Part One: The Ledger Has Two Halves

Few federal programs invite two completely different verdicts with equal confidence, and the national limited-access highway network built after 1956 is one of them. Ask an economist who studies productivity what the program accomplished, and the answer comes back in the language of measured gains: a one-time, unrepeatable boost to output per worker, concentrated in the vehicle-intensive industries that the new roads served best. Ask a historian who studies American cities what the same program accomplished, and the answer comes back in the language of measured losses: hundreds of thousands of homes destroyed, more than a million people displaced, with the damage concentrated in poor and Black neighborhoods that had little power to choose another route. Both answers are correct. Both rest on serious research. The honest accounting holds them at the same time, and this article is built to make that accounting possible.

The statute at the center of this assessment is the Federal-Aid Highway Act of 1956, designated in the legislative record as Public Law 84-627, together with the successive authorizations that carried the program forward. Construction was concentrated between the late 1950s and the early 1980s, the decades in which the core national network took shape. The findings below measure what that construction did, not what its sponsors hoped it would do. That distinction matters, because the series this article belongs to assesses each statute against two standards at once: the aims its drafters stated for themselves, and the consequences its drafters never measured. The 1956 act was sold principally as a national defense and commerce measure, a program to connect the country. Its economic returns turned out to be very large. Its urban costs turned out to be very large too, and nobody in Washington was keeping a ledger for those.

Public Law 84-627 arrived with three stated aims: national defense, interstate commerce, and the connection of the country’s principal cities. Defense came first in the law’s own framing and supplied the urgency; commerce and connection supplied the economic logic. Each of the three aims contained, unnoticed, the seed of a consequence the drafters never measured. A road program justified as a defense asset could command a scale of resources and a pitch of urgency that a purely commercial project could never have sustained, which made the program national rather than regional, federal rather than state-led, and fast rather than incremental. But a network built for the rapid movement of troops and materiel in a national emergency is also, by the same geometry, a network built for the rapid movement of everyone else in ordinary times. The defense rationale got the roads built. The civilian uses, which the commerce aim had already promised, are what filled them. Nothing in the defense aim required the drafters to ask who lived where the roads would go, because a defense asset is drawn on a map of the nation, not on a map of neighborhoods.

The commerce aim carried its own unexamined assumption. Interstate commerce, in the constitutional sense the drafters invoked, means trade across state lines, and a road network that serves it will naturally favor long-distance movement over local circulation. That priority shaped the engineering: limited access, high speeds, gentle grades, the separation of through traffic from local streets. But the same design that moves freight efficiently between cities moves commuters efficiently within metropolitan areas, and once the roads existed, the distinction between interstate commerce and local travel collapsed in practice. The trucks the commerce aim promised and the suburban commuters nobody promised ended up on the same pavement, generating the congestion the next generation would try, and fail, to build its way out of. The aim was achieved so thoroughly that it created the problem the program was later asked to solve.

The third aim, connecting the principal cities, is where the unmeasured consequences were most directly seeded. A network that connects cities must decide, at every city, whether to pass through or around, and the engineers of the 1950s, with the enthusiastic support of downtown business interests, chose through. The through routes delivered the downtown access the commerce aim wanted and the clearance opportunities the urban renewal era wanted, and they did so by converting the aim of connection into an instrument of demolition. Had the drafters chosen ring routes and bypasses as the default urban form, the displacement ledger of this article would look very different, and the suburbanization ledger might look different too. They did not, because connection, as they understood it, meant reaching the center, and the center was where the poorest residents lived. The aim was innocent. Its geometry was not.

The three aims were also in tension with one another from the start, and the tensions explain some of the program’s strangest outcomes. Defense wanted speed and redundancy: the ability to move heavy loads quickly along multiple routes, which argues for capacity beyond any civilian need. Commerce wanted efficiency: the cheapest reliable movement of goods, which argues for the routes and designs that minimized cost per ton-mile. Connection wanted reach: principal cities linked to one another, which argues for the through-city alignments that the other two aims might have been happy to route around. A program serving one master could have optimized. A program serving three had to compromise, and the compromises are visible in the record. The defense-grade capacity that commerce never needed became the room that induced demand filled. The connection-driven urban alignments that defense never required became the clearances the cities endured. Each aim got what it asked for, and each aim’s fulfillment complicated the others’.

This article organizes the evidence into five findings, each attached to the named researchers and publications that established it. Productivity rose, and the rise is documented by John G. Fernald in the American Economic Review in 1999. Suburbanization accelerated, and the causal claim is documented by Nathaniel Baum-Snow in the Quarterly Journal of Economics in 2007. Congestion proved resistant to widening, and that resistance is documented by Gilles Duranton and Matthew A. Turner in the American Economic Review in 2011. Displacement was enormous, and the numbers come from federal estimates, from highway historian Raymond Mohl, and from political scientist Clayton Nall. Safety improved, and freight was transformed, and both are documented by the Federal Highway Administration and by the congressional record marking the program’s fiftieth anniversary. Each finding carries its period, its authors, and its method, because the point of this article is not to assert conclusions but to show how they were earned.

The throughline, the single claim a reader should carry away, is this: the economic literature supports large gains from building a network that did not exist and small gains from widening one that does. High returns to the first network, low returns to the next lane. That asymmetry is why the same body of evidence can justify the original program and undercut most contemporary expansion proposals. The first network changed what was possible for the movement of goods and people across a continent. The next lane mostly changes how crowded the existing network feels, and the research on congestion explains why even that benefit does not last. Holding this asymmetry in mind is the practical payoff of the whole exercise, because every current debate about adding capacity to the existing system is, in effect, a debate about whether the returns to the first network apply to the next lane. They do not, and the researchers who measured the first network’s returns are the ones who say so.

The complication this article must handle is the existence of two reductive accounts, each of which erases half the evidence. One account treats the program purely as an economic triumph, a story of productivity and growth in which the destroyed neighborhoods are an unfortunate footnote. That account cannot survive the displacement numbers or the route-selection history. The other account treats the program purely as an instrument of destruction, a story of bulldozers aimed at Black communities in which the productivity and safety gains are dismissed as irrelevant. That account cannot survive Fernald’s productivity findings, the fatality-rate comparisons, or the simple fact that local officials, not only federal planners, chose most of the urban routes. The accurate story is messier than either reduction. It is a story of layered decisions, federal, state, and local, in which a program that genuinely raised national productivity also genuinely destroyed hundreds of thousands of homes, concentrated in the neighborhoods least able to resist.

Readers should also know what this article will not do. It will not relitigate the legislative history of the 1956 act in detail; the companion guide to that statute covers the law itself. It will not speculate about what a different transportation policy might have produced; the counterfactuals here are the ones the researchers constructed, and no others. And it will not treat the five findings as five separate stories. They are one story, told in five findings, because the same concrete that carried freight more cheaply also carried commuters out of central cities, because the same interchanges that cut travel times also cut neighborhoods in half, and because the same widening that promised congestion relief produced the traffic that filled the new lanes. The evidence table in Part Nine puts all five findings side by side so that the pattern is visible at a glance, and the closing part returns to the one claim worth keeping.

A word about the scale of the undertaking helps set the stage. The construction concentrated between the late 1950s and the early 1980s was one of the largest public works programs in human history, a continental network of limited-access roadways built across deserts, mountains, farmland, and, fatefully, cities. The rural segments are the ones most Americans picture when they think of the program: long ribbons of concrete crossing empty country, carrying freight and travelers at sustained high speeds. The urban segments are the ones this article spends most of its time on, because the cities are where the findings collide. The same mile of urban expressway that appears in Baum-Snow’s suburbanization estimates, in Duranton and Turner’s congestion estimates, and in Mohl’s displacement counts is a single piece of concrete doing three things at once: moving commuters outward, generating the traffic that congests it, and dividing the neighborhood it passes through. The five findings are not five stories about five different programs. They are five measurements of one program, and the measurements interlock because the concrete interlocks.

The series thesis deserves one more explicit statement before the evidence begins, because it governs what counts as a complete assessment. A statute can succeed on its own terms and still fail the people it was never designed to consider. The 1956 act’s drafters measured success in miles built, in connections completed, in the movement of goods and military convoys. They did not measure central-city population loss, or displaced households, or the racial distribution of the displacement, because those were not the program’s aims and nobody required the accounting. Assessing the law against its own aims yields a story of remarkable achievement: the network was built, the freight was transformed, the productivity gains were real. Assessing it against the consequences its drafters did not measure yields a story of remarkable damage: the neighborhoods cleared, the cities hollowed, the congestion manufactured. The complete assessment holds both, and it refuses the comfort of choosing. That refusal is the article’s method, and it is also its moral.

There is a final preliminary point about how to read the numbers that follow. Every major quantitative claim in this article carries its author, its publication, and its period, and the article will repeat those attributions rather than letting them fade into the background. This is deliberate. In public debate, findings shed their sources quickly: the 18 percent becomes “studies show,” the unit elasticity becomes “experts say,” the million displaced becomes “everyone knows.” The house rules for this article ban unattributed appeals to research, phrases like “research shows” with no source named, precisely because unattributed findings cannot be checked, qualified, or argued with. Naming Fernald, Baum-Snow, Duranton and Turner, Mohl, Nall, and the Federal Highway Administration is not academic decoration. It is what makes the ledger auditable, and an auditable ledger is the only kind worth keeping.

Part Two: How Transportation Economists Learned to Prove Causation

Before the findings, the methods, because the findings are only as strong as the methods behind them. The central problem in studying what roads do is embarrassingly simple: roads are built where people already are, or where planners expect them to go. A city that builds a highway and then grows has not necessarily grown because of the highway. It may have built the highway because it was already growing, or because its leaders expected growth, or because its employers demanded better connections to markets that were expanding for other reasons. Economists call this the problem of reverse causation, and for decades it made every claim about what highways caused into a claim that could be answered with a shrug. The correlation between roads and prosperity was obvious. The direction of the arrow was not.

The breakthrough in the suburbanization literature came from an old map. In 1944, more than a decade before the 1956 act, federal planners drew the Interstate Highway System Plan, a proposed national network of limited-access routes. Nathaniel Baum-Snow realized that this 1944 plan could serve as what economists call an instrument: a source of variation in actual highway construction that was decided long before the postwar suburbanization it would be used to explain. The logic runs like this. The 1944 plan marked certain cities and corridors for future routes. When the actual program was built after 1956, it followed the 1944 plan in many places but not all, for reasons that included later political bargaining, engineering constraints, and local opposition. If you compare cities whose 1944 plans called for highways through their centers with cities whose plans did not, and you find that the first group lost more central-city population, you have evidence that the highways caused the loss, because the 1944 plan itself could not have been drawn in response to suburbanization that had not happened yet. The plan is exogenous to the outcome. That is the whole trick, and it is worth understanding, because it is what separates the findings in this article from the anecdotes that surround them.

An instrument must satisfy two requirements, and the 1944 plan satisfies both. First, it must predict the treatment strongly: the plan’s proposed corridors must actually predict where highways were built, or the whole exercise is noise. Construction history confirms this, because where the plan drew its rays, the later network tended to follow. Second, it must affect the outcome only through the treatment: the plan must not correlate with the later forces that drove suburbanization on their own, or the estimate is contaminated. The plan’s 1944 vintage secures this, because a group of planners in the middle of a world war, drawing a network for military movement and intercity connection, could not have been responding to the suburban housing demand of 1970. What survives the filter is highway variation that is, for the economist’s purposes, as good as random, and that variation still predicts population loss. That is causation, not correlation. The plan is not a perfect instrument, and honest researchers say so, because plans reflect the thinking of their time. But it predates the great postwar suburban boom by years, and it was drawn for military and intercity purposes rather than as a response to metropolitan decentralization. When Baum-Snow reports that one new highway passing through a central city reduces its population by about 18 percent, the number means something different than a raw correlation would mean. It means the highway did it, or at least that the best available method for isolating causation says the highway did it.

A second methodological idea organizes the congestion finding, and it is the idea of elasticity. An elasticity measures how much one quantity changes when another quantity changes, in proportional terms. Duranton and Turner estimated the elasticity of vehicle-kilometers traveled with respect to roadway lane kilometers, and they found it to be roughly one. That is the unit elasticity at the heart of the fundamental law of road congestion: increase lane kilometers by ten percent, and vehicle-kilometers traveled rise by about ten percent. The one-for-one relationship is what makes the finding so arresting. It says that the traffic which fills a widened road is not some mysterious force but the predictable sum of three responses: current residents drive more, commercial traffic increases, and people move into the now-more-accessible area. None of these responses requires anyone to behave irrationally. Each is individually sensible. Together they erase the congestion relief the widening was meant to provide.

An elasticity of one is a knife-edge result with large consequences. If the elasticity were well below one, then widening would durably reduce congestion, because each new lane would carry mostly diverted traffic rather than new traffic. If the elasticity were well above one, widening would actively worsen congestion, generating more than enough new traffic to fill the added space and then some. An elasticity of roughly one means the system sits at the boundary: the road supply creates its own demand in matching proportion, like a container that expands to fit whatever is poured into it. The policy consequence is that congestion, measured as delay per traveler, becomes a roughly constant feature of the urban network regardless of how many lanes are built, because the quantity of travel adjusts to the quantity of capacity. This is why the finding is called a law: it describes an equilibrium tendency, not a one-time event.

The third methodological idea is the distinction between building a network and widening one, between the one-time effect of connecting places that were previously unconnected and the marginal effect of adding capacity to an existing network. Fernald’s productivity work turns on exactly this distinction. When no limited-access network exists, building one changes the economics of every vehicle-intensive industry at once: factories can serve larger markets, distributors can consolidate warehouses, just-in-time delivery becomes possible. Those are level changes, one-time shifts in what the economy can do. When the network already exists, adding a lane changes the economics at the margin: some trips get slightly faster, some freight moves slightly more cheaply, but nothing fundamental is transformed. The distinction explains why the same researcher can find that roads were enormously productive in the 1950s and 1960s and simultaneously find that additional road investment at the margin does not look unusually productive. The first network was a revolution. The next lane is a renovation.

The one-time versus marginal distinction has a parallel in every network technology, and the parallel helps fix the intuition. The first telephone network transformed communication; the ten-thousandth telephone line added convenience. The first electrical grid transformed production; the marginal substation added reliability. In each case the transformative gains belonged to the building of the network, and the later investments, however useful, operated within a world the network had already remade. Fernald’s finding places the highway program in this familiar pattern. The 1950s and 1960s built the network and harvested the transformation. Everything since has been the ten-thousandth line: valuable, worth doing on its own terms, but not transformative, and not to be justified with the rhetoric of transformation. The distinction does not devalue later investment. It correctly values it, which is a different thing.

Taken together, the three methodological ideas do something unusual for a policy literature: they make the evidence self-limiting. The 1944 plan instrument says what it says about suburbanization and no more. The unit elasticity says what it says about urban interstate widening and no more. The one-time boost says what it says about the build-out decades and no more. A lesser literature would stretch each finding beyond its bounds, turning the fundamental law into a universal claim about all roads or the productivity result into a blanket endorsement of concrete. This article will not do that, because the researchers did not do that, and the credibility of the whole enterprise depends on respecting the boundaries the methods impose. The reader who finishes Part Two should carry three tools: the instrument, the elasticity, and the one-time versus marginal distinction. Every finding that follows will be sharper for them.

It is worth dwelling a little longer on why the instrument strategy represented such a breakthrough, because the intuition behind it governs how much weight the suburbanization finding can bear. Before instruments, the literature on roads and cities consisted largely of correlations with stories attached. A researcher would observe that cities with more highways had more suburbanization and would tell a plausible story about the connection, and a critic would reply with an equally plausible story running the other direction, and there the debate would stall. The instrument breaks the stalemate by finding a moment when the treatment was assigned for reasons unrelated to the outcome. The 1944 plan is such a moment: a group of planners in the middle of a world war, drawing a network for military movement and intercity connection, could not have been responding to the suburban housing demand of 1970. When that plan predicts the highways of the 1960s, and the highways of the 1960s predict the population losses of the 1970s and 1980s, the chain of prediction carries causal force that no correlation can match. The plan is a window into a counterfactual world, the world in which highway assignment was random with respect to suburbanization, and looking through that window is what lets Baum-Snow speak of causation rather than association.

One more piece of institutional context belongs in this methodological opening, because it explains why the evidence looks the way it does. The program was a creature of layered government. Under the 1944 Federal-Aid Highway Act, routes were selected by joint action of the state highway departments. The Public Roads Administration called on the states to submit recommendations and worked with state and local officials to develop plans for the larger cities, and the federal Bureau then approved those plans. Federal money paid for the network, but state highway departments, working with local officials, selected the actual routes. That division of labor matters for everything that follows, especially the displacement story, because it means there is no single decision-maker to blame and no single level of government that can claim clean hands. The federal highway funding mechanics created a system in which Washington financed what state and local officials chose, and the choices those officials made are the ones the displacement record documents. That financing structure also shaped the evidence base itself: because different states built at different speeds and chose different urban routes, researchers gained the variation they needed to compare outcomes across places. The layered government that complicates blame is the same layered government that makes the econometrics possible.

How can economists tell whether a highway caused growth instead of merely following it?

They use instruments, sources of variation decided before the outcome they explain. The 1944 plan marked corridors for military and intercity reasons, twelve years before the 1956 act, so cities it favored did not choose their treatment. Comparing favored and unfavored cities isolates the highway’s own effect from the growth that would have happened anyway.

With the three tools in hand, the structure of the argument to come can be stated compactly. The productivity finding uses the cross-industry and over-time pattern of gains to identify the one-time boost of the build-out decades. The suburbanization finding uses the 1944 plan to identify the causal effect of highways on central-city population. The congestion finding uses historical instruments and the cross-city relationship between lane kilometers and vehicle travel to identify the unit elasticity. The displacement finding uses archival reconstruction and demographic analysis rather than econometric identification, because the question is not whether the highways caused the displacement but how much displacement the highways caused and whom it fell on. The safety and freight findings use official statistics, because the question is measurement rather than causation. Different questions, different methods, one ledger. The reader who keeps the method behind each finding in view will know exactly how much weight each finding can bear, and that knowledge is the difference between being persuaded and being merely impressed.

Part Three: The Productivity Finding

The strongest economic case for the program comes from John G. Fernald’s 1999 paper in the American Economic Review, “Roads to Prosperity? Assessing the Link between Public Capital and Productivity,” published in volume 89, issue 3. Fernald asked a question that sounds simple and is technically demanding: did public investment in roads raise the productivity of American industry, and if so, by how much and for how long? His answer has two parts, and both parts matter. Roads are productive, and they raise productivity disproportionately in vehicle-intensive industries, the industries whose production and distribution depend most on moving things by truck. But at the margin, road investments do not appear unusually productive. The paper’s most quoted sentence puts the asymmetry in plain language: “Intuitively, the interstate system was highly productive, but a second one would not be.”

To understand what Fernald measured, consider what productivity means in this context. Productivity is output per unit of input: how much the economy produces for each hour of labor, each dollar of capital, each gallon of fuel. When a new limited-access network lets a manufacturer serve a market three times as large from a single plant, or lets a distributor replace a dozen regional warehouses with a few national ones, measured productivity rises even though nobody works harder. The road is a form of public capital, and like any capital, it makes the other inputs more effective. Fernald’s contribution was to trace that effect across industries and across time, and to show that it was concentrated where theory says it should be concentrated: in the industries that use roads most intensively. A finding that roads raised productivity equally in all industries would be suspicious, because it would suggest the roads were merely correlated with some general boom. A finding that the gains concentrate in vehicle-intensive industries is the signature of a genuine causal channel.

The logic of Fernald’s test is the model for every identification strategy in this article. Consider what a skeptic would say about the naive story. The 1950s and 1960s were good decades for American productivity for many reasons: the postwar technological frontier, the educated workforce, the stable institutions. Roads were built in those decades. To watch the two trends rise together and declare the roads the cause is to commit the oldest fallacy in the empirical book. Fernald’s move was to stop asking whether productivity rose when roads were built and start asking where it rose most. If roads were doing real work, the work should show up where roads are used most intensively, in the sectors whose production depends on moving physical things by truck, and show up less where output is made of services that never touch a highway. The pattern he found matched that prediction. The skeptic’s alternative, that some general boom lifted everything, cannot explain why the lift was proportional to road use. A rising tide lifts all boats equally. Fernald found a tide that lifted the trucking-intensive boats highest, which is not how general booms behave.

The time pattern is the second half of the finding, and it is the half with the largest implications. The productivity gains from the road network were concentrated in the 1950s and 1960s, the decades when the network was being built out for the first time. After the network existed, additional investment stopped looking special. Fernald’s interpretation is that the construction of the original system delivered a one-time, unrepeatable productivity boost, and that the exhaustion of that boost helps explain the productivity slowdown that followed in the 1970s. The economy got the gains from connecting its regions with limited-access highways exactly once. There is no second connecting of the regions to be done.

This is the finding that does the most work in the article’s central claim, so it deserves to be stated with unusual care. Fernald did not find that roads are useless. He did not find that the 1956 program was a mistake. He found the opposite: the program was highly productive, unusually so. What he also found is that the productivity was a property of the first network, not of road investment in general. The distinction is everything. It means the evidence for the original program cannot be borrowed by the advocates of the next widening project. When a contemporary proposal promises that new lanes will pay for themselves in productivity gains, the correct response is to cite the very literature that established the original gains, because that literature is explicit that the gains were one-time. The researcher who most firmly established that the program was productive is also the researcher who most firmly established that building it again would not be.

The word “marginal” carries the whole policy argument, so it is worth slowing down over it. In economics, the margin is the next unit: not the road network as a whole, but the next lane mile, the next interchange, the next widening. The average return on the network blends the spectacular early gains, when the first limited-access connections replaced slow two-lane arterials, with the modest later gains, when another lane was added to a corridor that already had plenty. Fernald’s finding is that the average is high because the early gains were extraordinary, while the margin is ordinary, because the extraordinary circumstance, the absence of any network, no longer exists. This is also why the productivity slowdown after the build decades is partly an arithmetic story. A one-time boost to the level of productivity, spread over the construction years, shows up as faster growth while it happens and as slower growth when it stops, even if nothing else changes. The economy did not forget how to be productive. It finished building the thing that had made it unusually productive for a while.

It is worth pausing on why the one-time character of the gain makes intuitive sense, because the intuition guards against a common misreading. Imagine the first limited-access connection between two major metropolitan regions. Before it exists, every shipment between them moves on ordinary roads, through towns, at lower speeds, with higher costs and more delays. After it exists, the cost of moving goods between those regions falls permanently. That is a genuine, measurable gain. Now imagine adding a lane to that connection twenty years later. Some trips speed up slightly. Some congestion eases temporarily. But no new region is connected to any other region. No new market becomes reachable that was previously unreachable. The economics have changed at the margin, not at the level. The first project transformed the map. The second project edited it.

The quote that carries the whole finding repays slow reading: “Intuitively, the interstate system was highly productive, but a second one would not be.” The first clause concedes everything the program’s defenders claim. Highly productive, not marginally, not debatably, but highly, a strong word from a careful researcher writing in the profession’s flagship journal. The second clause withdraws everything the program’s defenders want to borrow. A second one would not be highly productive, because the conditions that made the first one productive, the absence of any comparable network, no longer exist. The sentence is doing double duty as both tribute and warning, and that doubleness is exactly the structure of this article’s central claim. Fernald wrote the tribute and the warning in a single sentence in 1999. The article’s job is to show that both halves of his sentence are supported by the rest of the evidence.

There is also a temporal irony in the productivity story worth naming. The gains were harvested in the 1950s and 1960s, the same decades when the urban segments were destroying the neighborhoods documented in Parts Six and Seven. The factory whose productivity rose because it could now serve a national market, and the neighborhood that was cleared to make room for the highway serving that factory, are two sides of a single historical moment. The productivity statistics do not record the clearance. The displacement statistics do not record the productivity. Each literature measured what it was designed to measure, and neither was designed to measure the other. This article exists because the two literatures need to be read together, and because the statute’s drafters, who celebrated the first and never counted the second, left the integration to later readers.

Finally, the productivity finding sets up the question that the congestion finding will answer. If the network was so productive, why did its builders and their successors keep promising that the next expansion would deliver similar gains? Part of the answer is political: concrete is popular, and ribbon-cuttings are photogenic. But part of the answer is intellectual: the memory of the transformative decades created an expectation that road investment is inherently productive, an expectation Fernald’s own work contradicts. The expectation survived because the first clause of his sentence, “highly productive,” traveled widely, while the second clause, “a second one would not be,” traveled less far. Restoring the second clause to public debate is one of the practical purposes of this article.

The productivity slowdown debate, which Fernald’s work helped reshape, deserves a brief coda. For years, economists treated the post-1973 slowdown as a puzzle demanding a villain: oil prices, regulation, the exhaustion of catch-up growth, the mismeasurement of services. Fernald’s contribution was to suggest that part of the puzzle was not a puzzle at all but an accounting artifact of the preceding boom. If the 1950s and 1960s were flattered by a one-time infrastructure dividend, then the 1970s look slower partly by comparison with an unsustainable baseline. The slowdown was real, but some of its depth was the mirror image of the earlier height. That reframing does not absolve the policies of the 1970s or explain every sector’s performance, but it disciplines nostalgia for the high-growth decades: some of what is remembered as normal dynamism was a tailwind that could blow exactly once, from a road network that could be built exactly once.

What does it mean to call the productivity boost a one-time gain?

It means the gains came from connecting previously unconnected regions, a transformation that can only happen once. After the network exists, new investment only edits capacity at the margin. Fernald found the 1950s and 1960s boost unrepeatable, which is why the original program’s returns cannot justify the next widening.

The One-Time Build and Its Financing

Before the article leaves productivity behind, one more piece of the first half belongs here: the program was a build, and a build is a different economic object from a maintenance regime. Construction under Public Law 84-627 and its successor authorizations ran hardest from the late 1950s into the early 1980s, a concentrated burst of civil works with no parallel in American history. The distinction matters because it is the physical counterpart of Fernald’s economic distinction. The large returns belonged to the burst, to the years when the network went from nonexistent to nearly complete. The small marginal returns belong to what came after, the long decades of operating, maintaining, and widening a finished system. An assessment that does not separate the build from the afterlife will misread both.

The financing arrangement made the burst possible. Congress paid for the construction phase through a dedicated federal trust fund fed by highway user charges, principally the taxes levied on motor fuel, rather than through annual appropriations from general revenues. The design was deliberate. A trust fund insulated the multi-decade build from the year-to-year politics of the appropriations process, and it tied payment to use: the drivers and truckers who benefited paid in proportion to their fuel consumption. The full mechanics of that arrangement belong to the companion account of federal highway funding mechanics, and this article will not duplicate it. What matters here is the consequence of the design. Dedicated revenues made the largest public works program in American history politically sustainable across administrations, and they did so by converting a political choice into an apparently technical one: the money was there, earmarked, so the building continued.

That continuity is worth pausing on, because it connects the financing to the second half of this article. A program that pays for itself through dedicated revenues develops a momentum of its own. State highway departments, which selected the actual routes in partnership with local officials, had every incentive to keep building, because the federal money was available and the local match was modest by comparison. The trust fund did not cause the routing decisions documented in Part Seven, but it funded the machine that made them, year after year, through the 1960s and 1970s, long after the original defense rationale had faded into ceremony. The one-time build was supposed to end. The financing made it easy to continue, and the continuation is where the marginal returns fell and the neighborhood costs accumulated. Fernald’s sentence applies to the money as well as to the concrete: the first network was highly productive, and a second one, bought with the same dedicated revenues, would not be.

The end of the build era is the hinge of the whole assessment. From the late 1950s into the early 1980s, the program was doing the transformative work: connecting unconnected cities, enabling the freight revolution, delivering Fernald’s one-time productivity boost. After the early 1980s, the transformative work was done, and what remained was the ordinary work of maintenance, reconstruction, and widening, at the ordinary marginal returns Fernald measured. The tragedy of the timing is that the heaviest neighborhood destruction, the 1957 to 1968 clearance years in Mohl’s records, came during the transformative phase, when the returns were highest and the resistance weakest. The farce of the timing is that the building continued past the transformation, sustained by the financing machine, into the decades when the returns were thin. The assessment has to keep the eras distinct, because confusing them is how the burst’s glory gets spent justifying the flow’s waste.

Part Four: The Suburbanization Finding

If the productivity finding is the strongest economic case for the program, the suburbanization finding is the strongest evidence that the program remade American life in ways its drafters never measured. Nathaniel Baum-Snow’s 2007 paper in the Quarterly Journal of Economics, “Did Highways Cause Suburbanization?,” published in volume 122, issue 2, used the 1944 Interstate Highway System Plan as an instrument for actual highway construction, exactly as described in Part Two. The headline result is stark: each new highway passing through a central city reduces that city’s population by about 18 percent. Not by 18 percent of the metropolitan area. By 18 percent of the central city itself, the dense urban core.

The “ray” in the finding is worth defining precisely, because the 18 percent figure is per ray and the per-ray framing is what makes it interpretable. A ray is a highway corridor radiating from the central city outward, the spoke of the metropolitan wheel. Most American cities received multiple rays, one for each direction the network reached. The 18 percent is the estimated population effect of one such ray, which means a city with three rays experienced roughly three times the decentralizing push, subject to the usual cautions about linearity. The per-ray framing matters because it converts the finding from a city-level anecdote into a dose-response relationship: more highways, more suburbanization, in proportion. Dose-response is the shape causal relationships take.

The counterfactual Baum-Snow constructed makes the finding concrete. Between 1950 and 1990, the central cities of American metropolitan areas lost about 17 percent of their population, even as their metropolitan areas as a whole grew by about 72 percent. The cities hollowed while their regions sprawled. Baum-Snow’s estimate implies that had the highway network not been built, aggregate central-city population would have grown by about 8 percent over the same period. The gap between the observed 17 percent decline and the counterfactual 8 percent growth is the measure of what the roads did. They did not merely accommodate a suburbanization that was happening anyway. They caused a large share of it.

That causal language deserves emphasis, because the alternative reading has been popular for decades. The accommodating account says Americans wanted suburban houses and cars, the highways merely served that demand, and the central cities would have emptied with or without the concrete. Baum-Snow’s instrument is designed to test exactly that account, and it rejects it. The 1944 plan, drawn before the suburban boom, predicts which cities got highways through their centers, and the cities it predicted lost population relative to the cities it did not. If suburbanization had been purely a matter of consumer preference, the 1944 plan would have no predictive power, because a twelve-year-old military and intercity plan would not track the geography of housing demand in the 1960s and 1970s. The plan’s predictive power is the evidence that the highways themselves pushed people outward.

The mechanism is not mysterious once it is stated plainly. A limited-access highway through or near a central city does two things at once. It makes distant suburban land accessible for commuting, which raises the value of that land and encourages its development. And it makes the central city itself a less pleasant place to live, because the highway brings noise, pollution, and physical division into the neighborhoods it passes through. The first effect pulls residents outward. The second effect pushes them. Together they produce the 18 percent figure, which is best understood not as a precise prediction for any single city but as an average causal effect across the cities in the study: on average, a highway through the center cost the central city nearly a fifth of its people.

It matters that the finding concerns central cities rather than metropolitan areas, because the distinction captures what was lost. The metropolitan areas grew enormously; the 72 percent figure is itself a product of the highway era, as new suburban development filled the land the roads opened. What shrank was the dense, walkable, transit-served core, the part of the metropolis where land was expensive, buildings were old, and residents were disproportionately poor and disproportionately Black. The suburbanization finding and the displacement finding are not two separate stories. They are the same story told from opposite ends: the people who could leave did, and the people who could not leave lived next to the construction.

Baum-Snow’s work also clarifies the limits of what the highway program can be blamed for. The counterfactual says central cities would have grown 8 percent without the network, not that they would have thrived in some imagined golden age. Other forces were pushing people outward too: rising incomes, the GI Bill’s mortgage subsidies, racial dynamics in housing markets, the decline of urban manufacturing. The highways were, in the careful language of the literature, a large share of the cause, not the sole cause. But a large share is enough to matter. It is enough to say that the geography of postwar America, the hollowed cores and the sprawling peripheries, is in substantial part a product of a federal construction program, not merely of consumer choice.

The suburbanization finding connects directly to the congestion finding that follows, because the same outward movement that emptied the central cities created the commuter sheds that the highways then had to serve. Each new suburb generated new trips. Each new trip added to the traffic on the network that had made the suburb possible. The program, in this telling, manufactured much of the demand it was then asked to satisfy, and the fundamental law of road congestion describes what happened when it tried. The findings interlock, which is why the article presents them as a single arc rather than a list.

The push and the pull deserve to be separated analytically, because they fell on different populations. The pull, newly accessible suburban land, newly valuable for development, newly convenient for commuting, acted on households with the means to move: the middle-class families who could afford a car, a down payment, and a longer commute. The push, the noise and pollution and physical division that a limited-access highway brings to the neighborhoods it traverses, acted on the households that remained: disproportionately poor, disproportionately Black, disproportionately renters without the resources to leave. The highway thus sorted the metropolis by class and race, rewarding mobility for those who had it and imposing immobility’s costs on those who did not. The 18 percent population decline is an average over this sorting process. Behind the average is a redistribution of who lived where, and the redistribution favored those who were already favored.

The counterfactual deserves one more look, because counterfactuals are where causal claims live or die. Baum-Snow’s estimate says central cities would have grown 8 percent without the network. That is a modest number, and its modesty is a strength. It does not claim that the highways destroyed thriving downtowns that would otherwise have boomed. It claims that the highways turned mild growth into steep decline, a swing of roughly 25 percentage points between the world with the network and the world without it. The modesty of the counterfactual growth makes the finding harder to dismiss: it does not require believing in a golden age of the central city, only in the ordinary growth that rising national population and income would have produced. The highways did not merely accelerate a trend. They reversed one.

It is also worth noting what the finding implies about the geography of opportunity in postwar America. The metropolitan growth of 72 percent was real prosperity, real houses, real jobs, real wealth accumulation, and much of it happened in the suburbs the highways opened. The suburbanization finding does not deny that prosperity. It says the prosperity was spatially reorganized by a federal program, and that the reorganization had losers as well as winners. The winners built equity in suburban homes. The losers watched their neighborhoods divided by concrete or were displaced outright, topics Parts Six and Seven take up in detail. The same federal dollars financed both outcomes. The ledger records both.

The mechanics of Baum-Snow’s comparison deserve one more layer of explanation, because the credibility of the 18 percent figure rests entirely on the instrument’s exogeneity. The study compares cities that the 1944 plan marked for central-city highways against cities it did not mark, and asks whether the marked cities lost more central-city population between 1950 and 1990. For this comparison to identify causation, the marking must be unrelated to the cities’ underlying propensity to suburbanize. The plan’s 1944 vintage is what secures that: drawn during wartime for military and intercity purposes, it reflected the strategic thinking of an earlier era, not the housing economics of the postwar decades. A critic might object that the planners of 1944 favored important cities, and important cities suburbanized more for other reasons. Baum-Snow’s specifications address such concerns by controlling for the observable correlates of importance, and the result survives. The instrument is not beyond criticism, no instrument is, but it is far stronger than the naive comparison of highway cities against non-highway cities, where the decision to build is entangled with everything else about the city.

What does an 18 percent population decline actually describe?

It describes the average causal effect of one highway through a central city, measured across American cities from 1950 to 1990. Without the network, Baum-Snow estimates central cities would have grown 8 percent; instead they declined 17 percent while their metros grew 72 percent. The figure captures hollowing, not regional shrinkage.

Part Five: The Congestion Finding

The most counterintuitive of the five findings is also the most robustly documented. Gilles Duranton and Matthew A. Turner’s 2011 paper in the American Economic Review, “The Fundamental Law of Road Congestion: Evidence from US Cities,” published in volume 101, issue 6, established what the title promises: a lawlike regularity in the relationship between road capacity and traffic. Their central estimate is that vehicle-kilometers traveled increase proportionately to roadway lane kilometers for interstate highways. In plain language, add ten percent more lane capacity and you get about ten percent more driving. The elasticity is roughly one, one-for-one. The road creates the traffic that fills it.

The finding deserves the name “fundamental law” because of its consistency across cities, time periods, and specifications. Duranton and Turner did not study one widening project in one city. They studied the cross-section of American cities, relating each city’s stock of interstate lane kilometers to the total vehicle-kilometers traveled on its roads, and they found the proportional relationship holding with remarkable stability. The result survives the obvious objections. It is not driven by a few large cities. It is not an artifact of how traffic is measured. It is not the result of confusing cause and effect, because the authors used the same kind of historical instruments, old plans and old routes, to isolate the causal direction. The roads cause the traffic. The traffic does not merely follow the roads.

Where does the extra traffic come from? The paper identifies three sources, and each is worth naming because each corresponds to a real human decision. First, current residents drive more. When capacity expands and speeds rise, trips that were previously not worth making become worth making: the errand across town, the visit to the distant shopping center, the commute from the farther suburb. Second, commercial traffic increases. Businesses that were constrained by congestion expand their delivery zones, run more trucks, and reorganize their logistics around the new capacity. Third, people move in. The improved accessibility raises the value of land near the expanded corridor, new development follows, and the new residents generate new trips. None of these responses is a market failure. Each is people and firms responding sensibly to a change in conditions. The lawlike aggregate is the sum of sensible individuals.

Two negative findings sharpen the result. Widening one type of road diverts little traffic from other types, which means the new capacity does not relieve the broader network so much as add to total driving. And the authors found no evidence that providing transit reduces vehicle-kilometers traveled, a finding that complicates the favorite alternative of congestion hawks and transit advocates alike. The paper’s scope condition matters here and should be stated exactly: the fundamental law concerns interstate highways in American cities. It is a finding about a specific road type in a specific setting, and its policy implication is about durable congestion relief, not a claim that no travel-time benefit ever occurs from any capacity addition anywhere. A widening can speed up trips for a while. The evidence says the speedup does not last, because the traffic grows to fill the space.

The congestion finding is the empirical backbone of the article’s central claim about the next lane. If vehicle travel rises one-for-one with lane capacity, then the standard political promise of highway expansion, that new lanes will reduce congestion, is a promise the evidence says cannot be kept in any durable sense. The lanes will fill. The congestion will return, now spread across a larger network serving more dispersed development. This does not mean widening never has any value. It means the value cannot be measured in lasting congestion relief, and proposals should be judged on other grounds, such as safety or freight movement or connectivity for underserved areas, rather than on a congestion promise that the fundamental law contradicts.

The finding also closes the loop with suburbanization. The outward movement Baum-Snow documented created metropolitan regions whose residents depended on cars for nearly every trip, and those regions then demanded more lanes to serve the trips the earlier lanes had enabled. The fundamental law describes the second half of that cycle: the new lanes generated new driving, which justified further dispersal, which generated further demand for lanes. The program’s defenders sometimes describe this as success, the network serving a growing nation. Its critics describe it as a treadmill, the network manufacturing the congestion it claims to cure. The evidence supports the treadmill description for durable congestion relief, while leaving open the genuine gains the network delivered in productivity, safety, and freight. Both things at once, again.

There is a temptation, when presenting the fundamental law, to make it sound like a counsel of despair: nothing can be done about traffic, so nothing should be tried. That is a misreading, and the authors’ scope condition is the guardrail against it. The law says that adding lane kilometers to urban interstates generates proportional traffic. It does not say that pricing, land use, transit investment, or demand management cannot change outcomes. It does not say that the original network should not have been built. It says one specific thing about one specific intervention, and it says it with unusual empirical force. The honest use of the finding is to retire one specific promise, durable congestion relief through widening, not to retire the idea that transportation policy can improve lives.

The three sources of induced traffic repay closer attention, because each reveals a different facet of how the network reshaped behavior. The first source, current residents driving more, is the most intuitive and the most democratic: when travel gets cheaper in time, people consume more of it. The trips that appear are not frivolous. They are the job taken in a farther suburb, the specialist doctor visited across the metro, the family member seen more often. Each represents real welfare, real value to the traveler. The fundamental law does not deny that value. It says the value comes at the cost of the congestion relief, because the new trips fill the space the widening created. The welfare gain is real and the congestion gain is illusory, and honest policy analysis counts both.

The second source, commercial traffic, connects the congestion finding to the freight finding of Part Eight. The same just-in-time logistics systems that made the network so productive for freight also made commercial traffic highly responsive to capacity. When lanes were added, distributors expanded delivery zones and ran more trucks, because the economics of their networks improved. This is the productivity story and the congestion story as a single mechanism: the network’s very efficiency at moving goods is what makes goods movement expand to fill new capacity. The trucks that lowered logistics costs from 16 percent to 10 percent of national product are the same trucks that appear in Duranton and Turner’s traffic counts. Success and congestion are joint products.

The third source, in-migration and new development, is the slowest and the most permanent. When a corridor is widened, land near it becomes more valuable, developers build on it, and the new residents generate trips for decades. This is the channel through which a widening project reshapes the metropolis itself, extending the commuter shed outward and locking in car dependence for another generation of development. It is also the channel that links the congestion finding back to Baum-Snow’s suburbanization finding: the highways first pulled population outward, and each subsequent widening ratified and extended the dispersal. The fundamental law is not just a fact about traffic. It is a fact about how transportation investment shapes the physical form of cities over half-centuries.

The negative findings deserve a final word, because they close off the easy escapes. That widening one road type diverts little traffic from others means the familiar hope, that a new bypass will drain the old arterial, is mostly misplaced; the bypass generates its own traffic. That transit provision showed no measurable reduction in vehicle-kilometers traveled complicates the standard alternative, though the authors’ scope and methods leave room for debate about what kinds of transit, in what settings, might do better. The point of rehearsing these negatives is not to foreclose policy imagination but to discipline it: whatever the alternative to widening is, it must grapple with the one-for-one regularity rather than wishing it away.

The scope condition deserves to be lingered on, because scope conditions are where honest findings are most often abused. Duranton and Turner studied interstate highways in American cities. They did not study rural interstates, where induced traffic behaves differently because the surrounding land use is different. They did not study ordinary arterials, where intersections and access points change the dynamics. They did not study the highways of other countries, where pricing, land use, and transit differ. The fundamental law is a finding about a specific facility type in a specific national context, and its policy implication, that widening cannot durably relieve congestion, is correspondingly specific. Anyone who cites the paper against a rural bypass project or a foreign metro is borrowing authority the authors did not claim. The article’s discipline throughout has been to respect such boundaries, and the congestion finding is where the discipline matters most, because the finding is strong enough to tempt overuse.

There is also a subtlety in what “durable” means that is worth making explicit. The paper does not claim that a widening’s benefits vanish overnight. In the months after new lanes open, travelers genuinely experience faster trips; the evidence allows that. What the unit elasticity implies is that the system drifts back toward its prior level of congestion as the three responses, more driving, more commercial traffic, more development, accumulate. The adjustment takes years, not days, which is why the illusion of success persists through ribbon-cuttings and the first few traffic reports. By the time the congestion returns, the widening is old news, and the next widening is proposed to relieve the congestion the last one was supposed to cure. The treadmill turns on a timescale longer than the political cycle, which is why the cycle repeats. Understanding the timescale is part of understanding the law.

Does widening ever buy any travel-time benefit at all?

Sometimes, and temporarily. Duranton and Turner found that new capacity can speed trips in the short run, but the one-for-one growth in vehicle travel erodes the gain. Their scope condition is precise: the fundamental law concerns interstate highways in American cities, and its policy lesson targets durable congestion relief, not every conceivable benefit.

Part Six: The Displacement Finding, Counted

The displacement finding is the one the productivity literature never measured and the statute’s drafters never budgeted for. When the network’s urban segments were driven through American cities in the late 1950s and the 1960s, they destroyed homes on a scale that is difficult to grasp without numbers, and the numbers come from three distinct sources that agree on the order of magnitude while differing on the exact count. Federal estimates, compiled by the Department of Transportation and reported in connection with the Reconnecting Communities program, put the figure at more than 475,000 households and over a million people displaced by construction of the national network. Highway historian Raymond Mohl of the University of Alabama at Birmingham, working from the construction record, estimated at least 330,000 urban housing units destroyed as a direct result of federal highway building between 1957 and 1968, and calculated that federal highway construction in the early 1960s dislocated an average of 32,400 families per year. Political scientist Clayton Nall, in The Road to Inequality published by Cambridge University Press, roughly estimated one million persons displaced, often in poor and ethnically or racially diverse neighborhoods.

The ranges overlap and the differences reflect different counting methods, not different realities. The federal figure counts households and people displaced across the whole program. Mohl’s figure counts urban housing units destroyed in a specific eleven-year window, the peak construction years. Nall’s figure is a rough estimate of persons displaced, emphasizing the demographic concentration. What all three establish is scale: the displacement was not a matter of a few blocks here and there. It was one of the largest episodes of forced residential relocation in American peacetime history, comparable in its demographic concentration to the urban renewal clearances of the same era, with which, as Part Seven will show, it was deeply entangled.

The aggregate numbers become concrete in the documented cases, and three of them are worth naming because they show the pattern in three different regions. In Nashville, the construction of Interstate 40 demolished more than 620 Black homes, cutting through a thriving Black business and residential district. In Miami, the routing through Overtown, then the center of Black life in the city, reduced the neighborhood’s population from about 40,000 residents to about 8,000. In St. Paul, Interstate 94 split the Rondo neighborhood, a prosperous Black community, in two. These are not the only cases; they are the cases the historical record documents most fully, and they illustrate what the aggregate statistics mean on the ground: not abstract “displacement” but specific streets, specific businesses, specific churches and gathering places, gone.

The demographic concentration is the finding’s moral center, and it must be stated with the precision the sources allow. Nall’s formulation is that the displaced were often in poor and ethnically or racially diverse neighborhoods. Mohl’s formulation is blunter: highway planners usually aimed construction at poor, Black neighborhoods, but not in every case. The qualifier matters, and Mohl supplied the counterexamples himself: ethnic neighborhoods in Chicago, Chinatown and the North End in Boston, working-class Jewish communities in New York. The pattern was racial and economic, but it was not exclusively so. What unified the targeted neighborhoods was not a single demographic category but a single political fact: they were the neighborhoods with the least power to resist. Poor Black neighborhoods were the most frequent targets because they combined racial vulnerability with economic vulnerability, but any community without political leverage could find a highway drawn through it.

The scale of the displacement also needs to be set against the residential patterns of the era to be fully understood. The neighborhoods the highways destroyed were, in many cases, the neighborhoods to which Black residents had been confined by decades of discriminatory housing practice, the same practices whose long-run consequences are traced in the companion account of the Fair Housing Act. Segregation had concentrated Black households in specific districts; the highway program then routed its urban segments through those districts, destroying the communities that segregation had created. The two policies were not coordinated in any conspiratorial sense, but they interacted with devastating consistency: one confined, the other cleared.

A final caution belongs with the numbers, and it comes from Nall himself. Interstate construction was, in his words, “a large, but only partial, contributor” to neighborhood destruction, alongside the concurrent practices of urban renewal and redlining. The highways did not act alone. They acted in concert with clearance programs that were already targeting the same neighborhoods, and attributing every lost home to the highway program alone would overstate the case and weaken it. The honest figure is the one the sources support: hundreds of thousands of homes destroyed by highway construction directly, within a larger landscape of neighborhood destruction in which highways were the largest single instrument but not the only one. Part Seven takes up how the routes were chosen and why the pattern took the shape it did.

The federal estimate’s provenance is worth noting, because it shows how the displacement record entered official memory. The figures, more than 475,000 households and over a million people, were compiled by the Department of Transportation and reported in connection with the Reconnecting Communities program, the federal effort to reckon with the highways’ urban damage. That the government’s own numbers now document the scale of the displacement represents a remarkable archival turn: the same department that built the network now publishes the count of those it displaced. The figures are not the product of hostile scholarship. They are the government’s own accounting, belated but official, and they set the upper end of the range that Mohl’s and Nall’s independent estimates corroborate from below.

The documented cases gain power from being set side by side. Nashville’s I-40, demolishing more than 620 Black homes, shows the pattern in a Southern city where a thriving Black business district was treated as expendable. Miami’s Overtown, collapsing from about 40,000 residents to about 8,000, shows the pattern at its most devastating, the near-erasure of a community that had been the center of Black life in its city. St. Paul’s Rondo, split by I-94, shows the pattern in a Northern city, a prosperous Black neighborhood divided by concrete in a way that maps the community’s destruction street by street. Three regions, three cities, one logic: the routes went where resistance was weakest and land was cheapest, and the residents who paid were those whom segregation had already made vulnerable.

That vulnerability had a history, and the history matters for understanding why the displacement concentrated as it did. The neighborhoods the highways targeted were not random collections of cheap land. They were, in many cases, the districts to which Black residents had been confined by decades of discriminatory practice: restrictive covenants, exclusionary lending, the redlining maps that marked Black neighborhoods as hazardous for investment. Confinement produced density, density produced the appearance of blight, and blight supplied the justification for clearance. The highway engineers who chose the cheapest routes were, whether they understood it or not, harvesting the consequences of segregation: the confined districts were cheap precisely because decades of discrimination had depressed their land values. The program did not invent the vulnerability. It priced it, routed through it, and paved it.

Mohl’s contribution, as a historian rather than an economist, was to reconstruct the destruction from the construction record itself: the plans, the acquisitions, the demolitions, year by year. His figure of at least 330,000 urban housing units destroyed between 1957 and 1968 is built from that paper trail, and his annual figure of 32,400 families dislocated in the early 1960s gives the destruction its tempo, not a single event but a sustained campaign, year after year, neighborhood after neighborhood. Nall’s contribution, as a political scientist, was to place the displacement in its demographic and political context: roughly a million persons, often poor and ethnically or racially diverse, displaced by a program whose route decisions reflected the distribution of political power. The three sources differ in method, archival, statistical, and political, and converge in magnitude. That convergence across methods is what makes the range trustworthy despite its width.

Mohl’s method deserves the same careful exposition this article has given the economists, because archival history has its own identification discipline. Where Fernald and Baum-Snow and Duranton and Turner used statistical instruments to separate cause from correlation, Mohl used the paper trail: the construction records, the right-of-way acquisition files, the planning documents, the federal reports that recorded, year by year, how many housing units were destroyed and how many families were dislocated. The 330,000 urban units between 1957 and 1968, the 32,400 families per year in the early 1960s, are not estimates from a model. They are counts from the records of the program that did the destroying, which kept track because it had to pay for the takings. The method’s strength is its directness: the documents say what happened, in the bureaucracy’s own hand. Its limitation is the one Mohl himself respected: the records count what the program recorded, and a program does not always record the full human cost of what it does. The ranges in this article, federal estimates alongside Mohl’s counts alongside Nall’s rough estimate, reflect that limitation honestly. Precision about the imprecision is the archival virtue.

Part Seven: Who Chose the Routes, and Who Stopped Them

The question of who chose the urban routes has a documented answer, and the answer distributes responsibility across levels of government in a way that defeats simple blame. Under the 1944 Federal-Aid Highway Act, the routes were, in the statute’s phrase, “selected by joint action of the State highway departments.” The Public Roads Administration called on the states to submit their recommendations and worked with state and local officials to develop interstate plans for the larger cities. The federal Bureau of Public Roads then approved the plans. In practice, this meant that state highway departments, working with city officials, drew the lines on the maps, and Washington financed and approved what they drew. The federal government supplied the money and the standards. The states and localities supplied the routes.

That division of labor is the essential context for Mohl’s finding that planners usually aimed construction at poor, Black neighborhoods. The planners in question were overwhelmingly state and local officials, not distant federal bureaucrats. They chose routes through the neighborhoods whose land was cheapest to acquire, whose residents had the least political influence, and whose clearance could be presented as serving two goals at once: building the highway and removing what the era called blight. The cheap-land logic was presented as fiscal prudence. The political logic was rarely stated aloud. Together they produced the pattern the displacement numbers record.

The cheap-land logic deserves scrutiny, because it reveals how fiscal prudence and racial harm could coincide without anyone stating the connection. State highway departments were stewards of public money, and routing through low-value land minimized acquisition costs. On its own terms, the logic was defensible: why pay more for land than necessary? But land values in American cities were not natural facts. They were the product of decades of segregation, disinvestment, and redlining, which had depressed values precisely in the Black neighborhoods the highways then targeted. The cheap-land criterion thus laundered a history of discrimination into a neutral engineering input. The engineers did not need to intend racial harm for the criterion to produce it. They needed only to treat market prices as innocent, and the market prices were anything but.

The entanglement with urban renewal is the second essential context, and it is where the honest accounting must resist attributing everything to the highways alone. The urban segments of the network were frequently routed through neighborhoods that were already targeted for clearance under the renewal programs of the same era, the programs authorized by the Housing Act of 1949. In city after city, the highway and the renewal project were planned together, sometimes by the same officials, sometimes with the highway serving as the renewal project’s boundary or its justification. Nall’s caution applies with full force here: the highway program was a large but only partial contributor to the destruction, operating alongside renewal clearance and the redlining practices that had shaped the neighborhoods in the first place. To say the highways destroyed Black neighborhoods is true. To say the highways alone destroyed them is not.

This layered causation is also why the complication stated in Part One insists on attributing route selection accurately. A purely federal story would blame Washington for routes that state highway departments drew. A purely local story would absolve the federal program that financed and approved those routes and set the standards they followed. The accurate story is that federal money, state engineering, and local politics combined to produce the outcomes, and that each level had opportunities to choose differently. Some did. The counterexamples Mohl documented, the Chicago ethnic neighborhoods, Boston’s Chinatown and North End, New York’s working-class Jewish communities, show that the pattern was a tendency enforced by power, not a law of nature. Where communities had leverage, routes moved.

Mohl’s counterexamples deserve individual attention, because each one illuminates the mechanism by which leverage deflected concrete. In Chicago, ethnic neighborhoods with established political organizations and aldermanic representation managed to redirect or blunt planned routes; their whiteness and their votes gave them a hearing that Black neighborhoods were denied. In Boston, Chinatown and the North End, communities with deep institutional roots, forced planners to look elsewhere. In New York, working-class Jewish communities with strong tenant organizations and labor ties mounted opposition that the highway builders found costly to override. None of these communities escaped entirely unscathed, but none suffered the near-erasure that Overtown did. The pattern across the counterexamples is consistent: where residents could impose political costs, routes moved; where they could not, routes stayed.

The federal role in this layered system needs one more precise statement, because it is easy to misread “state and local officials chose the routes” as exonerating Washington. The federal Bureau approved every plan, which means federal officials reviewed the urban alignments and signed them. Federal standards dictated the engineering that made urban highways so destructive to the neighborhoods they crossed. Federal money, covering the overwhelming share of costs, is what made the program possible at all; without it, no state could have afforded the urban segments. And the federal government set the rules of the game, including the absence of any requirement to count the displaced. To say the states drew the lines is accurate. To say Washington was a passive cashier is not. The layered account distributes responsibility; it does not dissolve it.

And some communities, famously, stopped the highways altogether. The freeway revolts of the 1960s and 1970s are the displacement story’s counternarrative, the evidence that the outcomes were contested and sometimes reversed. In San Francisco, after the Embarcadero Freeway opened and residents saw what an elevated highway did to their waterfront, the Board of Supervisors voted in 1959 to cancel seven of ten planned freeways, the first successful anti-freeway vote in American history. The fight continued into 1964, when the proposed freeway through the Panhandle and Golden Gate Park provoked another round of opposition and another cancellation. In New Orleans, the proposed route through the French Quarter, the Vieux Carre, was blocked after a sustained campaign. These were not isolated episodes. Across the country in the 1960s and 1970s, neighborhood campaigns stopped or redirected urban interstates, redrawing the maps that state highway departments had drawn.

The 1959 San Francisco vote deserves its full weight as a precedent, because first victories teach the tactics that later victories reuse. Before 1959, the assumption in every American city was that the highway plan, once drawn, would be built, and that neighborhood opposition was a delay to be managed rather than a veto to be feared. The Board of Supervisors’ cancellation of seven of ten planned freeways broke that assumption in the city where the freeway program was most advanced, and the news traveled. Organizers in other cities learned the San Francisco playbook: organize before the rights-of-way are bought, build the coalition beyond the immediately threatened blocks, and aim at the most vulnerable layer of the layered approval process. The 1964 Panhandle fight showed the playbook working a second time in the same city, which proved the first victory was not a fluke. By the time the French Quarter route was blocked in New Orleans, the revolt had become a recognized form of urban politics, with its own tactics, its own precedents, and its own win record. The program’s engineers had planned for geology and traffic. They had not planned for democracy.

The successful campaigns shared three features. First, they organized early, before the rights-of-way were acquired and the demolition contracts signed, because once the money was committed the political cost of reversal rose sharply. Second, they built coalitions across the usual divides, joining the directly threatened neighborhoods with the broader constituencies, waterfront users, park defenders, historic preservationists, who had their own reasons to oppose the roads. The French Quarter fight in New Orleans worked because the Vieux Carre mattered to the whole city, not just to its residents. Third, they exploited the layered decision process itself. Because routes required state action, local approval, and federal sign-off, a revolt that captured any one layer could stall the whole, and the federal approval at the end gave opponents a final venue when the state and local layers were captured by the highway interests. The process that had enabled the destruction, layered and diffuse, also multiplied the veto points available to its opponents.

The gaps the revolts left in the network are still visible, and they are worth reading as monuments. Every canceled urban freeway is a place where the map shows a stub, a missing link, a corridor that narrows where the plan said it would widen. Drivers experience these gaps as inconveniences, the traffic that the missing segment would have carried. But the gaps are also the physical record of a political choice: here, the neighborhood won. The assessment should count them on both sides. They represent congestion that was never relieved, which is a cost. They represent neighborhoods that were never cleared, which is a benefit. The fact that both statements are true about the same missing mile is the whole argument of this article in miniature.

The revolts changed engineering practice even where they did not stop construction. After the 1960s, no urban highway could be planned on the old assumption that the route was a technical matter for the engineers alone. Community input, environmental review, public hearings that opponents could actually use: these entered the planning process because the revolts had shown what happened without them. The change was not purely a victory for the neighborhoods. It also made highway building slower, more expensive, and more litigious, which is one reason the later decades produced fewer transformative projects and more incremental ones. But the procedural revolution is inseparable from the substantive one. A process that must listen to the neighborhood cannot route through it as casually as a process that need not listen.

The revolts matter for the article’s argument in two ways. First, they confirm the displacement finding by showing what residents expected the highways to do: people do not mount years-long campaigns against projects they believe will benefit them. The intensity of the opposition is itself evidence of the threat. Second, they qualify the determinism that sometimes creeps into histories of the highway era. The network was not the inevitable product of impersonal forces. It was the product of decisions, and decisions could be reversed when the people affected had enough organization, enough allies, and enough political standing to force a reversal. The tragedy of the displacement record is not that resistance was impossible but that it was unequally available. San Francisco’s supervisors could cancel seven freeways. The residents of Overtown and Rondo could not.

The policy response to the highway era came later, and it belongs in this part as the bridge to the closing argument. By the early 1990s, the construction era was over, the urban damage was documented, and the federal government turned to a different framework. The Intermodal Surface Transportation Efficiency Act of 1991, known as ISTEA, represented the policy response to the era this article assesses: a recognition, written into law, that the age of urban highway building had ended and that transportation policy had to answer to a broader set of aims. ISTEA did not undo the displacement or unbuild the highways. But it marked the moment when the federal government acknowledged, institutionally, that the ledger had two halves. The 1956 paradigm had asked how to build the most roads the fastest. The 1991 paradigm asked what transportation system the country wanted, with highways as one element among others, and with air quality, community consent, and multimodal choice as legitimate constraints on the building. The reauthorizations absorbed the revolts’ lessons into law.

Why did so many urban routes run through Black neighborhoods?

Because state and local officials chose routes where land was cheapest and political resistance weakest, and segregation had concentrated Black households in precisely those districts. Federal money financed the choices; state highway departments drew them. Mohl documents the pattern and its exceptions, which prove the routes reflected power, not engineering necessity.

Part Eight: Safety Gained, Freight Transformed

The ledger’s positive half does not end with productivity. Two more findings belong on it, and both are documented in official sources rather than academic journals: the safety record and the freight transformation. They are worth presenting together because they share a common root in the engineering of the network itself. Limited-access design, grade separation, the elimination of cross traffic and at-grade intersections, changed both how safely people could move and how cheaply goods could move. The same concrete served both ends.

The safety finding comes from the Federal Highway Administration. In the agency’s “Interstate Frequently Asked Questions,” the fatality rate on the national network in 2004 was 0.8 deaths per 100 million vehicle-miles traveled, compared with 1.46 for all roads. A 2000 FHWA analysis put the network rate at 0.85, a little more than one-half the rate on all highway systems, which stood at 1.53. The agency’s Conditions and Performance reports break the comparison down further: urban interstates recorded 0.45 fatalities per 100 million vehicle-miles against 1.06 for urban principal arterials and local roads, while rural interstates recorded 0.76 against 2.40 for rural local roads. The usable summary is that fatality rates on the limited-access network run roughly half the national all-roads average, with the gap widest against rural local roads, the two-lane country roads that the network replaced for long-distance travel.

The numbers deserve a moment of reflection, because they measure something the displacement story can obscure: the program saved lives, in large numbers, over decades. Every percentage point of traffic shifted from a rural two-lane road to a divided, limited-access highway represented a measurable reduction in the probability of death. The 2.40 rate on rural local roads against the 0.76 rate on rural interstates is not an abstraction. It is the difference between the road that existed and the road that replaced it for through travel, compounded over billions of vehicle-miles and half a century. An honest ledger records this alongside the homes destroyed, not as a justification but as a fact. The program that displaced a million people also cut the death rate for long-distance travel roughly in half.

The design features behind the safety numbers are worth naming, because they explain why the gap is structural rather than behavioral. A limited-access road removes the four most dangerous elements of ordinary driving: intersections, where crossing traffic collides; stoplights, where rear-end crashes cluster; oncoming lanes, where head-on collisions kill; and sharp curves, where run-off-road crashes happen. Grade separation, medians, gentle geometry, and controlled entry do the work that driver education campaigns can only wish for. The driver on the network is not necessarily more careful than the driver on the arterial. The road simply offers fewer ways to die. That is why the FHWA numbers are consistent across reporting years, 0.85 against 1.53 in the 2000 analysis and 0.8 against 1.46 in 2004. A behavioral effect would fluctuate with enforcement, vehicle fleets, and culture. A design effect persists, because the concrete does not change its mind.

The design-versus-behavior distinction carries a broader lesson about what kinds of safety interventions work. Behavioral interventions, enforcement campaigns, education programs, awareness months, ask drivers to be more careful, and drivers, being human, oblige for a while and then revert. Design interventions change what carefulness is required, and the concrete does not revert. The network’s safety record is almost entirely a design story: grade separation, medians, controlled access, forgiving geometry. The FHWA numbers are consistent across years precisely because they measure the road, not the driver. This is why the safety finding generalizes while the congestion finding disappoints. Congestion is behavioral, driven by travelers’ responses to price, and it defeats engineering. Safety per mile is physical, driven by the road’s geometry, and engineering masters it. The program that could not build its way out of traffic could build its way out of a large share of road death, and the difference between the two outcomes is the difference between trying to change travelers and changing the road.

The rural comparison carries the deepest historical meaning. Before the network, long-distance travel meant two-lane rural roads, narrow shoulders, farm equipment sharing the lane, and head-on exposure for every passing maneuver. The 2.40 fatality rate on rural local roads, against 0.76 on rural segments of the new network, measures what the replacement was worth on the corridors where Americans actually drove between cities. The network did not merely add a safer option alongside the old roads. It drained the old roads of their through traffic, which means the safety gain compounded: the dangerous roads got less dangerous because they got less crowded, while the new roads were safer by design. This is the rare case in the assessment where the program’s logic worked exactly as intended, with no perverse feedback. Unlike congestion, where new capacity induces new demand, safety per mile does not induce riskier driving on anything like the same scale. The gain held.

The rate-versus-aggregate distinction is the necessary discipline on the celebration. A fatality rate is deaths per mile, and the network roughly halved it. But total deaths are the rate times the miles, and the network multiplied the miles enormously, through the suburbanization of Part Four and the induced travel of Part Five. The arithmetic does not erase the achievement. Halving the rate while tripling the miles still leaves more total deaths than the old rate on the old miles, but it leaves far fewer than the old rate on the new miles, which is the correct counterfactual. The question is not whether the network killed fewer people than a world with no roads. It is whether it killed fewer than a world with the same travel on the old roads, and on that comparison the design advantage is decisive. Safety belongs on the aims side of the ledger without apology, and it belongs there with the qualification stated plainly, because the evidence article qualifies everything.

The urban safety comparison is the one with the most direct policy bite, and it deserves emphasis for that reason. Urban segments of the network recorded 0.45 fatalities per 100 million vehicle miles, against 1.06 for urban principal arterials and local roads. The city street, with its intersections, its turning traffic, its pedestrians and cyclists, its parked cars narrowing the lanes, is where most urban driving happens and where most urban dying happens. The limited-access road through the same city removes the driver from that environment entirely: no cross traffic, no pedestrians, no parking maneuvers, nothing but the flow. The more-than-halving of the fatality rate is the measure of what the street environment costs in lives, and it is an argument, on safety grounds alone, for separating through traffic from local circulation wherever the volumes justify it. The urban highways that destroyed neighborhoods, documented in Parts Six and Seven, are the same urban highways that saved lives per mile, documented here. The assessment does not resolve that contradiction. It reports it, because the contradiction is the truth.

The correct counterfactual, stated one final time with full precision, is this: compare the deaths on the network’s miles against the deaths those same miles would have produced on the roads the network replaced. On that comparison, the design advantage is decisive, roughly half the rate, wider against the rural locals. The incorrect counterfactual, the one that sometimes smuggles itself into the debate, compares the network’s total deaths against a world with less travel altogether, and finds the network wanting because it induced the travel. But the induced travel is the suburbanization of Part Four and the freight of the next section, the program working as designed, and a safety assessment cannot penalize the program for its own aims succeeding. Rate against rate, design against design: the network halved the dying per mile. That is the finding, bounded and durable.

The freight finding is documented in the congressional record marking the program’s fiftieth anniversary, published on congress.gov under the title “Celebrating 50 Years.” The record states that “the interstates facilitated a fundamental transformation of our freight and distribution system,” and it attaches numbers to the transformation. Truck utilization soared almost 200-fold in the fifty years after the program began. Virtually every item in American workplaces and households, in the record’s phrase, “has reached us via the interstate system.” The shift to truck-based distribution gave the United States, in the record’s assessment, “the world’s most efficient supply chain management system,” with total logistics costs declining from 16 percent of national product in 1980 to 10 percent circa 2001. The Federal Highway Administration’s 2006 Conditions and Performance report adds the mechanism: trucking growth was stimulated by just-in-time deliveries and the decentralization of business, both of which the network made economical.

The logistics-cost figure is the freight finding’s equivalent of Fernald’s productivity result: a single number that captures a transformation. When moving goods costs 16 percent of national product and falls to 10 percent, the difference, six percent of everything the economy produces, is the dividend of the distribution revolution. It showed up in lower consumer prices, in the viability of national retail chains supplied from centralized distribution centers, in the just-in-time manufacturing systems that treated the highway network as a rolling warehouse. The 200-fold increase in truck utilization is the physical counterpart: the network did not merely carry the freight that already existed but enabled a volume and pattern of freight movement that had been uneconomical before.

The freight finding also connects back to the productivity finding in a way that strengthens both. Fernald found that roads raised productivity disproportionately in vehicle-intensive industries. The freight transformation is what that disproportion looked like in practice: the industries that lived by moving things reorganized themselves around the new network, and the reorganization was productive. The just-in-time systems the FHWA described are the firm-level mechanism behind the industry-level productivity gains Fernald measured. The two findings, one from econometrics and one from the congressional record, corroborate each other.

There is an irony in the freight story that the article should not leave unstated. The trucks that made the supply chain the world’s most efficient also contributed to the congestion that the fundamental law describes, and to the wear on the pavements that later generations would be asked to maintain. The program’s successes generated its own maintenance burden and its own traffic. This is not an argument against the program. It is an observation about infrastructure: every network creates the conditions for its own congestion and its own decay, and the honest assessment counts those as costs of success rather than as evidence of failure. The ledger has two halves, and some entries appear on both.

The just-in-time mechanism, identified by the FHWA’s 2006 Conditions and Performance report, is the thread that ties the freight story to the productivity story and the congestion story at once. Just-in-time delivery treats the highway network as a rolling warehouse: parts arrive as needed, inventories shrink, capital is freed for other uses. That is a productivity gain, and Fernald measured it. It is also a traffic generator, because the rolling warehouse runs on trucks, and Duranton and Turner counted those trucks. The same mechanism appears on both halves of the ledger, productive and congesting, which is why the article keeps insisting that the findings interlock. The network did not have separate freight effects and traffic effects. It had one effect, the reorganization of movement, with both consequences.

The congressional record’s language deserves to be weighed as evidence rather than dismissed as celebration. “The interstates facilitated a fundamental transformation of our freight and distribution system” is a claim about causation, and the numbers attached to it, the 200-fold rise in truck utilization, the 16-to-10 percent fall in logistics costs, the “virtually every item” that reached its destination via the network, give the claim empirical content. Congress had institutional reasons to celebrate the program at its fiftieth anniversary, and a skeptical reader is right to discount the adjectives. But the quantities are harder to discount: they describe a reorganization of American economic life so complete that its results are invisible, priced into every good on every shelf. The freight finding’s strength is that it does not depend on the record’s praise. It depends on its numbers, and the numbers are large.

The falling logistics share, from 16 percent of national product in 1980 to 10 percent circa 2001, is the consumer’s-eye view of the same transformation, and it translates into the most democratic of the program’s achievements. Logistics costs are embedded in the price of everything: the food, the clothing, the appliances, the materials. When the logistics share falls by six percentage points, the prices of goods fall relative to what they would have been, and the savings accrue to every household, including the households the program’s construction displaced. This is the uncomfortable arithmetic the assessment has to hold. The same network that demolished Overtown and bisected Rondo also cheapened the goods on every American shelf, and the cheapening is not a small thing. It is one of the mechanisms by which the postwar economy raised living standards.

It is worth being explicit about what these two findings do not do, because their strength tempts misuse. The safety gains do not justify the displacement; lives saved on rural highways do not compensate families whose homes were demolished, and no one in this article’s evidence base claims they do. The freight gains do not justify the routing decisions; cheaper logistics do not excuse the targeting of Black neighborhoods, and the congressional record that celebrates the supply chain does not mention Overtown. The findings are entries in a ledger, not arguments in a trial. Their function is to ensure that the assessment of the program is complete, not to render a verdict of innocent or guilty.

How did the network change what freight meant in the American economy?

It made truck-based distribution the default, enabling just-in-time delivery and centralized warehousing on a continental scale. Truck utilization rose almost 200-fold in fifty years, and total logistics costs fell from 16 percent of national product in 1980 to 10 percent circa 2001, the dividend of the distribution revolution the network enabled.

Part Nine: The Five-Finding Evidence Table

The parts above have presented the evidence as a narrative arc: the methods that made causation credible, then productivity, suburbanization, congestion, displacement counted, route selection and resistance, safety and freight. The table below compresses the same evidence into a single view, so that the pattern across findings is visible at a glance. Each row carries the outcome, its direction, the period it covers, the principal research and the identification strategy behind it, and the current state of scholarly agreement. Readers who want to test the article’s claims should start here, because the table makes the sources checkable in a way that prose, however careful, cannot quite match.

Outcome	Direction	Period	Principal research and identification strategy	Settled or contested
Productivity	Sharply up for the first network; small at the margin	1950s to 1960s build-out; study published 1999	John G. Fernald, American Economic Review 89(3); cross-industry comparison of public capital and productivity, gains concentrated in vehicle-intensive industries	Settled that the first network was highly productive; debated how much of the later slowdown its completion explains
Suburbanization	Central-city populations fell substantially	1950 to 1990; study published 2007	Nathaniel Baum-Snow, Quarterly Journal of Economics 122(2); 1944 Interstate Highway System Plan as instrument for actual construction	Largely settled on direction and causation; exact magnitude varies by specification
Congestion	Vehicle travel rises one-for-one with lane capacity	Cross-section of US cities; study published 2011	Gilles Duranton and Matthew A. Turner, American Economic Review 101(6); estimated unit elasticity of vehicle-kilometers traveled to lane kilometers	Settled empirically for interstate highways in cities; scope limited to durable congestion relief
Displacement	Massive displacement, concentrated in poor and Black neighborhoods	1957 to 1968 peak construction; estimates compiled across decades	Federal DOT estimates; Raymond Mohl, highway historian; Clayton Nall, The Road to Inequality (Cambridge UP); archival and demographic reconstruction	Range unsettled (330,000 urban units to over a million persons); concentration pattern well documented
Safety and freight	Fatality rates roughly halved; freight costs and logistics transformed	1980 to 2001 for logistics costs; 2000 to 2004 for fatality rates	FHWA fatality-rate comparisons and Conditions and Performance reports; congressional 50th-anniversary record on freight	Settled on the direction of both effects; magnitudes documented by official sources

Three patterns in the table deserve explicit notice. First, the academic findings, productivity, suburbanization, and congestion, all rest on identification strategies designed to separate causation from correlation, and all three survived the scrutiny of the profession’s leading journals. The official findings, displacement, safety, and freight, rest on government statistics and historical reconstruction, and they are strongest where multiple sources converge. Second, the periods differ, and the differences matter: the productivity gains belong to the build-out decades, the congestion regularity is a standing feature of the mature network, and the displacement belongs to the peak construction years when urban segments were driven through cities. Third, the contested column is honest about what remains open. The displacement range is wide because counting the displaced depends on definitions. The suburbanization magnitude varies because instruments are imperfect. The congestion finding’s scope is limited because its authors limited it. An evidence table that hid these qualifications would be propaganda. This one is meant to be usable.

A fourth pattern, visible only when the rows are read together, is the one this article has been building toward. The first two rows, productivity and suburbanization, describe the network’s transformative decades: the economy remade, the metropolis reorganized. The third row, congestion, describes the mature network’s equilibrium: capacity generating its own demand, the transformative promise reduced to a treadmill. The fourth row, displacement, describes the cost exacted during the transformation, concentrated on those least able to refuse it. The fifth row, safety and freight, describes what the transformation durably delivered, the gains that persist regardless of how the other rows are judged. Read downward, the table is a life cycle: build, transform, congest, displace, deliver. The program’s whole history is in those five verbs, and the table’s virtue is that it refuses to let any one of them stand for the whole.

The table also serves a purpose the prose cannot: it makes the article checkable. Every row names its researchers or its sources, its period, and its method. A skeptical reader can take any row to a library and verify it, qualify it, or argue with it, and the article will have done its job if the argument is about the evidence rather than about unattributed assertions. That checkability is the standard the series sets for itself, and it is why the table sits where it sits: after the narrative has made the case, before the closing states the verdict, the evidence stands on its own, in rows, ready for inspection.

There is one more use for the table that the article should make explicit. It is a template for assessing any large public program, not just this one. List the outcomes, state their direction, date them, name the research and the identification strategy, and record the state of agreement. The discipline the table imposes, no unattributed claims, no missing periods, no blurred methods, is the discipline the series asks of every statute it assesses. The highway program is an unusually well-studied case, which is why its table is so full. Other statutes will have thinner rows and wider contested columns. The method does not change. Name the finding, name the researcher, name the period, and hold both halves of the ledger at once, whatever the ledger contains.

Part Ten: The First Network and the Next Lane

The article opened with a claim about what a reader should carry away, and it is time to state it in its final form. The economic literature supports large gains from building a network that did not exist and small gains from widening one that does. High returns to the first network, low returns to the next lane. Fernald established the first half: the system was highly productive, and a second one would not be. Duranton and Turner established the corollary: the next lane generates the traffic that fills it, so the congestion relief it promises does not last. Together they describe an asymmetry that should govern every contemporary debate about expanding the existing network. The evidence that justifies the original program is the same evidence that undercuts most proposals to widen it.

That asymmetry is the practical payoff, but the moral payoff is the both-things-at-once conclusion, and it needs to be stated without flinching. The initial network build-out generated large measured economic returns: the productivity boost Fernald documented, the freight transformation the congressional record celebrated, the safety gains the fatality rates demonstrate. The urban segments of that same build-out destroyed hundreds of thousands of homes, displaced over a million people, and concentrated the destruction in poor and Black neighborhoods whose residents had been confined there by segregation and whose political leverage was weakest. These are not competing narratives to be chosen between. They are two true descriptions of one program, and the series this article belongs to exists precisely for statutes like this one: laws whose consequences their drafters never measured, assessed against both their stated aims and the damage they did.

The complication the article promised to handle was the pair of reductive accounts, and the evidence now shows why both fail. The pure-triumph account cannot survive the displacement numbers, the route-selection history, or Mohl’s documentation of where the highways were aimed. The pure-destruction account cannot survive the productivity findings, the safety comparisons, or the fact that state and local officials, not only federal planners, chose the urban routes, which means the destruction was a product of layered American governance rather than a single malevolent design. The accurate account is harder to hold and more useful: a federal program, financed by Washington, routed by states and localities, that made the American economy markedly more productive and American cities markedly more divided, that saved lives on the open road and destroyed neighborhoods in the cities, that caused suburbanization and then could not pave its way out of the traffic suburbanization created.

For the policymaker reader, the two-purchases framing translates into a decision rule. When a new highway proposal arrives, the first question is which purchase it represents: the first network or the next lane. If the proposal connects the unconnected, serves a corridor with no limited-access alternative, and promises the transformative gains of the build era, then the 1956 evidence is relevant and the average returns apply. If the proposal widens the existing, adds capacity to a corridor the network already serves, and promises congestion relief, then the margin evidence is relevant, and Fernald’s finding and Duranton and Turner’s law apply. Most contemporary proposals are the second purchase wearing the first purchase’s rhetoric, and the decision rule strips the disguise. The rule does not say never build. It says: know which purchase you are making, cite the evidence for that purchase rather than the other, and present the congestion claim with the fundamental law attached. That is what evidence-based transportation policy looks like.

There is a final question the evidence raises but cannot answer, and the article should name it rather than dodge it. Knowing what the program did, what should the country do with the network now? The evidence constrains the answer without dictating it. It says the next lane will not pay for itself in productivity and will not durably relieve congestion. It says the neighborhoods the highways damaged are still there, in many cases still divided by the concrete. It says the safety and freight gains are real and worth preserving. Beyond those constraints, the choices belong to politics, not to research. The program’s fiftieth anniversary came and went with celebrations of the freight revolution. Its sixtieth year might be better spent on the other half of the ledger.

The namable claim deserves a final formulation, sharpened by everything the evidence has shown. High returns to the first network, low returns to the next lane. The first clause rests on Fernald: the 1950s and 1960s build-out delivered a one-time, unrepeatable productivity boost, concentrated in the vehicle-intensive industries, large enough to help explain the slowdown that followed its exhaustion. The second clause rests on Fernald and on Duranton and Turner together: at the margin, road investment does not look unusually productive, and new lane capacity generates proportional traffic, so the two standard justifications for widening, productivity and congestion relief, are the two justifications the evidence specifically undercuts. What remains for widening proposals are the narrower grounds: safety improvements on dangerous corridors, freight bottlenecks that genuinely constrain commerce, connections for communities the network bypassed. Those are real aims, and they can justify real projects. They are simply not the transformative aims, and they should not be sold with transformative rhetoric.

Readers who want to work through the five findings against the statute’s own stated aims, and to test each claim against its named source, may find a companion study notebook useful for organizing the comparison. The method of this article is portable: name the finding, name the researcher, name the period, and hold both halves of the ledger at once. Applied to any statute, that method disciplines memory. Applied to the highway program, it yields the verdict this article has argued for: the first network earned its returns, the next lane will not earn its, and the neighborhoods the concrete cleared deserve to be counted alongside the productivity it created.

The series this article belongs to will go on to assess other statutes by the same double standard: what the drafters aimed at, and what they never measured. The highway program is an ideal case for that method precisely because the two standards diverge so sharply. Against its own aims, the 1956 act is a triumph: the network was built, the country connected, the freight transformed, the productivity gained. Against the unmeasured consequences, it is a catastrophe: the neighborhoods cleared, the cities hollowed, the million displaced. Most statutes will not diverge so dramatically. But the method does not depend on drama. It depends on the discipline of counting what was never counted, and the highway program teaches that discipline in its starkest form. What gets measured gets remembered. What never gets measured gets paved.

Frequently Asked Questions

Q: Did the interstate system cause suburbanization?

Yes, according to the strongest available causal evidence. Nathaniel Baum-Snow’s 2007 paper in the Quarterly Journal of Economics used the 1944 Interstate Highway System Plan as an instrument for actual construction, a plan drawn twelve years before the 1956 act and before the suburban boom. He found that each new highway passing through a central city reduced its population by about 18 percent. His counterfactual implies central-city population would have grown about 8 percent between 1950 and 1990 without the network, against an observed decline of 17 percent while metropolitan areas grew 72 percent. The highways caused a large share of suburbanization rather than merely accommodating it.

Q: How many people were displaced by the interstate system?

Three sources agree on the order of magnitude while differing on exact counts. Federal Department of Transportation estimates, reported in connection with the Reconnecting Communities program, put displacement at more than 475,000 households and over a million people. Highway historian Raymond Mohl estimated at least 330,000 urban housing units destroyed as a direct result of federal highway building between 1957 and 1968, with early 1960s construction dislocating an average of 32,400 families per year. Political scientist Clayton Nall roughly estimated one million persons displaced, often in poor and ethnically or racially diverse neighborhoods.

Q: Did the interstate system destroy Black neighborhoods?

It destroyed neighborhoods in which Black residents were heavily concentrated, though the pattern was not exclusively racial. Raymond Mohl found that highway planners usually aimed construction at poor, Black neighborhoods, but not in every case, documenting counterexamples including Chicago ethnic neighborhoods, Boston’s Chinatown and North End, and New York working-class Jewish communities. Documented cases include Nashville, where I-40 demolished more than 620 Black homes, Miami’s Overtown, reduced from about 40,000 to 8,000 residents, and St. Paul’s Rondo, split by I-94. Clayton Nall cautions that highways were a large but only partial contributor alongside urban renewal and redlining.

Q: Did the interstate system raise productivity?

Yes, and the gains were large but one-time. John G. Fernald’s 1999 paper in the American Economic Review found that roads are productive and raise productivity disproportionately in vehicle-intensive industries, but that at the margin road investments do not appear unusually productive. His summary sentence is the key: “Intuitively, the interstate system was highly productive, but a second one would not be.” The productivity boost belonged to the 1950s and 1960s build-out, when connecting previously unconnected regions transformed what industry could do. Fernald argues the exhaustion of that boost helps explain the productivity slowdown that followed.

Q: Does adding lanes to the interstate system reduce congestion?

Not durably, according to the fundamental law of road congestion. Gilles Duranton and Matthew A. Turner’s 2011 paper in the American Economic Review found that vehicle-kilometers traveled increase proportionately to roadway lane kilometers for interstate highways, a roughly one-for-one relationship. New capacity generates new driving through three channels: current residents drive more, commercial traffic increases, and people move into the more accessible area. Widening one road type diverts little traffic from others, and the authors found no evidence that transit provision reduces vehicle travel. Short-run speed gains occur but erode as traffic grows to fill the new space.

Q: Is the interstate system safer than other roads?

Yes, by a wide margin. The Federal Highway Administration reported a 2004 fatality rate of 0.8 deaths per 100 million vehicle-miles traveled on the network versus 1.46 for all roads. A 2000 FHWA analysis put the network rate at 0.85, a little more than one-half the all-systems rate of 1.53. The detailed breakdowns show where the advantage is largest: urban interstates at 0.45 versus 1.06 for urban principal arterials and local roads, and rural interstates at 0.76 versus 2.40 for rural local roads. Limited-access design, grade separation, and the elimination of cross traffic account for the difference.

Q: What did freeway revolts do to the interstate system?

Neighborhood campaigns stopped or redirected urban interstates across the country in the 1960s and 1970s. In San Francisco, after the Embarcadero Freeway opened, the Board of Supervisors voted in 1959 to cancel seven of ten planned freeways, the first successful anti-freeway vote in American history, and the 1964 fight over the Panhandle and Golden Gate Park freeway produced another cancellation. In New Orleans, the proposed route through the French Quarter was blocked. The revolts redrew planned maps, proved that resistance could succeed where communities had political leverage, and marked the beginning of the end of the urban highway-building era.

Q: How did the interstate system change freight?

It enabled a fundamental transformation of freight and distribution. The congressional fiftieth-anniversary record states that truck utilization soared almost 200-fold in the fifty years after the program began, and that virtually every item in American workplaces and households has reached its destination via the network. The shift to truck-based distribution gave the United States what the record calls the world’s most efficient supply chain management system. Total logistics costs declined from 16 percent of national product in 1980 to 10 percent circa 2001. The FHWA’s 2006 Conditions and Performance report identifies just-in-time delivery and business decentralization as the mechanisms.

Q: What three aims did the drafters state for the interstate program?

Public Law 84-627, the Federal-Aid Highway Act of 1956, stated three aims for the national limited-access network: national defense, interstate commerce, and the connection of the country’s principal cities. Defense came first in the law’s framing and supplied the urgency that carried the program through multiple Congresses, making it national, federal, and fast. Commerce supplied the economic logic: faster, cheaper movement of goods and people across state lines. Connection supplied the geographic logic: linking the principal cities into one system. Each aim seeded consequences the drafters never measured. The defense rationale justified a scale that civilian traffic then filled. The commerce aim shaped limited-access engineering that served commuters as well as trucks, manufacturing the congestion the next generation tried to pave away. The connection aim produced the through-city alignments that demolished urban neighborhoods. The aims were also in tension: defense wanted redundant capacity, commerce wanted efficiency, and connection wanted reach.

Q: How was interstate construction divided between federal and state governments?

The program combined federal financing with state and local route decisions. Under the 1944 Federal-Aid Highway Act, routes were selected by joint action of the state highway departments. The Public Roads Administration called on states to submit recommendations, worked with state and local officials to develop plans for the larger cities, and the federal Bureau approved those plans. In practice, state highway departments chose the actual alignments in consultation with city officials, while Washington supplied the funding and the design standards. That division explains why responsibility for the displacement record is layered: federal money enabled what state and local officials drew, and no single level of government can claim either sole credit or clean hands.

Q: Why did interstate construction continue past the point of high returns?

Because the financing machine made stopping harder than continuing. Congress paid for construction through a dedicated federal trust fund fed by motor-fuel taxes, which insulated the multi-decade build from year-to-year appropriations politics and converted a political choice into an apparently technical one: the money was there, earmarked, so the building continued. State highway departments, which selected the routes in partnership with local officials, had every incentive to keep drawing on the fund, since the federal share was overwhelming and the state match modest. The result was a race to build that continued through the 1960s and 1970s, long after the original defense rationale had faded into ceremony, into the decades when marginal returns were thin and neighborhood costs were fully documented. The successive authorizations that extended Public Law 84-627 were each presented as completing the original vision rather than as new choices. The trust fund financed both the one-time productivity gains of the build years and the costs that accumulated as construction continued past the transformation.

Q: How did the 1944 Interstate Highway System Plan help researchers measure causation?

The 1944 plan gave economists an instrument, a source of variation in highway construction decided before the outcomes it would explain. Nathaniel Baum-Snow used the plan’s proposed corridors, drawn for military and intercity purposes more than a decade before the 1956 act, as exogenous variation for which cities actually got highways through their centers. Because the plan predated the postwar suburban boom, it could not have been drawn in response to suburbanization. Cities the plan favored lost central-city population relative to cities it did not, which is what lets researchers say the highways caused suburbanization rather than merely following it.

Q: What did John Fernald find about marginal returns to road investment?

Fernald found that roads were highly productive when the network was being built and ordinary at the margin afterward. His 1999 American Economic Review paper showed that public road capital raised productivity disproportionately in vehicle-intensive industries during the 1950s and 1960s build-out. But additional investment, once the network existed, did not look unusually productive compared with other uses of capital. His conclusion was that the original system delivered a one-time, unrepeatable boost that helps explain the later productivity slowdown. The policy implication is direct: the measured returns to the first network cannot be borrowed to justify the next lane, because the returns belonged to the act of connecting, not to the act of widening.

Q: What does Baum-Snow’s 18 percent population finding mean?

It means that, on average across American cities, one new highway passing through a central city caused that city’s population to fall by about 18 percent between 1950 and 1990. The figure is a causal estimate, not a raw correlation, because Baum-Snow instrumented actual construction with the 1944 plan. His counterfactual puts the result in context: without the highway network, aggregate central-city population would have grown about 8 percent over those four decades. Instead it declined 17 percent, even as metropolitan areas as a whole grew 72 percent. The 18 percent figure captures the hollowing of urban cores, driven both by the pull of newly accessible suburban land and the push of noise, pollution, and physical division in neighborhoods the highways crossed.

Q: What is the fundamental law of road congestion?

It is the empirical regularity, established by Gilles Duranton and Matthew A. Turner in their 2011 American Economic Review paper, that vehicle-kilometers traveled increase proportionately to roadway lane kilometers for interstate highways. The elasticity is roughly one: add ten percent more lane capacity and total driving rises by about ten percent. The new traffic comes from current residents driving more, increased commercial traffic, and in-migration to the more accessible corridor. The authors also found that widening one road type diverts little traffic from others and that transit provision showed no measurable reduction in vehicle travel. The law’s scope is specific to interstate highways in American cities, and its policy lesson concerns durable congestion relief rather than short-run effects.

Q: Why were the old rural roads so much more dangerous than the new interstates?

Before the network, long-distance travel meant two-lane rural roads with narrow shoulders, at-grade intersections, uncontrolled access, and opposing traffic separated by a painted line. The Federal Highway Administration’s data put the fatality rate on rural local roads at 2.40 deaths per 100 million vehicle-miles, against 0.76 on rural segments of the new limited-access network, a reduction of more than two-thirds on exactly the corridors where the danger was greatest. The network did not merely add a safer option alongside the old roads. It drained them of their through traffic, which means the safety gain compounded: the dangerous roads got less dangerous because they got less crowded, while the new roads were safer by design. Limited-access geometry, grade separation, medians, gentle curves, and the elimination of intersections, stoplights, and head-on traffic account for the gap. The driver on the new road is not necessarily more careful. The road simply offers fewer ways to die.

Q: What does Raymond Mohl’s research say about how urban routes were selected?

Mohl documented that state highway departments, working with local officials, drew the urban routes under a process in which the Public Roads Administration solicited state recommendations, collaborated with state and local officials on city plans, and the federal Bureau approved them. His central finding on targeting is quoted exactly: highway planners usually aimed construction at poor, Black neighborhoods, but not in every case. He supplied the counterexamples himself, including Chicago ethnic neighborhoods, Boston’s Chinatown and North End, and New York working-class Jewish communities. His work also quantified the destruction: at least 330,000 urban housing units destroyed between 1957 and 1968, with 32,400 families dislocated per year in the early 1960s.

Q: How did urban renewal programs interact with interstate construction?

The two programs were deeply entangled, which is why honest accounting does not attribute all neighborhood destruction to highways alone. Urban segments of the network were frequently routed through neighborhoods already targeted for clearance under the renewal programs authorized by the Housing Act of 1949, sometimes planned by the same officials and sometimes with the highway serving as the renewal project’s boundary. Clayton Nall cautions that interstate construction was a large but only partial contributor to neighborhood destruction alongside urban renewal and redlining. Segregation had concentrated Black households in specific districts, renewal programs targeted those districts for clearance, and highway engineers then drew their cheapest, least resisted routes through the same ground.

Q: How did San Francisco’s 1959 freeway vote change the interstate program?

Before 1959, the assumption in every American city was that the highway plan, once drawn, would be built, and that neighborhood opposition was a delay to be managed rather than a veto to be feared. After the Embarcadero Freeway showed San Francisco what an elevated highway did to the waterfront, the Board of Supervisors voted in 1959 to cancel seven of ten planned freeways, the first successful anti-freeway vote in American history. The news traveled, and organizers in other cities learned the playbook: organize early, before rights-of-way are bought; build coalitions beyond the immediately threatened blocks, joining residents with waterfront users, park defenders, and preservationists; and aim at the most vulnerable layer of the layered approval process. The 1964 fight over the Panhandle and Golden Gate Park freeway proved the first victory was not a fluke, and the French Quarter route in New Orleans was blocked next. The revolts also changed engineering practice: after the 1960s, community input, environmental review, and public hearings entered the planning process, making highway building slower and more expensive but less casual about the neighborhoods in its path.

Q: What happened when I-40 was routed through Nashville’s Black neighborhood?

Interstate 40’s construction through Nashville demolished more than 620 Black homes, cutting through a thriving Black business and residential district. The case is one of the best documented examples of the national pattern Raymond Mohl identified: state and local officials routing urban segments through the neighborhoods where land was cheapest to acquire and political resistance weakest. Nashville’s experience illustrates what the aggregate displacement figures mean in concrete terms, not abstract relocation statistics but specific streets, businesses, churches, and gathering places destroyed. It also illustrates the entanglement with segregation, since the district the highway cleared was one to which Black residents had been confined by the housing practices of the era.