Testing the stimulus

A Keynesian model cannot be used to demonstrate the benefits of a Keynesian stimulus  

One frequently hears statements that the stimulus package led to a lower level of unemployment or that it contributed to a higher level of GDP growth than would have occurred had no stimulus package been introduced. The notion is that as bad as our economic circumstances are already, economic conditions would have been worse had there been no stimulus. 

Such statements cannot, however, be substantiated by those who devise such policies. Since it is the same Keynesian model that has been the basis for the stimulus package that is also the basis for the modelling undertaken to provide the counterfactual, the result is that all that has been shown is that the model predicts any policy based on its own structural design will be successful. 

The actual world of events has not unfolded in two modes, one with a stimulus package, the other without. We do not know in any realistic way what would have occurred had the stimulus package not been introduced, or what would have occurred had some other set of economic policies been applied. 

What is normally done in such circumstances is, to state the technique in its simplest terms, is this. The model is run in which a baseline estimate is created which examines what would happen if no policy changes were adopted. Then the model is run a second time with various policy changes introduced. One can then say, based on the model’s results, that the policy would have a particular set of effects. 

That is the way in which policy options are generally assessed. But it is only a genuine test of the policy if either the model actually does replicate the way in which the world operates or, and this is much weaker, if everyone who uses the model agrees on its structure and is willing to accept its conclusions. 

In the present case, however, critics of the Keynesian approach to modelling are very unlikely to accept conclusions from a Keynesian model. Almost all such forecasting models are built using Keynesian parameters in which a stimulus will with certainty provide positive results. 

Therefore, when it is argued the stimulus has had a positive effect on the economy irrespective of the deterioration in our actual economic data, what is being said is that their modelling has shown that against the baseline projections, the economy should have had a positive response and our actual economic results would have been worse than those published. 

But for all that, there is no outside form of authentication that can be applied. And while it can be argued that the parameters on which the model was built are based on the actual previous contours of the economy, the fact remains our present set of circumstances are so unique that no model can have had its underlying structure based on the situation we now find ourselves in. 

An important instance of such wrong forecasting is found in the United States, where  the initial modelling of the effects of the stimulus unquestionably overestimated its effects. The Obama Administration provided early on a comparison of the rate of unemployment that would be reached if no stimulus were applied (the baseline case) and then the outcome if a stimulus was applied. 

The Innocent Bystanders blog has been comparing the actual unemployment rate with the original baseline projections and the projections made by the Obama “economic team”. These are shown in the chart below.

The pale line shows the unemployment projections made by the Obama administration if no stimulus plan were introduced. The darker line shows the projected unemployment rate after the introduction of the stimulus plan. 

The set of dots shows the actual unemployment rate in the United States. What is clear is that the unemployment rate not only has not fallen, but is well above the rate projected to occur even had there been no stimulus at all. 

Two conclusions may be drawn. Firstly, such projections are merely the results of a set of models built according to the economic concepts of those who write the programs. In this instance, they are Keynesian models. They therefore cannot be used to test the validity of their own basic structure if the core question is whether that structure is valid in the first place. This is circular reasoning and logically invalid. 

The second point is the more important. The Obama Administration indicated in advance the positive effects its stimulus package would provide. The stimulus has shown itself not to have worked, at least so far as its original projections are concerned. The Obama Administration has since stated conditions had actually been worse than originally thought, but that is only an after-the-fact rationale. 

It is just as plausible to argue the stimulus has not only not worked but has made conditions worse than they would have been had no stimulus been applied.

For those who argued the stimulus would actually cause harm, this is a confirming experience, while for those who think things should have improved they must now find explanations for this obvious failures in forecasting the effects they think the stimulus should have had.

Leave a Reply