Policy shocks

advertisement
Macroprudential Policy,
Countercyclical Bank Capital Buffers & Credit Supply Cycles:
Evidence from the Spanish Dynamic Provisioning Experiments
Gabriel Jiménez (Banco de España)
Steven Ongena (Tilburg & CEPR)
José-Luis Peydró (European Central Bank)
Jesús Saurina (Banco de España)
CEPR – UoV – OENB Bank Supervision and Resolution – 4th October 2011
Caveats:
1. Work in progress
2. These are our views and do not necessarily reflect those of the Bank of Spain,
the European Central Bank and the Eurosystem
The National Bank of Belgium and its staff
generously supports the writing of this paper
in the framework of the:
Colloquium of the National Bank of Belgium
“Endogenous Financial Risk”
October 11/12th, 2012
Brussels
The usual disclaimer is in effect: These are our views and do not necessarily reflect those of the European
Central Bank, the Eurosystem, the Bank of Spain and/or the National Bank of Belgium.
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Macroprudential policy and credit cycles
• Macroprudential policy aims at reducing the potentially strong
negative externalities from the financial to the real sector
• A key channel is “excessive” bank pro-cyclicality / credit cycles
due to financial frictions in:
– Banks (credit supply): Holmström and Tirole (QJE, 1997), Allen and Gale
(2000 and 2007), Diamond and Rajan (JPE 2001 and AER 2006), Adrian
and Shin (AER, 2010), Shleifer and Visnhy (JFE & AER, 2010), Kindleberger
(1978), Tirole (2011), Gersbach and Rochet (2011), …
– Non-financial sector (credit demand): Bernanke and Gertler (AER, 1989),
Kiyotaki and Moore (JPE, 1997), Lorenzoni (RES, 2008), Jeanne and
Korinek (2011), …
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
3
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Credit supply cycles
• “Excessive” bank pro-cyclicality /credit supply cycles due to
bank frictions
– In good times:
• Problem: seeds for the next crisis via too high credit supply
– In bad times:
• Problem: credit crunch by banks with low capital buffers
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
4
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
One solution: countercyclical bank capital buffers?
• Higher bank capital standards in good times (and lower standards
in bad times) can be beneficial both in good and bad times by
reducing “excessive” bank pro-cyclicality in credit supply
– In good times:
• Problem: seeds for next crisis via too high bank credit supply
• Solution: banks should hold more capital (“skin in the game”) to
internalize potential loan costs/externalities
– In bad times:
• Problem: credit crunch by banks with low capital buffers
• Solution: higher bank capital buffers built in good times to
support credit in bad times (without government help)
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
5
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Policy: the rationale of Basel III on
higher capital and countercyclical buffers
•
“The new [capital] standards will markedly reduce banks’
incentive to take excessive risks… lower the likelihood and
severity of future crises, and enable banks to withstand without extraordinary government support - stresses of a
magnitude associated with the recent financial crisis.”
G-20 Seoul Official statement, November 2010
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
6
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Question
• What are the effects of countercyclical bank capital buffers on
credit supply?
– More generally: Bank capital impact on credit supply?
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
7
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Theory
• The two complementary rationales of bank capital (better
incentives and buffers in crisis) highlighted by policymakers
are also present in theoretical models:
– e.g. Holmström and Tirole, QJE 1997; Morrison and White, AER 2005;
Diamond and Rajan, JF 2000-JPE 01-AER 06
• And even the countercyclical buffer, e.g. in models:
– with agency problems (e.g. Tirole, 2011; Gersbach and Rochet, 2011)
– without agency problems but with investor sentiment (e.g. Shleifer
and Vishny, JFE 2010 and AER 2010)
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
8
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Empirical identification
• To identify the effects of countercyclical bank capital buffers
on credit supply (in good and bad times) is needed both:
1. Shocks to countercyclical capital buffers
2. Comprehensive loan-level data
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
9
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Experimental setting: Spain and dynamic provisioning
• Spain 1999-2010 offers an almost ideal setting for identification:
1.
Policy experiments with dynamic provisioning exogenously changed banks’
retained profits in good times to be used during crisis times

Exploit policy shocks in good times: contractionnary introduction in mid 2000
and expansionary change in mid 2005

2.
Exploit provision buffers and a policy shock during the recent crisis
Comprehensive credit register (matched with bank and firm
characteristics) to identify credit availability

Difference-in-differences (banks more/less affected by shocks and before/after
shocks) controlling for time-varying observed and unobserved firm
heterogeneity with firm*time fixed effects
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
10
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Preview of the results
• Countercyclical capital buffers mitigate credit supply cycles
– They contracted credit availability (volume and cost) during
good times (4%), but expanded it during the recent crisis (6%)
– While bank-level effects are always economically strong, firms
are even more affected during crisis times when switching banks
is difficult (1.5% in good vs. 5% in crisis times)
– In the first full draft of the paper we want to further exploit firm
and bank heterogeneity
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
11
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Our contributions
• We exploit policy shocks to bank capital (countercyclical
buffers) both in good and bad times to identify the impact of
bank capital on credit supply:
1. Unique (in the world) policy experiments on countercyclical
capital buffers taking place before Basel III  key contribution
2. In Jiménez, Ongena, Peydró and Saurina (AER, forthcoming)
we find that credit supply is pro-cyclical in GDP and monetary
conditions and stronger for banks with a lower capital ratio
• We used lagged bank capital. But bank capital is a key
strategic variable and  likely endogenous
• Our innovation: to exploit the policy shocks affecting bank
capital: causality from bank capital to the supply of credit
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
12
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Outline for the rest of the talk
• Policy shocks: dynamic provisioning experiments
– How does it currently work?
– Different policy shocks (2000 and 2005 and 2008) and the crisis shock
• Empirical strategy and data
– Empirical strategy
– Loan, firm and bank datasets
• Results
– 2000 policy shock, the 2005 one and the 2007-2009 crisis
– Loan- and firm-level results
• Conclusions
– Implications for Basel III, bank bailouts, monetary policy and, in general, for
macroprudential policy
– Incomplete draft: things that we are doing
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
13
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
The introduction of dynamic provisions
• In July 2000, the Banco de España (Spain’s central bank,
banking supervisor and responsible for bank accounting) put
in place dynamic (statistical/ general) provisions due to:
– Spain had the lowest ratio of loan loss provisions to total loans
among all OECD countries in 1999
– A period of sizeable credit growth and difficulty in immediately
recognizing problem loans following a credit expansion
(see Saurina et al (2000), Saurina (2009a) and Saurina (2009b) for
all the details on dynamic provisioning)
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
14
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Dynamic provisions: policy shocks and basic idea
• Introduced in mid-2000
– modified in mid 2005 (for consistency with IFRS)
– modified in 2008:Q4 (to allow banks to use more the provision
funds built in good times)
• Spanish LLP try to cover the increase in credit risk during lending
expansions
• Forward-looking (provision before any loss arrives)
• Countercyclical: Build up a buffer in good times to be used in bad times
• Tier-2 Capital
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
15
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Current provisions
• Specific provisions cover incurred losses already identified in a
specific loan
• General (dynamic) provisions cover incurred losses not yet
individually identified in a specific loan through a collective
assessment for impairment
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
16
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
A simple countercyclical mechanism
• In periods of expanding credit, a buffer of provisions is being built up to
cover the increase in credit risk (the incurred losses not yet materialized in
specific loans)
– We analyze the introduction in 2000 and a modification in 2005
• In periods when specific losses materialize in individual loans, the banks
can draw down from the previously built-up buffer of provisions
– We analyze the built-up buffers in the recent crisis
• The Spanish general provision also includes an upper and lower limit in
the amount of the general fund being built
– The lower limit was relaxed in the crisis and we also exploit this
• There is a simple formula governing the process
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
17
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Empirical identification
• Using a difference-in-difference approach, we compare bank
lending before and after the different shocks:
– policy shocks in good times: introduction in mid-2000 (and change in
2005) of the new regulation
– crisis shock: provision funds at the start of the financial crisis in August
2007 and policy change of the lower floor of provision funds in 2008:Q4
• We differentiate across banks with varying susceptibility to the
shocks and employ firm*time fixed effects to control for timevarying observed and unobserved firm heterogeneity (see
Khwaja and Mian (AER, 2008) and Jiménez, Ongena, Peydró and Saurina
(AER, forthcoming))
– control also for other key bank characteristics
– we analyze all margins of lending
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
18
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Bank’s susceptibility to the shocks: Buffers
• For the policy shocks in good times:
– new formula applied to the existing loan portfolio for each bank
yielding a bank-specific amount of new funds to be provisioned
(over total assets)
– the 2005 policy shock changed the initial weights on different
loans
• For the crisis shock:
– how much each bank had built up as dynamic (general)
provisions just prior to the onset of the crisis (2006:IV) over
total assets
– policy change of lower floor of provision funds in 2008:Q4
affects more the banks with lowest provision funds in 2008:Q3
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
19
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Credit register
• Credit register from Spain matched with bank and firm relevant
information (2007:Q2-> 100,000 firms, 175 banks, 600,000 loans)
• Exhaustive loan (bank-firm) level data on all outstanding business
loan contracts at a quarterly frequency
• We calculate the total exposures by each bank to each firm in each
quarter from 1999:III to 2009:IV
– The sample period includes one year before the initial shock (to run placebo
tests) and we analyze 2.5 years of data on the crisis
• We analyze changes in (log) credit volume (commitment or drawn),
maturity, collateral and the cost of lending (proxied by the
percentage of drawing down to total committed loans)
– Intensive and extensive margin of lending
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
20
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Benchmark loan- and firm- level equations and hypotheses
• Shocks:
DCreditb,i,t = ai,t + btbDBuffersb +Controlsb,i,t-1 + eb,i,t
– Good times: policy shocks of mid 2000 and mid 2005
– Bad times: crisis (from 2007:Q3) and policy shock in 2008:Q4
• LHS is change (after-before shock) in credit: log credit volume (commitment or
drawn), short-term loans, collateralized loans and drawn to committed loans
•
Buffers is our main variable on dynamic provisions (def. on previous pages)
– Firm-time fixed effects to control also for time-varying unobserved heterogeneity
– Bank controls are bank size, capital, NPL, ROA, liquidity, real estate exposure and bank
type (commercial, saving and coop banks) (and bank fixed effects on level of credit)
•
Hypotheses under reduction of credit supply cycles due to capital buffers:
β<0 in the 2000 and 2005 policy shock & β>0 in the crisis shock
•
Estimate similar firm-level regression to check credit substitution & real effects:
DCrediti,t = at + btf FirmDBuffersi +Controlsi,t-1 + ei,t
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
21
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Policy shock of July-2000, loan level data &
difference in log credit volume
Similar results for extensive margin and for credit cost and maturity
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
22
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Policy shock of July-2000, loan level data &
time-varying coefficients of buffers on credit volume
Local Channel
ΔLog Commitment
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1.2
98Q4
99Q1
99Q2
99Q3
99Q4
00Q1
00Q2
00Q3
00Q4
01Q1
01Q2
01Q3
01Q4
02Q1
02Q2
02Q3
02Q4
03Q1
03Q2
03Q3
03Q4
-1
Also for credit drawn, extensive margin and cost and maturity with similar results
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
23
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Policy shock of July-2000, both loan & firm level data
time-varying coefficients of buffers on credit volume
Local Channel
Aggregate Channel
ΔLog Commitment
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1.2
98Q4
99Q1
99Q2
99Q3
99Q4
00Q1
00Q2
00Q3
00Q4
01Q1
01Q2
01Q3
01Q4
02Q1
02Q2
02Q3
02Q4
03Q1
03Q2
03Q3
03Q4
-1
Also for credit drawn, cost and maturity with similar results
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
24
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Policy shock of July-2000, loan level data &
time-varying coefficients of buffers on credit cost
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
25
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Economic effects and summary
• An increase of one standard deviation in buffers in 2000:II
reduces the committed volume of credit:
– at the bank (loan) level: by 4 percent
– at the firm level: by 1.5 percent
• Similar results for credit drawn, cost and maturity and for
extensive and intensive margin and for the 2005 shock (in this
case lower elasticities, probably due to an expansionary
shock)
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
26
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Summary of economic effects during crisis
• An increase of one standard deviation in buffers in 2006:IV
increases the committed volume of credit:
– at the bank (loan) level: by almost 7 percent
– at the firm level: by at most 5 percent
• An increase of one standard deviation in buffers in 2000:II
reduces the committed volume of credit:
– at the bank (loan) level: by 4 percent
– at the firm level: by 1.5 percent
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
27
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Conclusions and policy implications
• Identify countercyclical bank capital buffers effects on credit supply
• Experimental setting: Spain 1999-2010
– Dynamic provisioning experiments and complete credit register
• Countercyclical capital buffers strongly mitigate credit supply cycles
• Firms are more affected during crisis times when switching from
banks with low to high capital buffers is difficult
• Important policy implications for:
– Basel III, bank bailouts, monetary policy and, in general, for
macro-prudential policy
– Individual bank capital (not only aggregate) matters in crises!
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
28
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Incomplete version: things we are doing
• Extensive margin of new lending missing
• Firm heterogeneity: change of credit supply could be stronger for
riskier firms, smaller, with less tangible assets, with less relationship
banking
• Bank heterogeneity: weaker banks should be probably more
affected, e.g. banks with lower profits, lower capital buffers,
smaller, non-listed, with higher NPLs
• Real effects: implications for employment, sales, profits, investment
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
29
Thank you
30
Numbers
•
•
•



The ratio of general provisions to total credit subject to the general provision at the end of
2007 for individual balance sheets was 1.22%. If we exclude exposures with a 0% weighting,
the coverage ratio reaches 1.59%. For non-consolidated data in Spain, the general provisions
were 78.9% of total provisions at the end of 2007.
in 1999 the loan-loss provisions of Spanish banks were the lowest among OECD countries. In
2006, the Spanish banking system had by far the highest coverage ratio among Western
European countries, at 255 percent
Counter-cyclical provisions were included in Tier 2 capital i.e. up to 1.25 percent of riskweighted assets
Total loan loss provisions at a consolidated level at the end of 2007 were 1.33% of total
consolidated assets
The ratio of bank capital and those total assets was 5.78%
At the end of 2007, Spanish banks at a consolidated level had 1.20% of general provisions
over total credit granted
31
Introduction – Policy shocks – Empirical Strategy – Results – Conclusions
Summary statistics of buffers
• In 2000:II:
– Buffers has an average of 0.46 and the standard deviation is 0.09
– Only correlated to real estate exposure, bank-type and
collateralized loans. Not correlated to other bank, firm and loan
characteristics
• In 2006:IV:
– Average of pre-crisis buffers is 1.1 and the standard deviation is
0.21
– Not correlated to firm and loan characteristics, but to some
bank characteristics (not to bank type)
Gabriel Jiménez, Steven Ongena, José-Luis Peydró and Jesús Saurina
32
Download