Many Can Retire Later But Money Plays a Part

As longevity rises, many policy experts contend that people’s working lives will also lengthen. But this argument assumes all workers have experienced the same increase in life expectancy—without factoring in socioeconomic status.

“Does a Uniform Retirement Age Make Sense?,” just released by the Center for Retirement Research at Boston College, examines data on mortality and income. The brief finds that life expectancies for low socioeconomic status individuals have been improving more slowly than for individuals in a higher socioeconomic bracket in recent decades, causing the life expectancy gap to grow.

The brief estimates trends in mortality (the flip side of life expectancy) from 1979 to 2011 by education, a common measure of socioeconomic status. These estimates are then used to see how much longer each educational group can work today if the goal is to maintain the same ratio of retirement years to working years as existed in 1979.

Data from the National Longitudinal Mortality Study (NLMS) was used to estimate the increase in mortality inequality between 1979 and 2011. That study consists of individual-level observations from the Current Population Survey (CPS) matched to data from death certificates from the National Center for Health Statistics. For each individual, demographic and socioeconomic characteristics are obtained at the time of their CPS interview. Individuals are then followed from CPS interview through 2011. If they die, additional information on date, cause, and location of death is collected from death certificates.

The sample used in this study consists of individuals ages 25 or older in their sample year and includes 1.5 million observations. The study defines education by quartiles of educational attainment. Assigning individuals to any one quartile can be difficult. For example, individuals with exactly 12 years of education represent roughly the 40th to 60th percentiles of the education distribution and could be assigned to either the second quartile (25th to 50th percentile) or the third quartile (50th to 75th percentile).

NEXT: How does socioeconomic status impact mortality?

To address this problem, a regression-based approach assigns people to a quartile based on characteristics that are correlated with education level in the overall population (e.g., earnings, industry of employment, race, and family income). To estimate how mortality has changed over time across the education quartiles, the analysis adopts two assumptions. The first is that mortality increases exponentially with age. This assumption is based on research going back almost 200 years and is true until advanced ages.

The second assumption is that, within each gender and socioeconomic group, all ages experience the same annual percentage changes in their mortality rates. These two assumptions make it possible to estimate regressions to find out how much, on average, mortality has improved by socioeconomic status over the last three decades.

Two key findings emerged from the research. First, the expected pattern of growing mortality inequality by socioeconomic status exists: the least educated men and women saw improvements from 1979 to 2011 of 1.5% and 0.5% per year, respectively, compared with 2.5% and 1.2% per year for the most educated. Second, mortality has improved more for men than for women.

While mortality inequality is increasing, the analysis suggests that workers in all socioeconomic groups are likely to live longer today than in the past. As a result, assuming a reasonable health status is maintained, people can work longer while still spending similar proportions of time working and in retirement as those who retired 30 years earlier. Still, policies seeking to extend work lives that treat all workers the same will tend to cut into the retirement of low socioeconomic workers more than high socioeconomic workers. As a result, policymakers seeking to encourage working longer should be cautious about the potential effects that such policies could have on inequality.