Wednesday, August 3, 2016

Does Long Term Mean Reversion Still Work?

I  just read an article in Institutional Investor about the problems that Grantham, Mayo, Van Otterloo & Co (GMO) are having with their long-term mean reversion strategy:

GMOs Mean Reversion Strategy Is Testing In Today's Markets

Mean reversion works by trying to find times when certain markets are overextended to the upside or the downside. A mean reversion strategy would avoid markets that are overextended to the upside and buy markets that are overextended to the downside. They work because markets are constantly forming bubbles that subsequently burst. While this sounds intriguing, there is a problem. Bubbles can go for a long time before they burst so mean reversion strategies can be on the wrong side of things for a while until they ultimately pay off. 

If individual and institutional investors truly had long-term time horizons, then this wouldn't be a problem, but in real life the hardest thing for any investor to do is watch other people make money while they are not. It appears the stock market is in a bubble [PO1] and possibly has been for a while. However, it keeps going up and any corrections are short lived. At some point it will probably crash, but that could be years away. And that’s GMO’s big problem. GMO’s long-term strategies do not align with their clients’ shorter-term realities.

Is there a Better Mean-Reversion Approach?

Markets have always reverted to their mean, but since mean reversion can take a long time, a better approach could be to align the mean-reversion time frame to the investor’s time frame. That approach could create a better investor experience. Investors might talk about how they have a long time horizon, but when markets are moving against them all of that goes out the window. Mean-reversion strategies can be moved to monthly, weekly, daily, even intraday time frames. Instead of trying to only catch the major market turning points (i.e. bull and bear markets), these types of strategies can look for much shorter-term turning points, from intermediate-term corrections all the way down to intraday reversals. Unlike long-term approaches that would rely on valuation, which wouldn't work for shorter-term models, these approaches could use intermarket analysis, cyclic analysis, or traditional overbought/oversold indicators. 

Intermarket Analysis 

Intermarket analysis involves taking two or more markets that are related to each other and looking for divergences. For example, lets assume stocks and Treasuries move inversely and that bond traders tend to be quicker at predicting turning points than stock traders. An intermarket model could then look to buy stocks when they are going down while Treasuries are moving down at the same time. The model would sell stocks when they are moving up and Treasuries are moving up at the same time. 

Cyclic Analysis

Cyclic analysis would take some measure of market cycles and look to buy into short-term weakness and sell into short-term strength.

Traditional Overbought/Oversold Indicators

Traditional indicators like RSI can also help determine short-term overbought/oversold areas. Traditional RSI measures looked at 14 periods, but there has been some good work by Larry Connors and others looking at shorter-term measures.

Markets are mean reverting by their nature, so any well-thought-out mean-reversion strategy can be successful. Longer-term models are problematic, however, as they don't line up well with investors true time frames. Shorter-term models can still capture the longer-term moves while also lining up better with what investors want. At the end of the day it’s about meeting investors’ comfort zones and expectations. Sometimes traditional approaches do not align with these needs. Since central banks are dominating capital markets, this is one of those times.