Via John Quiggin, I came across (and really liked!) this piece by Scott Sumner defending Eugene Fama's Efficient Markets Hypothesis (EMH).
I interpret him as saying the following:
- The claim that a "bubble" exists when asset prices violate economic "fundamentals" is full of s*^#. Because there's no (algorithmic, mathematical) way of knowing what the "fundamental" price of an asset is. Hence, there's no sure-shot (again, read algorithmic) way of predicting a bubble; people who make it sound really easy (take asset prices, compare them with the "fundamental" price, find the difference, and voila, you know if there's a bubble) don't know what they're saying.
- Yes, people do predict bubbles and they have a 50/50 chance of being right. No one has been right 100% of the time, and certainly, no algorithm. It's only in retrospect, with 20/20 hindsight that one can know a bubble exists, because it's only in hindsight that one can know the factors that went into making a bubble.
- The reason that people do keep predicting bubbles is because a "cognitive illusion" is at work here. Because they think they can predict bubbles but they really can't because they end up being wrong 50% of the time. If they build models that predict bubbles, the models will only work 50% of the time too.
Economists are famously open* about the fact that they consider their discipline to be superior to most other social sciences, like sociology and political science. More rigorous, more empirical, etc. etc. Of course, sociologists disagree.
What Sumner's piece reveals is that simply using a lot of mathematics in your models doesn't really guarantee that your discipline is any more rigorous than, say, sociology. (I don't mean that sociology is not rigorous, I am just saying economics isn't more rigorous than sociology, like, say, physics is. Or, in other words, the social sciences and the natural sciences are different sorts of beasts and economics is a social science, despite all its pretensions to be otherwise.) Sumner seems to be saying that is impossible to construct a general-purpose algorithm (or mathematical model, take your pick) that is able to predict a bubble correctly every single time. To perceive that a bubble exists (and more importantly, to do it before it bursts and takes us all down with it) requires careful interpretation, not a derivation.
It turns out people do predict bubbles, even if they are wrong 50% of the time. Sumner thinks this is because they have a "cognitive illusion". This is wrong-headed. I think what this says is that bubble-prediction seems to be some kind of lived-in, embodied, practical skill that people acquire by living in the world and being a part of it (so of course, they think they can predict bubbles). So I can feel that housing prices are too high, and will crash, and of course I can use mathematical models to justify it to others, but that feeling (or rather, intuition) is based on far more than what that any mathematical model (or algorithm) can capture. In other words, knowing that a bubble exists is a tacit skill, to use Michael Polanyi's phrase, and therefore is something that just can't be captured in a general-purpose model that works every single time. It needs to be done on a case-by-case basis.
So consider this scenario. Person A predicts that our current situation is a bubble and person B says it is not. Person A has a model and person B, perhaps, has one too. Then it is impossible to say who is right, A or B, and be right, every single time. And the decision can certainly not be made just by looking at each model. One needs to see the assumptions the model is based on, the factors a model may overlook, or something else altogether. In other words, it is not only impossible to construct a general-purpose algorithm to predict bubbles, it is also impossible to construct an algorithm that can tell us whether A or B is right. (This, of course, is Thomas Kuhn's point in Structure, only he talks about competing scientific paradigms.)
Sumner's post, it seems to me, leads inexorably to the conclusion that economists stop looking for "generalities, principles" of the economy as a whole (or at the very least, stop pretending that they possess some methodology that is superior to the other social sciences). That a "unified social science" is not really going to materialize and we (all of us social scientists) are all better off looking into issues (in this case, bubbles) on a case-by-case basis. Seems to me that this is something that sociologists and anthropologists have long acknowledged.
Would this be a completely wrong interpretation?
* Yes, even that anti-EMH Paul Krugman):
As for social sciences other than economics, I am interested in their subjects but cannot get excited about their methods -- the power of economic models to show how plausible assumptions yield surprising conclusions, to distill clear insights from seemingly murky issues, has no counterpart yet in political science or sociology. Someday there will exist a unified social science of the kind that Asimov imagined, but for the time being economics is as close to psychohistory as you can get.