Tax Return Data Examples

Experience shows that the best data preparation approach is to rely on tax return data and supplement it as necessary. Note that any effort to supplement the core tax return data will require additional aggregate or micro data. The examples discussed in this section of the guide are in order of increasing amounts of additional data. Each of these examples have been encountered in development of microsimulation models over that past few years.

The first example assumes the tax return data are complete (so that they can be used to compute the tax liability of each filer) but are too large. No additional data are required to do the stratified random sampling that yields a smaller input data file that produces essentially the same tax results as when using the full input data file as model input. This problem was encountered when developing a model for Malaysia.

The second example assumes the tax return data are incomplete in that the tax administration system did not collect micro data on some types of income for a subset of taxpayers. This example shows how the missing incomes can be imputed to each taxpayer in the subset if we have two bits of information for each missing income: the number of taxpayers in the subset with zero income amounts and the total amount of missing income in that subset. This problem was encountered when developing a model for Albania.

The third example assumes the tax return data are complete but inadequate to support analysis of reforms (such as a refundable tax credit) that will expand the number of tax filers. In this case, we need income survey data to combine with the tax return data. This kind of problem was considered in Task 1 of the recent experimental work conducted with Malaysia data.