Which Financial Frictions? Parsing the
Evidence from the Financial Crisis of 2007-9
Tobias Adrian
Paolo Colla
April 2012
Hyun Song Shin
Adrian, Colla and Shin: Which Financial Frictions?
1
An Old Debate
• Do financial frictions operate through demand for credit?
— Shocks to borrower net worth, NPV of project, collateral value of
assets
— Bernanke and Gertler (1989), Kiyotaki and Moore (1997)
• Or, do financial frictions operate through supply of credit?
— Bank lending channel
— Kashyap, Stein and Wilcox (1993)
Adrian, Colla and Shin: Which Financial Frictions?
2
This Paper
• Examine evidence for 2007-9 crisis, pointing to:
— Inelastic demand for credit by firms
— Sharp contraction in supply of intermediated credit
— Shortfall made up by sharp increase in demand for direct credit
Adrian, Colla and Shin: Which Financial Frictions?
3
This Paper
• Examine evidence for 2007-9 crisis, pointing to:
— Inelastic demand for credit by firms
— Sharp contraction in supply of intermediated credit
— Shortfall made up by sharp increase in demand for direct credit
• Question: why is one dollar of credit through the banking system so
different from one dollar of credit that flows directly?
— Focus on behavior of banks
— Checklist of stylized facts
Adrian, Colla and Shin: Which Financial Frictions?
4
This Paper
• Examine evidence for 2007-9 crisis, pointing to:
— Inelastic demand for credit by firms
— Sharp contraction in supply of intermediated credit
— Shortfall made up by sharp increase in demand for direct credit
• Question: why is one dollar of credit through the banking system so
different from one dollar of credit that flows directly?
— Focus on behavior of banks
— Checklist of stylized facts
• Model of direct and intermediated credit from checklist
Adrian, Colla and Shin: Which Financial Frictions?
5
Non-corporate business sector total borrowing
(trillion dollars)
4.5
farm credit
system loans
4.0
US government
loans
3.5
other loans and
advances
3.0
2.5
bank loans
n.e.c.
2.0
commercial
mortgages
1.5
1.0
0.5
2011Q1
2009Q3
2008Q1
2006Q3
2005Q1
2003Q3
2002Q1
2000Q3
1999Q1
1997Q3
1996Q1
1994Q3
1993Q1
1991Q3
1990Q1
0.0
multifamily
residential
mortgages
home
mortgages
Figure 1. Credit to non-financial non-corporate businesses (Source: US Flow of Funds, tables L103, L104)
Trillion dollars
Adrian, Colla and Shin: Which Financial Frictions?
6
8.0
7.0
Corporate
bonds
6.0
Commercial
paper
5.0
4.0
Other loans
and
advances
3.0
2.0
Bank loans
n.e.c.
1.0
2011Q1
2009Q3
2008Q1
2006Q3
2005Q1
2003Q3
2002Q1
2000Q3
1999Q1
1997Q3
1996Q1
1994Q3
1993Q1
1991Q3
1990Q1
0.0
Total
mortgages
Figure 2. Credit to US non-financial corporate sector (US Flow of Funds, table L102)
Billion Dollars
Adrian, Colla and Shin: Which Financial Frictions?
7
800
600
400
Change
in
corporate
bonds
200
0
Change
in loans
-200
-400
2011Q1
2010Q1
2009Q1
2008Q1
2007Q1
2006Q1
2005Q1
2004Q1
2003Q1
2002Q1
2001Q1
2000Q1
1999Q1
1998Q1
1997Q1
1996Q1
1995Q1
1994Q1
1993Q1
1992Q1
1991Q1
1990Q1
-600
Figure 3. Changes in outstanding corporate bonds and loans to US non-financial corporate sector. Loans
are defined as sum of mortgages, bank loans not elsewhere classified (n.e.c.) and other loans (US Flow of
Funds, table F102)
Adrian, Colla and Shin: Which Financial Frictions?
8
Corporate Finance of Banking
A
L
Equity
Assets
Debt
Adrian, Colla and Shin: Which Financial Frictions?
A
L
9
A
L
Equity
Equity
Assets
Assets
Debt
Debt
Adrian, Colla and Shin: Which Financial Frictions?
A
L
10
A
Equity
L
Equity
Assets
Debt
Assets
Debt
Adrian, Colla and Shin: Which Financial Frictions?
11
Investment Banks (1994Q1 - 2011Q2)
Change in Equity & Changes in Debt
(Billions)
300
200
y = 0.9948x - 0.6713
100
0
y = 0.0052x + 0.6713
-100
-200
Equity
-300
-400
-400
Debt
-300
-200
-100
0
100
Change in Assets (Billions)
200
300
Figure 4. Scatter chart of {(∆  ∆ )} and {(∆  ∆ )} for changes in assets, equity and debt of
US investment bank sector consisting of Bear Stearns, Goldman Sachs, Lehman Brothers, Merrill Lynch and
Morgan Stanley between 1994Q1 and 2011Q2 (Source: SEC 10Q filings)
Adrian, Colla and Shin: Which Financial Frictions?
12
Commercial Banks (Call Reports) 1984Q1 - 2010Q2
Change in Equity & Changes in Debt (Billions)
800
600
y = 0.9779x - 9.5511
400
200
y = 0.0221x + 9.5511
0
-200
Equity
-400
-600
-400
Debt
-200
0
200
400
600
800
Change in Assets (Billions)
Figure 5. Scatter chart of {(∆  ∆ )} and {(∆  ∆ )} for changes in assets, equity and debt of
US commercial bank sector between 1984Q1 and 2010Q2 (Source: FDIC call reports).
Adrian, Colla and Shin: Which Financial Frictions?
13
BNP Paribas: annual change in assets, equity and debt
(1999 - 2010)
500
y = 1.0051x - 6.2
R2 = 0.9987
Change in equity and debt
(billion euros)
400
300
Debt
Change
200
100
Equity
Change
0
-100
-200
-200
-100
0
100
200
300
400
500
Asset change (billion euros)
Figure 6. BNP Paribas: annual change in assets, equity and debt (1999-2010) (Source: Bankscope)
Adrian, Colla and Shin: Which Financial Frictions?
14
Societe Generale: annual changes in assets, equity and debt
(1999 - 2010)
Annual change in equity and debt
(billion euros)
300
250
y = 0.996x - 3.15
R2 = 0.9985
200
150
Debt
change
100
50
Equity
change
0
-50
-100
-150
-200
-100
0
100
200
300
Annual asset change (billion euros)
Figure 7. Société Générale: annual change in assets, equity and debt (1999-2010) (Source: Bankscope)
Adrian, Colla and Shin: Which Financial Frictions?
15
Barclays: 2 year change in assets, equity, debt
and risk-weighted assets (1992 -2010)
2 year change in equity, debt and
risk-weighted assets (billion pounds)
1,000
800
y = 0.9974x - 0.175
R2 = 0.9998
600
2yr RWA
Change
400
200
2yr Equity
Change
0
-200
2yr Debt
Change
-400
-600
-800
-1,000
-1,000
-500
0
500
1,000
2 year asset change (billion pounds)
Figure 8. Barclays: 2 year change in assets, equity and debt (1992-2010) (Source: Bankscope)
Adrian, Colla and Shin: Which Financial Frictions?
16
Societe Generale: 2 year changes in assets, risk-weighted assets,
equity and debt (1999 - 2010)
2 year change in risk-weighted assets,
equity and debt (billion euros)
400
y = 0.9986x - 6.65
R2 = 0.9988
350
300
250
2yr debt
change
200
150
2yr equity
change
100
50
0
2yr RWA
change
y = 0.0014x + 6.65
-50
-100
-100
0
100
200
300
400
2 year asset change (billion euros)
Figure 9. Société Générale: 2 year change in assets, equity and debt (1999-2010) (Source: Bankscope)
Adrian, Colla and Shin: Which Financial Frictions?
17
Checklist for the Banking Sector
• Bank lending changes dollar-for-dollar through change in debt, with
equity “sticky”
• Equivalently, bank chooses leverage given pre-determined equity
• Implication: banking sector leverage is procyclical
Adrian, Colla and Shin: Which Financial Frictions?
18
Micro Evidence
• Sample: U.S. public firms 1998-2010
• Intersection between
— Compustat
— Loan Pricing Corporation (LPC) Dealscan database
— Securities Data Corporation (SDC) New Bond Issuances database
• 3,896 firms with new financing between 1998 and 2010 (out of 11,538
in Compustat sample)
Adrian, Colla and Shin: Which Financial Frictions?
19
Total Credit
New debt: Cost
0
100
50
200
bps
bln USD
100
300
150
400
200
New debt: Total amount
98
99
00
01
02
03
04 05
Time
06
07
08
09
10
98
99
00
01
02
03
04 05
Time
From Q2:2007 to Q2:2009: total amount 1/2, spread 4x
06
07
08
09
10
Adrian, Colla and Shin: Which Financial Frictions?
20
Split between Loans and Bonds
Bond financing: Total amount
0
0
20
50
bln USD
bln USD
40
100
60
150
80
Loan financing: Total amount
98
99
00
01
02
03
04 05
Time
06
07
08
09
10
98
99
00
From Q2:2007 to Q2:2009: loans 1/4, bonds 2x
01
02
03
04 05
Time
06
07
08
09
10
Adrian, Colla and Shin: Which Financial Frictions?
21
Pricing
200
bps
300
400
500
Bond financing: Cost
0
100
100
bps
200
300
400
Loan financing: Cost
98
99
00
01
02
03
04 05
Time
06
07
08
09
10
98
99
00
From Q2:2007 to Q2:2009: loans 4x, bonds 3x
01
02
03
04 05
Time
06
07
08
09
10
Adrian, Colla and Shin: Which Financial Frictions?
22
Firms with Access to Both Loans and Bonds
Firms with access to both types of credit can be used to identify demand
and supply shocks
Rated New issuer Crisis
(July 2007) Q2:2005 Q3:2005
Q2:2009
Q2:2007 Q3:2007
New debt (cum.)
New debt (cum.)
Sorting Before crisis
Crisis To qualify: obtained credit during crisis, positive assets before crisis,
non-missing firm characteristics
Sorting (rating and firm characteristics) based on Q2:2005
Adrian, Colla and Shin: Which Financial Frictions?
23
One Dimensional Sorts
Panel A: Amount
Total
Relative to
Before
median
Crisis
Crisis
Loan
t-stat
Before
Crisis
Bond
t-stat
Crisis
Before
Crisis
t-stat
Crisis
Size
Below
0.313
0.383
2.476**
0.280
0.288
0.290
0.033
0.095
5.520***
Above
1.791
1.709
-0.532
1.491
0.770
-6.122***
0.300
0.939
6.255***
Tobin’s Q
Below
1.066
0.989
-0.665
0.914
0.574
-3.475***
0.152
0.415
3.289***
Above
0.981
1.193
1.499
0.789
0.487
-3.153***
0.192
0.706
5.037***
(more...)
Adrian, Colla and Shin: Which Financial Frictions?
24
Panel B: Cost
Total
Relative to
Before
After
median
Crisis
Crisis
Loan
t-stat
Before
After
Crisis
Crisis
Bond
t-stat
Before
After
Crisis
Crisis
t-stat
Size
Below
129.26
262.11
8.185***
116.87
205.69
6.205***
218.40
449.84
6.222***
Above
73.55
263.02
11.855***
66.18
144.32
6.858***
112.98
376.87
9.394***
Tobin’s Q
Below
106.80
269.23
9.027***
94.88
193.67
7.290***
178.81
479.21
7.847***
Above
84.87
244.24
10.670***
77.10
151.18
5.526***
126.13
347.37
8.203***
Tangibility
Below
89.87
250.18
10.536***
81.33
170.47
7.265***
122.06
360.01
7.901***
Above
103.32
277.55
10.125***
92.00
190.32
7.086***
166.60
444.00
8.111***
Rating
Below
132.94
302.43
10.949***
120.87
216.40
7.999***
208.02
528.24
10.055***
Above
42.54
190.99
10.868***
34.40
92.80
6.261***
81.00
280.24
8.784***
(more...)
Adrian, Colla and Shin: Which Financial Frictions?
25
Large Firms' New Loans
80
70
60
Frequency
Before
50
After
40
30
20
10
More
14.0
13.5
13.0
12.5
12.0
11.5
11.0
10.5
10.0
9.5
9.0
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0
New loans ($ Billions)
Figure 10. New loans to large firms (upper tercile) before and after the crisis
Adrian, Colla and Shin: Which Financial Frictions?
26
Large Firms' Bond Issuance
120
110
100
Frequency
90
80
Before
70
60
After
50
40
30
20
10
More
7.0
6.5
6.0
5.5
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0
Bond Issuance ($ Billions)
Figure 11. Bond issuance by large firms (upper tercile) before and after the crisis in billions of dollars
Adrian, Colla and Shin: Which Financial Frictions?
One Dimensional Sorts: Main findings
• Flow of new credit: no change overall (!)
— New loans: sharp reduction;
— New bonds: sharp increase.
• Spreads: sharp increase.
27
Adrian, Colla and Shin: Which Financial Frictions?
28
Logit Analysis for Bond Issuance
• Determinants of new loans and bonds in regression framework (Denis
and Mihov 2003, and Becker and Ivashina 2011).
• Details:
— Sample: new issuers, rated the quarter prior to issuance
— Dependent variable:   = 1 if bond issued during
quarter , and = 0 if loan (robust to exclusion of simultaneous issuers,
191 obs.)
— Independent variables: firm characteristics (as before)
— Specification: logit model
Adrian, Colla and Shin: Which Financial Frictions?
Loan supply proxy
Crisis
Monetary
policy
29
BD
leverage
Lending
practice
Non-perf.
loans
EBP
Panel A: logit regressions (dependent variable: bond issuance)
Size
Tobin’s Q
Tangibility
Rating
Profitability
Leverage
Loan supply proxy
Observations
Pseudo R-squared
Loan supply proxy
0.337***
(0.042)
0.061
(0.075)
-0.285
(0.203)
0.052***
(0.019)
11.286***
(2.506)
1.007***
(0.255)
0.621***
(0.087)
0.337***
(0.042)
0.033
(0.074)
-0.270
(0.207)
0.061***
(0.019)
11.208***
(2.543)
1.039***
(0.257)
0.232***
(0.031)
0.366***
(0.042)
0.119
(0.076)
-0.236
(0.206)
0.045**
(0.019)
11.733***
(2.503)
0.798***
(0.263)
-0.009***
(0.001)
0.327***
(0.042)
0.070
(0.076)
-0.308
(0.205)
0.059***
(0.020)
11.192***
(2.537)
1.027***
(0.257)
0.015***
(0.001)
0.313***
(0.043)
0.174**
(0.078)
-0.211
(0.213)
0.087***
(0.021)
10.065***
(2.649)
0.974***
(0.270)
0.404***
(0.023)
0.354***
(0.042)
0.087
(0.075)
-0.244
(0.211)
0.039*
(0.020)
12.622***
(2.538)
0.796***
(0.259)
0.591***
(0.046)
4,276
0.075
4,276
0.077
4,276
0.079
4,276
0.085
4,276
0.128
4,153
0.095
Panel B: changes in implied probabilities
0.158
-0.179
0.257
0.422
0.238
0.140
Adrian, Colla and Shin: Which Financial Frictions?
30
Model Checklist
1. Direct and intermediated credit
2. In downturn, new loans contract but bond issuance increases
3. Spreads increase on both loans and bonds
4. Bank lending increases or decreases dollar for dollar with an increase or
decrease in debt, with equity being sticky
5. Bank leverage is procyclical
Adrian, Colla and Shin: Which Financial Frictions?
31
Model of Direct and Intermediated Finance
Intermediated
Credit
Banks
Ultimate
Borrowers
Claim
Ultimate
Creditors
Directly granted credit
• Banking sector
• Mean-variance investors who hold portfolio of (i) cash (ii) bank liabilities
(iii) risky loans
Adrian, Colla and Shin: Which Financial Frictions?
32
Bank Credit Supply
Notation for balance sheet of bank
Bank
E
1 r
C
1 f
L
Adrian, Colla and Shin: Which Financial Frictions?
33
Credit Risk
Vasicek (2002) model, backbone of Basel capital requirements.
Project  succeeds when   0, where
−1
 = −Φ
p
√
() +  + 1 − 
Φ () c.d.f. of standard normal,  and { } independent standard normals
³√
´
p
 + 1 −   Φ−1 ()
Pr (  0) = Pr
¡ −1 ¢
= Φ Φ () = 
Adrian, Colla and Shin: Which Financial Frictions?
34
Bank diversifies away idiosyncractic risk
Conditional on  , defaults are independent.
Keep  fixed but diversify: increase number of borrowers, reduce face value
of individual loans
In the limit, realized value of assets is function of  only
 ( ) ≡ (1 + )  · Pr ( ≥ 0| )
³√
´
p
−1
= (1 + )  · Pr
 + 1 −  ≥ Φ () |
´
³ √
 −Φ−1 ()
√
= (1 + )  · Φ
1−
(*)
Adrian, Colla and Shin: Which Financial Frictions?
35
12
15
ρ = 0.3
ε = 0.2
8
density over realized assets
density over realized assets
10
ε = 0.1
6
4
ε = 0.2
2
0
12
ρ = 0.01
9
6
ρ = 0.1
3
ε = 0.3
0
0.2
ρ = 0.3
0.4
0.6
z
0.8
1
0
0
0.2
0.4
0.6
0.8
1
z
Figure 12. The two charts plot the densities over realized assets when  (1 + ) = 1. The left hand
charts plots the density over asset realizations of the bank when  = 01 and  is varied from 0.1 to 0.3.
The right hand chart plots the asset realization density when  = 02 and  varies from 0.01 to 0.3.
Adrian, Colla and Shin: Which Financial Frictions?
36
Turning Credit Risk Model on Its Head
• Turn credit risk model on its head and think of it as credit supply model
— Fix . Determine credit supply 

 =

1+
1 − 1+
 (  )
 ∈ (0 1)
 is ratio of notional liabilities to notional assets to be derived below.
Adrian, Colla and Shin: Which Financial Frictions?
37
From (*), the c.d.f. of  is
 () = Pr ( ≤ )
¢
¡
−1
= Pr  ≤  ()
¡ −1 ¢
= Φ  ()
µ
¶¶¶
µ
µ
p

1
= Φ √ Φ−1 () + 1 − Φ−1

(1 + ) 
Common risk factor  determines shape of the density, with larger  implying
fatter tail.
Value-at-Risk (VaR) rule:
probability to   0
keep enough equity to limit insolvency
Adrian, Colla and Shin: Which Financial Frictions?
38
Bank credit supply  determined from
Pr (  (1 +  ) ) = Φ
Ã


√
−1
−1 (1+ )
Φ ()+ 1−Φ
(1+)
√

Notional liabilities (1 +  ) 
=
=Φ
Notional assets
(1 + ) 
where
 (  ) ≡ Φ
³√
!
=
µ √ −1
¶
Φ () − Φ−1 ()
√
1−
Φ−1 ()−Φ−1 ()
√
1−
´
(1)
Adrian, Colla and Shin: Which Financial Frictions?
39
Supply of Credit by Bank
Credit supply  and demand for funding  is obtained from (1) and balance
sheet identity  =  + 
=

1+
1 − 1+
·

=
1+
1+

· 1 − 1
Aggregation holds due to proportionality
Leverage =
1
1+
1 − 1+
·
Risk premium is well-defined
Risk premium = (1 − ) (1 + ) − 1
Adrian, Colla and Shin: Which Financial Frictions?
40
r
1 f
 1
C r 
 / 1   
0
E
1

1   1  f 
Supply of credit
Credit
Supply
Adrian, Colla and Shin: Which Financial Frictions?
41
Mean-Variance Investors
Loans are packaged into bonds that diversify away idiosyncratic risk.
Demand for bonds (supply of credit) by mean-variance investor with risk
tolerance 
 [(1 − ) (1 + ) − 1]
 2 (1 + )2
where  2 is variance of  ( ). There are  mean-variance investors, and
 =   . Aggregate supply of credit from mean-variance sector is
 =
We need to work out  2.
 [(1 − ) (1 + ) − 1]
2
 2 (1 + )
Adrian, Colla and Shin: Which Financial Frictions?
42
Normalized leverage
Variance of asset realization
1.0
0.10
0.9
α = 0.01
0.08
0.7
0.6
variance σ2
normalized leverage φ
0.8
ρ =0.1
0.5
0.4
0.06
ρ =0.3
0.04
0.3
ρ =0.1
0.2
0.02
ρ =0.3
0.1
0
0
0.2
0.4
0.6
0.8
default probability ε
1.0
0
0
0.2
0.4
0.6
0.8
default probability ε
1.0
Figure 13. Left hand panel plots the normalized leverage ratio  as a function of . The right hand panel
plots the variance 2 as a function of epsilon for two values of .
Adrian, Colla and Shin: Which Financial Frictions?
43
Market Clearing
 is risk premium, given by (1 − ) (1 + ) − 1
Credit market clears when total demand for credit is met by direct and
intermediated credit.

1
|
− 1+

1−
{z }

(1 − )2 
+
2 =  ()
2
 (1 + )
|
{z
}

Adrian, Colla and Shin: Which Financial Frictions?
44
Bank Iso-Lending Curves
Points in ( )-space with  constant
µ
¶
 1−
 () = 1 −
−1
  ()
(2)
Slope of the iso-lending curve tends to +∞ as  → 0
µ
¶∙
¸

1
1
−

0

() +
 0 () = − 1 −

2

since 0 () → −∞ as  → 0
(3)
Adrian, Colla and Shin: Which Financial Frictions?
45
Iso−lending curves for banks
α = 0.01
ρ = 0.3
E=1
α = 0.01
ρ = 0.3
T=2
CB =10
1.0
0.8
CB =2
risk premium π
risk premium π
1.0
Iso−lending curves for bond investors
0.6
0.4
0.2
0.8
0.6
C =3
H
0.4
C =2
0.2
H
C =1.5
B
0
0
0.1
0.2
0.3
default probability ε
0.4
0
0
0.1
0.2
0.3
default probability ε
0.4
Figure 14. Iso-lending curves in ( )-space for banks (left panel) and bond investors (right panel).
Parameter values are as indicated in the boxes.
Adrian, Colla and Shin: Which Financial Frictions?
46
Two Main Results
Proposition. Under mild regularity conditions, risk premium  is increasing
in .
Excess bond premium goes up in recessions
Proposition. For demand for credit not too elastic, an increase in  is
associated with a contraction of banking sector assets, both in absolute
terms and as a proportion of the total credit received by borrowers.
In recessions, bank lending contracts but bond lending expands.
Adrian, Colla and Shin: Which Financial Frictions?
0.0010
risk premium π
0.0008
α = 0.01
ρ = 0.3
E=1
T = 10
47
Bank iso−lending
curve, CB = 10
Region D
Bond investor
iso−lending
curve, CH = 10
0.0006
Region A
Region C
0.0004
Region B
0.0002
0
0
0.005
0.010
defaultmprobability ε
0.015
Figure 15. Crossing point for the iso-lending curves of banks and households.
0.020
Adrian, Colla and Shin: Which Financial Frictions?
48
Back to Our Main Question
Why is one dollar of credit through the banking system so different from
one dollar of credit that flows directly?
Answer: Size of banking sector proxies for banking sector risk-taking.
Procyclical behavior of banking sector drives the risk premium over the
cycle.
• Economic activity (esp. investment) sensitive to risk premium.
• Spike in excess spreads is followed by decline in economic activity
(Gilchrist, Yankov and Zakrajsek (2009), Gilchrist and Zakrajsek (2011))
Adrian, Colla and Shin: Which Financial Frictions?
49
Checklist for Macro Models
• Reconcile procyclical leverage with standard dynamic portfolio choice
• Incorporate explanatory power of balance sheet variables for asset pricing
(Adrian, Moench and Shin (2012))
• Explore quantitative impact of shifting composition of credit
• ...