Don't Go for the Exacta
by Clifford Asness, Ph. D. AQR Capital Management, Inc.
This issue comes up often, but recently I’ve seen more than the usual amount of “we forecast X, so we’re betting on Y” type stuff where X and Y are different events. The most common, but certainly not the only, example I’ve seen of such a two-step bet is having a strong currency view with a recommendation to buy, or sell, some specific companies that you hope will benefit, or get hurt, if your currency forecast comes true instead of directly buying or selling the currency.
Another example might be having a view that inflation and interest rates will come down, and recommending owning small growth companies because most of their cash flows are further out in time and so falling rates might (might!) benefit them more than other companies. In the first example you recommend betting on specific companies instead of betting on the currency itself, and in the second example you recommend small-cap growth stocks instead of, you know, just buying bonds.
This type of two-step bet is usually a bad idea. At the very least it is worse than the original, simpler, one-step idea.
It’s hard enough to be right. It’s much harder to be right multiple times in a row. Say you’re 70% sure the dollar is going to have a big down move. (Well, obviously, first ask yourself why you’re so sure of such a thing. I’m presuming the much bigger issue that one can have such views!) Then assume you are also 70% sure that if it moves as you predict, a certain set of companies will benefit (it’s hard to imagine companies that don’t have a lot of non-dollar-related variance, so 70% seems pretty impressive). Let’s also assume that if you’re wrong about the dollar (a 30% chance) there’s another 30% chance that these companies end up doing well for other reasons (these probabilities don’t all have to be 70% or 30%, but, hey, it’s my example).
You’ve just put yourself under a 50% chance of being right for your intended reason and a total chance of being right of 58%. Specifically, it is a 49% probability that you’re right about the dollar and then right about how the companies react (70% x 70%) plus another 9% you’re wrong about the dollar but accidentally right about the companies (30% x 30%). This is decently less than your original 70% likely currency view, so you’ve managed to turn a 70% likely advantage into a 58% likely one — presumably for no better payoff than making a single one-step bet!