The short-run impact of Covid-19 on retirement behavior in America

Two papers in this collection, by Goda et al. (Reference Goda, Jackson, Nicholas and Stith2023) and Quinby et al. (Reference Quinby, Rutledge and Wettstein2023), use the Current Population Survey to explore how the negative shock to employment following the Covid-19 pandemic affected labor force behavior. Accordingly, they address the question of whether the shock pushed older workers into retirement sooner than they might otherwise have planned and whether this might be one cause of the current backlog of unfilled positions. In addition, Goda et al. look at how the availability of enhanced unemployment benefits, which were paid for varying periods in different states in the initial stages of the pandemic, affected labor force behavior and also how the decrease in employment affected disability insurance claiming.

These two papers differ somewhat in the age groups and time periods they study. Goda et al. look separately at persons age 50–61 and 62–70, because the older group can claim social security retirement benefits as early as age 62. In addition, because disability rates rise with age, there may be more workers on the margin of claiming disability insurance in their 60s than in their 50s. Thus, there could have been an interaction effect between age and the pandemic shock to employment on rates of disability claiming. Quinby et al. use a more elderly sample than Goda et al., namely those aged 55–79. In addition to the age differences between the two studies' samples, the time periods differ between the two as well. Goda et al. use data from five years before the pandemic to two years after it (2015–2022), while Quinby et al. use data from two years before the pandemic and one year after.

Despite these differences, both of these studies reach similar conclusions with respect to labor force behavior during the first year of the pandemic. Specifically, they find that the negative shock to employment in the first few months of the pandemic led initially to a small increase in workers who reported themselves retired in the first year of the pandemic. This stability was consistent with the negligible change in the rate at which workers claimed social security retirement benefits in the pandemic's first year. During the second Covid year, however, Goda et al. report that social security retirement claiming rose about 3 percentage points from the previous year. They interpret this increase as older workers taking a ‘wait-and-see’ attitude, rather than claiming retirement benefits in the first year of the pandemic but as time passed more of them decided to retire. Quinby et al. show that those most affected by the initial increase in unemployment were those without a college degree, Asian-Americans and part-time workers.

Unlike earlier periods of high unemployment, Goda et al. detect a decrease, rather than an increase, in disability benefit claiming in pandemic years. The decrease was concentrated among those applying for supplemental security income (SSI), and those applying for SSI and social security disability income (SSDI) concurrently; claiming rates among those applying for SSDI alone were unaffected. Why this was so is a bit unclear, but Goda et al. point out that expanded unemployment benefits may have provided liquidity in the first 18 months of the pandemic, and so this lessened the pressure on households at the margin to claim disability benefits. That explanation is consistent with their finding that the expiration of more generous unemployment benefits increased disability insurance claiming, although the increase was only found among those concurrently applying for SSI and SSDI. Another, not mutually exclusive, reason for the decrease in disability claiming was the closure of social security field offices during the first two years of the pandemic, potentially raising the cost of claiming for those who would have otherwise applied for benefits in person.

Factors influencing where older persons find jobs

The proportion of older workers differs considerably across industries and individual firms. At the industry level, the prevalence of older workers depends, in part, on whether firms in an industry are relatively new or established long ago, whether the industry is growing or declining, the physical demands of jobs, and employer compensation policies. At the firm level, hiring and promotion practices of firms influence whether older workers are hired or retained at greater rates. The first paper in this section examines the long-run changes in the proportion of older workers across various industries. The second study focuses on hiring patterns and whether discrimination against the hiring of older workers is more prevalent in some industries and firms than others.

The long-run upward trend in employment rates for older men and women since the mid-1990s is well known. But underneath that slower-moving aggregate trend, there have been fast-shifting cross-industry movements. Allen and Wang (Reference Allen and Wang2023) examine the incidence of employment of older workers by industry and document which industries have growing and which have shrinking shares of older workers. Further, they compare firm and worker characteristics by industry, to detect associations of these characteristics with the upward or downward trends in older worker shares. The authors combine worker age and worker characteristics from the American Community Survey with capital data from the Bureau of Economic Analysis, along with firm and worker characteristics from the Current Population Survey.

Important potential economic drivers of the age structure of work are technology and international trade. With increasing use of technology such as robots, younger workers may have a greater ability to learn or adapt if the technology is complementary with labor. By contrast, if the technology substitutes for labor, then younger workers may be more likely to be displaced. Similarly, where trade impacts industries' growth or decline, those that shrink may end up with a higher share of older workers because younger employees tend to be the first to be displaced. Allen and Ting report that the share of workers age 55+ rose on average from 13% in 2001 to 23% in 2019. Of course, this also reflects labor supply factors (as the baby boomers' movement into the older worker category), and potentially labor demand factors as well.

This analysis yields several interesting results, with three that stand out. First, industries that offer pensions are associated with a larger increase in the share of older workers. This may derive from a ‘golden handcuff’ effect, with older workers less willing to leave their jobs when that entailed a loss of accrued pension benefits. Second, industries with higher high-tech capital to labor ratios had larger increases in older-worker shares. Thus, there is no indication that the high technology sector has a younger-worker bias. Third, industries more exposed to import substitution through trade with China showed larger increases in older-worker shares, particularly in the 2001–2007 period. Older workers may have been less mobile than younger workers in the face of the import shock, meaning that there was a larger shrinkage in younger worker headcounts than for older workers, and firms more exposed to import substitution may have reduced entry-level hiring.

The evidence coming from this paper shows much nuance regarding work patterns among older individuals. Some findings – like pension coverage being associated with a higher share of older workers – are not surprising, but the fact that high-tech industries and industries subject to disruption from trade shocks are less intuitive. Worker mobility, wages, and experience matter for understanding how the industrial age structure of employment reacts to a changing economic environment.

Age discrimination violates US law, yet it tends to be difficult to assess when age discrimination occurs. Typically, comparisons are made between hiring rates for older applicants compared to the size of the entire applicant pool. However, the size and composition of the applicant pool itself may be manipulated by the wording of job ads. Explicit age cutoffs in job ads are not allowed (except for special circumstances), though more subtle wording choices could potentially discourage applications from potential older workers to the same ultimate effect.

The study by Burn et al. (Reference Burn, Firoozi, Ladd and Neumark2023) investigates the possibility that language around job skill requirements in job ads are perceived by potential applicants as discriminatory. They start with a set of ageist stereotypes drawn from industrial psychology, such as being less physically able or less adept with technology. Next the authors identify language that evokes these stereotypes. They build language examples using machine learning techniques that search for text containing particular words together with a stereotype such as the ability to lift 40 pounds or to use software such as Microsoft Word. Finally, they conduct a survey to determine whether people perceive the chosen language as biased against older workers.

Their algorithm can identify several strings of language describing tasks associated with discriminatory stereotypes. With the survey sample, they test these phrases against a neutral phrase describing a similar job requirement. The empirical evidence suggests respondents find the discriminatory phrases to be biased against older workers. Interestingly, when asked if other people might find the phrases biased, the results are stronger. The results in this paper indicate a potential path forward for lawmakers and regulators interested in curtailing age discrimination in hiring. Identifying subtle discriminatory language in job ads could allow investigators to hone their hunt for potential manipulation of the applicant pool.

Key findings from this research

The papers in this special issue highlight a series of important issues associated with the aging of the population. The aging of the labor force is one of the most important long-run trends affecting the American economy. Allen and Wang show that workforce aging varies considerably across industrial sectors implying that future research needs to examine sectorial differences in explaining the impact of aging on firms. Burn et al. show that age discrimination continues to exist at the hiring stage. They focus on the wording of job announcements and how this can lead to fewer old workers applying for open positions. The impact of job announcements on the likelihood of older workers applying for certain jobs creates a discriminatory effect that is difficult to measure and affects the degree of measured discrimination.

The Covid-19 pandemic caused major disruptions to the labor market including large increases in unemployment, decline in monthly earnings, and modification in working conditions. Papers by Quinby et al. and Goda et al. examine how these economic shocks altered labor supply decisions of older workers. They conclude that in the short-run there was little change in retirement rates or claiming of social security benefits. The finding of limited impact on these choices during the first year of the pandemic may be due to the response of the federal government to provide significant monetary support to low-income households. The long-run impact on retirement rates remains to be seen.