tax evasion and inequality€¦ · evasion is small on aggregate but high at the top, strong...
TRANSCRIPT
Tax Evasion and Inequality
Annette Alstadsæter (Norwegian U. of Life Sciences)Niels Johannesen (U. of Copenhagen)
Gabriel Zucman (UC Berkeley)
October 2017
Introduction
How big is tax evasion in rich countries and how is itdistributed?
An important question for:
Study of income and wealth inequality
Tax policy
Tax enforcement
Main challenge in the literature:hard to capture evasion at the top
Widely used source to study tax evasion: random audits
Faces two key challenges:
Small number of rich individuals sampled
Hard to detect complex evasion involvingintermediaries (private banks, shell corp., etc.)
→ Random audits need to be supplemented by other datasources to capture evasion by the wealthy
We analyze new data capturing evasionby the wealthy
Massive leaks from HSBC Switzerland and MossackFonseca (“Panama Papers”)
Leaks random & from big, representative intermed.
Match to tax records in Norway, Sweden, Denmark
Combine with macro stats on wealth hidden in taxhavens, random audits, and amnesty data
→ First estimate of size & distribution of total evasion
Main result: tax evasion is small overallbut high at the top
0%
10%
20%
30%
P0-
10
P10
-20
P20
-30
P30
-40
P40
-50
P50
-60
P60
-70
P70
-80
P80
-90
P90
-95
P95
-99
P99
-99.
5
P99
.5-9
9.9
P99
.9-P
99.9
5
P99
.95-
P99
.99
P99
.99-
P10
0
% o
f tax
es o
wed
Position in the wealth distribution
Taxes evaded, % of taxes owed (stratified random audits + leaks)
Average: 2.8%
Tax Evasion by the Wealthy:
Evidence from Leaks
The HSBC leak: a unique source to studyevasion through intermediaries
The proba to have an unreported HSBCaccount rises sharply within the top 1%
0.0%
0.2%
0.4%
0.6%
0.8%
1.0%
P90-P95 [0.6 – 0.9]
P95-P99 [0.9 – 2.0]
P99-P99.5 [2.0 – 3.0]
P99.5-P99.9 [3.0 – 9.1]
P99.9-P99.95 [9.1 – 14.6]
P99.95-P99.99 [14.6 – 44.5]
Top 0.01% [> 44.5]
Net wealth group [millions of US$]
Probability to own an unreported HSBC account, by wealth group (HSBC leak)
HSBC evaders hide close to half of theirwealth at HSBC
0%
10%
20%
30%
40%
50%
P90-P95 [0.6 – 0.9]
P95-P99 [0.9 – 2.0]
P99-P99.5 [2.0 – 3.0]
P99.5-P99.9 [3.0 – 9.1]
P99.9-P99.95 [9.1 – 14.6]
P99.95-P99.99 [14.6 – 44.5]
Top 0.01% [> 44.5]
Net wealth group [millions of US$]
Average wealth hidden at HSBC, by wealth group (%oftotalwealth(includingheldatHSBC))
The Panama Papers confirm the sharpgradient in use of tax havens by wealth
0.0%
0.2%
0.4%
0.6%
0.8%
1.0%
1.2%
P90-P95 [0.6 – 0.8]
P95-P99 [0.8 – 1.8]
P99-P99.5 [1.8 – 2.7]
P99.5-P99.9 [2.7 – 8.1]
P99.9-P99.95 [8.1 – 13.3]
P99.95-P99.99 [13.3 – 41.4]
Top 0.01% [> 41.4]
Net wealth group [millions of US$]
Probability to appear in the "Panama Papers", by wealth group (Shareholders of shell companies created by Mossack Fonseca)
Amnesty data show widespread evasion atthe top
0%
2%
4%
6%
8%
10%
12%
14%
P90-P95 [0.6 – 0.8]
P95-P99 [0.8 – 1.8]
P99-P99.5 [1.8 – 2.7]
P99.5-P99.9 [2.7 – 8.1]
P99.9-P99.95 [8.1 – 13.3]
P99.95-P99.99 [13.3 – 41.4]
Top 0.01% [> 41.4]
Net wealth group [millions of US$]
Probability to voluntarily disclose hidden wealth, by wealth group (Swedish and Norwegian tax amnesties)
Hidden wealth is extremely concentrated
0%
10%
20%
30%
40%
50%
60%
P0-50 P50-P90 P90-P99 P99-P99.9 P99.9-99.99 P.99.99-P100
% o
f tot
al re
cord
ed o
r hid
den
wea
lth
Position in the wealth distribution
Distribution of wealth: recorded vs. hidden
Hidden wealth disclosed in amnesty
Hidden wealth held at HSBC
All recorded wealth
On aggregate, Scandinavian countriesown relatively little offshore wealth
0%
10%
20%
30%
40%
50%
60%
70%
Kor
ea
Pol
and
Chi
na
Den
mar
k Fi
nlan
d Ja
pan
Indi
a N
orw
ay
Indo
nesi
a C
anad
a Ira
n S
wed
en
Net
herla
nds
Bra
zil
Aus
tralia
M
exic
o U
SA
Aus
tria
Thai
land
C
olom
bia
Irela
nd
Spa
in
Sou
th A
frica
Ita
ly
Rus
sia
Fran
ce
Ger
man
y U
K
Bel
gium
Tu
rkey
P
ortu
gal
Taiw
an
Gre
ece
Arg
entin
a R
ussi
a (N
EO
) S
audi
Ara
bia
Vene
zuel
a U
AE
Offshore wealth / GDP (All countries with GDP > $200 billion in 2007)
World average: 9.8%
Even in countries with low total evasion,including hidden wealth ↑ inequality a lot
0%
1%
2%
3%
4%
5%
1930 1940 1950 1960 1970 1980 1990 2000 2010
Top 0.01% wealth share in Norway
Excluding hidden wealth
Including hidden wealth
The size & distribution of taxevasion in rich countries
Tax evasion on hidden wealth
0%
10%
20%
30%
40%
50%
P90
-95
P95
-99
P99
-99.
5
P99
.5-9
9.9
P99
.9-P
99.9
5
P99
.95-
P99
.99
P99
.99-
P10
0
% o
f tot
al ta
xes
owed
that
are
not
pai
d
Position in the wealth distribution
Offshore tax evasion, by wealth group
Lower-bound scenario
High scenario
Tax evasion detected in random audits
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
6.0% P
0-10
P10
-20
P20
-30
P30
-40
P40
-50
P50
-60
P60
-70
P70
-80
P80
-90
P90
-95
P95
-99
P99
-99.
5
P99
.5-1
00
% o
f tax
es o
wed
that
are
not
pai
d
Position in the wealth distribution
Taxes evaded, % of taxes owed (stratified random audits)
Macro average: 2.3%
Random audits detect a lot of errors ontax returns
0%
10%
20%
30%
40%
P0-
10
P10
-20
P20
-30
P30
-40
P40
-50
P50
-60
P60
-70
P70
-80
P80
-90
P90
-95
P95
-99
P99
-99.
5
P99
.5-1
00
Position in the wealth distribution
Fraction of households evading taxes, by wealth group (stratified random audits)
But random audits fail to capturesophisticated evasion at the top
0%
5%
10%
15%
20%
25%
30% P
0-10
P10
-20
P20
-30
P30
-40
P40
-50
P50
-60
P60
-70
P70
-80
P80
-90
P90
-95
P95
-99
P99
-99.
5
P99
.5-1
00
% o
f tot
al in
com
e (r
epor
ted
+ ev
aded
)
Position in the wealth distribution
Fraction of income undeclared, conditional on evading (stratified random audits)
Combining random audits and leaks
0%
5%
10%
15%
20%
25%
30% P
0-10
P10
-20
P20
-30
P30
-40
P40
-50
P50
-60
P60
-70
P70
-80
P80
-90
P90
-95
P95
-99
P99
-99.
5
P99
.5-9
9.9
P99
.9-P
99.9
5
P99
.95-
P99
.99
P99
.99-
P10
0
% o
f tax
es o
wed
that
are
not
pai
d
Position in the wealth distribution
Taxes evaded, % of taxes owed
Offshore evasion (leaks)
Tax evasion other than offshore (random audits)
Tax evasion makes the tax systemregressive at the top
25%
30%
35%
40%
45%
50% P
0-10
P10
-20
P20
-30
P30
-40
P40
-50
P50
-60
P60
-70
P70
-80
P80
-90
P90
-95
P95
-99
�P99
-99.
5
P
99.5
-99.
9
P
99.9
-P99
.95
P
99.9
5-P
99.9
9
P
99.9
9-P
100
% o
f tax
able
inco
me
Position in the wealth distribution
Taxes paid vs. taxes owed
Taxes paid
Taxes owed
The interplay between
evasion and avoidance
Substitution between evasion andavoidance
Can gov. increase tax collection on the wealthy byfighting tax evasion?
Depends on substitution between evasion and avoidance
We study substitution using sample of Norwegians whouse tax amnesty
They used to hide a lot of wealth
Decide to come clean
Do they start avoiding more?
Number of amnesty participants by year0
100
200
300
400
num
ber o
f dis
clos
ers
2006 2008 2010 2012 2014 2016year of first contact
Number of Norwegian households, excluding cases droppedFigure G.6: Number of amnesty participants, by year
Reported wealth increases by 60%post-amnesty
-.20
.2.4
.6le
vel r
elat
ive
to e
vent
yea
r
-6 -4 -2 0 2 4event time
Reported taxable income increases by20%
-.10
.1.2
.3le
vel r
elat
ive
to e
vent
yea
r
-6 -4 -2 0 2 4event time
Taxes paid rise in line with income &wealth: no sign of increased avoidance
-.10
.1.2
.3.4
leve
l rel
ativ
e to
eve
nt y
ear
-6 -4 -2 0 2 4event time
Conclusion
Main results
In rich economies with low self-employment, taxevasion is small on aggregate
But high at the top, strong gradient within top 1%
This can be explained by model where suppliers of taxevasion services internalize the costs of being caught
Model and evidence suggest collecting more revenuefrom the wealthy may be possible
Next steps
HSBC, Panama leaks and amnesty data available tomany tax authorities
Our method could be applied broadly to constructdistributional tax gaps in many countries
Ultimate goal is to correct global inequality statisticsin a systematic way
→ Tax evasion to be included in future DistributionalNational Accounts & WID.world inequality series
How offshore wealth affects inequality
0%
2%
4%
6%
8%
10%
12%
Spain UK Scandinavia France USA Russia
% o
f tot
al h
ouse
hold
wea
lth
The top 0.01% wealth share and its composition
Offshore wealth
All wealth excluding offshore
Supplementary Slides
Offshore wealth at HSBC, in all Swissbanks, and globally
World Scandinavia Sweden Norway Denmark
A. Wealth held offshore ($ billion)
At HSBC Switzerland Private Bank 118.4 1.01 0.49 0.32 0.20
In all Swiss banks 2,670 21.5 12.8 4.2 4.4
In all the world's tax havens (benchmark estimate) 5,620 51.0 28.4 14.1 8.4
- Bottom-up estimate 5,620 48.1 23.3 15.4 9.5
- Proportional allocation 5,620 108.8 49.0 24.0 35.9
B. Wealth held offshore (% of household wealth)
In all Swiss banks 1.5% 0.7% 0.9% 0.6% 0.4%
In all the world's tax havens (benchmark estimate) 3.3% 1.6% 1.9% 1.9% 0.8%
- Bottom-up estimate 3.3% 1.5% 1.6% 2.1% 0.9%
- Proportional allocation 3.3% 3.3% 3.3% 3.3% 3.3%
D. MemoHousehold wealth (2006-07 average) ($ billion) 172,356 3,229 1,453 711 1,064HSBC Switzerland as % of total offshore wealth 2.1% 2.0% 1.7% 2.3% 2.4%Total ICIJ (double-counting accounts assigned to multiple countries) 249,085 2,195 1,000 458 737Total ICIJ excluding tax havens 206,265 2,195 1,000 458 737Total wealth, ICIJ totals scaled to HSBC grand total 118,400 1,260 574 263 423Share of HSBC wealth (our estimate) 100% 0.86% 0.41% 0.27% 0.17%Share of HSBC wealth (ICIJ scaled down to HSBC) 100% 1.06% 0.48% 0.22% 0.36%Share of Swiss offshore wealth 100% 0.81% 0.48% 0.16% 0.17%
Sources: HSBC (2015), ICIJ, Zucman (2013), AJZ (2017), and this paper's Appendix I.
Notes: Total wealth held at HSBC Switzerland for Scandinavia, Sweden, Norway, and Denmark only include the accounts that could be matched to an individual taxable in Scandinavia; it excludes all unmatched accounts, non-resident account holders, and remove the double-counting of joint accounts.
HSBC was not the “go-to” place forScandinavians to hide their wealth
UAEArgentBelgiu
Brazil
Canada
German
EgyptSpain
UK
GreeceIndia
Israel
Italy
MexicoRussia
Saudi
Turkey
USA
Venezu
DenmarNorway
Sweden
0.0
2.0
4.0
6.0
8.1
Shar
e of
HSB
C w
ealth
0 .02 .04 .06 .08 .1Share of wealth in all Swiss banks
Data source: ICIJ and SNB.Note: In the full sample excluding tax havens (134 countries), a regression of the share of HSBCwealth on the share of Swiss deposits has slope b= 0.90 (se = 0.04) and R-square of 0.75.
HSBC wealth vs. wealth in all Swiss banks
HSBC evasion without re-ranking
0.0%
0.2%
0.4%
0.6%
0.8%
1.0% P
90-9
5
P95
-99
P99
-99.
5
P99
.5-9
9.9
P99
.9-P
99.9
5
P99
.95-
P99
.99
P99
.99-
P10
0
Net wealth group
Figure S.4: Probability to own an unreported HSBC account, by wealth group (All matched accounts, including vs. excl. account value)
HSBC wealth added to wealth
HSBC wealth excluded from wealth
Standard errors
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]
Wealth group % of all households
Test% of evaders'
wealthTest
% of all households
Test% of all
householdsTest
% of evaders' wealth
Test% of all
householdsTest
P0-90 0.00 35.08 A 0.00 0.03 36.52 C 0.03(0.00) (9.21) (0.00) (0.00) (1.86) (0.00)
P90-95 0.01 38.27 A 0.01 A 0.25 25.32 A 0.26(0.00) (4.45) (0.00) (0.01) (2.06) (0.01)
P95-99 0.03 39.34 A 0.01 A 0.78 27.42 AB 0.80(0.00) (3.51) (0.00) (0.02) (1.26) (0.02)
P99-99.5 0.07 42.32 A 0.04 B 2.83 31.02 B 2.89(0.01) (5.91) (0.01) (0.09) (1.95) (0.09)
P99.5-99.9 0.19 46.51 A 0.04 B 4.31 30.89 B 4.49(0.02) (3.77) (0.01) (0.12) (1.52) (0.12)
P99.9-99.95 0.38 A 36.19 A 0.16 B 8.16 31.26 ABC 8.51(0.08) (5.85) (0.06) (0.45) (2.79) (0.45)
P99.95-99.99 0.66 A 36.63 A 0.17 B 11.49 A 32.84 BC 11.76(0.12) (9.24) (0.07) (0.58) (2.92) (0.59)
P99.99-100 0.94 A 38.60 A 1.19 13.77 A 26.30 AB 14.83(0.30) (9.34) (0.39) (1.25) (4.51) (1.29)
Number of householdsNumber of tax evaders 8,233
7,547,1701,375
7,547,1708,571520
10,617,167300
7,547,170165
Intensive margin Extensive margin
HSBC + AmnestyAmnesty
10,617,167 7,547,170
HSBC Panama Papers
Intensive margin Extensive margin Extensive marginExtensive margin
Scandinavian macro aggregates andwealth distribution
Scandinavia Sweden Norway Denmark
Macroeonomic aggregates
Adult population (thousands) 14,711 7,179 3,434 4,097
National income per adult (US$) 60,977 49,949 87,119 58,387
Household wealth per adult (US$) 201,658 184,225 189,456 242,431
Household wealth / national income 331% 369% 217% 415%
Wealth shares (excluding offshore)
Bottom 50% 2.9% 5.8% -2.6% 2.7%
Middle 40% 43.8% 41.3% 52.8% 41.7%
Top 10% 53.3% 52.9% 49.9% 55.6%
Top 1% 21.8% 22.1% 17.9% 22.8%
Top 0.1% 10.6% 11.0% 8.9% 10.4%
Top 0.01% 5.3% 5.7% 4.6% 4.5%
Estimates of global offshore wealth
0%
2%
4%
6%
8%
10%
12%
14%
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
% o
f wor
ld G
DP
The global amount of household wealth in tax havens
Global offshore wealth (Our estimate)
Global offshore wealth (BCG)
Offshore wealth in Switzerland (Swiss National Bank)
Tax evasion in random audits:US. vs. Denmark
0%
5%
10%
15%
20%
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
3.5%
4.0%
P0-
10
P10
-20
P20
-30
P30
-40
P40
-50
P50
-60
P60
-70
P70
-80
P80
-90
P90
-95
P95
-99
P99
-99.
5
P99
.5-1
00
Position in the income (US) or wealth (Denmark) distribution
Figure S.23: Fraction of income undeclared (stratified random audits)
Denmark (left) average: 1.8%
US (right) average: 11%
Why is detected evasion higher in US?DCE multiplier + self-employment
0%
5%
10%
15%
20%
25%
30%
Nor
way
Ic
elan
d S
wed
en
Japa
n Lu
xem
bour
g D
enm
ark
Finl
and
Fran
ce
Irela
nd
Lith
uani
a B
elgi
um
Est
onia
U
nite
d K
ingd
om
Aus
tria
Net
herla
nds
Ger
man
y S
pain
H
unga
ry
Uni
ted
Sta
tes
Slo
veni
a C
zech
Rep
ublic
S
witz
erla
nd
Latv
ia
Por
tuga
l Ita
ly
Slo
vak
Rep
ublic
Tu
rkey
P
olan
d G
reec
e
The share of self-employment income in GDP in OECD countries (Gross mixed income as a % of factor-cost GDP)
Why does evasion seem to rise sharplywithin the top groups?
A model of the supply of evasion services
Population of mass one with wealth density f (y)
Monopolistic bank sells tax evasion services(historically, Swiss banks have operated as a cartel)
Charges θ per dollar of wealth hidden
Simplification: infinitely elastic demand at price θ →bank optimizes on the number of clients it serves
Manages k(s) in wealth when serves s = 1− F (y)and earns θk(s) in revenue
The bank’s problem
Bank has probability λs to be caught → fine φk(s)
Risk-neutral bank maximizes profits
π(s) = θk(s)− λsφk(s)
At interior optimum:
θ =
(1
εk(s)+ 1
)φλs
Where εk(s) = sk ′(s)/k(s) is elasticity of the amountof hidden wealth managed with respect to s
The Pareto case
If wealth Pareto-distributed, supply of evasion services is:
s =θ
(1 + b)λφ
b is the inverted Pareto-Lorenz coefficient (high b →high inequality)
Higher λ or higher φ → fewer & richer clients
If high inequality, bank will serve tiny fraction of the pop.
Stronger enforcement → fewer, wealthierclients
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
2006 2007 2008 2009 2010 2011 2012 2013 2014
Num
ber o
f clie
nts
Ave
rage
acc
ount
val
ue (m
illio
n U
S$)
Number of clients and average account value at HSBC Private Bank Switzerland
Number of clients (right scale)
Average account value (left scale)
How we estimate the impact of using theamnesty on tax avoidance
Event-study model
log(yit) = αi + γt + X ′itψ +∑
βkDkit + uit
yit : reported taxable wealth, income, taxes paid
αi : household fixed effects
γt : time fixed effects
Dkit : event-time dummies
X ′it : Controls: 10 bins of 2007 wealth × year, 10 binsof 2007 income × year, 6 bins of 2007 age × year
Amnesty participants:summary statistics pre-disclosure
Not amnesty participants
Amnesty participants
Number of individuals 3,807,898 1,307DEMOGRAPHICS
Age 46 58Male 50% 63%Number of children 2.3 2.3Foreign born or foreign national 13% 21%Married 41% 57%
INCOME AND WEALTH ($)Reported taxable wealth (tax value) 20,641 2,466,276True taxable wealth (tax value) 20,641 4,454,507Reported taxable income 55,717 211,407Reported taxable capital income 3,265 103,096
TAX AVOIDANCE INDICATORSMaximized dividend payments in 2005 0.7% 7.0%Owns a holding company 0.6% 9.3%Reports no taxable income 3.4% 1.1%Reports no taxable wealth 2.1% 0.2%Reports no capital income 44.4% 8.6%Reports no wage income 23.8% 29.5%Pays zero taxes 11.2% 2.4%80% wealth tax reduction 0.3% 7.3%Owns unlisted shares 3.9% 29.8%
All Norwegian residents (2007)
Pre-disclosure, amnesty participants avoidless taxes than similarly wealthy taxpayers
(1) (2) (3) (4)
Truetaxable
wealth
Maximized
dividend
paymentsin
2005
Ownsa
holding
company
80%wealth
taxreduction
Amnestyparticipant 0.0049 -0.0275*** -0.0433*** -0.0157***
(0.0064) (0.0038) (0.0035) (0.0023)
Observations 524,647 724,176 724,176 724,176
R-squared 0.9839 0.0595 0.1641 0.1357
Truetaxablewealth 100bins 100bins 100bins 100bins
Income 10bins 10bins 10bins 10bins
age 6bins 6bins 6bins 6bins
Standarderrorsinparentheses
***p<0.01,**p<0.05,*p<0.1
Summary of impact of disclosingpreviously hidden assets
(1) (2) (3)
Reported wealth
(in logs)
Reported income (in logs)
Taxes paid (in logs)
Post-disclosure (periods 0-2) 0.4571*** 0.1817*** 0.2296***relative to pre-disclosure (period -4 to -2) (0.0403) (0.0333) (0.0311)
Observations 5,820,893 7,956,464 7,771,735R-squared 0.8499 0.7255 0.8000Individual FE, wealth x year FE, income x year FE, age x year FE X X X
Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1
Compliance
No sign of rise in most obvious avoidancechannels
(4) (5) (6) (7) (8)
Founds holding
company (dummy)
Unlisted shares
(in logs)
Housing wealth
(in logs)
Zero capital income
(dummy)
Emigration (dummy)
Post-disclosure (periods 0-2) -0.0006 -0.1141 -0.0736 0.0110 -0.0001***relative to pre-disclosure (period -4 to -2) (0.0018) (0.1048) (0.0528) (0.0074) (0.0000)
Observations 8,176,582 900,957 6,142,102 8,176,582 8,176,582R-squared 0.0944 0.8617 0.7446 0.6063 0.2515Individual FE, wealth x year FE, income x year FE, age x year FE X X X X X
Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1
Channels of avoidance