대출회수율의 결정요인에 관한 연구： 신용보증 대출을 중심으로

While the modelling of the probability of default-henceforth PD-has been at the center of credit risk studies over the last several decades, a burgeoning literature on bank loan recovery rate only has emerged lately in line with the advent of new Basel Capital Accord capital guidelines, the prevalence of loan securitization, and the growth of credit derivative market. A majority of empirical studies concerning bank loan recovery rate-henceforth BLRr-could be fallen into one of two fields：the effect of negative correlation between PD and BLRr on credit loss of loan portfolio and the determinants of BLRr. However, to the best of our knowledge, a few reports are released to show basic characteristics of BLRr in Korea lending market, such as descriptive statistics of recovery rate. The aim of this paper is to empirically investigate the determinants of BLRr in Korea. The factors affecting BLRr documented from existing research can be classified into four categories：contract-specific characteristics,firm-specific characteristics, industry conditions, and macroeconomic conditions. None of those but collateral has drawn a consistent conclusion on BLRr. The variety of the empirical results shows that BLRr depends to a high degree on the dataset utilized, the differentiation between the practices of loan, and the calculation method, etc. In this context, after describing mean value and quantiles of BLRr in Korea lending market, we shed further light on the impact of collateral, the volume of loan, the size of firm, the state of the economy, the creditworthiness of the company,the age of firm, the intensity of the client relationship, and financial char-acteristics on the BLRr in Korea lending market. Our sample is from the Korea Credit Guarantee Fund’s recovery database. The data coverage is Korea small and medium sized enterprise default event of loan guaranteed. The dataset employed contains information about 30,797 companies, which are defaulted and recovered in the years from 1997 to 2008. Recovery rate is calculated as the nominal amount of recovery divided by the amount of defaulted loan. First of all, similarly to Calabrese and Zenga (2010), we found out a bi-modal distribution with peaks at the intervals from 0 to 0.05 and from 0.95 to 1. The fraction of recovery rate very close to 0 is the highest. By industry, the mean recovery rate of construction and service sectors is higher than that of manufacturing sectors. Meanwhile, as a result of analyzing factors influencing the recovery rate with the use of generalized linear regression model, dummy variable indicating whether the creditor has the collateral right on debtor’s asset or not appears to have a significantly positively effect on the recovery rate. The ratio of tangible asset to total asset as proxy variable for collateral is closely related to BLRr. This result reveals that collateral plays a dominant role in determining BLRr. In addition, we find that there is the negative relationship between the volume of loan and the recovery rate. On the contrary, the bigger the loan size is, the more the recovery amount is. This result implies that the recovery amount doesn’t necessarily increase in proportion to the loan size. A larger size of firm leads to a higher recovery amount. The size of firm is defined as the logarithm of total asset recorded on the most recent balance sheet before default. The age of firm is significantly positively related to BLRr and the amount of the recovery. Even though the impact of the intensity of client relationship on BLRr isn’t statistically significant, in view of positive correlation between client relationship and the age of firm,companies with intense client relationship with financial institution exhibits a higher recovery rate. It also appears to be the positive relationship between leverage and BLRr. Whereas, both PD and GDP growth rate do not have a significant effect on BLRr. We caution that these findings might be specific to the nature of our sample and this question is worth further examination on alternative samples in the future.