Why a poll is not an election prediction

The short answer

a poll is not an election prediction

A poll estimates opinion under a method and time window; an election records votes under an electoral system. Campaign events, turnout, sampling error and seat geography separate the two.

This guide helps verify public political material. It does not infer a person’s motive from one post and does not treat disagreement as proof of deception.

The useful question is not only “what is the rule?” but also “who administers it, which document controls it, and when might it change?” That distinction prevents an accurate general explanation from becoming wrong advice in a particular election, chamber or policy setting.

Evidence review

What a political poll actually measures

A political opinion poll is a survey of a sample of people designed to estimate how the broader population would answer the same questions at that moment. Polls do not predict election results. They measure stated voting intention, leader approval, issue salience or other attitudes at the time the survey was conducted. Polls are subject to sampling error: a poll of one thousand people with random sampling has a margin of error of approximately plus or minus three percentage points at the 95 per cent confidence level. This means the true population figure is likely to be within three points of the poll figure nineteen times out of twenty, and outside that range about one time in twenty. The margin of error applies to each party's estimate separately, so a gap between two parties carries a larger error range. These are statistical constraints that apply to every properly conducted poll, not defects in any particular pollster's methodology.

Another limitation is that polls measure voting intention at the national or state level but Australian elections are decided seat by seat. A poll showing a party at 52 per cent of the two-party-preferred vote nationally may conceal that the party is polling 60 per cent in safe seats it already holds and 48 per cent in marginal seats it needs to win. The geographic distribution of support matters more than the national aggregate, and polls generally lack the sample size at the electorate level to provide reliable seat-level estimates. Party strategists use polling data to target resources at marginal seats because they understand that the national number can be misleading. Voters should apply the same caution when reading national polls as predictors of the election outcome.

Evidence review

Why polls cannot translate directly into seats

Australia uses single-member electorates with preferential voting to elect the House of Representatives. A national two-party-preferred poll of fifty-two to forty-eight does not mean the leading party will win the election. The national vote is distributed unevenly across 150 electorates. A party can win the national two-party-preferred vote and still lose the election if its votes are concentrated in seats it already holds by large margins while the other party's votes are more efficiently distributed across marginal seats. The electoral pendulum published by the Australian Electoral Commission lists every seat by its margin, showing how many seats would change hands on a given uniform swing. But uniform swing is itself a simplification: real elections produce different swings in different states, regions and seat types. Polls cannot predict these variations.

Evidence review

Campaign events, undecided voters and turnout

A poll taken weeks before election day captures opinion that may shift by polling day. Campaign events such as debates, policy announcements, candidate controversies, endorsements and media coverage can change voter preferences. Undecided voters, typically around five to fifteen per cent of the electorate depending on the poll, are allocated or excluded according to each pollster's methodology, and their eventual decision can differ from any allocation. Turnout is another variable. Compulsory voting in Australia means turnout is typically above ninety per cent, so polls do not face the turnout-modelling challenge that polls in voluntary-voting systems face. Even with compulsory voting, however, who actually votes, including informal voters and those who fail to vote, can have marginal effects on close seats. Polls also cannot measure the effect of preference deals between parties, which can alter the flow of preferences in a way that changes the outcome in marginal seats without any shift in first-preference polling.

Evidence review

How to read a poll responsibly

When you encounter a political poll, check several things before drawing conclusions. Identify the polling organisation, the client who commissioned the poll and the date or date range when the survey was conducted. Check the sample size and the stated margin of error. Note whether the figures are first-preference votes or an estimated two-party-preferred result; the two-party-preferred figure involves modelling assumptions about minor-party preference flows. Note whether the poll is a national poll or a state-level or seat-level poll; national polls say nothing about individual seats. Compare the result with previous polls by the same organisation using the same methodology, not with a different pollster's numbers. Individual polls contain noise. Look for trends across multiple polls over time rather than reacting to a single poll.

Polling methodology varies between organisations in ways that affect the headline result. Some pollsters use live telephone interviewing, others use online panels, and still others use a mix of methods. Each approach has different challenges in reaching a representative sample. Online panels may over-represent politically engaged respondents. Telephone polls may under-represent younger voters who do not answer unknown numbers. The treatment of undecided voters also varies: some pollsters allocate undecided respondents proportionally to stated preferences, others exclude them from the headline figure, and some report undecided separately. These methodological choices are not right or wrong in the abstract, but they produce different numbers from the same underlying opinions. A poll consumer who does not know the methodology cannot properly interpret the result.

Evidence review

Official sources to consult instead of polls

The Australian Electoral Commission publishes authoritative information on past election results, electoral division margins and voting patterns at aec.gov.au. The AEC's Stop and Consider campaign encourages voters to question the source of electoral information they encounter rather than accepting it at face value. The Australian Communications and Media Authority regulates broadcasting content and provides guidance on media standards. Neither the AEC nor the ACMA endorses or rates opinion polls, but the AEC's electoral data provides the factual record against which poll-based claims can be checked after an election. The Parliamentary Library also publishes research on election results and voting patterns that provides context for interpreting polls. Past election results are facts. Polls are estimates. The distinction is important.

Common questions

Before you rely on the answer