Predicting Stock Returns When the VIX Is High

Summary

• A Journal of Financial Economics paper explains why market sentiment measures are most valuable to investors when uncertainty is peaking.
• The authors find an increase in investor sentiment predicts lower market returns over the subsequent three days. This predictive power is two to four times greater when the VIX is one standard deviation higher than its mean, as it is today.
• The paper shows how we can use measures of investor uncertainty like the VIX alongside popular sentiment measures, such as the University of Michigan Consumer Sentiment Index, to predict short-term returns.

Introduction

In uncertain times, asset prices can fluctuate wildly as investors lean more on their ‘animal spirits’ than market fundamentals. Uninformed traders, of which there are more in times of economic stress, increasingly look towards sentiment measures to validate their trading decisions. This opens the door to greater subjectivity in valuations, which increases the chance assets will be mispriced.

The notion that sentiment measures induce more noisy trading, greater mispricing and excess volatility was formalised in the 1990s. But a new Journal of Financial Economics paper wants to exploit this information to predict market returns. The idea is that sentiment exhibits stronger asset pricing effects at times when valuations are more subjective. The paper finds they do, meaning sentiment measures can predict market returns in uncertain times.

Uncertainty and Sentiment

Uncertainty is innately hard to measure but conceptually easy to understand. It reflects an inability to forecast probabilities, meaning investors have less of a rational foundation on which to base their decisions. In these times, a lack of concrete fundamentals pushes investors to overweight sentimental forces, i.e. beliefs. For this reason, irrationality has the greatest effect in uncertain environments.

To understand the interaction between uncertainty and sentiment, the authors need a measure of each.

Measuring Uncertainty

For uncertainty, the authors use two measures: the VIX and the mean stock-level idiosyncratic volatility (IVOL). The former captures the expected stock market volatility over the next 30 days for the S&P 500. The latter captures any movement in stock returns not explained by the market, firm size or value premiums (e.g., residuals from a Fama-French). The two measures are highly correlated (0.83).

Measuring Sentiment

For sentiment, the authors also use two measures. The first is the FEARS index, which was created in 2014 and covers 2004-2011. The measure was calculated from Google search volumes on terms that reflect households’ economic concerns. It is closely correlated with the University of Michigan Consumer Sentiment Index. The second sentiment measure (SENT) is the first principal component of five investor sentiment measures: the closed-end fund discount, the number of IPOs, the first-day returns of IPOs, the equity share in total new issues, and the dividend premium.

Methodology

The authors want to understand how investor sentiment affects stock returns in periods of high uncertainty. They test this at an aggregate level using returns on the S&P500 index (minus the one-month T-Bill) and at a firm-level using stock data from CRSP.

To answer how sentiment predicts returns, they focus on horizons of up to three days. Also, low, medium and high uncertainty corresponds to the 10^th, 50^th and 90^th percentile of the uncertainty distribution. We are currently in the 92^nd percentile (Chart 1).

The authors also want to understand the impact of sentiment on portfolios. The CRSP estimates the beta and volatility for NYSE/AMEX stocks using daily data in the past year. It does this to give the approximate risk of a stock. Investors can then sort stocks from low beta/volatility (low risk) to high beta/volatility (high risk), and create a portfolio that goes long low-beta stocks and short high-beta stocks. This is what the authors do to see how changes in sentiment affect the returns of such a portfolio.

S&P500 Returns and the VIX

At an aggregate level, the authors find that high uncertainty is associated with strong predictive ability of sentiment for S&P500 returns (Chart 2).

When uncertainty is at its mean level, a one-standard-deviation increase in FEARS (which is a lowering of investor sentiment) predicts an eight-basis-point increase in the following day’s market return. However, when uncertainty is one standard deviation above its mean, a one-standard-deviation increase in FEARS predicts a 23-basis-point increase in the following day’s market return.

In short, when VIX is high, a lowering of investor sentiment predicts higher S&P500 returns in the subsequent three days. The results show that the predictive ability of high-frequency sentiment for market returns is stronger when conditioning on uncertainty. The same is true at a monthly frequency.

Beta- and Volatility-Sorted Portfolios

Next, the authors look at the impact of sentiment on beta-sorted portfolios. They find that increasing investor sentiment predicts lower future returns to high-beta stocks relative to low-beta stocks. And one-day-ahead returns are more than three times larger when the VIX is one standard deviation above its mean relative to when it is at its mean.

For a strategy that is long low-volatility stocks and short high-volatility stocks, decreases in sentiment predict higher returns for high-volatility stocks relative to low-volatility stocks, implying lower portfolio returns. Also, sentiment-induced future returns of volatility portfolios are typically between two to three times larger when uncertainty is one standard deviation above its mean.

The results show that, when the VIX is high, changes in investor sentiment are very good at predicting three-day-ahead returns in beta- and volatility-sorted portfolios.

The Bottom Line

Greater volatility in asset prices can make trading daunting in uncertain times. In the short term, valuations begin to move away from their fundamentals as individuals struggle to see the wood for the trees. Sentiment measures, which uninformed traders are more likely to depend on to inform their decisions, start to play more of a role in asset pricing. This only adds to the noise. Or at least, so we thought…

The paper reveals a way of navigating markets in uncertain times. The very waves of pessimistic or optimistic sentiment that can lead us to incorrect valuations can help us predict returns. These waves act as signposts indicating where the market will head, which a savvy investor could begin to exploit. However, they only work at the noisiest of times, when valuations are furthest from their fundamentals and irrationality is at its peak. That happens to be now.