 |
 |
Among the earliest examples of such a behavioral model was Curley and Guttentag (1974).1 Their prepayment model utilized ordinary least squares (OLS) on mortgage cohorts because information on the individual loans or pools was unavailable.
The dependent variable in such behavioral models typically was the proportion of a given cohort that prepaid in a given period. The independent variables were limited to age and interest-rate related variables for each cohort. Thus, limited data availability resulted in fairly simple models.
3. Pool Level Models: 1980-1990
The birth of the secondary mortgage market in the 1980's provided the catalyst for better empirical models of mortgage termination. While Ginnie Mae issued its first MBS in 1970 and Freddie Mac its first MBS in 1971, the agency market remained small until Fannie Mae issued its first security in 1981. At that point, the agency market, as defined by conforming loan size and underwriting standards set by the three agencies, began its explosive growth.
As originating lenders began to sell many of their mortgages into the burgeoning agency market, investors demanded better information on the termination-related factors that dictated the returns on their mortgage-related investments. Because the agencies’ MBS were collateralized by pools of mortgages, the agencies began to disclose a basic set of attributes describing each pool at origination – weighted average characteristics such as coupon, maturity, etc. – and the overall termination factor for each pool on a monthly basis.
Peters, Pinkus, and Askin (1984) and Richard and Roll (1989) are good examples of the first behavioral models built upon this newly available agency data. Using Freddie Mac pool information, they utilized as dependent variables the aggregated characteristics of the mortgage pools, such as average loan size, average loan age, and weighted average coupon (WAC).
1 Curley, A. J. and Guttentag, J. M. (1974). Explorations in Economic Research, “The Yield on Insured Residential Mortgages”, pages 114–161. National Bureau of Economic Research.
