Using DSGE models to test/explain a hypothesis

The DSGE model can be used to test a hypothesis, right?? I mean not with data but in a theoretically consistent way. Maybe a better word is ‘explain’, not ‘test’. But here is an example:

A researcher formulates a hypothesis, ‘shock to trend growth causes large volatility in consumption’. And then uses the DSGE model to explain why that is plausible. In a sense, this is hypothesis testing, right? Leading to many different explanations for the same phenomenon, not necessarily conflicting with each other.

This is a rather deep philosophical matter. What do you mean with “test a hypothesis”? Following Popper, you can only reject a hypothesis, never actually confirm it. With a DSGE model, you can never conclusively rule out that a “shock to trend growth causes large volatility in consumption”. You can only present a model where you show that “shock to trend growth causes large volatility in consumption”. So in a sense you are rather formulating a hypothesis about a cause for business cycles that should later be tested.
That is the reason why people usually write down a model and then present tests of this model, usually in the hope that such tests do not contradict the model. But that is still not a confirmation of the model in a narrow sense.

1 Like

Many thanks, prof. Pfeifer, for the reply. I am reading on some phenomenon, and there are many papers formulating a hypothesis about its cause using the DSGE model. I kinda got confused with some papers using the phrase ‘we find that…’, while they are actually formulating a hypothesis. Perhaps a better phrase is ‘we show that…’. Some papers make this clear, others don’t. Thanks for the clarification.

Those papers should say that

We find in the context of our model that …

Whether you buy that depends on whether you find the model plausible in the first place.

1 Like