The whole mechanism on one page — with a working model you can set to Norwich, East Hampton, or Colchester
A revaluation does not change how much money a town collects — the budget does that. A revaluation changes how the total is divided. Every property’s bill is its slice of one pie: your assessment, divided by everyone’s assessments, times the levy. When assessments reset to market, the mill rate is cut so the pie stays the same size — and the only thing that matters for your bill is whether your slice grew or shrank relative to everyone else’s.
That is why the recent revaluations hit homeowners: home values boomed while commercial property, cars, and business equipment did not keep pace. Homes’ slice of the pie grew — in Norwich from 51% to 61.5% of the base, in East Hampton from 77% to 83.5% — so home bills rose even with no budget growth at all. And the size of the hit depends on how much non-home base a town has to shift taxes away from. The model below lets you move every lever yourself.
The town adopts a budget; the amount to be raised by taxes is the levy. The revaluation itself neither raises nor lowers it.
Each property’s assessment (70% of market value) determines its share of the levy. Between revaluations the shares are frozen; a revaluation resets them all at once.
Mill rate = levy ÷ total assessments × 1,000. When assessments jump ~40%, the rate falls ~30%. A falling mill rate is not a tax cut.
Your bill rises only if your assessment rose more than the average of the whole base (plus any budget growth). Rise less than that and your bill falls — even though your value went up.
A three-class model of a town’s tax base. Pick a preset — the three measured revaluations and the Colchester scenarios — or set the levers yourself and watch what happens to the average home’s bill. The chart and numbers update live.
Houses, condos, apartments. Norwich ~57%; Colchester ~74%; East Hampton ~79%.
The rest is cars & business equipment ().
The model shows the average home. Individual homes spread widely around it — in every measured town, lower-priced homes rose the most and the most expensive fifth the least, so increases concentrate at the entry level (see the Norwich and East Hampton pages for measured and estimated distributions, and the Colchester what-if to run your own home). Simplifications: one tax rate for all classes (Norwich’s fire districts and the statutory motor-vehicle cap are ignored), vacant land (~1–2% of these bases) omitted, shares normalized. Preset components are computed from CT OPM grand list data; each preset’s note states the town’s actual observed outcome for comparison.
Try this experiment with the machine above: keep every growth rate at the Norwich preset, and only move the “homes’ share” slider from 57% up to 79% (the commercial slider drops to 19% on its own, because the two shares together are capped at 98%). The average home’s increase falls from ~+26% to ~+14% — nothing about the housing market changed, only the base it sits in. Set the budget-change slider to zero as well, and what the revaluation does on its own falls from ~+19% to ~+8%.
When homes boom and the rest of the base doesn’t, the tax that used to be paid on commercial buildings, cars, and equipment has to land somewhere — and it lands on homes, in proportion to how much non-home base exists. A city like Norwich, with nearly half its base in those other classes, had a lot of tax to shift; a town like East Hampton or Colchester, where homes already carry three-quarters of the base, has little. Two towns can have identical housing booms and completely different homeowner outcomes.
The flip side: a town where commercial keeps pace with homes barely shifts at all — that is exactly what happened in Colchester’s 2021 revaluation (homes +23%, commercial +17%, cars and equipment +23%; average bill roughly flat), and it is the “quiet revaluation” preset above. Whether Colchester’s 2026 revaluation looks like Norwich, East Hampton, or its own 2021 depends on those relative moves — which is what the what-if scenarios bracket.
Everything above is the mechanism. This is the method: where the numbers on a town’s revaluation page come from, why one town carries more charts than the next, and what it means when a town carries none.
Counting parcels. Connecticut publishes every town’s assessor records. Take the file from before a revaluation and the file from after, match each property to itself by its parcel id, and the change in its assessment is measured rather than estimated. This is the strongest evidence available. It is also the narrowest, because it needs two files sitting on either side of the reset.
Reading sales ratios. Every arm’s-length sale is filed with the property’s assessment at the time. Sort those sales by assessed value, and the gap between assessment and sale price says how far each slice of the market has drifted. This works for all 169 towns, including towns whose revaluation has not happened yet, and it is an estimate. Where both methods run on the same town they agree closely. In Norwich the sales method put the cheapest fifth of homes at +74.3% and the most expensive fifth at +47.1%, against +74.9% and +52.6% counted parcel by parcel.
All 169 towns carry the drift chart, the flat-levy outlook and the sales-ratio view. For 46 towns the parcel-counted measurement has also been made, and it is drawn on that town’s Tax Outlook page. The figures behind it are published as data too, at /data/towns/<town>/revaluation_detail.json. The parcel measurement is demanding, and the honest summary is that most towns cannot support it:
Where a town has no parcel measurement its page says so rather than showing something weaker in the same frame. Nothing is estimated to fill the gap.
Each town’s assessor software writes its own labels, so the same kind of house is filed under different codes in different towns. Across the state there are about 9,900 combinations. Salisbury files most of its properties as “RES LAND”, which read literally means vacant land; taken at face value the town would appear to have almost no homes. Reading the code and the label together, and checking whether the property carries a building, sorts that out.
Sorting is not proof, so every town’s result is tested against a figure the state already published. The homes this site counts are added up and compared against the town’s residential grand list. Land within 10% and the town publishes. Miss and it does not. Sherman publishes no property use codes at all, so nothing can be classified there. The rest miss the published figure by between 11% below and 55% above, on their closest-matching file, which means this site is counting the wrong set of properties. Until that is understood, the honest thing is to publish nothing for them.
Which published figure is compared against turns out to matter. Connecticut reports houses and apartment buildings as separate assessment classes, and this site’s revaluation charts measure houses. Checking against the two classes added together made three towns, Greenwich, Rocky Hill and Vernon, look badly misclassified, at 28%, 45% and 61% above the published figure on the 2024 file. They were not. Their apartment stock is carried at values well above what the state counts in the apartment class, and merging the classes hid a question about one class inside a verdict about the whole town. Checked against houses alone, the same three towns on the same file come in at 0.99, 1.01 and 1.00.
Two further checks run before a town publishes, and both exist because a wrong answer here would look completely ordinary. First, a pair of files claims to bracket one revaluation, and the state already published what that revaluation did to the town’s houses. The pair has to reproduce that. Across the 46 towns the two agree to within 0.3% for the typical town and 7% at worst. Second, matching a property to itself needs an identifier meaning the same thing in both files, and the obvious one is not dependable: a few towns share not a single parcel number between their two files. A partly successful match is worse than an obvious failure, because the properties that keep their number through a renumbering are the ones that did not change, which would quietly understate how far apart the winners and losers landed. A town publishes only where at least 80% of its homes join; the typical town manages 99%.
Those two checks are not the same check, and Redding is why the second one exists. Falling back to its account numbers joined 84% of the town, comfortably past the threshold, and reported that the revaluation changed nothing at all: a factor of 0.999 where the state published 1.273. A join can pair a property with the wrong partner and still produce a tidy, complete-looking answer. Comparing whole-town totals cannot see that; comparing the measured result against the published step can, and does.
The sorting rules are in Data/cama_use.py; the check is Data/check_cama_coverage.py, which writes the per-town verdicts to Data/cama_coverage.json. The counts on this page come from that file. Both scripts re-run from the public data with no key.