Perhaps out of my own laziness, stupidity or naivety, I had always felt that many papers with a large theoretical focus were a bit of a waste of time, particularly when these then go on to run some basic OLS regressions. Perhaps because I often find them incomprehensible, I saw them as an excuse for people to show off, or some perverse requirement of the discipline (or both). Actually this feeling had been reinforced by sitting through a presentation of a paper on labor market rigidities and trade by Elhanan Helpman at ULB here in Brussels ( I know I'm treading dangerous ground here - I don't even have the courage to go back and look at the paper actually) where every question remotely concerned with reality met with the response that it was "not in the model".

However, having read Acemoglu's paper now, I have a far clearer view of its use. And it's quite subtle. A point he makes early on in the paper regards external validity. You may actually end up estimating an OLS equation, but if this is simply based on throwing some variables in there to see what happens, you cannot vouch for its external validity, while if you arrive at the equation to be estimated through some theoretical basis you can say a lot more - even if the estimated equation is the same and the results the same. Well I thought it was subtle...

Still, going back to my own ignorance-induced disdain for theory and a more cynical attempt to validate my own PhD thesis, I was relieved to read the following in a footnote:

"Many empirical equations that do not correspond to structural relationships may nonetheless contain useful information; they just cannot be used for counterfactual policy analysis."Indeed...

"We might simply be interested in uncovering correlations, which may help us distinguish between theories, since many relevant theories will have implications about what these correlations should look like.And in any case:

"our confidence in the implied answers to policy experiments crucially depends on our confidence in having captured the appropriate structural relationship with the model we are estimating."

Ahaaaaa. And here I began to feel my cyncism coming back:

"the proper use of economic theory... requires that we ... develop the case that economic theory robustly leads to the estimating equation in question, and that we clarify which important economic mechanisms and effects are being excluded from the model"And that is the major challenge. As the big man says: "no single model can include all of the relevant factors for all possible counterfactual exercises." And this is the point. My experience suggests that we invariably all have different views and explanations for what is driving a certain economic system, taking us back to square one, where every presentation of any theoretical paper has (often legitimate) holes poked in it for missing some key issue. So we might as well just go with the reduced-form estimation in the first place.

Anyway....

PS - Apart from that discussion, he also highlights the need to always take into account general equilibrium effects and political economy. Now try and do that in a theoretical model...

## No comments:

## Post a Comment