Over the past few weeks, spring has arrived, in what has been a very challenging period across many countries in the world. Looking outside, it’s a time of new beginnings, thankfully making a change from the past few months of winter. Markets have their seasons too. Some might actually mirror the actual seasons outside.
I remember many years ago, our team at Lehman Brothers had a model to look at FX volatility named after Cassandra, who in Greek mythology could tell the future, but was not believed. I’m not sure whether the model had a track record comparable to the Cassandra of Greek mythology, but it was nevertheless an insightful model. The basic idea was to model the seasonality of FX vol. At least at the time, there were notable patterns in FX vol which mirrored the seasons (and I expect if I were to repeat the exercise today, we might see something similar).
During the northern hemisphere summer there was typically a lull in FX vol, explained by the fact that market participants might be thin on the ground on trading desks, before a rise in Autumn as they got back to work. Interestingly, AUD and NZD FX vol didn’t quite fit this pattern, perhaps explained by the fact it was southern hemisphere winter during those periods?
This of course is one example of seasonality in markets. There are many other examples for seasonality, which we might wish to check (I don’t necessarily know if all of these work, but they sound reasonable hypotheses to me!), such as the pattern of equities vol around periods characterised by earning seasons. There are also numerous patterns in commodities, which impact the shapes of the futures curve in different assets. We might conjecture that seasonal demands in natural gas (eg. over winter) or gasoline (over the US summer driving season), would have an impact. Looking to cash market in FX, we might wish to check whether seasonal revenues from tourism dollars impact a local currency and so on.
In order to formulate these ideas, we need to have some domain knowledge. Whether we can profit from every seasonal hypothesis, depends on whether we can validate it (!) and how persistent it is. We can do seasonality analysis pretty easily with a tool like Python, but if haven’t got a very good explanation of why we might it in a certain asset, then perhaps it is just a matter of data mining. Another issue, with trying to identify seasonality is the relatively small number of points involved.