Which is more accurate for a population estimate, summing the weights or using a ratio of the weights?


Hello, So using 2013 ACS, I created a percentage of weights of disabled households to toal households to try and figure out how many households in Maine have a disabled HH member. I multiplied this by the number of total household in Maine from the 2013 ACS from American Factfinder.

To check my work, I poked through this forum and am seeing I could have just summed the weights to acheive the estimate I wanted. So my question is, which is more accurate? Adding the weights gives me ~159,000 while using the percentage gives me ~151,000.

Thank you!



If you did things right, these numbers should have matched. Without you posting the code and output, it is hard to say why these numbers differ. You may have missing data somewhere, which can through any of the sum of weights, numerator or denominator of the ratio.



I was not able to replicate your issue. I have a couple recommendations to help reconcile your numbers. First, make sure you are not including vacant households in your American FactFinder housing units count. In 2013, there are 116,339,928 occupied housing units (87.6% of total housing units). Second, in your IPUMS-USA data, you should exclude group quarters (i.e. GQ=3 or 4). Taking these steps should make the result you get from summing the IPUMS weights nearly identical to using the FactFinder housing unit count.

Hope this helps.