How do I balance my panel?

I’m interested in studying the personal income of veterans and non veterans from 2006 to 2012. Additional variables include the period of the veteran’s service, whether the person is self or wage-employed, and demographic variables such as race and sex.

Everytime I submit for my extract, however, I receive a panel filled with mostly veterans or vice-versa. How can I get a data set that has an even balance (50%) of my two key binary variables; class of worker and veteran status? I’ve tried tweaking the sample sizes to no avail.

IPUMS-USA does not currently support the kind of case selection you are describing. The ACS data you are downloading represents a sample of the entire U.S. population from each year. Within each sample, every person is given a weight (PERWT) calculated by the Census Bureau to account for sampling error and the like, meant to multiply that person by the amount of people they represent within the total U.S. population for the year they were sampled. Because creating a 50% veteran, 50% non-veteran subsample could introduce new sampling error, the weights may no longer be correct. You could use the IPUMS-USA extract of all persons from 2006-2012, and generate your own random sample of veterans and non-veterans, resulting in your desired sample size.

I hope this helps.