<%BANNER%>

Three Essays on Intermediary Lending

Permanent Link: http://ufdc.ufl.edu/UFE0044632/00001

Material Information

Title: Three Essays on Intermediary Lending
Physical Description: 1 online resource (167 p.)
Language: english
Creator: Wang, Xiaohong
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: bank -- base -- borrowing -- capital -- crisis -- financing -- intermediary -- investment -- lending -- loan -- shock -- structure
Finance, Insurance and Real Estate -- Dissertations, Academic -- UF
Genre: Business Administration thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: In this study, I examine three topics on intermediary lending. The first part of the study investigates a relatively common, but little-studied type of credit line, for which funds availability is limited by the firm’s time-varying asset composition. A “borrowing base” line of credit provides funds as a proportion of (e.g.) the amount of the borrower’s accounts receivable. I find that borrowing base (“BB”) revolvers are taken more often by borrowers with high risk and low profits or cash flow. Compared to other types of secured credit lines, the BB loan rate is relatively insensitive to the borrower’s initial risk profile because the lender has security and the loan terms automatically limit credit extended to unsuccessful firms. Borrowers use the BB to reduce their borrowing cost, gain a more generous credit limit, and operate with fewer financial covenants.  In the second part of the study, I investigate how a firm’s leverage ratio, debt maturity, and bank debt proportion are jointly determined.Using a system of simultaneous equations, I find that bank debt greatly increases firms’ leverage ratios, whereas a short debt maturity generally decreases firms’ leverage ratios. Debt maturity and bank debt can alter the relation between the leverage ratio and two types of borrowing costs: the potential conflicts between equity holders and debt holders, and financial distress risk. In the third part of the study, I examine how shocks to banks’ financial conditions impact corporate financing and investment decisions during the 2007-2009 financial crisis. I find that average firms relied more heavily on bank credit during the crisis. However, firms whose banks incurred more nonperforming loans used less bank credit when comparing their bank debt before and during the crisis. The reduction on bank debt weren’t replaced by alternative financing such as public debt or trade credit. There is some evidence that shocks on banks eventually affected corporate real activities; firms with more adversely affected banks invest less and hoard more cash during the crisis compared to their pre-crisis level. Overall, my results suggest that adverse shocks on the banking system can curtail bank lending and negatively affect the real sector.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Xiaohong Wang.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Flannery, Mark J.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2014-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044632:00001

Permanent Link: http://ufdc.ufl.edu/UFE0044632/00001

Material Information

Title: Three Essays on Intermediary Lending
Physical Description: 1 online resource (167 p.)
Language: english
Creator: Wang, Xiaohong
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: bank -- base -- borrowing -- capital -- crisis -- financing -- intermediary -- investment -- lending -- loan -- shock -- structure
Finance, Insurance and Real Estate -- Dissertations, Academic -- UF
Genre: Business Administration thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: In this study, I examine three topics on intermediary lending. The first part of the study investigates a relatively common, but little-studied type of credit line, for which funds availability is limited by the firm’s time-varying asset composition. A “borrowing base” line of credit provides funds as a proportion of (e.g.) the amount of the borrower’s accounts receivable. I find that borrowing base (“BB”) revolvers are taken more often by borrowers with high risk and low profits or cash flow. Compared to other types of secured credit lines, the BB loan rate is relatively insensitive to the borrower’s initial risk profile because the lender has security and the loan terms automatically limit credit extended to unsuccessful firms. Borrowers use the BB to reduce their borrowing cost, gain a more generous credit limit, and operate with fewer financial covenants.  In the second part of the study, I investigate how a firm’s leverage ratio, debt maturity, and bank debt proportion are jointly determined.Using a system of simultaneous equations, I find that bank debt greatly increases firms’ leverage ratios, whereas a short debt maturity generally decreases firms’ leverage ratios. Debt maturity and bank debt can alter the relation between the leverage ratio and two types of borrowing costs: the potential conflicts between equity holders and debt holders, and financial distress risk. In the third part of the study, I examine how shocks to banks’ financial conditions impact corporate financing and investment decisions during the 2007-2009 financial crisis. I find that average firms relied more heavily on bank credit during the crisis. However, firms whose banks incurred more nonperforming loans used less bank credit when comparing their bank debt before and during the crisis. The reduction on bank debt weren’t replaced by alternative financing such as public debt or trade credit. There is some evidence that shocks on banks eventually affected corporate real activities; firms with more adversely affected banks invest less and hoard more cash during the crisis compared to their pre-crisis level. Overall, my results suggest that adverse shocks on the banking system can curtail bank lending and negatively affect the real sector.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Xiaohong Wang.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Flannery, Mark J.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2014-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044632:00001


This item has the following downloads:


Full Text

PAGE 1

1 THREE ESSAYS O N INTERMEDIARY LENDING By XIAOHONG WANG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVER SITY OF FLORIDA 201 2

PAGE 2

2 201 2 Xiaohong Wang

PAGE 3

3 To my father and mother, and brothers and sisters at GCCC, and foremost Jesus Christ, my personal savior

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank my father and mother for always making any sacrif ice necess ary to love me and support me. Their love and devotion to me is what allowed me to achieve my personal and professional endeavors. I thank my Committee, Mark Flannery (Chair), Joel Houston, Nimalendran Mahendrarajah, and Sarah Hamersma, for thei r significant contributions as well as countless hours of support regarding my thank my fellow finan ce Ph.D. students for helping make the process enjoyable. I am grateful

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 9 ABSTRACT ................................ ................................ ................................ ................... 10 CHAPTER 1 INTRODUCTIO N ................................ ................................ ................................ .... 12 Overview of Chapter 2 ................................ ................................ ............................ 13 Overview of Chapter 3 ................................ ................................ ............................ 16 Overview of Chapter 4 ................................ ................................ ............................ 17 2 BORROWING BASE REVOLVERS: LIQUIDITY FOR RISKY FIRMS .................... 21 Borrowing Base Lines of Credit ................................ ................................ .............. 25 Sample Construction ................................ ................................ ........................ 26 Descriptive Statistics for Borrow ing Base Credit Lines ................................ ..... 28 Borrower, Lender, and Contract Characteristics: Borrowing Base vs. Other Secured Credit Lines ................................ ................................ ..................... 29 Multivariate Model Specification ................................ ................................ ............. 31 Endogenous Switching ................................ ................................ ..................... 32 Variable Definitions ................................ ................................ .......................... 34 Empirical Results ................................ ................................ ................................ .... 38 Contract Selection: Borrowing Base vs. Other Secured Credit Lines ............... 38 Spread, Credit Li mit, and Financial Covenants: Borrowing Base vs. Other Secured Credit Lines ................................ ................................ ..................... 40 Predicted Contract Terms by Selected Line of Credit Type .............................. 42 Chapter 2 Concluding Remarks ................................ ................................ .............. 44 3 THE JOINT DETERMINATION OF LEVERAGE RATIO, DEBT MATURITY, AND DEBT SOURCE ................................ ................................ ............................. 56 Agency Problem, Financial Distress and Capital Structure Decisions .................... 61 Leverage, Agency Problem, and Financial Distress ................................ ......... 62 Short term Debt, Agency Problem, and Financial Distress .............................. 63 Bank Debt, Agency Problem, and Financial Distress ................................ ....... 63 Leverage, Short Term Debt, and Bank Debt ................................ ................... 64 Existing Empirical Evidence ................................ ................................ ............. 65 Simultaneous Equations ................................ ................................ ................... 66 Data, Proxies, and Descriptive Statistics ................................ ................................ 68

PAGE 6

6 Data ................................ ................................ ................................ .................. 68 Proxies ................................ ................................ ................................ ............. 70 Descriptive Statistics ................................ ................................ ........................ 73 Joint Determinants of Leverage, Short term Debt, and Bank Debt ......................... 76 Estimation Results of Two equation, and Three equation systems. ................. 76 Differences in the Three equation System across Rated and Unrated Firms ... 85 Robustness Check ................................ ................................ ........................... 88 Chapter 3 Concluding Remarks ................................ ................................ .............. 91 4 THE EFFECT OF BANK SHOCKS ON CORPORATE FINANCING AND INVESTMENT: EVIDENCE FROM 2007 2009 FINANCIAL CRISIS .................... 100 2007 2009 Financial Crisis ................................ ................................ ................... 105 Data, Sample Description and Methodology ................................ ......................... 108 Data and Samp le Description ................................ ................................ ......... 108 Empirical Strategy ................................ ................................ .......................... 109 Model ................................ ................................ ................................ .............. 111 Descriptive Statistics ................................ ................................ ...................... 113 Empirical Results ................................ ................................ ................................ .. 114 Bank Debt ................................ ................................ ................................ ....... 114 Trade Credit, Public Debt, and Leverage ................................ ....................... 119 Investment and Cash Holdings ................................ ................................ ....... 1 20 Robustness and Extension ................................ ................................ ............. 123 Chapter 4 Concluding Remarks ................................ ................................ ............ 125 5 CONCLUSION ................................ ................................ ................................ ...... 144 APPENDIX A VARIABLES CONSTRUCTED IN CHAPTER 3 ................................ .................... 146 B THEORETICAL PREDICTIONS OF CONTROL VARIABLES IN CHAPTER 3 .... 148 C VARIABLES CONSTRUCTED IN CHAPTER 4 ................................ .................... 154 D DESCRIPTIVE STATISTICS OF WEIGHTED AVERAGE BANK CHARACTERISTICS IN CHAPTER 4 ................................ ................................ .. 156 E ROBUSTNESS TEST OF FIRM BANK MATCHING IN CHAPTER 4 ................... 157 F BANK NON PERFORMING LOANS IN CHAPTER 4 ................................ ........... 158 LIST OF REFERENCES ................................ ................................ ............................. 159 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 167

PAGE 7

7 LIST OF TABLES Table page 2 1 Sample selection procedure ................................ ................................ ............... 46 2 2 Borrowing base line of credit contractual characteristics by asset types ............ 47 2 3 Borrowing base line of credit contract ual characteristics by industry .................. 48 2 4 Borrowing base line of credit contractual characteristics by loan purpos e. ......... 49 2 5 Descriptive st atistics by line of credit type ................................ .......................... 50 2 6 Probit model for credit line selection ................................ ................................ ... 52 2 7 Switching regression of spread, credit limit, and financial covenants ................. 53 2 8 Predicted spread, credit limit, and number of financial covenants under actual and alternative line of credit contract (switching model) ................................ ..... 54 2 9 Predicted spread, credit limit, and number of financial covenants under actual and alternative line of c redit contract ( o rdinary lea st square model) .................. 55 3 1 Summary statistics of depe ndent and independent variables ............................. 94 3 2 Pearson correlation coefficient s between variables of interest .......................... 95 3 3 Joint determinants of leverage, short term debt, and bank debt wit h no firm and year fixed effect ................................ ................................ ........................... 96 3 4 Joint determinants of leverage, short term debt, and bank d ebt for rated and unrated firms ................................ ................................ ................................ ...... 98 3 5 Joint determinants of leverage, short term debt, and bank debt with firm and year fixed effect ................................ ................................ ................................ 99 4 1 Summary st atistics of firm characteristic ................................ .......................... 127 4 2 Summary statistics of bank characteristics ................................ ....................... 128 4 3 Firm f inancing and i nvestment b efore and a fter c risis c omparison ................... 129 4 4 Bank c haracteristics b efore and a fter c risis c omparison ................................ ... 130 4 5 Bank debt and bank financial conditions ................................ .......................... 131 4 6 Credit line drawdowns and bank financial conditions ................................ ....... 132

PAGE 8

8 4 7 Term loans and bank financial conditions ................................ ......................... 133 4 8 Trade credit and bank financial conditions ................................ ....................... 134 4 9 Public debt and bank financial conditions ................................ ......................... 135 4 10 Book leverage and bank financial conditions ................................ .................... 136 4 11 Market leverage and bank financial conditions ................................ ................. 137 4 12 Investment and bank financial conditions ................................ ......................... 138 4 13 Investment and bank financial conditions ................................ ......................... 139 4 14 Cash holding and bank financial conditions ................................ ...................... 140 4 15 Cash holding and bank financial conditions ................................ ...................... 141 4 16 Seemingly unrelate d regressions on firm financing ................................ ......... 142 4 17 Change in equity by bank financial condition quintiles ................................ ...... 143 B 1 Theoretical predictions of control variables in Chapter 3 ................................ 148 D 1 Descriptive statistics of weighted average bank characteristics in Chapter 4 ... 156 E 1 Robustness test of firm bank matching in Chapter 4 ................................ ........ 157

PAGE 9

9 LIST OF FIGURES Figure page F 1 Bank non performing loans in Chapter 4 ................................ .......................... 158

PAGE 10

10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THRE E ESSAYS O N INTERMEDIARY LENDING By Xiaohong Wang August 201 2 Chair: Mark Flannery Major: Business Administration In this study I examine three topics on intermediary lending The first part of the study investigates a relatively common, but little stu died type of credit line, for which ime varying asset composition. accounts receivable. I find that borrowi borrowers with high risk and low profits or cash flow. Compared to other types of profile because the lender ha s security and the loan terms automatically limit credit e xtended to unsuccessful firms. Borrowers use the BB to reduce their borrowing cost, gain a more generous credit limit, and operate with fewer financial covenants. In the second part of the stu dy maturity, and bank debt proportion are jointly determined. Using a system of whereas a short debt maturity generally and bank debt can alter the relation between the leverage ratio and two types of

PAGE 11

11 borrowing costs: the potential conflicts between equity holders and debt holders, and finan cial distress risk. In the third pa rt of the study, I examine impact corporate financing and investment decisions during the 2007 2009 financial crisis. I find that average firms relied more heavily on bank credit during the crisis. Howeve r, firms w hose banks incurred more nonperforming loans used less bank credit when comparing their bank debt before and during the crisis. The reduction on bank There is some evidence that shocks on banks eventually affected corporate real activities; f irms with more adversely affected banks invest ed less and hoard ed more cash during the crisis compared to their pre crisis level. Overall, my results suggest that adverse shocks on the b anking system can curtail bank lending and negatively affect the real sector.

PAGE 12

12 CHAPTER 1 INTRODUCTION In the three chapters that constitute this study, I examine three topics on intermediary lending The first part of the study, Chapter 2, investigates a relatively common, but little studied type of credit line, for which funds availability is limited by the varying asset composition, namely borrowing base lines. Whereas Sufi (2009) reports that firms with low cash flow are more likely to lose access to their lines of credit, high risk and low profits or cash flow. Compared to other types of secured credit lines, s initial risk profile because the lender has security and the loan terms automatically limit credit extended to unsuccessful firms. Borrowers use the BB to reduce their borrowing cost, gain a more generous credit limit, and operate with fewer financial c ovenants. debt maturity, and bank debt proportion are jointly determined. In particular, this study public bonds vs. bank loans on leverage. The substantial literature on capital structure decisions often examines one facet in isolation. The study of one capital structure decision in isolation is, however, at odds with the fact that firms simultaneously decide how much debt to take out (debt equity decision), how soon to repay the debt (debt maturity), and from whom to borrow (debt source) etc. Using a system of simultaneous equations, I find that debt maturity, debt source, and leverage ratio are jointly determined. Bank debt ratios. Debt maturity and bank debt can alter the relation between the leverage ratio and

PAGE 13

13 two types of borrowing costs: the potential conflicts betw een equity holders and debt holders, and financial distress risk. The results suggest that in addition to leverage ratio, debt maturity and debt source importantly affect the distortions associated with debt finance. In the third part of the study, I exam impact corporate financing and investment decisions using the 2007 2009 financial crisis as an experimental setting. Though bank credit is viewed as an important source of funding to corporations in the literat ure, when examining corporate financing and investment decisions, most of the empirical works analyzes these corporate decisions as a function of firm fundamentals (demand side). I explore the cross sectional difference in shocks across banks and over time estate market during the crisis and utilize this variation to measure how bank shocks impact corporate financing and investment decisions, when the nonfinancial sector was less directly affected by the financial crisi s. I find that average firms relied more heavily on bank credit during the crisis. However, firms whose banks incurred a larger amount of nonperforming loans used less bank credit when comparing their bank debt before and during the crisis. Their reduction credits such as public debt or trade credit. Shock on banks eventually affected corporate real activities that firms with more adversely affected banks invest ed less and hoard ed more cash during the crisis comp ared to their pre crisis level. Overview of Chapter 2 liquidity management decision, and found that credit lines provide reliable liquidity only for relatively high quality fir ms that can maintain high cash flows. He concludes that

PAGE 14

14 of credit take a variety of institutional forms, and I present here the first broad empirical my knowledge, this is the first empirical evaluation of borrowing base credit lines, despite a gro wing interest in corporate credit lines more broadly (e.g., Shockley and Thakor (1997), Dennis, Nandy, and Sharpe (2000), Roberts and Sufi (2009), Sufi (2009), Lockhart (2009), Flannery and Lockhart (2009), Yun (2009), Guo and Yun (2009), and Ivashina and Scharfstein (2010)). Borrowing base lines permit a firm to borrow the lesser of some absolute limit and a percentage of some specified type of asset. The most common borrowing base category is accounts receivable, but inventories provide another common b ase. by specific firm assets, thereby providing the lender with a means of repa yment even if the firm itself fails. Second, one can think of the BB constraint as providing a state dependent loan amount. The BB feature on a line of credit not only requires the borrower to pledge collateral, but also assures that the loan amount cann ot exceed the This study presents a systematic study of BB revolving lines of credit, for which I know of no precedent in the finance literature. In order to focus explicitly on the effects and implications of the BB restriction, I co mpare BB lines of credit to other secured credit lines. I construct a sample of 5,154 secured lines of credit by matching loans

PAGE 15

15 from Dealscan to financial information from Compustat over 1995 2008. I describe the d to evaluate the type of firms that operate other sort of secured line, I estimate an endogenous switching model as in Maddala (1983). I s riskiness, information asymmetry, and asset composition affect the terms of BB and non BB lines of credit. BB borrowers tend to be relatively poor credit prospects. Some practitioners therefore consider borrowing base loans to be very risky (Vinter (1 998, p. 371)), while others argue that BB loans can be quite safe if they are properly administered. This latter opinion is more consistent with my findings. Moreover, (spread) is less sensitive to firm size, profitability, cash flow volatility, Z score, investment opportunities, and leverage. These differences are both economically and statistically significant. Given that BB line borrowers tend to have worse credit quality, this seems puzzling. The explanation lies in the added se curity conveyed to lenders by are less important to a BB lender than to a lender of credit lines with fixed credit limits. Borrowing firms select into the type of loan that is more advantageous for them. According to my estimates, if BB borrowers were to give up the BB restriction and select another type of secured line, they would pay a substantially higher spread, receive a substantially lower credit limit, an d be subject to more restrictive covenants. Whereas Sufi (2009) reports that firms with low cash flow are more likely to lose access to their lines of credit, I find that one specific sort of credit line one tied to a borrowing base remains open even to firms with poor financials and low cash flows.

PAGE 16

16 These loans are taken primarily by smaller firms, whose open market borrowing alternatives are most limited. The BB restriction causes the amount lent to vary ess, making it a state dependent covenant condition weakens. Overview of Chapter 3 The substantial literature on capital structure decisions often examines one facet in isol ation. The study of one capital structure decision in isolation is, however, at odds with the fact that firms simultaneously decide how much debt to take out (debt equity decision), how soon to repay the debt (debt maturity), and from whom to borrow (debt jointly determined; maturity, in addition to leverage ratio, can be used to mitigate the underinvestment problem. Among the few papers that identify the joint determinatio n of capital structure jointly determined. He argues that maturity, in addition to leverage ratio, can be used to mitigate the underinvestment problem. Billett et al. (2 007) extends Johnson (2003) by including covenant structure into the endogenous decision system. They find that covenants can substitute for a short debt maturity and a low leverage to mitigate the agency problem of high growth firms. This study is the firs t empirical investigation that explores the source of debt in the joint capital structure decisions. In particular, I examine the role of bank financing in the joint capital structure decisions. Previous studies argue that bank debt is special. Bank financ

PAGE 17

17 information production and monitoring services. Bank debt can also provide flexibility in financial distress through loan renegotiation. James (1987), Lummer and McConnell (1989), Best and Zhang (1993), and Billett et al. (1995) all document a positive share price reaction to bank loan announcements, suggesting that bank debt is viewed favorably by the market. Using a system of simultaneous equations, I find that debt maturity, debt so urce, maturity and bank debt can alter the relation between the leverage ratio and two types of borrowing costs: the potential conflicts between equity holders and debt holders, and financial distress risk. The results suggest that in addition to leverage ratio, debt maturity and debt source importantly affect the distortions associated with debt finance. Overview of Chapter 4 Banks play an important role in providing financing to corporations. Bank lending is important not only to small firms with limited access to public debt market, but also to large and medium sized companies. Thoug h bank credit is viewed as an important source of funding to corporations in the literature, when examining corporate financing and investment decisions, f corporate financing and investment decisions. T hough some papers investigate how for example Bernanke (1983)), their method suffers from the critique that a reduction on bank lending may be driven by aggregate production or technology shocks from t he demand side. Recently, several studies (Peek and Rosengren (1997, 2000), Khwaja and Mian (2008), Schnabl (2011))

PAGE 18

18 address these critiques by identifying exogenous shocks on banks and examining the heterogeneous response of corporate borrowers to banks wi th different level of shocks. impact corporate financing and investment decisions using the 2007 2009 financial crisis as an experimental setting. First, t he recent financial cri sis is the largest shock to the banking system since the great depression and merits a systematic examination of whether and how shock s on banks impacted the real sector. Second, the recent financial crisis offers a nice setting to mitigate the confounded demand side effect. the lax lending standard during the boom period and a bust of real estate value during the crisis. Yet the deterioration of these real estate loan s was not expected by banks and created substantial variation across banks and over time. T he losses on the residential or commercial real estate loans made by banks before the crisis should not nd before and durin g the crisis and allow us examine how bank shocks impact corporate financing and investment decisions. To my best knowledge, this study is the most comprehensive study that investigates the impact of bank shocks on corporate financing and investment decis ions during the 2007 2009 financial crisis. I break down the investigati on into three steps. The first step aims to examine whether firms are given less bank credit when their relationship banks were more adversely affected. The second step investigates wh ether and how firms adjust ed their financing decisions in response to the negative shock from their relationship banks. The third step examine s whether shocks on the banking sector

PAGE 19

19 ultimately impact corporate b nd cash holding decisions. To disentangle the credit supply from the demand side, I identify an exogenous source shock on banks that affects banks heterogeneously but arguably has little to do with corporate demand of bank credit : melt down of the real estate market measured as bank non performing loans excluding the commercial and in industrial loans I explore the cross sectional difference in shocks across banks and during the crisis and utilize this variation to measure how bank shocks impact corporate financing and investment decisions, when the nonfinancial sector was less directly affected by the financial crisis. Overall, I find the following main results. First I show that firms increased their overall level of bank debt during the crisis. Both term loans and credit line drawdowns increased during the crisis. However, firms with more adversely affected banks did not use as much bank debt as firms with less affe cted banks. This effect manifested mainly affected the supply of bank credit to corporate borrowers during the crisis. Second, I find that firms with adversely affected bank s did not replace the reduced bank credit with other source of credit during the crisis. Public debt and trade credit did not increase for firms with more distressed banks with respect to other firms. The leverage of firms with more troubled banks decrease d more than that of firms with healthier banks. Third, I find some evidence of real effects associated with bank shocks. Firms with more troubled lenders invest less during the crisis compared to those with healthier banks.

PAGE 20

20 There is weak evidence that they tended to hoard more cash on their balance sheet during the crisis for precautionary reasons.

PAGE 21

21 CHAPTER 2 BORROWING BASE REVOLVERS: LIQUIDITY FOR RISKY FIRMS liquidity man agement decision, and found that credit lines provide reliable liquidity only for relatively high quality firms that can maintain high cash flows. He finds that 35% of his sample firms violate at least one credit line covenant, and that a violation is fol lowed by a 15% to 25% drop in the availability of total credit. Low cash flow is a strong contingent lines of credit that exist in the marketplace are distinct from the committed lines of credit tha t are described 1 However, lines of credit take a variety of institutional forms, and I present here the first broad empirical study of revolving credit lines whose maximum amount is tied to a my knowledge, this is the first empirical evaluation of borrowing base credit lines, despite a growing interest in corporate credit lines more broadly (e.g., Shockley and Thakor (1997), Dennis, Nandy, and Sharpe ( 2000), Roberts and Sufi (2009), Sufi (2009), Lockhart (2009), Flannery and Lockhart (2009), Yun (2009), Guo and Yun (2009), and Ivashina and Scharfstein (2010)). Borrowing base lines permit a firm to borrow the lesser of some absolute limit and a percent age of some specified type of asset. The most common borrowing base category is accounts receivable, but inventories provide another common base. 1 Theoretical studies ( e.g. Boot, Thakor, and Udell (1987), Berkovitch and Greenbaum (1991), and Holmstrom and Tirole (1998)) suggest market frictions by providing finance for valuable investment opportunities.

PAGE 22

22 secured revolver alternative when they cannot obtain an unsecured bank loan which, Barnett (1997) explains that based borrower generally does not have the A BB loan has two interesting contracting features. First, it is secured by spe cific firm assets, thereby providing the lender with a means of repayment even if the firm itself fails. Second, one can think of the BB constraint as providing a state dependent loan amount. The BB feature on a line of credit not only requires the borro wer to pledge nd to have higher accounts thus lends more only when the firm is doing well. By comparison, a troubled firm can run up the outstanding balance on an unsecured line ( or one with a general security condition. Although BB lines must be monitored more intensively than most other kinds of loans, they can be administered with relatively few restrictive covenants. 2 If the firm does well, it can borrow; if not, no funds are extended. (2008) of 291 firms in its Ultimate Recovery Database. They find that 237 of these firms 2 Because of its high reliance on collateral, an asset based lender needs the immediate ability to Therefore, this type of financing often requires an o ngoing monitoring process which includes periodic field examinations, frequent confirmation of accounts, control of proceeds of accounts receivable and e 82)

PAGE 23

23 had a revolving line of credit at default. As default approached, the proportion of bank average and median drawdown rates increase to 64% and 70% at default from 47% draw its line without posting additional collateral, protecting the lender from an unpleasant end game drawdown. My dataset, taken from Dealscan, indicates that the BB feature is quite commonly used in line of credit contracts. Between 1995 and 2008, 49% of all credit lines extended to U.S. borrowers were explicitly se cured by some sort of asset, of which, 38% were tied to a borrowing base formula. In dollar terms, BB lines constitute only 23% of all secured lines in my sample, suggesting that BB loans are particularly common at smaller firms. Indeed, among the sample firms smaller than $500 million, 45% of the secured line contracts include the BB feature. Consistent with the hypothesis that BB formula constitutes a powerful loan covenant, I also find that BB loans are provided with fewer covenants. This study prese nts a systematic study of BB revolving lines of credit, for which I know of no precedent in the finance literature. In order to focus explicitly on the effects and implications of the BB restriction, I compare BB lines of credit to other secured credit l ines. I construct a sample of 5,154 secured lines of credit by matching loans from Dealscan to financial information from Compustat over 1995 2008. I describe the with

PAGE 24

24 other sort of secured line, I estimate an endogenous switching model as in Maddala (1983). I comp osition affect the terms of BB and non BB lines of credit. BB borrowers tend to be relatively poor credit prospects. Some practitioners therefore consider borrowing base loans to be very risky (Vinter (1998, p. 371)), while others argue that BB loans ca n be quite safe if they are properly administered. 3 This latter opinion is more consistent with my findings. Moreover, (spread) is less sensitive to firm size, profitability, cash flow volatility, Z score, investment opportunities, a nd leverage. These differences are both economically and statistically significant. Given that BB line borrowers tend to have worse credit quality, this seems puzzling. The explanation lies in the added security conveyed to lenders by the BB restriction are less important to a BB lender than to a lender of credit lines with fixed credit limits. Consistent with the prediction of the costly contracting hypothesis that debt contrac ts are written to control the conflict between bondholders and stockholders, I find that riskier and less profitable borrowers are more likely to have the BB feature included in their contracts. Yet, borrowers of BB lines obtain lower spread, higher credit availability, and fewer financial covenants than what they would get under non BB secured lines. 3 To manage the credit risk of the borrowers, lenders request timely reporting of financials and other material information from the borrower and closely monitor the underlying borrowing base assets. The finance literature has stressed the informati on advantage of financial intermediaries (Diamond (1984), (1991)). One skill that controls lender losses is the ability to value and to dispose of borrowing base assets. For instance, a lender to a department store could discount their finance receivables more aggressively just after Christmas, when returns of sold merchandise are particularly high.

PAGE 25

25 The chapter is o rganized as follows. I first discuss the bo rrowing base contract, describe my sample, and present summary statistics about secured lines of credit. Since the literature contains relatively little information about BB loans, these summary statistics also describe the product and serve as stylized facts. Next, I describe the empirical methodology and model specification. I then present my emp irical results and conclude with a summary of the findings Borrowing Base Lines of Credit A borrowing base ( BB ) line of credit is similar to a typical line of credit in the sense that it allows borrowers to draw down the line to meet liquidity needs u p to a maximum amount during a certain time period. 4 And when an advance has been repaid, it becomes available for future draw down again. BB lines of credit, however, restrict the amount drawn to be equal to or less than a pre specified proportion of the borrowing base assets, most commonly accounts receivable or inventory. Under a BB line, the smaller of a stated maximum amount and the pre specified mple, United Natural Foods, Inc used eligible accounts receivable and inventory as the borrowing base for a line of credit that was amended in September 2001 to provide a credit facility 4 GE Capital explains a revolving credit facility as follows: A revolver is a loan which can be drawn down and repaid. In a business context, a revolver fre quently is secured by the borrower's receivables and/or inventory. This kind of asset based loan is designed to optimize the availability of working capital from the borrower's current asset base. Here's how it works. The borrower grants a security interes t in its receivables and/or inventory to the lender as collateral to secure the loan. This grant of security interest creates the borrowing base for the loan. As receivables are paid, the cash is turned over to the lender to pay down the loan balance. When the borrower needs additional working capital, the borrower requests another advance. The lender manages a revolving credit facility and the related collateral in order to offer the borrower the largest possible loan amount at any given time. ( http://www .gelending.com/Clg/Resources/lendingFAQs.html, last accessed on Oct. 7, 2010)

PAGE 26

26 of $150 million. In their 10Q for the fiscal year ending April 30, 2 002, they reported accounts receivable and inventories of about $90 and $140 million respectively, yet their borrowing base based on eligible subsets of those balance sheet items totaled only $143.5 million. Only by increasing its eligible receivables or inventory could United Natural Foods borrow the full committed amount of $150 million. Asset based loans in general were once the province of commercial finance companies. For example, when Carey, Post, and Sharpe (1998) compare bank and finance comp any loans, they conclude that finance company make riskier loans. In my sample, finance companies are relatively more concentrated in BB loans: they serve as lead lenders for 26% of BB tranches, as opposed to only 8% of non BB secured line tranches. Com mercial banks account for 70% of BB tranches, compared to 87% for non BB secured revolvers. Sample Construction My source of information about credit lines is the Dealscan database from Loan Pricing Corporation (LPC). Dealscan provides detailed informati on about commercial loans extended by commercial banks, investment banks, and nonbank institutions (e.g. insurance companies and finance companies). According to LPC, a majority of the loan tain additional information via private contact with the credit industry, borrowers, and lenders. contain several tranches, for example a senior term loan, a junior ter m loan, and a revolving line of credit. My terms, specifically the loan rate, line maximum, and the number of covenants.

PAGE 27

27 Table 2 1 describes the sample selection procedure. I start by ide ntifying all 67,398 tranches taken by U.S. borrowers between 1995 and 2008. 5 Using the method from Chava and Roberts (2008), I loan origination date with company names and their active dates in the Compustat d atabase. 6 This procedure leaves us 26,069 loan tranches taken by 5,138 identified publicly traded companies. Credit lines represent more than 70% of the dollars lent: $5.2 billion vs. $7.2 billion of all loans. From among the 17,640 lines of credit, I identify This yields 8,646 loan tranches that are explicitly identified as sec ured lines of credit. Unsecured lines are about four times larger than the secured lines and carry a much lower spread (79 bps vs. 243 bps). In order to gauge the effect of a borrowing base ( BB ) restriction, I compare BB lines to other secured lines of credit with no such restriction. By excluding unsecured lines of credit from my sample, I factor out decisions about loan security and focus on the incidence of the BB restriction. Of secured lines, 3,316 are BB lines of credit and 5,330 are secured but have only a pre specified loan maximum. After matching these borrowers to Compustat, my final sample constitutes 2,117 BB lines of credit and 3,037 non BB lines of credit to borrowers with available accounting information. 7 Among the 5 I start with year 1995 because the Dealscan database contains few borrowing base lines of credit prior to that date. 6 The matching file between Dealscan and Compustat is kindl y provided by Michael Roberts. 7 I exclude the line of credit contracts from my sample if any variable in T able 2 3 is missing. Most score, excess cash and volatility of cash flow.

PAGE 28

28 secured lines of cre dit, those with a BB feature are smaller and more expensive: BB lines can borrow only half as much as the non BB lines and require a 25% higher spread. A similar pattern holds among borrowers with assets below $500 million, which comprise 82% of all BB l ine borrowers. BB lines account for approximately 19% of publicly and the BB restriction appears in 19% of loan packages in Dealscan. 8 BB lines of credit, however, only account for 9% of the dollar amount of total lines of credit, because they are often offered in small size. This smaller loan limit reflects the typical of the total number of secured lines. This observation coincides w view that BB lines of credit are particularly appropriate for small firms. Descriptive Statistics for Borrowing Base Credit Lines Table 2 2 to 2 4 report descriptive statistics for my sample borrowing base ( BB ) loan features. Tabl e 2 2 demonstrates that the most commonly used borrowing base assets are accounts receivables (87% of all BB lines) and inventory (63%). Fixed assets, and cash and marketable securities only appear in 7% and 3% of my sample respectively. The advance rate s for different types of assets vary with their liquidity and opaqueness characteristics. Cash and cash equivalents have the highest advance rate of 97%, followed by accounts receivable, fixed assets, and lastly inventory. Although inventory is considered a current asset on the balance sheet, it has a substantially lower advance rate than other assets, including fixed assets. 8 In their sample of 1,000 loan deals, Roberts and Sufi (2009) also report that 19% of their loan contracts include the BB restriction.

PAGE 29

29 Table 2 3 shows that firms using BB lines of credit are more heavily concentrated in manufacturing, retail, and wholesale industri es than is the Compustat universe. On the other hand, firms in services, transportation, communications, and utilities, and mining industries are under represented in the BB sample. Table 2 4 exhibits the reported reason for taking BB lines and non BB lines. BB lines are mainly used to raise working capital, repay debt, or for unspecified corporate purposes. These three purposes account for up to 84% of my BB lines of credit sample. Compared to the loan purpose of non BB lines, BB lines of credit are taken more often to fund working capital and for unspecified corporate purposes. Non BB lines of credit are more frequently used to finance takeover and acquisitions. Borrower, Lender, and Contract Characteristics: Borrowing Base vs. Other Secured Credi t Lines Table 2 5 compares the borrower, lender, and contractual terms of borrowing base ( BB ) lines of credit vs. other sorts of secured lines. Borrower characteristics are measured at the end of the fiscal year preceding the contract starting date to mit igate the endogenous effect of new credit lines on firm characteristics (such as leverage). I first investigate whether the borrowers of BB lines of credit differ from those of non BB lines in terms of risk, information asymmetry, asset structures etc. I t hen report on the lenders associated with these types of loans. Finally, I examine the difference between tranche contract terms, including the spread over LIBOR, credit limit, number of financial covenants, and maturity. Firms that use BB lines tend to be smaller and more risky than firms using secured, non BB lines. To be specific, BB line borrowers have smaller assets ($0.4 vs. $1.4 billion) as well as lower cash flows (EBITDA), more variable cash flows, and lower

PAGE 30

30 Z scores. They also have lower mark et to book (asset) ra tios and excess cash. One comparative strength of BB line borrowers is that they have on average lower total book debt over assets, but this could just indicate that their risk is too high to be granted more debt. Firms using BB lines of credit also tend to have a higher degree of information asymmetry. They spend more (on average) on R&D, more often trade over the counter (as opposed to trading on the NYSE, AMEX, or NASD exchanges), and are less likely to be included in a major S&P i ndex. Not surprisingly, BB borrowers have a much higher proportion of current assets (accounts receivable and inventory), and a correspondingly lower proportion of property, plant, and equipment (PP&E) and intangible assets. 9 Turning to the creditor cha racteristics, Carey, Post and Sharpe (1998) argue that banks and finance companies specialize in different types of loans. They categorize t based loans were once the province of finance companies, some of the large U.S. banks have moved into the business (Caouette, Altman, and Narayanan (1998, pages 96 97)). In my sample, U.S. banks serve as lead lenders less frequently for BB lines than f or non BB secured lines (64% vs. 76%). Correspondingly, finance companies are the lead lenders for more BB lines (26% vs. 8% for non BB lines). Finally, the BB and non BB lines differ in their contract terms. BB lines carry a 56 bps higher contract rat e and a shorter maturity, but fewer financial covenants. Although non BB lines have a much larger average commitment amount, adjusting the line 9 In un significantly lower turnover rate and are more volatile.

PAGE 31

31 amount for firm size yields virtually identical line commitments: 30.8% of noncash assets for BB borrowers vs. 30.5% for non BB borrowers. To sum up, the univariate comparison results suggest that BB lines of credit and non BB lines of credit are quite different in terms of borrower, lender, and contractual characteristics. I now turn to multivariate tests to c ompare the determinants of loan contract terms (spread, credit limit, and covenants) in BB vs. non BB credit lines. Multivariate Model Specification I wish to explain three dimensions of credit line terms: the interest rate, the maximum loan amount, and the extent of financial covenant restrictions imposed on the borrower. I therefore begin my discussion by specifying a separate regression for each of these loan terms: Spread j j sj ( 2 1) Amount j j aj (2 2 ) NumCov j j cj ( 2 3) Where Spread j = the natural log of the loan rate spread over LIBOR on line j, Amount j noncash assets NumCov j = the log of 1 plus the number of covenants in the contract for line j, and. stry dummies. Loan contract terms have been frequently studied in isolation from one another, although clearly these variables are co determined (as emphasized by Melnik and Plaut (1986) and Dennis, Nandy, and Sharpe (2000)). I therefore regress each con tract term

PAGE 32

32 on the same set of exogenous or predetermined explanatory variables (X) and interpret my results as a reduced form. 10 In order to compare the determinants of borrowing base ( BB ) vs. non borrowing base ( non BB ) credit line terms, I would like to estimate separate equations of 2 1 through 2 3 for sub sample. Comparing the impact of firm characteristics on the loan terms offered on each loan type then implies the effect of a BB restriction on the other loan terms. How ever, OLS estimates of 2 1 th rough 2 3 are inappropriate in this context, because the loan type choice may be an endogenous decision. Endogenous Switching I estimate an endogenous switching model to compare the determinants of loan characteristics (spread, credit limit, and financi al covenants) between borrowing base ( BB ) vs. non borrowing base ( non BB ) lines of credit. Using spread as an example, I formalize the model here and des cribe the relevant variables later line or I = 0 for a non BB line. The value of this indicator variable depends on an unobserved underlying variable (I I f I i i i > 0 then I = 1 ( 2 4 ) I f I i i i i = 0 (2 5 ) w here Z i i i > i by Equation 2 4 i is normally distributed its condition al expected value is 10 The estimated coefficients of different borrower/lender characteristics can be viewed as the aggregate direct and indirect effect (via other endogenous contract terms) on a specific contract term. For instance, risk could increase the spread and reduce the credit limit, while lower credit limit could result in a better price. Risk could then affect the spread directly or indirectly through credit limit. We do not distinguish the direct and indirect effect here.

PAGE 33

33 > 0 (Maddala (1983)). Likewise, the non i i by Equation 2 5 and hence its conditional expected value is i makes the firm more (less) likely to choose a BB line. I estimate the i conditional on the sort of loan selected. I estimate regress ion (1) separately for the set of BB and non BB lines: = 1 ( 2 6) = 0 ( 2 7) Here, Sbi (Sni) is the log of spread for BB (non BB) line of credit f or loan i. T he 4 and 2 5 determines whether I observe Sbi or Sni. Equa tions 2 6 and 2 7 will not necessarily have zero means. OLS does not recognize this feature of the regression residuals, and would produce biased coefficient estimates if the unobserved information ias can be addressed by including the IMR from ((4) (5)) among the explanatory variables in Xi (Lee (1978), Maddala(1983), Flannery and Houston (1999), and Booth and Booth (2006)) : S bi b X i 1n* bi for the BB l ines (I i = 1) ( 2 8) S ni n X i 0n* ni for the non BB lines (I i = 0) ( 2 9)

PAGE 34

34 The residuals in Equations 2 8 and 2 9 have zero conditional means and the estimated coefficient on the IMR measures the cova i bi ) i ni ). If the i is associated with higher unobservable credit r isk. A positi ve coefficient on the IMR in Equation 2 8 i bi are positively correlated, so a firm that surprisingly chooses a BB line (with higher unobservable credit risk) will also pay a higher spread. A positi ve coefficient on the IMR in Equation 2 9 i ni are positively correlated: a firm that surprisingly chooses a non BB line (with lower unobservable credit risk) will pay a lower spread. The same lines of logic apply to the coefficients on the IMR in the loan limit and financial covenant regressions in equations 2 2 and 2 3 Variable Definitions This section defines the explanatory variables (X, Z) used in the endogenous switching model. I bor rowing base ( BB ) vs. non borrowing base ( non BB ) credit line: I i i i ( 2 10) where I = 1 (0) if the firm takes a BB (non BB) line of credit, and Z i is a vector of borrower characteristics and control variables. The explanatory variables include several distinct groups. Credit Quality, made up of firm size, EBITDA as a proportion of non cash assets, the volatility of EBITDA, market to book asset ratio, leverage, and excess cash holdings. Asymmetric information, proxied by the extent of R&D expenditures, firm age, a dummy variable for the fi rm being traded off the major exchanges (= 1 if the firm is not

PAGE 35

35 traded on NYSE, AMEX, NASD), and another dummy variable equal to unity for firms that do not belong to a major S&P indices (the S&P500, the S&P Midcap400, and the S&P Smallcap 600). Asset c omposition including the proportion of non cash assets made up of accounts receivable, inventory, PP&E, and intangible assets. I also include year dummies and one digit SIC industry dummies, for completeness. I expect riskier and more opaque firms to rely more on the BB lines because the BB restriction provides lenders with stronger protections than a more general secured line does. Asset composition e.g. the prominence of accounts receivable or inventory should also affect the choice between the se two loan forms. My dependent variables in Equations 2 1 through 2 3 are constructed in following previous researchers. The loan rates S bi S ni are measured by the drawn all in spread above LIBOR, as in Booth and Booth (2006) or Carey and Nini (2007). The drawn all in spread is measured as amortized fees plus the annual spread paid above LIBOR for drawing down the maximum allowance on the line. 11 ( I abbreviate the drawn all in I take the log of spread to allow for a non linear re lation between the spread and explanatory variables. I cash assets, as in Sufi (2009). 12 11 Dealscan converts the spread indexed to other rates such as prime rate, or bank CD rates into spread to LIBOR. 12 Sufi (2009) argues that the decisions to maintain cash and to take a line of credit are likely to be determined jointly. He removes cash from his total assets def lator to avoid a mechanical negative correlation.

PAGE 36

36 than measuring either the spread or the l oan limit. I group the (fifteen) specific covenants identified by Dealscan into five general categories, to avoid double counting covenants of the same nature. The five categories are as follows. 1. x. 2. and Min. EBITDA. 3. Worth, Max. Leverage Ratio, Max. Loan to Value, and Max. Senior Leverage. 4. 5. Coverage, Min. Cash Interest Coverage, and Min. Interest Coverage. I set the aggregate covenant variable t o unity if any of its associated specific covenants is included in the contract, and measure the number of covenants as log of (1 + the sum of the five aggregate covenant variables). 13 Firm risk is one of the main variables of interest in both the select ivity model 2 10 and in determining the contr act terms (Equations 2 1 through 2 3 ). My definition of risk includes both operating risk and financial risk. 14 The natural logarithm of noncash total assets and the volatility of cash flows measure operating risk. Financial risk is proxied measured as the 4 year, trailing standard deviation of annual changes in EBITDA scaled by average non cash assets in the lagged period. Pr ofitability, growth 13 Results are robust to different measures of the number of covenants, including un logged number of categorized (specific) financial covenants and the log of (1 + total number of specific financial coven ants). 14 The distinction between operating risk and default risk is rather vague because the two risks interrelate with each other.

PAGE 37

37 I measure these cash to sured as total assets less the book value of equity plus the market value of equity less cash, all divided by non cash total assets. I construct a measure of excess cash as the difference between the actual cash holdings and predicted cash holdings by the model developed in Opler et al. (1999). Information asymmetry may also affect credit line selection and loan terms. I select proxies for information asymmetry following Faulkender and Peterson (2006) and Sufi (2009). They are R&D expense, firm age (th e natural logarithm of 0.1 plus years since a firm first appears in the Compustat database), 15 a dummy variable indicating whether the firm is not included in a major S&P indices (the S&P 500, the S&P Midcap 400, and the S&P Smallcap 600). Following the convention, I code the observations with missing R&D expense as zeroes and construct a dummy variable to indicate a missing value. I also control for different asset components i n my credit line choice model and loan term regressions, as they could affect the contract selection and loan terms via the collateral channel. Since accounts receivable and inventory are the most commonly used borrowing base assets, I include asset scale d accounts receivable and inventory. I also control for net property, plant and equipment (net PP&E) and intangible assets. When the value of intangible assets is missing, I set it equal to zero and define a 15 I did not calculate firm age as years after IPO because the IPO date of many firms is missing.

PAGE 38

38 dummy variable equal to unity for this observa tion (otherwise zero). I scale all asset components with non cash total assets. tend to make riskier loans, I d ummy variables identifying whether the lead lender is a U.S. bank, a finance company, a foreign bank, or some other type of institution in my loan term regressions. 16 I omit the dummy identifying lead lenders that are U.S. banks. To identify the endogen ous switching model, I include the industry median of asset scaled accounts receivable and inventory by two digit SIC code in the credit line selection model, but not in the second stage regressions. Industry median of accounts receivable and inventory co uld be associated with the usage of BB line of credit in the industry, and therefore the contract selection decision of a borrower in the industry. Industry median of accounts receivable and inventory, however, should have reasonably low correlations with the contract terms of individual borrowers. I also include lender identi fiers in the regression 2 1 through 2 3 but not in the selection model 2 10 Empirical Results Contract Selecti on: Borrowing Base vs. Other Secured Credit Lines Table 2 6 presents t he estimated marginal effects of firm characteristics on the probability of choosing a borrowing base ( BB ) line of credit. To compare the economic 16 If any lender is define d as "Admin agent" or "Agent" or "Arranger" or "Book runner" or "Lead arranger" or "Lead bank" or "Lead manager" then he is defined as a lead lender. If the lead lender group has a higher proportion of U.S banks or foreign banks or other institutions, then the facility is defined as the U.S. bank or foreign bank or other institution initiated line of credit. I also code lender types as the percentage of U.S banks, financial companies, foreign banks, and other institutions among all lenders My results are a lmost identical.

PAGE 39

39 significance of variables measured in different scales, I standardize these marginal effects to indicate the change in the probability of choosing a BB line when a continuous independent variable changes by one standard deviation or a dummy variable changes from 0 to 1. Standard errors have been adjusted to account for heteroskedasticity and borrower level error correlations. Consistent with my hypothesis, borrowers of BB lines have worse credit prospects. Specifically, smaller firms with more volatile cash flows and worse financial ratios (lower score (higher degree of financial distress), fewer gr owth opportunities, and less excess cash) are more likely to take a BB line of credit. These effects are economically substantial. The unconditional probability that a firm has a BB line is 41%. A one standard deviation increase in log of non cash total assets is associated with nearly an 11% decrease in the probability of obtaining a BB line of credit. A one standard deviation increase in EBITDA is associated with an 8% decline in BB probability. Asymmetric information also plays a role: moving off a stock exchange into an OTC trading environment increases the probability of a BB loan by almost 6%. More assets. 17 Taken together, my probit results suggest that a BB line of credit is more suited to borrowers with a higher risk profile. 17 In another probit specification, I control for inventory and accounts receivable volatility and turnover rates and find no qualitative effect on my reported results.

PAGE 40

40 Spread, Credit Li mit, and Financial Covenants: Borrowing Base vs. Other Secured Credit Lines I characteristics affect the contract terms of borrowing base ( BB ) vs. non borrowing base ( non BB ) lines. The determinants of loan spread are estimated in columns (1) and (2) o f Table 2 7 Column (3) reports t statistics for the hypothesis that the BB and non BB coefficients are equal. Not surprisingly, the loan spread increases significantly with leverage, and decreases with firm size, EBITDA, Z score and market to book asset ratio. However, the BB spread is much less sensitive to firm risk characteri stics than is the non BB spread. (See the significant t statistics in column (3) for LN(non cash total assets), cash flow, cash flow volatility, Z score, and market to book). These differences are both economically and statistically significant. For exa mple, the coefficient on leverage (Total debt / total assets) in the non BB spread regression is more than twice as large as that in the BB regression (0.42 vs. 0.17). And CF volatility has a significantly positive coefficient (1.17) for non BB loans, but an insignificant coefficient (0.15) in the BB regression. The explanation for this apparent paradox may lie in the coefficients on accounts receivable and inventory, which are the most common borrowing bases. These variables significantly raise spread f or non BB loans, but (weakly) lower spread for BB loans. Among lead lenders, the (omitted) US banks seem to offer relatively low pricing on both types of loans. The IMR in the BB spread regression (column (1)) does not differ significantly from zero, in dicating no correlation between the unobservable determinants of line choice and the BB spread. However, the IMR coefficient in the non BB spread regression is significantly positive, indicating that borrowers choosing the non BB loan

PAGE 41

41 by surprise pay a lo wer spread, ceteris paribus. (Recall that the IMR for these borrowers are all negative.) The banking literature also maintains that riskier borrowers receive lower credit limits, and I find support for this hypothesis i n columns (4) and (5) of Table 2 7 For both sorts of secured credit line the credit limit declines with risk, as measured by Z score, market to book (growth opportunities), and leverage. The line limits exhibit lity, but not to other risk measures. Non BB line limits are negatively related to accounts receivable and inventory, and this effect differs significantly from the weakly positive and positive impact of accounts receivable and inventory on BB limits. Fi nance company lead lenders offer BB borrowers significantly larger limits than other lenders. The IMR coefficient is again zero for BB lines, but its coefficient for non BB line limits is significantly negative. This sign is consistent with the IMR coeff icient in the non BB spread regression: surprisingly choosing a non BB line indicates something positive Columns (7) and (8) of Table 2 7 report regression results for the number of covenants associated with a line. For non BB borrowers, larger firms with higher market to book are subject to significantly fewer covenants. These effects are significantly stronger than for BB lines. Fewer covenan ts are also present for secured loans to firms with more excess cash, more R&D expenditures, and for those traded over the counter, but these effects do not differ between the two loan types. A high concentration of accounts receivable and inventory has c ontrasting effects: raising the

PAGE 42

42 number of covenants for non BB lines and significantly lowering them for BB lines. (The coefficients differ significantly across equations). Coefficients in columns (7) and (8) indicate that finance companies impose relat ively fewer covenants than other lenders. This is consistent with them charging higher spread as in columns (1) and (2). The most surprising feature of the covenant regressions is that some weaker BB firms are subjected to fewer covenants. This is implied directly by the positive coefficients on EBITDA and Z score. The significantly negative coefficient on IMR also indicates that strong looking firms are subject to significantly fewer covenants when they choose a BB line. These results may reflect the re lative importance of collateral vs. future cash flows in repaying a BB line. For a firm with weak financial ratios, the lender looks primarily to the underlying collateral for repayment and hence does not care about restrictive covenants. Cash flows may be an important component of expected repayment for stronger firms, and financial covenants would permit the lender to re negotiate loan terms if cash flows fall short of expectations. To test this possibility, I divide BB borrowers into those below vs. ab ove median EBITDA and re estimate the regressions in columns (7) and (8). Consistent with my hypothesized explanation, only the low cash flow subsamples have significantly positive coefficients on EBITDA and Z score. For high cash flow firms, these var iables carry significantly negative or insignificant coefficients. Predicted Contract Terms by Selected Line of Credit Type In this subsection, I borrowing base ( BB ) vs. non borrowing base ( non BB ) lines seem ra tional. Following Booth and Booth (2006), I financial covenants) under their chosen type of credit line to the terms predicted if the

PAGE 43

43 borrower had chosen the other (counter factu al) type of c redit line. Table 2 8 is based on predictions from the endogenous switching model reported in Table 2 9. Table 2 10 presents the (inappropriate) OLS predictions for comparison. The first column of Table 2 8 reports the estimated mean contr act terms for BB lines among firms who actually take the BB contract. The second column reports the mean fitted contract terms, assuming that the firm had taken (counterfactually) a non BB line. For all three contract terms, the average BB borrower is be tter off with a BB loan: this choice gives them a lower spread (263 vs. 821), higher line limit (0.308 vs. 0.03), and fewer covenants (1.55 vs. 4.48). The same is true for the mean non BB borrower in Table 2 8 Their choice provides them with a signifi cantly lower spread, significantly higher loan amount, and significantly fewer covenants than they would have had with a BB line. In other words, I find that firms select into a line of credit contract rationally to get better contract terms. Even though BB borrowers are much riskier, the protection provided by a BB restriction affords them credit on terms similar to those provided to (less risky) non BB borrowers. Ta ble 2 9 reports similar calculations using OLS estimates of (1), (2), and (3) to provid e information about counter factual choices. Recall that these estimates are likely to be biased if the line choice reflects unobserved information about the borrowers. Nevertheless, I report these results for comparison with the (more appropriate) resul ts in Table 2 8 For non BB borrowers, Table 2 9 provides similar conclusi ons to those reported in Table 2 8 : these firms paid a lower spread, received a higher loan limit, and were subject to fewer covenants by eschewing the BB line. The OLS results fo r BB borrowers contradicts the results from endogenous switching model estimates,

PAGE 44

44 indicating that these firms could have had a lower spread and fewer covenants had they chosen the counter factual. I conclude, once again, that controlling for endogenous li ne choice is an important part of my estimation methodology. Chapter 2 Concluding Remarks I construct a sample of secured credit lines extended to U.S. corporations during the period 1995 2008. All lines of credit specify a maximum loan size, but some secured lines include a borrowing base (BB) restriction, which also ties available credit to a proportion of specified on book assets (most often accounts receivable or inventory). I find that firms choose purposefully between the two types of secured loa ns. BB loans are more likely to be taken by smaller and riskier firms with less cash flow, more cash flow volatility, and lower Z score. BB borrowers also tend to have a lower market to book ratio and cash balances, and a higher concentration of accounts receivable and inventory. Controlling for sample selectivity effects, I examine the determinants of three covenants. Loan spreads are less sensitive to the firm rise with leverage and fall with cash flow, Z score, and size, but these effects are significantly smaller for the BB lines. Th is is consistent with the popular/practitioner view that BB loans provide a borrower with improved collateral protection, although they also require higher monitoring costs. (Apparently, the cost of additional monitoring does not offset the reduction in ex pected losses associate with that monitoring.) The with the notion that lenders are looking primarily to collateral as the source of

PAGE 45

45 repayment. For example, cash flow increases the credit limit for non BB lines but has no effect on the limit for BB lines. Finally, I find that the BB restriction substitutes for more typical loan covenants. Borrowing firms select into the type of loan that is more advantageous fo r them. According to my estimates, if BB borrowers were to give up the BB restriction and select another type of secured line, they would pay a substantially higher spread, receive a substantially lower credit limit, and be subject to more restrictive cov enants. Whereas Sufi (2009) reports that firms with low cash flow are more likely to lose access to their lines of credit, I find that one specific sort of credit line one tied to a borrowing base remains open even to firms with poor financials and l ow cash flows. These loans are taken primarily by smaller firms, whose open market borrowing alternatives are most limited. The BB restriction causes the amount lent to vary dependent cov enant condition weakens.

PAGE 46

46 Table 2 1 Sample selection procedure This table presents the sample selection procedure. My sample covers only loan tranches for U.S. borrowers with an origination date between 1995 and 2008. Loans extended to financial firms are excluded in my sample My final sample consists of 5,154 secured lines of credit, whose borrowers can be identified in the Compustat database and have full ac counting inf ormation available Frequency Total Amount (in million dollars) Average Amount (in million dollars) Average Spread (in basis points) Tranches from Dealscan 67,398 13,685,844 203 238 Tranches merged with CRSP Compustat 26,069 7,230,3 85 277 205 Term loans 6,841 1,377,136 201 286 Others 1,588 610,360 385 238 Lines of credit 17,640 5,242,889 297 173 Lines of credit with unspecified security status 5,667 2,294,461 404 113 Unsecured lines of credit 3,327 1,830,328 550 79 Dealscan Compustat matched sample Secured lines of credit 8,646 1,118,100 129 243 Bo rrowing base line of credit 3,316 262,730 79 279 Firms with asset<$500 million 2,725 92,882 34 283 Non BB line of credit 5,330 855,370 160 224 Firms with asset<$500 million 3,386 200,373 59 239 Full accounting information sample Secured lines of credit 5,154 678,761 132 243 Borrowing base line of credit 2,117 160,744 76 275 Firms with asset<$500 million 1,740 61,279 35 284 Non BB line of credit 3,037 518,017 171 220 Firms with asset<$5 00 million 1,850 109,897 76 234

PAGE 47

47 Table 2 2 B orrowing base line of credit contractual characteristics by asset types. This table presents contractual characteristics of BB line of credit. Sample contains 2,117 lines of credit whose borrowers are identifi ed in Compustat database and have full accounting information available It shows frequency and advance rate for different ty pes of borrowing base assets. I categorize BB assets into four groups, cash & marketable securities, accounts receivable, inventory and fixed assets. The frequencies within each group do not add up to the group total because the group total counts a BB line of credit using multiple borrowing base assets within the same group as one observation Asset Type Frequency Frequency Percentag e Advance Rate Cash & cash equivalents 47 2.2% 97.3 Marketable securities 10 0.5% 77.8 Cash & marketable securities 55 2.6% 93.9 Accounts receivable foreign 79 3.7% 73.9 Accounts receivable domestic 80 3.8% 79.3 Eligible accounts receivable 1,755 82.9% 81.0 Accounts receivable 1,841 87.0% 80.6 Inventory raw material 146 6.9% 44.8 Inventory work in progress 53 2.5% 39.1 Inventory finished goods 142 6.7% 55.0 Eligible inventory 1,206 57.0% 55.4 Inventory 1,341 63.3% 53.8 Property, plant & equipment 99 4.7% 63.8 Eligible property value 35 1.7% 58.2 Oil & gas reserves 21 1.0% 83.1 Fixed assets 145 6.8% 65.2 Unknown 61 2.9% Total BB lines of credit 2,117 100%

PAGE 48

48 Table 2 3 B orrowing base line of credit contractual characteristics by i ndustry This table presents contractual characteristics of BB line of credit. Sample contains 2,117 lines of credit whose borrowers are identified in Compustat database and have full accounting information available. It reports the frequency of BB line of firms in Compustat database BB Firms Compustat Universe Firms Industry Frequency Frequency Percentage Frequency Frequency Percentage Manufacturing 1,084 51.2% 50,872 43.7% Services 352 16.6% 25,286 21.7% Trade retail 274 12.9% 7,285 6.3% Trade wholesale 183 8.6% 4,418 3.8% Transportation, communications, and Utilities 109 5.1% 14,176 12.2% Mining 62 2.9% 10,594 9.1% Construction 29 1.4% 1,331 1.1% Agriculture, forestry, and f ishing 17 0.8% 471 0.4% Nonclassifiable establishments 7 0.3% 2,034 1.7% Total 2,117 100% 116,467 100%

PAGE 49

49 Table 2 4 B orrowing base line of credit contractual characteristics by loan purpose This table exhibits the frequency of BB line of credit by lo an purpose in comparison with the frequency of non BB secured line of credit BB Non BB Tranche Purpose Frequency Frequency Percentage Frequency Frequency Percentage Working capital 647 30.6% 717 23.6% Debt repayment 603 28.5% 834 27.5% Corporate purpo ses 530 25% 586 19.3% Takeover and acquisition 128 6% 398 13.1% Commercial paper backup 70 3.3% 165 5.4% Debtor in possess 60 2.8% 54 1.8% LBO 25 1.2% 101 3.3% Others 54 2.6% 182 6% Total lines of credit 2,117 100% 3,037 100%

PAGE 50

50 Table 2 5 D escripti ve statistics by line of credit type This table displays the borrower, lender, and contractual characteristics of BB lines of credit versus non BB lines of credit My sample includes 2,117 BB lines and 3,037 Non BB lines. All firm characteristics are obt ained at the end of fiscal year preceding to the contract starting date. Variables ended with w5 are winsorized at the 5th and 95th percentile, following Sufi (2009). T test and Wilcoxon rank sum test are used to test the significance of mean and median be tween BB line of credit and non BB line of credit regarding borrower, lender, and contractual characteristics. ***, **, and denote statistically significan t at the 1%, 5%, and 10% levels BB NON BB Difference: BB Non BB Mean Median Std. Dev Mean M edian Std. D ev Mean Median Borrower Characteristics Total assets (million dollars) 402.0 117.6 1238.5 1355.9 321.3 4290.1 953.9*** 203.6*** Non cash total assets (million dollars) 370.1 105.2 1144.4 1276.4 292.0 4118.1 906.3*** 186.8** EBITDA/Non cash total assets w5 0.07 0.09 0.13 0.13 0.14 0.10 0.06*** 0.05*** CF volatility w5 0.11 0.07 0.10 0.07 0.05 0.08 0.04*** 0.02*** Altman's Z score w5 2.91 2.57 2.33 3.23 2.65 2.51 0.33*** 888 0.08*** Market to book asset ratio w5 1.65 1.23 1.10 1.80 1.46 1.03 0.15*** 0.23*** Total debt/Total assets w5 0.28 0.26 0.20 0.31 0.30 0.21 0.02*** 0.04*** Excess cash 0.77 0.73 1.47 0.64 0.53 1.43 0.13*** 0.20*** R&D/Non cash total assets w5 0.04 0 0.07 0.02 0 0.05 0.02*** 0*** R&D missed 0.43 0 0.50 0.51 1 0.50 0.08*** 1*** Firm age 12.24 8 11.66 13.47 9 13.28 1.23*** 1 Traded over the counter 0.35 0 0.48 0.19 0 0.39 0.16*** 0*** Not included in a major S&P index 0.85 1 0.35 0.74 1 0.44 0.11*** 0*** Accounts rec /Non c ash total assets w5 0.24 0.23 0.14 0.18 0.16 0.13 0.06*** 0.07*** Inventory/Non cash total assets w5 0.23 0.22 0.16 0.13 0.07 0.14 0.10*** 0.15*** Accounts receivable volatility 0.08 0.06 0.07 0.06 0.04 0.06 0.02*** 0.02*** Inventory volatility 0.06 0.0 5 0.06 0.03 0.02 0.04 0.03*** 0.03*** Accounts receivable/Sales 0.16 0.16 0.09 0.16 0.16 0.09 0 0 Inventory/Sales 0.14 0.14 0.10 0.09 0.07 0.09 0.05*** 0.07*** PP&E ./Non cash total assets w5 0.27 0.22 0.19 0.37 0.30 0.26 0.10*** 0.08*** Intang. asset s/Non cash total assets w5 0.11 0.03 0.14 0.16 0.08 0.19 0.05*** 0.05*** Intangible assets missed 0.13 0 0.34 0.13 0 0.34 0 0

PAGE 51

51 Table 2 5 Continued BB NON BB Difference: BB Non BB Mean Median Std. Dev. Mean Median Std. Dev. Mean Median Le nder Type Lead lender U.S. bank 0.64 1 0.48 0.76 1 0.43 0.12*** 0*** Lead lender finance company 0.26 0 0.44 0.08 0 0.27 0.18*** 0*** Lead lender foreign bank 0.06 0 0.24 0.11 0 0.31 0.05*** 0*** Lead lender other institutions 0.03 0 0.18 0.0 5 0 0.22 0.02*** 0*** Contractual Characteristics Loan spread 275.4 260 107.8 219.7 200 114.4 55.7*** 60*** Credit limit (million dollars) 75.9 25 172.9 170.6 75 337.1 94.7*** 50*** Credit limit/Non cash total assets w5 0.308 0.26 0.21 0.30 5 0.23 0.25 0.003 0.03*** Number of financial covenants 1.78 2 1.20 1.99 2 1.24 0.21*** 0*** Maturity 2.81 3 1.42 3.65 3.45 1.82 0.84*** 0.45***

PAGE 52

52 Table 2 6. Probit model for credit line selection This table presents the estimated marginal effects of lagged fiscal year end firm characteristics on the probability of using a BB line of credit. The marginal effects are calculated as the change in the probability of choosing a BB line for a standard deviation change in each independent, continuous variable and, the discrete change in the probability when a dummy variable switches from 0 to 1. All regressions are run using robust standard errors that allows for heteroskedasticity and within firm error term correlations. ***, **, and denote statistically significant at the 1%, 5 %, and 10% levels, respectively Dependent Variable Marginal Effects LN (Non cash total assets) 10.82*** EBITDA/Non cash total assets w5 8.24*** CF volatility w5 3.17*** Altman's Z score w5 2.98** Market to book asset ratio w 5 4.60*** Total debt/Total assets w5 0.07 Excess cash 1.85* R&D/Non cash total assets w5 1.10 R&D missed 2.73 LN(Firm age) 0.40 Traded over the counter 5.87*** Not included in a major S&P index 3.79 Accounts receivable/Non cash total assets w5 9.21*** Inventory/Non cash total assets w5 10.11*** Industry median accounts receivable 4.40*** Industry median inventory 5.01** Property, plant and equip./Non cash total assets w5 0.93 Intangible assets/Non cash total assets w5 3.50** Intangible as sets missed 0.69 Year dummies Y Industry dummies Y Pseudo R 2 0.216 # of Observations 5,154

PAGE 53

53 Table 2 7. Switching regression of spread, credit limit, and financial covenants Columns (1) and (2) exhibit the estimates in the second step of endogenous s witching model for log (drawn all in spread), column (4) and (5) are for credit limit/non cash total assets, and column (7) and (8) for log (1+number of financial covenants). Column (3), (6), and (9) display the t stats for the difference in coefficients o f BB and non BB spread, credit limit, and financial covenants. Inverse Mills Ratio is from the Probit model estimated in Table IV, which a re f( )/F( ) for the BB line of credit and f( )/(1 F( )) for non BB line of credit. All firm characteristics are obtained at the end of fiscal year proceeding to the contract starting date. Variables ended with w5 are winsorized at the 5th and 95th percentile, following Sufi (2009). All r egressions are run using robust standard errors that allows for heteroskedasticity and within firm error term correlations. ***, **, and denote statistically significant at the 1%, 5%, and 10% levels, respectively. LN(drawn all in spread) Credit limit/non cash total assets LN(1+number of fin. covenants) BB NON BB T stat BB NON BB T stat BB NON BB T stat (1) (2) (3) (4) (5) (6) (7) (8) (9) LN (Non cash total assets) 0.062*** 0.160*** 5.10 0.046*** 0.032*** 1.48 0.026 0.032** 2.81 EBI TDA/Non cash total assets w5 0.551*** 1.139*** 2.69 0.035 0.357*** 2.93 0.892*** 0.276 2.55 CF volatility w5 0.151 1.167*** 4.76 0.028 0.282*** 2.06 0.276 0.207 0.27 Altman's Z score w5 0.019*** 0.036*** 1.89 0.006* 0.013*** 1.55 0.021*** 0.0 06 1.43 Market to book asset ratio w5 0.004 0.076*** 3.87 0.022*** 0.033*** 1.06 0.017 0.049*** 2.88 Total debt/Total assets w5 0.174*** 0.420*** 2.72 0.096*** 0.103*** 0.15 0.096 0.029 0.66 Excess cash 0.003 0.006 0.80 0.002 0.004 1.08 0.031* ** 0.019** 0.95 R&D/Non cash total assets w5 0.384** 0.150 1.61 0.345*** 0.397*** 0.34 0.988*** 1.045*** 0.16 R&D missed 0.002 0.048* 1.39 0.049*** 0.012 2.20 0.052* 0.003 1.27 LN(Firm age) 0.013 0.006 1.26 0.010* 0.018*** 1.03 0.009 0.0 04 0.32 Traded over the counter 0.078*** 0.220*** 4.16 0.009 0.040*** 2.65 0.112*** 0.045 1.52 Not included in a major S&P index 0.080*** 0.105*** 0.61 0.024* 0.009 0.93 0.021 0.051** 1.66 Accounts receivable/Non cash total assets w5 0.125 0.517** 2.70 0.118 0.169* 2.39 0.354* 0.599*** 3.73 Inventory/Non cash total assets w5 0.226* 0.478** 2.81 0.317*** 0.147* 3.97 0.304* 0.288 2.25 Property, plant and equip./Non cash total assets w5 0.026 0.183** 1.32 0.059 0.007 1.09 0.106 0.029 0.53 Intangible assets/Non cash total assets w5 0.244*** 0.294*** 4.00 0.106* 0.061 2.26 0.276** 0.098 1.08 Intangible assets missed 0.017 0.056* 1.71 0.033** 0.022 2.33 0.049 0.061* 0.23 Lead lender finance company 0.115*** 0.287*** 4.2 6 0.028** 0.021 2.53 0.263*** 0.171*** 1.88 Lead lender foreign bank 0.013 0.086*** 1.97 0.009 0.019 1.11 0.094* 0.039 0.88 Lead lender other institutions 0.172*** 0.235*** 0.90 0.009 0.030* 1.25 0.080 0.084* 0.05 Inverse mills ratio 0.09 5 0.691*** 4.80 0.023 0.202*** 2.15 0.247* 0.498*** 4.20 Constant 5.903*** 6.281*** 1.34 0.356*** 0.331*** 0.19 1.454*** 0.579** 2.59 Year dummies Y Y Y Y Y Y Industry dummies Y Y Y Y Y Y R Sq uare 0.383 0.341 0.267 0.244 0.233 0.214 # Observations 2,117 3,037 2,117 3,037 2,117 3,037

PAGE 54

54 Table 2 8. Predicted spread, credit limit, and number of financial covenants under actual and alternative line of credit contract (switching model) This table reports predictions using the endoge nous switching model estimates of spread, asset scaled credit limit, and number of financial covenants. ***, **, and denote that pairwise t test of the differences is statistically significant at the 1%, 5%, and 10% levels, respectively BB Lines of Cre dit Non BB Lines of Credit Estimated mean value a under actual contract Estimated mean value if BB borrowers had chosen a non BB line Estimated mean value b under actual contract Estimated mean value if non BB borrowers had chosen a BB line Drawn all in spread 263 821*** 202.5 260*** Credit limit/non cash total assets 0.308 0.03*** 0.305 0.269*** Number of fin. covenants 1.55 4.48*** 1.73 3.37*** a, b, c, d .The mean estimated spread and number of financial covenants differ slightly from the means r eported in Table III because the estimates here come from regressions that fit the logs of these variables rather than the variables themselves.

PAGE 55

55 Table 2 9 Predicted spread, credit limit, and number of financial covenants under actual and alternativ e line of credit contract ( OLS model) This table reports predictions using the OLS model estimates of spread, asset scaled credit limit, and number of financial covenants. ***, **, and denote that pairwise t test of the differences is statistically signi ficant at the 1%, 5%, and 10% levels, respectively BB Lines of Credit Non BB Lines of Credit Estimated mean value a under actual contract Estimated mean value if BB borrowers had chosen a non BB line Estimated mean value b under actual contract Estima ted mean value if non BB borrowers had chosen a BB line Drawn all in spread 263 260*** 202.4 222*** Credit limit/non cash total assets 0.308 0.306 0.305 0.230*** Number of fin. covenants 1.55 1.40*** 1.73 1.88*** a, b, c, d .The mean estimated spread and number of financial covenants differ slightly from the means reported in Table III because the estimates here come from regressions that fit the logs of these variables rather than the variables themselves.

PAGE 56

56 CHAPTER 3 THE JOINT DETERMINAT ION OF LEVERA GE RATIO, DEBT MATUR ITY, AND DEBT SOURCE debt proportion are jointly determined. Although corporate debt has many dimensions, the substantial literature on capital structur e decisions generally examines only one facet of capital structure at a time. The debt equity decision is the most extensively studied area (Myers (1977), Myers (1984), Myers and Majluf (1984), etc.). Other scholars have extensively explored debt maturity ((Brick and Ravid (1985), Flannery (1986), Diamond (1991a), etc.) and debt source (Diamond (1984, 1991b), Rajan (1991), Gertner and Scharfstein (1991), etc.). The study of one capital structure decision in isolation is, however, at odds with the fact that firms simultaneously decide how much debt to take out (debt equity decision), how soon to repay the debt (debt maturity), and from whom to borrow (debt source). Among the few papers that identify the joint determination of capital structure decisions, Jo jointly determined. He argues that maturity, in addition to leverage ratio, can be used to mitigate the underinvestment problem. Johnson (2003) reports evidence that the negative rel ation between leverage ratio and growth opportunities is less substantial for firms financed with short term debt, relative to those that finance with long term debt. Billett et al. (2007) extends Johnson (2003) by including covenant structure into the end ogenous decision system. They find that covenants can substitute for a short debt maturity and a low leverage to mitigate the agency problem of high growth firms. While Johnson (2003) and Billett et al. (2007) advanced the literature on the joint decision of different capital structure facets, they left out one important aspect of the

PAGE 57

57 capital structure decision: whether to borrow privately or publicly (debt source). This is especially problematic since debt maturity and covenant structure are strongly corr elated with whether debt is publicly or privately placed. Barclay and Smith (1996) report that private debt, and in particular bank debt has a shorter maturity and more covenants. Thus, leaving debt source out of the joint determination system will not giv e This study is the first empirical investigation that explores the source of debt in the joint capital structure decisions. In particular, I examine the role of bank financing in the joint capital structure decisions. Previous studies argue that bank debt is special. Bank information production and monitoring services. Bank debt can also provide flexibility in financial d istress through loan renegotiation. James (1987), Lummer and McConnell (1989), Best and Zhang (1993), and Billett et al. (1995) all document a positive share price reaction to bank loan announcements, suggesting that bank debt is viewed favorably by the ma rket. Drawing inferences from the theoretical literature, I hypothesize that leverage ratio, debt maturity, and bank debt proportion can all be used to mitigate conflicts between 1 The un derinvestment problem in Myers (1977) and asset substitution problem in Jensen and Meckling (1976) are two types of agency problems. Both papers predict that firms may use less leverage or even no debt in their capital structure to avoid the potential conf licts. Alternatively, other theoretical papers argue that a shorter debt maturity can 1 I intend to add covenants into the joint system once I obtain a pro per proxy.

PAGE 58

58 reduce the underinvestment problem (Myers (1977)) and asset substitution problem (Leland and Toft (1996)). Furthermore, some theories suggest that bank debt alone can mi tigate these agency problems: the flexible renegotiation of bank debt can alleviate the underinvestment problem (Myers (1977)), and bank monitoring can reduce the asset substitution problem (Diamond (1984, 1991b), Fama (1985) etc.). Alternatively, Rajan (1 992) argues that banks with information monopolies can worsen conflicts between project return and can extract part of the surplus. This possibility limits entrepreneurial ef fort to an inefficiently low level. If these theories hold, firms with severe agency Additionally, the use of bank debt will play an important role in the leverage decision. Bank debt may either mitigate (Myers (1977)) or amplify (Rajan (1992)) the effect of agency problems on leverage reduction. I also posit that in addition to the leverage ratio and debt maturity, bank debt can be chosen to avoid financial distress. Firms i n financial distress, where they have trouble paying off their financial obligations, can incur direct bankruptcy costs and other indirect costs (e.g. impaired access to credit) (Opler and Titman (1994)). Therefore firms close to financial distress reduce their leverage to avoid these costs. Long debt maturity incurs less liquidity risk and rollover risk, and decreases the likelihood of financial distress. Bank debt also reduces the likelihood of financial distress by allowing for flexible loan renegotiatio n. The theory predicts that firms near financial distress do not need to reduce leverage by as much if they use debt with a longer ma turity and/or more bank debt.

PAGE 59

59 Following Johnson (2003), I use a system of simultaneous equations to capture the simultaneou s decisions about leverage, debt maturity and bank debt proportion. As in Johnson (2003), I measure leverage as the market leverage and debt maturity as the proportion of debt that matures within three years. I proxy bank debt as the proportion of debt rec orded in Capital IQ as revolving lines or term loans. I use growth opportunities as the proxy for the potential conflicts of interest between equity holders and debt holders, given that firms with more growth opportunities are more likely to be subject to the underinvestment problem and asset substitution. I use z likelihood of financial distress. I include interaction terms into the leverage ratio equation to test the agency problem and financial distress hypotheses. A positive coef ficient on the interaction of growth opportunities and short term (bank) debt suggests that short term (bank) debt can mitigate the agency problem and firms increase the leverage ratio accordingly. Similarly, a positive (negative) coefficient on the intera ction of z score and short term (bank) debt suggests that short term (bank) debt can increase (decrease) the probability of financial distress and firms reduce the leverage ratio accordingly. 2 The empirical results suggest that short term debt could reduc e conflicts between debt holders and equity holders, but increases the likelihood of financial distress. The coefficient on the interaction of growth opportunities and short term debt is significantly positive, consistent with the finding of Johnson (2003) The positive and significant coefficient on the interaction of short term debt and z score indicates that short term debt increases the likelihood of financial distress and reduces the leverage ratio. Firms 2 Recall that firms with a lower z score are more likely to be financially distressed.

PAGE 60

60 without a credit rating could have severe agenc y problems and are more likely to be financially distressed than rated firms due to information asymmetry and limited access to non bank debt markets. I show that rated firms can use short term debt to reduce agency problems without incurring too much fina ncial distress risk. Using more short term debt in general reduces the leverage ratio among the unrated firms because the excess liquidity and financial distress risk outweighs the benefit of agency problem reduction. Bank debt, however, amplifies the nega tive correlation between leverage ratio and growth opportunities, my proxy for agency problems. This effect is concentrated mostly among the unrated firms, who tend to have limited access to non bank funding due to their information asymmetry problems or h igh credit risk. These findings are consistent with the argument of Rajan (1992) that bank debt incurs additional conflicts of interest between equity holders and bond holders due to their information monopoly. Houston and James (1996) also find empirical support for Rajan (1992). They show that bank debt proportion decreases with growth opportunities among firms with single bank relationships; among firms borrowing from multiple banks, the relationship is positive My results show that there is only weak e vidence that bank debt reduces the negative The results suggest that debt maturity, debt source, and leverage ratio are jointly determined. Debt maturity and bank debt can alte r the relation between the leverage ratio and two types of borrowing costs: agency costs related to conflicts between equity holders and debt holders, and costs resulting from financial distress risk. The results

PAGE 61

61 suggest that in addition to leverage ratio, debt maturity, and debt source have an important effect on the distortions associated with debt finance. The remainder of this study proceeds as follows. I first present testable predictions for the relation between leverage ratio, short term debt, and b ank debt, and describe the econometric model used to test these relations. Next I discuss the data and proxies used in the analysis and reports descriptive statistics for the variables used in the econometric model. The following section presents results e stimated from the simultaneous equations of leverage, short term debt, and bank debt. Finally, I Agency Problem, Financial Distress and Capital Structure Decisions I organize this section as follow s. To get prepared to develop my main hypotheses, I will start out discussing how leverage ratio, debt maturity, and debt source emerge as endogenous variables to the costs and (or) benefits of borrowing respectively. In particular, I investigate how agenc y cost of debt and financial distress cost affect these endogenous decisions. Such discussions naturally lead us to my main hypotheses about the joint determination of leverage ratio, debt maturity, and debt source, and their predicted relation, given that these three variables are all endogenously driven by the similar cost and benefit considerations. To test my hypotheses, the econometric model must accommodate a system of three equations simultaneously. Towards this end, I will discuss how to estimate th e system of simultaneous equations econometrically.

PAGE 62

62 Leverage, Agency Problem, and Financial Distress Trade off theory argues that firms choose an optimal leverage ratio by trading off the costs and benefits of debt financing. 3 It suggests that firms wit h higher costs of debt will use less debt financing. Agency cost is an important component of debt financing costs proposed by theories. Agency cost of debt stems from the conflicts between shareholders and debtholders and distorts firm investment decision s. One type of agency cost is the underinvestment problem. Myers (1977) shows that firms financed with risky debt may pass by valuable investment opportunities because of wealth transfer from equity holders to debt holders, the so called debt overhang prob lem. Jensen and Meckling (1976) point out another type of agency cost, asset substitution problem. i.e. shareholders may substitute riskier assets for safer assets to maximize their residual claim even though projects are value decreasing. Firms with high er agency cost should use less leverage according to the trade off theory. Financial distress cost is another important type of borrowing cost. I define financial distress as the situation where the borrower has trouble paying off its financial obligations credit, increase competition, or lose nonfinancial stakeholder relationships (Opler and Titman (1994)). Firms close to being financially distressed should then reduce the leverage ratio to avoid bankruptcy and other indirect costs. Together, theories generate the following prediction. 3 Graham and Harvey (2001) show that 81% of surveyed firms have a target debt ratio or range when making their debt decisions.

PAGE 63

63 H1: A lower leverage ratio reduces the agency problem and the likelihood of financial distress. Short term Debt, Agency Problem, and Financial Distr ess Short term debt could reduce the agency cost of debt. Myers (1977) suggests that by having short term debt that matures before the growth options expire, managers on behalf of shareholders can exercise profitable investment opportunities and alleviate the underinvestment problem. 4 and Toft (1996) argue that using short commit asset substitution because increasing risk in short term does not benefit shareholders. Short term debt, however, could increase the likelihood of financial distress. Diamond (1991a) states short term debt exposes the firm to early liquidation risk even though short term debt allows for a reduction in borrowing costs when a fir m receives good news in the future. Together, theories generate the following prediction. H2: Short term debt can mitigate the agency problem but increase the likelihood of financial distress. Bank Debt, Agency Problem, and Financial Distress Theoretical papers argue that bank debt could either mitigate or exaggerate the agency problem, conflicts of interests between equity holder and bond holders. The flexible renegotiation of bank debt could alleviate the underinvestment problem proposed by Myers (1977 ). Extensive theoretical papers argue that banks have 4 By assuming short term debt m aturing sooner in the future but after the investment decision, Diamond and He (2010), revisit the debt overhang problem and argue that short term debt also could cause severe debt overhang problems, when firms got hit by significant negative shocks on t heir interim asset in place value. Firms choose an optimal maturity structure by trading off the long term debt overhang in good times and short term debt overhang in bad times.

PAGE 64

64 reduce asset substitution problem (Diamond (1984, 1991b), Boyd and Prescott (1986), Park (2000)). Fama (1985) sugges ts that the benefit of bank monitoring could spill over to bondholders by reducing the cost of public debt and allowing for a higher leverage Rajan (1992), however, argues that bank debt could bring in another conflict of interest between the borrower and the lender. Specifically, banks could gain information monopoly from their monitoring and control functions and demand a share of the rents from profitable projects of bor it is an empirical issue whether bank debt mitigates or exaggerates the agency problem. Gertner and Scharfstein (1991) and Chemmanur and Fulghieri (1994) argue that private debt is easier to ren egotiate during financial distress than public debt because of the concentrated ownership of private debt. Public debt, however, could incur too much and inefficient liquidation. Therefore, theories predict that H3: Bank debt may either mitigate or exagg erate the agency problem; bank debt reduces the likelihood of financial distress. Leverage, Short Term Debt, and Bank Debt The above discussions indicate that firms can choose not only leverage ratio but also short term debt and/or bank debt to control the agency problem and the likelihood of financial distress and maximize the net benefit of borrowing. Indeed, when facing short term debt or bank debt to control the age ncy problem. Similarly, firms with a

PAGE 65

65 can use less short term debt or more bank debt to control the likelihood of financial distress. Therefore, I hypothesize that H4: s hort term debt mitigates the negative relation between leverage ratio and agency cost of debt. Short term debt, however, exaggerates the negative relation between leverage ratio and the degree of financial distress. H5: bank debt either mitigates or exag gerates the negative relation between leverage ratio and agency cost of debt. Bank debt mitigates the negative relation between leverage ratio and the degree of financial distress. Existing Empirical Evidence There is a small but growing literature argui ng that capital structure decisions are jointly determined depending on firm characteristics and other borrowing considerations. Barclay et al. (2003) argue that maturity and leverage ratio are strategic complementarities and are jointly determined. Johns on (2003) shows that short debt maturity attenuates the negative effect of growth opportunities on leverage ratio while directly reducing the leverage ratio. Johnson interprets the results as firms trade off the cost of underinvestment problem against the cost of liquidity risk when choosing short maturity. Billett et al. (2007) extends the joint decision of leverage and maturity by including the endogenous decision of covenant structure. They find that covenant protection increases in growth opportunities, debt maturity, and leverage, and attenuates the negative relation between leverage and growth opportunity. The results suggest that in addition to leverage, and maturity, covenants can be used to mitigate the conflict of interest between equity holders an d bond holders. Despite the extensive literature on the special role of bank debt, there are few papers examining the effect of bank debt on leverage and other capital structure decisions. My study aims to shed light on the joint determination of leverage, short term debt, and bank debt usage with an emphasis on the role of bank debt.

PAGE 66

66 Simultaneous Equations structure decisions in isolation by using a single equation framework, I p ropose a system of simultaneous equations to account for the interdependence between leverage ratio, debt maturity, and debt sources. I argue that the simultaneous estimation of a system of equations has the following advantages over the single equation f ramework. First, examining the joint determination of leverage, short term debt, and bank debt rather than each one in isolation will produce consistent coefficients and capture the simultaneous and potentially offsetting effects that one endogenous decisi on variable (say bank debt) could have on another (for instance, leverage). It is not uncommon to see one endogenous capital structure decision variable being regressed on another. For instance, Stohs and Mauer (1996) regress debt maturity on leverage. H ouston and James (1996) and Johnson (1997) regress the fraction of bank debt on predetermined when analyzing how firms choose the other. The coefficients estimated Furthermore, simultaneous system of equations can test theories that produce ambiguous single predictions in the reduced form regression coefficients (Barclay et al. (2003)). Within a sim ultaneous system of equations, changing one of the exogenous variables can have both direct and indirect effects on the endogenous variables. For atio. However, a shift in investment opportunities also has an indirect effect on leverage through its effect on debt maturity more investment opportunities result in shorter debt maturity, and this shorter debt maturity can increase

PAGE 67

67 everage ratio because of reduced agency cost of debt. Therefore, the offsetting direct and indirect effects of investment opportunities on leverage produce ambiguous predictions for the reduced form regression coefficients. The simultaneous system of equat ions, however, is not a panacea to all econometrics issues. To identify the system of equations, one must impose more structure on the estimation process. Following Johnson (2003) and Billet et al. (2007), I examine theories to obtain guidance on the form of structural equations. I test these hypotheses by interacting the proxy for agency cost of debt, with the proportion of short debt and with the proportion of bank debt, respectively in the leverage equation. A positive interaction of agency cost of debt and proportion of short term (bank) debt would suggest that short term debt (bank debt) can moderate the negative relation between leverage ratio and agency cost of borrowing. Similarly, a negative (positive) interaction of financial distress and short t erm (bank) debt indicates that having less (more) short term (bank) debt moderates the relation between leverage ratio and degree of financial distress. The overall effect of short term debt and bank debt on leverage depends on their stand alone effect an d partial derivatives of the interaction terms. My simultaneous system of equations is specified as follows. Lev i,t+1 =F(ShortDebt i,t+1 BankDebt i,t+1 AG i,t ShortDebt i,t+1 *AG i,t BankDebt i,t+1 AG i,t ,FD i,t ShortDebt i,t+1 *FD i,t BankDebt i,t+1 *FD i,t X i, t i,t+1 (3 1) ShortDebt i,t+1 =F(Lev i,t+1 BankDebt i,t+1 AG i,t Lev i,t+1 *AG i,t BankDebt i,t+1 AG i,t, FD i,t Lev ,t+1 *FD i,t BankDebt i,t+1 *FD i,t Y i,t i,t+1 (3 2) BankDebt i,t+1 =F(Lev ,t+1 BankDebt i,t+1 AG i,t Lev ,t+1 *AG i,t ShortDebt i,t +1 AG i,t, FD i,t Lev i,t+1 *FD i,t ShortDebt i,t+1 *FD i,t Z i,t ) i,t+1 (3 3) Lev i,t+1

PAGE 68

68 ShortDebt i,t+1 year t+1 BankDebt i ,t+1 AG i,t FD i,t X i,t Y i,t and Z i,t termined or exogenous variables at year t in the regression of leverage,short term debt, and bank debt. i,t+1 i,t+1 i,t+1 : residuals. I use the two stage least squares to estimate the system of simultaneous equations. Two stage least squares is the most common method for estimating simultaneous equation models. Two stage least squares is more robus t to model misspecifications in the system and easier to compute and interpret than other estimation methods. Following Johnson (2003) and Billet et al.(2007), I estimate the simultaneous equations using pooled regression. Given that Lemmon et al. (2008) report that leverage ratio decision is largely driven by some time invariant firm fixed effect and mysample period covers the recent financial crisis, I also include firm and year fixed effects into the model for robustness check. Data, Proxies, and Descri ptive Statistics Data To empirically test the joint determination of leverage, short term debt, and bank debt, I need to gather data on these variables. While data on leverage and short term debt can be readily obtained from Compustat, researchers most gen erally collect bank

PAGE 69

69 5 Such high data collection cost constrains researchers to sacrifice either the breadth or the length of coverage. Capital IQ (CIQ) provides detailed debt structure informati on including outstanding bank debt for a wide range of firms from year 2002 to year 2009. 6 It groups total debt into seven categories: commercial paper, drawn revolving lines, term loans, senior bonds and notes, subordinated bonds and notes, capital lease some of these facilities are likely from other financial institutions. Yet, given that a vast majority of revolving lines and term loans are offere d by banks 7 I believe that such measurement error is unlikely to be severe. My sample consists of all nonfinancial firm year observations covered by both Compustat and Capital IQ (CIQ) between 2003 and 2009 matched by SEC CIK (Central Index Key). Year 20 02 data is excluded from the sample because lagged industry median bank debt information, a predicting variable for bank debt, is not available until year 2003. 8 As in Lemmon et al. (2008), I require that all firm years have nonmissing data for book asset s and that the book and market leverage ratio lie within the unit interval. Because both percentage of short term debt and bank debt usage is undefined for firms with zero debt, I exclude firm years with zero total debt. Following Johnson 5 For instance, Houston and James (1996) collect the ratio of bank debt over t SEC 10K filings. Johnson (1997) and Krishnaswami et al. (1999), and Cantillo and Wright (2000) gather 6 CIQ gathers data from sources such as company website; press release; a out year 2000 and 2001 because the sample coverage of CIQ for these two years is poor. 7 Carey et al. (1999) show that 82% of their loan sample are offered by a sole bank or bank syndication. 8 Recall from footnote 3 that ban k debt information is only largely available for most firms until year 2002. Some lagged industry median bank debt is not available until year 2003.

PAGE 70

70 (2003), I further remove firm years with more than 100% of their debt maturing within any specific period. To assure a good match between Compustat and CIQ, I discard firm years for which the difference between book assets recorded in Compustat and book assets recorded in CIQ exceeds 10% of book assets. To assure my data quality, I also filter out firm years whose total debt as reported in Compustat differs the sum of debt types as reported in CIQ by 10% or above. Finally, I constrain my sample to firm year observations wit h nonmissing values of all relevant variables. My final sample size is 10,560 firm year observations, representing 1,998 different firms. Proxies To examine the joint determination of leverage, short term debt and bank debt in E quations 3 1, 3 2, and 3 2 I need to find proxies to fit the system of simultaneous equations. I will first discuss proxies for the endogenous choice variables leverage, short term debt, and bank debt, followed by discussion of proxies for agency cost and financial distress cost. I will then specify other exogenous or predetermined variables included in each equation. Proxies for variables of interest Following standard convention in the literature, I measure leverage as the book value of total debt (long term debt plus debt in current liability) divided by the market value of assets (book value of total assets less the book value of equity plus market value of common equity). 9 Book leverage is defined as the book value of total debt divided by the book value of total assets. As in Barclay and Smith (1995), Barclay et al. (2003), Johnson (2003), and Billet et al. (2007), I use the proportion of total debt 9 Empirical researches that analyze market leverage ratio include Hovakimian et al. (2001), Johnson (2003), Welch (2004), Leary and Roberts (2005), Flannery and Rangan, (2006) etc.

PAGE 71

71 term debt. I measure the use of bank debt as the summation of d rawn revolving lines and term loans over total debt. Though proxies for agency cost of debt (AGi,t) are hard to obtain, theories suggest that firms with more investment opportunities are more prone to the debt overhang and asset substitution problem than f irms with fewer growth opportunities (Myers (1977), Jensen and Meckling (1976)). I measure investment opportunities using the market to book asset ratio. This is the most commonly used measurement of investment opportunities and Adam and Goyal (2003) find it the best proxy for investment financial distress (FDi,t). As in Mackie Mason (1990), Lemmon et al. (2008), and Sufi score by exclu ding leverage to ensure the exogeneity of the proxy. 10 A lower z score means that a firm is more likely to be financial distressed. Proxies for control variables Besides proxies for agency cost of debt (AGi,t) and the likelihood of financial distress (F Di,t), I need to include other exogenous or predetermined variables X, Y, an Z to identify the system of simultaneous E quations 3 1, 3 2 and 3 3 Exclusion restrictions are then required to generate credible two stage least square estimates. For instance, some variables in the vector X should only affect leverage, but not short term debt usage or bank debt usage, and therefore are excluded from vector Y and Z. Similarly, there should be variables included in vector Y (Z), but not in X and Z (Y). The simult aneous system of equations established by Johnson (2003) and Billet et al. (2007) 10 Specifically, this variable is calculated as z score=3.3 (pre tax income/total assets) + 1.0 (sales/total assets) + 1.4 (retained earnings/total assets) + 1.2 (working capital/tota l assets)

PAGE 72

72 serves as a basis for my analysis. I further survey the previous literature on leverage, debt maturity, and bank debt to improve my identification. My basic approach is to id entify the factors affecting leverage, debt maturity, and bank debt respectively from the predictions of theories (in some cases, strong empirical patterns). Naturally, the variables that have prediction power in one equation but not the others would serve as exclusion restrictions. For instance, both theoretical and empirical studies find that firms with higher asset maturity use more long term debt for maturity matching. Neither theories nor empirical studies generate predictions about asset maturity on l everage ratio and bank debt. Therefore, asset maturity is excluded in the leverage and bank debt equations. In this way, I identify my simultaneous system of equations. To preserve space, I leave the full discussion of control variable selections and their predic ted signs in E quation s 3 1 through 3 3 to Appendix B Interested readers can refer to that section for further insights. The control variables in my system of simultaneous equations are specified as follows. In the leverage equation, I include fixed assets over total assets, log of inflation adjusted net sales, EBITDA over total assets, cash flow volatility, abnormal earnings, investment tax credit, net operating loss, following Johnson (2003) and Billet et al. (2007). I also control for industry med ian leverage ratio and dividend payer given their (2009)). In the short term debt equation, I include firm size, the square of firm size, cash flow volatility, abnormal earnings, asset maturity, term premium, investment tax credit, net operating loss, below investment grade rating and no debt rating dummy as in

PAGE 73

73 Johnson (2003) and Billet et al. (2007). 11 I identify the bank debt equation by adding firm size, the square of firm size, fixed assets over total assets, cash flow volatility, abnormal earnings, below investment grade, and no debt rating dummy borrowing from Johnson (1997), Krishnaswami et al (1999), and Denis and Mihov (2003). I also include tightening of lendin g standards based on the Federal Reserve Senior Loan Officer Opinion Survey, and a market adjusted stock price index for banks in the bank debt equation following Becker and Ivashina (2010). I measure the endogenous capital structure variables leverage ra tio, short term debt, and bank debt at the end of fiscal year t+1, and measure the exogenous variab les at the end of fiscal year t except for the variable abnormal return, which is the unexpected EPS in year t+2. 12 13 Definitions of all variables used in t his study are listed in Appendix A All ratio variables (including my dependent variables) are winsorized at the 1st and the 99th percentiles to alleviate the influence of outliers. Descriptive Statistics Table 3 1 provides summary statistics for the endo genous capital structure variables and exogenous variables. In the last two columns, I report the mean and median values of the same variables for the Compustat nonfinancial firms of the same 11 The dummy variable below investment grade is not included in Johnson (2003) and Billet et al. (2007). But Diamond (1991a) predicts a non linear relation between the use of short term debt and firm credit, i.e, firms with the best and the worst creditworthiness use more short term debt, firms in the middle use more long term debt. I therefore group firms into three groups, those with an investment grade rating; those with a below investment grade rating; and those with no rating to allow for a po ssible non linear relation between credit quality and short term debt usage. 12 All financial variables are arguable endogenous choices of firms. I use the lagged variables to ensure that they are at least pre determined. 13 The motivation for including the lead EPS is specified in Appendix B.

PAGE 74

74 period, filtered with similar standards of the CIQ Compustat mer ged data. 14 Asterisks on the means and medians indicate whether my samples are significantly different from the Compustat universe firms. My sample is quite representative of the Compustat leveraged firms. Around 79% (10560 out of 13374 firm years) of th e Compustat leveraged firms are covered in my final sample. Firm characteristics of my sample are fairly similar to the Compustat sample. When comparing my key variables, leverage, short term debt, bank debt, market to book, and z score, my sampled firms u se slightly less short term debt, and are less financially distressed. In addition, my sampled firms are slightly bigger, more tangible, more profitable, less volatile, and more likely to pay dividends. Together they suggest that my sample includes more ma ture firms. But in general the differences in firm characteristics are not economically significant though some are statistically significant. Panel A of Table 3 2 contains a matrix of Pearson correlation coefficients among leverage, short term debt, bank debt, market to book ratio, and z score using all my sampled firms. These correlations reveal some simple relations between the key variables. Leverage ratio is negatively correlated with short term debt, and positively correlated with bank debt. This sug gests that bank debt can be distinguished from short term debt as bank debt and vice. 15 The correlation between bank debt and short term debt is 14 Filters include no missing total asset value, positive total debt, market and book leverage ratio within the unit, less than 100% of total debt maturing within any specified period, and no missing data for all relevan t variables. 15 growth firms that choose bank financing over public or nonbank private debt will have more short

PAGE 75

75 significantly positive, indicating that fi rms simultaneously use more bank debt and short term debts. Market to book ratio is negatively related to leverage and positively related to short term debt as predicted by Myers (1977). Market to book ratio is negatively related to bank debt, consistent w ith Rajan (1992). Z score is negatively related to leverage ratio, and short term debt, and positively correlated with bank debt, the opposite of my predictions. Theories argue that firms with low credit quality and severe information asymmetry problems m ight get credit rationed (Stieglitz and Weiss (1981)), or have no access to the long term debt market (Diamond (1991a)), or the public bond market (Diamond (1991b)). Faulkender and Petersen (2006), Guedes and Opler (1996), and Denis and Mihov (2003) find s upporting empirical results. Therefore, I examine the Pearson correlation coefficients among firms with and without an S&P long term debt r ating. Panel B and C of Table 3 2 report matrixes of Pearson correlation coefficients for rated firms, and unrated fi rms respectively. The correlation matrix of unrated firms and rated firms differs substantially in magnitude and even signs. For instance, the correlation between leverage and bank debt among rated firms is 0.23 in comparison to the 0.09 of unrated firms. The correlations between Z score and other variables of rated firms among rated firms have the opposite signs of those unrated firms. These results highlight the necessity to analyze the rated firms and unrated firms separately. Though the Pearson correl ation coefficients reveal some simple relation among the endogenous capital structure decisions and proxies for agency cost and financial distress cost, one has to take into account other covariates and the simultaneity issue. The next section seeks to tac kle these issues using simultaneous equation regressions.

PAGE 76

76 Joint Determinants of Leverage, Short term Debt, and Bank Debt This section presents the results of the simultaneous equation regressions for leverage, short term debt, and bank debt estimated by tw o stage least squares. I organize this section as follows. In part A, I will present the estimation results of the joint determination of leverage, short term debt, and bank debt. To facilitate the comparison with previous papers and to gauge the impact of adding bank debt into the system of endogenous decisions, I also fit a two equation system with leverage and short term debt endogenously decided as in Johnson (2003). In part B, I partition the whole sample into two groups: firms with and without an S&P long term issuer credit rating. I then re estimate the three equation system separately for the rated firm sample and unrated firm sample. This comparison allows us to access whether and how the relation among the endogenous variables are affected by firms debt markets. Part C offers several robustness checks for my main results. Estimation Results of Two equation, and Three equation systems. Table 3 3 presents the estimation results of system of equations. The fir st two columns report the joint estimation of leverage and short term debt equations, as in Johnson (2003), Barclay et al. (2003), and Billett et al. (2007). The next three columns report the joint estimation of leverage, short term debt, and bank debt. I will review the two equation systems briefly and compare my results with the previous papers. More detailed discussion will be given to the three equation system. Note that I estimate all three systems using pooled regressions without including firm and ye ar fixed effect. These pooled regressions were used by preceding studies and make my results more

PAGE 77

77 comparable with previous literature. 16 In Table 3 4 I add firm and year fixed effects into my model to take into account any effects from unobserved time invar iant firm characteristics or macro shocks in the robustness check section. 17 Standard errors are clustered by firms given that residuals of the leverage, short term debt, and bank debt are more likely to be correlated across observations from the same firm (Lemmon et al. (2008), Petersen (2008)). The joint estimation of leverage and maturity equations is largely consistent with Johnson (2003) and Barclay et al. (2003). Specifically, in the leverage equation the direct effect of short term debt is significa ntly negative, and the interaction of short term debt and market to book ratio is significantly positive. Johnson (2003) interprets these findings as short term debt could reduce the cost of underinvestment problems but introduce higher liquidity risk. Hol ding the market to book ratio and z score at the sample mean, the net effect of short term debt is negative, consistent with Barclay et al. (2003) and Johnson (2003). In the short term debt equation, the net effect of leverage on short term debt is also s ignificantly negative, as in Barclay et al. (2003) and Johnson (2003). 18 The negative relation between short term debt and leverage, however, is 16 Johnson (2003) reports the joint estimation of leverage and maturi ty with firm fixed effect, but not year fixed effect, and draws the main conclusion based on a pooled regression. Barclay et al. (2003) estimates the system of leverage and maturity with neither firm fixed effect nor year fixed effe ct. Billet et al. (2007) exclude firm fixed effect in the system of leverage, maturity, and covenants and use year fixed effect to instrument number of covenants, but not leverage nor maturity. 17 I also intend to run a cross sectional regression to take in to account the serial correlation in the error terms for robustness check in the future. 18 The partial derivative of leverage with respect to maturity is calculated as 0.626+0.114*1.773+0.009*1.14= 0.4136. This marginal effect is significantly different f rom zero at 1% level.

PAGE 78

78 more pronounced among firms with higher market to book ratio as in Johnson (2003). 19 The control variables in the leverage equation and maturity equations are largely consistent with Barclay et al. (2003) and Johnson (2003). My comparable estimation of the leverage and maturity system on a different data set facilitates the assessment of the significance of bank debt It reduces the possibility that my results on bank debt are driven by sample difference. The three equation system results suggest that it is important to include bank debt as an endogenous capital structure decision. The use of bank debt greatly impact s leverage equation, bank debt has a positive direct effect on the leverage ratio (the coefficient on bank debt is 0.571), suggesting that bank debt reduces the cost of borrowing and increases firm debt capacity. Surprisingly, the benefit of bank debt significant negative interaction term of bank debt and market to book ratio ( 0.06) is co nsistent with Rajan (1992), who argues that the monitoring and control function of bank debt could bring in a cost of bank finance. Specifically, Rajan argues that the informational monopoly of banks from their monitoring and control function could exploit negative and significant interaction term of bank debt and z score reveals that taking additional bank debt is more beneficial to the financial distressed firms (lower z sco re) than those healthy firms (higher z score). This is consistent with the conjecture that bank debt reduces the financial distress cost and therefore increases the leverage ratio. 19 Holding the market to book ratio and z score at sample mean, this net effect is calculated as 0.197 0.525*1.773 0.002*1.14= 0.7366. This marginal effect is significantly different from zero at 1% level.

PAGE 79

79 The effect of bank debt on leverage is economically significant. Fixing th e market to book ratio and z score at the sample mean, an increase from zero bank financing to 100 percent bank financing increases the leverage ratio by 42 percentage points. 20 The net effect of bank debt on leverage is contingent on the value of market to book ratio and z score. Holding the z score at its mean and the market to book ratio at its 90th percentile, an increase in bank financing from zero to 100 percent increases the leverage by 33 percentage points, suggesting that bank debt does not increas e as much leverage ratio among high growth firms than among low growth firms. 21 Similarly, holding the market to book ratio at its mean and the z score at its 90th percentile, an increase in bank financing from zero to 100 percent increases the leverage by 35 percentage points, suggesting that bank debt can increase the leverage ratio more among firms with a higher expected financial distress cost. Overall, bank debt increases the leverage ratio. Even for firms with a 90 percentile of market to book ratio and a 90 percentile of z score, switching from zero bank financing to 100 percent bank financing increases the leverage ratio by 26 percent. Short effect on leverage is very diff erent from bank debt, suggesting that short term debt and bank debt are not equivalences. Specifically, short term debt has a significant negative direct effect on the leverage ratio, which suggests that short term debt is costly possibly due to the increa sed liquidity risk, which reduces the leverage ratio. This is more evident 20 This partial derivative is calculated as 0.573 0.06*1.77 0.044*1.14 = 0.42. This marginal effect is significantly different from zero at 1% level. 21 This partial derivative is calculated as 0.573 0.06*1.77 0.044*3.21 = 0.33. This marginal effect is significantly different from zero at 1% level.

PAGE 80

80 given a positive interaction term of z score and short term debt, which indicates that short term debt reduces leverage more for firms with a higher likelihood of financial distress (lower z score). Consistent with Johnson (2003), short term debt attenuates the negative relation between the market to book ratio and leverage, consistent with the view that short term debt reduces the agency cost of debt. The net effect of short term d ebt on leverage depends on the market to book ratio and z score. For firms with sample mean market to book ratio and sample mean z score, changing a debt structure from 100 percent long term financing to 100 percent short term financing reduces the leverag e ratio by 17 percentage point. On the other hand, for firms with a 90th percentile market to book ratio and a 90th percentile z score, increase from zero short term debt financing to 100 percent short term financing weakly increases the leverage ratio by 1%. Intuitively, firms with high market to book ratio and high z score are precisely those who would find short term debt beneficial because of its reduction of the agency cost of debt without being concerned about too much liquidity risk. The relation between market to book ratio and leverage ratio is contingent on the amount of short term debt and the amount of bank debt. Short term debt attenuates the negative relation of market to book ratio and leverage while bank debt enhances the negative relation If a firm uses neither short term debt nor bank debt, then a standard deviation increase in market to book ratio reduces the leverage ratio by around 7.6 percent points. If a firm uses 100 percent short term debt but no bank debt, a one standard deviati on increase in the market to book ratio raises the leverage rat io by around 4.7 percent points If a firm borrows a 100 percent bank debt but no short term

PAGE 81

81 debt, one standard deviation increase in the market to book could reduce the leverage ratio by 14.6 percentage points. These results again suggest that short term debt could reduce the agency cost of debt, while bank debt surprisingly does not. Similarly, the net effect of z score on leverage ratio depends on the amount of short term debt and bank debt Higher short term debt increases the sensitivity of leverage to financial distress costs, while bank debt decreases the sensitivity, suggesting that short term debt increases the financial distress cost while bank debt reduces such cost. To interpret the magnitude, if a firm borrows 100 percent short term non bank debt, a one standard deviation increase in the z score raises the leverage ratio by around 3.6 percent points. If a firm uses 100 percent long term bank debt, a one standard deviation increase in the z score reduces the leverage ratio by around 19 percent points. This large negative relation between leverage ratio and z score may be because z score also captures firm profitability, which is negatively correlated with the leverage ratio, a patte rn consistently documented in the previous literature. Indeed, the pair wise correlation of z score and asset scaled EBITDA is around 0.73, suggesting that this conjecture is plausible. Coefficients on other control variables in the leverage equation are largely consistent with results of Johnson (2003), Barclay et al (2003), and Billet et al. (2007), except for the coefficient on firm size. Johnson (2003), Barclay et al (2003), and Billet et al. (2007) all report a surprising negative relation between s ize and leverage ratio. Frank and Goyal (2007), however, show that the positive relation between firm size and leverage is a well documented pattern in the capital structure literature. Barclay et al. (2003) admittedly agree that the negative sign on size

PAGE 82

82 possibility of model misspecification (P.161). My results, however, show that bigger firms use more leverage. The consistence of my results with the regularities of the capital structure literature provides supportive evidenc e to my model specification, which adds additional exogenous variables and includes debt source into the system of equations. term debt. This positive correlation is more pr onounced among firms with a higher degree of financial distress (a lower z score). This result is consistent with the view that bank debt complements short term debt to reduce the financial distress risk. Indeed, in practice, I observe that bank debt is of ten extended short term, possibly because this design could reduce financial distress cost. Higher market to book ratio directly increases the short term debt usage, consistent with the view that firms use short term debt to reduce the agency problem. Bank debt insignificantly attenuates this positive correlation (the interaction term of bank debt and market to book ratio is negative but insignificant.) This provides little support to the view that bank debt serves as a substitute for short term debt to con trol for agency cost of debt. The effect of leverage ratio on short term debt changes from statistically significant in the two equation system to statistically insignificant in the three equation system, suggesting that controlling for bank debt in the sy stem makes a difference for estimated results. Coefficients on the control variables are again largely consistent with the study of Johnson (2003) that I follow closely. The use of bank debt financing is greatly influenced by leverage ratio and short term debt. Specifically, a higher leverage ratio promotes more bank debt usage,

PAGE 83

83 suggesting that if a firm intends to borrow more it uses disproportionally more bank debt. This result is consistent with Johnson (1997) who finds that leverage ratio increases the use of bank debt, but inconsistent with Houston and James (1996) who find a negative relation. This positive relation between leverage ratio and bank debt is somehow weaker among firms with a higher degree of financial distress (a lower z score). At first glance, this is somehow at odds with the notion that the flexibility of bank debt renegotiation during financial distress makes bank debt more attractive to those financially distressed firms. It is possibly because as firms with a higher degree of financi al distress take on more debt, their likelihood of default increases. Risk averse banks, therefore, limit their risk exposure to the distressed firms by not granting them as much loan as to the healthy firms. Indeed, Sufi (2009) reports that bank lines of credit are only used for liquidity management among healthy firms with abundant cash flow. Mirrored with the maturity equation, short term debt increases the usage of bank debt, suggesting that short term debt and bank debt often complement each other. Thi s positive relation is less strong among firms with high growth opportunities. The effect of market to of leverage ratio and short term debt. Holding the leverage ratio and short term debt at thei r sample means, a one standard deviation increase in growth opportunities decreases the use of bank debt by 1 percent point. Holding the leverage ratio and short term debt at their 10th percentile, a one standard deviation increase in growth opportunities increases the use of bank debt by 5 percent. Holding the leverage ratio and short term debt at their 90th percentile, a one standard deviation increase in growth opportunities decreases the use of bank debt by 9 percentage points. Therefore,

PAGE 84

84 growth firm s with a high level of leverage and short term debt use less bank debt, while growth firms with a low level of leverage and short term debt use more bank debt. This is possibly because growth firms with higher leverage and short term debt are more likely t o be exploited by banks given that a higher leverage and short term debt Some control variables in the bank debt equation have their expected sign while er firms and firms that are below investment grade or have no rating use more bank debt, consistent with the special role of bank debt in reducing information asymmetry via either ex ante project screening or ex post project monitoring. Previous literature also reports similar empirical pattern (Houston and James (1996), Johnson (1998), and Krishnaswami et al. (1999)). Firms with high profitability and low cash flow volatility use more bank debt. Similar results are reported in Rauh and Sufi (2010) and Coll a et al. (2011). The result is consistent with Sufi easier to comply bank covenants and therefore use more bank debt. To sum up, my results show that it is importan t to recognize bank debt as an important endogenous capital structure decision. The use of bank debt greatly s that bank debt could increase the leverage ratio to a greater extend among more distressed firms suggesting that bank debt is able to reduce the expected financial distress cost for those firms. Bank debt, however, increases the negative correlation betw een market to book ratio

PAGE 85

85 agency cost due to its information monopoly. Bank debt increases with short term debt and vice versa, suggesting that overall the two types of d ebt often go together. Bank debt and short term debt, however, are not equivalent, as they have the opposite impact on the leverage ratio. Indeed, I find that short term debt reduces the leverage ratio in general. However, this magnitude is less pronounced among firms with high growth opportunities. These results suggest that short term debt could reduce the agency cost of debt while introducing higher liquidity risk and financial distress risk as in Johnson (2003). Bank debt increases as firms are more lev eraged. This relation is even stronger among firms with a lower degree of financial distress. This is possibly because risk averse banks tend to limit their exposure to the high risk firms. Differences in the Three equation System across Rated and Unrated Firms Theoretical models predict that firms with low credit quality and severe information asymmetry problems might get credit rationed (Stieglitz and Weiss (1981)), or have no access to the long term debt market (Diamond (1991a)) or the public bond market (Diamond (1991b)). Therefore, I partition the whole sample into two groups: firms with and without an S&P long term issuer credit rating. Rated firms are likely to have higher credit quality and less information asymmetry, and have full access to differen t debt markets. This section discusses the estimated three equation system for the rated fi rms and unrated firm in Table 3 4 To preserve space, only coefficients on the variables of interest are reported. The comparison of the estimated results between ra ted and unrated firms reveals some interesting patterns. First, bank debt has different impacts on the leverage ratio among rated vs. unrated firms. Though bank debt can significantly increase the leverage ratio among

PAGE 86

86 both rated and unrated firms, the sig nificant negative interaction term of market to book ratio and bank debt only appears among unrated firms. Indeed, among rated firms bank debt could attenuate the negative relation between the market to book ratio and leverage ratio, though the effect is n ot statistically significant. This further provides supportive evidence to the agency cost of bank hold up problem. This is because the unrated firms have limited access to non bank funding due to their information asymmetry problem or high credit risk and are more likely to be exploited by banks. The attenuation of bank debt on the relation between z score and leverage is also significant among unrated firms rather than the rated firm sample, suggesting that financial distress cost is a concern among unrat ed firms and bank debt could reduce such cost. The economic magnitude of the effect of bank debt on leverage differs between the rated and unrated firms. Holding the market to book ratio and z score at the sample means, changing the debt source from all nonbank debt financing to all bank debt financing increases the leverage by 47.3 percent points among rated firms. 22 The same change raises the leverage ratio by 33.3 percent among unrated firms. 23 The difference is significant at the 5% level. In contrary, holding the market to book ratio at its 90th percentile and z score at its mean, changing the debt source from all nonbank debt financing to all bank debt financing increases the leverage by 50.3 percentage points among rated firms in comparison to 20.8 pe rcentage points 24 among unrated firms. The 22 This marginal effect is calculated as 0.465+0.021*1.77+1.14*( 0.026) = 0.473. This magnitude is significant at the 1% level. 23 This marginal effect is calculated as 0.528 0.087*1.77 0.036*1.14 = 0.333. This magnitude is significant at the 1 % level. 24 This marginal effect is calculated as 0.528 0.087*3.21+1.14*( 0.026) = 0.208. This magnitude is significant at the 5% level.

PAGE 87

87 difference is statistical significant at the 1% level. This comparison demonstrates the different bank debt attenuation effects among rated firms and unrated firms. The overall positive effect of bank debt on lever age ratio suggests that borrowing from banks can somehow reduce the borrowing cost and therefore increase the leverage ratio. Second, the impact of short term debt on the leverage ratio differs between rated firms and unrated firms. Specifically, among ra ted firms short term debt has an insignificant negative direct impact on the leverage ratio and a larger attenuation benefits for high market to book firms in comparison to those among the unrated firms. Economically, with market to book and z score at the ir means, changing from exclusive long term debt financing to exclusive short term debt financing decreases the leverage ratio by approximately 3.5 percentage points among rated firms. 25 The same debt maturity structure change, however, reduces the leverage ratio by 18.9 percent among unrated firms. 26 Johnson (2003) reports similar finding. These results suggest that short term can be used to reduce the agency cost of debt without adding too much liquidity risk among rated firms. The unrated firms, with high er liquidity risk, would find the additional liquidity and financial distress cost from using short term debt outweighs the benefit of agency cost reduction. Third, my results show that bank debt and short term debt differ more substantially for the rated firms than for the unrated firms. Indeed, the complementary relation between short term debt and bank debt only appears among the unrated firms. This is 25 This marginal effect is calculated as 0.255+0.139*1.77+0.039*1.14 = 0.035. This magnitude is insignificant though. 26 This marginal effect is calculated as 0.286+0.046*1.77+0.014*1.14 = 0.189. This magnitude is significant at the 5% level.

PAGE 88

88 consistent with the fact that certain short term debt such as commercial paper are non bank debt and a re only accessible to rated firms. Fourth, high leverage increases the use of short term debt among rated firms in greater economic magnitude and statistical significance. Opler and Titman (1994) argues that a higher leverage ratio is associated with a hi gher degree of financial distress. The positive correlation between the short term debt and leverage ratio among rated firms suggest that financial distress cost is less of a concern for these firms. Given that firms with a higher leverage ratio have a hig her incentive to engage in asset substitution (Leland and Toft (1996)), these firms have higher agency cost and would use more short term debt to reduce this agency cost. Together, these results reveal that the joint decisions of leverage, short term debt, and bank debt are different between rated firms and unrated firms. It appears that these joint decisions are motivated by the concern of borrowing cost such as agency cost and financial distress cost. When making these joint decisions, rated firms are mor e concerned about reducing the agency cost of debt financing, unrated firms are more concerned for financial distress cost and the agency cost from bank information monopoly. Robustness Check Though the pooled regression with no firm and year fixed ef fect is often used in the previous studies, leaving out these fixed effects could potentially create omitted variable bias. This is especially problematic if there are any unobservable time invariant firm characteristics and unobservable macro economic sho leverage, short term debt, and bank debt. Indeed, Lemmon et al. (2008) show that the majority of variation in leverage ratio is driven by some time invariant effect. My sample

PAGE 89

89 period also covers the most recent financial cri sis, which is a big macro economic shock. In the worst case, if the excluded exogenous variables correlate with the unobserved fixed effects, estimations of the simultaneous equations are biased. However, the addition of firm and year fixed effects could lower the predictive power of excluded instrumental variables, in particularly those stationary variables, and introduce weak instruments problems. Bound, Jaeger, and Baker (1993, 1995) show that when the excluded instruments are only weakly correlated wit h the endogenous variables, the IV estimates are biased in the same direction as OLS and are inconsistent. Weak instruments also cause larger standard errors on IV estimates, making it harder to draw any inference. I therefore report the fixed effect estim ation results in Ta ble 3 5 to compare with the pooled regression results. Estimation results for the full sample are r eported in system (1) of table 3 5 In the leverage equation, the effect of short term debt and bank debt on leverage is largely consist ent with patterns in the pooled regression model. The only notable difference is that the negative coefficient on the interaction of z score and bank debt becomes statistically insignificant. The maturity equation, however, is less robust, with some estima tes flipping their signs. This is not surprising because the under idenfication test suggests that the maturity equation suffers from weak instruments problem. In the bank debt equation, the all coefficient estimates have the same signs as in the pooled re gression, although some coefficients lose their statistical significance (short term debt and the interaction term between z score and bank debt). System (2) and system (3) displays the estimated system of equations for rated and unrated firms. The estima ted results for bank debt equation for the fixed effect

PAGE 90

90 model are quite different from those of the pooled regression. The reason is that the rating dummy variable is an important predictor for bank debt, in particular in the fixed effect model. By splitti ng the sample into rated and unrated groups, the rating dummy is dropped out. The structural bank debt equation then changes accordingly. I use the underidenfication test to examine whether the excluded exogenous variables are relevant and the overidenfic ation test to evaluate the validity of the excluded restrictions. The underidentification test is usually satisfied in both pooled and fixed effect regressions, suggesting that the excluded exogenous variables are relevant to the endogenous variables. The overidentification test for the pooled and fixed effect regressions fails sometime, suggesting that either my model is misspecified or the specification fails the overi dentification test with my dataset. To improve my model take further. In summary, while I still need to further refine my econometric model, there are several observ ations robust to different estimation methods. debt, however, exaggerates the negative corr elation between market to book ratio and leverage. This effect is concentrated mostly among the unrated firms. These findings interest between equity holders and bond holders due to their information monopoly.

PAGE 91

91 Second, short term debt reduces the leverage ratio in general. However, this magnitude is less pronounced among firms with high growth opportunities. These results suggest that short term debt could reduce the agency cost of debt but introducing higher liquidity risk and financial distress risk as in Johnson (2003). The effect of short term debt differs among rated firms and unrated firms due to different borrowing cost concerns. It appears that short term debt can be used by the rated firms to control for the agency problem without incurring too much liquidity or financial distress risk. Some rated firms with high agency cost and low liquidity risk can use short term debt to reduce the borrowing cost and increas e the leverage ratio. Unrated firms, however, would find the short term debt too costly due to the excessive liquidity risk. Using more short term debt in general reduces the leverage ratio among the unrated firms. Third, bank debt increases with short ter m debt and vice versa, suggesting that overall the two types of debt often go together. Bank debt and short term debt, however, are not equivalences, as they have the opposite impact on the leverage ratio. Fourth, bank debt increases as firms are more lev eraged. This relation is even stronger among firms with a lower degree of financial distress. This is possibly because risk averse banks tend to limit their exposure to the high risk firms. Chapter 3 Concluding Remarks The substantial literature on capital structure decisions often examines one facet in isolation. The study of one capital structure decision in isolation is, however, at odds with the fact that firms simultaneously decide how much debt to take out (debt equity decision), how soon to repay the debt (debt maturity), and from whom to borrow (debt

PAGE 92

92 jointly determined; maturity, in addition to leverage ratio, can be used to mitigate the underinvestment problem. I fi nd that short term debt reduces the leverage ratio in general. However, this magnitude is more pronounced among firms with low growth opportunities and those that are closer to distress. These results suggest that short term debt can reduce the agency cost of debt, but at the cost of introducing higher liquidity risk as in Johnson (2003). It appears that the effect of short term debt differs among rated firms and unrated firms due to different borrowing cost concerns. Short term debt can be used by the rate d firms to control for the agency problem without incurring too much liquidity or financial distress risk. Using more short term debt in general reduces the leverage ratio among the unrated firms due to their excessive liquidity risk. Bank debt greatly inc growth opportunities, and bank debt increases this effect among (only) the unrated firms, who tend to have limited access to non bank funding due to their information asymmetry problem or high credit risk. These findings are consistent with Rajan holders and bond holders due to their information monopoly. The preliminary results suggest that debt maturity, d ebt source, and leverage ratio are jointly determined. Debt maturity and bank debt can alter the relation between the leverage ratio and two types of borrowing cost: the potential conflicts between equity holders and debt holders, and financial distress ri sk. The results seem to suggest that in addition to leverage ratio, debt maturity, and debt source importantly affect the distortions associated with debt finance.

PAGE 93

93 For future extension, I intend to improve the identification of the simultaneous equations and use alternative measurement for the agency problem and degree of financial distress. Given the strong pattern that bank debt increases the leverage ratio, I would like to investigate further the underlying reasons. Myers and Majluf (1984) suggest that firms prefer less information sensitive financing claims (e.g. internally generated returns) to more information sensitive financing claims (e.g. equity) to reduce the adverse selection problem. Bank debt, a type of informed debt, can better address the a dverse selection problem than public debt. Consistent with the view, James (1987), Lummer and McConnell (1989), Best and Zhang (1993), Billett, Flannery and Garfinkel (1995) all report a positive share price reaction to bank loan announcements, suggesting that bank financing is viewed favorably by the market. I plan to examine whether the reduction of adverse selection cost from bank debt enables firms to take on more leverage.

PAGE 94

94 Table 3 1. Summary statistics of dependent and independent variables This table reports summary statistics of all CIQ Compustat intersection nonfinancial firm year observations and Compustat only nonfinancial firm year observations from 2003 to 2009. Year 2002 data is excluded from the estimation because lagged industry median bank debt information is not available until year 2003. Variable defini tions are provided in Appendix A I use ***, ** and to denote significance at the 1% level, 5% level and 10% level for the mean and median comparison, respectively. CIQ Compust at Intersection (10,560 firm years) Compustat only (13,374 firm years) Variable Mean Median Std. Dev. Min Max Mean Median Market Leverage 0.186 0.151 0.157 0.000 0.885 0.185 0.149 Book Leverage 0.253 0.233 0.185 0.000 1.000 0.252 0.230 Short Debt 0.4 65 0.386 0.361 0.000 1.000 0.485*** 0.409*** Bank Debt 0.345 0.161 0.383 0.000 1.000 Market to Book 1.773 1.439 1.167 0.588 13.593 1.765 1.428 Z Score 1.140 1.590 3.037 40.462 5.117 1.041** 1.534** Fixed Assets 0.310 0.238 0.237 0.010 0.904 0.305* 0.232** Profitability 0.094 0.114 0.163 1.739 0.384 0.087*** 0.112*** Firm Size 6.351 6.491 2.145 6.344 12.764 6.144*** 6.247*** Firm Size Square 44.935 42.128 26.381 0.000 162.930 42.355*** 39.037*** CF Volatility 0.060 0.035 0.084 0.004 1.422 0.065 *** 0.038*** Abnormal Earnings 0.042 0.005 0.427 2.049 3.333 0.042 0.005* Stock Return 0.131 0.125 0.528 1.439 2.092 0.144* 0.133** Asset Maturity 10.706 7.054 10.623 0.315 67.156 10.416** 6.799 Industry Median Leverage 0.155 0.140 0.078 0.030 0.456 0.153 0.143 Industry Median Short term Debt 0.482 0.418 0.223 0.030 1.000 0.469*** 0.437*** Industry Median Bank Debt 0.195 0.172 0.154 0.000 0.972 Term Premium (%) 1.552 2.008 1.270 0.360 3.375 1.574 2.008 Tightening Bank Lending (%) 8.371 1.375 2 7.265 22.200 65.200 6.762*** 1.375*** Bank Index (%) 0.070 1.098 2.455 7.225 5.157 0.162*** 1.141** Dividend Payer 0.412 0.000 0.492 0.000 1.000 0.378*** 0.000*** Investment Tax Credit 0.106 0.000 0.308 0.000 1.000 0.106 0.000 Investment Tax Credit Missing 0.309 0.000 0.462 0.000 1.000 0.301 0.000 Tax Loss Carry Forward 0.442 0.000 0.497 0.000 1.000 0.440 0.000 Tax Loss Carry Forward Missing 0.367 0.000 0.482 0.000 1.000 0.370 0.000 Below Investment Grade 0.206 0.000 0.404 0.000 1.000 0.201 0.000 No Rating 0.574 1.000 0.494 0.000 1.000 0.608*** 1.000***

PAGE 95

95 Table 3 2. Pearson correlation coefficients between variables of interest. This table reports Pearson correlations between leverage, short term debt, bank debt, market to book, and z score of CIQ Compustat nonfinancial firm year observations from 2003 to 2009. Year 2002 data is excluded from the estimation because lagged industry median bank debt information is not available until year 2003. Panel A displays the correlations using all sampled firms, while Panel B and Panel C demonstrate the correlations using firms with/ without an S&P long term debt rating. Variable definitions are provided in Appendix A I use***, ** and to denote significance at the 1% level, 5% level and 10% level, respec tively. Market Leverage Short Debt Bank Debt Market to Book Z Score Panel A: All Firms (13,374 Firm years) Market Leverage 1.00 Short Debt 0.29*** 1.00 Bank Debt 0.04*** 0.24*** 1.00 Market to Book 0.34*** 0.10*** 0.06*** 1.00 Z Score 0.03*** 0.11*** 0.07*** 0.21*** 1.00 Panel B: Rated Firms (4,494 Firm years) Market Leverage 1.00 Short Debt 0.16*** 1.00 Bank Debt 0.23*** 0.04*** 1.00 Market to Book 0.45*** 0.14*** 0.05*** 1.00 Z Score 0.42** 0.13*** 0.03*** 0 .29*** 1.00 Panel C: Unrated Firms (6,066 Firm years) Market Leverage 1.00 Short Debt 0.24*** 1.00 Bank Debt 0.09*** 0.16*** 1.00 Market to Book 0.30*** 0.04*** 0.11*** 1.00 Z Score 0.02 0.11*** 0.14*** 0.26*** 1.00

PAGE 96

96 Table 3 3 Jo int determinants of leverage, short term debt, and bank debt with no firm and year fixed effect. This table presents the results of second stage simultaneous equations estimated by two stage least squares with no firm and year fixed effects. First stage re gressions results are omitted. Sample includes 10,560 firm year observations from year 2003 to year 2009. Year 2002 data is excluded from the estimation because lagged industry median bank debt information is not available until year 2003. Variables are de fined in Appendix A Standard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, **, and to denote significance at the 1% level 5% and 10% level, respectively. Two Equation System Three Equation Syste m Leverage Short Debt Leverage Short Debt Bank Debt Market Leverage t+1 0.197* 0.045 0.477*** (0.110) (0.114) (0.179) MB t *Market Leverage t+1 0.525*** 0.123 0.131 (0.168) (0.086) (0.168) Z score t *Market 0.002 0.035 0.111*** Leverage t+1 (0.041) (0.030) (0.037) Short Debt t+1 0.626*** 0.393*** 0.253** (0.046) (0.062) (0.121) MB t *Short Debt t+1 0.114*** 0.104*** 0.073** (0.012) (0.017) (0.034) Z score t *ShortDebt t+1 0.009 0.031*** 0.016 (0.008) (0.010) (0.013 ) Bank Debt t+1 0.573*** 0.481*** (0.071) (0.116) MB t *Bank Debt t+1 0.060** 0.050 (0.025) (0.038) Z score t *Bank Debt t+1 0.044*** 0.026* (0.016) (0.014) MB t 0.102*** 0.013* 0.065*** 0.026** 0.050* (0.008) (0.007) (0.01 3) (0.012) (0.027) Z score t 0.010 0.001 0.019* 0.005 0.010 (0.007) (0.004) (0.010) (0.005) (0.012) Fixed Assets t 0.003 0.124*** 0.086*** 0.176*** 0.066 (0.019) (0.036) (0.023) (0.034) (0.045) Profitability t 0.023 0.014 0.099** 0.073 0.23 2*** (0.025) (0.048) (0.043) (0.052) (0.055) Firm Size t 0.014*** 0.056*** 0.009*** 0.058*** 0.009 (0.002) (0.013) (0.003) (0.012) (0.014) Firm Size Square t 0.002** 0.004*** 0.002** (0.001) (0.001) (0.001) CF Volatility t 0.075** 0.135* 0.009 0.324*** 0.181** (0.030) (0.082) (0.043) (0.079) (0.084) Abnormal Earnings t 0.060*** 0.038*** 0.038*** 0.020** 0.006 (0.006) (0.010) (0.007) (0.009) (0.010)

PAGE 97

97 Table 3 3 Continued Two Equation System Three Equation System Leverage Short D ebt Leverage Short Debt Bank Debt Stock Return t 0.023*** 0.006 (0.004) (0.005) Asset Maturity t 0.001* 0.000 (0.001) (0.001) Industry Median Leverage t 0.274*** 0.277*** (0.038) (0.046) Industry Median Short term Debt t 0.1 06*** 0.174*** (0.033) (0.027) Industry Median Bank Debt t 0.247*** (0.039) Term Premium t 0.014*** 0.016*** (0.003) (0.003) Tightening Bank t 0.000** (0.000) Bank Index t 0.009*** (0.003) Dividend Payer t 0.027*** 0.009 (0.007) (0.007) Investment Tax Credit t 0.040*** 0.057*** 0.015 0.015 (0.008) (0.016) (0.010) (0.016) Investment Tax Credit Missing t 0.015*** 0.018* 0.008 0.000 (0.006) (0.010) (0.007) (0.011) Tax Loss Carry For ward t 0.031*** 0.035** 0.017* 0.025* (0.008) (0.014) (0.009) (0.013) Tax Loss Carry Forward Missing t 0.019** 0.027** 0.018** 0.028** (0.008) (0.013) (0.008) (0.013) Below Investment Grade t 0.076*** 0.152*** 0.116*** (0.023) (0.018) (0. 024) No Rating t 0.024** 0.112*** 0.083*** 0.047* 0.223*** (0.011) (0.017) (0.017) (0.028) (0.024) _cons 0.599*** 0.705*** 0.174*** 0.426*** 0.079 (0.032) (0.059) (0.052) (0.068) (0.112) Firm FE N N N N N Year FE N N N N N P value of Under Identi fication Test 0.0000 0.0000 0.0000 0.0000 0.0000 P value of Over Identification Test 0.0000 0.0007 0.0000 0.0002 0.0000

PAGE 98

98 Table 3 4. Joint determinants of leverage, short term debt, and bank debt for rated and unrated firms. This table presents the re sults of second stage simultaneous equations estimated by two stage least squares with no firm and year fixed effects. System (1) is estimated based on all firm year observations. System (2) and (3) are based on 4,477 and 6,018 firm year observations with and without S&P issuer credit rating respectively. First stage regressi ons results are omitted. Sample period is year 2003 to year 2009. Year 2002 data is excluded from the estimation because lagged industry median bank debt inform ation is not available until year 2003. Variables are defined in Appendix A Standard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, **, and to denote significance at the 1% level, 5% level, and 10% level, respectively. Full Sample (1) Rated Sample (2) Unrated Sample (3) Leverage Short Debt Bank Debt Leverage Short Debt Bank Debt Leverage Short Debt Bank Debt Market Leverage t+1 0.045 0.477*** 0.555*** 0.674*** 0.171 0.319 (0.114) (0.179) (0.164 ) (0.227) (0.232) (0.218) MB t *Market Leverage t+1 0.123 0.131 0.131 0.153 0.120 0.107 (0.086) (0.168) (0.100) (0.138) (0.136) (0.170) Z score t *Market 0.035 0.111*** 0.044 0.082 0.026 0.126*** Leverage t+1 (0.030) (0.037) (0.076) (0 .071) (0.040) (0.043) Short Debt t+1 0.393*** 0.253** 0.255 0.133 0.286*** 0.624*** (0.062) (0.121) (0.170) (0.266) (0.066) (0.165) MB t *Short Debt t+1 0.104*** 0.073** 0.139*** 0.138 0.046*** 0.075** (0.017) (0.034) (0.050) (0.091) (0.017) (0.038) Z score t *ShortDebt t+1 0.031*** 0.016 0.039 0.085 0.014 0.003 (0.010) (0.013) (0.059) (0.085) (0.009) (0.013) Bank Debt t+1 0.573*** 0.481*** 0.465*** 0.036 0.528*** 0.721*** (0.071) (0.116) (0.083) (0.180) (0.077) (0.1 69) MB t *Bank Debt t+1 0.060** 0.050 0.021 0.070 0.087*** 0.081* (0.025) (0.038) (0.046) (0.062) (0.023) (0.047) Z score t *Bank Debt t+1 0.044*** 0.026* 0.026 0.046 0.036*** 0.020 (0.016) (0.014) (0.035) (0.058) (0.011) (0.017) MB t 0.065*** 0.026** 0.050* 0.113*** 0.083*** 0.089* 0.018 0.031** 0.045 (0.013) (0.012) (0.027) (0.024) (0.022) (0.048) (0.017) (0.013) (0.030) Z score t 0.019* 0.005 0.010 0.039* 0.020 0.014 0.007 0.002 0.003 (0.010) (0.005) (0.012) (0.018) (0.025) (0.030) (0.006) (0.009) (0.015) Firm FE N N N N N N N N N Year FE N N N N N N N N N

PAGE 99

99 Tab le 3 5. J oint determinants of leverage, short term debt, and bank debt with firm and year fixed effect. This table presents the results of second stage simu ltaneous equations estimated by two stage least squares with firm and year fixed effects. System (1) is estimated based on all firm year observations. System (2) and (3) are based on 4,477 and 6,018 firm year observations with and without S&P issuer credit rating respectively. First stage regressi ons results are omitted. Sample period is year 2003 to year 2009. Year 2002 data is excluded from the estimation because lagged industry median bank debt information is not available until year 2003. Var iables are defined in Appendix A Standard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, **, and to denote significance at the 1% 5% and 10% level, respectively. Full Sample ( 1) Rated Sample (2) Unrated Sample (3) Leverage Short Debt Bank Debt Leverage Short Debt Bank Debt Leverage Short Debt Bank Debt Market Leverage t+1 0.076 0.580** 0.532* 0.730*** 0.058 0.588* (0.205) (0.228) (0.312) (0.246) (0.330) (0.331) MB t *Market Leverage t+1 0.355** 0.108 0.123 0.019 0.401** 0.185 (0.140) (0.158) (0.145) (0.119) (0.175) (0.200) Z score t *Market 0.029 0.024 0.119 0.022 0.000 0.054 Leverage t+1 (0.033) (0.029) (0.087) (0.118) (0.035) (0.037) Short Deb t t+1 0.139** 0.129 0.039 0.161 0.082 0.020 (0.065) (0.152) (0.139) (0.290) (0.066) (0.186) MB t *Short Debt t+1 0.030** 0.073** 0.068* 0.128** 0.011 0.044 (0.014) (0.028) (0.037) (0.063) (0.013) (0.038) Z score t *ShortDebt t+1 0.027 *** 0.019 0.068 0.154 0.011 0.025 (0.007) (0.015) (0.052) (0.132) (0.007) (0.017) Bank Debt t+1 0.429*** 0.502** 0.319*** 0.103 0.200** 0.162 (0.093) (0.246) (0.124) (0.369) (0.081) (0.291) MB t *Bank Debt t+1 0.043** 0.051 0.038* 0 .011 0.045*** 0.031 (0.018) (0.039) (0.020) (0.074) (0.017) (0.044) Z score t *Bank Debt t+1 0.011 0.035 0.014 0.110 0.013 0.022 (0.013) (0.029) (0.032) (0.083) (0.009) (0.027) MB t 0.013 0.030** 0.059** 0.061*** 0.045* 0.077** 0.018 0.023* 0.044 (0.011) (0.014) (0.023) (0.014) (0.025) (0.031) (0.013) (0.014) (0.029) Z score t 0.019*** 0.006 0.013 0.013 0.068*** 0.057* 0.004 0.005 0.017 (0.007) (0.010) (0.013) (0.018) (0.025) (0.030) (0.006) (0.009) (0.015) Firm FE Y Y Y Y Y Y Y Y Y Year FE Y Y Y Y Y Y Y Y Y

PAGE 100

100 CHAPTER 4 THE EFFECT OF BANK SHOCKS ON COR PORATE FINANCING AND INVESTMENT : EVIDENCE FROM 2007 2009 FINANCIAL CRISIS Banks play an important role in providing financing to corporations. Bank lending is important not on ly to small firms with limited access to the public debt market, but also to large and medium sized companies. Using the Kauffman Firm Survey, Robb and Robinson (2009) find that bank loans and business credit cards are the primary sources of financing for first year start up firms. Sufi (2009) show that around 85% of large publicly listed corporations obtain a bank line of credit. Theories motivate the importance of banks as they can provide delegated monitoring service (Diamond (1984)), screen out poor qua lity borrowers (Leland and Pyle (1977), Ramakrishnan and Thakor (1984)), provide flexible loan renegotiation (Berlin and Loyes (1988), Gertner and Scharfstein (1991)), and serve as an important liquidity cushion for systematic liquidity shocks (Kashyap et al. (2002) and Gatev and Stranhan (2006)). Although bank credit is viewed as an important source of funding to corporations in the literature, when examining corporate financing and investment decisions, most of the empirical work analyzes these corporat e decisions as a function of firm corporate financing and investment decisions. Though some papers investigate how (for example Bernanke (1983)), their method suffers from the critique that a reduction on bank lending may be driven by aggregate production or technology shocks from the demand side. Recently, several studies (Peek and Rosengren (1997, 2000), Khwaja and M ian (2008), Schnabl (2011)) address these critiques by identifying exogenous shocks on banks and examining the heterogeneous response of corporate borrowers to banks with different level of shocks.

PAGE 101

101 This study financial conditions impact corporate financing and investment decisions using the 2007 2009 financial crisis as an experimental setting. I am interested in the recent financial crisis for the following reasons. First, it is the largest shock to the banki ng system since the great depression and merits a systematic examination of whether and how shocks on banks impacted the real sector. 1 Second, the recent financial crisis offers a nice setting to mitigate the confounded demand side effect. During the crisi portfolios in general took a big hit because of the lax lending standard during the boom period and a bust of real estate value during the crisis. Yet the deterioration of these real estate loans was not expected by banks and crea ted substantial variation across banks and over time. Moreover, the losses on the residential or commercial real estate loans made by banks before demand before and during the crisis. The arguably impact corporate financing and investment decisions. To my best knowledge, my study is the most comprehensive study that investigates the impa ct of bank shocks on corporate financing and investment decisions during the 2007 2009 financial crisis. Although some evidences indicate that during the crisis more affected banks lent less or charged higher loan price (Ivashina and Scharfstein (2009), Co rnett et al. (2010), Santos (2010)), few papers provide direct 1 During the financial crisis, US banks wrote down $680 billion dollars on their balance sheets between the second quarter of 2 007 and the fourth quarter of 2009 1 and reduced their new loan issuance by 79% at the same time (Ivashina and Scharfstein (2010)).

PAGE 102

102 investment decisions My study intends to shed light on these issues. I break down my investigation into three steps The first part of the study examines whether firms were given less bank credit when their relationship banks were more adversely affected during the crisis. I hypothesize that firms whose banks experienced more negative shocks obtained less bank credit t hrough the bank lending channel (Bernanke and Blinder (1988)). I further break down bank debt into credit line drawdown and term loans and investigate how these two types of bank debt reacted differently to shocks on banks. Credit lines and term loans have different contractual features. Term loans have a fixed amortization schedule at the issuance, whereas credit lines can be report that during the recent crisis, banks w ith worse financial conditions experienced with a more troubled bank to take more ban k credit and therefore confounds with my loans because of their fixed amortization schedule. Therefore, I expect to observe that ore negatively to shocks on banks compared to credit lines. The second part of the study investigates whether and how firms adjust their financing decisions in response to the negative shock from their relationship banks. In the previous literature, trade credit and public debt are often viewed as alternative sources of credit to bank credit. Petersen and Rajan (1997) report that trade credit is an

PAGE 103

103 importance source of short term finance and a substitute of bank credit. The choice between bank debt and publ ic bond is also extensively studied in the previous theoretical and empirical studies (for example Diamond (1991), James and Houston (1997)). Therefore, I examine whether trade credit or public debt was used to replace reduced bank debt. I also investigate how firm leverage responded to shocks on banks to provide indirect evidence on whether bank debt was replaced by other types of debt. In the third part of the study I examine whether shocks to the banking sector ultimately impact corporate real activitie s. I first examine whether firms with more troubled banks reduced their investment more than firms with healthier banks. I further investigate whether firms with more troubled banks increased their cash holdings for precautionary motive so that firms can u se cash to finance valuable feature investment opportunities when bank credit is not available or is excessively costly. To test my hypotheses, I face an empirical challenge of disentangling the credit supply from the demand side. I tackle this issue by i dentifying an exogenous source of shock on banks that affects banks heterogeneously but arguably has little to do with corporate demand for bank credit. Specifically, I level of non performing loans excluding commercial and industrial loans. Bank non performing loan ratio reflects banks asset quality and is more difficult to manipulate than other accounting indicators because of the objective definitions (i.e. loans overdue for more than 90 days). Excluding the commercia l and industrial loans from the non performing loan measure further assures the exogeneity of the shock measurement. During the crisis period the share of commercial and industry loans in nonperforming

PAGE 104

104 loans portfolio was never above 12 percent. 2 This miti gates the reverse causality and impacted corporate financing and investment decisions. To further partial out the demand effect, I exclude construction related firms from my sample, because their investment opportunities can change with the boom and the bust of real estate market. I also control for a comprehensive set of firm characteristics financing and investment and include firm fixed e ffects to exploit within firm variation. Overall, I find the following main results. First, I show that firms increased their overall level of bank debt during the crisis. Both term loans and credit line drawdowns increased during the crisis. However, fir ms with more adversely affected banks did not use as much bank debt as firms with less affected banks. This effect manifested mainly affected the supply of bank credit to co rporate borrowers during the crisis. Second, I find that firms with adversely affected banks did not replace the reduced bank credit with other source of credit during the crisis. Public debt and trade credit did not increase for firms with more distressed banks with respect to other firms. The leverage of firms with more troubled banks decreased more than that of firms with healthier banks. Third, I find some evidence of real effects associated with bank shocks. Firms with more troubled lenders invest less during the crisis compared to those with healthier banks. There is weak evidence that they tended to hoard more cash on their balance sheet during the crisis for precautionary reasons. 2 This number can be compared with the 20% of commercial industrial loans out of total bank loan portfolio during my sample p on the assets and liabilities of US commercial banks. http://www.federalreserve.gov/releases/h8/

PAGE 105

105 I contribute to the literature by providing direct evidence on how b conditions affect corporate financing and investment decisions at the firm level. Ivashina and Scharfstein (2009) and Cornett et al. (2010) find that banks with severe liquidity shortage reduce their commercial and industrial lending during the 2007 2009 financial crisis. Nevertheless, these two studies remain silent on how such loan contractions affect corporate decisions. While Duchin et al. (2009) and Campello et al. (2010) investigate how financial constraints affect corporate investment during the recent Carvalho et al. (2011) investigate how bank stock performance is related to the borrower stock performance but provide limited evidence on firm financ ial decisions. Santos (2011) and Kwan (2011) examine the effect of bank financial conditions on loan pricing but remain silent on the overall effect of bank financial condition on corporate decisions. The remainder of the study is organized as follows. Fi rst I describe conditions during the 2007 2009 financial crisis. Next, I introduce the data, sample description, and the e mpirical methodology. I then present the results of my analysis and conclude with a summary of the findings 2007 200 9 Financial Crisis At the heart of the 2007 2009 crisis is the meltdown of the residential and commercial real estate market. US house prices on average declined 30% from the 2006 peak to early 2010 (Furlong and Knight (2009)). This broad based decline in house prices had been widely considered unlikely given the behavior of house prices over the post World War II period, despite of signs of overvalued real estate properties before the bust (Krainer (2004)).

PAGE 106

106 Banks took a big hit during the crisis from bad l oans and collateralized debt obligation. According to the Global Financial Stability Report, U.S. banks wrote down $680 billion dollars on their balance sheets between the second quarter of 2007 to the fourth quarter of 2009. 3 Banks also experienced liquid ity shortages as it got hard to finance themselves from the commercial paper and interbank borrowing market during the crisis. new commercial and industrial loans fell dramatically. At the same time, the total amount of outstanding bank loans increased. Ivashina and Scharfstein (2010) document a decrease of new loans to large borrowers by 79 percent from the second quarter of 2007 relative to the fourth quarter of 2009. They point out that the increase in bank loans duri ng the financial crisis is mainly due to increased drawdown of credit lines. I investment decisions using the 2007 2009 financial crisis as an experimental setting for the following re asons. First it is the largest shock to the banking system since the great depression. The exceptional magnitude of shock on banks and contraction of bank lending merits a systematic examination of whether and how bank shocks impacted the real sector. The shock on the banking sector was particular problematic given that some alternative financing sources were also limited. The commercial paper market collapsed after the bankruptcy of Lehman Brothers in September 2008 and junk bond issuance dropped 66 percen t due to flight to quality. 3 ng that time, this

PAGE 107

107 Moreover, the recent financial crisis serves as a convenient experiment setting in the sense that the crisis was not expected by banks ex ante, and had a detrimental but heterogeneous impact on banks ex post. While the bust of the housing bubble left banks sitting on mounting bad loans, banks were affected differently depending on their exposure to the housing bubble. Yet, when facing loses on their real estate loan portfolio, more affected banks may have become more risk avers e and reduced their overall lending. This reduction can include some commercial and industry loans, which presumably banks would be willing to lend, had they not being as risk averse from their capital losses. In this way, banks passed the shock from the r esidential or commercial market to the relatively healthier corporate sector. second quarter of 2008 supports the above statement. Neither corporate bond spreads nor bankrup tcy filings rose dramatically until September of 2008 (Huang (2010)). The level of bank non performing loans (NPLs) was 0.5 percent of total bank assets at the beginning of 2007 and jumped to 2.95 percent at the end of 2009, a five fold increase. In contra st, the component of bank commercial and industrial NPLs only increased from 0.11 percent of total bank assets at the beginning of 2007 to 0.36 percent at the end of 2009. In this study I explore the cross sectional difference in shocks across banks and o I utilize this variation to decisions, when the nonfinancial sector was less directly affected by the financia l crisis.

PAGE 108

108 Data, Sample Description and Methodology Data and Sample Description My main sample consists of yearly data on non financial and non utility firms drawn from Compustat. Specifically, I exclude firms with SIC codes between 6000 6999 (financial fir ms), and between 4900 4949 (utility firms). To mitigate the concern that changes in credit demand may drive my results, I exclude firms in the construction industry (SIC code 1500 1700) which were hit dramatically during the subprime crisis My sample peri od starts from 2003 and ends at 2009. 4 I take the following steps to construct my sample. First, I use the Loan Pricing bank lending relationship. According to Ivashina (2009) and Standard and Poor (2 006), lead banks are the primary banks responsible for ex ante due diligence and for ex post monitoring of the borrower, the types of services motivated in theory. Therefore, I identify a lending relationship only if a bank serves as a lead lender for a fi rm. I define lead lenders following Bharath et al. (2011) 5 I examine the Dealscan database to identify lead lenders in a five year rolling window for each firm year observation lagged by half a year. 6 For example, if I examine the lending relationships o f company X on December 31, 2005, I investigate all lead banks of company X from whom it took loans. I then assign weights to different lead banks according to the loan amount lead led by the banks. Assume bank A lead lent a 4 I use this sample period to alleviate the concern that during the 2000 2002 mild recession firms might have fewer investment opportunities and reduce their demand for bank cred it, which may reflect in my measurement of non performing loan. If I include year 2002 in my sample, my results are even stronger. 5 They define the lead lenders as those who are the sole lender on a loan, or are credited as more than 25% share of the loan. 6 I have the half year lag to allow for the new lending relationship to have some effect on corporate decisions.

PAGE 109

109 loan of 100 million dollars on Jun. 1, 2001. Bank A keep 40% of the loan and the remaining 60% is shared by other participant banks. Bank B and C lead lent a loan $200million on Sept. 2 2004. Bank B and C retain 20% and 30% of the loan. I assign 33.3% to bank A, 0.267 to bank B, and 0.4 to bank C. 7 My measurements of bank characteristics, including the bank shock variable, is calculated as a weighted average of all lead banks. I then link borrowing firms with Compustat to retrieve their accounting information and with CRSP to obtain sto ck information. 8 My bank debt and public debt data comes from Capital IQ 9 which reports detailed breakdown of corporate debt at a yearly basis. Houston and James (1997) find evidence that bank holding companies can form an internal capital market and all ocate their capital among various subsidiary banks. Therefore, I company parent (BHC) 10 My sample of lenders only includes US BHC 11 My f inal sample includes 4,442 firm year observa tions. Empirical Strategy Identifying a linkage between bank financial condition and corporate decisions is challenging, as the results could be confounded with the credit demand. That is, shocks to the banking sector during the 2007 2009 financial crisis were accompanied by 7 The weights are calcul ated as 100/300 for bank A; =0.267 to bank B, and =0.4 to bank C. 8 The link table is kindly provided by Michael Roberts. 9 To ensure my data quality, I exclude those firm year observations in m y bank debt and public debt sample, whose total debt reported on CIQ differs from the record in Compustat by 10%. 10 I 11 To ensure that the measurements of BHC financial conditions are representative of the financial conditions of overall lenders, I constrain my sample to those firm year observations whose majority lenders are subsidiary of US BHCs.

PAGE 110

110 changes in investment opportunities and therefore may just proxy for changes in credit demand. To disentangle the supply and demand effects on corporate decisions, I adopt the following identification strategies. First of all, I identi ty a source of shock on banks that is believed to be exogenous I use the level of bank non performing loan as the turned out to be proble matic (i.e. non performing) were extended during the credit boom and were largely unexpected by banks (Furlong and Knight (2010)). Such unexpected bank asset deterioration varied across lenders and over time. Additionally, the majority of nonperforming loa n portfolios is comprised of real estate loans, and the proportion of real estate non performing loans increased from around 80 percent of the total non performing loans before crisis to 9 0 percent after crisis. The figure in Appendix F illustrates these statements. This mitigates the endogeneity issue that the increase in Second, I demand for financing and inve stment. Third, my empirical models include firm fixed effects that exploit within firm variation. With this specification, the effect of bank non performing loans on associated same firm over time. This fixed effect specification removes all firm specific time invariant corporate financing and investment decisions. For example, firms with good pri vate projects may finance with banks of strong financials and these firms may also obtain

PAGE 111

111 more credit from banks for their positive prospect My results can be driven by the unobserved firm characteristics. By adding firm fixed effects in the model, I can mitigate the omitted time invariant firm characteristics and explore changes in the financing and investment decisions within the same firm. Model This section describes the empirical model My identification strategy is based on the premise that the fina financial positions, which induces banks to reduce credit supply to corporations and affect corporate financing and investment decisions. I test my hypotheses by interacting the crisis dummy with bank non performing loans. Specifically, I estimate the following regression model Appendix C specifies definitions of variables in the model Y i,t = b 0 + b 1 crisis + b 2 crisis*bank NPLs i,t 1 + b 3 bank NPLs i,t 1 + b 4 firm controls i,,t 1 +b 5 bank control s, t 1 + Firm FE i + e it ( 4 1) Y i,t : public debt, book and market leverage, trade credit, cash holding, and capital expenditures. Crisis : A dummy variable that equals 1 if the firm year observation is at or after 2008 and 0 otherwise. Bank NPLs : Bank NPL excluding commercial and industrial loans. NPL level variables due to its objec tive definition (i.e. loans overdue for more than 90 days). This variable is the weighted average across all lead lenders. The weights are described in the Data and Sample Description section.

PAGE 112

112 Firm controls : Firm fundamental variables that are determinan financing and investment decisions. Specifically, I identify the bank debt and public debt equation by adding log of inflation adjusted total assets, market to book asset, EBITDA over total assets, fixed assets over total assets, z score, as set volatility, and log of firm age following Petersen demand for credit, access to capital sources, information asymmetry, and credit worthiness. In the leverage equation I include log of inflation adjusted total assets, market to book asset, EBITDA over total assets, fixed assets over total assets, asset volatility, industry median leverage ratio, and dividend payout dummy given their strong theoretical and empirical pre (2008), Frank and Goyal (2009)). 12 In the cash equation, I include log of inflation adjusted total assets, market to book asset, EBITDA over total assets, net w orking capital to total assets industry ca sh flow risk, dividend payout dummy, and R&D to sales in the cash holding equation following Opler et al. (1999). As in the investment literature, I industry wide investment demand shifts over time, I also control for industry median capital expenditure. 12 I include additional firm controls such as z score, inves tment tax credit, tax loss carry forward. My main results still hold.

PAGE 113

113 Bank controls supply of credit 13 They include the Tier 1 capital ratio, core deposit ratio, liquid asset ratio, off balance sheet loan unused commitments and return on assets. Descriptive Statistics Table 4 1 to 4 4 present descriptive statistics of my sample. Table 4 1 provides summary statistics for corporate borrowers. Bank debt is an important source of financing for firms. In my sample, ba nk debt counts for 9.7 percent of total assets or around 38 percent of total debt. Term loans and credit line drawdowns each constitute 50 percent of total bank debt. Public debt and trade credit are two other important sources of funding, accounting for 1 3.9 percent and 8 percent of total assets respectively. The level of book leverage and market leverage is similar as documented in the previous literature (for example Lemmon, Roberts, and Zender (2008), Sufi (2007)). In an unreported analysis, I find that my sampled firms are bigger and older, more profitable, and have fewer investment opportunities than the Compustat universe. This suggests that my sample includes more mature firms. This partly comes from the fact that my sample requires a previous lendin g relationship with a bank recorded according to the issuance history from the Dealscan. Younger firms may rely more on equity than bank debt for financing (Bolton and Freixas (2000)). Given that my sample of firms is older and more mature, they can access other sources of funding relatively easily (for instance, to the public bond market). This can bias against rejecting the null. Table 4 2 reports the summary statistics for BHCs. My sampled BHCs in general are larger banks who play a major role in the sy ndication market. 13 I include these control variables to ensure that my of enhancing their capital and liquidity positions or capital/liquidity injection by the federal government.

PAGE 114

114 Table 4 3 compares borrower financing and investment before crisis (year 2003 to financial and investment decisions altered during the crisis comp ared to their before crisis level. During the crisis, firms relied more on bank debt. Asset scaled bank debt increased by 3.9 percentage points during the financial crisis compared to the pre crisis level period, a 46 percent increase. 64 percent of the in crease came from term loans, whereas drawdown on credit lines accounts for the other 36 percent. Public debt remained the same before and during the crisis whereas trade credit decreased marginally. During the crisis period, firm book (market) leverage inc reased by 4.5 (11.3) percentage points, suggesting that firms used more debt on their balance sheet. Firms reduced their capital expenditure and hold less cash during the financial crisis. Table 4 4 provides summary statistics for the BHCs in my sample, an d it shows that the financial condition of banks deteriorated during the subprime crisis My key variable bank noncommercial and industrial NPL increased by average of 143 percent, suggesting that this increase is widespread among banks. Bank liquid assets and core deposits also reduced and profitability dropped dramatically during the crisis period. Empirical Results Bank Debt I first present the sets of empirical results on bank debt. If shocks on the banks can affect the corporate sector, I should obser ve a first Table 4 5 presents the results on bank debt. I level before the crisis to the bank debt level during the cr isis, controlling for firm characteristics and fixed effects. Column (1) shows that the Crisis dummy is positive

PAGE 115

115 and highly statistically significant, indicating that firms relied on more bank debt during the crisis. On average, bank debt increased by 2.7 percentage points during the crisis. This indicates that other types of financing may be even harder to obtain than bank debt during the crisis. Larger firms with more investment opportunities and less asset risk use more bank debt, suggesting that compare d to other non bank sources of funding including equity, bank debt is used more often by mature firms. The coefficient on z score indicates that more distressed firms use more bank debt. This is consistent with the argument of Berlin and Loyes (1988) that bank debt can provide valuable financing flexibility to more distressed firms. Next, I investigate whether firms whose relationship banks incurred more NPLs obtained less bank debt during the crisis compared to firms with healthier banks. Column (2) inves tigates this hypothesis, controlling for firm specific controls and firm fixed effects. The coefficient on the interaction variable Crisis x NPL_noC&I s is negative and statistically significant, indicating that firms with more troubled banks obtained less bank credit during the crisis compared firms with healthier banks. The coefficient on NPL_noC&I is positive and statistically significant, suggesting that banks with a higher level of NPLs extended more credit before the financial crisis, an indicator of aggressive lending. However, during the crisis, banks with a higher level of NPLs lent less compared to healthier banks. 14 This suggests that banks lent less to corporate borrowers during the crisis after they experienced more asset deterioration in their n on C&I loan portfolio. 14 The point estimate is 3.052+2.272= 0.78, and is statistically significant at the 1% level.

PAGE 116

116 Column (3) shows that my main results continue to hold when I expand my controls to account for other bank characteristics that might affect their willingness to lend. Among the bank control variables, I notice that Bank Tier 1 Rati o is negative and statistically significant; indicating that bank with higher regulatory capital reduced its supply of credit more than other banks. I will defer the discussion of this effect to the section where I break down the total debt into credit lin e drawdown and term loans (see Tabl e 4 6 and Table 4 7 ). Column (4) demonstrates that my main findings still hold when I further control for year dummies to account for any macroeconomic shocks. 15 The economic impact of bank shocks on firm bank debt usage i s economically significant. According to column (4), one standard deviation increase in the level of NPLs leads to a decrease of 1.82 percentage points of bank debt during the crisis by corporate borrowers. 16 Compared to the 9.7 percent unconditional mean o f bank debt during the sample period, this is an increase of 18.8 percent. 17 These tests support my hypothesis that banks with greater asset quality deterioration lent less to their corporate borrowers during the crisis, compared to those healthier banks. I further break down bank debt into credit line drawdown and term loans and investigate how these two types of bank debt reacted differently to shocks on banks. Term loans have a fixed amortization schedule at the issuance and banks can cut off 15 Specifically, I include year dummy for 2004, 2005, 2006, and 2007. I 2008 and year 2009 to avoid perfe ct multicollinearity and to be able to estimate the crisis dummy. 16 This number is calculated as 2.282*0.008=1.83%. The standard deviation of the NPL comes from Appendix D which presents the weighted average of relationship banks for each firm. 17 Compare d to the standard deviation of bank debt, this leads to a 0.137 standard deviation change of bank debt. This is calculated as 0.0182/0.127=0.143.

PAGE 117

117 the supply o f new terms loans quickly if they wish to do so. In contrast, credit lines can by precautious corporate borrowers who drew down their credit lines at the concern of more troubled banks to take more bank credit and therefore confounds with my predict I expect to compared to credit line drawdowns during the crisis. Tables 4 6 and 4 7 present the results on credit line drawdowns and term loans, respectively. The Crisis dummy in column (1) of both tables is positive and highly statistically significant. Economicall y, the credit line drawdown (term loan) of an average firm increased by 1.3 (1.4) percentage points during the crisis. My empirical results confirm my hypothesis that term loans responded more negatively to bank shocks compared to credit line drawdowns du ring the crisis. In the regression of credit line drawdown, the coefficient on the interaction term Crisis x NPL_noC&I is negative but statistically insignificant across all specifications, providing no evidence that firms whose banks experienced more NPLs increased their drawdowns on credit lines during the crisis compared to firms with healthier banks. interaction term and bank supply shock effect predicts a negative sign, the insignificant sign can be driven by both effects. In comparison, Table 4 7 shows that the coefficient

PAGE 118

11 8 on the interaction variable Crisis x NPL_noC&I is negative and statistically significant across all specifications, suggesting that firms whose banks had more NPLs obtained less term loans during the crisis compared to firms with healthier banks. According to column (4) in Table 4 7 an increase of one standard deviation in NPLs leads to a decrease of 1.52 percentage point of term loans comparing the before and after crisis level. Compared to the 4.7 percent unconditional mean of term loan during the sample period, this is an increase of 32.3 percent, which is economically significant. 18 However, a Wald test of coefficient equality shows that the intera ction term Crisis x NPL_noC&I in the term loan regression and that in the credit line drawdown regression is not significantly different (p value is around 25 percent). As I pointed out earlier, Bank Tier 1 Ratio has a negative and statistically significa nt coefficient in the bank debt regression in Table 4 5 My results in Tables 4 6 and 4 7 show that this negative sign on Bank Tier 1 Ratio is driven by credit line drawdowns rather than term loans. Firms whose banks have a higher regulatory capital drew d own less on their credit lines than other firms. This is consistent with Ivashina likely to subject to run on credit lines by corporate borrowers. To sum up, my result s support the hypothesis that shocks on banks reduced bank credit supply. Although firms increased their reliance on bank debt during the recent financial crisis, firms with more troubled banks did not use as much bank debt as firms with healthier banks d uring the crisis. This effect manifested mainly on term loans rather than credit line drawdowns. 18 Compared to the standard deviation of term loans, this leads to a 0.137 standard deviation decrease of term loan. This is calculated as 0.0152/0.1=0.152.

PAGE 119

119 Trade Credit, Public Debt and Leverage As the investigation on bank debt suggests that firms with more adversely affected banks obtained less bank credit dur ing the crisis, I then investigate whether and how these firms adjusted their financing decisions in response to accommodate the negative shock from their relationship banks. Trade credit and public debt are often viewed as alternative sources of credit fo r bank debt. Therefore, I investigate whether firms with more troubled banks use relatively more of these two types of credit during the crisis. My results for trad e credit are reported in Table 4 8 Column (1) shows that the Crisis dummy is negative and statistically significant. During the crisis, average asset scaled trade credit decreased by 0.3 percentage points, a moderate decrease of 3.75 percent compared to its unconditional mean 8 percent. Across specifications in column (2) (4), the coefficient o n the interaction term Crisis x NPL_noC&I is negative but is not statistically different from zero. This suggests that firms whose banks experienced more statistically insignificant during both the pre crisis period and crisis period, indicating no impact of bank NPLs on firm trade credit. The signs on firm specific controls are generally consistent with the previous literature (for example Peterson and Rajan (1994)); s mall firms with more growth opportunities, fewer fixed assets, and higher degrees of financial distress use more trade credit. With respect to the pub lic debt, column (1) in Table 4 9 shows that the Crisis dummy is not statistically significant, indicating change during the crisis. The coefficient on the interaction variable Crisis x NPL_noC&I is negative but is not statistically different from zero across all specifications in column (2) (4). This indicates that fir ms whose banks experienced larger NPLs did not use more

PAGE 120

120 public debt during the crisis. The NPLs variable is statistically insignificant during both the pre crisis period and crisis period, indicating no impact of bank NPLs on firm public debt. I next inves tigate whether firms with more troubled banks reduced their leverage ratio, in order to provide indirect evidence on whether reduced bank credit was replaced by other types of debt for firms with more affected banks. Tables 4 10 and 4 11 present results on book leverage and market leverage, respectively. Column (1) of both tables shows that firm leverage increased during the recent financial crisis. However, such increase is less pronounced among firms with a higher level of NPL as the coefficient on the in teraction variable Crisis x NPL_noC&I is negative and statistically different from zero in column (2). These results hold even after I control for other bank characteristics and year dummies in column (3) and (4). According to column (4), an increase of on e standard deviation in the level of NPLs leads to a decrease of 2.45 (2.92) percentage point of book (market) leverage by corporate borrowers compared to their pre crisis level. Compared to the unconditional mean of 25.2 (24) percent, this is an increase of 9.7 (12.2) percent, indicating that the effect is economically significant. Together, these results are consistent with the view that firms with adversely the cri sis. Public debt and trade credit did not change for firms with more distressed banks compared to firms with healthier banks. Investment and Cash H oldings Given that firms associated with more troubled banks used less bank debt during the crisis and seeme d not to substitute the bank credit with other sources of credit, I proceed to investigate whether adverse shocks on the banking system ultimately impact

PAGE 121

121 corporate real activities. I first examine whether firms with more troubled banks reduced their invest ment than firms with healthier banks. I further investigate whether firms with more troubled banks increased their cash holdings for precautionary motive so that firms can use cash to finance valuable feature investment opportunities when bank credit are n ot available or are excessively costly. The results of investment, measured as capital expenditure, are reported i n Table 4 12 I during the crisis, controlling for a set of firm specific characteristics commonly used in the literature and firm fixed effects. Column (1) shows that the average capital expenditure scaled by total assets decreased by 0.2 percentage points during the crisis. With regards to firm controls, they are generally consistent with the previous literature. Firms with more investment opportunities and higher cash flows invest more on average. Next, I investigate whether firms whose banks experienced more NPLs during the crisis invested less. Column (2) (4) investigates this hypothesis, with different specifications. Column (2) only controls for firm characteristics and fixed effects. Column (3) further controls for bank characteristics. Column (4) adds year fixed effect and time varying industry median inves tment level to account for some unknown macroeconomic or industrial changes in firm investment. The coefficient on the interaction variable Crisis x NPL_noC&I is negative and weakly significant, indicating a possible real effect associated with bank shocks during the crisis. The coefficient on NPLs is insignificantly different from zero during the pre crisis period, but is weakly significantly negative during the crisis indicating that bank NPL did not have a major

PAGE 122

122 effect on firm investment before the finan cial crisis but decreased firm investment during the financial crisis. As data in the investment regression can be obtained at the quarterly frequency, which allows us to estimate the model with more statistical power and capture timely changes of both fir ms and banks, I rerun the above analyses using quarterly data and presents my results in Table 4 13 Results from the quarterly data suggest that bank shocks brought in real effects in the corporate sector. According to column (4) in Table 4 13 a change o f one standard deviation in the level of NPLs leads to a decrease of 0.16 per cent of investments by corporate borrowers. Compared to the unconditional mean of 1.6 percent, my results indicate a decrease by 8.75 percent. The results are therefore economica lly significant. Table 4 14 presents the results on cash. Column (1) shows that my Crisis dummy is positive and highly statistically significant, indicating that firms tend to hold more cash during crisis. Firm controls are generally consistent with the p revious literature (for example Opler et al. (1999)). Specifically, smaller firms with more investment opportunities and lower net working capital hold more cash. In column (2) (4) the interaction term Crisis x NPL_noC&I is positive and but statistically i nsignificant. As with the investment regression, I rerun the analyses using quarterly data. Results are presented in Table 4 15 The interaction term Crisis x NPL_noC&I is positive and weakly significant, providing some evidence that firms with more affect ed banks increased cash holdings more during the crisis compared to the pre crisis cash holding level. The NPLs variable is negative but statistically insignificant during the pre crisis period, but is weakly significantly positive during the crisis indica ting that bank NPL did

PAGE 123

123 not have a major effect on firm cash holding before the financial crisis but increased firm cash holdings during the financial crisis. Together, my results provide some evidence that shock on banks eventually transmitted to corporate real activities; firms with more adversely affected banks invest less and hoard more cash during the crisis compared to firms with healthier banks. Overall, my results suggest that adverse shocks on the banking system can curtail bank lending and negative ly affect the real sector. Robustness and Extension My results may be driven by a sample selection issue, i.e. riskier firms were matched with riskier banks, both of which were hit hard during the crisis. If this matching story holds, the reduction of firm bank debt usage and investment may be driven by reluctance of lending. To alleviate this concern, I compare borrower characteristics of banks with higher NPLs (above the mea n) to those of banks with lower NPLs (below the mean) during t he pre crisis period. Appendix E reports the comparison results My results show that firms with higher NPLs have slightly better credit quality than the firms with lower NPLs according to their pre crisis characteristics. Specifically, firms with riskier banks on average are older, bigger, have a higher cash flow and less asset volatility though they are slightly more distressed. These results are inconsistent with the firm bank matching story. Another concern is that there can be some measurement error in the borrower bank lending relationship. The lending relationship in my study is identified by the Dealscan loan issuance data and is assumed to hold within a five year period. However, a firm can endogenously switch to a healthier bank to refinance its loan if it is concerned

PAGE 124

124 relationship may exist in my sample adding noise to my measurement of lending relat ionship and bank shock. However, such noise should make my estimation less precise and bias us toward not finding any results. The fact that I still find significant results alleviates such measurement error concern. As in previous sections I find that f irms with more troubled banks used less bank of debt during the crisis. I then investigate whether these firms used other financing sources, such as equity, when the ir relationship banks were adversely impacted during the crisis. To answer this question, I categories: bank debt, public debt, trade credit, other liabilities, and equity. I run seemingly unrelated regre ssions to allow for correlated error terms across regressions due to the balance sheet constraint. Results are presented in Table 4 16 The results indicate that firms with more troubled banks reduced their reliance on bank debt, but increased their relat ive reliance on equity. I further investigate the change of absolute level of equity from the end of 2007 to the end of 2009 acr oss bank NPL quintiles. Table 4 17 displays the results. Firms with bank NPL in higher quintiles (quintile 4 and 5) increased mo re equity holdings compared to firms with bank NPL in lower quintiles (quintile 1 to 3). The last column in Table 4 17 reports the t statistics when comparing the mean of change in equity in quintile 5 vs. quintile 1 4. The results show some evidence that firms with more troubled banks used more equity during the crisis compared to those with healthier banks.

PAGE 125

125 Chapter 4 Concluding Remarks This study and investment decisions using the 2007 20 09 financial crisis as an experimental setting. I am interested in the recent financial crisis because the exceptional magnitude of shock on banks merits a systematic examination. This crisis was originated from the melt down of a housing bubble and banks were heterogeneously affected depending on their exposure to the real estate market. The nonfinancial sector was less directly affected until the later phase of the crisis. I utilize this cross sectional variation of shock on banks from the real estate mar ket to explore the different response of corporate financing and investment before vs. after the crisis. Specifically, I noncommercial and industrial nonperforming loans to measure shock on banks, given its relative exogeneity from corporate cr edit demand. My results suggest that the financial condition of banks is important to corporate financing and investment decisions. Overall, I find the following main results. First, I show that firms increased their overall level of bank debt during the c risis, the increase is not only due to the higher level of drawdown from their line of credit but also on term loans. This suggests that relative to alternative financing sources, bank debt is a relatively accessible source of funding. Second, I show that though firms in general relied more on bank debt during the crisis, those with more adversely affected banks did not use as much bank debt as firms with less affected banks during the crisis. This effect manifested mainly on term loans. Third, I find that for those firms with adversely affected banks, they did not replace the reduced bank credit with other source of credit during the crisis. Public debt and trade credit did not change for firms with more negatively affected banks in comparison to firms with healthier banks during the crisis.

PAGE 126

126 Indeed, firms with more adversely affected banks use less leverage during the crisis than those firms whose lenders were less affected. Moreover, I provide some evidence that shock on banks eventually transmitted to corp orate real activities; firms with more adversely affected banks invest less and hoard more cash during the crisis compared to firms with healthier banks. Overall, my results suggest that adverse shocks on the banking system can impact corporate financing a nd investment decisions.

PAGE 127

127 Table 4 1. Summary statistics of firm characteristics This table reports summary statistics of all nonfinancial firm year observations from year 2003 to 2009 Var iables are defined in Appendix C Variable Obs Mean Median Std. Dev. M in Max Dependent Variables Bank Debt 4,442 0.097 0.041 0.127 0.000 0.569 Drawdowns 4,442 0.048 0.002 0.082 0.000 0.391 Term Loans 4,442 0.047 0.000 0.100 0.000 0.519 Public Debt 4,442 0.139 0.117 0.145 0.000 0.657 Trade Credit 4,442 0.080 0.0 63 0.067 0.004 0.367 Book Leverage 4,442 0.252 0.229 0.182 0.001 0.916 Market Leverage 4,442 0.240 0.189 0.206 0.000 0.891 Capital Expenditure 4,442 0.016 0.011 0.018 0.000 0.104 Cash 4,442 0.105 0.058 0.126 0.000 0.646 Independent Variables S ize (Billion) 4442 5.102 0.988 13.454 0.019 101.308 Log(Size) 4,442 6.239 6.191 1.799 2.272 10.866 Market to book 4,442 1.412 1.146 0.910 0.361 5.650 Cash Flow 4,442 0.035 0.034 0.026 0.069 0.108 Z score 4,442 2.461 1.982 2.428 3.879 14.044 Fixed As sets 4,442 0.301 0.232 0.233 0.019 0.900 Firm Age (Year) 4,442 22.147 16.748 16.729 0.748 58.748 Log(Firm Age) 4,442 2.827 2.876 0.863 0.558 4.056 Assets Volatility 4,442 0.326 0.287 0.175 0.078 0.940 Industry Median Book Leverage 4,442 0.241 0.224 0.0 85 0.063 0.581 Industry Median Market Leverage 4,442 0.209 0.183 0.105 0.055 0.626 Industry Median Investment 4,442 0.012 0.008 0.011 0.013 0.070 Dividend Payer Dummy 4,442 0.436 0.000 0.496 0.000 1.000 R&D 1,596 0.013 0.007 0.019 0.000 0.109 Industr y Sigma 4,442 0.064 0.063 0.026 0.013 0.116 Net Working Capital Exclude Cash 4,442 0.086 0.075 0.154 0.303 0.524

PAGE 128

128 Table 4 2. Summary statistics of bank characteristics This table reports summary statistics of all matched bank holding companies from yea r 2003 to 2009 Var iables are defined in Appendix C Variable N Mean Median St. Dev Min Max Size 569 120.414 8.531 357.886 0.171 2268.347 NPL 569 0.012 0.006 0.018 0.000 0.172 NPL_noCI 569 0.010 0.004 0.016 0.000 0.164 Liq Assets 569 0.105 0.076 0.094 0 .007 0.519 Core Dep. 569 0.546 0.573 0.160 0.001 0.823 ROA 569 0.007 0.009 0.014 0.105 0.048 Tier1 569 0.112 0.105 0.039 0.000 0.368 Unused Commit 569 0.138 0.115 0.105 0.000 0.865

PAGE 129

129 Table 4 3. Firm f inancing and i nvestment b efore and a fter c risis c o mparison Before crisis period is defined as year 2003 to 2007 and crisis period i s defined as year 2008 and 2009 Var iables are defined in Appendix C Variable Before the Crisis During the Crisis Difference T stats Mean Mean Bank Debt 0.084 0.123 0.039 8.905 Drawdowns 0.044 0.058 0.014 5.024 Term Loans 0.039 0.064 0.025 6.904 Public Debt 0.139 0.138 0.001 0.210 Trade Credit 0.082 0.076 0.005 2.531 Book Leverage 0.237 0.282 0.045 7.379 Market Leverage 0.203 0.316 0.113 15.929 Capital Expenditu re 0.017 0.014 0.003 5.395 Cash 0.109 0.098 0.011 2.879

PAGE 130

130 Table 4 4 Bank c haracteristics be fore and a fter c risis c omparison Before crisis period is defined as year 2003 to 2007 and crisis period is defined as year 2008 and 2009 Var iables are define d in Appendix C Variable Before the Crisis During the crisis Difference T stats Size 108.199 160.457 52.258 1.252 NPL 0.008 0.025 0.017 7.742 NPL_noCI 0.007 0.022 0.015 7.472 Liq Assets 0.103 0.111 0.008 0.821 Core Dep. 0.555 0.518 0.037 2.305 ROA 0.011 0.006 0.016 8.684 Tier1 0.111 0.116 0.005 1.315 Unused Commit 0.142 0.125 0.017 1.842

PAGE 131

131 Table 4 5. Bank debt and bank financial conditions This table presents the results on bank debt and bank financial conditions. Sample includes 4,442 firm year observations from year 2003 to year 2009. Var iables are defined in Appendix C Standard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, **, and to denote significance at the 1% level, 5% lev el, and 10% level, respectively (1) (2) (3) (4) Crisis 0.027*** 0.047*** 0.038*** 0.033*** (0.005) (0.008) (0.008) (0.009) Crisis x NPL_noC&I 3.052*** 2.564** 2.282* (1.130) (1.162) (1.182) NPL_noC&I 2.272** 2.850** 2.657** (1.128) (1.184) (1.238) S ize 0.023*** 0.022*** 0.020** 0.018** (0.008) (0.008) (0.008) (0.008) Market to book 0.022*** 0.021*** 0.021*** 0.020*** (0.007) (0.007) (0.007) (0.007) Cash Flow 0.004 0.025 0.028 0.019 (0.128) (0.128) (0.129) (0.130) Z score 0.018*** 0.018 *** 0.018*** 0.018*** (0.003) (0.003) (0.003) (0.003) Fixed Assets 0.017 0.023 0.026 0.036 (0.047) (0.047) (0.046) (0.047) Log Firm Age 0.007 0.003 0.005 0.024 (0.017) (0.017) (0.018) (0.020) Asset Volatility 0.093*** 0.094*** 0.096*** 0 .097*** (0.014) (0.014) (0.014) (0.014) Bank Controls: Bank Size 0.003 0.004 (0.005) (0.005) Bank ROA 0.276 0.270 (0.497) (0.498) Bank Tier1 ratio 0.522** 0.699*** (0.252) (0.269) Bank CoreDep 0.050 0.014 (0.042) (0.042) Bank Liq Assets 0.021 0.161 (0.122) (0.134) Bank Unused Comm. 0.032 0.017 (0.044) (0.046) Constant 0.002 0.011 0.125 0.212 (0.059) (0.059) (0.133) (0.147) Firm FE Y Y Y Y Year FE N N N Y Number of O b servatio ns 4,442 4,442 4,442 4,442 Adj R 2 0.697 0.698 0.700 0.701

PAGE 132

132 Table 4 6 C redit line drawdowns and bank financial conditions This table presents the results on credit line drawdown and bank financial conditions. Sample includes 4,442 firm year observations from year 2003 to year 2009. Var iables are defined in Appendix C Standard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, **, and to denote significance at the 1% level, 5% lev el, and 10% level, respectivel y (1) (2) (3) (4) Crisis 0.013*** 0.025*** 0.022*** 0.019*** (0.003) (0.005) (0.006) (0.006) Crisis x NPL_noC&I 0.999 0.692 0.545 (0.902) (0.918) (0.943) NPL_noC&I 0.220 0.621 0.528 (0.937) (0.966) (1.010) Size 0.012* 0.011* 0.009 0.008 (0.007) (0.007) (0.007) (0.007) Market to book 0.011*** 0.010** 0.011*** 0.010** (0.004) (0.004) (0.004) (0.004) Cash Flow 0.140 0.169* 0.175* 0.171* (0.094) (0.093) (0.093) (0.094) Z score 0.009*** 0.008*** 0.008*** 0.008*** (0.002) (0 .002) (0.002) (0.002) Fixed Assets 0.038 0.044 0.038 0.043 (0.033) (0.033) (0.033) (0.033) Log Firm Age 0.015 0.010 0.012 0.022* (0.010) (0.010) (0.011) (0.012) Asset Volatility 0.042*** 0.042*** 0.044*** 0.045*** (0.009) (0.009) (0.009) ( 0.009) Bank Controls: Bank Size 0.002 0.001 (0.004) (0.004) Bank ROA 0.215 0.216 (0.316) (0.314) Bank Tier1 ratio 0.399** 0.503*** (0.180) (0.189) Bank CoreDep 0.037 0.017 (0.029) (0.030) Bank Liq Assets 0.099 0.201** (0.088) (0.098) Bank Unused Comm. 0.050 0.042 (0.035) (0.037) Constant 0.024 0.016 0.029 0.076 (0.044) (0.044) (0.104) (0.110) Firm FE Y Y Y Y Year FE N N N Y Number of Observations 4,442 4,442 4,442 4,442 Adj R 2 0.62 9 0.631 0.632 0.633

PAGE 133

133 Table 4 7 Term loans and bank financial conditions This table presents the results on term loans and bank financial conditions. Sample includes 4,442 firm year observations from year 2003 to year 2009. Var iables are defined in Appen dix C Standard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, **, and to denote significance at the 1% level, 5% lev el, and 10% level, respectively (1) (2) (3) (4) Crisis 0.014*** 0.022*** 0.017** 0.015 ** (0.004) (0.006) (0.007) (0.007) Crisis x NPL_noC&I 2.171*** 2.024** 1.900** (0.840) (0.859) (0.863) NPL_noC&I 2.222** 2.369** 2.278** (0.893) (0.934) (0.969) Size 0.011 0.011 0.011 0.010 (0.008) (0.008) (0.008) (0.008) Market to book 0.010* 0.010 0.010 0.009 (0.006) (0.006) (0.006) (0.006) Cash Flow 0.157 0.160 0.163 0.167 (0.102) (0.103) (0.104) (0.105) Z score 0.009*** 0.009*** 0.009*** 0.009*** (0.003) (0.003) (0.003) (0.003) Fixed Assets 0.019 0.021 0.011 0.007 (0.037) (0.038) (0.038) (0.039) Log Firm Age 0.007 0.007 0.006 0.002 (0.014) (0.014) (0.015) (0.017) Asset Volatility 0.050*** 0.051*** 0.050*** 0.051*** (0.012) (0.012) (0.012) (0.012) Bank Controls: Bank Size 0.004 0.005 (0.00 5) (0.005) Bank ROA 0.452 0.450 (0.415) (0.418) Bank Tier1 ratio 0.086 0.158 (0.201) (0.218) Bank CoreDep 0.011 0.004 (0.034) (0.035) Bank Liq Assets 0.126 0.050 (0.099) (0.106) Bank Unused Comm. 0.021 0.027 (0.039) (0.042) Constant 0.022 0.027 0.087 0.123 (0.052) (0.052) (0.116) (0.130) Firm FE Y Y Y Y Year FE N N N Y Number of Observations 4,442 4,442 4,442 4,442 Adj R 2 0.677 0.678 0.678 0.678

PAGE 134

134 Table 4 8 Trade credit and bank financial conditions This table presents the results on trade credit and bank financial conditions. Sample includes 4,442 firm year observations from year 2003 to year 2009. Var iables are defined in Appendix C Standard errors clustered by firm are reported in pare ntheses below the parameter estimates. I use ***, **, and to denote significance at the 1% level, 5% lev el, and 10% level, respectively (1) (2) (3) (4) Crisis 0.003** 0.002 0.001 0.000 (0.001) (0.002) (0.003) (0.003) Crisis x NPL_noC&I 0.217 0. 205 0.238 (0.389) (0.399) (0.413) NPL_noC&I 0.146 0.022 0.075 (0.390) (0.378) (0.403) Size 0.014*** 0.015*** 0.016*** 0.017*** (0.003) (0.003) (0.003) (0.003) Market to book 0.008*** 0.008*** 0.007*** 0.007*** (0.002) (0.002) (0.002) (0.002) Cash Flow 0.009 0.005 0.008 0.007 (0.045) (0.045) (0.045) (0.045) Z score 0.003*** 0.003*** 0.003*** 0.003*** (0.001) (0.001) (0.001) (0.001) Fixed Assets 0.028** 0.025* 0.024* 0.020 (0.014) (0.013) (0.014) (0.014) Log Firm Age 0.007 0.009** 0.006 0.000 (0.004) (0.004) (0.005) (0.005) Asset Volatility 0.004 0.004 0.005 0.004 (0.004) (0.004) (0.004) (0.005) Bank Controls: Bank Size 0.002 0.001 (0.001) (0.001) Bank ROA 0.019 0.041 (0.101) (0.101) Bank Tier1 ratio 0.054 0.098 (0.071) (0.076) Bank CoreDep 0.007 0.004 (0.011) (0.012) Bank Liq Assets 0.054 0.011 (0.034) (0.038) Bank Unused Comm. 0.001 0.004 (0.015) (0.016) Constant 0.157*** 0.154*** 0.14 5*** 0.183*** (0.022) (0.021) (0.039) (0.043) Firm FE Y Y Y Y Year FE N N N Y Number of Observations 4,442 4,442 4,442 4,442 Adj R 2 0.917 0.918 0.919 0.919

PAGE 135

135 Table 4 9 Public debt and bank financial conditions This table presents the results on pub lic debt and bank financial conditions. Sample includes 4,442 firm year observations from year 2003 to year 2009. Var iables are defined in Appendix C Standard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, *, and to denote significance at the 1% level, 5% lev el, and 10% level, respectively (1) (2) (3) (4) Crisis 0.006 0.009 0.017* 0.021** (0.005) (0.008) (0.009) (0.009) Crisis x NPL_noC&I 0.599 0.910 0.927 (1.141) (1.189) (1.203) NPL_noC&I 0 .501 0.048 0.220 (1.143) (1.207) (1.261) Size 0.013 0.013 0.017 0.020* (0.011) (0.011) (0.011) (0.011) Market to book 0.024*** 0.023*** 0.023*** 0.025*** (0.007) (0.007) (0.007) (0.007) Cash Flow 0.370** 0.374** 0.364** 0.373** (0.154) ( 0.154) (0.154) (0.153) Z score 0.016*** 0.016*** 0.016*** 0.016*** (0.003) (0.003) (0.003) (0.003) Fixed Assets 0.060 0.061 0.056 0.041 (0.052) (0.052) (0.052) (0.052) Log Firm Age 0.025 0.025 0.019 0.005 (0.016) (0.017) (0.017) (0.020) As set Volatility 0.060*** 0.060*** 0.058*** 0.059*** (0.015) (0.015) (0.015) (0.015) Bank Controls: Bank Size 0.004 0.002 (0.006) (0.006) Bank ROA 0.506 0.464 (0.395) (0.399) Bank Tier1 ratio 0.400 0.591** (0.259) ( 0.275) Bank CoreDep 0.022 0.004 (0.054) (0.056) Bank Liq Assets 0.148 0.055 (0.138) (0.155) Bank Unused Comm. 0.053 0.057 (0.053) (0.056) Constant 0.150** 0.148** 0.154 0.023 (0.072) (0.072) (0.170) (0.181) Firm FE Y Y Y Y Year FE N N N Y Number of Observations 4,442 4,442 4,442 4,442 Adj R 2 0.782 0.782 0.784 0.786

PAGE 136

136 Table 4 10. Book leverage and bank financial conditions This table presents the results on book leverage and bank financial conditions. Sample include s 4,442 firm year observations from year 2003 to year 2009. Var iables are defined in Appendix C Standard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, **, and to denote significance at the 1% level, 5% le v el, and 10% level, respectively (1) (2) (3) (4) Crisis 0.028*** 0.059*** 0.055*** 0.052*** (0.005) (0.008) (0.009) (0.009) Crisis x NPL_noC&I 3.765*** 3.433*** 3.067** (1.149) (1.178) (1.226) NPL_noC&I 2.154* 2.436** 1.940 (1.157) (1.194 ) (1.290) Size 0.029*** 0.027*** 0.028*** 0.029*** (0.010) (0.010) (0.011) (0.011) Market to book 0.050*** 0.048*** 0.049*** 0.049*** (0.008) (0.008) (0.008) (0.008) Cash Flow 0.214 0.274 0.270 0.267 (0.169) (0.168) (0.170) (0.170) Z score 0.038*** 0.038*** 0.038*** 0.038*** (0.004) (0.004) (0.004) (0.004) Fixed Assets 0.058 0.070 0.072 0.068 (0.047) (0.047) (0.047) (0.047) Log Firm Age 0.027 0.018 0.017 0.014 (0.017) (0.018) (0.018) (0.020) Asset Volatility 0.161*** 0.163* ** 0.162*** 0.165*** (0.016) (0.016) (0.016) (0.016) Ind.Median 0.241*** 0.251*** 0.248*** 0.238*** (0.044) (0.044) (0.044) (0.046) Divpay 0.008 0.006 0.006 0.007 (0.011) (0.011) (0.011) (0.011) Bank Controls: Bank Size 0.007 0.006 (0.006) (0.006) Bank ROA 0.056 0.078 (0.446) (0.447) Bank Tier1 ratio 0.294 0.249 (0.270) (0.286) Bank Core Dep. 0.041 0.031 (0.048) (0.051) Bank Liq. Assets 0.045 0.022 (0.131) (0.143) Bank Unused Comm. 0.002 0.016 (0.046) (0.049) Constant 0.141** 0.118* 0.287* 0.260 (0.070) (0.070) (0.149) (0.161) Firm FE Y Y Y Y Year FE N N N Y Bank Controls N N Y Y Number of Observations 4,442 4,442 4,442 4,442 Adj R 2 0.852 0.854 0.854 0.85 4

PAGE 137

137 Table 4 11 Market leverage and bank financial conditions This table presents the results on market leverage and bank financial conditions. Sample includes 4,442 firm year observations from year 2003 to year 2009. Var iables are defined in Appendix C St andard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, **, and to denote significance at the 1% level, 5% lev el, and 10% level, respectively (1) (2) (3) (4) Crisis 0.087*** 0.149*** 0.114*** 0.107*** (0 .008) (0.012) (0.013) (0.013) Crisis x NPL_noC&I 5.794*** 4.413*** 3.647** (1.517) (1.545) (1.583) NPL_noC&I 1.499 3.009** 2.007 (1.517) (1.510) (1.614) Size 0.027** 0.024** 0.021* 0.021* (0.012) (0.012) (0.012) (0.012) Market to book 0.0 10 0.013* 0.012 0.011 (0.007) (0.008) (0.008) (0.008) Cash Flow 0.848*** 0.999*** 1.003*** 0.991*** (0.224) (0.218) (0.219) (0.218) Z score 0.021*** 0.020*** 0.020*** 0.020*** (0.003) (0.003) (0.003) (0.003) Fixed Assets 0.177*** 0.207 *** 0.218*** 0.215*** (0.063) (0.063) (0.063) (0.064) Log Firm Age 0.022 0.004 0.003 0.000 (0.020) (0.020) (0.021) (0.024) Asset Volatility 0.265*** 0.269*** 0.272*** 0.275*** (0.023) (0.022) (0.022) (0.023) Ind.Median 0.276*** 0.336*** 0.340 *** 0.333*** (0.045) (0.045) (0.045) (0.049) Divpay 0.001 0.006 0.004 0.005 (0.015) (0.015) (0.015) (0.015) Bank Controls: Bank Size 0.011 0.010 (0.009) (0.009) Bank ROA 2.141*** 2.185*** (0.704) (0.706) Bank Tier1 ratio 1.437*** 1.394*** (0.343) (0.366) Bank Core Dep. 0.132** 0.101 (0.059) (0.062) Bank Liq. Assets 0.025 0.053 (0.159) (0.176) Bank Unused Comm. 0.047 0.019 (0.057) (0.061) Constant 0.176** 0.108 0.5 27** 0.514** (0.079) (0.078) (0.210) (0.227) Firm FE Y Y Y Y Year FE N N N Y Bank Controls N N Y Y Number of Observations 4,442 4,442 4,442 4,442 Adj R 2 0.797 0.807 0.813 0.814

PAGE 138

138 Table 4 12. Investment and bank financial conditions This table present s the results on investment and bank financial conditions. Sample includes 4,442 firm year observations from year 2003 to year 2009. Var iables are defined in Appendix C Standard errors clustered by firm are reported in parentheses below the parameter esti mates. I use ***, **, and to denote significance at the 1% level, 5% lev el, and 10% level, respectively (1) (2) (3) (4) Crisis 0.002*** 0.002 0.001 0.002 (0.001) (0.001) (0.001) (0.001) Crisis x NPL_noC&I 0.329* 0.343 0.350* (0.199) (0.212) (0.203) NPL_noC&I 0.122 0.131 0.261 (0.193) (0.191) (0.192) Market to book 0.004*** 0.004*** 0.004*** 0.003*** (0.001) (0.001) (0.001) (0.001) Cash Flow 0.091*** 0.083*** 0.084*** 0.057*** (0.022) (0.021) (0.021) (0.019) Industry Median Inv. 0.627*** (0.097) Bank Controls: Bank Size 0.000 0.001 (0.001) (0.001) Bank ROA 0.072 0.036 (0.062) (0.057) Bank Tier1 ratio 0.030 0.039 (0.039) (0.039) Bank Core Dep. 0.002 0.005 (0.006) (0. 006) Bank Liq. Assets 0.041*** 0.021 (0.014) (0.016) Bank Unused Comm. 0.002 0.002 (0.008) (0.008) Constant 0.008*** 0.008*** 0.020 0.013 (0.001) (0.001) (0.025) (0.025) Firm FE Y Y Y Y Year FE N N N Y Bank Controls N N Y Y Number of Observations 4,442 4,442 4,442 4,442 Adj R 2 0.636 0.640 0.640 0.661

PAGE 139

139 Table 4 13. Investment and bank financial conditions This table presents the results on investment and bank financial conditions. Sample includes 37,603 firm quarter observ ations from year 2003 to year 2009. Var iables are defined in Appendix C Standard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, **, and to denote significance at the 1% level, 5% lev el, and 10% level, resp ectively (1) (2) (3) (4) Crisis 0.001*** 0.002*** 0.001*** 0.002*** (0.000) (0.000) (0.000) (0.001) Crisis x NPL_noC&I 0.163* 0.189** 0.175** (0.090) (0.095) (0.087) NPL_noC&I 0.034 0.033 0.036 (0.088) (0.089) (0.085) Market to book 0 .003*** 0.003*** 0.003*** 0.002*** (0.000) (0.000) (0.000) (0.000) Cash Flow 0.023*** 0.022*** 0.022*** 0.020*** (0.004) (0.004) (0.004) (0.004) Industry Median Inv. 0.642*** (0.064) Bank Controls: Bank Size 0.001* 0.000 (0.0 01) (0.000) Bank ROA 0.071*** 0.065*** (0.014) (0.014) Bank Tier1 ratio 0.015 0.008 (0.011) (0.011) Bank Core Dep. 0.005* 0.006** (0.003) (0.003) Bank Liq. Assets 0.021*** 0.005 (0.005) (0.005) Bank Unused Comm. 0.004 0.004 (0.003) (0.003) Constant 0.009*** 0.010*** 0.010 0.009 (0.000) (0.000) (0.012) (0.011) Firm FE Y Y Y Y Year FE N N N Y Bank Controls N N Y Y Number of Observations 37,603 37,603 37,603 37,603 Adj R 2 0.641 0.645 0.6 46 0.658

PAGE 140

140 Table 4 14. Cash holding and bank financial conditions This table presents the results on cash holding and bank financial conditions. Sample includes 4,442 firm year observations from year 2003 to year 2009. Var iables are defined in Appendix C Standard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, **, and to denote significance at the 1% level, 5% lev el, and 10% level, respectively (1) (2) (3) (4) Crisis 0.011*** 0.008 0.008 0.008 (0.00 4) (0.006) (0.007) (0.007) Crisis x NPL_noC&I 0.982 0.634 0.594 (0.984) (1.010) (1.082) NPL_noC&I 0.426 0.113 0.165 (1.017) (1.036) (1.151) Size 0.015** 0.015** 0.013* 0.013* (0.007) (0.007) (0.008) (0.008) Market to book 0.022*** 0.022** 0.022*** 0.022*** (0.005) (0.005) (0.005) (0.005) Cash Flow 0.138 0.196 0.202 0.202 (0.143) (0.140) (0.140) (0.140) R&D 1.490 1.517 1.547 1.546 (1.057) (1.037) (1.037) (1.040) R&D Missed 0.008 0.007 0.005 0.005 (0.009) (0.009) (0.009) (0.0 09) Industry Sigma 0.045 0.224 0.256 0.253 (0.300) (0.301) (0.289) (0.298) Net Working Capital excash 0.086*** 0.082*** 0.080*** 0.080*** (0.031) (0.031) (0.031) (0.031) Dividend Payer dummies 0.009 0.008 0.009 0.009 (0.010) (0.010) (0.010 ) (0.010) Bank Controls: Bank Size 0.001 0.001 (0.006) (0.006) Bank ROA 0.751* 0.753* (0.406) (0.409) Bank Tier1 ratio 0.448** 0.449** (0.211) (0.225) Bank Core Dep. 0.019 0.016 (0.044) (0.048) Bank Liq. Assets 0.097 0.103 (0.114) (0.129) Bank Unused Comm. 0.039 0.038 (0.045) (0.048) Constant 0.172*** 0.152*** 0.148 0.145 (0.053) (0.054) (0.160) (0.164) Firm FE Y Y Y Y Year FE N N N Y Bank Controls N N Y Y Number of Observations 4,442 4,442 4,442 4,442 Adj R 2 0.862 0.864 0.865 0.865

PAGE 141

141 Table 4 1 5 Cash holding and bank financial conditions This table presents the results on cash holding and bank financial conditions. Sample includes 37,603 firm quarter observations fro m year 2003 to year 2009. Var iables are defined in Appendix C Standard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, **, and to denote significance at the 1% level, 5% lev el, and 10% level, respectively (1) (2) (3) (4) Crisis 0.007*** 0.008** 0.004 0.009 (0.002) (0.003) (0.004) (0.007) Crisis x NPL_noC&I 1.605** 1.534** 1.068 (0.654) (0.680) (0.704) NPL_noC&I 0.355 0.711 0.243 (0.647) (0.663) (0.696) Size 0.016*** 0.016*** 0.014*** 0.015*** (0.005) (0.005) (0.005) (0.005) Market to book 0.018*** 0.020*** 0.020*** 0.019*** (0.002) (0.002) (0.002) (0.002) Cash Flow 0.130*** 0.139*** 0.141*** 0.141*** (0.044) (0.043) (0.043) (0.043) R&D 0.322 0.319 0.323 0.326 (0.210) (0 .209) (0.210) (0.209) R&D Missed 0.001 0.002 0.002 0.002 (0.004) (0.004) (0.004) (0.004) Industry Sigma 0.051 0.362** 0.394** 0.380** (0.178) (0.185) (0.184) (0.185) Net Working Capital excash 0.045*** 0.043*** 0.042*** 0.042*** (0.015) (0.015 ) (0.015) (0.015) Dividend Payer dummies 0.007 0.006 0.006 0.007 (0.005) (0.005) (0.005) (0.005) Bank Controls: Bank Size 0.004 0.005* (0.003) (0.003) Bank ROA 0.052 0.101 (0.112) (0.118) Bank Tier1 ratio 0.333 *** 0.268*** (0.097) (0.096) Bank Core Dep. 0.006 0.001 (0.021) (0.022) Bank Liq. Assets 0.014 0.041 (0.053) (0.059) Bank Unused Comm. 0.033 0.020 (0.027) (0.029) Constant 0.191*** 0.166*** 0.210*** 0.235*** (0. 032) (0.034) (0.077) (0.079) Firm FE Y Y Y Y Year FE N N N Y Bank Controls N N Y Y Number of Observations 37,603 37,603 37,603 37,603 Adj R 2 0.853 0.854 0.855 0.855

PAGE 142

142 Table 4 16. Seemingly unrelated regressions on firm financing This table presents re seemingly unrelated regressions. I five categories. They are bank debt, public debt, trade credit, other liabilities, and equity. Sample includes 4,442 firm year observations from year 2003 to year 2009. Variables are defined in Appendix C Standard errors clustered by firm are reported in parentheses below the parameter estimates. I use ***, **, and to denote significance at the 1% level, 5% lev el, and 10% level, respectively. Bank Debt Public Debt Trade Credit Other Liab. Equity Crisis 0.033*** 0.021*** 0.000 0.006 0.060*** (0.007) (0.008) (0.002) (0.005) (0.007) Crisis x NPL_noC&I 2.282** 0.927 0.238 1.056 2.383* (0.997) (1.015) (0. 348) (0.734) (0.917) NPL_noC&I 2.657** 0.220 0.075 1.155 1.250 (1.044) (1.064) (0.340) (0.747) (0.932) Market to book 0.020*** 0.025*** 0.007*** 0.018** 0.062*** (0.006) (0.006) (0.002) (0.003) (0.004) Size 0.018** 0.020** 0.017*** 0.050*** 0. 025** (0.110) (0.01) (0.002) (0.004) (0.005) Cash Flow 0.019 0.373*** 0.007 0.206 0.289* (0.002) (0.129) (0.038) (0.072) (0.090) Z score 0.018*** 0.016*** 0.003*** 0.016*** 0.049*** (0.002) (0.003) (0.001) (0.001) (0.001) Fixed Assets 0.036 0.041 0.020* 0.006 0.064 (0.040) (0.044) (0.011) (0.023) (0.028) Log Firm Age 0.024 0.005 0.000 0.034** 0.016 (0.017) (0.017) (0.004) (0.008) (0.010) Asset Volatility 0.097*** 0.059*** 0.004 0.001 0.156*** (0.012) (0.013) (0.004) (0.008) (0.010) Bank Controls: Bank Size 0.004 0.002 0.001 0.003 0.004 (0.0046) (0.005) (0.001) (0.003) (0.004) Bank ROA 0.270 0.464 0.041 0.131 0.107 (0.420) (0.336) (0.085) (0.270) (0.337) Bank Liq Assets 0.161 0.055 0.011 0.044 0. 050 (0.113) (0.131) (0.032) (0.074) (0.093) Bank CoreDep 0.014 0.004 0.004 0.021 0.002 (0.036) (0.047) (0.011) (0.024) (0.029) Bank Tier1 ratio 0.699*** 0.591** 0.098 0.421** 0.609** (0.227) (0.232) (0.064) (0.150) (0.187) Bank Unus ed Comm. 0.017 0.057 0.004 0.035 0.082 (0.039) (0.048) (0.014) (0.026) (0.033) Constant 0.000 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.001) (0.001) Firm FE Y Y Y Y Y Year FE Y Y Y Y Y Bank Controls Y Y Y Y Y Number of Observations 4 ,442 4,442 4,442 4,442 4,442

PAGE 143

143 Table 4 17. Change in equity by bank financial condition quintiles This table presents results on the change of firm equity from the end of 2007 to the end of 2009 across bank NPL quintiles. Change in equity is scaled by tota l assets at the end of 2007. Bank NPL is measured as the average level of bank NPL from 2007 to 2009 for each firm. T stat is calculated as comparing quintile 1 4 to quintile 5. NPL_ noC&I Q uintile (Low to High) Change in Equity from Year 2007 to 2009 Mea n Std. Dev. T stat 1 0.0156 0.2010 1.59 2 0.0143 0.1781 0.61 3 0.0230 0.2057 1.82 4 0.0272 0.2167 0.13 5 0.0310 0.1916

PAGE 144

144 CHAPTER 5 CONCLUSION In the three chapters that constitute this study, I examine three topics on intermediary lending T he first part of the study, Chapter 2, investigates a relatively common, but little studied type of credit line, for which funds availability is limited by the ime varying asset composition, namely borrowing base lines. Whereas Sufi (2009) reports that firms with low cash flow are more likely to lose access to their lines of credit, I show remain open to borrowers with high risk and low profits or cash flow. Compared to other types of secured credit lines, the B lender has security and the loan terms automatically limit credit extended to unsuccessful firms. Borrowers use the BB to reduce their borrowing cost, gain a more gen erous credit limit, and operate with fewer financial covenants. debt maturity, and bank debt proportion are jointly determined. In particular, this study considers the public bonds vs. bank loans on leverage. The substantial literature on capital structure decisions often examines one facet in isolation. The study of one capital structure decision in isolation is, however, at odds with the fact that firms simultaneously decide how much debt to take out (debt equity decision), how soon to repay the debt (debt maturity), and from whom to borrow (debt source) etc. Using a system of simultaneous equations, I find that debt maturity, debt source, and leverage ratio are jointly determined. B ratios. Debt maturity and bank debt can alter the relation between the leverage ratio and

PAGE 145

145 two types of borrowing costs: the potential conflicts between equity holders and debt holders, and financial distress risk. The results suggest that in addition to leverage ratio, debt maturity and debt source importantly affect the distortions associated with debt finance. impact corporate financing and investment decisions using the 2007 2009 financial crisis as an experimental setting. Though bank credit is viewed as an impo rtant source of funding to corporations in the literature, when examining corporate financing and investment decisions, most of the empirical works analyzes these corporate decisions as a function of firm fundamentals (demand side). I explore the cross sec tional estate market during the crisis and conditions impact corporate financing and investment decisions, when the nonfinancial sector was less directly affected by the financial crisis. I find that average firms relied more heavily on bank credit during the crisis. However, firms whose banks incurred a larger amount of nonperforming loans used less bank credit when c omparing their bank alternative credits such as public debt or trade credit. There is some evidence that banks shocks eventually affected corporate real activities ; firms w ith more adversely affected banks invest ed less and hoard ed more cash during the crisis compared to their pre crisis level.

PAGE 146

146 APPENDIX A VARIABLES C ONSTRUCTED IN CHAPTER 3 This appendix details the variables constructed in the empirical analysis. All numbe rs in parentheses refer to the annual Compustat item numbers. Total Debt = short term debt (34)+long term debt (9) Market Equity = stock price (199)*shares outstanding (54) Market Assets = book assets (6) book common equity (60)+market equity Market Leverage = total debt t+1 /market assets t+1 Book Leverage = total debt t+1 /book assets t+1 (6) Short Debt = (short term debt t+1 (34)+debt due in 2 nd year t+1 (91)+debt due in 3 rd year t+1 (92))/total debt t+1 Bank Debt = (drawn revolving line t+ 1 +term loan t+1 )/total debt t+1 Market to Book = market assets/book assets (6) Z Score = (3.3*pre tax income t (170)+sales t (12)+1.4*retained earnings t (36)+1.2*working capital t (179))/book assets t (6) Fixed Assets = net PPE t (8)/book assets t (6) Profitability = EBITDA t (13)/book assets t (6) Firm Size = log(sales t (12)) deflated by the CPI deflator to convert to 2001 dollar Firm Size Square = square of firm size t CF Volatility = standard deviation of the first difference in EBITDA over the 5 years preceding scaled by the average of book assets during this period Abnormal Earnings = (EPS t+2 (58) EPS t+1 (58))/closing stock price t (199) Stock Return = Annualized average daily stock return t Asset Maturity = (gross PPE t (7)/depreciation exp ense and amortization t (14))*(gross PPE t (7)/book assets t (6))+(current assets t (4)/cost of goods sold t (41))*(current assets t (4)/book assets t (6)) Industry Median Leverage = Fama and French 48 industry median market leverage t Industry Median Short term Debt = Fama and French 48 industry median market short term debt t Industry Median Bank Debt = Fama and French 48 industry median bank debt t Term Premium = average10 year government bond yield t 6 month government bill yield t Tightening Bank Le nding = net percentage of domestic respondents tightening standard for commercial and industrial (C&I) loans to large and medium sized firms by Federal Reserve Senior Loan Officer t Bank Index = value weighted stock return for banks (industry #11) availabl t Dividend Payer = 1 if positive dividend per share by the ex date t (26); 0 otherwise

PAGE 147

147 Investment Tax Credit = 1 if positive investment tax credit t (51); 0 otherwise Investment Tax Credit Missing = 1 if missing inve stment tax credit t (51); 0 otherwise Tax Loss Carry Forward = 1 if positive tax loss carry forward t (52); 0 otherwise Tax Loss Carry Forward Missing = 1 if missing tax loss carry forward t (52); 0 otherwise Below Investment Grade = 1 if S&P long term debt rating below BBB t ; 0 otherwise No Rating = 1 if no S&P long term debt rating t ; 0 otherwise

PAGE 148

148 APPENDIX B THEORETICAL PREDICTIONS OF CONTROL VARIABLES IN CHAPTER 3 Table B 1. Theoretical predictions of control variables in Chapter 3 T his table specifies control variabl es X, Y, and Z included in the Equations 3 1 to 3 3 rationales to include them. Expected signs are given in the parenthesis. C ontrol Variables X/Y/Z Leverage Short term Debt Bank Debt Fixed Assets t Firms with more fixed assets are harder to engage in asset substitution, and h ave higher liquidation value for a lender to seize in default and therefore allow for a higher debt capacity (Williamson (1988)). (+) Firms with more fixed assets are harder to engage in asset substitution and need less monitoring from the short term deb t. They also have longer asset maturity and used more long term debt for maturity matching. (+) Firms with more fixed assets are harder to engage in asset substitution and need less bank monitoring. ( ) Profitability t Firms with higher in ternally generated cash flow prefer to use less external debt financing to avoid adverse selection cost associated with asymmetric information (Myers and Majluf (1984)). Alternatively, firms with higher internal cash flow may use more leverage because of h igher tax benefit of debt financing (Myers (1984). (+or ) Firms with higher profit have better credit quality and are less likely to be financially distressed and therefore use more short term debt. (+) Firms with higher internally generat ed cash flow prefer to use less information sensitive debt financing such as bank debt to avoid adverse selection cost from asymmetric information (Myers and Majluf (1984)). Alternatively, firms with more cash flow might find it easier to comply bank coven ants and use more bank debt (Sufi (2009)). (+or )

PAGE 149

149 Table B 1 Continued Control Variables X/Y/Z Leverage Short term Debt Bank Debt Firm Size t Trade off theory implies that bigger firms use more leverage because they are more transparent and m ore diversified and therefore have lower cost of debt financing. Pecking order theory, however, indicates that bigger firms with more asymmetric information use more equity and thus less leverage because of less adverse selection cost (Myers (1984)). (+ or ) Bigger firms have better credit quality and less liquidity risk, and therefore use more short term debt that are cheaper due to the incorporation of future favorable private information (Diamond (1991a)). (+) Bigger firms have bet ter credit quality and less likelihood of financial distress, and therefore use less bank debt for monitoring or financial flexibility. ( ) Firm Size Square t Diamond (1991a) predicts a non linear relation between credit quality and debt maturity. F irms with the best and the worst credit quality use short term debt and the middle ones use long term debt. Firm size square can be used as a proxy for credit quality and capture this non linear relation (+ or ) Diamond (1991b ) predicts a non linear relation between credit quality and public debt. Firms with best and worst credit quality use public debt and middle ones use private debt. Firm size square can be used as a proxy for credit quality and capture this non linear relat ion (+ or )

PAGE 150

150 Table B 1 Continued Control Variables X/Y/Z Leverage Short term Debt Bank Debt CF Volatility t More volatile cash flows introduce a higher likelihood of financial distress and reduce the probability that tax shields wi ll be fully utilized, and therefore reduce the leverage ratio. ( ) More volatile cash flows introduce a higher likelihood of financial distress and difficulty to repay debt, and therefore reduce the short term debt. ( ) More volatile cash flows incur a higher degree of financial distress and difficulty to repay debt, and increase the bank debt for financial flexibility. ( ) Abnormal Earnings t Firms with favorable private information use less information sensitive claim like debt rather t han equity to reduce the adverse selection cost. (+) Firms with favorable private information use short term debt to signal their type or reduce future borrowing cost (Flannery (1986) and Diamond (1991a)). (+) Firms with favorable private information borro w from banks to reduce adverse selection cost. (+) Stock Return t Market timing theories predict that managers actively time equity markets for mispricing and incur a negative relation between stock prices and leverage (Lucas and McDonald (1990) ). ( ) Asset Maturity t Firms match the liability maturity with their asset maturity to reduce financial distress cost and control for agency conflicts between bondholders and shareholders. ( )

PAGE 151

151 Table B 1 Continued Control Variables X/Y/Z Leverage Short term Debt Bank Debt Industry Median Leverage t Industry median leverage can target leverage ratio or proxy for omitted industry characteristics (Flannery and Rangan (2006)). (+) Industry Median Sho rt term Debt t Industry median short term debt can target their debt maturity or proxy for omitted industry characteristics. (+) Industry Median Bank Debt t Industry median bank debt can serve as a benchmark s target their use of bank debt or proxy for omitted industry varia bles (+) Term Premium t Tax shield from long term debt is higher when the yield curve is upward sloping (Brick and Ravid (1985)). So firms use less short term debt when the term pr emium is high. ( ) Tightening Bank Lending t Firms get less bank credit during period of tightening bank lending (Becker and Ivishina (2010)). ( )

PAGE 152

152 Table B 1 Continued Control Variables X/Y/Z Leverage Short term Debt Bank Debt Bank Index t Firms g et more bank credit during period of better bank performance (Becker and Ivishina (2010)). (+) Dividend Payer t Firms paying out dividend have more free cash flow and need less external debt financing. Alternatively, they might find that using more debt c ould reduce free cash flow and therefore reduce agency cost (Jensen (1986)). (+) or ( ) Investment Tax Credit t Firms with higher investment tax credit use less leverage because of less tax benefit from debt financing (DeAngelo and Masulis (1980)). ( ) Firms with more tax shield use more short term debt (Brick and Ravid (1985), Kane et al. (1985)). (+) Tax Loss Carry Forward t Firms with higher tax loss carry forward use less leverage because of less tax benefit from debt financing (DeAngelo and Masu lis (1980)). Alternatively, firms with higher tax loss carry forward could have lower profitability and will have to use more external debt financing. (+) or ( ) Firms with more tax shield use more short term debt (Brick and Ravid (1985), Kane et al. (1985 )). (+)

PAGE 153

153 Table B 1 Continued Control Variables X/Y/Z Leverage Short term Debt Bank Debt Below Investment Grade t Diamond (1991a) predicts a non linear relation between credit quality and debt maturity. Firms with the best and the worst credit qua lity use short term debt and the mi ddle ones use long term debt. B elow investment grade firms use less short term debt than investment grade firms. ( ) Diamond (1991b) predicts a non linear relation between credit quality and public debt usage. Firms with the best and the worst credit quality use public debt and the middle ones use private debt. Firms with below investment grade use more bank debt than those with above investment grade (+) No Rating t Firms with no public debt rating could have no acces s to the public debt market and therefore borrow less debt (Faulkender and Petersen (2006)). ( ) Diamond (1991a) predicts a non linear relation between credit quality and debt maturity. Firms with the best and the worst credit quality use short ter m debt and the middle ones use long term debt. There is no clear prediction about the comparison of short term debt usage between investment grade firms and no rating firms. (+ or ) Diamond (1991b) predicts a non linear relation between credit qualit y and public debt usage. Firms with the best and the worst credit quality use public debt ; middle ones use private debt. There is no clear prediction about the comparison of private debt usage between investment grade and no rating firms. F irms with no ratings have more information asymmetry and therefore use more private debt for monitoring than investment grade firms. (+ or )

PAGE 154

154 APPENDIX C VARIABLES C ONSTRUCTED IN CHAPTER 4 This appendix details the variables constructed in the empirical analysis in chapter 4 N umbers in parentheses for firm variables refer to the annual Compustat item numbers. Variables in parentheses for bank variables refer to the FR Y9 item numbers Total Debt = short term debt (34)+long term debt (9) M arket Equity = stock price (199)*shares outstanding (54) Market Assets = book assets (6) book common equity (60)+market equity Bank Debt = ( CIQ drawn revolving line t+1 + CIQ term loan t+1 )/ book assets t+1 Drawdown = CIQ drawn revolving line t+1 / book as sets t+1 Term Loan = CIQ drawn term loan t+1 / book assets t+1 Trade Credit = account payable (70)/book assets (6) Public Debt = CIQ public debt/book assets (6) Market Leverage = total debt t+1 /market assets t+1 Book Leverage = total debt t+1 /book a ssets t+1 (6) Capital Expenditure = (capital expenditure t+1 )/book assets t (6) Cash = cash and cash equivalent (1)/book assets (6) Market to Book = market assets/book assets (6) Z Score = (3.3*pre tax income t (170)+sales t (12)+1.4*retained earning s t (36)+1.2*working capital t (179))/book assets t (6) Fixed Assets = net PPE t (8)/book assets t (6) Cash Flows = EBITDA t (13)/book assets t (6) Firm Size = log(sales t (12)) deflated by the CPI deflator to convert to 2001 dollar Firm Age = log (1+nu mber of years in Compustat) Asset Volatility = stock return volatility t *(1 market leverage t ) Industry Median Book (Market) Leverage = I ndustry median book ( market ) leverage by 2 digit SIC code t Industry Median Capital Expenditure = I ndustry median c apital expenditure by 2 digit SIC code t Industry Sigma = I ndustry mean of cash flow standard deviation over 10 years by 2 digit SIC code t Net Working Capital = (current asset t (4) current liability t (5) cash and cash equivalent t (1))/ book assets t (6) R&D Expenditure = R&D expenditure t (46)/ book assets t (6) Bank Index = value weighted stock return for banks (industry #11) t Dividend Payer = 1 if positi ve dividend per share by the ex date t (26); 0 o therwise Bank Size = 1 if positive dividend per share by the ex date t (26); 0 otherwise

PAGE 155

155 Bank Non performing Loans (NPL) = loans pastdue90 (bhck5525) plus non accrual loans (bhck5526) over total assets over total assets (bhck2170) Bank Non perfo rming Loans Excluding Commercial and Industrial Loans (NPLs_noC&I) = C&I loans pastdue90 (bhck1607) plus C&I non accrual loans (bhck1608) over total assets (bhck2170) Bank Non performing Loans From Commercial and Industrial Loans = NPLs NPLs_noC&I. Liqu id Assets = sum of cash and balances due from depository institutions over total assets (bhck2170) (bhck0081+bhck0395+bhck0397), federal funds sold and securities purchased under agreements to resell (bhdmb987+ bhckb989), trading assets (bhck3545) over tot al assets (bhck2170) Core Deposit = noninterest bearing deposit (bhdm6631) + interest bearing deposit (bhdm6636), minus time deposits of $100,000 or more (bhcb2604) over total assets (bhck2170) Bank ROA = income (loss) before extraordinary items and othe r adjustments (bhck4300) over total assets (bhck2170) Unused Commitments = unused commitments (bhck3818) over total assets (bhck2170)

PAGE 156

156 APPENDIX D DESCRIPTIVE STATISTICS OF WEIGHTED AVERAGE BANK CHARACTERISTICS IN CHAPTER 4 Table D 1. Descriptive statistics of weighted average bank characteristics in Chapter 4 Vari able Obs Mean Std. Dev. Min Max Avg_NPL_noC&I 4,442 0.008 0.008 0.000 0.041 Avg_Size 4,442 20.261 1.358 15.316 21.535 Avg_ROA 4,442 0.006 0.005 0.020 0.016 Avg_Tier1 (%) 4,442 0.089 0.014 0.071 0.132 Avg_Core Dep. 4,442 0.373 0.145 0.079 0.730 Avg_L iq.Assets 4,442 0.064 0.026 0.027 0.150 Avg_Unused comm. 4,442 0.226 0.101 0.108 0.783

PAGE 157

157 APPENDIX E ROBUSTNESS TEST OF FIRM BANK MATCHING IN CHAPTER 4 Table E 1. Robustness test of firm bank matching in Chapter 4 Variable NPLs > Mean NPLs < Mean Difference T stats Size 6.307 5.969 0.338 5.120 Market to book 1.5 16 1.518 0.001 0.037 Cash Flow 0.036 0.034 0.003 2.678 Z score 2.607 2.783 0.176 1.869 Fixed Assets 0.303 0.301 0.002 0.218 Age 2.846 2.736 0.110 3.515 Assets volatility 0.295 0.313 0.018 3.096

PAGE 158

158 APPENDIX F BANK NON PERFORMING LOANS IN CHAPTER 4 Fi gure F 1. Bank non performing loans in Chapter 4 This figure plots the average bank non performing loans (NPLs) over total assets ratio over my sample period. The sample includes U.S. BHCs from the first quarter of 2003 to the fourth quarter of 2009. NPLs is the ratio of NPLs o ver total bank assets, NPLs_C&I is the ratio of commercial and industrial loans (C&I loans) that are non performing over total bank assets, and NPLs_noC&I is the difference between NPLs and NPLs_C&I. This figure shows the level of non C&I NPLs spiked durin g the financial crisis and level of NPL C&I loans were comparably more stable. 0.000% 0.500% 1.000% 1.500% 2.000% 2.500% 3.000% 3.500% NPLs NPLs_C&I NPLs_noC&I

PAGE 159

159 LIST OF REFERENCES Adam, Tim and Vidhan K. Goyal, 2008, Investment opportunity set and its proxy variables, Journal of Financial Research 31, 41 63. Altman, Edward I., 1968, F inancial ratios, discriminate analysis, and the prediction of corporate bankruptcy, Journal of Finance 23, 589 609. Altman, Edward I., 1977, The Z Score Bankruptcy Model: Past, Present, and Future (John Wiley & Sons, Inc., New York). Baker, Malcolm, and J effrey Wurgler, 2002, Market timing and capital structure, Journal of Finance 57, 1 32. Barclay, Michael J., Leslie M. Marx, and Clifford W. Smith Jr., 2003, The joint determination of leverage and maturity, Journal of Corporate Finance 9, 149 167. Barclay Michael J., and Clifford W. Smith Jr., 1995, The maturity structure of corporate debt, Journal of Finance 50, 609 631. Barnett, William, based lending, The Secured Lender 53, 80 82. Bharath, S reedhar S an deep Dahiya, A nthony Saunders, and Anand Srinivasan, 2007, Journal of Financial Economics 85, 368 419. Becker, Bo, and Victoria Ivashina, 2010, Cyclicality of credit supply: Firm level evidence, Working paper, Harvard University. Bernanke B en S., 1983, Non monetary effects of the financial crisis in the propagation of the great depression, NBER Working Paper No. 1054. Bernanke, B en S and A lan S. Blinder, 1988, Is it Money or Credit, or Both or Neither? Credit, Money and Aggregate Demand, American Economic Review 78 435 439 Berkovitch, Elazar, and Stuart I. Greenbaum, 1991, The loan commitment as an optimal financing contract, Journal of Financial and Quantitative Analysis 26, 83 95. Berlin, Mi tchell, and Jan Loeys, 1998, Bond covenants and delegated monitoring source, Journal of Finance 43, 397 412. Billet, Matthew, Mark Flannery, and Jon Garfinkel, 2006, Are bank loans special? Evidence on the post announcement performance of bank borrowers, J ournal of Financial and Quantitative Analysis 41, 733 752.

PAGE 160

160 Bolton, Patrick, and Xavier Freixas, 2000, Equity, bonds, and bank debt: Capital structure and financial market equilibrium under asymmetric information, Journal of Political Economy 108, 324 351. Bound, John, David A. Jaeger, and Regina Baker, 1993, The cure can be worse than the disease: A cautionary tale regarding instrumental variables, NBER Technical Working Paper No. 137. Bound, John, David A. Jaeger, and Regina Baker, 1995, Problems with inst rumental variables estimation when the correlation between the instruments and the endogenous explanatory variables is weak, Journal of the American Statistical Association 90, 443 450. Brick, Ivan E., and Abraham S. Ravid, 1985, On the relevance of debt m aturity structure, Journal of Finance 40, 1423 1437. Boot, Arnoud, Anjan V. Thakor, and Gregory F. Udell, 1997, Competition, risk neutrality and loan commitments, Journal of Banking and Finance 11, 449 471. Booth, James R., and Lena Chua Booth, 2006, Loan collateral decisions and corporate borrowing costs, Journal of Money Credit and Banking 38, 67 90. Campello, M urillo, Erasmo Giambona John R. Graham and Campbell R Harvey, 2011, Liquidity Management and corporate investment during a financial crisis, The Review of Financial Studies 24, 1944 1979. Cantillo, Miguel, and Julian Wright, 2000, How do firms choose their lenders? An empirical investigation, Review of Financial Studies 13, 155 189. Caouette, John B., Edward I. Altman, and Paul Narayanan, 1998 Managing credit risk: the next great financial challenge, Wiley & Sons, New York Carvalho D aniel R ., Miguel A. Ferreira and Pedro P. Matos, 2011, Lending relationship and the effect of bank distress: evidence from the 2007 2008 financial crisis, working paper. Carey, Mark, Mitch Post, and Steven A. Sharpe, 1998, Does corporate lending by banks and finance companies differ? Evidence on specialization in private debt contracting, Journal of Finance 53, 845 878. Carey, Mark, and Greg Nini, 2007, Is the cor porate loan market globally integrated? A pricing puzzle, Journal of Finance 62, 2969 3007. Charles, Hadlock J., and Christopher M. James, 2002, Do banks provide financial slack?, Journal of Finance 57, 1383 1419. Chava, Sudheer, and Michael R. Roberts, 20 08, How does financing impact investment? The role of debt covenants, Journal of Finance 63, 2085 2121.

PAGE 161

161 Chemmanur, Thomas J., and Paolo Fulghieri, 1994, Reputation, renegotiation, and the choice between bank loans and publicly traded debt, Review of Financ ial Studies 7 475 506. Colla, Paolo, Filippo Ippolito, and Kai Li, 2011, Debt specialization, Working paper, Universit Bocconi and University of British Columbia. Cornet, Marcia M. Jamie J. McNutt Philip E. Strahan and H assan Tehranian, 2010, Liquidit y risk management and credit supply in the financial crisis, Journal of Financial Economics 101, 297 312. Denis, David J., and Vassil T. Mihov, 2003, The choice among bank debt, non bank private debt, and public debt: Evidence from new corporate borrowings Journal of Financial Economics 70, 3 28. Dennis, Steven, Debarshi Nandy, and Ian G. Sharpe, 2000, The determinants of contract terms in bank revolving credit agreements, Journal of Financial and Quantitative Analysis 35, 87 110. Diamond, Douglas, 1984, F inancial intermediation and delegated monitoring, Review of Economic Studies 51, 393 414. Diamond, Douglas W., 1991a, Debt maturity structure and liquidity risk, Quarterly Journal of Economics 106, 709 737. Diamond, Douglas W., 1991b, Monitoring and reputa tion: The choice between bank loans and directly placed debt, Journal of Political Economy 99, 689 721. Diamond, Douglas W., and R aghumram G Rajan 2001, Liquidity risk, liquidity creation and financial fragility: A theory of banking, Journal of Political Economy 109, 287 327. Duchin, R an Oguzhan Ozbas and Berk A. Sensoy, 2010, Costly external finance, corporate investment, and the subprime mortgage credit crisis, Journal of Financial Economics 97, 418 435. anks?, Journal of Monetary Economics 15, 29 39. Faulkender, Michael, and Mitchell A. Petersen, 2006, Does the Source of Capital Affect Capital Structure?, Review of Financial Studies 19(1): 45 79. Flannery, Mark J., 1982, Retail Bank Deposits a quasi fixed factors of production, American Economic Review 527 536. Flannery, Mark J., 1986, Asymmetric Information and Risky Debt Maturity Choice, Journal of Finance 41, 19 37.

PAGE 162

162 Flannery, Mark J., and Joel F. Houston, 1999, The value of a government monitor for U. S. banking firms, Journal of Money Credit and Banking 31, 14 34. Flannery, Mark J., and K asturi P. Rangan, 2006, "Partial Adjustment and Target Capital Structures, Journal of Financial Economics 79, 469 506. Flannery, Mark J., and Lockhart, G. Brandon, 20 09, Credit line and the substitutability of cash and debt, Working Paper, University of Florida.Guo, Pengjie, and Hayong Yun, 2009, Commercial paper, lines of credit, and the real effects of the financial crisis of 2008: Firm level evidence from the manufa cturing industry, Working Paper, University of Notre Dame. Frank, Murray Z., and Vidhan K. Goyal, 2009, Capital structure decisions: Which factors are reliably important?, Financial Management 38, 1 37. Friedman M. and Schwartz A.J., 1963, A monetary hist ory of the United Stated, Princeton University Press. Furlong F red, and Zena Knight, 2010, Loss provisions and bank charge offs in the financial crisis: Lesson learned, FRBSF Economic Letter 2010 16 (May 24). Gao P engjie, and Hayong Y un 2011, Liquidity ba ckstop, corporate borrowings, and real effects, working paper. Gatev E van, and Phillip E. Strahan 2009, Liquidity risk and syndicate structure, working paper. Gertner, Robert, and David Scharfstein, 1991, A theory of workouts and the effects of reorganiza tion law, Journal of Finance 46, 1189 1222. Gilson, Stuart C., and Jerold B. Warner, 1998, Private versus public debt: Evidence from firms that replace bank loans with junk bonds, Working paper, Harvard University and University of Rochester. Graham, John R. and Campbell R. Harvey, 2001, The theory and practice of corporate finance: Evidence from the field, Journal of Financial Economics 60, 187 243. Greene, William H., 2002, Econometric Analysis (Prentice Hall, Englewood Cliffs, NJ). Guedes, Jose and Tim O pler, 1996, The determinants of the maturity of corporate debt issues, Journal of Finance 51, 1809 1833. Holmstrom, Bengt, and Jean Tirole, 1997, Financial intermediation, loanable funds and the real sector, The quarterly journal of economics 112, 663 691. Holmstrom, Bengt, and Jean Tirole, 1998, Private and public supply of liquidity, Journal of Political Economy 106, 1 40.

PAGE 163

163 Houston, Joel and Christopher James, 1996, Bank information monopolies and the mix of private and public debt claims, Journal of Finan ce 51, 1863 1889. Huang R occo 2009, How committed are bank line of credit? Evidence from the subprime mortgage crisis, working paper. Ivashina V ictoria 2009, Asymmetric information effects on loan spreads. Journal of Financial Economics 92, 300 319. Ivas hina, Victoria, and David Scharfstein, 2010, Bank lending during the financial crisis of 2008, Journal of Financial Economics 93, 319 338. James, C hristopher 1987. Some evidence on the uniqueness of bank loans. Journal of Financial Economics 19, 217 2 35. Jensen, MichaelC., andWilliam H. Meckling, 1976, Theory of the firm: Managerial behavior, agency costs, and capital structure, Journal of Financial Economics 3, 305 360. Johnson, Shane A., 1997, An empirical analysis of the determinants of corporate debt o wnership structure, Journal of Financial and Quantitative Analysis 32, 47 69. Johnson, Shane A., 1998, The effect of bank debt on optimal capital structure, Financial Management 27, 47 56. Johnson, Shane A., 2003, Debt maturity and the effects of growth op portunities and liquidity risk on leverage, Review of Financial Studies 16, 209 236. Kaplan S teven N and Luigi Zingales 1997, Do investment cash flow sensitivities provide useful measures of financing constraints ? Quarterly Journal of Economics 112, 16 9 215. Krainer, J ohn 2004. House prices and fundamental values." FRBSF Economic Letter 2004 27, October 1. Khwaja, Asim I., and Atif Mian., 2008, Tracing the impact of bank liquidity shocks, American Economic Review 98, 1413 1442 Kwan, Simon H., 2010, Fin ancial crisis and bank lending, working paper. Krishnaswami, Sudha, Paul A. Spindt, and Venkat Subramaniam, 1999, Information asymmetry, monitoring, and the placement structure of corporate debt, Journal of Financial Economics 51, 407 434. Leland, Hayne E. and David H. Pyle, 1977, Informational Asymmetries, Financial Structure, and Financial Intermediation, Journal of Finance 32, 371 387.

PAGE 164

164 Lemmon, Michael L., Michael R. Roberts, and Jaime F. Zender, 2008, Back to the beginning: Persistence and the cross sect ion of corporate capital structure, Journal of Finance 58, 1575 1608. Lee, Lung Fei, 1978, Unionism and wage rates: A simultaneous equations model with qualitative and. limited dependent variables, International Economic Review 19, 415 433. Lockhart, G. B randon, 2009, Adjusting to target capital structure: The effect of credit lines, Working Paper, University of Florida. MacKie Mason, Jeffrey K., 1990, Do taxes affect corporate financing decisions?, Journal of Finance 45, 1471 1493. Maddala G.S., 1983, Lim ited dependent and qualitative variables in econometric, Econometric society monographs No. 3 (Cambridge University Press, New York) Melnik, Arie, and Steven Plaut, 1986, Loan commitment contracts, terms of lending, and credit allocation, Journal of Finan ce 41, 425 435. revolver usage as firms approach default, December 2008. Myers, Stewart C., 1977, Determinants of corporate borrowing, Journal of Financial Economics 5, 147 175. Nini, Greg, 2009, How non banks increased the supply of bank loans: Evidence from institutional term loans, Working Paper, University of Pennsylvania. Opler, Tim, Lee Pinkowitz, Rene Stulz, and Rohan Williamson, 1999, The determinants and implications of corporate cash holdings, Journal of Financial Economics 52, 3 46. Opler, Tim and Sheridan Titman, 1993, The determinants of leveraged buyout activity: Free cash flow vs. financial distress costs, Journal of Finance 48, 1985 1999. Opler, Tim C., and She ridan Titman, 1994, Financial distress and corporate performance, Journal of Finance 49, 1015 1040. Park, Cheol, 2000, Monitoring and structure of debt contracts, Journal of Finance 55, 2157 2195. Paul, Asquith, Gertner Robert, and David, Scharfstein, 1994 Anatomy of financial distress: An examination of Junk Bond issuers, Quarterly Journal of Economics 109, 625 658. Peek, J oe, and Eric S. Rosengren 1997, The international transmission of financial shocks: The case of Japan. The American Economic Review 8 7, 495 505.

PAGE 165

165 Peek, J oe, and Eric S. Rosengren, 2000, Collateral damage: Effects of the Japanese bank crisis on real activity in the United States. The American Economic Review 90, 30 45. Petersen, M itchell A., 2009, Estimating standard errors in finance p anel data sets: Comparing approaches, The Review of Financial Studies 22, 435 4 80. Petersen, M itchell A., and Raghumram G. Rajan 1994, The benefits of lending relationships: Evidence from small business data, The Journal of Finance 49, 3 37. Rajan, Raghur am G., 1992, Insiders and outsiders: The choice between informed and arm's length debt, Journal of Finance 47, 1367 1400. Rajan, Raghuram G., and Luigi Zingales, 1995, What do we really know about capital structure? Some evidence from international data, J ournal of Finance 50, 1421 1460. Rauh, Joshua D. and Sufi Amir, 2010, Capital structure and debt structure, Review of Financial Studies 23, 4242 4280. Robb A licia M and David T. Robinson, 2009, The capital structure decisions of new firms, NBER working paper Roberts, Michael R. and Amir Sufi, 2009, Renegotiation of financial contracts: Evidence from private credit agreements, Journal of Financial Economics 93, 159 184. Santos, J ose 2011, Bank corporate loan pricing following the subprime crisis, Review of Financial Studies 24, 1916 1943. Schnabl, Philips 2011, The international transmission of bank liquidity shocks: Evidence from an emerging market, forthcoming Journal of Finance 006, A guide to the loan market, The McGraw Hill Comp anies, Inc.,NewYork,NY. Shockley, Richard L., and Anjan V. Thakor, 1997, Bank loan commitment contracts: Data, theory and tests, Journal of Money, Credit and Banking 29, 517 534. Smith, Clifford W. Jr., and Jerold B. Warner, 1979, On financial contracting: An analysis of bond covenants, Journal of Financial Economics 7, 117 161. Stiglitz, Joseph E. and Andrew Weiss, 1981, Credit rationing in markets with imperfect information, American Economic Review 71, 393 410. Sufi, Amir, 2009, Bank lines of credit in c orporate finance: An empirical analysis, Review of Financial Studies 22, 1057 1088.

PAGE 166

166 Vinter, Graham D., 1998, Project finance: A legal guide (Sweet & Maxwell, London). Yun, Hayong, 2009, The choice of corporate liquidity and corporate governance, Review of Financial Studies 22, 1447 1475.

PAGE 167

167 BIOGRAPHICAL SKETCH Xiaohong graduated from Dongbei University of Finance and Economics with a B achelor of Arts degree in finance After also earning her Master of Arts degree in e conomics and Master of Science degree in applied statistics from Bowling Green State University s he completed her PhD degree in finan ce at the University of Florida Xiaohong join ed Northeastern Illinois University as an assistant professor of finance in the fall of 2012