In the weeks before the last General Election a remarkable techicial innovation took place with almost no publicity. Several thousand households were telephoned by a computer and whoever answered was asked questions about their voting intention by a computer-generated voice to which they responded by using the phone’s keypad. Within hours the UK’s first ever completely automated opinion poll results were published.
The opinion poll establishment, if there is such a thing, was horrified. The BBC ruled that this new methodology was “not approved” and ordered that no mention of it or the results should on any BBC outlet throughout the whole campaign. This turned out to be the BBC’s and the public’s loss.
For by using a computer rather than a live person to ask the questions the polls, by Rasmussen, for the Independent, put the Tories at 33% – the actual result was 32.7%.
The expereince is that many Tories are reluctant to tell a stranger their political views, or are inclined to give what they think is the “correct” answer rather than what they actually believe. But the Rasmussen experiment demonstrated that reluctant interviewees are quite happy to declare their intentions to a remote machine – but not to a live interviewer.
There is a history of polls underestimating the Tories. In 1992 John Major’s Tories were returned to power after achieving an 8% lead over Labour in the popular vote even though all but one of the pollsters were forecasting Labour leads. Even the BBC’s exit poll had it wrong. In the London Mayoral race in 2000 the average polling figure for the Tory candidate was 13% – less than half the 27% that was achieved.
One measure conventional polls use to correct potential bias is to ask which way respondents voted last time and to weight the overall result accordingly. Thus, in broad terms, if the Tories got 30% in the actual election and only 20% of those you interview say they voted for the party then you adjust your survey figure for the party by upto a half.
The use of the internet by the polling firm YouGov has produced higher Tory percentages because, like the automated phone calls, it’s thought that people are less likely to lie to a computer than to a real interviewer. But only about half the population have access to the internet and this, it is argued, will skew your result.
Will there be more polls such as the automated ones? Will the BBC changed their view? Political gamblers have a big interest in this because the polls are the main way that we can find out how elections are going.
With the next General Election likely to be much tighter than the last interpreting opinion polls correctly could be the secret of success.