Recent studies across a number of countries evidence a substantial decline in household consumption expenditure around the time of retirement. This phenomenon, coined as the retirement consumption puzzle, brings a challenge to the traditional life-cycle model. The life-cycle model implies that household consumption should be continuous over time including the transition to retirement, provided that retirement is a foreseeable event.
To address the retirement consumption puzzle, this dissertation brings current evidence by carrying out four studies on U.S. household consumption behavior at retirement. Study 1 is a cross-sectional study using the latest available data wave from 2009 HRS CAMS. The main interest is to compare consumption behavior between non-retired and retired households. Based on the life cycle hypothesis, regardless of employment status, households sharing similar socioeconomic characteristics should exhibit similar consumption behavior. The empirical findings, however, show that the consumption behavior between non-retired and retired households is significantly different, holding all other factors constant. Spending for retired households is 8.5% lower than spending for non-retired households.
Study 2 is an aggregate-panel study using data from 2001-2009 HRS CAMS. The difference between Study 1 and Study 2 is that Study 2 tracks household consumption behavior over time and investigates whether there is a significant change in consumption pattern after retirement. Fixed-effects analysis is conducted to appropriately account for the effect of individual heterogeneity. The life cycle hypothesis predicts that when retirement is as planned by the household, there is no significant change in consumption after retirement. The fixed-effects regression indicates that there is an insignificant increase of 3% in household consumption after retirement. The discrepancy in the results from Study 1 and Study 2 regarding the retirement consumption puzzle comes from the lack of control of individual heterogeneity in the cross-sectional analysis.
Study 3 is a subsamples-panel study. Within the life-cycle framework, if retirement is voluntary and expected, there should be no significant change in household consumption after retirement. However, if the household head is laid-off or experiencing poor health, this household may be forced to be out of the labor market, resulting in early retirement. Because involuntary early retirement reduces the household’s lifetime resources, the household must reduce consumption and re-allocate to a new optimal consumption path. As predicted by the life-cycle model, the findings show that households with voluntary retirement have an insignificant increase in household consumption after retirement by 6%. Involuntary-retirement results in a decrease in consumption after retirement by 6-7%. The percentage difference in the household consumption change between these the two groups of households is 11-12% and statistically significant. Accordingly, these findings suggest that household consumption behavior differs by household type (voluntary versus involuntary retirement). These three studies use log consumption as the dependent variable. Alternatively, Study 4 uses consumption growth as the dependent variable to test the retirement consumption puzzle. The regression results provide no evidence of a retirement consumption puzzle, finding an insignificant 3% increase in consumption after retirement.