SDA—household total don't match census tables

Hello! I am trying to determine the number of households in each income bucket. I recoded “hhincome” into increments of 100,000—but am getting an overall number of households far higher (327mn) than what seems reasonable, or what is listed in the census tables from (131mn). Any ideas?

Oh, and to be clear: I am using the online SDA tool.

If you can provide some more information about your input into the SDA and the specific Census Bureau table you are referencing, I will be able to provide more targeted guidance. Without this information, my guess would be that you are creating person-level estimates, as there are roughly 330 million persons in the U.S., while the Census table you are looking at contains household-level estimates. Restricting your analysis to one person per household (filter PERNUM=1) would likely be a good starting point.

Hi Isabel,

Yes, that was exactly the mistake I was making.

Can I ask you for help with another problem: I am trying to map census data (in this case, the rate of homeownership). My map-making tool (DataWrapper) accepts, as the geocode, a county FIPS code. I therefore initially tried to extract the census data by county. But the census doesn’t have county information for many households.

My solution has been to extract the data by PUMA, and then aggregate these into counties, and then map it. Does that sound like a sensible solution?



IPUMS NHGIS may be better suited for this purpose. IPUMS NHGIS provides summary tabular data for select topics, such as home tenure, from the decennial census and American Community Survey (ACS) aggregated at various geographic levels, including at the county level. In the IPUMS USA ACS microdata, county (COUNTYFIP) is not identified directly, but imputed based on PUMA, the smallest geographic level explicitly available in the microdata. IPUMS USA only identifies a county when that county is coterminous with a single PUMA or if the county contained multiple PUMAs that did not extend into other counties. Therefore, your proposal to aggregate PUMAs into counties will still not identify counties that are contained across multiple PUMAs.

Using the NHGIS data finder tool, you can filter tables by topics such as tenure and mortgage status or median income at the county level. We also provide shapefiles which may be of use to you. In this brief video tutorial, you can learn how to use the IPUMS NHGIS data finder to create a custom extract of summary tables, shapefiles, or both.

Hi Isabel,

Thanks for all your help with the mapping issues—NHGIS is a terrific resource.

Can you help me with another problem? I am trying to link seasonal homes with incomes. In the census data, is a home enumerated as vacant (seasonal use) linked to the household income of the owners?


Thanks for your message. Isabel is out of the office, but I can answer your question.

Vacant units do not have person records associated with them as they are not occupied. Because income is asked of persons, there is no household income reported for these units. It is possible you could get at this using data from the American Housing Survey (see AHS codebook for reference), though I am not familiar with these data.

Thank you for the information.