In the run-up to the 2005 general election, there was some excitement as a rolling poll commissioned by ITV from Populus showed the Conservatives slipping to a result even worse than 1997. However, there was a sharp (5 point) recovery in the final pre-election day poll from Populus, which was a normal non-rolling poll.
Picking over the polling after the election, the explanation looked to be one aberrant sample on one day but which, due to the workings of a rolling poll, ‘infected’ the poll results over a longer period of time. It was only with the final pre-election poll, based on a wholly new sample, that this was cleansed from the figures.
That was a reminder of the perils of rolling polls – one unluckily atypical sample hangs around.
Roll the clock forward then to this week and we’ve had two consecutive days of rolling poll results from ComRes which are significantly out of line with all the other polling being conducted (placing the Conservatives in a clear lead). But once again those figures look to be down to just one day’s sample which has hung around (a point Anthony Wells has also made here).
And the commissioner of the rolling poll this time? The Independent in partnership with … ITV once again.
2 Comments
TONIGHT’S DEBATE
Speaking as a journalist, some of the bile being hurled Nick Clegg’s way in the run-up to tonight’s TV debate is beyond belief.
Nazi slurs, ‘expense’ smears.. one paper even suggests people shouldn’t vote for him because he doesn’t watch the soaps!
That said, I wonder if last week’s TV prime ministerial debate [being the first] will be the most significant, not only in terms of viewing figures, but in its effect on the campaign due to the sheer size of the TV audience.
If so, the fact that NC did so well in that first debate might compensate for the fact that the next two debates might see less decisive a result.
YM
That’s more a demonstration of dodgily applied statistics. There are ways to summarise and collate data that are less sensitive to aberrant polls like in that case, but they aren’t being used.