Infrastructure and Transport: what the evidence shows

The evidence on Infrastructure and Transport in the United States is strongest on the size of the systems and who owns them: about 4.2 million miles of public roads, more than 148,000 public water systems, and an EPA estimate that drinking water systems need $625 billion over 20 years. The best-known overall figures, such as a $3.6 trillion ten-year investment gap, come from advocacy-linked groups and depend on assumptions about what should be built. Research comparing construction costs across countries suggests that U.S. projects often cost far more than similar projects abroad, so the question is not only how much to spend but how.

In plain words

Some infrastructure facts are easy to count: miles of road, numbers of water systems, how much governments spend.

Other numbers are estimates. When someone says the country “needs” trillions of dollars for infrastructure, that number depends on what they think should be fixed or built, and on prices. Different groups get different numbers.

Researchers also ask why projects cost what they do. Some studies find that building a subway line in the United States costs many times more than similar lines elsewhere. If that is right, more money alone may not buy much more infrastructure.

This page sets out what is known, where analysts disagree, and what nobody knows yet. It is a companion to the main Infrastructure and Transport page and the Who decides page.

Defined terms

Needs estimate. A projection of what it would cost to bring a system to a chosen standard over a period. It reflects the standard chosen, not only the condition of the system.

Investment gap. The difference between a needs estimate and expected spending.

Lead service line. The pipe connecting a water main to a building, when made of lead.

Unit cost. The cost per unit of output, such as cost per kilometer of rail or per mile of road.

Hard costs and soft costs. Hard costs are construction labor, materials and equipment; soft costs are design, management, insurance and similar expenses.

Evidence on infrastructure size and ownership

The Federal Highway Administration reports 4,199,209 miles of public roads in the United States in 2023, of which 2,944,124 miles were rural and 1,255,085 urban.⁠¹ By ownership, counties owned 1,799,886 miles, towns, townships and municipalities 1,324,985 miles, state highway agencies 824,539 miles, and federal agencies 159,303 miles.⁠²

EPA reports that there are more than 148,000 public water systems in the United States, and that public water systems regulated by EPA and delegated states and tribes provide drinking water to 90% of Americans.⁠³

Interpretation. Ownership data show why national averages can mislead. Most road miles are rural and locally owned,⁠^1,2 and a rural mile and an urban mile can carry very different numbers of travelers. A statistic about “the condition of U.S. roads” depends heavily on whether it is weighted by miles or by use.

Infrastructure spending: what the data show

The Urban Institute reports that in 2021, state and local governments spent $206 billion on highways and roads, about 6% of their direct general spending; federal transfers covered $52 billion of that (25%), and state and local motor fuel taxes raised about $53 billion, about 26% of highway and road spending.⁠⁴

Federal funding rose under the Infrastructure Investment and Jobs Act, Public Law 117-58, approved November 15, 2021.⁠⁵ The law’s broadband component, the BEAD program, totals $42.45 billion; NTIA announced allocations to all 56 states and territories in June 2023 and reported approving the last final proposal on August 25, 2026, more than three years later.⁠⁶ NTIA also issued a policy notice restructuring the program during implementation.⁠⁶

The federal gasoline tax is 18.3 cents per gallon plus 0.1 cent for the Leaking Underground Storage Tank Trust Fund.⁠⁷ Because it is a fixed amount per gallon, its purchasing power erodes with inflation and its yield falls as vehicles use less fuel.

Interpretation. The time between authorization and construction is long. A broadband program authorized in 2021 completed approval of state plans in 2026,⁠⁶ so evaluations of what the 2021 law achieved will lag by years.

Drinking water: what the data show

EPA’s 7th Drinking Water Infrastructure Needs Survey and Assessment, reported to Congress in 2023, estimates that drinking water systems will need $625 billion over 20 years for pipes, treatment plants, storage tanks and other key assets.⁠⁸ The same survey estimates a total of about four million lead service lines nationwide.⁠⁸

In October 2024, EPA finalized the Lead and Copper Rule Improvements, which require water systems nationwide to “identify and replace lead pipes within 10 years,” along with more rigorous testing and a lower threshold for action.⁠⁹

Measurement caution. The lead service line count is an estimate, and the material of some lines can be uncertain until they are inventoried or dug up. Needs estimates also depend on which standards are assumed, so a new rule can raise the estimated need without any change in pipe condition.

Condition grades and the investment gap

The most widely cited summary of U.S. infrastructure is the American Society of Civil Engineers’ report card. The 2025 edition gave an overall grade of C, with roads at D+, transit at D, drinking water at C- and broadband at C+, and estimated a $3.6 trillion investment gap over the next ten years.⁠¹⁰

How to read it. Supporters treat the report card as an expert synthesis that draws attention to neglected systems. Skeptics note that ASCE is a professional society whose members design and build infrastructure, that grades are judgments combining many measures, and that the investment gap depends on assumptions about what level of service and how much new capacity the country should have. Both views can be true at once: the report card is useful for direction and less reliable as a precise number.

Where researchers disagree: why projects cost so much

The Transit Costs Project at the Marron Institute built a database of more than 900 rapid rail projects in 59 countries and found the United States the sixth most expensive country for rapid rail construction, a ranking it calls partly misleading because U.S. projects are less often tunneled than those in the five costlier countries (37% of U.S. track versus more than 65% in those countries).⁠¹¹ Its case study of New York estimates that Phase 1 of the Second Avenue Subway cost roughly 8 to 12 times a low-cost baseline drawn from Italy, Istanbul and Sweden.⁠¹¹

The project attributes the premium to several compounding factors: oversized stations, labor costs that make up roughly 40% to 60% of hard costs in the Northeast against about 19% to 30% in the low-cost cases, adversarial procurement, weak in-house agency capacity, and governance problems such as political micromanagement and early commitment before designs mature.⁠¹¹ It argues these multiply, so several premiums of 25% to 50% can produce an order-of-magnitude difference.⁠¹¹

Competing readings. Reform advocates read this as evidence that the United States could build much more for the same money by changing procurement, design standards and agency capacity. Some critics dispute the labor findings, and others argue that U.S. projects face real constraints, such as older utilities underground, stronger legal protections and different safety standards, that cross-country comparisons understate. The study itself notes that its New York figure is a range, not a precise value.⁠¹¹ Because most of the evidence comes from rail megaprojects in a few cities, it is uncertain how far the findings extend to roads, water or rural projects.

Where researchers disagree: who should pay

Funding evidence feeds a values dispute. User-fee advocates point to the Urban Institute’s finding that fuel taxes cover only about a quarter of state and local highway spending.⁠⁴ as evidence that drivers do not pay the full cost of roads. Others argue that roads serve everyone, including people who receive goods and services by truck, so general taxes are a fair source. Similar disputes arise over transit fares and water rates, where raising user charges improves cost recovery but can burden low-income households.

What is unknown about infrastructure and transport

• What the 2021 infrastructure law has achieved in condition, access or reliability, since much of the money is still being spent.
• How much infrastructure the country “needs,” which depends on standards and goals that are themselves contested.
• Whether cross-country cost findings for rail apply to other systems such as highways, water and broadband.
• How remote work and changing travel patterns will affect transit and road demand over the long term.
• How accurate lead service line inventories are, until systems complete the inventories required under the 2024 rule⁠⁹
• How to fund roads as fuel tax revenue declines, and how alternatives such as mileage-based charges would affect different groups.

Competing interpretations

An underinvestment interpretation holds that the main problem is too little money, pointing to the ASCE grades and gap.⁠¹⁰ and EPA’s water needs estimate⁠⁸

A cost-effectiveness interpretation holds that the main problem is how money is spent, pointing to the Transit Costs Project.⁠¹¹

A priorities interpretation holds that the problem is what gets built: too much new capacity and too little maintenance, or too much for one mode and too little for another.

These are not exclusive, and each supports different proposals.

Application to this framework

• My Vote Forward presents infrastructure evidence beside participants’ priorities, labeling official data, research findings and advocacy estimates differently. See Priorities.
• Needs and gap estimates are shown with their assumptions, because a number without its standard can mislead. See Measurement.
• Evidence does not settle the values behind funding choices, such as who should pay. See Tradeoffs.
• When a specific project or tax reaches the framework, its own cost and benefit evidence should be shown with the same labels. See Proposals.

Unresolved questions

• How should a public resource present advocacy-linked estimates, such as investment gaps, alongside government data without either dismissing or overstating them?
• Which outcomes (condition, reliability, travel time, safety, access, cost per unit) should matter most when communities compare infrastructure proposals?
• How should local infrastructure evidence be presented when most data are published at the state or national level?
• How quickly should evidence pages update as large federal programs are implemented and restructured?