Share of People Living in Metro Areas

The "Detailed Tables" from the ACS (i.e., the ACS Summary Files) include "geographic components" that allow you to breakdown data for certain levels further by metro / non-metro / micropolitan / urban / rural and a few other geographic subcategories. The set of available level-component combinations is listed in the "Appendix B" sheet in the Summary File Appendices, accessible through this page. That shows that you can get data for people "in metropolitan statistical areas" by state for the 1-year tables.

You could extract these data from the raw summary files, but they're a bit bulky. And unfortunately, as far as I can tell, there's no easy way via data.census.gov to get data for a single geographic component for all states at once. It seems you'd have to click on a separate box for the component of interest once for each of the states. (Ugh.) (EDIT: I now see that Glenn Rice pointed out the same problem in another response... so it's not just me that can't find another way to do this!)

Fortunately, we make it pretty easy (IMO) to get this type of data via IPUMS NHGIS. Through the NHGIS Data Finder, I'd select a "Geographic Levels" filter for "State", and then select a Years or Datasets filter for the ACS year I'm looking for. Then find and select the table of interest (in this case, B01003. Total Population), continue to the Options page, and there I can select which "spatial breakdowns" I'd like to include in my data request. I'd select the breakdowns for "Total area" and "In metropolitan statistical area" and continue on to request and download the data. Then I can get state-level total population and population in metro areas in one file, and just divide one by the other to compute the share.

Good solution, Jonathan.  NHGIS is an endless resource!
But how can one get an accurate answer using only 1-year tables?
By my reckoning, 461 of the 1180 major metro component counties are below the 65,000 population threshold to have 1-year data.  Wouldn't those counties be excluded from the calculations?
Or one could use Population Estimates for 2022, which would include all counties.

That's not a problem with the "geographic component" data because it break downs by state, not by metropolitan area parts within each state. While many of the individual parts have less than 65,000 residents, the total metropolitan population by state is greater than 65,000 for every state. I just downloaded the 2021 1-year data, and the smallest single total is 180,418 for Wyoming's metro population. All of the required data are there.

I'm quite sure they don't omit populations in smaller-population metro parts from these totals... That's the sort of thing one might do to protect privacy, but as I understand, the 65,000 limit is not about protecting privacy; rather, it's what's deemed necessary for an adequate sampling size. Adding up all metropolitan parts within each state gets there.

Very interesting.  Thanks for the insight, Jonathan.
I'll have to wrap my head around this.  What especially confounds me are the multi-state metros, and how they are able to apportion the population between states when some of the counties have pops below the 1-yr threshold of 65,000.
For example, Jasper (33k) and Newton (14k) are Indiana counties which comprise part of the Chicago-Naperville-Elgin, IL-IN-Wi metro.  I would think their population is too small to be included if we're only looking at 1-year data, since there are no 1-year pops for these counties.
Maybe the Census maintains 1-year data at a smaller level for internal analysis and aggregation.

Yes, exactly. The ACS collects data all over the place every year, including in small-population counties. It's a complete nationwide sample. Over the course of 5 years, this adds up to a large enough sample to report estimates even for areas with small populations (though often still with large margins of error). The reason they don't report 1-year estimates for small-population areas is not because there's no sample in those areas; it's because it's not an adequately large sample to produce generally useful estimates. So yes, they almost certainly collected many 2021 ACS responses in those Indiana counties you mention. The 1-year sample isn't big enough to publish 2021 1-year estimates for those counties, but they do publish 2017-2021 5-year estimates for those counties, which include the 2021 responses, and they publish 2021 1-year estimates of Indiana's metropolitan population, which also includes the 2021 responses in those smaller counties.

Yes, exactly. The ACS collects data all over the place every year, including in small-population counties. It's a complete nationwide sample. Over the course of 5 years, this adds up to a large enough sample to report estimates even for areas with small populations (though often still with large margins of error). The reason they don't report 1-year estimates for small-population areas is not because there's no sample in those areas; it's because it's not an adequately large sample to produce generally useful estimates. So yes, they almost certainly collected many 2021 ACS responses in those Indiana counties you mention. The 1-year sample isn't big enough to publish 2021 1-year estimates for those counties, but they do publish 2017-2021 5-year estimates for those counties, which include the 2021 responses, and they publish 2021 1-year estimates of Indiana's metropolitan population, which also includes the 2021 responses in those smaller counties.

Thanks very much!  I see what you mean, i.e. the collection of the 1-year data and how it's used even though it's not reported.
Very insightful.

Bert