Tax Evasion and Tax Compliance

52
6020
T
AX
E
VASION AND
T
AX
C
OMPLIANCE
Luigi Alberto Franzoni
University of Bologna, Italy
© Copyright 1999 Luigi Alberto Franzoni
Abstract
This chapter offers an overview of the theoretical and empirical research on tax evasion, delineating the variety of factors affecting noncompliance and examining possible remedies. Particular emphasis is placed on the institutional and procedural rules governing the tax enforcement policy.
JEL classification:
K34
Keywords:
Tax Enforcement, Compliance, Taxpayer’s Behavior, Tax Gap
1. Introduction
Tax evasion is said to occur when individuals deliberately fail to comply with their tax obligations. The resulting tax revenue loss may cause serious damage to the proper functioning of the public sector, threatening its capacity to finance its basic expenses.
Although tax compliance is a major concern for all governments and analytical investigation of tax evasion can be traced as far back as the work, one of the pioneers of ‘law and economics’, Cesare Beccaria (1764), the problem was long segregated from the main body of economics and left essentially to the attention of tax authorities and jurisprudence. The modern use of economic tools for the analysis of tax compliance can be credited to Allingham and
Sandmo
([1972] 1991), who extended the influential work of Becker (1968) on law enforcement to taxation using modern risk theory.
In
the decades since, the literature on tax evasion has blossomed (as witnessed by the voluminous bibliography enclosed). Probably no aspect of tax compliance has escaped at least preliminary scrutiny. Detailed introductions to this theme are now available, as in the monographs of Cowell (1990)
(theoretically
oriented) and Roth, Scholtz and Witt (1989) (an interdisciplinary perspective), and the surveys of Andreoni, Erard and Feinstein (1998)
(including
a thorough discussion of empirical results) and Slemrod and
Yitzhaki
(1998) (with large sections devoted to avoidance and administration).
As
a complex phenomenon, tax compliance can be addressed from a variety of perspectives. Taxpayers’ stance is influenced by many factors, including their disposition towards public institutions, the perceived fairness of the taxes,
6020
Tax Evasion and Tax Compliance
53
prevailing social norms, and the chances of noncompliance being detected and punished. Without questioning the relevance of ethical and sociological motivations, the economic analysis of tax compliance has focused mainly on how evasion can be deterred through detection and sanctions. The thesis is that the taxpayer’s behavior can be fruitfully seen as the result of a rational calculus, a careful assessment of the costs and benefits of evasion. Since even in the simple st tax and enforcement systems the incentives to comply are far from obvious, this economic perspective offers precious insights that can be used to derive suitable policy measures. Yet, given the complexity of the economic set-up in which the taxpayer usually makes compliance decisions, no simple policy prescription should be expected. In fact, as we will see, to date theoretical and empirical research has managed to establish very few firm points. Nevertheless, the general picture of tax compliance is much clearer now than just a few decades ago. At least the literature has shown that evasion is a serious problem, too complex to be solved by simple policy adjustments, and that the set of instruments for controlling it is vast.
This
chapter provides an overview of the findings of the theoretical and the empirical literature on tax evasion. Section 2 defines tax evasion, as opposed to tax avoidance and other unlawful activities. In Section 3, Allingham and
Sandmo’s
basic model of tax evasion is presented and discussed, with a brief review of its numerous extensions. Section 4 surveys the empirical evidence on taxpayer compliance. Section 5 deals with optimal tax enforcement policy, and investigates some possible strategies for combating evasion. Section 6 examines additional policy issues, connected with the institutional and procedural aspects of tax enforcement. Section 7 provides some concluding observations.
2. Definition and Extent of Tax Evasion
By
distancing effective payments from statutory taxes, tax evasion defines a specific revenue deficiency, known as the ‘tax gap’ (in the US, for example, the federal income tax gap has been estimated at 17 percent).
Let
us emphasize from the outset that the tax gap is not equal to the amount of additional revenue that would be collected by stricter enforcement, for perfect enforcement would significantly affect the economic scenario (some firms would go bankrupt, taxpayers would modify their labor supply, prices and incomes would change, and so on), so the tax base would surely be altered. As a result, at least in theory, net revenue could even turn out to be smaller. Thus standard measures of tax gaps must be interpreted cautiously. They are only roughly suggestive of the likely immediate effects of marginal improvements in enforcement. Also, one should be wary of the cliché that statutory taxes represent the ideal world and tax gaps an intrinsic evil. This is not only because
54
Tax Evasion and Tax Compliance
6020
taxes may not be ‘just’, but also because statutory taxes themselves are usually determined by a legislature that is perfectly aware that they will only be partially enforced and therefore differ from those that would be optimal under perfect enforcement.
On
closer scrutiny, therefore, estimation of the tax gap merely portrays the wedge between economic reality and a purely legal construct called ‘statutory taxes’. Reality and its legal representation may differ for any number of reasons, among which, as we shall see, the willful misrepresentation of tax liabilities is just one.
In
economic terms, evasion problems originate in the fact that the variables that define the tax base (incomes, sales, revenues, wealth, and so on) are often not ‘observable’. That is, an external observer cannot usually see the actual magnitude of an individual’s tax base, and hence cannot know his true tax liability. Sometimes this knowledge can be obtained by means of costly audits, in which case we say that the tax base is verifiable (at a cost). In other cases, as when it is related to cash payments, the tax base cannot be verified at all.
Taxpayers
can take advantage of the imperfect information about their liability and elude taxation.
A
related concept is tax avoidance (or reduction), by which individuals reduce their own tax in a way that may be unintended by tax legislators but is permissible by law. Avoidance is typically accomplished by structuring transactions so as to minimize tax liability. In some cases, avoidance is encouraged by legislation granting favorable tax treatment to specific activities in contrast to general taxation principles. From a legal standpoint, evasion differs from avoidance in being unlawful, and hence punishable (at least in theory). As far as economic function is concerned, however, evasion and avoidance obviously have very strong similarities; sometimes, indeed, they can hardly be distinguished (see for instance Feldman and Kay, 1981; Cowell,
1990;
McBarnet, 1992). This adds to the difficulty of interpreting the real implications of the tax gap.
Another
problem with the measurement of tax evasion relates to its proper delimitation within the broader set of the informal economy. No taxes are generally levied on transactions in the home and criminal sectors, which are usually beyond the reach of authorities and official statistics. Hence, proper determination of the boundaries of evasion is a formidable task, in that evasion is often inextricable from other illegal and unrecorded activities. What, one might ask, is the evaded tax of a hired killer?
Aggregate estimates of evasion must deal with all these problems, in addition to the classic problem of lack of direct data. Various estimation methods have been devised, some based on data collected by fiscal authorities, others - less reliable - on data derived from national accounts and surveys.
Their
application suggests that in the Western industrialized countries evaded
6020
Tax Evasion and Tax Compliance
55
ta xes amount to between 5 percent and 25 percent of potential tax revenue, depending on the technique adopted and the country (see monographs by Feige,
1989;
Pyle, 1989, and Thomas, 1992), with higher figures (up to 30-40 percent) for less-developed countries (Tanzi and Shome, 1994).
As
noted, one should not attach too much importance to such estimates, which essentially tell us that statutory taxes are not the whole story. What matters is effective taxation
,
that is, the net tax burden on individuals. This has major implications bearing on the economic consequences of evasion: the main question is not how evasion alters the shape of statutory allocation of the fiscal burden, but how it constrains the set of policies that can be implemented. When taxes can be evaded, taxation will prove to be an imperfect tool for pursuing government aims (be they redistribution, efficiency, or whatever), which will be only partly achieved. Indeed, effective taxation may turn regressive, as the more affluent usually have better opportunities to evade (or avoid) taxes. Also, evasion may be powerfully deleterious to horizontal equity, owing to unequal distribution of opportunities to evade and of the willingness to seize them. This in turn may induce production inefficiencies, because competition would be distorted by the unequal distribution of the tax burden among firms.
Th
e adverse consequences of tax evasion are sometimes exacerbated by laws, or even constitutions, drafted as if the tax base were observable, limiting the set of corrective instruments available to the government (which cannot, for instance, set tax rates according to their presumed degree of enforceability).
In
order to evaluate the way in which noncompliance affects the actual tax payme nt of individuals, one must examine taxpayers’ compliance behavior more closely. This can be done by developing a theoretical model to predict how taxpayers’ behavior is affected by the relevant variables. The following section reviews some models and assesses their fit with observed practise.
3. The Decision to Evade
Compliance
with the tax law typically means: (i) true reporting of the tax base,
(ii)
correct computation of the liability, (iii) timely filing of the return, and (iv) timely payment of the amounts due. The bulk of tax evasion involves the first point. Most evaders either do not declare their liability at all, or declare it only in part. In the following, we concentrate on the problem faced by an individual who has to decide how much of his tax aggregate to report, or whether to report it at all. The focus is on income taxes (which account for a large part of fiscal revenue in most countries). However, the insights provided can be applied to other taxes as well.
56
Tax Evasion and Tax Compliance
6020
3.1 Allingham and Sandmo
A
useful model of taxpayers’ evasion decision is that developed by Allingham and Sandmo ([1972] 1991) and Srinivasan (1973), and revised by Yitzhaki
(1974).
Evasion is viewed as a portfolio allocation problem: the taxpayer must decide what portion of his income y (postulated as exogenous) to invest in the risky activity labeled ‘tax evasion’. If the taxpayer does not want to take any risk, he reports his income in full; otherwise, he reports only a fraction of it and bears the risk of being caught and fined. The problem is thus to choose the optimal tax return, when the income reported is taxed at a fixed rate t and evasion is fined at a penalty rate f proportional to the tax evaded. The probability of an audit, that is, the probability that the true income level will be discovered, is a constant denoted by a . The taxpayer decides the amount to conceal so as to maximize his expected utility from net income. If we call y NA the net income when the evader is not audited (gross income less taxes on reported income) and y A the net income when he is audited (gross income less taxes on true income less the fine), we can write the taxpayer’s expected utility as EU
(
e
)
= (1
!
a
)
u
(
y
NA
) + a u
(
y
A
) = (1
!
a
)
u
[
y
!
t
(
y
!
e
)] + a u
(
y
!
ty
!
fte
),
where e denotes the amount of income concealed.
This
representation yields some interesting results from the standpoint of comparative statics. On the reasonable assumption that the taxpayer is risk averse, it can be shown that the amount of tax evaded, te *
, varies inversely with the audit rate a and the penalty rate f , while it depends negatively on the tax rate t and positively on income y if and only if the taxpayer’s utility function displays Decreasing Absolute Risk Aversion. Further, the proportion of tax evad ed, te
*
/y
,
increases with income if and only if taxpayer’s utility function displays Decreasing Relative Risk Aversion (see Cowell, 1990). Of these results, the least obvious is surely the inverse correlation between the amount of evasion and the tax rate (with DARA). This stems from the fact that both the direct gain from evasion (taxes saved) and the expected fine depend proportionally on t . Hence, an increase in the tax rate does not induce the
‘substitution’
of the risky asset for the safe one, but operates solely through the reduction in disposable income (Yitzhaki, 1974).
Once
the optimal amount of underreporting, e* , has been calculated, one can easily derive the ‘evasion rent’, defined as the monetary benefit accruing to the evader (more precisely, the amount of income that he would be willing to pay to switch from a virtual system of perfect enforcement to the actual, imperfect, one):
Evasion rent
= [1
!
a
(1 + f )] te* !
RP
( e* ), where RP
(
e*
)
is the risk premium associated with the audit lottery. The evasion rent is therefore equal to the net return on evasion (evaded taxes less expected sanctions) less the ‘loss’ due to the riskiness associated with random auditing
6020
Tax Evasion and Tax Compliance
57
(note, incidentally, that if taxpayers were risk-lovers, the risk premium would effectively represent a ‘gain’).
Several
observations are in order. First, for evasion rent to be positive, the net return on evasion has to be positive (the evasion gamble needs to be ‘better than fair’). That is, for evasion to occur at all, it is necessary that a (1
+
f
) < 1; that is, the penalty and audit rates must be sufficiently low. Second, when the net return on evasion is positive, the only reason why taxpayers may not evade their whole taxes is the fear of uncertainty (the risk premium loss). Indeed, if taxpayers were infinitely averse to risk, they would report their income in full even if the net return on evasion were positive (taxpayers are hyper-pessimists and behave as if they were to be audited for sure). Finally, the risk premium produces a differential between the rent to the taxpayer and the net revenue loss to the tax authorities. Hence, it provides a monetary measure of the ‘deadweight loss’ due to the randomness of tax enforcement (Yitzhaki, 1987).
3.2 Extensions
This
basic model gives an account of taxpayers’ evasion decisions in a very simple set-up: taxes and penalties are proportional, the audit rate is constant, only one form of evasion is available. In addition, the taxpayer is assumed to rely on expected utility theory and to be perfectly amoral, that is, to make compliance decisions with exclusive reference to the consequences for net income. All these assumptions are open to criticism, and models based on alternative assumptions have been developed. The following touches briefly on these contributions.
One
standard criticism of the Allingham and Sandmo model is grounded in the belief that compliance decisions depend on moral views. This is clearly a problematic issue, one that cannot be captured by the consequentialistic set-up of standard decision theory. Bordignon (1993) makes an interesting attempt to account for non-self-motivated decisions in tax evasion. He develops a compliance model in which taxpayers are guided by suitably defined ‘Kantian principles’, which determine the amount that each taxpayer considers fair to pay. Under this assumption, it turns out that tax evasion is generally lower than un der selfish behavior, that compliance depends on the level of public expenditure, and that evasion is likely to increase with tax rates.
Other
authors have stressed the ‘social’ factors at the basis of the taxpayers’ decision (see Roth, Scholtz and Witt, 1989, for an excellent account of the sociological research). Economists have emphasized the ‘stigma’ attached to the violation of social norms and shown that tax evasion may have strong spillover effects. Social stigma is likely to give rise to a multiplicity of possible equilibria: when most people evade, the stigma effect is small and evasion is not in fact discouraged; when few evade, the stigma effect is great and evasion is discouraged. The transition from one equilibrium to the other takes the form of a ‘noncompliance epidemic’: if, for some reason, more people start to evade,
58
Tax Evasion and Tax Compliance
6020
th e stigma decreases and evasion spreads to an ever larger fraction of the population (see Benjamini and Maital, 1985; Gordon, 1989, and Myles and
Naylor, 1996).
Alm
and McCallin (1990), Landskroner, Paroush and Swary (1990), Yaniv
(1990),
and Wrede (1995) have extended Allingham and Sandmo with models in which taxpayers face more complex ‘portfolio’ set-ups offering other risky activities and alternative forms of evasion. Wadhawan (1992) posits that audits detect only a fraction of taxpayers’ evasion, while Das-Gupta (1994) analyses the case in which taxpayers’ income derives from a multiplicity of transactions.
Scotchmer
and Slemrod (1989) and Scotchmer (1989) consider the effect of randomness in tax liability assessments. Among other things, both papers conclude that uncertainty over the true liability level or outcome of the audit increases net tax revenue, either because increased uncertainty makes evasion more costly (when taxpayers are risk averse) or because it may lead taxpayers to underreport their income and be subject to a fine (whereas overreporting only yields a rebate of the overpaid tax).
Several
authors have tried to extend Allingham and Sandmo’s model to include the labor supply decision, so as to endogenize taxpayers’ gross income
(see,
among others, Andersen, 1977; Pencavel, 1979; Isachsen and Strom,
1980;
Isachsen, Samuelsen and Strom, 1985, and Cowell, 1985). The problem is that as soon as the labor decision is factored in, the simple comparative statics of Allingham and Sandmo are lost. Depending on the taxpayer’s marginal disutility from labor and her risk-attitudes, all predictions become possible. This problem is usually overcome by imposing strong restrictions on the utility function. Cowell (1985) takes a different course, assuming that decisions are made in two separate stages: first, the taxpayer decides how many hours to work; then he allocates this total labor supply between legal and illegal activities (alternatively, between reported and unreported income). On this assumption, Cowell is able to show that Allingham and Sandmo’s results carry over (with some qualifications) to the extended set-up if taxpayer’s labor supply is forward rising. Perhaps more importantly, he shows that the comparative statics results are strictly dependent on the nature of the evasion choice, as it can be tied either to the amount of income to report (for the self-employed) or to the amount of time to spend in ‘off the books’ activities (for the moonlighter). The insights drawn from analysis of income tax evasion usually apply to other forms of evasion as well. Different considerations may be relevant, however, when the taxpayer is a firm subject to indirect taxation, as the evasion decision may affect output or pricing policy (tax shifting). However, Marrelli
(1984)
derives a separability result for the case of a monopolist: the evasion and shifting decisions are independent of one another as long as the audit probability is constant (see also Yaniv, 1995). The same result applies to oligopolistic markets when firms compete à la Cournot (Marrelli and Martina,