Scientific failures: particle physics vs. macroeconomics

In a comment thread over at Steve Williamson's blog (yes, I am avoiding work, why do you ask?), I wrote:

As for the reason for my pessimism, well, yes, [macroeconomists] can make tons and tons of models, but how do we choose between those models? We seem to have done a generally crappy job of that. It seems to me that if macro theory is not disciplined by empirical verification of the microfoundations, then theorists can just make a universe of theories to describe anything and everything, and the theories that get accepted will simply be picked by consensus, "common sense" intuition, and sometimes even politics. So I decided to focus on understanding how financial markets work, in the hope that that will help discipline macro models. 

Or, put in the pugilistic tone of the blogosphere: If I wanted to spend my life doing clunky math to describe worlds that don't exist, I would have stayed in physics and been a string theorist. ;)

(Of course, this is a bit tongue-in-cheek. I do intend to do some macro modeling at some point. And it's doubtful I would have been able to become a string theorist even if I'd wanted to, because almost zero people get jobs in string theory anymore, and I was good at physics but maybe not that good.)

But it's hard to ignore the parallels between what's been happening in particle physics and what's been happening in macroeconomics. In both fields, almost everybody reached a consensus on an approach to use (string theory in physics, DSGE in macro). In both fields, theory diverged from empirical verification and took on a life of its own. And in both fields, a small group of dissenters decided to take their grievances to the public, resulting in a lot of negative attention and outside pressure.

In physics, those dissenters were Lee Smolin, a theoretical physicist at the Perimeter Institute, and Peter Woit, a math lecturer at Columbia. In 2006, Smolin published The Trouble With Physics and Woit published Not Even Wrong. Both are critiques of string theory - both the theory itself and the academic culture that has grown up around it. If you are interested in this sort of academic dispute, or in philosophy-of-science in general, I'd recommend both books (if you already know some physics, start with Woit's; otherwise, start with Smolin's). Woit also writes a blog on the subject. (Update: Lawrence Krauss, a theoretical physicist at Arizona State, is also a public critic of string theory; his book, which I have not read, is called Hiding in the Mirror. A number of famous physicists have been publicly critical of string theory, including Richard Feynman, Sheldon Glashow, and Robert Laughlin.)

Long story short: The problem with string theory was really a problem with particle physics. Our existing theories work too well. Every single thing that's come out of particle experiments in the past 40 years has agreed with the Standard Model to the nth decimal place. Of course, the Standard Model is pretty ugly, so people have been working on ways to make it more mathematically and conceptually simple (i.e. "beautiful"). Problem is, without new unexplained experimental observations to guide us along, that's an insanely hard task. Physicists have tried one beautiful theory after another for 40 years, nothing worked, and string theory just happened to be the one that most theorists converged on and decided was the best bet.

Unfortunately, string theory hasn't worked out either. Maybe there will be a breakthrough, but right now the situation doesn't seem to be looking so good. For one thing, much of the math that has come out of string theory has been pretty ugly itself. (Update: Actually, plenty of beautiful pure math has come out of the search for string theory, leading to two Fields Medals that I know of. The "ugly" part is that string theory so far doesn't have simple, beautiful equations like successful theories in the past.)

More importantly, though, string theorists haven't been able to get their theory to make any testable predictions. Usually this kills a theory, but physicists have stuck with string theory for quite a while now. Why? Smolin and Woit attribute it to the culture of the profession. I have a more prosaic hypothesis, which is that theoretical physicists basically have nothing to do, and are using string theory to keep their skills sharp until the day when the world needs them again...

In any case, whether or not because of the public shaming by Smolin and Woit, string theory's star has dimmed quite a bit. The media has stopped hyping the theory so much. And there is a sense that confidence in the theory has also fallen within the profession itself - not because of the rebel authors, just because of the general going-nowhere-ness of the theory. Universities have basically stopped hiring string theorists. 

In macroeconomics, the public insurgency has been led by Krugman, DeLong, Quiggin, and a few others. The complaint against DSGE (or "Lucasian macro," as Krugman labels it) has been similar to that against string theory: Lucasian models, say the insurgents, failed to predict the crisis, and failed to predict which policies would be the most effective in combating the recession that followed. If theories can't predict things, say the rebels in both fields, they should be junked. In macroeconomics, like in particle physics, the rebels claimed that a stifling culture of intellectual conformity is inhibiting the search for new approaches.

Now here's another striking parallel. In both controversies, a handful of defenders of the existing paradigm have resorted to arguing that theories shouldn't need to make predictions. 

In physics, Lenny Susskind, one of the inventors of string theory, has invoked the "anthropic principle" to defend non-predictive theory. Remember back in elementary school, when you'd ask some kid some question - "Why did class get out early today?" - and he'd say "So you would ask! Hurr hurr hurr!" That's the anthropic principle. It says that the laws of physics are the way they are because if they were any different, you wouldn't be here to study them. Susskind has argued that string theory's failure to make definite predictions is OK, because the anthropic principle fills in the gaps. Even though string theory allows certain laws of the Universe to be dang near anything at all, those laws are the way they are just because they had to be that way for you to ask why! Most string theorists, I should mention, are not on board with this idea. The empirical culture of physics is very strong and very old. Nevertheless, the anthropic argument has won string theory some bad press among the wider populace, since if it's true (and hey, it might be true!), it basically means that all the high energy particle theorists are no longer needed and can just go home.

In macro, no one invokes the anthropic principle, but there have been several reasons given why DSGE models - or the DSGE paradigm - should not lead to definite predictions. The first big argument is that economic models should not have to be validated by empirical data. Models' purpose, some economists say, is not to explain observed phenomena, but to "clarify our thinking" - in other words, to work out the logical implications of assumptions that we choose a priori to accept as true, based on our common sense or whatever. The second defense is that since macroeconomists have created a huge number of models that describe a huge number of different possible ways that the economy could work, any outcome that happens was successfully predicted by macro theory. Obviously, neither of these arguments is very scientific! Other "supporting" arguments include A) the (logically false) notion that financial market efficiency makes economic predictions impossible and B) the gross misuse of George Box's quote that "all models are wrong."

As in the case of anthropic string theory, these protestations fail to hold much water with the public, which naturally asks "Then what are we paying you guys for?", while also pissing off any economist who believes in the Francis Bacon notion of science.

This cycle of theoretical stagnation --> comfortable professional consensus --> internal revolt --> outside pressure seems to be common to a lot of scientific disciplines. However, there are a few big differences between the two insurgencies I've described, all of which suggest that Lucasian macro will be around long, long after string theory has disappeared behind the event horizon of history.

For one thing, physics is suffering from being too good. Our existing theories may be ugly, but they work devilishly well. This means that if physicists decide to abandon string theory, they will just dream up something else, possibly using many of the very mathematical insights that they gained from working on string theory! In macro, on the other hand, we basically have no theories that work well (or if we do, we don't know which ones they are!), so if we give up on DSGE, we're stumped...and people don't like to be stumped. Second, while demand for high energy particle theorists appears to have collapsed, demand for macroeconomists continues to be incredibly strong. Finally, economics as a profession does not have the experimental tradition that physics has, which means that not making predictions is a much less severe public relations blow to a physics theory than to an econ one.

So Lucasian macro will survive even as string theory evaporates. Yet with millions unemployed, it seems a lot worse to be ignorant about the economy than to be ignorant about black holes and the birth of the Universe. The Macro Wars have higher stakes than the String Theory Wars, at least in principle.

OK, back to work. Posting will be pretty sparse for the next month, while I finish up my dissertation.