Have you looked at earlier studies that have measured the amount of grey area deductions taken as individual taxpayers switch from a refund position to tax owed?

In general, an often applied approach for measuring (corporate) tax agressiveness is to look at either book-tax differences or effective tax rates. Frank et al (2009) [The Acc Rev] have another measure. All have the following in common: You take the dependent variable (e.g. effective tax rate of a firm) and control for all factors that would logically explain the differences; what is unexplained is the measure for tax agressiveness. This residual can be related to variables of interest. [Note that this can be done in a separate or in the same regression as the one that is used to determine the residual (unexplained) part.] You need to control for variables that logically explain low effective tax rates such as the amount of R&D investment, loss carryovers etc. Many examples can be found in the tax aggressiveness literature.

Research contributes to empirically resolving this issue by estimating the relation between aggressive tax planning and the frequency of alleged accounting fraud.

https://msbfile03.usc.edu/digitalmeasures/clennox/intellcont/JAR2013-1.pdf