From Information to Mobility: What the Dutch Results Actually Show

The Dutch intervention was designed to change where unemployed people searched, not merely how motivated they felt while searching. This chapter tests that proposition against the outcomes: employment, hours, earnings, occupational destination, relevance for different job seekers, and the effects that might appear when the advice is scaled. The key judgment is therefore twofold. Did the information produce more work for the targeted population, and is there evidence that it did so through a changed search direction rather than through a generic increase in activity?

The answer is strongest in the main population for which the advice was relevant: people with no paid employment at the intervention date who were searching in occupations with structurally poor prospects. It is more qualified outside that group, and it becomes more model-dependent when the question shifts from individual effects to labor-market spillovers.

The employment effect was extensive, not intensive

The main analysis focused on 10,738 job seekers who were not working before the intervention and were looking in non-Covid occupations whose poor prospects were structural rather than a temporary consequence of pandemic restrictions. That restriction is not a technical footnote. It defines the problem the message was built to solve. Advice to leave a market that remains persistently slack is different from advice to leave a market that is likely to recover soon, and it is different again for someone who already has paid work while receiving unemployment insurance.

Eighteen months after the intervention, the information treatment increased monthly hours worked by about four hours. The control-group mean was 66.1 hours, so the increase was 6.0% relative to that baseline. Monthly labor earnings rose by about €72 from a control mean of €1,118, a relative increase of 6.4%. Employment rose by about 2.9 percentage points from a control mean of 55.9%, which the paper reports as a 5.2% relative increase. These are different descriptions of the same broad result: percentage points describe the absolute change in an employment rate, while a percentage increase describes that change relative to the control-group level.

The cumulative measures show why the time horizon matters. By the end of 2022, the treatment group had worked roughly 55 additional hours and earned about €1,100 more than the control group on a cumulative basis. A short message did not create an immediate transformation in every recipient’s labor-market position. Its effect accumulated as some people moved into work and continued working. The treatment did not significantly change unemployment-benefit receipt, showing that benefit dependence does not move in lockstep with hours and earnings in this experiment.

The pattern is concentrated on the extensive margin. When the researchers examined hours or earnings conditional on already having positive work or earnings, they found no treatment effect. The intervention therefore appears to have helped additional people obtain work rather than making already-employed recipients work substantially more or accept jobs with lower hours or earnings. That distinction matters for job quality.