Who Measures VR Success?

What this post is and what it isn’t

This post examines how “success” in the public vocational rehabilitation (VR) program gets measured and announced, using two states’ October 2026 National Disability Employment Awareness Month (NDEAM) documents and the federal record behind them. It draws only on the statutes, federal reports, state documents, and peer-reviewed research cited below.

It is not an accusation against Nevada, Alaska, or any other state. It does not claim that any figure cited here is false. Nevada’s headline number comes from federal performance data, and this post takes it at face value. Nevada and Alaska appear here as illustrations of a national pattern, not as targets.

It also does not compare like with like across years, and readers should notice when the system itself doesn’t either: the earnings figure discussed below is from Program Year (PY) 2024, while the most recent federal monitoring analysis of Nevada’s program covers PYs 2019 to 2021. That mismatch is part of the subject. Different instruments, different periods, and no rule requiring them to appear side by side.

This post is policy analysis, not legal advice, and not a ranking of states. Its question is a governance question: who selects the indicators of VR success, who frames the populations those indicators describe, and who gets to declare that the program is working.

The headline

On October 1, 2026, opening NDEAM, the Nevada Department of Employment, Training and Rehabilitation announced that, according to federal performance data from the Rehabilitation Services Administration (RSA) for PY 2024, VR Nevada participants working after leaving the program had median quarterly earnings of $6,307, 18% higher than the national median for VR participants. The same release reported that VR Nevada provided pre-employment transition services (Pre-ETS) to 2,571 students across 64 schools in 14 districts in Federal Fiscal Year 2025. Rehabilitation Division Administrator Drazen Elez framed the month as an opportunity to celebrate the value and talent that people with disabilities bring to Nevada’s workforce every day [1].

Nothing in that paragraph is invented. But notice what the release does not print: the national median itself. A reader who wants the benchmark has to leave the celebration and excavate federal performance tables. RSA’s own PY 2024 national report lists $5,331 in the Median Earnings column on the “Total Statewide, Actual” row [3], a figure arithmetically consistent with the release’s 18 percent. (RSA’s Nevada PY 2024 report lists $6,307 on the same row, about 18% above $5,331.) The point is not concealment. The point is genre. A press release selects, and selection is the subject of this post.

Notice, too, what the number is. Under Section 116 of the Workforce Innovation and Opportunity Act (WIOA), median earnings is one of the primary indicators of performance for the VR program, defined as the median earnings of participants who are in unsubsidized employment during the second quarter after exit [2]. Three quiet facts live inside that definition:

The denominator is the employed. Everyone who exited the program without a job is outside the calculation entirely. The median describes a subset, presented in the headline as the program.
The cohort is a prior one. A second-quarter-after-exit statistic is, by construction, a report on people who left the program in an earlier period. October’s celebration is always of people the system served some time ago.
A median is silent about distribution. Half of employed exiters earned less than $6,307 in the quarter. The figure says nothing about hours, benefits, job match, or whether the work uses the training the participant received.

None of this makes $6,307 a bad number. It makes it an incomplete sentence.

The view from the monitor

The last time federal monitors examined Nevada’s VR program closely, they read high earnings from the other end. RSA’s FFY 2023 monitoring report, analyzing the agency’s own RSA-911 case service data for PYs 2019 to 2021, found a training pipeline that was thin and thinning [4]:

Bachelor’s degree training reached 111 participants (2.9 percent) in PY 2019, 92 (2.6 percent) in PY 2020, and 82 (2.2 percent) in PY 2021.
Occupational or vocational training fell over the same period, from 373 participants (9.6 percent) to 196 (5.2 percent).
In PY 2021, the agency reported spending $274,477 on bachelor’s degree training, against $1,048,094 on assessment services.

RSA’s reviewers then connected the two halves of the ledger in one sentence. That explanation is the agency’s own, recorded by RSA, not RSA’s independent conclusion: the limited training services were consistent with the agency’s high quarterly median earnings, which the agency attributed to low-skilled jobs in the warehouse and casino industries [4]. The review team recommended the agency build training capacity for higher-skilled employment, including within those same industries [4].

The monitors also had findings about the measuring instrument itself: inconsistent reporting of essential data elements between the agency’s case management system and the RSA-911, and post-exit employment and earnings reported through the State Wage Interchange System without a written internal control process for data collection and validation of post-exit data and wage information [4].

Now the fence, stated plainly: PYs 2019 to 2021 are not PY 2024. Nevada’s training mix may well have improved since the monitoring review. I looked for a published PY 2024 service mix from RSA or the state that could be set beside the earnings headline, and did not find one. That absence is itself the measurement-governance point. The system produces an annual celebration built on the newest outcome figure. It does not require the structural numbers, meaning who got trained, in what, at what cost, to travel with it.

Celebration without numbers

Alaska shows the other pole of the same culture. On October 12, 2026, Governor Mike Dunleavy proclaimed October 12 to 16 as Employment First Awareness Week in Alaska. The proclamation notes that Alaska became an Employment First state in 2014, praises the collaboration of its disability employment network, recognizes youth training programs, and affirms competitive, integrated employment as the preferred option for people with disabilities [5]. It contains no outcome data at all: no earnings, no employment rates, no counts of people served, no comparison to any prior year.

Neither document is dishonest. Read together, they describe how NDEAM works as a measurement ritual. One state offers a number without its context. Another offers a celebration without a number. In both cases, the entity reporting success is the entity that delivered the service, and October arrives on schedule whether the underlying data does or not.

Who measures: the sociology of a success statistic

Every success statistic silently answers three prior questions, and each is answered by someone with interests in the answer.

Who selects the indicator? Congress made the first selection. WIOA Section 116 fixes the menu of primary indicators, including employment rates in the second and fourth quarters after exit, median earnings, credential attainment, and measurable skill gains, and performance levels on those indicators are negotiated between each state and the federal government [2]. But within that menu, the agency chooses which dish to serve in October. Nevada served median earnings. Alaska served none.

Who frames the denominator? By statute, the earnings median covers employed exiters only [2]. That is a defensible methodological choice, but it is a choice, a decision about who counts, and it vanishes from the headline. A median built on everyone who exited, employed or not, would tell a different story about the same program in the same year.

Who declares success? The agency that delivers the service also compiles the case data, submits it through the RSA-911, writes the press release, and picks both the comparison (“18% higher”) and the comparison’s level of visibility (the benchmark unprinted). The monitored and the monitor’s press office are the same institution.

Scholars of public administration have a vocabulary for what happens next. Strathern’s gloss on Goodhart’s law, that when a measure becomes a target it ceases to be a good measure, describes the pressure any celebrated indicator comes under [6]. The empirical literature on performance systems in public employment programs documents the predictable responses. Courty and Marschke found training agencies under the former Job Training Partnership Act timing participants’ program exits to the measurement calendar, which raised measured performance and incentive awards while the affected participants’ earnings suffered [7]. Heinrich and Marschke show that such responses are dynamic: organizations learn the incentive system over time, and their strategic behavior evolves with it [8]. Studying performance-based contracting in social welfare services, Koning and Heinrich identify cream-skimming, meaning preferential service to clients easiest to move across the measured line, and parking of those hardest to serve [9].

None of this requires bad faith, and this post alleges none. It requires two conditions: that the measured thing and the mission are not identical, and that the measurement is the part that gets rewarded, funded, or celebrated. VR adds a wrinkle the JTPA literature did not have to confront. In a state economy with strong entry-level wages in logistics and hospitality, median earnings can be genuinely high while the training pipeline is genuinely thin. A good median and a weak program are not a contradiction. RSA’s monitors said exactly that about Nevada in PYs 2019 to 2021 [4]. The number was real. So was what it concealed.

What a credible success regime would require

If the problem is measurement governance, the remedies are governance remedies. Four follow from the record above:

No orphan numbers. Any public earnings claim should be published with its benchmark, its cohort definition, and its denominator in the same document. If a reader cannot tell who is inside the statistic, the statistic is marketing.
Pair outcomes with service mix, same period. Earnings headlines should travel with the same year’s RSA-911 service data: how many participants received training, of what kinds, at what expenditure. A median without a training mix cannot distinguish a skilled workforce from a favorable industry mix.
Verification as routine, not event. RSA’s Nevada findings on data consistency and wage-record controls [4] argue for continuous, published data-quality auditing rather than episodic monitoring reviews separated by years.
Longer horizons. Second- and fourth-quarter snapshots reward rapid placement. Careers are longitudinal. A credible regime follows cohorts for years, tracking earnings trajectories, credential completion, and advancement, and reports what it finds even when the trajectory flattens.

A note on the SCII

This blog’s State Community Integration Index (SCII) scores states on community integration using published federal and state data rather than agency self-description. The discipline attempted there is the one argued for here: benchmarks printed next to claims, denominators visible, evidence periods labeled, and indicators chosen before results are known rather than after. Whether the measure belongs to an index like mine or to a governor’s proclamation, the questions do not change: who selected it, who is inside it, and who checked.

Three questions for any success claim

Which cohort? (Whose outcomes, from when?)
Which denominator? (Who is inside the number, and who was excluded before the counting started?)
Who verified it? (Anyone other than the agency announcing it?)

A claim that answers all three is evidence. A claim that answers none is a press release. Both will arrive this month, dressed alike. The difference is not the number. It is who measures.

Sources

Nevada Department of Employment, Training and Rehabilitation, “VR Nevada Highlights Services During National Disability Employment Awareness Month,” press release (Oct. 1, 2026), https://detr.nv.gov/content/media/National_Disability_Employment_Awareness_Month.pdf
Workforce Innovation and Opportunity Act § 116, 29 U.S.C. § 3141 (primary indicators of performance for the vocational rehabilitation program, including median earnings of participants in unsubsidized employment during the second quarter after exit).
U.S. Department of Education, Rehabilitation Services Administration, ETA-9169 Annual Report, PY2024 National, “Total Statewide, Actual” row, Median Earnings $5,331, https://rsa.ed.gov/sites/default/files/publications/annual-reports/2024/ETA-9169-AnnualReport-PY2024-National.pdf
U.S. Department of Education, Rehabilitation Services Administration, Federal Fiscal Year 2023 Monitoring Report on Nevada Vocational Rehabilitation Division Vocational Rehabilitation and Supported Employment Programs, March 27, 2024 (onsite review June 13 to 15, 2023), analyzing RSA-911 data for PYs 2019 to 2021, https://cms.detr.nv.gov/Content/Media/202405_Tab_11_FFY_23_Final_Monitoring_Report_RSA.pdf
Office of the Governor, State of Alaska, “Employment First Awareness Week 2026,” proclamation (Oct. 12, 2026), https://gov.alaska.gov/employment-first-awareness-week-2026/
Strathern, M. (1997). “‘Improving ratings’: Audit in the British university system.” European Review, 5(3), 305-321.
Courty, P., & Marschke, G. (2004). “An empirical investigation of gaming responses to performance incentives.” Journal of Labor Economics, 22(1), 23-56.
Heinrich, C. J., & Marschke, G. (2010). “Incentives and their dynamics in public sector performance management systems.” Journal of Policy Analysis and Management, 29(1), 183-208.
Koning, P., & Heinrich, C. J. (2013). “Cream-skimming, parking and other intended and unintended effects of high-powered, performance-based contracting in social welfare services.” Journal of Policy Analysis and Management, 32(3), 461-483.