The Non-performing Loans:

Some Bank-level Evidences

Yixin Hou1

Abstract
This paper looks into the non-performing loan problem in commercial banks. Using the threshold

regression technique, we found some evidences that non-performing loans have non-linear

negative effect on banks lending behaviour.

JEL Codes: G21 E44

Key Words: non-performing loan, credit crunch, threshold effect, capital ratio

I. Introduction

A simple definition of non-performing is: A loan that is not earning income and: (1) full payment

of principal and interest is no longer anticipated, (2) principal or interest is 90 days or more

delinquent, or (3) the maturity date has passed and payment in full has not been made.

The issue of non-performing loans (NPLs) has gained increasing attentions in the last few

decades. The immediate consequence of large amount of NPLs in the banking system is bank

failure. Many researches on the cause of bank failures find that asset quality is a statistically

1
Corresponding author Department of Economics, University of Birmingham Edgbaston Birmingham B15 2TT UK
Email address: [email protected]

1
significant predictor of insolvency (e.g. Dermirgue-Kunt 1989, Barr and Siems 1994), and that

failing banking institutions always have high level of non-performing loans prior to failure.

It is argued that the non-performing loans are one of the major causes of the economic stagnation

problems. Each non-performing loan in the financial sector is viewed as an obverse mirror image

of an ailing unprofitable enterprise. From this point of view, the eradiation of non-performing

loans is a necessary condition to improve the economic status. If the non-performing loans are

kept existing and continuously rolled over, the resources are locked up in unprofitable sectors;

thus, hindering the economic growth and impairing the economic efficiency.

In this paper, we focus on the impacts of non-performing loans on microeconomics, specifically, at

the bank level to empirically evaluate how non-performing loans affect commercial banks’ lending

behaviour. The rest of the article is organised as follows. Section 2 provides the definitions of non-

performing loans. Section 3 discusses some consequences of non-performing loans on the

economies. Section 4 explains the methodology for our empirical studies. Section 5 describes the

data. Section 6 gives out the empirical results. And section 7 gives out the conclusion.

II. Non-performing Loans

There is no global standard to define non-performing loans at the practical level. Variations exist

in terms of the classification system, the scope, and contents. Such problem potentially adds to

disorder and uncertainty in the NPL issues. For example, as described by Se-Hark Park (2003),

during 1990s, there were three different methods of defining non-performing loans in Japan: the

1993 method based on banking laws; the “Bank’s Self-Valuation” in March 1996; and the

“Financial Revival Laws-Based Debt Disclosure” in 1999. These measurements have gradually

broadened the scope and scales of the risk-management method. Similar to the trend in Japan,

more countries, regulators, and banks are moving towards adopting and adapting better and more

consensus practices. For example, in the U.S., federal regulated banks are required to use the five-

tier non-performing loan classification system according to BIS: Pass, Special Mention,

2
Substandard, Doubtful, and Loss. Presently, the five-tier system is the most popular risk

classification method, or, in some cases, a dual system of reporting according to their domestic

policy guidelines as well as the five-tier system. According to BIS, the standard loan

classifications are defined as follows:

(1) Passed: Solvent loans;

(2) Special Mention: Loans to enterprises which may pose some collection difficulties, for

instance, because of continuing business losses;

(3) Substandard: Loans whose interest or principal payments are longer than three months in

arrears of lending conditions are eased. The banks make 10% provision for the unsecured portion

of the loans classified as substandard;

(4) Doubtful: Full liquidation of outstanding debts appears doubtful and the accounts suggest that

there will be a loss, the exact amount of which cannot be determined as yet. Banks make 50%

provision for doubtful loans;

(5) Virtual Loss and Loss (Unrecoverable): Outstanding debts are regarded as not collectable,

usually loans to firms which applied for legal resolution and protection under bankruptcy laws.

Banks make 100% provision for loss loans.2

Non-performing loans comprise the loans in the latter three categories, and are further

differentiated according to the degree of collection difficulties.

In addition to the standardised system, efforts have been made to improve the classification of

loans. For example, more countries are shortening the period when unpaid loans become past due,

intending to put loans on lenders’ timetable sooner and require them to address these loans before

losses start to escalate. The International Accounting Standard 39 revised in 2003 focuses on

recognition and measurement of financial instruments and, most importantly, defines and

establishes the measurement and evaluation of impaired loans. As lenders usually make little or no

loss provision for impaired loans, they are at risk to be suddenly forced to reclassify such loans as

a loss and take a full write-down if the borrowers go bankrupt. The initiation of this standard is to

prevent lenders from being caught off-guard. In addition, many global economists, rating

agencies, and organisations such as the World Bank and the Asian Development Bank have begun

2
The details of the loan classifications are collected from various BIS documents.

3
to evaluate the effects of NPLs on GDP growth. They reduce growth estimates to reflect the time

and cost of resolving large non-performing loan issues.

In this paper, we use the standard BIS definition of NPLs, i.e., the last three categories of the BIS

classification. First, the standard definition makes it possible to compare the non-performing loan

problem across countries and banks. Second, the BIS definition is a prudential definition for

NPLs, which includes loans with uncertainty in addition to the virtual loss, thus, it enable banks to

address the NPL problems before it cause disasters.

III. The Effects of NPLs on Economics

Non-performing loans can lead to efficiency problem for banking sector. It is found by a number

of economists that failing banks tend to be located far from the most-efficient frontier (Berger and

Humphrey (1992), Barr and Siems (1994), DeYoung and Whalen (1994), Wheelock and Wilson

(1994)), because banks don’t optimise their portfolio decisions by lending less than demanded.

What’s more, there are evidences that even among banks that do not fail, there is a negative

relationship between the non-performing loans and performance efficiency (Kwan and Eisenbeis

(1994), Hughes and Moon (1995), Resti (1995)).

The phenomena that banks are reluctant to take new risks and commit new loans is described as

the ‘”credit crunch” problem. According to the United States Council of Economic Advisors

(1991), credit crunch is “a situation in which the supply of credit is restricted below the range

usually identified with prevailing market interest rates and the profitability of investment

projects”. A “credit crunch” is a disequilibrium phenomenon. It is present when banks are

unwilling to lend, especially when a firm with profitable projects cannot obtain credit in spite of

low interest rates (lower than the expected marginal products). Credit crunch results in excess

demand for credit and hence credit rationing, where loans are allocated via non-price mechanism.

Eventually, it imposes additional pressure on the performance of the monetary policy.

4
The idea of credit crunch has drawn attention when the traditional view failed to satisfactorily

explain the economy state for those countries that suffered from the South-East Asian financial

crisis in 1997. Under the traditional view, the link between the interest rate change and the real

economic activity occurs through investment and consumer durable expenditure. In response to

the currency crisis in 1997, the interest rate was raised. It was strongly believed by IMF that the

hike would help stabilise the foreign currency market and eventually induce banking reform by

crowding out low-profit projects. However, the persistent fall in economic growth rate and the

lasting economic recession cast doubt on the true benefits of the policy and the effectiveness of the

traditional view of the transmission mechanism. The idea of credit crunch addresses an alternative

explanation for the transmission mechanism.

During a crisis, in order to restore the credibility among creditors and depositors, failing financial

institutions not only try to expand their equity bases, but also reduce their risky assets or change

the composition of the assets portfolio. As a result of such defensive action, the corporate debtors

are always targeted, thus stalling the overall economic growth.

Specially, the reluctance of banks to lend can be caused by several reasons, such as the increased

capital adequacy requirement imposed by Basel Accords; impaired debt-servicing capacity,

especially small-to-medium enterprises (SMEs); risks of a further fall in collateral value, etc.,

which make the interest rate not to serve as the main determinant by banks in credit approval.

Non-performing loans have been viewed to constitute one of the most important factors causing

reluctance for the banks to provide credit. In a high NPL condition, banks increasingly tend to

carry out internal consolidation to improve the asset quality rather than distributing credit. Also,

the high level of NPLs requires banks to raise provision for loan loss that decreases the banks’

revenue and reduces the funds for new lending. The cutback of loans impairs the corporate sector

as they have difficulties in expanding their working capital, blocking their chances of resuming

normal operation or growing. Unavailability of credit to finance firm’s working capitals and

investments might trigger the second round business failure which in turn exacerbates the quality

of bank loans, resulting in a re-emerging of banking or financial failure. In a worse case, it triggers

5
an endless vicious liquidity spiral: As a result of poor economic condition and the depressed

economic growth, the level of NPLs increases  the weaker corporate sector makes banks more

reluctant to provide additional credits  with insufficient capital, the production sector is further

weakened, resulting in decreases in aggregate demand  again, even worse borrowers’ condition

creates more NPLs…

Krueger and Tornell (1999) support the credit crunch view and attribute the credit crunch in

Mexico after the 1995 crisis partially to the bad loans. They point out that banks were burdened

with credits of negative real value, thereby reducing the capacity of the banks in providing fresh

fund for new projects. Agung et. al. (2001) using the macro and micro panel data analyses to study

the existence of a credit crunch in Indonesia after the crisis. Both the macro and micro evidences

show that there was a credit crunch, characterised by an excess demand for loans, starting to

emerge in August 1997, one month after the contagion effects of the exchange rate turmoil in

Thailand spreading to Indonesia. They investigate the relationship between the loan supply and

real lending capacity, lending rates, real out put, bank’s capital ratio, and non-performing loan.

The results show that the coefficients on NPLs are negative and significant, which indicate that

bank credit supply declines with the worsening of the NPLs problem. Westermann (2003)

compares the cases of Germany after the credit boom of the late 1990s and Japan aftermath the

bubble burst in early 1990s. He argues that even though the German banks were in a better

condition than Japanese banks were, as the path of German’s aggregate credit looks so similar to

that of Japan, it is at least unlikely that the German credit slowdown was entirely driven by

demand, while that of Japan was mostly caused by a lack of supply. There must at least be some

supply side changes that affect the aggregate credit, and differences only exist in the magnitude of

the problem. He further points out that the one of the main reasons in Germany for the credit

crunch is the increased risk of non-performing loans after the credit boom.

IV. Empirical Specification

6
Empirical Methodology forTesting the Effect of Non-Performing Lonas

One implication from the Credit Crunch view is that increased non-performing loans can cause the

decline in commercial bank credits, as banks with high level of non-performing loans in their

portfolio may become increasingly reluctant to take up new risks and commit new loans.

In this paper, we use panel data from individual bank’s balance sheet to empirical assess whether

non-performing loans will negatively affect bank’s lending behaviour. The empirical model is

specified to capture the credit supply side factors. Normally, the supply of loans is determined by

banks’ lending capacity and factors that influence banks’ willingness to provide credits. In this

paper, non-performing loans are also taken into consideration. Based on Credit Crunch view, the

NPLs should have a negative effect on loans, implying that the higher NPLs in a bank’s portfolio,

the less credit that the bank can and is willing to supply.

For a simple commercial bank balance sheet, assets are mainly composed of commercial loans and

other earning assets; while on the liability side, deposits and capital are the main components.

Thus, we can conjecture that the loan growth is affect by deposit growth, capital growth and other

earning assets growth. In addition, we take the non-performing loan growth into consideration.

The basic model is as follows:

LGRi , t =a0 +a1 DGR i, t +a2 CGRi , t +a3 OEAGR i, t +a4 NPLGRi,t−1 (1)

where the index i is the index for individual banks and t is the index for time period.

LGR i,t is the loan growth rate,

DGRi ,t is the deposit growth rate in each time period t ,

CGR i,t is the capital growth rate,

OEAGR i,t is the other assets growth rate, and

NPLGR i ,t−1 is non-performing loan growth rate of the pervious year.

As financial intermediations, commercial banks’ main function is to receive deposits and make

loans to facilitate the flow of capitals. For most of the commercial banks, deposits are the main

7
funding sources for commercial banks’ assets. And loans take up the biggest proportion in the

asset portfolio. With the expansion of the asset size, banks will expand the volume of the loans to

re-balance the asset portfolio. Under the normal situation, loan growth rate is expected to move in

the same direction as the growth of deposits. The sign in front of

DGR i,t , thus, is expected to

be positive. But according to the non-performing loan hypothesis, when banks are in poor

condition ridden by high level of NPLs, the willingness for the banks to expand loans is decreased,

which implies that loan growth will not be consistent with the expansion of deposits.

Capital is the part of the long-term funds that banks can use with fewer restrictions than other

kinds of funds from outside sources. In addition, capital acts as the safety cushion for bank’s

lending. With a higher capital level, banks will feel more confident and less constrained to extend

loans. In order to maximise profits, which is the primary objective for most of the commercial

banks, banks are supposed to increase lending with the growth of capitals, which implies a

positive sign in front of

CGR i,t . However, under our hypothesis, such relationship between the

lending and capital is distorted, which suggests the possibility of the negative sign.

The sign in front of

OEAGR i ,t is expected to be negative as other earning assets are the

substitute for loans for a given amount of total assets. When the banks are heavily ridden by the

non-performing loan problem, they are unwilling to make loans for fear of the generation of new

non-performing loans; instead, they switch for securer assets, such as government bonds or

treasury bills, thus, the substitute effect will be greater. As a result of such greater substitution

effect, the other earning asset growth will have larger negative effects on loan growth. On the

other hand, higher credit risk may impede the growth of other earning asset. With higher credit

risk in portfolio, banks may have to retain more capital rather than investing in other earning

assets to improve the safety position. In addition, higher credit risk will reduce banks’ credit

ranking on the financial markets, which will also reduce their ability to engage in various

investments.

8
Under the non-performing loan hypothesis, the relationship between the loans and non-performing

loan conditions are negative, indicating negative sign associated with non-performing loan growth

rate. In our model, we use the lagged term of the non-performing loan growth. It is because that

the amount of non-performing loans is not normally known until the end of period; however, the

loan decisions are always made at the beginning of the period. Banks make decision according to

their existing non-performing loan situation, which can be roughly represented by the situation of

non-performing loans at the end of last period.

The Effect of Capital Adequacy

The Risk-Based Capital Regulation by the Basle Accord II has been playing increasing critical

role in commercial bank decisions. It mandates that banks hold capital in proportion to their

perceived credit risks. The Risk-Based Capital (RBC) is viewed as a regulatory tax that is higher

on assets in categories that are assigned higher risk weights. (Berger and Udell, 1994) As capital is

usually more expensive to raise than other assets, such as insured deposits, therefore, the

implementation of RBC is expected to further magnify the substitution effect, which encourage

banks to switch from the 100 percent risk credit category, such as commercial loans, to assets in 0

percent risk category, such as government bonds and treasury bills.

Under Basle Accord framework, banks are supposed to perform differently according to their

capital conditions. The reluctance in the supply of credit is expected to be more significant the

greater the proportion of non-performing loans held by the capital-deficient banks. Banks faced

with the requirement to raise their capital ratio to improve their risk position will make efforts to

meet the RBC standards either by raising expensive capital or by reducing risk-weighted assets

through substituting out commercial loans, where the latter enhances the negative effect on loans.

As a result, the capital ratio will affect the lending decision. In addition, it will further influence

the lending by interplaying with non-performing loans.

According to Basle Accord II, the target ratio of capital to risk weighted assets is set at 8%.

However, the mean capital ratios in our samples that we will show later are all above the required

9
8%, which suggests that banks may have more rigid internal capital ratio requirements. In

response to different characteristics and situations, different samples may have different levels of

effective capital ratio constraints. Thus, in our studies, we need to find out the effective capital

ratios which do play constraining roles for different samples. For each sample, we set

Dummyi ,t =1 if the capital ratio of the examining year t for bank i is equal or higher

than the effective capital ratio we find for that sample, otherwise,
Dummyi ,t =0 . The

adjusted regression equation which including the dummy effect on the constant and the dummy

effect on the non-performing loan is:

LGR i, t =a 0 +a1 DGR i, t + a2 CGR i, t + a3 OEAGR i, t +a 4 NPLGR i,t-1

+a5 Dummyi, t +a6 Dummyi, t NPLGR i , t−1 (2)

The Instrument Variable Method

The variables we use in the regression are potentially endogenous as they are simultaneously

determined through banks’ balance sheet constraints and are correlated with each other. So we

apply the method of two-stage least squares using instrumental variables. Wooldridge (2002)

argues that the two-stage least squares estimator is the most efficient estimator for endogenous

variables. We assume banks’ behaviours are continuous and they re-balance the portfolio at the

each period based on the portfolio of the previous period if nothing abnormal happens. So we use

the lagged values of the variables, i.e.,

OEAGR i,t−1 ,DGR i,t−1 ,CGRi,t−1 , as the instruments in

the first stage to get the estimations for the values of these variables in the current period, and

then, use these fitted values to perform the linear regression:

LGR i, t =a 0 +a1 DGR ′ + a2 CGR ′ +a3 OEAGR +a 4 NPLGR i,t-1

i, t i, t i,t'
+a5 Dummyi, t +a6 Dummyi, t NPLGR i , t−1 (3)

OEAGR ′ , DGR ′ ,CGR

where i,t i ,t i ,t ′ are the fitted values estimated in the first stage using the

instrument variables.

10
The Threshold Effect of NPLs

So far, we have emphasised the negative aspects of non-performing loans, i.e., they reduces the

efficiency of resource allocation of the financial system. But in reality, bad loans exist as a natural

consequence of lending behaviour when banks re-balance their portfolio. Banks decide the degree

of risk they will tolerate for a given level of expected return according to their risk preference.

When the level of non-performing loans goes beyond a certain level that banks can accept, the re-

balancing actions will be affected. It implies that the coefficient might change in reference to the

amount of NPLs; and hence, we conjecture that negative effect on lending is non-linear, i.e., there

is a critical threshold level decided by the rate of NPLs to total loans. Banks make lending

decision differently reacting to non-performing loans rate under or above a certain level; when

NPLs increase above the threshold, they start to cause negative effect on lending. The estimation

of the threshold, thus, is the main concern in this study and it will be interesting to compare banks’

behaviours below and above the threshold.

Threshold Method

Threshold regression techniques are used to address the question whether regression functions are

identical across all observations in a sample or fall into discrete classes. The threshold models

have a wide variety of applications in economics. Applications include separating and multiple

equilibria, sample split, mixture models, switching models, etc. Hansen (2000) argues that the

understanding of threshold models is a preliminary step in the development of statistical tools to

handle more complicated statistical structures.

The development of threshold regression models can go back to Dagenals (1969). He uses the

threshold regression technique to analysis the step-like-time-path discontinuous character of

durable goods. Hansen (1999) develops the panel threshold regression methods for non-dynamic

panels with individual-specific fixed effects. The basic structural equation for the model is:

μ i + β '1 xit +e it qit ≤γ

y it =
{
μ i + β '2 xit +e it qit > γ
(4)

where the subscript i indexes the individual and the subscript t indexes time; the scalar

11
 y it is the dependent variable, scalar
qit is the threshold variable, and
x it is regressor. In

the model, the observations are divided into two regimes depending on whether the threshold

variable
qit is smaller or larger than the threshold γ . The two regimes are distinguished by

differing regression coefficients

β 1 ,β 2 . He shows that for any given γ , the slope

coefficient β can be estimated by ordinary least squares (OLS).

The key issue of threshold method is how to determine the threshold point γ . A simple method

(Berthelemy & Varoudakis 1996) can be carried out as follows: For each possible threshold value,

denote by
σ^ 2j the sum of the squares of the residuals in regressions, which are estimated over

the two sub samples. j=1 and j=2 , which are defined by the breaking point. If
nj is

the size of each sub sample, the quasi log-likelihood of the data for this threshold level is defined

2 2
QL=−∑ j=1 n j ln( σ^ j )
by . The threshold value is chosen as the value which maximises

QL .

In our estimation model, there are two critical thresholds, the non-performing loan rate and capital

ratio. In specific, by applying Hansen’s threshold method, we determine the thresholds by

estimating the models as follows:

a 0 +a1 DGR ' + a2 CGR ' + a3 OEAGR ' +a 4 NPLGR i , t −1

{
i,t i,t i, t NPLRi ,t ≥ q
+ a5 Dummy i , t + a6 Dummy i , t NPLGR i , t −1
LGRi , t =
a'1 + a'1 DGR +a'2 CGR +a'3 OEAGR +a '4 NPLGR i ,t −1
i, t' i, t' i ,t ' NPLRi ,t <q
+ a'5 Dummy i , t + a'6 Dummy i , t NPLGR i , t −1
,

(5)

12
where q is the threshold level for the non-performing loan, and
Dummyi ,t =1 if the

capital ratio is greater than the effective capital ratio we find for that sample, and zero otherwise.

V. Data

The bank data are collected from BankScope Database by Bureau Van Dijk, which provides in-

depth comprehensive bank statistics from 1998 to 2005. The banks we have included in our study

are commercial banks. Taking deposits and making loans are their most fundamental functions and

they are all oriented to maximise their profits as the primary objective, which are suitable for the

purpose of our study.

The countries / regions we have looked into are as follows:

Region Country Number of Banks

U.S. 1214
Japan 155
3
Asian Crisis Countries Hong Kong 34
Philippines 30
Indonesia 33
Thailand 14
Republic of Korea 18
Western Europe France 39
Eastern Europe Poland 31
Croatia 11
Latvia 16
Romania 10
Serbia and Montenegro 7
Ukraine 20
Czech Republic 8
Bosnia 6
Hungary 13
Slovakia 9
Turkey 12
Table - 1 Sample countries and regions

3
No data available for Malaysia.

13
In our samples, the United States has a most developed financial system with sophisticated risk

management and control mechanism. We can use it as a benchmark to compare its results with

other countries and regions. Japan banks have experienced a prolonged stagnancy since the early

1990s, and the non-performing loan problem lies at the heart. By looking at Japanese banks, we

can find out how the extended non-performing loans affect banks lending behaviour. The Asian

Crisis is still remaining fresh in our mind when the financial systems in Southeast Asian countries

collapsed in 1997. The non-performing loans were mainly the results of over-heated lending in the

economic expansion period prior to the outburst of the financial crisis. Banks heavily burdened by

non-performing loans is one of the most important reasons of the crisis. The time period in our

study is from 1998 to 2005, just after the financial crisis, which enables us to study the lending

behaviour of the banks trying to recovering from the crisis. In the West Europe, we have France as

our sample country. Although France is a developed country as the U.S., it has a quite different

banking system. For the U.S., bank holding companies with larger sizes and wider business scopes

are playing main roles; while, in France, the banking system is characterised by many smaller

savings banks and building societies. As these banks generally do business at smaller scales and

scopes, they may have different attitudes towards risks, and hence, non-performing loans; thus the

lending behaviour may differ as well. We also study the banks in the Eastern European countries.

Most of the countries have experienced dramatic changes in political as well as financial system

since 1990s and some countries have joined or are joining the European Union. Economies in

these countries are growing fairly fast and banks are playing increasing important roles in the

economy. By studying these countries, we may find different pattern for these developing

countries. And the basic data descriptions are as follows:

Franc
U.S. Japan Asian Crisis Eastern Europe
e
13.87
0.41 14.01 5.73 37.21
Asset Growth Rate (%) (31.50
(6.30) (27.639) (21.22) (56.16)
)
1.93 4.11
7.68 1.73 4.05
OEA Growth Rate (%) (39.70 (10.18
(49.13) 24.57 (46.11)
) )
Loan Growth Rate (%) 1.12 -1.51 1.58 0.55 7.76

14
 (15.33
(4.94) (20.79) (13.33) (30.72)
)
-0.48
Deposit Growth Rate 0.76 1.49 0.53 1.53
(13.74
(%) (5.47) (21.43) (10.90) (15.35)
)
14.86 3.74
9.36 -12.81 1.13
NPL Growth Rate (%) (90.82 (35.82
(88.83) (42.19) (91.21)
) )
2.52 3.45
1.83 3.77 0.57
Equity Growth Rate (%) (23.20 (35.36
(45.35) (48.17) (55.41)
) )
0.83 6.90 12.18 8.75 8.88
NPLs to Loans Rate (%)
(2.38) (3.18) (15.71) (9.48) (10.92)
15.50 9.96
22.30 11.95 19.11
Capital Ratio (%) (14.61 (12.29
(17.83) (8.43) (12.11)
) )
Table - 2 Basic Data Description

Table -2 presents the mean values for the mean value for each of the yearly growth rates for the

variables, and the standard errors are presented in the brackets.

In commercial banks, equity is basically composed of common stocks, preferred stocks,

convertible stocks and retained earnings. Equity is the main and most stable components of the

capital which banks own and use for long term. Thus, we use equity as a proxy for the capital.

From the basic data description, we can get a general idea of the situations of banks of different

countries and regions. The United States, as a whole, has much lower non-performing loan levels

than the rest: its average NPL rate is only 0.83% compared to the highest 12.18% for Asian crisis

countries, which implies that commercial banks in the U.S. generally have sound risk managing

and controlling mechanism. While, Asian Crisis countries, are severely hit by the non-performing

loans problems and recovering from the financial crisis during the period we are studying. We can

expect different lending behaviours for them. The Eastern European banks have experienced fast

growth for asset and loans; however, the larger standard errors indicate high volatility and

imbalance of the growth in this sample, which is typical for many developing countries. Thus, we

can anticipate that lending behaviour of the developing countries will be different from those of

15
developed countries. Although the NPL rates are quite high for French banks, 8.75%, banks are

experiencing decreases in non-performing loans with the mean rate -12.81%, which suggests that

loan quality is improving in France in the period of study. Again, we may expect different lending

behaviours of banks with improved loan quality against those of the banks with worsening loan

quality.

For the risk-based capital ratio, all the samples have the mean values above the 8% required

capital ratio according to Basle Accord II. On one hand, it indicates that banks have put emphasis

on reserving adequate amount of capital to improve their risk position. Compared with other

countries, the Asian crisis countries have an especially high capital ratio with the mean value of

22.30%. It shows that banks in these countries are improving their asset quality and capital

adequacy since the crisis. However, we can argue that the increased capital ratio is basically the

result of government rescuing actions. In order to save banks from disasters, governments of these

countries have taken various measures, such as re-capitalising banks using public funds,

establishing special vehicles to move problem loans out of banks, etc.. By these administrative

measurements, the capital ratios are improved, while banks themselves have made fewer efforts.

VI. The Empirical Results and Analysis

U.S.

Before estimating the threshold for each sample, we first carry out the Hausman test to make sure

that the sample has specific individual effects so that we can use the Hansen threshold technique.

The test assumes the

H0 hypothesis is that the difference of the coefficients obtained from the

random effects and the fixed effects is not systematic; in other words, it assumes that
ci is

orthogonal to
x i,t , i.e., the random effects. Thus, if we can reject the
H0 hypothesis, we can

16
use the fixed effects method to determine the threshold. The Hausman test rejects the
H0

hypothesis and shows the evidence of the fixed effects. The major differences are in the other asset

growth rate, deposit growth rate and equity growth rate:

(b) (B) (b-B)

Sqrt (diag (V_b – V_B))
fix ran Difference
- -
OEAGR(t) 0.0310 0.0325
0.0968 0.1278
EGR(t) 0.1662 0.1057 0.0605 0.0658
DGR(t) 0.3338 0.2569 0.0769 0.0728
- -
NPLGR(t-1) 0.0001 0.0061
0.0073 0.0074
Test: H0: difference in coefficients not systematic
Chi2 (4) = (b – B)’[(V_b – V_B) ^ (-1)](b – B) = 0.00
Prob > chi2 = 1.0000
Table – 3 Hausman Test for U.S. sample

By using the threshold method, we find that the non-performing loan threshold for the U.S. sample

is NPL rate 0.6% and the effective capital ratio is 14.9%, as at such levels that the value of QL

is minimised. The regression result is as follows4:

Dependent Variable: LGR (t)

Coefficient Std. Error Coefficient Std. Error
Variable
NPL Rate >= 0.6% NPL rate < 0.6%
5
Constant 0.9010** 0.3560 1.3371*** 0.2228
DGR(t) 0.2782*** 0.0203 0.0816*** 0.0193
OEAGR(t) -0.1008*** 0.0052 -0.1494*** 0.0050
EGR(t) 0.0546*** 0.0112 -0.0003 0.0096
NPLGR(t-1) -0.0089** 0.0035 -0.0026 0.0021
Dm -2.8329*** 1.0287 1.0660* 0.6117
Dm*NPLGR(t-1) -0.0018 0.0060 -0.0058 0.0040
No. of Obs. 2570 3462
No. of groups 807 988
R-sq: Within = 0.2754 Within = 0.2758
Between = 0.1163 Between = 0.3650

4
5
*** indicates 1% significance, ** indicates 5% significance, and * indicates 10% significance.

17
 Overall = 0.2388 Overall = 0.2861
Wald chi2 117.57 235.79
Prob > Chi2 0.0000 0.0000
Table – 4 U.S. sample results

Generally, when banks have non-performing loans less than the threshold, the loan growth rate is

higher as suggested by the constants. Both of the cases have significant constants. When banks

have non-performing loan rate less than 0.6%, the constant is 1.3371; and when banks have non-

performing loan rate greater than 0.6%, the constant is lower at 0.9010. The results indicate that

banks with better risk position (lower non-performing loans) have higher inclination to extend

loans. However, when banks have higher non-performing loans in the portfolio, they may be more

concerned about the risky, and hence, have lower tendency to increase loans.

When we look at the coefficient associated with the non-performing loan growth rates, we find

statistically significant negative coefficient (-0.0089) when banks have non-performing loans more

than the threshold level; on the other hand, when banks have non-performing loans less than the

threshold level, the coefficient is not statistically significant. The results are consistent with our

initial hypothesis that higher non-performing loans than the threshold will hinder banks’ creation

of new loans. By decreasing loans, banks can reduce the possibility of new non-performing loans

generated from new lending. However, when banks’ non-performing loans are below the threshold

level, there is lower tendency for banks to reduce lending as the non-performing loans are under

banks’ acceptable level.

We find negative coefficients associated with other earning assets growth rates for both cases.

Other earning assets are substitute for loan given a fixed amount of assets for banks to use.

Although making loans are still the most important function for commercial banks; with the

innovation of financial instruments and less activity restriction, banks are expanding their business

scope other than lending in order to obtain higher profitability and larger market power. Especially

when banks have less risk constraints, i.e., less non-performing loans, they have more freedom to

engage in various businesses, which is indicated by a higher degree -0.1494 compared with

-0.1008.

18
When we look into the effect of capital adequacy, we find that higher capital ratios which are

above the effective capital ratio (14.9%) have different effects for the two cases. When banks have

non-performing loan rate below the threshold, the capital adequacy helps to accelerate the creation

of new loans, as the coefficients in front of dm are positive and statistically significant

(1.0660). Capital is the safety cushion for credit risk; as a result, higher capital level gives banks

more desire to increase loans. On the other hand, when the non-performing loan rate is higher than

the threshold level, the capital dummy has a statistically significant negative coefficient -2.8329. It

shows that capital plays a stabilising role to reduce the lending growth rate when banks have more

credit risk in the portfolio.

As a whole, we have found some evidences of our hypothesis that non-performing loans reduce

the incentive to increase lending when banks have non-performing loan rate above the threshold

level in the banks in U.S..

Japan

The Hausman test shows statistically significant evidence of fixed effects and the great difference

exits in deposit growth rate:

(b) (B) (b-B)

Sqrt (diag (V_b – V_B))
fix ran Difference
- -
OEAGR(t) -0.0004 0.0747
0.2023 0.2019
- -
EGR(t) -0.0004 0.1400
0.1236 0.1231
DGR(t) 1.0338 1.0358 -0.0020 0.1166
- -
NPLGR(t-1) 0.00001 0.0044
0.0183 0.0183
Test: H0: difference in coefficients not systematic
Chi2 (4) = (b – B)’[(V_b – V_B) ^ (-1)](b – B) = 0.00
Prob > chi2 = 1.0000
Table – 5 Hausman Test for Japanese sample

For the sample of Japanese banks, we find the threshold of NPL rate is 6.9%, much higher than

19
that of the U.S., which can be justified by a much higher average NPL rate of the raw data. In

additional, because Japanese banks have been suffering from banking problem for quite a long

period ever since the beginning the early 1990s, their ability to adjust lending objective according

to non-performing loan situations may be impaired. Thus, the threshold for them to change their

lending behaviour is quite high. And the effective capital ratio is 11.4%. The result is as follows:

Dependent Variable: LGR (t)

Coefficient Std. Error Coefficient Std. Error
Variable
NPL Rate >= 6.9% NPL rate < 6.9%
Constant -0.5232*** 0.2003 0.0265 0.1376
DGR(t) 0.0809 0.0519 0.0223 0.0419
OEAGR(t) -0.2490*** 0.0150 -0.3719*** 0.0132
EGR(t) 0.0174*** 0.0062 0.0175* 0.0099
NPLGR(t-1) 0.0021 0.0044 0.0031 0.0038
Dm 2.0530** 0.9937 0.5814 0.3718
Dm*NPLGR(t-1) -0.0819*** 0.0132 -0.0061 0.0072
No. of Obs. 392 409
No. of groups 119 126
R-sq: Within = 0.5755 Within = 0.7526
Between = 0.2914 Between = 0.7015
Overall = 0.4129 Overall = 0.7407
Wald chi2 14.03 23.49
Prob > Chi2 0.0507 0.0014
Table – 6 Results for Japanese sample

The constant is negative when banks have non-performing loan rate above the threshold level and

is statistically significant. It is consistent with our expectation that banks with poorer loan quality

will reduce the lending to consolidate their asset quality.

There is little evidence that the growth of non-performing loans will cause banks in Japan to

reduce their lending in either cases as neither of the coefficients associated with NPL growth is

statistically significant, which is to the contrary of the hypothesis that non-performing loans have

negative effect on bank’s lending. In fact, the results show that non-performing loan growth rate

does not significantly affect the lending growth rate.

20
When we look at the other earning asset growth, we find that there are negative coefficients which

are statistically significant for both cases. The negative signs suggest that there are obvious

substitute effects that with the growth of total assets, banks choose to increase other earning assets

rather than extending loans. However, such substitute effect is much larger when banks have non-

performing loans less than the threshold. It shows that banks have less intention to replace loans

by other earning assets when they have poorer loan quality. It may be explained that with bad asset

quality, banks have less ability to engage in other kind of activity as their credit ranking is not high

enough to support them; and what’s more, poorer asset quality may impose constraints for them to

use funds freely.

For banks having non-performing loan rate higher than the threshold level, capital adequacy

increases banks’ incentive to increasing lending. Although the coefficient for Dm⋅NPLGR is

negative (-0.0819), the coefficient for Dm is much larger (2.0530). As a result, the combined

effects shows that banks with more capital has higher tendency to increase lending as they are in a

safer capital position than otherwise.

To summaries, from the result of Japanese banks, we have found mixed result as to the effects of

non-performing loans and capital adequacy on banks’ lending behaviour. Although we haven’t

found direct impact that non-performing loans reduce lending, however, when banks face higher

non-performing loan level and lower capital ratio, they have the least incentive to increase their

lending.

We haven’t found clear evidence that non-performing loans have reduction effect on lending.

Alternatively, according to the view of “debt over-hanging” view, the prolonged distress of

banking problem in Japan is caused by the lack of demand rather than the supply. Banks still

increase their lending despite of their NPL problem; however, with the long-lasing stagnation of

the economy, companies have less capacity and incentive to borrow money as they have less

ability to make profit to pay the loans.

21
South-Eastern Asian Financial Crisis Countries

The Hausman test shows great difference in all the variables, which strongly suggests that we can

use the fixed effects estimation:

(b) (B) (b-B)

Sqrt (diag (V_b – V_B))
fix ran Difference
-
OEAGR(t) 3.8315 4.5032 22.1299
0.6716
-
EGR(t) 0.6058 -1.8054 6.1087
1.1996
- -
DGR(t) -0.6989 12.4331
1.6830 0.9841
- -
NPLGR(t-1) -0.0695 0.2784
0.0726 0.0031
Test: H0: difference in coefficients not systematic
Chi2 (4) = (b – B)’[(V_b – V_B) ^ (-1)](b – B) = 0.28
Prob > chi2 = 0.9912
Table – 7 Hausman Test for Asian sample

The threshold for Asian Crisis countries is 5.6% and the effective capital ratio is 13.5%. The

results are:

Dependent Variable: LGR (t)

Coefficient Std. Error Coefficient Std. Error
Variable
NPL Rate >= 5.6% NPL rate < 5.6%
Constant 4.5449** 2.1979 4.2506* 2.5045
DGR(t) -0.0310 0.1127 0.1595* 0.0947
OEAGR(t) -0.2879*** 0.0708 -0.2493*** 0.0717
EGR(t) 0.0593** 0.0278 0.0561 0.0592
NPLGR(t-1) 0.0078 0.0226 -0.0425 0.0333
Dm -2.7929 2.5710 1.3652 2.4936
Dm*NPLGR(t-1) -0.0315 0.0253 0.0237 0.0366
No. of Obs. 297 253
No. of groups 86 74
R-sq: Within = 0.4002 Within = 0.2869
Between = 0.3082 Between = 0.0765
Overall = 0.3548 Overall = 0.1813
Wald chi2 46.88 25.15

22
 Prob > Chi2 0.0000 0.0003
Table – 8 Results for Asian sample

Although this sample has the highest mean value of NPLs to loan rate, the threshold is not

relatively high compared with other samples. Most of the banks in this sample were severely

blown by the financial crisis and have been suffered from a lot of difficulty. As a result, they can

tolerate only low level of risk, and changes in the level of non-performing loans will greatly affect

their lending behaviours.

Different from other samples, the results of the constant terms show quite different patterns. When

banks face non-performing loan rate higher than the threshold, they generally have a larger

constant (4.5449) than the constant (4.2506) when they face non-performing loan rate lower than

the threshold, which suggest that “bad” banks increase lending more aggressively than “good”

banks.

We haven’t found statistically significant coefficients associated with non-performing loan growth

rate in either case, which means that although asset quality is a serious problem for banks in this

sample, banks’ lending behaviour are not affected greatly by the large amount of non-performing

loans. After the crisis, governments in South-East Asian countries have taken many rescue

activities such as injection of huge amount of government capital into problematic banks, moving

non-performing loans out of banks, and setting up special government organisations to deal with

non-performing loans, etc.. These activities have given banks disincentive to be worried about

their loan qualities as banks are hoping governments to take action to solve their non-performing

loans. Meanwhile, governments encourage banks to lend more to firms in difficulty, hoping that

more funds can help them get out of the difficulty and make profits to pay back loans. However, in

an economic recession as it was after the Asian Crisis, it is quite hard for firms to make profits

even with enough financing because of the decline in aggregate demand. Thus, it would lead to

more non-performing loans and ended up as a vicious circle. This can to some extent explain why

bank recovery is very slow in these countries after the crisis. By the end of 2005, the average non-

performing loan rate for Thailand, Korea, the Philippines and Indonesia were still around 10%.

23
The same reasons can be used to explain that equity growth rate have positive effect on loan

growth rate (0.0593) when banks have non-performing loan above the threshold. With the growth

of capital which is mainly from government injection, banks use these capitals to extend more

loans according to government policies.

One of the main sources of the capital for the banks in this sample during this period is from

government injection, which doesn’t need banks’ own efforts. Thus, banks are not constrained by

the capital requirements. Because of the government intervention, the improved capital adequacy

doesn’t indicate good bank internal risk management; thus, the capital dummy doesn’t show any

significant impact on banks’ lending behaviour.

The results are not consistent with that of Agung et. al (2001), who support the Credit Crunch

view. But our results support that of Ghosh and Ghosh (1999), who investigate the Credit Crunch

at the aggregate level for Indonesia, Korea and Thailand during 1997 – 98, and find little evidence

of credit crunch. They attribute the decline of credit in these countries mainly to insufficient

demands, not supply side reasons.

France

The Hausman test again shows the difference in deposit growth rate and suggests the fixed effects

model:

(b) (B) (b-B)

Sqrt (diag (V_b – V_B))
fix ran Difference
- -
OEAGR(t) 0.0025 0.0024
0.0096 0.0120
EGR(t) 0.0421 0.0487 -0.0066 0.0080
- -
DGR(t) 0.0476 0.0546
0.1383 0.1859
NPLGR(t-1) 0.0402 0.0353 0.0048 0.0114
Test: H0: difference in coefficients not systematic
Chi2 (4) = (b – B)’[(V_b – V_B) ^ (-1)](b – B) = 1.70
Prob > chi2 = 0.7907

24
 Table – 9 Hausman Test for French Sample

We have found that threshold for French banks is 5.75% of the NPL rate, and the effective capital

ratio is 9.8%. The regression results are shown below:

Dependent Variable: LGR (t)

Coefficient Std. Error Coefficient Std. Error
Variable
NPL Rate >= 5.75% NPL rate < 5.75%
Constant -0.6065 3.3826 7.9287 8.1714
DGR(t) -0.2394* 0.1324 0.0787 0.1201
OEAGR(t) -0.2623*** 0.0695 -0.0086* 0.0047
EGR(t) 0.0245 0.0252 0.2077** 0.0833
NPLGR(t-1) 0.0164 0.0495 0.5551 0.4644
Dm 3.2989 4.7084 -7.1216 8.6426
Dm*NPLGR(t-1) 0.0519 0.0624 -0.5672 0.4652
No. of Obs. 116 47
No. of groups 32 15
R-sq: Within = 0.2300 Within = 0.3462
Between = 0.5616 Between = 0.0356
Overall = 0.3021 Overall = 0.1457
Wald chi2 23.94 2.29
Prob > Chi2 0.0012 0.0655
Table – 10 Results for French sample

Both as developed countries, for the sample of French banks, we have a threshold much higher

than that of the U.S. sample. Compared with U.S, France is typically composed of smaller banks

such as savings banks and building societies, while U.S. has a large proportion of large banks and

bank holding companies. Bank holding companies and large commercial banks generally have

more alternative ways to invest in other than making loans. These alternatives make it possible for

banks to switch their business when the non-performing loans go up; thus, the lending behaviour

changes at a relatively low threshold. For smaller banks, loan is the main and the most important

asset. It is not very easy for them to change their behaviour of asset composition until a relatively

high non-performing loan rate.

The non-performing loan growth doesn’t have statistically significant effect on loan growth rate.

25
During the period we are observing, we find that French sample is experiencing high rate of

decreasing in non-performing loan with the mean rate -12.81%, while the other samples have

positive non-performing loan growth rate. Such improvement may be the result of better internal

risk management. With a better portfolio quality, there is less incentive for bank to reduce lending.

Rather than reducing loans, which enables banks to reduce the base for the new generation of non-

performing loans, banks use better risk management to control non-performing loan while keeping

their lending growth at the same time.

However, we still can detect from our results that banks with higher non-performing loan rate

have less incentive to increase lending than banks with lower non-performing loan rate. When

banks have non-performing loan rate higher than the threshold, facing with deposit growth, banks

reduce the lending, as we have negative coefficient -0.2394, which is statistically significant. The

coefficients suggest that with more funds generating from deposits which are available for banks

to use, banks intend to use the funds in other ways rather than to increase the loans when they

have many non-performing loans in the loan portfolio. Such evidence can also be found from the

coefficients associated with the other earning asset growth rates. We find that the coefficients are

negative and are statistically significant for both of the cases; however, the coefficient is much

higher -0.2623 when banks have non-performing loan rate higher than the threshold, compared

with -0.0086 when banks have non-performing loan rate below the threshold. When banks have

higher non-performing loans, they switch more of their funds to other earning assets; and such

tendency is lower when they have lower non-performing loans.

When the non-performing loan rate is less than the threshold, growth in capital gives banks

incentive to increase lending as the coefficient is 0.2077, which is statistically significant. When

banks have less non-performing loans which implies lower credit risk, there is less constraints for

banks to use the capital. Thus, they extend more loans in order to gain more profits. On the other

hand, when banks have higher non-performing loans, banks have to retain capital to achieve the

capital requirement and to improve risk position.

We don’t find statistically significant evidence whether the capital adequacy helps to increase

26
lending. It suggests that there is less incentive for banks to increase lending with adequate capital.

As French banks are mainly small banks which are more sensitive to risk and have fewer sources

for raising capital, capital adequacy mainly acts as a safety cushion rather than giving banks

incentive to increase loans. Increasing loans will lead to larger risky asset base; as a result, banks

main again face pressure to maintain or raise capital ratio.

Eastern European Countries

Finally, the Hausman test again confirms the fixed effects model:

(b) (B) (b-B)

Sqrt (diag (V_b – V_B))
ran fix Difference
- -
OEAGR(t) 0.2126 0.3172
0.1719 0.3845
EGR(t) 0.5296 0.0388 0.4907 4.2952
-
DGR(t) 2.6980 6.2865 15.1077
3.5885
-
NPLGR(t-1) 0.0485 -0.0954 0.1280
0.0469
Test: H0: difference in coefficients not systematic
Chi2 (4) = (b – B)’[(V_b – V_B) ^ (-1)](b – B) = 0.80
Prob > chi2 = 0.9383
Table – 11 Hausman Test for Eastern European sample

The threshold for the Eastern European sample is 4.3% of the NPL rate and the effective capital

ratio is 9.2%. The results are:

Dependent Variable: LGR (t)

Coefficient Std. Error Coefficient Std. Error
Variable
NPL Rate >= 5.6% NPL rate < 5.6%
Constant 13.4676** 6.8930 -31.0899* 18.3848
DGR(t) 0.1012 0.1179 -0.1454 0.1630
OEAGR(t) -0.3474*** 0.0277 -0.1172* 0.0715
EGR(t) 0.1033*** 0.0256 0.1483** 0.0736
NPLGR(t-1) 0.0595 0.0996 0.4909* 0.2553
Dm -8.3046 7.0490 39.7979** 18.3526
Dm*NPLGR(t-1) -0.0799 0.1003 -0.4761* 0.2553
No. of Obs. 435 265

27
 No. of groups 165 112
R-sq: Within = 0.4105 Within = 0.2022
Between = 0.1871 Between = 0.1036
Overall = 0.3061 Overall = 0.1356
Wald chi2 50.11 51.17
Prob > Chi2 0.0000 0.0000
Table – 12 Results for Eastern European sample

Different from other samples, we find quite different patterns of banks’ lending behaviour for East

European countries.

Generally speaking, the results show that banks with non-performing loans higher than the

threshold are more aggressive in lending behaviour. The constant term has a large positive value

13.4676, which is statistically significant within 5% level. It suggests that banks with higher non-

performing loans are less risk aversion and increase lending quickly. Again, the results show that

banks increase their lending quickly with increase in capital (0.1033) when banks have non-

performing loan higher than the threshold, and both of the coefficients are statistically significant.

The original data shows that loans are growing much more quickly in this sample. And the mean

non-performing loan to asset rate is second highest and is only lower than that of the South-East

Asian countries sample. In developing countries, banks typically increase their lending

aggressively in order to increase their business and obtain more market power. However, these

banks are always less sophisticated in risk control; as a result, such fast growth in lending will end

up in a vicious cycle that fast growth of non-performing loans is unavoidably accompanied with

high speed of loan growth. For the case of banks having non-performing loan rate less than the

threshold level, the constant is significantly negative (- 31.0899), which suggests that these banks

are much less aggressive in lending. This, in turn, may suggest that the low level of non-

performing loans are the result of prudential lending behaviour, and create a health cycle of

lending behaviour.

In addition, we find that the coefficient associated with other earning asset growth rate is a large

negative figure (-0.3474), which is statistically significant for banks with non-performing loans

higher than the threshold level. It suggests that other earning assets have a large substitute effect

28
on lending. At first glance, it seems to be inconsistent with the speedy growth of lending.

However, when we look at raw data, it shows that the asset growth rate is extraordinary high

40.39% and the loan growth rate only takes a small portion 9.59%. Thus, in this case, rather than

substituting loans out of total assets, other earning assets actually are the complementary for loans.

It indicates that banks are expanding their business widely with the fast development of their

financial markets. With the economic growth and evolution of financial system, banks are trying

to expand their business scope to play more roles in the financial sector rather than just focusing

solely on extending credits.

On the contrary, when banks have non-performing loans lower than the threshold, they are less

regressive in increasing lending as suggested by the estimated coefficients. However, when non-

performing loan rates are under the threshold level, non-performing loans have positive impacts

banks’ lending behaviour with a statistically significant positive coefficient 0.4909. It suggests that

banks may still increase their loans as the generation of non-performing loans is the natural result

of lending, especially for banks in the expansionary stage.

The results also suggest that the capital adequacy plays an important role in this sample. Although

this sample have a relatively high average capital ratio (19.11%), the effective capital ratio rate is

fairly low (9.2%), suggesting that that a relative low capital ratio can greatly affect banks’ lending

behaviour. When banks have non-performing loans less than the threshold and have capital

adequacy ratio above the effective value, i.e., the best case in all situations, they have very high

incentive to increase their lending, which is suggested by the very high positive coefficient

39.7979 associated with the capital dummy, and this coefficient is statistically significant.

Comparing the results from the other samples, the coefficient in this sample are of much larger

scale, which implies that banks in Eastern European sample are much more sensitive to capital

adequacy and have great aspiration to increase lending. Banks in expansionary stage are quite

speculative in lending behaviour. Once the capital ratio reaches slightly above the Basle Accord II

required 8% rate, they start to speed their lending. Such phenomenon is different from the other

samples. For developed countries such as the U.S., banks have less incentive to speed expansion

29
and keep the capital adequacy ratio high. Even they have enough capital, they don’t increase the

lending that much. For France and Japan, the capital adequacy doesn’t affect their lending

behaviour significantly, as they may regard the capital as safety cushion.

VII. Conclusion

We have studied how the non-performing loans affect banks’ lending behaviour in different

countries and regions. Our results suggest that non-performing loans have non-linear effects. To

our expectation, we have detected some evidences that higher level of non-performing loans

reduces banks’ aspiration to increase lending. However, countries with different situations have

shown different locations of the thresholds.

Evidence is less clear for South-East Asian Crisis countries as a result of distorted financial system

and government intervention. We have found that non-performing loans don’t deter banks’ risky

lending. For Japanese banks, we haven't found the negative influence of non-performing loans on

banks’ lending. The prolonged economic stagnation distorts the generally economic performance.

As a result, the lack of demand for loans may contribute significantly to the extended bank

problem in Japan. For developing countries such as Eastern European countries, banks are

experiencing fast growth, which makes it more difficulty to predict their lending behaviour.

We have also found that the risk-based capital ratio has played a significant role to restrict banks’

risky lending as its initial intention by BIS. Generally speaking, higher capital ratios give more

incentive to increase lending than lower capital ratios when banks have less credit risk in the

portfolio. Theses tendencies are most evident for banks in developed countries such as the U.S., as

it has a well-developed financial systems and sophisticate risk management mechanism. On the

other hand, implementation of risk-based capital requirement can also help to prevent risk-taking

behaviour by calming down over-heated lending behaviour for high risk banks.

30
 VIII. Reference
Agung, Juda et.al, “credit Crunch in Indonesia in the Aftermath if the Crisis facts, Causes and

Policy Implications”, directorate of Economic Research and Monetary Policy, Bank of Indonesia,

March 2001

Altman, Edward I. & Anthony Sauders, “An Analysis and Critique of the BIS Proposal on

Capital Adequacy and Ratings”, Journal of Banking and finance, 2001, 25-46

Agenor, P.R., J. Aizenman, A.Hoffmaister, “The Credit Crunch in East Asia: What Can Bank

Excess Liquid Assets Tell Us?”, NBER Working Paper 7915, Ocotber 2000

Arrow, K, “The Theory of Risk Aversion”, In Essays in the Theory of Risk Bearing, New York:

Markham, 1971, 90-120

Bank of Finland Institute for Economics in Transition, “BOFIT China Review Yearbook

2004”, 2005, www.bof.fi/bofit

Bank of Japan, “Japan’s Nonperforming Loan Problem”, Quarterly Bulletin November 2002;

Bank of Japan, “Bank of Japan Annual Review 2004”

Barr, R. L. Seiford & T. Siems, “Forcasing Banking Failure: A Non-Parametric Frontier

stimation Approach”, Researches Economiques de Lovain (60), 1994, 417-429

Bartel, John & Yiping Huang, “Dealing with the Bad Loans of the Chinese Banks”, Columbia

University, APEC Study Centre Discussion Paper Series, July 2000

Berger, Allen N. & Rpbert De Young, “Problem Loans and Cost of efficiency in Commercial

Banks”, Journal of Banking and Finance, Vol. 21, 1997

Berger, Allen, & Udell, Grogory F., “Did Risk-Based Capital Allocate Bank Credit and cause A

“Credit Crunch” in the United-Sates?’, Journal of Money, Credit and Banking, Vol.26, No.3, Part

2: Federal Credit Allocation: Thoery, Evidence, and History, 1994, 585-628

Batrh, James R., Gerard Caprio Jr., & Ross Levine, “Banking Systems Around the Globe: Do

Regulation and Ownership Affect Performance and Stability?”, 2001

http://econ.worldbank.org/docs/1079.pdf

Bernanke, Bens, Cara S. Lown & Benjamin M.Friedman, “The Credit Crunch”, Brookings

Papers on Econmic Activity, Vol.1991, No.2 ,1991, 205-247;

Berthelemy, J.C. & Varoudakis, A., “Economic Growth, Convergency Clubs, and the Role of

Financial Development”, Oxford Economic Papers, New Series, Vol.48 No.2, 1996, 300-328

31
Besanko, David & George Kanatas, “Regulation of Bank Capital: Do Capital Standards Promote

Bank Safety?” Journal of Financial Intermediation 5, 1996, 160-183

Borensztein, Eduardo & Jong-wha Lee, ‘Financial Crisis and the Credit Crunch in Korea:

Evidence from Firm-Level Data’, IMF working Paper WP/00/25, 2000

Calem, Paul & Rafeal Rob. “The Impact of Capital-Based Regulation on Bank Risk-Taking”,

Journal of Financial Intermediation 8,1999, 317-352

Central Bank of Malaysia, “Bank Negara Malaysia Annual Report 2004”, 23 March 2005

Chan-Lau, Jorge A. & Zhao Hui chen, “Financial Crisis and Credit Crunch as a result of

Inefficient Financial Intermediation – with Reference to the Asian Financial Crisis”, 1998;

Chiang, Alpha.C, “Funamental Methods of Mathematical Economics” Third Edition, McGraw-

Hill, 1984

Chen, Johnny P., “Non-performing Loan Securitization in the people’s republic of China’,

Stanford University, May 2004

Demirguc-Kunt, Asli & Enrica Detragiache, “The Determinants of Banking Crises in

Developing and Developed Countries”, IMF Staff Paper 45 (1), 1998, 81-109

Demirguc-Kunt, Asli, Enrica Detragiache, & Poonam Gupta, “Inside the Crisis: An Empirical

Analysis of Banking System in distress”, IMF Working Paper WP/00/56, October 2000

Domic, Ilker & Giovanni Ferri, “The Real Impact of Financial Shocks: Evidence from Korea”,

The World Bank, Policy Research Working Paper Series No.2010, October 1998

Dwight, Lawrence,”The Role of Non-Performing Loans in china: A Public Finance Approach”,

2004

Ernst & Young, “Nonperforming Loan Report: Asia 2002”

Ernst & Young, “2003 review Non-Performing Loans in China”

Ernst & Young, “Global Nonperforming Loan Report 2004”

Furlong, Frederick T. & Keeley, Michael C., “Bank Capital regulation and Asset Risk”,

Economics Review Number 2, Federal Reserve Bank of San Francisco, spring 1987, 20-40

Ghosh, Swati R & Atish R. Ghosh , “East Asia in the Aftermath: Was There a Crunch”’, IMF

working Paper WP/99/38, March 1999

Gonzalez, Francisco, “Bank Regulation and Risk-Taking Incentive: An International Comparison

of Bank Risk”, http://207.36.165.114/Dublin/fgonzaleFMA03.pdf (2002)

32
Hansen, Bruce E., “Threshold Effects in Non-Dynamic Panels: Estimation, Testing, and

Inference”, Journal of Econometrics 93, 1999, 345-368, 1999

Hansen, Bruce E.,”Sample Splitting and Threshold Estimation”, Econometrica, Vol.68, No.3,

2000,575-603

Hancock, Diana & James A Wilcox, “The ‘Credit Crunch’ and the Availability of Credit to Small

Business”, Journal of Banking and Finance 22 ,1998, 983-1014

Hellmann, Thomas F., Kevin C. Murdock, and Joseph E.Stiglitz, “Liberalization, Moral

Hazard in Banking, and Prudential Regulation: Are Capital Requirements Enough?” The

American Economics Review, Vol.90, No.1,Mar., 2000, 147-165

Hoffmaister, Alexander W. “The Credit Crunch in East Asia: What can Banks Excess Liquid

Assets Tell Us?”, Journal of International Money and Finance 23,2004, 27-47

Jacques, Kevin, & Peter Nigro, “Risk-Based Captial, Portfolio Risk and Bank Capital: A

Simultaneous Equation Approach”, Journal of Economics and Business 49,1997, 533-547

James, Christopher, “Some Evidence on the Uniqueness of Bank Loans”, Journal of Financial

Economics 19,1987, 217-236

James, John A., “Portfolio Selection with an Imperfectly Competitive Asset Market”, The

Journal of Financial and Quantitative Analysis, Vol.11, No.5, Dec. 1976, 831-846

Jensen, Michael C., & William H. Meckling, “Theory of the Firm: Managerial Behaviour,

Agency Costs and Ownership Structure”, Journal of financial Economics Vol.3, Issue 4., Oct.,

1976, 305-360

Kim, Daesik & Santomero, Anthony M., “Risk in Banking and Captial Regulation”, The

Journal of Finance, Vol.43, No.5, Dec., 1988, 1219-1233

Kim, dong Wang, Young Soo Lee & Kyung Sung Park, ‘Credit Crunch and Shocks to Firms”

Korean Experience under the Asian Financial Crisis’, Emerging Market Review 3, 2002, 195-210;

Koehn, Michael & Santomero, Anthony M., “Regulation of Bank Capital and Portfolio Risk”,

The Journal of Finance, Vol.35, No.5, Dec., 1980, 1235-1244

Kotikulo, Oramone, Runchana Pongsaparn & Pichit Patrawimolpon, “Mobilizing Domestic

Finance for Development and Credit Crunch Causes and Cures: Some Evidence from Thailand

and other East Asian Countries”

Krueger, Anne O., & Aaron Tornell, “The Role of Bank Restructuring in the Recovery from

33
Crises: Mexico 1996 – 98”, NBER Working Paper No. 7042, 1999

Lu, Ding, Shandre M. Thangavelu & Qing Hu, “The Link Between Bank Behavior and Non-

Performing Loans in China”, National University of Singapore, department of economics working

Paper No.0108, May 2001

Merton, Robert C., “An Analytic Derivation of the Efficient Portfolio Frontier”, The Journal of

Financial and Quantitative Analysis, Vol.7, No.4 (Sep., 1972), 1851-1872

Mitsuru Taniuch, “Recent Development in Japan’s Financial Sector: Bad Loans and Financial

Deregulation”; Journal of Asian Economics, Vol.8 No.2, 1997, 225-244

Montgomery, Heather, “The Effect of the Basel Accord on Bank Portfolios in Japan”, Journal of

the Japanese and International Economies, 2004

Park, Se-Hark, ‘Bas Loans and Their Impacts on the Japanese economy: Conceptual and

Practical Issues, and Policy Options’, Discussion Paper series A No. 439, The Institute of

Economic Research, Hitotsubashi University, January 2003

Revankar, Nagesh & Naoyuki Yoshino “An Empirical Analysis of Japanese Banking Behaviour

in a Period of Financial Instability”;

Santomero, Anthony M., “Modelling the Banking Firm: A Survey”, Journal of Money, Credit

and Banking, Vol.16, No.4, Part 2: Bank Market Studies, Nov. 1984, 576-602

Shih, Victor, “Dealing with Non-Performing loans: Political Constraints and Financial Policies in

China”, The China Quaterly 180, December 2004, 922-944

Stolz, Stephanie, “The Relationship Between Bank Capital, Risk-Taking, and Capital Regulation:

A Review of the Literature”, Kiel Working Paper No.1105, Kiel Institute for World Economics,

Feb., 2002

Suzuki, Kenji, ‘Rethinking Japan’s Bad Loan Management: Implications from a Comparison with

the Swedish Case’;

Taggart, Robert A, & Stuart I. Greenbaum, “Bank Capital and Public Regulation”, Journal of

Money, Credit and Banking, Vol.10, No.2, May, 1978, 158-169

Westermann, Frand, “The credit Crunch: A Comparison of Germany and Japan”, CSEifo Forum

2003 No.1, 2003

Ye, Yanfei, “The Way of Dealing with Non-performing Loans and Its effects on Maco-statistics in

China”, National Bureau of Statistics, China, 2003

34