Last week this site covered the most complete public dataset for a city-built open-access network at scale. Lehi measured the subscription business: 30% take rate against a 38% break-even forecast, published with the shortfall intact. The Colville study, published by Internet Society Pulse on September 15, measures something Lehi's numbers can't: the value of a network when almost nobody's household subscribes to it. The two datasets together answer the question every community network vote actually turns on — which model are you building, and what ledger should it be judged on?

What was built, and why nobody else built it

Until 2013, public facilities on the Colville Reservation ran on T1 lines delivering about 1.5 Mbps at high cost. The reservation spans more than a million acres of mountains in northeastern Washington. For a commercial carrier the arithmetic never worked: high fixed costs, thin subscriber base, and permitting that crossed tribal, county, state, and federal authorities. No carrier came.

So the Confederated Tribes of the Colville Reservation built the network themselves. Fiber to three mountain summits, radio relays the rest of the way, solar and wind power at sites the electric grid didn't reach. Community centers and government offices went from 1.5 Mbps to symmetric 100 Mbps. The build cost roughly $3.5 million, about $2.8 million of it federal grants, the rest tribal funds. The rollout was deliberately staggered — Nespelem and Omak connected in 2013, Inchelium and Keller in early 2014 — and that sequencing turned out to be what made measuring the impact possible.

The study, and the honest part of the method

Stratmann and Sharma (George Mason University) ran two comparisons and — this is the part worth noticing — reported both with different confidence levels.

The within-reservation comparison uses the staggered rollout: communities connected earlier versus later, same reservation, same government, same economy. Anything that arrived reservation-wide in a given year is excluded from the contrast. That is the causal estimate: income rose about 4 percent as communities came online. Paid employment at local establishments rose by more, with enough imprecision that the authors read it as direction rather than magnitude. The gains show up among existing tax filers — residents earned more, rather than new residents arriving.

The synthetic-control comparison builds a statistical "synthetic Colville" from similar reservations and finds per-capita income about a third higher — roughly $5,500 per person, labor force participation up 10.3 percentage points. The authors themselves say not to attribute that to broadband: the same window contains a new casino, regional wildfires, a federal trust settlement, and a tribal business reorganization. They treat the within-reservation 4% as the credible estimate and the larger figure as an upper bound on everything that changed.

That is the opposite of how infrastructure impact studies are usually marketed. The authors published the flattering number and the defensible number and labeled which is which. A community council weighing a broadband bond should ask for exactly this format from whoever hands them a projection.

Where the gains came from

The most useful finding is where the gains did not come from: home broadband subscriptions barely moved, and the population didn't grow. The benefit flowed through the institutions the network reached first — tribal administration and education. Government offices that could suddenly coordinate over long distances, process applications, and move records at modern speeds. In terrain where overlapping jurisdictions slow ordinary coordination, lowering the cost of coordination paid off.

This matches the build order. The network connected tribal government, schools, and health services first; households largely came later or not at all. The Hoopa Valley Tribe's data center is the same argument one layer up: the community's payoff arrived through institutional infrastructure the tribe owned and operated, not through retail subscriptions.

The two ledgers

Lehi and Colville are not competing evidence for the same claim. They are evidence for two different claims, and a community choosing between them should know which one it is making:

  • The retail subscription model (Lehi: city-built open-access fiber, five competing ISPs, published take rates) is judged on a telecom business ledger — take rate versus break-even, NPS, subscriber growth. Its honest numbers are published and they cut both ways: real customer pull, real financing discipline.
  • The institutional network model (Colville: connect the offices, schools, and clinics that coordinate the community) is judged on an institutional ledger — coordination costs, administrative capacity, service delivery. Its honest number is a 4% income effect measured against a staggered rollout, with the caveats attached.

A fire hall deciding whether to wire its stations, a small town deciding whether to fiber its offices and schools before its subdivisions, a tribal government staring down the same stacked permits: the Colville case says the institutional build is fundable at roughly $3.5M scale, measurable, and valuable even where household subscriptions never materialize. The market's refusal to build is not evidence that the network has no value. It is evidence the value doesn't flow through the market's preferred meter.

Two more numbers for the funding context: the Colville Tribes have since received a separate $48.4 million federal award to connect 2,867 unserved households — the residential build following the institutional one, in that order. And a new $790 million federal tribal broadband round was open with applications due September 17. Stratmann's own caution applies to every application in that pile: the money will buy less than what Colville built wherever the jurisdictional obstacles — permitting across tribal, county, state, and federal lines — are left standing. The binding constraint was never tribal capacity. It was the approval stack.

What this does not show

Keep the hedges the authors kept. This is one reservation, measured at ZIP-code level, with an employment estimate the authors themselves call too imprecise to state a magnitude. It is not evidence that any particular network pays for itself, and it does not transfer automatically to a town with different terrain, different institutions, and a market willing to build. What it does establish is narrower and harder to dismiss: where commercial carriers declined, a community with design authority and modest capital moved measurable economic outcomes — and the mechanism was institutional coordination, which is exactly the part of the stack a municipality, a fire district, or a tribe can own outright.