I am new to using the IPUMS online data system and am using excel, not statistical software. I want to build a profile of one subgroup by Race (a subgroup of the Asian population). I am working at the PUMA and MSA level. The numbers for age, educational attainment, occupation and industry, employment status all seem to make sense, with the population totals close to the ACS report. I am having trouble with income measures. Using the online system, I would like to compare an income measure of the subgroup to the total population. My first step was to compare the total population income measures of family or household income are very different from the ACS reports. I seem to get somewhat closer using total personal income to estimate the per capita income. The per capita income for three of the four PUMAs I calculated are within the published margin of error. The one that was not with in the margin of error had a 2744 difference and the MOE is 2448.
Do you have any suggestions for how I should proceed?
Although it is difficult for me to identify with certainty the true source of the discrepancy, I do have a couple of ideas that might explain your observations.
First, how are you dealing with the special codes for your income variable (e.g., household income HHINCOME, family income FTOTINC, or personal income INCTOT)? Depending on which variable you are using, there are different codes that do not actually represent real income values. These special codes identify respondents who are not eligible for the income question—such as individuals who live in group quarters for the household and family income variables or age restrictions for all income variables. In most cases, it makes sense to exclude these “N/A” values from your analysis.
Second, if you are using the family income (e.g., FTOTINC) variable, note that this variable uses a slightly different definition of “sub-families” than used by the Census Bureau. This could explain why your figures are close but slightly outside of the reported margin of error. See the discussion of this in the FAMUNIT variable description.
Here are the steps I took to get per capita income for 4 PUMAs. These steps resulted in 3 out of 4 PUMA’s per capita income within the margin of published error.
ROW: inctot
SELECTION FILTER: met2013(41180) PUMA (1804)
WEIGHT: perwt
Copy result into Excel
Multiply 2 columns (Labeled N * Distribution)
Exclude the N value 9999999
Total the calculated column
Divide by total distribution (Which equals the population for the PUMA)
If this is correct, I would like to rerun with an added filter selection for raced (610)
I would also like to estimate a measure of household income, but don’t understand how to do it using the online tool. This post makes me think I can’t: Matching ACS median household income?
D
Thank you for sending over the additional information. One detail that I think may be influencing your results is how you specify the geographic area. That is, typically, you should not filter on both MET2013 and PUMA at the same time. This is because (a) PUMA boundaries are state-dependent and should be filtered along with corresponding STATEFIP values and (b) MET2013 provide inexact correspondence to official delineations. I think this could be driving the discrepancies you are observing.
In order to estimate household income, the easiest approach is to use the HHINCOME variable. This variable sums up all values of INCTOT for all persons within a given household. In the on-line analysis tool you will want to make sure that your analysis is at the household level. So, select the household-level weights and filter on PERNUM==1 to ensure you are only analyzing one observation per household.