Does the Momentum Strategy Work at the Industry Level? Evidence from the Chinese Stock Market

ABSTRACT

This article examines the effectiveness of momentum strategy at the industry level in the Chinese stock market. We find that the intermediate-horizon momentum effect is stronger in industries with higher competition. This effect is consistent with the hypothesis that information contained in firms from highly competitive industries is vague and hence leaves more space for behavioral biases, which leads to the momentum effect. Alternatively, the measure of the Herfindahl–Hirschman index potentially captures the size effect in explaining this phenomenon. Moreover, concentrated industries experience a pronounced lead-lag effect of big firms on small firms, which is a potential explanation for the contrarian strategy. We do find that the short-horizon contrarian effect is pronounced in highly concentrated industries.

KEY WORDS:

• contrarian strategy
• industry concentration
• momentum
• portfolio investment

Acknowledgments

I am indebted to the editor and three referees for constructive comments and helpful suggestions. I am also indebted to Prof. Dr. Mei Wang, Prof. Dr. B. Burcin Yurtoglu, Dr. Dennis Dlugoush, and Guosong Xu for helpful comments. I also acknowledge that the data of this article were obtained from the library of the WHU-Otto Beisheim School of Management, Germany, when I was a PhD candidate at the Chair of Behavioral Finance.

Notes

1. According to the data from CSRC (China Securities Regulatory Commission), the stocks held by individual investors represent 25.33% of the total market value; however, 17.40% was held by institutional investors in 2012. Furthermore, individual investors generate 80.93% of the entire trading volume, compared with 15.19% of institutional investors.

2. Li (2011) uses five variables as proxies for financial constraints, the KZ index (Kaplan and Zingales 1997), the WW index (Whited and Wu 2006), the SA index (Hadlock and Pierce 2010), age, and size. Firm age and firm size are used as proxies for information uncertainty in our article.

3. One may argue that the slower diffusion of information in concentrated industries can lead to a pronounced momentum effect based on the HS model (Hong and Stein 1999), which is opposite to our above discussion of the market structure and the momentum effect. However, the information diffusion of the lead-lag effect here is in a relatively short time-period, i.e., from one week to one month. The intermediate-horizon information diffusion should be further discussed.

4. Initially, there are 24 industries. We exclude two industries, i.e., Aerospace (Industry code: 13) and Printing & Publishing (Industry code: 64), from the sample, because the number of stocks in these two industries is less than 15.

5. Compared with Datastream industry code, nearly 87% of the stocks are classified into the same industry as the SIC code of 20 industries (Moskowitz and Grinblatt 1999).

6. Although Hou and Robinson (2006) show that market capitalization is the smallest in the most concentrated industries, they further argue that the skewness in size distribution of firms within an industry takes responsible for this negative link.

7. We thank an anonymous referee for suggesting this explanation.

8. The mean value is more likely driven by certain large firms. Each innovation in the Oil, & Gas industry, or the automotive industry requires considerable R&D expenditures, e.g., the annual gross R&D expenditures for China Petroleum were 4.6 billion RMB during 2008–2012.

9. According to Gaspar and Massa (2006), industry leaders play a critical role in measuring momentum profits, especially for highly concentrated industries. Thus, we only present the value-weighted returns in the following parts. Our main results still hold when we use the equal-weighted returns.

10. The initial time period is from January 1993 to December 2014. However, the 6-month formation period and the required sufficient number of stocks in each industry, as well as the required number of industries each month, cause the beginning date here to be May 1994. Because of the availability of Sales, the ending date is December 2013.

11. We use the Shanghai composite index and the Shenzhen composite index as the market index for the stocks listed on the Shanghai and Shenzhen Stock Exchanges, respectively. Because the trading of Treasury is infrequent in China, we adopt the 3-month bank deposit rate as the proxy for the risk-free rate.

12. We thank an anonymous referee for suggesting this empirical investigation.

13. Hou (2007) suggests that monthly returns may reduce estimation accuracy, and daily returns could be influenced by microstructure, e.g., bid-ask bounce.

14. The result still holds when we adopt lagged returns in week t − 7, as well as one-lagged monthly return.

15. We thank an anonymous referee for suggesting this explanation.