The (comparatively) lilywhite states would be a good starting point. If the factor disappears there, then... but IIRC it doesn't.
But bear in mind that income and class are not identical. College students, unless their parents are really super duper hyper rich, tend to live in relative penury (and collect a lot of nontangible benefits from school and from home that aren't going to show up in an income question, making them appear poorer than they are.) And we know how they vote.
Of course, class in itself is almost impossible to identify in a society like America (which doesn't mean that it doesn't exist).
Another method could be to identify how the minority electorate breaks down in terms of income (ie which percentage is under 30k, which 30-50k, etc.), then infer a uniform vote regardless of income (a debatable premise, but probably not too far from reality) and make some math. I know that this kind of statistical tricks made on already dubious data can get pretty silly, but better than nothing.
You can probably do it with Black voters. The data I've seen indicates that Hispanic voters tend strongly to be affected by income though. I've seen plenty of data on this though so I'm sure it's available.
