That is an underexplored question, and it is considered in the job market paper of Chunzan Wu, from the University of Pennsylvania. Here is the abstract, which will unsettle many people:
Since the 1970s, income inequality in the U.S. has increased sharply. During the same time span, the U.S. federal income tax has become less progressive. Why? I examine this question in a Ramsey optimal tax policy framework. Within this framework, the tax policy is determined by: (1) a set of Pareto weights representing the government’s preference over different households; and (2) household lifetime utilities summarizing the effects of economic fundamentals. I first study the changes in economic fundamentals using an overlapping generations incomplete-markets life-cycle model with heterogeneous households. The model features both endogenous human capital accumulation and household labor supply and is calibrated to the U.S. economy in the 1970s and 2010s. Then I use this economic model to determine whether the change in income tax is the result of an optimal policy response to changing economic fundamentals or the consequence of a change in Pareto weights. I interpret the latter as changes in the political influences of various income groups. I find that: (1) changes in economic fundamentals alone induce a less progressive optimal income tax and can account for 40% of the reduction in progressivity we observe; and (2) the change in Pareto weights required to explain the remaining part of tax policy change favors high-income households and also implies less valued government services. Finally, using a stylized political economy model, I discuss potential explanations for this change in Pareto weights such as the lower cost of conveying information to swing voters and the rising inequality of voter turnout among different socioeconomic groups.
The paper is here (pdf).