PID Cost Methodology

There are three ways Niwot’s roads get paid for, and only three. This document prices all of them the same way, from the same parcel data, so they can be set side by side. Written for a reader who wants to check the arithmetic rather than take it on trust.

Purpose and Scope

Niwot’s roads are going to be paid for by the people who drive on them. That is not a campaign position; it is the only arrangement on offer. Boulder County stopped repaving subdivision streets in 1995, and the two incentives that once made a Public Improvement District affordable — a 30% county contribution toward the initial project and interest-free county financing — are no longer offered.

What remains is a choice among three structures:

• A neighborhood PID, in which one subdivision forms its own district and pays for its own streets.
• A townwide PID, in which the whole community forms one district and pays for all of them.
• Incorporation, in which a town funds the same road program out of a broader tax base that includes sales tax.

This document derives the cost of each, per household, at the same home. It does not argue for one. It shows the arithmetic and marks clearly where a judgment was made and in which direction.

Every figure here is reproduced by a script that reads the source data directly. Where this page and the script disagree, the script is right and this page is stale. Anyone reviewing the analysis is welcome to the working.

What Is Observed and What Is Chosen

A mill levy is a ratio: the dollars a district must raise each year, divided by the assessed value it can levy on. Four things determine it. Two are observed, one is set by engineering estimate, and one is an open question of district design.

The townwide row prices something nobody is offering. No proposal exists and nobody is assembling one. It has been put to this community twice — defeated at the ballot in 2013 (county Issue 5C) and withdrawn before the ballot in 2025. It is derived here because a household weighing the options is entitled to the figure, not because it is a choice on any ballot. The neighborhood row is the district a resident can actually start: the county has the mechanism and will help. See A Public Improvement District Will Not Fix Niwot.

The property side is checkable by any resident against their own assessment notice. The construction side rests on the County’s own cost figures, which matters: the County sets the level it will agree to fund, so the County’s number is the one that governs, not a contractor’s estimate.

The Source, and What Had to Be Corrected in It

The underlying work is a committee spreadsheet that prices all 23 Niwot neighborhoods individually against 2,015 County parcel records. Rebuilding it line by line turned up six errors, all of which are corrected here. Four are arithmetic. Two are a different and more consequential kind: the tax base was summed over the wrong parcels, so a neighborhood was charged for one set of streets while a different set of households was billed.

Four of the six understated the cost of a PID, which is to say they ran in the campaign’s favor. We corrected them anyway. A figure that only survives because nobody checked it is not worth publishing.

The fifth is worth dwelling on, because it is the one that would have done real harm. A neighborhood was being charged for a road whose residents were not in the district paying for it — fifteen households on a street that is more than half the pavement, absent from the tax base, while twenty-three households carried the entire bill. That produced a levy of 38.90 mills where the corrected figure is 31.41, and a letter quoting the wrong one would have told those households they faced roughly $780 a year more than they do. The corrected figure independently reproduces what an entirely separate model built ten weeks earlier gives for the same neighborhood, which is the strongest confirmation available that it is right.

One consequence is worth stating plainly. The same spreadsheet contains an earlier and a later version of the whole calculation, and on one neighborhood they appeared to disagree by a factor of 2.3 — which for a time looked like grounds to distrust the entire source. It was the first error in the table above: a single cell. Corrected, the two versions agree within 2.4 mills on that neighborhood, and within 3.8 mills on average across the nineteen that appear in both.

Neighborhood PIDs

For each neighborhood the calculation is the same. Measure the street area. Multiply by the County’s reconstruction cost. Escalate to the year the work would be done. Amortize over twenty years. Add the annual maintenance the roads will need afterward. Divide by that neighborhood’s own assessed value.

Streets already in good condition are treated differently from streets that are failing. A street projected to stay above a pavement condition index of 70 does not need full-depth reconstruction yet, so it is left out of the capital project — but it still enters the maintenance program, because every road eventually needs work. Four neighborhoods therefore rebuild less street than they maintain. Old Town Niwot is the clearest case: it has the best pavement in town and rebuilds roughly a third of its street area, which is why it sits near the bottom of the range.

Across the 24 neighborhoods the result runs from 13.77 mills at the low end, Monte Vista Avenue, to 47.24 mills at the high end, Centrebridge. The published figure is 14 to 47 mills, each end rounded to the nearest whole mill.

This table is the derivation, not the published figures. It carries the mill levies to the hundredth because that is what the parcel arithmetic produces. What gets published is the whole mill and the dollars that follow from it — so the floor publishes as 14 mills and $976 a year, not 13.77 and $960. Both are shown so the rounding is visible rather than silent.

A note on how we round, because we got it wrong once. These figures are rounded to the nearest whole mill. An earlier draft of this page truncated instead — 13.77 to 13, 14.68 to 14 — and defended it as rounding “against our own argument,” on the reasoning that understating a district makes our case harder to make. The reasoning is sound and the application was not. Being careful means publishing the cheapest scenario a district could plausibly be built on; it does not mean misstating a number. A figure that fails a reader’s own division does not read as caution, it reads as a mistake, and this page exists to be checked.

What separates one neighborhood from another is mostly street area per household. A subdivision with wide streets, long cul-de-sacs and few homes has more pavement to pay for and fewer people to pay for it. That is why neighbors comparing notes will find very different numbers, and why the spread is not a sign of error.

The low end matters more than its size suggests. Until September 2026 this range was published as 20 to 40+ mills, then as 18 to 45. The real floor is 13.77 mills on Monte Vista Avenue, and every earlier version of the floor made the cheapest possible district look more expensive than it is — rounding in the direction that favored our own argument. Both ends have since been rebuilt from the parcel data: four tax bases had been summed over the wrong parcels, Springhill was reset to the 30 mills the County itself quoted its board, Somerset was rebuilt from the streets its own project pays for, and Monte Vista Avenue entered the set. The published range is now 14 to 47.

The Townwide PID: Measuring the Base

A townwide district spreads the same road program across every property in Niwot. The annual requirement is $3,392,018 — twenty-year debt service on $25.8M of reconstruction, plus ongoing maintenance.

The question is what to divide it by. Rather than adopt a tax base figure from the spreadsheet, we measured one from the parcel records. The spreadsheet’s own townwide base is exactly $2,825,000 divided by 0.015 — back-solved from an assumed 15-mill target rather than counted. Dividing a fresh cost estimate by a base that an older mill target defined is circular, so we did not use it.

Counting instead, using the County’s own assessed values and excluding tax-exempt property, gives two boundaries that differ by one real question:

The 2025 proposal left the downtown business district out of its district — 23 parcels on Second Avenue and five in Cottonwood Square — because downtown already pays into the Local Improvement District. That is a defensible design choice, and it is also a real transfer: every dollar of assessed value outside the district is a dollar the rest of the community makes up.

Business Personal Property is the second question, and the spreadsheet never addressed it. Equipment at the Tech Center is taxable property assessed at 25%, and a townwide district would ordinarily levy on it. Leaving it out raises everyone else’s rate by about a mill.

Cross-check, and a gap we have not closed. The town budget model projects $833,000 of property tax at 4 mills, which implies a base of $208,250,000. The parcel count above, with downtown included and Business Personal Property added, comes to $200,438,957 — 3.8% short, about $7.8 million of assessed value this model cannot account for. An earlier version of this note reported the two agreeing within two tenths of one percent; that agreement came from reading the equipment figure as taxable value when it was market value, which made the check pass by absorbing the error it existed to catch. If the budget model’s base is the right one, every townwide levy on this page is about 4% too high — roughly two thirds of a mill at the bottom of the range. We publish it as measured rather than adjust toward a figure we cannot reproduce.

Those four cells are one question asked twice — is downtown in, and is business equipment in — and each answer is worth about a mill. But there is a third axis, larger than both, and it was missing from this table until 16 September 2026.

Fifteen parcels carry more than a fifth of the base. Boulder Tech Center and Monarch Park are $40.4 million of assessed value plus $5.4 million of business personal property, which is 22.8% of everything a townwide levy could fall on. Taking it out moves the levy about four mills, not one:

The spread is not uncertainty about the cost — the pavement is priced identically in all four rows — it is the unanswered question of who is in the district. That question moves both halves of the ratio, which is the part worth being careful about.

A district pays for the roads it contains

This correction was made on 18 September 2026, and the error it fixes is instructive. The two Tech-Center-out rows above used to divide the same bill as the other two. That is wrong: a district drawn without the Boulder Tech Center does not collect its property taxes, but it also does not repave its streets. Holding the bill fixed while shrinking the base charged the remaining 1,794 households for 1.4 miles of industrial parkway their district would have no responsibility for — and it overstated the top of the published range by about 1.4 mills, in the direction that flatters our own argument.

The Tech Center’s pavement is seven segments on the county survey: Dry Creek Parkway, East and West Dry Creek Parkway, and Monarch Park Place. 1.4 miles, 237,841 square feet, three of the seven already projected below a condition index of 25 by 2029. That is 6.2% of the town’s pavement carrying 7.1% of its capital cost — wide industrial roads in worse-than-average shape.

Removing the Tech Center cuts the bill 6.8% and the tax base 22.8%. That asymmetry — over three times as much taxable value as road — is precisely why its exclusion is expensive for everyone left, and it survives doing the arithmetic properly. The capital share is pro-rated rather than subtracted directly, because the community-wide capital figure is bucketed (1,750,000 sq ft at $11 plus 700,000 at $5.50) while the per-street column prices each segment off its own condition index; the two totals agree to 1.2%, and the choice moves the published levy by a hundredth of a mill. derive.py --roads prints the whole subtraction, segment by segment.

We do not publish these as a price. Each figure above is a first rung under one boundary, with every property inside paying, and none is a levy anyone would be charged. Until 12 October 2026 we quoted them as a range; since then the site states a townwide cost only as the build-up from the 2025 proposal’s 12 mills, which starts at 18 on that proposal’s own boundary and settles near 31 (the math).

The neighborhood pages test against 17 mills — the lowest first rung, the row most favourable to a district, and the only one levied on the same base as the town’s own 4 mills. A conclusion that survives the most generous assumption available to the alternative survives all four. Incorporation prices below a district in all 34 neighborhoods even at 17 mills. The unrounded low end is 16.92, a twelfth of a mill below the 17 we anchor to; at 16.92 the count is still 34 of 34, so the conclusion does not turn on the rounding. The closest two neighborhoods, both on low-value homes, are within a few dollars a month and are named on the districts page. At the old 15 mills a district came out cheaper in those two.

Why the bill rose in October 2026

Until 11 October 2026 this page divided a bill of $3,051,636, and the range was 15 to 20 mills. The difference is one factor in the source workbook. Every townwide row charges the district 21/25 of its yearly bond payment; no neighborhood row does. It is the county’s 2025 method — Boulder County Public Works told us the 2025 levy spread “total costs for 22 miles of road … over a 25-30 year life cycle” — and on the 2025 proposal’s calendar it was a real plan, not a mistake. Collect from 2026, build in 2029, and four years of collections at 84% of a payment bank exactly what twenty-one years at 16% short then use up: 4 × 0.84 = 21 × 0.16.

A district started now does not have those four years. It reaches a ballot in November 2027 at the earliest, first collects in 2028, and the county expects work in 2029. With one year in hand rather than four, the district has to meet its whole bond payment from the first repayment year. That is how every neighborhood district on this page is priced, and how the town’s own budget carries its road bond, so it is also the like-for-like basis. Both sides are priced on the same roads, in the same condition, on the same 20-year terms. A 25-year bond is the generous alternative, and the feasibility paper states what it would change.

This correction runs in our favour, so here is what it does not touch. The neighborhood levies, 14 to 47 mills, never carried the factor and do not move. The unit costs, the decay rule and the tax base are unchanged. And the base gap in the cross-check above still runs the other way.

Every published levy on this page is a whole mill, rounded to the nearest. The derived values are given in the tables above and in the conservatism ledger below, because a derivation should show its arithmetic; what gets published does not carry two decimal places it cannot support. The numerator is a county pavement survey priced at a generic $11 a square foot.

Why a Mill Levy Does Not Rise With Inflation

This is the part that is most often gotten wrong, including by us in earlier drafts, and it is worth setting out carefully.

A mill levy is a ratio. Construction costs inflate, and so do property values. If both rise 3% a year, the numerator and the denominator rise together and the rate does not move. A district needing 17 mills in today’s dollars needs 17 mills in 2029 dollars. Inflation cancels.

It follows that quoting a cost in one year’s dollars against a tax base in another year’s produces a number that means nothing. It is an easy mistake to make — the source spreadsheet amortizes its construction cost before inflation in one row and after it in the next — and it always makes the levy look smaller than it is.

What does not cancel is deterioration. As pavement falls below a condition index of 50, a street that could have been resurfaced now needs full-depth reconstruction, which costs several times as much. That grows the numerator without touching the denominator. Between the 2024 and 2027 condition assessments, 2.4 miles of Niwot street fell below 50 and another 1.6 miles below 70, adding $2.74 million of scope:

So the levy falls over time only if property values appreciate faster than construction inflation and road deterioration combined. In Niwot they do not:

Costs outrun values by roughly a point and three quarters a year. A PID quoted at 17 or 18 mills today is a floor, not a ceiling, and every year of deferral raises it. This is the same conclusion the road cost model reaches from the other direction: the price of waiting is not inflation, it is scope.

The Three Options, at the Same Home

All three figures below are for the same median Niwot dwelling, valued at $1,095,000, using the statutory $70,000 value reduction and the 6.8% residential assessment rate. Pricing one option at a median and another at an average is the most common way this comparison is made to mislead, and it is avoided here.

Three things follow from the table, and they are worth separating from any argument about them.

First, the cheapest possible PID outcome in Niwot still costs more than incorporation — and that cheapest outcome requires the whole town in one district, downtown included, business equipment in the base, and the work starting before the pavement degrades further.

Second, the reason is structural rather than rhetorical. A PID is funded by property tax alone. A town also collects sales tax, and a meaningful share of that is paid by people who do not live in Niwot — visitors, contractors, and deliveries into the boundary. The same road program spread across a broader base costs any one household less. That is the whole of the difference.

Third, no PID proposed in Niwot has included snow. Boulder County’s improvement districts have been capital programs, and neither the 2013 county-wide district nor the 2025 citizen proposal asked for a dollar of snow removal. A district’s levy buys what its petition asked for, and under every PID scenario on the table snow stays with the County.

Where We Rounded, and Which Way

A comparison that favors one side is worth nothing, so the rounding rule here is to round against ourselves. Where a judgment could go either way, we took the one that makes a PID look cheaper and incorporation look less advantageous.

The typical neighborhood figure was published at 32 mills for some months, about two mills below what the parcel data produced at the time, and was deliberately left there rather than raised. It is now 33.11, the home-weighted average across the 24 districts — weighted by homes rather than by district, because a district with 21 homes should not count as much as one with 140. Published as 33. A number that overstates what a neighbor would pay is the one kind of error this comparison cannot afford, even when correcting it would help the argument.

One caveat in the other direction, stated so a critic does not have to find it. The typical neighborhood figure of $192 a month prices the town median home at an average neighborhood rate. A household actually pays its own neighborhood’s rate on its own home, and across the 24 neighborhoods that averages closer to $258 a month. The published figure is the conservative one.

What This Does Not Settle

Three questions remain open, and none of them are rhetorical.

• Who would be in a townwide district. Whether downtown is in, whether business equipment is in, and whether the Boulder Tech Center is in together account for the whole spread between the four first rungs. All three are decisions about a boundary, not facts to be discovered — and the third is much the largest, because 15 parcels carry a quarter of the base. See Road Districts.
• What the County would agree to fund. Every figure here assumes County reconstruction standards at County cost estimates, with no contribution and no subsidized financing. If the County were to restore either incentive, every PID number falls.
• Pavement condition after 2027. The condition scores are engineering projections from a 2024 survey. If the roads degrade faster than 2 points a year, the levies rise faster than shown here. If slower, they rise more slowly. They do not fall.

We will publish revisions to this page when the answers change the figures, including when they move against us. Two corrections have already been made in public on the household cost figure, and both raised it.

Sources

• Boulder County Assessor — parcel-level actual and assessed values for all 2,015 parcels considered, with boundary and district membership flags
• Boulder County engineering — pavement condition index survey, 2024, scored street by street
• Boulder County — 2025 Public Improvement District cost schedule: $11.00/sq ft full-depth reconstruction, $5.00/sq ft mill and overlay
• Boulder County — Subdivision Paving PID program
• Piper Sandler — twenty-year amortization schedule at 4.62%
• Pro Forma Budget v1.5 — revenue note R1, assessed value base and growth assumptions
• The PID Proposal — the 2025 citizen proposal and its boundary
• Household Cost Methodology — the derivation of the $63 figure and the median home value used throughout

The calculations behind this document are kept as scripts rather than spreadsheets so they can be re-run against changed inputs, and so that an error in one of them can be found by reading it. That is how the four corrections above were found.