Marcel Prokopczuk

More than two-fifths of anomaly premia can be explained by a small set of proxy variables that capture key economic concepts. Using a decomposition approach that incorporates both the crosssectional and time-series dimensions of potential explanations, we analyze the return premia of a broad set of prominent anomalies in the cross-section of stock returns. The main drivers of anomalies are mispricing, limits to arbitrage, and lottery demand. Debt issuance, low-risk, profitability, short-term reversal, and value anomalies are explained relatively well, whereas accruals, momentum, profit growth, and quality anomalies are accounted for only to a limited extent.

Machine learning-based stock market beta estimators outperform established benchmark models both statistically and economically. Analyzing the predictability of time-varying market betas of U.S. stocks, we document that machine learning-based estimators produce the lowest forecast and hedging errors. They also help to create better market-neutral anomaly strategies and minimum variance portfolios. Among the various techniques, random forests perform the best overall. Model complexity is highly time-varying. Historical stock market betas, turnover, and size are the most important predictors. Compared to linear regressions, allowing for nonlinearity and interactions significantly improves predictive performance.

We comprehensively investigate the robustness of well-known factor models to altered factor formation breakpoints. Deviating from the standard 30th and 70th percentile selection, we use an extensive set of anomaly test portfolios to uncover two main findings: First, there is a trade-off between specification and diversification. More centered breakpoints tend to result in less (idiosyncratic) risk. More extreme sorts lead to greater exposure to the underlying anomalies and thus to higher average returns. Second, the models are robust to varying degrees. Hou et al.’s model [2015, Digesting Anomalies: An Investment Approach, Review of Financial Studies 28, 650–705] is much more sensitive to changes in breakpoints than the Fama–French models.

We analyze the relation between time-series predictability and factor investing. We use a large set of financial, macroeconomic, and technical variables to time-series-manage the market portfolio. A combination of the out-of-sample market excess return forecasts of all variables yields a managed market portfolio that generates alphas relative to cross-sectional factor models that exceed 5% per annum. More broadly, the relation between time-series evaluation measures and (multifactor) alphas is weakly positive but complex. The variables’ predictability for future returns is more important than that for volatility. Finally, we document that managed market portfolios based on lagged factor realizations also perform well. This paper was accepted by Lukas Schmid, finance. Supplemental Material: The online appendix and data are available at https://doi.org/10.1287/mnsc.2022.4459 .

We extract proxies for trader conscientiousness from plain text to explain the cross-section of individual and institutional trader performance. The literature claims conscientiousness to be a main determinant of workplace performance. We employ measures like text length, error frequency, sentence simplicity, and lexical diversity to construct a conscientiousness index. Our findings suggest that more conscientious traders outperform their peers on a risk-adjusted basis. This implies that personality affects investment decisions. Our data stems from a web-based platform that gives individuals the opportunity to set up small investment funds other users can invest in.

We study factor pricing and market integration across major asset classes. Factor models specializing in one asset class have limited pricing power for other asset classes. Thus, we reject perfect market integration. However, an optimal integrated factor model across asset classes can effectively characterize the returns of multiple asset classes and provide a useful benchmark for multi-asset, multi-factor investing. The optimal model includes several equity and corporate bond factors, suggesting the presence of multiple systematic return drivers. Despite this, there appears to be some degree of cross-market linkages, as the optimal model does not require factors from all asset classes.

Risk proxies such as beta, volatility or standard deviation are conventionally measured assuming that investors have symmetric risk preferences, with upside and downside deviations from the expectation being equivalently undesirable. Responsible investors, however, have the dual aims of enhancing upside potential while ensuring a sufficient level of downside protection. To measure this ambition, we relax the assumption of symmetric risk preferences and measure upper and lower partial moments separately. Specifically, we utilise a non-symmetry option pricing research design to test if responsible investors could live up to their ambition. We find that those who are simply Principles for Responsible Investment (PRI) members do not fulfil the ambition while financial service firms with highly rated responsible investment processes can actually achieve both aims: enhance upside potentials and protect from downside risks.

This paper evaluates the predictive performance of machine learning techniques in estimating time-varying betas of US stocks. Compared to established estimators, tree-based models and neural networks outperform from both a statistical and an economic perspective. Random forests perform the best overall. Machine learning-based estimators provide the lowest fore-cast errors. Moreover, unlike traditional approaches, they lead to truly ex-post market-neutral portfolios. The inherent model complexity is strongly time-varying. The most important predictors are various historical betas as well as fundamental turnover and size signals. Compared to linear regressions, interactions and nonlinear effects enhance the predictive performance substantially.

We confront prominent asset pricing models with the classical out-of-sample cross-sectional test of Fama and MacBeth (1973). For all models, we uncover three main findings: (i) the intercept coefficients are economically large and highly statistically significant; (ii) the cross-sectional factor risk premium estimates are far below the average factor excess returns; and (iii) they are generally not statistically significant. Thus, our findings show that the models do not only fail the equilibrium condition of the time-series test, but are also inconsistent with the weaker no-arbitrage condition. Overall, all new factor models cannot accurately explain the cross-section of stock returns.