In a comment on a recent post of mine about government spending on infrastructure, Ben H. wrote:

This seems to ignore the problem of public goods. It is, arguably, pretty easy to identify public goods for which spending $1 increases total welfare across the population by more than $1, but which no individual person has an incentive to spend their own $1 upon.

I didn’t think it was easy at all. I know that the theoretical argument for public goods is easy to make and I make it in every class I teach. But I’ve found that it’s much harder to identify specific public goods on which we can be quite confident that $1 of spending creates substantially more than $1 of benefits. I say “substantially more” because we need also to take account of the deadweight losses from the taxes to pay for those public goods, and the deadweight losses from those taxes tend to be 30% or more of the revenue raised.

Because I didn’t think it was easy to identify such cases, I challenged Ben H. to give cites. He replied:

To those demanding citations of specific papers, etc.: sorry, I’m not an expert in public goods theory and I’m not going to try to pretend that I am. It’s a vast, vast literature, and Google Scholar can provide you with thousands of citations in less than a second. If you want more curated cites to particular studies that are particularly rigorous and convincing, I’m sure there are lots of public goods theorists out there who you could ask, and who would happily provide you with their opinion on that. If you honestly want to know – if demanding cites is not just a rhetorical tactic intended to shut me up – then I suggest you pursue that avenue.

He’s right that the literature on public goods theory is vast. But his original claim was not about public goods theory–it was that it’s easy to identify cases where government can create value substantially above cost. That’s public goods empirics.

By the way, recall also that my original post was not about public goods per se. It was about infrastructure. With much infrastructure, though not all, it is technologically relative low-cost to exclude non-payers, which means that the infrastructure fails to satisfy one of the two criteria that Paul Samuelson laid down in the early 1950s for a public good.