Closed victoramoureux closed 6 years ago
@victoramoureux I do not think it is double counting. First, we use pil = pil + pxit
to convert the individual level data from net to gross so we can impute social security contributions. Then, once we are at the household level, hil
is still net of hxit
(because hil
comes directly from the survey data). Therefore we need to convert it from net to gross again.
This is exactly the fix I made last week. See commit https://github.com/matthewolckers/lis-tax-transfer/commit/775c4a7715587ee28e879ae396eb9670863fe1ac
Yes I'd noticed it was your commit! My mistake, I thought hil was obtained by aggregating pil by hid in the convert_ssc_to_household_level part but we only do that for psscee and psscer. I'll re-export Italy 2014 with the correct code then!
I have the impression that we are "too good" in the italian case because we add taxes two times to the definition of marketincome, hence artificially increasing the denominator in our tax shares:
first pxit to pil in the IT_gen_pvars (l.152) https://github.com/matthewolckers/lis-tax-transfer/blob/326c7329e8c32780b287db72379f8b0e0aded107/Stata/do/Extract_pctiles_0625.do#L152
then again : (recall tax = hxit + hsscer) https://github.com/matthewolckers/lis-tax-transfer/blob/326c7329e8c32780b287db72379f8b0e0aded107/Stata/do/Extract_pctiles_0625.do#L364
I think we should just do it once but I'm not completely sure... @matthewolckers what do you think? FYI, I accidentally ran it without the second line (hence with marketincome = hil + hic - hicvip + hsscer and hil = hil (original) + hxit) and IT seems to be an outlier again.