The role of costs in tax evasion : non-selfish attitudes or percuniary motivations?
In this thesis we examined the role of non-selfish attitudes in determining tax evasion. We contrasted the argument of the standard approach according to which tax payers choose the amount of income to declare (being motivated merely by monetary considerations) with the idea that some individuals may be inherently honest and would never evade even if it paid them to do so. The first two chapters set the scene for the ideas we developed in the rest of the thesis. We reviewed the standard portfolio model, and drew our attention to three predictions that appeared to be inconsistent with empirical evidence: a) A rise in the tax rate causes a decrease in the amount of tax evasion. This result is counter-intuitive and has little empirical support. b) If all agents maximised expected utility, it is difficult to explain why some evade and others do not. Assuming fine rates and detection rates to be the same across all agents, different degrees of risk aversion explain different amounts of tax evaded but do not explain why some people evade and others do not. c) There is a general view that, given actual fine rates and the probability of audit, the amount of evasion undertaken is too low, in relation to what the standard model predicts. This idea is not to our knowledge fully argued, in that it is unclear exactly how much evasion the standard model does predict. Also it is unclear how large empirical estimates of evasion actually are in the context of the portfolio model. Actual evasion must be compared with the amount of income that an agent is able to hide from the tax authority. We considered the economic literature on the role of morals in tax evasion together with some empirical evidence. In particular, we focused on the Myles and Naylor model which considers tax compliance as a social custom, and on Bordignon's analysis, which examines the role of fairness considerations in deterring tax evasion. In Myles and Naylor's model the issue of the effect of an increase in the tax rate on tax evasion is not completely resolved. It is possible to get an increase in tax evasion after an increase in the tax rate, but this will occur only because those who were previously honest start to evade, i.e. more people become evaders. Existing evaders would still decrease tax evasion after an increase in the tax rate, as in the Yitzhaki model. Hence the Myles and Naylor analysis offers only a partial resolution to this puzzling prediction. Bordignon's analysis is more satisfactory in this respect: in his model people can end up evading more, when the tax rises, even when they were previously evading. This happens when individuals perceive the increase in the tax rate as unfair and consequently are prepared to evade more than previously. Hence the puzzling prediction concerning the effect of a change in the tax rate is not completely resolved by the models on social interactions. We also pointed out that the argument that some people are inherently honest is vulnerable to evolutionary forces. Once we allow preferences to change and selfish attitudes to be present in the social environment, altruistic behaviour can be undermined by an increased tendency to "selfishness". Furthermore the simple arguments we gave suggested some difficulties in explaining how "honest" and "utility maximising" individuals could coexist in equilibrium. This general argument is not new to economics, having appeared as long back as 1959 in Tullock's famous paper on voting. The argument that some people may be inherently honest and not evade even if it would pay to do so is not completely satisfactory. In the models we considered people are assumed to be motivated by non-selfish considerations, some sort of predisposition to altruistic behaviour seems to be taken as given, rather than explained. There has been a large amount of literature that suggests some form of "altruism" motivates economic decisions. However the models and evidence that point in this direction suggest that the stability of altruistic or co-operative behaviour relies on some form of reciprocity. This argument for altruism is weaker in the context of tax evasion, however. Tax evasion is by its nature a hidden activity and individuals cannot observe each other. Moreover a single taxpayer can be reciprocal only with an anonymous group. Hence reciprocity considerations and punishment of cheaters are likely to be a lot less effective than in many of the cases studied. In the light of this, the question of why individuals should consider tax compliance as part of a social custom remains an open question. Empirical evidence on this issue is rather thin, in that most of empirical studies do not consider this question directly. From the studies we discussed, the evidence on the role of morals is not compelling. More compelling is the role of opportunities and costs of evasion in determining tax compliance. Curiously this issue has been neglected by the literature in the standard portfolio model. Here the only cost for evading is the fine, which is incurred only in case of detection. We explored the idea that costs or opportunities rather than willingness to evade might determine evaders' behaviour: some people may not evade simply because it is too costly or because they do not have the opportunity to do so. By introducing a minor modification to the standard approach, to allow for a cost element in evasion, we get some different results. In particular we show that it is possible to predict an increase in tax evasion after an increase in the tax rate, without relying on any particular psychological motivation. It is also possible to get a mixed outcome of evaders and non-evaders, even if individuals have the same income and face the same probability of detection. Typically, evaders are less risk averse than nonevaders. Different degrees of risk aversion lead to different choices on whether or not to evade. This was not a characteristic of the standard model. In Chapter 5 we analysed the optimal audit policy when taxpayers have different opportunities to evade. The analysis can be extended in various ways: 1) by considering different sizes for the two groups of tax payers; 2) by allowing individuals to choose the group to which they belong. This pursues the idea developed by Cowell (1985). 3) The choice of the tax rate could be endogenised, to allow more interaction among the taxpayer, the tax authority and tax enforcement agency. This would obviously require an explicit social welfare based approach. 4) It is also possible that the government could directly affect evasion by controlling the cost of compliance. In particular it could be interesting to examine how much of the compliance cost should be borne by the government and how much by the individual. These are directions for possible future research. Note The last chapter of the thesis is joint work with Thibaud Vergé at CMPO.