When a Temporary Public Job Becomes a Long Employment Trace
This series examines two large temporary public-employment programs: the U.S. Census Bureau's hiring for the 2010 Decennial Census and Stockholm Jobs, a set of municipal programs in Sweden. The focus is not simply whether participants earned wages while the programs were running, but why their later employment paths diverged—and what readers can learn about judging a temporary job by the bridge it creates to ordinary work and income security.
The puzzle is easy to state. A temporary job has a known end date, yet a new study of Census hiring finds employment effects that remain visible through 2023, roughly fifteen years after the hiring period. In Stockholm, temporary employment at a regular municipal workplace improved later employment, while work at a site created specifically for program participants produced negative employment effects. These results do not make “public job” a self-explanatory policy category. They suggest that what remains after the job depends on who received it, what the job made credible to future employers, where the work took place, and what route opened when the job ended.
The job ends; the labor-market trace may not
The U.S. Census operation was enormous: between June 2008 and May 2010, the Census Bureau recruited 3.9 million applicants for more than 850,000 temporary positions. The work included enumeration, office support, recruiting, and supervision. Stockholm Jobs operated on a smaller scale and within a different welfare system, offering six- to twelve-month municipal jobs to unemployed people with weak labor-market attachment. Its variants included regular placements in childcare, schools, nursing homes, and municipal administration, as well as Stockholm hosts, whose participants worked outdoors at a site created for the program.
The contrast matters because a temporary job can affect several things at once. It can make someone more likely to have a job later. It can change how much they earn when they are working, without changing the chance that they work at all. It can provide contacts, references, and familiarity with an occupation. It can also change eligibility for unemployment insurance or reduce reliance on social assistance without producing a regular job. Health and well-being may change as income, routine, purpose, and social connection change. These are related outcomes, but they are not interchangeable.
Consider the Census result. For applicants close to the minimum test score generally required for employment, being hired increased the likelihood of a positive W-2—evidence that an employer reported wages—in 2010 by about 61 percentage points. The effect did not vanish when Census work ended. The estimate remained about 28 percentage points in 2023. This does not mean that every Census worker received a permanent advantage, or that a short public job raised the wage rate of every later job. It means that, for a particular group of applicants near the hiring margin, the probability of working was substantially higher for years afterward.
Stockholm shows why the word “employment” needs a setting attached to it. Participants in Youth employments and Other municipal employments worked at regular workplaces and later had higher employment than comparable weighted controls. Participants in Stockholm hosts, by contrast, worked at a specially created workplace and later had lower employment than their comparison group. The program could therefore be temporary in both cases while leaving opposite employment traces.
The central question is not whether public employment is inherently beneficial. It is what changed in the worker's position after the job. A person may leave with a credible work history, a manager who can provide a reference, an employer who can offer another role, or a clearer route into a sector. Another person may leave with a completed activity but no employer connection, while being moved from one benefit system to another. The duration of the paycheck is only one part of the intervention.
One temporary job, several outcomes
A useful way to read these studies is to separate three ledgers. The first is the labor-market ledger: did the person work, and for how long? The second is the earnings ledger: did total annual income rise because the person worked more, because the wage rate rose, or because benefits changed? The third is the institutional ledger: which agency paid, which benefit became available, and what costs or responsibilities shifted between levels of government?
The Census evidence makes the first two ledgers especially clear. Among marginal applicants, the researchers find no lasting, statistically significant increase in W-2 wages conditional on having a positive W-2. Census work itself often produced a small annual wage record because it was part-time or part-year. After 2010, however, the later wage rates of people who worked were broadly comparable with those of people who were not hired. The long-run difference in total W-2 income came mainly from a higher likelihood of working. Nearly $6,000 more in annual W-2 wages can result from being employed more often even when the wage received during a job is not higher.
That distinction also changes how fiscal claims should be read. The U.S. paper calculates that, under an assumed 10 percent tax rate, a 3 percent discount rate, and a twenty-year horizon, the additional discounted tax revenue would be about $8,600 per marginal hire. The calculation is useful for showing why repeated employment matters, but it is not a complete social return. It does not observe all administrative costs, public output, displacement, supervision, or the value of benefits received by workers.
The Swedish cases make the institutional ledger visible. Sweden's municipalities finance social assistance, while unemployment insurance is paid from central funds. A Stockholm job paid a wage and could help a participant meet work requirements for unemployment insurance, although the highest level of earnings-related UI also depended on fund membership and other rules. Thus a reduction in social-assistance receipt could mean regular employment, a shift into unemployment insurance, or both. For the participant, UI could provide higher income and relief from repeated means testing. For a municipality, the same change could reduce local costs and move some responsibility to the central system.
This is why a fall in social-assistance receipt cannot stand alone as proof of labor-market success. For Other municipal employments, participants gained employment and earnings while receiving six fewer months of social assistance and 7.81 more months of UI over the study's post-program window. For Stockholm hosts, participants also received less social assistance and much more UI, but they had 1.7 fewer months of employment and SEK 39,100 less in earnings than their weighted controls. The benefit path and the employment path moved in different directions.
Health adds another dimension without resolving the mechanism. The Swedish study reports reductions in psychiatric-drug prescriptions and hospitalizations for some regular-workplace participants, with some effects persisting after the job. These records may reflect less stress, higher income, purpose, social integration, or other changes. They are important consequences, but they are not direct measures of well-being and do not show that one mechanism caused every improvement.
The diagram's two sides should not be collapsed. A public job can improve later employment and change benefit eligibility at the same time, but either outcome can occur without the other. It can also carry costs that are invisible in a simple employment statistic.
A bridge or a detour?
The evidence points toward a practical distinction. A temporary public job is a bridge when it changes the worker's credible next move. That may happen through a trusted signal, relevant experience, a useful network, a reference, or a benefit route that gives the person enough stability to search for work. A job is a detour when it occupies time without improving the next opportunity, locks the person away from employers, or mainly changes the payer of support.
The Census study offers a strong example of why access can matter most for people with weak alternatives. Applicants near the score-70 threshold had much lower prior employment and education levels than high-scoring applicants. For them, the job may have been the first public evidence that they could pass screening, complete a trusted assignment, and be accepted into a formal workplace. For applicants with stronger outside options, the same job produced a smaller short-run employment gain—about 22 percentage points at its 2010 peak—and no detected persistent effect after 2010. Most of those workers may have found work without the Census position.
But the Census authors cannot determine whether the lasting effect came from training and experience or from the signal of selection, background clearance, and trusted work. Marginal hires who worked fewer than 40 total hours had later employment trajectories similar to those who worked more, as did some workers terminated for misconduct or unsatisfactory performance. Those comparisons make a signal interpretation plausible, not proven. The Census setting also had unusual ingredients: testing, interviews, background checks, a trusted public function, a weak labor market, and many highly skilled applicants and workers. A different temporary job may not send the same message to employers.
Stockholm supplies a different test of the bridge idea. At regular workplaces, many later-employed participants remained in the same sector or even the same workplace. Among Youth employment participants, 32.8 percent were observed at the same workplace and 49.7 percent in the same sector during the later observation window. These are descriptive patterns, not proof that a particular manager caused a later hire, but they show a concrete route by which temporary work could become a continuing relationship. Stockholm hosts had no comparable ordinary workplace connection, and their employment outcomes were negative even as their UI receipt rose sharply.
The comparison still has limits. The Swedish estimates use rich administrative information and dynamic weighting, but they depend on the assumption that important differences between participants and nonparticipants were observed. Caseworker judgments about motivation, health, or likely success may not be fully recorded. Neither study randomly assigns people to every workplace design, and neither measures the full value of public output or the full cost of supervision and replacement. The two papers are therefore not a universal scorecard for public hiring. They are evidence that the afterlife of a temporary job is designed through selection, work setting, institutional rules, and the exit path.
A credible evaluation must consequently ask more than “Did participants later receive benefits?” It must ask: compared with what would have happened without the job, did they work more, earn more when working, or merely move between benefits? Did the workplace expose them to a real employer and a usable reference? Did the program end with a plausible next step, or with lock-in and no demand for the skills practiced? Who received the gain, who bore the performance cost, and which parts of the calculation remain unobserved?
The decisive comparison is therefore not between temporary and permanent labels. It is between a job that leaves a worker better positioned for the next opportunity and one that ends without creating that position. To tell those apart, we first need to examine how each study constructs the comparison between a participant's observed path and the path they might otherwise have taken.