How Middlesex pays for its machines.

Middlesex owns a fleet of trucks, fire engines, buildings, and paved roads that wear out on a schedule the town has already written down — at today's market prices, about $6.2 million of local spending over the next fifteen years. Every dollar figure on this page is in today's dollars: no general inflation is assumed. Future bills will be higher by whatever inflation brings — for every plan alike, with home values and the grand list rising alongside, so the comparisons stand. That figure counts only what the town can already see: the machines it owns today, on the cycles they have run. Fifteen years will add needs no one has priced yet, and every one of them arrives on top of the numbers here. How to pay is a choice, and the town's recent choices have a pattern. It starts with one machine.

One machine, eleven years of payments

In March 2024 the town voted to buy an excavator. The article authorized up to $200,000, financed over as long as fifteen years; that July the Select Board signed a $182,000 note at 5.97%, repaid in equal annual slices of principal plus interest on what remains. By the final payment in 2035 the note will have cost about $69,000 in interest. By the end of that first fiscal year, the reserve fund kept for exactly this kind of machine held half the purchase price — earning 0.7% while the note beside it charged 5.97%.

The excavator is the pattern

Every large machine the town runs was bought the same way: vote, sign, pay for years. The grader (2021), the Kenworth dump truck (2022), the Lo Pro truck (2019), the fire tanker (2019), and the fire station itself (a 2015 bond) each left a note behind. Alongside the notes, the town keeps six reserve funds, each bound by Vermont law (24 V.S.A. § 2804) to the purpose voters gave it — the Asset & Equipment Fund cannot fix the Town Hall roof. The funds are real money, $341,657 at mid-2025. But the pattern rarely lets them buy anything: in fiscal 2025 the town spent $258,327 on capital purchases and debt payments against a budgeted capital line of $126,000, with the difference covered by financing and one-off draws.

What the payments add up to

Each signature leaves a payment schedule behind, and the schedules stack. In fiscal 2026 the machine notes cost about $182,000 in principal and interest — five notes and a bond at once. The stack thins as notes mature: about $97,000 by FY2028, about $47,000 by FY2031, with the excavator running to FY2035 and the grader to FY2036. The flood-recovery debt — about $466,000 a year through FY2031 — sits on top of all of this and is identical under every plan on this page. It is a bridge, not a lasting burden: the town expects FEMA reimbursement to clear it by about FY2029, and the charted line is its contractual worst case.

What the town owns, minus what it owes

One number states where the pattern leaves the town as the planning window opens. Entering fiscal 2028, the six reserve funds hold $341,657 while the three debts still running — the excavator and grader notes and the fire-station bond — carry $446,900 of remaining balance. The town's capital accounts are $105,243 short of its capital debts.

Why the next fifteen years are harder

Three facts make the coming window different from the decade behind it. The bills are front-loaded: the overdue trucks, both fire-engine orders, and the Shady Rill road project all land by FY2031. Fire apparatus is bought years before it arrives: contracts sign roughly three years ahead of delivery, the price locks at signing, and prices have been rising about 7% a year. And the schedule is a floor, not a forecast: it lists only what is known today. The needs the town cannot yet see will land on top of it — under the borrowing plans, as interest on an already larger debt.

The order dates, not the delivery dates, are the deadlines.

Six ways to pay for the same schedule

Every plan below buys the identical equipment over FY2028–FY2042, on the town's real fund structure, at market prices, under Vermont borrowing law. Three rules shape all of them:

• Bonds sized to be spent. Federal tax law effectively forbids parking tax-exempt bond proceeds: an issuer must expect to spend 85% within three years to invest them without yield restriction, and 85% within five to keep the interest tax-exempt at all. Each plan therefore bundles its borrowing into a bond sized to the FY2028–31 crunch; plans that bond the mid-2030s Rescue order ask a second, smaller question when it lands (the seed plan's Rescue remainder fits a five-year note instead). Bond Bank access is not a constraint — pooled sales run twice a year, carry no fees, and the Bank lent to Middlesex four times in fourteen months at the peak of its flood borrowing. What each question spends is voter goodwill: the town's last bond question, $2.5 million for the town hall in November 2024, failed 610–534. A single consolidated bond that avoided the second question by parking twelve years of proceeds at 0.7% while paying 4% is priced in the stress tests.
• Heavy trucks stay out of the bond. Manufacturers' municipal installment programs finance them at about 5.5% on five-year terms, by ordinary warned article.
• Reserves earn 0.7%. That is the town's realized FY2025 yield, computed from the audit — in a year when Vermont money markets paid 3–5%. Moving the money to a market-rate account is tested separately in the stress tests.

At market prices, current practice is not the cheap option: it averages about $505,000 a year over the window and peaks near $635,000, as financed market-priced engines stack twelve-year notes on seven-year trucks. The town will pay capital bills of this size under every plan; the choice is what fifteen years of them leave behind. In absolute terms the plans ask between $135 and $154 a year per $100,000 of home market value — on a $300,000 home, they differ by at most a few hundred dollars a year. Every figure prices known needs only; whatever the window adds lands on top. The shaded years beyond FY2042 in the charts are a projection — the same machines recurring on their lifecycles — showing where each structure settles: the pre-funded ask falls toward the replacement bill with no interest in it, while the borrowing structures keep financing each cycle.

The rule underneath every plan

Each dollar not raised for capital is a dollar borrowed at 4% or more while the reserves earn 0.7%. The control below holds the plan structure fixed and varies only the size of a steady annual capital budget.

Checking the plans

Levy totals alone reward plans that end the window broke. The true cost of a plan is its fifteen years of tax levies, minus the change in the town's net position — all funds minus capital debt — from the −$105,243 carried into FY2028, plus the interest its remaining debt is scheduled to cost after FY2042. That last term matters: the borrowing plans sign notes to the very edge of the window, and current practice would still owe $84,000 of interest beyond it. (Post-window earnings on ending fund balances are not credited — debt interest is contractual; future earnings are not.) Both price bases are shown in the tables: market-adjusted is the lead; the committee's as-published prices are beside it for fidelity to the sheet. The base-case spread from current practice to full pre-funding is about $500,000, and full pre-funding is cheapest in every scenario tested. The structural finding in the middle of the field: solving for the smallest levy is not solving for cost. The break-even budget ($500,000) prices about $90,000 worse than the simpler seed plan, and every step up the steady-budget control above buys more equity per dollar. The step-up plan states the same point the other way: for about $26,000 more than the seed plan's true cost, holding $530,000 from FY2030 ends 2042 with about $390,000 more equity. Consolidating all borrowing into a single twelve-year bond — tested as a stress case — would add roughly $150,000–200,000 to the middle plans by parking proceeds at 0.7% while the bond costs 4%. Two results hold under every bond structure tested: full pre-funding is the cheapest of the six plans, and the front-loaded schemes below undercut even it.

Alternatives tested

Six alternative funding schemes were tested against the same rules — same schedule, same funds, same bond structure — in the same model.

The pattern is uniform: every scheme that beats the plans above shapes the tax ask to the front-loaded need — high through the engine-order years, stepping down after — and shrinks or eliminates the bond. The two cheapest run $420,000–430,000 below full pre-funding, and the note ladder never files a bond question at all, leaving the town's bond capacity in reserve against the next flood. Their price is concentration: roughly $900,000 a year asked of taxpayers through FY2031, on top of the flood debt line — about $1.4 million a year all in — and, for the note ladder, a warned article repurposing the idle Bridge and Court Resurfacing balances. The steady-$650k result restates the structural finding: solved for cost rather than for the smallest levy, the same structure undercuts every plan above. The state-loan ladder underperforms because the $150,000 annual cap is small against engine prices and awards are competitive, not promised.

The votes, not just the money

Each plan needs one or two Australian-ballot bond questions: the FY2028 crunch bond, plus — for plans that bond the Rescue purchase — a second, smaller question (about $265,000–450,000) in the mid-2030s. The seed plan and full pre-funding stop at one. Ordinary warned articles cover vendor truck installments and five-year bank notes (six to thirteen articles over fifteen years, listed in the tables above and the downloadable data). Because apparatus commits at contract, the first bond question effectively belongs to the March 2027 warning — before this window opens. A single consolidated bond would trade the second question for the roughly $150,000–200,000 of parked-proceeds cost priced above.

Assumptions and caveats

1. Prices. Apparatus at commercial-chassis market pricing: the sheet's 2023-vintage figures escalated 7%/yr to order year (Engine 6 ≈ $655,000 committed FY2028; Engine 1 ≈ $803,000 committed FY2030). The committee's as-published prices are shown throughout for reference; no contract will sign at them. Everything else is priced in today's dollars — no general inflation is assumed, so levies are floors, not forecasts: actual future bills rise with inflation for every plan alike, and the grand list and home values rise with them, leaving the comparisons and rankings intact. The apparatus escalation is a contract-market fact, not an inflation forecast.
2. Known needs only. The schedule prices the replacements the town can already see — the machines it owns, on their historical cycles. It carries no allowance for new needs, and fifteen years will bring them; every such dollar arrives on top of every figure shown, and under the borrowing plans as interest on a larger principal. The charts' shaded FY2043–52 region is a projection of exactly those known cycles, labeled as such; no metric is computed from it.
3. Earnings. The 0.7% base is the town's own realized yield (FY2025 audit); the 3.5% scenario requires actually moving the money and is labeled as such.
4. Rates. Bank notes 5.0%; the consolidated bond at a 4.0% Bond Bank assumption — the actual rate is set at the Bond Bank's sale (the town's May 2025 loan priced at 3.50%); a 5.0% stress column is shown. Vendor truck installments at 5.5% on five-year terms — the tax-exempt municipal band, with the town's 5.97% excavator note as the local datapoint.
5. Bond structure. Bonds are bundled so each is drawn within three to four years of issue — standard practice, and what federal tax law expects: 85% of proceeds spent within three years for unrestricted investment (Treas. Reg. §1.148-2(e)(2)), 85% within five to keep the interest tax-exempt at all (IRC §149(g)). A single bond sized for twelve years of projects cannot meet either test, so it appears only as a stress case — its parked-proceeds cost charged honestly — not as an issuable instrument. Bond Bank access is open (two pooled sales a year, no fees, no frequency limit; the Bank lent to Middlesex repeatedly through the flood years); what a second question spends is voter goodwill — the November 2024 town-hall bond, $2.5 million, failed 610–534. A modest spending tail beyond three years is routine for bond counsel; the consolidated twelve-year bond is not.
6. Grants. Class 2 paving awards assumed won at the statutory cap (min of $175,000 or 80%, 19 V.S.A. § 309b) — competitive annual awards, matching the sheet's own Shady Rill assumption; the no-grants column shows the exposure.
7. Shady Rill. The section 1 scope predates the 2023–24 flood reconstruction and must be re-scoped before any warning article; it remains the largest single line.
8. Missing work. The Bridge and Court Resurfacing funds have zero scheduled work in the window — the sheet, not the town, is missing culvert and court lines. Reform plans stop feeding those funds until the work is priced; current practice keeps taxing into them.
9. No major building project. No Town Hall or Town Garage renovation or replacement is included — the schedule carries only routine upkeep for both buildings ($151,000 and $19,500 over the window). Any major project, pending the Middlesex FIT proposal and Select Board endorsement or similar, would be new money on top of every plan shown.
10. Small items. Purchases at or under $15,000 are cash-flowed inside their fund rather than financed, which can carry a fund briefly below zero — a payable within a −$15,000 tolerance, never new borrowing. The workbook's fund tables flag any dip beyond it.
11. Flood debt. The flood-recovery debt is excluded from every plan and identical under all of them — and it is a bridge, not a lasting burden. Of the $6.4 million on the books at mid-2025, $4.0 million were notes that matured in 2025–26 while roughly $7.2 million of FEMA money arrived; the remaining $2.4 million of Bond Bank paper ($466,000 a year of debt service, stepping to $288,000 in FY2032) is expected to be swept by further reimbursements. More likely than not there is no net flood levy by about FY2029; the realistic residual is a one-time $400,000–700,000 of cost shares and compliance work. The charted flood line is the contractual worst case, standing in full only if the remaining federal approvals fail. Its debt service can be added to the payment chart above so no one mistakes the machine notes for the whole bill.
12. Unconfirmed with the town. The exact warned purpose language of each existing fund, the treasurer's interest-crediting practice, and a bridge/culvert program are open items; the model's working inferences are stated with the published data.
13. Annual votes. Reserve-fund funding votes bind for one year; every multi-year path here is an annually renewed intention, not a commitment.
14. The FY2027 base. Tax changes are measured against the $268,022 taxpayers already spend on non-flood capital and debt in FY2027 — the $126,000 capital line plus that year's four debt payments: fire station bond $46,817, grader note $25,572, excavator note $27,979 and the Kenworth note's final $41,654 — not against the capital line alone, because every plan absorbs those payments as they retire.

Sources and method

Every assumption is tied to a source: 24 V.S.A. §§ 2804, 1786a and 1789 and 19 V.S.A. § 309b for fund and borrowing law; VLCT's model town-meeting articles and guidance; East Montpelier's published Capital Reserve Fund practice; the Middlesex 2025 Town Report, the FY2025 audited financial statements, and the Budget Committee's Capital Asset Inventory FY2026.1; and current apparatus-market and municipal-rate data. The model, its inputs, and the code that produces every figure are published in the site repository; an independent reimplementation of the stated rules reproduces all results within $2. Fiscal years are July–June; FY2028 is the first budget a March 2027 town meeting can set.