It was my great pleasure – and honor – to attend this year’s Nobel prize ceremonies. It started with the Nobel prize lectures, which I found very thought provoking.
Shiller
I’ll work backwards, as it was thinking about Bob Shiller’s talk that taught me the biggest lesson. Preview: this will start pretty negative, but I learn a big lesson by the end. Hang in there, Shiller fans.
I thought I thought we had reached a consensus on volatility tests. Shiller (and others) brought us volatility tests, while Fama (and others), starting in 1975, showed that all sorts of returns are forecastable at long horizon. After sturm und drang, we – including Campbell and Shiller, but also a wider literature (I wrote a few papers) – realized that volatility tests are exactly, mathematically equivalent to return forecasting regressions. Expected returns (true measure) vary over time, a lot, and fully account for volatility tests.
The remaining question is whether time-varying expected returns are connected to macroeconomic quantities through marginal rates of transformation and substitution, or whether people misperceive probabilities and don’t know about time-varying expected returns.
There is a joint hypothesis theorem – probability and marginal utility always enter together in asset pricing formulas – so no amount of staring at prices will ever solve this interpretation question. We need models. Economic models (such as habit persistence) give a somewhat successful answer, but are also rejected. The great challenge for behavioral finance is to produce similar, scientific - looking models that tie irrational expectations to other data in a rejectable way, and thus rise above ex-post story telling.
Volatility tests were a deeply important, Nobel-worthy part of this story. They showed the economic importance of time-varying expected returns – and the as yet incomplete effort to understand those returns – in a way that t stats and R2 values did not.
Well, that’s what I thought the consensus was. What I found remarkable is just how much of that consensus Bob completely abjured.
At 1:12 Bob starts right in:
What is a bubble? You [Gene Fama] said nobody defines it. So I will define it. A speculative bubble is a fad. People get excited sometimes. Too excited… Prices start going up, they start talking, the newspapers start writing about it, more and more people pile in to a market and they push prices up more and it goes on for a while. eventually it breaks and the bubble bursts.
That’s not a “definition.” That’s an
explanation, a
theory. A definition tells you in an operational way what pattern in the
data describes “bubble.” An explanation is a theory that predicts the defined phenomenon.
That doesn’t answer Gene at all. Gene asked Bob how to
measure a price above “fundamentals.” how to
measure that a “fad” is underway? For example, in a previous podcast, Gene had offered to believe in bubbles if Bob could show a method that reliably forecast a negative market expected return.
Bob pointedly did not take even that olive branch, that chance to agree on a common language. If we can’t get straight what a definition is vs. an explanation, maybe the physicists are right that they shouldn’t give out economics Nobels. We’ll surely be at this another 35 years.
Next, Bob put up an update of the famous volatility graph, where he contrasts actual prices with ex-post dividends discounted at a constant rate. (1:15:45)
He called the dividend line “the actual market if everyone knew the future” and the “true value.”
(A minor thought. Really? Would the world really be working right if that’s what stock prices had all the return and no risk? If we have an equity premium puzzle now, imagine what it would look like with no risk! If prices have no risk so we should discount dividends with riskfree rates, the major failure of today’s markets is not the volatility of the price-dividend ratio, it’s the level, which should be many times higher?)
Admitting briefly that efficient markets allow some return forecastability, he showed us some graphs discounting dividends with interest rates and consumption growth raised to a power.
On this evidence, he concluded that we are "seeing repeated fads and fashions" though they are “integrated with the economy” in a way that is “difficult to understand.” Nonetheless, we can conclude that “The market is too volatile, people are a little crazy, there is a social psychological component.”
How do we we get from the failure of
one model (constant expected returns, or power utility) to the failure of
any possible model, to “people are a little crazy?”
More deeply, in the face of the joint hypothesis theorem, how do you get to claim victory for any view without a model at all?
More deeply, we’ve all been over and over this. The subsequent literature answered all this years ago. How could Bob not know that or even mention it?
At 1:23, he described the Campbell-Ammer variance decomposition, concluding “only about a half or a third of the fluctuations in the stock market could be explained by evidence about future dividends,” and concluding, “so most of the market doesn’t make sense”
This was really revealing. Bob’s Campbell-Ammer
slide says “excess [expected] return variation two to three times that of [expected] dividend innovation” His
words were “most of the market doesn’t make sense!”
Add this up and it’s all eye-popping. Bob is basically denying the 20 year old
theorem that volatility tests are equivalent to time-varying expected returns. I listened to the lecture and carefully to the video. You won’t find an admission of that theorem, or that
mechanically time varying expected returns account for these plots. That’s especially astonishing given that the Nobel committee cited him for discovering long-run return forecastability, ignoring Fama’s role! For example the
Nobel poster said
“Beginning in the 1960s Eugne Fama demonstrated that stock prices are extremely difficult to predict in the short run. .. If Fama’s results are right, then shouldn’t it be even harder to make predictions over several years? The answer is no, as Robert Shiller discovered in the early 1980s.”
Bob is denying the joint-hypothesis
theorem that probability and marginal utility always enter together, so we need a model of either to say anything. And Bob is denying the essence of what it means to supply a
definition. Bob closed with an overview of psychology and sociology concepts that inspire his views,
He urged economists to incorporate more ideas from psychology, sociology and other fields, “I think that in understanding speculative bubbles we have to be eclectic. .. population biology… epidemiology, neuro economics.. To understand complex phenomenal we need to take account of every kind of expertise.”
OK, "listen to psychologists" is good advice. Economics has benefitted from intellectual arbitrage many times in the past. But Nobel prizes are supposed to be given for past successes (typically, long-past!) not “maybe you can do something with this in the future.”
In an entire lecture, Bob did not give a single concrete example of how “listening to psychologists” produces one concrete positive step to understanding “bubbles.”
(There was a lot more in Bob’s speech, including description of his innovative work with Case in constructing a real estate price index. Curiously, he showed how today’s forward prices are forecasting another “bubble” –
this market price correctly forecasts “fundamentals,” unlike all the others? And he closed, advocating more markets, such as GDP futures, admitting they will have bubbles and fads too, but that they are useful anyway. “What I’ve done is present imperfect evidence…with the conclusion that’s maybe radically different about bubbles, but not about the general importance of our financial markets.”)
Deep Breath. Another viewIt slowly dawned on me though, that this is much too harsh an evaluation and an unsatisfactory theory. Bob is a smart and thoughtful guy. The theory that he doesn’t know the difference between a definition and an explanation, hasn’t read Fama’s 1970 definition of “efficiency” or “joint hypothesis,” doesn’t understand that volatility is exactly the same as return forecastability, and so on, just doesn’t make sense. I remembered my Kuhn (Structure of Scientific Revolutions) and McCloskey (Rhetoric of Economics). (If you’re an economist and haven’t read these, do so now.)
I realized just how deep and audacious Bob’s project is. He is telling us to abandon the “scientific” pretense. He wants us to adopt a literary style, where we look at the world, are inspired by psychology, and write interpretive prose as he has done. When he says that the definition of a a bubble is a fad, he isn’t being sneaky and avoiding the argument. He means exactly what he says and wants us to think and write this way too. A bubble, to Bob,
is defined as any time a time that he, writing about it, informed by psychology, and reading newspapers, thinks a “fad” is going on. And he invites us to think and write like that too. A model is, to Bob, wrapped up in one person’s judgement and not an objective machine. If I complain that this is ex-post story telling, he might say sure, stop pretending to be physics, write ex-post stories. If I complain that there are no rules and that this is no better than “the gods are angry,” he might say, no, read psychology not ancient theology, and the rules are you have to couch your story telling in their terms. He does not want us to try to construct models, either psychological or rational, that make quantitative predictions.
He wants to fundamentally remake how we do finance, how we talk about finance, how we write about finance. He wants to define a new rhetoric of finance. When he says we should read psychology and social psychology – and, implicitly, not physics or economics – he means exactly what he says. He (obviously) isn’t going to fall in the trap of writing rejectable models, making predictions and so forth. That’s like speaking Greek, and at his party, we speak Latin.
I am by nature a listener, an integrator. I wrote a paper on how volatility tests are the same as Fama French regressions. Bob has no interest at all in listening or integrating. He wants to redefine how we do things in his own style, as pure and simple as possible.
This is what scientific revolutions are all about. This is what Nobel Prizes are all about. They give them to people who strike out, write a novel language and methodology for conducting research, and convince others to follow and do it their way and talk their language. All previous revolutions – successful or not – have had these interminable debates where we can’t even seem to agree on the meaning of simple words (“efficiency,” “definition”, “model”) and talk past each other. The salient facts and classic tests are only written ex post by the winners. Bob wants a revolution of that sort, and listening to
economists is the last way to accomplish it.
Now that is an audacious project! And Bob has collected a lot of people who talk and write his way. Not me, so far – only one in ten attempted scientific revolutions catch on, and I’m placing my bets elsewhere. I still like to talk like a physicist. But I think I understand the audacity of the project, and why it is we seem to talk to cross purposes and not even agree on basic questions like what constitutes a definition, what’s a theorem, and whether the absence of quantitative rejectable behavioral models that tie expected returns to other data matters or not. And why trying to debate – to ask for a definition of bubble, for a quantifiable measure of “fundamentals”, to ask for a quantiative model of distorted expectations – will get nowhere.
HansenWith that thought in mind, I came to a similar different view of Lars Hansen’s talk. Lars isn’t in the middle of Gene and Bob; Lars is way off on the other end of Bob.
Lars chose to talk more about his current research and less about the research that got him the prize, a good technique for these lectures. He’s working on “ambiguity,” how to handle the fact that we don’t really know what the right model is, and, even more interestingly, how to construct models in which the people in the models don’t really know what the right model is. Typically for Lars, this is a very deep research program, which may lead to a fundamental difference in how we think about risk and information in economics.
At one point he described which he described models with "twisted expectations.“ Here’s the slide
In the first equation S with a tilde on it represents marginal utility, consumption to the gamma power in the usual formulation, X represents an asset payoff, and Q is then the price. This is the standard present value formula – except Lars wants to think about E as a "distorted” expectation. Following the usual theorems, in the bottom equation we can represent the same idea with the real expectation and an extra M term multiplying the stochastic discount factor. (Yes, everyone else uses M for Lars’ S, and P for his Q.) This is essentially the risk neutral valuation trick, that we can introduce a new “discount factor” M to represent the probability “twist.”
Seeing this, I would have been tempted to position it between Gene and Bob. Gene thinks of “efficiency” with true or rational expectations E. Bob thinks of inefficiency as “fads” meaning irrationally optimistic and pessimistic expectations. But Bob doesn’t show us how to link those irrational expectations to data. So I would have said this M, which Lars’ models do link to data, is a structured way to incorporate the non-rational distorted expectations that Bob thinks he sees into models, but in a disciplined, rejectable way.
Lars didn’t do that. In fact, when I suggested he position the talk as halfway between the “rational” and “behavioral” debate in this way, he said something deep, to the effect of he wished the whole rational-behavioral debate would just go away. Since it hasn’t gotten far in 35 years, he has a point.
But with Shiller behind me, I now understand Lars’ goal better. Lars, just like Bob, is setting forth a pure rhetoric, a pure language, a pure methodology for how we should think about finance and do finance. As Bob wants it to look like social psychology or maybe literary criticism, Lars wants it to look like physics. We write down the model, formally, and carefully. We test the model. We do not spend any time on loosely written ideas, either “rational” or “behavioral.” We don’t spend time on “alternative explanations” as is common in empirical finance. We don’t pretend that empirical work can say anything useful about whole classes of models, like “economic” or “rational” or “psychological.” In Lars’ world, the whole rational-irrational debate is a waste of time. Show us your models, or be quiet. A test can tell you something about
this model, period. At best a summary statistic like the Hansen-Jagannathan bound can tell you “this is what discount factors produced by any model must behave,” but that’s it.
This too is how Nobel Prizes are won. And looked at empirically – how many followers he has collected who write in his style – this is a successful language too.
FamaWhich brings me at last to Gene Fama, who came first. Gene gave a straightforward talk on efficient markets, long run forecastability and empirical finance. The one slight zinger was putting down some equations and citations to remind the world that indeed he started documenting long-run return forecasts in 1975. He apparently had some behavioral finance zingers in reserve, but didn’t get time to give them. The written version will be interesting.
Looked at in this rhetorical light, Gene can afford to be gracious. Gene also invented a language, a methodology, for empirical fiance. And his language and methodology did not just attract a small band of followers, but took over the finance profession, so thoroughly and completely that it’s easy to forget his influence. When Gene runs Fama MacBeth regressions, we run Fama MacBeth regressions – even if GLS might be more efficient, even if time series variation might be informative. When Gene sorts stocks into 10 portfolios, we sort stocks into 10 portfolios – even if 20 or smooth kernels might make sense. When Gene uses monthly returns, we use monthly returns. Gene writes beautiful paragraphs of prose to describe his theories, (no criticism, it’s just comparative advantage) so do we. When Gene defines terms like “efficiency” and “joint hypothesis” the rest of us use those definitions. When Gene points out differences between empirical finance and empirical economics, perhaps there you can see just how strong the Fama language effect has been.
