Academic journal article Journal of Money, Credit & Banking

Measuring the Default Risk of Small Business Loans: A Survival Analysis Approach

Academic journal article Journal of Money, Credit & Banking

Measuring the Default Risk of Small Business Loans: A Survival Analysis Approach

Article excerpt

CREDIT-RISK MODELS ARE used in a wide variety of lender applications, ranging from underwriting and pricing to account management and capital allocation. A number of techniques have been used to estimate one of the key components of these models--the likelihood of default over time. In this paper we demonstrate the advantages of explicitly capturing the effect of time on the probability of default for a sample of medium-maturity (i.e., seven years) Small Business Administration (SBA) guaranteed loans using a survival analysis/hazard model approach. We use a discrete-time hazard procedure--a computationally straightforward estimation procedure statistically similar to a Cox Proportional Hazard model--that is designed to make full use of the effect of changes in economic conditions over time. Our results show that the likelihood and timing of a medium-maturity, SBA loan defaulting is conditional on several key borrower, lender, and loan characteristics, as well as changes in the economic environment over time. Moreover, we find that as the medium-maturity loans season the likelihood of default increases initially, peaks in the second year after oligination, and declines thereafter.

Although our analysis is limited to data on loans originated under the SBA 7(a) loan guarantee program, our results are likely to be of general interest for several reasons. First, although there have been several studies that focus on small business credit accessibility (Scott and Dunkelberg, 2003, Berger, Frame, and Miller, 2002, Cole, Wolken, and Woodburn, 1996, Peek and Rosengren, 1996, Berger and Udell, 1995), data availability problems have limited research on the performance of small-business firms receiving loans. As a result, little is known about the performance of loans that represent a significant portion of the total commercial and industrial loans held by U.S. commercial banks--accounting for roughly 11% of the total. SBA loans make up a large percentage of the total volume of small-business loans originated by commercial banks. The SBA data then allow us to examine the default behavior of a large segment of the small-business loan market. We show that medium-maturity SBA 7(a) loans are concentrated in the relatively more risky segment of the loan market but they are comparable in quality to a large percentage of loans held by larger commercial banks

Second, using survival analysis techniques, we show that not all business credits are of equal default risk and that a bank's exposure to loss due to default is not constant, but varies significantly over the life of the loan. This has important implications for the recent proposed changes to the Basel Capital Accord that would impose a uniform capital charge against small business credits.

Finally, we outline a method of predicting defaults that fits well within a net cash flow modeling framework. Within that framework, the time path of default is as important as the event itself for measuring a lender's exposure to losses. The discrete-time hazard modeling approach allows us to explicitly account for time and capture the impact of changing conditions on the likelihood of default using time-varying covariates. Although we develop our approach within the framework of the 1990 Federal Credit Reform Act (FCRA), the method we use to estimate the default model is germane to the development of any model used for risk-based loan pricing, allocation of reserves and capital, and the valuation of pools of securitized loans. (1)

The remainder of the paper is organized as follows. In Section 1, we present a brief overview of the SBA 7(a) loan guarantee program including a comparison of the cumulative default rate experience of SBA loans with rated corporate bonds over the same time horizon. We outline several unique features of the guarantee program that characterize the SBA's exposure to losses and motivate our specification of a conditional default model. We also summarize the default history by loan attributes and cohort segments, and use a simple non-parametric hazard technique to identify the general pattern in the timing of default unconditional on other factors. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.