Why Supportive Messages Can Backfire
The Swedish experiment raises a harder question than whether one email tone is nicer than another. It asks how a message from a Public Employment Service is interpreted inside an existing relationship, and why the answer differs across job seekers. This chapter diagnoses the modest decline in applications and selected job-finding outcomes after supportive emails, weighing the evidence for a better-match explanation against the authors’ account of relational mismatch and the possibility that pressure crowds out self-directed motivation.
The diagnosis has to remain provisional. Random assignment establishes that the treatment package changed outcomes relative to the study control, but it does not directly reveal what recipients thought the messages meant. The most useful conclusion is therefore not that supportive language is inherently dangerous. It is that the same words can change behavior differently when they arrive from a monitoring institution, and that an average effect can conceal a concentrated response among particular groups.
The negative result needs a diagnosis
The supportive messages were designed to strengthen autonomous motivation. They emphasized choice, personal goals, competence, and understanding, and they acknowledged that repeated rejection can be frustrating. The intervention’s direction was theoretically plausible: searching for work is a prolonged, difficult activity, and a person may sustain it more effectively when they see it as meaningful rather than merely imposed.
Yet the observed pattern was the opposite of the intended one on some outcomes. Supportive assignment reduced formal applications by 0.128 per month in months 2–4 and by 0.094 per month in months 2–7. The longer-window reduction was about 1.1% relative to the control mean of 8.286 applications per month. Job finding within seven months did not differ significantly from control, but job finding within four and twelve months was slightly lower, with the twelve-month effect statistically significant at about 0.4 percentage points. Search-quality and job-quality measures did not show systematic improvement.
A first diagnostic question is whether this was really a harmful search response or a beneficial change in search strategy. Perhaps the supportive messages encouraged people to stop applying indiscriminately, take more time, and wait for a job that fit better. That would make a lower application count compatible with a better employment outcome. In principle, job-search quantity and job-search quality can trade off. A person might submit fewer applications but choose them more carefully, find a more suitable job, and stay longer.
The available results do not support that as the leading explanation. The supportive treatment did not produce systematic improvements in the study’s measures of search quality or job quality. The authors also considered job quality through administrative and survey outcomes rather than relying only on the number of applications. The absence of a quality gain cannot prove that no individual benefited from slowing down. It does, however, weaken the claim that the average reduction in search was an advantageous move toward better matches.