Push polling
There wasn't actually much more of interest in the ICM poll in the Sunday Telegraph. What did attract my attention is the allegation from the Boris Johnson campaign that he had been the victim of push polling. Now, I'm in no position to say whether that's true or not, and there wasn't much detail in the news reports, but it's a chance to have a look at what push polling is and isn't.
In this country a lot of people use the term push poll to refer to a poll with skewed questions aimed at getting the answers the client wants. In fact, a push poll as they occassionally surface in the USA is something much nastier than that.
A push poll isn't a poll at all, it isn't an attempt to measure public opinion or even to produce a fake skewed measure of public opinion. It is a way of spreading malicious rumours about an opponent under the guise of an opinion poll. It's relationship to actual polls is the same as the relationship of a chap in a British Gas uniform nicking jewelry from OAPs's relationship to a genuine British Gas meter reader.
Push polls rely on the psychological quirk that people tend not to question the assumptions within a question they are asked. In other words, if someone recieved a phone call that told them a candidate was corrupt or had secret plans to cut services or was having an affair or whatever they would be suspicious about it. In contrast, if they were phoned up by someone they thought was an opinion pollster and asked questions about their reactions to the fact that the candidate was corrupt, had secret plans to cut services and was having an affair they would be far more likely to accept the allegations at face value.
Because a push poll isn't intended to measure public opinion, it doesn't go out to a limited representative sample. Its purpose is to spread a negative rumour as widely as possible, so the phone call will go out to tens of thousands of people - this is one of the give away signs of a push poll, because they don't need a representative sample and just want to do as many phone calls in the shortest time possible, they don't tend to bother with all the demographic questions proper pollsters weight by (it should go without saying, no reputable polling company would ever be associated with actual push polling). Rather than being 20 minutes long, they will be just a couple of minutes long.
The classic example of a push poll was one used against John McCain in the Presidential primaries in 2000, when lots of people received a call asking how they would react if they knew John McCain had fathered an illegitimate black child (he hasn't!). These days people seem to be accused of push polling far more frequently than there is any evidence of it happening, probably because people are aware of them and they tend to get spotted and backfire. What tends to be mistaken for push polling is message testing - these are genuine polls, to a genuine representative sample of the public, that contain negative messages about a candidate. They aren't trying to spread a malicious rumour, they are testing out those negative messages before they are used in public to see if they work or if they backfire. If a party spreads negative rumours about a candidate openly that isn't push polling either - it might not be very nice, but push polling is all about pretending to be an independent polling company when you do it.
