For the variables in Tax–person (such as taxinc and fedtax), pre-2010 values include topcodes, while post-2010 values seem to have top codes replaced with their rank-proximity switched value (in accordance with the new topcoding procedure). There do not seem to be swapvalues or topcodes (Larrimore 2008) for the Tax-person variables. As such, how should researchers include variation in these top incomes as is seen for post-2010 values? Does the census bureau tax model have a way to deal with this discontinuity?
Sorry for the delay in responding to this question. We were discussing this with some of our contacts at the Census Bureau. A few notes may help clarify this issue. First, IPUMS CPS doesn’t do anything to the CPS tax variables when we get them. We simply report what is provided to us by the Census Bureau’s tax model. Second, the Census Bureau takes the already top-coded income values and runs these through their tax model. This is an important distinction between the Census Bureau running the non-top-coded income values through the tax model, and then top-coding the tax values. Therefore, the top-codes are essentially baked in to the values for the tax variables.
I hope this helps clarify things. If you have any additional questions feel free to email ipums@umn.edu.