Economic and Social Incentives for Tax Compliance: Evidence from a Field Experiment in Germany

Economic and Social Incentives for Tax Compliance: Evidence from a Field Experiment in Germany Nadja Dwenger (MPI) Henrik Kleven (LSE) Imran Rasul (UCL) Johannes Rincke (Univ. of Erlangen-Nuremberg) July 2014 NBER Summer Institute, Cambridge

A Tax Compliance Puzzle? Gap between theory and evidence: Theory predicts low compliance under low audit probabilities or penalties (Becker-Allingham-Sandmo) Evidence shows high compliance in modern tax systems despite low audit probabilities and penalties Closing the gap: Third-party information creates large detection probabilities conditional on evading Misperception of enforcement parameters Social incentives and psychology

Economic and Social Incentives Economic incentives (pecuniary): Deterrence and tax policy Social incentives (non-pecuniary): Recognition, duty, social norms, guilt, shame etc. Studying social incentives is di cult: measurement, variation We consider a setting in which we can distinguish economically and socially motivated individuals We vary economic and social incentives in a eld experiment

Local Church Tax in Germany: Key features Feature 1: tax and charity Legal obligation to pay for church members Overpayments are encouraged and represent donations Feature 2: true tax base is observable True tax base for church = reported taxable income to government Distinguish evaders, compliers & donors Feature 3: zero deterrence in the baseline Church never cross-checks reports against income tax returns Baseline compliance = social motivation (+ misperception)

Local Church Tax in Germany: Identifying Motivation Baseline behavior (p = 0) reveals motivation: Baseline evaders: economically motivated Baseline compliers: socially motivated Baseline donors: strongly socially motivated Compliance type is very persistent prior to experiment: We de ne types based simply on one pre-treatment year Each type could be subject to misperception (p > 0): Estimate misperception via (randomized) communication of p = 0 Separate misperception from motivation

Outline 1. Conceptual framework 2. Context, design and data 3. Findings (a) Baseline compliance under zero deterrence: Amount of social motivation, duty to comply (b) Randomized eld experiment: Economic incentives, social incentives 4. Conclusion

Warm-Glow Model of Tax Compliance Model that uni es two classic frameworks Becker-Allingham-Sandmo model: economic incentives Andreoni warm-glow model: social incentives Allows for coexistence of three types of individuals: Evaders: economically motivated Compliers: socially motivated [bunching due to both penalty and duty-to-comply preferences] Donors: strongly socially motivated Model generates type-speci c predictions on treatment responses

Local Church Tax in Germany Compulsory by law for all church members above 18 years of age Membership by baptism; possible to quit church 60% of the population are members Overpayments are encouraged and de ned as donations Our focus: Protestant church tax in a large metropolitan area in Bavaria Progressive schedule Concern with external validity: stakes are relativey small

Local Church Tax in Germany: Tax Schedule

Experimental Design Manipulations of the o cial tax noti cation from the Protestant church in a large metro area of Bavaria Individuals randomly assigned to di erent treatment letters: Tax simpli cation Zero deterrence Positive Deterrence (audit probabilities; audit probability notch) Social and monetary compliance rewards Social norms and moral appeal Randomization successful across all letter groups All treatments were truthfully implemented

Data Linked administrative panel data 2008-2012: Church District tax records State personal income tax records About 40,000 individuals per year with information on church tax paid, true taxable income, and personal characteristics Individual records of tax compliance for up to 4 years pre-treatment

Finding 1: Social Motivation is Substantial (or Not) Under zero deterrence, predicted compliance is zero absent social motivation But a signi cant fraction of individuals comply: 20% of individuals make payment true taxes owed 80% of individuals make payment < true taxes owed (and most of them pay zero) Implications for the compliance puzzle debate: The Becker-Allingham-Sandmo model is 20% wrong, 80% right

Compliance Under Zero Deterrence [Distributions of Payment Made vs Payment Owed] (Control Group)

Finding 2: Duty to Comply Sharp bunching at exact compliance under zero deterrence: No monetary discontinuity at exact compliance under zero deterrence ) Discontinuous social motivation to exactly comply with the law ) naturally labelled as duty to comply Reverse bunching approach : Normally bunching identi es response to discontinuous incentive, assuming smooth preferences Here bunching identi es discontinuity in preferences given smooth incentives

Duty to Comply [Bunching at Exact Compliance Under Zero Deterrence] (Control Group)

Finding 3: Simpli cation Matters, Misperception Less So Tax Simpli cation: Baseline evaders: strong positive e ect on compliance Baseline donors: no signi cant e ect Misperception of Audit Probability: Baseline evaders: weak positive e ect on compliance Baseline donors: no signi cant e ect

Tax Simpli cation: Total Response Impacts [Heterogeneity Across Compliance Types] (Tax Simpli cation vs Control Group) Full Sample Baseline Evaders (Economically Motivated) Baseline Donors (Socially Motivated) Probability of Evading Evasion Rate Probability of Evading Evasion Rate Probability of Donating Increased Donation Rate Tax Simplification 2.45** 3.17*** 2.66*** 3.36*** 4.04 30.37 (0.971) (0.910) (0.747) (0.732) (6.97) (27.88) Compliance in omitted reference group 79.29% 76.35% 94.98% 92.35% 62.34% 9.09% Number of Observations 5076 5076 4007 4007 476 476

Tax Simpli cation: Total Response Impacts [Distributional E ects on Baseline Evaders] (Tax Simpli cation vs Control Group)

Correcting Misperception: Total Response Impacts [Heterogeneity Across Compliance Types] (Zero Audit Probability vs Tax Simpli cation Group) Full Sample Baseline Evaders (Economically Motivated) Baseline Donors (Socially Motivated) Probability of Evading Evasion Rate Probability of Evading Evasion Rate Probability of Donating Increased Donation Rate Correcting Misperception 0.942 0.957 1.53** 1.41* 1.52 50.61 (0.889) (0.857) (0.715) (0.720) (5.78) (37.38) Compliance in omitted reference group 77.30% 73.37% 92.35% 88.69% 61.63% 5.71% Number of Observations 7641 7641 6049 6049 723 723

Correcting Misperception: Total Response Impacts [Distributional E ects on Baseline Evaders] (Zero Audit Probab vs Tax Simpli cation Group)

Finding 4: Deterrence Works Baseline evaders: compliance increases with audit probability Signi cant increases for uniform audit probabilities of 10%, 20%, 50% Stronger e ects for notched audit probability Baseline donors: weak e ect of audit probability Weak e ect on probability of donating but no e ect on donation amount Crowd-in of social motivation (social image concerns, perception of fairness, relative warm-glow e ects)

Positive Audit Probability: Total Response Impacts [Heterogeneity Across Compliance Types] (Positive Audit Probability Groups vs Zero Audit Probability Group) Full Sample Baseline Evaders (Economically Motivated) Baseline Donors (Socially Motivated) Probability of Evading Evasion Rate Probability of Evading Evasion Rate Probability of Donating Increased Donation Rate Deterrence 3.13*** 2.80*** 3.12*** 2.81*** 7.07* 25.59 (0.660) (0.627) (0.536) (0.534) (4.22) (22.20) Compliance in omitted reference group 78.04% 74.52% 93.80% 90.43% 61.72% 8.16% Number of Observations 12692 12692 9979 9979 1261 1261

Positive Audit Probability: Total Response Impacts [Distributional E ects on Baseline Evaders] (Positive Audit Probability Groups vs Zero Audit Probability Group)

E ect of Audit Notch Treatment (Audit Probability Notch vs Zero Audit Probability Group)

Finding 5: Compliance E ects of Perceived Penalty Audit results suggest a perceived penalty of being caught evading But we can provide direct evidence using a novel bunching approach: Asymmetric treatment of evaders (expected penalty proportional to p) and donors (no monetary penalty) Penalty kink at the point of exact compliance! bunching But there is also bunching due to duty-to-comply. Hence we consider Change in bunching when moving from p = 0 (only duty-to-comply) to p > 0 (duty-to-comply + perceived penalty)

Perceived Penalty [Bunching at Exact Compliance] (Positive Audit Probability Groups vs Simpli cation Group)

Finding 6: Recognition Works on the Socially Motivated, Not on the Economically Motivated Social and monetary rewards create opposite e ects on the two types: Baseline donors increase donations Baseline evaders further reduce compliance Average e ect is strongly attenuated Interpretation: Rewarding compliers (rather than punishing evaders) highlights voluntary contribution aspect and downplays mandatory tax aspect

Recognition: Total Response Impacts [Heterogeneity Across Compliance Types] (Reward Groups vs Tax Simpli cation Group) Full Sample Baseline Evaders (Economically Motivated) Baseline Donors (Socially Motivated) Probability of Evading Evasion Rate Probability of Evading Evasion Rate Probability of Donating Increased Donation Rate Compliance Rewards 0.259 0.447 1.27* 1.20* 2.02 83.24** (0.821) (0.789) (0.664) (0.666) (4.95) (32.80) Compliance in omitted reference group 77.30% 73.37% 92.35% 88.69% 61.63% 5.71% Number of Observations 12632 12632 9909 9909 1247 1247

Compliance Rewards: Total Response Impacts [Distributional E ects on Baseline Evaders] (Social and Monetary Reward Groups vs Tax Simpli cation Group)

Conclusions Compelling evidence on the role of social incentives for tax compliance Direct evidence on social motivation from a zero deterrence baseline: Becker-Allingham-Sandmo model is 20% wrong, 80% right Study interaction between individual motivation and policy First integrated treatment of tax compliance and charitable giving Imperfectly enforced tax systems have an element of voluntary giving Does the unique setting pose a trade-o with external validity? Church members are not inherently more compliant than the overall population (Kleven et al. 2011)