Why are there missing data for black median HHI, tenure status, and poverty rates at the county level?


I am trying to compare black-white disparities in median household income, homeownership, and poverty rates at the county level using the 2012 - 2016 five-year ACS estimates, but there is a significant amount of missing data for those three variables. The source tables are as follows:

  • B19013B: Median Household Income in the Past 12 Months (in 2016 Inflation-Adjusted Dollars) (Black or African American Alone Householder)
  • B25003B: Tenure (Black or African American Alone Householder)
  • B17020B: Poverty Status in the Past 12 Months by Age (Black or African American Alone)

I should add that 38% of counties have missing data for black median HHI, 5% missing for black poverty, and 14% missing for black home ownership at the county level. Moreover, this does not seem to be a suppression issue because there are counties where the black population count is smaller, but black median HHI is reported compared to counties where the black population count is larger, but black median HHI is not reported.


I am just looking at these tables now and not seeing any missing data. I think I understand where the confusion is coming from. The table includes a bunch of columns with no information in any of the cells. These include the columns: “COUSUBA,” “PLACEA,” TRACTA," “BLKGRPA,” etc. These indicate other levels of geographic identification such as the county subdivision code, the place code, the census tract code, the block group code, etc. that do not relate to your county level file. The codebook that came along with your table indicates what each of these columns identify. The reason they are blank in your tables is because you are searching for county level data, rather than data on these sub-county levels. The County Name column and the columns with the desired information (i.e.,median household income) are both filled.


I find missing data in the median household income table, but for tenure and poverty, I find cells with a value of zero but no cells that are completely empty. Empty cells (missing data) differ from cells with a zero.

Missing data are relatively common in the ACS 5-year estimates, even for counties. The Census Bureau requires a minimum number of unweighted cases in order to generate counts/medians for geographic units (you can find details in the ACS data suppression document). My guess is that the Bureau doesn’t have enough responses from African American households to generate a median value for the county.

Cells containing zeroes should be treated exactly that way - the Bureau estimates that no African Americans own homes in that county or no African Americans are below the poverty line in that county. Most of the counties with zero values have fairly small African American populations, so these zeroes are plausible.