Inflation Adjustment

I'm trying to compare income figures between years and having trouble inflation adjusting in a way that matches the Census Bureau's adjustments. I'm hoping others have had better luck and can point me in the right direction.

What I want to do is look at changes in household income for quintile levels (table B19080, which only reports nominal values) from 2007 to 2013. To adjust the values to real 2013 dollars I used the annual average CPI-U-RS estimates from the Bureau of Labor Statistics (www.bls.gov/.../cpirsdc.htm), which from what I read is the price index used by the Census Bureau. As a test to make sure I was doing it right, I adjusted the nominal median HH income figures in table B19013 and compared to the real values in CP-03. But my own estimates of the real income value in CP-03 are off (only about $200 or so, but still a noticeable difference). I don't think it's my math: 2012 nominal income * (CPI2013 / CPI2012). Has anyone had better luck replicating the ACS adjustments? Where am I going wrong?
Parents
  • I have an even more basic question - where did you find the CB's adjustments on income?

    FWIW I intend to use the BLS calculator to adjust for inflation
    www.bls.gov/.../inflation_calculator.htm
  • The income/earnings questions don't go by calendar years, they go by "earnings/income in the past 12 months." I think what they are doing is somehow averaging the current survey year's and the previous year's CPI-U-RS (although...not everyone would be urban, so not sure). If you're interviewed in June 2011, then your earnings for the past 12 months are partially from 2010.

    Are you working with PUMS? Whenever I work with PUMS multiyear data, I have to use the "adjinc" variable in the Data Dictionary (found here: www.census.gov/.../documentation.html) to express earnings/income in the most recent year's dollars (e.g., 2015 dollars for the 2011-2015 file).

    Here's how I do it in SAS:

    *define variable called earnings_2015 for earnings expressed in 2015 dollars;
    earnings_2015 = .;
    earnings_2015 = pernp * (adjinc/1000000);

    Similarly, there's a variable called "adjhsg" for home value, etc.

    Hope that helps!
Reply
  • The income/earnings questions don't go by calendar years, they go by "earnings/income in the past 12 months." I think what they are doing is somehow averaging the current survey year's and the previous year's CPI-U-RS (although...not everyone would be urban, so not sure). If you're interviewed in June 2011, then your earnings for the past 12 months are partially from 2010.

    Are you working with PUMS? Whenever I work with PUMS multiyear data, I have to use the "adjinc" variable in the Data Dictionary (found here: www.census.gov/.../documentation.html) to express earnings/income in the most recent year's dollars (e.g., 2015 dollars for the 2011-2015 file).

    Here's how I do it in SAS:

    *define variable called earnings_2015 for earnings expressed in 2015 dollars;
    earnings_2015 = .;
    earnings_2015 = pernp * (adjinc/1000000);

    Similarly, there's a variable called "adjhsg" for home value, etc.

    Hope that helps!
Children
  • Thanks, Diana. I'm working with counties of 20-30K population so not working with PUMS.

    I realized after posting that it is problematic using the BLS calculator since it requires that you enter a month as well as a year to adjust from and to.

    I found this on the page about comparing 2015 5yr ACS to 2000 Census, under Income section:

    The Census Bureau recommends using CPI-U-RS adjustment factors published annually by the Bureau of Labor Statistics (BLS) to adjust 1999 median, mean, and per capita income dollar amounts shown in Census 2000 Summary File 3 to 2015 dollars by multiplying the 1999 dollar amounts by the CPI-U-RS factor of 1.4229669.For CPI-U-RS inflation adjustment factors for other years see Updated CPI-U-RS, All items, 1978-2015 [PDF - <1.0 MB].

    and there's similar guidance for other years.