Data & Sources

Source Update Log

• 2026-08-07 Two open questions about who pays what are now closed — and the sanitation fee turned out to be charged per dwelling unit, not per parcel. Both had been queued as questions to put to the City Assessor. Neither needed an ask; both were answerable from published records.
  
  The homestead option. Our city property-tax payer split assumes Elmira taxes every class at one rate. Until now that rested on absence — the word “homestead” appearing in none of the city’s budgets, audits or minutes — which is weak evidence, because it is a claim about every document we did not read. There is a positive source instead: New York publishes the register of municipalities that file to use homestead and non-homestead rates under RPTL § 1903 (archived copy), keyed by SWIS code. Elmira (070400) is not on it, and no Chemung County municipality is. Nor could Elmira adopt the option without first being certified an “approved assessing unit”, which follows a state-approved revaluation — and Elmira has not revalued since 1995.
  
  District coverage. One tax bill showed a library district, an Elmira sewer tax and a $365 sanitation fee, but could not show who else pays them. The Special District Summary on the city-totals page of the 2025 City of Elmira Final Assessment Roll (archived summary pages) counts them, against that roll’s own total of 9,555 city parcels: the sewer tax is on 9,512 parcels (99.5%), the sanitation fee on 6,725 (70.4%) with a further 1,057 parcels opted out, and the library district is not a city special district at all — it is billed within the county tax. The four structural lines ($65.19) are unchanged and remain what our calculations use.
  
  The correction. We had called the sanitation fee a flat $365. It is $365 per dwelling unit: the roll’s four fee codes are tiers at 1, 2, 3 and about 4 units per parcel, 8,226 billable units in all, so a two-family is billed $730. The sampled bill read as flat because that parcel has one unit. The per-unit reading reconciles — 8,226 × $365 = $3,002,490 against the city’s reported 2025 refuse-and-garbage revenue of $3,064,905, 2.0% apart, where a per-parcel reading lands 20% low. The fee’s regressivity is unchanged and still measured per unit. The summary also turned up a Downtown Development district (363 parcels ad valorem, 150 by unit) that this site had never recorded.
• 2026-08-07 The homepage now cites the 2025 Comptroller ranking, and two pages caught up with corrections the rest of the site had already published. Nothing new was measured here. Three claims were moved onto sourcing that already existed.
  
  The tax-rate claim. The homepage described Elmira as having "one of the highest property tax rates in New York State," sourced to the 2016 Financial Restructuring Board review and a 2020 Empire Center ranking. Both are a decade old. It now states what the Comptroller's CY2025 tax tables show: $18.49 per $1,000 of full value, second of the 61 cities that reported, behind only New York City. The two limits recorded when we added that source on 28 July travel with the claim in the footnote — it is the city levy alone (about $24.86 with Chemung's levy, before the school district's), and Ithaca filed no rate for CY2025, so it is unranked and is not a comparison this data can make. The older rankings are kept in the footnote as earlier evidence.
  
  The sales-tax framing. The homepage still summarised the sales-tax story as "the city's share fell from 12.33% to 8.2%" — the framing corrected on this page on 28 July, and already fixed on The City–County Relationship. The homepage card now leads with what actually changed: the county distributed 37.5% of its pool every year through 2014 and now distributes about 25%. The $22.8M is stated as what Elmira did not collect, not as evidence that the formula was rewritten against it.
  
  The PILOT percentages on the talk. The rate correction of 5 August moved Elmira College's $5,000 from 0.3% of a combined bill to 0.22%, and Arnot Health's $17,357 from 0.6% to 0.5%. The PILOT analysis and this log were updated then; the talk slides were not, and carried the superseded pair for two days. They now match. The percentages there are still hand-typed rather than generated, which is how they drifted; build_talk_data.py already computes the Arnot figure (0.49) but does not yet emit it into the slide.
• 2026-08-06 ✅ Rebuilt — the reassessment model is back, and it disagrees with the withdrawn one substantially. Elmira's homes are assessed at about 46% of market value, not the 78% we published for a year. New York State's own figure settles it in the new model's favour.
  
  What replaced what. The withdrawn curve took assessment ratios measured in bands of sale price and applied them as bands of assessed value. The replacement does not convert between the two at all. It asks the question the calculation actually needs — a home is assessed at $47,000; what does a home like that sell for? — and answers it directly: sort the city's sales by assessed value, then average the sale price in each group. Never the ratio. A ratio has assessed value on top and sale price underneath, so sorting on either one drags it; the average sale price within a group of assessments does not have that problem, because the thing being sorted on is no longer inside the statistic.
  
  The external check we should have found sooner. New York publishes a Residential Assessment Ratio for every municipality — its own estimate of what fraction of market value a place assesses houses at, separate from the all-property equalization rate. For Elmira in 2025 it is 52.08%, against a 56% equalization rate. Our rebuilt model says 46.5%. Those two sit beside each other; the withdrawn model's implied 78% sits nowhere near either. This series has been public the whole time (data.ny.gov bsmp-6um6) and is now downloaded by scripts/download_assessment_equity.py.
  
  What changed on the site. The reassessment page has its outcomes table and calculator back, and the presentation has its slide back. The rate pair is now $65.19 → $35.64 per $1,000 of assessed value: the base nearly doubles, so the rate nearly halves. Outcomes are far more compressed than before — the largest modelled change is a few hundred dollars a year, not $29,933 — because estimating from the roll alone can only give you the average for homes like yours, and the page says so.
  
  Two things we are deliberately not publishing. First, a single "X% of homes would pay less" figure. Elmira's assessed values pile up exactly where the model's winners and losers divide, so across equally defensible versions of the model that share runs from 41% to 49%, and under cruder ones to 74%. It is not identifiable, and it was a reader's refusal to believe the old 69% that uncovered the original error in the first place. We publish the range. Second, a direction for the $40–50K band, which straddles the crossover and holds more homes than any other — the table says "too close to call".
  
  Two claims elsewhere on the site were wrong and are fixed. The grievance advice on the reassessment page said a home assessed at $40,000 was worth about $35,000 and that its owner should file. Such a home typically sells for about $77,000; that advice would have sent people to lose a hearing. The page now explains the ground that does fit Elmira — unequal assessment, being taxed on a larger share of your home's value than your neighbours are, measured against the state's published RAR. And the rebate's $45,000 threshold loses one of its three justifications: the assessed-to-market ratio never reaches 100%, so nobody is "taxed on value that doesn't exist". The threshold itself does not move — the other two measures still land on $45,000 exactly.
  
  The Fair-Share Map is a different map now. The old one estimated each parcel's market value from its assessed value and coloured it by assessed ÷ that estimate — both halves derived from the same number, so every parcel with the same assessment got the same colour. It looked like a fairness map and was underneath a recolouring of the assessment roll. A better curve would not have fixed that; only a sale reveals what one particular house is worth. The new map therefore shows the ~1,400 city homes that actually sold since 2018, with older prices restated in today's dollars, and every dot is a measurement rather than an estimate. It carries addresses but no owner names.
  
  New: scripts/reassessment_model.py (the estimator), scripts/audit_reassessment_figures.py (re-derives every figure on the page and fails on drift, including a check that the rate pair is revenue-neutral over the same base — the check that would have caught $53.50 → $51.27 the day it was written), scripts/download_assessment_equity.py and scripts/visualize_assessment_fairness_map.py. The MV_PTS constant has been deleted from all three scripts that carried a copy of it, and the auditor fails if any script declares one again.
• 2026-08-06 New source — the state grades Elmira's assessment roll every year, and has since 2004. It has failed the professional standard every year since 2007.
  
  ORPTS publishes a coefficient of dispersion and a price-related differential for every New York municipality, split residential and all-property (data.ny.gov 4sut-q3dt). This is the regulator running the same two statistics the regressivity page computes from sales, on its own data by its own method. For 2024 it puts Elmira's residential COD at 46.2 and its PRD at 1.26; we compute 47.9 and 1.276. Two independent measurements, effectively one answer.
  
  The series also shows something we could not have shown from sales alone: Elmira's roll used to be fine. Its COD was 12.9 in 2006 — inside the IAAO limit of 15 — and has failed that standard by a widening margin every year since. Nothing was done to the roll to cause that. It is simply what happens to a roll left alone while a housing market moves underneath it. The chart is now on the regressivity page.
• 2026-08-05 ⚠️ Correction — the J-curve is now a City of Elmira study, and the headline gap is restated. The old figure was "72% more tax per dollar of value"; it should have been about twice as much, measured on the city alone. This is a correction in two directions at once, so it is worth being precise about what was wrong and what was not.
  
  What was wrong. First, the analysis pooled all 6,491 arm's-length single-family sales in Chemung County. But an assessment ratio only means something inside a single assessing unit, and the county has eleven, with state equalization rates running from about 1% in Ashland and Baldwin to 100% in Big Flats and Catlin. Pooling them mixed differences between town rolls into what reads as unfairness within one roll. Second, homes were grouped by sale price, which is also the ratio's denominator — so a house that happened to sell below its worth landed in a lower group and showed a higher ratio, tilting the curve downward on its own. Third, once restricted to the city the old eleven price bands were far too fine: the top three held 14, 12 and 2 sales, and the generator's minimum-count floor of 10 let two of them through. The up-curl of the J rested on about 26 sales.
  
  What was right, and stays. An internal review had proposed re-grouping by assessed value instead. That is the mirror of the same problem — assessed value is the ratio's numerator, so it tilts the curve the other way. The correct null is not a flat line in either direction. We now measure the tilt rather than assume it: 510 properties in the data sold twice within three years, which gives the dispersion of an individual sale around a home's true worth directly (0.232 in log terms, an upper bound since some pairs are renovations). Simulating a city with that much sale noise and no regressivity at all produces a gap of 1.27×. The observed city gap is 2.57×, so the corrected figure is 2.02× — smaller than the raw number, larger than the retired county-wide 72%.
  
  What the page leads with now. The IAAO Standard on Ratio Studies statistics, which are the assessment profession's own limits and are far less sensitive to this problem: a coefficient of dispersion of 47.9 against a standard of 15, a price-related differential of 1.276 against an acceptable 0.98–1.03, and a price-related bias of −0.572 against ±0.05. All three hold across every window tested. As an independent check, the median ratio for 2023–25 city sales is 0.500 against the state's published 56% equalization rate for Elmira.
  
  A new auditor, scripts/audit_jcurve_figures.py, now re-derives every figure on the page from the source data and fails if any drifts, including a guard that no price band falls below 30 sales. The Fair-Share Map and the rebate threshold depend on a separate assessed-to-market curve that has its own labelling error; both now carry warnings and are pending rebuild.
• 2026-08-05 ⚠️ Withdrawn — our model of who wins and who pays under a reassessment. It inverted a statistic, and manufactured the answer we were looking for.
  
  The reassessment page carried a table of household outcomes and a calculator; the presentation carried a winners-and-losers slide. All of it is down. It is the second-largest error we have had to correct, and unlike the Arnot one it flattered our own argument, which makes it worse.
  
  What went wrong. To model whose bill moves, you must estimate each home's market value from its assessed value. We did that with a curve lifted from our own J-curve analysis. The J-curve itself is sound — but it sorts sales into bands by sale price, and we applied its ratios as though they were bands of assessed value. That is not a valid inversion. Sorting on sale price selects homes that sold cheaply relative to their assessment, so the lowest band shows assessments above market almost by construction; reading it backwards as "a home assessed at $40,000 is worth $35,400" is a known statistical trap and we walked into it.
  
  How wrong — and why we are not simply publishing a corrected number. Re-sorting the same 1,689 sales by assessed value gives very different answers (homes under $40,000 assessed land at 0.825 of market rather than the 1.130 the model assumed; the modelled split moves from 69% of homes paying less to 39%). But that estimator is also biased, in the opposite direction, and we are not going to swap one flawed figure for another.
  
  The reason is structural. An assessment ratio is assessed value ÷ sale price, so whichever of those two variables you sort on also sits inside the ratio and drags the curve with it: sorting by sale price tilts it toward looking regressive, sorting by assessed value tilts it toward looking progressive. The correct baseline is therefore not a flat line, and judging either curve against "flat" is its own error. Calibrating it properly means separating sale-price noise from assessment error — estimable from repeat sales — and simulating what each statistic would show if assessments were perfectly fair.
  
  What survives — and it is the main thing. Read against a correctly simulated baseline rather than an eyeballed flat line, Elmira's sales show strong regressivity on the assessment profession's own measures. Lower-value homes really are taxed on a larger share of what they are worth. The roll has not been revalued since 1995. That finding is not weakened by any of this — if anything the first attempt at a fix understated it.
  
  What is withdrawn is narrower. The step from a sales-ratio curve to a prediction for an individual house. That is a change of variable, not a relabelling, and it is the step we got wrong. Until it is rebuilt we will not tell you what your own bill would do.
  
  The fairness map uses the same curve and is flagged in its generator pending a rebuild. data/tax-rates.json deliberately publishes no reassessment figures, so no page can pick one up by accident; the withdrawn variants are kept under reassessmentWithdrawn for the record. A rebuilt model, conditioned on assessed value, is tracked in the open-issues ledger.
  
  Found because a reader refused to believe that 69% of a city could get a tax cut. They were right — and the second half of this notice exists because the first attempt to fix it was wrong too.
• 2026-08-05 ⚠️ Correction — we published a combined tax rate of $53.50 per $1,000. It was a miscalculation, and it named no basis. There are two correct rates, and $53.50 was neither.
  
  A property tax rate is meaningless without saying which value it applies to. Elmira has two, and they differ by the equalization rate:
  
  • $66.81 per $1,000 of assessed value — what a city tax bill actually charges. City $29.98, county $8.06, community college $1.53, library district $1.07, Elmira sewer $0.55, school $25.61, plus a $365-per-dwelling-unit sanitation fee. Four of those lines are carried by every city parcel ($65.19); the sewer tax reaches 99.5% of parcels and the library district is billed on the county rather than the city roll, so the lower figure is what our calculations use.
  • $39.64 per $1,000 of full market value — the basis on which one municipality can honestly be compared with another, because every place assesses at a different fraction of market.
  
  $53.50 sat between the two. It was about 18% low against the bill and 35% high against full value. It entered the site in May 2026 as a constant in the reassessment calculator, spread to five pages, the glossary tooltip and five scripts, and was never checked against anything external. Worse, three files applied it to assessed values although it was closer to a full-value rate — so every "what this exempt property would owe" figure was understated by roughly a fifth. Those figures have moved: Arnot Health's PILOT is 0.5% of what its exemption shelters, not 0.6%, and Elmira College's is 0.22%, not 0.3%. The direction of every argument is unchanged; the magnitudes were wrong.
  
  A second error, larger in effect. The reassessment calculator ran $53.50 → $51.27. Those two numbers were mutually impossible for the model they described. Checking that is what led us to the model itself, which has now been rebuilt — see the notice above.
  
  Two things this fixes that we had not noticed. The tax-history table on decline.html ran assessed-value rates for 2003–2012 straight into full-value rates for 2013–2025 under a single heading labelled "combined," making 2013 look like a rate cut. It wasn't — New York's ORPTS changed how it reports, not what Elmira charged — and the figures under that heading were city-only, not combined. That table is now generated with both bases shown side by side. And we had assumed the city has no special districts, because the Comptroller's city table has no column for them. It does have them; a real bill shows three.
  
  Every rate on the site now comes from data/tax-rates.json, generated by scripts/build_tax_rates.py from the state's own rate dataset (data.ny.gov iq85-sdzs), the Comptroller's workbooks, the city's TAN official statement and an actual 2025 city tax bill, which the build reconciles to the cent. The build fails rather than publish a rate if two independent derivations of it disagree. Nothing about the rate is hand-typed any more, which is the only durable fix for an error of this kind.
• 2026-07-30 Two city zoning documents added to the catalogue — including the official zoning map, which we had wrongly believed did not exist. While double-checking the zoning pages before release we found that the city's Zoning Board of Appeals page carries the certified Official Zoning Map of the City of Elmira — adopted 21 December 1998 by Resolution 98-516, last printed 27 January 2021, attested by the City Clerk and Mayor — and that it already carries a "2020 Downtown Zoning" legend showing the form-based districts. The adopted form-based code PDF is on the same page. Our earlier research had checked only the DRI Zoning Update page, which still offers the February 2020 proposed map, and we generalised from it. Both documents are now archived under data/source-docs/zoning/ and cited on the zoning page.
  
  The zoning source cards above are corrected accordingly. What remains true: the city's posted ordinance is still the February 2010 printing, the official map is a scanned image that cannot be queried or joined to parcel data, and it predates Ordinance 2022-144. Our form-based boundary still comes from the February 2020 draft, because that is the only version published with georeferencing — so downtown assignments remain "close, not certain," though for a different reason than we first gave.
• 2026-07-30 ⚠️ Correction — we said Arnot Health paid nothing. It pays $17,357 a year. This is the most significant error we have had to correct, and it stood on the site for as long as the PILOT page has existed.
  
  The city's 2026 adopted budget worksheet itemises the transactions inside account 410810 — Other Payment in Lieu of Tax. Two of those lines are Arnot Ogden 1 OG ($16,016.14) and Arnot Ogden 2 OG ($1,341.19), together $17,357.33 in the 2026 adopted column. We had written, on several pages, that Arnot paid nothing at all.
  
  We missed it. The payments were sitting in a public document we had already archived and cited on this page. We have changed how we check these budget worksheets so that itemised detail like this is read directly in future.
  
  What changed, and what did not. Corrected on pilot.html (including a correction notice on the page itself), matters.html, elmira-equity.html, the presentation page, scripts/build_talk_data.py and the repository README. The underlying argument is unchanged and, we think, better stated: $17,357 against $54.4M of exempt assessed value — roughly $3.55M a year at the combined rate of $65.19 per $1,000 of assessed value — is 0.5%. The case was never that a zero existed; it is that the contribution is a rounding error against the exemption. But "nothing" and "almost nothing" are different claims and we published the wrong one.
  
  Three further corrections found in the same pass. (1) The "~$324K/yr the city collects in PILOTs" figure was the 2021 actual column; the 2026 adopted total is ~$569,146 (410800 $138,335 + 410810 $430,811). (2) The $54.4M is held by 18 wholly exempt parcels, not the 47 we cited — Arnot's other 29 city parcels are ordinary taxable property ($914,150 of taxable value, mostly houses near the campus) on which it pays normal property tax, which we had never mentioned. (3) elmira-equity.html carried a "~$90M in assessed value / $4.8M theoretical liability" figure for Arnot that reconciles to neither the city total ($56.4M) nor the county-wide one ($105.5M); it has been replaced with the verified city figures.
  
  Still open, and not guessed at: which parcels or agreement the two Arnot payments arise from — nothing matching "Arnot" appears in the county IDA's PARIS project data, so these are not IDA deals — and whether Arnot pays anything to the county or the school district, since the city's budget is the only book we can see. A records request is open. We re-ran the same check against Elmira College and LECOM: neither appears anywhere in the PILOT accounts, so the existing claim that the college's $5,000 sits in account 412890 rather than a PILOT line stands.
• 2026-07-30 The presentation now carries the zoning figures. site/talk-data.js (the data blob behind the unlisted presentation page) gains a zoning section lifted directly from zoning.json by scripts/build_talk_data.py — nothing recomputed, the page's own generated numbers re-shaped, so the talk cannot disagree with Zoning, Explained. Same denominator discipline as the page: every lot judged by the code that governs it (84.6% of all 8,286 residential lots held to one home; 93.7% outside the 2020 form-based code, 0.9% of the 811 residential lots inside it). The presentation itself was reframed around what it always practiced: one site where the scattered public record — Comptroller filings, the assessment roll, SalesWeb, PARIS, council minutes, the zoning ordinance, county GIS — is lined up, explained, and linked back to its sources.
  
  And an answer to "why is Ithaca missing?" The tax-rate source card above used to say only that Ithaca "did not file," which invites the wrong inference — that filing is optional, or that Ithaca's rate is too low to bother reporting. Filing is required of every city by General Municipal Law § 30. Ithaca's state reporting has run years behind: OSC audit 2025M-114 (27 March 2026) found no annual report filed for 2022–2024, independent audits years in arrears, and no fiscal-stress score since 2017. The card now says so, and cites the audit.
• 2026-07-29 New: zoning. And a correction to something we had implied — Elmira did modernise its zoning code. We had treated Elmira as a city that never reformed its 1998 zoning. That is wrong. On 17 August 2020 the City Council adopted a form-based code (Ordinance 2020-241 § 38, now Appendix B-1), which legalises two-family, three- and four-family housing and accessory apartments by right downtown, cuts minimum lot sizes, and removes parking minimums in the Central Business District. The same ordinance abolished the old Business B district.
  
  The substantive finding is narrower and sharper than "the zoning is bad." Reading the use table alone understates the problem badly, because the bulk table is what actually binds. Applying both to every residentially-zoned lot, each judged by the code that governs it: 85% of Elmira's 8,286 residential lots (7,010) can hold exactly one home. The split is the finding. Outside the 2020 form-based code it is 93.7% (7,003 of 7,475). Inside it, 0.9% — 7 lots of 811. Same city, same housing stock, same century of construction; the difference is which side of a 2020 boundary a lot falls on. Residence B is named "One–Two Family," but a two-family there needs 10,000 sq ft and 60 ft of frontage, and only 381 of its 1,788 lots outside the 2020 code qualify — 316 of them have the frontage but not the land.
  
  Denominator note, because we got this wrong twice before publishing. A first draft said "87%," counting business and industrial districts in the denominator. A second said "91.1% of 8,286," counting every Residence-district lot under the base code — wrong, because 811 of them sit inside the 2020 form-based code, where two-family is permitted by right on a 1,000 sq ft lot, so a base-code cap of one is not the operative answer there. The published version keeps all 8,286 lots but scores each under its own governing code, which is both complete and correct. The worked example was replaced for the same reason: it had used 201 Columbia St, which is inside the 2020 code — the ordinance names Columbia Street in the Neighborhood Mixed-Use intent — so that duplex can be rebuilt, the opposite of what we had said. It now uses 1008 Pratt St, which is outside.
  
  The 2020 code covers 1,603 parcels, roughly a tenth of the city's homes. So the practical question is not whether to write a reform, but where the boundary of the one already adopted should go.
  
  New page: Zoning, Explained, with a parcel-level zoning map that answers "what may I build here?" for every lot in the city and names the rule that sets the limit. Both are generated by scripts/visualize_zoning_map.py into zoning.json. Two source cards added above.
  
  Two things we deliberately did not publish. First, assessed value per acre by housing type does not support a headline here: the median apartment parcel is $324K/acre against $290K/acre for single-family, and two-family parcels assess below one-family. That is the frozen 1995 roll talking, not the market, so we left it out rather than lean on it. Second, we do not model parking: inside the form-based code the ordinance says off-street parking "may not be required" and the Planning Board sets it case by case, so there is no fixed number to compute; outside it the old one-space-per-dwelling rule applies but we do not subtract parking from buildable area. The unit counts reflect the use and bulk tables only.
• 2026-07-28 New source — the Comptroller's full-value tax-rate tables, and what they say about where Elmira ranks. The site has long described Elmira as having "one of the highest property-tax rates in New York State," sourced to the 2016 Financial Restructuring Board review and Empire Center rankings. The State Comptroller's annual local government tax tables let that be stated precisely instead: for CY2025 the city's full-value tax rate is $18.49 per $1,000 — second of the 61 cities that filed, behind only New York City.
  
  Recorded here with the two limits that go with it. The figure is the city levy alone; adding Chemung's county levy on city property brings it to roughly $24.86 per $1,000 before the school district's levy. And the ranking covers only cities that filed — the City of Ithaca did not file for CY2025, so it is absent from the table and must not be used as a comparison on this data.
  
  Workbooks for 2003–2025 are archived at data/source-docs/tax-rates/, and the 2025 workbook is served directly from this site so the ranking can be checked against the Comptroller's own figures without hunting for it. First used to build site/talk-data.js.
• 2026-07-28 Correction of framing — the city's share wasn't cut. The county stopped sharing. We had described Elmira's sales-tax problem the way the city does: its slice of the countywide pool fell from 12.33% to about 8%. That is arithmetically true, and we verified it to the dollar. But it invites a conclusion the data does not support — that the sharing formula was rewritten against Elmira.
  
  Checking the county's own books year by year shows something different. Chemung distributed exactly 37.5% of its sales-tax pool to municipalities every year from 2007 through 2014 — a round three-eighths, held eight years running. From 2015 it steps down: 31.5%, 29.7%, 27.4%, settling near 25% from 2022. Meanwhile the City's share of the money actually distributed never breaks — 31–35% across the whole period, with no discontinuity at 2015 — and towns and villages hold ~65% throughout.
  
  So the formula did not change; the county's retention did, from 62.5% to about 75%. This is what NY Tax Law § 1262 permits: it governs only revenue “not used for county purposes,” so a county can enlarge its own retention and shrink the shared pot without touching the formula. It is also how both sides can be telling the truth at once — the county that its population-based formula is unchanged, the city that its receipts collapsed. City & County now leads with that finding and a new chart of the retention share, and the $22.8M figure carries a note explaining exactly what it measures.
• 2026-07-28 The sales-tax Sankey now names every recipient. The countywide pool chart used to end at a single “Towns & villages” block worth $12.8M, because the county files its whole municipal distribution as one line. Each town's and village's own filing fills that in: Town of Horseheads $2.57M, Southport $2.15M, Big Flats $1.73M, Village of Horseheads $1.43M, Town of Elmira $1.35M, then seven smaller recipients.
  
  A trap worth recording, because it nearly shipped. The obvious way to pull this is account code A1120 — that is what the City and the County both use. Towns don't: they split the distribution across four funds (general, part-town, and two highway funds), so an A1120-only query returns 35% of the money and silently omits the two largest recipients, including the Town of Elmira. It was caught because the reconciliation was run before anything was built on the numbers: $4.5M against $12.8M is a 65% gap, and a gap that size means the query is wrong, not that the money is missing. Corrected, the municipalities' filings reconcile to 95.7% of the county's figure, and the remaining 4.3% is drawn as its own node rather than quietly spread across the named towns.
• 2026-07-28 New — seventy years of Elmira's assessment roll, in one line. Frozen Assessments has a new chart built on the state's equalization-rate series: the roll drifted from 48% of market value in 1970 to 13.8% by 1994, was reset to 112.9% by the 1995 revaluation, and has slid back to 56% since. The same graphic now anchors “Why the Freeze Persists” on Reassessment, which also gains the state's own price for fixing it — $341,000 gross, $224,000 after aid, in 2009 dollars.
  
  Two judgement calls worth stating. The series exists back to 1954, but the earliest years repeat a single rate for up to five years at a time, so the chart starts in 1970, where annual measurement begins — we would rather lose sixteen years than draw a trend through carried-forward numbers. And the Two Streets, Up Close section was retired to make room: it showed two hand-picked streets as bar charts, which the parcel map above it already covers in full and without the selection question. A chart of every parcel beats a chart of two streets we chose.
• 2026-07-27 Correction — the sales-tax divergence chart now starts at 2007, and the countywide pool is verified against the State. The chart on City & County plotted the county's retained sales tax from 1995. It shouldn't have: Chemung changed how it presents that number in 2007. Through 2006 it booked the collection net of what it passed to the city and towns, reporting no distribution at all (A19854 = $0 every year); from 2007 it books gross. A line spanning both regimes shows a 66% jump at 2006–07 that is pure bookkeeping — the distribution share implied by the last net year (37.7%) is the same as the first gross year (37.5%). Nothing happened to the money. The chart now begins at 2007 and says why. The page's argument is unaffected: it was always about the divergence after the mid-2010s, well inside the clean period.
  
  Separately, the pool itself is no longer taken on the county's own word. The NYS Tax Department's distribution record is an independent account of the same money, and it agrees with our figure to a median ratio of 0.996 over the 18 comparable years (worst year 0.963). That check now runs on every build.
• 2026-07-27 Sourced at last — Elmira's last citywide reassessment was 1995, 31 years ago. In July we flagged our own claim that the roll had gone “decades” without a genuine citywide reassessment: it was probably true, but nothing in our holdings pinned the date, so we softened the wording to what we could prove (99% of residential assessments unmoved across the five available rolls). The date now has a state source. New York's most-recent-reassessment list shows two undistinguished “Elmira” rows, 1995 and 2013, because the city and its parent town share a name — the same collision that has tripped this project before. ORPTS's 2009 Chemung County assessment study resolves it: Table A-4 is keyed by SWIS code and lists 70400 City of Elmira, Latest Reassessment 1995, with the Town of Elmira (73000) separately at 1992 and a project planned for 2009 — so the 2013 row is the town's. The claim we hedged was correct, and matters.html now states the year and cites both documents. The 92% level of assessment the study records for the city in 2008, against roughly 56% today, is consistent with no reset since.
• 2026-07-27 Correction — the value-per-acre page was counting Town of Elmira parcels as city parcels. per-acre.html is a city-only argument, but the script built its “City of Elmira” set by matching the municipality name “Elmira” on the ORPTS roll. New York files parcels under the parent town, so that also swept in the Town of Elmira and its special franchises — larger-lot suburban parcels that are not in the city and pay no city levy. The page’s parcel set has been rebuilt on SWIS code 70400, the city’s own assessing unit: 14,398 → 9,486 parcels, and single-family parcels 9,508 → 5,854. A second flaw in the same join — parcel keys matched without their municipality, producing duplicate rows — was fixed at the same time.
  
  Figures that moved, all now city-only: downtown common-wall rows $1.06M → $1.12M per acre; single-family $319K → $292K; vacant land $18K → $21K; value per foot of frontage for downtown rows $2,333 → $2,291 and for parking $268 → $276. The lot-size gradient steepened once the town’s large lots were removed: value per acre now falls from about $318,000 on a 0.1–0.2 acre lot to about $96,000 on a 0.75–1 acre lot. The page’s conclusion is unchanged — on several measures it is now stronger — but the numbers it was stating were not the ones it claimed to be stating. Every named parcel on the page (Wegmans, Aldi, Weis, the downtown blocks, the Water St lot) was verified unaffected.
  
  A premise was corrected at the same time. That section used to open “a house is a house,” treating large-lot and small-lot homes as equivalent. Once the town’s parcels were removed, the city’s own roll showed they are not: median assessed value rises with lot size, from $46,000 on a 0.1–0.2 acre lot to $83,000 at 0.75–1 acre, and to $168,500 above an acre. The page now says so, and makes the stronger claim the data actually supports — the large-lot home is worth nearly twice as much and still returns under a third the value per acre. The chart is capped at one acre: above that, Elmira has only ~34 single-family parcels, and at 3.6× the typical city home their higher value per acre reflects the houses, not the land use.
• 2026-07-27 Method — the city property-tax payer split is now labelled estimated, and it is checked on every build. The City Budget Explorer splits the ~$17M property-tax line by who pays it. That split is an apportionment by each class’s share of the taxable base, not a figure anyone reports, and it rests on one assumption: that Elmira taxes all property classes at a single rate. We checked. The city’s own 2025 TAN official statement (p. 34) prints one rate per year, not two; “homestead” appears in none of the adopted budget worksheets, audited financial reports, or rate-setting council minutes; and applying the city’s published rate to the taxable base we compute reproduces the published levy to within 2% in every year, 2021–2025. The build script now runs that reconciliation and records the per-year workings in proptaxSplit inside city-budget.json. One thing the check surfaced: the dollars being split are property tax collected, not levied — and in 2024 the city collected 90% of its $16.17M levy, a $1.6M shortfall, against 100% collection in every prior year shown.
• 2026-07-24 Correction — Elmira College does pay a $5,000/yr PILOT, frozen since 1994. We previously listed the reported 1994 college agreement as "under review." The City Chamberlain's office has since confirmed the payment and identified where it sits: it is not in the city's PILOT accounts (410800/410810, the federal and IDA/housing PILOTs) but bundled under 412890 Other General Department Income in the adopted budget — which is why every PILOT-line, audit, and state-series search missed it. pilot.html and matters.html now state the $5,000 figure and set it against the scale of the exemption: the college holds $34.8M of exempt assessed value (58 city parcels, SWIS 70400, 2025 roll) — the value of about 741 median Elmira homes ($47,000 each) — while paying roughly what one and a half such homes pay in combined property taxes (~$3,064 each at the combined rate of $65.19 per $1,000 of assessed value; the college's exemption covers all three levies, the same three the homeowner pays). For scale, the pages note that 1% of the college's exempt value would be ~$348,000/yr. The college is framed as nearly (not fully) exempt; Arnot Health still pays $0 [corrected 2026-07-30 — Arnot pays $17,357/yr; see the entry at the top of this log]. The executed 1994 agreement has not yet been obtained; a records request is open.
• 2026-07-21 Value-per-acre page rebuilt to make the argument on two measures. The page had become a list of caveats that made no claim. It now states the Strong Towns / Minicozzi case directly, still using City of Elmira parcels only: (1) value per acre by land use and for named parcels — including three of the city's own supermarkets (Aldi, Weis, the Wegmans plaza), which carry far less value per acre than downtown blocks; and (2) a new value-per-foot-of-frontage measure as an explicit, hedged proxy for the street/pipe/sidewalk the city maintains per property (downtown row ~$2,333/ft vs. parking ~$268/ft), plus a single-family "same house, more pipe" chart (small lots ~50 ft of frontage / ~$342K per acre vs. large lots ~200 ft / ~$146K per acre). Frontage comes from the roll's front field and is reported only where meaningful — large-footprint supermarkets (recorded front = 0) are kept off the frontage charts because a tiny deed frontage would flatter them, the same trap that retired the earlier front-foot page. Regenerated per-acre.json from scripts/visualize_per_acre.py.
• 2026-07-21 Removed the smallest-lot band from the Elmira lot-size curve. The prior page highlighted a $474,000-per-acre median for 595 single-family parcels at or below 0.1 acre. That median was not driven by one outlier, but the band includes tiny and unusual parcel geometries where boundary or acreage errors are greatly magnified by division. The entire band is now excluded from the page and generated chart data; the analysis begins with 5,092 parcels between 0.1 and 0.2 acres. No individual observations were selectively removed.
• 2026-07-20 Analytical correction — per-acre page narrowed to the City of Elmira. The previous version compared raw assessed value per acre across municipalities with different reassessment histories and treated lot acreage as evidence of infrastructure cost. The revised per-acre analysis removes the Big Flats, Horseheads, county big-box, and municipal-ranking comparisons; regenerates per-acre.json from City of Elmira parcels only; and describes assessed value per acre as a land-intensity measure rather than a service-cost calculation. Infrastructure claims now state that frontage, network length, capacity, and maintenance costs require separate data.
• 2026-07-20 Claim under review — reported 1994 Elmira College PILOT. A reader reported that Elmira College entered a $5,000 agreement with the City in 1994. We have not located the executed agreement or an approving Council resolution in reachable public records. The archived 2025 and 2026 budget worksheets itemize federal, housing, and IDA-related PILOT receipts but do not name Elmira College; the City's 2020–2024 audited reports do not identify such an agreement either. The categorical “no agreement / $0” wording on the PILOT page and Why It Matters has been replaced with that narrower finding. Confirmation requires the 1994 agreement or Council resolution and the current City revenue account, if any. → Resolved 2026-07-24: the Chamberlain confirmed the payment and its revenue account; see the top entry.
• 2026-07-16 Added the IDA section — five new pages and a new dataset. The Chemung County IDA, every project, the board, the city lens and the agency's own property are built from the sources catalogued above: the NYS Authorities Budget Office's PARIS filings (9rtk-3fkw, FY2017–2024) and the IDA's footprint on the ORPTS assessment rolls (exemption code 18020), joined project-to-parcel by hand and published as ida.json (scripts/build_ida_json.py), plus the CCIDA document center, the UTEP, the county charter and the news reporting cited on the pages. Every dollar is generated by a script whose assertions pin it to the sourced project-by-project table; project categories and the two mixed-use deals were verified against the executed agreements, not PARIS's coarse purpose field. Headline findings: 49 active deals, ~$105M of assessed property paying almost no property tax, a net $3.2M/yr subsidy of which school districts bear 53% with no vote on any of it. Two things we state plainly because they cut against the argument: PILOT compliance is essentially perfect (due vs. paid match to within $271 across eight years and $23.5M of payments), and agreements do end on schedule — CVS's $54.5M distribution center returned to full taxation in 2025. All document links were verified to return a PDF.
• 2026-07-14 Restructured the City of Elmira page; withdrew the "within-street inequity" framing. The Elmira fiscal page previously presented street-level assessment spreads (e.g. "57× spread on Clinton St") as evidence of inequity between neighbors. That framing compared each home to its street's median assessment — but different homes on one street genuinely differ in value, so a wide spread is not by itself evidence of unfair assessment, and we don't have the parcel-level market data to say which neighbors are treated unequally. The claim was removed. Problem 3 on that page now makes the equity argument the way this site can actually support it: the sales-ratio (J-curve) analysis of 6,491 real sales, which shows the frozen roll taxes the cheapest homes hardest. The equalization-ratio bar chart and the blight histogram were also removed from that page in the same trim (the underlying data remains in the published JSON), and the 3D city map moved to the top of the page.
• 2026-07-14 Added an assessed ⇄ market-value toggle and a City of Elmira 3D map. Because municipalities assess at different fractions of market value (City of Elmira's frozen roll is at a 56% equalization rate; some towns are near 1%), raw assessed value per acre overstates towns that reassess. The county 3D map now offers ORPTS full market value per acre as a toggle (the equalized, town-to-town-fair figure, from the same assessment roll), and a new City of Elmira 3D map on the city page shows the city alone, where one shared roll makes parcels directly comparable. Same parcel-boundary source as below.
• 2026-07-14 Added the 3D value-per-acre map and its parcel-boundary source. The per-acre page now embeds an Urban3-style 3D map: every parcel footprint extruded so bar height = assessed value per acre. Boundaries come from the Chemung County GIS public feature service (geometry vintage 9/2021), joined to the 2025 roll by print key + SWIS code. Bar heights are capped at $8M/acre because a few sliver parcels carry institutional assessments on tiny polygons; tooltips always show the true value. Color runs red (least productive) to green (most productive), and fully tax-exempt parcels — county-taxable value of zero on the 2025 roll, the same test used by the Elmira exemption analysis — render blue regardless of value. The 2D dot map remains available as a lighter fallback.
• 2026-07-07 Added a note explaining why maps show owner names. The three maps whose parcel popups include owner names (all parcels, fair share, reassessment status) and their host pages now carry a short note: the names come from the county's public assessment roll, the same record available on the Chemung County parcel search (Beacon), and ownership patterns (LLCs, absentee landlords, exempt institutions) are part of the public record these maps exist to show. Nothing was added or removed from the data itself.
• 2026-07-07 New page: Contact Your Officials. Names, emails, and phone numbers for the Mayor, all six district councilmembers, the Interim City Manager's office, the City Clerk, and the City Assessor — all drawn from the City of Elmira's website and staff directory, verified July 2026. Notable: council email addresses are district-based aliases (e.g. [email protected]), and Charmain Cattan has been Interim City Manager since June 13, 2026.
• 2026-07-07 Fixed a dead NYS grievance link; added the City Assessor's Office as a source. The Grievance Day section linked to a tax.ny.gov page that no longer exists; it now points to the state's current Contest Your Assessment guide, and to the City of Elmira Assessor's Office page, which confirms the Grievance Day details published here (third Tuesday in July, 4–8 PM, City Hall; RP-524 forms accepted after July 1).
• 2026-07-04 Correction — the Water Board's parcel footprint was undercounted. The Water Board page's "what it pays" section originally counted only parcels titled "Elmira Water Board," which understated the system by about half and left the audited $236,946 in suburban real-estate taxes un-reconcilable. Per City Charter § 166-n the Board acquires property "in the name of the City of Elmira," so much of the system — including the filtration plant at 1 Fountain Dr ($9.8M) — is on the roll under "City of Elmira." Counting the water-infrastructure classes under both owner names: the exempt in-city infrastructure is ~$19.65M (not $9.56M), the taxable suburban base is $6.26M (not $4.05M), and the audited tax reconciles at a normal ~$37.9/$1,000. The parcel query in scripts/build_budget_json.py was updated accordingly and water-board.json regenerated.
• 2026-07-04 Water Board page — added the charter and election-turnout records. Archived the Elmira Water Board charter (Article X-A) and the certified 2025 and 2024 election canvasses. These document a striking, underreported fact now on the Water Board page: the board's commissioners are elected in a city-run June special election in which only City resident taxpayers on the assessment roll may vote, and recent races drew just 8 votes (2024) and 24 votes (2025), both unopposed. The charter also fixes an earlier gap — commissioner pay is set at $75/meeting (§ 166-a), not "undisclosed."
• 2026-07-03 New page: The Water Board. Added The Water Board (under the Elmira nav group) — an investigation of the city-owned Elmira Water Board's surplus, its unused fair-return mechanism, and the live dispute over it. New primary sources catalogued: the NY Attorney General's 2007 opinion on the city charter and the Water Board's 2025 audited statements. The chart and parcel figures are generated into water-board.json by scripts/build_budget_json.py from the OSC water enterprise-fund lines (EW8029/EW628/EW99019) and the 2025 assessment roll; the net-position series is asserted equal to the audited statements at build time. The "City" nav group was renamed "Elmira."
• 2026-07-02 Strong Towns page retired — a correction. The page's headline metric (assessed value per foot of street frontage) did not support its claim: downtown row buildings ($2,525/ft) and shopping centers ($2,574/ft, n=5) are statistically tied, because a big-box parcel's internal parking and drives never front a public street — the metric flattered exactly what the page argued against. Rather than keep it, we retired the page (it redirects to Tax Value per Acre, where the per-acre framing makes the argument correctly). The defensible charts (value per square foot of lot; retailer land-share) moved there; the top-10 property lists moved to the City of Elmira page and now cover city parcels only. strong-towns.json was renamed land-productivity.json. Also split the Employee Benefits chart's look-alike band colors.
• 2026-07-02 Top-10 lists now generated; benefits deep-dive added. The "Top 10 by Assessed Value" list on the Strong Towns page is now generated directly from the 2025 assessment roll (topProperties in strong-towns.json, built by scripts/visualize.py) instead of hand-maintained — this corrected one error: the Horseheads Real Prop LLC warehouse (120 Wygant Rd, $23.2M) is exempt on the county roll, so 7 of the top 10 are exempt, not 6. Added a companion "Top 10 by taxable value" list. On the city budget page, added an "Inside Employee Benefits" chart: OSC child accounts normalized across the 2013 renaming ("Hospital & Medical (dental) Ins" → "Hospital", "Police & Firemen Retirement" → "Police Retirement", etc.) into five buckets via a mapping table in scripts/build_budget_json.py; the build verifies the buckets sum to the published Employee Benefits total every year.
• 2026-07-01 Plain-language pass. Added a site-wide glossary (tap any dotted-underlined term for a definition; glossary.js is the single source for all definitions), rewrote the densest explanatory passages (equalization rate, audited-accounting terms, tax certiorari), expanded shorthand in chart axis labels, and re-worded the map legends ("changed owners" instead of "transferred", etc.). No data, numbers, or sources changed — wording only; the three affected maps were regenerated from the same 2025 roll.
• 2026-06-28 Stage 2 chart retrofit. Replaced the pre-rendered matplotlib images on the property-tax pages (Regressivity, Value per Acre, Trends, Elmira fiscal health, Frozen-With-Exceptions, Strong Towns) with interactive ECharts driven by JSON. Documented the NYS ORPTS SalesWeb sales and parcel-acreage sources, and added the six derived chart-data files (jcurve / per-acre / trends / elmira-fiscal / equity / strong-towns .json), each built by its scripts/visualize_*.py.
• 2026-06-26 Added direct links to the city–county source documents: the City's Oct 6 2025 sales-tax presentation (County CivicClerk + local copy), and the county-side records as individual cards — the 2017 Shared Services Report, the June 30 2025 DPW termination notice, the Sept 24 2025 timeline letter (each with its official Chemung County DocumentCenter link and a local copy), and the Oct 2 2025 legislature meeting summary (research note).
• 2026-06-26 Rebrand to Open Elmira. Added City Budget Explorer, County Budget Overview, and City–County Relationship pages driven by a new OSC line-item pipeline (scripts/build_budget_json.py → data/*.json). Consolidated all data sources onto this central page; the old decoder/sources.html now redirects here.
• 2026-06-25 Added City of Elmira FY2020 and FY2021 audited reports from MSRB EMMA; extended the decoder to a five-year series (FY2020–2024).
• 2026-06-25 Archived EURA reports (FY2023–2025), the 2025 Budget Worksheet, the 2025 TAN Official Statement, and the 2026 Adopted Budget worksheet.
• 2026-06-25 Initial decoder build: FY2022–2024 audited financials, FRB 2016 review, OSC audit follow-up (2020), OSC fiscal profile (2013).