Is Bank Debt Special for the Transmission of Monetary Policy? - - PowerPoint PPT Presentation

Is Bank Debt Special for the Transmission of Monetary Policy? Evidence from the Stock Market

Filippo Ippolito - Ali Ozdagli - Ander Perez

UPF, Barcelona GSE & CEPR - FRB of Boston - UPF & Barcelona GSE

18 October 2013

• 16th Annual DNB Research Conference

1 / 52

Introduction

Questions

I Is bank lending to …rms special for monetary policy transmission? I If so, why is it special?

2 / 52

Introduction

Questions

I Is bank lending to …rms special for monetary policy transmission? I If so, why is it special?

Strategy

I Compare stock price reaction to monetary policy shocks of

bank-dependent and non-dependent …rms

2 / 52

Introduction

Questions

I Is bank lending to …rms special for monetary policy transmission? I If so, why is it special?

Strategy

I Compare stock price reaction to monetary policy shocks of

bank-dependent and non-dependent …rms Challenges

I Measuring bank-dependence ) answer: new detailed debt structure

data

I Identifying transmission mechanisms

• 1. bank lending channel ) answer: bank-…rm match
• 2. pass-through channel ) answer: interest rate hedging activity by

…rms

2 / 52

Transmission Mechanisms

3 / 52

Transmission Mechanisms

4 / 52

Overview of Results

Is bank lending to …rms special for monetary policy transmission?

I Yes I Daily stock price response to 1% surprise increase in fed funds target

I around 4% on average I around 1% more if 2 std dev increase in (bank debt/assets) 5 / 52

Overview of Results

Is bank lending to …rms special for monetary policy transmission?

I Yes I Daily stock price response to 1% surprise increase in fed funds target

I around 4% on average I around 1% more if 2 std dev increase in (bank debt/assets)

Why is it special?

5 / 52

Overview of Results

Is bank lending to …rms special for monetary policy transmission?

I Yes I Daily stock price response to 1% surprise increase in fed funds target

I around 4% on average I around 1% more if 2 std dev increase in (bank debt/assets)

Why is it special?

I Bank lending channel

I Bank dependent …rms borrowing from …nancially healthy banks less

responsive

5 / 52

Overview of Results

Is bank lending to …rms special for monetary policy transmission?

I Yes I Daily stock price response to 1% surprise increase in fed funds target

I around 4% on average I around 1% more if 2 std dev increase in (bank debt/assets)

Why is it special?

I Bank lending channel

I Bank dependent …rms borrowing from …nancially healthy banks less

responsive

I Interest rate pass-through channel

I Interest rate hedgers are less responsive 5 / 52

Data Sources and Sample

I Sample: U.S. publicly listed …rms, 2003-2008

I No detailed …rm debt structure data pre 2003 I No conventional monetary policy post 2008

I Firm characteristics: Capital IQ and Compustat, annual level I Stock returns: CRSP I Monetary policy surprises: calculated as in Kuttner (2001) and

Bernanke and Kuttner (2005)

I Bank characteristics: Call Reports I Bank-Firm matching: LPC Dealscan (syndicated loans) I Interest rate hedging by …rms: annual 10-K …lings with SEC

6 / 52

E¤ect of Monetary Policy Surprises Across Subsamples

Similar response of stock prices to federal funds rate surprises across sample periods

7 / 52

Is Bank Debt Special?

8 / 52

A New Bank-Dependence Measure

I Capital IQ (CIQ) is provided by Standard and Poor’s and contains

detailed debt structure as follows:

I Commercial paper (CP) I Bank debt = Drawn Credit Lines (RC) + Term loans (TL) I Bonds = Senior (SBN) + Subordinated (SUB) I Capital leases (CL) I Other debt (Other): other short-term borrowings, trade credit,

deferred credits and trust-preferred securities

9 / 52

A New Bank-Dependence Measure

I Capital IQ (CIQ) is provided by Standard and Poor’s and contains

detailed debt structure as follows:

I Commercial paper (CP) I Bank debt = Drawn Credit Lines (RC) + Term loans (TL) I Bonds = Senior (SBN) + Subordinated (SUB) I Capital leases (CL) I Other debt (Other): other short-term borrowings, trade credit,

deferred credits and trust-preferred securities

I We de…ne bank dependence as (Total Bank Debt)/(Total Assets)

9 / 52

Descriptive Statistics

10 / 52

Descriptive Statistics

11 / 52

Descriptive Statistics

12 / 52

Is Bank Debt Special?

I Speci…cation

Reti,t = β0 + β1Surpriset + β2 (BankDebt/At)i,t1 +β3Surpriset (BankDebt/At)i,t1 +γControlsi,t1 + λSurpriset Controlsi,t1 + εi,t,

13 / 52

Is Bank Debt Special?

I Speci…cation

Reti,t = β0 + β1Surpriset + β2 (BankDebt/At)i,t1 +β3Surpriset (BankDebt/At)i,t1 +γControlsi,t1 + λSurpriset Controlsi,t1 + εi,t,

I Bank debt specialness: β3 6= 0

13 / 52

Is Bank Debt Special?

14 / 52

Is Bank Debt Special?

Bank dependent …rms are more responsive to monetary policy shifts

15 / 52

Further Robustness: Short Term Debt E¤ect?

Higher sensitivity of bank debt users not due higher exposure to short-term debt

16 / 52

Further Robustness: Omitted Variables Bias?

I Hausman test based on asset pricing theory

I Compare …xed-e¤ects with OLS I any variable that makes stock more responsive to mon pol should

directly a¤ect …rms’ expected returns

I OVB should a¤ect not only coe¢cient of bank dependence

interacted with surprise, but also uninteracted

I Hausman test does not reject null of exogeneity 17 / 52

Further Robustness: Omitted Variables Bias?

I Hausman test based on asset pricing theory

I Compare …xed-e¤ects with OLS I any variable that makes stock more responsive to mon pol should

directly a¤ect …rms’ expected returns

I OVB should a¤ect not only coe¢cient of bank dependence

interacted with surprise, but also uninteracted

I Hausman test does not reject null of exogeneity

I Analyze determinants of bank debt usage

17 / 52

Determinants of Bank Debt Usage

Bank debt users not more likely to be risky or interest rate sensitive

18 / 52

Why is Bank Debt Special?

19 / 52

Bank Centered Transmission Mechanisms

I Bank lending to …rms is special for monetary policy transmission

I Bank dependent …rms are more responsive 20 / 52

Bank Centered Transmission Mechanisms

I Bank lending to …rms is special for monetary policy transmission

I Bank dependent …rms are more responsive

I Bank centered channels of monetary policy transmission that predict

stronger responsiveness

20 / 52

Bank Centered Transmission Mechanisms

I Bank lending to …rms is special for monetary policy transmission

I Bank dependent …rms are more responsive

I Bank centered channels of monetary policy transmission that predict

stronger responsiveness

I Bank Lending Channel X 20 / 52

Bank Centered Transmission Mechanisms

I Bank lending to …rms is special for monetary policy transmission

I Bank dependent …rms are more responsive

I Bank centered channels of monetary policy transmission that predict

stronger responsiveness

I Bank Lending Channel X I Interest Rate Pass-Through Channel X 20 / 52

Bank Centered Transmission Mechanisms

I Bank lending to …rms is special for monetary policy transmission

I Bank dependent …rms are more responsive

I Bank centered channels of monetary policy transmission that predict

stronger responsiveness

I Bank Lending Channel X I Interest Rate Pass-Through Channel X 20 / 52

The Bank Lending Channel

21 / 52

The Bank Lending Channel

I Loan supply of …nancially constrained banks more sensitive to

monetary policy

I Theory I Bernanke and Blinder (1988), Bernanke and Gertler (1995), Stein

(1998), Bolton and Freixas (2006), Gertler and Karadi (2009)

22 / 52

The Bank Lending Channel

I Loan supply of …nancially constrained banks more sensitive to

monetary policy

I Theory I Bernanke and Blinder (1988), Bernanke and Gertler (1995), Stein

(1998), Bolton and Freixas (2006), Gertler and Karadi (2009)

I Evidence I bank size: Kashyap and Stein (1995), Kashyap and Stein (2000) I liquidity: Kashyap and Stein (2000), Jimenez, Ongena, Peydró and

Saurina (2012)

I capitalization/leverage: Kishan and Opiela (2000), Jimenez,

Ongena, Peydró and Saurina (2012)

I a¢liation with holding company: Ashcraft (2006) 22 / 52

The Bank Lending Channel

I Loan supply of …nancially constrained banks more sensitive to

monetary policy

I Theory I Bernanke and Blinder (1988), Bernanke and Gertler (1995), Stein

(1998), Bolton and Freixas (2006), Gertler and Karadi (2009)

I Evidence I bank size: Kashyap and Stein (1995), Kashyap and Stein (2000) I liquidity: Kashyap and Stein (2000), Jimenez, Ongena, Peydró and

Saurina (2012)

I capitalization/leverage: Kishan and Opiela (2000), Jimenez,

Ongena, Peydró and Saurina (2012)

I a¢liation with holding company: Ashcraft (2006)

I Firms …nd it hard to substitute bank lending

I Khwaja and Mian (2008), Chava and Purnanandam (2011), Jimenez,

Ongena, Peydró and Saurina (2012)

22 / 52

Challenge: Matching Firms to Banks

I Use LPC Dealscan database of syndicated lending

I wide coverage

I Calculate for each …rm-bank-year:

I share lent by each bank in all lending received in previous 5 years I robustness: only leads, vs. leads and participants

I Calculate for each …rm-year:

I weighted average …nancial health of all banks (weight: share lent

by each bank)

23 / 52

Matching Firms to Banks

24 / 52

Bank Lending Channel - Lender’s Health Variables

I Size: total assets

I Motivation: Kashyap and Stein (1995, 2000)

I Capital Ratio: tier 1 capital / total risk-weighted assets

I Motivation: Kishan and Opiela (2000), Jimenez, Ongena, Peydró

and Saurina (2012)

25 / 52

Bank Lending Channel - Test

I Speci…cation

Reti,t = β0 + β1Surpriset + β2 (BankDebt/At)i,t1 + β3BankHealthi,t1 +β4Surpriset (BankDebt/At)i,t1 BankHealthi,t1 +(2nd order interactions) +γControlsi,t1 + λSurpriset Controlsi,t1 + εi,t

26 / 52

Bank Lending Channel - Test

I Speci…cation

Reti,t = β0 + β1Surpriset + β2 (BankDebt/At)i,t1 + β3BankHealthi,t1 +β4Surpriset (BankDebt/At)i,t1 BankHealthi,t1 +(2nd order interactions) +γControlsi,t1 + λSurpriset Controlsi,t1 + εi,t

I Bank lending channel: β4 > 0

26 / 52

Bank Lending Channel - Bank Size

FULL SAMPLE

Bank dependent …rms borrowing from large banks are relatively less responsive

27 / 52

Bank Lending Channel - Bank Size

SUBSAMPLES ACCORDING TO ACCESS TO ALTERNATIVE SOURCES OF FINANCE

Bank dependent …rms borrowing from large banks are relatively less responsive, particularly if unrated/young

28 / 52

Bank Lending Channel - Capital Ratio

FULL SAMPLE

Bank dependent …rms borrowing from well-capitalized banks are relatively less responsive

29 / 52

Bank Lending Channel - Capital Ratio

SUBSAMPLES ACCORDING TO ACCESS TO ALTERNATIVE SOURCES OF FINANCE

Bank dependent …rms borrowing from well-capitalized banks are relatively less responsive, particularly if unrated/young

30 / 52

Selection Bias

I Possible omitted variables bias due to endogenous matching of

banks to …rms

I For example: large and well capitalized banks specialize in riskier

borrowers.... (double interaction term)

I ....but bias towards risky borrowers decreases with bank dependence

(triple interaction term)

31 / 52

Selection Bias

I Possible omitted variables bias due to endogenous matching of

banks to …rms

I For example: large and well capitalized banks specialize in riskier

borrowers.... (double interaction term)

I ....but bias towards risky borrowers decreases with bank dependence

(triple interaction term)

I To deal with endogeneity concerns

I evaluate existence of endogenous matching of banks to …rms (on

basis of observables)

I instrumental variable estimation 31 / 52

Matching of Banks to Firms

Bank dependence does not increase propensity of healthy banks to lend to safe …rms

32 / 52

Instrumental Variables Estimation

I Bank Size IV Estimation

I Instrument: log median size of banks in the Metropolitan Statistical

Area (MSA) …rm is located (Berger et al. (2005))

I Rationale: …rm more likely to form lending relationships with banks

located close (relevance condition)

I Median size of banks in a region unrelated to stock returns

(exclusion restriction)

33 / 52

Instrumental Variables Estimation

I Bank Size IV Estimation

I Instrument: log median size of banks in the Metropolitan Statistical

Area (MSA) …rm is located (Berger et al. (2005))

I Rationale: …rm more likely to form lending relationships with banks

located close (relevance condition)

I Median size of banks in a region unrelated to stock returns

(exclusion restriction)

I Capital Ratio IV Estimation

I Instrument: corporate income tax rate in the state in which a …rm is

located (Ashcraft (2008) and Berger and Bouwman (2013))

I Interest on debt is tax-deductible, but dividends are not (relevance

condition)

I Corporate income tax rate in a state is unlikely to directly a¤ect the

stock returns of …rms in that state (exclusion restriction)

33 / 52

Instrumental Variables Estimation

Instrumental variables results con…rm OLS regression results

34 / 52

The Interest Rate Pass-Through Channel

35 / 52

Interest Rate Pass-Through Channel

I Floating vs …xed-rates

I Widespread use of ‡oating-rates in bank loans I Prevalence of …xed-rates in nonbank liabilities 36 / 52

Interest Rate Pass-Through Channel

I Floating vs …xed-rates

I Widespread use of ‡oating-rates in bank loans I Prevalence of …xed-rates in nonbank liabilities

I Novel transmission channel: interest rate pass-through channel I Mechanism:

I Floating rates calculated as spread over reference rate (LIBOR, prime

rate,...)

I Monetary policy actions ) reference rates ) cost of existing bank

loans for …rms

36 / 52

Interest Rate Pass-Through Channel

I Floating vs …xed-rates

I Widespread use of ‡oating-rates in bank loans I Prevalence of …xed-rates in nonbank liabilities

I Novel transmission channel: interest rate pass-through channel I Mechanism:

I Floating rates calculated as spread over reference rate (LIBOR, prime

rate,...)

I Monetary policy actions ) reference rates ) cost of existing bank

loans for …rms

I Importance

I Absence of …nancing frictions: impact is present value of variation in

interest expense

I Financing frictions: impact ampli…ed 36 / 52

Transmission Mechanisms

37 / 52

Floating Rate Debt

Most bank debt is ‡oating-rate, most nonbank debt is …xed-rate

38 / 52

Interest Rate Pass-Through Channel: Test

I Test: all else equal, bank-dependent …rms that engage in interest

rate risk hedging should be less responsive to monetary policy

39 / 52

Interest Rate Pass-Through Channel: Test

I Test: all else equal, bank-dependent …rms that engage in interest

rate risk hedging should be less responsive to monetary policy

I Use text-search algorithm to collect ‡oating-to-…xed rate hedging

from SEC 10-K …lings

39 / 52

Pass-through Channel - Hedging De…nition

FILED AS OF DATE: 20060331 COMPANY NAME: NETSMART TECHNOLOGIES INC "The term loan bears interest at LIBOR plus 2.25%. We have entered into an interest rate swap agreement with the Bank for the amount outstanding under the term loan whereby we converted

• ur variable rate on the term loan to a …xed rate of 7.1% in order

to reduce the interest rate risk associated with these borrowings."

40 / 52

Interest Rate Pass-Through Channel: Test

I Test: all else equal, bank-dependent …rms that engage in interest

rate risk hedging should be less responsive to monetary policy

I Use text-search algorithm to collect ‡oating-to-…xed rate hedging

from SEC 10-K …lings

I Regression speci…cation:

Reti,t = β0 + β1Surpriset + β4Surpriset (BankDebt/At)i,t1 Hedgei,t1 + β5Surpriset (BankDebt/At)i,t1 FinConstrainti,t1 + (second order terms) + γControlsi,t1 + λSurpriset Controlsi,t1 + εi,t

I Pass-through channel: β4 > 0

41 / 52

Pass-through Channel - The Role of Hedging

Bank dependent …rms that hedge are relatively less responsive to monetary policy surprises

42 / 52

Pass-through Channel - Robustness

I Firm’s decision to hedge clearly endogenous I Most theories for widespread use of hedging to do with …nancing

constraints.

I Controversy I Financially constrained more likely to hedge (Froot, Scharfstein and

Stein [1993], Stulz [1984])

I Financially constrained more likely to hedge (Rampini, Su…, and

Viswanathan [2012])

I Evidence: …rms more likely to face …nancial constraints (small

…rms,...) less likely to manage risk (Stulz [1996])

I Important to control for several measures of …nancial constraints

I rating, size and age (Almeida, Campello, and Weisbach [2004],

Hadlock and Pierce [2013]).

43 / 52

Pass-through Channel - Robustness

Impact of hedging robust to controlling for …nancial constraints

44 / 52

Pass-through Channel - The Cost of Debt

Cost of debt of non-hedgers more sensitive to monetary policy

45 / 52

Conclusion

I Bank-dependent …rms are more responsive to monetary policy shocks

I controlling for leverage, …nancial constraints, interest rate sensitivity,

maturity, asset pricing factors

I Evidence of a Bank Lending Channel

I …rms borrowing from healthy banks less responsive I especially if they can’t access bond markets

I Evidence of a Pass-Through Channel

I interest rate hedgers are less responsive 46 / 52

APPENDIX

47 / 52

Firm Size and Sensitivity to Monetary Policy

I Evidence that small and unrated …rms react more to monetary policy

I Measured by inventory investment, sales I Gertler and Gilchrist (1994) I Measured by stock price reaction I Ehrmann and Fratzscher (2004)

I Our evidence for all sample (1989-2008)

I large …rms are relatively more sensitive 48 / 52

Firm Size and Sensitivity to Monetary Policy

Robustness checks on the sign of controls across di¤erent samples

49 / 52

The Risk Taking Channel

I Low interest rates might increase banks’ incentives to take on more

risks

I Theory: Bernanke (2003), Stiglitz and Greenwald (2003), Diamond

and Rajan (2006), Borio and Zhu (2007)

I Evidence: Maddaloni and Peydro (2011), Jiménez, Ongena, Peydró

and Saurina (2011), Dell’Ariccia, Laeven and Suarez (2013)

50 / 52

Analysis of Outliers

Outliers included

51 / 52

Analysis of Outliers

Outliers excluded

52 / 52