Much of the disagreement about the truth or falsity of the efficient market hypothesis seems to stem from a disagreement about what constitutes identifiable. Scott Sumner, and a lot of other EMH proponents, define bubbles as identifiable only if the method of identification can reliably be profited from. Ryan Avent, Andrei Shleifer, and other EMH critics, argue that because of the risk in betting against a bubble, and the lack of perfect capital markets, one could have a relatively certain belief that a bubble exists but not be able to profit from it.A third route is to agree with Sumner that are identifiable and the fact that you profit from them reliably proves it, thus EMH is false.
I think this is part of the reason Sumner is frustrated by the anti-EMH crowd: he isn’t distinguishing between those that identify profitability as sufficient proof from those who don’t, so he is faced with some people arguing that “people reliably profit off of bubbles, therefore EMH is false!” and others arguing that “market irrationality prevents reliably profiting off of bubbles, and EMH is still false”. There may be people who hold both of these contradictory points at the same time, and they are wrong. But the existence of two separate arguments against EMH that happen to contradict each other is not evidence against either theory or for EMH.
Aside from the insistence of profitability, the other problem I have with Sumner’s definition of identification is the claim that identifying one bubble is not enough, you must also able to identify them. This is a very econometrician-centric way of thinking of identification: you get a particular model of asset prices and estimate a set of parameters, which you can then use to forecast. If your forecast is not reliable, then your model or your parameters are wrong, and your identification is false. But why should a model that can identify a housing market bubble be capable of identifying a tulip bubble or a tech bubble? For that matter, why should an individual with a particular set of knowledge that has allowed him to idenfity a bubble in one market be held to the burden of proof of identifying the next bubble, which will likely be in a completely different market of which he has no knowledge? Take Calculated Risk, for instance. He clearly and specifically identified the housing bubble using his extensive knowledge of the entire housing industry- from realtors to securitizers. Are we really going to demand that he identify the next bubble in gold, oil, Beanie Babies, or some other market he knows nothing about until we accept that he identified the housing bubble? I see no reason to expect this to be true. Nor do I see any reason to expect that all bubbles be equally identifiable. Using Sumner’s criteria, I’m not sure how you tell the difference between a world in which 1/5 bubbles are identifiable and a world in which 0/5 bubbles are identifiable.
Look, I’m not coming down strongly on one side of this debate- my instinct actually tends to be that, in general, you can’t beat market prices, and we should have humility about disagreeing with them. I also understand why a definition of bubble identification that does not include a higher burden of proof is unsatisfactory; it lends itself to people like Dean Baker claiming they identified the current housing bubble in, I think, 1948. Nonetheless, I’ve obviously got disagreements with the way Sumner and other EMH proponents define bubble identification. I’m still open to persuasion on this.