How to improve tax compliance: Evidence from population-wide experiments in Belgium

Jan-Emmanuel De Neve, Clément Imbert, Johannes Spinnewijn, Teodora Tsankova, Maarten Luts 20 May 2021



Truthful reporting of taxable income and timely payment of tax dues are cornerstones of the healthy functioning of societies. While growth in third-party reporting has limited the ability to misreport income (Kleven et al. 2011, 2016, Jensen 2019), tax administrations continue to devote considerable resources to the collection of taxes. The annual cost of non-compliance with individual income taxes is about $319 billion in the US (IRS 2016) and £35 billion in the UK (HMRC 2019). Improving our understanding of the drivers of tax compliance and the cost effectiveness of further interventions is critical for our ability to reduce this ‘tax gap’. 

In 2014 we started a close collaboration with the Belgian tax authority (FPS Finance) to better understand the drivers of tax compliance. This led to a series of experiments that varied the communication of the tax authority with all income tax-filers and payers in Belgium. This collaboration has become a pioneering example of evidence-based policy making in Europe. Having adopted the letters that the experiments showed to be the most effective in increasing tax compliance, the Belgian tax authority set up an internal unit and started an international collaboration to continue experimenting and evaluating innovative tax policies. Experimentations ‘at scale’ have several advantages (Muralidharan and Niehaus 2017, DellaVigna and Linos 2021). They overcome concerns about the representativeness of the experimental sample or the scalability of the policy while providing the statistical power to precisely estimate the returns to different interventions and explore the heterogeneity of responses to treatment in the population. Our close collaboration with the tax administration also guaranteed access to the data needed to estimate the long-term effects of the nudge-interventions, and allowed us to compare their effectiveness with standard enforcement actions currently in use (e.g. bailiffs). 

In De Neve et al. (2021), we report on four experiments we ran during fiscal years 2014 to 2016. To provide a comprehensive evaluation of factors determining tax compliance, we changed the communication of the tax administration at four stages of the tax process. First, in the Tax Payment experiment, we varied the design of the tax bill sent to all taxpayers. Second, in the Payment Reminder experiment, we changed the design of the reminder letter sent to taxpayers who were late. Third, in the Tax Filing experiment, a pop-up pie chart highlighted public services paid for by taxes when individuals filed their taxes online. Finally, in the Tax Filing experiment, we varied the reminder letter sent to individuals who filed their taxes late. We distinguish between three types of treatments. The ‘simplification’ treatments shortened letters and highlighted action-relevant information (e.g. the amount or the deadline). We also tested a total of 16 different additional messages, broadly grouped into ‘deterrence’ messages (e.g. highlighting penalties) and ‘tax morale’ messages (e.g. highlighting public value of tax revenues, appealing to social norms). These three types of treatments cover the three key drivers of tax compliance: deterrence, tax morale, and information frictions (Besley et al. 2019, Benzarti 2017, Slemrod 2018).

Simplifying communication has a consistently positive effect on tax compliance

Figure 1 Main experimental results

Note: The figure presents treatment effect estimates from baseline specifications including individual control variables for four experiments. The outcomes are partial payment probability at deadline (a) and at day of enforcement (b), reported taxable income (c), and filing probability at day of enforcement (d). 

Figure 1 presents our baseline estimates for the simplification, deterrence, and tax morale treatments. The figure conveys a clear and strong pattern across all four experiments. In the three trials in which communication with the taxpayer was simplified (a, b, d), it had a positive effect on tax compliance. The magnitude of the effect is as much as 23% in the Payment Reminders experiment. The effect is larger than in other experiments, which suggests a stronger impact in contexts where communication used to be more complex. In the same three experiments, the deterrence messages had an additional positive, significant effect of around one to three percentage points. In the three experiments in which the tax administration tried to increase tax morale (a, b, c), it had either no effect or even reduced compliance. 

Standard enforcement measures make the control group catch up but are less cost-effective

Using the payment history, we can estimate treatment effects at any time after treatment. Figure 2a shows that for the late payers, the simplified reminders had a strong and immediate effect on payment probability, which peaked around the time when enforcement actions started at 12 percentage points. As regular enforcement actions began, the control group caught up with the treatment group, but not completely, so that six months later there was still about one percentage point higher compliance among the treatment than the control group. 

To shed more light on the relative role of enforcement and simplification, we exploit an enforcement threshold, a tax liability level at which the probability of prosecution actions increases discontinuously. As Figure 2b shows, to the left of the threshold where enforcement intensity is low, there is still a large difference in tax compliance at 180 days between simplification treatment group and the control; to the right of the threshold, where enforcement is high, there is almost no difference.

Figure 2 (a) Dynamic of the treatment effects and (b) enforcement discontinuity in the late payment experiment.


Note: The figure (a) presents simplification treatment effect estimates by days since letter receipt for the Payment Reminder FY2014 experiment. The outcome is (partial) payment probability. The figure (b) presents probability of payment at 180 days to the left and the right of the enforcement discontinuity for the control group or the simplification treatment group.

The combination of our experimental design and a regression discontinuity at the enforcement threshold allows us to estimate the effect of enforcement and simplification for the same taxpayers and compare their cost-effectiveness. We estimate that at the enforcement threshold, the cost of raising €1 of tax revenue through enforcement actions is €0.31, against €0.05 for simplified communication. Regardless of the method used, simplification turns out to be highly cost-effective.

Nudge-type interventions have lasting but heterogenous effects

We studied the dispersion of the treatment effects across taxpayers using machine learning techniques (Wager and Athey 2018, Chernozhukov et al. 2018). The effect of simplification treatment is positive for all observable subgroups. The deterrence treatment, however, has negative effects on some taxpayers. Using causal forests estimates, we can determine that simplification is most effective among taxpayers with children, while deterrence is the most effective among younger taxpayers and those with lower outstanding tax liability. Interestingly, invoking tax morale turned out to be counterproductive for most taxpayers.

We also studied the long-term effects of simplification and whether repeated interventions remain effective (Allcott and Rogers 2014, Cronqvist et al. 2018). We find a positive and significant effect of simplification on tax compliance in the next fiscal year. Repeating the experiment in 2015 for taxpayers who were late twice shows that simplification had a positive effect on payment in 2015 and was no less effective for taxpayers who had already received a simplified letter in 2014. We interpret our results as providing suggestive evidence against any fatigue effect, at least for recidivist taxpayers.

Concluding remarks

Our analysis shows that simplifying communication by the tax administration consistently improves tax compliance, nudges taxpayers to pay taxes on time, and makes both late filers and payers comply more swiftly. Making compliance as easy as possible therefore deserves even greater attention, as communication is an inherent part of any tax administration.

The results of this close collaboration between academics and government speak for themselves and are a testimony to the importance of evidence-based policy making. Addressing the Federal Parliament, the Belgian Minister of Finance confirmed that €4 million additional revenue was received in the 2015 tax year, together with €30 million revenue collected more quickly compared with previous years (when simplified letters were not used), and €1 million saved in enforcement costs (Senate of Belgium 2019). Over the ensuing years, the cumulative impact of these nudges is estimated to have accelerated the payment of over €100 million in late taxes, leading to structural gains in tax revenue and administrative cost savings. Not only do public finances benefit from these interventions, but so do the individual taxpayers who would have otherwise incurred steep penalty fines and interest rates.


Benartzi, S, J Beshears, K L Milkman, C R Sunstein, R H Thaler, M Shankar, W Tucker-Ray, W J Congdon and S Galing (2017), “Should governments invest more in nudging?”, Psychological Science 28(8): 1041–1055.

Benzarti, Y (2017), “How taxing is tax filing? Using revealed preferences to estimate compliance costs”, NBER Working Paper 23903.

Besley, T, A Jensen and T Persson (2019), “Norms, enforcement, and tax evasion”, NBER Working Paper 25575.

DellaVigna, S and E Linoz (2021), “RCTs to Scale: Comprehensive Evidence from Two Nudge Units”, Working Paper.

De Neve, J-E, C Imbert, J Spinnewijn, T Tsankova and M Luts (2021), “How to Improve Tax Compliance? Evidence from Population-wide Experiments in Belgium”, The Journal of Political Economy 129(5): 1425–1463.

HM Revenue & Customs (2018), “Measuring tax gaps 2019 edition”, Technical report, HM Revenue & Customs.

Internal Revenue Service (2016), “Tax gap estimates for tax years 2008–2010”, Technical report, Internal Revenue Service.

Jensen, A (2019), “Employment structure and the rise of the modern tax system”, NBER Working Paper 25502.

Kleven, H J and W Kopczuk (2011), “Transfer program complexity and the take-up of social benefits”, American Economic Journal: Economic Policy 3(1): 54–90.

Kleven, H J, C T Kreiner and E Saez (2016), “Why can modern governments tax so much? an agency model of firms as fiscal intermediaries”, Economica 83(330): 219–246.

Muralidharan, K and P Niehaus (2017), “Experimentation at Scale”, Journal of Economic Perspectives 31(4): 103–124.

Senate of Belgium (2019), Question écrite n° 6-2279.

Slemrod, J (2018), “Tax compliance and enforcement”, NBER Working Paper 24799..



Topics:  Taxation

Tags:  tax compliance, Belgium, nudges

Associate Professor of Economics, Saïd Business School, University of Oxford

Associate Professor, University of Warwick

Associate Professor in Economics, London School of Economics; Research Fellow, CEPR & IFS

Assistant Professor, Tilburg University

Advisor – Tax Compliance & Behavioural Insights, Federal Public Service Finance – Belgium


CEPR Policy Research