Ten Years Later, the Training Payoff Looks Different

A workforce program can look different depending on when you measure it. Long-term research offers a more demanding question than whether graduates got a first job: what happens to the advantage after that?

Seven years after people entered the WorkAdvance study, the Per Scholas program had increased average annual earnings by 14 percent relative to its control group. The other three providers showed no statistically significant effect on average earnings that year. MDRC published those findings in March 2022.[1]

Then the researchers kept following the participants.

The September 2025 follow-up found a different pattern in Year 10: St. Nicks Alliance increased average annual earnings by 32 percent relative to its control group. The other three programs, including Per Scholas, showed no statistically significant effect on either of the study’s two main earnings outcomes that year.[2]

For program developers, the interesting part is the changing picture. An earlier gain need not last indefinitely; a later gain need not appear early. We should treat these findings as an invitation to examine a career over time, rather than award a permanent success-or-failure label after the first favorable result.

What the comparison actually measures

WorkAdvance was evaluated through random assignment: people were offered access to its services or placed in a control group that could seek other help in the community. The study involved four nonprofit providers. It did not sample community colleges nationally.[4]

The practical distinction is to compare access to the program with the available alternatives. College leaders should distinguish an earnings impact from a graduate’s salary, an individual’s raise, or a guarantee for a new cohort. Nor should a statistically nonsignificant estimate be treated as proof that the true effect is exactly zero.

Year 10 covers 2021 through 2023 across study cohorts. The prespecified main outcomes were annual earnings and the share earning at least $45,000. These are historical follow-up results, not measurements of people completing training today.[2]

MDRC reports that all four programs increased earnings at some point during follow-up, though at different times. A result for a single year should therefore not stand in for the entire earnings history.[2]

The report’s administrative earnings records do not show participants’ employment sectors. Without that information, the role of the employment sector remains uncertain.[4]

For a college reader, those limits are useful. Consider the studies as evidence about a service model and its changing results, not a ranking of today’s providers or a forecast for a local certificate.

The program was larger than the course

The WorkAdvance model combined sector-focused career preparation, occupational training connected to an industry-recognized credential, employer-linked placement, and services after employment. Its design included coaching for up to two years after random assignment, addressing retention, advancement, reemployment, and problems involving employers.[3]

The clock matters: that support period started at study entry, not graduation. The evaluation tested access to the combined program; it does not isolate the effect of a particular coach, course, or employer relationship.[3]

Our reading is that this changes the program-design conversation. A course map describes what a learner will study. Who helps when a first job proves unsuitable, when advancement calls for another skill, or when returning to college becomes difficult? The design team should consider those questions even when it cannot fund every answer.

Completion and career progress deserve distinct places in that discussion. Treat a credential as an educational achievement. Assess an earnings claim against evidence about economic outcomes and a credible comparison. Do not ask either measure to do the other’s work.

A useful exercise for the next program review

Consider an industrial-maintenance certificate as a planning example, not a program evaluated in these studies. Before adding another credential to its pathway diagram, a college team could work through the transitions the diagram is meant to represent.

Start with the first job. Ask employer partners which responsibilities they would entrust to an entry-level completer, what supervision would be available, and how they would assess readiness. The partnership team should document the difference between an employer’s interest in the program and an actual hiring commitment.

Then examine the next move. What would qualify a worker for a more demanding assignment or a better-paid role? Ask whether the answer involves additional instruction, demonstrated workplace performance, a license, experience, or some combination. Build the proposed next step around that answer rather than assume another certificate is the solution.

Finally, plan for a setback. The college should identify a contact for graduates who lose a job, cannot use their training, or want to return. Decide what assistance is feasible, who owns it, and how the team will learn whether anyone uses it. These are implementation choices to test locally, not a claim that WorkAdvance established the best design for every college.

The tradeoff deserves an honest discussion: continuing support takes staff time that could otherwise serve new learners. A college could start with a limited follow-up effort and examine participation, referrals, and unresolved needs before making a larger commitment. Define what would justify expanding it—and what would prompt a redesign.

Keep the question open after graduation

For the next outcomes review, teams should keep separate records of completion, employment, earnings, and subsequent learning. Ask the research team to identify the cohort, observation period, missing information, and comparison behind each finding. A graduate survey can inform service design; it should not be presented as a causal evaluation merely because the responses are encouraging.

The takeaway is not to wait a decade before improving a program. The opportunity is to make room for evidence that complicates the first answer. A pathway diagram can describe a sequence; colleges should investigate whether people can actually move through it. Training may be short. The question of what it makes possible deserves a longer life.