5. Discussion
The present analysis responds to the long-lasting call to examine “whether there is any specific deterrent effect of an audit and to uncover the reasons for the presence or absence of such an effect” (Andreoni et al., 1998, p. 844). It also responds to calls from multilateral organizations to use standardized country-level data in order to evaluate the effectiveness of enforcement strategies employed in practice (IMF, 2015; OECD, 2010). Building on behavioral theories of tax compliance, we hypothesize that compliance increases until a certain auditing level is reached, and decreases beyond that level. Using country-level data from 2003-2014, we obtain strong support for a U-shaped association between the auditing level and tax evasion. In lay terms, this implies that more auditing is not necessarily better. Seen from a different angle, tax authorities may benefit from trusting taxpayers to some extent, as high levels of distrust may be undesirable (Mendoza and Wielhouwer, 2015). In particular, tax authorities may have incentives to investigate how taxpayers perceive different auditing levels. Over two decades ago, researchers were already finding evidence in favor of a positive – yet decreasing – association between auditing and compliance, which raises the question of whether there are limits on what auditing can achieve (Alm et al., 1992). To the best of our knowledge, however, this is the first study to examine a possible U-shaped association between auditing and tax evasion. An important point of discussion is whether the measures employed in this study have an adequate degree of validity and can be considered as “credible” (Alm, 2012; Slemrod and Weber, 2012). One advantage of the auditing level measure is that it is based on hard (externally verifiable) data. The disadvantage, however, is that with the available data it is not possible to distinguish between different types of verification actions. It would be ideal to control for the fraction of verification actions that are, for example, random vs. targeted, assigned to firms vs. individuals, uniquely vs. repeatedly targeted at the same taxpayer, or intrusive and inquisitive vs. educational and supportive. Controlling for these attributes would shed light on why or when a possible backfiring effect may take place.
