Imagine there are two groups of 30-year-olds: A) Those who were offered university places, but decided not to go. B) Those who were not offered university pla

Imagine there are two groups of 30-year-olds: A) Those who were offered university places, but decided not to go. B) Those who were not offered university places. I would expect group A to have a significant 'earnings premium' over group B. Even though neither group attended university, group A are more likely to be smart, hard working, lucky, or whatever drives career success. If you want to estimate the earnings premium of a university degree, you have to compare the folks who went to university with other folks just like them, i.e. with folks in group A. Not with a mixture of groups A and B. > Is there another thing you would attribute it to? Yes, some of the earnings premium is due to the students pre-university chances of success. > It's not hard to simply ask "of this profession, > what is the salary of those w/ a degree, and those w/o" at the national level This wouldn't measure the thing we're trying to measure, because it suffers from selection bias.