I work in finance, and there's a class of models that works something like this:

Market A -> Market B

Very simply, you theorise that market A predicts market B, and you build a model along the lines of "regress market B returns on A" and so you decide whether to buy or sell B depending on what A did. Pretty much as simple as you can get.

However, suppose the real dynamic is this:

C -> Market A

C -> Market B

In other words, the real driver is another effect causing both A and B to move together. Let's say there's also autocorrelation in C.

In this case, A and B would still look mutually informative. When A goes up, B goes up, and so on. Furthermore, because there's a tendency for C to continue in the same direction, not only will A and B move together, a regression of B's next period returns would also look like it's correlated to A, confusing the issue of whether A is driving B or C is driving B. Of course this will come to a head when C is not doing anything and you try to make sense of why B is not responding to the noise in A.

I'm looking for terms to describe this situation, papers, readings, and so forth.

I would think it's pretty common, as it looks a lot like the simple graphs you see when studying causality, but I never found a convenient handle for the situation.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.