Why all the hating on DSGEs?
Over at Greg Mankiw’s blog he links to a comment on DSGE models by the legendary Robert Solow. Surprisingly, the comment is relatively negative.
If I am honest, I found Solow’s attack on DSGE a little strange, and fairly inconsistent. Here is why:
Update: (Before saying why I think this, I should say) This initially came from an Arnold Kling post, and the lingo regarding biology has moved into some of Brad’s negative writing about economists he disagrees with (ht Economist View). My own view is that these authors are attacking a straw man when they attack DSGE’s (the straw man I’m talking about is the 1980’s RBC model, as I discuss later) – instead of attacking the inappropriate assumptions of some of the practitioners they have decided to attack the method. This disappoints me.
Yes – all models have weaknesses. But it is about seeing what to use and where. And contrary to Brad’s language he is being far from objective in his attack on this type of modeling.
Now, back to the post as it was a few days ago 😉
On unemployment and “the model”
In one paragraph he states that we have frictions such as rigid prices for goods and labour – then in the following paragraph he says that there is no such thing as unemployment in a DSGE model, which is obviously false from the first point.
However, this is merely a small point compared to what I see as his broad critique of DSGE’s, which is that they treat the economy like a couple of decision makers acting optimally, when the economy isn’t like that.
This is a strange criticism. A DSGE model does have few agents, often with infinitely long lives. However, these are simplifying assumptions – and many of the results can be generalised to heterogeneous agents who have finite lives (but over-lapping generations of people who value their descendants). There is definitely value in researching the impact of heterogeneity and the such – and that is exactly what macroresearchers appear to be doing – at least judging by the mailing lists I’m on for macro papers.
Furthermore, what is Solow’s alternative. Is he suggesting that me model each individual in society? I don’t see how that could provide tractable results, or even if it is remotely possible.
On the use of the model
What ultimately kicks me is the conclusion, which is absolutely at odds with what people use DSGE models for!
The point I am making is that the DSGE model has nothing useful to say about anti-recession policy because it has built into its essentially implausible assumptions the “conclusion” that there is nothing for macroeconomic policy to do
This is abjectly false. The scope for policy intervention relies on the assumption of “rigidities” which are implicitly proxies for the very market failures he is begging for from economic models. There are a swath of papers coming out of universities every year using DSGE modeling to justify all sorts of policy interventions, and justifying all sorts of responses to economic events.
The problem isn’t that the model always tells us to do nothing – the issue is trying to fish through the differing assumptions that are being made about the aggregate economy, and trying to find which assumptions we think are most appropriate for discussing the real economic situation.
You seem confused
I am a touch confused – as Solow was effectively the godfather of this form of modeling. His long-run growth theories were one of the major building blocks of what has become DSGE models. As a result, the man does know what he is talking about.
But his critique of DSGE models that
- They don’t allow for unemployment and
- They immediately dictate that there is no role for stabilisation policy,
Doesn’t seem to follow from the current set of DSGE modeling, but instead from critiques of old school Real Business Cycle theories from the 1980s.
Is there a lot of work still to be done on DSGE models to ensure that we have a clearer description of the economic situation, yes. Do practitioners need to be mindful of the limitations of DSGE models when forming policy, yes. But in no way does DSGE modeling act like all unemployment is voluntary and that there is no role for government.
Now, I am assuming that the critique is about the use of DSGE models to form policy and describe the economy – if the critique is actually about the predictive power of DSGE models I think the case is even weaker, as predictive weakness is an issue that holds for the entire discipline – and it is not a game killer.
Matt,
I suspect it is because DSGE models are too often used in the wrong place to think about problems they are not particularly suited for. It is true they are getting better, in the sense they are now incorporating aspects of the real world that until recently have been incorrectly ignored as irrelevant. But it still beggars belief that they are used to anaylse monetary policy issues when :
they don’t have adequate agent heterogeneity – even the overlapping generatins models don’t really capture the lifecycle aspects of real economies (eg retirement saving): See Clarida 1991 for a long ignored indictment of this problem;
they have a hugely simplistic approach to how firms really set prices;
they have a very simplistic approach to how banks and financial markets work;
they ignore almost all complexities of the tax system.
Personally, I struggle to think of interesting question in monetary economics that DSGE models are suited to answer. But I was taught that most of the interesting questions in monetary economics concern the ways financial markets organise themselves to deal with the twin problems of bounded rationality and opportunism (I’m dumb; you’re scum), be they bank runs, counterparty arrangements, or the provision of liquidity, and my education completely ignored the idea that monetary economics was all about using interest rates to manipulate output gaps and phillips curves to fine tune the economy. One gets the sense Solow also grew up in a tradition that thought the interesting questions in monetary economics concerned the way the price, quantity and terms and conditions of nominal debt contracts affected the economy, and that models that incorporate all these aspects of financial contracts are needed to be generally useful. Possibly the most prescient modelling work by a macroeconomist before the events of 2007 was done by Hyun Sung Shin – using models of banks that weren’t DSGE models.
But Solow is at least being consistent in his opposition. Do you recall his description of why he didn’t want to discuss macroeconomics with Lucas or Prescott using real business cycle models in the 1980s”
“Suppose someone sits down where you are sitting right now and announces to me that he is Napoleon Bonaparte. The last thing I want to do with him is to get involved in a technical discussion of cavalry tactics at the Battle of Austerlitz. If I do that, I’m getting tacitly drawn into the game that he is Napoleon Bonaparte.”
It all comes down to balance. DSGE models aren’t useful for everything, and often-times – particularly when considering a collapse of banking relationships – they aren’t useful at all.
The last few years have been one of those times.
Andrew
@Andrew Coleman
“I suspect it is because DSGE models are too often used in the wrong place to think about problems they are not particularly suited for”
I agree strongly with this statement.
“they don’t have adequate agent heterogeneity – even the overlapping generatins models don’t really capture the lifecycle aspects of real economies (eg retirement saving): See Clarida 1991 for a long ignored indictment of this problem;
they have a hugely simplistic approach to how firms really set prices;
they have a very simplistic approach to how banks and financial markets work;
they ignore almost all complexities of the tax system. ”
I agree those are all problems when it comes to implementing the model, my impression was that these issues had been noted – and central banks tend to fiddle their results from DSGE models to represent their views on these issues. It isn’t an ideal situation for sure – but I see the DSGE approach as something that is still being created rather than a final product.
“Possibly the most prescient modelling work by a macroeconomist before the events of 2007 was done by Hyun Sung Shin – using models of banks that weren’t DSGE models”
Interesting. I had not heard of this, but will look into it.
“But Solow is at least being consistent in his opposition. Do you recall his description of why he didn’t want to discuss macroeconomics with Lucas or Prescott using real business cycle models in the 1980s”
Indeed I do. However, I had the impression that he was in the current “New Keynesian” camp – when this really says that he is not.
Although, I guess the New Keynesians of the 1980’s were a very different breed to the modern New Keynesian – guess they need a new name.
“It all comes down to balance. DSGE models aren’t useful for everything, and often-times – particularly when considering a collapse of banking relationships – they aren’t useful at all.
The last few years have been one of those times.”
I agree with this also – by construction a DSGE model isn’t very good at handling large shocks and non-linear dynamics, which have to be the centrepoint of such events.
Personally, I have a number of concerns with the use of DSGE models, and I can definitely see how an over-reliance of “one model” when trying to understand the macroeconomy is dangerous. However, I felt that some of the claims Solow made here went too far – and were just too inaccurate, even with the understanding that he was explaining these issues to an audience of intelligent lay people.