Saturday, June 11, 2011

Bob Murphy sees economic science very differently from how I see it, I think

In his post pointing out my misreading of the graphic in this post, Bob writes: "OK, so again: You said that Keynesian theory predicts the data would say one thing. You didn’t say this, but I assume you would agree, that Austrian theory is agnostic on that point; those theories do not predict the data would say that same thing. So, if the data conformed to the Keynesian prediction, that would be a feather in John Maynard’s cap.

Ah, alas, it turns out the data say the opposite of what you told us Keynesian theory would predict. And now you are saying, this has no bearing on the validity of Keynesian theory vis-a-vis Austrian theory, because the latter is agnostic on this point?

How can this be? If the data turned out one way, it would help Keynesian theory, but if it turns out to be the opposite, it’s a tie?"

When I was working off a misunderstanding of the graphic, I talked about what we "expect to see" according to Keynesianism, and I said it was an interesting graph (I still think it's an interesting graph - it's surprising these discrepancies would end up being anything other than noise). I didn't say it was proof of Keynesianism. But I did have an expectation going into it, and that expectation ended up being wrong.

So is that a reason to discount Keynesianism? Maybe. But I don't understand why Bob is so quick to jump on that. He jumped on it with 1920-21. He jumped on it with his recent Mises Daily article where he declared the Austrian school was right on the stimulus and Keynesianism was wrong. And he appears eager to tally this to the anti-Keynesian camp here. I don't understand this perspective.

My reaction is to say "Huh - that's weird. I wonder why that is happening?" (non-Keynesians might not find it quite as weird because their expectations weren't dashed like mine were). It seems to me we oughta try to answer that question before considering this a mark against anyone. Commenters here offered some ideas (including Nick Rowe who just said it's impossible to interpret). In Bob's comment string, I suggested a life cycle consumption hypothesis explanation - an explanation Bob apparently came to forty comments after me. That seems like a pretty reasonable explanation to me, which solves our conundrum. Is life-cycle consumption inconsistent with Keynesianism? It wasn't the consumption theory Keynes offered, but it is not inconsistent with Keynesian macroeconomics. Is it inconsistent with the Austrian school? Not that I'm aware of. So the best explanation that the crowd-sourced minds of my blog and Bob's blog can come up with for this data is once again something that is consistent with multiple macroeconomic theories. That's where I come out at the end of this, and that's what I call scientific progress. We have a lot of ideas. We take it to the data. I make mistakes. I get my mistakes corrected by my peers. I get curious and confused. We figure out an explanation of the data, and we draw conclusions. For a scientific or an empirical mind, this has all been a great experience, and I am plenty happy to admit I was wrong.

But there seems to be this tendency in some corners - instead of doing science - to do proofs and falsifications. It makes sense, of course, that Austrians think in terms of "proving" things. That's the deductive mindset after all. It shouldn't surprise regular readers to learn that I really don't think we're equipped to do that. If you're interested in abstract philosophy and epistemology, Popper is among the best. If you're actually interested in doing something useful you can largely forget Popper (well - at least you can forget Popperian epistemology - I don't want to rip on Popper himself, who may very well have agreed that epistemology is of limited use in the practice of science, except as a vague mirage). I have serious doubts that Keynesian theory is falsifiable. I have serious doubts that Austrian theory is falsifiable. I have serious doubts that any truly interesting theory, at least of a highly complex phenomenon, is falsifiable. This point is crucial for scientists to understand.

We can talk in terms of observations that are consistent with or corroborate our ideas. If we collect a lot of consistent observations and corroborations, we can be more and more confident of the value of our ideas. We'll never have knowledge of it as epistemologists perceive it, but we'll have a more and more useful understanding of our world. When you think of science as corroboration rather than science as proof, it becomes very hard to talk in terms of theories being "wrong". I've run into this in two places in my short career: critiquing Buturovic and Klein on economic enlightenment, and critiquing Woods, Murphy, and Powell on 1920-21. Both made what appeared to me to be pretty extraordinary claims: that liberals were less economically enlightened than conservatives and libertarians, and that 1920-21 supported the Austrian school over Keynesianism and monetarism. What was my response? That Buturovic and Klein's data supported a number of conclusions (I offered which of those seemed more plausible to me), and that 1920-21 supported a number of conclusions (I didn't really advocate Keynesianism as more plausible there so much as defend it as being quite consistent with the data).

It's relatively easy to demonstrate how different theories are consistent with the data. Why? Because the economy is a very complex subject of study and lots of different processes can be operating at once. For the exact same reason, though, it's very hard to find inconsistent evidence that can be used as a disproof of a theory. The data on GDP(E) and GDP(I) didn't match up with my initial expectations, but they do match up with a prominent and well regarded consumption theory. So does that inconsistency mean (1.) that Keynesianism has a mark against it, or (2.) that the mechanisms governing consumption dominate in the data? Who knows? I certainly don't and Bob certainly doesn't. Now, if it were consistent with a rival macroeconomic theory, then Keynesianism would be in more trouble, I'd think. But I'm not aware of any macroeconomic theory that would expect what was observed in the data. Bob says the Austrian school is "agnostic" on this question, but it's really not. The Austrian school cannot be used to conclude that "if expenditure exceeds income, unemployment increases". Nothing says that, so we look elsewhere, and consumption theory gives us some insight.

My advice is this - do science by racking up corroborations and consistencies. Don't fall into the trap of trying to disprove other ideas. That's a deductionist mentality. That's the mentality of what Keynes called a "remorseless logician". It's damn near impossible, so you're only going to look silly trying to do it. Look at past posts on this blog to see how many times I've said "Austrian theory is wrong" or "X disproves the Austrian school". I try to put the heat on some elements of Austrian theory, and I do quibble with methodological and philosophical points. But you don't see me proclaiming the falsity of the Austrian school. Why? Because disproof is hard (as is proof).

So in a way Bob's characterization of me is accurate. I just want to explain why I approach things that way. Consistent data and corroboration is a good thing. Collect that stuff and use it to make judgements. Inconsistent data should first be explained - it may have nothing to do with what you were originally thinking about. I guess the best way to sum it up is with these two points:

1. In science, inconsistencies are much more likely to bring you a deeper understanding of what's going on than they are to bring disproof. In logical deduction, inconsistencies are much more likely to provide a disproof but not a deeper understanding of what is going on. Science is not deductionism, and

2. Theories succeed in science by being capable of explaining a wide range of observation. Theories are usually surpassed by other theories that do a better job at explaining observation. If other theories can't explain the observation better, they offer no more consolation*.

*Kuhn makes this point repeatedly - science doesn't progress unless a better explanation emerges. Inconsistency with one theory without consistency with an alternative theory means very little. The Austrian school and Keynesianism (minus life-cycle consumption) are equally incapable of grappling with observation. Keynesianism plus life cycle consumption theory and the Austrian school plus life cycle consumption theory do do better. We have made progress by determining that amongst ourselves.


  1. Daniel, can you point us to where you criticized Krugman's deductionist viewpoint in the numerous posts he's made, like this? If I didn't know anybody, I would say Krugman feels vindicated since Keynesian theory predicted interest rate and inflation movements better than the Chicago School and Austrian guys. But surely Krugman isn't a follower of non-scientific Austrianism.

    To drop the snarkiness: Of course a single thing doesn't refute a whole theory. I'm not asking you to tatoo "I HEART MISES" on your chest. I'm simply asking you to admit that it's a little problematic that what you took to be a prediction of Keynesian theory--and an important one at that, we're not talking about CDS spreads in Mongolia but unemployment--was exactly backwards in the data. That's freaky.

    So to wrap up: I don't fault Krugman for running victory laps, except for his tone. He has every reason to think the last two years are a feather in John Maynard's cap. I'm not throwing out Austrian theory, but I'm at least admitting that it's problematic that my own predictions on CPI thus far have been wrong.

  2. Oops should be, "If I didn't know any BETTER, I would say Krugman feels..."

  3. 1. Krugman is too declarative a lot. I don't think I've ever denied this. Just the other day I agreed with you and Nick Rowe about Krugman on the CPI.

    2. What Krugman points out is consistent with Keynesianism, inconsistent with loanable funds theory of the interest rate. What we were looking at with GDP(E) and GDP(I) was inconsistent with both Keynesianism and the Austrian school (Austrians are not "agnostic" on it - they don't predict it!). On Krugman's post, Keynesianism doesn't need to invoke additional details to explain the data (although clearly the details don't hurt). In our case, both Keynesianism and the Austrian school need to invoke additional details (life cycle hypothesis) to explain the data.

    Is this a mark against my initial version of Keynesianism? ABSOLUTELY. I've pretty much said that when I tried to figure out what was going on, didn't I? But life cycle hypothesis wins on that point - and only life cycle hypothesis, and I have bent over backwards not just to admit this - but to highlight this alternative to my readers and yours.

  4. The concept of corroboration depends on the concept of falsification. A statement about evidence is only corroborating if its negation is falsifying. Therefore, if economic hypotheses are corroborable, then they are also falsifiable.

    You're right, economic hypotheses are corroborable, but only because they are falsifiable.

    When saying you "have serious doubts that any truly interesting theory, at least of a highly complex phenomenon, is falsifiable," you're not talking about the logic of falsification, but something like the method of falsification. These really aren't the same thing. The former is just a matter of formal transformations and fiat designations of truth and falsity; the latter concerns the norms and conventions of the actual practice and institutions of science.

  5. "deductionist" means, "you are correct but I am still right". Liberals arts stuff.


All anonymous comments will be deleted. Consistent pseudonyms are fine.