Hi all,
New here! I've been lurking for a couple of months as I try to get myself familiarized with the quirks of working with ACS data. But I have run into a problem with some perplexing estimates in some key variables used in my research, and I am hoping someone here might have some insight.
Specifically, my research team and I are interested in the impact of county-level, race-specific characteristics on county child welfare performance indicators. Race-specific unemployment rates are expected to explain some of the variability in county performance. The problem we are running into is that the rates drawn from S2301 contain a substantial percentage of extreme estimates for county-level black rates (0s and, to a far lesser extent, 100s). See the first table below. I'm guessing this is a coverage issue, as 90% of the cases have margins of error above 23%.
I suppose my question here is three-fold:
- Am I correct to assume that many of these extreme values are due to limited coverage?
- If so, are there established best practices when using the ACS estimates to distinguish between valid and problematic estimates? I want to avoid recoding valid 0's as missing just as much as I want to avoid including problematic estimates.
- If the ACS data are simply not ideal for obtaining race-specific estimates, does anyone happen to know if there other data sources that would provide county level estimates by race?
Thanks so much for your help (and your patience!).
~ Miranda
Example of Raw Data