Hi,
I computed shares of Black residents in poverty by state in 2015 in IPUMS USA: 1850-2023 and IPUMS CPS: ASEC 1962-present and am getting vastly different figures. I ran a query to compute the general US population and still am getting different figures. Was wondering why?
Noting down what I did below.
IPUMS USA:
Row = statefip
Column = spmpov
Weight = perwt
Filter = year(2015),race(2),gq(0-2)
Here’s what I got - 75.4% not in poverty and 24.6% in poverty.
IPUMS CPS:
Row = statefip
Column = spmpov
Weight = sdawt
Filter = year(2015),race(200) - did not include GQ since ASEC only surveys non-institutionalized residents anyway
Here’s what I got - 76.5% not in poverty and 23.5% in poverty. [Note: The Census uses ASEC data and puts it at 24.1% - closer to what I’m getting in my IPUMS USA query]
The state-level figures are also pretty different. Since ASEC asks questions about the previous year, I also tried filtering for year(2016) and still got figures that are very different from the 2015 IPUMS USA numbers.
I also ran a general query on 2015 population on IPUMS CPS and got ~316.2 million. Using the same variables on IPUMS USA, I got ~321.4 million and after including a GQ(0-2) filter, got ~313.3 million.
I know the two surveys are completely different and common variables don’t necessarily mean the same thing in either dataset but shouldn’t the final figures (especially that of population) come close? Why are the numbers so different? Am I doing something wrong?
While measuring something like poverty or any percentile income threshold (like median), what dataset is better?
Thanks!