When comparing various markets or asset classes I often use an approach based on volatility rebalancing. The idea of this approach is to make daily position adjustments to keep the position size balanced in accordance to local volatility. When volatility increases, we decrease the position size, and vice versa. We can use a rolling window of N last days, and calculate the volatility as an average of logarithms of daily returns.
After such a normalization, we have a synthetic security with a chart resembling the original one, but with a volatility more or less steady across all the time span. Why is it steady? Because a recent historical volatility is a good estimate of a future volatility. It is possible to make more complicated models and predict future volatility a little bit better, but there's no point in complicating things. Naive approach “volatility most probably won’t change” works fine most of the time.
So what’s good about the volatility-balanced security?
First, you control your risks. If you enter “buy & hold” position on a calm year like we have now with VIX around 13, next year you may encounter choppy markets with VIX above 40, and then you may find out you’re not ready for this, go panic and start doing stupid things. If you keep your position volatility-balanced, you have pretty much the same volatility exposure all the time and the level of this exposure you set yourself.
Second, you benefit from a “leverage effect”, which is pretty good on stock indexes, for example. You get increased position on calm markets when they usually grow and decreased position on volatile markets when they usually fall. As a result, overall portfolio risk decreases, returns increase, Sharpe ratio is getting better.
So, we can make synthetic securities that are better than originals. What can we do with it? The same things we do with originals. For example:
1. Long-term portfolio: here is a “better” version of a classic investment portfolio with 60% in stocks (SPY) and 40% in treasuries (TLT):
2. Pair trade! Go long one security and go short another. Today’s ETF/ETN universe offers so many proxies to so many markets and sectors and asset classes, a number of ideas you can implement with such kind of a pair trading is limited only by your imagination. Both components of your pair are volatility-normalized, so you have your pair automatically balanced.
And even better. CAPM says that returns to volatility for different markets/assets should be the same because the return is a risk premium, and risk has a good correlation with volatility. As a result, pair trading with both sides of a trade volatility-balanced should have long-term mean-reverting characteristics. If one side goes too far from an equilibrium - that change its “return to risk” property and this breaks the CAPM rule. We can bet that this trend will reverse when enough investors flood this better side.
A recent example of a pair: TLT long vs. SPY short
Using this kind of charts we can easily estimate assets/markets imbalances. Just watch for a divergence from long-term average levels. This is how you can use volatility-rebalancing for market research.
Lately, we’ve finally built a tool to work with volatility rebalancing: Cognitum Rebalancer.
Most useful part of the software - it counts transactional costs. There are many liquid ETF on the market, but not-very-liquid ETFs are much more numerous. Some of them are very interesting, but their bid-ask spreads are far from 0.01. In pair trading, when rebalancing is particularly active, costs may become an issue. If transactional costs may kill your idea, you’d better know this in advance.