Previous Studies on Factors Affecting Bad Debt

Profits will decline as high levels of bad debt require banks to increase their credit risk provisions, which will reduce banks' revenues (Athanasoglou, 2008).

2.1.4.2. Impact of bad debt on bank capital safety

Maybe you are interested!

The bank balance sheet channel is related to the traditional bank lending and bank funding channels. The traditional bank lending channel focuses on deposits - limited reserves for bank lending. However, the traditional bank lending model ignores the role of bank capital and endogenous credit risk by assuming that all loans will be repaid. Holmstrom and Tirole (1997) point out the importance of bank capital to finance lending because it provides incentives for banks to monitor borrowers and overcome the moral hazard problem. However, a decrease in bank capital due to a decrease in debt repayments following a shock that weakens the net worth of firms will reduce the supply of loans. Blum and Hellwig (1995), Borio et al. (2001) and Goodhart et al. (2004) study the bank capital channel in the context of regulatory requirements. Increased credit risk in a recession leads to a reduction in bank capital ratios and therefore banks face higher capital needs to meet regulatory requirements. However, raising capital becomes more difficult and costly as banks’ profits and ability to build reserves decline and they are likely to write down assets. This means less credit available to businesses and households, which in turn reduces borrower spending and reduces aggregate demand in the economy.

The moral hazard hypothesis also reflects the relationship between NPLs and bank capital. Diwan and Rodrik (1992) argue that high NPLs increase uncertainty about banks’ capital position and, therefore, limit their access to capital for mobilization. This in turn increases banks’ lending rates and contributes to a decline in credit growth. Two additional mechanisms mentioned in the literature are the high costs associated with managing high NPLs

(Mohd et al., 2010) and lower capital resulting from provisioning. Both contribute to low credit supply, which in turn affects economic activities.

2.1.4.3. Impact of bad debt on credit growth

The borrower balance sheet channel relates to the borrower’s capital position (or net worth) that affects their access to credit. The financial accelerator theory discussed above also explains bank lending behavior in relation to cyclical fluctuations in the economy. As the economy grows, net worth improves, and the external capital surplus becomes lower, as lenders perceive less risk in lending to higher net worth borrowers. The reverse shock suggests that lower current cash flows to borrowers lead to a decrease in net worth and an increase in the external capital surplus. The increase in the borrower’s cost of capital discourages them from investing more, and as a result, affects the demand for credit, amplifying and amplifying the effect of the initial shock. Kiyotaki and Moore (1995) develop a dynamic equilibrium model to describe how borrowers’ net worth is sensitive not only to changes in cash flows, but also to changes in the value of their real financial assets. In this model, assets play a dual role in the economy: (i) Producing goods and services and (ii) Providing collateral for loans. When a temporary shock causes asset values to fall, a direct effect occurs because the change in collateral leads to a change in credit availability. In addition, the reduction in production and consumption due to a shock to the real economy also causes a subsequent fall in asset prices, leading to a transmission of the shock over time.

The credit cycle model (Figure 2.5) is an economic model developed by Kiyotaki and Moore that shows how small shocks to the economy can be amplified by credit constraints, leading to large output fluctuations. The model assumes that borrowers cannot be forced to repay their loans. Therefore, in equilibrium, lending occurs only if there is collateral. That is, the borrower must possess a sufficient amount of assets or capital that can be seized.

in case they default. This collateral requirement amplifies business cycle fluctuations because in a recession, the income from capital falls, causing the price of capital to fall, which makes capital less valuable as collateral, discouraging business investment by forcing them to reduce their borrowing and thus worsening the economic downturn.

Figure 2.5. Credit cycle model

Present Future

The company's asset needs are limited.

The company's net worth is subject to a limited reduction.

Day t day t+1 day t+2

Temporary negative shock

Asset costs decrease

The company's net worth is subject to a limited reduction.

The company's asset needs are limited.

The company's asset needs are limited and reduced.

Asset costs decrease

Falling asset prices

Source: Kiyotaki and Moore (1995)

2.2. Previous experimental studies

2.2.1. Previous studies on factors affecting bad debt

Next, the thesis summarizes some empirical studies in countries around the world and Vietnam on the relationship between factors and bad debt. From there, consider the application of these factors in the study.

2.2.1.1. Empirical evidence on specific factors

- Credit growth

First, studies on the impact of credit growth factors have inconsistent results. One of the pioneering studies on the relationship between bad debt and credit growth is Clair (1992) when conducting research for commercial banks in Texas in the period 1980-1990. The author uses the bad debt ratio as a credit quality variable, and the explanatory variables are divided into 3 groups, including: (i) credit growth; (ii) financial characteristics and (iii) business conditions. Three types of credit growth are used simultaneously in the model including internal growth, growth from mergers and growth from acquisitions. The impact of credit growth has a lag of 0, 1, 2 and 3 years. The method used is least squares regression. The results show that internal and acquisition credit quality improvements are achieved at lags 0 and 1. However, merger credit only reduces bad debt and does not affect the rate of non-recoverable debt. In addition, financial characteristics and business conditions variables act as good controls in the model.

Keeton (1999) conducted a study on the relationship between credit growth and bad debt in the two-period time series 1967-1983 and 1990-1989. The author used two vector autoregressive VAR models. One model included credit growth, credit standards and GDP. The other model included income, loan capital and default rate. The first model showed that credit growth did not affect credit standards in the period 1990-1998 but tightened in the period 1967-1983.

In addition, credit standards tend to constrain credit growth in both periods. The second model suggests that high past lending will lead to high default rates in the future.

Salas and Sarina (2002) mentioned two types of TTTD: own growth and branch network growth. The study used dynamic econometric models and the difference GMM regression method. The lagged variables of the factors were used at 2, 3 and 4 years. The general results showed that TTTD increased bad debts. Specifically, growth in branch networks caused bad debts in the commercial bank type after 3 years. For savings banks, own growth and branch network growth increased the bad debt ratio after 3 and 4 years, respectively. Macroeconomic factors and bank characteristics played a good controlling role in the research model.

Foos et al. (2010) used the cost of credit loss provision and the ratio of bad debts to net interest income as proxy for bank risk. The method used is to estimate the OLS model and system GMM. The results of the study show that past credit loss provision increases the cost of credit loss provision and the ratio of bad debts to net interest income in the future. In addition, this study also presents the relationship between credit loss provision and bad debts along with the interaction of the effects of mergers and acquisitions. The impact of growth is still positive but mergers and acquisitions have reduced this impact.

- Asset size

Studies on the impact of asset size are summarized in Table 2.4. Most studies found a negative correlation between asset size and bad debt or did not find any impact of asset size on bad debt (Louzis et al., 2012; Salad and Saurina, 2002; Jimenez and Saurina, 2006; Nguyen Thi Ngoc Diep and Nguyen Minh Kieu, 2015).

Louzis et al. (2012) argue that banks can improve their credit quality if they have more opportunities to diversify their investment portfolios. If banks are able to find good projects or invest in companies with growth potential, credit quality will improve. Then, capital used for lending will be lower, reducing the possibility of lending to defaulters. Moreover, using capital to diversify investment forms requires a long time to recover capital and profits while bank capital is mostly short-term capital. Liquidity risk can occur when a large number of customers withdraw money. Therefore, banks with large asset size will have more opportunities to diversify while still controlling liquidity. Meanwhile, a positive correlation between size and bad debt is also found in Le (2016)'s study on the macro-financial linkage with banking performance for East Asian commercial banks and Das and Ghosh (2007) on Indian commercial banks. This is explained by the fact that large banks take risks when expanding their size, leading to increased bad debt.

- Profit efficiency and banking cost efficiency

Most of the empirical studies that have been accessed have found that bad debts and cost efficiency of banks are negatively related (Berger and Humphrey, 1992; Wheelock and Wilson, 1995; Karim et al., 2010). Karim et al. (2010) used both stochastic frontier analysis to measure cost efficiency and Tobit regression model to determine the relationship between bad debts and cost efficiency of commercial banks in Singapore and Malaysia. The research results showed that the relationship between bad debts and cost efficiency is negative. Tsai and Huang (1999) used a cost function to test the relationship between quality management and cost efficiency in the banking industry of Taiwan. They found that asset quality and cost efficiency are positively related. By taking into account risk and quality factors in measuring the cost efficiency of Japanese commercial banks during the period 1993-1996, Altunbas et al. (2000) determined that the level of debt

NPLs are positively related to bank inefficiency. Moreover, banks often experience a decline in efficiency after implementing risk management. On the other hand, Fan and Shaffer (2004) found that although NPLs are negatively related to efficiency, the results are not statistically significant.

Podpiera (2008) tested the relationship between NPLs and cost efficiency using GMM method for Czech commercial banks in the period 1994-2005. The research results support the hypothesis that poor management or a decrease in cost efficiency increases NPLs, but reject the unlucky hypothesis of reverse causality.

Berger and DeYoung (1997) investigated the relationship between loan quality, cost efficiency, and capitalization on a large sample of US commercial banks over the period 1985–1994. Loan quality was measured as the ratio of nonperforming loans to total loans. Cost efficiency was calculated using a nonparametric marginal approach to calculate the annual efficiency score of each bank. The Granger causality test consisted of two equations with each of the three main variables regressed with its own lag and with the other two variables, while the other source was time and cross-sectional data for control. Each equation was estimated by OLS and the sum of the lagged coefficients of each variable generated causal information. The results showed that the relationship between cost efficiency and nonperforming loans was inversely bidirectional, supported by the bad luck hypothesis and the bad management hypothesis.

Williams (2004) presents a more detailed validation of Berger and DeYoung (1997) on a sample of European banks over the period 1990-1998. Loan quality is measured by the ratio of loan loss provisions to total loans. The method of determining CE is similar to Berger and DeYoung. The study concludes that decreases in cost efficiency and profits tend to decrease loan quality, consistent with the poor management hypothesis. Meanwhile, Rossi et al. (2005) extend Williams' study to other European countries on a sample of 278 banks over the period 1995-2002. The study concludes that it is consistent with the bad luck hypothesis, that is, decreases in loan quality reduce cost efficiency and profits.

- Capital structure and operational safety

Keeton and Morris conducted a study on US commercial banks during the period of 1979-1985 by selecting research variables such as return on equity, bank size, and bank risk tolerance expressed through variables such as equity over total assets, and outstanding loans over total assets, to test this hypothesis. The research results showed that bad debts increased for banks with relatively low equity-to-asset ratios. The results of Le (2016) were consistent with this relationship on a sample of commercial banks in 8 East Asian countries during the period of 2005-2014.

The negative relationship between NPLs and capital ratios has also been found in the studies of Berger and DeYoung (1997) and Salas and Saurina (2002). However, there is no consensus on the sign of the regression coefficient between NPLs and the above factors in the research literature. The research results show that the relationship between NPLs and equity is not statistically significant in countries such as India and Greece (Das and Ghosh, 2007; Louzis et al., 2012).

Table 2.3 summarizes previous studies on the impact of specific factors on bad debt of commercial banks in the world as well as in Vietnam. This table also shows that with different samples of different countries, the relationship between factors and bad debt is not uniform.

Comment