How effective is investor sentiment in predicting stock market returns? In his October 2017 paper entitled “Measuring Investor Sentiment”, Guofu Zhou reviews various measures of equity-oriented investor sentiment based on U.S. market, survey and media data. He highlights the Baker-Wurgler Index (the most widely used), which is based on the first principal component of six sentiment inputs: (1) detrended NYSE trading volume; (2) closed-end fund discount relative to net asset value; (3) number of initial public offerings (IPO); (4) average first-day return on IPOs; (5) ratio of equity issues to total market equity/debt; and, (6) dividend premium (difference between average market-to-book ratios of dividend payers and non-dividend payers). Based on the body of research and using monthly inputs for the Baker-Wurgler Index during July 1965 through December 2016, three sets of investor sentiment survey data since inceptions (between Dec 1969 and July 1987) through December 2016 and two sets of textual analysis data spanning Jan 2003 through December 2014 and Jul 2004 through Dec 2011, he finds that:
- Desirable properties for an investor sentiment measure are:
- Its ups and downs should match asset price manias and crashes.
- It should relate to other measures of investor overreaction and underreaction such as fund flows.
- More speculative and less liquid assets should be more sensitive to it.
- Its extreme values should eventually revert to some persistent mean.
- Regarding the Baker-Wurgler Index:
- Individual inputs to the index exhibit slight power to predict U.S. stock market returns at a monthly horizon, with R-squared statistics ranging from 0.000 to 0.013 (explaining no more than about 1% of monthly market return variations).
- The index itself has R-squared only 0.003.
- Equally weighting the six inputs instead of using principal component analysis increases R-squared to a still slight 0.006.
- A partial least squares approach to combining the six inputs boosts R-squared to a still small 0.019.
- Regarding market volume and breadth data:
- Changes in advancing volume, total volume and equity option put and call volumes have little or no power to predict monthly market returns, with R-squared statistics ranging from 0.000 to 0.005. Combining volume metrics via partial least squares works best, but R-squared is still only 0.006.
- Fraction of advancing stocks and fraction of stocks reaching new highs work better than volume metrics, but R-squared statistics still range only from 0.001 to 0.007. Again, combining breadth metrics via partial least squares works best, but R-squared is still only 0.014.
- Regarding investor/advisor surveys:
- Bullish/bearish survey data from the American Association of Individual Investors, Chartcraft and Market Vane have little or no power to predict monthly market returns, with R-squared statistics ranging from 0.000 to 0.007. Again, combining survey metrics via partial least squares works best, boosting R-squared to a still small 0.026.
- Surveys may have greater predictive power at longer forecast horizons.
- The Financial and Economic Attitudes Revealed by Search index (based on queries from millions of U.S. households regarding recession, unemployment and bankruptcy) and the Manager Sentiment index (based on aggregate tone of corporate financial disclosures) exhibit some power to predict monthly stock market returns, with R-squared statistics 0.027 and 0.096, respectively. Principal component analysis of the two indexes boosts R-squared to 0.122.
In summary, the body of research, with updates, indicates that investor sentiment metrics have none to small power for predicting stock market returns at a short (1-month) horizon.
Cautions regarding conclusions include:
- Analyses are statistical, not economic. Applying findings to construct equity market timing strategies would incur portfolio constraints and trading frictions.
- The strongest findings of predictive power come from textual analysis datasets that are relatively short and may be heavily influenced by the rare 2008-2009 equity market crash.
- The survey addresses return prediction only for the U.S. stock market.
See also the Sentiment Indicators research category for a collection of indicators/tests.
