Hmm, I’m not sure why you wouldn’t be able to associate data from time series tables with your pre-existing dataset in the same way you could associate data from NHGIS source tables. In both types of tables, NHGIS includes the state and county FIPS and NHGIS codes, which are the usual ways to link county data together.
As for the “boundary changes issue”, note that the NHGIS time series for poverty data are “nominally integrated,” meaning that the tables match counties across time based only on their names with no adjustments to control for boundary changes. I.e., the counties in the time series tables should match up exactly with the counties in source tables.
It might be clearer for you if you request the time series tables in the “Time varies by file” layout instead of the default “Time varies by column” layout. This option is available on the Review and Submit page, just before you submit a data request. In the “Time varies by file” layout, each year’s data is supplied in a separate file, so you’d get the 1970 data in its own file–very similar to how you get source table data.
Finally, if you’d still like to know exactly how to construct denominators for poverty rates, that information is provided in the codebook file that comes with the AX6 time series table in a data extract ZIP file. If you scroll to the bottom of the codebook, you’ll see exactly which source variables were summed together to produce each time series statistic (including Persons for Whom Poverty Status Is Determined).