Content uploaded by Jonathan Lewellen
Author content
All content in this area was uploaded by Jonathan Lewellen on May 07, 2014
Content may be subject to copyright.
A preview of the PDF is not available
The Cross-section of Expected Stock Returns
Abstract and Figures
This paper studies the properties and predictive ability of return forecasts from Fama-MacBeth cross-sectional regressions. These forecasts mimic how an investor could, in real time, combine many firm characteristics to get a composite estimate of a stock's expected return. Empirically, the forecasts exhibit significant cross-sectional variation and have strong predictive power for subsequent stock returns. For example, using ten-year rolling estimates of Fama-MacBeth slopes and a cross-sectional model with 15 firm characteristics (all based on low-frequency data), the expected-return estimates have a cross-sectional standard deviation of 0.90% monthly and a predictive slope for future monthly returns of 0.77, with a t-statistic of 10.17.
Figures - uploaded by Jonathan Lewellen
Author content
All figure content in this area was uploaded by Jonathan Lewellen
Content may be subject to copyright.
... In general, the rapidly growing amount of literature can be divided into two strings: The application of ML techniques to crosssectional data in order to validate advantages in the context of large data sets and the application of ML techniques to time series data to improve forecasting accuracy (see Karolyi and Van Nieuwerburgh 2020). In this regard, econometric benchmarks are given by studies of Fama and French (2008) and Lewellen (2015) who provide fundamental discussions on cross-sectional data and Welch and Goyal (2008) and Koijen and Nieuwerburgh (2011) who provide elementary findings regarding econometric benchmarks for time series data. This paper draws on the first string of literature and addresses the interpretability of machine learning applied to economic data and we also demonstrate that our implications are highly relevant for time series forecasting, the second string. ...
We simulate economic data to apply state-of-the-art machine learning algorithms and analyze the economic precision of competing concepts for model agnostic explainable artificial intelligence (XAI) techniques. Also, we assess empirical data and provide a discussion of the competing approaches in comparison with econometric benchmarks, when the data-generating process is unknown. The simulation assessment provides evidence that the applied XAI techniques provide similar economic information on relevant determinants when the data generating process is linear. We find that the adequate choice of XAI technique is crucial when the data generating process is unknown. In comparison to econometric benchmark models, the application of boosted regression trees in combination with Shapley values combines both a superior fit to the data and innovative interpretable insights into non-linear impact factors. Therefore it describes a promising alternative to the econometric benchmark approach.
Identifying risk factors that have significant explanatory power for the cross‐sectional asset returns is fundamental in asset pricing. We adopt a novel automatic debiased machine learning (ADML) method proposed by Chernozhukov, Newey, and Singh (2022) to robustly estimate partial pricing effect of a certain factor controlling for a large number of confounding factors under a nonlinear stochastic discount factor (SDF) assumption. The ADML resolves biased estimation, non‐robustness, and overfitting issues that are common to traditional machine learning approaches. We find that the most significant factors selected by the ADML outperform the Fama–French sparse factors and factors identified via the double‐selection LASSO method under a linear factor model assumption. Out of a high‐dimensional zoo of US stock market factors commonly tested in the finance literature, we identify approximately 30 to 50 factors having significant but declining pricing power in explaining the cross‐section of stock returns. Our findings are robust to hyperparameter settings and choices of test assets and machine learning methods.
Cel artykułu. Celem niniejszego opracowania było zbadanie jak decyzje podejmowane przez walne zgromadzenia akcjonariuszy w zakresie dywidendy wpływają na kształtowanie się notowań akcji spółek publicznych w Polsce w perspektywie długoterminowej. Metoda badawcza. W ramach badania przeprowadzono analizę wpływu decyzji dywidendowych na kształtowanie się długoterminowych stóp zwrotu z akcji przy wykorzystaniu metody kalkulacji ponadprzeciętnych stóp zwrotu „kup i trzymaj” (buy-and-hold abnormal returns approach). Ponadto zbadano rolę wskaźnika stopy dywidendy w kształtowaniu się przekrojowej zmienności stóp zwrotu. Przeprowadzone badanie odnosiło się do spółek publicznych notowanych na rynku głównym Giełdy Papierów Wartościowych w Warszawie. Wyniki badań. Wyniki badania można podsumować następująco. Po pierwsze, przeprowadzane badanie empiryczne pozwoliło na określenie możliwości osiągnięcia nadwyżkowych stóp zwrotu z inwestycji w portfele składające się z akcji spółek dywidendowych. Uzyskane wyniki pokazują pozytywny wpływ decyzji dywidendowych na wyceny spółek publicznych w długim okresie utrzymania akcji w portfelu. Portfele akcji spółek inicjujących i wznawiających wypłaty dywidendy pozwalają na osiągnięcie dodatnich nadwyżkowych stóp zwrotu w długim okresie, jednakże osiągane stopy zwrotu są statystycznie istotne tylko dla wybranych okresów utrzymania akcji w portfelu. Po drugie, przeprowadzone badanie przekrojowych stóp zwrotu pozwoliło na potwierdzenie możliwości osiągania w długim terminie nadwyżkowych stóp zwrotu z akcji spółek publicznych o wysokiej stopie dywidendy. Osiągnięte w ramach badania wyniki są zgodne z wynikami badań prowadzonych dla rozwiniętych rynków akcji w zakresie kierunku zmian kursów akcji po ogłoszeniu decyzji w zakresie dywidendy. Przeprowadzone badania rozszerzają prowadzone dla polskiego rynku akcji rozważania w zakresie wpływu decyzji dywidendowych na stopy zwrotu akcji poprzez uwzględnienie w analizie długiego okresu.
We implement a dynamically regularized, bootstrapped two-stage out-of-sample parametric portfolio policy to evaluate characteristics’ efficacy in the conditional stock return-generating process in the metric of expected power utility. Traditional characteristics, such as momentum and size afforded large utility gains before 1999. These opportunities have since vanished. Overfitting—imprecision in weight estimation—is correlated with the optimal portfolio’s variance. Therefore, it is not a problem for power utility investors with coefficients of relative aversion greater than four. For more risk-tolerant investors, we successfully reduce estimation error by increasing the curvature of the loss function relative to the investor’s utility function. (JEL L200; C110; C350)
This paper evaluates various explanations for the profitability of momentum strategies documented in Jegadeesh and Titman (1993). The evidence indicates that momentum profits have continued in the 1990s, suggesting that the original results were not a product of data snooping bias. The paper also examines the predictions of recent behavioral models that propose that momentum profits are due to delayed overreactions that are eventually reversed. Our evidence provides support for the behavioral models, but this support should be tempered with caution.
Previous work shows that average returns on common stocks are related to firm characteristics like size, earnings/price, cash flow/price, book-to-market equity, past sales growth, long-term past return, and short-term past return. Because these patterns in average returns apparently are not explained by the CAPM, they are called anomalies. We find that, except for the continuation of short-term returns, the anomalies largely disappear in a three-factor model. Our results are consistent with rational ICAPM or APT asset pricing, but we also consider irrational pricing and data problems as possible explanations.
We examine (1) how value premiums vary with firm size, (2) whether the CAPM explains value premiums, and (3) whether, in general, average returns compensate β in the way predicted by the CAPM. Loughran's (1997) evidence for a weak value premium among large firms is special to 1963 to 1995, U.S. stocks, and the book-to-market value-growth indicator. Ang and Chen's (2005) evidence that the CAPM can explain U.S. value premiums is special to 1926 to 1963. The CAPM's more general problem is that variation in β unrelated to size and the value-growth characteristic goes unrewarded throughout 1926 to 2004.
Market efficiency survives the challenge from the literature on long-term return anomalies. Consistent with the market efficiency hypothesis that the anomalies are chance results, apparent overreaction to information is about as common as underreaction, and post-event continuation of pre-event abnormal returns is about as frequent as post-event reversal. Most important, consistent with the market efficiency prediction that apparent anomalies can be due to methodology, most long-term return anomalies tend to disappear with reasonable changes in technique.
ABSTRACT Two easily measured variables, size and book-to-market equity, combine to capture the cross-sectional variation in average stock returns associated with market {3, size, leverage, book-to-market equity, and earnings-price ratios. Moreover, when the tests allow for variation in {3 that is unrelated to size, the relation between market {3 and average return is flat, even when {3 is the only explanatory variable. THE ASSET-PRICING MODEL OF Sharpe (1964), Lintner (1965), and Black (1972)