What a Clearer Tax System Can—and Cannot—Change
The subject of this series is the Dutch randomized experiment on whether mistaken effective marginal tax rates change how much people work. Its final judgment is conditional: a clearer marginal return can change work hours, most clearly for people who initially think taxes and withdrawn benefits take too much and who have room to add hours. This conclusion connects the experiment’s individual response to its possible distributional consequence, separates it from broader tax-wedge statistics, and gives a practical way to distinguish misunderstanding from an unattractive return or an infeasible choice. It also marks the boundaries beyond which the evidence should not be stretched.
The strongest answer is conditional
The experiment supports a specific answer to the question in its title. Tax misperceptions can change how much people work, but the effect is not a universal reaction to information or to taxes in the abstract. The path runs through a sequence of conditions: the tax-and-transfer system creates an actual marginal return; a person holds a belief about that return; information can move the belief; the revised perceived net wage can matter to the person; and the person must be able to adjust work hours.
The evidence is strongest at the point where these conditions meet. Overestimators reduced their perceived effective marginal tax rates and increased work hours by about 1.45 hours per week from a baseline of 19.1 hours. The estimated effect was statistically significant. Underestimators moved in the predicted direction on hours, but their reduction was less than half an hour from a baseline of 29.14 hours and was not statistically significant. People whose initial beliefs were about right did not show a material directional change. The pattern is consistent with a belief correction changing the perceived marginal net wage, especially for the group whose starting belief made work look much less rewarding than the calculated schedule suggested.
That wording matters. The researchers did not lower the statutory tax rate, change a credit, increase an hourly wage, or give respondents additional disposable income directly. They supplied coarse information about what most people in broad income ranges generally gain from an extra €100 of gross income. When overestimators revised their perceived marginal tax rate downward, the extra hour appeared to pay more. Their reported hours then rose. The result is evidence that perception is one route through which tax design reaches labor supply; it is not evidence that information is equivalent to a tax cut.
It is equally important not to read the asymmetry as a universal law about different kinds of workers. Overestimators had larger average errors, were strongly represented among lower-income respondents, and responded more clearly in the study. Underestimators were closer to the actual rate on average, and the estimated hours response was small and imprecise. The difference could reflect the size of the initial error, the treatment’s broad ranges, subgroup composition, attention, constraints, or statistical precision. The study supports the observed pattern, not one definitive explanation for every part of it.
The result is also a reminder that accuracy and behavior are separate outcomes. The intervention improved measured accuracy but left much of the original misperception in place. A person can become less wrong without becoming fully informed. A person can understand that an extra €100 yields more net income than they thought and still decide not to work more. The marginal return competes with time, fatigue, care, health, study, employer arrangements, and the value of leisure. The experiment shows that the perceived return can matter; it does not claim that it governs every decision.
Why the distributional implication follows
The possibility of a distributional effect comes from the location of the errors, not from a separate redistribution program. Lower-income respondents were more likely on average to overestimate their marginal tax rates. The hours response was concentrated at the bottom of the individual-income distribution: people earning below €2,000 per month increased hours by 1.8 per week from a baseline of 12.5 hours, a 14.3% increase, while hours did not change in the higher income categories.
Put together, those observations describe a plausible pathway. A lower-income person who had believed that little of an additional hour’s gross pay would survive taxation and benefit withdrawal receives information that makes the return look better. If that person can add hours, the perceived return rises and hours increase. If the additional hours translate into additional earnings without a larger percentage reduction in hourly wages, income at the bottom can rise relative to income higher up. The paper therefore concludes that the intervention can improve individual labor-supply efficiency while reducing income inequality.
The conclusion is meaningful precisely because it is conditional. The paper states that the income implication assumes the increase in hours does not reduce hourly wages, or at least does not reduce them by a larger percentage than hours increase. The study does not directly estimate long-run household income, poverty, consumption, or an inequality index. It observes the hours pattern and combines it with the income distribution of misperceptions.