I am analyzing the effect of immigration on native-born workers in the construction sector. I am analyzing only those individuals who have occ2010 codes 6210 through 6765. I want to obtain the average incomes (and unemployment rates) for these native-born construction workers in each Metropolitan Statistical Area (MSA) in each year so that I can use these as dependent variables to run regressions. Do I need to use the perwt variable to obtain MSA-level averages that are representative of all native-born construction workers for each MSA? More specifically, if I take a weighted average using the perwt variable to calculate the average income (unemployment rate) for natives in each MSA in each year, will this weighting procedure give me statistics that are representative of the native construction workers in each MSA or will this give me statistics that are potentially representative of some other group of workers (possibly of workers in non-construction professions or of construction workers at the national level instead of the MSA level)? I am currently using the simple average and need to know if using the weighted average will provide more reliable estimates of construction workers in each MSA in each year.
If I want to obtain MSA-level averages of annual income for construction workers, do I need to use perwt?