EXAMINING OF DETERMINANTS OF NON-PERFORMING LOANS

It is common knowledge that financial institutions play a vital role in the economy by allocating capital from surplus agents to deficit agents in various economic sectors (Fukida, Dahalan, 2012). This means that a sound banking sector is necessary for economic growth because it ensures macroeconomic stability and develops sound financial institutions (Jovovic, 2014).

However, during the last two decades, the deregulation process has strengthened competition among banks. Competition increased banks’ credit risk, i.e. affecting their loan portfolios in terms of bad loan screening procedures and relaxing borrowing criteria (e.g. Manove et al. (2001), Bolt, Tieman (2004), Jeong, Jung (2013)).

This has led to a significant increased level of non-performing loans (NPLs), which affects the liquidity and profitability of banks and thereby the financial stability of the banking systems and, in general, the macroeconomic stability.

Many indicators are used to measure banks’ lending activity, but the most commonly used indicators to identify credit risk are non-performing loans to total loans (NPLs) and loan loss provision to total loans (LLP).

Prior to the financial crisis in the last decade, the quality of the loan portfolio credit remained relatively stable. Thereafter, the quality of the banks’ lending activity exacerbated sharply. The deterioration of the quality of banks’ loan portfolios caused distress in the banking sector in both developed and emerging economies.

The problem of the increase of NPL ratio is evident in the banking sector in many countries. Saba (2012) states that, since 2008 the level of NPLs has remarkably increased and the link between NPLs and decline of banks’ credibility is considered as a main factor of the failure of credit policy.

It is well-known that the stability of the financial sector and its likelihood of distress depend highly on the portion of the NPLs; thus NPLs serve as an indicator of defaults in the financial sector. A number of studies have shown that excessive credit growth often precedes financial crisis.

The aim of this paper is to carry out deep panel data analysis despite of detection of the relationship between macro and micro environment over NPLs.

This paper contributes to the literature by investigating the determinants of the aggregate loan quality in 25 emerging countries using the former indicator of banks’ credit quality – NPLs. The paper has the following structure: The first section contains the introduction.

Section 2 provides evidence on the relative studies and theoretical patterns based on the previous studies. Section 3 contains the description of econometric methodology used in the paper. In the Section 4 the final empirical research can be found with the presentation of the outcomes. The final section summarizes the conclusions.

1. Literature Review

There are many studies conducted on the problem of default rate of loans in the banking sector, most of them based on the relationship between bank-specific variables and macroeconomic factors. The earliest empirical study is conducted by Keeton and Morris (1987).

They found that the risk that banks were taking was one of the reasons for loan losses and credit failures. Lis et al. (2000) econometrically identified loan losses through various banking and macroeconomic factors using a panel data of Spanish commercial and savings banks for the period 1985–1997.

They found that gross domestic product (GDP) growth rate has a negative effect on problem loans, that bank size is negatively related to problem loans, while loan growth, collateral loans, net interest margin and market power are positively related to that.

Nkusu (2011) analysed the linkage between non-performing loans and macroeconomic performance of 26 advanced economies from 1998 to 2009. He used only macroeconomic variables in his study.

His findings revealed that a poor macroeconomic performance (i.e. slower GDP growth, higher unemployment or decreasing asset prices) could be related to the increasing non-performing loans in advanced economies.

The findings of the study of De Bock and Demyanets (2012) showed that GDP growth rate, exchange rates and loan growth are the main determinants of NPLs in the examined countries. In contrast to Nkusu they analysed the determinants of bank asset quality in 25 emerging countries during 1996–2010, by examining only aggregate macroeconomic and credit indicators.

Makri and Papadatos (2012) econometrically identify the determinants of credit risk in Greek banking sector by using LLP as a proxy of loan defaults, for the period from 2001 to 2012. They found that LLP is positively affected by unemployment rate and public debt, but negatively affected by capital adequacy ratio. There are many studies that are based on panel cross-country analysis.

The regression model of Mileris (2012) indicates that there is a strong relationship between changes in the macroeconomic environment and the performance of loans. He states that unemployment rate and interest rates have a strong influence on the quality of loans in the banks. Saba (2012) argues that there is significant dependency among NPL ratio and interest rates and between total loans and NPL rate.

Moinescu (2012) studied determinants of NPLs in Central and Eastern European countries (CEE) during the period between 2003 and 2011. Econometric results of his research confirm that GDP growth is the prominent macroeconomic explanatory variable of NPLs developments among CEE economies.

He found that real GDP growth and the change in output gap were almost equally important. His analysis reveals a strong determination influence with a short-term impact of the economic performance on the non-performing loans ratio dynamics across the CEE banking systems.

Stylized representation of the change in the banking book quality as a function of the amplitude of business cycle suggests that the larger the difference between peak and depth of economic growth, the higher the NPL ratio jump during recession period (Moinescu, 2012, 56).

Similar findings were reported by Mileris (2012). Makri et al. (2014) identified the factors affecting the NPL rate in the Eurozone’s banking sector for the period between 2000 and 2008.

Their findings reveal strong correlations between NPLs and various macroeconomic (public debt, unemployment, annual percentage growth rate of GDP) and bank-specific factors (capital adequacy ratio, rate of nonperforming loans of the previous year and ROE).

Precisely, they found statistically significant and negative correlation between NPLs and ROE. They also found statistically significant and positive relationship between:

the dependent variable NPLs and its lagged value;
NPLs and public debt;
NPLs and the unemployment rate.

A notable branch of literature is related to the appraising role of bank-specific factors based on balance sheets of banks. Tabak (2005) employs semi-annual data from the Balance Sheet and the Income Statements of Brazilian banks for the period from 2000 to 2005 and finds that the credit risk is the major source of operational banking risk.

He concluded that a high level of NPLs indicates that banks have high credit risk and if not regularly managed may induce banking failures.

The similar approach was done in Greece. Precisely, Louzis et al. (2012) used panel data methods to examine the determinants of non-performing loans in the Greek banking sector, separately for each loan category.

According to Louzis et al. (2012) the determinants of NPLs should not be seen only among macroeconomic variables considering the fact that they are external to the banking sector. Thus, the characteristic features and the choice of policy of each bank are predicted to have an impact on NPLs rate.

The results indicate that the Greek banking sector can be explained by GDP, unemployment, interest rate, public debt and bank-specific factors such as management quality and performance. The leverage has a statistically positive influence on business and mortgage NPLs.

In addition, the ROE indicator is statistically significant for mortgage and consumer NPLs while insignificant for the business NPLs. The impact on loan categories is obvious with mortgages being the least reactive to changes in the macroeconomic environment.

Lu et al. (2005) explored the relationship between banks’ lending behaviour and NPLs by using the financial data from annual financial reports of all publicly listed companies. The results of this study show that the Chinese banks have a structured lending bias towards state-owned enterprises (SOE), especially those with high default risk.

The study observes that the high-risk SOE were able to borrow more than the low-risk SOEs and non-SOEs (Lu et al., 2005). Using panel regression analysis, Pain (2003) investigates the factors that can help explain the increases in loan-losses in the UK for crucial banks.

He points out that macroeconomic variables can indeed affect banks’ NPLs, but he states that bank-specific factors are also important considering the fact that a great percentage of lending to riskier sector, such as commercial companies, has been usually linked to higher provisions (Pain, 2003).

Espinosa and Prasad (2010) concluded that the NPLs ratio aggravates as economic growth weakens and interest rates rise by using a dynamic panel data from 80 banks in the Gulf Cooperation Council (GCC).

Their model suggests that bad macroeconomic conditions may indeed affect the future levels of NPLs ratios. Vector autoregressive (VAR) effects were also examined in order to look for the feedback that increasing NPLs might have on economic growth.

Economic activity should be strongly, statistically and significantly affected by credit according to this panel VAR model. Finally, according to Figlewski et al. (2012) the following macro factors are believed to have a great impact on banks´ creditworthiness:

Factors related to general macroeconomic environment (unemployment rate, inflation, etc.); Factors related to the direction in which the economy is moving (real GDP growth, the change in consumer sentiment, etc.); Factors of financial market conditions (interest rates, stock market returns, etc.).

2. Data and Methodology

The paper examines 25 emerging countries (Albania, Algeria, Bulgaria, Bosnia and Hercegovina, Croatia, Chile, Cyprus, Ghana, Greece, Hungary, Island, Ireland, Kazakhstan, Mauritania, Moldova, Montenegro, Macedonia, Slovenia, Senegal, Serbia, Pakistan, Romania, Tunisia, Ukraine and Zambia) for the period from 2000 to 2011.

The data were collected from Federal Reserve Saint Louis Bank’s, World Bank’s, Bank of International Settlements’s official websites and Eurostat database. Panel data approach is used to examine the results. Panel data analysis deals with data in which behaviour of entities is observed over time.

The main purpose of panel data analysis is to note whether there is any pattern in the data collected over time and different entities (cross-sectional between different countries). Considering the structural breaks for some of the data, an unbalanced panel data set is used.

The effect over the dependent variable in pre-financial crisis period and mostly during the financial crisis is examined within this time span. The period after the financial crisis from 2011 to 2014 is not considered due to unavailability of the data.

a. Data description

The dependent variable in the study is the non-performing loans (NPLs) ratio, which represents the sum of borrowed money upon which the debtor is not able to make his or her scheduled payments for at least 90 days.

The gross domestic product (GDP), the unemployment rate (UNR), the inflation rate (INF), the nominal effective exchange rate (NEER), the house price index (HPI) and the return on equity (ROE), the return on assets (ROA), the bank’s capital to assets (CAR) ratio, the loan loss provision (LLP) and the net interest margin ratio (NIMR) are used as explanatory variables.

The determinants of NPL based either on macroeconomic conditions or bank specific factors were the subject of many studies done by numerous researchers.

According to Boudriga et al. (2009), in order to fairly reduce non-representativeness of the sample, an aggregate data for the whole banking sector of each country (in contrast to the analysis of individual data for each bank) should be more suitable to use.

Researchers often use panel data across many countries, in order to explain effects of macroeconomic factors, whereas for bank specific factors there is a frequent handling of panel data analysis from the sample of the biggest banks in one region or one country (Jovovic, 2014).

As a matter of fact, both types of variables in one model are used in the paper, along with aggregate bank-specific factors which are gathered for the whole financial sector of the chosen country. Selection of the variables was guided by the following:

that variables should be the ones significant for NPLs, especially for emerging markets, due to their specific characteristics, level of development, situation on the financial market and the like
that there are available data on all the variables for all the selected countries;
that conflicting results on their impact on NPLs can be found in literature.

The empirical researches show that NPLs are closely related to the economic and business cycle, i.e. behind every financial crisis there are macroeconomic factors, such as downturns in aggregated economic activity.

When growth slows or turns negative, borrowers reduce their cash inflows, which in turn makes it difficult for them to pay the interest and principal on bank loans. Under these circumstances borrowers will face liquidity shortages and the delays in the fulfillment of their financial obligations to banks will likely increase.

We used the GDP to measure the aggregated economic activity, because the GDP is highly informative on other relevant macroeconomic variables. For the purpose of the paper, the GDP variable has been transformed into logarithmic function and represents the GDP growth over years.

It is expected that the rise of the GDP will cause a decline in the NPLs ratio. According to the related literature, UNR is one of the most important indicators of the NPLs ratio. For this reason, this variable is included into the research.

The variable UNR represents the number of the unemployed workforce as a percentage of total labour force. It is expected that higher unemployment rate will lead to a rise in NPLs rate.

High inflation rate is one of the biggest issues of the central banks in many emerging countries; therefore, their goal is to stop severe inflation and to keep the increase in prices to a minimum level. The higher the inflation rate, the higher the NPLs rate.

This is because wages are assumed to be sticky, so they adjust less quickly than the overall increase in the prices in the economy. Therefore, people are more confronted with difficulties in paying back loans while prices are higher.

It is expected that higher inflation rate will provoke rise in the NPLs ratio. However, Rinaldi and Sanchis-Arellano (2006) find a positive relationship between the inflation rate and NPLs. Theoretically, inflation should reduce the real value of debt and hence make debt servicing easier (Skarica, 2013).

In addition to these variables, which are standard empirical determinants of NPLs, two more macroeconomic variables are included in the study due to their importance for the selected countries: NEER and HPI. The influence of the exchange rate depreciations on NPLs can be of dual nature.

The growth in foreign currency has a negative impact on the possibility of private borrowers to pay loans, because it is known that in the selected emerging countries borrowers take (long-term) loans in foreign (convertible) currency, but get their salaries in local currency.

On the other hand, literature emphasizes that depreciation of national currency can have a positive impact on the decrease of NPLs, through an increase in export volumes and thus an improvement of the financial position of the corporate sector.

NEERs have been used for the purpose of studying the influence of depreciation on national currency. For the purpose of the paper the NEERs2 are calculated as geometric weighted averages of bilateral exchange rates taken from the Bank of International Settlements statistics.

Swings in property prices3 have been extremely large and frequent in selected emerging countries. The impact of real estate price changes on NPLs can also be twofold. On the one hand, the growth of real estate price can lead through decreased NPLs through collateral.

Moreover, the pledging of collateral decreases borrower’s moral hazard. On the other hand, pledging collateral could increase problem loans, because banks have fewer incentive to screen and monitor properly (they will take over the collateral) and the worthier the collateral, the higher the optimism of banks (Manove et al. (2001), Salas and Saurina (2002)).

In order to capture these effects NPLs, the HPI (2000 is equal 100) is included in our research. The main criterion for the selection of micro variables, in addition to their availability of data and their importance to the business of banks, was that conflicting results about the variables' impact on the rate of NPLs could be found in literature.

Variables selected are ROA, ROE, CAR, LLP and NIMR. According to Godlewski (2004), Garsiya and Fernandez (2007) and Stakic (2014), ROA and ROE are the most important indicators of efficiency, profitability, i.e. the quality of bank management.

ROA indicates the management’s ability to use its assets effectively in order to generate higher profits. The ROA ratio is displayed as a percentage which is calculated by dividing company’s annual earnings with its annual assets.

On the other side, the ROE measures the amount of net income compared to shareholders equity. It indicates banks’ profitability.

It is expected that banks with high efficient (ROA) and high rate of profitability (ROE) have less pressure regarding profit making, hence also a lower dependency on venturing into the risk exposed placement.

At the same time, low-cost efficient and less profitable banks are having a higher degree of problems with a high rate of NPLs. Studies conducted by Godlewski (2004) and Stakic (2014) confirm these expectations.

Low-cost efficiency and low ROE are positively associated with an increase in the NPLs rate failure. However, research conducted by Garsiya and Fernandez (2007) showed that high ROA and ROE are followed by a higher risk exposure.

Also, Boudriga et al. (2009) identified positive correlation between the aforementioned variables and NPLs. In addition to these two variables, CAR was included in the research as a standard empirical determinant of NPLs.

Capital to assets represents the ratio of bank’s capital, reserves and total assets including all nonfinancial and financial assets. The CAR is a good indicator of bank’s solvency, and at same time it is highly correlated with leverage.

In literature, there is no consensus about sign of relationship between the capital adequacy and NPLs ratio. It is expected that the higher CAR will cause NPLs rate to decease, but there are arguments in favour of the fact that banks with higher CAR are venturing into higher risk- taking activities, thus creating higher risk for credit portfolio (Rime, 2001), and therefore high NPL rates.

It is similar with LLP. Usually it stands in direct relationship with the level of NPLs. When predicting the high level of capital losses banks are allocating higher provisioning in order to reduce simultaneously the unpredictability of profitability and strengthen their medium-term liquidity (Stakic, 2014).

However, managers are free to use provisioning in order to demonstrate financial strength of the bank, in view of the fact that readiness to allocate higher reserves may be linked with a robust belief in the bank’s future performances.

The NIMR represents a good indicator of how optimal the investment decisions that a bank makes are. A negative value denotes that the bank did not make an optimal decision. However, the research conducted by Salas and Saurina (2002) shows that this variable does not affect NPLs rate.

On the other hand, Espinoza and Prasad (2010) point out that there is a significant relationship between the NIMR and the NPLs. A decrease in the NIMR can bring about a change in the credit policy, making it riskier.

The risk increase will create a loan portfolio with higher default probability in the future, which is why the variable has been lagged few periods. For this reason, one and two lags of this variable are used in the research. A negative sign is expected for this variable.

Location: Indonesia