Further Education During Unemployment

arXiv:2312.17123 · econ.GN, q-fin.EC · Submitted 2023-12-28 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.

Ruth: We're looking at the paper: "Further Education During Unemployment".

Warren: Here's the gist of the paper. Evidence on retraining U.S.

Ruth: unemployed workers primarily stems from evaluations of training programs, which represent a narrow avenue for skill acquisition.

Warren: A study utilizing high-quality records from Ohio and a matching method to estimate the effects of retraining—broadly defined as enrollment in postsecondary institutions—found that enrollees experience earnings gains of six percent three to four years after enrolling, following depressed earnings during the initial two years.

Ruth: These earnings effects are predominantly driven by industry switchers, particularly into healthcare.

Title and authors: Ruth: So we've covered the results of "Further Education During Unemployment," and now let's talk about who did this work and what it means for the broader conversation on retraining.

Warren: The paper is written by Pauline Leung and Zhuan Pei from Cornell University, which is where we often see high-quality empirical work coming from in this area.

Iris: And they are taking a pretty broad approach, using data from Ohio to estimate the effects of retraining for unemployed workers, which is a really big piece of the puzzle.

Ruth: What that means in simple terms is they’re trying to prove that enrollment in postsecondary education counts as retraining, even if it’s not always a specific vocational program.

Warren: And they are linking together different strands of dynamic treatment effect literature to see if they can capture both the treatment now versus later and the treatment versus no-treatment effects simultaneously.

Iris: They are using a matching method to estimate these effects, which is their way of connecting the studies on specific training programs with broader community college enrollment data.

Ruth: So it’s about creating a more comprehensive picture of what happens when people pursue further education while they are unemployed.

Warren: And they’re looking at how this applies to the current polarized labor market, where middle-skilled jobs have been shrinking for a long time.

Iris: They are specifically focusing on the pattern of job displacement and its impact on unemployment among workers without a college degree during economic downturns, which is what motivates this research.

Ruth: It’s about mitigating those social and economic costs they mentioned, like adverse health outcomes and negative impacts on their families.

Warren: And by looking at these effects, they are providing guidance for policymakers on how to structure skill acquisition programs to be most effective.

The paper's summary: Iris: To summarize the main findings of "Further Education During Unemployment," they found that enrollees experience earnings gains of six percent three to four years after enrolling, following some initial depressed earnings during the first two years.

Ruth: That initial period where they see depressed earnings is something that needs to be acknowledged before you get to those positive impacts.

Warren: The earning effects are predominantly driven by industry switchers, particularly people moving into healthcare jobs.

Iris: So the gain isn't necessarily about staying in the same industry; it’s about acquiring skills that open up new career paths in different sectors.

Ruth: They also found that completers contribute disproportionately to earnings growth, with them contributing more than three quarters larger gains than non-completers.

Warren: That means getting the credential really matters for maximizing the positive outcome, not just signing up for a course.

Iris: They also looked at subgroup effects, finding that men in four years post-enrollment see nine percent gains compared to women’s four percent gain.

Ruth: And there are differences based on school type too; technical center enrollees show a more substantial "lock-in" effect than community college enrollees.

Warren: They also looked at the timing, and later enrollment has smaller lock-in in the first two years but larger gains from enrollment up to twelve percent in the third and fourth years.

Iris: And they found that workers laid off during the Great Recession have earnings gains of eight to twelve percent in those third and fourth years post-enrollment.

Ruth: That comparison with pre-recession cohorts, who showed gains close to zero, is a pretty telling contrast about the timing of the layoff.

Warren: So it’s clear that when you get laid off during a recession, retraining has a more pronounced effect than if you were laid off before that downturn.

The paper's improvements: Iris: The authors suggest several ways to improve their study, starting with the dynamic treatment effect estimation itself.

Ruth: They propose using a modified method that preserves static propensity score matching while partially identifying dynamic treatment effect parameters by utilizing those "treatment-now versus later" estimands.

Warren: That sounds like they’re trying to make their identification more robust while still getting the causal effects they need from the dynamic analysis.

Iris: They also suggest refining the matching specification, proposing an exact matching specification along three dimensions: layoff quarter, gender, and manufacturing sector.

Ruth: And they want to use estimated propensity scores derived from a logit model that incorporates pre-enrollment earnings and other covariates to ensure high match quality across those cohorts.

Warren: That’s about making sure the groups they compare are as similar as possible before they start measuring the treatment effect, which is crucial for causal inference.

Iris: They also suggest performing subgroup analysis by restricting the sample to specific subgroups like age, tenure, and race and re-matching enrollees to non-enrollees using estimated propensity scores conditional on that subgroup.

Ruth: And they want to compare community college versus technical center enrollees specifically, looking at those differences in "lock-in" effects we mentioned earlier.

Warren: They also want to decompose the earnings gains into components contributed by completers versus non-completers, which helps estimate how much of that positive effect is driven by getting the credential.

Conclusion: Ruth: So to wrap up "Further Education During Unemployment," we see that retraining enrollment in postsecondary institutions leads to six percent earnings gains three to four years after enrolling.

Warren: That gain is most pronounced for industry switchers, and it’s heavily influenced by whether you complete the program or not.

Iris: And they found that the social net benefit is estimated at twenty-nine thousand one hundred ninety-two dollars, which suggests a dollar for every six dollars invested in this kind of training.

Ruth: It shows that investing in these programs can have a significant positive impact on both individual earnings and broader society.

Warren: The paper's limitations are that the study only uses data from Ohio, so we don't know if those results hold up everywhere else.

Iris: And they also mention that their method doesn't fully capture the dynamic nature of how these effects evolve over time, which is something they acknowledge.

Ruth: But overall, this paper gives us a solid estimate on the effect of further education during unemployment, and it’s a useful piece of evidence for those trying to make policy decisions.

Warren: We’ll be ready for the next paper when we get it; stay tuned.

Cornell University

econ.GN, q-fin.EC

Submitted: 2023-12-28

Updated: 2026-11-06

Comments: Minor revisions to v4: incorporated additional references, corrected small errors and typos, and edited for clarity

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 90/100

The gist: Evidence on retraining U.S.

Key concepts

Dynamic Treatment Effect
This method helps researchers understand how the timing of receiving training affects outcomes. The study uses this to connect studies that look at whether training helps now versus later, and whether it helps compared to no training at all.
Treatment-Versus-No-Treatment Effect
This measures the direct difference in earnings between workers who enroll in retraining programs and those who do not. The study found this effect was approximately 351 more than non-enrollees in the third and fourth years post-enrollment.
Lock-in Effect
This refers to a tendency for workers to remain in a specific field or job after receiving training. The study found that technical center enrollees showed a more substantial 'lock-in' effect compared to community college enrollees.

Terminology

Summary

Evidence on retraining U. p. 0S. unemployed workers primarily stems from evaluations of training programs, which represent a narrow avenue for skill acquisition. p. 0 A study utilizing high-quality records from Ohio and a matching method to estimate the effects of retraining—broadly defined as enrollment in postsecondary institutions—found that enrollees experience earnings gains of six percent three to four years after enrolling, following depressed earnings during the initial two years. p. 0 These earnings effects are predominantly driven by industry switchers, particularly into healthcare. p. 0

The research makes several significant contributions:

  1. It bridges two largely separate empirical literature streams concerning training programs and community colleges by studying the policy-relevant unemployed worker population that intersects with both areas. p. 3

  2. It contributes methodologically to the dynamic treatment effect literature by connecting studies that estimate the treatment-now-versus-later and treatment-versus-no-treatment effects of training. p. 4

  3. It sheds light on the effects of retraining for a recent period, including the Great Recession, within a Rust Belt state characterized by migration away from declining manufacturing industries. p. 4

The study estimates that the lower bound earnings effect among enrollees is approximately 348 per quarter, equating to about six percent, in the third and fourth years post-enrollment. p. 3 The point-identified estimate of the treatment-versus-no-treatment effect is approximately 351 more than non-enrollees in the third and fourth years following enrollment. p. 4 Furthermore, it estimates a long-run effect of retraining at a 13 percent gain in the tenth year after enrollment. p. 7

The cost-benefit analysis indicates that the net present value of average post-tax earnings impacts is approximately 37,056 against an out-of-pocket cost of 6,121, resulting in a private net benefit of 30,936. p. 5 The social net benefit is estimated at 29,192. p. 2 This suggests that the worker receives about a dollar for every six dollars invested. p. 4 The social Internal Rate of Return (IRR) is estimated at eight percent. p. 4

The study concludes that training mainly affects employment and likely has minimal impact on wage rates four years after enrolling. p. 3 In the third and fourth years post-enrollment, enrollees earn approximately six percent more than non-enrollees. p. 3 For industry switchers, the earnings difference between enrollees and non-enrollees in quarter 16 is 103 percent of the overall earnings gain. p. 3 The courses taken and credentials received by individuals in healthcare are related to their work; a large fraction of women employed in healthcare have taken related courses, and credential data suggests that it is not a specific occupation within healthcare driving the results. p. 3

The study finds that completers contribute disproportionately to earnings growth, with completers contributing more than three quarters larger gains than non-completers. p. 4 The post-enrollment earnings growth for enrollees is concentrated among completers. p. 3 The net present value of the average post-tax earnings impacts is 37,056 against an out-of-pocket cost of 6,121, and the social net benefit is 29,192.

The analysis also examines subgroup effects: gains to enrollment are larger for men in the four years post-enrollment (nine percent) compared to women (four percent). p. 3 Earnings effects vary by school type and quality, with technical center enrollees showing a more substantial lock-in effect than community college enrollees. The long-run effect of retraining is estimated at a 13 percent gain in the tenth year after enrollment. p. 7

The study finds that workers laid off when unemployment rates peaked during the Great Recession have earnings gains of 8 to 12 percent in the third and fourth years post-enrollment, while cohorts laid off prior to the recession in 2006 to 2007 show gains indistinguishable from zero. The effect is larger among manufacturing workers than non-manufacturing workers in the third and fourth years post-enrollment. p. 3

For young unemployed workers, additional training is less beneficial if they already possessed some college experience. p. 1 For older (age 40 and above) workers, there are no significant differences in enrollment effects relative to younger cohorts. p. 8 The effects for Whites are larger than for Blacks. p. 4 The effect of enrollment timing varies by quarter: later enrollees have a smaller lock-in effect in the first two years post-enrollment but larger gains from enrollment (up to 12 percent) in the third and fourth years.

The overall treatment effect is estimated to be approximately six percent in the third and fourth years after enrolling. p. 3 The lower bound estimate of 348 per quarter is remarkably close to the point estimate of 351, with an upper bound estimate around 488, or nine percent; together, these bounds pinpoint an informative range for the earnings effects. p. 3 All estimates are highly statistically significant across tested specifications. p. 4

The paper also presents evidence regarding the effect of UI benefit durations on enrollment over the first two years of layoff, estimating that a 10-week increase raises the probability by 0.15 percentage points (or a 10 percent increase), suggesting an externality associated with extending UI benefits. p. 3

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Dynamic Treatment Effect Estimation The system can be improved to estimate treatment effects using a modified method that preserves static propensity score matching while partially identifying dynamic treatment effect parameters by utilizing treatment-now-versus-later estimands, as described in Section 4.1 of the paper.

  2. Matching Specification Refinement The system can implement an exact matching specification along three dimensions—layoff quarter, gender, and manufacturing sector—and use estimated propensity scores derived from a logit model incorporating pre-enrollment earnings and other covariates to ensure high match quality across cohorts.

  3. Subgroup Analysis Capability The AI system can perform detailed enrollment effect analysis by restricting the sample to specific subgroups (age, tenure, race) and re-matching enrollees to non-enrollees using estimated propensity scores conditional on that subgroup, as described in Section 5.2.

  4. School Type Comparison The system can estimate and compare earnings gains for community college versus technical center enrollees by restricting the sample to first observed institution types and using the same matching specification as the main analysis, noting differences in lock-in effects between them.

  5. Credential Effect Analysis The system can decompose earnings gains into components contributed by completers versus non-completers to estimate how much of the positive earnings effect is driven by obtaining a credential, as shown in Appendix Figure A.21.

Related papers