Transcript Slide 1

UNDERSTANDING INDIVIDUAL TAX
COMPLIANCE
Gareth D. Myles
University of Exeter and Tax Administration Research Centre
In collaboration with
Miguel Fonseca
Shaun Grimshaw
Nigar Hashimzade
Tim Miller
Matthew Rablen
Exeter and TARC
Exeter and TARC
Durham and TARC
Exeter and TARC
Brunel and TARC
The financial support of ESRC/HMRC/HMT is gratefully
acknowledged.
INTRODUCTION
 An
understanding of the individual tax
compliance decision is important for
revenue services
 It is necessary for designing good policy
interventions that reduce the tax gap
 Tax compliance is an area where orthodox
analysis has been challenged by
behavioural economics
 This talk explores the limitations of the
orthodox analysis and suggests
improvements
STARTING POINT
A
natural starting point is to consider
non-compliance as a gamble
 A non-compliant taxpayer is gambling on
not being audited and discovered
 Let the taxpayer have income Y and
declare income X, with 0 ≤ X ≤ Y
 Income when not caught is
Ync = Y – tX
 If the fine is F then income when caught is
Yc = [1 – t]Y – Ft[Y – X]
ORTHODOX ANALYSIS
 If
income is understated the probability of
being caught is p
 Applying expected utility theory implies
the optimal declaration X solves
max{X} E[U(X)] = [1 – p]U(Ync) + pU(Yc)
 There


are two states of the world:
In one state the taxpayer is not caught evading
and income is Ync
In the other state they are caught and income
is Yc
EVASION DECISION
• The choice problem is
Yc
shown in Figure 1
• The optimal declaration
achieves the highest
indifference curve
1  t Y
• The taxpayer chooses to
locate at the point with
declaration X*
• This is an interior point 1  t 1  F Y
with 0 < X* < Y
• Some tax is evaded but
some income is declared
X Y
X*
X 0
1  t Y
Y
Figure 1: Interior choice:
0 < X* < Y
Y nc
NON-COMPLIANCE
 Non-compliance
occurs when the indifference
curve is steeper than the budget constraint
at X = Y
 This is true when p < 1/[1 + F]
 If this condition is satisfied the taxpayer
should be non-compliant
 It is independent of preferences
 When F = 1 the taxpayer will evade if p < ½
 The model predicts that for realistic
parameter values every taxpayer should be
non-compliant
TAX EFFECT
• An increase in the tax
Yc
rate moves the budget
constraint inward as in
Figure 2
• The outcome is not
1  t Y
clear-cut
1 tˆY
• If taxpayers are more
X old
willing to take on a fixed 1  t 1  F Y
X new
gamble as income
increases then a tax

1  tˆ1  F Y
increase reduces tax
evasion
1 tˆY 1  t Y Y
• This is because the fine
is Ft so an increase in
Figure 2: Tax rate increase
the t raises the penalty
Y nc
TESTING THE RESULTS
 The
model could be tested by comparing
its predictions against data
 The publicly available data is very limited
and has not been adequate to test the
model
 An alternative strategy has been to use
experiments to test the model
 How does the behaviour of experimental
subjects compare to the predictions?
EXPERIMENTS
 Most
experiments have been run in
experimental labs using students as
subjects
 TARC has gone beyond this by using
online experiments with large numbers of
actual taxpayers
 The results of the experiments are not
supportive of the orthodox analysis
 The experiment in which you participated
will illustrate this
WINNER
 The
lowest payoff was 122700
 The highest payoff in the experiment was
215000
 The winner of the prize is:
Mark Phillips
University of Southern
California
STRUCTURE
 You
were enrolled randomly in one of two
experiments
 In one experiment Part A involved tax
compliance
 In the other experiment Part A involved
an investment decision
 For both experiments Part B tested
attitude to risk
 The tax experiment will be discussed first
COMPLIANCE EXPERIMENT
 What
does the model predict about
behaviour?
 For all sets of parameter it was the case
that p < 1/(1 + F)
 So the model predicts every participant
should have been non-compliant
 Non-compliance might vary between
participants
 But the optimal strategy to maximise
expected income is to declare nothing
COMPLIANCE EXPERIMENT
 The
data do not match these predictions
 10 participants out of 50 declared honestly
 Only 4 declared nothing every time
(including me!)
 Some participants were partially noncompliant
 The choices are summarised in the
histograms that follow
COMPLIANCE EXPERIMENT
12
Tax - % undeclared
10
8
6
4
2
0
10
20
30
40
50
60
70
80
90
100
COMPLIANCE EXPERIMENT
Tax - Payoff
12
10
8
6
4
2
0
130000
140000
150000
160000
170000
180000
190000
200000
210000
222000
INVESTMENT EXPERIMENT
 The
investment experiment involved the
allocation of saving
 There was a risky asset and a safe asset
 The payoffs were structured so that the
risky asset was a better-than-fair bet
 The optimal strategy to maximise
expected income is to put everything into
the risky asset
 The histograms summarise the responses
INVESTMENT EXPERIMENT
Investment - % risky
12
10
8
6
4
2
0
10
20
30
40
50
60
70
80
90
100
INVESTMENT EXPERIMENT
Investment -Payoff
12
10
8
6
4
2
0
130000
140000
150000
160000
170000
180000
190000
200000
210000
222000
COMBINATION
 Why
did we run two versions of Part A?
 The compliance experiment and the
investment decision had the same payoffs
 If tax compliance were just a gamble then
the experiments should have the same
choices
 This was the reason for randomising
participants and experiments
 The comparison of histograms shows the
pattern of choices are very different
COMBINATION
12
12
Tax - % undeclared
10
10
8
8
6
6
4
Investment - % risky
4
2
2
0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
50
60
70
80
90
100
OBSERVATIONS
 These
results are not explained by
attitudes to risk
Tax- Lottery Switch Point
Investment - Lottery Switch Point
10
10
9
9
8
8
7
7
6
6
5
5
4
4
3
3
2
2
1
1
0
0
0
1
2
3
4
5
6
7
8
9
10
11
0
1
2
3
4
5
6
7
8
9
10
11
OBSERVATIONS
 This
experiment was first reported by
Baldry in 1986
 It always works!
 He concluded that tax compliance was not
just a gamble
 The comparison shows that the orthodox
analysis is not correct
 Recent research has explored how it
should be revised
 Some of this research is now reviewed
OPPORTUNITIES
 Not
all taxpayers have an opportunity to
be non-compliant
 Employment income is usually subject to
third-party reporting or withholding
 Self-employment opens the opportunity for
non-compliance
 Occupational choice should be modelled
 The potentially non-compliant self-select
into occupations where non-compliance is
possible
OCCUPATIONAL CHOICE
 Self-employment
can be successful (S) or
unsuccessful (U)
 For optimal evasion, Ei*, the payoff from
self-employment is
EU = (1–q) EUu (Eu*) + qEUs (Es*)
 The
choice of occupation is determined
(partly) by risk aversion
 Low risk aversion implies selfemployment and significant noncompliance
BEHAVIOURAL APPROACH
 The
next issue is why be honest if it does
not pay?
 The problem that confronts modelling is
how to maintain rationality but reach
different conclusions
 This issue has had to be addressed in
many areas of economics
 “Anomalies” are observed decisions that
do not fit theoretical predictions
 These have lead to the development of
behavioural economics
BEHAVIOURAL APPROACH
 Behavioural
economics can be seen as a
loosening of modelling restrictions
 Two different directions can be taken:
(i) Revise the assumption about
information underlying the decision
(ii) Reconsider the private nature of the
compliance decision
 This allows additional factors to be
incorporated in the evasion decision
INFORMATION
 In
the orthodox model the taxpayers use
the objective probability of audit and know
the fine
 Two criticisms
The probability is not public information
2. The fine is not widely known
1.
 There
is evidence that subjective beliefs
about unknown variables inflate the
probability of bad events
NON-EXPECTED UTILITY
w1(p, 1 – p) and w2(p, 1 – p) be weighting
functions that depend on p and 1 – p
 More weight is given to the bad outcome
so w1(p, 1 – p) > p
 The general form of non-expected utility is
V = w1(p, 1 – p)U(Yc) + w2(p, 1 – p)U(Ync)
 The inflation of the probability will raise
the rate of compliance
 Let
ALTERNATIVES
 Some
of the alternatives that have been
applied to the compliance decision are:
Rank Dependent Expected Utility imposes
structure on the translation of probabilities
 Prospect Theory translates probabilities,
changes payoff functions, and uses a reference
point
 Non-Additive Probabilities do not require the
normal consistency of aggregation for
probabilities
 Ambiguity focuses on uncertainty over the
probability of outcomes

SOCIAL CUSTOMS
 Attitudes
to compliance also matter
 Some taxpayers will always be fully
compliant
 This can be explained by a social custom
(an informal rule on behaviour)
 If the social custom is broken there is an
additional loss of utility

S
U if followed, U – S if broken
can also be interpreted as a
psychological cost of non-compliance
SOCIAL CUSTOMS
S = mciEi where m is the proportion of
population who are compliant
 Choose either to be compliant with payoff
UNE = U(Y[1 – t])
 Or to be non-compliant with payoff
UE = E[U] – mciEi
 Let
with high ci (individual concern
about custom) will be compliant
 People
Non-Compliant
0
Compliant
c
SOCIAL INTERACTION
 How
can we explain the formation of
attitudes and beliefs?
 Both can be the outcome of social
interaction
 This can be modelled using a social
network that governs the interaction
between individuals
 Individuals meet with their contacts in
the network and exchange information
 Information affects compliance
SOCIAL NETWORK
A
network is a
symmetric matrix A
of 0s and 1s (bidirectional links)
 The network shown
is described by
0
1
A
0

0
1
0
1
0
0
1
0
1
0
0
1

0
1
2
3
4
SOCIAL NETWORK
 Social
networks can be studied using
agent-based models
 We have done this to look at audit rules
and predictive analytics
 Information transmission can sustain a
subjective probability above the objective
probability
 Attitudes can differ among occupational
groups
 Compliance can be increased by fostering
attitudes
CONCLUSIONS
 The
talk was titled “Understanding
individual tax compliance”
 When viewed as an individual decision
the orthodox model makes incorrect
predictions
 More accurate predictions can be made by
understanding compliance as a social
decision
 We need to take into account attitudes,
beliefs, and opportunities
CONCLUSIONS
 Occupational
choice links with risk
aversion to self-select those willing to be
non-compliant into a position where noncompliance is possible
 The process of social interaction is central
to the formation of attitudes and beliefs
 A stronger social custom can give higher
compliance
 Unknown audit rules force the formation
of a subjective probability