DGSE stands for Dynamic General Stochastic Equilibrium, basically it’s very complicated models that try to explain what’s going on in the economy. These models are nowadays widely used in policy relevant institutions such as the central banks. For several years, their reliability has however been put into question, notably for their lack of forecast of the last financial crisis. This article is the first of a series which will talk about the relevance of DGSE models in modern macroeconomics. As you will see, opinions with regards to DGSE models are extremely heterogeneous, I will thus try to summarize the different points of views and highlight the different arguments.
Here goes the first article about this topic;
Some years ago, Nobel Prize winner Paul Krugman announced that the field of macroeconomics went too far in mathematical modeling and that we should stop relying on such models and instead go back to more classical models such as the Keynesian ones. He then went on arguing that highly sophisticated models such as the famous DGSE were hence irrelevant ant shouldn’t be trusted. Even though I agree to some extent regarding the relevance of Keynesian models since they give a clear hindsight on the intuition behind the mechanisms of the economy without bearing any mathematical sophistication, I also believe that this remark doesn’t take into account all the complexity induced by the topic. Let me explain where I believe Krugman’s argument fails.
As we have seen, macroeconomists have been put under siege for their lack of predictions concerning the 2008 financials crisis. DGSE models being one of the most widely used models among academics and central banks, it naturally was the first model to be put under the fire of critics. Does DGSE deserve it? That is what we are going to investigate.
To answer this question, Wieland and Wolters have developed a tool that allows to run simulations using 50 different macroeconomics model that are supposed to be representative of the population of macroeconomics models. It is interesting to say that the tool includes models that Krugman talks about. They use their models in order to check whether the models can, first, correctly predict recession and economic slowdowns, and secondly whether these models can adapt to new parameters and predict economic recovery.
The graph is pretty much self-explanatory, but it cannot harm to emphasize what it shows. Basically what we can see on left graph (before the collapse of Lehman Brothers) is that all these models suck at predicting financial crises. But what matters here is to see that none of the models correctly predicted the 2008 financial downturn. At first sight, it therefore looks unfair to condemn DGSE models only. The second graph shows the predictions of the models, once the financial crisis started. All models seem to understand correctly the direction of the slope (except two outliers – if I were you, I might not rely on those to make economic predictions) but they tend to systematically underestimate the severeness of the recession. It looks like they have a clear tendency to be overoptimistic concerning the capacities of the economy to recover.
Under the light of these graphs, it appears that Keynesian models or DGSE models don’t do significantly better than one another. The only conclusion we can draw is that, if we are to condemn macroeconomics models, it is unfair to condemn DGSE models only. This feature could shed light on another question. After all, can macroeconomics models predict what is unpredictable? Should we expect from them to forecast financial bubbles? We will talk about these issues.
So, to answer the question ; do DGSE models suck? Well, it seems that they do suck at forecasting financial crises, but one shouldn’t forget that most models didn’t see it coming neither. Besides this lack of forecasting power, one can however give them some credit since they appear to be relatively reliable concerning economic recovery. More on this in upcoming articles.