I started to study about macroeconomic modeling and I am interested in the following questions. I understand that there are no definite answers to these questions. But I am interested in the opinion of local experts.

Is there a future for DSGE models after the 2007 crisis?

Will DSGE models be replaced by agent-based models?

Does it make sense to study and use in practice the econometric systems of (simultaneous) structural equations after Lucas critique in today's realities?

I would appreciate a reasoned answer and a constructive discussion.

  • $\begingroup$ With regards to the econometrics tags, how does macroeconomic modelling use statistics? $\endgroup$
    – ahorn
    Dec 27, 2019 at 9:30
  • $\begingroup$ @ahorn The topic of structural econometric models (SEM) and Lucas critique is directly related to econometrics $\endgroup$
    – Vitalii
    Dec 27, 2019 at 18:06
  • $\begingroup$ Vitalii, unfortunately, that fails to answer my question. $\endgroup$
    – ahorn
    Dec 27, 2019 at 18:14
  • $\begingroup$ @ahorn Okay, I'll try one more time. When studying simultaneous econometric models in textbooks, an example is the construction of simple models that are used to measure quantitative relationships at the macro level. For example, the relationship between the volume of gross national product and productive resources: fixed assets, labor, material resources. This model should include the Cobb-Douglas function. Also, this type of models is called macroeconometric models. See the literature on this topic to find out more. $\endgroup$
    – Vitalii
    Dec 27, 2019 at 20:23

1 Answer 1

  1. Is there future for DSGE?

Well, DSGE stands for dynamic stochastic general equilibrium models. You can’t really model time series appropriately without the dynamic stochastic part, and when it comes to general equilibrium there are economists who think that this approach does not really capture all interesting questions, as in many cases it might be more interesting to see what happens when economy is in transition between equilibria. But I don’t think in mainstream macro DSGE will be gone any time soon.

However, this does not mean that the DSGE models are not being improved or changed. There is a lot of research trying to include more heterogeneity, for example in agents expectations (see this paper) or imperfect information (see here) and I recall seeing at a conference some models that try to add behavioral micro foundations to the DSGE.

I don’t want to make long term forecasts because they are inherently inaccurate- but as I see it DSGE will be here at least for next 10 years and I would not be surprised if it would be longer.

  1. Will DSGE be replaced by agent based models?

Well this is hard to say (assuming you are talking about agent-based computational (ACE) economics - since there can be agents in DSGE as well). So far the research on AEC is mostly very recent so it’s hard to judge the literature. It’s definitely possible they would eventually replace DSGE but on the other hand AEC has its own methodological problems (see here - although also note that the paper is mainly criticism of DSGE but it contains also some useful info on potential pitfalls of AEC). It’s also not impossible that both approaches will be used concurrently in the future.

  1. Does it make sense to use simultaneous equations?

Yes, it does not just make sense it’s a necessity if you make any analytical model of macroeconomic activity. You need to use simultaneous equations if your system is endogenous to properly describe it and in macroeconomics most variables are well known to be endogenous.

I also don’t see how would this be affected by “today’s realities”. It’s not like let’s say income was endogenous in 90s but stopped being endogenous in 2007 or 2019.

Even ACE tries to model endogenous interactions, so there is implicit simultaneity in them. So that’s not even an issue here. The only reason why ACE models are not technically simultaneous equations models is that most of them don’t include any explicit analytical expression but are performed via Monte Carlo simulations or some set of if then statements or some other computational approach.

You could only say that simultaneous equations don’t make sense in macroeconomics only if you want to claim that either macroeconomics systems are not endogenous or that macroeconomic systems can’t be described by analytical expressions (or both). I think you would have serious trouble finding credible economist that would get behind any of these. I think even most ardent AEC proponents would not dispute that theoretical macro models based on simultaneous equations are useful to get intuition or “deeper knowledge” behind macroeconomic variables and relationships between them.

  • $\begingroup$ I am very grateful for your comprehensive answer. Perhaps I did not ask my last question correctly. By a system of simultaneous equations I mean econometric systems of structural equations. As far as I understand, there is strong criticism regarding such models (for example, criticism of Lucas). But on the other hand, such models solve the forecasting problem well. $\endgroup$
    – Vitalii
    Dec 22, 2019 at 15:14
  • $\begingroup$ Perhaps there are ways to solve the problems described in criticism. For example, by using dummy variables. Or is this a bad idea? And I wonder if it makes sense to investigate such models in modern times? $\endgroup$
    – Vitalii
    Dec 22, 2019 at 15:14
  • 1
    $\begingroup$ @Vitalіі well even outside theory if you have endogenous variables and you use any equation based modeling then you have to use simultaneous equations. However, you could find some people saying that the whole equation part is wrong because macroeconomic systems are too complex to be even described by parametric models to begin with but it’s not really widely accepted idea - at least not now. Using dummy models does not make the model not simultaneous equation model as that’s still in the equation/parametric modeling family $\endgroup$
    – 1muflon1
    Dec 22, 2019 at 15:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.