My point was that sample size is too small. According to Standard Occupational Classification [
https://www.bls.gov/soc/] there are 459 broad occupations.
Multiply that by age group and you still haven't tossed a coin even a single time for every combination of occupation/age group.