What we do know is that we can explain about 50% of the difference in earnings through observable factors, variables that we collect. Those would include, for example, years of experience. Women tend to have about 3.1 fewer years of experience. Even the kind of experience they have is very different. It's often part-time part-year or part-time full-year, whereas men's is full-year full-time. So we know that we can explain some of the difference that way.
We also know that the types of occupations make a difference. That can account for about 5% of the difference. Unionization rates are slightly lower for women. That can make up another 1% of that difference. But for 50% of that difference, we don't know. That is from unobservable factors.
One thing we didn't look at for that study was that when women work full-time, they often work fewer hours than men do. You're comparing full-time workers, but women, for example, work about 600 fewer hours per year than men do, even when you're looking at full-time workers.
So there are many things wrapped up in it, but we figure that we can explain about 50% of the gap, and the other 50% is caused by unexplainable variables, or at least variables that we didn't have.