Revisiting the “Smart Money”
Ray Micaletti was recently on the Excess Returns podcast. Ray has written several interesting papers about “relative sentiment”, which measures how “smart money” managers are positioned relative to “dumb money”. I have tracked some of these metrics personally in the past and, honestly, not had a great experience with them. In contrast to that, during the same period of time, I had a great experience with simple trend following strategies (see here, for example). As a result, my impression of relative sentiment has been fairly negative.
In the podcast, however, Ray mentioned that he has found that relative sentiment actually pairs nicely with trend following. That pushed me to take another look at these ideas, so I dug into the data a second time. To my surprise, this time, I did find some strategies that seem worth using.
Measuring Sentiment
In his papers, Micaletti measures sentiment using the Sentix index, a poll of German investors’ views about various markets, including the US. My preference is to use the views of US investors about US markets. Additionally, I prefer to look at how investors are positioned rather than their opinions because, in my experience, opinions are almost purely a function of whether the market went up or down last week. So, rather than using Sentix, I decided to build a “relative sentiment” indicator using the AAII and NAAIM allocation surveys. The former is (by definition) a survey of individual investors, while the latter is a survey of professionals.
The AAII has published their allocation survey monthly since 1987. NAAIM, in contrast, has only been collected since 2006. A further difficulty with the NAAIM data is now noisy it is. The following picture shows the weekly allocation to equities by active investment managers since 2021:
“Active investor managers” appear to change their allocation to stocks at an alarmingly rate. In one week in 2023, for example, they went from a 45% allocation to stocks to a 65% allocation. In contrast, the largest move that individual investors have made in one month, going back to 1987, was 13%.
To form my own “relative sentiment” indicator, I first built a monthly NAAIM data set by averaging over the prior 5 weeks. Even then, the allocation of active investment managers was extremely volatile. To combat this, I ended up using a six-month average of the monthly averages.
I then defined the “relative sentiment” (or more accurately, the “relative allocation”) as the difference between the AAII monthly survey and my 6-month average of NAAIM weekly survey results.
The following picture shows how that metric has changed over time. The orange curve is the relative sentiment, while the grey curve is the average value minus 0.25 standard deviations over the data set until that point.
While the allocation survey is quite volatile, this relative sentiment metric tends to stay on one side for long periods of time. The orange line was only below the grey line from 2007–09, 2011–12, 2015–16, and May 2022 to present. The latter has been an especially good call about when to stay out of the market.
Trend Following vs Relative Sentiment
Of course, trend followers also would have gotten out of the US market in the middle of 2022. So it is worth investigating whether relative sentiment adds anything beyond simple trend following.
Looking at the data since 2006, in most down markets, trend following would have exited the market earlier than relative sentiment. The only counter-example was in 2007, when relative sentiment exited first.
That said, exiting first is not always a winning strategy. After all, stock prices usually go up, so being out of the market is, on average, a net negative. The data on this was mixed: during the period from 2008 to 2016, relative sentiment underperform by exiting and entering more slowly; however, from 2017 to present, relative sentiment has benefited from moving more slowly.
While the overall performance comparison between the two strategies (trend following and relative sentiment) has been a wash, relative sentiment has certainly had lower turnover. The trend following strategy (using a 10-month simple moving average) has entered or exited the market more than once per year on average, whereas the relative sentiment strategy has entered or exited less than once every other year.
Like trend following, the benefits of the relative sentiment strategy show up primarily in volatility rather than average returns. Both strategies significantly decrease volatility by avoiding the increased volatility that comes during down markets.
As a result of decreased volatility, like trend following, relative sentiment outperforms the S&P 500 by 0.7% per annum since mid-2007. More importantly, it decreases the maximum drawdown to just 21% versus 56% for the S&P 500. That said, this measure of outperformance is relatively small compared to the decrease in volatility. As we discussed previously, the reduction in volatility is something we should believe will continue, much more so than the minor improvement in returns.
Combining Trend and Sentiment
Given the general tendency of trend following to enter and exit the market more quickly, and given the fact that the stock market usually goes up, a reasonable strategy would be to wait for sentiment to exit the market but re-enters it as soon as trend following does. That would generally maximize the time spent in the market.
That strategy outperformed by 2% per annum since 2007 and reduced the maximum drawdown a tiny bit more to 20%. However, building a state machine of that sort, and deciding how to handle the rare cases where sentiment exits or re-enters before trend, is complicated.
A simpler strategy is to just reduce the equity allocation to 50% when one of the two strategies signals to exit. After all, a disagreement between the strategies is probably a good indicator that markets will be more volatile, even if returns are still positive, due to the differing opinions about where the economy is headed in the near term.
That simpler strategy decreases the volatility even further, bringing the maximum drawdown since 2007 to just 17%. It has also outperformed by 2% per annum during that time.
If I had to pick one, that last strategy would be the one I would follow. As we have discussed previously, ensembles of models generally outperform individual models, so a strategy that combines both models in a simple way like this is likely to outperform either individual strategy.
Who is the Smart Money?
Reducing the maximum drawdown since 2007 to just 17% while outperforming is a notable achievement, and certainly data that should make us consider using relative sentiment in our process.
That said, I want to push back against the idea that institutional investors are the “smart money” in some deep sense. The evidence from the AAII vs NAAIM surveys suggests to me that the primary difference between individual and institutional investors is that institutional investors are simply quicker to change their allocations. You can see what I mean in the following chart, which compares the two:
The blue line shows the fast-changing institutional allocation to stocks, while the orange line is the slower-changing individual allocation. Keep in mind that the blue line is a 6-month moving average of institutional investor allocations. The month-to-month swings are even more volatile! Individual investors change their allocations much more slowly, by comparison.
Looking at the two curves, you can see that they tend to reach major tops and bottoms around the same time. To me, this is inconsistent with the idea that institutional investors are leveraging an informational advantage. Plus, the blue line itself changes too frequently from high to low to be primarily a reflection of information about the state of the economy because the state of the economy certainly does not change so drastically in such a short period of time.
Rather, the relative sentiment strategy, which looks at the difference between institutions to individuals, seems primarily like a clever way to trend follow the institutional investors, i.e., to pick out when the herd has changed direction and move along with them. (Here, the individual investors are playing the role of the “moving average” of institutional investors allocations. We buy when the blue line goes above this “moving average” and sell when it drops beneath it.)
The papers on this topic like to point out that the allocations changes made by institutional investors are predictive of future market moves. That suggest that they know what is coming before the rest of the market. However, I think there is a simpler explanation.
I think this correlation is a consequence of two facts: (1) institutional investors tend to herd and (2) institutional investors move the markets. The academic papers on these topics do mention the tendency of institutional investors to herd. And when some part of the herd starts moving in one direction, that predicts that the rest of the herd will soon follow. Furthermore, institutional investors move enough money around to take market prices with them, while individual investors do not. The idea that institutional investors are constantly watching each other and moving like as a group, pulling market prices along with them, is not behavior that I would describe as “smart money”.
That said, the evidence we saw above indicates that you and I can actually outperform by watching how the herd of institutional investors is moving relative to the slower moving collection of individual investors. So while I am not ready to crown institutions as the most clever, I am happy to use the signal they generate to help me outperform.
