Case Study

Forecasting at the USDA’s Foreign Agricultural Service

Challenge: Per a World Trade Organization dispute settlement case, the Foreign Agricultural Service (FAS) at the U.S. Department of Agriculture (USDA) is required to make several changes to its General Sales Manager 102 (GSM-102) Loan Guarantee Program, including, but not limited to, converting GSM-102 into a break-even loan program over a 10-year period. To achieve this goal, FAS required information related to how the various policy changes would affect the government’s budget. Additionally, to achieve a break-even loan program, FAS must adjust the GSM-102 fee rates to be variable risk-adjusted fee rates, or fee rates equal to a borrower’s probability of net default. FAS also requires a fee calculator that calculates risk-adjusted fee rates unique for each borrower’s risk characteristics.

Solution: Summit completed an actuarial study to determine the effect of the necessary policy changes on the GSM-102 Program. The analysis sought to econometrically forecast the impact on the credit subsidy rate of certain loan guarantee characteristics, including the size of the loan guarantee by port value, the terms of the guarantee, importing country’s characteristics, importing bank characteristics, sovereignty status, and Interagency Credit Risk Assessment System (ICRAS). Using historical cash flow data for GSM-102 loan guarantees obligated from 1992 to 2004, the team constructed a Tobit econometric forecasting model to determine whether and the extent to which observed Loan Guarantee (LG) characteristics affect the default probability and the loss (i.e., the un-recovered cash outflow due to claims). This analysis was helpful to various offices within the USDA to determine the monetary impact of changes to the GSM-102 Program. Summit also developed a risk-adjusted fee calculator for the GSM-102 Program that makes use of a historical training data set and a Fractional Logit econometric model to forecast a default probability specific to each unique combination of borrower characteristics, and applied Kaplan-Meier Analysis to historical recovery data to forecast a marginal recovery probability. The fee calculator seeks to assign break-even fees to each unique combination of borrower (risk) characteristics by solving for fees equal to the net present value of expected loss net of recovery. Summit is currently assisting in the deployment of a Web-based version of the fee calculator based on the approved prototype.