A comprehensive model for managing credit risk on home mortgage portfolios

USE OF THE MODEL FOR CREDIT MANAGEMENT The first use of the model is to project aggregate losses over the life of the current portfolio in order to determine the adequacy of reserves for financial reporting and regulatory purposes. Default rates and losses are highly dependent on changes in prices of housing; prepayment rates of fixed-rate mortgages are highly dependent on interest rates. The model, as implemented, provides for the exogenous setting of interest rates, regional changes in unemployment rates, and regional rates of change in housing prices. Each quarter, projections of these key indicators are revised by corporate economists. The model is run with the base (most likely) economic scenario, an optimistic scenario (with lower unemployment and higher appreciation in housing), and a pessimistic scenario (with higher unemployment and lower rates of appreciation in housing). That way, a range is determined for the expected performance of the portfolio with three basic computational passes on a file containing data for each individual mortgage. This analytical approach is much more efficient than using stochastic simulation to produce random samples from models of the joint distributions for the economic indicators, applying the recursive model to the resulting scenarios, and then producing summary statistics for the expected outcomes. In employing the model, we thus opt for analytical efficiency over completeness in representation of the range of possible outcomes.

The model is also used to compare the credit risk associated with different segments of the portfolio and to estimate the credit risk on new loans in light of historical experience. In this application, the beginning state of each active loan is set appropriately (as current, delinquent 30 to 89 days, or delinquent 90+ days), and the model is applied recursively to project the likelihoods of alternative states in annual intervals over the remaining term of the loan. The statistics for the portfolio segments are simply the aggregates of the statistics for individual loans in the segment. In Table 6, we show how the model is used to compare the projected performance of loans according to the size of the original loans and estimated current loan-to-value ratios. The risk summary for each segment includes:

number of active accounts,

total principal amount outstanding,

expected number of defaults before maturity,

expected aggregate loss,

percentage of accounts projected to terminate in default, percentage of outstanding dollars to be realized in credit loss, average magnitude of loss on default,

percentage of the average outstanding amount to be lost on a defaulted loan, average age of the loans (in years) since origination, average remaining term to maturity,

average time that the loans are projected to remain on the books (net of defaults and prepayments),

projected average loss rate per year expected to be on the books, mean of estimated current loan-to-value, and average interest rate for loans in the category. These are the performance statistics on which the credit managers focus in assessing risk and evaluating performance. Presented in this tabular format, the statistics give a summary profile of the characteristics of the chosen portfolio segments. Note how the summaries clearly demonstrate the systematically increasing rates of default and severity of loss as the loan-to-value ratio increases in both size categories. The data also reveal a systematically higher rate of default and loss severity associated with larger loans.