School after job loss: earnings gains take time

Returning to school after a layoff can mean postponing earnings while trying to improve future prospects. Does that trade-off pay? In the arXiv preprint Further Education During Unemployment, Cornell economists Pauline Leung and Zhuan Pei estimate a modest earnings advantage several years after enrollment—and a substantial wait before education costs are recovered.

Their study looks beyond government-funded training programs to a broader route into retraining: enrollment in Ohio’s public community colleges and technical centers. It measures the effects of entering education, not just earning a qualification.

Following workers through job loss and school

The researchers linked unemployment insurance claims, quarterly earnings, public college and technical-center records, and Workforce Investment Act training records. Their analysis covers roughly 1.9 million claims from 1.3 million people who filed eligible claims between 2004 and the third quarter of 2011.

An enrollee was someone who began at a community college or technical center within two years of layoff. Workers enrolled during the two years before layoff were excluded, as were those who initially entered a four-year institution. The final matched analysis included 70,879 enrollees.

These were workers with enough prior employment to qualify for unemployment insurance, not a representative sample of everyone unemployed. Most attended community colleges. About 90 percent took at least one occupational course, but only about 26 percent obtained a credential within four years.

Building a comparison, not running an experiment

Workers were not randomly assigned to school. To estimate what enrollees might otherwise have earned, the authors matched them with claimants who did not enter the observed institutions during the two-year window.

Matches shared the same layoff quarter, gender and prior manufacturing status. A statistical model also used demographics, job tenure, local unemployment and detailed earnings histories—including earnings between layoff and enrollment. That recent history mattered: leaving it out produced markedly different results.

Timing creates another complication. Comparing someone enrolling now with someone not enrolling now can inadvertently compare early schooling with later schooling. The authors instead seek the effect of enrolling versus not enrolling anywhere in the two-year window.

Their method estimates a lower bound under explicit assumptions. One is that, after accounting for observed history, enrollment decisions do not reveal further differences in potential earnings without schooling. Another is that later enrollees would have earned no more without training than comparable workers who never enrolled during the window. Pre-enrollment earnings comparisons support the latter assumption, but cannot establish every unobserved counterfactual. “Lower bound” is therefore conditional, not a guarantee against all bias.

An initial loss, then a gain

During the first two years after enrollment, enrollees earned an estimated $381 less per quarter than their matched comparison group. The authors describe this as a “lock-in” effect: studying may constrain job search and work.

In years three and four, the estimated difference turned positive, reaching $348 per quarter, or about six percent. All these amounts are in 2012 dollars. An alternative, more complex method produced a similar estimate of $351.

The later advantage mainly reflected employment rather than higher weekly earnings. Four years after enrollment, enrollees were about seven percentage points more likely to have recorded employment and nine percentage points more likely to work outside their pre-layoff industry. Neither group, on average, had regained its earlier earnings level.

An accounting breakdown attributed the earnings advantage at that point to industry switchers. Healthcare stood out for both women and men; construction also mattered for men. Related coursework and healthcare credentials aligned with these employment patterns. That is suggestive evidence about mechanisms, not a separate causal test proving which courses produced the gains.

Longer horizons, larger assumptions

For 15,833 workers who enrolled by the third quarter of 2007, the researchers could follow earnings for ten years. The estimated advantage reached 13 percent in year ten. This earlier, pre-recession enrollment group is not interchangeable with the full sample.

Using its earnings trajectory, the authors calculated that private education investment broke even after about eight years, and total investment including public subsidies after about fourteen. Their lifetime return calculations assume a 30-year remaining working life and that year-ten earnings gains persist through year thirty. Those later gains are projected, not observed; the calculations also omit individual uncertainty and nonmonetary study costs.

Important gaps remain. Private-school enrollment and out-of-state earnings are unobserved. Matching cannot eliminate every difference in motivation, anticipated job offers or other circumstances. The records also describe one state and an earlier labor market, with earnings followed only through 2017.

The study’s message is consequently narrower than “retraining works for everyone.” For these Ohio claimants, further education was associated with an initially costly but ultimately stronger earnings recovery, especially alongside movement into new industries. Whether a particular worker benefits depends on more than the average—and the payoff may require years of financial patience.