Z. With the target determined in a first stage, the difference between the target and the current de bt ratio is known. t i t i t iMDR MDR Deviation, 1 , .* (11) The actual change in leverage also is known. t i t i t iMDR MDR Change, 1 , . (12)
51 Substituting Equations 11 and 12 into Equati on 8 shows that only the adjustment speed coefficients, 0 and 1, are estimated. 1 , , 0 .) )( ( t i t i j j t iDeviation Z Change (13) There is no dynamic component to the sec ond stage, so LSDVC is unnecessary and I estimate an ordinary least square s regression correcting for the cl ustering of standard errors at the firm level. I provide two checks of robustness. First, the first-stage partial adjustment model is calculated using GMMBB instead of LSDVC. For comparison, I report the coefficient estimates based on both the LSDVC and GMMBB targets. The results are overwhelming similar. Further, as the first stage assumes a uniform adjustment sp eed, I also fit a tobit model including both firm dummy va riables and leverage factors, X, to calculate the target leverage. The tobit model circumvents the a ssumption of uniform adjustment and leads to qualitatively similar results, but assumes fi rms are at equilibrium. For brevity, the empirical section reports onl y the results for the estimations based on LSDVC and GMMBB targets. Table 3-4 shows the base specifi cation for each methodology. The average speed of adjustment is 21.3% using LSDV C-based targets and 18.5% with GMMBB. These estimates closely match the one-stage estimates presented in Table 3-3, indicating that the two-stage methodology does in troduce any distortion. Empirical Evidence on the Costs and Be nefits of Rebalancing Leverage Having reviewed the econometr ics of estimating a dynamic pa nel with fixed effects, I can examine the relationship between the cost s and benefits of adju stment and the speed of capital structure rebalancing. While both the target leverage and the adjustment speed
52 vary with firm characteristics, the determinants are not assumed to be necessarily the same for both the level and speed. Therefore, the ex isting financial literature guides the choice of adjustment speed factor proxies. Table 3-5 summarizes the proxi es, their relationship with each factor, and their e xpected impact on the speed of adjustment. The variable definitions are presented in the Appendix. From Table 3-5, it is clear that several pr oxies fall under more than one of the factors in adjustment costs and benefits. Frank and Goyal (2006) note, â€œOver time we have improved our understanding of th e factors that are empirica lly related to leverage. Interpreting the evidence has remained diffi cult. Many variables could reasonably be interpreted as representing different theories of capital structure.â€ This holds true as well for adjustment speed factors. As noted prev iously, I hypothesize that adjustment benefits increase with potential distre ss costs (such as high underinve stment costs) and lead to faster adjustment to the optimal debt ratio. Even so, growth oppor tunities also may be correlated with financial constraints and this could slow the rebalancing process. Since the expected impact on the adjustment speed is unclear, I present bot h possibilities. When available, multiple proxies are presented for each adjustment speed factor. I discuss the literature behind each proxy and the empirical ev idence supporting the role of adjustment costs and benefits in the capital structure adjustment process. Adjustment Costs and Financial Constraint Proxies The cost of capital structure adjustment reflects financial constraints, external financing costs, and stock pri ce movements. Tables 3-6 and 3-7 present the impact of financial constraints. Cash out flows of dividends, taxes, a nd investments convey decreased flexibility while profitability and asset sale s represent increased flexibility. Firms are
53 categorized as high or low in terms of dividend payment, marginal tax rate, net investment and profitability based on wh ether they are in the top (Q4) or bottom (Q1) quartile. Asset sales are defined with an indicator variable equaling unity if a sale exceeding 5% of total assets occurred. Fama and French (2002) sugge st that dividends reduce fi nancial flexibility and find dividend payers adjust toward their optimal debt ratio more slowly. I document similar results. Tables 3-6 and 3-7 show slower adjustment for fi rms that pay high dividends (Q4) and faster adjustment when th at constraint is removed (Q1). Taxes are another possible cas h constraint that would sl ow adjustment. However, for a profitable and underleveraged firm, highe r taxes increase the potential benefits of debt tax shield benefits. The marginal tax rates were generously supplied by John Graham (see Graham (1996) for a detailed explanation of this data). This rate incorporates factors such as non-debt tax shields. Neither hypothesis dominates when Z is limited to taxes. Firms with high tax obligations adjust toward their optimal debt ratio more quickly as predicted given the tax benefits. However, fit ting with the financial constraint story, the least constrained firms (Q1) also adjust more quickly than the average firm. Following Jensen (1986) and Lang, Ofek, and Stulz (1996), inve stments provide a third measure of cash constraints. Net inve stment is measured as capital expenditures minus depreciation relative to net sales as pe r Ahn, Denis, and Denis (2006). An above average rate of rebalancing fo r firms with low investments (Q1) indicates that financial flexibility permits more rapid rebalancing. And there is some evidence that constrained firms (with high investments) adjust more slowly.
54 For financial flexibility, pr ofitability seems like a reasona ble indicator on the surface. In fact, asset substitution concerns are base d on that very flexibility (Jensen and Meckling 1976). The theory literature, according to Harris and Ra viv (1991), is unclear on the relationship between profitability and leverage. In terms of the speed of adjustment, free cash flow removes or reduces the cost of external financ ing. Further, increased profitability may increase the value of debt tax shields or minimize asset substitution concerns if the firm is underleveraged. Bo th the increased flexibility and adjustment benefits imply a faster adjustment speed. The empirical results corroborate this. Firms with low profits (Q1) adjust more slowly. However, there is no evidence that very profitable firms adjust rapidly. Asset sales also might proxy for financia l flexibility. Leary and Roberts (2005) recognize the sale of plant, property and e quipment and the sale of investments as offsetting investment constraints. The last co lumn of Tables 3-6 and 3-7 show asset sales (defined as sales of PPE and investments that exceed 5% of total assets) are associated with increased adjustment speeds. These proxies indicate that fina ncial constraint and flexibility is a salient adjustment factor. However, the economic significance is relatively small. The average impact on the adjustment speed is approximately 0.2% for th e first five factors. Dividends appear to have the largest constraint and that stil l is less than a 1% difference in a year. Adjustment Costs and External Financing Proxies As the cost of external financing fluctuates with intertemporal variation, information asymmetry, and access to the capital markets, I expect these factors to affect the capital structure adjustment process. The impact of basic financing costs on adjustment speeds is
55 examined using the time-series variation in interest rates. Information asymmetry is another substantial hurdle to financing. My ers and Majluf (1984) a nd other advocates of pecking order theory believe it to be a la rge barrier â€“ causing a permanent preference for free cash and then debt. I proxy for informati on asymmetry with price volatility and firm age. Lastly, as Faulkender and Petersen (2006) note, access to the capital market is expected to be relevant. To evaluate access to the capital markets, I examine the importance of firm size, debt capacity, and public debt on adjustment speeds. As general macroeconomic conditions may affect external financing costs (Korajczyk and Levy 2003), I examine the time -series variation in interest rates by categorizing firm-year observations as having prior year interest rates that are in the top and bottom quartile relative to the sample period. There is no evidence that capital structure adjustment varies with in terest rates (Tables 3-8 and 3-9). Price volatility is considered a proxy for asymmetric information. The cost of adjustment should increase with informati on asymmetry and the speed of adjustment should decrease. For firms with at least eight ye ars of historical stock price data, I evaluate how volatility affects the speed of adjustment. Tables 3-8 and 3-9 show slightly slower adjustment follows hi gh price volatility (Q4). Firm age may mitigate information asymmetry. Younger firms, with less reputation, are more likely to engage in asset substitution (Diamond 1989). While this implies that younger firms have a higher cost of financing, asset substituti on concerns lead to smaller and more frequent financing act ivity. This expected return to the capital markets reduces the level of information asymmetry and the co st of external financ ing (Welch 1996). Alti (2003) also notes that younger firm s are generally are higher grow th firms. Therefore, it is
56 unclear whether younger firms have more info rmation asymmetry and if they adjust differently. To evaluate this, young firms are defined as those w ith four years or less in the public markets following Welch (1996) and Leary and Roberts (2005). The evidence implies firm age does mitigate information asymmetry but the economic significance is small. Larger firms frequently have lower info rmation asymmetry, which would imply a lower cost of financing and faster adjustment . But they generally have less cash flow volatility, which reduces the pot ential costs of dist ress (Flannery and Rangan 2006) and the expected speed of adjustment. The em pirical evidence supports the information asymmetry argument. Smaller firms adjust more slowly although, like firm age, the magnitude is small. Debt capacity is a recent addition to the capital structure disc ussion. Lemmon and Zender (2004) believe recognizing debt capacity is an important f actor in understanding the pecking order eviden ce. While Leary and Roberts (2005) find pecking order fails even with the inclusion of this variable, it is a r easonable measure of access to external capital. Therefore, debt capacity (debt relative to fi xed assets) is used to measure external financing costs. High debt rela tive to collateral leads to fast er adjustment â€“ perhaps due to binding loan covenants. Both Ta bles 3-8 and 3-9 show that fi rms with a low debt relative to collateral (Q1) adjust ~1.5% more slowly each year . These findings support the distress story since low relative levera ge firms have limited potential for distress and less need for rapid adjustment. And it also correspond with the conclusion of Leary and Roberts (2005) that debt capacity limitations do not determ ine capital structure adjustment.
57 For the final measure of access to the capital markets, I look at whether a firm has public debt. Faulkender and Pe tersen (2006) note that the ab ility to issue public debt reduces the cost of external financing and is an important leverage determinant. Firms with public debt face lower financing costs that would increase the speed of adjustment. Conversely, private debt may ha ve more covenants or more active monitoring than public debt, adding a cost to appro aching distress. This increases the adjustment benefits and implies faster adjustment for private debt. This empirical eviden ce supports the distress story. There is strong evidence that firm with public debt adjust more slowly than those with only private debt. Adjustment Costs and Stock Price Movements Stock price movements can affect the cost of rebalancing capital structure. Not only do price movements toward the target (that is, price increases when overleveraged or price decreases when underleveraged) provide cos tless adjustment, price increases may reduce the cost of external financing (Graham and Ha rvey 2001). In the current framework, these two price movements should increase the adjust ment speed. I compare stock price changes that move a firm toward its target against firms with no price changes as well as price increases versus price decreases. Stock price changes that exceed five percent are included in the analysis. Tables 3-10 and 3-11 presen t these results. Faster adjustment follows stock price movements toward the target debt ratio. The evidence is mixed on the relationship between price shif ts and speed. Small positive coefficients are documented for both price increase and decreases. Large increases provide flexibility while large decreases, most likely, relate to the costs of distress.
58 Adjustment Benefits Next, I examine the benefits of adjustme nt. The costs of financial distress, managerial benefits, and taxes a ll affect the value of maintaining the optimal debt ratio. To capture the potential costs of distress, I exam ine relative leverage, fixed assets, and asset volatility in addition to the already mentione d profitability, price decreases, growth, and private debt. For managerial benefits, job secu rity is an important feedback mechanism. Therefore, I examine the threat of takeover. Relative leverage affects the probability of distress. Overleveraged firms, by definition, can exceed the bounds where debt has a positive impact on firm value. In this realm, the costs of potential di stress mount. Research by Ho vakimian, Opler, and Titman (2001) document faster adjustment speeds for overleveraged firms us ing a sample of new issues and repurchases. As Fischer et al. (1989) conclude s there is region of no capital structure rebalancing and Ju et al. (2005) defines this range as a 10% deviation from the optimal debt ratio, I examine relative levera ge defined as 10% above or below the LSDVC calculated target debt ratio. Table 3-12 and 3-13 show that adjustment speeds for overleveraged firms are ~22% versus 20% fo r average firms. However, underleveraged firms also adjust more quickly than the averag e, but not as quickly as overleveraged firms. Meanwhile, collateral is gene rally thought to reduce the co sts of distress (Harris and Raviv 1990). As the costs of deviating from the target should fall with distress costs, I expect a slower speed of adjustment for fi rms with higher tangibility. Alternatively, tangibility lessens the asset substitution problem (Jensen and Meckling 1976). This implies faster adjustment. Slower rebalanc ing occurs in firms with high fixed assets
59 (Tables 3-12 and 3-13). Thus, the reduction in costs of distress appears to be more important that the information asymmetry. Growth is frequently consid ered an important leverage factor. Growth firms may hold less debt due to asset substituti on fears (Jensen and Meckling 1976) or underinvestment (Myers 1984, Stulz 1990). Grow th also is expected to affect the adjustment speed. Growth measures are ofte n associated with highe r distress costs (Opler and Titman 1994) and this could lead to a fast er adjustment process. I examine growth using the market to book ratio a nd report these results in the third column of Tables 3-12 and 3-13. High growth firms (Q4) exhibit slower adjustment (~ 1.0% different than average firms each year). This corresponds to the financial constraint story not the costs of distress. Asset volatility also increases the potential likelihood of di stress and, therefore, is a possible proxy for this adjustment speed fact or. Firms with high income volatility are more likely to encounter cas h flow problems and experien ce distress. Therefore, management should adjust leverage more quickly given the expected co sts of distress and I expect higher income volatility to reduce the speed of adjustment. In my paper, annual volatility of income is measured over the prior 8-12 years. While low volatility firms have slower adjustment, the results are economically marginally. As noted earlier, the empirical evidence docum ented with debt and price decreases is consistent with distress costs factoring actively into the capital structure adjustment process. Coupling this with the relative leve rage and fixed assets results in Tables 3-12 and 3-13 indicate that, to the ex tent my proxies capture the co sts of distress, these costs appear to be a significant determinant of capital structure adjustment speeds.
60 Next, I examine managerial benefits. Th e benefits of greater efficiency and corporate discipline may vary for managers. Since the threat of ta keover and the ensuing possibility of job loss may encourage managers to stay closer to the optimal debt level (Stulz 1990), I use the take over wave of 1983 to 1989 (similar to Holmstrom and Kalpan 2001) as a period of heightened takeover risk. Faster adjustment is expected during this period of higher deviation costs. The last colu mn of Tables 3-12 and 3-13 suggest that this was not the case, or the proxy is inappropriate. If anything, firms during this period adjust their leverage more slowly. Lastly, the traditional finance theory, as well as the substantial work by Graham, would indicate a strong role for debt tax shields in th e speed of capital structure adjustment. As noted in the discussion of Ta bles 3-6 and 3-7, high marginal tax rates are correlated with more rapid rebalancing. This suggests that taxes are a significant determinant of the capital stru cture adjustment process. Dominant Adjustment Factors My paper has identified and provided em pirical evidence on a number of capital structure adjustment speed f actors. Given the variety of adjustment speed magnitudes documented with the various proxies, it seem s worthwhile to explore whether certain factors dominate others and are most important to rebalancing. Table 3-14 presents a cumulative analysis of the proxies presented in my paper. It should be noted that volatility factors are excluded as they are available only for firms w ith at least eight years of historical data. On the whole, the Table 3-14 results support the conclusions drawn from the individual factor anal ysis. Financial constraint proxies are significant determinants.
61 Dividends, a constraint proxy, decrease the speed of adjustment while profitability, representing financial flexibility, increases the pace of rebalancing. Fi rm size is significant proxy within the external financing cost proxies and stock price movements remain relevant with faster adjust ment following both price increa ses and movements toward to the target. The benefits of adjustment also are important in this cumulative analysis. The costs of distress continue to be an adjustme nt speed determinant w ith overleveraged firms rebalancing rapidly. The tax be nefits of debt are another si gnificant determinant. Firms with high marginal tax rate adjust more quickly. Conclusion My study presents evidence on the speed of capital structure adju stment. First, I address the econometric issues involved in documenting an av erage rebalancing speed. Using the newly developed bias-corrected least squares dummy variable approach, I document that firms move towards their target debt ratio relatively quickly, adjusting approximately 22% each year on average. Similar results are found with the Blundell Bond GMM methodology. This su pports the growing literature that has documented rapid adjustment to an optimal leverage and corrobor ates the revised tradeoff theory of capital structure. I also present a theory of capital structure adjustme nt, hypothesizing that firm attributes affect more than the optimal level of debt. They also affect the cost of adjusting the debt ratio and the value of reaching and/or maintaining that target. The process of leverage adjustment is a tradeoff between th e benefits of the targ et and the costs of rebalancing. Employing various proxies for the costs and benefits of adjustment, I examine the impact of these factors on the speed of adjustment. There appears to be strong
62 empirical support for the hypothesis. Clearly, the interpretation is only as reliable as the proxies themselves but financial constraints, external financing costs, the costs of distress, and the tax benefits of debt a ffect the speed of adjustment in a manner consistent with my hypothesis. Managerial benef its appear to have less of an impact on the adjustment process. That could reflect th is factor being of secondary im portance or it could indicate a weaker proxy. Overall, there is ample evid ence that adjustment costs and benefits are significant determinants of capital structure.
63 Table 3-1. Summary Statistics. Variable Obs. Mean Std. Dev. Min. Max. MDRt+1 111,371 0.287 0.248 0.000 0.999 MDRt 111,371 0.275 0.243 0.000 0.996 EBIT_TA 111,371 0.052 0.258 -17.846 10.657 MB 111,371 1.552 2.173 0.010 102.735 DEP_TA 111,371 0.046 0.045 0.000 3.265 ln(TA) 111,371 18.307 2.039 13.346 25.440 FA_TA 111,371 0.323 0.218 0.000 1.000 R&D_Dummy 111,371 0.461 0.499 0.000 1.000 R&D_TA 111,371 0.033 0.091 0.000 4.291 Ind_Median 111,371 0.230 0.137 0.000 0.806 Panel Length 111,371 10.620 8.000 2 38 The sample consists of annual CRSP/Compusta t data from the years 1968 through 2004. Excluded from the sample are financial fi rms, regulated utilities, and firm-years with missing data for total assets, long-term debt, or debt in current liabilities, and negative book value of equity as we ll as those without a lag market debt ratio. Firm-years with data format code s 4, 5, and 6 are dropped. The log of total assets is deflated to 1983 dollars. All variables are winsorized at the 1st and 99th percentiles.
64Table 3-2. Panel Length Sensitivity. Full 30+ Years Ten Year Subgroups Five Year Subgroups FE IV AB BB LSDVC FE IV AB BB LSDVC FE IV AB BB LSDVC MDR 0.784 0.939 0.846 0.819 0.561 0.850 0.825 0.773 0.348 0.881 0.820 0.745 (0.000) (0.000) (0.000) 0.000 (0.000) (0.000) (0.000) 0.000 (0.000) (0.000) (0.000) (0.000) EBIT_TA -0.020 0.334 -0.113 -0.033 -0.070 0. 315 -0.090 -0.028 -0.073 0.007 -0.089 0.049 (0.082) (0.000) (0.003) (0.761) (0.000) (0.000) (0.047) (0.847) (0.000) (0.000) (0.173) (0.740) MB -0.001 0.027 -0.004 -0.002 -0.001 0.025 0.000 0.003 0.000 0.400 0.003 0.009 (0.270) (0.000) (0.078) (0.882) (0.580) (0.000) (0.929) (0.817) (0.870) (0.218) (0.722) (0.637) DEP_TA -0.480 0.058 -2.243 -0.489 -0.497 0.02 4 -2.248 -0.521 -0.497 -0.239 -1.818 -0.460 (0.000) (0.750) (0.000) (0.400) (0.000) (0.910) (0.001) (0.497) (0.000) (0.001) (0.082) (0.547) ln(TA) 0.016 -0.019 -0.001 0.015 0.046 -0.034 0.001 0.047 0.067 0.320 0.026 0.049 (0.000) (0.011) (0.761) (0.354) (0.000) (0.007) (0.854) (0.158) (0.000) (0.502) (0.174) (0.241) FA_TA 0.055 -0.010 0.173 0.048 0.093 -0.012 0.416 0.087 0.128 0.027 -0.073 0.091 (0.000) (0.750) (0.134) (0.580) (0.000) (0.730) (0.035) (0.469) (0.000) (0.412) (0.839) (0.568) R&D_Dummy 0.000 0.001 0.033 0.001 -0.002 0.005 0.139 0.002 -0.004 -0.027 0.282 0.006 (0.930) (0.850) (0.511) (0.966) (0.620) (0.530) (0.085) (0.950) (0.370) (0.000) (0.034) (0.902) R&D_TA 0.023 0.390 0.084 0.042 0.017 0.355 -0.360 0.051 0.093 0.363 -0.641 0.256 (0.630) (0.110) (0.808) (0.921) (0.770) (0.120) (0.581) (0.936) (0.240) (0.001) (0.590) (0.737) Ind_Median -0.015 -0.227 -0.385 -0.003 0.010 -0 .240 -0.157 -0.002 -0. 038 -0.052 0.437 -0.055 (0.260) (0.000) (0.057) (0.985) (0.540) (0.000) (0.684) (0.992) (0.052) (0.247) (0.489) (0.809) Constant -0.265 -0.001 0.139 -0.802 0.000 -0.135 -1.161 0.001 -0.682 (0.000) (0.140) (0.302) (0.000) (0.620) (0.490) (0.000) (0.276) (0.168) Year Dummies Y Y Y Y Y Y Y Y Y Y Y Y # Obs 19,140 18,901 19,140 18,502 19,140 16,349 19,140 17,226 19,140 11,484 19,140 15,312 # Groups 638 638 638 638 1,914 1,914 1,914 1,914 3,828 3,828 3,828 3,828 Sargan/Hansen 0.340 0.108 0.000 1.000 0.000 1.000 2nd Order Autocorr. 0.000 0.003 0.000 0.000 0.001 0.001 Adj Speed ( ) 22% 6% 15% 18% 44% 15% 18% 23% 65% 12% 18% 26% The panel length bias is compared for firms surviving at least 30 years. The first set of columns includes the full 30+ year p anel for each firm. The second (third) sets divi de each firm into ten (five) year su bgroups. FE IV presents fixed effects instrumenting for the lagged depende nt variable, AB presents Arellano -Bond GMM, BB presents the Blundell-Bond GMM, and LSDVC presents the bias-c orrected LSDV approach. P values are listed in parentheses. 1 , , , 1 ,) 1 ( ) ( t i t i t i t iMDR X MDR
65 Table 3-3. Baseline Adjustment Speed Estimation. LSDVC GMMBB MDRt 0.782 0.786 0.835 0.825 (0.000) (0.000) (0.000) (0.000) EBIT_TAt (0.013) -0.018 0.006 0.001 (0.000) (0.000) (0.000) (0.771) MBt 0.001 0.000 0.008 0.004 (0.000) (0.927) (0.000) (0.000) DEP_TAt (0.230) -0.190 -0.605 -0.575 (0.000) (0.000) (0.000) (0.000) ln(TA)t 0.013 0.017 0.009 0.006 (0.000) (0.000) (0.000) (0.000) FA_TAt 0.043 0.035 0.162 0.122 (0.000) (0.000) (0.000) (0.000) R&D_Dummyt (0.001) 0.001 -0.001 0.067 (0.088) (0.261) (0.796) (0.000) R&D_TAt (0.025) -0.020 -0.046 0.011 (0.000) (0.000) (0.000) (0.094) Ind_Mediant 0.038 0.025 -0.056 -0.054 (0.000) (0.000) (0.000) (0.046) Constant -0.133 -0.175 (0.000) (0.000) Year Dummies N Y N Y Avg # Obs 6,609 6,609 6,246 6,246 Avg # Groups 713 713 713 713 Adj Speed ( ) 22% 21% 17% 17% The average speed of adjustment is estimated using LSDVC and GMMBB. The later is limited to continuous panels. Given co mputational constraints, coefficient estimates are the average based on 1 00 replications, drawing a random 5% sample of the data each time. Coeffi cient estimates are presented with and without year dummies. P values, base d on the variance of the 100 replications, are in parentheses. 1 , , , 1 ,) 1 ( ) ( t i t i t i t iMDR X MDR
66 Table 3-4. Two-Stage Estimates of the Speed of Adjustment. LSDVC GMMBB 0 0.213 0.185 (0.000) (0.000) Constant 0.032 0.033 (0.000) (0.000) # Obs 111,371 106,314 R2 0.934 0.934 The speed of adjustment is estimated with a two-stage methodology. The target, MDR*, is generated in the first stage using either LSDVC or GMMBB. Deviation is the difference between MDR* and the current market debt ratio. Change is the actual change in leverage. Robus t p values in parentheses. 1 , , . t i t i t iDeviation Change
67Table 3-5: Adjustment Speed Factor Proxies. Adjustment Costs Adjustment Benefits Proxy Financial Constraint Financing Costs Stock Price Movements Expected Speed Distress Costs Manager Benefits Debt Tax Shields Expected Speed Expected Net Impact Overleveraged + NA NA Slower + + NA Faster ? Dividends + NA NA Slower NA NA NA NA Slower Marginal Tax Rate + NA NA Slower NA NA + Faster ? Investments + NA NA Slower + NA NA Faster ? Profitability NA Faster NA + ? ? Asset Sales NA NA Faster NA NA NA NA Faster M/B NA NA NA Slower + NA NA Faster ? Fixed Assets NA NA Faster NA NA Slower ? Income Volatility NA NA NA NA + NA NA Faster Faster Takeover Threat NA NA NA NA NA + NA Faster Faster Price Increases NA Faster NA NA Slower ? Price Movement to Target NA Faster NA NA NA NA Faster Interest Rate NA + NA Slower NA NA NA NA Slower Firm Size NA NA Faster NA NA Slower ? Price Volatility NA + NA Slower + NA NA Faster ? Debt Capacity (Debt/FA) NA + NA Slower + NA NA Faster ? Young Firms NA NA Faster NA NA NA NA Faster Rated Debt NA NA Faster NA NA NA NA Faster Adjustment cost and benefit fact ors are listed. As the proxies may influence more than one factor, this table summarizes their primary relationships and expected impact on the speed of capital structure adju stment. A positive sign (+) indicates a positive correlation between the proxy and the factor. On the cost side, a positive implies higher adjustment costs and a slower adjustment. On the benefit side, a positive implies in creased benefit to reaching the target leverage and therefore a faster adjustment.
68 Table 3-6. LSDVC: Adjustment Cost s and Financial Constraints. Dividends Marg Tax Rate Investments Profits Asset Sale 0 0.212 0.212 0.212 0.214 0.213 (0.000) (0.000) (0.000) (0.000) (0.000) Quartile 1 0.002 0.003 0.002 -0.002 (0.000) (0.000) (0.000) (0.000) Quartile 4 -0.007 0.002 -0.001 0.000 (0.000) (0.000) (0.140) (0.540) Asset Sale 0.002 (0.038) Constant 0.032 0.032 0.032 0.032 0.032 (0.000) (0.000) (0.000) (0.000) (0.000) R2 0.934 0.934 0.934 0.934 0.934 # Obs 111,371 111,371 111,371 111,371 111,371 The speed of adjustment is estimated with a two-stage methodology. The target, MDR*, is generated in the first stage using LSDVC. Deviation is the difference between MDR* and the current market debt ratio. Change is the actual change in leverage. The adjustment sp eed factors, which vary by column, are Z. This table examines firms in th e top and bottom quartile for dividends, Graham's marginal tax rate, investments, and profits. It also presents the impact of asset sales on adjustment sp eed. Robust p values in parentheses. 1 , , 0 .) )( ( t i t i j j t iDeviation Z Change
69 Table 3-7. GMMBB: Adjustment Costs and Financial Constraints Dividends Marg Tax Rate Investments Profits Asset Sale 0 0.185 0.185 0.185 0.186 0.185 (0.000) (0.000) (0.000) (0.000) (0.000) Quartile 1 0.002 0.003 0.002 -0.002 (0.001) (0.000) (0.000) (0.000) Quartile 4 -0.005 0.002 -0.001 -0.001 (0.000) (0.000) (0.087) (0.200) Asset Sale 0.002 (0.086) Constant 0.032 0.033 0.033 0.033 0.033 (0.000) (0.000) (0.000) (0.000) (0.000) R2 0.934 0.934 0.934 0.934 0.934 # Obs 106,314 106,314 106,314 106,314 106,314 The speed of adjustment is estimated with a two-stage methodology. The target, MDR*, is generated in the first stage using GMMBB. Deviation is the difference between MDR* and the current market debt ratio. Change is the actual change in leverage. The adjustment sp eed factors, which vary by column, are Z. This table examines firms in th e top and bottom quartile for dividends, Graham's marginal tax rate, investments, and profits. It also presents the impact of asset sales on adjustment sp eed. Robust p values in parentheses. 1 , , 0 .) )( ( t i t i j j t iDeviation Z Change
70 Table 3-8. LSDVC: Adjustment Costs and External Financing. Interest Rate Price Vol. Age Size Debt Capacity Public Debt 0 0.213 0.214 0.214 0.213 0.214 0.213 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Quartile 1 -0.001 0.001 -0.001 -0.017 (0.120) (0.200) (0.025) (0.000) Quartile 4 0.000 -0.004 0.000 0.003 (0.260) (0.000) (0.290) (0.000) Young -0.002 (0.000) Public Debt -0.008 (0.000) Constant 0.032 0.025 0.032 0.032 0.032 0.032 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) R2 0.934 0.960 0.934 0.934 0.934 0.934 # Obs 111,371 60,568 111,371 111,371 111,371 111,371 The speed of adjustment is estimated with a two-stage methodology. The target, MDR*, is generated in the first stage using LSDVC. Deviation is the difference between MDR* and the current market debt ratio. Change is the actual change in leverage. The adjustment sp eed factors, which vary by column, are Z. This table examines the top and bottom quartile of interest rate and firms in the top and bottom quartile for price volatili ty (for firms with at least 8 years of data), fixed assets, size, and debt capaci ty. It also presents the impact firm age and public debt dummy variables. Robust p values in parentheses. 1 , , 0 .) )( ( t i t i j j t iDeviation Z Change
71 Table 3-9. GMMBB: Adjustment Costs and External Financing. Interest Rate Price Vol. Age Size Debt Capacity Public Debt 0 0.186 0.186 0.186 0.186 0.186 0.186 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Quartile 1 0.000 0.001 -0.001 -0.014 (0.310) (0.087) (0.064) (0.000) Quartile 4 0.000 -0.004 0.000 0.003 (0.350) (0.000) (0.660) (0.000) Young -0.001 (0.001) Public Debt -0.007 (0.000) Constant 0.033 0.025 0.033 0.033 0.033 0.033 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) R2 0.934 0.961 0.934 0.934 0.935 0.934 # Obs 106,314 57,703 106,314 106,314 106,314 106,314 The speed of adjustment is estimated with a two-stage methodology. The target, MDR*, is generated in the first stage using GMMBB. Deviation is the difference between MDR* and the current market debt ratio. Change is the actual change in leverage. The adjustment sp eed factors, which vary by column, are Z. This table examines the top and bottom quartile of interest rate and firms in the top and bottom quartile for price volatili ty (for firms with at least 8 years of data), fixed assets, size, and debt capaci ty. It also presents the impact firm age and public debt dummy variables. Robust p values in parentheses. 1 , , 0 .) )( ( t i t i j j t iDeviation Z Change
72 Table 3-10. LSDVC: Ad justment Costs and Price Movements. 5% Price Change Price Move to Target 0 0.211 0.212 (0.000) (0.000) Increase 0.002 (0.002) Decrease 0.003 (0.000) To Target 0.003 (0.000) No Change 0.000 (0.029) Constant 0.032 0.032 (0.000) (0.000) R2 0.934 0.934 # Obs 111,371 111,371 The speed of adjustment is estimated with a two-stage methodology. The target, MDR*, is generated in the first stage using LSDVC. Deviation is the difference between MDR* and the current market debt ratio. Change is the actual change in leverage. The adjustment speed factors, which vary by column, are Z. The first column examines firms that experienced a 5% change in price in the prior period. The second column ev aluates the adjustment speed following 5% price movements toward the ta rget (ie, 5% price decrease if underleveraged). Robust p va lues in parentheses. 1 , , 0 .) )( ( t i t i j j t iDeviation Z Change
73 Table 3-11. GMMBB: Adjustment Costs and Price Movements. 5% Price Change Price Move to Target 0 0.184 0.185 (0.000) (0.000) Increase 0.002 (0.001) Decrease 0.002 (0.000) To Target 0.002 (0.000) No Change 0.000 (0.009) Constant 0.033 0.032 (0.000) (0.000) R2 0.934 0.934 # Obs 106,314 106,314 The speed of adjustment is estimated with a two-stage methodology. The target, MDR*, is generated in the first stage using GMMBB. Deviation is the difference between MDR* and the current market debt ratio. Change is the actual change in leverage. The adjustment speed factors, which vary by column, are Z. The first column examines firms that experienced a 5% change in price in the prior period. The second column ev aluates the adjustment speed following 5% price movements toward the ta rget (ie, 5% price decrease if underleveraged). Robust p va lues in parentheses. 1 , , 0 .) )( ( t i t i j j t iDeviation Z Change
74 Table 3-12. LSDVC: Ad justment Benefits. Rel. Leverage Fixed Assets M/B Asset Vol. Takeover 0 0.200 0.213 0.214 0.214 0.213 (0.000) (0.000) (0.000) (0.000) (0.000) Overleveraged (10%) 0.021 (0.000) Underleveraged (10%) 0.014 (0.000) Quartile 1 0.000 0.000 -0.001 (0.760) (0.340) (0.001) Quartile 4 -0.002 -0.010 -0.001 (0.000) (0.000) (0.140) Threat -0.002 (0.000) Constant 0.032 0.032 0.032 0.025 0.032 (0.000) (0.000) (0.000) (0.000) (0.000) R2 0.938 0.934 0.934 0.960 0.934 # Obs 111,371 111,371 111,371 60,568 111,371 The speed of adjustment is estimated with a two-stage methodology. The target, MDR*, is generated in the first stage using LSDVC. Deviation is the difference between MDR* and the current market debt ratio. Change is the actual change in leverage. The adjustment speed factors, which vary by column, are Z. This table examines firms that are ove r or underleveraged by 10% or in the top and bottom quartile for market-to-book, and asset volatility (for firms with at least 8 years of data). The last column examines the effect of a takeover threat (defined as 1983 to 1989). Robust p values in parentheses. 1 , , 0 .) )( ( t i t i j j t iDeviation Z Change
75 Table 3-13. GMMBB: Adjustment Benefits. Rel. Leverage Fixed Assets M/B Asset Vol. Takeover 0 0.172 0.186 0.187 0.187 0.186 (0.000) (0.000) (0.000) (0.000) (0.000) Overleveraged (10%) 0.021 (0.000) Underleveraged (10%) 0.014 (0.000) Quartile 1 0.000 0.000 -0.001 (0.290) (0.380) (0.001) Quartile 4 -0.002 -0.009 -0.001 (0.000) (0.000) (0.009) Threat -0.002 (0.000) Constant 0.033 0.033 0.033 0.025 0.033 (0.000) (0.000) (0.000) (0.000) (0.000) R2 0.935 0.934 0.934 0.961 0.934 # Obs 106,314 106,314 106,314 57,703 106,314 The speed of adjustment is estimated with a two-stage methodology. The target, MDR*, is generated in the first stage using GMMBB. Deviation is the difference between MDR* and the current market debt ratio. Change is the actual change in leverage. The adjustment speed factors, which vary by column, are Z. This table examines firms that are over or underleveraged by 10% or in the top and bottom quartile for market-to-book, and asset volatility (for firms with at least 8 years of data). The last column examines the effect of a takeover threat (defined as 1983 to 1989). Robust p values in parentheses. 1 , , 0 .) )( ( t i t i j j t iDeviation Z Change
76 Table 3-14. Dominant Factor Analysis. LSDVC GMMBB 0 0.209 0.182 (0.000) (0.000) High Dividends -0.009 -0.007 (0.000) (0.000) High Profits 0.001 0.001 (0.046) (0.091) High Investment 0.001 0.001 (0.210) (0.160) Asset Sale 0.001 0.001 (0.260) (0.410) High Size (TA) 0.002 0.002 (0.001) (0.002) High IR 0.000 0.000 (0.820) (0.780) Young 0.000 0.000 (0.590) (0.710) High Debt Capacity 0.006 0.005 (0.000) (0.000) Public Debt -0.006 -0.005 (0.000) (0.000) Price Increase 0.005 0.004 (0.000) (0.000) Toward Target 0.004 0.003 (0.000) (0.000) Overlev 10% 0.012 0.010 (0.000) (0.000) High M/B -0.010 -0.009 (0.000) (0.000) High FA 0.001 0.000 (0.210) (0.440) High Marg Tax 0.004 0.003 (0.000) (0.000) Threat -0.002 -0.002 (0.000) (0.000) Constant 0.032 0.033 (0.000) (0.000) # Obs 111,371 106,314 R2 0.935 0.935 The speed of adjustment is estimated with a two-stage methodology. The target, MDR*, is generated in the first stag e using either LSDVC or GMMBB. Deviation is the difference between MDR* and the current market debt ratio. Change is
77 the actual change in leverage. The ad justment speed factors are Z. Robust p values in parentheses. Firms in the top quartile are categorized as "High". 1 , , 0 .) )( ( t i t i j j t iDeviation Z Change
78 CHAPTER 4 CONCLUSION Risk management is a multi-faceted activity. Firms, and managers, increase firm value by minimizing risks that do not offe r any upside potential or that create high potential distress costs. My dissertation examines acquisi tions that provide operational hedging and the capital structure adjustment process as two such avenues for valuegenerating risk management. Both chap ters examine trade-offs made in risk management decisions. First, I present a theory of optimal he dging. Corporate finance decisions that reduce the overall volatility of the firm provide operational hedging. In this chapter, I examine whether firms recognize this contribut ion and rebalance their use of derivatives hedging. Acquisitions that provide operational hedging are used to empirically examine this relationship between operational and fina ncial hedging. I document a substitution between the risk management alternatives . This supports the hypothesis that firms manage risk across a variety of available tools. Next, a model of optimal capital structure adjustment is introd uced. I hypothesize that a firmâ€™s leverage choice reflects both its ta rget debt ratio and the costs and benefits of adjustment. This chapter discusses the complex econometric issues involved with dynamic panel estimation and then shows empi rical evidence on the capital structure adjustment process. There is considerable ev idence that adjustment costs and benefits are importance determinants of the capital structure adjustment speed.
79 APPENDIX VARIABLE DEFINITIONS CRSP/Compustat Industrial Annual Data references are in parentheses. D Long-term debt (9) + debt in current liab ilities (34) E Common shares outstanding (25) * share price (199) EBIT_TA Income before extraordinary it ems (18) + interest expense (15) + total income taxes (16) / total assets (6) MB (D + E + preferred stock (10)/ total assets DEP_TA Depreciation and amortization (14)/ total assets LnTA ln(total assets) deflated by th e consumer price index to 1983 dollars FA_TA Net PPE (8)/ total assets R&D_Dum 1 if research and developm ent expense (46) > 0, else zero Income Operating income before depreciation (13) Asset Sales Sales of PPE and sales of investments (213) Overleveraged 10% above LSDVC calculated target debt ratio Underleveraged 10% below LSDVC calculated target debt ratio Profits Net sales (12) Investments Capital expenditures (1 28) â€“ depreciation / net sales Payer 1 if dividends (26)>0, else zero Debt Capacity 1 if (D/ Net PPE) >1, else zero Tangibility Net PPE
80 LIST OF REFERENCES Aggarwal, Rajesh K., and Andrew A. Samw ick, 2003, â€œWhy Do Managers Diversify Their Firms? Agency Reconsidered,â€ Journal of Finance 58, 71-118. Ahn, Seoungpil, David J. Denis, and Diane K. Denis, 2006, â€œLeverage and Investment in Diversified Firms,â€ Journal of Financial Economics 79, 317-337. Allayannis, George, Jane Ihrig, and Jame s P. Weston, 2001, â€œExchange-Rate Hedging: Financial versus Operational Strategies,â€ American Economic Review 91, 391-395. Allen, Linda, and Anthony Saunders, 1992, â€œBank Window Dressing: Theory and Evidence,â€ Journal of Banking and Finance 16, 585-623. Amihud, Yakov, and Baruch Lev, 1981, â€œR isk Reduction as a Managerial Motive for Conglomerate Mergers,â€ Bell Journal of Economics 12, 605-617. Baker, Malcolm and Jeffrey Wurgler, 2002, â€œMarket Timing and Capital Structure,â€ Journal of Finance 57, 2-32. Baltagi, Badi H., 2005, Econometric Analysis of Panel Data, John Wiley and Sons, West Sussex, England. Benston, George J., William C. Hunter, and Larry D. Wall, 1995, â€œMotivations for Bank Mergers and Acquisitions: Enhancing th e Deposit Insurance Put Option versus Earnings Diversification,â€ Journal of Money, Credit, and Banking 27, 777-788. Berger, Allen N., Nathan H. Miller, Mitc hell A. Petersen, Raghuram G. Rajan, and Jeremy C. Stein, 2005, â€œDoes Function Follow Organizational Form? Evidence from the Lending Practices of Large and Small Banks,â€ Journal of Financial Economics 76, 237-269. Berger, Philip G., Eli Ofek, David L. Ye rmack, 1997, Managerial Entrenchment and Capital Structure Decisions, Journal of Finance 52, 1411-1438. Billett, Matthew T., 1996, â€œTargeting Capital Structure: The Relationship between Risky Debt and the Firmâ€™s Likelihood of Being Acquired,â€ Journal of Business 69, 173192. Blundell, Richard and Stephen Bond, 1998, â€œIni tial Conditions and Moment Restrictions in Dynamic Panel Data Models,â€ Journal of Econometrics 87, 115-143.
81 Brewer, Elijah, William E. Jackson III, and James T. Moser, 2001, â€œThe Value of Using Interest Rate Derivatives to Manage Risk at U.S. Banking Organizations,â€ Federal Reserve Bank of Chicago Economic Perspectives, 49-66. Brewer, Elijah, Bernadette A. Minton, & James T. Moser, 2000, â€œInterest Rate Derivatives and Bank Lending,â€ Journal of Banking and Finance 24, 353-379. Bruno, Giovanni S.F., 2005, â€œApproximating the Bias of the LSDV Estimator for Dynamic Unbalanced Panel Data Model,â€ Economic Letters 87, 361-366. Cebenoyan, A. Sinan, and Philip E. Strahan, 2004, "Risk Management, Capital Structure and Lending at Banks," Journal of Banking and Finance 28, 19-43. Coase, Ronald, 1937, â€œThe Nature of the Firm,â€ Economica 4, 386-405. Diamond, Douglas W., 1984, â€œFinancial In termediation and Delegated Monitoring,â€ Review of Economic Studies 51, 393-414. Diamond, Douglas W., 1989, â€œReputa tion Acquisition in Debt Markets,â€ Journal of Political Economy 97, 828-862. Donaldson, Gordon, 1961, Corporate Debt Capacity, Division of Research, Graduate School of Business Administration, Harvard University, Boston, MA. Editors, 2004, â€œRepairing Morgan,â€ Wall Street Journal, January 16, A10. Esty, Benjamin, Bhanu Narasimhan, and Pe ter Tufano, 1999, â€œInterest-Rate Exposure and Bank Mergers,â€ Journal of Banking and Finance 23, 255-285 Fama, Eugene F. and Kenneth R. French, 1997, â€œIndustry Costs of Equity,â€ Journal of Financial Economics 43, 153-193. Fama, Eugene F. and Kenneth R. French, 2002, â€œTesting Trade-Off and Pecking Order Theories of Capital Structure,â€ Review of Financial Studies 15, 1-33. Faulkender, Michael and Mitchell A. Petersen , 2006, â€œDoes the Sour ce of Capital Affect Capital Structure?,â€ The Review of Financial Studies 19, 45-79. Fischer, Edwin O., Robert Heinkel, and Jo sef Zechner, 1989, Dyna mic Capital Structure Choice: Theory and Tests, Journal of Finance 44, 19-40. Flannery, Mark J., and Christopher M. James, 1984, â€œThe Effect of Interest Rate Changes on the Common Stock Returns of Financial Institutions,â€ Journal of Finance 39, 1141-1153. Flannery, Mark J. and Kasturi P. Rangan, 2006, â€œPartial Adjustment toward Target Capital,â€ Journal of Financial Economics 41, 41-73.
82 Frank, Murray Z. and Vidhan K. Goyal, 2006, â€œTrade-off and Pecking Order Theories of Debt,â€ Handbook of Corporate Finance: Empirical Finance (edited by B. Espen Eckbo), Burlington, MA. Froot, Kenneth A., David S. Scharfstein, a nd Jeremy C. Stein, 1993, â€œRisk Management: Coordinating Corporate Invest ment and Financing Policies,â€ Journal of Finance 48, 1629-1658. Froot, Kenneth A., and Jeremy C. Stein, 1998, â€œRisk Management, Capital Budgeting, and Capital Structure Policy for Financial Institutions: An Integrated Approach,â€ Journal of Financial Economics 47, 55-82. Geczy, Christopher C., Bernadette A. Minton, and Catherine Schrand, 1999, â€œChoices Among Alternative Risk Management Stra tegies: Evidence From the Natural Gas Industry,â€ The Rodney L. White Center for Financial Research Paper 028-99. Graham, John R., 1996, â€œDebt and the Marginal Tax Rate,â€ Journal of Financial Economics 41, 41-73. Graham, John R. and Campbell Harvey, 2001, â€œThe Theory and Practice of Corporate Finance,â€ Journal of Financial Economics 60, 187-243. Graham, John R., Michael Lemmon, and Jack Wolf, 2002, â€œDoes Corporate Diversification Destroy Value?,â€ Journal of Finance 57, 695-720. Graham, John R., and Daniel A. Rogers, 2002, â€œDo Firms Hedge in Response to Tax Incentives?,â€ Journal of Finance 57, 815-839. Guay, Wayne, and S.P. Kothari, 2003, â€œHow Much Do Firms Hedge with Derivatives?,â€ Journal of Financial Economics 70, 423-461. Harris, Milton and Artur Raviv, 1990, â€œCapital Structure and the Info rmational Role of Debt,â€ Journal of Finance 45, 321-349. Harris, Milton and Artur Raviv, 1991, â€œT he Theory of Capital Structure,â€ Journal of Finance 46, 300-355. Haushalter, G. David, 2000, â€œFinancing Po licy, Basis Risk, a nd Corporate Hedging: Evidence from Oil and Gas Producers,â€ Journal of Finance 55, 107-140. Hayakawa, Kazuhiko, 2005, â€œSmall Sample Bias Properties of the System GMM Estimator in Dynamic Panel Data Models,â€ Research Unit for Statistical Analysis in Social Sciences Paper. Holmstrom, Bengt and Steven N. Kaplan , 2001, â€œCorporate Governance and Merger Activity in the United States: Making Sense of the 1980s and 1990s,â€ Journal of Economic Perspectives 15, 121-144.
83 Hovakimian, Armen, Tim Opler, and Sheridan Titman, 2001, â€œThe Debt-Equity Choice,â€ Journal of Financial and Quantitative Analysis 36, 1-24. Hughes, Joseph P., William W. Lang, Lore tta J. Mester, and Choon-Geol Moon, 1999, â€œThe Dollars and Sense of Bank Consolidation,â€ Journal of Banking and Finance 23, 291-324. Jensen, Michael C., 1986, â€œAgency Costs of Free Cash Flows, Corporate Finance and Takeovers,â€ American Economic Review 76, 323-339. Jensen, Michael C. and William H. Meckling, 1976, â€œTheory of the Firm: Managerial Behavior, Agency Costs, and Ownership Structure,â€ Journal of Financial Economics 3, 305-360. Jin, Yanbo, and Philippe Jorion, 2005, â€œFirm Value and Hedging: Evidence from U.S. Oil and Gas Producers,â€ Journal of Finance 61, 893-919. Ju, Nengjiu, Robert Parrino, Allen M. Po teshman, and Michael S. Weisbach, 2005, â€œHorses and Rabbits? Trade-Off Theory and Optimal Capital Structure,â€ Journal of Financial and Quantitative Analysis 40, 1-23. Judson, Ruth A. and Ann L. Owen, 1999, â€œE stimating Dynamic Panel Data Models: A Guide for Macroeconomists,â€ Economic Letters 65, 9-15. Kiviet, Jan F., 1995, "On Bias, Inconsistency, and Efficiency of Various Estimators in Dynamic Panel Data Models," Journal of Econometrics 68, 53-78 Korajczyk, Robert A. and Amnon Le vy, 2003, â€œCapital Structure Choice: Macroeconomic Conditions a nd Financial Constraints,â€ Journal of Financial Economics 68, 75-109. Lang, Larry, Eli Ofek, and Rene M. Stulz, 1996, â€œLeverage, Investment, and Firm Growth,â€ Journal of Financial Economics 40, 3-29. Leary, Mark T. and Michael R. Roberts, 2005, â€œDo Firms Rebalance their Capital Structure?,â€ Journal of Finance 60, 2575-2619. Lemmon, Michael L. and Jaime F. Zender, 2004, â€œDebt Capacity and Tests of Capital Structure Theories" AFA Washington, DC Meetings Paper. Lemmon, Michael L., Michael R. Roberts, and Jaime F. Zender, 2006, â€œBack to the Beginning: Persistence and the Cros s-Section of Capital Structure,â€ Social Science Research Network Paper. Lewellen, Wilbur, 1971, â€œA Pure Financial Rationale for the Conglomerate Merger,â€ Journal of Finance 26, 521-537.
84 MacKay, Peter and Gordon M. Phillips, 2005, â€œH ow Does Industry Affect Firm Financial Structure,â€ Review of Financial Studies 18, 1434-1466. Mehrotra, Vikas, Wayne Mikkelsen, Megan Pa rtch, 2003, â€œDesign of Financial Policies in Corporate Spinoffs,â€ Review of Financial Studies 16, 1359-1388. Modigliani, Franco and Merton H. Miller, 1958, â€œThe Cost of Capital, Corporation Finance and the Theory of Investment,â€ American Economic Review 48, 261-297. Mian, Shehzad, 1996, â€œEvidence on Corporate Hedging,â€ Journal of Financial and Quantitative Analysis 31, 419-439. Myers, Stewart C., 1984, â€œThe Capital Structure Puzzle,â€ Journal of Finance 39, 575592. Myers, Stewart C. and Nicholas S. Majluf, 1984, â€œCorporate Financ ing and Investment Decisions When Firms Have Informa tion that Investors Do Not Have,â€ Journal of Financial Economics 13, 187-224. Nain, Amrita, 2005, â€œThe Strategic Motiv es for Corporate Risk Management,â€ AFA Philadelphia Meetings Paper No. 5060. Nance, Deana R., Clifford W. Smith Jr ., and Charles W. Smithson, 1993, â€œOn the Determinants of Corporate Hedging,â€ Journal of Finance 48, 267-284. Nickell, Stephen, 1981, â€œBiases in D ynamic Models with Fixed Effects,â€ Econometrica 49, 1417-1426. Opler, Tim C. and Sheridan Titman, 1994, â€œFinancial Distress and Corporate Performance,â€ Journal of Finance 49, 1015-1040. Roberts, Michael R., 2002, â€œThe Dynamics of Capital Structure: An Empirical Analysis of Partially Observed System,â€ Duke University Fuqua School of Business Paper. Samor, Geraldo, 2004, â€œCVR D Has Investors Fretting; Acquisitions Would Risk Exposure to Commoditiesâ€™ Fluctuations,â€ Wall Street Journal, July 9, C14. Santomero, Anthony M., 1997, â€œCommercial Bank Risk Management: an Analysis of the Process,â€ Wharton Financial Institutions Center Paper. Smith, Clifford W., and Rene M. Stulz, 1985, â€œThe Determinants of Firms' Hedging Policies,â€ The Journal of Financial and Quantitative Analysis 20, 391-405. Stein, Jeremy, 2003, â€œAgency, Inform ation, and Corporate Investment,â€ Handbook of the Economics of Finance (edited by George Constantin ides, Milt Harris, and Rene Stulz), Elsevier Science, 111-165.
85 Strahan, Philip E., 2003, â€œThe Real Effects of U.S. Banking Deregulation,â€ Federal Reserve Bank of St. Louis Review, 111-128. Stulz, Rene, 1984, â€œOptimal Hedging Policy,â€ Journal of Financial and Quantitative Analysis 19, 127-140. Stulz, Rene, 1990, â€œManagerial Discre tion and Optimal Financial Policies,â€ Journal of Financial Economics 26, 3-27. Stulz, Rene, 2004, â€œShould We Fear Derivatives?,â€ Journal of Economic Perspectives 18, 173-192. Tufano, Peter, 1996, â€œWho Manages Risk? An Empirical Examination of Risk Management Practices in th e Gold Mining Industry,â€ Journal of Finance 51, 10971137. Villalonga, Belen, 2004, "Does Diversification Cause the 'Diversification Discount'?," Financial Management 33, 5-27. Welch, Ivo, 1996, â€œEquity Offerings following the IPO: Theory and Evidence,â€ Journal of Corporate Finance 2, 227-259. Welch, Ivo, 2004, â€œCapital St ructure and Stock Returns,â€ Journal of Political Economy 112, 106-131. Windmeijer, Frank, 2005, â€œA Finite Sample Correction for the Variance of Linear Efficient Two-Step GMM Estimators,â€ Journal of Econometrics 126, 25-51. Wooldridge, Jeffrey M., 2002, Econometric Analysis of Cross Sectional and Panel Data, The MIT Press, Cambridge. Xu, Yexiao, and Burton G. Malkiel, 2003, â€œInve stigating the Behavior of Idiosyncratic Volatility,â€ Journal of Business 76, 613-644.
86 BIOGRAPHICAL SKETCH Kristine received her BA in political science from the University of Chicago and an MS in statistics from Rutgers University. Recently, she was awarded the American Association of University Women American Dissertation Fellowship. For her first four years at UF, she held the Marshall Criser Pr esidential Fellowship. Her research interests include corporate finance, risk management , payout policy, and investments. Before coming to UF, Kristine worked in New York City with both Merrill Lynchâ€™s Private Equity Group and PricewaterhouseCoopersâ€™ Fi nancial Securities Litigation practice.