When Bartels compared the policy preferences of the rich and poor to actual policy results (with controls) his results were disturbing. He finds that low-income preferences had virtually no effect on policy outcomes.
Then Bartels, in a deeply original and important contribution to the literature, estimates what the effect of equal representation would be on social spending, and uses that measurement to conclude that, by contrast, biased responsiveness reduces real social spending per capita by 28 percent on average. In the United States, he finds that the gap is around 40 percent.
To repeat: Social spending in the United States is 40 percent lower than it would be if policymakers didn't disproportionately respond to the rich.