This paper examines the determinants of non-performing loans (hereafter NPLs) in the perspective of the macroeconomic and bank-level variables. Our contributions to the literature are in three folds. First, we employ the fixed effect model controlling for bank-level heterogeneity and the dynamic panel model controlling for autocorrelation to examine the determinants of NPLs growth in the Korean banking sector. On top of that, we extend the sample to commercial banks and savings banks. Including savings bank is crucial since those banks were distressed during the sample periods. Third, this paper analyzes the feedback effects of the banking sector on the real economy using panel vector autoregressive model.
The results show that, among other macroeconomic factors, ‘growth rate of GDP’, ‘growth rate of KOSPI’, ‘growth rate of HPI’, ‘inflation’, ‘risk-free rate’, ‘corporate sector leverage’, ‘government spending-to-taxes’, ‘non-financial sector credit-to-GDP’, and ‘growth of nominal exchange rate’ support the macroeconomic hypotheses. For the bank-level hypotheses, it was found that 1) ‘non-interest expenses-to-total assets’ and ‘total expenses-to-total income’ satisfy the skimping hypothesis, 2) ‘bank size in log’ satisfies the too-big-to-fail hypothesis, 3) ‘liquid assets-to-total deposits’ satisfy the liquidity hypothesis, and 4) ‘total loans-to-total assets’, ‘total loans-to-total deposits’, and ‘interest income-to-total loans’ satisfy the lending supply hypothesis.
Furthermore, as for the impact of the banking sector on the real economy, a simultaneous negative relationship was found between bank's NPLs and GDP. The causal relationship between the NPLs and GDP, on the other hand, was only found in the inflation and credit channels. According to the impulse response functions estimated, shocks in the banking sector result in negative consequences of the real economy for about one year. This implies that there exists the positive macroeconomic effect if individual banks effectively manage their NPLs.