Reassessing the Predictability of the Investor Sentiments on US Stocks: The Role of Uncertainty and Risks

Abstract

We examine the predictive power of US investor sentiments on US sectoral returns and aggregated S&P 500 index in the presence of different risk and uncertainty indices. Investor sentiments are measured using the sentiment index proposed by Baker and Wurgler (2006). We also use bearish and bullish investor sentiment indices i.e., AAII sentiment indices, published by the American association of individual investors. We also use different risk and uncertainty indices i.e., economic policy uncertainty, financial uncertainty, geopolitical risk, and US equity market volatility. Results of the baseline model (i.e., the single model where the only predictor is sentiment index) show that the forecasting power of the predictor is weak. Augmenting the baseline model to account for uncertainty measures shows that the uncertainty's transmission impact is more accurate for out-of-sample forecasts than the single-factor model. We also highlight that the performance of the multi-factor predictive model incorporating USS B&W is superior to the benchmark model. Policy implications of the results are also discussed.

Keywords:

• Investor sentiments
• US sectoral returns
• S&P500
• Predictive model

Acknowledgement

This research is partly funded by the University of Economics Ho Chi Minh City, Vietnam.

Notes

1 The statistics are sourced from https://www.cnbc.com/2020/03/31/stock-market-today-live.html and https://www.statestreet.com/content/dam/statestreet/documents/ici/ICI-HistDat-Eng-2020Nov.pdf

2 SP100, DJIA and Nasdaq stock returns.

3 These factors include number of initial public offerings (IPOs), the average first-day returns of IPOs, the dividend premium, the closed end fund discount, the New York Stock Exchange (NYSE) turnover, and the equity share in new issues.

4 NYSE stock returns.

5 Baker and Wurgler (2006) sentiment index.

6 These firms are listed on NYSE, AMEX and NASDAQ.

7 These sectors include Basic Materials, Consumer Goods, Consumer Services, Health Care, Industrials, Oil & Gas, Technology, Telecommunications, and Utilities.

8 These sectors are Automobiles, Consumer Discretionary, Financials, Health Care, Industrials, Information Technology, Materials, Telecommunications, and Utilities.

9 Due to the inter-linkage between Theil’s U-statistics and CT test and for ease to understand the results tabulation, we refrain from presenting the CT test results. In situations where the U-statistics is less than 1, mathematically, it is expected that the CT test would be positive and vice-versa.

10 AAII mails 100 questionnaires on each weekday which are tallied on every Thursday, not to be dated earlier than two weeks. Unlike investors sentiment survey, the AAII conduct an asset allocation survey on monthly basis asking individual investors to highlight their actual allocations in portfolios through stocks, cash and bonds by mailing 600 questionnaires at the beginning of the month. These questionnaires are tallied with responses received during the month. These responses return consistently on monthly basis with central point being the middle of each month allowing us to consider the mean of weekly AAII sentiment figures during the entire month. 

11 Data for geopolitical risk is sourced from https://www2.bc.edu/matteo-iacoviello/gpr.htm, geopolitical uncertainty from https://www.philadelphiafed.org/research-and-data/real-time-center/partisan-conflict-index, financial uncertainty from https://www.sydneyludvigson.com/data-and-appendixes/, and US economic policy uncertainty from http://www.policyuncertainty.com.

12 Values of persistence and endogeneity tests are N/A for US sectoral returns because they are meant for the predictors, as argued by Westerlund and Narayan (2015).

13 Generally, the results here contradict Shahzad et al. (2017) who found stronger evidence for the predictability of bearish sentiment compared to the bullish sentiments. Their study, however, consider the commodities markets.

14 Their predictors include both the managerial sentiments and the consumer confidence indicators.

15 See the Appendix for the break dates