Designing the Bridge Beyond a Temporary Job

The two studies answer the title's question more precisely than a simple “yes” or “no.” The 2010 U.S. Decennial Census hiring program left a long employment trace for applicants near the hiring margin, while the Swedish Stockholm Jobs evaluation found positive employment effects at regular workplaces and negative effects at a specially created worksite. The final judgment this series supports is conditional: a temporary public job is most likely to last in the labor market when it changes a worker's credible next opportunity, not merely their short-term income, and when the program measures the employment, benefit, health, and operating consequences together.

That judgment combines the studies without pretending they are one experiment. The U.S. evidence comes from a threshold-based design that is strongest for applicants near a score of 70. The Swedish evidence comes from dynamic weighting of administrative records and is strongest as a comparison of program-specific paths under selection-on-observables assumptions. Their settings differ, but the contrast is still informative because both ask what remains after a public job ends.

What makes the trace last?

The most durable effect occurs when the temporary job changes the counterfactual that matters. For a worker who would almost certainly have found another job, a public placement may replace one job with another and leave little additional employment trace. For someone near the edge of formal work, the same placement can change whether they obtain a recent work record, a reference, a routine, or a credible route back to employment at all.

This is why the U.S. Census results differ so sharply by applicant position. For marginal applicants near the score-70 threshold, Census hiring increased positive W-2 employment by about 61 percentage points in 2010, and the estimate was still about 28 points in 2023. In the broader within-score comparison, the peak 2010 increase was about 22 points and dissipated by 2011. The effect was graded by score, from roughly 46 points for scores 70–80 to 16 points for scores 92–100. The public job did not have one fixed value. Its incremental value depended on the employment path it replaced.

That distributional pattern has an important implication for design. Reaching people with weak alternatives can produce the largest social gain, but it also requires more than a low entry bar. The Census study finds that low-scoring hires were more than twice as likely to be terminated for misconduct or unsatisfactory performance as higher-scoring hires. More than one in ten hires below 75 were terminated for those reasons, compared with roughly one in twenty above 95. A program that targets the people who stand to gain most may need stronger matching, supervision, and support rather than simply accepting the added operational cost as unavoidable.

The second condition is credible information. A temporary job can leave a trace if future employers can understand what the worker did and what the selection means. The Census opportunity was preceded by an employment test, a background name check, job-specific ranking, interviews, training, fingerprinting, another background check, and employment-eligibility verification. The job was part of a recognizable national public function. That sequence may have made “worked for the Census” more informative than an indistinct period in a program.

The evidence does not tell us that employers consciously rewarded the Census name. The authors consider two possible mechanisms: work experience and training, or the signal of passing screening and performing trusted work. Marginal hires who worked fewer than 40 total hours had later employment trajectories similar to those who worked more, and hires terminated for misconduct or unsatisfactory performance had similar trajectories to those not terminated. These patterns make the act of selection and clearance look important, but they do not directly measure employer beliefs or separate signal from confidence, attachment, and experience.

A third condition is contact with the labor market the worker hopes to enter. Stockholm Jobs makes this visible by placing some participants in regular municipal workplaces and others at a worksite created for participants. The regular workplaces already had managers, coworkers, services, and sector practices. The program-only workplace had organized work and supervision, but a weaker line to an existing employer's future vacancy.

The later employment outcomes follow that distinction. Youth and Other municipal employments at regular workplaces produced positive employment and earnings effects relative to their weighted controls. Between 18 and 36 months after program start, 32.8 percent of Youth employment participants with a workplace were at the same workplace and 49.7 percent were in the same sector. For Other municipal employments, the corresponding figures were 15.5 percent and 35.9 percent. At 36 months, about half of employed participants in both regular-workplace programs worked in education or health, sectors that also hosted many temporary placements.

These figures are not proof that a particular contact caused a later hire.