Via Kevin Drum and Greg Sargent a study from Thomas Hungerford of the Congressional Research Service (PDF) indicates that in the 1991-2006 period the distribution of wages, taxes, and social insurance benefits all reduced the level of inequality in the United States. This was offset by a big pro-inequality rise in the returns to capital, as witnessed by capital gains income, retirement income, and business income.
What inclinations one draws from this empirical finding are going to be strongly colored by your priors. If you believe the tax preference for investment income serves no useful pro-growth purpose, then the obvious upshot is to double-down on advocating higher taxes on investment income. Those inclined to believe the reverse are going to say the upshot is that there's no viable policy remedy to growing inequality.
My judicious middle ground take is that this highlights the importance of disaggregating "capital" as it's construed by the actual functioning of the American economy. "Capital" for these purposes includes what the classical economists would have considered "land"—the valuable dirt that hosts San Francisco houses buildings and North Dakota oil wells. It also includes intellectual property. It also includes, of course, the stuff that a classical economist would have recognized as "capital"—machines used in the production process. I continue to think that conservatives are write to believe that the tax code should in fact favor the accumulation of production equipment rather than the accumulation of consumption goods. But land and natural resource extraction should almost certainly be taxed much more heavily (it's in Ricardo) and we should be acting more decisively to make municipal policy less biased toward creating land scarcity. On intellectual property, the agenda shouldn't be so much to tax the profits more heavily as to reduce or eliminate them entirely by weakening or eliminating these protections.