Block Group level poverty by race data: I cannot find it.

I am gathering data for social equity/racial justice research. The ACS provides block group level data for race/ethnicity for persons, and also on poverty for persons. But it does not give “persons in poverty by race.” This information would be very valuable, but I was told by the Census staff that it cannot be distributed for privacy reasons. It is frustrating that we cannot locate “neighborhoods” (aka block groups) with low-income people of color. If we cannot locate them, that hampers our ability to target services to assist them. It seems like a Catch 22.

I was able to find municipal level data by HOUSHOLD which has income levels by race. That was somewhat helpful. But it seems like the existing census bureau practices are limiting our ability to do rigorous research on social justice, which is a very pressing issue.

Do any of you have any thoughts on this? Am I missing something? Is there some other means by which I can get neighborhood level data on poverty by race?

  • Would census tract level data suffice?  Tracts are composed of from 1 to 5 blockgroups.  Table B17001 is available for race and ethnic groups at the tract level.

  • In the case of ACS data, the issue with statistics like these is not just privacy but also sample size. The margins of error are generally high for block group counts, and even more so for smaller subpopulations, so poverty rate by race/ethnicity would often not be very reliable at this scale.

  • Correct. The problem is that we have small communities of mostly Mexican origin farm workers scattered through rural parts of our County.  So if we look at a tract level, they sort of "disappear" from the data. I had a resident who was doing a study on improving internet access asking me where the disadvantage communities were. I could give a good answer for urban centers, but not for out in rural areas. It's frustrating.

  • I hear this concern and have found various ways of dealing with it -- none perfect or national in scope but maybe within a county it might help --  for one thing SNAP ("food stamp") data is available from states at the zip code level and often can point you to populations struggling to get by in almost-real time since they're way ahead of ACS in time frame. For those not able or afraid to claim these benefits because of status there's food pantries and those who run them will be up on where people are living since they usually collect some information on participants. Once I found a Spanish Mass at a local church and talked to people there, and the priest was well aware of settlement patterns in that population

  • Dear Jake,

    One way to try to do this us to use Small Area Estimation where you use a tabulation from a "large area" in your case say the county level (or larger),  a statistical model for the outcome of interest, in this case poverty, and relate poverty to variables that you do have at the census tract or block group level.  For the outcome poverty a lot of work has been done on this for different geographies as support for many benefit programs (Federal and State) depend on the poverty level in a particular geography. You may be able to find a reference where someone has done this analysis for the variables that you are interested in. To build a model from scratch requires a lot of statistics and a good feel for what the variables are telling you. The devil is in the details. Perhaps you can collaborate/partner with an organization or a statistician who can help out.

    The UN does this type of analysis at larger geographies.


    Dave Dorer