On tax evaders and corrupt auditors

Abstract

This paper takes cognizance of existence of bribery-type corruption in the tax system. The study seeks to analyze the strategic interaction between a tax evader and a corrupt auditor within a given tax situation. An equilibrium bribe rule is derived for the situation where a tax evader comes face to face with a corruptible auditor. In our stylized model, situations are found where both parties adopting the agreed bribe rule, benefit from cheating the system In addition, the existence of an equilibrium point is established in a two-person fixed threat Nash bargaining situation. A comparative static exercise brings out some intuitively appealing findings. Lastly, when the probability of a super audit is internalized within the system, besides other results, we find that, in a bribe situation, the effect of neither the increase in penalty rate for evasion nor the vigilance over the activities of a tax auditor on their behavior is straightforward. Rather they are ambiguous or at best situation specific. From the policy point of view, we find that if the government is successful in keeping the value of the expected probability of super audit above some value q_o, bribery will vanish from the system. The age old method of the carrot and stick policy to obtain a desirable state also gets established in our setting. Bribery can be made unprofitable by adopting an appropriate carrot and stick policy towards the corruptible auditors.

Keywords:

• tax evader
• corrupt auditor
• bribe
• super audit
• equilibrium point
• comparative statics

Acknowledgements

This is a revised and updated version of a chapter of the author's PhD thesis (unpublished). The author gratefully acknowledges the University Grants Commission for the Research fellowship during his PhD program and the Centre for Economic Studies and Planning, School of Social Sciences, Jawaharlal Nehru University, New Delhi for inviting the author to a three-month visiting fellow program, January–April 2006, during which this paper was completed.

Notes

1. Polinsky and Shavell (2001), Hindriks etal. (1999) had looked into corruption as comprising of bribery, extortion and framing. Our study of corruption is restricted to bribery wherein a tax evader bribes a tax auditor to mitigate the actual penalty due to him for under-reporting his income.

2. Cowell (1990) provides a good survey on tax evasion literature.

3. A significant and growing literature on economics of tax evasion and corruption has occurred since the pioneering work of Virmani (1987). Notable works on this line are Goswami etal. (1991), Besley and McLaren (1993), Mookherjee and Png(1995), Hindriks etal. (1999), Polinsky and Shavell (2001). Quite a comprehensive overview of economics of corruption is provided in Mishra (2004).

4. By interior solution, we mean the existence of an equilibrium solution (α∗, e∗) where 0 < α∗, e∗ < 1; α being the proportion of income declared by the TP and e being the proportion of income of the TP that TA exposes.

5. We have not really delved into the revenue and the distributional aspect of tax policy as we have tried to restrict our analysis to unfolding the behavior patterns of the TP and a corruptible auditor in a bribe situation. Hindriks etal. (1999) has dealt with tax revenue and distributional issues of tax policy in a corrupt system. Polinsky and Shavell (2001) analyses the issue of corruption in a general welfare framework.

6. Basu etal. (1992) deals with situations where the SAs are also corruptible.

7. The declared income αY of the TP could consist of two components: taxable and avoidance/tax shelter income. The latter income is that portion of his income that gets preferential tax rates for it. Thus, tax avoidance in contrast to tax evasion is the legal reduction in tax liabilities that occurs by taking advantage of various provisions of income tax codes. Our present concern is with the tax evasion issue, not denying the fact that a very significant leakage from expected tax revenue takes place through schemes of tax avoidance. For simplicity, all of αY is taken as taxable income, here.

8. It is the probability that the auditor will get to know the true income and secure evidence for successful prosecution. If p is the random probability of audit of a TP, then p may be considered as p  – δ(C) where C is the chosen concealment scheme by the TP. Once a TP has decided on a particular C, δ(C) is fixed and hence for a given p , p gets fixed.

9. In the later section we also consider the case when q is a function of the exposure level, e.

10. In our model, under-exposing is considered to always be a deliberate act (and not an act of error and omission) on the part of the TA.

11. The sanction and reward structure in our model is quite similar to that of Singh, (1995), Mookherjee and Png (1995), Hindriks etal. (1999), Polinsky and Shavell (2001).

12. W may be looked upon as the reservation wage of Besley and McLaren (1993) but in our paper W has no role in the decision process of the TA. Here, the penalty on the auditor for corruption is not related to W. For papers dealing with wage and remuneration of corrupt auditors refer to Mookherjee and Png (1995), Besley and McLaren (1993), Wane (2000).

13. β could be positively correlated with the degree of cynicism against the government and negatively with a well-entrenched ethic of loyalty and honesty within the civil service (McLaren, 1996). In our setting, β is an exogenous variable and it represents the probability of being audited by a corruptible auditor in the case of an audit. Our corruptible TAs bear the same characteristics of dishonest tax inspectors of Besley and McLaren (1993).

14. It is assumed that, during the bargain, both parties take due account of the possibility of occurrence of super audit.

15. The expected pay-offs of the parties in the bargain will be different for the two cases.

16. Here, the bribe transaction is not undone in contrast to that of Polanski and Shavell (2001). In our setting, in the super audit case, if e < 1 is found it is presumed that bribery has taken place. But the bribe amount B is still a privateinformation known to the TP and his TA. The penalty structure of corruption in our model has an inbuilt mechanism to take account of the unknown bribe amount in the hand of the TA. Also in our case, the fine for corruption for the TP as well as TA has not been related to their ability to pay, as our fine is not restricted to monetary sanctions but imprisonment could also be included.

17. A > π implies that the government considers the dishonest act of the TA to be more grievous and detrimental than that of the TP.

18. It is another matter that q ≥ ½ seems to very high for practical purposes.

19. r _o is similar in spirit to r _B of Polinsky and Shavell (2001).

20. Polinsky and Shavell (2001) found optimal fines for bribery and framing to be maximal. In their case the maximal fines are bounded by their initial wealth. A in our case is not bounded. It could comprise a fine plus imprisonment.

21. Increases in π and A are to be within the constraint of .

22. It is assumed that the agents in our setting will always go for a deal wheneverthereis a positive expected gain from it, irrespective of the magnitude of the gain.

23. There are two similar situations even under (a2). But from the TP's EY viewpoint, those two situations are exactly identical and hence not considered here.

24. It may be pointed out that the TP's seemingly honest behavior in such situations is just by default.

25. Here α∗ is the optimal level of income declaration of the TP and e∗ is the optimal exposure level of a corrupt TA. This type of situation arises because the TP in his calculation of EY takes due account of the possibility of his TA being honest as well and the possibility of getting super audited etc, whereas e∗ is the optimal exposure level of a corruptible auditor in a bribe scenario.

26. A higher p and q make the cost of evasion rise and hence the TP raises his α level.

27. EY for the case where p = 1 will be

28. Here, for simplicity, we have taken C(Y _H ) = 0.

29. Z = (1−q) (1−r) + q′{(1 + 2r)(1−e) + 3r(e−α)}.