Summer Faculty Research Proposals 2002-2003

Project title: “Household Decision-Making and the Causes and Effects of the Post-Retirement Consumption Decline”

Investigators:

Steven Stillman Research Fellow, RAND Corporation stillman@rand.org

Jennifer Ward-Batts Assistant Professor of Economics Claremont McKenna College jennifer.ward-batts@mckenna.edu

Project Description:

Evidence from several countries indicates that households reduce consumption expenditures substantially around the age of retirement. This pattern has been documented for the U.S. by Hamermesh [1984]; Mariger [1987]; Bernheim, Skinner, and Weinberg [2000]; and Lundberg, Startz, and Stillman [2001]; for Canada by Robb and Burbridge [1989]; and for the U.K. by Banks, Blundell, and Tanner [1998]. The consumption decline appears to be fairly widespread across consumption categories, rather than concentrated on work-related expenses, and to take the form of a discrete drop at the year of retirement.

This behavior is puzzling, since life-cycle consumption models predict that households will want to smooth consumption when they experience a predictable drop in income, as at retirement. After examining alternative explanations that are consistent with forward-looking life-cycle behavior, most researchers have attributed this consumption drop to myopic behavior or to the systematic arrival of discouraging information at retirement. However, an alternative possible explanation is suggested by a collective model of household behavior. Most wives expect to live several years longer than their husbands, and therefore prefer, absent perfect altruism, for the household to consume less as the couple ages than husbands do. Given this, and assuming that the husband's bargaining power depends upon his current income or employment status, the husband's retirement from a career job should cause a deterioration in his relative influence on household decisions and a decline in the couple's consumption spending.

Lundberg, Startz, and Stillman (LSS) test this hypothesis by comparing the post-retirement consumption change of married couple households to single households using food consumption data from the Panel Study of Income Dynamics (PSID) for the years 1979 - 1986 & 1989 - 1992. They find that expenditures drop at retirement by 8 to 10 percent for married couples, but do not decrease significantly for single-person households. The magnitude of the consumption drop is also found to be increasing in the relative age of the husband. These results lend some support to a collective rather than unitary approach to the decisions of older couples, and suggest that changes in relative bargaining power explain the commonly observed post-retirement drop in the household consumption of married couples.

Lundberg and Ward-Batts (2000) present additional evidence that the relative bargaining power of wives may affect the amount a couple saves for retirement. They find that, among married couple households in the Health and Retirement Study (HRS), those where the husband has substantially more education than the wife have lower net worth.

This project proposes to repeat the analysis in LSS using data from the HRS, and to explore the value and feasibility of doing the same with Family Expenditure Survey (FES) and/or British Household Panel Study (BHPS) data from the UK. This proposed project has three main goals: (1) to examine whether the post-retirement consumption decline previously found for married couples in other data also occurs for couples in the HRS; (2) to validate the results in LSS using a different sample of retiring US households; and (3) to ascertain whether further research should be pursued using UK data. Further research will be pursued, and additional funds sought from the National Institute on Aging, in the following cases: (1) If results indicate some impact of a shift in bargaining power at retirement, HRS data will be further used to analyze whether the relative health status of spouses changes post-retirement, possibly due to the wife's relative increase in bargaining power. In addition, HRS data will allow us to test some implications of the marital bargaining model that LSS were unable to examine using PSID data. (2) If we determine that either or both UK datasets can be used to do the same tests proposed in this study using HRS data, we will pursue that additional research. Using UK data would be an important additional contribution both to test whether the collective model framework is partly responsible for the unexplained portion of the consumption drop found by Banks, et al, and to show that the result is not particular to the US.

The HRS is a national longitudinal survey of the 1931-41 birth cohort of Americans and their spouses or partners, with oversampling of blacks, Hispanics, and Florida residents. The survey began in 1992 with a sample of over 12,600 persons in 7,600 households, 81% who were either married or living with a partner. Follow-up surveys have taken place biannually beginning 1994. As of the 1998 survey, the age-eligible participants were between the ages of 57 & 67 and approximately 12% of the household heads were retired. The survey includes, in each wave, data on household food expenditures, and extensive data on household assets and savings. Individual health-related measures in each wave include self-reported height, weight, ordinal health status, life expectancy, smoking, and alcohol consumption, in addition to doctor's diagnoses of conditions such as diabetes, stroke, heart attack, psychological problems, and arthritis. The data also include physical functioning measures for ADLS and IADLS, and cognitive tests, such as word recall and counting backwards by 7.

The HRS has three major advantages over the PSID for analyzing consumption and health changes surrounding retirement: (1) The HRS, because it focuses on a cohort approaching retirement age, should have a larger sample size of individuals with at least one observation both before and after retirement than the PSID. This should improve the precision of the estimates and will allow us to test whether the post-retirement consumption drop is more pronounced for married couples where the husband is older, has a shorter life expectancy, or is in worse health than the wife. Data limitations in the PSID combined with the small sample size made it difficult for LSS to test these important implications of the marital bargaining model. (2) The HRS contains extensive data on household assets and savings, which will allow us to examine whether savings patterns change with retirement and whether household wealth affects post-retirement consumption changes separately for both singles and married couples. (3) The health and behavioral questions available in HRS will allow us to examine whether the hypothesized change in marital bargaining power leads to relative changes in the welfare of spouses. Health problems are likely to be important concerns for individuals near retirement age. If the husband's retirement really reduces his marital bargaining power, we should find improvements in the wife's health relative to her husband's, and under certain assumptions decreases in the husband's smoking and drinking relative to his wife's.

The analysis in this project will use a simple fixed effects regression model to examine how household consumption and savings, and individual health and behavior, change after household heads and their wives retire. In all cases, the post-retirement change for married couples will be compared to the change for singles to control for other common factors which may cause post-retirement changes in the dependent variables. We anticipate that this project will result in a publishable paper which further examines the causes of the post-retirement consumption drop for married couples. The analysis of post-retirement changes in health is more exploratory in nature. A finding that the relative health of spouses changes after retirement will lead to a more comprehensive R03 grant proposal to further examine this important issue. That proposal would plan to take advantage of exam-based health measures, which may be included in future rounds of the HRS, to better measure changes in health status after retirement.

References

Banks, James, Richard Blundell, and Sara Tanner, "Is There a Retirement Savings Puzzle?" American Economic Review, September 1998, 88:4, 769-88.

Bernheim, B. Douglas, Jonathan Skinner, and Steven Weinberg, "What Accounts for the Variation in Retirement Wealth Among Households?" 2000, forthcoming American Economic Review.

Hamermesh, Daniel S., "Consumption During Retirement: The Missing Link in the Life Cycle," Review of Economics and Statistics, February 1984, 66:1, 1-7.

Lundberg, Shelly and Jennifer Ward-Batts, "Saving for Retirement: Household Bargaining and Household Net Worth," Michigan Retirement Research Center Conference Paper 00-02.

Lundberg, Shelly, Richard Startz, and Steven Stillman, "The Retirement-Consumption Puzzle: A Marital Bargaining Approach," Journal of Public Economics, forthcoming.

Mariger, Randall P., "A Life-Cycle Consumption Model with Liquidity Constraints: Theory and Empirical Results," Econometrica, May 1987, 55:3, 533-57.

Robb, A. L. and J. B. Burbridge, "Consumption, Income, and Retirement," Canadian Journal of Economics, August 1989, 22:3, 522-42.

Additional related previous research by investigators:

Ward-Batts, Jennifer, "Health, Wealth, and Gender: Do Health Shocks of Husbands and Wives Have Different Impacts on Household Wealth?" Michigan Retirement Research Center Working Paper 01-06.

Ward-Batts, Jennifer, "Out of the Wallet and into the Purse: Modeling Family Expenditures to Test Income Pooling," Population Studies Center Research Report No. 00-466.