Should investors buy or sell stocks experiencing unique (idiosyncratic) volatility spikes? In their August 2016 paper entitled “Unusual News Flow and the Cross-Section of Stock Returns”, Turan Bali, Andriy Bodnaruk, Anna Scherbina and Yi Tang investigate relationships among sudden increases in stock idiosyncratic volatility, unusual firm news, changes in analyst earnings forecast dispersion, short selling and future returns. They identify idiosyncratic volatility shocks as large deviations from the volatility predicted out-of-sample by a regression model that accounts for market, size and book-to-market effects. They identify unusual news flow using Thomson-Reuters News Analytics data (covering 41 media) by comparing the number of stories about a firm in the current month to the average monthly coverage the prior four months, measured overall and separately for positive, negative and neutral stories. They measure changes in analyst earnings forecast dispersion (standard deviation divided by mean) based on data from I/B/E/S as the difference between current dispersion and dispersion two months ago. They measure data on shorting demand and utilization (shares borrowed divided by shares available for lending) using data from Markit. Using monthly values of the specified data from various inceptions through December 2012, they find that:
- The predicted idiosyncratic volatility is reasonable:
- The monthly correlation between predicted and realized values is significantly positive at 0.33.
- Predicted volatility tends to be higher for small stocks, growth stocks, for stocks with high levels of analyst disagreement, high turnover, and low liquidity and for young firms.
- Also, because the shocks of interest are large, findings are robust to choices of idiosyncratic volatility metric and prediction model.
- Idiosyncratic volatility shocks relate positively to:
- Unusual firm news flow (most strongly to unusual negative news flow, followed by neutral and then positive news flow).
- Temporary increases in analyst earnings forecast dispersion that tend to revert over the subsequent two months.
- Idiosyncratic volatility shocks relate negatively to shorting demand and utilization (presumably due to fear of margin calls).
- Average number of shares borrowed by short sellers is 1.49 million lower and utilization is 4.22% lower between the fifths of stocks with the highest and lowest volatility shocks.
- This relationship persists after controlling for past returns and a number of other variables.
- With elevated sentiment divergence and increased shorting constraints after idiosyncratic volatility shocks, optimistic investors first drive prices up as pessimistic investors sit on the sidelines. Subsequently, as investors learn more about consequences of the news, sentiments converge and prices decline.
- The tenths (deciles) of stocks with the lowest and highest idiosyncratic volatility shocks this month have, respectively:
- Average idiosyncratic volatility shocks -9.39% and +14.09%.
- Average next-month equal-weighted gross returns 1.67% and 0.64%, with lowest-minus-highest deciles generating 4-factor (market, size, book-to-market, momentum) alpha -0.93%.
- Average next-month value-weighted gross returns 1.24% and 0.27%, with lowest-minus-highest 4-factor alpha -0.98%.
- Effects hold for 1964-1988 and 1989-2012 subsamples, are not driven by small, low-priced and illiquid stocks and are robust to univariate and multivariate controls for many variables, including market volatility.
In summary, evidence indicates that unusual firm news coverage increases investor disagreement about its stock, thereby inducing an idiosyncratic volatility shock that makes shorting costly and initially drives price higher but ultimately drives price lower.
Cautions regarding findings include:
- Reported returns are gross, not net. Accounting for monthly portfolio reformation costs would reduce all returns. Also, inherent in the methodology, costs of shorting stocks with large idiosyncratic volatility shocks are high, reducing the alphas of lowest-minus-highest hedge portfolios.
- Making parameter selections for multiple variables (in specifying unusual activity levels) introduces the possibility of compounded data snooping bias, thereby overstating strength of findings. Some robustness tests mitigate this concern.
- Tracking stock idiosyncratic volatility shocks with precision and managing the associated portfolio are beyond the reach of many investors, who would bear fees for delegating the process to a fund manager.
