Misperceptions About Tax Audits

For some entities, the utility-maximizing evasion rate depends substantially on tax audit features, such as audit probabilities and penalty rates. Bergolo et al. (2017) document large misperceptions about these features. In this paper, we expand the analysis of survey data to explore potential sources of these misperceptions. Of all the channels that we explore, recent contact with audits best explains differences in misperceptions.

For some entities, such as self-employed individuals reporting income taxes or firms reporting value-added taxes, the optimal evasion rate depends substantially on audit features like audit probabilities and penalty rates (Allingham and Sandmo, 1972).
Whereas it is easy for firms to find other important information such as inflation rates or exchange rates, it is difficult to find information about the probability of being audited and penalty rates. Indeed, Bergolo et al. (2017) show evidence that firms have large misperceptions about these audit features. 1 In 1 Previous studies have used surveys to measure misperceptions about audit probabilities (e.g., Harris et al., 1988;Hessing et al., 1992;Sheffrin and Triest, 1992). The evidence from Bergolo et al. (2017) has two main advantages. First, compared to previous surveys that use samples of the general population, Bergolo et al. (2017) focus on a population whose stakes of misperceiving the auditing process can be substantial. Second, Bergolo et al. (2017) combine survey and internal administrative data from the tax agency, which provides a more straightforward comparison of perceptions and reality. this paper, we expand their analysis to explore the sources of these misperceptions.

I. Data
See Bergolo et al. (2017)  in penalties." After each question, we elicited certainty in the response using a 1-to-5 scale, from "Not sure at all" to "Very sure." To estimate the "actual" audit probability and penalty rate, we use administrative data from the IRS for the sample of firms invited to the survey. The actual audit probability is calculated as the percentage of firms that were audited at least once in 2011-2013. The average penalty rate is calculated among firms who were caught evading in 2011-2013. Moreover, there is a systematic positive Bias: the average perceived probability (39.50%) is substantially higher than reality (7.98%), with a statistically significant difference (p < .001).  One potential explanation for this overestimation of audit probability is given by the availability heuristic model (Kahneman and Tversky, 1974): audits are arguably salient events and thus seem more frequent than they are.

III. Heterogeneity Analysis
In this section, we present some heterogeneity analyses to explore the sources of these misperceptions. Table 1   and firms that did not. By construction, the 3 For instance, among firms that report an audit probability of 50%, the fraction that report "Not sure" or "A bit sure" is 50.92%; in comparison, among firms that report an audit probability different from 50%, the fraction that reports "Not sure" or "A bit sure" is 29.06%. This difference is statistically significant (p <.001).
Bias and MAE are higher for firms reporting 50%. However, the misperceptions are still substantial even after dropping responses of 50%. In other words, although this issue with the survey responses may inflate the degree of misperceptions, it is far from explaining the whole puzzle.

IV. Conclusions
We present measures of perceptions about the auditing process in a sample of 6,181 firms from Uruguay. We find large misperceptions about audit probabilities and penalty rates. We also find a systematic overestimation of audit probabilities but no such overestimation for penalty rates. Of all the channels that we explore, recent contact with audits best explains differences in misperceptions.