Nordic Economic Policy Review 2026: Advancing Policy Through Randomised Experiments

Nordic Economic Policy Review 2026

Comments on Simen Markussen and Marie Bjørneby: The Norwegian Tax Experiment

Kaisa Kotakorpi

1 Introduction

Markussen and Bjørneby (2026) present the design of a tax experiment to be carried out in Norway. The experiment involves a randomised natural field experiment, or RCT, in which 8% of 20–35-year-old Norwegian taxpayers are eligible for an earned income tax allowance, while all other taxpayers see no change to their tax schedule. The experiment affects the participants’ marginal tax rates and participation tax rates. I will not review the details of the experiment here, but refer to the account provided in Markussen and Bjørneby (2026). The purpose of this comment on the experiment is twofold. First, to look at the Norwegian tax experiment in the context of the development of research involving randomised field experiments in public economics. Second, to comment on the attractive features of the experiment and raise a few questions about the design that deserve consideration when the findings are discussed and interpreted in due course.

2 Context: field experiments in public economics

Natural field experiments or randomised controlled trials (RCTs) conducted in real-world settings have proliferated as part of the credibility revolution in empirical economics. Public economics and, in particular, research on taxation have lagged behind somewhat. RCTs appear to have been more common, e.g., in development, education, and labour economics. This methodological development has been documented in the context of public economics by Kleven (2014). I will not attempt a comprehensive review of the literature or a treatment of the multitude of issues that have to be considered when conducting this type of research. Rather, I will make a few remarks that put the Norwegian tax experiment into context.

One prominent reason why field experiments have not been commonly used to analyse the effects of taxation is probably that randomising tax rates in the real-world has usually been considered unethical or inequitable from the point of view of horizontal equity, i.e., against the principle of equal treatment. However, this argument should not lead to an outright ban on tax experiments. A comprehensive discussion is beyond the scope of this comment. One point worth considering is that any policy change that does not apply to the entire population will require some lines to be drawn, and this will entail at least quasi-randomness in the treatment of similar individuals, depending on which side of the eligibility line they happen to fall. In addition, while compromising some equity is a cost, experimentation also has benefits, e.g., generating reliable knowledge about the effects of major policy programmes and hence avoiding misguided choices that may prove costly. Therefore a cost-benefit calculus on the use of experimentation may be warranted. (Of course, there may also be principles that should not be compromised at any cost – a type of argument that requires clear justification.)

However, the question of the ethics of randomisation does not mean that no field experiments have been conducted in the fields of public economics or tax research. One particular strand of research in which they have been prominent and useful is the study of tax evasion and enforcement. Conducting a randomised experiment on these topics has probably been viewed as less problematic from an ethical point of view, as randomised (tougher) enforcement should affect only those who have not complied with their legal duties as taxpayers. One benefit of using field experiments to analyse tax evasion is that innovative research designs are particularly valuable in uncovering evasion. It is something that, by definition, the individuals concerned want to hide, so it is not immediately observable in administrative data. However, reactions to randomised enforcement measures can provide information about tax evasion in the baseline. Disadvantages of this approach include the need for careful interpretation of the findings, as real-life enforcement, by its very nature, targets a heavily selected group of taxpayers. As such, lessons about the effects of randomised tax audits, for instance, are unlikely to be directly generalisable to tax administrations’ real-life risk-based audit programmes.

A review of this strand of literature is provided by Pomeranz and Vila-Belda (2019), and a meta-analysis by Antinyan and Asatryan (2024). In the Finnish context, papers that report results from randomised field experiments related to tax enforcement include Kosonen and Ropponen (2015) on providing information about a VAT reform; Harju et al. (2020) on evasion of taxes on imported used cars; and Eerola et al. (2025) on evasion of rental income tax. Insights into cooperation between researchers and tax administrators in Finland and the UK are discussed in Almunia et al. (2019).

The papers mentioned above applied randomised variation to other design features of the tax system, but not tax rates. Experiments explicitly randomising tax rates are rare. A prominent Finnish example that involves randomisation of effective tax rates is the Finnish Basic Income Experiment. The design and effects of the experiment have been reported in a series of papers, Verho et al. (2022), Hämäläinen and Verho (2022) and Hämäläinen et al. (2025). In applying randomised eligibility to basic income for a subset of unemployed Finns, the experiment created randomised variation in participation tax rates, i.e., the fraction of earnings individuals get to when they move from unemployment to work. This provides a unique setting in which to examine the causal effects of work incentives on outcomes such as labour force participation.

Implementing the Finnish basic income experiment was a major achievement from both scientific and societal perspectives, and it has profound implications for what is feasible for future generations of researchers and public administrators. The design of the experiment and the ethical considerations behind this type of experimentation were reviewed and approved by the Constitutional Committee of the Finnish Parliament. This showcases the importance of the experiment and the process, as the Committee assessed whether this type of randomised experiment is consistent with citizens’ constitutional rights. Indeed, the experiment has paved the way to other important endeavours that have followed, such as Finland’s Two-Year Preschool Experiment, also reviewed in this volume (Izadi and Sarvimäki, 2026).

3 Comments: design of the Norwegian tax experiment

The Norwegian tax experiment is important in the context of the background described in Section 2. It applies randomised variation to the earned income tax allowance to provide credible causal evidence on the effects of an important feature of income tax systems in many countries. Discussions during the planning stage of the experiment involved careful consideration of ethical issues and the benefits and drawbacks of experimentation, as discussed above. One feature of the experiment I will comment on below is that it combines variation in effective tax rates with a personalised information treatment, making the allowance’s effect on effective tax rates salient to the participants.

The experiment has multiple attractive features. First, as should be clear by now, this type of research is very rare, and an RCT with randomised variation in effective tax rates is exceptional. Second, a key feature is that the experiment induces variation in both marginal tax rates and participation tax rates, which allows for the study of both extensive and intensive margin effects on labour supply. Third, the experiment is on a very large scale, involving 8% of taxpayers in a wide age bracket. Fourth, the experiment lasts for a relatively long time, i.e., five years, allowing time for the effects to emerge if participants take a while to react, e.g., if new jobs cannot be found immediately when the incentives change. Fifth, there are no other restrictions on the target group beyond the age restriction, providing good generalisability of the results to a broad, well-defined subgroup of the population. Sixth, the researchers intend to measure perceptions of tax rates before and after the treatment. This is very important for interpreting the results of the experiment: any effects will be conditional on successfully changing individuals’ perceptions of tax rates. In other words, the strength or impact of the treatment depends on changes in perceptions; this is the first stage variation that induces (or does not) changes in the final outcomes of interest. Finally, the experiment provides an opportunity to study the effects on multiple other outcomes, not just direct labour market outcomes.

Are there potential caveats or complications that should be kept in mind when discussing and interpreting the results of the experiment in due course? First of all, it is worth considering whether introducing a tax allowance rather than changing nominal tax rates directly is an innocuous choice. Referring to a “tax-cut experiment” perhaps carries with it the connotation of cutting nominal tax rates, whereas in practice the experiment involves introducing an allowance. This, of course, implies a cut in effective tax rates, but given that a tax allowance are highly complex, the salience of cuts implemented to tax rates directly vs. indirectly via an allowance may be different. On the other hand, combining the tax allowance with explicit information about participants’ actual tax rates will make the tax changes implied by the experiment salient. This makes cuts implemented via an allowance look more like a cut to nominal tax rates.

A second and related question concerns the external validity of this type of combined treatment, i.e., a tax cut/allowance combined with highly personalised information. In real life, policy changes are typically not combined with the provision of such extensive and personalised information directly to individual taxpayers. In this sense, non-compliers in the combined treatment who receive both the tax cut and the information but do not access the information are of interest, as they are closest to a real-life situation in which policy changes are implemented without a simultaneous change in information. This, however, is a highlys elected group.

The plans for the experiment do not include a separate treatment with only a tax change and no information. In my view, this type of pure tax treatment would have been of interest. The argument appears to be that having a combined treatment with information and a tax change, as well as a separate treatment with information only, would allow for the separation of the effects of the two. However, as all participants in this constellation do receive information that makes tax rates salient, this does not necessarily allow researchers to pin down the precise effects of tax changes in the absence of such information. Rather, all effects of the experiment are measured conditional on receiving the information treatment; whether they would be the same in the absence of information is not clear a priori, or at least it warrants discussion. One intriguing line of thought in this context is whether the problem lies with the experiment or with the real world – why not provide better information in the real world as well? Could tax rate changes routinely be accompanied by personalised information to increase their salience to taxpayers? Why this is not done in practice may, of course, be due to multiple reasons, such as implementation costs, potential problems with information accuracy related to complex policy changes, and, more fundamentally, it is not even obvious whether full salience is always optimal (see Goldin (2015) for a discussion on optimal tax salience).

Third, the control group in the experiment is not fully “business-as-usual”: the individuals in the control group receive an invitation to log in and find out whether they were selected for the experiment. This may trigger behavioural reactions to the possibility of being eligible for a tax allowance, only to find out that, in fact, you are not receiving it.

Fourth, an important consideration is the size of the changes in incentives implied by the experiment. For example, the changes in participation tax rates in the Finnish Basic Income experiment were very big. The experiment lowered participation tax rates by 23 percentage points for full-time employment, yet no significant change was found in the number of days in employment (Verho et al. 2022). Even though the sample size in the Norwegian experiment is an order of magnitude larger, meaning greater power to detect even small effects, the magnitude of the treatment may not be solely a power issue; reactions to small vs. large incentive changes may be genuinely different, e.g., due to labour market frictions.

All in all, the Norwegian tax experiment is an important endeavour both scientifically and from a societal and public administration perspective. The results and careful discussion of their interpretation and implications will be of interest to both the scientific audience and policymakers.

References

Almunia, M., Harju, J., Kotakorpi, K., Tukiainen, J., & Verho, J. (2019). Expanding access to administrative data: The case of tax authorities in Finland and the UK. International Tax and Public Finance, 26(3), 661–676.

Antinyan, A., & Asatryan, Z. (2024). Nudging for tax compliance: A meta-analysis. The Economic Journal, 135(668), 1033–1068.

Eerola, E., Kosonen, T., Kotakorpi, K., & Lyytikäinen, T. (2025). Tax compliance in the rental housing market: Evidence from a field experiment. American Economic Journal: Economic Policy. Advance online publication.

Goldin, J. (2015). Optimal tax salience. Journal of Public Economics, 131, 115–123.

Harju, J., Kosonen, T., & Slemrod, J. (2020). Missing miles: Evasion responses to car taxes. Journal of Public Economics, 181, 104096.

Hämäläinen, K., Simanainen, M., & Verho, J. (2025). Health effects of cash transfers: Evidence from the Finnish basic income experiment. Journal of Public Economics, 250, 105480.

Hämäläinen, K., & Verho, J. (2022). Design and evaluation of the Finnish basic income experiment. National Tax Journal, 75(3), 573–596.

Izadi, R., & Sarvimäki, M. (2026). Learning to experiment: Practical lessons from Finland’s two-year preschool experiment. Nordic Economic Policy Review.

Kleven, H. J. (2014). How can Scandinavians tax so much? Journal of Economic Perspectives, 28(4), 77–98.

Kosonen, T., & Ropponen, O. (2015). The role of information in tax compliance: Evidence from a natural field experiment. Economics Letters, 129(C), 18–21.

Pomeranz, D., & Vila-Belda, J. (2019). Taking state-capacity research to the field: Insights from collaborations with tax authorities. Annual Review of Economics, 11, 755–781.

Verho, J., Hämäläinen, K., & Kanninen, O. (2022). Removing welfare traps: Employment responses in the Finnish basic income experiment. American Economic Journal: Economic Policy, 14(4), 501–522.