CGE (computable general equilibrium) models are nowadays achieving a huge improvement in terms of dynamic vs static; multiregional vs single region; many industries vs just a few; many final demand accounts vs just a few, etc.
What do you believe was the driver for CGE modelling enhancement in recent decades?
1. Software availability (GAMS, GTAP, Gempack; Matlab, R, other);
2. Hardware advances (faster computers; memory sizes etc);
3. Data availability (electronic data, detailed data, open data, other);
4. Theory (techniques, approaches, etc);
5. Other...
Without question, CGE models have great appeal: they look complex, require a certain level of mathematical knowledge, there are good software packages nowadays and all that you mention in your discussion post. But the certain mathematical elegance and the supposedly nice results should not tempt us to take them very seriously. Because the real world is not in a state of general equilibrium, but is a chaos of billions and millions of imbalances. Even though thousands of economists are messing around with CGE - what we need instead are disequilibrium models. But they are much more difficult to model and even more difficult to solve. So many CGE approaches simply remain what is often found in empirical economics: GIGO - garbage in, garbage out.
simply great answer Hans-Georg Petersen .
I would say they give some feelings... though, I prefer econometrics...
This article in the Journal of Economic Perspectives, which must be understandable, summarises the recent progress in the area:
Article On DSGE Models
Thanks José-Ignacio Antón , for sure the DSGE models are an interesting progress which I believe will go even further.
Econometrics is not an economic model, but a statistical method to test hypotheses, which are parts of or derived from economic models. DSGE models are CGE models, too. Therefore, Peterson's critique does also apply to them. One of the main drivers for CGE was that Neoclassical hardliners (e.g. Hayek) wanted to rewin their position against Keynesianism. New Keynesian (NK) models are, of course, not Keynesian, but New Classical models with some slightly different assumptions. GIGO is the right short description. You can read my short article on a simple NK model ("Nonsense Economics").
This is a combination if all together. You cannot single out one factor.
However you did not mention data quality.
The more complex and detailed such models become the more vulnerable they are and their propagations in calculations.
Georg Erber, for sure this is all together. I decided to initiate this discussion because a friend was emphasizing too much on hardwares and I believe that data quality and open data has a lot to do with the advance in CGE. Anton Rainer is also correct when we should look the DSGE models to have more information. I do not agree to look at only one approach, but I also recognize that is very hard for the researcher to gather econometrics with CGE. But academics are improving skills and we're having better works.
Hello all,
As revealed by Professors Hans-Georg Petersen and Anton Rainer , the big challenge of these models is about the the practical utility of its results; can we really develop real economic policies (and decisions) based on the results of these models?
I believe that complexity theory has not yet been sufficiently acknowledged as an instrument of economic modelling. I am particularly attracted to Jim Rickard's book:
https://www.amazon.co.uk/Road-Ruin-Global-Elites-Financial/dp/1591848083
I would welcome feedback on the subjects he covers, especially the new, but controversial 'behavioural economics'.
It's only an IMF push for finding growth rates. These models are a result of great push for more complex blackbox type models.