Thank you for responding to my earlier question on county data. I need some cross tabs that are not available in the general acs tables. I am filtering the data extract from ipums that I pulled and as a test I am trying to match ACS table B23025 armed forces numbers by county. I am filtering on the following fields and my totals are 1,176 when I sum the person weights. The B23025 table lists 1,172 for armed forces in Sacramento County. Would you consider these filters as correct to identify active duty service members. I will add age and education to my cross tabs if the filters are appropriate. When I look at an age range of 16-44 the extract gives me 1,079 and the table b23001 has 1,043. Does a difference of 36 seem resonable. These are from the ACS 2016 5 yr tables.
First, those numbers seem reasonably close to me. The microdata samples cover a smaller portion of the population than the complete ACS sample on which the summary tables are based, so we expect some discrepancies. That said, it would be good to look at a larger set of numbers to determine if there’s any systematic bias: if your numbers are consistently higher or lower than those in the summary tables across many categories or counties, then that would suggest your weighting or codes may be off.
Second, if you’re trying to match B23025 specifically, then I believe you need only use empstatd (along with state and county codes) and not ind or vetstatd. The structure of B23025 is very simple and its labeling brief, with no mention of specific industries or veteran status, suggesting to me that its categories most likely correspond directly to empstatd categories. I’m not an expert on this subject, however, so you may want to consult the ACS Subject Definitions for more information on how concepts in the summary tables are defined.