²¹An earlier version of the estimation allowed the splurge factor to differ across consumers, either between education groups or within groups via a common-distribution structure. The lottery-MPC and wealth-distribution moments did not provide evidence against the null of a single splurge common to all consumers. Imposing this restriction reduces the parameter dimensionality of the joint estimation and tightens identification of the discount-factor distribution; we therefore use a single $\varsigma$ throughout.