Capital Gains Overhang with a Dynamic Reference Point (with B Summers & D Duxbury)
Management Science, Volume 66, Issue 10 (October 2020)
Abstract: Financial models incorporating a reference point, such as the Capital Gains Overhang (CGO) model, typically assume it is fixed at the purchase price. Combining experimental and market data, this paper examines whether such models can be improved by incorporating reference point adjustment. Using real stock prices over horizons from 6-months to 5-years, experimental evidence demonstrates that a number of salient points in the prior share price path are key determinants of the reference point, in addition to the purchase price. Market data testing is then undertaken using the CGO model. We show that composite CGO variables, created using a mix of salient points with weights determined in the experiment, have greater predictive power than the traditional CGO variable in both cross-sectional US equity return analysis and when analyzing the performance of double-sorted portfolios. In addition, future trading volume is more sensitive to changes in the composite CGO variables than to the traditional CGO, further emphasizing the importance of adjusting reference points.
Sequential Reference Point Adaptation: Evidence from a Five-Year Stock Chart Experiment (with D Duxbury)
Finance Research Letters, forthcoming
Abstract: Investor reference points determine whether stock positions are coded as gains or losses, shaping selling behavior and reference-dependent phenomena such as the disposition effect and capital gains overhang. Prior experimental estimates rest on end-of-horizon designs that elicit a single reference point only after the full price path is observed. We use a sequential elicitation design that records reference points month by month as experienced investors follow realistic five-year U.S. stock-price paths, yielding 39,173 reference points from 767 investors. Sequentially elicited reference points update continuously, with 50 to 57% of each monthly price change passing through. Updating is pulled toward both the purchase price and the current price, with a larger estimated coefficient on the current price. Running maxima and minima, prominent in end-of-horizon estimates, provide no evidence of positive adjustment toward historical extrema, and the estimated sequential specification implies lower mean absolute deviation from the current price than the two end-of-horizon formulas. Under sequential elicitation, reported reference points are path dependent and closely tied to recent prices, and differ materially from values implied by end-of-horizon formulas.
Shareholder-Weighted Returns and Stock Return Predictability (with Y Zhou)
International Review of Financial Analysis, forthcoming
Abstract: We ask whether weighting a stock's return path by its current shareholder base reveals information that price only variables overlook. Using daily volume to reconstruct the current shareholder base, we estimate three intuitive variables from a stock's return path: a cumulative return that reflects the profit or loss of current holders, an average return that captures the typical daily experience of holding the stock, and a scale-invariant gain-loss ratio that records the net fraction of shareholders who are in profit. The shareholder-weighted variables forecast one, three, and six-month returns, subsume capital gains overhang, and complement classic momentum. Evidence from portfolio sorts and cross-sectional regressions remains strong after controlling for standard characteristics and across exchanges, price segments and sub-periods, including the post-2000 era. For practitioners, the new shareholder-weighted variables improve risk-adjusted returns relative to traditional momentum strategies and are easy to implement.
Corporate ESG performance, mispricings and gains from mergers and acquisitions (with Y Zhou, W Rudkin, Y Zeng)
Working Paper
Abstract: Previous research shows that realized and unrealized outcomes influence subsequent risk-taking behavior, albeit in opposite ways. We further decompose unrealized outcomes into adapted and unadapted components, which have distinct implications for risk-taking. Adapted returns are internalized as individuals adjust their reference point, while unadapted returns remain externalized. Our empirical evidence shows that adapted and unadapted returns exert opposing influences on future stock returns. Notably, the highest-performing stocks have a distant past return that is low (reflecting an adapted loss) paired with a recent past return that is high (reflecting an unadapted gain).
Decomposing Unrealized Returns into Adapted and Unadapted Components Using Reference Points (with D Duxbury)
Working Paper
Abstract: Previous research shows that realized and unrealized outcomes influence subsequent risk-taking behavior, albeit in opposite ways. We further decompose unrealized outcomes into adapted and unadapted components, which have distinct implications for risk-taking. Adapted returns are internalized as individuals adjust their reference point, while unadapted returns remain externalized. Our empirical evidence shows that adapted and unadapted returns exert opposing influences on future stock returns. Notably, the highest-performing stocks have a distant past return that is low (reflecting an adapted loss) paired with a recent past return that is high (reflecting an unadapted gain).
New Project: Distorted Returns- The Influence of Linear Scales (with B Summers)