Pro Forma Stress Testing

The pro forma answers what a town would cost. This answers a different question — how wrong could it be, and how wrong would it have to be before the answer changed. Every budget line is tested to the point where it would exhaust the town’s reserves, and the result is published whether it flatters the model or not. Written for a reader who wants to attack the numbers rather than accept them. The first part is the answer and the conclusions we draw from it; everything after it is the working. 19 September 2026; every model figure regenerated for version 1.13 of the pro forma, 15 October 2026.

The Short Version

An incorporated Niwot would run on about $4.6 million a year and hold a reserve alongside it: about $2.2 million at its low point in 2028, growing to $6.6 million by 2040. That reserve is the town’s cushion, and it is what this document tests.

For every line in the budget we asked one question. How far would this number have to be wrong before the cushion was gone? The pro forma already asks that of six lines. We asked it of all thirty-six.

Twenty-nine of them cannot matter. Each would have to be wrong by more than its own entire size before the reserve ran out. Five of the twenty-nine could not empty it even at twenty times their budgeted size.

Seven lines could matter. They are the road bond’s repayment and its construction cost, ongoing road maintenance, the Sheriff’s contract, property tax, use tax, and sales tax, and they are taken in that order below. Six of the seven already have an answer. The seventh is the one to look at.

Both road bond lines are already answered

The repayment is capped by the voters. Reserves would be threatened if debt service came in 45% above the model, but the repayment limit written into the ballot question binds first, at about 15%. The town cannot lawfully issue a bond expensive enough to reach the threshold.

On construction cost, the reserve is the last line of defence rather than the first. An 18% overrun would exhaust it only if the town built the full road programme at the higher price and paid the whole difference from the cushion. A council has other options first — scope, timing and grants — and the budget already carries contingency above the priced cost of the work.

Two would be absorbed rather than paid for

Ongoing road maintenance is mostly not spending at all. Most of that line is money set aside toward rebuilding roads decades from now, and it can be deferred. That slows the reinvestment cycle, which is a real cost, but it does not empty a reserve.

The Sheriff’s contract already carries its own margin. It funds two full-time equivalents of Sheriff service against the roughly two thirds of one that Niwot’s call volume draws today, and even so, Niwot’s modelled public safety cost is the lowest among its peer towns. The threshold sits on top of a margin we have already built in.

Two would need something without precedent

Property tax would have to fall by nearly half. The largest single concentration in the base is the town’s two commercial parks, Boulder Tech Center and Monarch Park, at about a quarter of its taxable value. Losing both outright would be a little over half of what is needed.

Use tax would have to fall by 87%. Construction and vehicle purchases in Niwot would have to stop almost entirely, and stay stopped.

Which leaves sales tax

Sales tax is the real one. The reserve would be exhausted by a permanent 19% fall in the town’s whole taxable base. Everything then turns on what that 19% is measured from — and the base has two parts, each set low for a different reason.

The local third, measured from the downturn rather than the good years

The pro forma takes local retail at about $25 million. That is not a forecast. It is the Local Improvement District’s actual 2025 receipts, recorded after the decline this committee has spent two years documenting: district activity peaked at $32.3 million in 2022 and had fallen to $25.0 million by 2025, a real-terms drop of about 23%, leaving revenue slightly below its 2019 level.

So the 19% is measured from the trough, not the peak. And because local retail is only about a third of the base, a 19% fall in the whole of it means losing roughly three-fifths of all local retail, on top of the 23% already gone. Measured from its 2022 peak, Niwot’s commercial district would have to be down about 70%, and stay there. That is not a recession or a bad decade. It is the commercial district substantially ceasing to exist.

The remote two-thirds, set below what comparable towns actually collect

The larger part of the base is online, delivered and contracted purchasing, which does not depend on any storefront staying open. We could not measure it for a town that does not yet collect tax, so we estimated it from two Front Range towns that publish their own figures — and assumed less than either of them reports.

Niwot is modelled between the two: above Lyons, below Louisville, the town it most resembles on income. Had we assumed Louisville’s published rate, the threshold would be about 29% rather than 19%, and the town would collect roughly $272,000 more a year. We did not take it, and the report sets out the structural reason rather than leaving it as caution.

That is the shape of the number this whole document turns on. One third of it is a measured figure taken at its low point; two thirds of it is an estimate set beneath two published observations.

What it adds up to

Twenty-nine lines cannot move the answer, and five cannot move it even at twenty times their size. Of the seven that could, two are foreclosed by a cap the voters themselves set, two would be absorbed rather than paid for, and two would need falls with no precedent. The last would take a commercial collapse several times larger than the one Niwot has already lived through and has already budgeted from.

On this test, the budget is robust. Not because we are confident about every line in it, but because almost no line can matter, and the one that can is measured from a depressed base rather than a hopeful one.

That is a claim about structure rather than a promise about the future. Its limits are set out in full below, including a defect we found in our own simulation and the places where the budget is set in our favour rather than against us.

A reader content to take the summary on trust can stop there. What follows is the working underneath it: the budget being tested, the method and how it was checked against the pro forma’s own published results, the full thirty-six-line ladder rather than the seven, the road bond in detail, where the budget is deliberately set against us, what a council can do in response, and the questions we cannot answer.

What the Pro Forma Says

This document tests a budget, so the report starts with the budget itself. Every threshold in it is measured against these figures, and none of them means much without them.

An incorporated Niwot is modelled as a town of about 4,300 residents running on roughly four and a half million dollars a year. Taking 2030, the first fully settled year of operations:

Figures on this page are in thousands, as the pro forma presents them — $4,581K means $4,581,000.

The town also accumulates a reserve. At the end of 2028 it stands at $2,177K, its lowest point in the whole projection. It slips back slightly in 2030, the one year the budget runs a deficit ($4K), and grows in every other year to $6,613K by 2040.

What “reserves reach zero” means, and what it does not

Every threshold in this document is the point at which one error, on its own, would drive those cumulative reserves to zero at some point before 2040. That is the failure condition used throughout, so it is worth being exact about what failing would actually mean.

It is not bankruptcy, and it is not the town ceasing to function. Reserves are the cushion a town holds above its spending. Exhausting them means a council has run out of cushion and must bring spending into line with revenue in that year’s budget — which, under Colorado’s Local Government Budget Law, it is required to do in every year regardless. The practical consequence is reduced services and deferred road work. It is a bad outcome. It is not insolvency.

We use it as the test because it is a bright line, and because the pro forma already uses it. It is a demanding standard rather than a catastrophic one, and a reader should hold it in mind on every row below.

What This Adds

A budget projection faces three separate questions, and they are easy to confuse. Most objections we receive are really an objection to one of them aimed at another.

The pro forma’s stochastic testing is a thorough answer to the first question. It says nothing about the second or third, and a reader who takes its 5.10% as covering all three is reading more into it than it claims. That gap is what this document exists to close.

The method here is the pro forma’s own. Its Break-Even Sensitivity table already asks “how far would sales tax have to fall before reserves are exhausted” and answers it for six variables. We extended the identical calculation to every line in the budget — thirty-six of them — and it reproduces all five of the published thresholds to within a tenth of a percentage point. That reproduction is the only reason the rest of the table is worth reading.

How Wrong Would Each Line Have to Be?

Each row below asks a single question: if this one line were permanently wrong, in the direction that hurts, by how much would it have to be wrong before the town’s cumulative reserves reached zero at any point between 2028 and 2040? Everything else is held at the pro forma’s values.

This is the most useful thing in this document, because it converts a disagreement into arithmetic. A skeptic need not accept our probability distributions or our economic assumptions. They need only answer, for themselves, whether they believe the town manager will cost four times what we budgeted.

The seven lines that can move the outcome

These are the only lines in the budget that could exhaust reserves if wrong by less than their own size. They are where scrutiny belongs.

† This is the fourth of four cushions, not the first. The capital budget already carries contingency, and a council has cheaper responses than spending reserves. Set out in full below.

‡ Reserves break at this level, but the voter-approved repayment cap binds long before — again, see the next section.

Two of these deserve a note. The Sheriff line is budgeted at 2.0 full-time deputies against the roughly 0.67 FTE that Niwot’s call volume draws today, so it is already carrying three times current service before this test begins, and the +94% is on top of that. That margin is a service-level choice rather than padding for its own sake, and Appendix D of the pro forma shows where it lands: at 2.0 FTE, Niwot’s modelled public safety cost of $88 per resident is the lowest in a peer set of four Front Range towns, against $150 for Berthoud on the same sheriff-contract model. And ongoing maintenance is mostly not spending at all — the majority of that line is an annual set-aside toward a rebuild cycle decades away, which is why it appears again in the response section below as the largest single thing a council could defer.

The lines that would have to be wrong by more than their own size

The five lines that cannot change the answer at any realistic size

Mayor ($12K), codification and legal notices ($15K), materials costs ($19K) and election costs ($13K) cannot exhaust reserves even if each came in at twenty times its budget. The Road and Bridge Fund ($18K) is revenue, and losing every dollar of it would not exhaust reserves either; since version 1.12 of the pro forma halved it to the statutory 50% share, it is among them.

This is the least intuitive result here and the easiest to misread as a modelling artefact. It is not one. These lines are simply small enough that the reserve balance absorbs any multiple of them. It is worth stating because public argument about a municipal budget tends to concentrate on the lines that are easiest to picture — what the mayor is paid, what an election costs — and those are precisely the lines that cannot matter. We have spent real effort ourselves arguing about lines that cannot move the answer.

How Likely Is Any of This?

A threshold is not a forecast. Saying that sales tax would have to fall 19% tells you nothing about whether it will. Those are different questions, and the second one cannot be answered by a model at all — it has to be argued. Here is the argument for the two that carry the most weight. The road bond, which is the tightest line of all, has a section of its own after this one.

Sales tax: the threshold is larger than the worst thing that has actually happened

This is the line that genuinely matters, so it deserves the most scrutiny rather than the most reassurance.

The modelled base is about $80 million a year, in two parts:

A 19% fall is about $15 million of that base, permanently. Two ways to picture what that would take.

From downtown alone, it is not reachable. All local retail is $25 million. Losing $15 million of it means downtown Niwot permanently losing roughly three-fifths of its entire commercial activity — not a bad few years, a structural collapse of the commercial district.

And we know what a bad stretch looks like, because Niwot has just had one and it is documented in the district’s own revenue records. In 2025 dollars, taxable activity in the district peaked at $32.3 million in 2022 and fell to $25.0 million by 2025 — a decline of about 23%, which put real revenue slightly below its 2019 pre-pandemic baseline. That is a genuine downturn and this committee has written a great deal about it.

That decline is already inside the budget. The $25 million above is not an estimate of what downtown might earn; it is the 2025 receipts, after the fall. The pro forma projects forward from the trough rather than from the 2022 peak, so a reader weighing the −19% threshold is weighing it against a revenue base that has already absorbed the worst stretch in the district’s recent record.

It is also only about 37% of the way to this threshold. A decline of that same size again, from today’s level, would remove about $5.7 million — roughly 7% of the total base, against the 19% that would be needed. The downtown decline that motivated much of the case for incorporation is, on this specific measure, well inside what the budget absorbs.

From remote sales, it means households changing their behaviour. Reaching $15 million there would require Niwot households to permanently reduce online, delivered and contracted purchasing by about 28%. That is not a business-closure risk. It is a claim about sustained household consumption in one of the higher-income communities in the state.

How much confidence the $55 million carries

It is the larger half of the base and it is an estimate, so it is the right thing to press on. It was derived per-resident against two Front Range towns that publish or provided the equivalent category:

The conservatism is in the mismatch between those two columns. Niwot has the highest per-capita income of the three and is modelled at 87% of Louisville’s observed rate — below the community it most resembles on income, and only modestly above the smaller, lower-income anchor. Since remote purchasing is driven substantially by household purchasing power, the model assumes Niwot behaves less like its income peers than its income would predict.

Both benchmarks are public documents, and you can check them. Louisville’s figures come from its Finance Committee’s Sales & Use Tax Report, which reports an “Outside City” category for vendors with no location in the city. Lyons publishes a monthly local-versus-non-local split in the Finance Update attached to its Board of Trustees packets, defining non-local in its own words as “online businesses, mobile vendors/catering, deliveries from out-of-town companies, natural gas (Xcel), etc.”

On both, we assumed less than the town itself reports:

And we found two errors in our own arithmetic on Lyons while checking this, both of which had made the town look less remote-heavy than its published figures show. Our working paired Lyons’ 2025 monthly shares with its 2024 annual total, about $187,000 lower than 2025’s, and it averaged the monthly percentages rather than weighting each month by its own revenue — which matters, because Lyons’ busiest months are also its most local. July is 61% local, August 62%, January 46%. On the Town’s full-year 2025 figures, published in its April 2026 finance update — $2,232,733 in sales tax at 4.0% — Lyons’ non-local share is 47.2% and the per-resident figure is the $12,382 above. We have left the lower number in the budget.

So what if the estimate is wrong?

This is the one input where we can answer that with observed numbers rather than invented ranges, because two real towns publish theirs. Niwot is modelled at $12,777 per resident, which sits between them — above Lyons’ $12,382, below Louisville’s $15,302. So rather than guess, we can simply run the budget at each town’s actual rate and see what moves.

Two things follow, and the second is the more useful one.

The downside is bounded by something real. If Niwot turns out to shop like Lyons — a town with a third less income per resident — the threshold tightens from about 19% to about 17%, and reserves still reach roughly $5.9 million. That is the pessimistic benchmark, and it is not a cliff.

The upside is larger than the downside, and we did not take it. Niwot’s per-capita income is $85,000 against Louisville’s $71,000, and remote purchasing tracks household purchasing power more than anything else. If Niwot behaved like the town it most resembles on income, the remote base would be about 20% larger, the town would collect roughly $272,000 more a year, and the sales tax threshold would loosen to about 29%. We budgeted below that on purpose, and every figure in this document is computed on the lower assumption.

Why we did not simply take Louisville’s number

There is a structural reason, and it matters more than caution. Louisville’s “Outside City” is not a consumer category. It counts every vendor with no location in the city, which includes business-to-business supply, contractor purchasing and wholesale alongside household shopping.

The City told us how that happens. Asked how it classifies remote sellers, Louisville’s sales tax technician explained that classification comes from “the information they send us,” and that “we get lots of payments through the Colorado SUTS program, most of those accounts default to being general merchandise in our system.” SUTS is the state portal out-of-jurisdiction sellers use to file into self-collecting cities — which is to say, the very population that makes up the category. Remote business supply and remote household shopping land in the same bucket, undifferentiated. How much lands there is now bounded. The City told us on 21 September that about two-thirds of Outside City collections arrive through SUTS, against a General Merchandise line that is 32.0% of the base — so the default bucket holds at most a third of the category, and the majority of what files through the portal sits in lines the City did classify.

Louisville has roughly five times Niwot’s population and a substantially larger commercial and employment base. A residential community with a small commercial district should not expect to generate that business-supply half in the same proportion, however similar its household incomes. So the $272,000 above is a ceiling we would probably not reach, not revenue we are forgoing. We cannot size the part of Louisville’s rate that is business supply, because the City cannot either — it told us plainly that it has no way to know who buys what without auditing a vendor one at a time.

That is the honest shape of it. We are not claiming the higher number, and we are not pretending the gap between us and Louisville is all money left on the table. What we are saying is narrower and firmer: this input is bracketed by two towns’ published accounts rather than by our own judgment, we sit inside that bracket, and we sit in its lower half.

One caveat stands and should not be smoothed over. The two towns are not the same instrument. Louisville is home rule and self-collecting, and its “Outside City” category sweeps in business-to-business, contractor and wholesale activity. Lyons is a statutory town whose tax the state collects, and its non-local category explicitly includes utilities. The pro forma calls this a directional validation of magnitude rather than a category-identical reconciliation, and that description is right. This remains the least-measured major input in the budget — but it is now the one where the modelled figure sits between two independently published observations, in the lower half of that bracket.

Property tax and use tax: what would have to happen

Property tax would have to fall 45%. Two mechanisms could move it, and they are worth separating.

The assessment rate. Colorado has changed residential assessment rates repeatedly in recent years — the Gallagher repeal in 2020, then SB 22-238, SB 24-233 and HB 24B-1001 — so this is a live mechanism rather than a theoretical one, and the pro forma says plainly that it does not model future statutory changes. But none of those changes is remotely of this size. Property tax is also under a quarter of modelled revenue, and the line lags the market by the county’s biennial reassessment cycle, so it moves slowly enough to be budgeted around.

Concentration. The more specific risk is that the base is not evenly spread. Boulder Tech Center and Monarch Park are fifteen parcels holding roughly a quarter of the town’s taxable value, and they share a boundary with the City of Boulder, which makes the question of whether they stay a real one rather than a hypothetical. We would rather put that in front of a reader than have it raised for us.

So it is worth sizing. Losing both parks outright is a little over half of this threshold — a serious blow to a young town’s finances, and not on its own an exhaustion of its reserves. It would also not happen quietly or quickly: it would be a visible process, in public, with a council able to respond in its budget as it unfolded.

Use tax would have to fall 87% — construction activity and vehicle purchases in Niwot stopping almost entirely, and staying stopped for a decade. It is the smallest of the seven lines and the one furthest from its threshold.

The Road Bond, and Where the Authorization Binds

The tightest line in the table above is the road bond capital outlay at +18%, and it is not in the pro forma’s published six-row table. We found it by running this analysis. It is the number an opponent would most like us to have left out, so here is what is behind it.

Read bare, the row says that if the road programme costs 18% more than budgeted, reserves reach zero. That reading is wrong in one specific way: it measures the last cushion as though it were the only one. There are four, and reserves are the fourth.

First: the budget already carries contingency

The $15.725 million is not an engineering estimate of the work. It is an engineering estimate plus two layers of allowance, and the ladder is published on our roads page:

Inside the cost model those allowances are explicit parameters: a 10% contingency applied to the treatment total, and a further allowance of about $1.1 million for bridges, drainage, sidewalks and similar assets, to be scoped in the engineering phase.

So the +18% is measured against a budget that already sits well above the priced scope. A programme would have to exceed the engineering estimate by considerably more than 18% before it exceeded the budget by 18%.

Second and third: a council reduces scope, extends timing, or seeks grants

These are the pro forma’s own stated responses to a cost or rate surprise, and none of them costs reserves anything. A road programme is not an obligation to spend a fixed sum — it is a list of streets, and a council that receives high bids can build fewer of them this cycle.

That has a real cost, and we would rather name it than imply it is free: deferring a street extends the reinvestment cycle and raises what that street eventually costs to rebuild. It is a worse outcome. It is not a solvency event.

Fourth: reserves, and that is what +18% measures

Only if a council chose to build the full scope, at the higher price, funding the difference entirely from reserves and doing none of the above, does the +18% threshold come into play.

Two further things are worth knowing. The town cannot simply borrow the difference. The ballot question caps principal at $15,000,000 and the model already issues the entire authorization — the ~$16 million appearing as 2029 revenue is what $15 million yields after issuance premium and underwriter costs, not spare borrowing capacity. Covering an overrun with a larger bond would mean going back to the voters.

And the town’s reserves are at their lowest in 2028, at about $2.18 million — before the bond is issued in 2029 at all. Reserves rise in every year after that except 2030, when they slip back by a few thousand dollars.

The cap binds before the finances do

Testing the bond turned up something more useful than the thing we set out to check. If interest rates at issuance came in above what the pro forma models, three separate limits would be approached in this order:

The town cannot lawfully issue a bond expensive enough to threaten its own reserves. The cap voters approve stops it at roughly a third of the way there. If rates at issuance were high enough to press against that ceiling, the town would have to reduce the size of the borrowing or restructure it — which is what the cap is for.

Unlike almost everything else on this page, this one does not depend on trusting our model. It is arithmetic on the certified ballot title and the fiscal filing, both public documents.

Where We Overstated Our Own Costs

The word “conservative” appears throughout this document and throughout the pro forma, and a reader is entitled to ask whether it is doing any work. So here is the ledger: every place we know of where the budget is deliberately set against our own interest, and then the places where it is set in our favour.

A budget built of conservative estimates rather than best estimates has a consequence that matters for reading the ladder: several of these lines can only be wrong in one direction. A threshold that says the Sheriff line would have to be understated by 94% is measuring from a figure already set at three times current service.

Set against us

Set in our favour

A ledger with only one column is a sales document. These run the other way, and both are load-bearing:

We have not attempted to net these against one another into a single number. The conservative items are mostly specific and sized; the two against us are uncertainties rather than quantities, and adding a sized figure to an unsized one to produce a reassuring total is the exact move this document exists to avoid.

The Property Tax Base

Three assumptions stand behind the property tax line. All three are stated in the pro forma. None has been tested in print before. That is what this section does.

How these figures are produced. Every table in this section is generated by the property tax scenario model, which rebuilds the line from Assessor values for 1,838 parcels, varies one assumption, and measures the change against the pro forma’s own published row. The probabilities are run on the same 100,000 simulated economies as the headline, so each as-modelled row reproduces its 5.10% exactly and every other row is directly comparable with it.

How much rests on them. Property tax is $1,422K of $6,373K — about 22% — of modelled 2040 revenue, and less than a quarter in every year. Sales and use tax together are more than twice as large. That ratio is the reason the answers below come out the way they do.

Assumption 1 — Market value growth of 4.4%

The residential figure comes from the S&P CoreLogic Case-Shiller Home Price Index for Denver. Measured to August 2025, that index has compounded at 5.39% a year since August 1989 and 4.55% a year since January 2000. The national index returns 4.15% and 4.77% over the same two windows. The modelled 4.4% is below both Denver figures — below the realised long-run rate for the market Niwot is actually in, over two windows that each contain a full housing cycle including the 2007–2011 crash.

Assume property values in Niwot never rise again in nominal terms for fourteen consecutive years — against 3% inflation, a 34% decline in real terms — and cumulative reserves are still positive in every year, ending 2040 at $3.3M. The lowest point of the projection does not move at all, because it occurs in 2028 before compounding has had time to matter.

Reserves are exhausted only if residential and commercial values fall about 5.4% a year, every year, through 2040 — a 49% cumulative decline with no recovery. Denver has never done that.

Assumption 2 — The incorporation lift

The model applies an additional 5% a year for two years, on top of baseline growth, to both residential and commercial value. On a home worth $100 at the start of that reassessment, values would reach $108.80 without the lift and $118.80 with it. That is $10, or 10% of its starting value, and it leaves property tax 9.2% higher than it would be without the lift. Both describe the same assumption; the tables below compare against no lift, so they use the second. The pro forma has always disclosed this. Assumption A3 states that “an additional up to 5% growth adjustment is applied to property market values in the first two years following incorporation as a planning assumption,” that “actual impacts may be lower or zero,” and that the pro forma “remains structurally solvent … if this post-incorporation adjustment is 0%, as it is not relied upon to fund core services.” What has not been published before is the evidence the assumption rests on, and the arithmetic behind that solvency claim.

Incorporation changes who sets land use, who maintains the street and who answers for local services, and housing markets price that. The closest published evidence is Patrick and Mothorpe (2017), Regional Science and Urban Economics 67(C):78–89, studying seven new cities incorporated in metropolitan Atlanta from Sandy Springs (2005) onward. Using difference-in-differences hedonics with matched controls, parallel pre-trends and placebo tests, they find property values inside a new city rose 4–5% in the two years following incorporation and 12–13% over the full analysis period, with the effect strongest where values inside the boundary were high relative to the surrounding unincorporated county — Niwot’s situation. Oates (1969) is the underlying result: local spending capitalises positively, local taxes negatively, and the market prices the net.

It is not universal. Hodges and Stephens (2022), Growth and Change 53(1):200–220, analyse four incorporations in Riverside County, California using transactions from 2000 to 2014 and find, “in contrast to previous studies in other regions, … a generally negative capitalization effect observed post-incorporation.” Whether an incorporation raises values depends on what the government does and what it costs — which is why this section tests the assumption rather than resting on it.

Two things about how the model uses it. The lift sits between the study’s two estimates — above the two-year figure, below the full-period one. And the timing is conservative: it is applied at the 2031 revaluation, so it first reaches collected tax in 2033, rather than at the first revaluation after incorporation. Applying it there instead would raise property tax by $344K and 2040 reserves by $344K.

Assume incorporation has no effect whatever on property values and reserves are positive in every year, the low point does not move, 2040 reserves are $5,755K, and the probability of reserves falling below zero is 6.61%. One qualification we owe the reader: Patrick and Mothorpe estimate residential capitalisation. Applying the same lift to commercial property is an extension their work does not directly support, which is why the residential-only rung is published above.

Assumption 3 — The residential assessment rate

This is the largest of the three exposures and the only one nobody in Niwot controls. The model assumes 6.7% of market value, after the statutory reduction of 10% of value or $70,000, whichever is less. Colorado has changed this rate four times since 2020 — the Gallagher repeal, SB 22-238, SB 24-233 and HB 24B-1001 — and can change it again in any session.

A cut to 5.5% — a larger move than any single change in the recent sequence — costs about $1.86M of 2040 reserves and leaves them positive in every year.

All three wrong at once

The scenarios above vary one assumption at a time, which understates risk because real adverse periods move several together. So here is the case where every one of them goes against the town simultaneously: no incorporation lift, zero nominal appreciation for fourteen years, and the residential assessment rate cut to 5.5%.

Reserves remain positive in every year of the projection. But the shape changes, and that is worth saying plainly: in this case the lowest point moves from 2028 to 2040, which means reserves are no longer accumulating at the end of the horizon. That is the signature of a structural problem rather than a startup dip, and it is the case a Council would be watching for.

The finding

Tested individually and together, none of these three assumptions determines whether the town is solvent. The deterministic projection survives all three going wrong at once, and the break-even analysis above already says why: this line would have to fall 45% before reserves are exhausted, and property tax is only about 22% of revenue.

What these assumptions do change is the size of the reserve the town builds, and the width of the tail. The property tax base determines how comfortable the town is, not whether it survives. The line that determines the second question is sales tax.

What Happens If the Economy Turns

The pro forma tests the budget against 100,000 simulated economic futures over 2027–2040, varying revenue and expense growth together as they move together in a real downturn. Its published results:

The honest reading of the ladder alongside this is that single-line results are optimistic. Real downturns do not move one line at a time — sales tax, use tax and permit fees fall together, and a salary error is rarely confined to one position. That is exactly what the simulation is for, and it is why the two belong on the same page rather than being offered as alternatives.

Two errors found in this simulation, and corrected

15 October 2026: the contingency draw ran the wrong way. The simulation is meant to draw more of the budgeted contingency as the economy weakens, from 45% of it in a normal year toward all of it in a deep downturn. Through version 1.12 of the pro forma it keyed that rise to the wrong side of the economic factor, so it drew the contingency down in strong years and held it at 45% in weak ones. A review of the published code found it. Corrected in version 1.13, the probability of reserves going negative at some point is 5.10%, up from 2.15%, and every simulated figure on this page has been regenerated with the correction. The break-even ladder is unaffected: it is a deterministic calculation with no economic shocks in it.

29 September 2026: the building department fee moved the wrong way. SAFEbuilt’s fee is 85% of permit revenue, by contract, so the two building department lines move together. The simulation applied one rule to every expense line — expenses rise under economic stress — so it cut permit revenue in a downturn and raised SAFEbuilt’s fee in the same year. Disclosed on this page on 19 September and corrected in version 1.9 on 29 September; it had overstated the risk. Both corrections are dated in the pro forma’s version history.

What a Council Does About It

The pro forma’s headline figure gives a council one crude tool: once reserves fall below four months of operating costs, it cuts half of its discretionary spending, whatever the size of the gap. Real councils budget year by year. Under the Local Government Budget Law of Colorado, C.R.S. 29-1-103, every local government adopts an annual budget, and no budget may provide for expenditures in excess of available revenues and beginning fund balances.

So the question is not whether a council responds, but how much reducible spending it would have. On the pro forma’s numbers, roughly $1.49 million in 2030, and $20.6 million across 2029–2040. That capacity is mostly the road reinvestment set-aside — an accrual toward a rebuild cycle decades away rather than work performed in the year — together with downtown events and marketing, trees and landscaping, signs and striping, materials, the margin by which public safety is funded above current service, and routine road maintenance capped at 30%.

Held out of that pool entirely: snow removal and street sweeping, bond debt service, insurance, the statutory annual audit, and election costs. Public safety enters the pool only as the margin above today’s service and is capped at half of it, so what a council could reduce never reaches the service Niwot has now. A council could in practice reduce other contracted service levels too. This does not assume it, which makes the capacity above an understatement rather than a ceiling.

How that capacity would be allocated is a decision for a future council, not an assumption of this model. The figures below measure how much of the pool a downturn consumes, not which line it comes from. They are run on the Combined Stress budget described above — every competitive grant removed and the Town Manager at $130,000 — so they test the response under pressure, not the budget as modelled.

In 38.2% of simulations no reduction is required at any point. In the median case a council reduces spending in two years out of twelve and uses 3.6% of the capacity available to it.

No revenue response is assumed anywhere in this. A council under sustained pressure would also have municipal fee capacity beyond what is modelled, and — subject to voter approval under Article X, Section 20 of the state constitution — could seek a higher mill levy or sales tax rate. On 2030 figures one additional mill would raise about $224,000 and half a point of sales tax about $464,000, each larger than a typical adverse-year shortfall. None of it is counted above.

The response itself matters more than any budget line

This is the most important limitation on any stochastic result in the pro forma, including our own, and it deserves to be stated rather than buried. All three rows below are run on the Combined Stress budget. The first sizes each year’s response to that year’s gap. The other two apply a fixed cut whatever the gap; the last is the rule behind the pro forma’s headline figure. They show how much the answer depends on the response chosen, not three forecasts.

A fixed reduction over-reduces in mild years and under-reduces in severe ones. The lower rows are published because a reader is entitled to know how much of the headline figure is the economy and how much is the response rule we chose to model.

Sales tax would have to fall roughly 19% permanently to exhaust reserves. Response policy moves the number further than that does — which is the finding, and the reason this section exists.

Would a Council Cut Downtown and Events?

The pool above names downtown, events and marketing among the lines a council could reduce, and it is the line residents ask about by name. The aggregate figures cannot answer that question: they report how much of the pool a downturn consumes, not which line gives it up. Because the pool is drawn in a deliberate order, those are not the same answer. So here is the line on its own.

It is sixth of seven, and the order is not incidental. A council closes the gap it has, starting with what hurts least: the road reinvestment set-aside first, then trees and landscaping, then signs and striping, then materials, then the sheriff margin above current service, then downtown, and only last a capped slice of routine road maintenance.

Three clarifications, because this list gets read on its own. The road reinvestment set-aside is not the road repair programme: it is an annual accrual toward a rebuild cycle decades away, not work performed in the year. The deferred repairs the bond pays for — the actual fixing of Niwot’s roads — are held constant in every simulation and are not in this pool at all, along with the bond repayments themselves. Nothing here reduces them. And routine road maintenance is last, and capped at 30%, for a reason: deferring pavement work is what the roads measures exist to stop, so the model treats it as a last resort rather than an easy saving. And the sheriff margin is exactly that — a margin, not the service. E3 funds 2.0 FTE against roughly 0.67 FTE of patrol allocated to Niwot today — three times current service. The pool may draw on at most half of it, so the floor is 1.0 FTE — still about half again more patrol than Niwot has now. What a council can reduce here is the margin above today’s service, never the service itself. T14 of the pro forma designates discretionary programmes for reduction before core services; that is why this line sits where it does rather than higher, and why what the pool may take from it is capped at the cushion rather than the line.

The first line alone is $750K in 2030 against downtown’s $265K. A shortfall has to exhaust it and the four lines behind it — $1,066K in 2030 — before downtown is reached at all.

Among the 7.6% that do reach it, the median simulation reduces the line in one year out of twelve, by about 43% of it in that year, for a cumulative $152K against $3,906K budgeted across 2028–2040 — 3.9% of what the line is scheduled to spend. At the 90th percentile of that group: two years, and $470K.

How much comes out of the events budget

Those are conditional figures — they describe the runs that reach the line, about one in 13. The question most people are actually asking is simpler: across everything that could happen, how much comes out of the downtown and events budget? Every simulation, including the 92.4% that take nothing:

The average reduction is $16K spread across thirteen years — about $1,200 a year against a line of roughly $300,000.

Why the two sets of figures differ so much. The conditional median of $152K and the unconditional average of $16K describe the same model. The first answers “how deep does it go once it is reached,” the second “how much comes out in total.” Because 92.4% of the distribution is zero, quoting the first as though it were the second overstates the exposure about 9 times. Both are published here so that neither can be mistaken for the other.

What this does not decide. The model reduces a budget line; it does not allocate within one. How a council would split any reduction between community events, marketing, flowers and beautification is a decision for that council and, as under the LID today, for a resident-led body. E10a’s illustrative split puts community events at $78,000 of the $250,000 — but it is illustrative, and nothing here should be read as a figure for events alone.

The unfavourable end of this, stated rather than omitted: in 1.2% of all simulations the line pauses for a full year. In the rest it does not.

Measured against what downtown has today

The question behind the question is not whether a budget line moves. It is whether downtown ends up with less than it has now. Those have different answers, because the town line is set above what downtown currently spends.

The LID’s adopted 2026 programme is $199,742. Its receipts have exceeded that in every year on record — $249,635 in 2025, $252,477 in 2024, and as much as $295,031 in 2022, a surplus of between $49,893 and $95,289 a year — and the difference accumulates as the roughly $515,000 of reserves the district holds. This line is therefore set against what downtown spends, not what the district collects, and then adds approximately $50,000 for scope beyond the LID boundary.

So $250,000 in 2028 is about 25% more than today’s programme, and 15.2% of the line is headroom above today’s standard in real terms — more in nominal dollars, rising to 44% by 2040, which is why the real comparison is the one used here. A reduction inside that headroom changes the budget and not the programme.

The headroom is worth about 1.5 points, and it should not be sold harder than that. It absorbs shallow reductions, but when the downtown line is reached at all it is usually cut past the cushion. Measured properly, the line stays at or above today’s programme in real terms in 93.9% of simulations, against the 92.4% that never touch it at all.

The comparison runs the other way as well. The status quo has no stress test, and it has already had the downturn. LID revenue peaked at $295,031 in 2022 and was $249,635 in 2025 — a fall of about 23% in real terms that has not recovered, on the index published in the Business Vitality analysis. This is the same decline described under sales tax above, where it is measured on the district’s taxable base rather than its receipts; note there that the fall is already inside this budget, because the pro forma projects forward from the 2025 trough and not from the 2022 peak. Downtown has already absorbed a larger sustained reduction than anything this simulation does to the town line, and it absorbed it with no council required to respond, no budget law compelling one, and no reserve policy. A 1% sales tax on one commercial district is a narrow and volatile base. A municipal sales tax spreads it across the whole community, and puts a body in the room that is obliged to do something when it falls.

What the line is, and what it is not

It is not the town taking over the events. Rock & Rails, the holiday parade, the Jazz Festival and the rest are organised by the Niwot Community Association, the Niwot Business Association, the Niwot Cultural Arts Association and volunteers. The Local Improvement District reimburses programmes brought to it today; it does not run them, and neither would a town. This line funds that work rather than replacing it.

It is not a reduction from what downtown has now. The LID is funded by a 1% sales tax on businesses inside its boundary and its adopted 2026 budget is $199,742, of which roughly 13% pays an administrative coordinator it must hire because it has no staff. On incorporation the municipal sales tax replaces the LID tax and this line takes its place at $250,000 in 2028, serving the whole town rather than the Second Avenue commercial blocks, with the coordination absorbed into town payroll. More money, a wider boundary, and more of each dollar reaching programme rather than administration. The derivation is at E10a of the pro forma.

The honest summary is that a discretionary line is genuinely discretionary — it can be reduced, and in a severe enough downturn it would be. It appears on the list because we would rather publish the list than claim there isn’t one. What the model says is that in 92.4% of futures it is never reached, and that the first $1,066K of pressure lands somewhere else.

What We Still Do Not Know

Publishing this list is the point of the exercise. A document that tested its own numbers and reported no remaining doubt would not be worth reading.

• The specification layer is not modelled. The ladder asks how far one line could move before the answer changes. It does not attach a probability to each line actually being wrong by that much, because doing so honestly requires a defensible range for every line and we do not have one for every line. Our own prototype of that layer suggests it would matter more than the macroeconomic uncertainty does — but its inputs are placeholders, so no figure from it is published here.
• The break-even method tests one line at a time, permanently. Real errors are correlated and often temporary. The simulation addresses the first; nothing here addresses the second.
• The projection stops at 2040; the bond is repaid through 2048. Reserve figures on this page cover 2028–2040 only. Statements about total repayment cover the full schedule. Mixing the two would badly misstate how much headroom the repayment cap leaves.
• Grant revenue is not guaranteed. Competitive grant awards are modelled on eligibility and peer award history, not on commitments.
• The remote sales base is an estimate for Niwot, not a measurement of it. It is the largest single revenue input in the budget. The two towns it is benchmarked against publish their own figures and the model sits between them, but neither town is Niwot, and they do not collect their tax the same way as each other.
• The maximum annual repayment figure of $1.42 million is filed, not certified, and is subject to bond counsel review. The certified $28 million total binds at essentially the same point, so the conclusion above does not rest on it.
• The simulation has had two errors, both now corrected. The building department coupling (corrected 29 September, version 1.9) and the contingency draw (corrected 15 October, version 1.13) are described above. Both were found by reading the model rather than running it; no number of simulations would have surfaced them, because the simulation was faithfully doing the wrong thing. We do not claim to have found every such error, which is why the code is published.
• Who can check this. Every test on this page is published with the code that produces it, and one command reruns them. Corrections that readers have found are dated on How Our Numbers Have Changed. If you think a number here is wrong, the model is the place to show it.

How This Was Checked

The break-even engine validates against the pro forma’s own published table before it will print anything: if it ever stops reproducing those five thresholds, it exits with an error rather than producing a number. The underlying budget figures are parsed directly out of the published pro forma rather than retyped, so this analysis cannot quietly drift from the document it describes. Every model figure and table on this page is generated from the model, and a check runs before the website is built that fails the build if any of them stops matching what the model computes.

Every figure on this page traces to the pro forma, which owns them, or to the certified ballot title and the fiscal impact filing, which are public record. The model itself is published. Download it (a zip of Python source, the budget tables and the data they read) and run python3 verify.py to reproduce the pro forma’s published figures, or change an assumption and see what moves. If you would rather we ran your inputs, send them and we will publish the result.

If you think a number here is wrong, tell us and we will either correct it or publish why we disagree. That is the standard this document is asking to be held to, and it is not much use unless someone takes it up. Write to contact@niwot.town.