Learning from Experience in the Stock Market

Working Paper: CEPR ID: DP9845

Authors: Anton Nakov; Galo Nuo

Abstract: New evidence suggests that individuals "learn from experience," meaning they learn from events occurring during their own lifetimes as opposed to the entire history of events. Moreover, they weigh more heavily the more recent events compared to events occurring in the more distant past. This paper analyzes the implications of such learning for stock pricing in a model with finitely-lived agents. Individuals learn about the rate of change of the stock price and of dividends using a weighted decreasing-gain algorithm. Information is dispersed across age cohorts with older agents having larger information sets than younger ones. In the model, the stock price exhibits stochastic fluctuations around the rational expectations equilibrium due to successive waves of optimism and pessimism. We demonstrate how this heterogeneous-beliefs model can be approximated by an economy with a representative agent who updates his beliefs following a constant-gain learning scheme. The aggregate gain parameter of the approximation is a nonlinear function of the survival rate and of the individual gain parameters.

Keywords: asset pricing; constant-gain learning; dispersed beliefs; heterogeneous beliefs; OLG

JEL Codes: D83; D84; G12


Causal Claims Network Graph

Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.


Causal Claims

CauseEffect
Learning from experience (C90)Oscillating dynamics in stock prices (C69)
Loss of historical knowledge (N00)Bias in beliefs about dividends and stock prices (G40)
Low retirement rate (J26)Non-convergence of asset price to rational expectations equilibrium (G19)
Stronger discounting of older information (D15)Momentum in asset pricing (G19)
Asset exposure constraints bind (G12)Trend reversal in asset pricing (G19)
Heterogeneous beliefs model (D80)Representative agent model using constant-gain learning scheme (C73)
Learning from experience (C90)More volatility in price-dividend ratio compared to rational expectations model (G19)

Back to index