Working Paper: NBER ID: w31081
Authors: Nicholas C. Barberis; Lawrence J. Jin
Abstract: Motivated by neural evidence on the brain's computations, cognitive scientists are increasingly adopting a framework that combines two systems, namely “model-free” and “model-based” learning. We import this framework into a financial setting, study its properties, and use it to account for a range of facts about investor behavior. These include extrapolative demand, experience effects, the disconnect between investor allocations and beliefs in the frequency domain and the cross-section, the inertia in investors’ allocations, and stock market non-participation. Our results suggest that model-free learning plays a significant role in the behavior of some investors.
Keywords: investor behavior; model-free learning; model-based learning
JEL Codes: D03; G02; G11
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| model-free learning (C51) | investor behavior (G41) |
| previous allocations (D61) | future allocations (G31) |
| positive stock market returns (G17) | future allocations (G31) |
| distant past returns (N00) | future allocations (G31) |
| weighted average of past returns (G12) | demand for risky assets (G19) |
| recent market returns (G17) | probability of good returns (G17) |
| beliefs (D83) | allocations (H77) |
| model-based system (C53) | beliefs (D83) |