Free cash: is Lunenburg hoarding, or rebuilding?

Two claims are being made about the same number.

• That the town is too conservative — sitting on money it could comfortably spend, and making the budget tighter than it needs to be.
• That free cash is "not up to standard", so the town is in a rebuilding state and the balance is not available to spend down.

Both are testable. Neither is tested here, and the reason is worth stating before any figure: the data required to settle it is a ratio, and what we hold is a numerator.

What we now hold

The Division of Local Services publishes a Free Cash Proof for every community — the year-end calculation, broken into its components. sources/dls/ holds it for Lunenburg and eight comparable towns for 2021 to 2025, and sources/data/free-cash-proof.csv is the extract.

It reconciles to itself twice over: the components sum to the total the sheet prints, and each year's calculation equals the next year's opening figure. 81 checks across nine towns and five years, all tie to the dollar. scripts/extract_free_cash.py refuses to write otherwise.

This closes notes/DATA-WANTED.md §5, which had been blocked because DLS sits behind bot protection.

Certified free cash, 2021 to 2025

Do not read down this table. These are absolute dollars for towns of very different size, and nothing in the source carries a population, a budget or a levy. Littleton having four times Lunenburg's balance is a fact about Littleton's size before it is a fact about Littleton's policy.

What the table does support is reading across. Lunenburg's balance fell 36% between 2022 and 2023, and has risen every year since to its highest point in the five. Shirley (−61% over the period) and Townsend (−48%) went the other way.

The standard, and who says so

The Town states it, in its own FY27 budget press release (town-budget/text/4090-click-here-for-a-release-on-quot-understanding-lunenburg-apos-s-fy27-budget-how-.txt, page 6):

"DLS also recommends that communities understand the role free cash plays in sustaining a strong credit rating and strive to generate free cash in an amount equal to 5-7% of its annual budget. This year, Lunenburg certified a record $3.354 million in free cash - 6.65% of the operating budget - well within DLS recommendations. In the last 10 years, Lunenburg has been below DLS free cash recommendations for seven years, only meeting this recommendation in 2022, 2023, and 2026."

That single paragraph contains both sides of the disagreement, and both are true.

• This year is strong. A record balance, 6.65%, inside the range.
• The record is weak. Below the recommendation in seven of the last ten years.

They are not in conflict. They describe different windows. Somebody saying the town is sitting on money is describing 2026; somebody saying the town is rebuilding is describing 2017–2025. Neither has to be wrong for the other to be right.

The standard itself is single-sourced, and it matters

The 5-7% figure appears in exactly one document in this archive — that press release, and its copy on the town site. It is the Town quoting DLS. We hold no DLS publication saying it, and this project does not restate a threshold it has not read.

That is not pedantry, because the threshold is load-bearing in a way the balance is not:

• At 5-7%, 6.65% is comfortably inside, and the ten-year record is as described.
• At any lower band, 6.65% is above the range rather than within it, and the count of years the town fell short would fall too — because that count is entirely a function of where the line is drawn.

So the conclusion "this year is healthy" survives either way. The conclusion "we have been short for seven of ten years" does not survive independently of the threshold, and the threshold is currently one sentence in a document written by one party to the argument. Reading DLS's own guidance is the single highest-value thing left to do here.

Three denominators, three answers

Free cash is certified as of 1 July, so the balance identified 1 July 2025 is what was available at the start of FY26 — which makes FY26's budget the denominator.

All three are "the operating budget" and none of them is the same number. The spread does not change any conclusion — every version is inside 5-7% — but it is worth knowing that the denominator is soft before quoting a ratio to two decimal places. We cannot reproduce the Town's $50.44M from anything we hold.

We still cannot compute this for any peer, because we hold no budget figure for them. So the ratio places Lunenburg against a standard; it does not place Lunenburg against its neighbours.

A year-labelling trap

DLS labels the proof by the calendar year of the 1 July certification. The Town labels the same money by the fiscal year it can be spent in. They are off by one.

This is checkable rather than assumed. Lunenburg's three largest certified balances in the DLS file are 2021, 2022 and 2025. Add one to each and you get 2022, 2023 and 2026 — exactly the three years the Town says it met the recommendation. The mapping is confirmed by the data rather than by reading the labels.

So the table above and the Town's statements are one year apart. The record year is "2025" here and "2026" in the press release. Anyone setting the two side by side without noticing will conclude they disagree when they agree.

What the composition shows, and this part does compare

A share has no size, so the make-up of each town's free cash can be compared even when the totals cannot. The largest component in most years is money that was appropriated and never spent.

Unspent and unencumbered appropriations, as a share of identified free cash:

Lunenburg's 2025 proof in full:

 2,457,761  Unencumbered/unexpended appropriations (CL#11)
 1,225,720  Excess local receipts (CL#6)
   270,669  Prior year free cash not appropriated (CL#12)
   149,455  Excess cherry sheet receipts (CL#8)
    14,256  Other adjustments
     7,139  Revenue received but not estimated (CL#7)
  -170,954  Net change in adjustment to free cash
  -237,764  Outstanding receivables
 ─────────
 3,716,282  Identified free cash, 1 July 2025
 3,354,370  Certified

Two thirds of Lunenburg's free cash in 2025 is money the town appropriated and did not spend — $2,457,761, against $1,225,720 from local receipts coming in above estimate. That share has risen from 31% in 2023 to 60% and then 66%.

Where the money already goes: the capital programme

Free cash is not spare. It is the capital programme's largest single funding source, and the FY27 capital plan publishes its own ten-year table saying so.

Source: sources/town-budget/text/4111-article-13-fy-2027-capital-plan.txt, funding history table, and ATM 2026 warrant Article 13. Extracted to sources/data/capital-funding-history.csv, which reproduces the plan's own printed Average row to the cent — model/freecash.py refuses to export if it does not.

The redirect ceiling is larger than the whole of what free cash gave capital in seven of those ten years, and 1.21× what it gave last year. FY27 is the outlier on both sides at once: the record certification and the record contribution to capital are the same money. This is the capital-side twin of the normal-year finding — the draw is affordable in this year, and in few others.

A third of the programme is not money the schools could have had

The FY27 programme has three funding sources and they are not interchangeable:

The stabilization fund was adopted at the 2017 Annual Town Meeting for "Funding Future Capital Needs for Vehicles and Equipment", requires a two-thirds vote, and holds $2,641,397.10. Money not spent from it stays in it. Cancelling a project it pays for frees nothing for the schools, and a model that lets a reader cut those projects for money back is inventing dollars.

The plan does not publish a project-by-project funding table, but it footnotes exactly two projects "\Funded from Vehicle Use Special Purpose Stabilization Fund" — Engine 2 at $335,000 (p.15) and the Front End Loader at $259,000 (p.23). Those two sum to $594,000, which is the figure the plan's own funding page shows against that fund. That is how the assignment is known rather than guessed*, and model/freecash.py reconciles it in both directions before it will compute anything: the marked projects must total the stabilization line, and the rest of the funded programme must total free cash plus taxation.

So the programme a free cash redirect can actually strand is $1,236,203, not $1,830,203 — and rank 11, the $259,000 front end loader, is not in it. An earlier version of this model took projects off the bottom of the whole funded list, which stranded the loader with free cash that never paid for it.

What it costs capital is two questions, and only one of them is arithmetic

Redirect $500,000 and the programme is funded by $500,000 less. That needs no assumption.

Which projects stop does. The plan ranks 22 projects, funds the first 12, and leaves $1,437,005 of ranked, costed work below the line. So there is a queue: nothing gains slack, and a dollar removed is a dollar of requested work not done. But which dollar depends on how the Capital Planning Committee behaves, and the two readings are far apart:

The rigid reading overshoots because the items are indivisible and rank 7 is a $494,500 roof: taking anything at all off the bottom reaches that roof by the fifth item, so a $300,000 draw and a $500,000 draw produce the same $693,949. At $300,000 that overstates the loss by 131%.

That overshoot is an artifact of the assumption, not a cost of the policy. A committee losing $300,000 of funding and holding ten unfunded projects in front of it can defer the roof and fund $393,949 of smaller queued items instead, or take the four items above the roof and stop. Nothing published says whether it would. The site reports both ends and calls neither the answer; notes/DATA-WANTED.md §3e names the one document that would settle it — any prior year's recommended list beside its funded list.

What this does not show. It does not show that capital work is lost forever: a project deferred is not a project cancelled, and this project holds no evidence either way about what happens to items that fall below the line in one year. It also says nothing about the merit of any project against any school dollar. It establishes only that the money is already committed to something ranked, that a third of the programme could not have gone to the schools under any vote, and that "available within the guideline" is not the same sentence as "available".

What this does and does not establish

It does not mean waste, and it is not the schools. Every town turns money back; departments do not spend to the penny, and a balance of zero would mean a town had budgeted with no margin anywhere. The figure is town-wide across all 67 departments, not a school figure, and nothing here apportions it.

Uxbridge is the unusual one, not Lunenburg. At 1% in 2024 and 12% in 2025, Uxbridge's free cash comes almost entirely from revenue rather than underspending. Lunenburg at 60–66% sits with Ayer, Shirley and Townsend.

What it does establish is the shape of Lunenburg's free cash, and the shape bears on the argument. A balance built mainly from underspending is a different thing from one built from revenue outperformance. The first says the operating budget is being set above what departments use; the second says revenue is beating forecast. They imply different remedies, and Lunenburg's is increasingly the first.

That is directly relevant to budget-vs-actual.md, which asks how far appropriations diverge from spending and could only answer it for the school side, from restatements.

And the Town explains part of it, in a sentence that belongs to another argument

The paragraph immediately after the free cash figures in that same press release:

"At the end of FY25, the LPS District experienced a larger-than-normal budget surplus (more than double the historical average of $200,000). The surplus was primarily caused by two factors: significant turnover and unfilled positions in the facilities department resulted in unspent salaries and stalled maintenance projects and several paraprofessional salaries were ultimately covered by newly identified grants. However, the previous administration did not adjust the general operating budget to reflect these savings. These unspent funds become part of the town's certified 'free cash'."

Two things follow, and the second is bigger than free cash.

On free cash: the Town names a mechanism for its own underspending — budget lines that were not adjusted when the money stopped being needed. That is a statement of cause by one party, not a measurement, and nothing here tests it. But it is the Town's own account of why its free cash looks the way it does.

On special education: "several paraprofessional salaries were ultimately covered by newly identified grants" is the Town stating, in a published document, the exact mechanism notes/DATA-WANTED.md §3b calls load-bearing and says the budget cannot distinguish. The in-district special education escalator is built on a paraprofessional line, and that line rising or falling because a grant appeared or ended looks identical, in the budget, to the district employing more or fewer people.

This does not settle it. It is one sentence, it names no amount, no year beyond FY25 and no number of positions, and it concerns the general operating budget rather than the special education line specifically. What it does is move "a grant could be doing this" from a hypothesis this project raised to a thing the Town has said happened. sped-and-funds.md should carry it.

What would settle the disagreement, in order:

1. DLS's own published guidance on free cash levels. The 5-7% band is the hinge of the whole argument and we have it only as the Town's paraphrase. One document.
2. Lunenburg's Financial Policies Manual (April 2024). The Town says it "has its own policies regarding how Free Cash should be spent and prioritized". Those policies are not in the archive.
3. Operating budget or total general fund revenue for the eight peers, so the ratio can be compared rather than just computed. DLS Schedule A would carry it — DATA-WANTED.md §6, also behind bot protection.
4. The department-level detail behind the $2,457,761. The proof gives a town-wide total and no breakdown. Which departments turned money back, and whether it is the same ones each year, is the difference between a structural pattern and a run of one-offs.

How to reproduce

python3 scripts/extract_free_cash.py     # 81 reconciliations, all must tie

Sources in sources/dls/ with PROVENANCE.md beside them. The denominator is sources/data/town-ledger-fy26-q3.csv, extracted by scripts/extract_town_ledger.py, which reconciles to the report's own GRAND TOTAL before it will write.