total population


I understand that the IPUMS sample consists of observations at the individual level so if you sum up the observations by a specific geographic boundary (ie PUMA), then you have a count of the sample for that geographic boundary.

However is there a total population variable by geographic boundary for normalization purposes? For example, I want to divide the “hcovany” health insurance variable by the total population in order to get a percentage of those with health insurance rather then using the raw numbers.



The only variable currently available in IPUMS-USA that gives total populations for geographic areas is CITYPOP. However, this variable only gives populations for Cities that are “Identifiable” within the sample. A city is identifiable within a sample if the boundaries of the smallest available geographic unit (PUMAs in 1990, 2000, and 2005-onward samples) match perfectly with the boundaries of the sample. This is discussed further in the Comparability Statement for CITY.

You can also retrieve population totals for various geographic regions in 2010 and some ACS samples from American Fact Finder and NHGIS.

I hope this helps.


I actually figured out that if you simply summed the Person weight variable, then you can get the total population and aggregate it to whatever geographical boundary you need, city, puma, county, etc.

Is this correct to do?


Summing the weights is a good way of getting an estimated population size. I simply thought you wanted official population estimates, rather than estimates generated from the data itself. I am glad you found a solution.