When Better Information Is Not Enough

The Dutch experiment shows that a corrected tax belief can reach the workweek, but only when the person has a workable route from information to action. This chapter examines that boundary. It distinguishes a mistaken belief about the marginal return from a constrained choice, interprets the study’s comparison of constrained and unconstrained respondents, and uses the Dutch labor market and job-preference results to identify what the intervention did not solve. The practical lesson is diagnostic: a weak hours response can reflect misunderstanding, infeasibility, or a choice that is not governed by the measured marginal return.

A constraint is not a tax error

The study measured constraints before the information was given. Respondents were classified as constrained if they disagreed that their health was good enough to work more, reported that the care of their children or grandchildren could not be managed if they worked more, reported a problem managing care for loved ones, or agreed that their studies could not be combined with more work. People who were neutral or agreed with the relevant statements were classified as having no reported constraint. In the final sample, 38% were classified as constrained and 62% as unconstrained; the corresponding group sizes were 709 and 1,169.

This measure does not ask whether a person would like more hours in an abstract sense. It asks whether a particular barrier limits the possibility of increasing them. That distinction matters for the mechanism. An overestimator who believes the marginal tax rate is too high may revise that belief after receiving information. If the person can change their schedule, the higher perceived net return can enter a work decision. If the person’s health, care responsibilities, or study schedule makes more hours unavailable, the corrected calculation may have no immediate place to operate.

The subgroup results fit that feasibility story. Among overestimators who reported no constraint, work hours increased by slightly more than two hours per week, with p = 0.03. Among constrained overestimators, the estimated increase was 0.45 hours and was not statistically significant, with p = 0.58. The difference is not just a matter of the final outcome. Unconstrained overestimators reduced their misperception by about nine percentage points, while constrained overestimators reduced it by less than two points, an estimate statistically indistinguishable from zero.

That pattern supports a connection between relevance and response. Information about a work incentive is more likely to matter when the recipient can plausibly use it. A person who is deciding whether to add an hour may pay attention to the net return; a person who has already ruled out additional hours because of a health or care barrier may have less reason to retain a detailed marginal-rate message. The information can be true and understandable yet still be behaviorally inactive for that person.

The paper discusses rational inattention as one possible interpretation: people may devote less attention to information that is not relevant to an action they can take, or may forget it more easily. This is a plausible mechanism, not a demonstrated explanation. The constrained-versus-unconstrained analysis was not preregistered, and the constraints were self-reported.