According to Thomas Piketty [1], it's mainly a result of a bifurcation in education, and dramatically rising managerial compensation helped with tax cuts.
I personally also think this is the main answer. In other words, all our massive productivity increases have come from people with a ton of education, and they reap all the rewards, helped with lower taxes. The median worker isn't any more educated and hasn't reaped anything.
> ...I certainly do not believe that r > g is a useful tool for the discussion of rising inequality of labor income: other mechanisms and policies are much more relevant here, e.g., supply and demand of skills and education. For instance, I point out in my book (Piketty 2014a, ch. 8–9) that the rise of top income shares in the United States over the 1980–2010 period is due for the most part to rising inequality of labor earnings, which can itself be explained by a mixture of two groups of factors: rising inequality in access to skills and to higher education over this time period in the United States, an evolution which might have been exacerbated by rising tuition fees and insufficient public investment; and exploding top managerial compensation, itself probably stimulated by changing incentives and norms, and by large cuts in top tax rates...
[1] http://webcache.googleusercontent.com/search?q=cache:x918MB5...