Taxation and corruption: theory and firm-level evidence from Uganda

Abstract

This article develops a simple framework to analyse the negotiation over bribe and tax payments during the tax collection process. We show that the larger the bribe a firm offers to a tax collector, the larger the tax rebate it gets. More particularly, we show that the negotiation over bribe and tax payments hinges on four other factors: firms’ official liabilities, detection, firms’ negotiation power and red tape costs imposed on firms. Some of the predictions from the theoretical model are tested using firm-level data from Uganda. We find that bribe and tax payments are inversely related, thereby supporting the hypothesis of a negotiation taking place between firms and tax collectors. In particular, a 1% point increase in average bribe payments per employee is associated with a 7% point reduction in average amount of tax payments per employee. Results are robust to various instruments dealing with the endogenous relationship between bribes and taxes.

Keywords:

• corruption
• tax evasion
• tax administration
• red tape
• bargaining power

JEL Classification:

• D73
• H21
• H26
• H32
• D82

Acknowledgements

We thank an anonymous referee and seminar participants at the various conferences and seminar series for helpful discussions. All errors remain our own.

Notes

¹ See Svensson (2005), Mishra (2006), Banerjee et al. (2012) and Olken and Pande (2012) for reviews of the literature on corruption. See Slemrod and Yitzhaki (2002) for a review of the theoretical literature on tax evasion.

² For example, Sequeira and Djankov (2008) show that bribes amount to 14% of shipping costs in the port of Maputo and 4% in Durban and argue that this difference is due to officials having more discretionary power in the former location than in the latter. For a discussion and examples of tax collectors’ discretionary power over the private sector, see, for instance, Tanzi and Davoodi (2002) or Rose-Ackerman (2004).

³ Bertrand et al. (2007) consider the role of negotiation power in the context of licenses provision.

⁴ Tax rates and wages could also be endogenized. See, for instance, Hindriks et al. (1999) on tax schemes or Besley and Mclaren (1993) on wage schedules.

⁵ Tax collectors’ capacity to extract rents may be associated with different factors: a tax collector’s degree of dishonesty, the ability to interpret complex and confusing tax laws to negotiate larger bribes (Tanzi and Davoodi, 2002), firms’ capacity to pay (Svensson, 2003), or the level of discretion over the task performed (Shleifer and Vishny, 1993; Sequeira and Djankov, 2008).

⁶ Gauthier and Goyette (2012) examine the theoretical implications of bureaucrats’ heterogeneity in various administrative settings.

⁷ As in Besley and McLaren (1993), we assume that tax collectors have perfect knowledge of a firm’s profits and tax liabilities. Assuming uncertainty over firms’ profits, firms could report lower profits and would hence pay lower bribes and tax payments. This would imply that our empirical results are a lower bound on the effect of corruption on tax evasion. In the empirical section, we explain how we deal with these measurement issues.

⁸ In a situation where the firm can be audited by another public employee after the bribe and tax payment negotiation, we can include a fine, A, in Equation 1. This does not qualitatively affect our results.

⁹ The model could easily be amended so that the outside option of a tax collector is different from zero. This would only affect the level and not the qualitative nature of the results.

¹⁰ Firms’ bargaining power could arise from the extent to which the capital of a firm is sunk, as less mobile firms could be forced to pay higher bribes (Svensson, 2003). Other sources of bargaining power could be associated with the age of a firm, foreign ownership and exporting status. We discuss these variables in the empirical analysis.

¹¹ There exists a corner solution where T_pi = 0. However, in what follows, we focus on an interior solution. Indeed, we could observe positive amounts of taxes paid because tax officials fear triggering their superiors’ suspicion if not enough taxes were collected, or because of some intrinsic motivation, guilt, etc. The data does not allow testing for these hypotheses, so we leave this for future research.

¹² The added complexity of a convex cost function does not affect the qualitative results below.

¹³ See, for instance, Svensson (2003), Gauthier and Reinikka (2006), Fisman and Svensson (2007) and Goyette (2014, forthcoming).

¹⁴ The 1998 enterprise survey specifically asked firms about taxes paid in 1995 through 1997.

¹⁵ For instance, the CIT rate has remained at 30% and the VAT rate, which was at 17% in 1997, is now at 18%. For both taxes, the tax threshold has remained at $50 000.

¹⁶ See Svensson (2003) and Reinikka and Svensson (2006) for details on the strategy to collect corruption data.

¹⁷ Other taxes include import duties, withholding tax, presumptive tax on small businesses, local property tax, etc.

¹⁸ See Gauthier and Reinikka (2006), for further details.

¹⁹ Svensson (2003) shows that there is no observable evidence that the sample of firms reporting bribes is not representative.

²⁰ Some entrepreneurs may have underreported their level of sales used in constructing tax obligations. In such a case, the constructed tax obligation would be a lower bound on the true measure of taxes officially owed by firms. On the other hand, tax payments are prone to measurement errors in the opposite direction, i.e., taxes paid are likely to be over-reported by entrepreneurs by fear of reporting evasion (see Fisman and Svensson, 2007). If the true level of sales is higher than reported and the true level of taxes paid is lower than reported, then our results on the effect of tax evasion on bribes are a lower bound.

²¹ Comparing the 1998 data set with the 2006 Uganda World Bank Enterprise Survey (WBES) data, we note that the rate of tax evasion is relatively similar at about 50% of sales. For the subset of African country firms in the overall 2006 WBES data set, on average firms declare only about 57% of their sale values for tax purposes.

²² Results using tax per sales, bribe per sales and tax obligations per sales.

²³ Variables could have been scaled in terms of sales or profits. However, the number of employees is less prone to misreporting by firms than sales or profits.

²⁴ There are various reasons why the probability of detection would vary among firms, in particular geographical location, sector, types of tax inspector, etc.

²⁵ The calculation is as follows. The coefficient on taxes paid in Table 2 Column 1 is −0.009. This implies that a $100 decrease in taxes paid per employee, which corresponds to a 7.4% change from the average amount of taxes paid per employee ($1355), is associated with an increase of $1 in bribes paid per employee, which corresponds to a 1.4% increase from the average bribe per employee ($71).