Tax Compliance and Advance Tax

Payments: A Prospect
Theory Analysis

Abstract - While obligatory advance tax payments do not inter-

fere with the taxpayer's evasion decision under expected utility
theory, they do affect the decision to evade under prospect theory.
The present paper applies prospect theory to a simple model of tax
evasion, exploring the role that advance tax payments may play in
enforcing tax laws. The paper demonstrates, as empirically found
in the United States, that advance tax payments may substitute for
costly detection efforts in enhancing compliance. However, con-
trary to a recent claim in the tax evasion literature, deliberate high
advance tax payments are unlikely to eliminate the incentives for
noncompliance.

INTRODUCTION

E mpirical and experimental evidence suggests that ad-

vance tax payments may play an effective role in tax au-
thorities' enforcement strategy. Cox and Piumley (CP, 1988)
found that voluntary compliance rates in the United States
increased consistently with the amount of refund that tax-
payers expected to receive upon the filing of a tax retum and
decreased consistently with the amount of taxes that they still
had to pay.' Chang et al. (1987) and Robben et al. (1990) con-
firmed experimentally the empirical findings for the United
States and several European countries, whereas CarroU (1992),
documenting diaries of taxpayers' tax-related thoughts and

' Voluntary compliance has been estimated using data from the Taxpayer
Compliance Measurement Program (TCMP) of the Internal Revenue Ser-
vice for tax year 1982, which consisted of a random selection of 50,000 tax
retums subjected to a thorough inspection solely for research purposes.
The measure of voluntary compliance used is the Voluntary Compliance
Rate (VCR), defined as the percentage of total tax liability (as established
by a TCMP examination) that is voluntarily reported. With few exceptions,
the VCR for all types of income has been found to be the lowest among
Gideon Yaniv those who owed the most with their retums (71.6 and 90.7 percent among
National Insurance individuals whose corrected balance due exceeded $1,000 and whose pri-
mary source of income was business and wages, respectively), the highest
Institute, Jerusalem
among those who were due the largest refund (95.2 and 96.6 percent among
95437, Israel individuals whose corrected refund exceeded $1,000 and whose primary
source of income was business and wages,respectively),and to gradually
increase as the balance due decreases or the refund increases. A detailed
National Tax Journal description of CP's (1988) study may also be found in Chang and Schultz
Vol. Ul, No. 4 (1990) and Robben (1991).

753
 NATIONAL TAX JOURNAL

behaviors, found that taxpayers thought ability of nonrisky gains to induce hon-
primarily in terms of the out-of-pocket esty is not unrelated to the risk involved
gains and losses at the time offiling, con- in gaining more through behaving dishon-
cluding that whether taxpayers expect to estly. Why should a skillful evader, who
receive a refund or have to supplement is aware of the low probability of getting
their prepaid taxes is important for the caught and punished for tax evasion
understanding and control of taxpaying (which is less than one percent in the
behavior. Unites States), switch to safety when sub-
In a recent contribution, Elffers and jected to an advance payment that is
Hessing (EH, 1997) suggest, following slightly (or even significantly) higher than
Robben (1991) and Webley et al. (1991), her true tax liability? Just because she is
that prospect theory, developed by manipulated to play in the risk-aversion
Kahneman and Tversky (1979), may help domain? Intuitively, it may be possible
explain taxpayers' observed behavior un- that prospect theory supports CP's find-
der obligatory advance tax payments. ing that compliance increases as the ad-
Elffers and Hessing argue that when pre- vance payment is raised above the true tax
paid taxes are greater than the true tax li- liability and decreases as the advance pay-
ability, the taxpayer expects a gain from ment is lowered below the true tax liabil-
filing a return, whereas if prepaid taxes ity, but to find out whether this is so, as
are less than the true tax liability, they ex- well as whether advance tax payments
pect a loss. Hence, in the spirit of pros- may effectively eliminate the incentive for
pect theory, the taxpayer is risk averse noncompliance, requires a more rigorous
with respect to the former case and risk application of prospect theory to the tax
seeking with respect to the latter. Conse- evasion problem than the qualitative con-
quently, he will opt to avoid risk in the siderations pointed out by EH.
former case and to take his chances in the The present paper applies prospect
latter. Elffers and Hessing conclude that theory to a simple model of tax evasion
the incentives for noncompliance can be
with the purpose of inquiring into the
eliminated if the tax authorities deliber-
taxpayer's decision of whether and to
ately set the advance pa)m\ents slightly
what extent to underreport their taxable
above taxpayers' true tax liability so as to
income if obliged to pay a tax advance
ensure a gain from filing a retum.
prior to the filing of a tax return.^ While
Elffers and Hessing's conclusion might, the tax advcince is in part determined by
however, be a bit hasty; after all, risk aver- the taxpayer's tax rate and estimated tax-
sion does not imply that the individual able income and in part by the taxpayer's
will always avoid risk, the same as risk choice (as some tax systems allow taxpay-
seeking does not imply that she will al- ers a certain degree of flexibUity as regards
ways pursue risk. More specifically, the their prepaid amount), it may also include

The rationale underlying thisrequirementis twofold: first, taxes become due when incomes are eamed, rather
than when tax retums are filed; second, people with low and middle incomes tend to use their income as it
becomes available and may find it difficult, if their income falls, to pay a tax debt accrued in the previous year
(Pechman, 1983). Analytical treatment of advance tax payments in the tax evasion literature has so far been
restricted to wage eamers whose taxes are withheld at source by the employer. Under approximate withhold-
ing with obligatory end-of-year filing (the American system), wage eamers may evade taxes simply by failing
to file a retum (Yaniv, 1988). Under exact withholding without obligatory filing of an end-of-year tax retum
(the British system), wage eamers may still evade taxes by splitting their work efforts between two (or more)
jobs, failing to inform their employers that they are also employed elsewhere, thus enjoying lower tax brack-
ets (Yar\iv, 1998). The withholding system might also generate incentives for the employer to evade his employ-
ees' taxes by remitting to the tax collector less than the amount withheld (Yaniv 1988,1995) or by collaborat-
ing with his employees in withholding less than required (Yaniv, 1992).

754
Tax Compliance and Advanced Tax Payments

an exogenous component, permitted to However, only if taxpayers substantially

the tax authorities by a variety of policy overweight the low probability of detec-
options. Allowing for such a component, tion prevailing in most countries may
the paper subsequently inquires into the EH's conclusion hold as well; otherwise,
implied trade-offs between advance pay- if taxpayers hold realistic assessments of
ments and conventional enforcement in the actual probability level, reasonably
enhancing compliance. The paper, how- high advance payments, even if ensuring
ever, abstracts from taxpayers' own dis- a refund upon filing a retum, are unlikely
cretion in the determination of the prepaid to eliminate the incentives for noncompli-
amount, because this implies that taxpay- ance. The paper subsequently examines
ers may have control over whether they the relationship between tax evasion and
owe taxes or have taxes due at the time of the income tax rate under obligatory ad-
filing, rather than view their tax balance vance payments, revealing that had the
at that time as exogenously given. Hence, tax advance been independent of the in-
the application of prospect theory to ana- come tax rate, an increase in the latter
lyzing the relationships between tax com- would unambiguously decrease compli-
pliance and advance payments, which ance. Because the tax advance increases
rests on the assumption that potential tax with the tax rate, it acts oppositely to en-
evaders think primarily in terms of out- courage compliance. Still, when compli-
of-pocket gains and losses at the time of ance is sufficiently high, the former effect
filing, may not be appropriate. is shown to dominate, in sharp contrast
The paper begins with constructing the with the literature's result that compliance
formal model, showing that expected util- increases with the income tax rate.
ity theory, which has frequently been ap-
plied to the tax evasion problem, yields THE MODEL
no relationship whatsoever between ad-
vance tax payments and the taxpayer's Consider a taxpayer whose actual in-
evasion decision, whereas if the come less deductible expenses during a
taxpayer's behavior conforms with pros- given tax year, W, is subject to a constant
pect theory assumptions, the entry con- tax rate, ft Suppose that the taxpayer is
dition into tax evasion, as well as the ex- required by law to declare her income and
tent of evasion, becomes dependent on the deductible expenses to the tax agency by
size of the advance payment.' The paper filing a tax retum at the end of the tax
proceeds to investigate the relationships year.* Suppose also that the tax agency
between tax evasion and advance tax pay- requires the taxpayer to pay a tax advance
ments, demonstrating, in support of CP's of a given size, D, prior to filing a retum,
empirical finding, that obligatory advance to be offset later against the taxes due on
pajmients may indeed substitute for costly her taxable income.^ The tax advance is
detection efforts in enhancing compliance. assumed to consist of two components: an

Prospect theory has failed so far to attract the attention of economists as a possible tool of analyzing tax
evasion, an exception being Aim and Beck (1990), who applied it to the analysis of tax amnesties.
The tax law may actually allow taxpayers certain margins of time for filing their tax retums (in the United
States, for example, retums must be filed not earlier than January 31 and not later than April 15 of the follow-
ing tax year). This gives rise to the question of when to file one's retum (i.e., whether to file promptly at the
earliest possible day or to delay filing), a decision modeled by Slemrod et al. (1997), which we ignore here.
Practically, wage and salary eamers have their tax advances deducted at source by their employers in the
form of tax withhoidings. Self-employed and recipients of property income usually pay their tax advances in
monthly or quarterly installments during the year in which their income is received. The tax advance, D,
defined above, may be viewed as the sum of all tax prepayments, abstracting, within the single-period frame-
work, from timing considerations.

755
 NATIONAL TAX JOURNAL

endogenous one, which is a positive func- respectively. The taxpayer is assumed to

tion of estimated tax liability, S; and an choose a level of declaration, X", so as to
exogenous one determined by the tax maximize a given target function. If ex-
agency_as part of its tax enforcement pected utiUty, EU{I) = (1 -p)U(I*) + pU{I-),
policy, D. Hence, is assumed to be her target function,
the tax advance, D, is canceled out in
[1] D = a(S) + D both equations 2 and 3, and the model
collapses to the well-known AUingham
where a' (S) > 0. Estimated tax liability will and Sandmo (AS, 1972) model, with
be assumed to equal 6b, where b is some the penalty function suggested by
income base serving as an estimate of the Yitzhaki (1974), implying that the impo-
taxpayer's actual income (e.g., last year's sition of an obligatory advance payment
income or average income in the would play no role in the taxpayer's eva-
taxpayer's profession). sion decision.
Suppose now that the taxpayer consid- Consider, however, prospect theory,
ers the possibility of declaring to the tax which replaces the utility function by a
agency less than her true taxable income, "value function," v(»). The value attached
X (SW), through either underreporting by the taxpayer to each possible outcome
true income or overreporting true ex- of the risky decision is assumed to depend
penses. In this case, her evaded taxes, 6(W on the change in net income from some
- X), will be taxed, if detected, at a pen- reference point, rather than on the level
alty rate X (>1). The probability of being of net income itself. Most importantly, the
detected evading taxes is assumed to be value function is assumed to be concave
independent of the taxpayer's activity, set for gains but convex for losses, so that the
by the tax agency at the level of p (<1). taxpayer is risk averse with regard to the
While taxpayers in general may have sub- former but risk seeking with regard to
jective perceptions of the probability of the latter.' Applying these assumptions
detection, which differ from its tme value, to the evasion decision at hand first re-
we begin by assuming that the potential quires determination of the reference
evader is fully informed of the actual point from which changes in net income
probability level. are measured. Elffers and Hessing suggest
Underreporting of actual income gives that the reference point should be income
rise to two possible levels of final net in- after the payment of the tax advance and
come: I*, if the taxpayer's evasion is not prior to the filing of a retum, W- D7 The
detected; and I', if her evasion is detected. changes in the taxpayer's net income
These are given by therefore are

[2] 7- = W - D + (D-0X) [2'] M* = D-eX

and and
[3] [D-0X-A6»(W-X)] [3'] = D-eX-Xe(W-X)
' Also, prospect theory replaces the probability of the risky occurrence, p, with a "weighting function," 0(p),
that depends positively on p but that overweights low probabilities and underweights high ones. As noted
above, we abstract first from the possible discrepancy between the actual and the perceived probability of
detection, assuming therefore that 0(p) = p.
' While this is the most plausiblereferencepoint for examining the effect on behavior of out-of-pocket gains or
losses at the time of filing, it also seems to be the only possiblereferencepoint that maintains D as a relevant
parameter in the taxpayer's problem. For other possible reference points, such as income before taxes, W, or
income after taxes due, (1 - 6) W, D will be canceled out in both equations 2' and 3'.

756
Tax Compliance and Advanced Tax Payments

in case of nondetection and detection, re- TAX COMPLIANCE AND THE

spectively. Notice that equations 2' and 3' ADVANCE PAYMENT
imply that the evading taxpayer antici-
pates a certain refund ofD-OX (or, given Suppose first that the advance tax pay-
that D < OX, a certain supplementary tax ment, D, is set above the taxpayer's true
payment of OX - D), irrespective of tax liability, 6\N, so that the taxpayer ex-
whether they are caught, and a probabi- pects a gain from reporting honestly. Nev-
listic penalty of Xe(W-X), due to the pos- ertheless, she might opt to report dishon-
sibility of detection. estly (i.e., choose X* < W) if it serves to
increase the value of her prospect. A suf-
The taxpayer now chooses X" so as to ficient condition for doing so is that
maximize the value, V, of their prospect dV/dX < 0 at X = W, or that
[4] + pv[-xe(w - x)]
v'(D -
[7]
where, contrary to expected utility theory,
the certain refund (or supplementary pay-
ment) is valued as v(D - OX) and weighted where v'(0)" denotes the left-hand deriva-
by a probability of unity.^ Maximizing tive of the value funcdon at the origin,
equation 4 now withrespectto X, the first- which is assumed by prospect theory to
and second-order conditions for an inte- be steeper than the right-hand derivative,
rior optimum are v''(0)*. Contrary to Yitzhaki's (1974) model,
and to its numerous extensions, where the
entry condition into tax evasion, pX < 1,
[5] §^=0|-v'(D-0X) depends on the probability of detection
and the penalty rate alone, the entry con-
dition in the present model depends also
on the size of the tax advance, D, as well
and as on the income tax rate, 0. It is also
stricter than Yitzhaki's condition, because
v'(D - dW)/vXO)- < 1. Quite surprisingly,
[6] ^ = " = S'iv"( a deterrent to tax evasion may exist even
if detection does not involve an effective
[-A0(W-X)]1<O penalty but just the payment of the taxes
due (i.e., if A =1).'
respectively. Because both v'(D - 6X) and Consider now the implications of the
v'[-X9(W - X)] are positive, there are no entry condition for tax enforcement. Tax
apparent restrictions on satisfying equa- agencies usually face exogenously given
tion 5. However, because v"[-A0(W - X)] tax and penalty rates (stipulated in the tax
is positive (reflecting risk seeking for laws), having discretion over the extent
losses), equation 6 may only be satisfied and intensity of tax audits, which deter-
if D - 0X is positive, so that v"(D - GK) is mine the probability of detection, and,
negative (reflecting risk aversion for presumably, some discretion over the size
gains). of the advance tax payment through its

' That is, contrary to expected utility theory, V is not written as (1 - p) v(A/ *) + pv(Af"). Because D -flXis due in
both states of the world, it is grouped imder the certain value of v(D - SX), an adjustment called "editing"
by Kahneman and Tversky (1979). The taxpayer then faces a modified prospect of v(D - eX) + (1 - p)v(O) +
pv[-Afl(W - X)], the middle term of which drops out, because v(0) = 0 in prospect theory.
' This, however, is in line with the fiscal psychology literature, which suggests that taxpayers may perceive
even the payment of taxes due as a penalty (Hessing et al.,1992).

757
 NATIONAL TAX JOURNAL

exogenous component, D. The former is tax payments may substitute for costly de-
evidently resource cor\suming, whereas tection efforts in inducing full compliance,
the latter presumably entails no resource they must be set at a sufficiently high level,
costs. To allow the tax agency full flexibil- which is inversely related to the level of
ity in setting the advance payment as low detection efforts. Figure 1 (part II) dem-
as it wishes above the taxpayer's true tax onstrates that when D is relatively small,
liability, it will be assumed that a{S) < GW. approaching OW from above, v\D - OW)/
Condition 7 now implies that the incen- vXO)- approaches v'(O)V v'(0)-, which has
tive for tax evasion can be eliminated not been shown experimentally by
only through a costly increase in detec- McClelland et al. (1986) to equal approxi-
tion efforts (which increases pX), but also mately one-third. Hence, when pX is rela-
through a noncostly raise of D [which in- tively high, approaching one-third from
creases D and decreases v'(D - OW)], so below, D may be set only slightly above
as to increase the gain expected from hon- OW to induce honesty. As pX is reduced,
est declaration. However, while advance D must be increased,'" and as pX ap-

Figure 1. Evasion/No Evasion Regions

vXD-OW)

Mathematically, holding equation 7 as an equality and differentiating pX with respect to D yields d(pX)/dD =
v"(D - eW) / V '(0)- < 0, so that pA is inversely related to D along the full-compliance contour.

758
Tax Compliance and Advanced Tax Payments

proaches zero, D must be raised to infin- laration for which D - 0X < 0, such a solu-
ity to induce honesty. In view of the tion would represent a minimum rather
substantially low levels of costly enforce- than a maximum. Thus, the taxpayer
ment prevailing in most countries," and would either report their true taxable in-
given that taxpayers who opt to engage come or report no taxable income at all,
in tax evasion hold realistic assessments depending on whether "value" is greater
of the actual probability level, reasonably for X = W or X = 0, respectively. Substi-
high advance payments, even if ensuring tuting X = VV and, altematively, X = 0
a refund upon filing a retum, are unlikely into equation 4, rearranging, and setting
to eliminate the incentive for noncompii- v(0) = 0 (as prospect theory assumes), the
ance. Still, totally differentiating the taxpayer will report no taxable income if
first-order condition 5 with respect to X v[-{OW - D)] < v(D) + p\i-XOW) and will
and D reveals that as long as evasion is report all taxable income if this inequal-
practiced. ity is reversed. Linearly approximating
the changes in "value" around v(0), the
[8] dlC^ entry condition into evasion becomes
dD a
so that higher advance payments, while
not necessarily eliminating evasion, v'(0)-/ ew
would at least reduce evasion. This sup- Hence, the entry condition in this case is
ports CP's empirical finding that compli- stricter than Yitzhaki's (1974) condition as
ance increases consistently with the well and dependent on the size of D. Given
amount of refund due to taxpayers upon that the first-order condition 5 represents
filing a tax return.'^ a minimum, advance tax payments may
Suppose now that a{S) < OW and that thus substitute for costly detection efforts
the advance payment is set below (or ex- in inducing full compliance even when set
actly at) the taxpayer's truejax liability. below the true tax liability. Figure 1 (part
That is, suppose that a{S) + D = D< dW. I) demonstrates that when D = Q,pX must
Hence, the taxpayer expects a loss (or nei- be raised to 1 - [1 - v1(0)V v'(O)-][fl(S)/0W]
ther a loss nor a gain) from declaring hon- to induce honesty (which would collapse
estly. The second-order condition 6 im- to unity had D been free of an endogenous
plies that an interior solution may stiU be component). As D is increased, pX may be
obtained at a sufficiently low level of dec- lowered, and when D = OW, pX may be set
laration for which D - 0X > 0 (although as low as v'(O)*/v'(O)" to induce honesty.
D - ew < 0). In this case, result 8 would Still, because the latter expression has been
hold just the same, implying, again in sup- estimated to equal one-third, which far ex-
port of CP's findings, that compliance is ceeds the values of pX in real life, advance
greater the lower the amount of taxes still payments set below (or exactly at) taxpay-
due to the tax collector, OW - D. Given, ers' true tax liability are unlikely to elimi-
however, that the first-order condition 5 nate the incentive for evasion of fully in-
is solved at a sufficiently high level of dec- formed taxpayers.

" Aim et al. (1992) point out that in the United States less than one percent of individual income tax retums are
subject to a thorough tax audit and that the penalty on fraudulent evasion is only 75 percent of impaid taxes.
Hence, the expected penalty rate, pX, is as low as 0.0175.
" While taxpayers in the United States have some flexibility in determining their advance payments, it does not
necessarily follow that they also determine ahead whether they owe taxes or have taxes due at the time of
filing, because final income and deductibles may unexpectedly differ from those anticipated when the pre-
payment decision was made (see Martinez-Vazquez et al., 1992). Hence, taxpayers' tax balance when filing a
retum may still be viewed as exogenously given and prospect theory applied to explain their compliance
behavior.

759
 NATIONAL TAX JOURNAL

A crucial assumption of the analysis is, in the income tax rate. Totally differenti-
of course, that taxpayers' perceived prob- ating the first-order condition 5 with re-
ability of getting caught and penalized for spect to X and 6 yields (recalling that D
tax evasion coincides with the true prob- contains an endogenous component, a(S),
ability of detection. However, recent em- where S = Ob)
pirical and experimental evidence seems
to suggest, in line with prospect theory
presumptions, that individuals often [10] ^ = -^([X-fl
overweight the true probabilities of low-
probability events and, in particular, that - pA2(W-X) v"[-A0(W-X)])
taxpayers tend to overweight the low
probability of audit that they in fact face." the sign of which is negative if X > a'{S)b
Such evidence may help explain why and ambiguous otherwise. It may be rea-
people pay far more in taxes than sug- sonable to assume that a{S) = c6, so that
gested by the standard expected utility the endogenous component of the tax ad-
theory of compliance, as the latter implies vance is some fraction, 0 < a < 1, of esti-
that the real-life parameter values of the mated taxable income, S. This implies that
tax evasion gamble are not likely to deter the sign of equafion 10 will be negative
'rational' taxpayers from reporting dis- if X > ab, which is definitely satisfied if
honestly. Allowing for taxpayers' a=0, but may also be satisfied if a = 1.''
overweighting of the low probability of Hence, had the tax advance been indepen-
detection would obviously moderate the dent of the income tax rate, an increase in
pessimistic conclusion of the present dis- the latter would unambiguously decrease
cussion regarding the effectiveness of declaration. Because the tax advance in-
advance tax payments to eliminate the in- creases with the tax rate, it acts oppositely
centive for noncompliance: given that tax- to encourage declaration. Still, as long as
payers substantially overweight the actual declaration is sufficiently high, the former
probability of detection, tax advances set effect will dominate. While this result ac-
only slightly above taxpayers' true tax li- cords with common sense and intuition,
ability may indeed suffice to induce hon- it is in sharp contrast with Yitzhaki's
esty, as argued by EH (1997)," Moreover, (1974) result dominating the tax evasion
given substantial overweighting, taxpay- literature, that declared income always
ers could be induced to report honestly increases with the tax rate. Furthermore,
even if tax advances are set below their true a negative relafionship between declared
tax liability, income and the income tax rate would
arise in this case even if the penalty on
tax evasion is imposed on the concealed
TAX EVASION AND THE INCOME income, W- X, as assumed by AS (1972),
TAX RATE This is so because the first-order condi-
tion 5 would then become (assuming
Given an interior optimum, consider
now the taxpayer's response to a change

" See Aim etal, (1992),

" Formally, this implies that the probability of detection, p, should be replaced by a weighting function, ip(p),
such that (j>{p) > p. Substituting into condition 7 and setting D only slightly above 8W, the inequality sign may
be reversed if (pip) is sufficiently high so that (t>(p)X approaches one-third,
" Notice that, at an interior solution, D-9X must be positive to ensure that v"(D - 6X) is negative (otherwise,
the second-order condition 6 does not hold and equation 4 is maximized at a comer solution). For a = 1, this
requires that X<b + D/d, still allowing for X>b.

760
Tax Compliance and Advanced Tax Payments

[5'] ^ = - fect donunates (or is exactly offset by) the

latter, supporting the substitufion effect in
generating a negafiverelafionshipbetween
declared income and the income tax rate.
yielding In Yitzhaki's (1974) model, the subsfi-
tution effect disappears, because an in-
crease in the income tax rate increases the
[10'] -i[-v(D- marginal cost of tax evasion {pX9) by the
same proportion as its marginal benefit
v"(D-0X)] (&). The remaining income effect is thus
responsible for the positive relafionship
which for X > ab is negative, in contrast between declarafion and the income tax
with the ambiguous prediction character- rate identified in that model. In the present
izing AS's model. model, however, the income effect consists
How can we account for the differences of two components: the certain refund
in implications between expected utility component (as discussed above) and the
theory and prospect theory in this regard? uncertain penalty component, captured
Consider first AS's expected-utility model by the first and second terms of equation
of a risk-averse taxpayer, where an increase 10, respecfively. The latter component af-
in the income tax rate has been interpreted fects 'value' in the loss domain where the
by AS to generate opposing substitufion taxpayer is a risk seeker, and where a tax
and income effects on declarafion: on the rate increase, which increases the loss ex-
one hand, a tax rate increase makes eva- pected at a given level of evasion,
sion more profitable at the margin, induc- compensatively discourages declaration.
ing the taxpayer to reduce declaration; on Finally, considering the taxpayer's re-
the other hand, a tax rate increase makes sponse to a change in the traditional law
the taxpayer less wealthy (regardless of enforcement parameters, the penalty rate,
whether detected), which, assuming de- and the probability of detecfion, we obtain
creasing absolute risk aversion,reducesthe
tendency to take risks, thus inducing the
taxpayer to increase declaration. In the
prospect theory model, the substitution ef-
fect is still preserved,refiectedthrough the -X) v"[-
first term of equation 10'. However, as evi- and
dent from equation 5', a tax rate increase
affects just the certain component of the
value funcdon (which, contrary to expected [12] ^ = --^v'[-
dp il
utility theory, is weighted by a probability
of one). Consequently, the income effect in
respectively. While the sign of equation 12
the present model, captured by the second
is unambiguously posifive, implying, in
term of equation 10', is independent of risk-
line with the expected ufility model, that
aversion behavior. It refiects, on the one
an increase in the probability of detecfion
hand, the taxpayer's attempt to cushion the
will increase declarafion and decrease eva-
resulting increase in tax payments by sub-
sion, the sign of equafion 11 is ambiguous,
jecting less income to taxafion and, on the
in contrast with the posifive and intuitively
other hand, her greater tendency for com-
expected sign generated by the expected
pliance when being subjected to a higher
utility model. The reason for this is that the
tax advance. When X> oib, the former ef-
income effect of an increase in the penalty

761
 NATIONAL TAX JOURNAL

rate is positive in the latter model but nega- The application of prospect theory to
tive in the present model where a penalty analyzing the taxpayers' compliance be-
rate increase affects 'value' in the loss do- havior under advance payments rests on
main, acting to encourage risk taking. Re- the assumption that the taxpayer cannot
grettably, prospect theory does not always self-select into a tax-refund or a tax-due
generate a more plausible prediction than status, thus viewing their tax balance at
does the expected utility model, the time of filing as exogenously fixed. Re-
laxing this assumption suggests that eva-
sion and prepayment may be interrelated;
CONCLUDING REMARKS hence, a decision to evade may have led to
While advance tax payments play no a given tax status rather than being in-
role in the taxpayer's evasion decision vin- duced by it," Modeling the taxpayer's
der expected utility theory, they do affect joint evasion-prepayment decision neces-
the decision to evade under prospect sitates, however, a different analytical
theory. The present paper has applied pros- framework, as it implies that the taxpayer
pect theory to a simple model of tax eva- does not think merely in terms of out-of-
sion, inquiring into the relationships be- pocket gains and losses at the time of fil-
tween tax compliance and advance tax ing. It also involves weighing the altema-
payments. The results support the empiri- tive cost of self-selecting into a refund
cal and experimental evidence that ad- status against the benefit of doing so,
vance tax payments, even if set below tax- which is rather obscure. As recently noted
payers' true tax liability, may substitute for by Slemrod et al. (1997), why taxpayers
costly detection efforts in enhancing com- choose to grant the govemment an inter-
pliance. Moreover, sufficiently high advance est-free loan is a fascinating question,
payments may induce full compliance. Highfill et al. (1998) suggest that taxpay-
However, while advance payments may be ers may overwithhold so as to avoid a
set only slightly above taxpayers' true tax penalty in case that withholding does not
liability to induce full compliance when reach a certain percentage of their current
detection efforts are relatively high, they (or last year) tax liability, if facing income
must be raised considerably as detection and deductibles uncertainty at the time of
efforts are reduced, approaching infinity as withholding. Given this motive, Highfill
detection efforts approach their low real- et al. model the prepayment decision of a
life levels. Hence, as long as taxpayers risk-neutral taxpayer who does not seek
hold realistic assessments of the actual to avoid or evade taxes. Allowing for tax
probability of detection, it is unlikely that evasion would substantially complicate
reasonably high advance payments, al- the model because of the additional source
though ensuring a tax refund, will suffice of uncertainty involved.
to eliminate tax evasion. This pessimistic Acknowledgments
conclusion is, however, moderated if tax-
payers are known to overweight the low I wish to thank Henk Elffers, Thea Herz,
probability of detection, in which case, if Dick Hessing, Shlomo Yitzhaki, three
overweighting is substantial, tax advances anonymous referees and the editor, Dou-
set even helow taxpayers' true tax liability glas Holtz-Eakin, for helpful comments
may suffice to induce honesty. and suggestions. The remairung errors are
solely my responsibility.
" This problem is discussed by Martinez-Vazquez et al, (1992), who used an experimental approach to examine
whether underwithheld taxpayers are more likely to evade taxes, controlling for the possible interrelation-
ship between underwithholding and evasion through distinguishing between intended and unintended
underwithholding and concentrating on the impact of the latter.

762
Tax Compliance and Advanced Tax Payments

REFERENCES Highfill, Janet, Douglas Thorson, and William

V. Weber.
AUingham, Michael G., and Agnar Sandmo. "Tax Overwithholding as a Response to Un-
"Income Tax Evasion: A Theoretical Analy- certainty." Public Finance Review 26 No. 4
sis." Journal of Public Economics 1 No. 3/4 (July,1998): 376-91.
(November, 1972): 323-38. Kahneman, Daniel, and Amos Tversky.
Aim, James, and William Beck. "Prospect Theory: An Analysis of Decision
"Tax Amnesties and Tax Revenues." Public Under Risk." Econometrica 47 No. 2 (March,
Finance Quarterly 18 No. 4 (October, 1990): 1979): 263-91.
433-53. Martinez-Vazquez, Jorge, Gordon B. Harwood,
Aim, James, Gary H. McClelland, and William and Ernest R. Larkins.
D. Schulze. "Withholding Position and Income Tax
"Why Do People Pay Taxes?" Journal of Pub- Compliance: Some Experimental Evidence."
lic Economics 48 No. 1 Qune, 1992): 21-38. Public Finance Review 20 No. 2 (April, 1992):
Carroll, John S. 152-74.
"How Taxpayers Think About Their Taxes: McClelland, Gary H., William D. Schulze, and
Frames and Values." In Why People Pay Taxes: Don L. Coursey.
Tax Compliance and Enforcement, edited by Valuing Risk: A Comparison of Expected Utility
Joel Slemrod, 43-63. Ann Arbor: Michigan with Models from Cognitive Psychology. Univer-
University Press, 1992. sity of Colorado Working Paper. Boulder:
Chang, Otto H., Donald R. Nichols, and Joseph University of Colorado, 1986.
J. Schultz. Pechman, Joseph A.
"Taxpayer Attitudes Towards Tax Audit Federal Tax Policy. Washington, D.C: The
Risk." Journal of Economic Psychology 8 No. 3 Brookings Institution, 1983.
(September, 1987): 299-309. Robben Henry S.J.
Chang, Otto H., and Joseph J. Schultz. A Behavioral Simulation and Documented Be-
"The Income Tax Withholding Phenomenon: havior Approach to Income Tax Evasion.
Evidence from TCMP Data." Journal of the Deventer: Kluwer, 1991.
American Taxation Association 12 No. 1 (Feb- Robben, Henry S.J., Paul Webley, Russell H.
ruary, 1990): 88-93. Weigel, Karl-Erik Wameryd, Karyl A. Kinsey,
Cox, Dennis R., and Alan H. Plumley. Dick J. Hessing, Francisco Alvira Martin,
Analysis cf Voluntary Compliance Ratesfor Dif- Henk Elffers, Richard Wahlund, Luk Van
ferent Income Source Classes. Unpublished Langenhove, Susan B. Long, and
report. Washington, D.C: Internal Revenue John T. Scholz.
Service, Research Division, 1988. "Decision Frame and Opportunity as Deter-
Elffers, Henk, and Dick J. Hessing. minants of Tax Cheating: An Intemational
"Influencing the Prospects of Tax Evasion." Experimental Study." Journal of Economic Psy-
Journal of Economic Psychology 18 No. 3 chology 11 No. 3 (September, 1990): 341-64.
(April, 1997): 289-304. Slemrod, Joel, Charles Christian, Rebecca
Hessing, Dick J., Henk Elffers, Henry S.J. London, and Jonathan A. Parker.
Robben, and Paul Webley. "April 15 Syndrome." Economic Inquiry 35
"Does Deterrence Deter? Measuring the No. 4 (October, 1997): 695-709.
Effect of Deterrence on Tax Compliance Webley, Paul, Henry S.J. Robben, Henk Elffers,
in Field Studies and Experimental Stud- and Dick J. Hessing.
ies." In Why People Pay Taxes: Tax Compli- Tax Evasion: An Experimental Approach. Eu-
ance and Enforcement, edited by Joel ropean Monographs on Social Psychology.
Slemrod, 291-310. Ann Arbor: Michigan Cambridge, UK: Cambridge University
University Press, 1992. Press, 1991.

763
 NATIONAL TAX JOURNAL

Yaniv, Gideon. Yaniv, Gideon.

"Withholding and Non-Withheld Tax Eva- "Tax Evasion by Misinforming Withholding
sion." Journal of Puhlic Economics 35 No. 2 Agents." Public Finance Review 26 No. 4 (f uly,
(March, 1988): 183-204. 1998): 362-75.
Yaniv, Gideon. Yitzhaki, Shlomo.
"Collaborated Employee-Employer Tax "A Note on 'Income Tax Evasion: A Theo-
Evasion." Public Finance/Finances Publiques retical Analysis.'" Journal of Public Econom-
47 No. 2 (May, 1992): 312-21. ics 3 No. 2 (May, 1974): 201-2.
Yaniv, Gideon.
"ANote on the Tax-Evading Firm." National
Tax Journal 68 No. 1 (March, 1995): 113-20.

764