How These Numbers Work

Audited actuals

What are audited actuals, and how do they differ from a town budget?

Audited actuals are what a town actually spent and received across all its governmental funds, as reported after the year closed. They are not the budget the town adopted.

Every Connecticut municipality files audited financial statements with the state. The Office of Policy and Management crosswalks them to Connecticut's Uniform Chart of Accounts, so that the same line item means the same thing in all 169 towns, and publishes the result as the Municipal Fiscal Indicators dataset. That common accounting is what makes a town-to-town comparison possible at all.

Three properties of these figures change how they should be read:

They are actuals, not appropriations. An adopted budget is what a town votes to appropriate before the year starts. These figures are what was spent and received once it ended. A town can spend less than it appropriated, or more, and either is ordinary.

They cover all governmental funds, and are gross of grant pass-throughs. Federal and state grant money that a local budget never appropriates — ESSER, Title I, IDEA, food service — is counted here. Audited totals therefore generally run above the adopted general-fund budget for the same year. The gap is definitional. It is not overspending and it is not a saving.

Who compiled them changed partway through. OPM compiled the figures itself through FY 2019; from FY 2020 forward, municipalities self-report their own audited data. OPM discloses this in the dataset description. It is a break in how the series was produced, not a break in what it measures.

Because the two bases answer different questions, this site never sums them, indexes them together, or draws them as one line. Where both appear in one chart, the point of that chart is that they differ.

The adopted budget

What does a town's adopted budget include?

The adopted budget is what a town votes to appropriate in its general fund before the year begins. It is a different measure from audited spending, on a different calendar.

Each year OPM collects a summary of every town's adopted budget: the revenue it expects, the spending it authorizes, and the mill rate it sets. This is the number a town meeting or referendum actually votes on.

It is narrower than audited spending in two ways. It covers the general fund rather than all governmental funds, and it excludes grant money the town does not appropriate. It is also earlier: it is a plan for a year that has not happened yet, while audited actuals describe a year that has.

Comparing a town's adopted budget with its audited actuals for the same year is a reasonable thing to do, and this site does it — but as a comparison of two measures, never as one series. A difference between them is not by itself evidence of anything.

Mill rate

What is a mill rate, and why can't two towns' rates be compared?

A mill is $1 of tax per $1,000 of assessed value. Connecticut assesses property at 70% of market value, so a posted rate is not comparable between towns.

A town's tax bill is its mill rate multiplied by the assessed value of its property, divided by 1,000. One mill is one dollar of tax per $1,000 of assessment.

Assessed value is not market value. Connecticut assesses property at 70% of appraised market value. A home a buyer would pay $400,000 for is assessed at roughly $280,000, and the mill rate applies to that.

A posted rate mixes two things together. The rate is a ratio between what a town needs to collect and the size of its grand list. A town with a large tax base can fund the same services at a lower rate than a town with a small one. Comparing two posted rates compares spending and tax base at once, and separates neither.

Some towns post a separate motor vehicle rate, which state law caps. Where a town does, this site says so beside the figure.

The measure that is comparable between towns is the equalized mill rate.

Equalized mill rate

What is an equalized mill rate?

The equalized rate restates a town's tax levy against the market value of its property. It is the rate that can be compared between towns and across a revaluation; the posted rate cannot.

The state estimates what each town's grand list would be at full market value rather than at its assessed 70%, and publishes that as the equalized net grand list. Dividing a town's total tax levy by that figure gives its equalized mill rate.

This removes two distortions at once. It removes the difference between towns that revalued recently and towns that did not, and it removes the step a revaluation puts into a town's own posted rate.

Because of that, the equalized rate is what this site uses whenever a rate is compared between towns or across years. The posted rate is still shown — it is the number on the tax bill — but it is labeled as not comparable wherever it appears.

The equalized rate is an estimate built on the state's sales-ratio work, so it carries the uncertainty of any estimate. It is the best available basis for comparison, not a measurement.

Revaluation

Why does a town's mill rate drop in a revaluation year?

Every Connecticut town revalues its property on a five-year cycle. When values rise, the grand list grows and the posted mill rate steps down — even if the town collects exactly the same amount of tax.

State law requires each town to revalue all its real property every five years, on its own schedule. A revaluation resets assessments to current market value.

In a rising market the grand list grows, and a town that wants to collect the same amount of tax must set a lower mill rate to do it. The rate falls; the levy does not. A lower posted rate, on its own, does not mean a lower tax bill.

What a revaluation does change is how the bill is distributed. Properties whose value rose faster than the town average pay a larger share afterward, and properties that rose more slowly pay a smaller one, whatever happens to the total.

Two towns compared in a year when one has just revalued and the other has not are being compared on different footings. The equalized mill rate is the fix; a posted rate has no fix.

The grand list

What is a town's grand list?

The grand list is the assessed value of all taxable property in a town. It is dated October 1 and pays for the fiscal year starting the following July, so it always runs a year ahead of the budget it funds.

The grand list is compiled each October 1 and divided into classes: residential, commercial, land, motor vehicle and personal property. The net grand list is the total after exemptions.

Its year is not the fiscal year. The grand list dated October 1 of one year sets the taxes for the fiscal year beginning the following July. A grand list year and a fiscal year named on the same page are not the same period, and this site labels which is which wherever both appear.

The class mix matters for who pays. A town whose grand list is mostly residential collects most of its property tax from homeowners; a town with a large commercial or industrial base does not. That is a fact about the tax base, not a judgment about either town.

Values here are assessed values, at 70% of market value. See mill rate.

Peer groups

Which towns is a town compared with, and who chose them?

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 site shows both.

Three kinds of group appear on this site, each with a different claim behind it.

Surrounding towns are the municipalities sharing a border with the subject town. They share a labor market, a weather pattern and often a regional school district. They are not matched on size or wealth, and often differ by an order of magnitude in population.

District Reference Groups are the CT State Department of Education's grouping of school districts by family income, parent education and occupation, family structure, poverty, home language and enrollment. They were built for education questions and have nothing to do with municipal services. They were last reassigned in 2006, on roughly 2000-census inputs; a town whose population or income has shifted since may fit its group less well now than it did then.

Rule-based peer sets are built here, by a published formula rather than by hand: standardize the chosen variables across all 169 towns, log-transforming the skewed size and wealth measures, then take the nearest towns by Euclidean distance. Two sets are built, because the two questions have different drivers — a municipal set on population, grand list per resident, unemployment and school-age share, and a school set on enrollment, grand list per resident and school-age share. Variables are equally weighted, which is a stated choice rather than a fact. Change the formula and you get a different list; that is the point of publishing it.

No group is the right one. Each answers a question, and naming the question is the only honest way to use one.

Medians, and why the town is left out of its own

How is a group of towns summarized?

A group is summarized by its median — the middle town, not the average. The subject town is excluded from its own group median, because a town cannot be part of the benchmark it is measured against.

Line the towns in a group up from lowest to highest and take the one in the middle; where the group has an even number of towns, take the midpoint of the two middle ones. That is the median. It is less sensitive to one unusual town than an average, and it is not necessarily the same town from one year to the next.

Each town's own value is computed first, and the median taken of those. The alternative — dividing the group's median numerator by its median denominator — produces a number that describes no town in the group, and can sit several percentage points away from the median town's actual figure.

The subject town is excluded from its own group median. A town that is part of its own benchmark moves that benchmark toward itself, which understates every difference. Group rosters in the source data include the subject town; this site filters it out and says so.

A median is a middle, not a target. Being above or below one is not by itself a finding.

Per-pupil spending

Why do per-pupil figures differ between sources?

There is more than one per-pupil number, and they are not interchangeable. What this site derives from OPM spending and average daily membership is not the state Department of Education's Net Current Expenditures Per Pupil.

The numerator differs. OPM education expenditures cover what a town spent on education across all governmental funds, grant money included. The state Department of Education's Net Current Expenditures Per Pupil deliberately excludes several categories — capital, debt service, and certain reimbursed costs among them. A figure derived from the first will not match one published as the second, and neither is wrong.

The denominator differs too. Average daily membership averages enrollment across the year; the October 1 headcount is a single day. ADM runs below the headcount for most districts.

A regional district splits the count. A town in a regional district reports only the grades it runs itself; the regional district reports the rest and bills its member towns. A town's own per-pupil figure therefore describes only part of what it pays for educating its children, and towns in regional districts are not directly comparable to towns that run K–12 alone.

Wherever this site prints a derived per-pupil figure, it says which numerator and which denominator produced it, next to the number.

Census survey estimates

What is an ACS 5-year estimate?

American Community Survey figures are rolling five-year averages with a margin of error, not counts of a single year. Consecutive releases overlap and should not be read as year-over-year change.

For places the size of most Connecticut towns, the Census Bureau publishes only the 5-year form of the American Community Survey. Each release pools five years of responses to get a sample large enough to report.

Consecutive releases share four of their five years. The difference between the 2023 and 2024 releases is not one year of change; it is one year swapped at each end of a five-year window. Treating the difference as annual change overstates how fast anything moved.

Every estimate has a margin of error. For a small town, the margin on median household income can be wide enough that two towns whose point estimates differ are not distinguishable at all.

Median home value is what owners tell the survey their home is worth. It is neither an assessment nor a sale price, and it will not match either.

An ACS estimate set against a single audited fiscal year is a comparison of two different kinds of thing. This site labels the release year wherever one appears.

State aid

How much of a town's revenue comes from the state?

Intergovernmental revenue is money from the state and federal governments, education aid included. A town's share of it reflects state formulas as much as anything the town decides.

Education Cost Sharing is the largest single state grant to most towns, distributed by a statutory formula weighted toward towns with lower property wealth and higher student need. Other statutory grants cover roads, PILOT payments for tax-exempt property, and specific programs.

Two towns with similar budgets can have very different aid shares because the formulas treat them differently, not because either chose anything. Aid share is a fact about the state's distribution rules first and about the town second.

OPM publishes both an estimate of statutory aid by town and, separately, the intergovernmental revenue each town actually reports receiving in its audited statements. The two are on different bases and will not match.

Reserves, debt and long-term liabilities

What do fund balance, bonded debt and net pension liability mean?

Fund balance is what a town has not spent; bonded debt is what it has borrowed; pension and OPEB liabilities are what it has promised. The three are different kinds of obligation and do not add up.

Total fund balance is the accumulated difference between what a town took in and what it spent. Unassigned fund balance is the part of it not already committed to a purpose — a subset, not an addition.

Bonded long-term debt is borrowing outstanding, typically for capital projects. Annual debt service is what is paid on it this year. A town that has just finished paying off a school shows low debt and low debt service; so does a town that has not built one.

Net pension liability and net OPEB liability are actuarial estimates of promised future benefits not covered by assets set aside. They are sensitive to the discount rate assumed, which is set by each plan, so figures are less comparable between towns than the dollar amounts suggest.

None of these is a score. A reserve level that is prudent for a town with volatile revenue may be more than another town needs, and the same number reads differently against different obligations.

Reported school staffing

What do EdSight staffing counts actually count?

Staffing figures are positions each district reports to the state under fixed assignment categories. They count reported roles, not workload, and a category boundary can move a person between lines without anyone changing jobs.

Connecticut districts report their staff to the State Department of Education each year, assigned to the state's categories, and EdSight publishes the result as full-time equivalents. A person working half time counts as 0.5.

A category is a reporting choice, not a job description. Districts assign the same real role differently, and the state has revised its categories over the period covered. A line that steps up or down in one year across many districts at once is usually a change in how the question was asked, not in who was hired.

A count is not workload. Two districts with the same number of reported teachers can have very different class sizes, caseloads and course offerings. The count says how many positions were reported, and nothing else.

Central office is not school level. Administrators reported at school level and administrators in a district's central office are separate lines; a district that is thin in one may not be in the other, and comparing only one is comparing part of the picture.

A town in a regional district reports only the staff it employs directly — see per-pupil spending for the same split on the money side.

How much of a district is administration

What does “administrators as a share of staff” actually measure, and what can it not settle?

The state's administrator count bundles coordinators and department chairs with principals and the superintendent's office, a small district needs a superintendent as much as a large one does, and the headcount share and the spending share do not give the same answer.

Two different measures on this site answer the question “how much of this district is administration?”, and they are built from different state collections. The headcount share is the administrator positions a district reports to EdSight divided by every staff position it reports, on an October 1 school-year basis. The spending share is general administration plus school-based administration divided by total expenditures, from the district's fiscal-year filing under the state's uniform chart of accounts. Neither is derived from the other and no chart on this site combines them.

“Administrator” is the state's word, and it is broad. The two categories are “Administrators, Coordinators and Department Chairs — District Central Office” and “… — School Level”. A department chair who teaches most of the day is inside the count, next to the superintendent and the principals. The state publishes no narrower split, so no page here can produce one.

The spending functions are wider than the people. General administration and school-based administration pay for the clerical staff in those offices, and their supplies and purchased services, as well as for administrators. That is one reason a district can sit low on the headcount share and high on the spending share.

Size does part of the work, and it is measurable. One superintendent is a fixed cost whether a district has 400 students or 4,000, so the smallest districts sit higher on every administration share before any decision is taken. How much higher is not a matter of opinion: the scatter on the administration ranking page fits the share against enrollment and prints both the fit and the observed median in each quarter of districts by size, so a reader can see how much of the spread size accounts for and how much it does not.

What none of it measures is pay. Connecticut publishes no administrator or teacher salary figures by district. A district with fewer, better-paid administrators can spend more on administration than one with more of them, and nothing published by the state can currently tell those two apart.

A regional district's administrators serve every member town alike, so these figures belong to a district and are never split between its towns — see peer groups for how a town is compared, and reported school staffing for what a position count is.

Businesses and jobs located in town

Do job counts describe where people work or where they live?

Establishment and job counts come from employers' quarterly filings and describe workplaces located in the town. They say nothing about where residents work, and the two are very different in most Connecticut towns.

The Quarterly Census of Employment and Wages counts establishments and jobs from the unemployment-insurance filings employers make. A job is counted in the town where the worksite is.

This is workplace, not residence. A town where few jobs are located is not a town where few people work — most Connecticut residents commute out. A low job count describes the local employment base, not the employment of residents.

Coverage has gaps. The self-employed, most farm labor, and some religious organizations are outside the unemployment-insurance system and so outside these counts.

Small towns are suppressed. Where too few establishments exist in a category to publish without identifying an employer, the figure is withheld. A missing value means the state did not publish it, not that the count is zero.

Statewide ranks

What does a rank out of 169 towns tell you?

A rank is a position, not a grade. Which end of a list is preferable depends entirely on the measure and on what the reader is trying to find out.

Where this site prints a rank, it prints the position, the number of towns it is out of, and the direction of the sort — lowest to highest — alongside the statewide median. Anything less than all four is not interpretable.

No adjective is attached, deliberately. Low debt per resident can mean a town has finished paying for its schools or that it has not built them. A high equalized rate can mean high spending or a small tax base. The rank is the same number either way, and the page does not tell the reader which story to prefer.

Ranks on this site are computed across all 169 towns for a single fiscal year, named wherever the rank appears. A rank from one year is not a trend.

Groton reports twice in the state's table — the town and the separately reporting city — so the statewide table carries 170 entities for 169 towns. The non-town entity is excluded from counts and rankings here.

The outlier flag. A headline figure on a town page is printed in plain ink, with its rank and the statewide median beneath it. It is flagged as an outlier only when it falls in the lowest or highest tenth of all towns — the 17 lowest or 17 highest of 169 — and the flag states that position and the distance from the median, nothing else. Amber there means unusual, not bad: which end of the list is preferable still depends on the measure and on what the reader came to find out.

Spending per need-weighted pupil

Why divide school spending by something other than the student count?

The state's own school-aid formula counts a student who is poor or still learning English as more than one student. Dividing spending by that weighted count, using the weights the General Assembly wrote into law, compares districts that serve different students on a common footing.

The weights are the state's, not this site's. Connecticut's Education Cost Sharing formula (CGS §10-262f(25)) defines a town's total need students as its resident students plus 30% of those eligible for free or reduced-price meals, plus a further 15% of those eligible students beyond 60% of enrollment, plus 25% of its multilingual learners. Spending divided by that count is spending per need-weighted pupil.

What it does and does not adjust for. It adjusts for poverty and language need in the proportions the statute chooses. It carries no weight for special education, whose cost can be dominated by a few out-of-district placements, so this site also shows the figure with special-education tuition removed. The weights were changed in 2021; the current ones are applied to every year so that years compare, which is a stated choice.

Two bases. By operating district the numerator is the state's EFS per-pupil report (self-reported, unaudited, excludes debt and capital). By town the numerator is CSDE's Net Current Expenditures, which folds a regional district's cost back onto its member towns. The two are shown apart and never combined.

Academic growth (percentage of target achieved)

What is growth, and why show it before test scores?

Growth measures how far a student moved in a year toward a target set from their own prior score; the achievement index measures where students ended up. Growth depends less on family circumstances than the level does — less, not not at all — so this site shows it first.

How the state computes it. Each student in grades 4–8 who took the Smarter Balanced test in two consecutive years gets a growth target from their prior score alone — the same target for every student at that prior score, with no adjustment for demographics. The district figure is the average share of the target its students reached, so 60 means students reached 60% of the way on average; 100 would mean every student hit an ambitious one-year target. It is not a proficiency rate.

Why it leads. On this site's own build, six family-circumstance variables explain roughly half of the difference in growth between Connecticut districts but roughly three quarters of the difference in the achievement index. Growth is the measure less tied to who the students are. It is still tied: a growth figure is not free of demographics, and the pages say so.

Limits. Growth exists for grades 4–8 only, so a high-school district has none; no growth was computed for 2019-20 or 2020-21; and a district with fewer than 500 tested students moves noticeably from year to year, which is why the site uses three-year means and marks small districts.

Results relative to what demographics predict

What does it mean that a district is above or below the line?

A regression across all districts predicts a district's achievement index from six family-circumstance variables. The difference between what a district earned and what was predicted is its residual: a position relative to other Connecticut districts, printed with an interval, never a grade.

What goes in. The share of students eligible for free or reduced-price meals, of English learners and of students with disabilities (from the October 1 enrollment count), and the median household income, share of adults with a bachelor's degree and share of children living with one parent (from the Census ACS, with margins of error). These are the variables the state used to build its District Reference Groups in 2006, on current data. Race is not a variable.

What the residual is. Earned minus predicted, in index points. Because the fit includes an intercept, the residuals average to zero across districts, so a large share sit below the line; being below it is a position, not a finding. Each residual comes with a 90% interval; inside it, a district cannot be told apart from the line. A district with fewer than 500 tested students is marked because its results can move more from year to year; the mark does not change its interval.

What it cannot say. It is correlational. It does not say why a district sits where it does, and it does not say whether spending caused anything. The same regression is run on growth, where the variables explain much less, and those residuals are published labeled as a disclosure.