The EMH Isn't Testable and That's OK - Part II

In Part I, we defined the Efficient Markets Hypothesis and introduced the joint hypothesis problem. I don't want to primarily post on zombie debates, but John Quiggin's book gives me a chance to discuss a little bit about what a lot of modern finance is built on, before I start blogging about whether or not stochastic volatility is a risk factor in a linear factor model (stay tuned!).

I'll let Quiggin start us off with his view:

“But as a string of philosophers of science, being with the late Karl Popper, have shown, a theory that can’t be refuted by any conceivable evidence isn’t really a theory at all…The global financial crisis, along with the earlier dotcom crisis has shown that, on any ordinary understanding of its terms, the efficient markets hypothesis can’t be right…So supporters of the efficient markets hypothesis have sought a redefinition that would make it invulnerable to refutation…This argument in one form or another has been put forward by all the leading defenders of the EMH, notably including Eugene Fama and John Cochrane of Chicago and Scott Sumner of Bentley University.”

Quiggin argues in the chapter that trying to squirm out of irrefutable evidence EMH defenders have changed the theory over time to render it useless. That narrative is completely false. Here is Eugene Fama in 1970 in a seminal paper on the EMH, before he was the “father of modern finance” and was just a young professor trying to sort out the theory, “the theory of efficient markets is concerned with whether prices at any point in time "fully reflect" available information. The theory only has empirical content, however, within the context of a more specific model of market equilibrium, that is, a model that specifies the nature of market equilibrium when prices "fully reflect" available information.”

It’s not nearly as elegantly stated as it is in the previously linked podcast, but that's the joint hypothesis problem in its first formulation in 1970. I am really surprised it isn’t front and center in Quiggin’s account of EMH, because it is certainly front and center in any finance course discussing the testing of EMH. The joint hypothesis problem isn't new, and it isn't something that proponents of EMH have been hiding for the last forty-two years.

Is this a sham? In finance, It's pretty humbling to try to create the "right" theory for prices. The world is complicated and modeling that is hard. How many essential types of risk are there? Can I really model them all? Even the most important ones? What about cash flows? Am I certain I can predict how the growth of the internet and technology will effect firm's various business activities? Is it possible to imagine a world where the dotcom bust doesn't happen, a world with a few more Amazon's, a world where the biotech boom that was so desired actually pays the dividends we dreamed about? Believe me, I'm open to the idea that bubbles can occur. Here is a really good case that the dotcom crash was a classic bubble/mania, but let's admit that the endeavor of diagnosing a bubble even ex post is not simple, much less ex ante. We only observe one outcome of many possible futures.

So I've argued that Quiggin's narrative is wrong. Testing EMH may be impossible, but that isn't something Quiggin discovered. Academics have been discussing the joint hypothesis problem, since the EMH's inception, rather than stuck in a self-erected cell built by constant squirming from critics. In part 3, I'll try to argue all is not lost, if the EMH isn't testable.