An acknowledgment at the outset: I have been a strong advocate of infrastructure investment for years and have even argued that debt-financed infrastructure is likely to reduce debt-to-gross domestic product ratios because of the growth bump from infrastructure. This argument has been taken up by the International Monetary Fund and many others. So I am not neutral on this subject.
My views come in part from a simple calculation. Imagine an infrastructure project that costs $1 and yields a modest 5 cents a year in real return forever, in terms of higher GDP. The project thus has a 5 percent social rate of return. Tax collections would, by extension, rise by about 1.5 cents a year. With the indexed bond market suggesting federal real borrowing costs that are negative for 10 years and 50 basis points for 30 years, the government will come out ahead on the investment. Now, as Brad DeLong and I pointed out a few years ago, matters are more complex than this particularly because capital depreciates, and because infrastructure projects have other effects, but the basic point continues to hold.
So, I was very surprised to read the June CBO report on the consequences of federal investment. As its central base case, it assumes, from its reading of the literature (also here), what it calls a 5 percent rate of return on infrastructure investment. But it then finds that the payback effect of infrastructure investment on the federal deficit is very small, contradicting my earlier claims and more importantly raising doubts about the desirability of a big infrastructure push.
Why the discrepancy? After carefully reading the report and speaking at some length with CBO officials, I think I understand at least one aspect of it. CBO uses the term “rate of return” in a way that is very different than what I would regard as standard usage. They take a 5 percent rate of return to mean that one dollar more capital produces five cents more output. This turns out to be quite different from assuming that investments in public capital earn a 5 percent return.
For example CBO does not subtract depreciation in calculating their rate of return unlike what is standard in the private sector. Nor, contrary to what I would have assumed was natural, do they suppose that rates of return are reduced by the fact that on their (questionable) assumptions it takes many years for dollars invested to turn into capital. (Think of R&D as an example here). Adjusting for these factors, CBO is actually assuming an economic return a little above 2 percent on public investment and then concluding it will not help much.
If one assumes that public investment is basically unproductive, it is no surprise that it does not yield large economic benefits. There is still the question of if the 5 percent rate of return is misdescribed, whether or not CBO has made a reasonable judgement about the productivity of public investment. I'm pretty skeptical. They rely on aggregate econometric estimates of an elasticity of GDP with respect to public capital. Such estimates are plagued by all sorts of problems of heterogeneity, simultaneity, limited variation in the exogenous variables and so forth.
I would prefer to build up from individual projects by looking, for example, at estimates of the return on fixing potholes, building dams, or repairing water systems. I think anyone taking this kind of ground up approach will conclude that the social return to public investment is far higher than 2 percent. This is, of course, a matter of judgement.
There are other, probably lesser, problems with CBO’s work. Much of the adverse effect of public investment on the budget in their work comes from an assumed increase in interest rates resulting from budget deficits. This assumption would have been natural once, but is much less compelling in the current liquidity trap, possible secular stagnation environment. They also haircut the return to federal investment by assuming that it crowds out state and local investment but do not recognize explicitly the benefits of less state and local borrowing. And they focus on deficit impacts rather than the more appropriate question of impacts on debt-to-GDP ratios.
On balance, I do not think anyone should change her mind about public investment based on CBO’s analysis. Great umpires never change their minds. But they do learn from their mistakes. I hope CBO will do better next time out.
Lawrence H. Summers, the Charles W. Eliot university professor at Harvard, is a former treasury secretary and director of the National Economic Council in the White House. He is writing occasional posts on Wonkblog about issues of national and international economics and policymaking.