Abstract: We introduce forecast-agnostic (FA) portfolios that exhibit out-of-sample market-timing ability without relying on estimated predictive coefficients. These portfolios go long or short the market based on the level of a predictor variable, thereby avoiding the instability and estimation error that undermine traditional market-timing strategies. Despite using predictor variables that typically deliver negative out-of-sample (OOS) R-squared values, FA portfolios deliver significantly positive alphas on average. We explain these seemingly contradictory phenomena by interpreting regression coefficients as portfolio returns: genuine predictability is necessary for high portfolio returns, whereas achieving a positive OOS R-squared additionally requires the ability to forecast the returns on the FA portfolios themselves. Simulations show that FA alphas have power to detect predictability that extends beyond in-sample diagnostics and the OOS \(R^2\).
Patrick Adams, Massachusetts Institute of Technology
Abstract: Do temporary stock price crashes matter for long-term investors? I use over 25 years of U.S. income tax data to characterize the savings behavior and risk exposures of high-income working-age households. Aggregate stock price crashes coincide with persistent declines in wage and private business income for many of these households, who take large drawdowns from their liquid assets – including stocks – in response. I develop a life-cycle model with consumption adjustment frictions to match this observed savings behavior and determine its portfolio choice implications. Investing in stocks is risky when falling income and rigid expenditures may force investors to liquidate their holdings at temporarily-depressed prices, resulting in low optimal portfolio shares. These results challenge the conventional wisdom that the stock market is relatively safe for long-term investors.
Abstract: Macroeconomic announcements lead to the repricing of previous firm-specific earnings news, generating cross-sectional heterogeneity in risk compensation. When firms announce earnings, investors form joint beliefs about firm-specific and aggregate conditions. Subsequent macroeconomic announcements reveal information about the aggregate state of the economy, prompting investors to reassess the firm-specific component of prior earnings news. We develop a dynamic general equilibrium model in which investors rationally learn from both earnings and macroeconomic announcements to quantify this repricing channel. Empirical evidence supports the model's predictions: on macroeconomic announcement days, firms with recent earnings news earn a lower risk premium relative to those without, and this effect is stronger for firms whose earnings announcements were more informative about aggregate conditions.
Discussant: Harjoat Bhamra, Imperial College London
Abstract: A wide range of empirical techniques cannot accurately estimate a policy event’s causal effects, because agents adjust decisions in advance based on beliefs about future policy outcomes. We show how researchers can measure anticipation bias and refine estimates, by integrating reduced-form and structural estimation. Our novel procedure estimates agents’ beliefs by comparing model-predicted outcomes to reduced-form estimates, and only requires a single policy change to implement. We illustrate the importance of this approach by applying it to the Paris Agreement, which is frequently used to understand how agents respond to an increase in climate regulatory risk. We find that before Paris, agents assigned a 77% likelihood to an agreement with some form of emissions penalties being reached. Our estimates imply that anticipation led high-emissions firms to reduce investment and increase cash holdings, relative to low-emissions firms. Thus, reduced-form studies of the Paris Agreement may understate its causal effects by up to 50%.
