Do We Really Need to Spend a

Trillion Dollars on Infrastructure?


matthew a. turne, is a professor of economics at Brown and a Kenen Fellow at Princeton University’s International Economics Section. This article draws heavily from “Transportation Infrastructure in the U.S.,” by Gilles Duranton, Geetika Nagpal and Matthew A. Turner (NBER working paper 27254).

Published January 24, 2021


In an era of hyperpartisan politics, it seems the two parties can’t agree about anything. Well, not quite anything: for years, Congress and the White House have discussed (sometimes amicably) the pressing need to repair America’s aging transportation infrastructure. And while talk has never quite turned to deed, there seems to be widespread agreement that the country needs a nine-figure infusion of federal cash to rebuild infrastructure worthy of the richest country on earth.

But that “widespread agreement” is built on a fragile frame. The case for a gigantic infrastructure spending package usually rests on three legs. First, that U.S. infrastructure is falling apart. Second, that infrastructure spending has the ancillary benefit of contributing to economic development and the creation of high-paying jobs. Third, that the spending would help to relieve traffic congestion. Each of these three claims is subject to empirical testing, and none holds up well to scrutiny.

Graceful Aging

The cluster of figures on pages 69 and 70 offers objective measures of the condition for the interstate highway system, all bridges, subway cars and public transit buses. Start with the highways.

Each year, the Federal Highway Administration records the number of inches of “suspension travel” — literally, the distance the vehicle bounces up and down on its suspension — experienced by a typical car as it travels a mile on the interstate system. This measure of pavement quality is called the “International Roughness Index” (IRI) and is more or less a census of potholes. Panel A reports the average IRI values between 1990 and 2008 for urban interstates, rural interstates and for the whole network.

The FHA classifies a highway segment’s condition as “acceptable” or “good” according to whether its IRI value is above or below 95 inches of suspension travel per mile. As the dotted line shows, the average condition of the interstate system improved from about 108 inches of suspension travel per mile in 1990 to about 88 inches in 2008. The pace of improvement was similar for rural and urban road segments, although the condition of rural roads was (and is) significantly better than their urban counterparts’.

Now for bridges. As part of what it calls the National Bridge Inventory, the FHA reports an index of condition for all bridges in the U.S. every year. Index scores run from zero to nine, with higher scores indicating progressively better condition and a score of seven indicating good condition.

Panel B reports the share of U.S. bridges in each category for three years: 1992, 2003 and 2017. In all three years, the modal score (that is, the most frequent score) is a seven. Moreover, the proportion of bridges in this category increases over time. Likewise, the share of bridges in the bottom three categories is small and declines over time. There were more bridges in good condition and fewer crumbling bridges in 2017 than in 1992.

Meanwhile, public transit ridership is highly concentrated in a handful of cities. Indeed, according to the Federal Transit Agency, there were just 14 “heavy rail” systems (aka subways and elevated rail) in the country in 2017. New York City’s subway system dwarfs the others. It carries about 70 percent of all the urban heavy rail riders in the U.S., while the next six largest systems carry most of the rest. Bus service is spread much more widely. About 1,100 transit districts provide service, although New York again overshadows the others; it accounts for 40 percent of all bus riders.

Look first at the buses. Panel C reports the average age of public transit buses, with one line showing the numbers for the six largest systems (New York, Chicago, Washington D.C., Los Angeles, Boston and Philadelphia) by ridership. A second line reports the average age of buses in the remaining smaller districts, while the third line reports the average age for all public transit buses.

Although buses in the larger systems are somewhat older than in the smaller ones, the overall trend is modest improvement. In 1992, an average public transit bus was about eight and a half years old. By 2017, it was younger than eight years.

Panel D reports on subway car age, with parallel statistics for New York City, the other 13 systems and the total. The average age of a U.S. subway car turns out to be about 20 years. While there is considerable variation in average fleet ages across time, there’s no obvious trend suggesting aging.

That said, it’s important to remember that rolling stock makes up a modest portion of the infrastructure of a subway system. So we are skeptical about using subway car age as a measure of overall maintenance quality – particularly in light of anecdotal evidence of the deterioration of the New York system and those in other big cities.

The conclusion to be drawn from all four panels is still pretty clear: U.S. transportation infrastructure is not crumbling. Over the past generation, the condition of the interstate highway network has been consistently improving, even as the miles of roadway modestly increased. The same can be said for bridges, where the stats suggest slow improvement. The stock of public transit buses is younger than it was a generation ago.

Subways are more problematic. The mean age of a subway car stayed about the same from 1992 to 2017. But at more than 20 years old, the average car is way past its prime. And there simply is no systematic evidence about changes in the condition of subway tracks and stations.

All told, investment in the interstate, in bridges and in public transit buses has matched or exceeded depreciation over the past generation; heavy rail is probably the exception.

RJ Sangosti/The Denver Post via Getty Images
Is More Better?

Of course, even if transportation infrastructure investment has exceeded depreciation, as the numbers suggest, we may still want to plow funds into more and better infrastructure — for example, if such investments contributed to economic growth. But again, the research coming from academia (as opposed to interest groups with a stake in construction, bond finance and maintenance) does not bear this out.

This research typically compares city-level measures of infrastructure quality to citylevel measures that can serve as proxies for economic activity. Such comparisons immediately raise the question of whether economic activity follows infrastructure investment, or the converse. In fact, the academic literature is preoccupied with this question. Much of the research effort is devoted to implementing statistical techniques that simulate clinical trials in which the “treatment” (more infrastructure) is randomized across “patients” (localities), and the direction of causation can be determined with confidence.

Research based on this approach generates pretty clear conclusions: economic activity does, indeed, migrate to take advantage of transportation infrastructure. In rural areas, economic activity migrates toward highways, and in urban areas, toward subways and other transit. In urban areas, people spread out from the center in response to both more/ better highways and subways. Employment also decentralizes in response to new highways, although not as fast as population.

The literature does not, however, demonstrate clearly that improvements in transportation lead to economic growth. Rather, conclusions vary with the statistical methods used, and there is no compelling basis for preferring one method to another.

What we can say is that transportation infrastructure plays an important role in determining where economic activity takes place. However, we do not have strong evidence that infrastructure drives sustainable growth or creates “good” jobs. Most likely, the overall effect of transportation infrastructure on economic development is small.

On the contrary, highway construction is a machinery-intensive process that employs relatively few people. Second, infrastructure projects are slow to design and build.

Indeed, infrastructure spending does not even seem to be particularly effective as economic stimulus to promote employment during recessions. This somewhat surprising finding reflects two facts. First, modern infrastructure construction — of highways in particular — is not labor-intensive. On the contrary, it is a machinery-intensive process that employs relatively few people. Second, infrastructure projects are slow to design and build. Earmarking spending for highway construction is thus more likely to increase employment in the next recession than in the current one.

There is also considerable evidence about the relationship between the expansion of road capacity and traffic congestion. The experience of Los Angeles is typical: efforts to relieve congestion by adding capacity have repeatedly been undone by increased demand for highway space. In other words, if you build it, they will come.

In short, investments in transportation infrastructure intended to boost economic activity or reduce congestion are, at best, a risky proposition.

Digging Deeper

So, our infrastructure is not crumbling, spending on infrastructure does not appear to be a cost-effective way to promote economic growth, and capacity expansions most likely have only a small impact on congestion in the long run. With the common arguments in its favor discredited, why is the idea of spending more — much more — on infrastructure so appealing to both politicians and the public?

For clues, look more closely at the congestion issue. The U.S. Census conducts major surveys of travel behavior every seven years, which allow us to estimate the speed of an average trip taken by car, bus or subway. And the trends are sobering. Between 1995 and 2017, the speed of an average car trip declined from about 26.5 mph to about 23.5 mph. In the same period, the speed of an average bus trip declined from about 15.1 mph to 11.1 mph, and the speed of an average subway trip declined from about 15 mph to about 10.5 mph.

Meanwhile, the level of usage has increased sharply for both highways and subways. The figures to the right describe the amount of service provided by the three transit modes over the past generation. Panel E reports annual average daily traffic (AADT) for an interstate lane from 1980 to 2008. This is the standard measure of usage and is simply the count of vehicles traversing the roadway on an average day. Traffic on the urban portion of the interstate grew by about 50 percent, to about 15,000 vehicles per day. The second line shows the corresponding trajectory for rural traffic, which increased at about the same pace, from about 3,000 to about 5,000 vehicles per day. The third line reports the national average. (Since about two-thirds of lane-miles are rural, the national average AADT is nearer the rural than the urban.)

Panel E invites two further observations. First, the capacity of an interstate lane is a little over 2,000 vehicles per hour, for a theoretical daily capacity of around 48,000. This is far above even urban AADT values in 2008. Thus, traffic congestion really reflects a shortage of peak-hour capacity — not overall highway capacity.

Second, like public transit, the interstate system overwhelmingly serves urban travelers. Note the resulting disconnect here between perception and reality. The allocation of funds between highways and mass transit is often posed as a contest between rural drivers and urban transit users. In fact, both mass transit and the interstates overwhelmingly serve urban areas.

Panel F shows subway ridership from 1992 to 2017. One line reports on the New York subway, while the second line reports on the remaining 13 smaller systems and the third line reports the national total. U.S. subways carried about 2 billion riders per day in 1992; the number had roughly doubled by 2017. Most of this increase was on the New York subway, though ridership on the other systems increased roughly proportionately.

Panel G shows ridership on U.S. public transit buses over the same time period. Here, one line describes the seven largest systems; the second line, the 1,000-plus smaller transit districts; and the third line, national total. Curiously, in spite of an apparently large increase in the demand for mobility, ridership (which ranged between 4.5 and 5.5 billion passengers per year) shows no clear upward trend.

These data suggest why large infrastructure funding proposals are so popular: in spite of improvements in objective measures of transit condition, peoples’ experience on the highways and on public transit has deteriorated due to crowding and an associated decrease in speed.

RX for Congestion

Proposals for transit-oriented development, transit expansion and what economists call congestion pricing are often advanced as responses to congestion. And each has been investigated.

Not surprisingly, concentrating people into denser housing (transit-oriented development) does little to reduce peak-hour traffic congestion and (yes, surprisingly) has little effect on their total miles of driving. Doubling residential density probably decreases per capita driving by less than 10 percent. Meanwhile, the evidence that increases in public transit capacity decrease congestion is weak.

Indeed, Panel F makes this clear. Over the period from 1992 to 2017, the stock of public transit buses increased by about 30 percent, while ridership hardly budged.

On the other hand, evidence for the efficacy of congestion pricing programs, like the tolling cordon around central London, the congestion pricing scheme in Singapore and various experimental programs in the U.S. is strong. The programs provide financial incentives to change the timing of trips or abandon them altogether. And it works: they increase the speed of travel.

Allen Brown/DBimages/Alamy Stock Photo
Who Pays — and Who Doesn’t

If transportation outlays are to be cost-effective and fairly distributed, the way we allocate expenditures across transportation modes requires scrutiny — as does the way we collect revenues to pay the bills. Two issues in particular deserve attention.

All vehicles are not equal offenders when it comes to road wear and tear. The damage caused to a highway increases very rapidly with the weight borne by each vehicle axle. In fact, a single, fully loaded tractor-trailer rig does about as much damage to the roadway as 2,000 passenger cars (no misprint). To put this in perspective, a single lane of limitedaccess highway can carry about 2,000 cars per hour before it becomes congested. Thus, a single, fully loaded tractor-trailer rig causes as much wear as a whole hour of rush-hour commuting traffic.

Transportation economists have long argued that we should shift the burden of highway finance to the most damaging traffic by taxing trucks on the basis of weight per axle. Such a policy has the further advantage of creating an incentive for truckers to distribute weight over more axles, reducing damage per ton of freight moved as well as helping to pay for highway maintenance.

Second, the organization of bus-based public transit deserves a hard look. Over the past generation, both highways and subways have roughly doubled capacity, as measured by lane-miles and subway cars. Yet bus systems have barely been able to keep ridership constant.

This is important. Total government expenditure on public transit buses and the interstate highway system are about the same — around $20 billion annually. However, the interstate system provides about 700 billion vehicle-miles of travel services per year, while buses provide about 20 billion rider-miles per year.

U.S. in frastructure is not crumbling. On the contrary, except for our subways, investment is pretty clearly outpacing depreciation.

Thus, government expenditure per bus passenger-mile is about 35 times higher than it is for each private vehicle mile. If we refine this calculation to account for private expenditure and the value of time spent traveling, the cost of a bus-mile remains about double that of a mile by car.

It may be that buses must ply routes that are intrinsically more costly to service than the average car trip, and/or it may be that buses generate substantial external benefits in terms of providing mobility for low-wage workers who can’t afford cars. However, it may also be the case that we are just not allocating our resources wisely.

Back to the Original Question

Do we really need a trillion-dollar transportation infrastructure bill? Probably not. U.S. infrastructure is not crumbling. On the contrary, except for our subways, investment is pretty clearly outpacing depreciation. Moreover, such investments probably do not promote economic growth or create good jobs. We might expect them to have important effects on where economic activity ends up. What, then, should our infrastructure policy be?

First, we spend about as much public money on public transit buses as the interstate, and it is thus tempting to conclude that we should reduce funding for buses. This is probably not the right conclusion. Many of our cities will struggle to function without public transit, and the alternative to buses is light rail and subway systems. Because the capital costs for such systems are astronomical, a better alternative would be to experiment with different ways of providing bus service to try to develop bus-based transit that more people will want to use.

Second, the financial burden of maintaining the interstate rests largely on general government revenue and on gas taxes that affect passenger cars, while most of the damage to the roads is caused by trucks bearing too much weight on too few axles. Shifting the financial burden from cars to these trucks would quite possibly shift some long-haul freight to railroads and would certainly create incentives for less damaging practices in the trucking industry. Finally, there is only one way to manage traffic congestion: congestion pricing.

None of this would be easy in political terms. Indeed, that reality partly explains why infrastructure policy stubbornly resists rational thinking. But the alternative is more of the same – more money spent, with evershrinking payoffs. We could probably find better ways to spend a trillion dollars.

main topic: Infrastructure