Following a link e-mailed by Ken Houghton I read this at “the Ambrosini critique” (thanks Ken)
The need for more math is also related to the increase in the empirical relevance of theory. I’m convinced the only standing legacy of the Real Business Cycle literature, besides method, is its insistence on bringing the models to the data. In modern macro, its simply not enough to identify the existence of some effect or other. For example, real business cycles were relevant because they proved to be quantitatively important… a large chuck of business cycle fluctuations are driven by supply shocks. And RBCs have been supplanted because they didn’t explain enough of the data. The empirical relevance of real shocks couldn’t have been tested without out explicit mathematical models of the phenomenon.
This is what frustrates me about Kling, Krugman, et al’s ad-hoc theorizing. They seem contented to identify that certain macroeconomic features exist, but they don’t bother to quantify the importance of those features.
and also a comment
January 29th, 2009 at 3:03 pm
To be honest, yours is an interesting request. How does one show a discipline is empirically relevant?
What sort of evidence would convince you?
(I am not the referent of the pronoun “you” in the comment).
I comment after the jump
Now Krugman does appeal to quantitative models. For example he notes the CBO predictions. Now you might not consider the CBO model to be an economic model, because I suppose it lacks micro foundations. The fact is that Krugman makes quantitavie calculations. To you they don’t count, because the theoretical argument is just that a causal effect isn’t zero, then it is estimated reduced form. Does this approach yield worse predictions than those generated by DSGE models ? I am familiar with the Lucas critique. So was Marshak who stated it long before Lucas (who made “no claim of originality” in “Econometric Policy Evaluation: A Critique and cited Marshak). However, Marshak also attempted to forecast with models which he knew were vulnerable to the Lucas critique.
He thought that was the best approach available. Since Marshak we have accumulated a good bit of data on forecasts and outcomes. Was he wrong ?
Or to put it another way, does anyone whose employment depends on getting macroeconomic predictions right use DSGE models ? If not what is the market failure ?
Pushmedia it is easy to show that a discipline is empirically relevant. The acid test is out of sample forecasting. Models can be tweaked to fit the past. A more empirically relevant model gives better out of sample forecasts.
In particular, DSGE are an empirical advance if and only if out of sample forecasts based on DSGE models are better than those based on, say, VARs, old Keynesian macro models or something else without micro foundations.
Now you will notice that the methodology described by Ambrosini is based on the ability to match summary statistics.
OK so I took a model which I am absolutely sure has nothing to do with reality and tried to see how blatantly I had to cheat to get it to match summary statistics. My sense was that it was about average plausible for DSGE (had spillovers, the labor market cleared, no exogenous technology shocks). I conclude that the approach is not fruitful.
If you are interested, I can send you the paper presenting the model and the gauss file which does the simulations. If you can convince me I fudged more than the average macro theorist, I will thank you. If you can convince me that the model and program have any scientific value at all, I will be very very grateful.
Consider astronomy. It is like macro in that it is non experimental. It is unlike macro in that astronemers can predict, for example, where planets will be in the sky. Astronomers are clearly empirically successful, they can predict things that you see in the sky better than you and I.
Can we predict what will happen to the economy better than astronomers ?