State highway data reaches Washington on a staggered calendar that almost nobody outside the reporting offices has read. Interstate pavement and its related items are due to FHWA on April 15 of the year after collection. Certified public road mileage follows on June 1. Non-Interstate pavement, the non-pavement items, the sample sections and the statewide summaries arrive on June 15. The field manual setting those dates is still the December 2016 edition, and it requires the submittal to represent the road network as it stood on December 31 of the collection year.
Those dates are why “latest available data” carries no fixed meaning across federal highway datasets. Each has its own reference period, release lag, and answer to whether a figure was counted, sampled, or modeled. Six carry most of the weight in American highway argument.
HPMS is a census of miles wrapped around a fixed sample
FHWA describes the Highway Performance Monitoring System as a national level highway information system covering the extent, condition, performance, use and operating characteristics of the nation’s highways. It carries administrative and extent of system information on all public roads. Everything else arrives as a mix of universe and sample data for the arterial and collector functional systems. The system was developed in 1978 as a continuing database, replacing the special biennial condition studies run since 1965.
The internal architecture decides more than the release date does. Full Extent items are reported for a whole roadway system, such as the National Highway System or an entire functional class. Sample Panel items are reported only for designated sections, and the field manual is explicit that those sections form a fixed panel monitored year to year which, when expanded, represents the Full Extent of the systems sampled.
Roughness is collected at full extent on the entire National Highway System and every principal arterial section, and only expanded from a sample panel below that, mandatory for rural minor arterial sample sections and optional for urban minor arterials and the classes below them. A state-level roughness figure for a rural minor arterial is therefore a panel expansion rather than a measurement of every mile, and the manual separately permits factored or estimated data where it says so for a given attribute. Neither fact appears on the face of a published HPMS table, and both change what a gap between two states can support.
The bridge inventory runs on a rolling clock
Under 23 CFR 650.315, every owning agency, whether a state transportation department, a federal agency or a tribal government, must prepare and maintain an inventory of all bridges subject to the subpart, and that data goes to FHWA annually or whenever requested. The annual submission is well known; the entry deadlines beneath it decide what a given file contains.
For all inspection types, changes go into the owning agency’s inventory within three months after the month when the field portion of the inspection is completed. The same three-month window governs modifications that alter previously recorded data, newly constructed bridges after opening to traffic, and changes in load restriction or closure status, and each agency has to establish and document a process ensuring those constraints are met. An annual National Bridge Inventory file is therefore a snapshot of a database that was moving while it was taken, and the inspection behind any single record can be a quarter older than its entry.
Two documents supply the definitions, and 650.317 incorporates both by reference. Element level inspection data for National Highway System bridges is collected in accordance with the Manual for Bridge Element Inspection, second edition, 2019. How to collect and report that data is set out separately, in FHWA-HIF-22-017 of March 2022, titled Specifications for the National Bridge Inventory and abbreviated SNBI. FHWA’s inventory page describes a changeover still in progress: reporting continues under the older Recording and Coding Guide until the annual submittal date specified in an implementation memo, and subsequent submittals move to the newer specifications. A trend line built across that changeover mixes two data specifications rather than tracking one, and what the resulting numbers can and cannot show is taken up in the state of the bridge inventory.
FARS is a census assembled by state employees out of state paperwork
The Fatality Analysis Reporting System became operational in 1975 and holds a census of fatal traffic crashes across all fifty states, plus the District of Columbia and Puerto Rico. Each crash year’s data reaches the public in two releases: an Annual Report File the following year, then a final file about a year after that with updates folded in. The tests that decide which deaths qualify for it are set out in what the roadway fatality data shows. What matters for reading the file is how it gets built. NHTSA’s National Center for Statistics and Analysis directs the system under a cooperative agreement with an agency in each state government. Trained state employees, called FARS analysts, gather, translate and transmit their state’s data in a standard format, their number varies by state, and from the source documents they code more than 140 data elements.
The source documents are worth knowing before any FARS field is read as an observation. The manual lists police crash reports, death certificates, state vehicle registration files, coroner and medical examiner reports, state driver licensing files, state highway department data, emergency medical service reports, and vital statistics.
A sharp recent example: three automated driving system elements were added in 2019, and collection proved difficult, because most states have no such field on their crash report and identification was possible only through the report narrative. The elements were largely coded as not reported. Quality control on the 2019 and 2020 data then highlighted inconsistencies the agency judged capable of leading to varying or misleading results, and the three were removed from those two years while further research is conducted. An empty field in a federal file is sometimes a finding about the paperwork rather than a fact about the road.
The freight framework publishes three futures and gets quoted as one
The Freight Analysis Framework is produced through a partnership between the Bureau of Transportation Statistics and FHWA. Version 5 starts from the 2017 Commodity Flow Survey and Census Bureau international trade data, incorporates agriculture, extraction, utility, construction, service and other sectors, and delivers tonnage and value by regions of origin and destination, commodity type and mode for base year 2017 with a thirty-year forecast at five-year increments.
Those increments arrive under three economic growth scenarios, and the arithmetic is not small. FHWA’s forecast study puts baseline domestic tonnage growth at a 1.1 percent compound annual rate from 2017 to 2050. The high-growth scenario runs 1.3 percent a year and yields 26.5 billion tons of domestic commodity flow in 2050, which the study puts 1.6 billion tons above baseline. The low-growth scenario runs 0.8 percent and reaches 22.5 billion tons, 2.4 billion below. A single 2050 tonnage quoted without its scenario label has discarded a four billion ton spread. The framework’s other well-known trap, the treatment of imports inside a tonnage total, is examined in the freight corridor rankings.
Highway Statistics is the oldest series and the most heavily footnoted
The Highway Statistics Series has appeared annually since 1945 and collects motor fuel, motor vehicle registrations, driver licenses, highway user taxation, highway mileage, travel and highway finance. Most of it is submitted by the states directly to FHWA, and each state’s data is then analyzed for completeness, reasonableness, consistency and compliance with the reporting instructions in A Guide to Reporting Highway Statistics.
The footnotes are the load-bearing part. Table VM-2 in the 2024 edition carries a note that travel for the rural minor collector and the rural and urban local functional systems is estimated by the states based on a model or other means and provided to FHWA on a summary basis, while travel for all other systems is estimated from state-provided HPMS data. A second note, attached to individual states, records that the state modified its process for estimating summary vehicle miles traveled. A break in a state series can sit inside a national total with nothing but a superscript to mark it, and calendar year 2024 travel reached the public in a table dated October 2025. Superscripts are the working material of the research and resources section.
The cost index measures what states paid, not what producers charged
The National Highway Construction Cost Index measures the average change over time in the prices paid by state transportation departments for roadway construction materials and services. FHWA contrasts it with the Bureau of Labor Statistics producer price index for streets and highways, which measured selling prices received by producers and which BLS discontinued in July 2010. The two sit on opposite sides of the transaction, and the NHCCI includes the contractor’s mark-up over the original producer’s price.
Its construction is a chained Fisher Ideal index. The formula runs on pay-item-level bid quantities and estimated bid prices from the Oman Bid-Tabs database, produces a single change between adjacent periods, and multiplies those changes into a series. Because the Fisher index has no base year, and other indexes use different methodologies or base years, FHWA states that direct comparison will not yield an accurate picture and that an accurate one requires converting both into a percent change first.
There is no way to separate price increases due to inflation from price increases due to increased productivity without additional data, specifically data on labor utilization in relation to the construction materials.
The published record lags further than most users assume. FHWA’s narrative series ends at 2024 Q3 as of August 2026, and that installment is dated May 14, 2025. It reports a 6.5 percent increase over 2024 Q2, 4.1 percent after seasonal adjustment, and 7.4 percent year over year against 2023 Q3. That installment reached readers roughly seven and a half months after the quarter closed, and the lag underwrites every argument about what a mile of interstate costs.
What the documentation already says
Each of these six publishes a number in one place and explains what it is made of elsewhere, in smaller type, under a heading nobody clicks. That asymmetry is the real problem, and it is not a problem of secrecy. The sample panel and the removed automation fields are as public as the totals built from them, and so is the absence of a base year in an index that gets quoted as if it had one. All of that documentation comes from the agency that published the figure. A claim that has survived contact with its own documentation is worth more than one that has never been near it.