NY-21 · House

The district, in brief

New York's 21st covers the North Country and the Adirondacks with parts of the Mohawk Valley and Capital District, including Glens Falls, Lake George, Plattsburgh, Potsdam, Herkimer and Rome, on the map first used in 2024. The incumbent is Republican Elise Stefanik, re-elected with 62.1% in 2024, who ran for governor, later withdrew, and is not seeking re-election.

The Republican nominee is Anthony Constantino, a businessman and candidate in the canceled 2025 special election, who beat assemblymember Robert Smullen 59.2% to 40.3% in the primary after trailing him in an April McLaughlin poll. The Democratic nominee is Blake Gendebien, a dairy farmer and the nominee in the cancelled special, who won his primary 64.6% to 35.0% over Stuart Amoriell; he is endorsed by Joe Morelle, John Mannion, Bill Owens, Jamie Raskin and the Working Families Party. Independents Richard Grayson and Christopher Schmidt are also running.

Every rater listed calls the seat R, all in September 2026: Decision Desk HQ (September 29), Cook (September 25), Inside Elections (September 22) and Sabato's Crystal Ball (September 16). As of June 30, 2026 Constantino had raised $12,060,254 with $3,036,386 on hand; Gendebien $5,461,528 with $2,622,585. Five general-election polls are listed: Public Policy Polling (D), September 9–10, 2026, Constantino 48%, Gendebien 43%; Impact Research (D), August 30 – September 2, 47–44; J.L. Partners, August 15–17, 50–40; SurveyUSA for WNYT, August 4–8, Gendebien 43%, Constantino 39%, Schmidt 5%; Impact Research (D), May 26–31, 45–44. The facts put the district's 2024 presidential two-party margin at -20.8 for the Democrat.

A paragraph quoting a model forecast from the day this brief was written is omitted; the current forecast is at the top of this page.

The facts, each beside its sentence

Not in the source: Five general-election polls are in the text but the model's basis says no district polls.; The Republican primary endorsement blocks are not labelled by candidate, so Trump's and others' endorsements cannot be attributed with certainty.; No presidential result percentages for the district in the text.; The text does not say why the 2025 special election was canceled.; No general-election ratings finer than 'R' are given.

Read from the district's section of Wikipedia's state page (CC BY-SA), archived as cfb3d3b04173; written by claude-fable-5-1 (batch reader) on 2026-09-30 and checked mechanically: every quote occurs verbatim in the archived text, every number in the brief occurs in the inputs. No probability here is the reader's.

Who holds the seat now

Elise M. Stefanik (Republican), in this chamber since 2015, 6 terms. Committees: House Committee on Armed Services; House Permanent Select Committee on Intelligence; House Committee on Education and Workforce.

From the congress-legislators dataset (CC0), archived as ae8cb9477c48. For a Senate race this lists both of the state's senators; the one whose term ends in the election year holds the seat on the ballot. For a House district redrawn for 2026 it is the member elected under the old lines with this number, who may be running elsewhere; the dossier above names who is running here.

How it has voted for president

A shift of 5.3 points toward the Democrats from 2016 to 2024. county returns placed on the current lines by 2020 block population; MEDSL county file (2016), an MIT-licensed county file (2020, checked against certified state results) and MEDSL precinct returns (2024). For a district split across counties this is an estimate, and where the state published the exact 2024 figure for these lines, that exact figure is the one the model uses.

Who are the voters?

Of about 543,265 adults 25 and over. In 2020 the district voted 42.0% Democratic (two-party) for president. With the same preferences and a 2022-style midterm turnout, it would vote 43.2% Democratic: the midterm electorate alone moves the margin 2.4 points toward Democrats. Group shares are estimated from how results vary across the state's census tracts (ecological inference, anchored to Pew's 2020 validated-voter figures) and calibrated to this district's actual 2020 vote; turnout is the Census survey's state rate by race and national rate by education, whose state samples carry margins of several points.

Not answered here: how each group will vote in 2026. These are 2020 preferences; this profile is description, not yet an input to the forecast.

The money

FEC bulk candidate summaries through June 30, 2026; $324,743 of independent expenditures for or against these candidates this cycle (FEC Schedule E, types 24E and 24A), which is the parties' and outside groups' own revealed view of whether this seat is in play.

The district in numbers

ACS 5-year 2023, the Census Bureau's own tabulation of this district (118th Congress lines). Unemployment is the BLS county series placed on the district by block population, not seasonally adjusted, so changes are same-month comparisons. Income, transfers and jobs are BEA county accounts placed the same way, compared with the average district. Every other figure is a named column of a named table in the Bureau's summary file, archived before it was read. This is what a reader knows about a district before looking at a poll; the model does not yet use it, and whether it should is a registered question.

Who is favoured, and by how much?

The 90% interval runs from -25.7 to +4.2 points, D-R. It is wide because this district has never been polled.

What is this built on?

A model trained on every contested House race 2012-2024 predicts this district's lean from its own record: presidential and House results on the current lines, incumbency, both candidates' prior offices, outside and party money before 1 October, demographics and public-assistance shares from the Census release before the election, local unemployment and income, and religious tradition. The national margin comes from presidential approval, the economy, the midterm penalty and the generic ballot corrected for its measured lean. Draws share national and state errors, so the seat total and this district come from the same simulation.

Not answered here: district polls, which are not yet folded in.

Why does the model say this?

Each figure is how far this district's predicted lean moves when that one feature is set to the typical value across 2026 districts, everything else held: 2024 presidential lean (R+19.3): 7.0 pts toward the Republican; 2024 House lean (R+21.6): 3.5 pts toward the Republican; asian share (1.0%): 0.8 pts toward the Republican; hispanic share (3.8%): 0.4 pts toward the Democrat; religion: mainline (5.9%): 0.4 pts toward the Democrat.

How sure is it, and of what?

Total sd 9.04 points, made of the seat's own volatility 91%, correlated polling error 9%. The decomposition is stored with the forecast rather than reconstructed, so a probability whose dominant input is a placeholder is visibly that.

Not answered here: a nowcast. 'If ballots were cast today' needs something that observes today, and for a House district nothing does. Publishing the election-day figure under both labels would make one forecast look like two.

How was this number made?

Run 6f211e6f-2bdc-4dda-8f72-81fc628dab0b, house_district_prior/0.2.0, as of 2026-10-02. Promoted by a pre-registered gate: on 2016-2022 its Brier score was 0.036 against 0.049 for the previous method, and on a single 2024 run 0.023 against 0.032 (experiments/district_gate_result.md, with an independent review).