Saturday, December 21, 2013

What is an Econometric Model? Objectivity vs. Reflexivity

In response to my recent post, titled, "The History of Econometrics - An Alternative View", Judea Pearl  sent me a thoughtful and intriguing comment. The comment is posted already, but I think that it deserves more than just being tucked away at the bottom of another post.

So, I am giving Judea's comment additional attention here. I hope that you'll find it interesting, and that it will provoke some much-needed discussion.

Here's Judea's comment in its entirety:

"Thanks for posting this interesting thesis of Annie Cot.

I agree with the author that the history of econometrics, when written by econometricians, suffers from the problem of "reflexivity", or lack of objective perspectives. I therefore ask permission to interject an outsider's perspective into this discussion and express my opinion, as an amateur philosopher, that the history of economics need be written after economists converge on a consensus regarding the aims of their endeavor and what methods advance them toward those aims. 

My readings into the current economic literature and my conversations with numerous economists told me that such convergence has yet to be achieved. I have documented my observations in the three articles below, one of which (2) was discussed on this blog.

2.       [See my earlier post,  here, DG]
As I argue in (1), there is an objective logic, and an objective language, within which one can characterize the aims and achievements of a field like econometrics. This logic has yet to be understood, let alone accepted, by economists and, until this happens, the history of economics will continue to suffer from "reflexivity".

Just to demonstrate how far we are from convergence, consider the question "What is an economic [or econometric, DG] model?". I would like to raise this question for discussion on this blog, and observe the diversity of opinions.

To start the discussion, the Wikipedia's definition of 'Econometric Models' reads (February 18, 2012):
"An econometric model specifies the statistical relationship that is believed to hold between the various economic quantities pertaining to particular economic phenomena under study.'' "
Thanks, Judea.

So, readers, what are your responses?

Oh yes - you should definitely check out Judea's own blog, "Causality Analysis in Theory and Practice".

© 2013, David E. Giles


  1. Hi: Maybe this is too simplistic but I think an econometric model is one that provides a way of quantitatively
    connecting a response to one or more predictor variables.

    I come from a statistics background ( but I happen to be using/working on an econometric model ) so to me, an econometric model is not different from a statistical model except in the fact that the variables usually happen to be related to economics. In fact, in the model I'm working on, the variables are not related to economics. I'm using an econometric model in a non-economics application so I would argue that econometric models are really just statistical models.

    Disclaimer: My economics background is extremely limited ( I took intro to micro and intro to macro ~
    30 years ago ) so if below is way off the mark for econometricians, I just think it adds to the discussion.

    I believe that "econometrics" is really just statistical modelling or time series modelling in disguise.


    1. About 50% of economists subscribe to the opinion that economic models are but a statistical models.
      Some venture to go as far as stating that it is a parsimonious way of encoding a set of statistical models, one for each instantiation of the parameters. But this cannot be the whole story, because
      this would not explain why economists labor so hard to IDENTIFY structural parameters. If prediction is all that the model gives us, why not do optimal predictions directly from the covariance matrix. Moreover,if prediction is all that we seek, how can we ever guide decision makers on policy questions?
      Thus, there is more to an econometric model than what the Wikipedia definition tells us.
      What is it?

  2. Judging from the citations in Pearl's articles to articles by Pearl, his analysis is about as reflexive as you can get.

    1. Dismissal by citation counting does not make my analysis reflexive.
      Please have a glance at the content and judge scientifically whether the logic
      proposed is reflexive or objective. Alternatively, how about citing
      an alternative to the logic proposed: What is an econometric model?

  3. This is quite interesting. In my opinion some key concepts are missing in thw Wikipedia definition, for instance, the relationship shall be a causal one. Also, it seems to imply that these "economic quantities" represent variables establised in an economic model, but I would like the terms "variables" and "economic model" to be spell out.

    1. Lara,
      I totally agree with you that causal relations is the key to econometric models, else they
      would be useless for policy making. Some go even further (e.g., Heckman) and claim that
      the model should also encode counterfactual relations (e.g., what the salary level would be
      of those trained, had they not been trained).
      But if this is what an econometric model supposed to do, how do we read causal effects
      or counterfactuals from the model?
      Or, to be more concrete, how many economic instructors did you know who can answer
      such questions? and how many econometric textbook do you know that tell readers how
      to answer such questions?
      So, where are we?
      If causes and counterfactuals are the heart and soul of econometric models, how can historians
      of econometric write history without attending to its heart and soul, namely the evolution of causal and counterfactual thinking in 20th century econometics.
      This is what triggered my reaction upon reading "The History of Econometrics - An Alternative View"