In fairness, it's a bit of both. The position of parents on the income scale vs. the position of children on the income scale is (too) highly correlated all acr

In fairness, it's a bit of both. The position of parents on the income scale vs. the position of children on the income scale is (too) highly correlated all across the income scale. The mechanisms for this correlations are different ones at different points of the scale, as you say, and certainly the very top is the part that needs the most fixing.