Extrinsic vs Intrinsic Motivations for Tax Compliance

Extrinsic vs Intrinsic Motivations for Tax Compliance:
Evidence from a Field Experiment in Germany
Nadja Dwenger, MPI for Tax Law and Public Finance
Henrik Kleven, LSE
Imran Rasul, UCL
Johannes Rincke, University of Erlangen-Nuremberg
May 2014
Introduction
2
The Tax Compliance Puzzle
Gap between theory and evidence:
◾ Theory of tax evasion (Becker-Allingham-Sandmo) predicts low
compliance under low audit probabilities or penalties
◾ Evidence shows high compliance in modern tax systems despite very
low audit probabilities and penalties
Closing the gap:
◾ Third-party information creates large detection probabilities
conditional on evading (Kleven et al. 2011)
◾ Misperception of enforcement parameters (Scholz-Pinney 1993)
◾ Social motives and psychology (Andreoni et al. 1998)
Contribution
3
Extrinsic vs. Intrinsic Motivations
Extrinsic motivations:
◾ Much is known about the compliance eects of deterrence
Intrinsic motivations:
◾ Little is known about the compliance eects of morale, norms, and
psychology
Our study:
◾ Extrinsic and intrinsic motivation can be identied and measured in
the baseline
◾ Injects deterrence and recognition into the system and estimates the
eects on the extrinsically and intrinsically motivated
Institutional Background
Local Church Tax in Germany: Three Unique Features
Feature 1: tax and charity
◾ Legal obligation to pay for church members
◾ Overpayments represent church donations
Feature 2: true tax base is observable
◾ True tax base = 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 = intrinsic motivation + misperception,
and we can separate the two
4
Institutional Background
Local Church Tax in Germany: Identifying Motivation
Baseline behavior (p = 0) reveals motivation:
◾ Baseline donors: charity motivated (strongly intrinsic)
◾ Baseline compliers: compliance motivated (intrinsic)
◾ Baseline evaders: enforcement motivated (extrinsic)
Each type could also be aected by misperception (p > 0)
◾ Estimate misperception via (randomized) communication of zero
deterrence
◾ Separate misperception from motivation
5
Institutional Background
Local Church Tax in Germany: Tax Schedule
6
Theoretical Framework
Theoretical Framework
Model of tax compliance that unies
◾ standard Becker-Allingham-Sandmo model: extrinsic motivation
◾ Andreoni-warm-glow model of pro-social behavior:
intrinsic motivation
Gives predictions on three types of individuals:
◾ evaders: underpay
◾ compliers: pay the amount due
◾ donors: overpay
Model generates type-specic predictions on responses to shifts in
extrinsic & intrinsic motivation
7
Randomized Field Experiment
8
Deterrence and Recognition in a Field Experiment
Letter treatments in a randomized eld experiment
Deterrence treatments (extrinsic) and recognition treatments (intrinsic):
◾ Tax salience
◾ Zero audit probability
◾ Positive audit probability and audit probability notch
◾ Recognition (social and monetary rewards)
Social reward: Lottery for local newspaper ad naming winners. This and
all other treatments truthfully implemented
Study heterogeneity between the extrinsically and intrinsically motivated
Data
Data
Administrative tax data 2008-2012
Individual records of tax compliance for up to four years pre-treatment
Treated individuals in the experiment: about 48,000 individuals/year
Randomization
9
Results
10
Finding 1: Intrinsic Motivation Is Substantial
With zero deterrence, the predicted compliance rate is zero absent
intrinsic motivation
But a signicant fraction of individuals do comply
◾ 20% of the full sample makes payment ≥ true payment
Implications for the compliance puzzle debate:
◾ Intrinsic motivation is substantial, but still the majority behaves as
rational, selsh taxpayers
◾ The Becker-Allingham-Sandmo model is 80% correct
Results
11
Baseline Distribution of Payment Made vs. Payment Owed
60
80
Full Sample
0
20
Share in %
40
People Tax Gap: 79.6 (.757)***
Money Tax Gap: 76.7% (.764)***
0
5 10
25
45
Payment in Euro
Tax Payment Made
70
100+
Tax Payment Owed
Results
12
Finding 2: Tax Salience Matters, Misperception Does Not
Tax Salience:
◾ Baseline evaders: simplication and highlighting legal obligation
increases compliance
◾ Baseline compliers and donors: very small and insignicant eects
Misperception of Audit Probability:
◾ Compliance unaected by salient, zero audit probability letter
◾ Little misperception of audit probability on average
Results
13
Eects of Tax Salience
Salience vs. Control Letter
Table 1: Effects of Tax Salience
Extrinsically Motivated
(Baseline Evaders)
Intrinsically Motivated
(Baseline Compliers/Donors)
(1)
(2)
A: Effect on Probability for Increased Payment (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
64.8***
-17.5
(13.7)
(14.6)
6.23%
14.6%
B: Effect on Tax and Donation Gap (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
Number of Observations
-3.36***
3.51
(.732)
(7.62)
92.3%
40.6%
4007
1069
Results
13
Eects of Tax Salience
Salience vs. Control Letter
Table 1: Effects of Tax Salience
Extrinsically Motivated
(Baseline Evaders)
Intrinsically Motivated
(Baseline Compliers/Donors)
(1)
(2)
A: Effect on Probability for Increased Payment (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
64.8***
-17.5
(13.7)
(14.6)
6.23%
14.6%
B: Effect on Tax and Donation Gap (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
Number of Observations
-3.36***
3.51
(.732)
(7.62)
92.3%
40.6%
4007
1069
Results
14
Eects of Misperception
Zero Audit Probability vs. Salience Letter
Table 2: Effects of Correcting Misperception
Extrinsically Motivated
(Baseline Evaders)
Intrinsically Motivated
(Baseline Compliers/Donors)
(1)
(2)
A: Effect on Probability for Increased Payment (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
-11.0
-11.6
(7.55)
(14.2)
10.5%
12.4%
B: Effect on Tax and Donation Gap (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
Number of Observations
1.41**
-5.33
(.720)
(5.96)
88.7%
43.6%
6049
1592
Results
14
Eects of Misperception
Zero Audit Probability vs. Salience Letter
Table 2: Effects of Correcting Misperception
Extrinsically Motivated
(Baseline Evaders)
Intrinsically Motivated
(Baseline Compliers/Donors)
(1)
(2)
A: Effect on Probability for Increased Payment (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
-11.0
-11.6
(7.55)
(14.2)
10.5%
12.4%
B: Effect on Tax and Donation Gap (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
Number of Observations
1.41**
-5.33
(.720)
(5.96)
88.7%
43.6%
6049
1592
Results
15
Finding 3: Deterrence Works on Evaders, Not on Donors
Baseline evaders: compliance increases with audit probability
◾ Signicant increases for uniform audit probabilities 10%, 20%, 50%
◾ Much stronger eects for notched audit probability
Baseline compliers and donors: audit threats leave behavior unaected
Deterrence increases compliance of the extrinsically motivated, but not
the intrinsically motivated
Results
16
Eects of Audit Threats
Higher Audit Probability Letters vs. Salience Letter
Table 2: Effects of Higher Audit Probabilities
Extrinsically Motivated
(Baseline Evaders)
Intrinsically Motivated
(Baseline Compliers/Donors)
(1)
(2)
A: Effect on Probability for Increased Payment (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
29.8***
18.6
(6.64)
(12.0)
9.00%
11.0%
B: Effect on Tax and Donation Gap (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
Number of Observations
-2.81***
5.62
(.534)
(4.81)
90.4%
39.6%
9979
2713
Results
16
Eects of Audit Threats
Higher Audit Probability Letters vs. Salience Letter
Table 2: Effects of Higher Audit Probabilities
Extrinsically Motivated
(Baseline Evaders)
Intrinsically Motivated
(Baseline Compliers/Donors)
(1)
(2)
A: Effect on Probability for Increased Payment (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
29.8***
18.6
(6.64)
(12.0)
9.00%
11.0%
B: Effect on Tax and Donation Gap (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
Number of Observations
-2.81***
5.62
(.534)
(4.81)
90.4%
39.6%
9979
2713
Results
17
Excess Bunching by Audit Probability Notch
Notch Treatment: 50% audit probability if payment
≤ 10 Euro, zero audit probability if payment > 10 Euro
Excess bunching = causal eect of audit probability on payments
Difference
in density
Difference in densities
between audit notch
treatment and control
group
0
10 €
Payment
Results
18
Finding 4: Eect of Rewards Depend on Motivation
Recognition (social and monetary rewards) has opposite eects on
baseline evaders and donors:
◾ Baseline evaders: reduce tax payment
◾ Baseline compliers: unaected
◾ Baseline donors: increase payment
Whether a reward for compliance works or backres depends crucially on
what motivates taxpayers in the rst place
Results
19
Eects of Social & Monetary Rewards
Rewards Letters vs. Salience Letter
Table 4: Effects of Social and Monetary Rewards
Extrinsically Motivated
(Baseline Evaders)
Intrinsically Motivated
(Baseline Compliers)
Strongly Intrinsically Motivated
(Baseline Donors)
(1)
(2)
(3)
A: Effect on Probability of Increased Payment (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
-15.6**
-7.72
48.3*
(6.90)
(15.4)
(25.3)
10.5%
15.5%
8.57%
B: Effect on Tax and Donation Gap (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
1.20*
-5.13
-.198
(.666)
(15.2)
(5.10)
88.7%
17.6%
75.1%
C: Effect on Increased Donation Gap (in %)
Controls for past compliance, demographic, and strata variables
83.2**
Compliance in omitted reference group
5.71%
(32.8)
Number of Observations
9909
1476
1247
Results
19
Eects of Social & Monetary Rewards
Rewards Letters vs. Salience Letter
Table 4: Effects of Social and Monetary Rewards
Extrinsically Motivated
(Baseline Evaders)
Intrinsically Motivated
(Baseline Compliers)
Strongly Intrinsically Motivated
(Baseline Donors)
(1)
(2)
(3)
A: Effect on Probability of Increased Payment (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
-15.6**
-7.72
48.3*
(6.90)
(15.4)
(25.3)
10.5%
15.5%
8.57%
B: Effect on Tax and Donation Gap (in %)
Controls for past compliance, demographic, and strata variables
Compliance in omitted reference group
1.20*
-5.13
-.198
(.666)
(15.2)
(5.10)
88.7%
17.6%
75.1%
C: Effect on Increased Donation Gap (in %)
Controls for past compliance, demographic, and strata variables
83.2**
Compliance in omitted reference group
5.71%
(32.8)
Number of Observations
9909
1476
1247
Conclusions
20
Conclusions
Unique test of the Allingham-Sandmo (1972) model and the impact of
social and economic incentives on tax compliance
◾ Directly measure evasion in an AS eld setting (with baseline p = 0)
◾ Randomized letter treatments manipulating deterrence and
non-deterrence channels
We nd that
◾ 80% of taxpayers evade absent enforcement
◾ deterrence increases compliance among the extrinsically motivated
◾ rewards are ineective or backre for the extrinsically motivated...
◾ ...but work for the intrinsically motivated
Thank You
21
THANK YOU
[email protected]
[email protected]
[email protected]
[email protected]
22
Appendix
Successful Randomization
Table A1: Taxpayer Characteristics by Treatment Assignment
Number of
individuals
Male
Age
(1)
(2)
(3)
Married
Number of
Children
Joint Filer
[yes=1]
Wage Income Capital Income Liable for
[yes=1]
Above Tax
Trade Tax
Exempt Level
[yes=1]
[yes=1]
(4)
(5)
(6)
(7)
(8)
p -value of the F -test
on Joint Significance
Relative to
Income
(in Euro)
Probability
of paying in
2010
(9)
(10)
(11)
(12)
(13)
{.557}
Control
Salience
Letter
Control Group
2836
0.500 44.424
0.418
0.478
0.365
0.878
0.004
0.028
42,426.560
0.270
-
Simplified Letter
2872
0.504 44.000
0.414
0.463
0.363
0.873
0.003
0.026
43,225.510
0.285
[.557]
-
p=0
5681
0.497 44.050
0.403
0.453
0.353
0.879
0.002
0.025
42,933.980
0.278
[.613]
{.970}
p=.1
2863
0.501 43.739
0.393
0.473
0.348
0.893
0.003
0.026
41,082.650
0.284
[.200]
{.319}
p=.2
2844
0.496 43.786
0.404
0.445
0.353
0.879
0.001
0.021
41,220.260
0.300
[.009]
{.516}
p=.5
2819
0.510 44.089
0.415
0.473
0.362
0.878
0.003
0.023
42,288.990
0.285
[.748]
{.997}
Threshold
2852
0.504 43.730
0.412
0.484
0.352
0.884
0.003
0.023
41,768.010
0.278
[.378]
{.683}
Social Reward
2846
0.514 43.978
0.408
0.466
0.359
0.870
0.004
0.025
42,883.080
0.283
[.362]
{.967}
Small Monetary Reward
2831
0.495 43.842
0.409
0.469
0.356
0.881
0.004
0.026
40,865.230
0.285
[.369]
{.668}
Large Monetary Reward
2836
0.503 44.234
0.438
0.496
0.390
0.887
0.002
0.023
42,649.880
0.287
[.300]
{.530}
Social + Monetary Reward
2848
0.513 43.647
0.408
0.483
0.349
0.880
0.003
0.026
41,975.310
0.280
[.193]
{.521}
Local
2795
0.503 43.736
0.412
0.441
0.352
0.876
0.005
0.022
42,301.670
0.278
[.132]
{.893}
Average Payment 2011
2859
0.288
[.567]
{.656}
Back to Data
0.491 44.304
0.402
0.456
0.346
0.883
0.002
0.025
42,303.270