Mechanisms vs models

Simon Wren-Lewis asks a good question about Robert Lucas‘s and John Cochrane‘s apparent misunderstanding of the balanced budget multiplier: how can very clever people make silly errors?

He suggests two good answers. I’d like to suggest a third. There are two different ways of thinking about economics - the model paradigm and the mechanism paradigm, and the former has crowded out the latter.

Simon says:

If you spend X at time t to build a bridge, aggregate demand increases by X at time t. If you raise taxes by X at time t, consumers will smooth this effect over time, so their spending at time t will fall by much less than X. Put the two together and aggregate demand rises.

This is clear and true. And it would be obvious to anyone using the mechanism paradigm. If you ask “What is the mechanism whereby higher taxes reduce consumer spending?” you pretty much walk into the notion of consumption smoothing. Equally, though, the paradigm also leads one to other reasons to be sceptical of the practical efficacy of the balanced budget multiplier. 

But lots of brilliant economists don’t think merely in terms of mechanisms but rather build impressive models. And like photographers, they tend to fall in love with their models which distracts them both from others’ models and from mechanisms.

This matters, because the importance of particular mechanisms varies from time to time. The social sciences, wrote Jon Elster:

can isolate tendencies, propensities and mechanisms and show that they have implications for behaviour that are often surprising and counter-intuitive. What they are more rarely able to do is state necessary and sufficient conditions under which the various mechanisms are switched on. This is [a] reason for emphasizing mechanisms rather than laws. Laws by their nature are general…Mechanisms by contrast make no claim to generality. (Nuts and bolts for the social sciences, p 9-10)

A good example of this lies in the idea of expansionary fiscal contraction. The virtue of this idea is that it draws our attention to mechanisms (a falling exchange rate, better corporate animal spirits, whatever) whereby fiscal contraction might boost the economy. The drawback is that these mechanisms are just unlikely to operate here and now. Yes, there’s a model that tells us that expansionary fiscal contraction can work. And there are models that say it can’t. But arguing about competing models misses the practical point.

Now, there is an obvious reply to all this. Models have the virtue of ensuring internal consistency, and thus avoiding potentially misleading partial analysis. However, I’m not sure whether this is an argument against mechanisms so much as against poor thinking about them.

When I was a student (back in the 80s!) I learned lots of models (OK, a few), but when I became a practising economist, I found them to be less useful in thinking about the economy than Elsterian thinking about mechanisms.

This is not to dismiss models entirely. I’m just saying that, insofar as they have uses other than as mental gymnastics for torturing students, it is because of the mechanisms contained within them. The parts might be more useful than the sum.

Comments

[Luis Enrique]

hang about - don't mainstream models show Lucas and Cochrane are wrong? I'm sure Krugman would tell you they've gone wrong because they are relying on "common sense" rather than a model.

[Luis Enrique]

oh and great to see a UK academic macroeconomist of his stature blogging

[chris]

Yes - mainstream models do say L and C are wrong. But this is only because they contain the consumption smoothing mechanism. You don't need loads of equations to show they're wrong - just that one point.
And mechanisms are NOT necessarily common sense. Some (eg perverse incentives, selection effects, self-refuting beliefs etc)are non-commonsensical.

[Luis Enrique]

OK. It occurs to me that a common criticism of mainstream economic modelling is that it typically shuts down (simplifies away) many aspects of reality in order to focus on a single mechanism, which is the subject of the model (paper). Perhaps the problem in doing so is that this necessarily creates a modelled environment in which the mechanism is operable, thereby pushing aside exploration of conditions under which it may not be?

[I didn't (mean to) equate mechanisms with common sense]

[gappy]

The premise of your post ("how can very clever people make silly errors?") is plainly incorrect. Very clever people make silly error all the time of course, but ascribing one to Lucas shows *your* ideological bias, if anything.

I say this because Simon Wren-Lewis' dismissal of Lucas argument is very cursory, and just instrumental to his main point: that his argument is motivated by ideology. Things are not that simple, however, and the superficiality of his counterargument shoudl have raised a red flag. Lucas' statement was heavily debated and if you read other comments on the web by many other economists, at the very least you have to concede that Lucas' argument is not as trivially wrong as Wren-Lewis' makes it to be; and it's not primarily one of Ricardian equivalence as Krugman read it.

I am not into credentialism, but Lucas started thinking about these problems in the early 60s, and this is his very research area. He is aware (and has quoted) literature different from his. To dismiss his argument simply by pointing out to smoothing effects is presumptuous.

There is no doubt he has his biases. The interesting questions are whether a) his theoretical framework is at least consistent, b) it has higher predictive power than the alternatives.

There is another statement that is very, very wrong in your post: "But lots of brilliant economists don’t think merely in terms of mechanisms but rather build impressive models."

I say this is wrong because if you were well-acquainted in at least *some* area of economic theory, you would have written: "Mediocre economists build [seemingly] impressive models, which quickly fall into obscurity. Brilliant economists build models that have a strong interpretability and use the required toolkit". Otherwise stated, great economists are not obsessed with technique, but with meaning. Occasional readers get stuck on the technical details, and think the authors must have been too.

Talk about bias.

[Nick Rowe]

Good post Chris.

[Keith]

Your readers might want to read the lengthy criticism of Cochranes opposition to fiscal stimulus by Prof Bradford Delong at UC Berkeley. Which is a dismissal of a "Nobel winner" as talking nonsense. Not that Nobel actually created a prize in economics. But the dismissal is based on textual examination of published interviews. Enjoy!!

[BenP]

"The interesting questions are whether a) his theoretical framework is at least consistent, b) it has higher predictive power than the alternatives"

Neo-classical can be consistent (with some 'calibration') but the assumptions required are laughable.

Also if you knew Lucas's work you would know crisis is unpredictable within his model.

Neo-classical economics is ideology, Capitalism failed and had to be bailed out.

[Keith]

Well BenP you could argue that deregulated finance capitalism has gone tits up as it has often before; but the regulated welfare state form has been quite successful compared to Soviet style communism and un regulated greed is good capitalism. The thing that is odd is how there has been a huge ideological move away from the successful humanistic form of Capitalism towards versions that are both less humane and less successful in purely economic terms. As Prof Krugman points out economic growth has declined as social democratic institutions are weakened. Which also seems to go along with the rise of various odious reactionary ideas and unpleasant political movements. Nothing succeeds like failure.

[John West]

Chris - an interesting post. It led me to wonder whether you'd come across Prof. Emanuel Derman's book 'Models Behaving Badly', which covers similar ground. I found a WSJ review (they, rather unsurprisingly, think Derman's too harsh on the Efficient Market Model).

Of more interest is this podcast where he's interviewed on what he means by models behaving badly. A former physicist, he says metaphors work well in science and economics/finance but that models act like physical laws - except the 'science' of these laws is based on constantly shifting human behaviour and are consequently doomed to failure.

http://www.scienceofbetter.org/podcast/derman.html

[rjw]

very interesting post Chris

Models and mechanisms are not really complete alternatives, of course - it's just that (as Luis above highlights) formal economic models are generally "thin" in the sense that they typically focus on a quite limited number of mechanisms, and that users of the models are sometimes not predisposed - either politically or methodologically- to continually ask themselves whether their models actually capture the relevant mechanisms in particular cases.

Lucas, of course, is a good example of why economists should read more philosophy of science. I distinctly recall reading as an undergraduate something he wrote in which he claimed that any model not based on rational optimising inidividual agents could not possibly be "scientific". It was a valuable thing to read. It taught me that very smart people sometimes have ideological (or cognitive) blinkers on that are so enormous, they end up saying incredibly stupid things. A lesson I've never forgotten.

[BenP]

"As Prof Krugman points out economic growth has declined as social democratic institutions are weakened"

But you recognise this is the wrong way round?