Expected Utility Maximization Approach

Allingham and Sandmo (1972) assume that tax declaration is a decision under uncertainty. The taxpayer has a choice between two main strategies: (1) to declare his actual income, or (2) to declare less than his actual income. If he chooses the first option, he will have to pay the full amount of the tax. However, if he chooses the second option, he can pay less than the full amount, but he will have to face the probability of being detected and penalized. His problem is to maximize the expected utility derived from his income after tax and penalty (if any). Allingham and Sandmo assume that the taxpayer’s behavior conforms to the von Neumann-Morgenstern axiom for behavior under uncertainty. The cardinal utility function of the taxpayer has income as its only argument and the marginal utility is assumed to be everywhere positive and strictly decreasing so that the individual is risk averse.³ The taxpayer’s objective function is

View Expanded

where W is the taxpayer’s actual income which is exogenously given and is known to the taxpayer but not to the tax authority; θ is a constant tax rate, θ > 0; X, the income reported to the tax authorities, is the taxpayer’s decision variable, X > 0; p is the probability that the evasion will be detected; n is the penalty rate on unreported income, Π > 0; U is the utility from disposable income; and E(U) is the expected utility. The taxpayer chooses his declared income (X) so as to maximize his expected utility [E(U)].

Under the assumption of decreasing absolute risk aversion,⁴ Allingham and Sandmo show that the sign of ∂X/∂θ is indeterminate, which means that the effect of the tax rate on the reported income is indeterminate. Allingham and Sandmo explain that this indeterminate effect is the result of the interaction between the substitution and the income effects. The substitution effect is negative because an increase in the tax rate makes it more profitable to evade taxes on the margin. The income effect is positive because an increased tax rate makes the taxpayer less wealthy, and, under decreasing absolute risk aversion, this tends to reduce evasion.

The effect of a penalty rate on the reported income, ∂X/∂Π is shown to be unambiguously positive, which implies that an increase in the penalty rate will always increase the reported income. The effect of the probability of detection is also shown to be unambiguously positive, which implies that an increase in the probability of detection will always lead to a larger income being declared.

Expected Income Maximization Approach

The expected income maximization approach of Srinivasan (1973) assumes that the objective of the taxpayer is to maximize his expected income after taxes and penalties. Thus, the taxpayer’s problem is to maximize the following expected income function:

View Expanded

where y is an individual’s true income; T(y) is the tax based on true income; λ is the proportion by which income is understated; P(λ) is the penalty multiplier—thus, P(λ)λy is the penalty on the understated income λy; Π is the probability of detection; and A(y) is the expected income after taxes and penalty.

Srinivasan assumes that T(y) > 0, 0 < T^′(y) < 1. and T^′′(y) ≥ 0 for all y > 0, which means that the tax is a positive, increasing, and convex function of income. He also assumes that P(λ) ≥ 0, P^′(λ) < 0, and P^′′(λ) ≥ 0, which means that the penalty multiplier is a positive, increasing, and convex function of the proportion by which income is understated, λ.

Srinivasan denotes the optimum proportion of understated income by λ^* and shows that ∂λ^*/∂Π < 0, which means that the optimum proportion of understated income, (λ^*), decreases as the probability of detection (Π) increases. The higher the probability of detection, the lower the proportion of income being evaded. He also shows that ∂λ^*/∂y is > 0 if T′′ is positive. This means that given a progressive tax function, and a probability of detection, Π, independent of income, y the richer the person, the larger the optimal proportion by which he will understate his income. Under a progressive tax structure, as income increases, the proportion of underreported income also increases. This result is true only when the probability of detection is independent of income. If the probability of detection is an increasing function of income and the tax rate is proportional, then the proportion of understated income will decrease as income increases.

Comparison of the Two Models

The two models are different in the assumptions concerning the objective function (including the utility specification), the tax rate, and the penalty rate (Table 1). The objective function of Allingham and Sandmo’s model is an expected utility function, whereas in Srinivasan’s model it is an expected income function. Allingham and Sandmo’s utility function implies that the taxpayer is risk averse, while the expected income function of Srinivasan implies that the taxpayer is risk neutral. Allingham and Sandmo assume a proportional tax rate, but Srinivasan allows the tax rate to be either proportional or progressive. Allingham and Sandmo’s penalty rate is proportional, but Srinivasan’s penalty rate is a positive, increasing, and convex function of the proportion of unreported income (λ).

^{Table 1.}

Comparison of the Assumptions of the Allingham and Sandmo Model and the Srinivasan Model of Tax Evasion

Sources: Compiled from Michael G. Allingham and Agnar Sandmo, “Income Tax Evasion: A Theoreti-cal Analysis,” Journal of Public Economics (Amsterdam). Vol. 1 (November 1972), pp. 323-38. and T.N. Srinivasan, “Tax Evasion: A Model,” Journal of Public Economics (Amsterdam), Vol. 2 (November 1973), pp. 339-46.

^{Table 1.}

Comparison of the Assumptions of the Allingham and Sandmo Model and the Srinivasan Model of Tax Evasion

	Allingham and Sandmo’s Model		Srinivasan’s Model
1.	Assumptions concerning the utility function	1.	Assumptions concerning the utility function
	The taxpayer’s behavior conforms to the von Neumann-Morgenstern axiom for behavior under uncertainly.		None.
	The taxpayer’s cardinal utility function
	has income as its only argument.
	The marginal utility is everywhere positive and strictly decreasing, which implies that the individual is risk averse.
2.	Assumptions concerning income	2.	Assumptions concerning income
	Actual income, W, is exogenously given		Implicitly, actual income, Y. is exogenously
	and is known to the taxpayer but not to		given and is known to the taxpayer but not
	the government’s tax collector.		to the government’s tax collector.
	The taxpayer chose X, the declared income,		The taxpayer chose λ, the proportion by
	so as to maximize his utility function		which income is understated, so as to
			maximize his expected income.
3.	Assumptions concerning the tax rates	3.	Assumptions concerning the tax rates
	Tax is levied at a constant rate, θ.		The tax paid is a function of income and T(X) > 0.
			The marginal tax rate is positive and strictly
			less than unity, that is, 0 < T′(Y) < 1.
			T′′(Y) ≥ 0, which means that the tax rate could be either proportional or progressive. If T′′(Y) = 0 for all Y, we get a constant marginal tax rate, which together with T(0) = 0 will yield a proportional tax rate. If T′′(Y) > 0, then it will correspond to a progressive tax structure.
4.	Assumptions concerning the probability of detection	4.	Assumptions concerning the probability of detection
	Tax evasion may he detected at a probability of n.		lax evasion may he detected at a probability of n.
5.	Assumptions concerning the penally rate	5.	Assumptions concerning the penalty rate
	If the tax evasion is detected, the penalty		The penalty multiplier, P(λ), is a positive
	is at the rate of Π and is imposed on the		increasing, and convex function of λ,
	undeclared income (W — X)		that is, for all λ > 0.
			$\begin{array}{cc}P\left(\lambda \right)\ge 0& \left(5.1\right)\end{array}$
			$\begin{array}{cc}{P}^{\prime }\left(\lambda \right)>0& \left(5.2\right)\end{array}$
			$\begin{array}{cc}{P}^{\prime \prime }\left(\lambda \right)\ge 0& \left(5.3\right)\end{array}$
			This penalty multiplier P(λ) is imposed
			on the undeclared income λy.
6.	Assumptions concerning the objective function	6.	Assumptions concerning the objective function
	The objective function is the expected utility function		The objective function is the expected income function
	of the taxpayer, which is specified to be		of the taxpayer, which is specified to be
	E(U) = (1 - P)U(W - θX) + P[U(W - θX − Π(W - X))].		A(Y) = Π[y-T(y)-λP(λ)y] + (1-Π){y - T[(1-λ)y]}.

Sources: Compiled from Michael G. Allingham and Agnar Sandmo, “Income Tax Evasion: A Theoreti-cal Analysis,” Journal of Public Economics (Amsterdam). Vol. 1 (November 1972), pp. 323-38. and T.N. Srinivasan, “Tax Evasion: A Model,” Journal of Public Economics (Amsterdam), Vol. 2 (November 1973), pp. 339-46.

View Table

^{Table 1.}

Comparison of the Assumptions of the Allingham and Sandmo Model and the Srinivasan Model of Tax Evasion

	Allingham and Sandmo’s Model		Srinivasan’s Model
1.	Assumptions concerning the utility function	1.	Assumptions concerning the utility function
	The taxpayer’s behavior conforms to the von Neumann-Morgenstern axiom for behavior under uncertainly.		None.
	The taxpayer’s cardinal utility function
	has income as its only argument.
	The marginal utility is everywhere positive and strictly decreasing, which implies that the individual is risk averse.
2.	Assumptions concerning income	2.	Assumptions concerning income
	Actual income, W, is exogenously given		Implicitly, actual income, Y. is exogenously
	and is known to the taxpayer but not to		given and is known to the taxpayer but not
	the government’s tax collector.		to the government’s tax collector.
	The taxpayer chose X, the declared income,		The taxpayer chose λ, the proportion by
	so as to maximize his utility function		which income is understated, so as to
			maximize his expected income.
3.	Assumptions concerning the tax rates	3.	Assumptions concerning the tax rates
	Tax is levied at a constant rate, θ.		The tax paid is a function of income and T(X) > 0.
			The marginal tax rate is positive and strictly
			less than unity, that is, 0 < T′(Y) < 1.
			T′′(Y) ≥ 0, which means that the tax rate could be either proportional or progressive. If T′′(Y) = 0 for all Y, we get a constant marginal tax rate, which together with T(0) = 0 will yield a proportional tax rate. If T′′(Y) > 0, then it will correspond to a progressive tax structure.
4.	Assumptions concerning the probability of detection	4.	Assumptions concerning the probability of detection
	Tax evasion may he detected at a probability of n.		lax evasion may he detected at a probability of n.
5.	Assumptions concerning the penally rate	5.	Assumptions concerning the penalty rate
	If the tax evasion is detected, the penalty		The penalty multiplier, P(λ), is a positive
	is at the rate of Π and is imposed on the		increasing, and convex function of λ,
	undeclared income (W — X)		that is, for all λ > 0.
			$\begin{array}{cc}P\left(\lambda \right)\ge 0& \left(5.1\right)\end{array}$ View Expanded
			$\begin{array}{cc}{P}^{\prime }\left(\lambda \right)>0& \left(5.2\right)\end{array}$ View Expanded
			$\begin{array}{cc}{P}^{\prime \prime }\left(\lambda \right)\ge 0& \left(5.3\right)\end{array}$
			This penalty multiplier P(λ) is imposed
			on the undeclared income λy.
6.	Assumptions concerning the objective function	6.	Assumptions concerning the objective function
	The objective function is the expected utility function		The objective function is the expected income function
	of the taxpayer, which is specified to be		of the taxpayer, which is specified to be
	E(U) = (1 - P)U(W - θX) + P[U(W - θX − Π(W - X))].		A(Y) = Π[y-T(y)-λP(λ)y] + (1-Π){y - T[(1-λ)y]}.

Sources: Compiled from Michael G. Allingham and Agnar Sandmo, “Income Tax Evasion: A Theoreti-cal Analysis,” Journal of Public Economics (Amsterdam). Vol. 1 (November 1972), pp. 323-38. and T.N. Srinivasan, “Tax Evasion: A Model,” Journal of Public Economics (Amsterdam), Vol. 2 (November 1973), pp. 339-46.

View Expanded

In Allingham and Sandmo’s model, the assumption of decreasing marginal utility (which implies risk aversion) is so crucial to the model that if it had not been made the model would not have an optimal solution.⁵ The assumptions of constant tax and penalty rates are also used throughout the derivation. However, it is not easy to analyze the effects of comparing Allingham and Sandmo’s assumptions concerning tax and penalty rates with Srinivasan’s assumption because to do so would involve reformulation of the model. Thus, such effects are not analyzed here.

The indeterminate effect of the tax rate is based on an additional assumption, namely, the decreasing absolute risk aversion. This assumption is crucial to the result because if the absolute risk aversion had been assumed to be either constant or increasing, the tax rate would have had a negative effect on the declared income. Nevertheless, the decreasing absolute risk aversion seemed to be the most attractive assumption because it is supported by everyday observation. The effects of the penalty rate and the probability of detection are free from the assumption of decreasing absolute risk aversion. Thus, whether the absolute risk aversion is assumed to be decreasing, constant, or increasing, the effects of both the penalty rate and the probability of detection on the declared income remain positive.

In Srinivasan’s model, the assumptions of the nonregressivity of the tax rates (T^′′(Y) ≥ 0), and the convexity of the penalty function [P^′′(λ) ≥ 0], are important to the model because without either of them there is no guarantee that the mode) will have an optimum solution.

Based on the tax rate assumption, Srinivasan’s model is more general than Allingham and Sandmo’s model because it allows one to analyze a situation where there is a proportional tax rate as well as a progressive tax rate, while Allingham and Sandmo’s model deals with the proportional tax rate only. Based on the penalty rate assumption, Srinivasan’s model is also more general than Allingham and Sandmo’s model because it allows one to analyze a situation where the penalty rate is progressive as well as where it is proportional, while Allingham and Sandmo deal only with proportional penalty rates. The significance of the difference in these two assumptions lies in the fact that Srinivasan’s model is built of tax and penalty functions, whereas Allingham and Sandmo’s model is built of tax and penalty rates. The key words are functions as opposed to rates. A model with tax and penalty functions is more flexible than a model with tax and penalty rates because the former allows one to manipulate different rate structures whereas the latter does not.

However, based on the objective functions, the model of Allingham and Sandmo is more general than that of Srinivasan because it allows the taxpayer to have his own utility function and because it takes account of the taxpayer’s risk-bearing behavior while Srinivasan’s model does not. Allingham and Sandmo’s expected utility approach allows them to assume that the taxpayer is risk averse, which is more realistic than the risk-neutral assumption implicit in Srinivasan’s model. However, since Allingham and Sandmo’s utility function has only one argument—which is income—then, in terms of the coverage of the relevant factors affecting the taxpayer’s utility, Allingham and Sandmo’s model is not different from that of Srinivasan. And because Allingham and Sandmo’s utility function has income as its only argument, ceteris paribus, the qualitative result of the two approaches would have been the same if the taxpayer’s utility function had been a monotonic transformation of the expected income. However, as assumed by Allingham and Sandmo, the taxpayer’s utility function is not a monotonic transformation of the expected income. This explains the differences in the qualitative results of the two models.

The Effect of the Tax Rate

It is frequently claimed that high tax rates induce tax evasion. This section reviews the theoretical literature to see whether there are grounds for such a claim. The effect of the tax rate on tax evasion is discussed in Allingham and Sandmo (1972), Yitzhaki (1974), McCaleb (1976), Sandmo (1981), and Koskela (1983a). As stated in the preceding section, Allingham and Sandmo (1972) have shown that the effect of the tax rate on tax evasion is indeterminate. However, their result is based on the assumption that the penalty is imposed on the understated income. The effect of the tax rate when the penalty is imposed on the evaded tax is discussed in Yitzhaki (1974), who uses the same model as Allingham and Sandmo to show that if a penalty is imposed on the evaded tax as is the case for Israel, Thailand, and a number of other countries, the indeterminacy disappears and the tax rate has a negative effect on tax evasion. Yitzhaki’s result contradicts the general belief that high tax rates stimulate tax evasion. His economic explanation of this result is that once the penalty is imposed on the evaded tax, the ordinary tax rate as well as the penalty rate increases proportionally with λ Therefore, there is no substitution effect and we are left with a pure income effect, which is positive. As the tax rate increases, the taxpayer increases his reported income and reduces the amount of tax evasion.

Based on Allingham and Sandmo’s model, McCaleb (1976) shows that the effect of tax rates on tax payment, ∂(θ)/∂θ, can be either positive, zero, or negative, which implies that an increase in the tax rate may cause the tax payment to increase, remain the same, or decrease.

Sandmo (1981) also shows that on purely theoretical grounds, one cannot easily prove the popular claim that high rates of tax stimulate activities in the hidden economy. This conforms with the results of his first paper which he co-authored with Allingham (Allingham and Sandmo (1972)).

Koskela (1983a) found that if the penalty rate is imposed on the amount of income evaded, a compensated increase in the tax rate will increase tax evasion. However, if the penalty rate is imposed on the evaded tax, a compensated increase in the tax rate will reduce tax evasion. The compensated increase in the tax rate is an increase in the tax rate that is accompanied by a lump-sum transfer so that either the expected tax revenue of the government or the expected utility of the taxpayer remains unchanged. Like Allingham and Sandmo, Koskela uses the expected utility approach and assumes a decreasing absolute risk aversion. However, he explains that when the penalty rate is imposed on the undeclared income, the increase in the lump-sum transfer neutralizes the negative income effect on tax evasion and we are left with the positive substitution effect, which increases tax evasion as tax rate increases. On the other hand, if the penalty rate is imposed on the evaded tax, then the amount of penalty will increase proportionally with the tax; thus, there is no substitution effect and the increase in the tax rate has only a negative income effect on tax evasion. This income effect is now stronger than in the case where the penalty is imposed on the undeclared income because the evaded tax also increases as a result of the increase in the tax rate. Koskela’s result does not provide grounds for the popular claim that an increase in the tax rate stimulates tax evasion. In the first place, Koskela, like Yitzhaki (1974), produced just the opposite result, that is, if the penalty is imposed on the evaded tax, an increase in the tax rate reduces tax evasion. In the second place, where Koskela provides grounds for stating that an increase in the tax rate increases tax evasion, it is under the condition that the tax rate increase is accompanied by an increase in the lump-sum transfer (in the form of tax exemptions) from the government to the taxpayer in an amount such that either the government’s expected tax revenue or the taxpayer’s expected utility remains unchanged. However, such a condition is not part of the claim.

To summarize, the theoretical literature does not support the claim that an increase in the tax rate will lead to an increase in tax evasion. Theoretical investigation shows that if the penalty rate is imposed on the evaded tax, an increase in the tax rate will lead to a reduction in tax evasion rather than to an increase as claimed. If the penalty rate is imposed on the understated income, the tax rate has an indeterminate effect on tax evasion. Only under the conditions that the penalty rate is imposed on the understated income and that such an increase in the tax rate is accompanied by a compensatory increase in the lump-sum transfer so that the expected tax revenue of the government remains unchanged will the increase in the tax rate lead to an increase in tax evasion. However, the second condition is not a part of the claim because the purpose of increasing the tax rate is to raise more revenue and not to keep the tax revenue unchanged.

The Effect of the Shape of the Tax Rate Schedule

Given the ability of an individual to pay tax, it is believed that a progressive tax rate schedule will generate more tax revenue than a proportional tax rate schedule. However, such a belief is based on the assumption that tax evasion does not exist. If tax evasion exists, it is uncertain whether a progressive tax rate schedule will generate more tax revenue. However, in recent years it has been popularly claimed that a progressive tax rate schedule stimulates tax evasion and consequently will yield less tax revenue to the government than a proportional tax rate schedule. Therefore, it is useful to investigate whether such a claim has theoretical grounds.

The effect of the shape of the tax rate schedule on tax evasion is considered in Srinivasan (1973), Nayak (1978), and Koskela (1983b). Srinivasan, assuming a progressively increasing penalty multiplier and a probability of detection, Π, independent of the level of income, y, shows that a progressive tax function that yields the same revenue as the proportional tax function in the absence of understatement of income will yield less expected revenue and penalties in the presence of understatement of income. Nayak, using a model along the lines of Srinivasan and assuming that the penalty is a progressive function of the fraction of income understated, λ, finds that a regressive tax function, which yields the same total tax revenue as a proportional tax function in the absence of understatement of income, will yield a larger expected revenue in the presence of the optimum understatement of income.

The results of the analyses of Srinivasan and Nayak imply that if the progressive, proportional, and regressive tax rate structures, which are designed to generate the same amount of revenue in a situation where tax evasion does not exist, are enforced in a situation where tax evasion does exist, a regressive tax rate structure will yield the highest tax revenue. The analyses also show that a proportional tax rate structure will yield higher revenue than a progressive tax rate structure.

Koskela (1983b), using the expected utility maximization approach and assuming a general tax function, shows that given the decreasing absolute risk aversion, the direction of the change in the fraction of income reported when the actual income increases is ambiguous for progressive and regressive taxation. However, when the taxation is linear progressive,⁶ the non-decreasing relative risk aversion is a sufficient, but not a necessary, condition for the negative relationship between the fraction of income not being declared and the actual income. When the taxation is linear regressive,⁷ the nonincreasing relative risk aversion is a sufficient, but not a necessary, condition for the positive relationship between the fraction of income not being declared and the actual income.

The intuitive explanation of the ambiguity is that under the progressive and regressive tax rates, the substitution effect and the income effect work in opposite directions. In the case of progressive taxation, on the one hand, the fraction of income reported tends to decrease as the actual income increases due to the substitution effect because it is more profitable to evade taxes. On the other hand, a rise in the income reported induced by an increase in the actual income raises the average taxes and thus makes the taxpayer worse off. This income effect tends to increase the fraction of income reported as the actual income increases. Thus, the effect is ambiguous. In the case of regressive taxation, on the one hand, the fraction of income reported tends to increase as the actual income increases due to the substitution effect because, at a lower marginal tax rate, it is less profitable to evade tax. On the other hand, a rise in the fraction of income reported, induced by an increase in the actual income, lowers the average taxes and thus makes the taxpayer better off. This income effect tends to decrease the fraction of income reported as the actual income increases. The end results depend on the size of the parameters which are unknown on a priori grounds. If the marginal tax rates are constant, then there is no substitution effect because there is no change in the relative prices, and the fraction of income not being declared depends on the behavior of the average taxes and the relative risk aversion. Thus, Koskela concluded that, given any hypothesis on the relative risk aversion, the linear regressive tax rate schedule, and not the linear progressive tax rate schedule, tends to induce a rise in the fraction of income not being declared when the actual income increases.

At first glance, it seems that Koskela’s conclusion contradicts the conclusions of Srinivasan, as well as those of Nayak. However, a closer examination indicates that the two sets of conclusions are not contradictory. Srinivasan and Nayak deal with the aggregate revenue under progressive, proportional, and regressive tax rate structures, while Koskela deals with the effect of the change in the actual income of a single taxpayer on the fraction of income that he reported. In Srinivasan and Nayak’s analyses, there is a restriction that the three tax rate structures will yield the same amount of revenue to the government, whereas there is no such restriction in Koskela’s analysis. Koskela shows that under the linear progressive tax rate structure, an increase in the actual income would lead to a reduction in the fraction of income not being declared, whereas under the linear regressive tax structure, an increase in the actual income will lead to an increase in the fraction of income not being declared. This is where the seeming contradiction arises because it implies that under the linear regressive tax rate structure, as the income of an individual increases, the fraction of income reported decreases, which tempts one to think that the amount of revenue will be reduced. This is not necessarily the case because it depends on the relative change in the fraction of income reported with respect to the relative change in the actual income. Only if the percentage reduction of the income reported is greater than the percentage increase of the actual income will there be a decrease in the total tax revenue to the government. The tax rate does not have any effect here because, first, the marginal tax rate is imposed on the marginal income and, second, under the linear regressive tax rate structure, the marginal tax rate is constant. Thus, as long as there is an increase in the absolute amount of reported income, there will always be an increase in the revenue no matter how much the fraction of income reported changes.

The regressive, the linear progressive, and the linear regressive tax rate schedules will not be discussed further because they exist only in theory and are not likely to be implemented. We will now consider the progressive and the proportional tax rate schedules because the former exists in most countries and the latter appears frequently in the tax policy literature and may replace the former.

Based on the two types of tax rate schedules, although Srinivasan’s results are not exactly comparable with those of Koskela, they do suggest the same area of policy concerns, although they suggest them differently. According to Srinivasan, a progressive tax should not be implemented because it yields less revenue in the presence of tax evasion. However, Koskela’s result suggests that whether a progressive tax will yield more or less revenue is uncertain because it is not certain whether the fraction of income reported will increase or decrease when the actual income increases. The source of the differences lies in the assumption concerning the risk-bearing behavior of the taxpayer. Srinivasan’s assumptions imply that the taxpayer is risk neutral, while Koskela assumes that the taxpayer has a decreasing absolute risk aversion. If the assumption of a risk-neutral taxpayer were incorporated in Koskela’s analysis, there would be no income effect, leaving only a positive substitution effect of an increase in income on tax evasion; thus, as income increases tax evasion will increase because it is more profitable to evade tax at a higher rate of tax. Consequently, under the progressive tax rate structure, an increase in income will lead to a decline in the fraction of income reported. In that case, there will be no differences in the policies suggested by Srinivasan and Koskela.

Thus, it can be concluded that the popular claim that a progressive tax rate schedule stimulates tax evasion and consequently will yield less tax revenue to the government than the proportional tax rate schedule has theoretical support. However, such support is based on the assumption that the taxpayer is risk neutral. If the taxpayer is assumed to have a decreasing absolute risk-aversion behavior, then the claim has no theoretical support.

The Effect of the Penalty Rate

The penalty is incorporated in the tax system to discourage tax evasion. It is generally believed that the effect of a penalty is to deter tax evasion. This section will survey the literature to see whether such a belief has theoretical grounds.

The effect of the penalty rate on tax evasion is discussed in Allingham and Sandmo (1972), Singh (1973), McCaleb (1976), Sandmo (1981), and Koskela (1983b). In Section I we have shown that Allingham and Sandmo (1972) demonstrated that an increase in the penalty rate will always increase the fraction of actual income declared. Singh (1973), using the maximization of expected income approach, shows that the higher the penalty rate, the lower the magnitude of tax evasion.

McCaleb (1976) finds that an increase in the penalty rate will lead to an increase in the tax payment. This result is based on the assumption that all policy parameters, namely, the tax rate (θ), the penalty rate (Π), and the probability of detection and conviction (P) are independent. However, if the probability of detection and conviction (P) depends negatively on the penalty rate (Π), then the effect of an increase in the penalty rate on the amount of unreported income and on the amount of tax is indeterminable. McCaleb states that the reason for assuming that the probability of detection and conviction depends on the penalty rate is because it is likely that the probability of conviction by a judge or jury may vary inversely with the severity of the punishment.

Sandmo (1981)⁸ also shows that an increase in the penalty rate will lead to a decrease in the supply of hours worked in the informal market, which implies that the proportion of unreported income will decrease and the proportion of reported income will increase.

Koskela (1983b), using the expected utility maximization approach and assuming a general tax function (i.e., the structure of the tax rate could be either progressive,⁹ proportional, or regressive¹⁰), as well as the endogenous probability of detection, found that a penally is a deterrent to tax evasion.

To summarize, theoretical results indicate that the penalty has a negative effect on tax evasion except in cases where the probability of detection is a negative function of the penalty. In such cases, the effect of the penalty on tax evasion is indeterminate. Thus, it can be concluded that, generally, the belief that the penalty is a deterrent to tax evasion has theoretical support.

The Effect of the Probability of Detection

The effect of the probability of detection on tax evasion is discussed in Allingham and Sandmo (1972), Srinivasan (1973), Singh (1973), McCaleb (1976), Christiansen (1980), Sandmo (1981), and Koskela (1983b).

Allingham and Sandmo (1972) have shown that the effect of an increase in the probability of detection is to increase the reported income. Srinivasan (1973) also shows that an increase in the probability of detection will lead to a decrease in the optimal proportion of understated income. On the basis of the tax structure of India, Singh (1973) shows that at a constant penalty rate, the higher the value of probability of detection (Π), the lower the proportion of income understated (λ). This means that an increase in the probability of detection is a deterrent to tax evasion.

McCaleb (1976) shows that an increase in the combined probability of detection and conviction would lead to an increase in the tax payment. Christiansen (1980) discussed the relative effectiveness of the penalty rate and the probability of detection as deterrents of tax evasion, which implies that an increase in the probability of detection reduces tax evasion. Koskela (1983b), assuming an endogenously determined probability of detection, shows that an increase in the probability is a deterrent to tax evasion. He assumes that the probability of detection depends positively, but linearly, on the ratio of the undeclared income to the actual income, that is, P = P(W - X)/W; P′; P′ > 0, P′′ = 0. The reasons for assuming that P is endogenously determined and is a function of X and W is because the tax authorities might base their audit and investigation policy on a statistical hypothesis that in the absence of any knowledge about the actual income, a person with a low reported income is more likely to be an evader than a person with a higher reported income; thus, the former is more likely to be subject to audit and investigation than the latter. Moreover, the authorities may have some crude indication from the general observation that the actual income of some specific taxpayers in certain sectors has increased but the reported income is unchanged. This will lead to an audit of the taxpayer’s account, which suggests that P is related to W.

To summarize, all theoretical analyses indicate that an increase in the probability of detection will lead to a decrease in tax evasion.

The Effect of Fiscal Inequity

So far, we have focused our attention on four factors, namely, the tax rate, the progressivity of the tax system, the penalty rate, and the probability of detection. All these four factors were analyzed within the framework of the conventional tax evasion models. However, these models have been criticized for being overly simplified because they view tax evasion as only a special form of gambling, namely, the gambling for extra income in light of the likelihood of detection and penalties imposed on detected tax evaders. This view takes account of only one element in the relationship between the government and the taxpayer. According to Spicer and Lundstedt (1976), the relationship between a taxpayer and his government contains at least three elements—the element of coercion, the element of internalized norm or role expectation, and the element of exchange. The conventional tax evasion model takes account of only the coercive element of the relationship. The element of exchange is discussed below. The element of internalized norm is not considered in this paper since it belongs to the realms of human psychology and sociology rather than economics.

In the element of exchange, a taxpayer is seen as exchanging purchasing power in the market in return for government services. Thus, the taxpayer’s behavior is affected by his satisfaction or lack of satisfaction with his terms of trade with the government. Tax evasion is seen partly as an attempt by the taxpayer to adjust his terms of trade with the government in response to dissatisfaction stemming from a perceived inequity in his terms of trade when compared with other taxpayers. Thus, another factor affecting tax evasion is the perceived relative inequity in the taxpayer’s terms of trade with the government. This idea has been developed further by Spicer and Becker (1980) as the link between (horizontal) fiscal inequity and tax evasion. In this context, the taxpayers’ utility functions are assumed to be interdependent so that the utility derived from extra income accrued through tax evasion depends on the taxpayer’s sense of equity regarding his relationship with the government. If the taxpayer perceives himself to be a victim of inequity, his anger increases the marginal utility that he derives from an extra dollar of tax evasion income and hence increases tax evasion. On the other hand, if he perceives himself to be the beneficiary of fiscal inequity, his guilt feelings reduce his marginal utility of income from tax evasion and hence decrease evasion. Proponents of this hypothesis acknowledge that it is not possible for the taxpayer to assess the exact value of what he pays and what he receives from the government in return. However, they argue that it seems reasonable to assume that the taxpayer has general impressions and attitudes concerning his own and others’ terms of trade with the government. As far as evidence is concerned, a number of survey research and simulation studies have reported positive correlations between the perception of fiscal inequity and tax evasion.¹¹

Simulation or Experimental Approach

Arguing that theoretical models of optimal evasion, though yielding interesting insights, are often beset by key, ambiguously signed derivatives, Friedland and others (1978) conduct a tax evasion study using the game-simulation approach. They simulate taxpaying situations and conduct experiments on 15 subjects. Their objectives are to find out (1) how sensitive income tax evasion is to changes in tax rates; (2) which socioeconomic variables are related to tax evasion; (3) whether the decisions to evade tax and the extent of tax evasion are separate and distinct decisions; and (4) whether large fines are a more effective deterrent than frequent audits. Their subjects are seven male and eight female Israeli undergraduate psychology students whose average age is 25, Regression and correlation analyses based on data from the experiments indicate that the relation between underreporting and the tax rate can be experimentally determined (at t = 25 percent, the incidence of underreporting occurred 50 percent of the time, and at t = 50 percent, the incidence occurred 80 percent of the time). The decision to evade tax (P) and the extent to which taxes are un-derreported (X) are distinct and separate decisions. Large fines with a small probability of detection are a more effective deterrent than small fines with a large probability of detection. Women are more likely to evade, but understate a much smaller fraction of their income than men, and those who habitually buy lottery tickets are not more likely to evade than those who do not. However, lottery ticket buyers conceal much more income when they do evade.

Spicer and Becker (1980) use a simulation approach to test the relationship between tax evasion and perceived fiscal inequity. Their hypothesis is that tax evasion increases for victims of fiscal inequity but decreases for beneficiaries of fiscal inequity. Their experiment is conducted on 57 subjects, 21 of whom are male and 36 female. Inequity is simulated by providing some participants with false information regarding relative tax rates. The result is that the perceived high-tax group has an evasion rate of 33 percent, the perceived medium-tax group a rate of 25 percent, and the perceived low-tax group a rate of 12 percent. A regression analysis based on data from the experiments indicates that the perceived relative tax rates, sex, and age are significant in explaining the variation in the percentage of tax evaded, but the income is not significant. This regression result supports the hypothesis that perceived inequity is another factor that affects tax evasion and that the amount of taxes evaded increases for victims of fiscal inequity but decreases for beneficiaries of fiscal inequity.

Survey Approach

The survey approach is less restrictive than the theoretical model and the simulation approach in that it allows the investigator to consider a wider range of factors. However, the survey approach is subject to sampling biases and sampling errors as well as to the problem of the reliability of the responses. Factors affecting the tax evasion examined in the survey studies can be classified into four types, namely, factors related to the degree of sanctions, administrative capabilities, fiscal inequity, and social norms. Spicer and Lundstedt (1976) investigate these four types of factors. In particular, their hypotheses are (1) tax evasion is less likely when sanctions against it are perceived to be severe; (2) tax evasion is less likely when the probability of detection is perceived to be high; (3) tax evasion is more likely when a taxpayer perceives his terms of trade with the government to be inequitable compared with other taxpayers; and (4) the more tax evaders a taxpayer knows, the more likely he is to evade taxes himself. Spicer and Lundstedt conduct a survey on 130 households selected from two suburbs in a large metropolitan area in central Ohio. From the survey data, a “tax resistance scale,” a tax evasion index, and an inequity index are constructed. The scale measures a relative propensity to evade taxes rather than tax evasion itself. The tax evasion index provides a way to assess the extent to which a taxpayer evades taxes. The inequity index measures the perceived fiscal inequity. Regression analyses indicate that the inequity index and the number of tax evaders known personally (a proxy of factors related to social norms) are significantly and positively related to tax evasion. The perceived probability of detection is significant in the equation using tax resistance as a dependent variable but not in the one using tax evasion. The perceived severity of sanctions is not significant in any equation.

Spicer and Lundstedt find that the level of family income is negatively related to tax resistance and that the proportion received in wages, salaries, and pensions is positively related. The first finding is in conflict with Srinivasan’s theoretical result¹² and with Mork’s (1975) finding, which is reviewed below. The second finding contradicts the general belief that tax evasion opportunities decrease as the proportion of income received in wages, salaries, and pensions increases. Spicer and Lundstedt explain that these discrepancies may have arisen because those with higher incomes or smaller proportions of wage income have greater opportunities to resort to tax avoidance—which is legal—and, therefore, the motivation to undertake risky acts of tax evasion is reduced. They also find that age is significant and negatively related to tax resistance, whereas the experience with tax audits is significant but positively related to tax resistance.

Mork (1975) uses interview data from the Norwegian Occupational Life History Study, which are compiled by the Institute of Applied Social Research in Oslo. This data source, which is composed of 3,479 observations, is extraordinarily extensive and includes data on the 1970 income of the interview respondents. The income data from the survey are then compared with the assessed income¹³ and pension-giving income¹⁴ from the income tax return of each respondent. These income data are tabulated by income classes, and the average reported assessed income (X₁) and the average reported pension-giving income (X₂) for each class are then divided by the interview income (X)obtaining X₁/W and X₂/W. If we consider the interview income (W) as actual income, then X₂ W and X₂W are the proportion of income reported in the context of the standard tax evasion models, Mork found that as income moves to the higher brackets, the proportion of income reported is smaller. Thus, Mork concludes that empirical evidence indicates that as income increases the proportion of income reported decreases. This, in the framework of Allingham and Sandmo (1972), implies that the relative risk aversion is a decreasing function of income. This result is consistent with Srinivasan’s theory but inconsistent with Spicer and Lundstedt’s survey result.

Enrick (1963), using the survey approach, investigates U.S. taxpayers’ income tax consciousness or awareness. In particular, his purposes are to examine (1) how accurately individuals estimate their own tax liabilities, apart from the particular day near April 15 when a final return must be filed; (2) if they tend to make errors in estimation, in what direction are these likely to fall; and (3) whether the withholding has a demonstrable effect on the accuracy or inaccuracy of estimation. Enrick argues that the study of tax consciousness or awareness is important because if we do not know peoples’ tax consciousness it does not make much sense to claim that we know the extent to which changes in their tax burden will affect their behavior. In Enrick’s survey, first, without referring to his income tax return, each participant is asked to guess and write down the total amount of federal income tax he thought he had paid in the previous year and then after looking it up, write down the actual amount he paid. Enrick’s survey result indicates that the respondents do not accurately know the amount of federal income taxes they paid. In addition, a slight tendency to underestimate rather than overestimate the amount of tax paid is noted. And finally, there is no demonstrable effect of withholding on the degree of income tax awareness. Schmölders¹⁵ conducts a similar study for the Federal Republic of Germany but reports a moderate degree of tax overestimation. According to Enrick, the difference between his result and that of Schmolders is possibly due to the difference in questionnaire technique. The German taxpayer is asked to estimate his tax burden as a percent of gross income. The taxpayer’s error is then evaluated based on statistical data on average income taxes paid by various craft, business, and professional groups of taxpayers. Enrick argues that by being asked only one figure, the respondent may possibly have felt inclined to exaggerate his tax burden. In Enrick’s study, the taxpayers are asked both the estimated and the actual tax paid in juxtaposition.

Forms of Tax Evasion

There are four possible cases of tax evasion: (1) providing false values of the socioeconomic variables that enter into the determination of the tax base such as underreporting income or claiming nonexistent dependents; (2) intentionally misinterpreting the law so that the tax liability is minimized; (3) doing both (1) and (2); and (4) failing to file a tax return. The first case includes actions such as underreporting or unreporting of income, claiming nonexistent dependents, claiming fictitious expenditures, and engaging in any other illegal activities that have the effect of reducing the tax base (net income). The second case includes actions such as intentionally applying wrong tax rates, intentionally claiming unentitled tax credits, and engaging in any other illegal activities that would lead to the reduction of the legal tax liability.¹⁶ The theoretical literature covers only the first case and part of the third case, which involve underreporting of net income. It does not cover tax evasion under the second case, the remaining part of the third case, and the fourth case. Thus, the scope of the literature is rather limited.

Types of Income Evaded

In general, taxation is evaded on two types of income: income derived from illegal activities and income derived from legal activities. For the former, the decision to evade taxes arises because the activities are illegal and need to be kept a secret; therefore, the income derived therefrom is not reportable. This decision may seem to be independent of tax policy and tax administrative parameters but is actually not.¹⁷ For the latter, the decision to evade the taxes stems from the expected utility maximization or the expected income maximization behavior of the taxpayer. In this case, the decision to evade taxes is affected by tax policy and tax administrative parameters. The literature reviewed deals with tax evasion of income apparently derived from legal activities without reference to that derived from illegal activities.

Individual Versus Aggregate Tax Evasion Behavior

Tax evasion behavior is similar to consumption behavior in the sense that it can be analyzed either at the individual or micro-level, or at the aggregate or macro-level. Literature on the theoretical and empirical analysis of tax evasion reviewed in the previous sections is at the micro-level,¹⁸ because it deals with the decision making of individual taxpayers rather than the aggregate pattern of tax evasion in the economy as a whole. Thus, it cannot directly provide an answer to the question why tax evasion in one country differs from that in another. It can only provide some indirect suggestions explaining the differences.

Types of Factors Affecting Tax Evasion

Two types of factors affect tax evasion: tax and nontax. The former are factors within the tax system, such as the tax rate, the tax base, the tax structure, the penalty system, the probability of detection, and the probability that the penalty will be applied once evasion is detected. Nontax factors arise outside the tax system and influence the decision of an individual to evade tax. Nontax factors are of two types: micro-nontax factors and macro-nontax factors. Micro-nontax factors pertain to the individual taxpayer, such as sex or educational background, which influence his decision to evade tax. Macro-nontax factors relate to the characteristics of the economy as a whole, such as the price policy and the incomes policy of the economy. Whereas tax factors are more relevant to tax evasion on income derived from legal activities and nontax factors are more relevant to evasion on income derived from illegal activities, both factors have certain influences on evasion of both types of income. On the one hand, tax factors, such as excessively high tax rates and overly progressive tax rate schedules, may constitute a cause for illegal activities from which the income derived is not reportable. On the other hand, nontax factors may constitute a cause for tax evasion on income derived from legal activities. For example, an excessive government expenditure leakage through corruption may influence the taxpayer’s decision to evade tax by underreporting income from legal activities.

Theoretical literature reviewed in this paper deals only with tax factors, but not with nontax factors, while empirical literature deals with tax and micro-nontax factors. Macro-nontax factors are not covered in the literature. Thus, the literature is limited in the sense that it takes account of only a part of the whole range of factors that affect tax evasion.

To summarize, the literature reviewed in Sections I, II, and III is limited insofar as (1) it deals only with tax evasion through underreporting of income; (2) it covers evasion of legal income only; (3) it is a micro analysis; and (4) it takes account of only some of the factors that affect tax evasion.

The Tax Rate

The effectiveness of the tax rate as a policy instrument for the deterrence of tax evasion depends on the relationship between the tax rate and tax evasion, that is, whether the high tax rate is a cause of tax evasion. Unfortunately, such a relationship has not been definitely established one way or the other. On the one hand, it is generally believed that a high tax rate is the main cause of tax evasion. Such a belief is supported by observations, reports, and studies of experts and committees on taxation as well as by statements of tax policymakers. For example, Kaldor (1956), in his well-known report on Indian tax reform, states that the incentive to evade taxes depends on the marginal rates of taxation, because these govern the gains from evasion as a percent of the sum evaded. Kaldor’s statement is reinforced by the Direct Taxes Administration Enquiry Committee (1960) of India, which finds that high personal income tax rates are one of the major reasons for tax evasion. The Committee observes in its report that many witnesses state that the prevailing high rates of taxation are one of the main causes for tax evasion and that the high rates of tax in the high-income brackets are said to be tolerated only because of the considerable evasion that takes place. Moreover, the then Finance Minister of India, in his budget speech, stated that he had come to the conclusion that in India, the tax rate of direct tax at top levels encouraged large-scale evasion. Tanzi (1983a, p. 5), in discussing the erosion of tax bases in developing countries, states that,’High tax rates, of course, make the problem of evasion worse.” Herschel (1978), in commenting on the relationship between tax evasion and tax avoidance, states that both of them can be considered as alternative means of facing extremely high tax rates. This implies that high tax rates are one source of tax evasion. In addition, a simulation study by Friedland and others suggests that beyond some rate of tax, the fraction of reported income becomes very elastic with respect to the tax rate and that the positive relationship between the decision to underreport is increased as the tax rate increases. Moreover, survey data used by Mork (1975) indicate that taxpayers in higher-income brackets who pay tax at higher rates tend to have a higher proportion of underreported income. Furthermore, in an analysis of individual tax returns, Clotfelter (1983) finds that higher tax rates tend to stimulate tax evasion.

On the other hand, as reviewed in Section II. the results of theoretical analyses do not directly confirm the belief that a high tax rate is the cause of tax evasion. Such theoretical results are supported by the experience of tax administrators, tax committees, and some empirical observations. For example, tax administrators in many developing countries believe that, as far as farmers, traders, and professionals are concerned, tax evasion exists at all income levels and not only in the high-income brackets where the marginal tax rate is high. Moreover, the Working Group on Central Direct Taxes Administration of the Administrative Reforms Commission of India (1969), which investigated the causes of tax evasion by studying the amounts of detected concealed incomes, reported that an increase in the rate of taxation was not followed by an increase in tax evasion, nor did a decrease in the rates bring about a higher tax responsiveness. Furthermore, the experience of tax administration in India suggests that lower tax rates do not secure better tax compliance; for example, during 1970/71 to 1978/79 when the marginal tax rates were steeply reduced, income tax collection as a proportion of national income in the nonagricultural sector was also reduced. All these are evidence that there is no relation between high tax rates and tax evasion.

The empirical results for both sides of the argument are sketchy and indirect, which indicate that further such studies are needed before making a definitive conclusion regarding the effect of the tax rate on tax evasion. Thus, one may conclude that the effectiveness of the tax rate as a policy measure for the deterrence of tax evasion is unclear.

The Penalty Rate and the Probability of Detection

Unlike the tax rate, the theoretical result of the effect of the penalty rate on tax evasion is much clearer. According to the tax evasion model, a penalty has a negative effect on tax evasion¹⁹ except when the probability of detection is a function of the penalty; in such a case, the effect of a penalty on tax evasion is indeterminate. For the probability of detection, the message from the theoretical result is even clearer because all theoretical analyses, without any exception, indicate that an increase in the probability of detection is a deterrent to tax evasion. The next legitimate question is, if both the probability of detection and the penalty rate are deterrents to tax evasion, what is their relative effectiveness? A number of theoretical analyses have addressed this issue, which are reviewed below chronologically.

Singh (1973) finds that, at a given proportion of underreported income, there is a negative relationship between the penalty rate and the probability of detection. He also finds that, for India, if the penalty rate is at 200 percent of the evaded income, then the probability of detection must be one third for the proportion of underreported income to be zero, that is, no evasion at all, McCaleb (1976) argues on theoretical grounds alone that the penalty rate and the probability of detection are superior instruments to tax rates in generating increased tax payments through the reduction of tax evasion because the tax rate gives indeterminate results, whereas both the penalty rate and the probability of detection give a definite result. Further, he argues that if the implementation cost is also taken into account, then the penalty rate is superior to the probability of detection. Thus, he draws a general conclusion that only the penalty rate on unreported income is a preferred instrument on the grounds of both certainty of the effect and the cost of implementation. Christiansen (1980) finds that, given a constant expected gain from tax evasion, if the penalty is increased and the probability of detection is adjusted downward accordingly, risk averters will always reduce their tax evasion. This implies that a high penalty rate is a more effective deterrent to tax evasion than the high probability of detection. Further, if the penalty rate is “small enough,” then an increase in the penalty rate will increase tax evasion. However, if the initial penalty rate is “large enough,” an increase in the penalty rate will reduce tax evasion. This indicates that when the penalty rate is initially low, then an increase in the probability of detection is more effective than an increase in the penalty rate. When the penalty rate is high enough, then an increase in the penalty rate is more effective than an increase in the probability of detection.

Koskela (1983b), introducing cost of detection into his model, finds that provided the ratio of the marginal cost of detecting tax evasion is “small,” large penalties seem to be more effective deterrents to tax evasion than high probability of detection. He explains that an intuitive interpretation of this result is that an increase in the penalty rate and a decline in the probability of detection that results in a constant expected gain from a given tax evasion induces a greater loss for the taxpayer in the ease of detection. Greater risk will thereby be involved and tax evasion will decline assuming that the taxpayers are risk averse. Further, he demonstrates that a high penalty rate is a more effective deterrent to tax evasion than large lump-sum fines. This is because the lump-sum fine does not vary with the amount of undeclared income and thus leaves more incentive to risk-taking than the penalty rate, which varies with the declared income. Friedland and others (1978) report that their simulation study indicates that a high penalty rate with a low probability of detection is a more effective deterrent than a low penalty rate with a high probability of detection.

To summarize, at least in theory, both the probability of detection and the penalty rate are effective deterrents of tax evasion and the penalty rate is relatively more effective than the probability of detection.²⁰

Factors Relating to Tax Administration

The Withholding Scheme

Evidence from the Tax Compliance Measurement Program (TCMP)²¹ indicates that the wage and salary income which is under a withholding scheme has a high rate of compliance or a low rate of evasion.²² Moreover, statistics in many developing countries indicate that a sizable amount of income tax is collected from wages and salaries and is deducted at the source. (For example, in Thailand 78 percent of individual income taxes is collected from wages and salaries; the corresponding figures are 73 percent for Cyprus, 70 percent for Indonesia, and 67 percent for Turkey.) This indicates that a tax withholding scheme is an effective means of preventing tax evasion. However, the problem with the withholding scheme is that it cannot be applied to many types of income because, for the withholding scheme to be effective, there must be a relatively small number of easily identifiable payers of the income. Many types of income, for example, rental income, income of professionals, and income from small businesses, do not lend themselves to withholding schemes because there are more payers than receivers of such income. However, capital income, such as dividend and interest paid by banks, does lend itself to withholding schemes because there are a lot fewer payers of such income than receivers. Agricultural income of farmers, whose products are mainly for export or for further processing by domestic factories (for example, sugarcane in Thailand and rubber in Indonesia) also lends itself to withholding schemes. However, care should be taken in introducing withholding schemes on these types of income because if the schemes are not correctly instituted and administered, the withholding tax becomes the final tax and acts more like an export tax or a sales tax on agricultural products rather than as an income tax on farmers. This is because the tax is based on turnover rather than on income and the tax does not vary with the income of the taxpayer. Farmers who experience losses still have to pay this tax.

The Self-Enforcement Tax System

Economists have attempted to design an interlocking tax system in such a way that it will rely on the self-interest of taxpayers to encourage them to provide information to the tax department. Kaldor (1956), in his well-known report on India, suggests that the five taxes, namely, the income tax, the capital gains tax, the wealth tax, the personal expenditure tax, and the gift tax be filed in a single comprehensive return and should be assessed simultaneously. Kaldor explains that the advantage of filing a comprehensive single return and simultaneous assessment is that the taxes are self-checking in character for two reasons: first, concealment or understatement of items in order to reduce liability in some taxes may increase liability in other taxes, and, second, the information furnished by a taxpayer in order to prevent overassessment of his own liabilities automatically reveals the receipts and gains made by other taxpayers. Higgins (1959) carries Kaldor’s idea further by introducing a self-enforcing incentive tax system for developing countries. Higgins’s system includes (1) a personal income tax (including capital gains), (2) a corporation income tax, (3) a general sales or turnover tax, (4) a wealth tax, (5) a tax on excess inventory, and (6) a personal expenditure tax. Theoretically, the Kaldor-Higgins system is self-checking because personal expenditure is defined as the excess of income over savings, and savings are equal to the increase in net wealth. Thus, taxpayers who underreport their expenditure by overstating their savings increase their wealth tax liability. A seller of a property who understates his capital gains hurts the buyer because the buyer cannot claim the full amount of the investment, thereby forcing him to declare higher expenditures and increasing his expenditure tax liability. The excess inventory tax is designed to discourage underreporting of sales, thus helping to enforce sales and income tax. However, the system is not practical and cannot be implemented. Goode (1981, p. 266), commenting on some economic aspects of tax administration states:

Other Administrative Factors

Other tax administrative measures that can possibly prevent tax evasion include the simplification of tax laws and procedures, taxpayer education, the publicizing of names of tax law offenders, and the honesty of the tax enforcing department. These measures are practiced in a number of countries with varying degrees of success.

Nontax Factors²³

As pointed out in Section IV, the theoretical models of tax evasion deal only with micro-analysis of tax evasion and account only for tax factors affecting tax evasion, while some empirical studies under review account for tax and micro-nontax factors but not macro-nontax factors. Such models and studies do not directly or fully explain why tax evasion varies from one country to another. We now consider several macro-nontax factors that can affect tax evasion in developing country circumstances. No attempt is made here to establish the relative importance of the various macro-nontax factors.

Micro-nontax factors are relevant only to the micro-analysis of tax evasion behavior but are not relevant when macro-tax evasion behavior (such as intercountry comparison of tax evasion behavior) is considered because their effects tend to be canceled when the variables are aggregated. Macro-nontax factors are relevant to both micro-analysis and macro-analysis of tax evasion behavior. Since the focus here is on the intercountry comparison of tax evasion, only macro-nontax factors will be explored.

Price Control Policies

Price control tends to generate black markets and tax evasion, irrespective of whether the control is in the goods and services market, the factor market, the money market, or the foreign exchange market. In the goods and services market, price control imposed on any good or service usually leads either to the understatement of sale receipts in the accounts that are prepared for the inspection of the tax authorities, or to the disappearance of such goods or services from the regular market; they can be purchased only in the black market at higher prices. In the factor markets, price control usually leads to the declaration of false prices of the factor being controlled. For example, rent control usually leads to tax evasion either through the understatement of rental income in the tax return of the landlord, or through an unreported amount of premium²⁴ (which is usually substantial) that the landlord receives in cash before renting the property at the controlled rent. In an economy where the minimum wage is artificially set higher than the equilibrium wage, one is likely to find overstatements of wages paid in the tax returns of employers, unreported employment on the part of employees, and even illegal manufacturing plants because the plant owners could not register their plants since all or almost all employees receive salaries lower than the minimum wage. In the money market, the ceiling imposed on the interest rates often drives funds from the officially organized market to the unorganized market because interest rates in the organized market are closely controlled, while it is not possible to enforce such interest ceilings on loans in the unorganized market. The result is that a sizable amount of interest income from loans in the unorganized market is not taxed because it has not been reported to the tax authorities. In the foreign exchange market, if exchange rates are fixed artificially and differently from equilibrium rates, then a black market for foreign exchange is likely to develop. Such a market is not only illegal in itself but it also facilitates the operation of other illegal activities such as drug trafficking, smuggling, and illegal trade in arms and ammunition by making available foreign currencies needed for such activities, which otherwise would be available only through official channels.

Government Rules and Regulations

The preponderance of rules and regulations imposed by the government tends to increase tax evasion. This is because in an economy where there are too many and too complicated rules and regulations governing business practices, it is generally difficult, often not profitable, and sometimes impossible to do business legally. Thus, businesses have to find some ways to get around such rules and regulations and often have to do business illegally. For example, rules laid down to restrict trade—such as import quotas, import licensing, and import prohibitions—often lead to the problem of smuggling, which in itself is an evasion of customs duties and which generates other black markets. Once the black markets are started, tax evasion follows automatically. Other examples of rules and regulations governing business practices are prohibitions, rationing, forced sales of commodities to the government or to the marketing board, town and city planning regulations, factory acts, permits to produce certain goods, and licenses required to start a shop.

Public Sector Salaries

Public officials have certain responsibility and authority which can either be used for providing public services or be abused for private benefit. If the salaries of such officials are not commensurate with their responsibility and authority, the temptation for them to abuse such authority and responsibility is high. Moreover, if their salaries are so low that they cannot maintain a reasonable standard of living, it is likely that they will find ways to supplement their income via legal or illegal means. If illegal means are used by officials other than tax officials, these are likely to have an indirect effect on increasing tax evasion. If such means are used by tax officials, these are likely to have both indirect and direct effects in increasing tax evasion.

Government’s Expenditure Policy

Taxes finance government expenditures, which in turn are expended for the well-being of taxpayers. This is the justification for taxes and government expenditures. However, when taxpayers start to question the justification for either the size or the pattern of expenditures, or both, then they are unwilling to pay taxes, at least in the amount for which they are liable. For example, taxpayers may see no justification for the large size of the budget because they do not see why the government has to be so luxurious, or why they have to pay tax to support public servants who live much more lavishly than they do. Or, taxpayers may disagree with the government’s spending policies such as defense spending or spending on welfare. In such a case, they will be unwilling to pay tax and such an unwillingness leads them to find ways of reducing their tax liability through legal or illegal means. This leads to either tax avoidance or tax evasion. Such evasion is based purely on political considerations.

Other Factors

Other nontax factors affecting tax evasion are, for example, the level of education of the population and the stage of development of the economy. The level of education of the population influences tax evasion in two ways. First, a better educated population is more likely to have a better understanding of the tax system and its purposes; thus, the chance that they will violate tax laws because of ignorance or because not enough attention is paid to the tax system, is reduced. Second, a well-educated population is more likely to know how the tax system works and thus can reduce its tax liabilities through utilization of various relief provisions available in the tax laws, thereby reducing the necessity to evade tax.

The stage of development of the economy can indirectly affect tax evasion. In developing countries, economic activities are conducted in a manner that does not lend itself to easy tax enforcement. In such countries, the agricultural sector is relatively large, the share of the self-sufficient economy in the total economy is high, the share of the nonmonetized economy to total gross domestic product is high, and small and scattered production units are more prevalent. Such circumstances hinder the effectiveness of the tax enforcement mechanism and create greater temptation and opportunity for tax evasion.

It would seem that not only the validity of the above argument concerning tax and nontax factors affecting tax evasion but also the extent to which such factors influence tax evasion can be tested empirically by using the regression technique. Such tests would enable us to explain why and how tax evasion differs between developing and developed countries. However, the absence of a well-defined and generally accepted measurement of tax evasion²⁵ prevents us from making such a test. Thus, we have to rely on qualitative evaluation of such factors and we will do so for the developing countries in the next subsection.

Tax Evasion Under Developing Country Circumstances: Some Explanations

We have investigated both tax and nontax factors that affect tax evasion. We can now evaluate them with a view to shedding some light on the extent of tax evasion in developing country circumstances.

Until recently, the rate structures of the income tax in most developing countries have been highly progressive with respect to the absolute amount of income. Such high progressivity is likely to lead to high tax evasion, since, as has been shown, a progressive tax rate structure encourages tax evasion. The high progressivity of the income tax in developing countries is a result of historical developments in the tax systems of these countries.

The income tax laws of most developing countries are adapted from the tax laws of well-developed societies. As a result, on the one hand, one can find a very sophisticated income redistribution scheme as well as built-in stabilizing measures in most of the tax laws—regardless of the stage of development of the countries. On the other hand, one also finds generous exemptions and relief for the promotion of development. The combination of generous exemptions and relief while there is pressing need for revenue to finance development and income redistribution concerns have led most developing countries to enforce highly progressive income tax systems.

Penalty rates in developing countries are quite varied. Some countries have rather high rates, while others have low rates. In many developing countries, only fixed fines are imposed, which are often out of date. With high rates of inflation, unless fixed fines are changed frequently, they become obsolete very rapidly. Many developing countries do not distinguish between penalties and interest, and charge only penalties without charging interest on the evaded tax. The penalties are charged as a percentage of either the evaded income or the evaded tax without any time dimension. Thus, the absence of interest charged on the evaded tax further weakens the effect of penalties. Since high penalty rates are rarely enforced, the deterrent effects of penalty rates in developing countries are likely to be low and therefore tax evasion is likely to be high.

The probability of detection depends on the efficiency of tax administration. Since tax administration in the developing countries is relatively inefficient, the probability of detection is likely to be low. Thus, tax evasion is likely to be high.

Price control is normally implemented because of equity considerations and is generally designed either to make goods or services that are subject to such control available to the public at the controlled low prices or to guarantee a certain minimum income level (in the case of the price of labor—the minimum wage). Price control is practiced more frequently in developing countries than in developed countries because of the greater need in developing countries to take into account equity considerations in conducting public policy. This is because, in general, personal income is more unevenly distributed in developing countries.²⁶ In the goods and services market, one generally finds that prices of essential goods are controlled either because of shortage of supply due to low productive capacity (in the case of goods domestically produced and consumed), or because of higher prices in foreign markets (in the case of domestically produced goods for domestic consumption as well as for export), or because of the lack of foreign exchange (in the case of imported goods). These occurrences are more frequent in developing countries than in developed countries. Thus, price controls are more prevalent in developing than in developed countries. In the factor market, rent control and the minimum wage are widely practiced in developing countries. In the money market, interest rates in the organized market are relatively inflexible, but the unorganized or parallel money market is very popular, whereas such a market is insignificant in developed countries. In the foreign exchange market, most developing countries use some form of fixed exchange rates, while the exchange rates in the major industrial countries are floating. Furthermore, most developing countries often experience balance of payments deficits and foreign exchange shortages because their foreign exchange earnings depend mainly on exports of primary products that fluctuate greatly either because of weather conditions or because of the fluctuation in the world market prices, over which they have no control. Fixed exchange rates, coupled with a high frequency of balance of payments deficits, lead to a black market for foreign exchange, which in turn leads to tax evasion.

On the extent to which the government regulates the economy through rules and regulations, there are no a priori reasons for stating whether it is more or less widely practiced in the developed or the developing countries. However, in some specific areas, for example, in import control, one can find a higher frequency of controls in developing countries than in developed countries, because foreign currency shortages occur more often in the developing countries.

In developing countries, there is often a large gap between the salaries of public employees and the income level needed for a reasonable standard of living, while such a gap is rarely observed in developed countries. This gap usually leads to either legal or illegal supplementary income. The unex-plainable supplementary income of public officials contributes directly or indirectly to higher tax evasion.

Concerning the level of education and the stage of economic development, it is obvious that these are lower in developing countries than in developed countries. Thus, tax evasion due to these factors is likely to be higher in the former than in the latter.

To summarize, the developing country circumstances are characterized by a highly progressive nominal tax structure, ineffective penalty deterrents, low probability of detection, impaired horizontal fiscal equity, prevailing price controls in all markets, a moderate degree of government rules and regulations, a large gap between public sector salaries and a reasonable standard of living, low level of taxpayer education, and low stage of development. Since such circumstances stimulate tax evasion, it is likely to be higher in developing countries than in developed countries.