I think this sounds like a fun project for avoiding my personal research. Perhaps I will go crunch some data.
However, as a sociologist, I think income is probably not the best measure to use for this question. First of all, income fluctuates from year to year and (hopefully) increases substantially over the life course. Also, once a couple gets married, the fact that they are married to specific individuals likely shapes their decisions about their own career path in the future. (I.e., lots of women cut back on careers to raise kids.) To use income, then, we’d need income from something like a year before the couple gets engaged. (That is possible to come by, but not anytime soon.) Further complicating things is that men are often older than the women they marry. Since age is correlated with income, that makes the picture less clear.
I’d use level of education, maybe in five rough groups: high school drop out, high school diploma/GED, some college, BA, graduate/professional degree. Level of education is often used as a measure of social class in sociology. Is it perfect? No. But it’ll do. (What would be awesome for this question would be some sort of data on parents’ education/occupation/income, but I don’t know where to find that quickly.)
I’d do my analysis separately for men and women. My suspicion is that, especially for older generations, it was more common for women to marry men with higher levels of education than vice versa.
And speaking of age, I’d probably do some age groups. Something like: 20-29, 30-39, 40-49, 50-59, 60-75.
Finally, I think I’d want to throw race/ethnicity into the model. Why? Because I know that compared to black men, black women are way over-represented in higher education, but things are much more even for whites.
Anything else that should go into this analysis?