This page presents historically comparable Current Population Survey (CPS) microdata weights from 2010 to 2024. They are consistent with the official weights published in the 2025 CPS basic monthly microdata and make it possible to calculate historically comparable time-series of labor-market measures.
The Bureau of Labor Statistics (BLS) warns against using published weights to create multi-year time-series. This is because official published weights are not revised, even though the Census population estimates that are the “controls” for these weights are revised annually. Because published weights are not revised, there is an annual population adjustment to the CPS in every January report. The January 2025 population adjustment was unusually large because Census changed its methodology for estimating immigration, which revised population estimates significantly upward (see linked blogposts for more background).
My historically comparable CPS microdata weights are based on the most recent available population estimates for the national population and for 96 demographic groups by age, sex, and race and ethnicity.
Each year, the latest annual Census population estimates are released or revealed in multiple stages: the national estimates in December, then with some demographic detail in February, and the complete final set of estimates with full demographic detail in June. Each stage permits an increasingly accurate method of calculating current population estimates for demographic groups, and using those estimates to create historically comparable CPS weights. Here are the stages for this round of annual Census population estimates – the vintage 2024 estimates.
METHOD #1 (doable in December 2024): Infer demographic-group estimates. National population estimates were released in December 2024 for the period April 2020 (the month of the most recent Decennial census) to December 2025. No demographic detail was released at that time. But it turned out that the huge revision to national estimates was due to an improved methodology for estimating immigration, so a reasonable assumption was that the demographic groups with the largest population adjustment were the demographic groups that include more recent immigrants (e.g. Hispanic/Latino men age 25-34, but not White women age 75+). The 2023 American Community Survey (ACS) provides detailed demographic information on recent immigrants, and in fact Census uses the 2023 ACS in calculating its 2024 estimate of immigration. I used 2023 ACS data to generate population estimates for demographic groups for the period April 2020 to December 2024, with results in this blogpost.
METHOD #2 (doable in February 2025): Calculate demographic-group estimates using published January 2025 CPS weights. The January 2025 CPS is controlled to the 2024 Census population estimates for demographic groups even though the full 2024 demographic-group estimates have not been published. Thus, the latest population estimates for demographic groups can be derived from the January 2025 CPS microdata, which were published in mid-February. I compared these estimates from January 2025 with the population trend for demographic groups as published in 2024 to calculate the recent annual population adjustment for each demographic group. With this adjustment, I generated population estimates for demographic groups for the period April 2020 to December 2024; these are the basis for the factors available for download, below.
METHOD #3 (doable in June 2025): Use complete final 2024 population estimates. In June 2025, Census will release the complete, final 2024 population estimates, including for demographic groups, for the period April 2020 to December 2025. Those estimates can be used directly to create factors for revising historical weights back to April 2020. I plan to do this once the estimates are released.
For all methods, the historically comparable weights can be extended back to April 2010 by adjusting the April 2010-April 2020 final Census estimates (published in 2021, with a revision coming later in 2025) to align with the latest population estimates by demographic group for April 2020, following this method.
For all methods, the 96 demographic groups include every combination of men and women; ages 0-15, 16-24, 25-34, 35-44, 45-54, 55-64, 65-74, and 75+; and six racial/ethnic categories:
- non-Hispanic White
- non-Hispanic Black
- non-Hispanic American Indian / Native American
- non-Hispanic Asian / Pacific Islander
- non-Hispanic two or more races or some other race
- Hispanic of any race.
Instructions for downloading and using the historically comparable CPS microdata weights
Here are the factors from method #2, for 2024 Census population estimates. There is a factor for each of the 96 demographic groups, in each of the 177 months from April 2010 to December 2024. These are the most accurate factors available prior to the June 2025 release of the final 2024 population estimates for demographic groups.
To use the microdata weights, start with a dataset of CPS monthly microdata downloaded from IPUMS-CPS, including the variables COMPWT, SEX, AGE, RACE, and HISPAN.
Generate the race group and age group variables as follows (in Stata):
- recode race (100=1) (200=2) (300=3) (650/652=6) (700/999=9) (809=6), gen(raceg)
- replace raceg=10 if inrange(hispan, 100, 699)
- egen ageg=cut(age), at(0, 16, 25, 35, 45, 55, 65, 75, 999)
Merge the file (linked above) of factors onto the microdata and multiply the published weights:
- merge m:1 year month ageg sex raceg using historical_microdata_weights, nogen
- recode factor24_m2 (.=1) if year==2025 & ageg>=16
- gen double wt24_m2=compwt*factor24_m2
The resulting weight variable can then be used to calculate historically comparable time series.
Note that the variable name factor24_m2 means that this is the factor for the vintage 2024 population estimates, following method #2.
Please note: these are experimental. I will update these factors and might revise the description. Questions and feedback most welcome here.