ô Planning Paper I Toward Informed Decision-Making On Retirement Policy: A Portfolio Of Studies Related To The Retirement Decision pages 17-18 III. METHODOLOGICAL ISSUES AND SUBSTANTIVE APPLICATIONS õ The goals emerging from this portfolio (see p. 6-7) are ambitious ones: to improve forecasts of retirement and related trends, and to evaluate options for influnecing when people choose to retire. But what problems need to be solved to improve forecasts, and what options should be evaluated? This section considers these two questions. In addition, it can be applied to help resolve issues in each of the four substantive areas corresponding to federal objectives for the retired elderly. ô Forecasting the Retirement Decision: What Problems Should Be Addressed? õ In our review of the literature on retirement, we identified over 50 factors that appear to influence the retirement decision. These factors can be organized into three categories. The first consists of government programs such as social security, the personal income tax, and various regulations governing employer practices. The second category is comprised of individual-level factors, such as demographic and family-related characteristics, private pension coverage and other financial matters, health, and attitudes toward work and retirement. Finally, there are factors related to the employer and the market place, such as the employer's size and industry, type of pension plan provided, policies about retaining older workers, and concerns about productivity, costs, and upward mobility for younger employees. Also included in this last category are macro-economic conditions such as inflation and unemployment. Research about the effect of these factors on the retirement decision is summarized in Appendix B. What is the importance of these factors (and the retirement decision in general) in making predictions about retirement trends, the needs and status of retirement systems, and the like? As described earlier (page 3), the retirement decision figures most heavily in long-range forecasts, since these are highly sensitive to shifts in the demographic characteristics of the population, whereas short-term forecasts, in contrast, are most sensitive to economic changes. Long-range forecasting models tend to be micro-simulations, while the short-range ones are macro-economic or actuarial in nature. (Specif examples of all three types of models are provided in Appendix C). One example of how the factors influencing the retirement decision are typically handled in these microeconomic simulations is seen in the DYNASIM 2 model developed by the Urban Institute. From existing survey data (in this case, the Longitudinal Retirement History Survey), a regression anaalysis is done to model the extent to which factors such as demographic characteristics, health, and financial status have affected current retirement decisions. This model is then used in conjunction with other models which were developed to exlain or predict employment patterns, occupational and demographic changes, and pension coverage and benefits. The final set of models, DYNASIM, is then applied to a new sample of individuals to forecast interatively annual changes, such as in the decision to withdraw from the labor force.