The “devalue and dismiss” fallacy, methodological pluralism, and DSGE models

One of the most common fallacies in the economics blogosphere — and elsewhere — is what I call “devalue and dismiss.”  That is, a writer will come up with some critique of another argument, let us call that argument X, and then dismiss that argument altogether.  Afterwards, the thought processes of the dismisser run unencumbered by any consideration of X, which after all is what dismissal means.  Sometimes “X” will be a person or a source rather than an argument, of course.

The “devalue” part of this chain may well be justified.  But it should lead to “devalue and downgrade,” rather than “devalue and dismiss.”

“Devalue and dismiss” is much easier of course, because there then will be fewer constraints on what one can believe and with what level of certainty.  “Devalue and downgrade” keeps a lot of balls in the air and that can be tiresome and also unsatisfying, especially for those of us trained to look for neat, intuitive explanations.

Enter DSGE models.  There are plenty of good arguments against them.  Still, they provide a useful discipline and they pinpoint rather ruthlessly what it is they we still do not understand.  We can and should devalue them in a variety of ways, and for a variety of reasons, but still we should not dismiss them.  Better yet than “devalue and downgrade” might be “devalue, downgrade, and…yet…de-dogmatize,” because these models usually point out the limits of our understanding.  Those models defeat us, and thus it is odd when we attempt to portray the situation as us defeating them.

Note that very smart people are often good at “devalue and dismiss” because they can come up with a lot of good reasons to devalue the arguments or frameworks of others.  But still they should not leap so quickly to the “dismiss.”

I would mention that Alex, while he did criticize DSGE models yesterday, also appreciates their uses.

Addendum: Here is Chris House, defending DSGE models.


DSGEs try to understand how the economy actually works based on fundamental principles like utility maximization and profit maximization in the presence of shocks. If one doesn’t believe that firms and individuals try to do that to an order of approximation, then please take all that stuff out of the microeconomics texts!! There won’t be much left.

The defense by Chris House fails it point. It is only in the limit of a frictionless DGSE model that there is no forecasting power for fundamentals such as output and inflation. ANd in this limit the DGSE model is useless for most policy applications anyway (i.e. we immediately have Ricardian equivalence, No real effect of monetay policy etc)

And of course criticizing DSGE does not mean criticizing ay model which is stochastic, dynamic and general equilibrium. Just those sort of models that exist out there for analyzing economic policy

So is it ever justified to outright dismiss an argument / person? Or are we stuck in a perennial succession of downgrades?

1. Flat Earthers?
2. Intelligent designers that deny Darwinism (not 100% dismissed! but no rabbit bones from the Pre-Cambrian yet.)

I don't dismiss Keynesianism because of its crudeness, but I downgrade it more than crude Homo Economicus. And while I love Kahneman, the whole 'behavioral revolution' in economics carries a strong itch to dismiss Homo Economicus, tho this is still the standard and most useful model in all of the social sciences.

In other words, DSGE is (sorta) to Keynesianism as 'behavioral economics' is to traditional micro, so let's be consistent people.

A sufficiently downgraded view is essentially indistinguishable from dismissal with regard do decision-making. That is, whether you see a view as having 0% likelihood or 0.001% likelihood, you're not going to make decisions based on it. But the difference is that dismissal prevents you from considering new evidence that arises; you're so extremely biased against the possibility against a view that it's impossible to rationally adjust.

So to answer your question: I'd argue it's never justified to completely dismiss an idea, no matter how outlandish, but that extreme downgrade is a more useful approach anyway.

Reminds me of Cromwell's rule. .....leave a little probability for the moon being made of green cheese; it can be as small as 1 in a million, but have it there since otherwise an army of astronauts returning with samples of the said cheese will leave you unmoved.-Dennis Lindley

The periodic recurrence of this debate in the blogs amuses me. Some of the participants (e.g. Yglesias) appear to have limited experience with DSGE, so it's interesting that they feel confident enough to opine on the subject. Some of them demonstrate a total lack of familiarity with the massive task of trying to model an economy, as when Matthew Klein tweeted "wouldn't a model that genuinely reflects how world works predict better than other models?" So we have journalists who believe a totally realistic model is attainable. If that is their belief, they have nothing useful to contribute to this particular debate.

DSGE models provide a sort of discipline that is hard to achieve in other models, even harder to achieve in serious narrative/heuristic theorizing, and almost totally absent from the blogs. We have a few people operating under the delusion that their approach to econ is superior to formal DSGE because they use fewer silly assumptions. That's the great thing about narrative econ: your assumptions are easily hidden from others and, often, yourself. It's all part of the illusion that macroeconomics is simple, an illusion that some bloggers implicitly endorse wholeheartedly even though it does their readers a great disservice.

In any case, more thoughts here:

I don't think anyone denies that DSGE models provide discipline. The important question is whether they deserve to be called science. A model that makes seemingly ridiculous assumptions can be valuable if it makes useful predictions. The criticism of DSGE models is that they make unreasonable assumptions about the economy and also fail to make useful predictions. That they are nevertheless a mainstream tool in macroeconomics means that macroeconomists should be extremely humble about what they claim to know. What I don't think is helpful is to dismiss criticism from people outside the field on the grounds that they lack credentials or years of experience in the field. It doesn't take advanced math skills to read Karl Popper and then ask whether DSGE models -- as judged by their ability to make "risky" predictions -- do not somewhat resemble his definition of pseudoscience.

I agree with much of that, Ricardo, and I'm all for economists being humble (see One way to demonstrate that humility is to make assumptions and proposed mechanisms totally transparent, as formal modeling forces us to do. But, of course, we should also be humble about the results we get from the models and our studies generally. It's important to be aware of the limits of our knowledge.

I definitely don't think outsiders have nothing useful to say, and I don't mean to dismiss all criticism from outside. I do take issue with the ones who make criticisms of DSGE that apply equally to every other form of economic inference (like the Klein comment I quoted) or make criticisms of DSGE that don't even apply to DSGE (see here and my comment on it). I do find it puzzling that pundits who've never worked with these models feel impelled to talk about them, but if their criticisms acknowledge the difficulty of economic inference generally (and consequent importance of assumption and simplification) and accurately describe the aspects of DSGE that they don't like, I think we should take them seriously.

Finally, on the topic of humility for economists, can we also say that this mandate applies to econ commentators generally? In my anecdotal experience, economists are far more aware of the limits of our economic understanding than are the pundits who daily tell us they can explain this or that economic phenomenon "in one chart!"

Isn't the value of DSGE the very fact that they are extraordinarily complex and inaccurate, and pretty useless for forecasting?

Back to the meteorology illustration. If a hurricane is approaching the coast, the meteorologists can quantify how uncertain they are of the exact landfall. The closer it gets the more certain they are as variables get removed from the models. It isn't a matter of better science; it is a limitation of reality.

And yet the jackasses at the CBO make predictions 10 years out.

A couple anecdotes. In my business over they years there have been instances where the phone stops ringing for short periods, 3 weeks usually. Not seasonal, just everything seems to stop. We check the phones, no they are ok. I talk to suppliers and others in the industry and they experience the exact same thing. Then it starts again. Two months later the news will have a blurb on how the economy slowed during that period. In retrospect the story can be constructed.

A large retailer in Canada when they order home air conditioners from a manufacturer have an arrangement that the manufacturer takes back unsold stock at the end of the year. The risk is shared, and the full attention of both sides is put to coming up with a number. In 2009 the excess unsold were dumped on the market and I got some screaming deals. No one saw it coming. A good DSGE model may have put a probability on such a thing happening, rather than sales trends from past years as a guide, but ultimately it was someone taking a risk and hoping it paid off.

Long-run predictions tend to be easier to make than short-run ones. This is the result of the Law of Large Numbers.

There are many applications of this: the efficient markets hypothesis, the fact that climate forecasts tend to be more accurate than weather forecasts, and the fact that we can confidently predict there will be GDP growth in the long-run even if we can't predict when the next recession will be. The fact that CBO makes predictions 10 years out but doesn't call recessions is a good sign, not a sign that they are "jackasses" as you so eloquently put it.

Most of those aren't predictions as much as 'tomorrow will be the same as yesterday'. That usually is the case until it isn't.

As for the CBO, you should look at their predictions and how things actually turned out.

Tyler: I agree.

But note that there is the Bayesian prior to consider. If you started out thinking DSGEs are awesome, then critiques should merely temper your enthusiasm. But there are people who started out thinking that DSGEs are worthless; these people dismissed DSGEs even before reading the critiques.

So when we downgrade, there's the question of where the downgrade began, and where it ends up!

To beg the question, how do the doubters get to the belief that DSGEs are worthless in the firstplace, if not through a previous devalue and dismiss.

This reminds me of a quip by my teacher Paul Grice. I was sitting on the steps of the college's Philosophy Building ( I was very good at this ) reading F.H. Bradley ( Because of T.S. Eliot ) when Prof. Grice ambled up the steps, looked at what I was reading, and said something like , "Good Lord, .Bradley. Well, everyone comes back around sometime."

When models are used in disingenuous ways to support belief systems that are essentially ideological, it is fair to dismiss the proponents of the model who use them in this context. Dismissing a model outight is silly, as it depends on the application, whether it is being used appropriately, etc.

The noise in macro-economics these days isn't about the models per se, it's about how they are being hijacked for ideological ends.

I'm curious that this blog doesn't seem to want to entertain the notion that the argument over the models is just a subtext for an ideological battle....

The current debate as near as I can tell. DSGE works just fine when the economy is going down hill, because everyone knows we are going downhill. It fails during growth when knowledge is asymmetric.

DSGE stands for "dynamic stochastic general equilibrium."

A model is not true or false, it is useful or not useful depending on the problem/question.

With DSGE the question is whether it is the best use of so many resources.

Does this mean we should devalue and downgrade the devalue and dismiss fallacy itself?

DSGE models definitely have their place, and are used in the financial sector. And models which are (in all but name) DSGE are used widely, specifically in the prediction of uptake of new technology.

Noah Smith's critique ultimately proves too much. If we apply the market test to statistical physics, it too fails. Doesn't mean it's not a useful and correct theory.

Sorry, Chris Stucchio, nobody is using DSGE models for anything in the private sector as you claim they are. Your weasel wording of "in all but name" is presumably what is letting you think that you have actually said something accurate or even useful, but not, you have not. You are making the mistake of calling a model of supply and demand that may have more than one market in it as DSBE "in all but name," but not close. I suggest you name one company out there that is using real DSGE models, just one.

While it is easy to think this is abou ideology, it is not. The old classical vs Keynesian battle simply moved to be inside the DSGE framewokr some time ago, as in the RBC versus New Keynesian versions. This is what led people like Olivier Blanchard over a decade ago to declare that unity had been found in macro, although that is clearly not the case.

Also, saying that one should throw out macro models that do not assume that people max U and profits is silly. There are aggregations problems, even in micro when we move from those max U and profits individuals up to individual market supply and demand curves. If one wants to model a particular market, one estimates market supply and demand from market data. One may impute some cost function or U function from that, but one does not go out and estimate individual U functions or individual firm cost curves and then aggregate up to the market. That is not how it is done.

Also, it is an open secret that real micro general equilibrium theorists consider DSGE to be pathetic joke. That macroeconomists nod at them and invoke them to defend DSGE is an even bigger joke.

What makes this whole discussion difficult and messy is that even as policymakers and private consultants disabuse DSGE models, they completly domiinate the top ranks of macro in terms of journal publishing. I know for a fact that while quite a few people have tried, no alternatives to DSGE have been alliowed into print in the top 4 journals in recent years. Now one can say, "Ah ha! That proves that those papers are no good! We shall take the alternatives seriously when they can pass muster at the top 4 journals!" But this starts to become a circle jerk closed shop story after awhile. Everybody knows that the emperor has no clothes except for those working in the emperor's court, and they continue to control the official media about what the emperor is wearing. This leads to a lot of heated denunciations and carryings on by many involved.

Thanmks for finally ttalking about >The

Barkeley Rosser,
So, by these "alternatives to DSGE," I am sure you mean Agent Based Models (ABM). You are essentially claiming that most top macroeconomists are in a self-preservation conspiracy to keep ABMs out of top journals. This is plain JFK-grassy-knoll conspiracy paranoia!
I think top macro faculty have looked at ABM papers and see them for what they are: black box models with arbitrary behaviors imputed to some agents that have no economics in them.
The best models that we have, DSGEs, have won the day, until you can beat them, deal with it.

Sorry, Euripides, but the reason there is this argument is precisely that a lot of us do not accept this judgment, particularly in light of the uselessness and awful performance of the DSGEs, despite the increasingly bizarre set of bells and whistles being added to them. The front room people at the Fed and other places are thoroughly disgusted with what they are being given, but the back rooms will not hire people doing ABMS.

"Arbitrary behaviors." Really? But assuming that the economy looks like an optimization based on some assumed utility function of a homogeneous agent calibrated to fit past data is equally arbitrary. It is clearly not what is really going on, which involved micro interactions of the heterogeneous agents. Yes, I am fully aware that all kinds of pompous loudmouths go on about how DSGEs now use heterogeneous agents and how this is even a contrast to ABMs, but they way they do so is to assume an interval that may have some variation over risk or initial wealth, but then in effect acts as a large homogeneous agent, as an interval, which may then change shape as the model proceeds.

A virtue of ABMs is that in fact the behaviors of the individual agents are clearly specified and may involve interactions with each other in ways that DSGE models do not allow for. The criticism is that they do not fit this established paradigm of DSGE. The papers get rejected because "there is no equilibrium in this model." This becomes theology, and ABM is heresy, not an approach rejected on scientific grounds. BTW, I have now posted on this on Econospeak.

BTW, it has been my observation for quite some time now that the people who most loudly devalue and dismiss ABMs do so anonymously. That includes you, "Euripides."

The problem with the devalue and dismiss metaphor is the question if there's anything to devalue. Is there any value there to begin with and where's the evidence for it? As always with those things, burden of proof rests on the proponent. Why should anyone take those models seriously?

Comments for this post are closed