Volume 10 Issue 1 - January 2009
Retirement Wealth Across Cohorts: The Role of Earnings Inequality and Pension Changes by Ann Huff Stevens
Wealth and income at retirement are the result of many forces that accumulate over the lifetime: labor force participation, wage levels, public policies, and savings decisions, among others. As a result, changes in any of these factors may affect the retirement wealth of cohorts in future years. Major changes in the United States’ labor market over the past 30 years - most notably increasing wage and earnings dispersion - predict changes in the resulting distribution of wealth for cohorts now on the verge of retirement. This study investigates how changes in earnings during the 1970s, 80s, and 90s translate into differences in the level and distribution of retirement wealth for birth cohorts now approaching retirement.
This study uses data from the Health and Retirement Study (HRS) to compare two birth cohorts whose prime earnings years differ substantially in terms of the level and dispersion of earnings. The first cohort, initially observed in 1992, was born between 1936 and 1941. The second cohort, the youngest of the baby boom generation, is initially observed in 2004, and was born between 1948 and 1953. For each of these cohorts, I use lifetime earnings measures from the HRS averaged over ages 36 to 51 and estimate the relationship between lifetime earnings and three measures of retirement wealth. The study asks what the wealth distribution of the later cohort would have been if they had retained the same distribution of lifetime earnings that prevailed during the working years of the earlier cohort. This provides an estimate of the degree to which changing earnings distributions can explain changes in wealth.
I examine three different components of overall retirement wealth: (1) private, nonpension wealth, (2) the sum of private pension wealth and non-pension wealth, and (3) expected wealth from Social Security benefits. Both the underlying mechanisms (how do changing earnings lead to changes in wealth?) and the size of the effect (how much of the wealth changes can be attributed to earnings changes?) depend on the specific component of wealth considered.
Comparing Wealth Earnings Across Cohorts
Starting with the wealth measure that includes only non-pension, non-Social Security wealth, Table 1 shows the increased dispersion in the distribution of wealth among the second cohort, and the corresponding increase in dispersion in lifetime earnings. Among men, the ratio of wealth at the 25th percentile of the distribution to that of the 75th is approximately .17 among the earlier cohort, and falls to .07 in the later cohort. Among women, dispersion also increases, with the ratio of wealth at the 25th to 75th percentile falling from .14 to .07.
Earnings and Non-pension Wealth
Next, I ask how much of this change in wealth is related to the changes in earnings that occurred over the last three decades. To do this, I predict what the distribution of wealth for cohort 2 would have been if they had faced the same distribution of lifetime and current earnings as cohort 1. Table 2 summarizes this analysis. The first row shows that 23% of cohort 1 men had wealth below $19,300, but among cohort 2 this figure had risen to nearly one-third, or 32%. The third column shows that, if cohort 2 had experienced the earnings distribution of the earlier cohort, the fraction with very low wealth would still have risen, but not by as much, to 29%. Looking at the fraction with wealth levels below $53,000, there continues to be an increase across cohorts (from 48 to 54%), but now holding earnings constant at those of cohort 1 suggests that half of the observed change is related to earnings changes. Among women, there is a similar increase in the fractions with low wealth across cohorts, but very little of this is explained by changing earnings. This is because women’s lifetime earnings rose substantially between these cohorts, and so it is not surprising that rising earnings are not connected with falling levels of wealth.
Earnings and Total Wealth
The next panel of table 2 summarizes the same exercise, but adds the present value of expected pension wealth to the previous non-pension wealth measures. When pension wealth is included, earnings play an even larger role in explaining the increased fraction of baby boomers with low wealth. Almost the entire change in those men with total wealth below $24,000 (and more than the observed change in those with total wealth below $82,500) is explained by the changes in earnings. While we would expect individuals to adjust their non-pension savings behavior to earnings changes, it may seem surprising that the effect of earnings grows when pension wealth is added. The probable mechanism here, however, is unlikely to involve individual savings or investment behavior. Instead, this reflects the strong correlation between earnings and non-wage compensation, including pensions. Men in cohort 2 who faced declining earnings also appear to have faced reductions in their employer-based pension wealth.
Earnings and Social Security Wealth
Finally, how might Social Security wealth be affected by changing earnings distributions? Note that, holding lifetime earnings constant, Social Security benefit levels will increase across cohorts, because the benefit formulas are designed to hold replacement rate roughly constant. Thus, as average earnings rise, Social Security benefits rise as well. This is reflected in the final row of Table 2, which shows that the fraction of men and women with Social Security wealth below $188,000 will fall from 49% (for men) and 73% (for women) among cohort 1 to just 38% (for men) and 50% (for women) among cohort 2.
The story is somewhat different when we look at the fraction of men with lower levels of expected Social Security wealth. Twenty-three percent of men in each cohort are expected to have total Social Security wealth of less than $147,000. The lack of a reduction in the fraction of men with very low wealth (compared with rising Social Security wealth at higher points in the distribution) reflects the fact that declining earnings among this segment of men is just offset by the growth in benefits expected over time with constant real earnings. Among women, in contrast, there is a substantial reduction in the fraction of the cohort with very low wealth, driven by both rising earnings and the benefit formulas.
It should come as no surprise that the major expansion in inequality over the last three decades of the twentieth century would show up in the accumulated wealth levels of individuals whose working lives were centered around these years. On the other hand, much economic research shows that relatively little of the variation in wealth can be related to individuals’ lifetime earnings. This study shows that earnings changes are, in fact, strongly associated with changes in wealth for those cohorts who are now close to retirement. For men in the bottom half of the earnings distribution, reduced earnings (relative to earlier cohorts) are reflected in lower levels of wealth accumulated prior to retirement. Changes in pension values appear to be strongly correlated with these earnings changes, so that the effects of earnings are even larger when pensions are added to the wealth measures.
Finally, Social Security wealth continues to grow across cohorts, but earnings changes also affect this source of wealth. For males at the bottom of the earnings distribution, real growth in Social Security benefits expected at a given level of lifetime earnings is just offset by the decline (in real terms) in their earnings relative to earlier cohorts.
Ann Huff Stevens is Associate Professor of Economics at the University of California, Davis. Her past research on retirement and aging includes studies of the responsiveness of retirement expectations to pension plans and Social Security, the effects of worker knowledge about pensions on retirement timing, and the effects of job loss on older workers’ employment, earnings and wealth.