tax evasion at the top of the income distribution: theory
TRANSCRIPT
Tax Evasion at the Top of the Income Distribution:Theory and Evidence
John Guyton (IRS)Patrick Langetieg (IRS)
Daniel Reck (London School of Economics)Max Risch (Carnegie Mellon University)
Gabriel Zucman (UC Berkeley)
September 20211
1The views expressed here are those of the authors and do not necessarily reflect the official view of the Internal Revenue
Service. This project was conducted through the Joint Statistical Research Program of the Statistics of Income Division of theIRS. All data work for this project involving confidential taxpayer information was done on IRS computers, by IRS employees,and at no time was confidential taxpayer data ever outside of the IRS computing environment. Reck and Risch have been IRSemployees detailed under agreements made possible by the Intragovernmental Personnel Act of 1970 (5 U.S.C. 3371-3376) andthe IRS Student Volunteer Program. 1 / 17
Motivation
What is the distribution and composition of tax evasion in theUnited States?
. Important for the study of inequality, distributional nationalaccounts, and tax enforcement policy
. Key data source for empirical study of individual tax evasion:risk-stratified random audits (National Research Program, NRPin the US)
. Important challenge: difficulty capturing top-end evasion
. Audits do not detect all evasion
. Likely esp. difficult at the top: complex finances andpass-through ownership structures
2 / 17
This paper
. Combine data on risk-stratified random audits, operationalaudits, and targeted offshore enforcement
. Demonstrate that standard examination procedures for individualaudits miss some top-end evasion
. e.g. offshore financial accounts, pass-through structures
. Use conservative methods to make the basic point, then attempt moreambitiously to correct aggregates
. These types of audits generate the risk-stratified random audit dataused in estimating the tax gap
. Propose adjustments to IRS estimates of income under-reporting,tax gap for sophisticated tax evasion
. Adjustments modest on aggregate: adjusted tax gap = 1.1x currentestimate
. ...but large at the top: x1.5 for top 1%, >2x for top 0.1%
. Increases estimated top income share by ≈ 1 pp
3 / 17
This paper
. Combine data on risk-stratified random audits, operationalaudits, and targeted offshore enforcement
. Demonstrate that standard examination procedures for individualaudits miss some top-end evasion
. e.g. offshore financial accounts, pass-through structures
. Use conservative methods to make the basic point, then attempt moreambitiously to correct aggregates
. These types of audits generate the risk-stratified random audit dataused in estimating the tax gap
. Propose adjustments to IRS estimates of income under-reporting,tax gap for sophisticated tax evasion
. Adjustments modest on aggregate: adjusted tax gap = 1.1x currentestimate
. ...but large at the top: x1.5 for top 1%, >2x for top 0.1%
. Increases estimated top income share by ≈ 1 pp3 / 17
Background: NRP risk-stratified random audits
. Auditors assess compliance across the full individual tax return(F1040)
. Same procedures as individual audits done in the SmallBusiness and Self-Employed (SBSE) division of IRS
. We pool 2006–07, 2008–09, 2010–13 NRP waves (IRS 2012,2016, 2019)
. Risk-stratified random sample. Over-samples top & negativeincomes
. → 105,160 obs, 12,053 in the top 1% by reported income
. Used weighted data throughout to represent full population
4 / 17
Background: NRP risk-stratified random audits
. Auditors assess compliance across the full individual tax return(F1040)
. Same procedures as individual audits done in the SmallBusiness and Self-Employed (SBSE) division of IRS
. We pool 2006–07, 2008–09, 2010–13 NRP waves (IRS 2012,2016, 2019)
. Risk-stratified random sample. Over-samples top & negativeincomes
. → 105,160 obs, 12,053 in the top 1% by reported income
. Used weighted data throughout to represent full population
4 / 17
Background: NRP risk-stratified random audits
. Auditors assess compliance across the full individual tax return(F1040)
. Same procedures as individual audits done in the SmallBusiness and Self-Employed (SBSE) division of IRS
. We pool 2006–07, 2008–09, 2010–13 NRP waves (IRS 2012,2016, 2019)
. Risk-stratified random sample. Over-samples top & negativeincomes
. → 105,160 obs, 12,053 in the top 1% by reported income
. Used weighted data throughout to represent full population
4 / 17
NRP data: Methods
. Conditional on income, estimate:. Income under-reporting gap = total under-reporting /
total true income
. Tax gap = total evaded tax / total true tax
. Rank individuals by corrected market income (re-rankingmatters!)
. Initially: under-reporting detected in risk-stratified random auditsonly
. Then examine, add sophisticated evasion offshore and inpass-through businesses
. Then account for “Detection Controlled Estimation” adjustments. Used in official tax gap statistics to account for un-detected evasion
due to differential auditor effectiveness
. Add DCE last, account for potential overlap with sophisticated evasion
. Note: DCE adjustments here are based on the multiplier method of Johns & Slemrod (2010).
. Note: Extensive sensitivity analysis in paper, more in progress
5 / 17
NRP data: Methods
. Conditional on income, estimate:. Income under-reporting gap = total under-reporting /
total true income
. Tax gap = total evaded tax / total true tax
. Rank individuals by corrected market income (re-rankingmatters!)
. Initially: under-reporting detected in risk-stratified random auditsonly
. Then examine, add sophisticated evasion offshore and inpass-through businesses
. Then account for “Detection Controlled Estimation” adjustments. Used in official tax gap statistics to account for un-detected evasion
due to differential auditor effectiveness
. Add DCE last, account for potential overlap with sophisticated evasion
. Note: DCE adjustments here are based on the multiplier method of Johns & Slemrod (2010).
. Note: Extensive sensitivity analysis in paper, more in progress
5 / 17
NRP data: Methods
. Conditional on income, estimate:. Income under-reporting gap = total under-reporting /
total true income
. Tax gap = total evaded tax / total true tax
. Rank individuals by corrected market income (re-rankingmatters!)
. Initially: under-reporting detected in risk-stratified random auditsonly
. Then examine, add sophisticated evasion offshore and inpass-through businesses
. Then account for “Detection Controlled Estimation” adjustments. Used in official tax gap statistics to account for un-detected evasion
due to differential auditor effectiveness
. Add DCE last, account for potential overlap with sophisticated evasion
. Note: DCE adjustments here are based on the multiplier method of Johns & Slemrod (2010).
. Note: Extensive sensitivity analysis in paper, more in progress5 / 17
Risk-stratified random audit total incomeunder-reporting - No DCE
-1.0%
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
6.0%
7.0%
P1
0-2
0
P2
0-3
0
P3
0-4
0
P4
0-5
0
P5
0-6
0
P6
0-7
0
P7
0-8
0
P8
0-9
0
P9
0-9
5
P95-9
9
P9
9-9
9.5
P9
9.5
-P99
.9
P9
9.9
-99.9
5
P9
9.9
5-P
99
.99
P9
9.9
9-1
00
incom
e u
nder-
report
ing,
% o
f tr
ue incom
e
(report
ed +
unre
port
ed)
Position in the income distribution (exam-corrected income)
6 / 17
Risk-stratified Random Audits Detect Mostly Self-Employment (Sch. C) Non-Compliance Table
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
6.0%
7.0%
P1
0-2
0
P2
0-3
0
P3
0-4
0
P4
0-5
0
P5
0-6
0
P6
0-7
0
P7
0-8
0
P8
0-9
0
P9
0-9
5
P9
5-9
9
P9
9-9
9.5
P9
9.5
-P99
.9
P9
9.9
-99.9
5
P9
9.9
5-P
99
.99
P99.9
9-1
00
incom
e u
nder-
report
ing,
% o
f tr
ue incom
e(r
eport
ed +
unre
port
ed)
Position in the income distribution (exam-corrected income)
Schedule C income
F1040 "other" income
Partnerships and S-corporation income
Rental income
Capital Gains
Other
Wages
DividendsInterests
7 / 17
Risk-stratified random audits do not detect offshore evasion that we
know was happening Rare Detection of FBAR Non-Compliance
8 / 17
Adding unreported offshore income to risk-stratifiedrandom audit estimates (No DCE) Sensitivity
0%
1%
2%
3%
4%
5%
6%
7%
P1
0-2
0
P20-3
0
P3
0-4
0
P4
0-5
0
P5
0-6
0
P6
0-7
0
P7
0-8
0
P80-9
0
P9
0-9
5
P9
5-9
9
P9
9-9
9.5
P9
9.5
-99.9
P9
9.9
-P99
.99
P99.9
9-P
100
incom
e u
nder-
rep
ort
ing
, %
tru
e incom
e(r
eport
ed +
unre
port
ed)
Position in the income distribution (exam-corrected income)
NRP (No DCE) + Offshore
NRP (No DCE)
4.0 ppor$22.5b
10 / 17
Pass-through evasion: Background
. NRP risk-stratified random audits primarily examine individualtax returns
. Pass-through businesses (S-corps, partnerships) file own taxreturns, distribute income to their owners on Schedules K-1
. Importance of pass-through income at the top of the US incomedistribution is high and growing (Smith et al 2019)
. Standard individual audits have limited resources to examinepass-throughs
. Entity-level returns are very rarely examined in NRP audits
=⇒ NRP risk-stratified random audits likely do not detect allevasion via pass-through business entities
11 / 17
Pass-through evasion: Background
. NRP risk-stratified random audits primarily examine individualtax returns
. Pass-through businesses (S-corps, partnerships) file own taxreturns, distribute income to their owners on Schedules K-1
. Importance of pass-through income at the top of the US incomedistribution is high and growing (Smith et al 2019)
. Standard individual audits have limited resources to examinepass-throughs
. Entity-level returns are very rarely examined in NRP audits
=⇒ NRP risk-stratified random audits likely do not detect allevasion via pass-through business entities
11 / 17
Rarity of Business-Level Audits =⇒ Rarity of DetectedUnder-Reporting for S Corps and Partnerships
12 / 17
Adding Pass-Through Evasion To Risk-Stratified RandomAudit Estimates (No DCE) Sensitivity
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
6.0%
7.0%
8.0%
9.0%
P1
0-2
0
P2
0-3
0
P3
0-4
0
P4
0-5
0
P5
0-6
0
P6
0-7
0
P7
0-8
0
P8
0-9
0
P9
0-9
5
P9
5-9
9
P9
9-9
9.5
P9
9.5
-99.9
P9
9.9
-P99
.95
P9
9.9
5-P
99
.99
P9
9.9
9-1
00
% o
f to
tal in
com
e (
report
ed +
unre
port
ed)
Position in the income distribution (exam-corrected income)
NRP (No DCE)
NRP (No DCE) + pass-through businesses
13 / 17
Implications for Inequality
. Accounting for federal income tax evasion increases theestimated top 1% share
. Conservative adjustments for offshore and pass-through evasionincrease the top 1% income share
. ≈ 1 pp relative to top 1% reported income share
. rise in top 1% share is robust to alternative specifications
. Large effect on distribution of unreported income:
. >20 pp increase in top 1% share of unreported income whenincluding offshore and pass-through evasion
. Benchmark: 31% of unreported income, 36% of unpaidfederal income tax attributable to top 1%
16 / 17
Implications for Inequality
. Accounting for federal income tax evasion increases theestimated top 1% share
. Conservative adjustments for offshore and pass-through evasionincrease the top 1% income share
. ≈ 1 pp relative to top 1% reported income share
. rise in top 1% share is robust to alternative specifications
. Large effect on distribution of unreported income:
. >20 pp increase in top 1% share of unreported income whenincluding offshore and pass-through evasion
. Benchmark: 31% of unreported income, 36% of unpaidfederal income tax attributable to top 1%
16 / 17
What Have We Learned?
. Naive interpretations of risk-stratified random audit data suggestthat the rate of evasion is decreasing with income throughout thedistribution
. Other evidence =⇒ sophisticated evasion is common at the top
. At least two types of sophisticated evasion are prevalent enoughto matter for macro aggregates:
. Concealed offshore financial accounts (at least pre-FATCA)
. Evasion via pass-through businesses
. Detection of these is so rare that DCE corrections cannotplausibly capture most of this.
. These two forms of evasion add ≈ $50 Billion per year to the taxgap in our sample period, virtually all from the top 1%
. Theory of costly concealment can explain why such evasion isconcentrated at the top
17 / 17
What Have We Learned?
. Naive interpretations of risk-stratified random audit data suggestthat the rate of evasion is decreasing with income throughout thedistribution
. Other evidence =⇒ sophisticated evasion is common at the top
. At least two types of sophisticated evasion are prevalent enoughto matter for macro aggregates:
. Concealed offshore financial accounts (at least pre-FATCA)
. Evasion via pass-through businesses
. Detection of these is so rare that DCE corrections cannotplausibly capture most of this.
. These two forms of evasion add ≈ $50 Billion per year to the taxgap in our sample period, virtually all from the top 1%
. Theory of costly concealment can explain why such evasion isconcentrated at the top
17 / 17
What Have We Learned?
. Naive interpretations of risk-stratified random audit data suggestthat the rate of evasion is decreasing with income throughout thedistribution
. Other evidence =⇒ sophisticated evasion is common at the top
. At least two types of sophisticated evasion are prevalent enoughto matter for macro aggregates:
. Concealed offshore financial accounts (at least pre-FATCA)
. Evasion via pass-through businesses
. Detection of these is so rare that DCE corrections cannotplausibly capture most of this.
. These two forms of evasion add ≈ $50 Billion per year to the taxgap in our sample period, virtually all from the top 1%
. Theory of costly concealment can explain why such evasion isconcentrated at the top
17 / 17
NRP - No DCE Decompositions Back
Full Population Top 1%Total Total under- Total under- Total under- Total Total under- Total under- Total under-
income of reported reported reported income of reported reported reportedthis type/ income of income of income of this type/ income of income of income of
Total this type/ this type/ this type/ Total this type/ this type/ this type/income (%) Total Total Total income (%) Total Total Total
under-reported income (%) income of under-reported income (%) income ofincome (%) this type (%) income (%) this type (%)
Capital Gains 5.8 7.1 0.28 4.8 21.3 18.8 0.43 2.0Dividends 3.9 2.8 0.11 2.9 8.6 3.9 0.09 1.0Interest 1.9 0.7 0.03 1.5 3.0 2.0 0.05 1.6Line 21 Other Income 0.2 11.9 0.47 253.6 2.6 8.5 0.43 7.5Partnerships and S Corp 5.6 6.5 0.26 4.6 21.7 18.9 0.43 2.0Rental 0.7 8.9 0.35 48.3 1.6 5.4 0.12 7.9Schedule C 5.3 49.3 1.95 36.8 4.2 35.0 0.79 18.7Wages 72.4 3.5 0.14 0.2 38.2 2.9 0.07 0.2Other 4.1 9.3 0.37 0.1 -1.0 4.6 0.10 -0.1Total 100.0 100.0 3.96 100.0 100.0 2.27
1 / 17
The extent of third-party reporting declines with incomeat the very top Back
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
80.0%
90.0%
100.0%
P1
0-2
0
P2
0-3
0
P3
0-4
0
P4
0-5
0
P50-6
0
P6
0-7
0
P7
0-8
0
P8
0-9
0
P9
0-9
5
P9
5-9
9
P9
9-9
9.5
P9
9.5
-P99
.9
P9
9.9
-99.9
5
P9
9.9
5-P
99
.99
P9
9.9
9-1
00
Incom
e S
hare
by I
nfo
rmation C
ate
gory
Position in the distribution (exam-corrected income)
Substantial Information and Withholding
Some Information
Little or no Information
2 / 17
Operational audits of 10% of top 0.01% taxpayers detectmore than NRP estimates for the full population Details
Back
3 / 17
Assumptions for Offshore IncomeCorrections
Parameter Lower-bound Preferred Upper-boundscenario scenario scenario
Amount of US offshore wealth (in billion $) 750 1,058 1,500Fraction of offshore wealth concealed 85% 95% 100%Rate of return on offshore wealth 4.65 % 6% 11%Distribution of offshore wealth FBAR Average of FBAR and Nordic NordicAverage Marginal Tax Rate 20% 25% 30%
Back
6 / 17
FBAR non-compliance is virtually never detected in NRPrisk-stratified random audits Back
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
3.5%
4.0%
4.5%
5.0%
p0-p50 p51-p90 p91-p99 p99-p99.9 p99.9-p100
Pro
babili
ty o
f dete
cte
d F
BA
R fili
ng r
equirem
en
t
Position in the exam-corrected income distribution
Compliant
Non-compliant
7 / 17
Adding unreported offshore income torisk-stratified random audit estimates Back
0.0%
2.0%
4.0%
6.0%
8.0%
10.0%
12.0%
14.0%
16.0%
P1
0-2
0
P2
0-3
0
P3
0-4
0
P40-5
0
P5
0-6
0
P6
0-7
0
P7
0-8
0
P8
0-9
0
P9
0-9
5
P9
5-9
9
P9
9-9
9.5
P9
9.5
-99.9
P9
9.9
-P99
.99
P9
9.9
9-P
10
0
incom
e u
nder-
report
ing,
% tru
e incom
e(r
epo
rted
+ u
nre
po
rted
)
Position in the income distribution (exam-corrected income)
NRP
NRP + Offshore (Lower-bound)
NRP + Offshore (Benchmark)
NRP + Offshore (Upper-bound)
8 / 17
Details on allocation of pass-throughincome
. Business income shares for partnerships from Cooper et al(2016), TY2011
. Business income shares for S corps from SOI tabulations of Scorp returns (Saez and Zucman 2020)
. Allocate financial capital income for pass-throughs as follows:
. Total pass-through interest from SOI / Total interest inTY2011
. Assume pass-through share is constant through incomedistribution (conservative wrt concentration)
. =⇒ distribute total pass-through interest according tooverall individual interest income shares
. Similarly for dividends, cap gains
Back
9 / 17
Summary and Sensitivity Back
. In aggregate, the pass-through correction (1.3% of true income)≈ 2x larger than the offshore correction (0.7%)
. Top 1% inc under-reporting rate increases from 2.3% (NRP noDCE) to 6.6% (c.f. ≈ % for full pop)
. These estimates are conservative in a number of respects
. Dropoff within top 0.1% due to assumed low evasion rates forcap gains, could assume higher
. Other sensitivity checks in the paper:
. Disallow 20% of pass-through losses and allocate to the topFig
. Add 2/3 of income in circular partnerships Fig
. Vary under-reporting rates: 12%-28% business income0-10% cap gains, 0-6% dividends/interest Fig
10 / 17
Summary and Sensitivity Back
. In aggregate, the pass-through correction (1.3% of true income)≈ 2x larger than the offshore correction (0.7%)
. Top 1% inc under-reporting rate increases from 2.3% (NRP noDCE) to 6.6% (c.f. ≈ % for full pop)
. These estimates are conservative in a number of respects
. Dropoff within top 0.1% due to assumed low evasion rates forcap gains, could assume higher
. Other sensitivity checks in the paper:
. Disallow 20% of pass-through losses and allocate to the topFig
. Add 2/3 of income in circular partnerships Fig
. Vary under-reporting rates: 12%-28% business income0-10% cap gains, 0-6% dividends/interest Fig
10 / 17
Sensitivity analysis: partnershipmisreporting rates Back
0.0%
2.0%
4.0%
6.0%
8.0%
10.0%
12.0%
P1
0-2
0
P2
0-3
0
P3
0-4
0
P4
0-5
0
P5
0-6
0
P6
0-7
0
P7
0-8
0
P8
0-9
0
P9
0-9
5
P95-9
9
P9
9-9
9.5
P9
9.5
-99.9
P9
9.9
-P99
.95
P9
9.9
5-P
99
.99
P9
9.9
9-1
00
% o
f to
tal in
com
e (
report
ed +
unre
port
ed)
Position in the income distribution (exam-corrected income)
NRP
NRP + Pass-throughs (low-end scenario)
NRP + Pass-throughs (benchmark)
NRP + Pass-throughs (high-end scenario)
13 / 17
Combined Estimates: Sensitivity AnalysisBack
0%
5%
10%
15%
20%
25%
30%
P1
0-2
0
P2
0-3
0
P3
0-4
0
P4
0-5
0
P5
0-6
0
P6
0-7
0
P7
0-8
0
P8
0-9
0
P9
0-9
5
P9
5-9
9
P9
9-9
9.5
P9
9.5
-99.9
P9
9.9
-P99
.95
P9
9.9
5-P
99
.99
P9
9.9
9-1
00
incom
e u
nder-
report
ing,
% o
f tr
ue incom
e(r
eport
ed +
unre
port
ed)
Position in the income distribution (true income)
Lower-bound sophisticated evasion
Aggregate sophisticated evasion high end
Aggregate sophisticated evasion low end
Benchmark
Upper-bound sophisticated evasion
14 / 17
Importance of Complex Partnership Structures IncreasesWith Income At the Very Top Back
0%
10%
20%
30%
40%
50%
60%
70%
P1
0-2
0
P2
0-3
0
P30-4
0
P4
0-5
0
P5
0-6
0
P6
0-7
0
P7
0-8
0
P8
0-9
0
P9
0-9
5
P9
5-9
9
P9
9-9
9.5
P9
9.5
-99.9
P9
9.9
-99.9
5
P9
9.9
5-9
9.9
9
P9
9.9
9-1
00
Share
of to
tal part
ners
hip
incom
e th
at
derives fro
m
com
ple
x p
art
ners
hip
s
Position in the income distribution (reported income)
Any income from other partnerships
At least 50% of income from other partnerships
15 / 17
Most Top 0.1% Partnership Owners Receive Income fromComplex Partnerships Back
0%
10%
20%
30%
40%
50%
60%
70%
80%
P1
0-2
0
P2
0-3
0
P30-4
0
P4
0-5
0
P5
0-6
0
P6
0-7
0
P7
0-8
0
P8
0-9
0
P9
0-9
5
P9
5-9
9
P9
9-9
9.5
P9
9.5
-99.9
P9
9.9
-99.9
5
P9
9.9
5-9
9.9
9
P9
9.9
9-1
00
Sha
re o
f p
art
ners
hip
ow
ne
rs r
eceiv
ing
incom
e f
rom
com
ple
x p
art
ne
rship
s
Position in the income distribution (reported income)
Any income from other partnerships
At least 50% of income from other
partnerships
16 / 17
Benchmark Combined Estimates:Tax Gap (% of Taxes Owed) Back
0%
5%
10%
15%
20%
25%
30%P
10
-20
P2
0-3
0
P3
0-4
0
P4
0-5
0
P5
0-6
0
P6
0-7
0
P7
0-8
0
P8
0-9
0
P9
0-9
5
P9
5-9
9
P9
9-9
9.5
P9
9.5
-99.9
P9
9.9
-P99
.95
P9
9.9
5-P
99
.99
P9
9.9
9-1
00
Position in the income distribution (true income)
17 / 17