If you see a massive queue around the block outside a restaurant, what does it imply? Well, for one it’s popular. If it’s popular, you’d assume the food is also pretty good. Is this always the case? Sometimes it may well be the case. Other times, it’s simply because of hype, and the food is mediocre at best. How do we reduce the chance of it being a case of hype? Typically, we try to identify those places where locals are queuing rather than tourists. Essentially, we assuming there is the positive correlation between food quality and queue length.
In financial markets (yes, yes, I can hear correlation is not causation being uttered here), correlation is a key part of making decisions. The difficulty is that correlations are not always stable, just like our burger example. Let’s take the example of one of the simplest strategies, trend following.
For trend following to “work”, we are implicitly making the assumption that there is some autocorrelation in the price. If prices rise, we expect them to rise. Conversely, if we are trading mean-reversion, we are assuming a negative autocorrelation, namely that prices rising will be followed by falling prices and so on. When markets are ranging, trend following strategies will underperform and when they’re trending, mean-reversion won’t work well. If we run both strategies together, they can complement each other.
Then there are cross market correlations. Perhaps unsurprisingly, during periods of risk aversion, correlations will increase and for example AUD/USD will end up being a proxy for S&P 500 futures. Similarly, bitcoin also often behaves like a risk on/risk off proxy. More broadly vol assets will tend to sell off during periods of market turbulence. However, it’s not all risk on/risk off correlations. In FX for example, changes in the interest rate differential, a proxy for monetary policy divergence, can be correlated to FX spot, and it is these changes which are important rather than the absolute levels (Brent Donnelly has written about this recently in his AM/FX newsletter).
If we have a fundamental view to make a trade, we are taking a view about a correlation. If we expect inflation to fall, and we are buying bonds as a result, we are assuming that inflation rates are correlated to yields. But even with fundamental data, the market reaction can vary. At certain points, inflation can be a key market mover, at other points the trade deficit can be a key number. One of the most frustrating things, can be having the correct view about a certain data release or broader fundamentals, but price action seemingly ignores it.
Correlations are unstable, and they will depend on the market regime. Some correlations may well be spurious and best avoided. However, taking a view on the correlation, is a key part of trading (not just in terms of instruments like correlation swap), even if sometimes those correlation assumptions are implicit.