Purpose

The purpose of this paper is to examine the relationships between the higher moments of returns (realized skewness and kurtosis) and subsequent returns at the industry level, with a focus on both empirical predictability and practical application via trading strategies.

Design/methodology/approach

Daily returns for 48 US industries over the period 1970–2019 from Kenneth French’s data library are used to calculate the higher moments and to construct short- and medium-term single-sort trading strategies. The analysis adjusts returns for common risk factors (market, size, value, investment, profitability and illiquidity) to confirm whether conventional asset pricing models can capture these relationships.

Findings

Past skewness positively relates to subsequent industry returns and this relationship is unexplained by common risk factors. There is also a time-varying effect in which the predictive role of skewness is much stronger over business cycle expansions than recessions, a result consistent with varying investor optimism. However, there is no significant relationship between kurtosis and subsequent industry returns. The analysis confirms robustness using both value- and equal-weighted returns.

Research limitations/implications

The calculation of realized moments conventionally uses high-frequency intra-day data, regrettably unavailable for industries. In addition, the chosen portfolio-sorting method may omit some information, as it compares only average group returns. Nonetheless, the close relationship between skewness and future returns at the industry level suggests variations in returns unexplained by common risk factors. This enriches knowledge of market anomalies and questions yet again weak-form market efficiency and the validity of conventional asset pricing models. One suggestion is that it is possible to significantly improve the existing multi-factor asset pricing models by including industry skewness as a risk factor.

Practical implications

Given the relationship between skewness and future returns at the industry level, investors may predict subsequent industry returns to select better-performing funds. They may even construct trading strategies based on return distributions that would generate abnormal returns. Further, as the evaluation of individual stocks also contains industry information, and stocks in industries with better performance earn higher returns, risks related to industry return distributions can also shed light on individual stock picking.

Originality/value

While there is abundant evidence of the relationships between higher moments and future returns at the firm level, there is little at the industry level. Further, by testing whether there is time variation in the relationship between industry higher moments and future returns, the paper yields novel evidence concerning the asymmetric effect of stock return predictability over business cycles. Finally, the analysis supplements firm-level results focusing only on the decomposed components of higher moments.

中文翻译：

---

更高的时刻和美国行业回报：已实现的偏度和峰度

目的

本文的目的是检验较高的回报时刻（已实现偏度和峰度）与行业层面的后续回报之间的关系，重点是通过交易策略的经验可预测性和实际应用。

设计/方法/方法

Kenneth French 数据库中 48 个美国行业在 1970 年至 2019 年期间的每日回报用于计算较高时刻并构建短期和中期单一排序交易策略。该分析调整了常见风险因素（市场、规模、价值、投资、盈利能力和流动性不足）的回报，以确认传统资产定价模型是否可以捕捉这些关系。

发现

过去的偏度与随后的行业回报呈正相关，这种关系无法用常见的风险因素来解释。还有一种时变效应，其中偏度对商业周期扩张的预测作用比衰退要强得多，这一结果与不同的投资者乐观情绪一致。然而，峰度与随后的行业回报之间没有显着关系。该分析使用价值加权回报和等加权回报来确认稳健性。

研究限制/影响

已实现时刻的计算通常使用高频日内数据，遗憾的是行业无法使用。此外，所选择的投资组合排序方法可能会忽略一些信息，因为它只比较平均组回报。尽管如此，行业层面的偏度与未来回报之间的密切关系表明，常见风险因素无法解释回报的变化。这丰富了市场异常的知识，并再次质疑弱形式市场效率和传统资产定价模型的有效性。一种建议是，可以通过将行业偏度作为风险因素来显着改进现有的多因素资产定价模型。

实际影响

鉴于行业层面的偏度与未来回报的关系，投资者可通过预测后续行业回报来选择表现较好的基金。他们甚至可以根据会产生异常回报的回报分布来构建交易策略。此外，由于个股的估值也包含行业信息，业绩较好行业的股票收益较高，行业收益分布的风险也可以为个股选择提供参考。

原创性/价值

虽然在公司层面有大量证据表明更高时刻与未来回报之间的关系，但在行业层面却鲜有证据。此外，通过测试行业较高时刻与未来回报之间的关系是否存在时间变化，本文提供了关于股票回报可预测性在商业周期中的非对称效应的新证据。最后，分析补充了公司层面的结果，只关注较高时刻的分解成分。