Working Paper: NBER ID: w27820
Authors: Cosmin L. Ilut; Rosen Valchev
Abstract: We develop a tractable model of limited cognitive perception of the optimal policy function, with agents using costly reasoning effort to update beliefs about this optimal mapping of economic states into actions. A key result is that agents reason less (more) when observing usual (unusual) states, producing state- and history-dependent behavior. Our application is a standard incomplete markets model with ex-ante identical agents that hold no a-priori behavioral biases. The resulting ergodic distribution of actions and beliefs is characterized by “learning traps”, where locally stable dynamics of wealth generate “familiar” regions of the state space within which behavior appears to follow past-experience-based heuristics. We show qualitatively and quantitatively how these traps have empirically desirable properties: the marginal propensity to consume is higher, hand-to-mouth status is more frequent and persistent, and there is more wealth inequality than in the standard model.
Keywords: Costly reasoning; Bounded rationality; Wealth inequality; Marginal propensity to consume
JEL Codes: C11; D83; D91; E21
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| costly reasoning (D91) | state and history-dependent behavior (D91) |
| state and history-dependent behavior (D91) | learning traps (C92) |
| learning traps (C92) | locally stable dynamics of wealth (C62) |
| costly reasoning (D91) | wealth distribution (D31) |
| costly reasoning (D91) | consumption patterns (D10) |
| persistent hand-to-mouth (HTM) status (C41) | reasoning framework (B41) |