How nudges can be used to minimise tax evasion.

How nudges can be used to minimise tax evasion.

1. You’ve Got Mail: A Randomized Field Experiment on Tax Evasion Kristin Bott, Alexander Cappelen, Erik Sorensen, Bertil Tungodden , 2020

2. Introduction and Background 01 Prior research and how this has motivated the aim of Bott et al. (2020) design.

3. Prior Research Previous research shows conflicting findings regarding the effectiveness of tactics aimed at reducing tax evasion. Tax morale has been associated with higher level of tax compliance (versus penalty).₁ Controlled lab experiments find this effect of moral suasion, as well as the influence of increased detection probability. 3 Field experiments do not find this.₂ Cross cultural differences? 1: Alm et al., 1992; Andreoni et al., 1998; Bobek and Hatfield, 2003 2: Blumenthal et al., 2001 (Minnesota - income tax); Castro & Scartascini, 2015 (Argentina - property tax); Dwenger et al., 2016 (DE - church tax), Fellner et al., 2013 (AT - TV licensce) 3: Alm et al., 1992; Dal Bó & Dal Bó, 2009

4. Aim of the Study Identify the causal effects of introducing moral suasion and increasing the perceived detection probability on tax evasion, both for the follow-up year and long-term. “By directly testing alternative policy tools and providing estimates of the causal effects of these tools, our study should provide policymakers with more effective means to combat tax evasion ." Dependent Variables (“Tax Evasion”) Intensive margin : How much foreign income is self-reported by taxpayers who have self-reported some foreign income in the absence of treatment (2011) Extensive margin: How many taxpayers self-report any foreign income. Independent Variables Moral suasion (letters): Tactics used to motivate moral behavior, eg. intrinsic motivation (non pecuniary) to pay taxes Detection probability (letters): Tactics to increase the perceived probability of getting caught evading taxes Baseline year: 2011 Follow-up year: 2012 One year later: 2013

5. Design and Methodology 02 The sample characteristics and experimental design setup of Bott et al. (2020).

6. Sample A randomized field experiment with the Norwegian Tax administration (NTA) was conducted to evaluate how different messages affect self-reported foreign tax compliance. Inclusion criteria: citizens who were flagged by NTA due to discrepancies in self reported data and ACA reports N= 15,708 (60% male) Masking: “Practical Policy Experiment” ACA report : reports from (some) other tax administrations about Norwegian tax residents income and wealth in respective countries (2011) Baseline Sample vs. General: Nonrepresentative sample? Measurement error? What is a discrepancy?

7. Procedure Pre-populated tax return from the NTA to fill out for 2012. Treatment groups received different messages (from NTA) dependent on their condition. Deadline for the pre-populated tax return for 2012. Pre-populated tax return from the NTA to fill out for 2013. 2013 wk. 14 2013 wk. 15 2013 wk. 18 2014 wk. 14 2014 wk. 18 Deadline for the pre-populated tax return for 2013. Baseline year: 2011 Follow-up year: 2012 One year later: 2013 Intention-to-treat design? Long-term Attrition? Short-term

8. Treatment Group Characteristics 03 Participants were randomly assigned to one of five conditions: the no letter group, the base letter, the fairness letter, the societal benefits letter, and the detection letter condition.

9. Treatment Conditions Group Contents Theories No letter condition (control) No supplementary letter was received before the filing deadline. – Base letter condition (active vs. passive voice treatment → N.S.) Explained reason for the letter (due to increase in int. taxpayers) How foreign income is taxed How to report foreign income Reminder of liability for foreign income Instructions with weblink to NTA and call center number (blind/scripted)₁ Knowledge effect : Examines if providing tax filing knowledge affects behavior (vs. no letter). Moral arguments condition → visual with base letter vs. visual w/out base letter Received the identical contents of the base letter, plus one of two statements: Fairness treatment: reminded taxpayers that most Norwegians report correctly Social benefits treatment: highlighted taxation benefits eg. publicly financed services in health, edu., etc. Fa irness motive : People sacrifice gains to align with fairness and conform to social norm s (focusing effect & information effect) Reciprocity motive: Encourages tax payment by highlighting reciprocity (motive) for state-provided benefits Detection probability condition Received the identical contents of the base letter, replacing the underlined point: Statement regarding NTA’s knowledge of foreign income, emphasizing detection risks. Updates the subjective beliefs of being audited by tax authorities. 1: Detection treatment called the call center more than base (no diff between base and moral).

10. The Base Letter Explained reason for the letter (due to increase in int. taxpayers) How foreign income is taxed How to report foreign income Reminder of liability for foreign income Instructions with weblink to NTA and call center number (blind/scripted)₁ Active Voice (eg. “You must report your foreign income..” Passive Voice (eg. “Foreign income must be reported..” Knowledge effect : Examines if providing tax filing knowledge affects behavior (vs. no letter). ∨ Contents Subconditions (2)

11. Explained reason for the letter (due to increase in int. taxpayers) How foreign income is taxed How to report foreign income Reminder of liability for foreign income Instructions with weblink to NTA and call center number (blind/scripted)₁ Reminded taxpayers that most Norwegians report correctly The Fairness Letter Fa irness motive : People sacrifice gains to align with fairness and conform to social norm s → focusing effect & information effect Contents

12. The Societal Benefits Letter Reciprocity motive: Encourages tax payment by highlighting reciprocity (motive) for state-provided benefits Societal Benefits Letter Alone Societal Benefits & Visual Base Letter & Visual ∨ ∨ Subconditions (3) Contents

13. The Detection Letter Updates the subjective beliefs of being audited by tax authorities. Explained reason for the letter (due to increase in int. taxpayers) How foreign income is taxed How to report foreign income Reminder of liability for foreign income Instructions with weblink to NTA and call center number (blind/scripted → detection)₁ Statement regarding NTA’s knowledge of foreign income, emphasizing detection risks. Contents

14. Intensive and Extensive Margin Short-term 04 How did the letters influence the intensive and extensive margins of taxpayers relative to ACA report discrepencies?

15. Total Margin Treatment groups (base, moral arguments, detection probability) report twice as much foreign income as the no letter group. No letter group < base group (N.S) Active vs. passive subconditions → N.S. 8,155 17,030 Result 1. A letter from the tax authorities has a large and statistically significant effect on the self-reported foreign income. Knowledge effect : Examines if providing tax filing knowledge affects behavior (vs. no letter).

16. 8,155 11,342.9 + 10,351 Total Margin Moral and detection groups (pooled) report twice as much foreign income as the base group. Moral suasion increases self-reported foreign income by 70%. Fairness: base + 10,372 NOK Societal benefits: base + 6,345 NOK 3 visual subconditions → N.S. The detection treatment increases self-reported foreign income by 80%. Detection: base + 10,351 NOK Result 2. Including moral suasion or information that increases the detection probability in the letter from the tax authorities has an economically and statistically highly significant effect on the amount of self-reported foreign income. (Pooled) + 7,671 18,022

17. Intensive Margin Moral treatments increase conditional average foreign income by 50% compared to the Base treatment. Moral treatments affect the intensive margin, increasing income by 37,900 NOK (standard error 14,300 NOK). Result 2. Including moral suasion or information that increases the detection probability in the letter from the tax authorities has an economically and statistically highly significant effect on the amount of self-reported foreign income.

18. Extensive Margin Control group versus base (treatments combined): significant impact increasing the share of people that report foreign income 11% to 22% (robust for all subgroups.) Moral letters: Had a (small) and statistically insignificant effect on the extensive margin. Detection letter: Had the most significant impact , increasing the share of people that report foreign income from 20% (base) to 33%. Result 3. The detection letter causes a large and statistically significant increase in the share of individuals who self-report foreign income, whereas the moral letters on average have no effect on the extensive margin. 11% 22%

19. Foreign Income Reporting Demographics Overall. Inclusion of background variables does not alter the estimated treatment effects, but these variables do predict self-reported amounts in the follow-up year.. Measurement error? Report higher levels & more likely to report foreign income Report higher levels & more likely to report foreign income Report higher levels More likely to report higher levels of income (2011 vs 2012) & income (ACA) ‖ self report @ ext.

20. Heterogeneity Analysis Short-term 05 An analysis of how the findings interact with demographic characteristics of the participants.

21. Reported significantly higher amounts of income (p = 0.016, int.). Moral and High Socioeconomic Status Reporting higher income compared to non-citizens (p = 0.010, ext.). Detection and Norwegian Citizenship Reporting significantly higher income (p>0.01, ext.) Detection and Age Above 60 Taxes Increased the amount of foreign income reported by individuals who reported some at baseline (p = 0.026), but not for those who did not (p < 0.10). Increased the likelihood of reporting foreign income for both those who didn’t report any at baseline and those who did (p < 0.001.). Increased foreign income reported for baseline non-reporters (p = 0.003) and baseline reporters (p < 0.001). Moral and Baseling Reporting Detection and Baseline Reporting Detection and Baseline Reporting Interactions with Demographics Result 4. The heterogeneity analysis provides evidence consistent with the moral letters strongly affecting the intensive margin and the detection letter strongly affecting the extensive margin of taxpayer behavior. Overall. Moral letters and the detection letter generally increase the level of self-reported foreign income. The detection letter has a large and highly statistically significant effect on the extensive margin in all subgroups, whereas the moral doesn’t.

22. Intensive and Extensive Margin Long-term 06 Following up with participants one year (2014) after the intervention to assess long-term effects of the letters on their reporting behavior.

23. Intensive & Extensive Margins 0.326 Marginally significant effect of base letter on extensive margin in the long run (p = 0.046) No significant impact of the moral arguments treatment long-term. Large and statistically significant effect of the detection letter on the extensive margin. Share of taxpayers reporting a positive foreign income increases from 25.4% in the base treatment to 32.6% in the detection treatment (p < 0.001) Result 5. The detection letter has a significant long-term effect on the extensive margin, whereas we do not find any significant long-term effects of the moral letters. 0.254 N.S.

24. Mechanism of Action Stop behavior: absence of detection system motivating foreign investments and information on detection possibility making types of investments less attractive in long run Change behavior: change income-generating activity Legal loophole: seek legal ways to reduce taxable income N.S

25. Discussion and Conclusion 07 Possible discussion on why we see the differential effects on intensive and extensive margins.

26. Crowding out of moral motivation? Eg. penalty fears > moral motivation.₁ Future Research - Mechanisms of Action “Crowding Out” Moral appeals work when taxpayers find tax evasion morally wrong, especially when detection is difficult (shown due to baseline reporting interaction). Detection is effective in all environments. Why? Cultural Aspects Did moral appeals work by changing the perception of the tax administration and then affecting willingness to pay? ₂ Changing Perception Did moral appeals work through changing the mood which affects prosocial behavior? . 3 Changing Mood Overall . The moral appeals mainly worked by making the moral arguments salient when the taxpayer made the decision of how much to report but did not work at a fundamental level by increasing the weight taxpayers assign to the moral cost of misreporting.. 1: Gneezy and Rustichini, 2000 2: Feld and Frey, 2007 3: Drouvelis and Grosskopf, 2016

27. Conclusions Effectiveness of Intervention Strong effect on the (intensive) and extensive margin of tax reporting. Detection letters increased perceived detection risk, leading to substantial changes in both immediate and long-term compliance. Influence of Detection Significant effect on the intensive margin of tax reporting for baseline reporters. Moral letters made moral arguments salient, impacting immediate reporting behavior. Influence of Moral Suasion Detection letters have a lasting impact on taxpayer behavior, increasing the share of individuals reporting foreign income. Moral letters do not have statistically significant long-term effects on reporting behavior. Long-term effects 1 2 3 4 ~122M NOK in the follow-up year.

28. Implications Use a mix of moral appeals and detection strategies . Moral appeals effective when taxpayers find tax evasion morally wrong and detection strategies effective in broader settings, regardless of moral motivation. Demonstrated that moral motivation influences behavior in real-world settings with significant financial stakes. Insights applicable to any organization where compliance is crucial, suggesting moral appeals should be part of the compliance toolkit.