Walking Inside the Potential Tax Evader’s Mind: Tax Morale Does Matter

Abstract

We conduct an empirical study on the determinants of the psychological costs of tax evasion, also known as tax morale. As a preliminary step, we build a model of tax evasion including non-monetary considerations, show the relationship between tax compliance and tax morale. In the empirical analysis of tax morale we find, using a binomial logit model, that the justification of tax evasion can be explained by the presence of grievance in absolute terms (those who feel that taxes are too high, those who feel that public funds are wasted, and those who accept underground economic activities); and grievances in relative terms (the suspected level of others’ tax evasion). The sense of duty and the level of solidarity are also relevant factors, but to a lesser extent.

Appendices

Annex 1: A Model Integrating Psychological Costs Associated with Tax Evasion

In this annex we will introduce tax morale considerations inside a traditional model of tax evasion, to provide a clear theoretical linkage between the two notions. We propose as reference a notion of tax evasion based on the Allingham and Sandmo (1972) model, because this was used as starting point for subsequent models. We will depart from it by including two new elements in the utility function to take into account the impact of non-economic effects.

The original baseline Allingham and Sandmo (AS) model corresponds to a decision under uncertainty about whether or not to evade, reached by maximizing the utility function. The potential evader has a level of income Y legally submitted to taxes, which is known only by him. He considers declaring an amount X, lower than Y. If the tax rate is t and he or she is not caught out, he or she will save an income equal to t(Y − X). This decision is made under uncertainty, as there is a probability p of being audited. If caught cheating, the evader will have to pay the undeclared amount Y − X, at a penalty rate π which is higher than the tax rate t. The potential evader tries thus to maximize the following expected utility function:

$$ E\left[ U \right] = \left( {1 - p} \right)U\left( {Y - tX} \right) + pU\left( {\left( {Y - tX} \right) - \pi \left( {Y - X} \right)} \right) $$

(9)

Our proposal is to modify the AS model to include what are strictly non-economic considerations that a taxpayer can take into account before deciding whether to cheat or not, and to what extent. To be honest about the original contribution by Allingham and Sandmo (1972), it has to be stated that the authors also considered an alternative model which included a “reputation factor” affecting the utility function. Nevertheless, the authors use the simplified version of their model to extract their main theoretical propositions.

The “reputation factor” was incorporated in the AS model in an analogous way as we propose with the last term in Eq. 10, the factor λ.

$$ E\left[ U \right] = \left( {1 - p} \right)U\left( {\left( {1 - t} \right)Y + t\left( {Y - X} \right)\left( {1 - \alpha } \right)} \right) + pU\left( {\left( {1 - t} \right)Y + t\left( {Y - X} \right)\left( {1 - \alpha } \right) - \pi \left( {Y - X} \right) - \lambda } \right) $$

(10)

In comparison to Eq. 9, we break down in Eq. 10 the total income between net income and payment to the inland revenue. Factor λ corresponds to the disutility generated by the reputation loss produced when the tax evader is discovered by the tax inspectors. It is estimated ex ante by the potential evader and incorporated into the maximization rule.

It can be noticed that Eq. 10 differs from Eq. 9 also in the element (1 − α), which is present in both terms of the equation. α refers to the psychological cost, which is more directly linked to what authors identify as “tax morale.”

From Eq. 10, we can calculate also what the conditions of equilibrium associated with full tax compliance are, where the quantity declared for tax purposes corresponds to the actual personal income: X = Y.

$$ \mathop {\left. {\frac{\partial E\left[ U \right]}{\partial X}} \right|}\nolimits_{X = Y} = - t\eta \left( {1 - p} \right)U^{\prime } \left( {\left( {1 - t} \right)Y} \right) - \left( {t\left( {1 - \alpha } \right) - \pi } \right)pU^{\prime } \left( {\left( {1 - t} \right)Y - \lambda } \right) < 0 $$

(11)

Equation 11 can be rewritten as

$$ p\pi < t\left( {1 - \alpha } \right)f\left( \lambda \right) $$

(12)

with f(λ) = β, and β equal to

$$ \beta = p + \left( {1 - p} \right)\frac{{U^{\prime } \left( {\left( {1 - t} \right)Y} \right)}}{{U^{\prime } \left( {\left( {1 - t} \right)Y - \lambda } \right)}} $$

(13)

Concerning the relationship between λ (the reputation factor) and β (which appears in the main text as f(λ), it is easily seen that if λ takes value 0 (absence of feeling of shame), then β takes value 1, and this factor disappears from Eq. 12. If the reputation factor influences the taxpayers’ decisions (λ > 0), then β < 1 as far as we accept that the marginal utility of income is decreasing, which is a common and reasonable economic assumption. The presence of a reputation factor diminishes the utility of each tax income evaded.

We can identify several reference points by giving specific values to the factors α and λ which are interesting for our purposes in Eq. 12. In the first case, we eliminate all references to non-pecuniary considerations in the maximization exercise. This happens when α is 0 (total absence of tax morale) and λ is zero (taxpayers do not care about reputation if caught cheating). In this case, we reach the basic findings obtained by the AS baseline model, where the taxpayer fulfills correctly his tax duties as far as the expected tax sanction on undeclared income (pπ) is at least equal to the loss of regular income due to taxes (t). If α remains equal to zero, but λ > 0, the threshold of tolerance to cheating behavior is reduced, all other things being equal. This result is similar to the alternative AS model which included the “reputation factor.” If we take into account ethical concerns (0 < α < 1), keeping λ > 0, the conditions of equilibrium in this case are those already presented under Eq. 12. As α is positive but smaller than one, the marginal benefit associated with tax evasion when taking into account tax morale and reputation factors is lower than when guided only by reputation factors. Again, the space where tax evasion yields net utility to a potential tax evader is still reduced t(1 − α)β < tβ < t. Finally, if tax morale concerns are completely dominant (α = 1), this means, as already stated, that the additional consumption generated by tax under-reporting does not produce any increment of utility at all, as the punishment feeling completely overrides the satisfaction of additional revenues. In this case, Eq. 12 becomes: pπ < 0, where there is no compatible tax evasion threshold under a full tax morale assumption. Even if the probability of being caught were null, there is no place for a cheating behavior under full tax morale beliefs, as the cheating space under the last assumption needs a negative value for pπ, a solution which is not feasible in the real world.

The model proposed in this article provides some theoretical support for a number of empirical puzzles and findings concerning the determinants of tax evasion behavior or, conversely, of tax compliance. We can consider three groups of taxpayers, according to their attachment to tax morale considerations. The first group of taxpayers never evade their fiscal burden, independently of their income, even though it is easy for them to underreport income because there are low or even negligible chances of being caught. The model also captures the behavior of a second group of taxpayers who take into account the effects of tax evasion on their tax morale feelings and/or on their reputation. Individuals in this group can enter the tax evasion territory if the difference between the economic benefits and costs is higher than the anticipated psychological costs. Whatever be the truth of this, they will evade a lower amount of tax than if tax morale considerations were completely absent. Finally, there is a third group of individuals who only take into account the probability of audit and the penalty they will pay if caught, as against the potential profit generated by the under-reported income, to decide whether or not to hide all or a share of their taxable income from the tax administration. This last group of individuals generally concentrates the attention of researchers into tax-evasion behavior. In contrast to this, we have concentrated our analysis on the first group of taxpayers, who never cheat because tax morale concerns override any economic potential benefit from tax evasion.

Annex 2: Questionnaire

We would like to express our sincere gratitude for your participation and completion of this survey.

• 1. Foreword

To ensure the scientific nature of this survey the identity of individual respondents will remain anonymous.

• 2. Answering the Questionnaire

For this survey we have formulated a series of questions about a variety of subjects.

• Section A. We request respondents to formulate an opinion and answer the questions (Yes/No). It is essential that the respondent’s opinion be expressed honestly (remember the survey is anonymous). A respondent may choose not to answer a question if they feel uncomfortable with the question or response. It is better to leave a question blank than to respond untruthfully.

• Section B. We ask respondents to answers the questions by estimating a corresponding percentage of the population that you think agree/disagree to each issue in the questionnaire. We understand in most cases you will not have a clear idea of the right population percentages. In these cases you are to answer with percentages you believe reasonable. Percentages are to be awarded in graduated values of 5%, ranging from 0 to 100% (0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100%).