The Dynamic Industry Return Predictability: Evidence from Chinese Stock Markets

ABSTRACT

This paper examines the dynamics, direction, and determinants of industry return predictability in Chinese stock markets during the period 1993–2015. Using the dynamic approach, we find that industry portfolio predictability is time varying and has wide variations across industries. Lagged returns in four industries (banking, real estate, leasing, and information technology) are positively associated with aggregate market returns, while lagged returns for traditional industries are largely inversely associated with market returns. Our findings are consistent with gradual information diffusion across economically-linked industries. The likelihood of industry predictability increases by 4.5–8% in a bull market over that in the bear market. Our results advise investors to distinguish industries and stock market conditions to better time the market.

KEY WORDS:

• bull and bear markets
• Chinese stock markets
• crisis
• industry return predictability

Acknowledgments

Gaiyan Zhang acknowledges the receipt of funding support from the Natural Science Fund of China (71571112 & 71572161) and International Studies and Programs of the University of Missouri-St. Louis.

Supplementary material

Supplemental data for this article can be accessed here.

Notes

1. Fama (1991) surveys three categories of work on market efficiency, (1) tests for return predictability, (2) event studies, and (3) tests for private information. Our study adds to the first category by studying the return predictability of industry portfolio returns.

2. The slow diffusion of information hypothesis stresses that limited cognitive ability or limits to arbitrage makes investors participate in limited markets and cannot obtain information on the market in which they are not participating, leading information in one market to slowly spread to another.

3. Jorgenson and Nomura (2005) use industry-level data to show that the economic growth in Japan has been dominated since 1995 by investments and productivity growth in information technology for the revival of productivity and an economic resurgence.

4. For example, Hong et al. (2018) examine the predictability of the Shanghai Composite, the Shenzhen Composite, and the Hang Seng China Enterprise index returns during the period 1993 to 2010 and find that considering structural breaks in model parameters improves stock return predictability.

5. RESSET is a leading financial database vendor that specializes in providing financial information and services in China. RESSET has been widely used in Chinese stock market research, e.g., Yang and Luo (2014), Yang and Jia (2016), Chen et al. (2017), Weng and Wang (2017), Qu, Liu, and He (2018), Lao, Tian, and Zhao (2018), and Li et al. (2018).

6. According to the China Securities Regulatory Commission (CSRC) (established in April 1998), listed firms with losses over two consecutive annual periods, with a firm value lower than the par value, where auditors hold disclaimer opinion or adverse opinion about the firms, or if they cannot operate due to natural disasters or man-made damage, must be classified as ST firms. ST stocks have a 5 percent daily limit up or down, unlike the 10 percent for normal stocks. It is a common practice to exclude ST firms in stock market studies on China. See, for example, Ding, Zhang, and Zhang (2007), Tsai, Lin, and Hung (2015), Gu, Kang, and Xu (2018), and Qu, Liu, and He (2018).

7. The Shanghai index recorded its largest single-day return of 105% on May 21, 1992, as investors considered the elimination of the daily price limit a sign of government support for the stock market. Moreover, data before 1993 are not available for most industries.

8. We thank an anonymous referee for suggesting that the relationship could be complicated because China heavily imports all kinds of raw materials and energy.

9. In the crisis regression, we omit the industrial growth rate and the bank loan expansion rate because they have very high correlations with crisis.

10. The industry rate of return has a high correlation of 31.4% with the bull market dummy. So it is dropped in the regressions where the bull market dummy is included.

11. We omit the industrial yield rate in the bull market regression because it has high correlation with the bull market dummy.

12. The regression tables with interaction terms have similar results. They are omitted due to space constraints but are available from the authors upon request.

Additional information

Funding

This work supported by the Natural Science Fund of China (71571112 & 71572161) and International Studies and Programs of the University of Missouri-St. Louis.