Tuesday, April 20, 2010

Fallacious Fallacies and the Great Recession

Aggregate Demand and the Great Recession
On April 17th, the chair of Obama's Council of Economic Advisors chairwoman (and W&M alum) Christina Romer set off a firestorm in the economics blogosphere with a relatively benign rendering of mainstream economics through the medium of a shout out to 1990s political culture. Namely, that: "it's the aggregate demand, stupid". Romer's case is that aggregate demand is driving unemployment as opposed to unemployment insurance. The Economist provides a general overview of the question, while a paper prepared for a Brookings panel and a San Francisco Fed paper provide some empirical structure to the debate.

Bloggers have lined up on either side. Menzie Chinn, Michael Derby (WSJ), Mark Thoma (here and here), and Brad DeLong take Romer's side. Arnold Kling and Megan McArdle are opposed to her position. Tyler Cowen and Bryan Caplan are also opposed, but their posts go down this strange rabbit hole of nominal wage rigidities, which is very unusual because (1.) Romer never mentions nominal wage rigidities, and (2.) nominal wage rigidities have exactly zero to do with aggregate demand deficiencies, although some economists are fond of referencing them.

Fallacious Fallacies
Some of the more interesting posts on the issue get into alleged fallacies committed by the pro-Romer aggregate demand crowd. I'm going to call these "fallacious fallacies", because they miss the mark by a fair margin, they threaten clear thinking, but you hear them a lot because human beings (particularly bloggers) love to play "gotcha". So I'm gonna call "gotcha" on the "gotcha" guys. The fallacious fallacy in question is the "aggregation fallacy". While most economists content themselves with the fact that aggregate demand is real and job search intensity (which is lowered by generous UI benefits) is real, and we have to arbitrate between their relative importance, some economists flatly reject the validity of aggregating anything. Robert Higgs recently listed the "aggregation fallacy" first in his list of six alleged fallacies. The argument is that by aggregating a variable like income into a variable like GDP, or an individual price into an aggregate like the CPI, complex processes at a lower level of aggregation are glossed over. This critique can come in several varieties, but it's disconcerting for me that it is ever expressed in the all-encompassing way that Higgs presents it. Peter Boettke recently brought this aggregation debate into Romer's aggregate demand debate. In his comment section, after some debate on the question, Peter writes that he is "unpersuaded that aggregate concepts do much of anything to improve our understanding of an economic system".

The problem, as I see it, with this common "fallacious fallacy" is that it takes a kernel of truth and explodes it completely out of proportion, to the detriment of clear thinking. Of course haphazard aggregative thinking can lead to erroneous conclusions. But we already have a name for that: the ecological fallacy. Ecological fallacies are committed when you infer things about individuals based on the behavior of aggregates. That's an entirely valid thing to look out for, and it's something that macroeconomists who do trade in aggregates obviously have to take special care to watch out for. But Boettke and others, by taking the risk of committing an ecological fallacy to the extreme, themselves are at grave risk of a fallacy of composition. I would argue that the rejection of the paradox of thrift, the rejection of the prospect of a "general glut", or the rejection of the possibility of a wage-price spiral are all such fallacies of composition. If I were to commit a "fallacious fallacy" of my own, I would say that Boettke and others are committing a "disaggregation fallacy" (but I won't, because I have no methodological hang-ups about micro-data, which I personally use far more often than macro data).

Why do these fallacies persist? I'm not quite sure. A couple weeks ago I would have attributed it to the insularity of the Austrian school. Like Darwin's finches, eccentricities around something as mundane as an aggregate serve a function in the Austrian school (it's an easy to understand whipping boy for Keynesianism that is accessible to new initiates) that would not have evolved outside of that niche but can persist in that particular insular environment. Since transitioning from a few less illuminating Austrian blogs to more illuminating Austrian blogs, though, I have a much harder time making this argument. Then what is it? Why do smart guys with good things to say like Peter Boettke insist on peddling "fallacious fallacies" like the "aggregation fallacy"? Why is it so hard to see that there are two very real fallacies - the ecological fallacy and the fallacy of composition - that do apply in certain instances, but that you can't simply assume apply in all instances? I'm not sure - any ideas? I can't imagine how these people justify the co-existence of psychology and sociology, or organic chemistry and biology.

Quick Conclusion on UI, AD, and the GR
I want to point out that the empirical work I link to above shows that extended UI benefits have marginally increased unemployment (by 0.4 percentage points). We obviously don't have the data or the identification strategy to get an especially rigorous estimate, but that one seems to hold water with the Fed and the NBER. This suggests that Romer is right that UI benefits aren't a huge component of the story. But it also suggests that the economists who argue that the aggregate demand impact of the UI program in particular is a net positive are wrong. The argument goes that if you give unemployed workers more spending power it boosts aggregate demand, which helps the economy. A negative net impact of UI extensions - even a small negative impact - suggestst that any AD effect is swamped by the incentive effects. In other words, tentative, preliminary empirical evidence suggests that we're getting to the point where UI extensions may be justifiable on a humanitarian basis, but are getting harder and harder to justify on a stimulus basis. That does not mean that UI is keeping us in quasi-depressionary conditions. That doesn't even mean we shouldn't extend UI. And it certainly doesn't mean what Jerry O'Driscoll seemed to be implying: that you can't simultaneously say that AD and UI are important determinants of the unemployment rate, but that AD is more important than UI right now (as opposed to, say, the late 1990s when UI might have been a more important factor). What it means is we need to be careful about how we justify UI extensions, and seriously consider cutting off the extensions.

This has been fun. Future "fallacious fallacies" I might tackle are the Broken Window Fallacy, the accounting-identity-as-behavioral-law fallacy and reductio ad absurdum. Each of these, like the "fallacy of aggregation" has a kernel of truth that is habitually blown way out of proportion by zealous "gotcha" bloggers.


  1. In this vein, does anyone know anything about Kevin Hoover's "Causality in Macroeconomics" (http://www.amazon.com/Causality-Macroeconomics-Kevin-D-Hoover/dp/0521002885/ref=sr_1_1?ie=UTF8&s=books&qid=1271767466&sr=8-1). It's one of a shelf full of books I've been wanting to read for a long time but haven't had the opportunity to. It addresses these questions and looks quite good, but I'm not really familiar with the author.

  2. Pete Boettke's comments are mildly embarrassing to me, since I normally hold him in such high regard. Even a generous reading suggest an unhelpful concern with methodological purity.

    The situation is that aggregates (and averages) are deduced from sequences of elements, thus their logical content is always less than the sequence from which they are deduced. In other words, everything that can be validly deduced from the aggregate can be validly deduced from the sequence, but not vice versa. The consequence is a loss of empirical content, i.e. testable consequences, because aggregates are consistent with more possible sequences than the sequences from which they are deduced.

    The loss of empirical content, however, should not be mistaken for a complete absence of empirical content. Aggregates do limit possible sequences, because not all combinations of measurable consequences are consistent with them. The result is that, logically, aggregates imply true statements about microeconomic arrangements.

    Aggregates should be handled with care, and too often important properties of the sequence are hidden by excessive aggregation, but Boettke's comments seem quite disappointing and absurd to me.

  3. I think that's well stated, Lee. That's the logical side of it. The empirical counterpart, of course, is that aggregates are a lot easier to report, keep track of, and think about than the "possible sequences" of micro arrangements you speak of. So what do we do? Abandon aggregate work because it poses risks, or take advantage of the more accessible macro data and work with it carefully, always in dialogue with the micro data available? I'll opt for the later. The transaction costs involved are not negligible, and we need to take that into account.

    I think it's also worth noting - and I think a critical rationalist would agree - that just because the micro arrangements offer less logical dead ends in theory doesn't mean that irresponsible inference from these micro-arrangements won't introduce a variety of false conclusions. In other words, aggregates lose you empirical content, and not recognizing this risks committing an ecological fallacy. But making proper use of the empirical content of micro arrangements, while fine in theory, may be tough in practice, leading to a fallacy of composition. Does that make sense?

    I suppose I'm simply saying that after we talk about the logical consistency and the empirical content of these data, we also have to think about the mistakes that people will inevitably make in practice. And just because working from micro arrangments has less logical pitfalls, doesn't mean it has less practical pitfalls.

    btw - I appreciated your comments and Steve Horwitz's comments on the post. I'm not sure whether excess demand for money is the only thing that's going on, but it's an important part of it.

  4. I didn't mean to suggest that an excess demand for money (plus reserves) is the only problem; it is, however, the most pressing concern and most within the power of the authorities to remedy.


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