When Job-Search Messages Move the Wrong Behavior
This series examines digital interventions delivered by Public Employment Services to unemployed job seekers: what the messages ask people to do, what they actually change, and whether that change leads to work. The central focus is the gap between visible engagement and effective job search. A person may open an email, submit a required activity report, or even click a link without applying in a more productive market or finding a job. Understanding that gap gives the reader a better way to judge both the promise and the limits of low-cost employment policy.
The question matters because unemployment agencies can reach large populations through ordinary digital channels. A short message costs little to send, arrives quickly, and can be timed to a decision or a deadline. That makes it tempting to treat communication as a light-touch substitute for more expensive counseling or service redesign. But the low cost of delivery does not make the behavioral problem simple. Job search can decline as unemployment lasts longer, partly because repeated rejection makes sustained effort difficult. It can also be misdirected: a job seeker may keep searching in an occupation with few vacancies while overlooking realistic alternatives.
Those are different problems. Someone who lacks persistence may need a different intervention from someone who is willing to move but does not know where the opportunities are. A message that addresses the wrong problem can produce an observable response while leaving the underlying bottleneck untouched. Worse, the wording may carry an unintended signal when it comes from an institution that also monitors eligibility and can trigger sanctions.
The same channel, different questions
Two large randomized field experiments make this distinction concrete. The first, conducted with the Swedish Public Employment Service, tested six-month sequences of short motivational emails. One version used a supportive tone intended to strengthen autonomous motivation: it emphasized choice, personal goals, confidence, and understanding. The other used a controlling tone that stressed obligations, deadlines, monitoring, and the risk of sanctions. Both were attempts to increase the effort needed to keep searching for work.
The second experiment, conducted with the Dutch Public Employment Service, began from a different diagnosis. It targeted job seekers whose primary occupations had poor labor-market prospects. Instead of mainly asking recipients to try harder, the intervention showed them how their current occupation compared with attainable alternatives. The suggestions combined evidence from previous occupational transitions with current indicators of demand, along with information about relevant skills and the occupations themselves. Recipients were shown options; they were not forced to abandon their original search.
These experiments should not be read as a direct contest between Sweden and the Netherlands, or between “motivation” and “information.” Their countries, institutions, populations, timing, and outcome measures differ. The value of putting them together is more precise. The Swedish experiment shows what happens when a message changes an administrative behavior but does not improve the search that matters. The Dutch experiment shows what can happen when a message changes the information available at the point where a job seeker chooses where to look.
A message has to travel through a chain
A useful way to think about employment communication is as a sequence rather than a single treatment effect. First comes exposure: did the person receive and notice the message? Then comes an immediate response, such as opening a link or submitting a required report. Next comes a search decision: which occupations, vacancies, or employers does the person consider, and with what intensity? Only after those steps can a labor-market outcome appear—employment, hours worked, earnings, job duration, or the quality of the match.
The stages are related, but none can stand in for all the others. An open rate tells us that the message had an opportunity to matter, not that the recipient understood it. A report submission tells an agency that a formal requirement was met, not that the person searched more effectively. More applications may indicate greater effort, but applications sent in a severely congested market need not produce more jobs. A rise in employment is more consequential, but it still needs interpretation: did more people obtain work, did already-employed people work more, or did people accept less stable jobs?
The two studies were designed with this problem in mind. The Swedish researchers could observe monthly reporting across a population of about 200,000 job seekers, as well as applications and job finding. The Dutch researchers combined administrative records on hours, earnings, benefits, and employment with surveys about the occupations in which new jobs were found. These additional measures make it possible to ask not only whether a message “worked,” but how far its effect traveled and where it stopped.
Random assignment strengthens that inquiry by making treatment and control groups comparable, on average, before the intervention. It allows a difference in later outcomes to be interpreted as the causal effect of being assigned to a message package within that study’s population and setting. It does not, by itself, reveal the psychological mechanism. Nor does it turn a result from one institutional environment into a universal law. Mechanisms need to be inferred from the pattern of outcomes, the design of the intervention, and the context in which recipients encountered it.
The diagram’s most important feature is the gap between reporting and effective search. Reporting is not irrelevant; it may be part of a larger process. But it is an intermediate behavior, and it can move independently of applications, occupational choice, or employment. A serious evaluation therefore follows the path rather than stopping at the first visible response.
What the two interventions put in front of the job seeker
The Swedish intervention tried to alter the reason for searching. Its supportive messages treated job search as something a person could endorse because it served a meaningful goal. Its controlling messages treated search as an obligation reinforced by external and internal pressure. The distinction came from Self-Determination Theory, which separates autonomous motivation from controlled motivation. The theory offered a coherent design language, but the intervention still operated inside an existing relationship: the Swedish agency required monthly activity reports and monitored whether search effort was sufficient.
That relationship matters before a recipient reads a single sentence. A supportive message from a counselor whose role is understood as helping may be received as genuine assistance. A similar message from a highly monitoring institution may be read against the background of rules and enforcement. “You can choose” may sound like support, or it may imply that the agency has relaxed its control. “You must report” may prompt compliance, yet say little about how to conduct a better search. The sender and the surrounding system help determine what the wording means.
The Dutch intervention put a different decision in front of the recipient. It made a market comparison concrete: the occupation currently being targeted had few vacancies relative to job seekers, while several alternatives were both more promising and plausibly reachable from the person’s experience. This approach addressed an information problem without assuming that unwillingness was the main obstacle. The baseline evidence was revealing. Many job seekers said they were willing to consider alternatives and felt that their skills could transfer, yet their beliefs about the prospects of different occupations were poorly aligned with actual differences in tightness.
The contrast is therefore not between a warm email and a cold email. It is between trying to alter motivation and trying to alter the choice set with credible, usable information. A motivational message asks a person to sustain a behavior whose direction may already be uncertain. An occupational recommendation can change the direction of search while leaving the final decision with the job seeker. Both can be sensible in the right circumstances. Neither should be evaluated by the same first-stage measure alone.
The question this evidence makes unavoidable
The Swedish experiment makes the title’s apparent paradox possible. Both motivational treatments increased submission of activity reports, so the messages clearly had a behavioral response. Yet controlling messages did not increase applications or job finding, while supportive messages modestly reduced applications and produced weakly negative job-finding effects in selected periods, especially among men. The result is not that every supportive message harms every unemployed person. It is that visible compliance did not provide evidence of improved search, and a supposedly helpful frame could be counterproductive in the relationship where it was delivered.
The Dutch experiment points to the complementary possibility. When the limiting problem is not simply effort but poor information about where a feasible opportunity lies, a message can help by changing the search problem itself. Personalized alternatives, demand signals, and evidence of real occupational transitions give a recipient something more actionable than encouragement. The later employment and mobility results are examined on their own terms, with their own limits; they do not erase the Swedish findings.
The practical judgment supported by this series is consequently conditional. Before sending a reminder, pressure cue, or supportive appeal, an employment service needs to ask what is actually preventing progress. Is the immediate problem attention, formal follow-through, persistence, knowledge of demand, confidence in transferable skills, or congestion in the chosen market? Which observable behavior should change if the diagnosis is correct? And is that behavior close enough to employment that a rise in the agency’s records will mean a rise in the job seeker’s prospects?
Those questions keep the subject in view: digital communication between a public employment institution and a person trying to return to work. They also impose a discipline on the evidence. A message is not successful because it is opened, because it sounds encouraging, or because it produces a compliant record. It is successful only when the behavior it changes is the behavior that the job seeker needs, in a relationship and a labor market that allow that change to lead somewhere.