Predicting Aggregate Taxpayer Compliance Behavior

The IRS seeks to be able to estimate the impact that its service and enforcement interventions have on the voluntary compliance of taxpayers. A key method for doing that is to conduct field experiments among taxpayers as they respond to their real tax obligations. A typical experimental design includes tracking the behavior of both a test group and a control group, which are sufficiently identical in all relevant respects except that the intervention is applied only to the test group. We believe, however, given the nature of most IRS interventions and the diversity and geographically non-uniform distribution of the taxpayer population, that it will often be very difficult to construct a control group that is sufficiently identical to a test group in all relevant respects, except for the intervention (or in preventing the control group from being affected by the intervention). The goal of this research, therefore, was to develop the capability to control statistically for factors that influence taxpayer behavior, supplementing the role of control groups in future field experiments. To do this, we develop econometric models to predict aggregate reporting behavior among individuals. That is, we seek to estimate what taxpayers would have done in the absence of an intervention. The difference between what they would have done and what they actually did in any given time period is a measure of the change in voluntary compliance. If that measure of behavioral change improves as a result of an intervention, then the intervention is considered to have improved compliance.