I’m working on a paper on income inequality and would like to have income inequality by county for the entire US. There are older papers that have used this and referenced the Census
IPUMS NHGIS provides tabulated US Census data on a wide variety of topics. One such topic is income inequality. Using the IPUMS NHGIS data finder, you’ll be able to find many tables (with GIS boundary codes) that provide information on the GINI index at various geographic levels.
I hope this helps and you find what you are looking for. Let us know if you have any further questions.
It looks like you can calculate this with IPUMS data. The first way is presented in this YouTube tutorial. The problem, which I’d throw to the IPUMS staff is sorting based on income. This method would require the weighted income for each observation. Is it possible to do that? Or am I misunderstanding the relationship between observations and weighting.
Second, someone created a program in Stata “ginidesc” that calculates the gini index. I’m playing around with it now. Using this program my gini coefficient for Illinois (using 2018 5 year data) is 1.5% lower than what the census reports (https://www.census.gov/content/dam/Census/library/publications/2019/acs/acsbr18-01.pdf). I’m not sure why this is - I haven’t dug into how the program works. But my guess is that census Gini estimates use a full dataset whereas the Illinois figures are top coded, thus attenuating the inequality. Here’s the code I used in stata:
ginidesc hhincome [fw = hhwt] if pernum == 1 & hhincome != 9999999,by(year)
You are correct that top-coding is likely driving some of this discrepancy as well; this is one important reason to expect differences between estimates derived from the public use data versus official estimates from the Census Bureau. I suspect this is what is driving the differences you are seeing. However, it is worth noting that the Census Bureau document you linked indicates that the table is limited to the household population (e.g., excludes persons in group quarters); you can restrict on group quarters status with the GQ variable.
To your question about weighting income values, you can treat each individual (or household) as representing N individuals (or households), where N is the weight.