I am a little confused about how IPUMS adjusts ACS income variables for inflation for the 5-year data. My understanding based on an earlier forum post is that the income data are inflation adjusted to the last year in the file. For example, all income variables in the 2019-2023 IPUMS file are in 2023 dollars. This does seem to be the case.
A related forum post claimed that the original income variables in the 5-year ACS PUMS data that the Census releases are already adjusted for inflation. However, the ACS user guide seems to imply that the data need to be adjusted using the variable ADJINC (see page 17).
I compared the wage income variable in the Census PUMS (2019-2023 5-Year) data downloaded via the Census FTP with the corresponding IPUMS file. I was able to find the same individuals in both files. Based on the person-level data, it seems clear that the Census PUMS data are not inflation adjusted. When I apply the ADJINC variable, I get close to the IPUMS wage figures at the person level, but the match isn’t exact.
I also noticed that the adjustment factors on the IPUMS website match ADJINC for 2023 but not for previous years.
My question is if I should expect the IPUMS data to match the Census PUMS data after applying ADJINC to the Census PUMS data? How exactly does IPUMS adjust the 5-year data for inflation?
The IPUMS documentation explaining the inflation adjustment is confusing and don’t really explain how the adjustment works.
In the original PUMS data released by the Census Bureau, dollar amount variables are not adjusted for inflation. IPUMS adjusts dollar amount variables in the multiyear ACS samples to report values in constant dollars, adjusted to the last year of the multiyear file. For example, we adjust dollar amount variables in the 2019-2023 5-year ACS sample into 2023 dollars. I apologize for any confusion caused by the 2021 forum post you linked to. We will revise this post.
The Census Bureau provides an inflation adjustment factor for users to manually adjust dollar amount values for inflation. The variable for the adjustment factor is called ADJINC in the original PUMS data. The value of ADJINC is constant within single-year ACS data files, and within each single year of multi-year ACS data files (i.e., within each value of MULTYEAR). This adjustment factor makes adjustments to account for the rolling survey schedule of the ACS. In the multiyear files, it also adjusts dollar values to the final year. Because IPUMS automatically adjusts multiyear files to report dollars to the final year value, our version of the ADJINC variable (ADJUST) provides only the component that adjusts for the rolling survey schedule. This means that the IPUMS variable ADJUST and the original Census Bureau PUMS variable ADJINC will not match in the multi-year files. Note also that the IPUMS adjustments for inflation in the multiyear files do not include the adjustment for the rolling data collection of the ACS.
The ACS is conducted on a rolling basis throughout the year. The reference period for all income and earnings variables in the ACS is the previous 12 months. As a result, different respondents have different absolute reference periods for reporting their income and earnings. For example, someone surveyed in January 2023 reports income earned between January 2022 and December 2022, while someone surveyed in August 2023 reports income earned between August 2022 and July 2023. This means that income and earnings are a mix of calendar year dollars (income earned in the calendar year of the survey), and of dollars that need to be adjusted to be expressed in calendar year dollars.
Ideally, an adjustment factor to account for this rolling design would be unique to each month of data. However, month-specific adjustment factors would make it easier for individuals to be identified (month of survey is not in the public use data), so the Census Bureau does not provide them. Instead, the Census Bureau averages the twelve month-specific adjustment factors to create a single adjustment factor that is constant for all cases within each year of the ACS. Census Bureau documentation advises users to multiply all income and dollar amount variables by the ADJUST constant, which converts numbers into calendar year dollars. However, a constant adjustment factor cannot adjust dollar amounts to calendar-year dollars. Dollars earned during the previous year (as for January respondents) need to be adjusted the most; dollars earned mostly during the current year (as for December respondents) are essentially in calendar-year dollars already and need to be adjusted only slightly. Therefore, IPUMS does not apply the adjustment factor but does offer it as a stand-alone variable that users can apply manually if they wish.
See also information on our adjust monetary values (AMV) feature, which allows users to automatically adjust dollar amount variables to 2010 dollars as they create their data extract.
Hi Isabel, Thank you for this detailed explanation. Here’s my recap for the ACS 5-year data. I hope I’ve got this right, and I’m sharing in case it is useful for others.
- The original Census ACS 5-year PUMS income variables are not already adjusted for inflation.
- The ACS PUMS User Guide instructs users to adjust income and earnings dollar amounts by multiplying the relevant variables by ADJINC / 1,000,000.
- ADJINC consists of two components. The first component is the Census Bureau’s adjustment for the ACS rolling survey design: households report income for the prior 12 months, but interviews occur throughout the year. The second component is the inflation adjustment to the final year in the multiyear file. This second component uses the CPI-U-RS to convert income values to final-year dollars. For example, the Census API shows the 2019 ADJINC value as 1.010145 * 1.19558276, where 1.010145 is the rolling-survey/calendar-year adjustment component and 1.19558276 is the CPI-U-RS-based adjustment to 2023 dollars.
- IPUMS does not apply the Census rolling-survey adjustment to its income variables because the rolling-survey adjustment would ideally require month-of-interview-specific factors, but interview month is not available in the public-use data. Census instead provides a single annual factor, which IPUMS does not view as adequate for truly converting each respondent’s prior-12-month income to calendar-year dollars.
- This explains why IPUMS income variables do not match Census PUMS income variables multiplied by the full Census ADJINC / 1,000,000. IPUMS adjusts ACS multiyear income variables for inflation by using the CPI-U-RS to convert income values from each survey year to dollars for the final year of the multiyear file. However, IPUMS does not also apply the Census rolling-survey component embedded in ADJINC.
- Users trying to reconcile Census PUMS and IPUMS values should adjust the Census PUMS income variables using only the final-year CPI-U-RS component of Census ADJINC, not the full ADJINC factor. Equivalently, users could deflate the IPUMS values using that same CPI-U-RS-based final-year component.
It would be helpful if IPUMS could make these distinctions more clear in the documentation. The current explanation doesn’t mention the CPI-U-RS or explain what IPUMS actually does to the raw data.
Yes, this is all correct. In answering your forum post we have come to the same conclusion that our documentation could be much clearer and consolidated in a way that’s visible and comprehensible for users. We are working to update our documentation and provide more information about adjustments to dollar amount variables in one location on our website.
Thank you for helping me work through this. I appreciate everything the IPUMS team does to make the data as accessible and user-friendly as possible, including efforts to improve documentation in response to user feedback. It is a tremendous resource for the research community.