As we approach the closest election in years, I have been repeatedly asked two questions about political polls.

One is how polls taken at about the same time can contradict each other and the second is how the growth of cellphone use impacts the results of these polls.

Hawke's Bay Regional Councillor Peter Beaven and I successfully launched telephone based market research in New Zealand in the 1980s and the company we began; (originally called Insight Market Research) still exists as UMR, a successful polling company which does work for the Labour Party.

In those days political polling was a simpler proposition. Electors had only one vote for a candidate in the electorate in which they lived. Parties were elected by winning more electorates than any other.


Most electorates were "safe" for one party so the focus was on the electorates which were "marginal" and would decide the outcome of a general election.

As a key marginal seat, Hastings (now absorbed into the larger MMP seat of Tukituki) was more politically important then than it is now.

With MMP, electors have two votes, one for a candidate and one for a party. It is the party vote which determines the proportion of seats allocated to any given party so nationwide party vote polls have almost completely supplanted the marginal seat polling that we pioneered all those years ago.

It is not only the electoral system that has changed. When telephone polling began, cellular phones didn't exist and virtually every voter could be reached via a landline that was listed in a printed phone book.

Despite these dramatic changes, polling in New Zealand has generally been able to predict election outcomes with a moderate degree of accuracy and politicians watch them with great interest as we saw in the lead-up to the resignations of Metiria Turei, Andrew Little and Peter Dunne.

To explain how polls conducted at the roughly the same time produce differing results, I often recount an experience I had in 1999.

In that year Auckland University got funding to undertake a tracking poll in the lead-up to the election.

This poll repeatedly showed a landslide to Labour while all other polls at the time were predicting a close result.

I took an interest in why this poll was so positive for Labour when compared with the others and got a copy of its questionnaire. When I compared it with the one used by the Labour Party's own pollsters I found only one difference.

The Auckland University questionnaire included an offer of going in a draw for $250 worth of groceries if the interview was completed. This reduced the poll's refusal rate and captured more Labour voters.

With this incentive removed, the Auckland University poll fell into line with all of the others.

This anecdote demonstrates just how hair-trigger sensitive polls can be and why all should be treated with suspicion.

Contradictory polls, as we saw last week, often simply demonstrate very high levels of volatility in an electorate and it's perfectly possible given the recent changes in the leadership of three of the parliamentary parties that a lot of voters are swaying between a vote for Labour and one for National.

The move from landlines to cellphones has proven difficult for pollsters as many voters, especially younger people, simply cannot be contacted via a landline. It's hard to find someone under 40 who bothers with a landline these days.

Pollsters handle this serious communication blockage in a number of ways. One method, used by TV3's Reid Research is to capture an internet panel and use this to augment the sample of young people in poll.

This is not a great idea.

On-line panels are self-selecting and are never truly random sample of voters. They obviously don't include poor people who can't afford an on-line presence but whose votes are just as good as yours and mine.

It is likely that these most difficult to poll youngsters will decide the outcome of the general election next week should they decide to get involved.

This week's TVNZ poll indicated that 67% of voters aged between 18 and 35 are planning to vote Labour which would normally mean the end of the National government, however there's a risk that this group just won't enrol and vote.

The most recent enrolment update from our Electoral Commission tells us that close to a quarter of these voters are yet to appear on the electoral roll even though there is now the option of enrolling and casting a special vote in one go.

The leader's debates on TVNZ and TV3 have drawn some of the biggest audiences ever recorded in this country so it remains possible that many of these latecomers will still get involved and participate.