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