AAII Investor Sentiment as a Stock Market Indicator
September 25, 2015 • Posted in Sentiment Indicators
Is conventional wisdom that aggregate retail investor sentiment is a contrary indicator of future stock market returns correct? To investigate, we examine the sentiment expressed by members of the American Association of Individual Investors (AAII) via a weekly survey of members. This survey “measures the percentage of individual investors who are bullish, bearish, and neutral on the stock market for the next six months; individuals are polled from the ranks of the AAII membership on a weekly basis. Only one vote per member is accepted in each weekly voting period.” Survey results are apparently available the market day after the polling period. We define aggregate (net) investor sentiment as percent bullish minus percent bearish. Using outputs of the weekly AAII surveys and prior-day closes of the S&P 500 Index from July 1987 through early September 2015 (1,469 surveys and 56 independent 6-month forecast intervals), we find that:
Two weekly survey reports are missing, one in January 1996 and one in June 2000. One weekly level of the S&P 500 Index is missing in September 2001. We ignore calculations affected by these missing data.
The following chart shows net investor sentiment and contemporaneous S&P 500 Index level over the available sample period. Visual inspection suggests that individual investors are usually net bullish in aggregate and that net sentiment may move with the stock market. The average net sentiment is 8.5%.
For precision, we relate net investor sentiment to future stock market return.
The next chart summarizes correlations between net investor sentiment and S&P 500 Index returns over various past and future intervals, using the stock market close on the day before publication of survey results to separate past and future returns (the last close that could influence poll participants). Since polling takes place over a week, 1-week and 2-week past returns are most immediate for survey participants. Results suggest that:
- Sentiment relates positively to recent lagged returns. In other words, an advancing (declining) stock market indicates relatively high (low) net investor sentiment.
- Sentiment relates negatively (but weakly) to future returns over short to intermediate horizons. At the surveyed horizon of six months, the correlation between net investor sentiment and future return is -0.12. The R-squared statistic is therefore about 0.015, indicating that sentiment explains just 1.5% of the variability in future returns.
It may be that the contrarian stance of polled investors represents failure to anticipate some mean reversion tendency of recent stock market returns. In other words, aggregate sentiment is a proxy for market mean reversion. See “Purifying Stock Market Sentiment Indicators” for elaboration of this interpretation.
For the the weekly survey/report frequency, return measurements for intervals longer than one week overlap. These overlaps can distort statistics (and mean that an investor could not exploit all signals).
Is the contrary relationship with future returns stable over time?
The next chart shows rolling lagged 52-week correlations between between net investor sentiment and S&P 500 Index 26-week future return over the available sample period. The correlations are mostly negative (individual investors in aggregate usually lean the wrong way) but are occasionally positive. The upward sloping dashed trend line suggests that investors improve their understanding of stock market behavior over the sample period.
As a separate test of trend, the correlation between net investor sentiment and S&P 500 Index 26-week future return is -0.25 (-0.04) over the first (second) half of the sample period. The break point between the first and second halves is about mid-2001. In other words, individual investors lean less the wrong way after mid-2001 than before.
What drives the contrary indication?
The final chart summarizes the average 1-week, 4-week and 26-week S&P 500 Index future returns by ranked tenth (decile) of net investor sentiment measurements over the available sample period. Results suggest that:
- For 1-week future return, there is no reliable relationship with sentiment.
- For a 4-week future return, sentiment is generally a weak, unsystematic contrary indicator.
- For 26-week future return, the negative relationship with sentiment derives from extreme deciles. Sentiments in middle deciles are uninformative.
- Even the most extreme bullish decile does not indicate shorting.
The range of net investor sentiment for the most bearish (bullish) decile is < -15% (> 33%). Returns in the most bearish decile exhibit high variability compared to other deciles.
Considering even more extreme net sentiment distribution tails:
- The average 26-week S&P 500 Index return after the 1% most bearish net sentiment readings (15 observations < -35%) is 15.1%. However, these observations occur in clusters, supporting only seven independent trades in 28 years.
- The average 26-week S&P 500 Index return after the 1% most bullish net sentiment readings (15 observations > 49%) is -2.1%. Again, these observations occur in clusters, supporting only eight independent trades in 28 years.
In summary, evidence suggests that investors may be able to exploit extreme values of AAII net investor sentiment as contrary signals, but strong (extremely bearish sentiment) signals are rare.
Cautions regarding findings include:
- The above analyses are in-sample. An investor operating in real time during the sample period would have only past data to estimate any net sentiment thresholds for signals.
- As noted, overlapping return measurement intervals can distort statistics (or, said otherwise, result in signals that are not exploitable).
- The cutoff dates for return calculations (closes before report dates) delineate what survey participants may know, but specify future returns that are not fully exploitable by about one trading day.