Buy the dip or short the market? AAII bears have some answers
The AAII produce a very useful survey of bullish, neutral and bearish sentiment. Members are asked where they think the market will be in the next six months, with the % answering "up" being the bullish proportion, the % answering "down" being the bearish proportion and the percentage answering "same" being neutral. The survey data is used to produce a diffusion index (% bullish - % bearish) and an 8 week moving average of bullish sentiment. This last item is positively correlated with stock market returns measured over the eight weeks (+0.22). Bullish sentiment tends to increase as the SP500 rises.
The direction of causality is, unfortunately, from stock returns to sentiment. Respondents tend to become more bullish because the SP500 rises, not the other way around. While the survey data has some detectable predictive power it is tiny with the effect on week-ahead prices being in the region of 0.01 to 0.03%.
These features have led the survey to become well known by professional and active investors as a contrarian indicator. When the diffusion index is very positive or bullish sentiment is very high, it is probably time to exit stocks. When the diffusion index is very negative, it is probably a good time to buy.
The survey also contains another quantifiable feature, the degree of persistence. It takes a short time for sentiment trends to shift. This means that at times sentiment turns overly bullish or bearish relative to the underlying prices that mainly drive that sentiment. Despite the survey's lack of predictive power, this over- or under- bullishness or bearishness can be exploited to indicate turning points over relatively short trading horizons.
A bear index
The survey data can be used to calculate a bear index as follows:
%bearish/(%bullish+%neutral) - 25%
The value of the bear index could range from -25% (when 0% of respondents are bearish) to infinity (100% of respondents are bearish).
The reasons for preferring this measure to the diffusion index are:
- The most frequently occurring value of the ratio %bearish/(%bullish+%neutral) is 25%. There are always bearish respondents so the index measures response relative to the "common" proportion of bears (Whether 25% is deducted from the index does not alter results range).
- The AAII diffusion index loses potentially useful information. An equal % of people moving out of extreme positions (becoming less bullish and less bearish) increases the % of neutral responses. In this case, the diffusion index will not change, but sentiment in the market has: there are both fewer bulls and fewer bears.
- A 1 percentage point fall in %bearish lowers the index by a smaller amount than a 1 percentage point increase in %bearish raises the index. This asymmetry mirrors the asymmetry in stock returns: % changes in price falls (which would trigger an increase in %bearish/decrease in % bullish) tend to be larger than % changes in price rises.
- Changes in the bear index are more sensitive to changes in stock returns than the AAII diffusion index.
- There appears to be stronger (negative) correlation between the bear index and stock returns, compared to using the AAII bullish indicator. The correlation between the 13 week moving average bear index (13wmabear) and stock returns over the same 13 weeks is -0.37. For comparison with the 8 week % bullish moving average correlation (+0.22), the 8 week bear index correlation is about -0.29.
The graph below shows the 13 week moving average bear index plotted against the 13 week moving average SP500 price, along with the weekly SP500 closing prices and the difference between the bear index and its 13 week moving average.
As the graph shows, the rise and fall of stock prices is mirrored in the bear index moving average. The contrarian nature of the AAII survey data is clearly visible and suggests for example, buying and selling signals as follows:
- BUY when the bear index is above its 13wma trend but that trend is falling ( bear index cuts falling trend from below)
-SELL when the bear index falls below a rising 13 wma trend (bear index cuts the rising bear trend from above).
For reference, the graph below plots the bear index and its 13 week moving average (right hand scale) along with the sp500 (left hand scale) back to 1987 (the period for which the AAII survey has downloadable data).
Don't be a Perma-Bear
Exploiting the causal relationship between past sentiment, past stock price returns and current sentiment, it is possible to "predict" a level of current sentiment. Comparing that with actual current sentiment presents an (expected) error in calculation which would arise because current prices have moved away from their past position (also in a mildly predictable way), but will remain correlated with current sentiment. As a result, over- or under-bearishness relative to the current prices and sentiment, together with other indicators of price trend, present a method for estimating week ahead SP500 prices.
There will of course be errors between estimated and actual prices and these differences allow for possible short term trading opportunities.
The graph below provides the sample of actual sp500 closing prices and "bear adjusted" closing prices since end of May 2012 to present. The latter effectively represent the closing prices the bears would expect in the week ahead, once they realize they have been over- or under bearish. It is when these bear adjusted closing prices are above actual closing prices (sell), so that overly bullish sentiment remains, or below actual closing prices because overly bearish sentiment lingers, (buy) that short term trading opportunities arise.
A variety of trading strategies can be employed, and for illustration I provide the returns from following a single "long only" strategy mechanically from 1987 to date, with a 1% stop loss at each buy. The details are not important (similar results are obtained from slight variations to the strategy), what is key is the timing.
The graph below compares the returns from one of the better strategies for exploiting the information with the buy and hold strategy from October 1987 (after black Monday) to date.
As can be seen in the graph, the strategy performs quite well for much of the time, picking clear troughs as entry points. However, the strategy takes money out of stocks prior to the two great peaks, missing the full potential gains. Over the period, the strategy just beats the stock market.
Any number of minor modifications to strategy parameters could easily lead to underperforming the market.
The point of the above graph is that buying low and selling high only in response to contrarian bear signals alone (being a Perma-bear) runs the risk of losing the upside gains of over-extended markets: the Bears fall into bear traps.
Nevertheless, the strategy is also very good at avoiding price falls.This last point is the perhaps the most relevant at this current time.
Consider now if the individual enters at the first of the two peaks, with the same strategy as before.
The graph below compares the buying and holding from December 2000 with the "long only" bear strategy. The buy and hold investor would after 13 years just begin to show a positive gain in capital whereas the "long only" bear would have achieved a 100% return.
Despite a number of false starts, the stop losses prevent capital erosion. The bear avoids the bull pits, and is able to hold onto much of the upside gains.
Conclusions for portfolio positioning
Many investors will be tempted to "buy the dip" on the basis of "improving fundamentals", other pieces of "good news", and speculation about the Fed tapering soon and whether it will be reversed.
My advice is to consider the following:
- The market has flushed out all the bears. They are on their way back. The bear index briefly dipped below 0% during July but is now around 14%. For every one bear there are about 2 people either bullish or neutral, up from 1 bear for every 3 bullish/neutral.
- The bear adjusted target close for 22-August sits at 1686.03. Friday's close at 1655.83 will very likely trigger a spike in bearish sentiment and a reversal of the 13wma bear index.
- Much good news and extended Fed help for markets has been priced in. If there are no further announcements, the market will increasingly price in policy tightening.
- The market is not at a bottom. Many signs point to trading at or close to a peak. Don't be tempted to "buy high" and end up selling low.
- The worst lost opportunity at the moment by moving money to the sidelines is that you miss an over-extended stock price bubble. The best case is that you lock in gains from the past few months.
- If however you decide to buy the dip, put tight stop losses in place. Don't fall into a Bull Pit.
- It takes a brave investor or analyst to call a market top. Many factors unknown could still affect the price path. Absent such events, I expect some price volatility at this level of prices, encouraging a temporary reversal of the bear index as prices spike. Should such a reversal fall below the 13wma of the bear index, it would be a clear sell signal - for me - to go short.