Fiscal multipliers measure the effect of changes in government fiscal policy on GDP. They aim to capture the behavior of a multitude of individual housholds and firms in a single number. As such, it should come as no surprise that the ultimate fiscal multiplier does not exist. Consumers and firms react differently to different policy measures. Consequently, different types of policy measures have different multipliers. Consumers and firms behave differently under different economic circumstances. Multipliers may therefore depend on these circumstances.
More specifically, one may conjecture that in an open economy part of the fiscal stimulus may just increase imports. Or that government stimulus will have a larger effect if private agents face (financial) constraints. And that multiplier eﬀects may be larger when the central bank accommodates ﬁscal policy or is bound by a zero interest rate.
And this is indeed what empirical research seems to show. A working paper by Sebastian Gechert and Henner Will looks at 89 studies on fiscal multipliers. The graph below shows the frequency distribution of multipliers found in studies using different types of models ( in technical terms – New Classical RBC (or D(S)GE) models, New Keynesian DSGE models, structural macro-econometric models, VAR models, and all kinds of single equation estimation techniques. Check the paper if you want to know more). The graph shows substantial variation between these different methods.
This has two consequences. First, it matters which model economists use to assess the potential benefits of additional government spending. If a model doesn’t include the possibility of monetary policy being at the zero lower bound, or does not incorporate agents that are – for example – credit constrained, it will underestimate the effect of additional government spending. Second, the choice for a particular model that economists make may itself become the subject of a political discussion. If commentators don’t like the outcome of a particular analysis, they may well argue that this is due to the subjective choice for a particular model of the economist in question.