Working Papers

Risk Revisited (SSRN)

with Yucheng Liang

This study explores how the historical price path of a stock shapes investors' perceptions of risk. In a series of experiments, we present participants with real and fabricated stock price charts and elicit both incentivized and unincentivized measures of their risk perceptions. We document that a parsimonious set of three conceptually independent features, i.e., recency, cluster, and sign, can explain large proportions of the variation in risk perceptions. We find that rational inferences about the return data-generating process cannot fully explain the effect and that perceptions both related and unrelated to subjective standard deviation of returns are affected in a consistent way. Using real stock market data in the U.S., we show that these three features help explain cross-sectional variations in returns, and predict trading volume and future volatility.

Non-fungible Cash in the Stock Market (SSRN)

with Ning Zhu

Investors perceive cash in their savings accounts differently from cash recycled in their stock brokerage accounts. We propose a novel “temperature” framework for financial resources, where the former is labeled “cold cash” and the latter “hot cash.” We find that individual investors buy stocks more cautiously with colder cash. Exploiting the quasi-natural experiment of the 2016 Chinese IPO lottery reform, we show that the effect of cash temperature on investors' cautiousness in stock selection is causal. An online experiment, where cash temperature is randomly assigned, indicates that the differential loss aversion is a potential channel for the cash-temperature effect. Building on the observational and experimental findings, we propose a portfolio choice model that features preferences with temperature-dependent sensitivity to future gains and losses. The model generates the empirical patterns documented in this paper and provides a cash-temperature perspective for other puzzles in the literature.

Differential Origins and Impacts of Extensive and Intensive Flows: Theory and Evidence (SSRN)

with Allen Hu

We decompose the trading volume of individual stocks into extensive and intensive flows, where extensive flows are trading volumes of investor entry and exit while intensive flows are position adjustments by continuing holders. To distinguish their differential origins and impacts, we model extensive and intensive flows in a variant of Grossman-Stiglitz. The model features a group of constrained investors, who can hold only one of the two risky assets, and a group of unconstrained arbitrageurs, who can hold both assets. Constrained investors endogenously move between two assets, creating extensive flows in the market. After shocks to the payoffs of both assets, only continuing holders can observe the true signal, while others need to infer the signal from price changes. Since the extensive flows are unobservable, they are not accurately incorporated into signal inference, which leads to the Subjective Flows Equilibrium (SFE). Compared to the benchmark of Rational Expectation Equilibrium, SFE predicts differential effects of extensive and intensive flows on asset returns and volatility. We provide evidence to confirm these predictions using daily transaction records of a large group of individual investors.

Dual Beliefs: Subjective Learning of Trading Talent (SSRN)

This paper studies how investors evaluate themselves and its implications. In my model, investors form subjective beliefs about both the stock currently held in portfolio and their trading talent and update their beliefs through learning with fading memory. I calibrate the memory decay parameters to account-level trading records and show that beliefs updating for talent is about 7 times more sensitive to return signals than that for the stock in portfolio. Consequently, the model predicts that stock replacing typically happens after a good performance of the existing stock, providing a dual learning perspective on the disposition effect. This framework also accounts for the widely documented performance-contingent trading intensity and investor attrition, which cannot be reconciled with the decreasing-gain property of the standard Bayesian learning.