Where should Missoula grow?
The typical Missoula home gained 37.9% in real value between 2019 and 2024, while the City permitted about 750 homes a year over the same period.
The instinct is to build where land is cheap, which means the edge. This walks through what that choice actually costs, and where to grow instead.
Start with every low-intensity parcel
City parcels that are vacant, single-family, or mobile home, with no more than one home on them. 18,571 parcels on 11,317 acres — 69% of the city's parcels and 39% of its land.
No filtering yet on zoning, hazard, or transit. Those attach later, as separate layers of evidence rather than as filters on this one. Shown here: the Rattlesnake, upper and lower — 1,962 of the 18,571.
Why not just work in parcels?
A parcel is the wrong size for a policy. They range from a tenth of an acre to forty, no two neighbors are comparable, and a council cannot draw a zoning boundary around 18,571 of them individually.
So the land is re-measured onto a grid of equal cells, each about 26 acres — roughly a quarter-mile across, or the distance most people will walk to a bus. Every measurement from here on is per cell, and every map after this one is this same grid.
How the land gets onto the grid
A parcel that straddles two cells is split between them by area, so six acres and four acres of a ten-acre parcel land in the cells that actually contain them. The lazy alternative — assigning the whole parcel to whichever cell holds its center — would move all ten acres to one side of a line, which matters most for exactly the big edge parcels this analysis cares about.
Floodway and steep slope take 920 acres
The supplied taxlots record what share of each parcel sits in the floodway and what share is too steep to build. Those shares come off within each parcel, so a lot that is a third floodway keeps its other two thirds.
The alternative — dropping any parcel with a constraint — would have thrown away 545 whole parcels to remove a median of 29% of each.
A built lot is not an empty one
An empty acre and an acre with a house on it are not equally available. Each parcel is scaled by how much of it is already covered by a building: a fifth of its area counts if a building covers 35% or more of the lot, all of it counts if the lot is empty, sliding between.
That removes another 2,143 acres, leaving 8,177. It is the largest assumption in the analysis and it is listed at the end with the others.
What the code already permits
Everything after this point is counted in homes, not acres. Missoula's residential zones state a minimum lot area per dwelling — R5.4 is 5,400 square feet, RT10 is 10,000 — so the homes a lot permits is its area divided by that figure, net of what already stands.
13,164 homes, against 17,168 already on the same lots. The largest zone by far, R5.4, permits a median of zero more. Two exclusions: parcels over five acres are subdivisions rather than infill, and PUDs, commercial and unzoned land state no residential density at all.
Opportunity arrives in neighborhoods
Policy is written for areas, not for scattered lots, so it is worth checking whether the opportunity actually forms areas. It does: places with room to build sit next to other places with room to build far more often than chance allows.
The map shows the four ways a place can relate to its neighbors. The two outliers matter as much as the clusters — they are where a blanket area policy would be wrong about a specific block.
What the four colors mean
- Room, surrounded by room
- A genuine opportunity area. The raw material for a growth designation.
- Built out, among built out
- Already developed or genuinely constrained. Nothing to decide here.
- Room, surrounded by built
- An isolated pocket inside a developed area. Infill, not an area policy.
- Built, surrounded by room
- A developed island inside an opportunity area — often exactly where the preservation and displacement concerns sit.
How this was measured
Moran's I is a single number for the whole city answering "is this pattern clustered, or could it be random?" It runs from −1 (perfectly alternating, like a chessboard) through 0 (random) to +1 (perfectly clustered). Missoula scores 0.66. To check that is not a fluke, the values are shuffled at random 999 times; not one shuffle produced a pattern this clustered.
Local Moran's I then asks the same question of every cell individually rather than the city as a whole, comparing each one to its own neighbors. That is what produces the four categories above, and it is the reason to use it instead of a plain hot-spot measure — a hot-spot map shows you the clusters but hides the two kinds of exception.
Not every address reaches the same city
The question is not "is there a bus stop nearby" but "how much of Missoula's employment can someone actually get to from here on a weekday morning, walking and riding?"
Answered for every cell against every block of jobs in the county. The typical place reaches 59 jobs within half an hour. The best reaches 42,017.
How the travel times were built
Walking and riding sit in one network, so the shortest path chooses between them rather than being told which to use. Walking uses the OpenStreetMap footpath network at 3 mph. Boarding costs half the route's waiting time — the average wait if you turn up without a timetable. Riding costs the actual scheduled running time from Mountain Line's published feed.
This is why a place beside a bus stop can still score badly: if that route runs every 40 minutes and reaches nowhere anyone works, the stop is not access.
Why nearby jobs count for more
A job forty minutes away is not worth as much as one ten minutes away, so jobs are weighted by how long they take to reach rather than counted inside a cut-off. The weight halves every thirty minutes — a job half an hour out counts as half a job, an hour out as a quarter.
This is the standard gravity model, borrowed from physics: interaction falls off with distance. The alternative, drawing a thirty-minute circle and counting everything inside, treats a job at 29 minutes as full value and one at 31 minutes as worthless.
$803 a year separates the two ends of the map
The same commute question, answered by car. Expected distance to work comes from the road network and where the jobs actually are; two trips a day, 250 days, priced at the IRS rate of 67 cents a mile — a full ownership cost, not just fuel.
A household in the areas this analysis recommends growing spends $1,121 a year getting to work. In the areas it says avoid, $1,924. Every figure elsewhere on this site is that $803 difference.
The edge is cheaper to build. It has other costs.
Land at the edge costs $42,000 an acre against $805,000 where this analysis recommends growing, so building there is far cheaper.
Four costs come with it, and they recur. Each year, a household out there drives 1,199 more miles, spends $803 more reaching work, and emits half a ton more carbon — and the City maintains 1,690 more feet of sewer to serve them.
What "invest first" actually means
Some places have the land and the demand but not the service. Housing opportunity multiplied by reachable jobs, decayed by travel time, is accumulated onto the streets between them; the streets carrying the most of it while running infrequent service are the ones where an operating increase buys the most.
13.8 miles of street qualify. Every one sits on a route Mountain Line already runs, so the ask is operating budget rather than new coverage, and none of it reaches past the built edge.
The capacity and the risk are the same ground
Renter share, low income, and severe rent burden, from the 2020–2024 American Community Survey. Displacement risk is claimed only where vulnerability and active development pressure overlap — 120 acres, holding 1,800 of the homes the code permits.
Vulnerability on its own describes a neighborhood; it does not show anyone being displaced. Two thirds of the homes this city could add sit in exactly these neighborhoods.
Four things the City can do
The cells are gathered into 70 contiguous areas, each holding at least 100 acres of opportunity so no area is too small to write a policy about. An ordered rule then assigns each one a single action.
Click to hide a strategy from the map.
How the areas were drawn
Max-p is a regionalization method: it groups small units into the largest number of areas it can while guaranteeing two things — every area is geographically contiguous, and every area clears a stated minimum. Here the minimum is 100 acres of opportunity, so no area exists that is too small to justify a policy of its own.
"The largest number it can" matters. Forcing a fixed number of areas, the usual alternative, makes the answer depend on a number chosen in advance; Max-p lets the data decide how many areas Missoula actually has. It found 70. The same grid was also run through SKATER, a different method, as a check that the boundaries were not an artifact of the technique.
Three routes. Eight buses.
Only two of twelve Mountain Line routes come every fifteen minutes. Only two areas in the city have land, jobs and a frequent bus at once. Six more have the land and the jobs and are waiting on the bus.
Raising Routes 12, 5 and 11 reaches 792 of those homes, plus 652 in the core. Nine more vehicles, no new alignments. Then the homes build the ridership that pays for the service that makes the homes work.
77% matches Our Missoula 2045
The adopted plan's intensity was held out of the decision entirely so it could serve as an independent check. Fifty-four of seventy areas land where the plan already puts them.
The other sixteen are the useful part, and they disagree in two different directions.
What the two disagreements mean
- Plan says low, evidence says opportunity — 9 areas
- The plan designates these for little or no additional housing, but they have room to build, reach jobs well, and carry no serious hazard or displacement risk. These are the candidates for an intensity change.
- Plan says grow, evidence says caution — 7 areas
- The plan marks these for growth, but they score badly on hazard, access, or the cost of serving them. Not an argument against growth there, but a reason to ask what has to happen first.
Neither list is a recommendation to overrule the plan. They are the sixteen places where an evidence-led read and an adopted policy disagree, which is where a conversation is worth having.