Thursday, August 18, 2011

Public consumption and the business cycle

One aspect of government purchases the current crisis has highlighted is how volatile they can be. Quite obviously, they are influenced by politics, to the point of complete reversal between massive spending and severe belt-tightening within months as in the US and the UK. But there could also be a more systematic component that is linked to the business cycle. After all, the government may be trying to improve the welfare of its constituents and for example substitute public consumption for lacking private consumption, or the same for investment.

Ruediger Bachmann and Jinhui Bai look at this using an augmented real business cycle model. They claim that 25-40% of the variance of public consumption can be accounted for by shocks to total factor productivity once implementation lags and costs of public consumption, as well as taste shocks to public vs. private consumption. I am no particular fan of taste shocks, as they are the symptoms of a modeler who is giving up on trying to explain something and simply equates the error term in the Euler equation to a shock. Then much is driven by how this shock is calibrated, in this case to match a four year electoral cycle and some data moments. When I think about shocks in this context, I think indeed about who is in power to decide on public expenditures. But that is not completely exogenous. Indeed, the state of the economy has an impact on who gets elected or reelected. And this can be calibrated without trying to match the data moments one is trying to explain.

1 comment:

Anonymous said...

Thanks to the Economic Logician for discussing our work.

But he is also somewhat misrepresenting the nature of the exercise (I will come to that).

First, let's not forget the state of the quantitative macro-PF business cycle literature, which is really in its infancy. All we did was to see how far you can get with a minimally altered standard model to explain the cyclical dynamics of public policy variables (not many have done that). The answer we gave is: somewhat, but not too far. You need frictions and, potentially, other shocks that are at least somewhat orthogonal to the shocks that hit the economic subsystem (it's true, as a benchmark, in the paper we made them completely orthogonal). You can call them taste shocks, you can call them power shocks (i.e. to the identity who is in power), they are in our model (almost) isomorphic.

But we also have a positive message. And the message comes from a variance decomposition exercise, for which it is totally fine to have somewhat reduced form shocks in your model (the so-called DSGE literature does exactly this in every paper with more fancy econometrics). A sizeable fraction of public policy fluctuations CAN be explained as endogenous reactions to economic shocks (in most macroeconomic models today this fraction is zero). We are totally open to the idea that this is a lower bound and that economic shocks also influence the state of demand for public goods through the identity of the party in power. This could be an interesting next step. But our paper is also pretty clear that you almost surely cannot get the business cycle dynamics of public consumption right in a representative agent framework and a one-shock model (by shocks I mean something orthogonal to each other).

That we believe that "who is in power" and "how that changes with the state of the economy" is important is shown by our companion paper (, where we use economic and political heterogeneity, and which helps in some dimensions, but not in others.

The bottom line is: we do not have a fully satisfying quantitative model that integrates public policy variables into business cycle theory. We are the first to admit that. But we do believe we have made a start. We and hopefully others will take it up from there.