The United States Mortality DataBase county mortality series includes period life table data indexed by calendar year for all US counties, not including the District of Columbia (DC).
For the time being, the life tables for all counties within all fifty states are available by state and sex as zipped, bundled archives. You must register and create a username and password to download any USMDB data products. To register or manage your account, click here. Data are available in two formats: comma-separated variable (.CSV) text files and R programming environment single-object (data-frame) binaries (.rds)
There are separate files for each state and sex (female "f", male "m", and both-sexes-combined "b") and the files are organized by county/county-grouping, age, and calendar year. Life tables are provided in the 5x1 format of age and time (by five year age group; 0, 1-4, 5-9, ..., 80-84, 85+ - and by single calendar year).
We group low-population counties and demographically similar neighbors (1084 total) together into county-groups (401 total) with historically consistent boundaries that exceed a minimum population threshold of 10,000 at any time point in our series. County groupings never cross state borders. 2071 counties are left ungrouped. A comma-separated variable lookup table linking the 1084 individual counties (by FIPS code) to the 401 groups can be found here. The individual counties in the life table files are listed by FIPS code.
Read the Methods Protocol (.pdf) for explanation of the file format and for more details about the Bayesian model and the life table calculations. The R and Stan codes to run the Bayesian model will soon be available here, as well. We also link to the companion reference, A Flexible Bayesian Model for Estimating Subnational Mortality (Alexander, Zagheni, Barbieri, Demography 2017) (link to pdf).