I am conducting a study that examines the impact of a state-level policy on individuals living in counties that have physician shortages. My main data source for the study is individual-level survey data for the years 1996-2015 but I also use county-level data for the same time period to control for county-level characteristics such as the number of physician assistants per 100,000 population. However, I am not certain as to how to obtain descriptive statistics (means) for the county-level per capita characteristics (i.e. physician assistants per 100,000 population). Do I obtain this at the individual-year level using the individual-level survey weights or do I obtain this at the country-year level? If it is the county-year level do I weight by county population or not? And if I am to weight, do I weight using the county-year population (i.e. allow the county population to vary across years) or do I weight using the population in a particular year (e.g. 2000)?

P.S. The county-level per capita characteristics were obtained for each year using the county population in that year.

Thanks