Three microeconomic studies of pensions and retirement savings

This thesis comprises three essays using micro-data from the American Survey of Income and Program Participation. The Survey of Income' and Program Participation (SIPP) is a set of independent short panels collected by the U.S. Census Bureau.; In the first essay, I reexamine the empirical support of some implications of the Life-Cycle Hypothesis. In particular; I estimate age-wealth profiles for US households allowing for cohort effects. The analysis employs a data set which has not been previously analyzed in this way: the Survey of Income and Program Participation (SIPP) which contains wealth information from 1984 to 1995. The main findings suggest that elderly households continue accumulating assets after retirement, which contradicts the standard lifecycle model. However, my descriptive results also reveal an overall lack of substantial savings by most US households. Therefore, more than the observed lack of dissaving upon retirement, it is this lack of savings which undermines standard Life-Cycle models.; The second essay studies the impact of occupational pension portability on interfirm mobility in the United States. I explore the role of employer provided pensions on job mobility choices using data from the Survey of Income and Program Participation. Defined benefit plans are found to have a significant negative effect on mobility. However, I find no significant evidence that potential pension portability losses deter job mobility among workers covered by these plans. I also find that the portability policy change implemented by the Tax Reform Act of 1986 had only minor effects on mobility. Puzzlingly, defined contribution plans, although fully portable, are found to have an impact similar to defined benefit plans. Evidence of compensation premiums accruing to workers in pension, union and health insurance covered jobs supports the view that workers are less likely to leave “good jobs”.; My last essay extends the general framework of Arellano and Bover (1995) for efficiently estimating models with individual effects to the case of repeated short panels. Many data sets such as the SIPP, CPS and CEX, have the structure of a repeated short panel (RSP). A linear error-components (or fixed-effect) model can be estimated on such data either by exploiting the genuine (short) panel structure or on a pseudo-panel of cohorts. We propose combining both sets of moments in a General Method of Moments framework. If the panels of a RSP overlap in time, then the panel and pseudo-panel moments are not redundant. In this case, combining them is not only more efficient, iv but also, the minimized GMM criterion then gives an (over identification) test of whether the individual and aggregate (cohort) level movements in the data can both be accommodated by the error-components model. As an application we estimate conventional age-financial wealth profiles on the U.S. Survey of Income and Program Participation. Equality of the panel and pseudo-panel estimates is strongly rejected by the data. In particular, the true-panel estimates imply substantial post-retirement dissaving, while the pseudo-panel estimates do not.