Labor Market Scarring from COVID-19: Evidence from the Labor Force Survey

Viet Nam’s headline labor market recovery from COVID-19 looks complete. Aggregate unemployment returned to pre-pandemic levels within about two years, and employment counts recovered even faster. That headline conceals a slower-moving story underneath it. Tracking the same individuals over time using Labor Force Survey panels from 2019 through 2023 shows that a meaningful share of workers displaced early in the pandemic never returned to their prior earnings trajectory, and the scarring is concentrated in ways that ordinary aggregate statistics do not reveal.

Two recoveries, not one

The employment count recovery and the earnings recovery are different processes, and conflating them overstates how complete the adjustment has been. Most displaced workers did find new employment relatively quickly, which is what shows up in the aggregate numbers. What the panel data show is that a large share of that reemployment happened in occupations below the worker’s pre-pandemic position, particularly for informal workers who had been employed in manufacturing and services roles that did not reopen on the same terms.

This occupational downgrading matters because it is sticky. Workers who took a lower-rung job as a stopgap during 2020 and 2021 were, by 2023, still disproportionately likely to be in that lower-rung job rather than back on their original trajectory. The pattern is consistent with a scarring mechanism in which the initial displacement erodes job-specific skills, network access, or employer signaling value in ways that a simple return to full employment does not repair.

Who bore the cost

The earnings losses are not evenly distributed. Women who were displaced during the pandemic show a larger and more persistent earnings gap relative to their pre-pandemic trajectory than men with comparable pre-pandemic characteristics, consistent with women having absorbed a larger share of household care responsibilities during closures and having faced more constrained reentry options as a result. Younger workers, particularly those who were early in their careers when the shock hit, show a distinct pattern in which the damage is less about lost earnings in absolute terms and more about a lost trajectory, meaning the gap between their actual and counterfactual earnings widens over time rather than closing as it typically would for an early-career worker gaining experience.

Informality is the other dividing line. Formal-sector workers who lost jobs during the pandemic recovered earnings substantially faster than informal-sector workers with similar pre-pandemic characteristics. Formal employment appears to have offered a kind of insurance against scarring, whether through severance protections, unemployment insurance access, or simply stronger attachment to firms that survived the shock, that informal workers did not have.

Implications for policy

The findings argue for treating labor market recovery as a distributional question rather than an aggregate one. A policy response calibrated to the headline unemployment rate would have concluded that support was no longer needed well before the scarring documented here had a chance to resolve, because the aggregate numbers recovered while the underlying earnings and occupational trajectories of the most affected workers did not.

Two implications follow directly from the data rather than from general principle. Active labor market programs targeted at informal workers displaced during the shock would have addressed a gap that formal unemployment insurance does not reach, since the informal workforce is largely outside that system by construction. And because the earnings gap widens over time for young workers rather than narrowing, early intervention after a shock is likely to be more effective, and cheaper, than support offered several years later once occupational downgrading has become entrenched.

The broader lesson for crisis response in a labor market with a large informal sector is that employment recovery and earnings recovery can decouple for years, and only individual-level panel data, rather than aggregate labor force statistics, will show that decoupling while it is still possible to address.