Saving, Wealth and Retirement: Evidence from the Health and Retirement Study

This dissertation examines household saving and retirement decisions using data from the 1992 to 2000 Health and Retirement Study (HRS). After a brief introduction, the second chapter examines the relationship between children and household net worth around the time of retirement. Restricting the sample to first-time married couples and controlling for many factors including lifetime earnings and unearned income, I find that households with more children generally have less net worth than households with fewer, but the difference is small. Moreover, there is no strong evidence that the difference is a result of the adverse effect of college financial aid rules on savings. In addition, there is little evidence that children affect parents' subjective assessments of retirement security. These findings are consistent with the basic principle of the life-cycle framework where parents plan optimally for their retirement consumption. The third chapter examines how wealth changes affect retirement decisions. I focus on the differences between self-reported expected retirement dates in one survey and either the actual or expected retirement dates in the next survey for the same individuals, and relate these to changes in household net worth. Moreover, I recognize that institutional constraints could deter individuals from adjusting their retirement dates to accommodate unanticipated events, such as wealth shocks. I find that the relationship between wealth changes and retirement is generally significant in the expected direction, and more pronounced for the respondents unlikely to face binding institutional constraints. The fourth chapter examines whether Americans are preparing adequately for retirement. My coauthors and I construct a life-cycle model that incorporates detailed HRS data on family structure and age of retirement and allows for uncertainties in earnings, social security benefits, and defined benefit pension receipts. Then, we solve every household's life-cycle problem and, using data on the household's entire history of earnings realizations, derive optimal wealth at given ages. The model explains over 70 percent of the variation in wealth for married households, and over 80 percent for never-married single households. Comparing the derived optimal wealth targets with the observed amount of net worth, we find little evidence that HRS households have under-saved.

Retirement Behavior

12722