Magazine article The RMA Journal

Economic Capital for Consumer Loans

Magazine article The RMA Journal

Economic Capital for Consumer Loans

Article excerpt

No single economic capital approach has yet emerged as best practice for consumer lending. This article discusses mean variance, asset value correlation, and econometric models. Work continues on model development as lenders seek the best framework for analyzing the risk of consumer loans.

Consumer lending has been the late bloomer in the family of credit risk capital models. It's not surprising that commercial loans were first to receive attention from capital modeling pioneers. After all, larger loans to corporations represented bigger individual risks, and commercial loan losses had historically been more volatile than consumer loan losses. Further, the size of commercial loans justified the collection and analysis of more borrower-specific information, and the market provided observable information about the estimated ability of public companies to generate cash flow to repay debt as they built value for shareholders.

Modeling for consumer lending, on the other hand, developed in a different direction. Although lenders were reluctant to rely on commercial loan default models, consumer lending presented an ideal environment for the application of scorecards and other default-prediction techniques. Most banks had far more consumer loans than commercial accounts and numerous defaults. Many banks wanted to avoid the expense of individually analyzing the information they collected on their consumer borrowers, and this data was easily amassed in databases where analysts could identify patterns and link behavior to characteristics of the borrower or loan. Such expected loss models were especially important to support securitization activities, which developed both to fund these loans and to avoid the regulatory capital required for low-risk consumer assets.

Indeed, understanding expected behavior remains by far the most important element in managing the risk of consumer lending. Consumer loan losses are not nearly so vulnerable to swings in the economy as commercial loans, but large, unexpected losses have unfortunately cost quite a few banks' shareholders plenty. In most cases, these losses occurred not because the economy hurt otherwise healthy borrowers, but because the borrowers were not as healthy as originally believed (or hoped). Adequate modeling identifies the characteristics that differentiate performance and segments accordingly, so that the lender knows when performance is likely to change. Good risk management is aware of the competitive environment, sensitive to changes in acceptance rates, and on guard against adverse selection. Changes in product offerings are tested in a disciplined fashion so that high losses are limited to test cells. It's also necessary to understand the timing of losses to avoid being misled by the fact that most consumer loans have lower loss rates in their first months on the books than in their second and third years.

The expected behavior of consumer loans must also consider the role of prepayments--a factor that is not as important in commercial lending. Estimating lifetime loss rates requires a projection of future losses and future balances. Further, as many lenders discovered in 2003, higher-than-expected prepayments do not come evenly from all parts of the portfolio. Good customers are often more likely to refinance, leaving a shrinking pool of paying customers to offset charge-offs. In such cases, a bank must extract additional revenues from the remaining portfolio--through additional usage, lees, or increased interest rates--or risk having losses overtake their margin income. Even if prepayments have only a modest effect on loss rates, the value of high-quality consumer loan portfolios may be driven more by prepayment rates than by credit losses.

It is difficult to overstate the importance of modeling expected behavior for consumer lending. The best capital models will produce meaningless results if the inputs are wrong. But good analytics provide a solid basis on which capital models can be built. …

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.