Pandemic Labor Force Participation and Net Worth Fluctuations

Abstract

The US labor force participation rate (LFPR) experienced a record drop during the early pandemic. While it has since recovered to 62.2 percent as of December 2022, it was still 1.41 percentage points below its pre-pandemic peak. This gap is explained mostly by a permanent decline in the LFPR for workers older than 55. This article argues that wealth effects driven by the historically high returns in major asset classes such as stocks and housing may have influenced these trends. Combining an estimated model of wealth effects on labor supply with micro data on balance sheet composition, we show that changes in net worth caused by realized returns explain half of the drop in LFPR in the 2020-21 period and over 80 percent of "excess retirements'' during the same period.

Introduction

The COVID-19 pandemic threw the US economy into a short but deep recession, during which the labor force participation rate (LFPR) fell by 2.2 percentage points (pp)—its largest drop on record. While the LFPR quickly rebounded, it remained 1.41 pp below its pre-pandemic peak as of the end of 2022. Researchers, market practitioners, and policymakers alike have argued that the drop in the LFPR may be contributing to a shortage of workers and excessive tightness of the labor market, which in turn may be contributing to inflation remaining high (Powell, 2022). However, a closer inspection of the data indicates that while the LFPR for prime-age workers (under age 54) has mostly recovered, the LFPR for older workers seems to have permanently fallen and failed to recover. 

Applying the pre-pandemic peak LFPR to the level of the civilian noninstitutional population aged 16 and over as of December 2022 suggests that there were about 3.73 million "missing workers'' in the US economy. Additionally, the share of the retired population has increased considerably, over and beyond what long-run demographic and shorter-run business cycle trends would predict. Comparing actual retirements with those predicted by a statistical model that allows for such trends (Montes, Smith, and Dajon, 2022) leads us to conclude that there were 3.27 million "excess retirees'' in the US economy as of December 2022. 

At the same time as the US economy was recovering from the short pandemic recession, real returns boomed across various asset classes, namely stocks and housing. Driven partly by the reopening of the US economy after the 2020 lockdowns and partly by the robust monetary and fiscal policy responses to the macroeconomic effects of the pandemic, real returns for many assets were abnormally high during the years 2020 and 2021, at least when compared with the historical record. The cumulative real return on a diversified index of stocks, the S&P 500, surpassed 35 percent between December 2019 and December 2021, versus a historical average of 17.5 percent for a two-year period. The real return on housing was close to 20 percent over the same period, versus a historical average of 7.3 percent. 

Neoclassical theory offers a natural connection between these two patterns: wealth effects on labor supply. Standard models of labor supply assume that leisure is a normal good, with its desired consumption rising in response to an increase in income or wealth. There is broad empirical evidence for this effect, especially when such increases in wealth are unexpected (Imbens, Rubin, and Sacerdote, 2001). Moreover, these effects are likely to be more salient and relevant for older individuals, who are nearing retirement age, and for whom extensive-margin labor force participation decisions may be more elastic with respect to unexpected changes in wealth (Cheng and French, 2000; Zhao, 2018).

In this article, we try to quantify what share of the missing workers and excess retirees may be plausibly attributed to rising asset values during the pandemic. We proceed in three steps: First, we use the 2019 wave of the Survey of Consumer Finances (SCF) as a representative sample of the balance sheet composition of US individuals at the beginning of the pandemic. In particular, we can estimate exposures to major asset classes such as stocks, housing, government bonds, and corporate bonds. Second, we impute realized returns on these asset classes to compute how the net worth for each individual changed during the pandemic given their initial portfolio composition. Finally, we use an empirical model of wealth effects on labor supply (Benson and French, 2011) to estimate the impact of the estimated changes in net worth on labor force participation decisions at the individual level, and we aggregate these estimates using the appropriate weights. The final output is an estimate of the number of people who left the labor force due to changes in asset values. Since these estimates are generated at the individual level and are then aggregated, we can produce them for different demographic groups. 

In our baseline, most conservative exercise, we focus on people aged 55-70, whose retirement decision is plausibly more sensitive to wealth effects. We find that the predicted change in the LFPR for this group accounts for almost 30 percent of the drop in aggregate LFPR between 2020 and 2021. If we expand the analysis to all those 55 and older, we can explain over 50 percent of the drop in aggregate LFPR over the same period. In terms of "excess retirements,'' we can explain close to half of the observed excess retirements by considering only the 55-70 age group and close to more than 80 percent when focusing on all those 55 and older.

There are many reasons why people may have chosen to retire early or leave the labor force during the pandemic period other than unexpected changes in wealth. Older people were at greater risk of severe illness and death from COVID-19, which undoubtedly played an important role for those with occupations involving greater physical contact. Many news stories also reported on older members of the household being responsible for taking care of loved ones as childcare facilities or other daycare institutions closed due to government-mandated lockdowns. Wealth effects may not have been the only reason why people chose to retire but may have rather compounded these other reasons by allowing people to retire (as opposed to causing the retirement). We also report estimates for the 2020-2022 period, which include 2022, a year of declining asset valuations. While asset values declined during this year, there were no significant changes in terms of the LFPR, so our model explains a smaller share of the drop during this period. There are several other explanations for why the LFPR has failed to recover, such as reduced immigration flows starting in 2020 (Peri and Zaiour, 2022).

Our work aligns with several studies that put forward rising (declining) asset values as a driver of early (late) retirement decisions and declines (increases) in the LFPR. Coronado and Perozek, 2003 find that individuals who benefited from the bull stock market in the 1990s retired earlier than those who did not. Benson and French, 2011 argue that sudden and unexpected declines in asset values, especially housing, during the Great Recession led to delayed retirements and higher than the expected LFPR. Goda, Shoven, and Slavov, 2011 find that individuals exposed to stock market declines during the Great Financial Crisis of 2007-08 delayed retirement but that this effect was partly attenuated by worsening labor market conditions. In more closely related work, Favilukis and Li, 2023 argue that increases in housing wealth can fully explain the "Great Resignation''  among older workers and that metropolitan statistical areas with more substantial price growth tend to have a lower LFPR for homeowners around retirement age. 

Our work is also related to the literature that dissects the post-pandemic drop in workers and hours worked. Lee, Park, and Shin, 2023 focus on the decline in aggregate hours worked post-pandemic and use micro data to decompose it into intensive and extensive margins. They find that more than half of the decline in total hours worked is due to a decline in the intensive margin, i.e., workers who remained in the labor force but reduced the number of hours they worked. Furthermore, they find that those whose hours declined tend to be prime age, educated men who tended to work long hours and had high earnings before the pandemic. They argue that this helps explain why the labor market remains tight even after the partial recovery in the LFPR. However, in our study, we abstract from the intensive margin and focus on extensive-margin decisions only. 

Hobijn and Sahin, 2022 argue that the decline in the LFPR may be overstated as it does not account for either (i) the fact that the LFPR was probably above trend pre-pandemic due to business cycle factors or that (ii) there are natural long-term downward trends on the LFPR due to demographics. Furthermore, Garcia and Cowan, 2022 show that both women and men saw a reduction in work hours and the likelihood of working full-time in response to school closures. However, only women were less likely to work at all. These effects were concentrated among uneducated parents in occupations less likely to be compatible with telework.