Bethel, CT — How Towns Are Compared

Any comparison of towns is a choice, and the choice changes the answer. Measured against Connecticut's highest-spending municipalities, most towns' spending per resident reads as low; measured against the lowest-spending municipalities, the same figure reads as high. Neither comparison is wrong, and neither is informative on its own. This page lists every comparison group used for this town, says plainly where it came from, what it is good at, what it is bad at, and how you can check or replace it.

The answer to "who picked these towns?" is not "trust the author." It is: here is the group, here is its origin, here is its known weakness, and here is a rule you can run yourself, whose inputs are stated and whose selection of towns is arithmetic.

One rule governs everything below: a peer group is a tool for a specific question, not a verdict. A group that is right for comparing school spending can be wrong for comparing snow plowing. Where two defensible groups disagree, this page shows both.

The four comparison groups on this site

1. The District Reference Group — the state's grouping of similar school districts

… districts, including this town · CT State Department of Education · created 1996, last reassigned 2006

The CT State Department of Education sorted school districts into groups with similar family characteristics: family income, parents' education and occupation, family structure, poverty, home language, and enrollment. This town is in its District Reference Group along with ….

Good at: education comparisons. Its inputs are the family-background variables that genuinely predict what a district costs and how students do — exactly the variables you would want held roughly constant before comparing school spending.

Bad at: town-side comparisons, and size matching. Its inputs describe student and family background; none of them measures road mileage, public-safety staffing, or the other drivers of municipal cost. The group spans a range of population sizes, and per-resident municipal cost varies with scale. Assignments were last made in 2006, on roughly 2000-census inputs.

2. Surrounding towns — the ones that share a border

…

The towns a resident actually compares themselves to: same labor market, same regional services, same weather, same road-salt contracts, often the same newspaper.

Good at: regional and labor-market context, and the comparison residents intuitively care about — what the town next door does.

Bad at: scale and school structure. Neighbors are chosen by geography alone; the group spans a range of population sizes. Not every Connecticut town runs its own K-12 district: where a town belongs to a regional school district for some or all grades, its reported education spending covers only the grades it runs and is not on the same basis as a K-12 town's. Any chart that compares education spending has to say which towns those are.

3. All … towns — the statewide distribution

Every Connecticut municipality · used in the ranking chart below

Instead of choosing peers, rank this town against everyone and report where it lands.

Good at: being unarguable. No selection choice is involved, so no one can claim the comparison towns were picked to flatter. It is the honest denominator when a headline number needs context.

Bad at: explaining why a town sits where it does. A statewide equalized-rate ranking largely sorts towns by taxable property per resident, which follows from the property market rather than from any budget decision. It answers "where does this town sit?" but not "is that reasonable for a town like this one?"

4. Rule-based peers — picked by formula, not by hand new

… towns per set · derived from CT OPM data by Data/build_peer_sets.py · latest matched year match year

A person set this rule up: which variables go in, that size and wealth are put on a log scale, that the variables carry equal weight, and how many towns to keep. What a person does not do is pick the towns. Each town's characteristics are standardized across all … towns, and the towns closest to this town in that standardized space become the peer set. That selection step is arithmetic: change the rule and the list changes; run it unchanged and you get the same list every time.

There are two sets, because the two questions have different drivers — using one peer group for both is the flaw this is meant to correct.

Good at: being checkable. The formula, its inputs and every town's distance are published, so a disagreement can be about the formula rather than about the author's judgment. Anyone can re-run it and get the same list; the sets shown here were run on the latest matched year data.

Bad at: intuition and completeness. The result can include towns no resident here thinks about. It carries no measure of median household income — the Census ACS API now requires a key this project does not hold — and no measure of service levels, so two towns can match on every input while providing different services. It matches on one year at a time, currently the latest matched year.

How the rule works

For each variable, every town's value is converted to a standard score — how many standard deviations above or below the statewide average it sits. Each town then becomes a point in that space, and the peers are the towns at the shortest straight-line (Euclidean) distance from this town. Two choices are worth stating openly, because they change the result:

Size and wealth variables are log-transformed first. Population, enrollment, and tax base are heavily lopsided — a handful of very large or very wealthy towns would otherwise dominate the arithmetic and flatten every real difference among mid-sized towns. On a log scale the unit of distance becomes "twice as big," which is the comparison that makes sense for town size.
Variables are weighted equally. That is a choice, not a fact. It was not tuned to produce a particular list, and the data file records the inputs so anyone can weight them differently.

Wealth is measured as equalized net grand list per resident — the town's taxable property restated to full market value, divided by population. It is the tax base per person — the quantity that constrains what a mill rate can raise. Median household income would be the more direct measure of community wealth, but the Census ACS API now requires a key this project does not hold, so the tax base per resident stands in for it.

python Data/fetch_ct_ucoa.py && python Data/build_peer_sets.py

Inputs come from CT OPM Municipal Fiscal Indicators (ej6f-y2wf), the latest matched fiscal year. The output records every town's distance, so the cut can be moved or checked:

Who the rule picks

These peer sets are built by this site, from CT OPM data, by Data/build_peer_sets.py. They are not an official State of Connecticut grouping and carry no standing beyond the rule described above.

The … nearest towns for each question, with how each one relates to the hand-built groups. Median household income would be the more direct wealth measure and is not among the inputs. neighbor DRG both neither

Loading…

Source: CT OPM Municipal Fiscal Indicators (ej6f-y2wf), the latest matched fiscal year, run through Data/build_peer_sets.py in this repository. These sets are this site's own construction, not an official state grouping.

Where this town sits among all … towns

Percentile rank on each measure, latest matched year. 0 is the lowest town in Connecticut, 100 the highest, 50 the statewide median. No peer group is involved, so no selection choice affects these.

Loading…

Source: CT OPM Municipal Fiscal Indicators (ej6f-y2wf), the latest matched fiscal year. Higher is not better or worse — it is only higher. A high rank on reserves and a high rank on mill rate mean very different things, and this chart takes no position on either. The per-pupil figure here is OPM education expenditures ÷ enrollment (audited, all funds); it is not the CSDE Net Current Expenditure Per Pupil series, and the two are not interchangeable. OPM education expenditures cover only the grades a town itself operates, so for a town that belongs to a regional school district for some of its grades the per-pupil rank is not on the same basis as a town running its own K-12 district.

How the groups compare on the same question

Total expenditures per resident, latest matched year. The same figure for this town against four different reference points — the reason this page exists.

Loading…

Source: CT OPM Municipal Fiscal Indicators (ej6f-y2wf), total_expenditures ÷ population, the latest matched fiscal year. Every group median excludes this town, so the town is never part of its own benchmark. These are audited all-funds actuals, gross of grant pass-throughs — not an adopted general-fund budget. District Reference Group assignments were last made in 2006. The rule-based bar uses a peer set built by this site from CT OPM data by Data/build_peer_sets.py, not an official state grouping.

Two things worth stating plainly

A label has to match the group. This site got that wrong once: an earlier version described a short list of often-compared towns as a District Reference Group, when the state's own assignment placed several of them in other groups. The correction is recorded in this repository's CHANGELOG.md under 28 June 2026. The rule adopted since: every group is labeled by what it is and by who assembled it — a District Reference Group is the state's assignment, surrounding towns are the ones sharing a border, and a rule-based set is identified as this site's own construction.

One peer set cannot do two jobs. Town-side spending and school spending have different drivers, so the rule-based sets are split accordingly — one matched on the variables that drive municipal cost, one on the variables that drive school cost. Where a comparison of municipal operations has to lean on an education-derived grouping, that limitation is stated rather than assumed away.