ACS Data Users Group

I am using 2019-2023 ACS PUMS person level data (pusa-pusd, all the 4 .csv files). Using these 5 years data my weighted frequencies for each year varies between 31 to 33 million in case of workers and from 34 to 35 million for the rest of the population (children and non-working adults), yielding a total of about 65 to 67 million, as shown below. These numbers are very low compared to the actual working population for these years and also for the entire population. I have looked into my coding and everything, it is pretty simple to get these tables, so I do not think that I am doing anything incorrect. Is it that the five years data together yield actually an annual value in weighted frequencies and I should be using the percentages to get the actual estimates for working and entire population for each year separately? I looked into the data dictionary and the questionnaires and did not come across anything that states this explicitly, any help is appreciated to get this resolved.

Hello, it sounds like you're using the ACS-5-year PUMS data to produce 1-year estimates. The 5-year PUMS data are designed to produce estimates over a 60-month period, not for individual years, which is why the weighted frequencies for each year are about 1/5 of the total. If you want a 5-year estimate from 2019-2023 you can just produce weighted tabulations from the 5-year file and they will represent the entire period. Or, if you want data for each year, I'd suggest using the 1-year PUMS data.

I hope this helps.

Thank you Mark, given Census doesn't prefer us to use the 2020-one year ACS data, I was planning to extract the yearly weighted frequencies using the 5-year data. But now I am thinking of using the one-year data for 2019, 2021, 2022, and 2023 separately.