Research · When the polls missed

The young voters who stayed home: British Columbia's 2013 polling miss

Abstract. Every pollster had the NDP ahead, by 7 points on average; Christy Clark's Liberals won 44.1% to 39.7%. The firms' own reviews found they had weighted young, NDP-leaning voters to their share of the population rather than of the people who vote, and missed a late Liberal turnout edge.

Polling Forecast · 2 November 2026 · 6 min read · Canada, 2013 British Columbia provincial election, polling error, final polls, inquiry, pollster methods

Every election the polls got badly wrong →

Summary

Every pollster in the 2013 British Columbia campaign had Adrian Dix's NDP ahead, and the final polls averaged NDP 42.9, Liberals 35.9: a 7-point lead. On 14 May Christy Clark's Liberals won 44.1% to 39.7% and took 49 of 85 seats. The error on the lead was 11.5 points, a year after Alberta's miss. The pollsters' main finding was turnout: their samples gave young voters, who leaned NDP and mostly did not vote, about twice the weight they had at the ballot box.

What the polls said

The last poll from each firm with fieldwork ending within ten days of polling day:

Pollster	Fieldwork end	Mode	Liberal	NDP	Green	Conservative
Ipsos Reid	13 May	Online	37	45	9	6
Angus Reid	13 May	Online	36	45	9	7
EKOS	12 May	IVR	34.5	40.5	13.0	9.3
Justason	9 May	Online	31	45	14	8
Forum Research	8 May	IVR	41	43	8	6
Hill & Knowlton	8 May	Online	34.6	41.1	13.6	7.5
Oraclepoll	7 May	Live phone	37	41	12	10
Average			35.9	42.9	11.2	7.7

The NDP's lead had been larger. Angus Reid's first poll of the campaign had it 17 points ahead, and Abacus's late-April poll had it 10 points up. The leads narrowed over the last fortnight but every firm still had the NDP in front at the end. Forum was closest at 2 points, Justason furthest at 14.

What happened

Elections BC's Statement of Votes: Liberal 44.13%, NDP 39.72%, Green 8.15%, Conservative 4.75%, independents 2.36%. Seats: Liberal 49, NDP 34, Green 1, independent 1. Turnout was 55.32% of estimated eligible voters (57.10% of registered voters).

Against the average, the Liberals were 8.3 points too low and the NDP 3.2 too high. The Greens were 3.1 points too high and the Conservatives 2.9 too high. The polls overstated every party except the one that won.

Why it happened: what the pollsters found

There was no formal industry inquiry. The firms published their own reviews.

Ipsos Reid interviewed about 1,400 voters on polling day. 11% said they had decided in the booth. John Wright said the NDP vote did not show up on the day. According to Coletto and Breguet's later review, the exit survey matched the result once it was weighted by age to Elections BC's turnout estimates, and Ipsos concluded that turnout modelling caused its pre-election miss.
Angus Reid, writing in Maclean's about two months after the vote, put his firm's miss at 13 points and the industry average at about 12. Under-35s were about 30% of British Columbia's population, but Elections BC's research on 2009 put them at closer to 15% of voters. His polls weighted young people to their population share. That one change, he said, would have cut the final NDP lead to about three points. A follow-up found no evidence of last-minute switching. He also described a late surge in Liberal turnout in the Interior and in multicultural ridings in Metro Vancouver.
EKOS (Frank Graves, "An Unapologetic Analysis of the BC Polling Debacle", 29 May 2013) said its final poll was an unweighted tie and had the NDP share right but not the Liberal one. Graves blamed an incumbent "risk aversion premium" that pollsters cannot see, and a large Liberal turnout advantage. EKOS's likely-voter questions had no predictive value. A re-poll with the same method two days after the vote "captured the election result", which he took as evidence that sampling was not the problem. He saw a steady drift away from the NDP rather than a sudden late shift.
Abacus Data (David Coletto) called the miss "a wake-up call". Reweighting its mid-campaign poll (NDP 43, Liberal 33) by age to Elections BC turnout data cut the lead to about three points.
Barb Justason said online panels favour younger voters: "We're dealing in a world where we cannot produce a probability sample." The Liberals' own pollster, Dimitri Pantazopoulos, said the online polls had failed.

Coletto and Breguet's 2015 study of nine provincial elections drew this together: pollsters weighted under-35s to about 30% of their samples when they were about 15% of 2009's voters, while over-60s were under-represented.

What our data shows

Our reading of the published figures:

The miss was mostly the incumbent's share. The Liberals were understated by 8.3 points, more than twice the NDP's overstatement of 3.2. In Alberta a year earlier the governing PCs were 10.0 points too low. Both times the polls caught the challenger's support more accurately than the government's.
Six points went to small parties that did not deliver. The Greens and Conservatives together were overstated by 6.0 points. In our reading that is where a surplus of young, irregular voters in the samples showed up.
Age explains about half. Angus Reid's and Abacus's reweighting each took 6 to 7 points off the NDP lead. Applied to the average, that leaves the parties about level, so age accounts for roughly half of the 11.5-point miss. The other half is the Liberals' 4.4-point win: an incumbent premium, late deciders and a Liberal turnout edge that weighting by age alone does not catch.
Online polls missed by more. The four online final polls had the NDP ahead by 9.4 points on average. The two IVR polls averaged 4 and the one live phone poll had 4. These seven polls are too few to settle it, and Coletto and Breguet found no clear online penalty across provinces once other factors were controlled.
The NDP overstatement is familiar. At Canada's federal elections the NDP has been overstated every time since 2000, by 1.50 points on average (Canada polling error paper). The largest federal error on the gap between the top two parties since 2000 is 5.8 points. British Columbia's was twice that.

What the pollsters changed

Angus Reid proposed three changes: weight younger voters to their share of actual voters, publish estimates for different turnout levels, and sample British Columbia's regions in more detail.
EKOS said it would invest in better likely-voter tools and called for an industry and media consortium to provide a benchmark during campaigns. It ran no media polls in British Columbia in 2017 and polled for the Green Party instead.
Ipsos Reid's John Wright said the firm was satisfied with its methodology. Its own exit survey pointed to turnout weighting by age as the fix.
Publishing. The Globe and Mail reported in 2017 that some firms would publish fewer horse-race numbers. Mainstreet Research, which polled privately in 2013, forecast a Liberal win and did not publish it, ran a weekly tracker with Postmedia.

The 2017 election was the test. The result was Liberal 40.9, NDP 39.9, Green 16.7. Five pollsters' final-week figures had the NDP at 40 or 41 and the Liberals between 39 and 41. Insights West, at 41 each and Greens 17, was closest. Éric Grenier's headline for CBC was "Pollsters vindicated in B.C. election after 2013 miss."

What it means for our forecast

Our Canadian correction is learned from federal elections. Half of each party's average past miss comes off today's polls, and the rule is kept because it beat the plain average at seven of nine elections when each was predicted without knowing its result (Canada polling error paper). That correction already lowers the NDP, which leans the same way British Columbia's polls did. Provincial misses are not in the federal record, and one province's single election cannot pass the test alone.

For provincial forecasts the simulations draw a province-wide error for each of the two main parties with a standard deviation of 4 points by default, plus regional and district noise. An 8.3-point miss on the governing party is about two standard deviations, so it sits inside the simulated range. Firm-level records are on our Canadian pollster ratings.

Sources

How to cite

Polling Forecast (2026). The young voters who stayed home: British Columbia's 2013 polling miss. Retrieved 2 November 2026 from https://www.pollingforecast.com/research/missed-ca-2013-bc. Live figures change as new polls arrive; the date above is when the text was last revised.