Rational inattention and the dynamics of consumption and wealth in general equilibrium

Abstract

We use a recursive utility version of a basic Huggett (1993) model to study the cross-sectional dispersion of consumption and wealth (relative to income). The basic model implies too little dispersion compared to the data, whereas a one-parameter extension to include rational inattention (limited information processing) delivers a better fit to both facts in general equilibrium.In particular, intertemporal substitution plays an important role in determining the two key dispersion moments via affecting the degree of optimal attention in equilibrium. Alternative models that rely on habit formation, incomplete information about current income, or borrowing constraints are not consistent with the facts we document.

Forward Guidance without Common Knowledge

“…48 Although this assumption is separate the information-flow constraint (100), it is standard in the literature (e.g., Woodford, 2003a, Mackowiak and Wiederholt, 2009, Luo et al, 2017 and seems appealing if one interprets the noise as the product of cognitive limitations. It is also broadly consistent with experimental evidence (e.g., Khaw, Stevens and Woodford, 2016).…”

A Model of Expenditure Shocks

“…Facts 1 and 2 hold even when focusing exclusively on nondurable consumption, which suggests that the durability of goods is not a driving force.2 Relatedly,Luo et al (2017) argue that borrowing constraints have difficulty replicating the feature of the PSID that the cross-sectional dispersion in consumption relative to income is nearly constant across the income or wealth distribution.…”

The Optimal Amount of Attention to Capital Income Risk and Heterogeneous Precautionary Saving Behavior

“…As argued in Sims (2006Sims ( , 2010, ex-post normal distribution is optimal in rational inattention models only under linear quadratic Gaussian setup. However, for the reason of tractability, the present paper and many studies in the rational inattention consumption-saving models with labor income risk, such as Luo et al (2017), assume a normal ex post distribution. In order to show that our main findings are valid without such approximation, in the next section we will relax the assumption of ex post Gaussianity and solve the model numerically for an optimal joint distribution of the unobservable state variable and the control variable as in Sims (2006).…”