Upload
karlo
View
215
Download
2
Embed Size (px)
Citation preview
Do Bailouts Cause Moral Hazards or Franchise Valuein Banking?
Karlo Kauko*
I. INTRODUCTION
Moral hazard is said to be an inherent problem in banking. Profits are alwaysprivate, but during financial crises governments needs to socialise banks’ lossesbecause not doing so would jeopardise the flow of lending and paralyse theeconomy. Hence, the expected value of shareholders’ wealth is a monotonicallyincreasing function of risk taking. Boyd (1999) even argued that the resultingrisk appetite might spread to other industries because banks with distortedincentives favour risk bearing customers. There may be a time inconsistencyproblem of banking crisis management; it may be cheaper for the government tosave banks than to suffer the consequences of a full-blown banking crisis, but thissolution encourages risk taking in the future (Gale and Vives 2002). Capitalrequirements imposed on banks are useful partly because they reduce incentivesto gamble by increasing shareholders’ stakes (Rochet 1992).
Cordella and Levy Yeyati (2003), instead, proposed in their purely theoreticalcontribution, that a franchise value effect may dominate in a continuous timesetting. If there is a crisis, the central bank may save banks by providing emergencyliquidity assistance. Possibilities to be saved in future crises increase banks’ fran-chise values, i.e. the discounted value of future profits. Rational shareholders donot want to take excessive short term risks that might cause the bank to fail underrelatively normal circumstances if the bank has got good possibilities to makeprofits during a very long future lifespan, which is likely if survival in crises ismade possible by a lenient lender of last resort. If there is the probability of thebank failing in the next recession, not much is lost if poor gambling luck results inimmediate bankruptcy. Cordella and Levy Yeyati assumed that the governmentalbody is the central bank offering emergency liquidity, but other interpretations ofthe model are possible. In the model of Keeley (1990), imperfect competitionboosts banks’ franchise values and reduces risk appetite; the analogies are clear.
* Dr.Sc., adjunct professor, Bank of Finland and Aalto University; Bank of Finland, PO BOX 160, 00101Helsinki, Finland; [email protected]. The views expressed are those of the author and do not neces-sarily reflect those of the Bank of Finland or the Eurosystem.
KYKLOS, Vol. 67 – February 2014 – No. 1, 82–92
82 © 2014 John Wiley & Sons Ltd.
Both the mainstream view of safety nets as a source of moral hazard and thefranchise value hypothesis are theoretical constructions. Their relevance to thereal world is an empirical issue. There are relatively few econometric contribu-tions that would explicitly analyse the impact of safety nets on banks’ risk taking.Dam and Koetter (2012) used a large panel data on German banks in 1995–2006.Most banks of the sample were fairly small. The authors estimated how differentbank and location specific variables affected the probability that a distressedbank would be rescued by the government, and how the estimated probabilitiesto be bailed out in case of emergency affected risk taking. These probabilities hadan economically and statistically significant impact on risk taking, especiallyamong cooperative banks; safety nets were found to encourage risk taking. Thesefindings are not inconsistent with the theoretical results of Cordella and LevyYeyati because there was no financial crisis in Germany during the sampleperiod, whereas Cordella and Levy Yeyati focused on the behavioural conse-quences of bailing out banks during large scale macrofinancial jitters. Groppet al. (2010) found further evidence in favour of the moral hazard hypothesiswith data on German savings banks; when these credit institutions lost theirgovernment guarantees, they reduced risk taking. Gropp et al. (2011) studiedthousands of banks based in OECD countries. If Fitch/IBCA assessed that thebank has got good possibilities to be bailed out, this increased risk taking of otherbanks in the same country. According to the authors, this effect is probably dueto competition; public guarantees offered to one bank lower the franchise valueof competing banks. The bailout probability had hardly any impact on risk takingof the protected bank itself. Moreover, several contributions provide some resultson one closely related topic, namely deposit insurance systems and their impacton risk taking (Bartholdy et al. 2003, Demirgüç-Kunt and Detragiache 2002,Hoggarth et al. 2005).
There is a small yet growing field of empirical literature analysing cross-national differences in the severity of the recent financial crisis. (See Aizenmanand Pasricha 2012, Kamin and DeMarco 2012, Kauko 2012, Rose and Spiegel2011, Artha and de Haan 2011, Peev and Mueller 2012) In this paper, this newresearch approach is applied to the consequences of crisis management policies.A number of countries had been hit by national or regional financial crises beforethe 2008 turbulence. Crisis management policies had differed, and countries mayhave built different reputations as crisis managers. How did these reputationsaffect banks’ incentives to incur risks? In the following sections, the main focusis on the consequences of blanket guarantees offered by governments. Theblanket guarantee is comparable to emergency liquidity by the central bank,which is the crisis management tool assumed by Cordella and Levy Yeyatibecause even shareholders are rescued in both arrangements. Instead, by defini-tion, shareholders lose the franchise value if banks are nationalised. Old share-holders may or may not be saved in recapitalisations.
DO BAILOUTS CAUSE MORAL HAZARDS OR FRANCHISE VALUE IN BANKING?
© 2014 John Wiley & Sons Ltd. 83
As will be seen in following sections, the mainstream view of bailouts as asource of pernicious moral hazard is strongly rejected. Instead, governmentalblanket guarantees in past crises lowered the relative share of non-performingloans in 2009, indicating that having being saved in the past probably encouragedrisk aversion during the pre-2007 build-up phase. The effect is robust to theinclusion of a number of control variables. This finding is consistent with thefranchise value hypothesis.
Cordella and Levy Yeyati did not derive substantial results on the impact ofvarying degrees of depositor protection, but if depositors have incurred losses inprevious crises, the recent crisis was worse in this sample. Instead, nationalisa-tions and recapitalisations in past crises have no detectable impact, possiblybecause of insufficient variation of these variables.
The 2nd section presents the data. The 3rd section presents the results. The 4th
section concludes.
II. DATA
The main sources of data are the World Bank database and Laeven and Valencia(2008), hereafter L&V. The data consists of two sub-samples. First, all the 29countries with a crisis in 1981 – 2003 described in detail in Table 4 of L&V areincluded.1 In a couple of cases listed by L&V, a crisis began in 2007, but thesecases are now omitted. In addition to these crisis ridden countries, there is areference group of 20 countries;2 this group includes each country for which thedata is available, unless the country is mentioned by L&V in the long yet lessdetailed list of crises. On average, reference group countries are more advancedthan crisis-ridden countries.
Pre-2007 risk taking is proxied by LN(NON-PERFORMING 09), definedas Ln(100*non-performing loans/total loans) in 2009. A loan is classified asnon-performing when interest payments and repayments are late but the loan isnot yet completely written off. Willingness to fund even high-risk customersmust be highly dependent on banks’ pre-crisis risk taking incentives, and thedegree of pre-crisis risk taking certainly correlates with the deterioration ofthe quality of the loan portfolio during the crisis. This variable is not particu-larly skewed. (See Table 1) Because the concept of non-performing loans(NPLs) is not internationally harmonised, the control variable LN(NON-PERFORMING 06) is included. It probably captures most effects due to
1. Argentina, Bolivia, Bulgaria, Chile, Colombia, Croatia, Czech R, Dominican R, Ecuador, Estonia,Finland, Ghana, Indonesia, Japan, Korea, Latvia, Lithuania, Malaysia, Mexico, Norway, Paraguay,Philippines, Russia, Sweden, Thailand, Turkey, Ukraine, Uruguay, Venezuela.
2. Australia, Austria, Belgium, Canada, Denmark, France, Gabon, Germany, Greece, Guatemala, Ireland, Italy,Malta, Namibia, Pakistan, Portugal, Singapore, Switzerland, United Arab Emirates, United Kingdom.
KARLO KAUKO
84 © 2014 John Wiley & Sons Ltd.
Tabl
e1
Des
crip
tive
Stat
istic
son
Var
iabl
es
WO
LE
SAM
PLE
,N=
49C
RIS
ISR
IDD
EN
CO
UN
TR
IES,
N=
29R
EFE
RE
NC
EG
RO
UP
CO
UN
TR
IES,
N=
20
MIN
IMU
MM
EA
NM
AX
IMU
MST DE
VM
INIM
UM
ME
AN
MA
XIM
UM
ST DE
VM
INIM
UM
ME
AN
MA
XIM
UM
ST DE
V
LN
(NO
N-P
ER
FOR
MIN
G09
)−0
.92
1.34
2.96
0.79
−0.5
11.
412.
960.
81−0
.92
1.24
2.53
0.76
LN
(NO
N-P
ER
FOR
MIN
G06
)−1
.61
0.76
2.37
1.06
−1.6
10.
802.
161.
08−1
.20
0.70
2.37
1.05
BL
AN
KE
TG
UA
RA
NT
EE
00.
181
0.39
00.
311
0.47
00.
000
0.00
DE
POSI
TO
RL
OSS
ES
00.
221
0.42
00.
381
0.49
00.
000
0.00
NA
TIO
NA
LIS
AT
ION
00.
391
0.49
00.
661
0.48
00.
000
0.00
RE
CA
PITA
LIS
AT
ION
00.
431
0.50
00.
721
0.45
00.
000
0.00
GD
PC
API
TA19
800
564
942
903
746
20
283
015
891
439
629
49
738
4290
39
080
EN
GL
ISH
LA
W0
0.20
10.
410
0.10
10.
310
0.35
10.
49G
ER
MA
NL
AW
00.
181
0.39
00.
211
0.41
00.
151
0.37
SCA
ND
INA
VIA
N0
0.08
10.
280
0.10
10.
310
0.05
10.
22L
AT
INA
ME
RIC
AN
00.
221
0.42
00.
341
0.48
00.
051
0.22
EX
CO
MM
UN
IST
00.
141
0.35
00.
241
0.44
00.
000
0.00
RU
LE
OF
LA
W−1
.38
0.47
1.99
1.07
−1.3
80.
101.
980.
97−1
.12
1.01
1.99
0.99
NO
PRE
VC
RIS
IS0
0.41
10.
500
0.00
00.
001
1.00
10.
00O
UT
PUT
LO
SS0
19.0
998
.228
.03
032
.26
98.2
30.1
20
0.00
00.
00N
OIN
FOL
OSS
00.
161
0.39
00.
281
0.45
00
00.
00C
UR
RA
CC
OU
NT
2006
−22.
70.
525
.99.
2−2
2.7
0.5
16.4
9.5
−11.
30.
625
.99.
0C
RE
DIT
TO
GD
P20
06−1
3.4
75.0
219.
655
.3−1
3.4
48.7
122.
231
.59.
311
3.1
219.
660
.7E
CO
NG
RO
WT
H01
-06
−0.6
3.9
8.9
2.1
−0.6
4.5
8.9
2.0
0.9
2.9
6.2
1.7
CR
ED
ITG
RO
WT
H00
-06
01.
293.
850.
670
1.34
3.85
0.84
0.74
465
61.
211.
93G
DPC
API
TA20
0678
919
465
7296
018
711
921
1207
172
960
1638
378
930
186
5414
116
970
DO BAILOUTS CAUSE MORAL HAZARDS OR FRANCHISE VALUE IN BANKING?
© 2014 John Wiley & Sons Ltd. 85
national differences in banks’ accounting systems. These data are from theWorld Bank database.
The dummy variable BLANKETGUARANTEE equals +1 if the governmentoffered a blanket guarantee in the last pre-2008 crisis (10 cases), zero otherwise.The dummy variable DEPOSITORLOSSES equals +1 if depositors experiencedlosses in the previous crisis (11 cases). NATIONALISATION equals +1 ifat least some banks were nationalised (21 cases). RECAPITALISATIONequals +1 if banks were recapitalised (23 cases). All these variables are fromL&V.
Moreover, data on the legal system, the degree of economic development andthe political background are used as control variables. These long-lasting struc-tural characteristics may have affected both past crisis management policies andthe severity of the recent crisis, causing endogeneity problems. The severity ofthe previous crisis is measured by the resulting loss of output, but unfortunatelythis information is available for a very limited number of countries. The follow-ing structural control variables are tested.
• LATIN AMERICAN = Dummy for Latin American countries, 1 if located inthe area, zero otherwise
• SCANDINAVIAN = Dummy for Nordic countries (Norway, Sweden,Finland, Denmark)
• RULE OF LAW; World Bank database; higher values denote better rule oflaw
• EX COMMUNIST = Dummy for countries with a centrally plannedeconomy before 1990 (7 countries)
• GDP CAPITA 1980 = the 1980 value of GDP per capita (WB database); the1980 value is chosen to make sure that no crisis management policydescribed by L&V has affected it. This variable is 0 if EX COMMUNIST=1.
• NO PREV CRISIS equals +1 if the country belongs to the referencegroup
• ENGLISHLAW = Dummy for legal systems of British origin (Classificationby La Porta & al 2007, 10 countries)
• GERMANLAW = Dummy for legal systems of German origin (La Porta &al 2007, 9 countries)French and Scandinavian legal systems are used as the reference case.
• OUTPUTLOSS = the estimated loss of output in percent of GDP caused bythe previous crisis (Laeven & Valencia 2012, Table A1), zero if the infor-mation is missing or the country belongs to the reference group.
• NOINFOLOSS equals 1 if no output loss is reported (8 cases).
It may not be meaningful to include pre-crisis macroeconomic data becausesuch variables are highly endogenous. If banks’ reckless lending causes risks
KARLO KAUKO
86 © 2014 John Wiley & Sons Ltd.
by creating property price bubbles and excessive indebtedness, it would bepointless to control for real estate prices and loan stocks. However, a fewregressions are run with the following macroeconomic variables from the WorldBank database.
• CURR ACCOUNT 2006 = The current account as a percentage of the GDPin 2006. Positive numbers denote surpluses and negative deficits. Previousresearch has found that a current account deficit made the recent crisis worse(See Aizenman and Pasricha 2012, Kauko 2012).
• GDP CAPITA 2006 = GDP per capita in 2006 in USD. Rose and Spiegel(2011), Aizenman and Pasricha (2012) and Kauko (2012) found that low-income countries were only moderately affected by the Subprime-Lehmancrisis.
• CREDIT TO GDP 2006 = Bank lending divided by the GDP in 2006 as apercentage. According to Acosta-Gonzalez et al. (2012), the credit to GDPratio is a good predictor of cross-national differences in the severity of therecent crisis.
• ECON GROWTH 01-06 = Average growth rate of real GDP in percent in2001–2006. Rapid growth before the outbreak of the crisis may have been asign of over-heating. If one focuses on the most advanced economies of eachera, an international crisis is probably very severe in countries where eco-nomic growth was faster than average two to four years before the crisis(Jordà et al., 2011).
• CREDITGROWTH 00-06 = (Credit-to-GDP ratio in 2006)/(Credit-to-GDPratio in 2000). Jordà et al. (2011) observed that over the period 1870–2008,in both national and international financial crises, the credit-to-GDP ratioused to grow fast four years before the outbreak of a crisis.
Descriptive statistics on the variables are presented in Table 1.
III. RESULTS
When the relative amount of NPLs is regressed on BLANKETGUARANTEEand different control variables, BLANKETGUARANTEE, obtains a negativeand statistically significant coefficient in each equation, consistently with thefranchise value hypothesis. (See Table 2) The inclusion or exclusion of data onother crisis management tools and country specific institutional or structuralcharacteristics does not affect this main conclusion. Financial stability in 2009probably indicates that lending was risk averse before the global turbulence.
RECAPITALISATION and NATIONALISATION probably reflect protectionof wholesale creditors, but these variables have no detectable impact, possibly
DO BAILOUTS CAUSE MORAL HAZARDS OR FRANCHISE VALUE IN BANKING?
© 2014 John Wiley & Sons Ltd. 87
Table 2
Determinants of Non-Performing Loans in 2009
Cross-sectional OLS, institutional and structural control variables only
Cross national samples, explained variable = Ln(Non-performing loans/total loans in 2009)Higher values of the explained variable indicate higher levels of distress.
1 2 3 5 6 7N = 29 N = 29 N = 29 N = 49 N = 49 N = 49
Constant 1.500 1.070 1.438 1.317 1.320 1.048(6.2)*** (2.7)*** (4.5)*** (3.7)*** (4.7)*** (2.5)**
LN(NON-PERFORMING06)
0.210 0.271 0.014 0.283 0.284 0.345(1.5) (−2.0)* (0.1) (2.3)** (2.4)** (2.6)**
BLANKETGUARANTEE −0.885 −0.782 −0.551 −0.497 −0.500 −0.524(−3.8)*** (−3.5)*** (−2.4)** (−2.1)** (−2.2)** (−2.4)**
DEPOSITORLOSSES 0.598 0.426(2.2)** (1.9)*
NATIONALISATION 0.172 0.124(0.5) (0.5)
RECAPITALISATION 0.009 0.047(0.0) (0.2)
GDP CAPITA 1980 2.43E-05 −1.42E-05 −1.42E-05 −1.63E-05(0.5) (−1.4) (−1.4) (−1.5)
ENGLISHLAW 0.828 0.174 0.174 0.149(2.3)** (0.7) (0.8) (0.6)
GERMANLAW −0.123 −0.312 −0.312 −0.232(−0.3) (−1.1) (−1.1) (−0.8)
SCANDINAVIAN −0.844 −0.031 −0.030 0.083(−1.2) (−0.1) (−0.1) (0.2)
LATIN AMERICAN −0.404 −0.466 −0.460 −0.515(−1.6) (−1.8)* (−1.8)* (−2.0)*
EX COMMUNIST 0.947 0.978 1.043 0.827(2.6)** (2.3)** (2.8)*** (2.3)**
RULE OF LAW −0.214 −0.108 −0.107 −0.066(−1.2) (−0.8) (−0.8) (−0.5)
NO PREV CRISIS 0.021 −0.026 0.170(−0.1) (−0.1) (0.5)
OUTPUTLOSS 0.000(0.0)
NOINFOLOSS 0.068(0.2)
Adj R squared 0.287 0.342 0.646 0.516 0.542 0.544F-test 6.64*** 3.91** 6.67*** 5.27*** 6.67*** 5.40***Jarque-Bera of residuals 0.527 0.213 0.985 0.390 0.385 0.779
*** = 1 % signific, ** = 5 % signific, * = 10 % significance t values in parenthesesWhite heteroskedasticity-consistent standard errors & covariance 29 countries with a financialcrisis in 1981–2003 in eqs 1–3; 20 reference countries (with no recent crisis history) included ineqs 4–6OUTPUTLOSS, BLANKETGUARANTEE, NATIONALISATION, RECAPITALISATION andDEPOSITORLOSSES refer to government policies in previous national and regional crises
KARLO KAUKO
88 © 2014 John Wiley & Sons Ltd.
because of insufficient variation of these factors. Depositors’ losses in past criseshave made the situation worse, inconsistently with the moral hazard hypothesis;retail financiers who had suffered losses in previous crises did not force banks tokeep risks under control by withdrawing their savings from banks with impru-dent risk management. Instead, the franchise value hypothesis may explain thisfinding; helping banks to collect cheap deposit funding by always protectingdepositors may have increased banks’ franchise values and discouraged risktaking.
Most control variables have no detectable effect. However, countries that hadgot a centrally planned economy before 1990 were particularly vulnerable. LatinAmerican countries may have been less and countries with a British legal systemmore affected than others, but the evidence is weak (equations 3 – 6 in Table 2).There is absolutely no evidence that the severity of the previous crisis wouldsomehow affect the severity of the 2009 crisis (eq. 5).
Controlling for the pre-crisis macrofinancial environment weakens the impactof BLANKETGUARANTEE. The regression coefficient falls substantially butthe variable remains statistically significant (See Table 3). Consistently withprevious findings, rapid loan growth during pre-crisis years and, above all, acurrent account deficit made the crisis worse. Interestingly, the current account issignificant only if reference group countries are included (eq. 4). Banks’ recklessrisk taking may have boosted aggregate demand because fewer customers havebeen financially constrained, resulting in more net imports. Nevertheless, prelimi-nary attempts to explain credit growth and current account balances with dummyvariables on past crisis management policies proved largely unsuccessful.
IV. DISCUSSION
According to Cordella and Levy Yeyati (2003), emergency finance offered duringfinancial crises increases the franchise value of banks and may discourage risktaking. Consistently with this argumentation, countries where blanket guaranteeshad been offered in past banking crises were relatively moderately affected by therecent global crisis. This finding is in strict contradiction with the popular moralhazard hypothesis. Moreover, countries where depositors had incurred losses inprevious crises were particularly seriously affected. Both findings indicate thatdifferences in the amount of moral hazard did not drive cross-national differencesin the severity of the recent crisis. The franchise value approach offers a muchbetter explanation to observed regularities. Instead, there is no evidence thatsaving bondholders by recapitalising or nationalising banks would have affectedfinancial stability during the worst phases of the recent crisis.
A central problem of this empirical analysis is the small size of the sample.There are only 29 countries for which detailed data on past crisis management
DO BAILOUTS CAUSE MORAL HAZARDS OR FRANCHISE VALUE IN BANKING?
© 2014 John Wiley & Sons Ltd. 89
was readily available. Nevertheless, the results are statistically significant androbust to the inclusion of numerous control variables.
These findings have clear policy implications. Instead of promising harshpolicies in a possible crisis, which is often an unconvincing commitment
Table 3
Determinants of Non-Performing Loans in 2009
Cross-sectional OLS, institutional, structural and pre-crisis macrofinancial control variables
Cross national samples, explained variable = Ln(Non-performing loans/total loans in 2009) Highervalues of the explained variable indicate higher levels of distress.
1 2 3 4N = 29 N = 29 N = 29 N = 49
Constant −0.431 −0.098 −0.215 0.162(−1.4) (−0.2) (−0.4) (0.3)
LN(NON-PERFORMING 06) 0.406 0.398 0.423 0.490(4.8)*** (4.1)*** (4.1)*** (4.9)***
BLANKETGUARANTEE −0.324 −0.327 −0.357 −0.382(−2.2)** (−1.8)* (−2.1)* (−2.3)**
DEPOSITORLOSSES 0.260 0.294 0.562(1.2) (1.3) (3.0)***
NATIONALISATION 0.120 0.118(0.8) (0.8)
RECAPITALISATION 0.095 0.138(0.6) (0.8)
ENGLISHLAW 0.471 4.88E-01 0.357(1.3) (1.2) (1.5)
LATIN AMERICAN 0.001 0.016 −0.208(0.0) (0.1) (−0.9)
EX COMMUNIST 0.375 0.393 0.353(2.1)* (1.8)* (1.5)
GDP CAPITA 2006 6.1E-06 6.5E-06 6.4E-06(1.0) (1.0) (0.7)
CURR ACCOUNT 2006 −0.012 −0.017 −0.019 −0.035(−1.0) (−1.4) (−1.4) (−3.5)***
CREDIT TO GDP 2006 0.000 −0.002 −0.002 −0.001(−0.2) (−0.6) (−0.7) (−0.4)
ECONGROWTH 01-06 0.274 0.155 0.141 0.049(4.9)*** (1.9)* (1.9)* (0.8)
CREDITGROWTH 00-06 0.213 0.234 −0.002 0.189(1.3) (1.6) (1.5) (1.2)
NO PREV CRISIS 0.164(0.8)
Adj R squared 0.803 0.826 0.816 0.741F-statistic 17.26 13.12 10.57 10.83Jarque-Bera of residuals 5.054* 0.293 2.307 1.122
*** = 1 % signific, ** = 5 % signific, * = 10 % significance; t values in parenthesesWhite heteroskedasticity-consistent standard errors & covariance 29 countries with a financial crisisin 1981–2003 in eqs 1–3; 20 reference countries (with no recent crisis history) included in eq4.OUTPUTLOSS, BLANKETGUARANTEE, NATIONALISATION, RECAPITALISATION andDEPOSITORLOSSES refer to government policies in previous national and regional crises
KARLO KAUKO
90 © 2014 John Wiley & Sons Ltd.
anyway, it may not be harmful to explicitly promise to guarantee the stability ofthe banking system in such an extreme situation, especially if the promisedmeasures enable old shareholders to have a claim on profits made during futurebenign eras. Not only would harsh policies cause welfare losses during the crisis.They might even encourage risk taking in the future, contrary to cliché argu-ments on moral hazard. Moreover, compromising depositors’ protection seemsnon-advisable. These empirical findings and their policy implications are prob-ably not applicable to hedge funds, mutual funds, life insurance companies orbond issuers. Moreover, the findings tell nothing about the consequence ofbailing out banks during relatively normal times.
The moral hazard argumentation has become so well established that fewpeople are willing to challenge it. The point made in by Cordella and LevyYeyati is seldom if ever referred to in policy discussions. The whole ideamay be unpopular because of political and even emotional reasons. The Rightwants to believe in non-regulated market forces whereas the Left has got asuspicious attitude towards the financial industry, let alone towards wealthtransfers to banks’ shareholders. The idea of subsidising banks and socialisingrisks is equally unpleasant for both. Nevertheless, not suiting any mainstreampolitical ideologies should not weaken the credibility of a theory in academiceconomics.
Instead, other measures than pre-commitment to harsh policies might beefficient in discouraging excessive risk taking. For instance, as proposed byRochet (1992), higher capital adequacy requirements during benign times wouldincrease the amount of crisis time losses borne by shareholders instead oftaxpayers without affecting the franchise value of future profits.
REFERENCES
Acosta-González, Eduardo, Fernando Fernández-Rodríguez and Simon Sosvilla-Rivero (2012). Onfactors explaining the 2008 financial crisis, Economics Letters. 115: 215–217.
Aizenman, Joshua and Gurnain Kaur Pasricha (2012). Determinants of financial stress and recoveryduring the great recession, International Journal of Finance and Economics. 17: 347–372.
Artha, I Kaded Dian Sustrina and Jacob de Haan (2011). Labor Market Flexibility and the Impact ofthe Financial Crisis, Kyklos. 64: 213–230.
Bartholdy, Jan, Glenn W Boyle and Roger D. Stover (2003). Deposit insurance and the risk premiumin bank deposit rates, Journal of Banking and Finance. 27: 699–717.
Boyd, John H. (1999). Expansion of commercial banking powers . . . or universal banking is the cart,not the horse, Journal of Banking and Finance. 23: 655–662.
Cordella, Tito and Eduardo Levy Yeyati (2003). Bank bailouts: moral hazard vs. value effect, Journalof Financial Intermediation. 12: 300–330.
Dam, Lammertjan and Michael Koetter (2012). Bank bailouts and moral hazard: Evidence fromGermany, The Review of Financial Studies. 25: 2343–2380.
Demirgüç-Kunt, Asli and Enrica Detragiache (2002). Does deposit insurance increase bankingsystem stability? An empirical investigation, Journal of Monetary Economics. 49: 1373–1406.
DO BAILOUTS CAUSE MORAL HAZARDS OR FRANCHISE VALUE IN BANKING?
© 2014 John Wiley & Sons Ltd. 91
Gale, Douglas and Xavier Vives (2002). Dollarization, bailouts and the stability of the financialsystem, Quarterly Journal of Economics. 117: 467–502.
Gropp, Reint, Christian Gruendl and André Guettler (2010). The impact of public guarantees on bankrisk taking: evidence from a natural experiment. European Central Bank Working Paper 1272.
Gropp, Reint, Hendrik Hakenes and Isabel Schnabel (2011). Competition, risk-shifting and publicbail-out policies, Review of Financial Studies. 24: 2084–2120.
Hoggarth, Glenn, Patricia Jackson and Erlend Nier (2005). Banking crises and the design of safetynets, Journal of Banking and Finance. 29: 143–159.
Jordà, Òscar, Moritz Schularick and Alan M. Taylor (2011). Financial crises, credit booms andexternal imbalances: 140 years of lessons, IMF Economic Review. 59: 340–378.
Kamin, Steven B. and Laurie Pounder DeMarco (2012). How did a domestic housing slump turn intoa global financial crisis? Journal of International Money and Finance. 31: 10–41.
Kauko, Karlo (2012). External deficits and non-performing loans in the recent financial crisis,Economics Letters. 115: 196–199.
Keeley, Michael C. (1990). Deposit Insurance, Risk and Market Power in Banking, AmericanEconomic Review. 80: 1183–2000.
Laeven, Luc., Fabian Valencia (2008). Systemic Banking Crises: A New Database. IMF WorkingPaper WP/08/224.
Laeven, Luc, Fabian Valencia (2012). Systemic Banking Crises: An update. IMF Working PaperWP/12/163.
Peev, Evgeni and Dennis C. Mueller (2012). Democracy, Economic Freedom and Growth in Tran-sition Economies, Kyklos. 65: 371–407.
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Schleifer (2007). The Economic Consequencesof Legal Origins (Forthcoming in the Journal of Economic Literature).
Rochet, Jean-Charles (1992). Capital requirements and the behavior of commercial banks, EuropeanEconomic Review. 36: 1137–1170.
Rose, Andrew K and Mark M Spiegel (2011). Cross-country causes and consequences of the crisis:an update, European Economic Review. 55: 309–324.
SUMMARY
Governmental safety nets are said to be a major source of moral hazard in banking. If profits are private butlosses are borne by taxpayers, banks have perverted incentives to take excessive risks. Cordella and LevyYeyati (2003) presented an opposing hypothesis. Safety nets offered during turbulent times help banks tosurvive and may increase banks’ franchise value. This induces stronger risk aversion. This unorthodoxhypothesis is tested with a cross-national comparison. National banking crises in 1981–2003 were handledby national authorities in different ways, and countries may have built reputations as tough or generous crisismanagers. The recent financial crisis was less severe in countries where the government had offered blanketguarantees in previous crises, which may be interpreted as a sign of stronger risk aversion during thepre-2007 build-up phase. Moreover, losses suffered by depositors in past crises made the recent crisis worse.These findings are consistent with the franchise value hypothesis but in strict contradiction with themainstream moral hazard argumentation. The finding is politically unpopular and based on relatively fewobservations, but it is statistically significant and robust to the inclusion of numerous control variables.
KARLO KAUKO
92 © 2014 John Wiley & Sons Ltd.