Imperfect Information, Lending Standards and Macroprudential

Imperfect Information, Lending
Standards and Macroprudential Policies
Pedro Gete and Natalie Tiernan
Georgetown University
June 2012
An undisputed cause of the crisis
I
Banks gave "too much" credit to low
quality borrowers. Banks did not screen
"enough"
I
I
This happened in U.S., Spain, U.K.,
Iceland, Ireland, Greece...
Many of these economies had
boom/bust in credit, output and the
current account
This paper:
I
I
Study banks' screening and lending
decisions
Focus on banks' expectations
Rational overoptimism or
overpessimism
I
I
Study three macroprudential tools:
Capital requirements
I
I
Taxes on banks' borrowings
I
Taxes on banks' lending
Example of Rational Overoptimism
"Spain's economic success over the past
years has been most impressive...
GDP growth is likely to remain above the
euro-area average of just below 2% for
several more years, allowing Spain to
climb past Italy and Germany in the
rankings of GDP per capita by 2020"
Research Department of Deutsche Bank (2007)
Banks face two information problems when
lending:
1)
Asymmetric information about
borrower's type
I
I
Banks screen borrowers to mitigate
problem
Screening is costly
2)
Imperfect information about the
aggregate economy
I
I
I
Sometimes growth is persistent,
sometimes not
Banks need to forecast
Leads to rational overoptimism or
overpessimism
Summary of results
I
Dynamics of bank expectations =)
dynamics of lending standards
(screening)
I
I
Higher volatility in the Great
Moderation period
Rational Overoptimism =)lax lending
standards=) boom/bust
I
New role for macroprudential regulation:
I
I
if banks less risk averse than society
=) lax lending standards
Imperfect information magni es the
problem
I
The three policy tools help achieve right
lending standards
They alter costs/bene ts of screening
I
I
I
I
They should be time-varying
Taxes on borrowing or lending: between
0.1% and 1%
Capital requirements: between 4% and
13%
The Model
I
There are banks and borrowers
I
Borrowers need credit Lt to produce
I
Credit is not-collateralized (commercial
loans)
I
Borrower's income
yt (!; st ; Lt ) = st ! Lt
I
Two sources of risk in a loan:
aggregate productivity shock
I
st =
I
!=
I
idiosyncratic productivity
Borrowers are heterogeneous
!
U [0; 1]
Banks
I
I
Banks need to pay screening cost to
discover !
Screening cost modeled as an
"opportunity cost":
"
I
screening=)# sales
Loan of cers checking credit records
could be salesmen attracting
customers
Imperfect information
I
is observable, but not its components
st
st = zt +
t
is persistent:
zt follows 2-state Markov Chain
I
zt
I
t
is not persistent:
t is i:i:d: Normal
2
2
;
2
I
Banks base period t decisions on prior
pt
I
Pr(zt = zH j
Once st observed at end of period t,
compute posteriors using Bayesian lter
Pr(zt = zi j
I
t 1)
t)
=
f (st jzt
=zj ;
t
f (st jzt =zi ;
j
1 ) Pr(zt =z j t
i
t 1 ) Pr(zt =z j
i
1 )+f (st jzt =z ;
t 1)
i
t 1 ) Pr(zt =z j
t 1)
Next period's prior is the posterior
updated with transition matrix
Pr(zt+1 = zi j
t)
= Pr(zt = zi j
t )Pii
+ Pr(zt = zj j
t )Pji
Banks make two decisions:
1)
How many resources to allocate to
screening?
I
I
Choose ; the probability of
successfully matching with a
borrower
(1
) is probability of successfully
discovering a borrower's type
2)
Matched bank (informed or uninformed):
to give credit or not?
I
Two ways to nance a loan:
banks' capital (retained earnings)
external funding (partial eq'm
view)
I
I
Lt = K t +Bt
I
Capital requirements limit loan size
Kt
Lt
Bank's problem at each period t
I
We focus on quantity of credit, not on
price of credit:
I
I
I
Banks can observe yt (!; st ; Lt )
0
is an unseizable fraction of output
Banks receive remaining portion:
Rt Lt = (1
) yt (!; st ; Lt )
Results
Result #1
I
I
I
Dynamics of lending standards depend
on information content of economic
news 2
In noisy times, smaller changes
Model prediction matches new fact:
larger volatility in lending standards
since Great Moderation
IRs to TFP shock for different amounts of noise
Lending Standards
Signals and Beliefs
1.012
0
1
1.008
0.9
1.006
0.85
1.004
0.8
1.002
0.75
1
0.998
-2
0.7
0
2
4
6
8
0.65
H
Aggregate Productivity Level
t
p if high noise (right scale) 0.95
p if low noise (right scale)
1.01
Prior of being in z
Change in Screening Intensity (%)
exp(s ) (left scale)
-0.2
-0.4
-0.6
-0.8
-1
-1.2
High noise
Low noise
-1.4
-2
0
2
4
6
8
A new fact
I
I
I
Two samples for lending standards:
1967Q1-1983Q4 and 1990Q2-2008Q3
1990Q2-2008Q3 is less noisy (estimated
two state Markov switching model à la
Hamilton 1989)
Identify technology shock using long run
restrictions in VAR (Blanchard-Quah)
only tech shocks have permanent
effect on the level of output
I
In data: higher reaction in less noisy times
Result #2
I
I
I
Fluctuations in beliefs=) uctuations in
lending standards
Expect higher aggregate
productivity=)lower screening intensity
Problem: noise shocks lead banks to
overestimations/underestimations
Observations +-2 standard deviations from mean?
PDF of Conditional Densities
50
H
f(s | z )
t
45
L
f(s | z )
t
40
35
30
25
20
15
10
5
0
-0.05
0
s
t
0.05
A period of overoptimism leads to sudden stop
Rational Overoptimism
1.02
1
exp(z ) (left scale)
t
exp(s ) (left scale)
H
1
0.98
0
0.5
1
2
3
4
5
6
0
7
Prior of being in z
Aggregate Productivity Level
t
p (right scale)
Lax screening, Boom/bust in credit & output
Lending Standards
Credit
22
20
-0.2
18
-0.4
Loans (in Levels)
Change in Screening Intensity (%)
0
-0.6
-0.8
-1
-1.2
14
12
10
8
-1.4
-1.6
16
6
0
1
2
3
4
5
6
7
4
0
1
2
3
4
5
6
7
Result #3: Macroprudential Policy
I
I
I
I
Assume banks are risk neutral but
households are risk averse
Risk neutral banks are "too much"
exposed to agg productivity shocks
Imperfect information magni es that risk
Taxes on banks' borrowings or lending,
or capital requirements help achieve
social optimum
Lending Standards
0.485
Screening Intensity
0.48
0.475
Risk Neutral
Risk Averse
0.47
0.465
0.46
0
0.2
0.4
0.6
p
0.8
1
Rational Overoptim ism
1.02
1
exp(z ) (left scale)
t
exp(s ) (left scale)
H
1
0.98
0
0.5
1
2
3
4
5
6
0
7
Prior of being in z
Aggregate Productivity Level
t
p (right scale)
Lending Standards
Credit
22
0
20
18
-0.4
Loans (in Levels)
Change in Screening Intensity (%)
-0.2
Risk Neutral
Risk Averse
-0.6
-0.8
-1
-1.2
Risk Neutral
Risk Averse
14
12
10
8
-1.4
-1.6
16
6
0
1
2
3
4
5
6
7
4
0
1
2
3
4
5
6
7
Capital
2.4
2.2
Capital (in Levels)
2
1.8
1.6
Risk Neutral
Risk Averse
1.4
1.2
1
0
1
2
3
4
5
6
7
I
The three policy tools help achieve right
lending standards
They alter costs/bene ts of screening
I
I
Macroprudential tools should lean
against banks' beliefs
I
When optimism, tighten policy
I
When pessimism, loosen policy
Capital requirements
Capital Requirem ent
13
Capital Requirement (%)
12
11
10
9
8
7
6
5
4
0
0.2
0.4
0.6
p
0.8
1
Tax on bank lending
l )Rt Lt
(1
Lending Tax
1
0.9
0.8
Tax (%)
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.2
0.4
0.6
p
0.8
1
Tax on bank borrowing
(1 +
b )(1
+ ib )Bt
Borrow ing Tax
1.4
1.2
Tax (%)
1
0.8
0.6
0.4
0.2
0
0
0.2
0.4
0.6
p
0.8
1
Conclusions
I
I
Wm. McC. Martin, Jr. (1955):
"The job of the Federal Reserve is to
take away the punch bowl just as the
party gets going"
Our version:
"One job of the macro regulator is to
announce clouds when banks think it is
sunny, and sun when it rains"
Appendix
Quantitative Properties of Calibrated Model:
I
For U.S. banking system, 1987-2010, the
model matches:
I
Average return to capital,
Capital/asset ratio, Net interest
margin, Ratio of losses to total loans
I
Volatilities quality/quantity credit
I
Correlations quality/quantity credit