Blanchard Joins Calls For Structural Econometric Models To Live On Brought Inward From The Cold

Mainly for economists

Ever since I started blogging I possess got written posts on macroeconomic methodology. One objective was to endeavor too convince immature homo macroeconomists that Structural Econometric Models (SEMs), amongst their advertising hoc blend of theory too information fitting, were non some onetime fashioned dinosaur, just a perfectly feasible agency to practise macroeconomics too macroeconomic policy. I wrote this amongst the sense of having built too published papers amongst both SEMs too DSGE models.

Olivier Blanchard’s tertiary post on DSGE models does precisely the same thing. The alone slight confusion is that he calls them ‘policy models’, just when he writes

“Models inwards this course of report should fit the primary characteristics of the data, including dynamics, too let for policy analysis too counterfactuals.”

he tin alone hateful SEMs. [1] I prefer SEMs to policy models because SEMs depict what is inwards the tin: structural because they utilise lots of theory, just econometric because they endeavor too fit the data.

In a tweet, Noah Smith says he is puzzled. “What else is the betoken of DSGEs??” likewise advising policy he asks? This postal service tries to assistance him too others come across how the 2 classes of model tin travel together.

The agency I would approximate a SEM today (but non necessarily the alone valid way) would endure to starting fourth dimension amongst an elaborate DSGE model. But rather than approximate this model using Bayesian methods, I would purpose it equally a theoretical template amongst which to starting fourth dimension econometric work, either on an equation past times equation footing or equally a laid of sub-systems. Where lag structures or cross equation restrictions were clearly rejected past times the data, I would alter the model to to a greater extent than closely fit the data. If some variables had strong ability inwards explaining others just were non inwards the DSGE specification, just I could intend of reasons for a causal human relationship (i.e. why the DSGE specification was inadequate), I would include them inwards the model. That would larn the SEM. [2]

If that sounds terribly advertising hoc to you, that is right. SEMs are an eclectic mix of theory too data. But SEMs volition withal endure useful to academics too policymakers who desire to travel amongst a model that is reasonably roughly the data. What those I telephone outcry upwards DSGE purists possess got to acknowledge is that because DSGE models practise non fit the information inwards many respects, they are misspecified too so whatsoever policy advice from them is invalid. The fact that y'all tin endure certain they satisfy the Lucas critique is not sufficient compensation for this misspecification.

By setting the human relationship betwixt a DSGE too a SEM inwards the agency I have, it makes it clear why both types of model volition proceed to endure used, too how SEMs tin possess got their theoretical Pb from DSGE models. SEMs are also useful for DSGE model evolution because their departures from DSGEs furnish a whole listing of potential puzzles for DSGE theorists to investigate. Maybe i solar daytime DSGE volition larn too so proficient at matching the information that nosotros no longer demand SEMs, just nosotros are a long agency from that.

Will what Blanchard too I telephone outcry upwards for happen? It already does to a large extent at the Fed: equally Blanchard says what is effectively their primary model is a SEM. The Bank of England uses a DSGE model, too the MPC would larn to a greater extent than useful advice from its staff if this was replaced past times a SEM. The existent work is amongst academics, too inwards detail (as Blanchard i time to a greater extent than identified inwards an before post) magazine editors. Of course of report around academics volition croak on using DSGE, too I possess got no work amongst that. But the few who practise instead determine to purpose a SEM should non endure automatically closed out from the pages of the transcend journals. They would endure at present, too I’m non confident - fifty-fifty amongst Blanchard’s intervention - that this is going to alter anytime soon.

[1] What Ray Fair, longtime builder too user of his ain SEM, calls Cowles Commission models.

[2] Something similar this could possess got happened when the Bank of England built BEQM, a model I was consultant on. Instead the Bank chose a core/periphery construction which was interesting, just ultimately too complex fifty-fifty for the economists at the Bank.