Transit Brief · City of Asheville

The bus trilemma: the working detail

Every figure behind the paper, with its basis, its margin and what it cannot show. This is the page for anyone who wants to check the work.

Prepared 2026-08-23 · Figures from the National Transit Database (Time Series 2.1, npsm-38gk; Complete Monthly Ridership, 8bui-9xvu, which runs through June 2026), the City of Asheville's adopted budget books, and its council documents of 24 March and 28 July 2026 · Peer band is rule-defined, not hand-picked

On August 18 the redesigned bus network was pulled from the city council's August 25 agenda, and the vote is now delayed indefinitely. The public argument is about money: a system that "costs three times as much to run as it did in 2018." That framing is measuring the wrong thing. On three of the four standard cost rulers, Asheville runs a cheaper bus system than most comparable American cities. What changed is not the price of the service. It is how many people are on it.

1 · The problem, stated exactly

Four rulers exist for what a bus system costs. The council's argument, and the coverage of it, has been conducted almost entirely on the fourth one.

FY2024, fixed-route bus	Asheville	Peer median	Reading
Cost per rider	$7.92	$10.10	22% below
Cost per revenue hour	$123.28	$130.86	6% below
Cost per passenger mile	$2.16	$2.26	4% below
Cost per revenue vehicle mile	$9.96	$9.36	6% above
Riders per revenue hour	15.56	13.09	19% above median

The band is 145 agencies, chosen by rule rather than by hand: every US fixed-route bus system in an urbanized area of 150,000 to 500,000 people, carrying at least 150,000 trips a year. Thirty of them report no passenger miles, so that median is over the other 115. Two limits of that rule are worth stating, since the rule is the whole defence of the comparison. It counts agencies, not areas. Thirteen urbanized areas contribute more than one agency, and 28 of the 145 sit in them, so a town-scale operator can share a class with a city-scale one. Durham carries three. And ART runs only 19 vehicles in peak service, which would let it file a reduced NTD return. It files a full one anyway.

The one ruler where Asheville looks expensive divides by vehicle miles, and Asheville's buses are slow: 12.4 miles in a revenue hour against Roanoke's 16.3. Roanoke's buses cost within one cent of Asheville's per rider.

One caution about the year. FY2024 is the latest full annual release, and it is also a local high. ART carried 1,377,810 trips in FY2023 and 1,607,423 in FY2024, then 1,416,025 and 1,458,089 in the two years after. FY2024 sits 16.7 percent above FY2023 and 10.2 percent above FY2026. The peer comparison above is like-for-like, because every agency in the band is measured in the same year. But a cost per rider computed on a peak year flatters the agency, and the FY2026 productivity figures used later in this brief are the harder number. The direction survives the check: the same FY2024 cost over FY2026's lower ridership is $8.73 a rider, still 14 percent below the peer median.

2 · One cliff, then seven flat years

Riders per revenue hour is the cleanest measure of whether a bus is doing its job, and the federal database publishes it monthly. Asheville's own July-to-June fiscal years, every bar from the same file:

Riders per revenue hour, ART fixed-route bus. All nine years computed from the same monthly file and reconciling with the annual release to the digit. Peer median is FY2024.

The loss is one year wide. Between FY2019 and FY2020 ridership fell 28.8 percent while service grew 13.1 percent. That is the pandemic, and it happened to every transit system in the country. What is specific to Asheville is what came after. Since FY2021 riders are up 1.8 percent on 3.0 percent fewer hours. That is a slow improvement, and it is very small next to what was lost. FY2019 to FY2020 cost 570,197 trips. The gain since FY2021 is 25,900. Five years of recovery has clawed back under 5 percent of one bad year.

One year in that run needs naming. Helene falls in FY2025, and it shows: trips fell 11.9 percent from FY2024 and service fell 8.5 percent in the same year. FY2026 took both back up, trips by 3.0 percent and hours by 5.5 percent, which is why productivity kept drifting down while ridership recovered. The storm is a dip in this series. It is not the story of it.

So the honest version of the headline is not that the buses got expensive and not that they keep emptying. It is that Asheville bought a third more service around a collapse it did not cause, has never filled it, and has never published a target for filling it.

3 · The trilemma

Two lines of arithmetic govern every option anyone can propose:

Which leaves four things anyone can actually move: hours bought, price per hour, riders per hour, and who pays. Fares are not one of them. In FY2024 they were $601,698 against $12,732,832 of operating expense, 4.7 percent. Spread over 1,607,423 boardings that is 37 cents a ride actually collected. That is the number to reason from in either direction. A fare rise starts from 37 cents, not from the posted fare.

The price per hour is close to fixed: Asheville is already below its peer median there, and the operations contract with RATP Dev is signed through the four base years. The 24 March 2026 council packet prices those years at $12,671,301, $13,309,046, $14,102,049 and $14,817,546. It states no escalation clause and no index, so the 5.35 percent a year used here is the compound rate across those four published amounts, computed for this brief rather than quoted from the contract. Holding the city's own costs flat, total operating cost grows about 4.62 percent a year. So ridership has to rise 4.62 percent a year for cost per rider merely to stand still, about 14.5 percent from the first base year to the fourth. That is the bar before any improvement counts as improvement.

Coverage, service hours, and the current local contribution. Pick two. Every proposal anyone has made is one of these three rows, and only one of them has been put in front of council.

The one thing that moves the triangle instead of trading along it Cost per revenue hour. Asheville's is $123.28, already below the band median of $130.86. Among the 47 agencies in the band that contract their operations out it is the 16th cheapest, against a contracted median of $140.98. There is no easy saving there, and the base contract runs through FY2030. Which is why Step 5 matters more than it looks.

The second row has never been put in front of council, so here is what it is worth.

ART added 18,730 revenue hours between FY2019 and FY2026, which is 18.8 percent of what it runs now. What removing them would save depends entirely on which cost you charge them at, and that choice moves the answer by nearly half.

NTD publishes Asheville's operating expense by function. In FY2024, vehicle operations ($4,826,429) and vehicle maintenance ($2,130,859) came to $6,957,288 of $12,732,832, or 54.6 percent. Facility maintenance and general administration are the rest, and most of that does not go away when a bus runs fewer hours.

Removing 18,730 hours, priced at…	Per revenue hour	Annual saving	Break-even rider loss
Fully allocated cost	$161.26	$3,020,379	18.8%
Vehicle operations and maintenance only	$88.11	$1,650,352	10.3%

Rates are the FY2027 adopted budget over FY2026 revenue hours, and the second row applies FY2024's function split to it. The answer sits between the two rows, and which end it sits nearer depends on how the operations contract splits fixed cost from variable. The March packet does not publish that split. Treat the first row as an upper bound rather than an estimate.

What it would cost in riders is the part nobody can answer:

If shrinking back to FY2019 hours costs…	Annual trips	Per rider, all costs cut	Per rider, running costs only
no riders at all	1,458,089	$8.96	$9.90
10 percent of them	1,312,280	$9.96	$11.00
18.8 percent, in proportion to the hours	1,184,313	$11.03	$12.19
26.3 percent, the share ridership fell FY2019 to FY2026	1,074,466	$12.16	$13.44
Today, for comparison	1,458,089	$11.03	$11.03

The break-evens are the whole finding, and they are far apart. Cut every dollar with the hours and shrinking beats today's cost per rider if the rider loss stays under 18.8 percent. It beats the redesign's own best case of $9.35 only under 4.2 percent. At variable cost the first bar falls to 10.3 percent and the second disappears: losing no riders at all still lands at $9.90, above the redesign's best case. Two honest cautions on top of that. The added service did not cause the lost ridership: the pandemic did, and nothing here shows the marginal hours carry nobody, only that in aggregate they never filled. And cutting hours without cutting destinations means worse frequency everywhere, so this is the coverage fight in a less honest form. What it really is, is the cheapest experiment available, and it needs the route-level data in Step 1 before anyone could run it responsibly.

4 · What the fight is actually over

The draft network removes service to seven destinations. Council's stated alternative is to add some of them back, at the cost of the fifteen-minute frequency the redesign is built to deliver. The $16 million gets spent either way. What stalled was the decision about how to arrange it, and here is the scale of the thing that stalled it:

Average daily riders, from the city's own 28 July 2026 council deck. The seven are the Outlets/Brevard Road/Transformation Village segment (40), West Haywood between Brevard and Patton (34), and the Grove Park Inn (30). Then MAHEC (11), the Social Security office (8), Kenilworth (7) and Beaverdam (2). The deck's Option 2 offers council four of the seven to add back: the Outlets segment, West Haywood, MAHEC and the Social Security office. The two smallest of those four are the Social Security office and MAHEC, at 8 riders a day and 11: a federal benefits office and a safety-net clinic, 19 riders between them. Kenilworth and Beaverdam are smaller still, at 7 and 2, but the deck does not offer them for add-back.

5 · What other cities got when they tried this

Asheville is not first. The relevant question is not whether a frequency-first redesign is a good idea in principle, but what happened to ridership in the cities that ran one.

City	What it did	Riders after
Knoxville	KAT Reimagined, launched 26 August 2024. The consultant cited an 18% gain here.	−4.1% on +22.8% service, first year after launch
Wilmington	Cost-neutral redesign, January 2025. Its plan protected coverage in writing, so the easier half of Asheville's trade.	−3.2% calendar; +2.8% fiscal
Winston-Salem	Approved unanimously, April 2026. Had already cut its three worst routes separately 15 months earlier, and bought microtransit as an explicit coverage substitute.	Too recent to measure
Roanoke	No redesign. Cited as cheaper per vehicle mile.	Costs within a cent of Asheville per rider

Four cities are not a law, and none of them is Asheville. Read the rows as evidence that the promised gain is not automatic, not as a forecast. Knoxville and Wilmington are computed from the monthly federal file, and the window is the thing to check. Knoxville is given on the twelve months either side of its 26 August 2024 launch, which is what isolates the redesign: 2,730,918 riders on 199,562 revenue hours before, 2,619,353 on 245,060 after. On Knoxville's own July-to-June fiscal years the same data gives −4.0% on +19.9% service, and a second year of −4.5%; a launch-aligned second year cannot be computed yet, because the monthly file ends in June 2026, two months short. Do not pair a launch-aligned ridership change with a fiscal-year service figure: they are different windows, and 4.1% belongs with 22.8%, not with 19.9%. Wilmington's two figures are the same data on two windows, and both are given because the choice of window changes the sign; Wave's own staff have told their board ridership rose 8.6 percent, which the federal file does not reproduce on any window tested. The Knoxville launch date and the Wilmington board actions come from a research pass, not from a primary document read here.

6 · What is actually available, cheapest first

There is no single solution. There is a sequence, and it is ordered by price, which puts the biggest number last because it is a decision rather than a purchase.

Step 0, what a target is worth. Against the FY2027 budget of $16,086,060 and FY2026's service:

Cost per rider implied by each productivity target, at the FY2027 adopted budget over FY2026 revenue hours. This is a budget divided by an actual, which is why today's figure reads $11.03 against the $7.92 actual for FY2024. Because the federal file is monthly, any target set here is checkable within about 90 days and any failure is visible inside a year. Knoxville's redesign was measurably failing in twelve months.

Step 2, the legal point that belongs with it.

One legal point belongs with it. Under 49 CFR 37.121, the obligation to run complementary paratransit attaches to operating a fixed route system at all. A companion rule, 49 CFR 37.131, sets the corridor at three-quarters of a mile either side of every route. Asheville is not proposing to abolish fixed route, only to trim seven tails off it, so every surviving route keeps its paratransit corridor, and a van on a removed tail is added to that obligation rather than substituted for it.

That cuts both ways, and both belong in the decision. Cutting a tail shrinks the mandated corridor around it, which is a saving nobody has claimed and also a service reduction for a protected group, and it changes what Buncombe County carries, since the county files Asheville's ADA paratransit under its own agency. Against that, a van run as complementary paratransit takes on the service criteria of 49 CFR 37.131, and they are not a short list. The entity has to schedule any eligible trip requested the previous day, at any requested time on that day. It may not require a rider to start a trip more than an hour before or after their desired departure time. It may not charge more than twice the fixed route fare, and may not charge a personal care attendant at all. It has to run the same hours and days as the fixed route. And it may not restrict trips by purpose, cap the number of trips, or keep a waiting list. A plain cost per trip prices none of that. One caution the other way: 37.121(b) lets the 37.131 criteria be modified for undue financial burden, so they are not immovable.

Step 5, the fleet.

The fleet is worth a line here, because it is the one long-horizon cost that moves with this decision and because "the buses are old" gets said. They are not old, by the only comparable measure. ART runs 35 buses at an average age of 7.51 years, against a band median of roughly 7.6 to 7.7 depending on how the fleet band is drawn. Either way it sits in the middle, so the low operating cost is not bought by running old buses. Fourteen of the 35 are twelve years or older, which is true and unremarkable for a fleet that age. What matters for the option years is that any change of vehicle type, toward smaller buses on thin segments for instance, is a nine-year commitment if both option pairs are exercised and a five-year one if they are not.

7 · What is not a solution

• The redesign as a budget fix. The plan says so itself. The Draft Network Report tells readers: "Remember, this is cost-neutral, so if you want more service on your street, that must be paid for by cutting something else!" The alternative Growth Network, which uses 60 percent more resources, is labelled "not a proposal." So the redesign was never offered as a saving, and on the arithmetic it is not one: at the consultant's own +18 percent it cuts cost per rider 15.3 percent, and escalation takes all of that back in 3.7 years. Adopt it for reliability, transfers and job access if those are worth having.
• Replacing the network with vans. 126 out of 126, above.
• Fare-free, in either direction. Fare revenue was $601,698 in FY2024, the same dollars as in section 3: 4.7 percent of that year's actual cost, or 3.7 percent of the FY2027 adopted budget. Both are true and they are different ratios, which is this brief's own point turned on itself. And the money at stake is 37 cents a boarding, not the posted fare. Either way it is an access decision, not a cost decision.
• "Efficiency" without a denominator. The ruler the public argument is being conducted on is the only one of four where Asheville looks worse than its peers, and it looks worse mostly because the buses are slow.

8 · Three things nobody has established, and each decides the answer