Goldman Sachs CEO Lloyd Blankfein tweeted Tuesday: “Arrived in China, as always impressed by condition of airport, roads, cell service, etc. US needs to invest in infrastructure to keep up.”

This raises an interesting question which I consider in my recently released paper: how do we know how much infrastructure investment is needed in the US?

From an economic perspective, the answer is certainly not “invest to the point where our airports feel as high quality as China’s.” But absent real markets, the amount “needed” is difficult to quantify - an example of the “knowledge problem” associated with central planning.

What level of congestion would drivers tolerate before they were willing to finance road expansion, for example? Eliminating all congestion would be prohibitively expensive. So how far should expansion go? How often should a road be repaired? How much transportation should be by train?

Markets are good at finding the optimal mix of infrastructure spending over time and rewarding those that are better at satisfying demand. Governments, even with the best of intentions, lack the necessary knowledge to find that mix.

Cost–benefit analysis of new projects can be undertaken to decide where scarce resources deliver the highest returns. But that is a very imperfect methodology, and does not tell us anything about “how much” should be invested overall.

In this vacuum, different proxies for how much “should” be invested are cited:

• The American Society of Civil Engineers estimates how much it would cost to improve the infrastructure to a set standard as measured by eight different criteria
• Public investment levels are compared with other countries, or to previous periods of US history
• International surveys of infrastructure quality or capacity are compared
• Quality indicators are tracked over time

But all of these measures have problems from an economic perspective.

There is an obvious self-interested component to the estimates of engineers. But more importantly, sub-measures on quality tell us little about demands and future needs for infrastructure. Part of the survey may be associated with improving the quality of transit systems, for example. But what if driverless cars render them obsolete?

Public investment levels across countries or historically likewise tell us little about what we need to invest today in this country in particular. And international surveys tend to ignore important considerations: many cross-country comparisons of transport infrastructure, such as the World Economic Forum’s Global Competitiveness Report, involve subjective survey answers across countries, meaning that results are shaped by expectations. More objective indices, such as the capacity measures examined by the Kiel Institute (where the US ranks very highly), ignore the quality of that infrastructure.

Of course, you can track quality measures over time too. We know, for example, that the proportion of bridges which are structurally deficient has been falling over the past two decades, but measures of congestion have been rising. This could be highly indicative and used to inform spending decisions. But in some situations allowing physical infrastructure to age or deteriorate may be sensible given the falling demand for its use. As I outline in detail in my new paper, there are many examples of politicians prioritizing things other than economic growth in deciding levels and distribution of transport funds too.

No, in order to really get more responsive infrastructure investment to need, we need more markets in infrastructure provision and to remove the artificial barriers which prevent private investments. A greater use of user fees, such as tolling and congestion pricing, for example, could lead to a greater link between demands and funding.

Read more about all this in my paper here.