A MICROSIMULATED MODEL OF RETIREMENT AND SOCIAL SECURITY POLICY

This is a study of retirement and its impacts on the outlays of the social security system. A microsimulation model has been constructed to integrate a variety of research findings pertaining to the economics of aging. The model was used to estimate a retirement probability function and to simulate the effects on the retirement level and total social security outlays of two policy changes: increasing the mandatory retirement age from 65 to 70 and improving employments markets for unemployed older workers of pre-retirement age.;The population of the simulation consists of a random sample of approximately 1,900 households headed by a person 55 years of age or older, drawn from the March 1971 Current Population Survey (CPS). Eight CPS variables are used for each head and spouse in the household: age, sex, marital status, total income, earned income, social security income, employment status, and occupation. Values for other variables used in the simulation are assigned to the population on the basis of correlations between these values and the personal and financial characteristics of the individuals it consists of. Tabulations of the Retirement History Study were used to produce probability distributions of the number of years an employed person had been on the job. Private pension coverage was assigned using this job tenure variable and published findings of the relationship of job tenure and other characteristics to private pension coverage (Kolodrubetz and Landay 1973). This private pension coverage variable was in turn used to assign the incidence of mandatory retirement, again based on published findings (Reno 1975). Finally, the characteristic of poor health is assigned.;The estimated retirement function generates reasonably good tracking of actual retirement levels and average benefit levels over the 1970 to 1975 simulation period and implies that recent estimates of the power of social security benefits as an inducement to early retirement simulation indicated that an age 70 mandatory retirement age throughout the simulation period would have reduced the number of retirements by 1.8 percent and social security outlays by 2.5 percent.;The second policy simulation "improved" the employment prospects for unemployed workers of pre-retirement age. Such a situation might occur if efforts to reduce age discrimination were successful. The simulation indicated that such a change essentially had no effect on either the retirement level or social security outlays. Although this outcome may be partially the result of underdeveloped labor market modeling, the results of this simulation do not support the view that an "employment strategy" of this sort would significantly reduce either retirement rates or social security outlays.;Given these initial characteristics and assignments, each month the population is aged, subjected to the probabilities of death, incurring a health problem (Zedlewski, 1974), unemployment (Hall 1972; Parnes, King 1976), employment if unemployed (Smith 1977; Parnes, King 1976), and retirement. The parameters of the retirement probability function reflect an attempt at synthesizing the results of five major studies of the retirement decision (Barfield and Morgan 1969; Boskin 1977; Quinn 1977; and Burkhouser 1976, 1977). These parameters are refined using the simulation model itself. Earnings are calculated using a recent estimate of the older worker wage function (Quinn 1977). The retirement decision involves the social security amount, calculated from recent published estimates of the replacement rate provided to persons of various ages at various points in time over the simulation period (Fox 1979). If covered by a private pension, another published estimate of the benefit amount as a function of earnings and tenure is used to calculate the income of the person from this source (Kolodrubetz 1973).