Zigang Li

Professional house price forecast data are consistent with a rational model where agents must learn about the parameters of the house price growth process and the underlying state of the housing market. Slow learning about the long-run mean generates overreaction to forecast revisions and a modest response of forecasts to lagged realizations. Heterogeneity in signals and priors about the long-run mean helps the model account for cross-sectional dispersion in forecasts. Introducing behavioral frictions helps improve the model's predictions for short-horizon overreaction and dispersion. Using a cross-section of forecasters and a term structure of forecasts are crucial for inference.

We show that powerful autonomous learning does not ensure trading dominance in competitive markets, even when humans hold systematically biased beliefs. We characterize how information structure and market sophistication bound reinforcement-learning traders' performance relative to humans. Humans receive private signals, potentially including soft information unavailable to algorithms, and differ in their ability to extract information from prices and anticipate others' trading, as captured by level-k reasoning. AI traders learn demand schedules from prices and realized profits and may also receive private signals. Despite heterogeneous initial conditions and persistent differences in learning parameters, all AI traders converge to a common probabilistic trading policy, generating algorithmic herding, which amplifies the equilibrium-price response to aggregate AI demand and compresses profits. Without an informational advantage, AI is outperformed by the most sophisticated humans because it learns from market data generated by average, rather than frontier, human sophistication. Even the least sophisticated humans can outperform AI when their private information is either very precise or very imprecise. AI sophistication rises with that of the humans who compete with and train it.

We decompose investor demand into its extensive margin—initiating or liquidating a position—and intensive margin—scaling an existing position. Extensive demand is economically large, accounting for 40% of institutional and 80% of retail flows, and its estimated price elasticity is closer to zero and often positive. Extensive demand is primarily driven by past returns and attention, whereas intensive demand reflects fundamentals. To account for these findings, we develop a Grossman-Stiglitz-style model of inelastic markets in which extensive flows act as unobservable endogenous demand shocks and investors learn from prices based on their subjective expectations about these flows. Learning from prices leads to a more inelastic, and even upward-sloping, extensive demand curve. Embedding the model in an asset demand system, we structurally estimate both elasticities under alternative specifications of flow expectations and find that extrapolative expectations best fit the data. Reduced-form evidence and counterfactual analysis both show that as the share of extensive demand rises, prices, volatility, and firm-specific price informativeness increase, while market-wide informativeness declines.