The numbers I work on are from the Labour Force Surveys (called, in America, the Current Population Survey) which are household surveys carried out on a large sample of the population. These count people, not jobs, (which the payroll statistics would do). So there shouldn't be a double count when one person holds two (or more) jobs.
The next part of this 3-part epic is about hourly rates of pay. I'll try to bring second jobs in there, since it seems like the right place.
