Did the polling companies skew their results ?
I usually operate on the basis that nine times out of ten a "cock-up theory" will be closer to the truth than a "conspiracy theory." As the saying sometimes called "Hanlon's Razor" puts it,
"Never attribute to malice what can be adequately explained by incompetence."
Yesterday in his Telegraph blog Dan Hodges accused the polling companies of lying, saying that they "clustered" at the end of the campaign.
As he quite rightly says, the question after the election was "why were the polls so wrong?" but during the campaign itself the question we were asking was "why are the polls all over the place?"
His answer is that towards the end of the campaign they deliberately made sure they were producing similar answers as close as possible to each other and to the "too close to call" space so that each company would minimise the risk of looking uniquely incompetent if that position were wrong.
I don't know that I agree with the word "lying" as they may have genuinely been in doubt about why their numbers were at variance with each other. For example, the polling company who said just after the election that they had declined to publish a poll on the eve of the election which subsequently turned out to be almost exactly right, because they thought it was an "outlier," may have been telling the truth about their motives.
However, it would have been better had they been more open about the degree of doubt they felt in that case. And one lesson of the election should be that this kind of suppression or adjustment of data may be a very bad idea.
Whether or not you think there was deliberate cheating involved, it has to be said that looking at the range of outcomes predicted in the early stages of the campaign, and how they converged at the end, Dan makes an extremely strong argument.
The polling data appears to have been, shall we say, "adjusted" in a way which may or may not have been deliberately dishonest but was certainly unwise.
"Never attribute to malice what can be adequately explained by incompetence."
Yesterday in his Telegraph blog Dan Hodges accused the polling companies of lying, saying that they "clustered" at the end of the campaign.
As he quite rightly says, the question after the election was "why were the polls so wrong?" but during the campaign itself the question we were asking was "why are the polls all over the place?"
His answer is that towards the end of the campaign they deliberately made sure they were producing similar answers as close as possible to each other and to the "too close to call" space so that each company would minimise the risk of looking uniquely incompetent if that position were wrong.
I don't know that I agree with the word "lying" as they may have genuinely been in doubt about why their numbers were at variance with each other. For example, the polling company who said just after the election that they had declined to publish a poll on the eve of the election which subsequently turned out to be almost exactly right, because they thought it was an "outlier," may have been telling the truth about their motives.
However, it would have been better had they been more open about the degree of doubt they felt in that case. And one lesson of the election should be that this kind of suppression or adjustment of data may be a very bad idea.
Whether or not you think there was deliberate cheating involved, it has to be said that looking at the range of outcomes predicted in the early stages of the campaign, and how they converged at the end, Dan makes an extremely strong argument.
The polling data appears to have been, shall we say, "adjusted" in a way which may or may not have been deliberately dishonest but was certainly unwise.
Comments