Working Paper: NBER ID: w23917
Authors: Ian Dewbecker; Charles G. Nathanson
Abstract: We study an ambiguity-averse agent with uncertainty about income dynamics who chooses what aspects of the income process to learn about. The agent chooses to learn most about income dynamics at the very lowest frequencies, which have the greatest effect on utility. Deviations of consumption from the full-information benchmark are then largest at high frequencies, so consumption responds strongly to predictable changes in income in the short-run but is closer to a random walk in the long-run. Whereas ambiguity aversion typically leads agents to act as though shocks are more persistent than the truth, endogenous learning here eliminates that effect.
Keywords: ambiguity aversion; income dynamics; directed attention; nonparametric learning; consumption behavior
JEL Codes: C14; D83; E21
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Directed attention to low-frequency income dynamics (E25) | Stronger response in consumption to predictable income changes in the short run (E21) |
| Insufficient learning about high-frequency income characteristics (C58) | Strong response in consumption to predictable changes in income (D11) |
| Agent's learning process (D83) | Balances biases introduced by ambiguity aversion (D81) |
| Ambiguity aversion (D81) | Systematic errors in consumption growth at high frequencies (E21) |