The Tax Rate People Act On
The subject of this series is a randomized information experiment in the Netherlands that asks whether people’s mistaken beliefs about their effective marginal tax rate change how much they work. Robert Dur, Job Harms, and Arjan Non do not study a tax cut or a change in the law. They study what happens when people learn something about the share of an additional €100 of gross income that they can actually keep. That distinction supports a practical judgment: tax policy influences behavior not only through the rules governments write, but also through the marginal returns that people believe those rules create.
The experiment matters because the intuitive question facing someone who considers taking an extra shift is not “Which statutory bracket am I in?” It is closer to “If I earn another €100, how much will reach my household after taxes and withdrawn benefits?” In a complicated system, the answer can be difficult to calculate, and a person can make a reasonable-looking decision using an inaccurate answer. This series follows the evidence for that mechanism—from the structure of the tax-and-transfer system, through mistaken perceptions, to work hours and the limits imposed by real-life constraints—while keeping separate what the experiment directly measured from what it only suggests about policy.
The next €100 is the relevant unit
A headline tax rate describes one part of the system. It may tell us the statutory percentage applied to taxable income in a band, but it does not necessarily tell us what happens to disposable income when gross earnings rise. Tax credits may increase or decrease. Housing, health-care, childcare, or other allowances may be withdrawn as income rises. A benefit that is stable over one range can begin phasing out in another. The marginal return to work is the combined result at that particular point on the schedule, for that particular household.
That is why the relevant concept is an effective marginal tax rate. If a worker’s gross income rises by €100 and disposable income rises by €60, the effective marginal tax rate is, conceptually, 40%: €40 of the increment has been absorbed by taxes or reduced transfers. The rate is about the change caused by the next unit of income, not the share of the person’s entire income that goes to the government. Two households with the same gross wage can face different marginal returns because their household circumstances place them on different credit and allowance schedules.
The Dutch study uses this incremental idea because it connects directly to a work-hours decision. A worker who is deciding whether an extra hour is worthwhile cares about the additional disposable income generated by that hour, alongside fatigue, care responsibilities, preferences, and workplace conditions. If the worker believes that most of the extra gross pay will disappear, the opportunity may look less attractive than it really is. If the worker believes the marginal return is generous when it is not, the same decision may look more attractive. Nothing in the official schedule has to change for the decision to change when the belief changes.
This is a more precise claim than saying that “high taxes reduce work.” A high burden may matter, but the experiment is aimed at the link between an individual’s marginal return and their understanding of it. It also does not assume that every person can freely choose hours. A corrected belief can influence a feasible choice; it cannot create childcare, improve health, alter a fixed employer schedule, or remove a study obligation. Those conditions are part of the question, not footnotes to it.
What the Dutch experiment is testing
Dur, Harms, and Non examine this relationship among the Dutch working-age population. Their two-wave survey was conducted through the LISS panel in August and October 2024. In the first wave, respondents reported what they believed would happen to their net disposable income if their gross monthly income increased by €100. They also reported their work hours and whether health, childcare, caregiving, or study limited their ability to work more. A randomly selected group then received information about the net gain associated with a €100 increase for broad income bands. Two months later, the researchers asked again about beliefs and hours.
This structure lets the study ask more than whether people with certain incomes tend to work more or less. Income, education, family circumstances, and hours are related in many ways, so an ordinary comparison cannot tell us whether a mistaken tax belief caused a work decision. Random assignment creates a treatment and a comparison group that can be examined after the information was offered. The researchers can then ask whether people who began with similar kinds of beliefs changed differently when one group received the information.
The baseline evidence gives the test a particular shape. People with lower incomes substantially overestimated their effective marginal tax rates on average, while people with higher incomes tended to underestimate them slightly. The result is not that all low-income respondents were wrong in the same direction, nor that high-income respondents possessed accurate tax knowledge. Errors existed within each income group. But the pattern matters because overestimating the rate makes an additional hour appear to produce too little net income, while underestimating it makes the return appear larger than it is.
The study reports that the information intervention made beliefs more accurate two months later. It also reports an increase of about one and a half work hours per week among people who had initially overestimated their marginal tax rate. That headline is important, but it is not the whole conclusion. The intervention did not erase most misperceptions, the response was much weaker for people who faced constraints on increasing hours, and the evidence does not show that every kind of work decision changed. The useful question is therefore not simply whether information “works.” It is under what conditions a correction reaches behavior, for whom, and with what distributional consequence.
Three distinctions will remain important throughout. First, an actual effective marginal rate is a feature of the tax-and-transfer schedule, whereas a perceived rate is a belief that may be wrong. Second, a change in perceived net pay is not the same as a change in statutory tax rates or actual wages. Third, a change in work hours is not automatically a measured change in long-run household income. Keeping these distinctions intact is what makes the evidence more informative than a general argument about incentives.
Why this is bigger than one tax code
The Dutch case sits inside a wider debate about labor taxation. The OECD’s 2026 release reports that, for a single worker earning the average wage, the average tax wedge across OECD countries was 35.1% of labor costs in 2025, and that the wedge increased in 24 of 38 countries. Those figures help explain why the relationship between labor taxes and work remains politically salient. But a tax wedge is not the same measurement as the Dutch experiment’s effective marginal tax rate. The OECD measure combines taxes paid by employees and employers, subtracts cash benefits, and expresses the result as a share of total labor costs for a specified household type and wage point. The experiment asks what an individual gains from the next increment of gross income.
The difference is not a technical nuisance. An average burden can rise while the marginal return in a particular income range falls, or the reverse. A household with children can face a different net return from a single worker at the same wage. A country can have a moderate average wedge and still contain sharp marginal withdrawal rates for some people. Cross-country averages describe the landscape; they do not tell us what a particular worker thinks an additional hour will pay.
The experiment also raises a distributional question. Because overestimation is concentrated more heavily among lower-income respondents, correcting beliefs may increase work hours at the bottom more than at the top. If those additional hours do not cause a larger percentage fall in hourly wages, the researchers’ evidence supports the possibility that income differences narrow. That is a meaningful implication, but it is not a direct estimate of poverty, household income, or a long-run inequality index. It depends on how hours translate into earnings and on whether other parts of the tax-and-benefit system respond.
The practical judgment this series supports is consequently conditional. When weak labor supply reflects a misunderstanding of the marginal return, clearer information may recover part of the incentive that the schedule was intended to provide. When the return is genuinely low, information can make the trade-off clearer without making it better. When a person cannot add hours because of health, care, study, or workplace conditions, information alone may have little effect. Those are different problems, and treating them as one problem called “low labor supply” risks choosing the wrong remedy.
The first question, then, is not whether people are rational enough to respond to taxes. It is whether the relevant tax incentive is visible in the form that a person must use to make a decision. That requires looking closely at the hidden slope created by taxes, credits, and withdrawn benefits. Once that slope is visible, the experiment’s pattern of mistaken beliefs—and its unequal consequences—becomes much easier to understand.