Investor Sentiment and Bond Risk Premia: Evidence from China

ABSTRACT

This article shows the statistical significance of a set of variables related to market sentiment and uses them to predict the risk premium embedded in China’s sovereign bonds. We construct a composite index of market-wide investor sentiment as a linear combination of proxies for a degree of market participation and risk appetite of investors. Further, we show that these sentiment-related factors can be summarized in a single-return forecasting factor, similar in a spirit of Cochrane and Piazzesi (2005). Our empirical results show that this sentiment factor has predictive power beyond that contained in the yield curve and macroeconomic variables, and this predictability is robust for out-of-sample testing. In addition, the predictive power of the sentiment factor shows relevance during the 2008 global financial crisis, indicating that the forecasting ability of investor sentiment is mainly derived by a sentiment-induced “flight-to-quality.”

KEY WORDS:

• bond risk premia
• China bond market
• investor sentiment
• return-forecasting factor

JEL CLASSIFICATION:

• G02
• G12
• G14

Notes

1. Kessler and Scherer (2009) cover Australia, Canada, Germany, Japan, Switzerland, the U.K., and the U.S. Sekkel (2011) uses nominal zero-coupon government bond yields constructed by Wright (2011), and the sample covers Australia, Canada, Germany, Japan, Norway, New Zealand, Switzerland, Sweden, the U.K., and the U.S. Those studies do not cover China.

2. The flight-to-quality is a market phenomenon associated with investors’ herding behavior and it involves reallocation of investors’ money to what appear to be “safe” assets from “risky” assets, without making rational and sophisticated decisions during large dips in sentiment. According to Baker and Wurgler (2012), we can easily find evidence of “flight-to-quality” in anecdotes. For instance, during the largest one-day crash in the U.S. stock market history in October 1987 and the collapse of Long-Term Capital Management (LTCM) in August 1998, government bond prices surged.

3. These stocks are characterized as being long-mature, low-volatility, profitable, from dividend-paying firms, and are neither high growth nor distressed.

4. Xu and Green (2013) show that investor sentiment affects stock returns using Fama-French 25 portfolios, and Bu and Pi (2014) find the significant predictive power of the proxies of investor sentiment similar to that of Baker and Wurgler (2006) for HS300 index.

5. Chen et al. (2007) investigate China’s brokerage data and conclude that investors suffer from all of the three representative behavioral biases in capital markets, including overconfidence, disposition effects, and representativeness heuristic. Also for herding behavior, see Tan et al. (2008), Chiang, Li, and Tan (2010), and Lao and Singh (2011).

6. The World Bank website (http://www.worldbank.org) provides global ranking for nominal Gross Domestic Product in million USD. In 2015, China was ranked the second-largest economy in the world after the U.S.

7. Tan et al. (2008) show that domestic individual investors dominate the A-share market.

8. Spyrou (2013) points out that when financial crises arise from an increase in risk aversion and loss of confidence among investors, monetary policy interventions are completely ineffective in resolving the problem.

9. Recently, Wurgler dropped NYSE Turnover as one of the sentiment indicators due to the explosion of institutional high-frequency trading in his website. Nevertheless, we use turnover as one of the sentiment indicator since The Chinese stock market is still a less developed market compared to developed countries and most of the participants in the Chinese stock market are individual investors. Also, the recent study regarding sentiment index in China uses turnover as one of proxy for investor sentiment (e.g., Han and Li 2017). Despite this issue, our empirical results are quantitatively and qualitative similar when we drop the turnover.

10. If there are more capital market-oriented data vendors such as Wind Info, we want to check the robustness of our results with these databases.

11. There are many other studies considering the number of newly opened accounts for individual investors as a proxy for investor sentiment in Chinese stock market. See Han and Wu (2007), Yi and Mao (2009), Pan, Tang, and Xu (2015), and Han and Li (2017).

12. In this process, we pick one between the current and lagged series for each proxies, which has higher correlation value with the first-stage index.

13. Several past studies use the annual holding excess returns as bond risk premia such as Fama and Bliss (1987), Campbell and Shiller (1991), Cochrane and Piazzesi (2005), Laborda and Olmo (2014), among others.

14. Fan, Tian, and Zhang (2012) investigate the dynamics of quarterly bond excess returns, but we mainly deal with the annual bond risk premium, as done by Fama and Bliss (1987), Cochrane and Piazzesi (2005), and Ludvigson and Ng (2009).