I have noticed that fedtax and statetax both have negative values. This doesn’t seem to make sense to me, as these variables amount for taxes before any credits so they should always be positive. Could you tell me why are there negative values? Thank you
In 2004, the Census Bureau introduced a new tax model to simulate the federal and state tax liabilities for CPS respondents. This new model permitted tax credits to be refundable, specifically the additional child tax credit and the EITC, which allowed tax liabilities to go negative for the first time. In other words, the values of FEDTAX for 2004 are not truly a measure of before-credit tax liability.
Accordingly, all of the negative values for FEDTAX, and nearly all of the negative values for STATETAX, occur in 2004. In 2005, the Census Bureau began to report before- and after-credit tax estimates separately with refundable tax credits being encompassed by the FEDTAXAC variable instead of FEDTAX. We will update the online IPUMS-CPS documentation for FEDTAX to note this difference in 2004 more clearly.
Hope this helps.
Thank you for your answer. Things are a bit clearer now, but I still have some questions. In many of the cases that fedtax and/or statetax is negative the values of both the child credits and the EITC are zero, which means that there were no deductions, unless there are other besides these two. So how can the tax be negative in this case?
Also, is there a way to make the two series (fedtax and statetax) positive and consistent over time? In principle I would add the tax credits for the years where the series already include the deductions (2004 and 2005), but given what I noted above I don’t think this would solve the problem.
For those respondents with a negative FEDTAX who did not receive the EITC, the refundable Additional Child Tax Credit must have been responsible for the negative amount. Indeed, over 97% of those cases have at least one child present in the household. Unfortunately, ACTCCRD values are not available prior to 2005.
As for the negative values of STATETAX, you will notice that only 24 states are represented amongst these respondents, as opposed to all states having respondents with negative federal tax liabilities. Looking at Appendix Two in this document, you will notice that the 24 states where respondents have negative values for STATETAX are those states with at least one refundable tax credit.
Unfortunately, I do not see an obvious way to make FEDTAX or STATETAX perfectly consistent over time, as the data necessary to do so is not available. One way to get fairly consistent numbers over time may be to simply make the negative tax liability values zero. This would approximate the methodology of the CPS tax model pre-2004; however, this is ultimately your decision.
Hope this helps.
Thank you for your answer, which has helped me clarify things further. However, I still have a couple of doubts:
I managed to replicate fedtaxac as fedtax-eitcred-atccrd (-mwpval for 2010 and 2011) for all years after 2005 with the exception of 2012. For this year I get significant diferences for some observations. For e.g. for serial=96 and pernum=1 fedtax=5565 and fedtaxac=4765 but eitcred=actccrd=0 so “my” fedtaxac is just equal to fedtax, i.e. 5565. Do you know why this happens?
I’m still confused about what is deducted from statetax to get statetaxac. In the documentation it says that statetaxac=statetax-eitcred-atccrd. But this doesn’t make sense as the deduction of these credits is already considered when computing fedtaxac. And indeed, when I try to replicate statetaxac this way I do not get the right figures. So where are the diferences between statetax and statetaxac actually coming from? I suppose it comes from somestate-level deductions but this is not something we can see in the CPS data, right? Put it another way, there is not enough information in the available data to replicate statetaxac, right?
I was able to replicate the discrepancy you found in 2012 where FEDTAXAC is less than FEDTAX without any tax credits having a positive value. Note that only 0.64% of the age 15+ sample is affected and over 60% of these are instances where the discrepancy is exactly $400 or $800. In fact, none of the differences are greater than $800. This leads me to believe these are persons assumed to be receiving the Making Work Pay tax credit.
As discussed in the above answers, the tax variables in the CPS data are estimated using a tax simulation model. It is plausible that an assumption was included in the tax model for 2012 that accounted for a small subset of MWP-eligible persons receiving a delayed refund; however, I am unable to locate any documentation that includes the specifics of the 2012 CPS tax model. The only way to confirm my assumption about the MWP tax credit in 2012 would be to contact the Census Bureau directly.
As for the state tax question, you are correct that the relevant tax credits are actually state-level credits, which differ from state to state. These state-level credits are not individually included in the CPS data, so it is not possible to replicate STATETAXAC. I would recommend referring to the state credit table provided in my previous answer as a starting point for understanding what types of credits are included.
Hope this helps.
Thank you very much for your answers, Tim, they were very helpful!