Congress wrote a service life into the statute that built the Interstate, and it was nineteen years. Section 108 of the Federal-Aid Highway Act of 1956 required the system’s design standards to be sized against a traffic forecast for one named year, 1975, with right-of-way widths wide enough to build to them. Traffic in 1975 was the target. Most of that pavement is still under traffic five decades past the forecast it was sized against.
Nothing went wrong. The gap between 1975 and now is the ordinary distance between two different quantities that the word “last” collapses into one: the design life an engineer computes, and the service life an agency actually gets.
Federal policy declines to name a number
Anyone expecting a required service life in the regulations will be surprised by how short 23 CFR Part 626 is. Its operative sentence reads in full: pavement shall be designed to accommodate current and predicted traffic needs in a safe, durable, and cost effective manner. The definitions section lists the factors a pavement design weighs, naming materials, traffic, climate, maintenance, drainage and life-cycle costs. There is no minimum term anywhere in the part.
The number that federal guidance does fix sits one level up, in economics rather than engineering. FHWA’s September 1996 policy statement on life-cycle cost analysis, restated in the agency’s technical bulletin on the subject, recommends an analysis period of at least 35 years for every pavement project, whether new construction, total reconstruction, rehabilitation, restoration or resurfacing. The bulletin adds the rule that makes the distinction sharp: the analysis period should generally always be longer than the pavement design period, except for extremely long-lived pavements, and as a rule of thumb it should be long enough to include at least one rehabilitation.
Read those together and a 35-year horizon with a 20-year design inside it is not a contradiction. It is the intended arrangement. The design life is a milestone within the analysis, not the end of the asset.
Design life ends at an inconvenience, not a failure
The clearest definition in the federal literature comes from a 2013 Turner-Fairbank report on remaining service life. Design life, it says, is the predicted time it will take for the structure to reach a minimum acceptable condition value, and then it adds the qualifier that decides everything downstream: all current pavement design methods base that minimum on functional service criteria, not on an extreme condition representing severe structural defects or preventing vehicle passage.
A road at the end of its design life has become rough, cracked or rutted enough to fall below a comfort and safety standard. It has not stopped carrying trucks. Every argument about whether a highway is “worn out” is therefore an argument about where somebody set a threshold, and the honest way to read a design life is as a forecast attached to a particular choice of thresholds.
Mechanistic-empirical design makes the arbitrariness explicit. Under the AASHTO mechanistic-empirical method, design life is defined as the shortest time it takes for any one of the modeled distresses to reach its terminal limit. The prediction is a weakest-link result. Add a distress model to the analysis and the design life can only fall or stay level. Drop one and it can only rise. Two agencies modeling the same section with different distress sets will publish different design lives without either being wrong.
What the thresholds are, in numbers
Federal reporting has settled the question for the National Highway System, and the tolerances are looser than the rhetoric about crumbling roads suggests. Under 23 CFR 490.313, a section’s roughness rating is Good below an International Roughness Index of 95 inches per mile, Fair from 95 to 170, and Poor above 170. Asphalt cracking is Good below 5 percent of surface area and Poor above 20 percent. Asphalt rutting is Good below 0.20 inches and Poor above 0.40. On jointed concrete, faulting is Good below 0.10 inches and Poor above 0.15, and slab cracking is Poor above 15 percent. Continuously reinforced concrete crosses into Poor at 10 percent.
The composite rule is where the tolerance shows. Asphalt and jointed concrete carry three ratings, and a section earns an overall Good rating only if all three are Good; it falls to overall Poor only when two or more of the three are Poor. Continuously reinforced concrete is scored on just two of those measures, roughness and cracking, and the same logic scales down with it: overall Good requires both to be Good, and overall Poor requires both to be Poor. A pavement can be measurably terrible on roughness and remain officially Fair. The federal instrument is calibrated to catch pavements that have failed in more than one way at once, which is a defensible choice for a national measure and a poor guide to when a specific road needs work.
Loads, not years
Pavement design has never really been denominated in time. It is denominated in axle passes, and time enters only because traffic accumulates.
FHWA’s Traffic Monitoring Guide sets out the traditional unit. One equivalent single-axle load is the damage from an 18,000 pound single-axle load, and designers project cumulative equivalent loads from every vehicle expected across the pavement’s life. The guide’s own worked illustration shows how steeply the conversion bends: a 10,000 pound single axle carries an equivalence factor of 0.09, a 16,000 pound axle 0.61, and an 18,000 pound axle 1.00 by definition. An axle 80 percent heavier than the 10,000 pound case does roughly eleven times the damage. Doubling from 8,000 to 16,000 pounds moves the factor from 0.04 to 0.61, a factor of about fifteen.
That curve explains why Interstate design life is a freight question before it is a materials question. Combination trucks ran 195,758 million vehicle-miles nationally in 2023, and roughly 110,338 million of those, about 56 percent, fell on Interstate lanes that make up a little over one percent of public road mileage. The consequences of that concentration for geometry and structure are taken up in how trucking shapes highway design; the consequence for design life is that a forecast of tonnage matters more to the answer than any material choice.
No load forecast captures what climate does to a pavement, and the authors of the Transportation Research Board’s circular on perpetual pavements are blunt about how fast it moves: unless proper attention is given to frost protection, ride quality may degenerate from good to poor in a matter of months because of differential frost heave in the subgrade. Months, not decades. Nor does a forecast reach what lies underneath, since a thick structure on a poor foundation is a thick structure that will crack from the bottom. The freeze-thaw mechanism has its own treatment in freeze-thaw damage and potholes.
The perpetual pavement claim, and who makes it
The most ambitious answer on offer is that a pavement need not have a terminal date at all. A perpetual pavement, as FHWA’s remaining-service-life report defines it, is an asphalt pavement designed and built to last longer than 50 years without major structural rehabilitation or reconstruction, needing only periodic surface renewal in response to distresses confined to the top of the pavement. The premise is that a thick enough asphalt section on a stable foundation keeps damage out of the layers where repair is expensive, so renewal becomes a recurring surface operation rather than an eventual rebuild.
The concept paper behind that definition appeared in Transportation Research Circular 503 in December 2001, written by David Newcomb of the National Asphalt Pavement Association, Mark Buncher of the Asphalt Institute, and Ira Huddleston of the Asphalt Pavement Association of Oregon. Their target was performance beyond 50 years with the top 25 to 100 millimeters replaced roughly every 20 years. The authors sell the material, which does not make the mechanism false and does put the burden of evidence on them.
Their evidence is observational and worth having. A review of thick asphalt sections on Interstate 90 across Washington State found pavements between 23 and 35 years old, none of which had ever been rebuilt for structural reasons. West of the Cascades the average age at resurfacing was 18.5 years. East of the mountains, first resurfacing came at an average of 12.4 years and second resurfacing 12.2 years after that. The structure lasted; the surface did not, and the surface was cheap by comparison. That is the perpetual pavement argument in its honest form, and the design implications run through long-life pavement design.
The Pennsylvania schedule is the more honest question
There is no national figure to quote, and an agency that publishes one is describing its own assumptions rather than a property of asphalt or concrete. FHWA’s technical bulletin is candid about where those assumptions come from. Some agencies derive performance lives from their pavement management data and historical experience. Others do not.
Some SHAs develop performance lives based on the collective experience of their senior engineers.
That sentence is not an indictment. Expert judgment is a legitimate input, and the same bulletin recommends treating design life as a distribution rather than a point value, running Monte Carlo simulation over the uncertain inputs. For one of the two alternatives in its worked example, the correlation plot puts initial agency cost first as a driver of net present value, at 0.72, and the design life of the initial construction second at negative 0.52. The sign is the interesting part. A longer assumed life lowers the present value, which is exactly why the assumption deserves scrutiny rather than deference.
A state’s pavement design manual states the design period it uses by facility class, and that number is the least informative of the figures the agency actually holds. Its life-cycle cost procedure states the assumed performance period for each treatment, which shows what the agency plans to do and when rather than how long the first structure is supposed to last. The bulletin reproduces one such schedule from the Pennsylvania Department of Transportation for new and reconstructed concrete: joint sealing at years 5, 10 and 15, concrete pavement restoration with full-width diamond grinding at year 20, more sealing at 25, a structural overlay at 30, and shoulder work at 35. Nothing on that list is a failure. Each entry is a planned intervention against a pavement still carrying traffic. The bulletin’s own note under that schedule describes the year 20 concrete pavement restoration as “the CPR strategy slated for year 20,” a description of a plan, not a failure. The state’s pavement management records show what the previous pavement on that same alignment actually did, which is the only figure in the set derived from measurement rather than prediction, and those records recur throughout highway design coverage.
A road built to a 20-year design and rehabilitated on that Pennsylvania schedule is serving traffic in year 35 in good condition, and it was never a 20-year road. The number worth asking an agency for is not how long the pavement will last. It is what the plan is for year 20, whether it is funded, and what happens to the schedule if it is not.