Bank Risk-Taking, Securitization, Supervision, and Low Interest Rates: Evidence from Lending Standards potx

Bank Risk-Taking, Securitization, Supervision, and Low Interest Rates:
Evidence from Lending Standards



Angela Maddaloni and José-Luis Peydró
*





July 2009



Abstract
We analyze the root causes of the current crisis (Allen, 2009; Diamond and Rajan,
2009) by studying the determinants of bank lending standards in the Euro Area. We
use the answers from the confidential Bank Lending Survey where national central
banks request banks quarterly information on their lending standards. We find robust
evidence that low short-term interest rates soften standards for both businesses and
households and – by exploiting cross-country variation of Taylor-rule implied rates –
we find that rates too low for too long soften standards even further. The softening is
over and above an improvement of the borrower’s collateral risk and outlook, thus
suggesting higher loan risk-taking by banks. In addition, we find that weaker banking
supervision standards and higher securitization activity amplify the softening of
lending standards due to low short-term rates, even when we instrument securitization.
Finally, low short-term rates have a stronger impact than long-term rates on the
softening of standards. These results help shed light on the origins of the current crisis

and have important policy implications.

*
The authors are at the European Central Bank, Kaiserstrasse 29, D 60311 Frankfurt am Main,
Germany. Contact information: , and / jose-
We thank Lieven Baert and Francesca Fabbri for excellent research
assistance. Any views expressed are only those of the authors and should not be attributed to the
European Central Bank or the Eurosystem.

2
“One (error) was that monetary policy around the world was too loose too long. And that created this
just huge boom in asset prices, money chasing risk. People trying to get a higher return. That was just
overwhelmingly powerful We all bear a responsibility for that” … “The supervisory system was just
way behind the curve. You had huge pockets of risk built up outside the regulatory framework and not
enough effort to try to contain that. But even in the core of the system, banks got to be too big and
overleveraged. Now again, here’s an important contrast. Banks in the United States, even with
investment banks now banks, bank assets are about one times GDP of the United States. In many other
mature countries - in Europe, for example – they’re a multiple of that. So again, around the world,
banks got to just be too big, took on too much risk relative to the size of their economies.”
Timothy Geithner, United States Secretary of the Treasury, Charlie Rose’s PBS, May 2009
“I believe the causes cannot be found in any one market, such as the US. Nor are they limited to a
particular business, such as subprime mortgages. These triggered the crisis, but they did not cause it.
The causes are the same as in any previous financial crisis: excesses and losing the plot in an
extraordinarily favourable environment. Indeed, some fundamental realities of banking were forgotten:
cycles exist; lending cannot grow indefinitely; liquidity is not always abundant and cheap; financial
innovation involves risk that cannot be ignored”
Emilio Botín, Chairman of Bank Santander, Financial Times, October 2008
I. Introduction
The current financial crisis has had a dramatic impact on the banking sector of
most developed countries, it has severely impaired the functioning of interbank

markets, and it has triggered an economic crisis in these same countries.
What are the root causes of this crisis? Several commentators and academics
have suggested that the global financial crisis was originated by an excessive
softening of lending standards. Three key elements were mentioned as drivers: too
low levels of interest rates, high securitisation activity, and weak bank regulation
supervision standards.
2
Therefore, the crisis that started in the subprime mortgage
market in the US may have been the manifestation of deep rooted problems, which
were not peculiar to one financial instrument and/or country but were present
globally, albeit to different degrees (Allen, 2009, and Diamond and Rajan, 2009).
3

Moreover, these root causes may have also been interrelated and mutually amplifying
in affecting the risk-taking of financial institutions (Rajan, 2005). In this paper, we
empirically test these hypotheses.

2
See for example Allen (2009), Brunnemeier (2009), Calomiris (2008), Taylor (2007 and 2008), Engel
(2009), and numerous articles in The Financial Times, The Wall Street Journal, and The Economist.
Nominal rates were the lowest in almost four decades and below Taylor rates in many countries while
real rates were negative (Taylor, 2008; and Ahrend, Cournède and Price, 2008).
3
Allen (2009), Rajan (2009) and Diamond and Rajan (2009) distinguish between proximate versus
root causes of the current crisis.

3
Low interest rates, weaker bank regulation supervision standards and high
securitization activity may imply higher loan risk-taking by banks. Moral hazard
problems are severe in the banking industry due to e.g. deposit insurance, potential

bail-outs and very high levels of leverage; hence, high levels of liquidity increase
incentives for bank risk-taking (Allen and Gale, 2007).
4
Without bank agency
problems, excess of liquidity would be given back to shareholders or central banks,
but – due to bank moral hazard problems – banks may over-lend the extra-liquidity in
bad projects. Allen (2009) and Allen and Gale (2007 and 2004) connect too high
levels of liquidity with too low levels of short-term interest rates.
5
In fact, the level of
overnight rates is a key driver of liquidity for banks since banks increase their balance
sheets (leverage) when financing conditions through short-term debt are more
favourable (Shin, 2009; Adrian and Shin, 2009; and Brunnermeier et al., 2009).
Therefore, low short-term interest rates increase bank risk-taking through this channel
and, hence, stronger banking supervision standards – via reducing bank agency
problems – should reduce the higher risk-taking associated to low rates.
6
Finally, low
levels of both short and long-term rates may induce a search for yield from financial
intermediaries due to their moral hazard problems. Securitization of loans result in
assets yielding attractive returns for investors, but the social cost may be lower
screening and monitoring of securitized loans. Hence, the impact of low rates on the
softening on standards is stronger with higher securitization activity (Rajan, 2005).


4
For the link between liquidity and loan risk-taking by banks, see Chuck Prince, Citigroup Chairman,
when describing why his bank continued financing leveraged buyouts despite mounting risks, said:
“When the music stops, in terms of liquidity, things will be complicated. But, as long as the music is
playing, you’ve got to get up and dance. We’re still dancing.” (Financial Times, July 9, 2007).

5
Due to bank agency problems, short-term rates soften lending standards either by abating adverse
selection problems in credit markets thereby increasing bank competition (Dell’Ariccia and Marques,
2007), or by reducing the threat of deposit withdrawals (Diamond and Rajan, 2006).
6
There are other channels by which low levels of interest rates affect bank (loan) risk-taking. First,
lower risk-less rates increase the attractiveness of risky assets in a mean-variance portfolio framework.
Second, in habit formation models agents become more risk-averse during economic slowdowns
because their consumption decreases relative to their status-quo (Campbell and Cochrane, 1999), thus a
tightening in monetary policy by depressing real activity may increase investors’ risk aversion. Third,
low rates may decrease banks’ intermediation margins (profits), thus reducing banks’ charter value,
increasing in turn incentives for risk-taking (Keeley, 1990). Fourth, there could also be monetary
illusion with low levels of short-term rates which would make banks to desire riskier products to
increase returns (Shiller, 1997; and Akerlof and Shiller 2009). Fifth, an environment in which central
banks focus only on consumer goods price stability may make monetary policy rates too low, fostering
in turn asset price and credit bubbles (Borio 2003; Borio and Lowe, 2002). For Acharya and
Richardson (2009) the fundamental cause of the crisis was the credit boom and the housing bubble.
These were largely developed by too low levels of monetary policy rates (Taylor, 2007).

4
We empirically analyze the following questions: Do low levels of both short and
long-term interest rates soften bank lending standards? Are the effects stronger with
higher securitization activity or weaker banking supervision standards? And, do
results differ across type of loans, i.e. across business, house purchase, and
households’ consumption loans?
There are four major challenges to identify the previous questions. First,
monetary policy rates are endogenous to the (local) economic conditions. Second,
banking supervision standards may be endogenous to monetary policy especially in
cases when the central bank is responsible for both. Third, securitization activity is
endogenous to monetary (bank liquidity) conditions, since these affect the ability of

banks to grant loans. Fourth, it is very difficult to obtain data on the pool of potential
borrowers approaching a bank, to know their quality, and then to know whether, why
and how banks change their lending standards to customers.
Our identification strategy relies upon the data we use, the Euro Area Bank
Lending Survey, which allows us to tackle the four previous identification
challenges.
7
First, we use data from the Euro Area countries, where monetary policy
rates are identical. However, there are significant cross-country differences in terms of
GDP growth and inflation, implying in turn significant exogenous cross-sectional
variation of Taylor-rule implied rates (Taylor, 2008). Second, banking supervision
standards in the Euro Area are responsibility of the national supervisory authorities
(often national central banks), whereas monetary policy is decided by the Governing
Council of the Eurosytem.
8
Third, there is significant cross-sectional variation in
securitization activity partly stemming from cross-country differences in the
regulation of the market for securitization. Fourth, we have access to the confidential
Bank Lending Survey database of the Eurosystem. National central banks request
quarterly information on the lending standards they apply to customers and on the
loan demand they receive. The rich information allows us to analyze whether banks

7 Banks are not only the key financial intermediaries that reduce the information problems which are
crucial for the real effects of monetary policy through credit markets (Bernanke and Gertler, 1995), but
banks are also the main providers of credit in most economies and, in particular, in the Euro Area (see
for example Hartmann, Maddaloni, Manganelli, 2003, and Allen, Chui and Maddaloni, 2004).
8 International guidelines like Basel are also very important for bank regulation, but there is rule for
discretion in supervision standards, in particular for banking capital see Laeven and Levine (2009) and
Barth, Caprio and Levine (2006).

5
change their lending standards, to whom (average or riskier borrowers), how (loan
spreads, size, collateral, maturity or covenants), and why (due to changes in borrower
risk, or in bank balance-sheet strength, or in bank competition).
The evidence we provide suggests that banks take on higher risk when overnight
rates are lower. This conclusion arises from the combination of the following robust
results: First, a softening of lending standards is associated to low overnight rates. The
association is highest for business loans.
9
Second, higher GDP growth also implies a
softening of standards, i.e. standards are pro-cyclical. Our findings are economically
relevant: taking into consideration the standard deviation of overnight rates and GDP
growth, the impact of a change in the overnight rate doubles a change in GDP growth
both for business and households’ consumption loans, but it is similar for loans for
house purchase. Third, by exploiting cross-country variation of Taylor-rule implied
rates, we find – especially for loans for house purchase – that a softening of lending
standards due to short-term rates too low for too long (measured as the number of
periods when short-term rates are lower than Taylor-rule implied rates). Fourth, low
overnight rates have a stronger impact than low long-term rates on the softening of
standards.
10
Fifth, all the lending standards are softened when short-term rates are low,
both for average and for riskier borrowers. The softening implies lower loan margins,
lower collateral requirements, longer loan maturity, less covenants and larger loan
size. Finally, and more importantly, there is a softening of standards even when
changes in standards are not associated to improvements in borrowers’ credit-

9
Jiménez, Ongena, Peydró and Saurina (2009) and Ioannidou, Ongena and Peydró (2009) are the first
to investigate the impact of short-term (monetary policy) rates on loan risk-taking by banks. They use

comprehensive credit registers for business loans from Spain and Bolivia respectively. They find that
low levels of overnight rates increase loan risk-taking by banks. Our results complement these papers
by analyzing not only business loans but also loans for house purchase and consumption, and also by
analyzing all Euro Area countries. In addition, we do not have the comprehensive credit register but we
have the potential pool of borrowers and we know whether, how and why banks change their lending
standards. For indirect evidence on short-term rates and risk-taking, see Bernanke and Kuttner (2005),
Rigobon and Sack (2004), Manganelli and Wolswijk (2007), Axelson, Jenkinson, Strömberg and
Weisbach (2007), Den Haan, Sumner, and Yamashiro (2007), and Calomiris and Pornrojnangkool
(2006).
10
Due to the saving glut and the existence of current account “imbalances”, savers were looking for
investment opportunities abroad. However, they were seeking short-term assets (Gross, 2009) and, in
fact, Brender and Pisani (2009) reports that about one third of all foreign exchange reserves are in the
form of bank deposits. Little is known about the maturity composition of the remainder, most of which
is invested in interest-bearing securities. The scarce available data on the composition of USD foreign
exchange reserves that can be gleaned from the US Treasury International Capital data suggests that
over half of foreign official holdings of US securities had a maturity of less than three years (see Gross,
2009).

6
worthiness, but due to improvements in bank balance-sheet constraints (higher bank
liquidity or capital, or better bank access to market finance) and also due to higher
banking competition, especially from non-banks and market finance. Hence, the
softening of standards is over and above an improvement of the borrower’s collateral
risk/ value and outlook, thus suggesting higher loan risk-taking by banks when short-
term rates are low.
11

Using time-varying measures of banking supervision standards for bank capital,
we find that the impact of low short-term rates on the softening of lending standards

over and above improvements in borrower’s credit-worthiness is larger when
supervision standards are weaker, both for loans for house purchase and for
consumption. However, the level effect of supervision standards is not as significant
as the direct effect of rates on lending standards. The lack of strong significance of the
direct effect of banking supervision on lending standards may be consistent with the
arguments made by Allen (2009) and Rajan (2009) concerning the need for “good”
supervision regulation, which does not necessarily mean more stringent supervision.
12

Finally, in contrast with short-term rates, we don’t find that the impact of low long-
term rates on the softening of standards depends on banking supervision standards.
Finally, we analyze securitization activity.
13
We find that the impact of low short-
term rates on the softening of standards for all type of loans is larger when
securitization activity is higher. However, if we consider the tightening of standards
only related to bank balance-sheet’s constraints, securitization activity in conjunction
with overnight rates has a significant impact only for loans for house purchase and
consumer credit. Moreover, we find significant effects for short-term rates but not for
long-term rates. Finally, instrumenting securitization activity by the regulation of the
market for securitization in each country does not significantly modify the results,
where the instrument has a t-stat higher than 8 in the first-stage regression and, hence,

11
That is, the effect of low rates on the softening of standards is over and above the firm balance sheet
channel of monetary policy (Bernanke and Gertler, 1995). Because of imperfect information,
incomplete contracts and imperfect bank competition, expansive monetary policy increases banks’ loan
supply (Bernanke, Gertler and Gilchrist, 1996; Bernanke, 2007; Kashyap and Stein, 2000; Diamond
and Rajan, 2006; Stiglitz, 2001; and Stiglitz and Greenwald, 2003).
12

See also Barth, Caprio and Levine (2006).
13
For evidence of excessive softening of lending standards due to securitization, see Keys et al. (2008),
Mian and Sufi (2009), and Dell'Ariccia, Laeven and Igan (2008).

7
it does not suffer from weak instrument concerns (Staiger and Stock, 1997). In
consequence, the previous set of results suggests higher loan risk-taking by banks
when securitization activity is high and short-term rates are low.
The analysis of conditions and terms of the loans suggest that when short-term
rates are low and securitization activity is high margins on loans to riskier firms are
not affected while margins on riskier households – either for house purchase or for
consumption – are softened. In addition, collateral requirements, covenants, maturity
and loan-to-value ratio restrictions are softened.
When analyzing at the factors by which banks change the standards, in a situation
with low short-term rates and high securitization activity, the results we find
highlight: (i) the importance of the “shadow banking system” in setting of the
softening of bank lending standards (both stemming from higher competition from
non-banks and from market finance, which both have a different level of regulation
and supervision than banks), (ii) the importance of bank balance-sheet liquidity in the
softening of standards, and (iii) the risk transfer effects linked to securitization in the
sense that the collateral risk and value may matter less for banks when granting loans
if banks can transfer these risks off.
All in all we find that low short-term rates soften lending standards. The
softening is over and above improvements in borrower’s collateral risk/ value and
outlook, and all the relevant standards are softened. Hence, our results suggest that
banks take on higher loan risk when overnight rates are low. In addition, the impact of
low short-term rates on the softening of standards is stronger when securitization is
high or banking supervision standards are weak, thus suggesting that low levels of
short-term rates may create excessive bank risk-taking (Allen and Gale, 2007; Rajan,

2005).
14
Finally, low short-term rates matter statistically and economically more than

14
From Allen and Gale (2007) for example, low short-term interest rates create high risk-taking by
banks because of moral hazard problems in banks. Hence, since we find that softer banking supervision
standards make stronger the impact of low rates on higher risk-taking, the evidence suggests excessive
risk-taking due to low short-term rates.
In addition, since – as explained earlier – the higher impact of low rates on the softening of standards
due to high securitization, that we find, may also be due to moral hazard problems (Rajan, 2005); our
evidence, therefore, suggests excessive risk-taking when short-term rates are low as compared to a
situation with low agency problems in the banking sector (i.e., banks grant more loans when short-term

8
low long-term rates on the softening of standards, both directly and indirectly via
softer bank supervision standards and higher securitization levels. These results help
shed light on the root causes of the current global crisis (Allen, 2009; Rajan, 2009;
and Diamond and Rajan, 2009) and have important policy implications for monetary
policy, banking regulation and supervision, and for financial stability.
The rest of the paper proceeds as follows. Section II describes the data,
introduces the variables employed in the empirical specifications, and reviews the
empirical strategy. Section III discusses the results and Section IV concludes.
II. Data and Empirical Strategy
A. The Bank Lending Survey (BLS) dataset
The main dataset used in the paper are the answers from the Euro Area BLS. The
national central banks of the Eurosystem request information on bank lending
standards since 2002 through the Euro Area BLS, a quarterly survey on banks' lending
practices based on the answers of a representative sample of banks in each country.
The questions asked were chosen on the basis of theoretical considerations related to

the monetary policy transmission channels and on the experiences of other central
banks running similar surveys, in particular in the US and in Japan. The main set of
questions did not change since the start of the survey in 2002:Q4.
15

The survey contains 18 specific questions on past and expected credit market
developments. Past developments refer to credit conditions over the past three
months, while expected developments focus on the next three months. There are two
main borrower sectors that are the focus of the survey: enterprises and households.
Loans to households are further disentangled in loans for house purchase and for
consumer credit, consistently with the classification of loans in the official statistics
for the Euro Area.


rates are low and may care less about the standards they set if they can securitize the loans thereby
transferring the risk off).
15
Berg, van Rixtel, Ferrando, de Bondt and Scopel (2005) describe in detail the setup of the survey.
Sauer (2009) provides an update of the most recent developments and the few changes to the survey
(request of additional information via ad-hoc questions).

9
The backward-looking questions cover the period from the last quarter of 2002 to
the first quarter of 2009. In order to use a balanced panel, the analysis is restricted to
the 12 countries which comprised the Euro Area in 2002:Q4. Over this period we
consistently have data for 12 Euro Area countries (Austria, Belgium, France, Finland,
Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, and Spain). The
sample of banks is representative of the banking sector in each country. Therefore it
comprises banks of different size, although national guidelines on the formation of the
sample expressed a preference to include the largest banks in each country.

The questions imply only qualitative answers and no figures are required. The
survey is carried out by the national central banks of the Euro Area countries.
Typically the questionnaire is sent to senior loan officers, like for example the
chairperson of the bank’s credit committee. The response rate has been virtually 100%
all the time.
The questionnaire covers both supply and demand factors. Concerning the supply
factors, which are addressed in ten different questions, attention is given to changes in
lending standards, to the factors responsible for these changes and to the credit
conditions and terms applied to customers.
Lending standards are defined as the internal guidelines or criteria that guide a
bank's loan policy and are addressed in two questions, each referring to a different
borrower (enterprises and households, further disentangled in loans for house
purchase and consumer loans).
16
The question asked is: “Over the past three months,
how have your bank’s lending standards as applied to the approval of loans (to
enterprises or to households) changed?” Banks can chose their answers among five
choices, ranging from “eased considerably” to “tightened considerably.” (See
Appendix for a detailed description of the questions used in the paper).
17

The successive set of questions gives respondents the opportunity to assess how
specific factors affected lending standards as applied to the approval of loans to both

16
In cases when foreign banks are part of the sample, the lending standards refer to the loans' policy in
the domestic market which may differ from guidelines established for the headquarter bank.
17
See for all the information related to the
BLS.

10
enterprises and households. In particular whether the changes in standards were due to
changes in bank balance-sheet strength (either bank liquidity, capital, or access to
market finance), to changes in competitive pressures (either from other banks or from
non-banks), to changes in expected general economic conditions or changes in
borrower risk (borrower’s collateral risk/ value or borrower’s outlook). We will use
this information to assess whether the changes of standards are just due to changes in
borrower credit-worthiness to assess bank risk-taking and also to understand the
channels by which low rates, high securitization and weak supervision affect loan
risk-taking by banks.
Finally, we have information on the changes in the terms and conditions of a
loan. These are the contractual obligations agreed upon by the lender and the
borrower, such as the interest rate (both for average and for riskier borrowers), the
loan collateral, size, maturity and covenants. We use this information to assess
whether the different standards are adjusted for the risk taken.
Concerning demand factors, which are the topic of seven questions, various
factors related to financing needs and the use of alternative finance are mentioned.
Three questions look at borrowing demand from enterprises and four at demand from
households. Finally, the survey also allows participating banks to give free-formatted
comments in response to an open-ended question.
18

The Euro Area results of the survey (which are a weighted average of the results
obtained for each Euro Area country) are published every quarter on the website of
the ECB. In few countries, the answers from the Survey at the national level are
published by the respective national central banks. However, the overall sample
including all the answers at the country and bank level is confidential.
For the purpose of this paper we concentrate only on few questions from the BLS
that we describe in detail in Appendix I. The questions are related either to the

previous three months or to the expected change in lending standards for the next
three months. We find very similar results using either of the two set of questions and

18
For the purpose of this paper we do not use the answers to questions related to the demand; however,
in non-reported regressions we control for demand answers. The results are qualitatively similar.

11
opt to report only the results related to actual changes since these ones are based on
actual standards applied to and actual loan demand received.
B. Macroeconomic and financial variables
We use several macroeconomic and financial variables in our analysis. We lag by
one quarter the macro and financial variables, hence we use information from
2002:Q3 to 2008:Q4.
19
All the series have quarterly frequency to be consistent with
the answers from the BLS. The main proxy for the monetary policy rate is the
quarterly average of the EONIA overnight interest rate, as published by the ECB. The
main macroeconomic controls we use are: the annual real GDP growth rate and the
inflation rate.
20
The inflation rate is defined as the quarterly average of monthly
inflation rate expressed in annual terms.
To exploit the cross-sectional differences in the stance of monetary policy at each
moment in time and to get exogenous variation of monetary policy, we calculate for
each country a Taylor-rule implied rate over the sample period and then use the
difference between this rate and the actual EURIBOR 3 months rate as explanatory
variable (we follow Taylor, 2008, and Ahrend, Cournède and Price, 2008). These
rule-implied rates are calculated following simple country Taylor rules with
coefficients 0.5 for inflation and output gap (see Taylor, 1993). Output gap and

inflation are country specific, while the natural rate has been set at 2.1 and the
inflation target at 1.9.
21
We also define the periods of “expansive” monetary policy as
the number of consecutive quarters in which the nominal short-term rate, measured by

19
See Appendix II for a detailed description of the variables used in the paper.
20
In non-reported regressions we have used as macroeconomic variables also expectations of GDP
growth and inflation from Consensus Forecast or ECB forecast. The results are qualitatively similar,
but these variables are not available for all Euro Area countries over the whole period considered. In
addition, we have also used variables that proxy for country risk, as for example the difference between
the long-term rates for each country (based on the 10 year Treasury bond) and the corresponding long-
term German rate. The results are qualitatively similar.
21
The output gap for each Euro Area country is the average of the output gap estimates from the
European Commission, the OECD and the IMF. As a robustness check we have also used the Taylor
rule specification in Gerdesmeier, Mongelli and Roffia (2007) with interest-rate smoothing. In
particular, we use the estimated coefficients for the Euro Area and we plug them in a different Taylor-
rule equation for each country. The results are qualitatively similar to the ones obtained with simple
Taylor rules from Taylor (1993 and 2008) and Ahrend, Cournède and Price (2008).

12
the EURIBOR 3 months, was below the Taylor-rule implied rate starting in 1999:Q1,
i.e. when the euro was implemented.
To assess the impact of long-term rates, we also use 10 year Treasury bond yield
different for each country. We also control in some robustness regressions for the
term spread, which is calculated as the difference between the 10 year rate and the 3-
month rate, house prices growth and credit growth (the source is ECB, see Appendix

for details).
One of the most notable innovations in financial markets over the last few years
has been the use of securitization. Thus, we also construct a variable which measures
securitization activity. This is the ratio between the volume of all the deals involving
asset-backed securities and mortgage-backed securities in each quarter, as reported by
Dealogic, normalized by the outstanding volume of loans.
22
The securitization
variable is country-specific since we have information about the nationality of the
collateral.
23
The volume of loans is available from the official ECB statistics.
In addition, since securitization is endogenous to the financial system regulation
and to the business cycle, in particular to the level of short-term interest rates, we
instrument the level of securitization activity with a time invariant indicator based on
the legal environment for securitization in each country. We construct this indicator
from country information contained in the report Legal Obstacles to Cross-Border
Securitization in the EU, European Financial Markets Lawyers Group, 2007. This
measure assesses the legal environment for securitization. The view taken is that a
more regulated environment can be conducive to a framework of “legal certainty”
which may be attractive for investors. The indicator results in ample cross-country
variation in the Euro Area.
24


22
As a robustness check we use also gross volumes of new loans issued. In addition, since it can be
presumed that loans are securitized by the banks with a quarter lag after they have been granted, we lag
the numerator of the ratio by one quarter. In both cases results do not change.
23

We are taking into account only deals for which the underlying collateral resides in one of the Euro
Area countries. Thus, we do not include securitization from Euro Area banks of loans granted outside
the Euro Area.
24
See Appendix II for details.

13
We also use a capital stringency index which assesses the capital supervision
standards for bank capital (see Laeven and Levine, 2009). Capital stringency is an
index of regulatory oversight of bank capital. This index is based on the following
questions: Is the minimum capital asset ratio requirement risk weighted in line with
the Basel guidelines? Does the minimum ratio vary as a function of market risk? Are
market values of loan losses not realized in accounting books deducted from capital?
Are unrealized losses in securities portfolios deducted? Are unrealized foreign
exchange losses deducted? What fraction of revaluation gains is allowed as part of
capital? Are the sources of funds to be used as capital verified by the regulatory or
supervisory authorities? Can the initial disbursement or subsequent injections of
capital be done with assets other than cash or government securities? Can initial
disbursement of capital be done with borrowed funds? Thus, capital stringency does
not measure statutory capital requirements. Instead, it measures the regulatory
approach to assessing and verifying the degree of capital at risk in a bank (Laeven and
Levine, 2009).
25

Table 1 shows the summary statistics of the main variables used, including the
correlations across variables and across countries.
26
Table 1, Panel A, shows that the
average overnight rate (common across countries) was 2.87 with a standard deviation
of 0.81, whereas long-term rates had an average of 4.05 and 0.45 as standard

deviation. Average GDP growth was 2.44% while its standard deviation was 2.09,
ranging from a minimum of -7.39 to a maximum of 8.46. Inflation average was 2.51
with 0.99 as standard deviation. Average Taylor rate differences were -1.37, which
indicates that on average monetary policy was expansive, with a standard deviation of
2.59, a minimum value of -9.58 and a maximum of 3.47, i.e. there is ample variation
of Taylor-rule implied rates over the sample. Securitization has an average of 1.76
with a maximum value of 1.62, ranging from 0 to 10.33. Capital index that measures
regulation supervision standards for bank capital ranges from 3 to 7 and the
securitization instrument based on regulation of the securitization at the country level
varies from 1.5 to 14. In Table 1, Panel B, C and D, we can see that there is ample
variation of lending standards applied to non-financial firms and to households with

25
See Appendix II for details.
26
The average of the country variables are not weighted by the size of the countries.

14
maximum variation for loans for house purchase. It is also interesting to note that the
average lending standards were positive, which implies tightening.
The correlations among the variables are not very strong. As we can see in Table
1 there is significant cross-country variation between Taylor-rule implied rates (GDP
growth and inflation were high in some countries whereas in others were low). In
addition, there is also ample cross-sectional variation of securitization in the Euro
Area, and also in banking supervision standards. All in all, the dynamics of the
lending standards, of the securitization activity, supervision and of the business cycles
show significant heterogeneity across Euro Area countries from 2002 to 2009.
C. Empirical strategy
We want to empirically address the impact on the softening of lending standards
of both short and long-term rates, of securitization and of banking supervision

regulation. Moreover, we want to analyze whether the softening of standards implies
higher loan risk-taking by banks.
There are four major challenges to identify the previous questions. First,
monetary policy rates are endogenous to the (local) economic conditions. Second,
banking supervision regulation may be endogenous to monetary policy especially in
cases when the central bank is responsible for both. Third, securitization activity is
endogenous to monetary (bank liquidity) conditions, since these affects the ability of
banks to grant loans. Fourth, it is very difficult to obtain data on the pool of potential
borrowers approaching a bank, to know their quality, and then to know whether, how
and why banks change their lending standards.
Our identification strategy to tackle the four previous identification challenges
relies upon the data we use, the Euro Area Bank Lending Survey dataset.
First, with regard to the monetary policy measure, there is an identical monetary
policy (overnight) rate for all the countries, which show some significant time
variation between 2002 and 2009.
27
At the same time, in our sample the Euro Area
comprises 12 countries with imperfect business cycle synchronization and thus with


27
In Bernanke and Blinder (1992), and in Christiano, Eichenbaum, and Evans (1996), among others,
the overnight interest rate is the indicator of the stance of monetary policy. The ECB targets the
overnight rate as a measure of the stance of its monetary policy.

15
different levels of GDP growth and inflation.
28
Therefore, we can exploit exogenous
cross-sectional variation of the stance of monetary policy. For example, Spain and

Ireland have grown at a much higher rate and with a higher inflation rate than
Germany and France, the two largest Euro Area countries during the 2002-06 period.
Hence, the monetary policy rates may have been too low for Ireland and Spain and
maybe high for Germany and France (Taylor, 2008).
Second, banking supervision regulation in the Euro Area is decided by the
national supervisory authorities and/or national central banks, whereas monetary
policy is decided by the European Central Bank and Eurosystem as a whole.
Therefore, banking supervision and regulation standards are exogenous to monetary
policy rates in the Euro Area.
Third, there is significant cross-sectional variation in securitization activity partly
arising from cross-country differences in the regulation of the market for
securitization.
Fourth, we have access to the confidential Bank Lending Survey dataset of the
Eurosystem. As explained earlier, national central banks request from banks quarterly
information on the lending standards they apply to customers and on the loan demand
they receive. The survey started in 2002:Q4 and it collects the answers from a
representative sample of around 90 banks. This rich information allow us to analyze
not only whether banks change the lending standards for the pool of borrowers they
receive, but also to whom they change the standards (average or riskier borrowers),
why (due to changes in borrower risk, or in bank balance-sheet strength, or in
competition) and how (loan spreads, size, collateral, maturity or covenants).
These data overcome some of the problems inherent to data on actual credit
granted, which do not allow to infer conditions offered to the pool of potential
borrowers, i.e. those that were either rejected by the banks or that found the terms of
the loans too expensive and decided not to take the loans. In addition, the BLS data
contain information for all type of loans (loans for business, for house purchase and
for consumption) and on all type of standards (loan spreads for average or riskier
borrowers, loan size, maturity, covenants, etc). Finally, the survey data also contain
information on why banks changed their standards: is it because of an improvement in

28
See for example Camacho, Pérez-Quiros and Saiz (2006).

16
the borrower credit-worthiness, or because the bank has more liquidity to lend, or
because there is higher competition for banks (both stemming from the banking sector
or from the non-banking sector, e.g. the shadow banking system)? All this information
helps us to disentangle loan demand and supply (i.e. the sample selection issue) and to
analyze bank loan risk-taking.
We run a series of GLS panel regressions where the left-hand side variable is the
net percentage of banks that have tightened their lending standards over the previous
quarter – i.e., the difference between the percentage of banks in each country
reporting an increase in the tightening of standards minus the percentage of banks
reporting a softening of lending standards.
29
Therefore, a positive figure indicates a
net tightening, and a negative figure a net easing of lending standards.
30
On the right
hand side we have measures of monetary policy rates, long-term rates, securitization
activity and banking regulation, all lagged by one quarter. The monetary policy is
measured either by the quarterly average of overnight rates (EONIA) or by the
differences between the rates implied by a Taylor rule and the 3-month Euribor.
31
The
normal panel we use is (country, quarter) with country fixed-effects that capture time-
invariant differences in the banking, economical and legal structure of each country.
We also control for GDP growth and inflation and, in some specifications, we control
for time (quarter) fixed effects, country risk, credit growth, term spread, house prices
and bank size and fixed-effects.

III. Results
We start analyzing in Table 2 the impact of overnight rates (EONIA) on lending
standards for loans to non-financial firms, for house purchase and for consumption

29
This measure has a high predictive power both for credit and output growth both for the US and for
the Euro Area (see Lown and Morgan, 2006; and Ciccarelli, Maddaloni and Peydró, 2009).
30
The use of this statistics has the consequence that no distinction is made for the degree of
tightening/easing of lending standards in the replies. Diffusion indexes can be calculated to take care of
this issue. They are balanced statistics where also the intensity of the tightening is taken into account.
In particular, a weight of 0.5 is given to the percentage of banks answering that they have tightened
somewhat and a weight of 1 to the percentage of banks that have tightened considerably. We do not
report results relative to the use of diffusion indexes, but they do not differ qualitatively from the
results obtained with net percentages.
31
Taylor-rule implied rates for Euro Area are normally calculated with 3 months rates (See Taylor,
2008; and Ahrend, Cournède and Price, 2008).

17
(Question 1 and 8 of BLS, see Appendix). The dependent variable “Changes in
lending standards” is the net percentage of banks which have reported to have
tightened their lending standards for the approval of loans.
32
In column 1, we see that
the coefficient on overnight rates is equal to 24.889***. Therefore, a higher level of
overnight rates implies higher lending standards for non-financial firms, i.e. a
tightening of lending standards.
33
Once we control in column 2 with real GDP growth

and inflation rate (i.e. the main determinants of overnight rates), results are still highly
statistically significant: the coefficient on overnight rates is 22.152***. The
coefficient on GDP growth is negative and equal to -3.287***. Higher GDP growth
softens lending standards to non-financial firms; hence, lending standards are pro-
cyclical. On the other hand, the coefficient on inflation is 5.283***, which indicates
that a higher inflation rate tightens lending standards to non-financial firms.
From column 3 to 6 we report the results of the same regressions for lending
standards to households for house purchase and for consumption. Results are very
similar to column 1 and 2 both for overnight rates, GDP growth and inflation.
However, the effect of overnight rates on standards is stronger for loans to non-
financial corporations than for loans to households (22.152***, 11.544*** and
8.249*** respectively).
Results are also highly economically significant: a one sigma decrease of
overnight rates softens standards to enterprises almost three times more than a one
sigma increase of real GDP growth (17.89 and 6.87 respectively). In addition, since
the standard deviation of lending standards for firms is 33.73, 1 sigma EONIA moves
more than half sigma of standards for firms (17.89/33.73). Concerning loans to
households, both overnight rates and GDP growth have similar economic significance
(9.32 and 9.56 respectively) for loans for house purchase, while for loans for
consumption overnight rates have almost double economic importance than GDP
growth (6.66 and 3.95 respectively).

32
We run GLS panel regressions with country fixed-effects and standard errors corrected for within
autocorrelation and also for heteroskedasticity between countries.
33
*** means significant at 1% level, ** significant at 5%, and * significant at 10%. For convenience
we also indicate the significance levels of the coefficients in the text.

18

Banks may soften their lending standards when overnight rates are lower because
of improvements in borrowers’ net worth and quality of collateral as shown by
Matsuyama (2007), Bernanke and Gertler (1995), or Bernanke, Gertler, and Gilchrist
(1996 and 1999). At first instance we have used GDP growth to control for the
improvement of borrowers’ net worth. From column 7 to 12 we make a further
identification step. The left hand side variable is now defined as the change in lending
standards due to changes in banks’ balance sheet constraints, i.e. changes in standards
not associated to borrowers’ quality (it corresponds to Question 2 and 9 of BLS, see
Appendix). From column 7 to 12 we see that low overnight rates softens lending
standards because of improvements in banks’ balance sheet position. In this case
lending standards are softened because of “pure” bank-supply factors and thus they
reflect an increase in loan risk-taking by banks. Results are both statistically and
economically highly significant. Moreover, now EONIA is economically more
important than GDP growth for all type of loans including loans for house purchase.
In Table 2 we have used the level of overnight rates as an indicator of monetary
policy. This measure of monetary policy is time-varying. However, to get cross-
sectional variation in the monetary policy stance and to assess the level of short term
rates against a benchmark, we calculate the difference between the rate implied by a
country-specific Taylor-rule and a nominal short-term rate.
34
Note also that this
measure provides exogenous cross-sectional variation of monetary policy stance (due
to different levels of inflation and/or GDP growth) as the deviation of the stance in a
country at a given point in time with respect to the Euro area average is due to
common monetary policy rates decided for the whole Euro Area.
In Table 3, column 1 to 3, we see that lower values between Taylor-rate
differences (i.e. more expansive monetary policy or low levels of monetary policy
rates) imply a softening of standards for loans to non-financial firms, for house
purchase and for consumption. Moreover, in column 4, 6 and 8 we see that the
softening is over and above an improvement of borrowers’ collateral risk/ value and

outlook – i.e., the softening also stems from pure bank-supply factors. Next, we
introduce an additional variable measuring for how long the stance of monetary policy

34
Another way to do it is through real short-term interest rates. In this case, negative rates are low. In
non-reported regressions, we find virtually the same results if we use real rates.

19
has been expansive in each country by counting the number of consecutive quarters in
which nominal short-term rates have been lower than Taylor-rule implied rates. As we
can see in column 5, 7 and 9, rates too low for too long imply an even further
softening of standards. Results are significant for all type of loans but are stronger for
loans for house purchase (-0.336*** for loans for house purchase as compared to
-0.174 for business loans).
Short-term rates versus long-term rates
In Table 4 we include long-term rates in the analysis. In column 1, 2 and 3, we
find that short-term rates continue to be highly significant both economically and
statistically as determinants of total lending standards for all type of loans. Long-term
rates are only significant determinants of lending standards to households for house
purchase, which are generally long-term loans. However short-term rates have a larger
impact than long-term rates in explaining standards, even for loans for house purchase
– note that the standard deviation of overnight rates is 0.81 whereas for long-term
interest rates is 0.46.
In column 4, 5 and 6 we see that low short-term rates are more important than
low long-term rates in the softening of standards due to improvements in bank’s
balance-sheet conditions (better bank liquidity, capital and access to market finance)
and, thus unrelated to improvements in borrower’s risk. In addition, in Table 5 Panel
A and B, we can see in detail the impact of both short-term and long-term rates on the
specific factors affecting lending standards: better expected economic conditions,
better borrower collateral risk/value and outlook, better bank capital or liquidity

position or market access to finance and, finally, competitive pressures stemming
either from the banking system itself or from non-banks. We find that the softening is
due to all factors. It is interesting to note that for business loans the softening is due
more to higher bank competition or to improvements in bank balance-sheets position
than to improvements of the borrower’s risk of collateral.
In Table 6 (Panel A to C) we look at the impact of low interest rates on credit
conditions and terms of loans. Low short-term rates soften all type of standards for all
type of loans (price and non-price terms). It is interesting to note that the softening of
standards to average and riskier borrowers is similar for loans for house purchase,

20
whereas low short-term rates imply softer standards for average borrowers than for
riskier borrowers for business loans and consumer loans. Instead, low long-term rates
soften more standards for riskier rather than average businesses.
All in all, our results suggest that low short-term rates imply higher loan risk-
taking by banks. Banks soften the standards not only because of borrowers’ better
conditions but also because banks’ balance-sheet constraints are relaxed. Moreover,
banks relax all standards including margins on both average and riskier loans,
collateral, covenants, size, and maturity. In addition, our results suggest that low
short-term rates have a stronger impact on the softening of standards than long-term
rates.
Banking regulation and supervision standards
Now we analyze whether the impact of low rates on the softening of standards is
lower with stronger regulation supervision standards for bank capital. In Table 7 we
introduce capital stringency which is an index of regulatory oversight of bank capital
(Laeven and Levine, 2009). This measure has cross-sectional variation and also some
time variation (see Appendix and Section II). Hence, we use Taylor-rate differences
as the monetary policy measure to fully exploit the cross-sectional variation of both
monetary policy rates and banking regulation supervision standards.
We find that the impact of low short-term rates on the softening of lending

standards (over and above improvements in borrower’s collateral risk and outlook) is
larger with weaker supervision standards, both for loans for house purchase and for
consumption (Table 7, Panel B). At the same time, the direct effect of short-term rates
on lending standards is more significant than the level effect of supervision standards.
Finally, in contrast with short-term rates, the softening impact of low long-term rates
on lending standards does not depend on banking regulation supervision standards.
The results we find are not very strong though, maybe because there is not
enough variation of banking regulation supervision across Euro Area countries or
maybe because it is difficult to capture how good regulation or supervision is with
measures based on stringency of requirements. This is consistent with the arguments
made by Allen (2009) and Rajan (2009) concerning the need for “good” supervision
regulation, which does not necessarily mean more stringent.

21
Securitization
In Table 8 we introduce the securitization activity at the country level. First, we
note that higher securitization activity tends to softer lending standards, a result
similar to the ones obtained by Mian and Sufi (2009) and Keys et al. (2008) for the
US mortgage market. More importantly, when we add an interaction term of
securitization and overnight rates, we find that higher securitization activity amplifies
the impact of low short-term rates on the softening of standards. The results are
similar for all type of loans. We do not find, however, similar results when we interact
long-term rates with securitization.
In Table 9 we consider changes in lending standards due to bank balance-sheet
constraints. We find that higher securitization activity makes stronger the impact of
low short-term rates on the softening of lending standards both for loans for house
purchase and for consumption. This implies that there is softening of standards over
and above improvements of borrower risk due low short-term rates and high
securitization activity, thus suggesting higher loan risk-taking by banks when both
securitization activity is high and short-term rates are low, as suggested by Rajan

(2005).
In Table 10 we analyze the impact of securitization activity on all the factors that
affect lending standards. In Panel A we see that for loans to non-financial firms, the
interaction between overnight rates and securitization activity has a significant
coefficient when the softening of lending standards is due to more non-bank or market
finance competition and also due to risk of collateral. In Panel B we see that for loans
to households for house purchase, higher securitization makes stronger the impact of
low short-term rates on the softening of lending standards due to more non-bank
competition, better bank balance-sheet constraints, and also from better risk of
collateral. In Panel C, we find similar results for consumer credit except for
change in standards due to competition.
In Table 11, we analyze terms and conditions of loans and we find that the
softening of higher securitization in conjunction with low short-term rates is for the
following standards: (i) for firm loans: margins on average loans, collateral
requirements, covenants, and maturity; (ii) house purchase loans: margins on both

22
average and riskier loans, collateral requirements, and loan-to-value ratio restrictions;
and (iii) consumer credit: margins on both average and riskier loans, collateral
requirements, maturity, and non-interest rate changes.
In a situation with low short-term rates and high securitization activity, the
previous results suggest higher loan risk-taking by banks. Moreover, when analyzing
at the factors by which banks change the standards, the previous results highlight: (i)
the importance of the “shadow banking system” in setting of the softening of bank
lending standards (both stemming from higher competition from non-banks and from
market finance, which both have a different level of regulation and supervision than
banks), (ii) the importance of bank balance-sheet liquidity in the softening of
standards, and (iii) the risk transfer effects linked to securitization in the sense that the
collateral risk and value may matter less for banks when granting loans if banks can
transfer these risks off.

Given that presumably the securitization activity in each country not only
depends on the regulation and development of the financial system of that country,
but also on short-term rates and the business cycle, securitization activity is
endogenous to monetary policy. Therefore, we instrument securitization by using an
indicator of the regulatory environment for securitization in each country (see Section
2 and the Appendix for the explanation of the instrument). As shown in Table 12
Panel A, we find that the securitization activity is highly correlated with securitization
regulation (t-statistic in the first stage regression is 8.10), thus the instrument does not
suffer from weak instrument concerns (Staiger and Stock, 1997). Moreover, from the
second-stage regression (Panel B), the estimates suggest that the impact of low short-
term rates on the softening of standards is larger when the component of actual
securitization predicted by the securitization regulation is higher. We do not find,
however, similar results when we interact long-term rates with securitization.
Our results suggests not only that low levels of short-term rates increase loan
risk-taking but they may create excessive risk-taking since we find that the impact of
low short-term rates on the softening of standards is stronger when securitization is
high or banking supervision standards are weak. It suggests excessive risk-taking as
compared to a situation with zero or low bank agency problems: From Allen and Gale
(2007) for example, low short-term interest rates create high risk-taking by banks

23
because of moral hazard problems in banks. Hence, since we find that softer banking
supervision standards make stronger the impact of low rates on higher risk-taking, our
evidence suggests excessive risk-taking. In addition, since as explained earlier, we
find higher impact of low rates on the softening of standards due to high
securitization, and there is theory that suggests that this is due to moral hazard
problems in banks (Rajan, 2005), our evidence suggests excessive risk-taking when
short-term rates are low, as compared to a situation with low agency problems in the
banking sector.
35

All in all, we find that low short-term interest rates, high securitization, and
weaker banking regulation directly and in conjunction soften the lending standards.
This softening is over and above improvements in the borrowers’ credit-worthiness,
thus indicating higher loan risk-taking by banks. Moreover, the impact of short-term
interest rates on the softening of lending standards is stronger with higher
securitization and weaker banking supervision standards, thus suggesting excessive
loan risk-taking by banks. Finally, low short-term rates are more important than long-
term rates for the softening of standards. All these results, therefore, have implications
for the origins of the current global financial crisis (see Rajan, 2005 and 2009; Allen
2009; Calomiris, 2008; Diamond and Rajan, 2009; Brunnemeier, 2008; Taylor, 2007
and 2008; and others).
36

IV. Conclusions
We analyze the root – and not the proximate – causes of the current crisis (Allen,
2009; Diamond and Rajan, 2009) by studying the determinants of bank lending
standards in the Euro Area, where there are identical monetary policy rates but there is
ample cross-sectional variation in GDP growth, inflation, securitization and bank
regulation. We use the answers from the confidential Bank Lending Survey where


35
As explained earlier, banks grant more loans when short-term rates are low and banks may care less
about the lending standards they set if they can securitize the loans thereby transferring the risk outside.
In addition, the competition from the non-banking sector may intensify this behavior when both short-
term rates are low and securitization activity is high.
36
Bank risk problems may transmit through the system through interbank contagion and other
mechanisms, see Iyer and Peydró (2009) and Bandt, Hartmann and Peydró (2009). Once the banking

system is in trouble, a credit crunch stemming from low bank capital and liquidity has a high likelihood
to happen (see Jiménez, Ongena, Peydró and Saurina, 2009b) in turn affecting the real economy (see
Ciccarelli, Maddaloni and Peydró, 2009).

24
national central banks request from banks quarterly information on the lending
standards they apply to customers and on the loan demand they observe.
We find robust evidence that low short-term interest rates soften standards for
loans to both business and households, and the softening is over and above an
improvement of the borrower’s collateral risk/ value and outlook, thus suggesting
higher loan risk-taking by banks when overnight rates are low. Moreover, lending
standards are pro-cyclical and – by exploiting cross-country variation of Taylor-rule
implied rates – we find that short-term rates too low for too long soften standards even
further, especially for loans for house purchase.
In addition, using a time-varying measure of supervision standards for bank
capital, we find that weaker supervision standards increase the impact of low short-
term rates on the softening of standards. Moreover, we find that higher securitization
activity increases the impact of low short-term rates on the softening of standards,
even when we instrument securitization by the level of regulation in the market for
securitization of each country. Moreover, the impact of higher securitization and
weaker supervision standards in conjunction with low short-term interest rates softens
the lending standards over and above an improvement of the borrower’s collateral
risk/ value and outlook, thus suggesting higher (and even excessive) loan risk-taking
by banks. In particular, the results suggest that the softening is due to higher
competition from non-banks and market finance (the shadow banking system), from
credit risk-transfer (collateral risk is less important when the risk can be transferred
off easily), and from better bank balance-sheets (more bank liquidity and better bank
access to market finance).
Finally, low short-term rates are more important than long-term rates on the
softening of standards directly and indirectly via softer bank supervision standards or

higher securitization.
All in all, the results suggest that all these root causes of the current crisis have a
direct effect and also reinforce each other in the softening of lending standards set by
banks (see Rajan, 2005 and 2009; Allen 2009; Calomiris, 2008; Diamond and Rajan,
2009; Brunnemeier, 2008; Taylor, 2007 and 2008; and others). These results help shed

25
light on the origins of the current crisis and have important policy implications for
monetary policy, banking regulation and supervision, and for financial stability.