Polls

Why are pollsters and forecasters getting it wrong?

Last Thursday at the Oldham and West Royston by-election the results were:-

Labour 17,209   [majority 10,722]
UKIP 6,487
Conservative 2,596
Lib Dem 1.024
Green 249
Monster Raving Looney 141

In May 2015 the general election looked like:-

Labour 23,630   [majority 14,738]
UKIP 8,892
Conservative 8,187
Lib Dem 1,589
Green 839

During the 4-6 weeks running up to 3-Dec-2105 Oldham election it was being reported the 14k Labour majority could be as low as 3k, in the final week some said Labour might lose the seat albeit narrowly, but this was not believed to be accurate, the low majority figure was however.

As it turned out, the forecasters got it wrong.

It also happened in the 2105 general election where pollsters, forecasters and bookies had being saying for literally weeks before the election "it was too close to call", in the end the results speak for themselves:-

Conservative 11,334,576  seats 331
Labour 9,347,304  seats 232
UKIP 3,881,099  seats 1
Liberal 2,415,862  seats 8
Green Party 1,157,613  seats 1

So why are they getting it so wrong now?

There are several possible reasons for this, including people making up [or changing] their minds very late, Labour voters not turning out in the numbers that they told pollsters they would [lying - hardly] not telling pollsters which party they would vote for [this can always happen] and problems assembling accurate data from small selections, polls are usually 1,013 or 1.043 or something similar. I have often wondered if there was not a better solution to small number polls.

If times change and people change with the times, are the questions the pollsters asking the same as they have always asked and are they still relevant today, perhaps the pollsters should up their game by modernising their questions to meet the market?