It is often comforting to look at an opinion poll and see that you are enjoying a lead. Ican help you to relax a little, or to assume that what you are doing is working. These assumptions can sometimes be followed by total shock as the polls are proven wrong and the result is different to that predicted, the campaign team may ask what did I do wrong. While often there are clear reasons, there is one factor that should not be discounted and that is the effect of the poll itself. Here is how it might work.

The best example of a poll that got it wrong was in the Presidential primary contests and the Democrat fight for the New Hampshire nomination. Polls predicted an Obama Landslide but in the end it was Hillary Clinton that just managed to win through. So what may have gone wrong?

1. Polls are context specific
While rolling polls can iron out problems and identify changes in opinion, any poll is related entirely to the moment when the question was asked. The Obama surge may have been linked to his appearance as a winner on the media, so before the New Hampshire campaign got more intense, or there may have been a post-Iowa sympathy surge to Hilary Clinton; either way the results can be outdated and theoretically can change very quickly in response to actual events or media predictions and commentary.

2. Polls effect turnout
Polls can offer an indication of support but they are not predictive of actual behaviour. The disparity between poll and result may have been a reflection of individuals who support Obama, would go to a caucus if their vote mattered, but assumed it did not; so they stay indoors in the warm and so actual votes are lower than support. Alternatively, or equally, the polls may have mobilised the Clinton supporters in the fear that she would lose. Either way, anything but a marginal result can either mobilise or depress turnout; a marginal result, in theory anyway, should just mobilise all those keen on one particular outcome. If you have any doubt on the latter point, why do you think that TV presenters when running a phone poll tell viewers the results could not be closer, they haven't a clue but it encourages voting!

3. Polls effect results
But polls may also make people question their support. Supporting Obama, during that early stage of the campaign, may be different than wanting him to be President. Voters may have considered whether another Democrat (Clinton) had a better chance of winning the White House, or they don't actually want him as President, or they do want Hilary; whatever the various reasons people may have changed their minds between poll and vote.

4. The mystical caucus
While voting can always prove unpredictable, the US caucus system is something different. Here it is about face to face persuasion by campaigners for each of the candidates. It is here that prejudices can be played upon, beliefs eroded, hard arguments can have impact; all the tricks come into play here. Maybe evidence of Hillary's record and experience made some Obama supporters question their allegiance; or maybe Edwards' supporters were more swayed by the Clinton team. Whatever it makes for a very unpredictable outcome.

So for those four reasons, all of which may well be happening at the same time to skew results one way or the other, polls need to be taken with a pinch of salt and a degree of intelligent scepticism until the results actually emerge. If not you get headlines wrong, assume too much and look very daft when the results emerge.