본 연구는 가계대출의 뇌관으로 지적되고 있는 저신용자 다중채무자 과다채무자 등 신용 취약계층의 대출행태에 대한 분석을 실시했다. 즉 신용 취약계층 별로 대출금액 연체금액 대출자 수에 대한 금융기관 간 상호연계성을 살펴보았으며 이들의 대출금액이 시스템적 리스크에 미치는 영향도 분석했다. KCB가 제공한 50만 명 대출자의 2006년 1월부터 2015년 6월까지의 통합자료 분석 결과, 신용 취약계층 차주는 은행보다 비은행금융기관을 보다 많이 활용하며 이들의 대출과 관련된 금융기관 간 연계도도 전체대출자에 비해 높은 것으로 밝혀졌다. 취약계층별로는 저신용자가 다중채무자나 과다채무자에 비해 금융기관 간 연계도가 높은 것으로 나타났다. 또한 과다채무자의 대출금액이나 다중채무자 및 과다채무자의 연체금액의 경우 은행은 주로 타 금융기관에 영향을 주고 비은행은 영향을 받는 것으로 나타났다. 개별 금융기관의 연계성 분석에서는 다중채무자 및 과다채무자의 대출금액의 경우 상호저축은행이, 저신용자와 다중채무자의 연체금액의 경우 보험의 금융기관 간 연계성이 가장 높게 나타났다. 신용 취약계층이 시스템적 리스크에 미치는 영향에 대한 분석에서는 신용 취약계층의 총 대출금액이 은행권 및 비은행권에 관계없이 시스템적 리스크에 4개월 선행하여 영향을 미치는 것으로 밝혀졌다. 또한 결합재무곤경확률(JPoD)로 측정된 시스템적 리스크에 대해서는 다중채무자의 은행 및 비은행권 대출금액이, 한계기대부족액(MES)으로 측정된 시스템적 리스크에 대해서는 과다채무자의 비은행권 대출금액이 시스템적 리스크에 대해 그랜저 인과관계가 존재하는 것으로 나타났다.
This study examines loan behavior of vulnerable households with low credit scoring, multiple loans, or over-indebtedness that poses downside risks to economic growth. The purpose of the paper is to measure the interconnectedness among financial institutions regarding loan balance, delinquency balance, and the number of borrowers by vulnerable household type and to evaluate the impact of loan balance by vulnerable household type on systemic risk.
Interconnectedness among financial institutions in loan behavior by vulnerable household type is analyzed using Granger-causality networks by Billio et al. (2012). Financial institutions include seven individual commercial banks and five non-bank financial industries such as card, capital, insurance, union, and mutual saving bank. I also observe interconnectedness among financial institutions regarding loan behavior of the entire households for normal borrowers. In network analyses, this paper computes the DGC (Degree of Granger Causality) presenting the fraction of statistically significant Granger-causality relationships among all pairs of financial institutions, the closeness centrality measuring the averaged shortest path between a financial institution and all other institutions reachable from it, and the eigenvector centrality showing the importance of financial institutions based on how connected they are to the rest of the network.
Systemic risk is measured by the joint probability of default (JPoD) that estimates the probability of all the banks in the system becoming distressed using CDS spread of individual banks. This paper also employs the marginal expected shortfall (MES) to compensate the defect of the JPoD, excluding non-bank financial institutions in estimation of systemic risk. Out of four methods for the MES by Brownless and Engle (2012), I utilize the RSF(Rolling Static Factor) model based on EWMA (Exponential Weighted Moving Average) to measure the size of market capital shortfall. In this model, an entire market systemic risk is the integration of individual financial institutions` capital shortfall calculated by the marginal expected shortfall of 43 non-bank financial institutions as well as ten commercial banks.
Using the monthly data from the KCB for the loan behavior of 500,000 household borrowers from January 2006 to June 2015, this paper shows that vulnerable households use more commercial banks than non-bank institutions and that the financial institutions used by vulnerable households display a more severe degree of interconnectedness than those used by entire households. It is also observed that low credit scoring borrowers lead to higher interconnectedness among financial institutions than multiple loan borrowers or over-indebted borrowers. In terms of the granger causality between financial institutions regarding loan behavior used by vulnerable household type, this paper finds that commercial banks mainly granger-cause other institutions while other institutions granger-cause non-bank financial institutions with respect to the loan balance and the delinquency balance of over-indebted borrowers as well as the delinquency balance of multiple loan borrowers. In addition, mutual saving banks rank the highest degree of interconnectedness among financial institutions regarding the loan balance of multiple loan borrowers and over-indebted borrowers. Insurance ranks the highest regarding the delinquency balance of low credit scoring borrowers and multiple loan borrowers.
This paper finds that the loan balance in banks as well as that in non-banks of the vulnerable households granger-causes the systemic risk. The systemic risk by JPoD is granger-caused by the loan balance of multiple loan borrowers, whereas the systemic risk by MES is granger-caused by the loan balance in non-banks institutions of over-indebted borrowers. This study also discovers that the increase of systemic risk is followed by four months of the loan balance increase in vulnerable households. In particular, systemic risk by JPoD rises three months subsequent to the shock on the loan balance of multiple loan borrowers, while systemic risk by MES increases three months after the shock on loan balance in non-bank institutions of over-indebted borrowers.
This study provides various substantial policy implications for the design of financial market stability. The result that the vulnerable households utilize more non-bank financial institutions implies that financial authorities should pay more attention to household debt from non-bank institutions. It is especially suggested that authorities need to introduce new loan regulations based on the total loan balance of non-bank institutions and to consider revised loan policies or financial support systems for the vulnerable households.