New York Election Polls Understate Democratic Margins

New York’s November 3 general election returned Democratic incumbents to the state’s top offices, with preliminary results showing Governor Kathy Hochul winning 59.72% against Bruce Blakeman’s 40.07%. The October 28 Siena Research Institute poll had projected a narrower 54% to 40% contest.

Siena Polls Projected Democratic Leads

The final October 28 survey by the Siena Research Institute placed incumbent Governor Kathy Hochul ahead of Republican challenger Bruce Blakeman by 14 percentage points among likely New York voters.

Siena’s October 28 polling showed Hochul maintaining a double-digit advantage over Blakeman entering Election Day. The preliminary results subsequently showed a wider gubernatorial margin than Siena had anticipated.

In the attorney general race, incumbent Democrat Letitia James led Republican Saritha Komatireddy by 54% to 39%. The preliminary attorney general results subsequently exceeded that projected advantage.

The comptroller survey placed incumbent Democrat Thomas DiNapoli ahead of Republican Joseph Hernandez by 59% to 32%. DiNapoli subsequently recorded a wider preliminary margin than the survey indicated.

Across the three contests, Siena’s polling consistently identified Democratic advantages. The discrepancies concerned the size of those advantages rather than which candidates were leading before Election Day.

These surveys measured preferences among likely voters. Their estimates depended on sampling and assumptions about participation, unlike election returns derived from ballots actually counted.

Preliminary Results Exceeded Polling Margins

With 93.06% of election districts reporting as of November 7, Hochul had received 3,639,098 votes against Blakeman’s 2,441,594.

Their reported vote shares produced a 19.65-point margin, approximately 5.65 percentage points wider than Siena’s forecast. Hochul exceeded her polling estimate by 5.72 points, while Blakeman’s preliminary 40.07% share remained close to the projected 40%.

Attorney General James received 3,593,711 votes (60.2%), compared with 2,368,991 votes (39.7%) for Komatireddy. The resulting 20.5-point difference exceeded Siena’s projected margin by approximately 5.5 percentage points.

DiNapoli recorded 3,833,740 votes (64.6%), while Hernandez received 2,092,847 votes (35.3%). Their 29.3-point difference exceeded the forecast margin by approximately 2.3 points.

The comptroller contest differed from the gubernatorial race because both candidates exceeded their projected vote shares. DiNapoli’s increase was greater, resulting in a wider margin despite Hernandez also outperforming his polling estimate.

The figures remain preliminary rather than certified. County election boards must complete their canvasses, including the consideration of applicable provisional and challenged ballots, before New York’s statewide certification process concludes.

Pew Identifies National Polling Differences

The polling differences in New York coincided with a broader national pattern examined by Pew Research Center.

In its November 6 analysis, Pew Research Center examined polling performance across 27 competitive gubernatorial and U.S. Senate contests.

The organization found that pre-election surveys underestimated Democratic margins by at least two percentage points in 23 of 27 races. The average understatement across those contests was approximately four percentage points.

New York’s gubernatorial and attorney general discrepancies exceeded that national average. The comptroller contest showed a smaller difference between its projected and preliminary victory margins.

Pew’s national analysis provides context for understanding why correctly identifying election leaders does not necessarily mean that polling precisely captures their eventual level of support.

The national polling comparisons also involve different political populations. Results from competitive Senate and gubernatorial races cannot automatically explain voting patterns in every state or legislative district.

RealClearPolitics polling averages had indicated a Democratic advantage approaching nine percentage points on the national generic congressional ballot before Election Day.

The national generic-ballot average measures broader party preferences but cannot establish vote margins in individual congressional districts.

The reported national polling differences do not establish which methodological factors contributed most to the discrepancies.

Finger Lakes Outcomes Need Local Data

Statewide results cannot establish how candidates performed within individual Finger Lakes legislative districts.

The reported statewide figures do not include sufficient Finger Lakes district-level vote totals to establish individual legislative outcomes.

New York’s statewide races reflect votes cast across the entire state, while legislative elections involve smaller geographic constituencies. Political support measured statewide may therefore differ from the distribution of support within particular districts.

The distinction is relevant when interpreting the Democratic margins recorded for governor, attorney general, and comptroller. Those results cannot independently establish regional legislative outcomes.

District-level results would provide a more precise comparison of regional political preferences with statewide voting patterns.

Independent Voter Polls Showed Differences

Differences between major polling organizations became especially apparent in their findings about voters outside New York’s major political parties.

During October, Blakeman’s campaign challenged public polling as surveys produced conflicting estimates of independent voter support.

Both Siena and Marist placed Hochul ahead statewide, but their findings for smaller demographic groups differed.

Siena’s combined independent-and-other category favored Hochul, while Marist’s self-identified independent sample favored Blakeman.

These categories were not equivalent. Siena combined independents with other voters, whereas Marist examined respondents specifically identifying as independents.

Differences in classification complicate comparisons between the surveys. Smaller demographic samples also generally carry greater statistical uncertainty than estimates based on a survey’s full statewide sample.

The conflicting subgroup findings therefore cannot independently establish which candidate held greater support among all unaffiliated New York voters.

For election analysts, the disagreement demonstrates why subgroup estimates require careful interpretation even when different polling organizations identify the same statewide leader.

Election Models Face Methodological Limits

The New York results illustrate several limitations affecting attempts to estimate election outcomes before ballots are counted.

Likely-voter models depend on assumptions about participation. Differences between the expected electorate and the people who ultimately vote can affect polling accuracy.

Nonresponse bias represents another potential source of uncertainty when people with different political preferences participate in surveys at unequal rates.

However, the preliminary New York results do not establish that nonresponse bias caused the observed Democratic underestimation.

Survey weighting can also influence estimates. Polling organizations adjust samples to reflect demographic and political characteristics, but those adjustments depend on assumptions about the electorate.

The independent-voter disagreement between Siena and Marist illustrates a related challenge: differently defined demographic groups can produce results that appear contradictory without measuring identical populations.

National polling comparisons introduce geographic limitations because statewide and nationwide figures cannot independently establish localized voting patterns.

The available figures identify discrepancies between polling estimates and preliminary returns. They do not determine whether participation patterns, sample composition, or changes in voter preferences explain the differences.

Political Trust and Online Debate

New York’s election analysis also raises questions about political information circulating through social media platforms, traditional news organizations, and online communities.

Public trust in these institutions cannot be measured from candidate vote totals. Assessing confidence in social media companies and news organizations requires separate research into public attitudes.

Specific numerical measures of institutional trust and online political-community retention are not established in the reported findings, preventing reliable comparisons across institutions or political groups.

Likewise, the preliminary election results do not establish whether online political discussions affected turnout, candidate preferences, or participation in public surveys.

Polling performance, institutional trust, and online engagement involve different measurements. A discrepancy between forecast and actual vote margins cannot independently demonstrate changes in political confidence or digital behavior.

Final certification will establish the official statewide margins, providing the figures against which the October polling estimates can ultimately be assessed.