Determinants of Non-Performing
Loans in the European Union
Yumeng Yuan S2439085
University of Groningen Faculty of Economic and Business
Supervisor: dr. R.M. van Dalen June 2014
ABSTRACT
The purpose of this thesis is to identify determinants of non-performing loans of banking systems of the European countries from 2000 to 2011. The main variables are macroeconomic factors and bank-level factors. According to dynamic panel estimates, real GDP growth, inflation (GDP deflator), exchange rate, share prices and bank return on equity have a significantly negative effect on non-performing loans. The non-performing loans in last year and the financial crisis significantly positive influence non-performing loans in this year. However, there is no significant relationship between bank capital-to-assets ratio and non-performing loans.
1. Introduction
Banks are playing a vital role in the financial system of today’s market economy. As the financial system affects almost all industries in the world, a stable development of banks is very important for economic growth. However, non-performing loans (NPLs) are a serious problem for banks all over the world.
Especially, after the Subprime Crisis in the U.S., the amount of NPLs increased rapidly and many famous banks in the U.S. went bankrupt. This also influenced the economic development in the European Union. That is to say, the European debt crisis broke out in the European Union. An increasing number of firms could not repay the loans because of the economic depression. This caused an increase in the amounts of NPLs of banks. In other words, the development of almost industries, especially the banking system，was not as good as before the crisis. It is therefore interesting in examine the determinants of NPLs.
According to data of NPLs from The World Bank and Federal Reserve Economic Data, Fig.1 in Appendix A shows the change of average NPLs to total gross loans in European countries from 2000 to 2011. The broken line graph shows that NPLs to total gross loans decreased from 2000 and touched bottom at 2007.Then, it increased rapidly after 2007 because of the crisis and global economic recession. Moreover, there is no downtrend for NPLs to total gross loans in 2011. That is to say, the crisis is far from over.
Council (GCC) countries1 from 1995 to 2008 respectively. Klein (2013) treats both macroeconomic variables and bank-level variables as determinants of NPLs in Central, Eastern and South- Eastern Europe (CESEE) from 1998 to 2011. Besides, Louzis et al. (2011) select macroeconomic and bank-specific variables as independent variables and investigate determinants of NPLs only in Greece from 2003 to 2009.
This thesis continue the previous research by empirically studying the determinants of NPLs in all European countries, except Cyprus and Malta, from the first year of 21st century to 2011. More specifically, NPLs are explained by both macroeconomic variables and bank-level variables. The dataset consist of almost all European countries from 2000 to 2011. Moreover, this study makes a comparison of NPLs between pre-crisis and during crisis. The main estimation approaches are the fixed effects model and the dynamic panel data estimation method that is proposed by Arellano and Bond (1991).
The findings of this study are as follows, a decrease in real GDP growth, inflation (GDP deflator), exchange rate and share prices tends to increase NPLs. Secondly, lower quality of banks’ management, which is measured by lagged profitability, leads to higher level of NPLs. Thirdly, lagged NPLs and financial crisis have a positive effect on NPLs. Lastly, bank capital-to-assets ratio does not influence NPLs.
The thesis proceeds as follow: Section 2 starts with a brief review of the previous literature. Section 3 focuses on the empirical methodology and the hypotheses. Section 4 discusses the data and shows descriptive statistics. Section 5 points out the results from the models and estimations. Section 6 shows the conclusion.
2. Literature Review
Non-performing loans（NPLs）are a measure of asset quality in financial institutions, especially in banks. Many scholars regard NPLs as a key element of banking or financial crises. For example, Li (2003) and Fofack (2005) show that there is a significant relationship between credit risk, which is represented by the level of NPLs, and financial crises. Therefore, the level of NPLs is increasingly important in the management of banks.
In recent decades, the economic ties among different countries and areas in the world are becoming ever closer due to the acceleration of global economic integration. Additionally, the banking industry is a vital element for economic development in all countries and regions in the world. Specifically, during the American Subprime Crisis and European Debt Crisis, a large number of famous banks were suffering a bad time. For instance, Lehman Brothers Holdings Inc. went bankrupt in 2008. These crises influence the level of NPLs and the profitability of the majority of banks. The level of NPLs, which is used to measure the stability of a banking system, is becoming a hot topic among numerous scholars. They put forward different hypotheses about the determinants of the NPLs.
2.1 Literature about hypotheses
share prices, the exchange rate and lending interest rate respectively. Klein (2013) focuses on inflation, the exchange rate and change in unemployment rate. In addition to unemployment rate and interest rate, Louzis et al. (2011) study the link between NPLs and public debt in Greece.
The second group of the hypotheses pays attention to the bank-level variables that influence the NPLs. Espinoza and Prasad (2010) hypothesize that firm-specific determinants, including the size of capital, credit growth and efficiency (noninterest expense/assets) could affect the NPLs. Klein (2013) considers bank-level variables, such as equity-to-assets ratio, return on equity, loan-to-assets ratio and the loans growth rate to explain NPLs. Similarly, Louzis et al. (2011) also consider bank-level variables of ownership concentration, management ability and credit policy. In addition, Bonin and Huang (2001), Istrate et al. (2007), Espinoza and Prasad (2010) and Louzis et al. (2011) hypothesize that moral hazard is a major systemic cause of the increase in NPLs. They argue that the moral hazard can cause risk-taking behavior in banks. For example, Louzis et al. (2011) treat owned capital to total assets ratio as the measure of moral hazard. Keeton and Morris (1987) point out that the managers of banks would like to increase the riskiness of loan portfolio for purpose of getting more profit and as a consequence banks may have relatively low capital. Fofack (2005) uses equity over total liquid assets and interbank loans over total assets as a proxy of moral hazard.
2.2Literature about methodology
drawbacks of the fixed effects model. For example, Beck et al. (2013), Klein (2013) and Louzis et al. (2011) use GMM method with instrumental variables to deal with the problem of correlation among errors in their studies to investigate determinants NPLs. In addition, Klein (2013) introduces and compares different GMM (generalized method of moments) methods and a system GMM method.
Other methods include panel VAR estimation and logit transformation. For example, Espinoza and Prasad (2010) use logit transformation and both Espinoza and Prasad (2010) and Klein (2013) use panel VAR methodology to evaluate the magnitude and duration of the effects.
2.3 Literature about results
Beck et al. (2013), Klein (2013) and Louzis et al. (2011) find that a rise in (contemporaneous) real GDP growth leads to a decline in the level of NPLs. Beck et al. (2013) also show a significant relationship between the level of NPLs and share price, exchange rate and lending interest rates. A decrease in share price and exchange rate lead to an increase of NPLs, while lending interest rates prove to have a significantly positive impact on the NPLs. When it comes to other macroeconomic variables, Klein (2013) concludes that inflation has a negative and significant impact on NPLs, while unemployment is positive and significant related to NPLs. Nkusu (2011) shows that an increase in the house price leads to an increase in NPLs.
cause of NPLs. Istrate et al. (2007) also conclude that moral hazard that result from the use of cash infusions to payoff NPLs or from deposit insurance leads to an increase in NPLs. In particular, Louzis et al. (2011) divide NPLs in the Greek banking system into three categories, including mortgage NPLs, business NPLs and consumer NPLs. Their results show that leverage has a significant and positive effect on mortgage NPLs and business NPLs. Return on equity is only significant and negatively related to the mortgages NPLs and consumer NPLs, but insignificant in explaining the business NPLs. Specifically, the change in the lending rates affects consumer loans deepest and the NPLs of business loans are most sensitive to an increase in real GDP growth.
2.4 Differences and limitations of literature
Based on the analyses mentioned above, there are no obvious differences among the results in the literature about the determinants of NPLs in banks. The most obvious differences in the literatures are the methodology and country analysis versus bank analysis. Particularly, Louzis et al. (2011) only focus on the dynamic panel estimation, while Klein (2013) and Beck et al. (2013) start with the fixed effect approach and then select the Arellano and Bond estimation. After using two different approaches, they also make a comparison between results from the fixed effect model and results from dynamic panel model. Besides, Espinoza and Prasad (2010) use logit transformation and panel VAR methodology to evaluation the magnitude and duration of the effects.
Additionally, Beck et al. (2013), Klein (2013) and Espinoza and Prasad (2010) do a cross-country analysis. They study determinants of NPLs in different countries. While, Louzis et al. (2011), Lu et al. (2001), Allen et al. (2011) and Istrate et al. (2007) just focus on only one country’s banks.
Although the time periods of previous literatures include crisis year, they seldom take the Subprime Crisis or European Debt Crisis explicitly into account or make a comparison of NPLs between pre-crisis period and crisis period when analyzing their results and making conclusion. It is essential to relate the analysis of determinants of NPLs to the economic background, because NPLs are an important measure to manage the banking system and to control the financial stability in a country. Specially, during the crisis, it is important to make a comparison of NPLs before the crisis and during the crisis, which could provide a better intuition for management of banks.
3. Hypothesis and Methodology
This section has two parts. The first part introduces the hypotheses and explains the reasons for making these hypotheses in detail and the relationship with previous literatures. The second part of this section focuses on the methodology that is used to test the hypotheses. To be specific, it introduces and explains the statistical relationship between dependent variable and independent variables and the estimation approaches.
3.1 Hypothesis
According to the literature review in Section 2, there are two main types of hypotheses. The first set of hypotheses focuses on the macroeconomic variables, whereas the second set of hypotheses pays close attention to bank-level variables. For instance, Beck et al. (2013) only investigate the relationship between NPLs and macroeconomic variables, including real GDP growth, leading interest rate, share price and exchange rate at a country level in 75 countries. Besides the macroeconomic variables, Klein (2013), Espinoza and Prasad (2010) and Louzis et al. (2011) also studies the link between NPLs and bank-level variables, including equity-to-assets ratio, return on equity, loan-to-assets ratio, not-interest income to total income ratio, total liabilities to total assets ratio and so on.
therefore have to face an increasing amount of NPLs. So, the first hypothesis concerns the relationship between NPLs and real GDP growth.
Hypothesis 1: There is a negative relationship between NPLs to total gross loan and real GDP growth.
Secondly, inflation, which is measured by the growth rate of GDP, is a well-known measurement of macroeconomic performance. The implicit deflator points out the rate of price changes in the economy of a country. Basically, inflation influences the ability of general cash flows of firms and also impacts the interest rate. In general, appropriate inflation is good for economic growth, while unbridled inflation is dangerous, which leads to bad growing trend of economic development. Moreover, borrowers may not repay the loans and banks face the risk of credit default. Therefore, the second hypothesis is related to inflation.
Hypothesis 2: There is a negative relationship between NPLs and inflation (GDP deflator).
Hypothesis 3: Exchange rate depreciations lead to an increase of NPLs.
Another determinant of NPLs is the share price. Although it may less obvious to link share prices to NPLs, they are correlated with the value of companies and reflect the future development of firms. A dramatic drop in share prices may lead to a decline in confidence of all stakeholders. For instance, staff may plan to quit and find a new job, suppliers may no longer allow on credit and small shareholders who are unaware of the truth may sell stocks. This would make firms get into trouble and firms may not have enough ability to repay these loans. In addition, share prices are related to house prices. Kakes and van den End (2006) point out that equity is not only a leading indicator for the housing market, but also has a causal influence on the housing market as well. When share prices are dropping, the values of collateral for housing loans also decrease and the loan quality of consumer loans therefore can be affected.
Hypothesis 4: Share prices have a negative impact on NPLs.
Finally, in explaining NPLs, bank-level variables should not be omitted. Management ability and capital ratio are important factors that determine NPLs. Even in economic prosperity, bad management of banks can also cause loan quality reduction. The profitability, which is measured by return on equity, is a measure of the quality of bank management. Additionally, the capital-to-asset ratio indicates the ability of banks to protect against expected and unexpected losses of loans. In particular, the minimum capital-to-asset ratio is eight percent in accordance with the Basel I rules for banks.
Hypothesis 5: Higher quality of the bank’s management as measured by profitability (return on equity) leads to lower NPLs.
Additionally, NPLs depend in its previous values. The one year lagged NPLs are therefore included as an independent variable.
3.2 Methodology
Typically, macroeconomic variables and bank-level variables are considered as possible factors in empirical models that trying to explain NPLs. Considering the fact that the problems of omitting variables and heterogeneity can be overcome by panel data techniques, the impacts to the level of NPLs from macroeconomic variables and bank-level variables are analyzed and quantified by panel data approach.
The panel regression is ran as follows:
𝑦i,t= 𝜏𝑦i,t-‐1+∝ 𝑥1i,t+ 𝛽𝑥2i,t+ 𝛾𝑥3i,t+ 𝛿𝑥4i,t+ 𝜃𝑥5i,t-‐1+ 𝜑𝑥6i,t-‐1+ 𝜌𝐷t+ 𝜇i,t （1）
Where, 𝑦i,t denotes NPLs to total gross loans (%) for country i at year t. 𝑥1i,t, 𝑥2i,t, 𝑥3i,t and 𝑥4i,t denote real GDP growth (%), inflation (%), nominal effective exchange rates (%) and S&P global equity indices (%) respectively for country i at
year t. In addition, 𝑥5i,t-‐1 and 𝑥6i,t-‐1 represent bank-level variables, bank return on equity (%) and bank capital-to-assets ratio (%) in last year, for country i at year t. These two bank-level variables measure the profitability and the capital adequacy of banks. The lagged values are included because the profitability and the capital adequacy of banks in last year can influence managers’ decisions in this year. In addition, Dt is a dummy variable that is equal to 1 during the economic crisis and 0 otherwise.
at the year of 2008. Others, as Lowe (2010), think that the crisis started at the end of 2007 and the beginning of 2008 (2007/2008). As there is no agreement about the exact date of the crisis, this thesis follows Klein (2013) and selects 2008 as the beginning of crisis. This thesis also estimates the panel regression without a crisis dummy variable and with a crisis dummy variable that has the year 2007 as the start of crisis to make a clear comparison.
Before estimating equation (1) several tests are performed, firstly, fisher unit root test is selected to check whether data are stationary. In accordance with the argument from Maddala and Wu (1999), fisher unit root test has the advantage that it can be used for both balanced panel data set and unbalanced panel data set.
Secondly, this thesis uses the Hausman test to check whether the random effects approach or the fixed effects model is more appropriate. The random effects model deals with panel data by proposing a global intercept, which is constant over time, and a random disturbance for each entity. This approach is only valid when all explanatory variables are not correlated with error term. If they are correlated, the fixed effects approach is preferable. The Hausman test assesses whether the coefficients of the random effects model are jointly significantly different from the consistent fixed effects estimator. If the null hypothesis of the Hausman test is rejected, the fixed effects model is preferred.
However, there are some problems with the fixed effects model. First of all, the time period covers 12 year, while individuals are 26 countries. Roodman (2006) calls it as “small T, large N” panels. Secondly, NPLs in the left-hand-side of equation depends on its own past value. Thirdly, independent variables may be correlated with past or current value of the error. Fourthly, lagged dependent variable cause autocorrelation. Finally, there are unobserved country-specific effects, which may be correlated with the explanatory variables. Therefore, only using the fixed effects model is simple and can be biased and inconsistent because of the lagged variable and error term.
The GMM method of Arellano and Bond (1991) can overcome the problems mentioned above. Anderson and Hsiao (1981) propose a method to first differencing the model and to apply an instrumental variable in order to get rid of the error term. However, this method is not efficient. A more efficient approach is put forwarded by Arellano and Bond (1991). They argue that additional instruments can be obtained by taking advantage of the orthogonally conditions between the error term and the lagged variables. In other words, Arellano and Bond (1991) utilize first differences to remove the individual affects and treat all the lagged dependent variable as instruments. There are two kinds of GMM approaches, which are the difference GMM approach and the system GMM approach. Although drawbacks of the fixed effects model can be overcome by the both two GMM methods, this thesis selects the system GMM method. Because if samples have a limited time period and high persistence, results of difference GMM approach are not as precise as the results of system GMM method.
growth and nominal effective exchange rates as endogenous. The reason is that the causality can run in both directions and they therefore are correlated with the error term. The lagged bank-level variables are treated as pre-determined. The other independent variables and lagged dependent variable are treated as instruments. Inflation and S&P global equity indices do not depend on the error term and are not related to endogenous variables.
4. Data
The initial dataset consists of the bank-level variables and macroeconomic-level variables in the European countries, covering the years 2000 to 2011. The time period starts from the first year of the 21st century and ends in the 2011 because of the data availability. During this period, the American Subprime Crisis happened and influenced the European countries deeply, which was called European Debt Crisis. It is therefore possible to make a comparison of banks’ non-performing loans between pre-crisis and during crisis by selecting this time period.
This thesis exclude Cyprus and Malta in the sample, because there are no data available for these countries about NPLs to gross loans, and bank capital-to-assets ratio and S&P global equity indices. In other words, this study focuses on 26 European countries.
selected to take into account the different size of banking system in different countries. Another advantage of using the ratio is that it avoids the multicollinearity between variables. The data of bank non-performing loans to total gross loans comes from The World Bank and Federal Reserve Economic Data.
Table I Definitions and source of variables
Variables Definitions Source
Bank Non-Performing Loans (NPLs) to Total Gross Loans (%)
Bank nonperforming loans to total gross loans are the value of nonperforming loans divided by the total value of the loan portfolio (including nonperforming loans before the deduction of specific loan-loss provisions).
The World Bank Federal Reserve Economic Data
Real GDP Growth (%) Annual percentage growth rate of GDP at market prices based
on constant local currency. UNdata
Inflation, GDP Deflator (annual %)
Inflation as measured by the annual growth rate of the GDP implicit deflator shows the rate of price change in the economy as a whole.
UNdata
Nominal Effective Exchange Rate (NEER) (%)
The unadjusted weighted average value of a country's currency relative to all major currencies being traded within an index or pool of currencies.
Bank for International
Settlements S&P Global Equity Indices (%)
S&P Global Equity Indices measure the U.S. dollar price change in the stock markets covered by the S&P/IFCI and S&P/Frontier BMI country indices.
The World Bank
Bank Capital-to-Assets Ratio (CTA) (%)
Bank capital-to-assets is the ratio of bank capital and reserves to total assets. Capital and reserves include funds, provisions, and valuation adjustments. Capital includes tier 1 capital and total regulatory capital. Total assets include all nonfinancial and financial assets.
The World Bank Federal Reserve Economic Data
Bank Return on Equity (ROE) (%) The amount of net income returned as a percentage of
shareholders equity. It is a measure of banks’ profitability.
The World Bank Federal Reserve Economic Data
Table I gives an overview of the definitions and sources of all variables that are included in this thesis2. The major data sources for the bank variables are The World Bank and Federal Reserve Economic Data and the major data source for the macroeconomic variables is UNdata.
The second group of independent variables consists of bank-level variables, including bank capital-to-assets ratio and bank return in equity. It is meaningful to find out the link between bank non-performing loans to total gross loans and bank capital to asset ratio. Since the bank capital-to-assets is used to analyze the financial condition of banks and to gauge health of financial statements. Managers in banks also use it to measure whether banks have enough equity to support loans. In accordance with the Basel I rules, banks need to meet the requirement that the minimum capital to asset ratio is eight percent. As usual, a higher ratio means that the bank has better ability to manage risk. Put differently, a bank with a higher capital-to-asset ratio is better protected against expected and unexpected losses. The other bank-level variable is bank return on equity, which is used to measure the profitability of banks. More specifically, the profitability of banks can be used to represent the ability of management of banks. Specifically, both bank-level variables are lagged one period, since banks’ performance in last year may influence decisions of managers for loan portfolio in this year.
Table II Descriptive Statistics
Variables Mean Median Maximum Minimum Std.Dev Obs.
Panel A（2000-2011）
NPLs to Total Gross Loans (%) 4.78 3.05 29.30 0.10 4.65 305
Real GDP Growth (%) 2.36 2.69 12.23 -17.95 3.78 311
Inflation, GDP Deflator (%) 3.33 2.49 44.25 -4.64 4.30 311
NEER (%) 99.47 99.89 191.66 65.69 10.16 312
S&P Global Equity Indices (%) 9.75 10.03 189.23 -73.02 36.99 306
Bank CTA Ratio (%) 7.01 6.50 15.30 2.70 2.42 305
Bank ROE (%) 9.28 10.95 36.36 -102.89 13.89 310
Panel B
Bef. Dur. Bef. Dur. Bef. Dur. Bef. Dur. Bef. Dur. Bef. Dur. NPLs to Total Gross Loans (%) 3.67 5.77 2.50 4.25 29.30 19.70 0.10 0.20 4.07 4.53 201 104
Real GDP Growth (%) 4.05 -0.34 3.81 0.94 12.23 8.28 -0.91 -17.95 2.41 4.48 207 104
Inflation, GDP Deflator (%) 4.10 2.19 2.88 1.82 44.25 15.54 -0.78 -4.64 5.04 2.90 207 104
NEER (%) 98.68 101.62 98.99 100.47 191.66 116.55 65.69 93.33 12.80 3.21 208 104
S&P Global Equity Indices (%) 19.92 -10.95 18.21 -16.28 189.23 76.73 -48.79 -73.02 34.29 37.78 202 104
Bank CTA Ratio (%) 7.15 6.75 6.70 6.45 15.50 13.80 2.00 3.20 2.64 2.27 203 102
Bank ROE (%) 12.88 4.17 12.80 7.00 36.36 33.43 -24.18 -102.89 7.51 19.19 206 104
Notes: Bef. means before crisis (2000 to 2007) and Dur. means during crisis (2008-2011). All data in this table are in percentages. The number of observations differs due to data availability. In detail, NPLs to total gross loans in 2000 and 2006 of Denmark, 2006 and 2007 of Netherlands, 2001 of Poland and 2000 and 2002 of Romania are not available. Real GDP growth and Inflation in 2000 of Ireland are also not available. Equity indices of Luxembourg from 2000 to 2005 also are not provided. For bank capital-to-assets ratios in 2000 of Denmark, in 2006 and 2007 in Hungary, in 2011 of Slovenia and in 2000, 2010 and 2011 of Sweden are not provided. Bank return on equity of Greece in 2000 and 2001 are not available.
Table III Correlation Matrix
(1) (2) (3) (4) (5) (6) (7) (1) Lagged NPLs (%) 1 (2) Real GDP Growth (%) 0.14 1 (3) Inflation, GDP Deflator (%) 0.23 0.34 1 (4) NEER (%) -0.65 -0.31 -0.02 1(5) S&P Global Equity Indices (%) -0,16 0.14 -0.06 -0.07 1
(6) Bank CTA Ratio (%) 0.38 0.05 0.45 0.06 0.05 1
(7) Bank ROE (%) -0.09 0.36 0.22 -0.04 -0.01 0.09 1
Table III shows the correlation among independent variables. It is clear that the correlation coefficients between two variables are below 0.7. In other words, there is no multicollinearity in independent variables.
5. Empirical Results
This section introduces the empirical results and analyzes the economic meanings of the empirical results.
First of all, Appendix B presents the main results of the panel unit root test. The fisher unit root test rejects the null hypothesis of a unit root in the data. In other words, the most of the variables are stationary.
Table IV Determinants of non-performing loans (Fixed Effects estimation)
Dependent Variable: Non-Performing Loans to Total Gross Loans Panel A (Crisis starts in 2008)
Independent Variables: Coefficient Std. Error p-value
Lagged NPLs 0.59 0.05 0.00
Real GDP Growth (%) -0.23 0.07 0.00
Inflation, GDP Deflator (annual %) -0.21 0.06 0.00
NEER (%) -0.05 0.03 0.08
S&P Global Equity Indices (%) -0.01 0.00 0.00
Bank CTA Ratio (%) -0.15 0.10 0.12
Bank ROE (%) -0.05 0.01 0.00
Dummy (Crisis starts in 2008) 0.30 0.29 0.29
Adjusted R2 0.84
Panel B
Crisis starts in 2007 No Crisis dummy variables
Independent Variables: Coe. Std. Err p-value Coe. Std. Err p-value
Lagged NPLs 0.59 0.05 0.00 0.58 0.05 0.00
Real GDP Growth (%) -0.23 0.64 0.00 -0.25 0.06 0.00
Inflation, GDP Deflator (annual %) -0.22 0.06 0.00 -0.22 0.06 0.00
NEER (%) -0.06 0.03 0.07 -0.05 0.03 0.08
S&P Global Equity Indices (%) -0.01 0.00 0.00 -0.01 0.00 0.00
Bank CTA Ratio (%) -0.13 0.10 0.16 -0.16 0.10 0.11
Bank ROE (%) -0.04 0.01 0.00 -0.05 0.01 0.00
Dummy variable 0.48 0.24 0.04 - - -
Adjusted R2 0.84 0.84
inflation are quite larger than that of S&P global equity indices. That is to say, the changes of the real GDP growth and inflation have greater influences to NPLs than that of S&P global indices.
Additionally, the nominal effective exchange rate is also statistically significant at the 10% level. The coefficient is negative, which means that exchange rate depreciation leads to an increase of NPLs. Therefore, hypothesis 3 can also be confirmed. In other words, when home country’s currency is worth less, NPLs tends to increase. But considering the small coefficient, the effect of NPLs from exchange rate is relatively small in economy.
For the bank-level variables, bank return on equity is negative and statistically significant. That is to say it supports hypothesis 5. Bank return on equity measures the profitability of banks and in turn profitability is an important indicator of the quality of the management of banks. It can be therefore concluded that the bad management of banks leads to an increase in NPLs. The other variable, bank capital-to-assets ratio, is not statistically significant since its probability is larger than 0.1. Put differently, the relationship between bank capital-to-assets ratio and NPLs cannot be confirmed. The capital adequacy of banks in European countries does not influence NPLs from 2010 to 2011.
The coefficient of lagged NPLs is positive and probability is less than 0.01. The level of NPLs in last year influences the level of NPLs in this year. Particularly, the coefficient of lagged NPLs is the largest. That is to say lagged NPLs has the most important effect on NPLs in banking system. The dummy variable is not statistically significant when the crisis starts in 2008. Lastly, the R-squared and adjusted R-squared are relatively close to 1，which means the goodness of fit statistics is good.
real GDP have the greatest impact on NPLs.
According to the disagreement that the start year of subprime crisis is 2007 or 2008 mentioned in Section 3.2, panel B shows results with dummy variable that crisis starts in 2007 and results without dummy variable.
The main results remain unchanged for estimation with dummy variable that crisis starts in 2007 and estimation without dummy variable. The independent variables of real GDP growth, inflation, S&P global equity indices, nominal effective exchange rate and bank return on equity are statistically significant. NPLs in last year also influence NPLs in this year. One difference is dummy variable that crisis starts in 2007 is positive statistically significant. The coefficient of dummy variable is quite large, which means it has a major impact on NPLs.
However, there are some problems that cannot be dealt with the fixed effects approach perfectly. For instance, there is a “small T, large N” panel and the dependent variable depends on its own past value. Therefore, the Arellano and Bond estimation is selected, since it has advantages that can deal with these problems. Table V reports the empirical results of GMM.
of NPLs are proved. From an economic perspective, it shows that a booming and steady growth economy, prosperity of stock market and a steady exchange rate lead to a decline in NPLs. However, the coefficients of the exchange rate and the equity indices are still relatively small. This means that the impact of the exchange rate and the equity indices on NPLs are not as large as the other macroeconomic determinants of NPLs.
Table V Determinants of non-performing loans (Arellano and Bond)
Dependent Variable: Non-Performing Loans to Total Gross Loans Panel A (Crisis starts in 2008)
Independent Variables: Coefficient Std. Error p-value
Lagged NPLs 0.69 0.55 0.00
Real GDP Growth (%) -0.22 0.02 0.00
Inflation, GDP Deflator (annual %) -0.14 0.07 0.05
NEER (%) -0.04 0.02 0.07
S&P Global Equity Indices (%) -0.00 0.00 0.01
Bank CTA Ratio (%) 0.13 0,17 0.45
Bank ROE (%) -0.02 0.01 0.00
Dummy (Crisis starts in 2008) 0.54 0.20 0.01
AR (1) 0.03
AR (2) 0.93
Sargan test 1.00
Panel B
Crisis starts in 2007 No Crisis dummy variables
Independent Variables: Coe. Std. Err p-value Coe. Std. Err p-value
Lagged NPLs 0.68 0.06 0.00 0.69 0.05 0.00
Real GDP Growth (%) -0.21 0.32 0.00 -0.23 0.04 0.00
Inflation, GDP Deflator (annual %) -0.18 0.85 0.03 -0.18 0.07 0.01
NEER (%) -0.05 0.03 0.03 -0.02 0.03 0.46
S&P Global Equity Indices (%) -0.00 0.00 0.00 -0.01 0.00 0.00
Bank CTA Ratio (%) 0.21 0.17 0.24 0.17 0.13 0.20
Bank ROE (%) -0.02 0.01 0.00 -0.02 0.00 0.00
Dummy variable 0.77 0.11 0.00 - - -
AR (1) 0.03 0.04
AR (2) 0.84 0.86
The bank return on equity is negative statistically significant at 1% significant level. This implies that a better management of banks，which measured by higher profitability leads to lower NPLs. Similar to the result of the fixed effects method, the bank capital-to-assets ratio is not statistically significant related to NPLs. The capital adequacy of banks not affects NPLs.
The coefficient of lagged NPLs is positive and statistically significant at the 1% level. This means prove that NPLs in previous years affect NPLs in this year. Opposite to the result of dummy variable in the fixed effects model, the dummy variable that crisis starts in 2008 is significant at 5% significant level and this confirm that the crisis influence the NPLs in banks.
Similar to the results of the fixed effects model, the coefficients of lagged NPLs and dummy variable are relatively larger. This implies that lagged NPLs and the crisis have the most important influence on NPLs. The contributions of bank return on equity, nominal effective exchange rate and S&P global equity indices are still relatively low. In the result of system GMM of Klein (2013), lagged NPLs also have the largest coefficient and impact NPLs greatest.
Panel B shows the results with the dummy variable that the crisis starts in 2007 and without the crisis dummy variable. Comparing to results in panel A, main results still remain unchanged. The probabilities of bank capital-to-assets ratio and nominal effective exchange rate in estimation without the crisis dummy variable are larger than 0.1 and not significant. That is to say, other variables, including real GDP growth, inflation, equity indices, exchange rate, bank return on equity and crisis influence the level of NPLs.
6. Conclusion
This thesis studies determinates of non-performing loans of banks in European countries (excluding Cyprus and Malta) from 2000 to 2011. The two main perspectives of determinants of NPLs in this thesis are macroeconomic variables and bank-level variables. The macroeconomic factors consist of real GDP growth, inflation (GDP deflator), nominal effective exchange rate and S&P global equity indices. The bank-level factors are bank capital-to-assets ratio and bank return on equity. The main approaches of estimation are the fixed effects model and the Arellano and Bond estimation. The major findings in this thesis are as follows: first of all, there is a statistically significant negative link between NPLs to total gross loans and four macroeconomic variables, which are real GDP growth, inflation (GDP deflator), nominal effective exchange rate and S&P global equity indices. Secondly, higher quality of banks’ management, which is measured by bank return on equity of previous year, leads to lower NPLs. Thirdly, lagged NPLs and the crisis have relatively larger coefficient. They have a great effect on the level of NPLs. Finally, there is no evidence to confirm that lower capital-to-assets ratio tends to worsen NPLs.
2010 and they conclude that a rise in real GDP growth leads to a decline in NPLs ratios and share prices depreciations cause an increase in the level of NPLs. In addition, they also conclude that there is a statistically significant negative relationship between exchange rate and NPLs. Klein (2013) studies banks in Central, Eastern and South-Eastern Europe from 1998 to 2011 and concludes NPLs are sensitive to bank-level variables. Put differently, higher profitability leads to lower NPLs and low equity tends to worsen NPLs. Comparing the results of this thesis to the previous literatures, this thesis gets the same results of the relationship between NPLs and macroeconomic factors. With regards to bank-level variables, the result of relationship between bank return on equity and NPLs is similar to the result of Klein (2013). However, the result of link between bank capital-to-assets ratio and NPLs is different from that of Klein (2013). In this thesis, the result of bank capital-to-assets ratio is not significant. In contract, Klein (2013) finds a statistically significant effect of bank capital to assets on NPLs.
However, this thesis has two main limitations. On the one side, different countries have different accounting method and regulations, which lead to inconsistent classification of NPLs. On the other side, government policies are different in different countries; specially refer to economy and finance. This thesis ignores the country special policy factor in different countries.
References
Allen, F., Qian, J., Zhang, C., Zhao, M., 2012. China’s financial system: opportunities and challenges. Unpublished working paper. National Bureau of Economic Research, Boston,United Stated.
Anderson, T. W., Hsiao, C., 1981. Estimation of dynamic models with error components. Journal of the American statistical Association, 76(375), 598-606.
Arellano, M., 1987. Practitioners’ corner: computing robust standard errors for within groups estimators. Oxford bulletin of Economics and Statistics, 49(4), 431-434.
Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The review of economic studies, 58(2), 277-297.
Beck, R., Jakubik, P., Piloiu, A., 2013. Non-performing loans: what matters in addition to the economic cycle? Unpublished working paper. European Central Bank, Frankfurt am Main.
Bonin, J. P., Huang, Y., 2001. Dealing with the bad loans of the Chinese banks. Journal of Asian Economics, 12(2), 197-214.
Brooks, C., 2008. Introductory econometrics for finance, 2nd edition. Cambridge University Press, Cambridge, UK
Publications, Washington, DC.
Espinoza, R. A., Prasad, A., 2010. Nonperforming Loans In The GCC Banking System And Their Macroeconomic Effects. International Monetary Fund, Washington, DC.
Fofack, H., 2005. Nonperforming loans in Sub-Saharan Africa: causal analysis and macroeconomic implications. Unpublished working paper. World Bank Policy Research, Washington DC.
Guy, K., 2011. Non-performing loans. Research and Economic Analysis, 37.1, 10.
Istrate, E., Das Gupta, D., Weissburg, P., 2007. Toward developing a structured approach to the diagnosis and resolution of nonperforming loans: the case of China and India. Review of Policy Research, 24.4, 345-365.
Kakes, J., Van Den End, J. W., 2004. Do stock prices affect house prices? Evidence for the Netherlands. Applied Economics Letters, 11.12, 741-744.
Keeton, W. R., Morris, C. S., 1987. Why Do Banks’ Loan Losses Differ?. Economic Review 5, 3-21
Klein, N., 2013. Non-performing loans in CESEE: determinants and impact on macroeconomic performance. Unpublished working paper. International Monetary Fund, Washington, DC.
Li Yang, 2003. The Asian financial crisis and non-performing loans: evidence from banks in Taiwan. International Journal of Management, 3,69-74
bank-specific determinants of non-performing loans in Greece: A comparative study of mortgage, business and consumer loan portfolios. Journal of Banking & Finance, 36.4, 1012-1027.
Lu, D., Thangavelu, S. M., Hu, Q., 2001. The link between bank behavior and non-performing loans in China. Unpublished working paper. National University of Singapore, Singapore.
Mileva, E., 2007. Using Arellano-Bond dynamic panel GMM estimators in Stata. Unpublished working paper. Fordhan University, New York.
Nkusu, M., 2011. Nonperforming loans and macrofinancial vulnerabilities in advanced economies. Unpublished working paper. International Monetary Fund, Washington, DC.
Podpiera, J., & Weill, L., 2008. Bad luck or bad management? Emerging banking market experience. Journal of Financial Stability, 4.2, 135-148.
Reinhart, C. M., Rogoff, K., 2009. This time is different: Eight centuries of financial folly. Princeton University Press, Princeton.
Roodman, D., 2009. How to do xtabond2: An introduction to difference and system GMM in Stata. Stata Journal, 9.1, 86.
Schwert, G., 2009. EViews 7 User's Guide II. Quantitative Micro Software, LLC, Irvine, California.
0 1 2 3 4 5 6 7 8 9 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Appendix A
Appendix B
Table 1. Panel unit root tests (Fisher)
Variables Fisher-ADF (Prob.) Fisher-PP (Prob.)
Real GDP Growth (%) 0.03 0.00
Inflation, GDP deflator (annual %) 0.00 0.00
NEER (%) 0.00 0.00
S&P Global Equity Indices (%) 0.00 0.00
Bank CTA Ratio (%) 0.02 0.00
Bank ROE (%) 0.04 0.13
Lagged NPLs 0.00 0.06
Appendix C
Table 2. Determinants of non-performing loans (Hausman Effects estimation)
Correlated Random Effect- Hausman Test
Test Summary Chi-Sq. Statistic Chi-Sq. d.f. p-value
Cross-section random 50.05 8 0.00
Independent Variables Fixed Random Var(Diff.) p-value
Lagged NPLs 0.65 0.76 0.00 0.00
Real GDP Growth (%) -0.17 -0.17 0.00 0.87
Inflation, GDP Deflator (annual %) -0.17 -0.01 0.00 0.00
NEER (%) -0.04 0.00 0.00 0.00
S&P Global Equity Indices (%) 0.00 0.00 0.00 0.00
Bank CTA ratio (%) -0.10 0.14 0.01 0.00
Bank ROE (%) -0.04 -0.04 0.00 0.31
