The piece of the book that we saw has two distinct things going on: attempting to explain the determinants of Americans' changing attitudes towards taxes without regard to how this affects political outcomes, and exploring why tax progressivity has recently (at least in the last 7 years) declined.
For the former, Campbell, while still weighing aspects of her approach, at present uses "perceived cost-benefit theory," an adaptation of rational choice theory that takes into account how taxes' form and structure can dramatically shape perceptions wholly independently of substance. I was basically okay with this, although her use of the perceived benefit side can be (and was) questioned. Others argued that ideology plays a bigger role than she gave it credit for.
As the paper acknowledges, rational choice has a tough road to hoe in public policy even if one believes it governs behavior in, say, financial markets, given the voting paradox. If it isn't rational, in a narrowly economic sense, for me even to inform myself (much less vote) in a mass society where I can't significantly affect the outcome, then we may be using the wrong theory. What makes self-interest almost work a little better here is that people have some inclination to believe that things that are good for themselves are also good policy. I attribute this to the Pleistocene incentive to be self-interested yet genuinely convinced of the broader merits when arguing "policy" (e.g., where should we go tomorrow) with the other hunter-gatherers. But there is no evolutionary impetus to deploy one's most advanced cognitive tools and efforts to solve complicated problems that one has no power to decide anyway.
Insofar as rational choice (almost) works in spite of itself with respect to simple choices that would unmistakably be better rather than worse for oneself, it may seem paradoxical that the median voter has let the system become so much less progressive in the last few years. No policy judgment about optimal progressivity is needed for one to posit that most voters would rather have rich people pay a bit more tax if this meant that they themselves could pay a bit less. But this basis for considering reduced progressivity paradoxical would only apply in a zero-sum framework, where all current voters are paying a fixed amount of tax and the question is who will pay more and who less. When you can simply run up the tab at the expense of future generations, as the Bush Administration has done, then current voters both (a) have no reason to bother figuring it out any more (especially if agenda control means they aren't comparing it to alternative changes that lose the same revenue) and (b) could rationally say fine, I don't care about future generations.
So perhaps there is no paradox after all, even leaving aside the role of ideology, normative issues, complicated tax incidence questions, and changed worldwide economic conditions ithat arguably reduce the optimal level of progressivity from where it was 40 years ago.