I show that if it is costly for households to process information about asset returns, a model with ex-ante identical households features persistent inequality. The steady state has a two-agent structure, with inequality maintained by a complementarity between attention and wealth: wealthy households have stronger incentives to pay attention to asset choices, and so earn higher returns than asset-poor households.Fiscal expansions are less powerful in this model than in a standard model with heterogeneous discount factors, because when an expansion causes poor households to start saving, they also increase their attention. They therefore earn higher interest rates, and so save even more, smoothing the windfall from the policy over a longer time period. I provide evidence for this increase in attention using crossstate variation in uncertainty about savings interest rates in the aftermath of the 2017 Tax Cuts and Jobs Act in the US. The effects of fiscal policy therefore depend not just on the existence of inequality, but also on the cause of that inequality.