I am calculating Gini coefficients for the universe of places using 5-year ACS estimates (2008-2012). The calculation requires the number of households per place in each income category and the mean income of each category. I am using the Census Bureau data aggregated by place to estimate the number of households in each category. In order to more precisely estimate the mean value of each household income category, I would like to use the microdata.
The page describing IPUMS adjustment techniques indicates that IPUMS pre-applied the Census Bureau’s adjustment factors to FTOTINC “[b]etween November 2009 and March 2010.” It’s not clear what the practice has been since March 2010.
Is the IPUMS FTOTINC variable for the 2008-2012 5-year sample calculated using the same adjustments as the aggregate ACS data provided by the Census Bureau?
Census place information is generally consistent with the IPUMS-USA variable, CITY. Cities are not originally available in the public-use data, but some can be identified through low-level geography. The smallest unit in the ACS microdata is the PUMA, which must contain at least 100,000 residents in order to retain the anonymity of respondents. A PUMA’s boundaries frequently straddle city boundaries or encapsulate multiple smaller cities. Because of this, it is often not possible to determine the city of residence for a respondent. At the bottom of the comparability section for CITY you will find more information about how cities are identified as well as a list of documents containing details about each IPUMS CITY code and its corresponding FIPS state and place codes.
For the treatment of FTOTINC, there are a couple things to say. In the multi-year ACS files, IPUMS-USA standardizes the amounts of FTOTINC to the dollar value of the final year of data included in the file. In other words, FTOTINC is expressed in 2012 dollars for the 2008-2012 5-year file.
The adjustment provided by the Census Bureau has not been applied to the current values of FTOTINC but is available to users in ADJUST. These values will still be slightly different from what is used in the published summary tables you are finding, since the Census Bureau applies a monthly adjustment value that accounts for month of response. In the public-use files, the Census Bureau did not release monthly adjustment values for confidentiality reasons, but instead released an annual average.
I hope this helps!