Since I moved to British Columbia last year I have been told many times that the NDP were going to win the 2013 provincial election. Tonight they lost – comprehensively. Almost everyone except the most optimistic Liberal was surprised. But those predicting an NDP win were backed up by polls predicting they should win comfortably. With support at 48 per cent, the NDP entered the campaign 19 per cent clear of their Liberal rivals according to an Ipsos-Reid poll.

During the campaign the polls clearly tightened. By the start of May the Liberals had chopped that lead in half, a new Ipsos Reid poll showed. They were polling at 35 per cent (up six per cent) while the NDP was at 45 per cent (down three per cent), as the Liberals at this stage appeared to benefit from the drop away in support for the BC Conservatives (down three per cent to seven per cent).

The day before polling day, Ipsos-Reid found a further tightening of race, with the NDP on 45 per cent (no change) and the Liberals up two per cent to 37 per cent. This is not to pick on Ipsos-Reid. An Angus-Reid poll conducted in the two days before election day was remarkably similar, putting the NDP on 45 per cent and the Liberals on 36 per cent.

Today, at the polls that actually decide the election, the result was totally reversed. At the time of writing, with the last few votes still being counted, the Liberals are on 44 per cent and the NDP are on 39 per cent. So what went so wrong for the pollsters?

That is something that will be debated at length in the days and weeks to come, but an initial look at the polling data offers some avenues for investigation.

The last Ipsos-Reid poll noted: “The NDP currently has the support of 45% of decided voters in BC.”

Angus-Reid’s final poll said: “Across British Columbia, 45 per cent of decided voters and leaners (unchanged since Friday) would cast a ballot for the BC NDP candidate in their riding in the provincial election.”

The big unknown I cannot see in the headline data provided is how many undecided voters were sampled and excluded? Did those undecided voters break disproportionately for the Liberals?

Another question to consider is did the young people that took part in the surveys actually go out and vote? Traditionally older electors tend to be better at getting down to the polling station and converting their voting intentions into actual votes. The final Angus-Reid poll found voters aged 18 to 34 were intending to vote for the NDP by 54 per cent to 22 per cent for the Liberals, while the same poll found support for the two parties almost evenly split among the over 55 per cents.

The polls also had quite large margins of error (+/- four per cent for Ipsos-Reid’s final poll, +/- 3.5 per cent for Angus-Reid), but it is also perfectly possible that the pollsters were accurately reporting what people were telling them. There can be a tendency for voters to not declare their support for a party if it is not perceived as being the most popular or trendy. And then there is also the possibility that the Liberal campaign won voters, or the NDP turned voters off, during the final day of campaigning.

Clearly something happened and the autopsy is only just beginning.