Cross-posted from The Wardman Wire:
The 1992 polling debacle
The 1992 general election was a bad one for the British political polling industry. During the campaign, the vast majority of polls put Labour ahead and of the final round of polls three put Labour ahead, one put Labour and the Conservatives neck-and-neck and only one – Gallup – gave the Conservatives a lead, but even that was a mere 0.5%. The actual result? A Conservative lead of 7.6%.
A declining number of opinion polls
The response of the polling industry was a series of post-mortems and experiments with changes in methodology. Amongst those who commissioned polls, though, the response was also one of greater scepticism of the value of commissioning polls. Add in first the economic pressures of the 1990s and then the widespread seemingly inevitability of a Labour general election victory after Tony Blair become Labour leader, and it is no surprise that during the 1992-97 Parliament the number of opinion polls was consistently lower than in 1987-1992.
Individual periods of political excitement and then the approach of the general election did result in burst of extra polls, but consistently the number of polls conducted ran at a lower level than in the previous Parliament.
In the next two Parliaments – 1997-2001 and 2001-5 – there was very little sign of the number of opinion polls recovering. Perhaps no surprise again as in the polls that were conducted Labour held a large lead, with only the occasional brief periods of exception. With one party consistently largely ahead, the interest in individual poll results was understandably muted except at those moments such as the petrol price crisis of autumn 2000, when the Conservatives very briefly went into the lead in the polls.
Number of opinion polls recovers
Politics since 2005 has, however, been far from consistent and predictable and, indeed, the number of polls commissioned has picked up once again, as can be seen from the graph*. Across 2008, the number of polls was only 10 less than in 1990.
Although only two firms were doing political polling in both 1990 and 2008 (ICM and MORI), the overall number of firms has changed little. Five firms polling in 1990 were no longer political polling by 2008 (ASL, Gallup, Harris, NMR and NOP), but this is nearly balanced out by the four firms which were not polling in 1990 but were doing so in 2008 (BPIX, ComRes, Populus and YouGov).
Is the increasing number of political opinion polls a good thing?
Generally, the more data you have, the better, but this comes with caveats. Perhaps the most important is that thanks to the vagaries of sampling, around 1 in 20 opinion polls can be expected to be a “rogue” poll, that is with the support for a party off by more than 3%. (This is leaving aside the question of whether polls have any systematic error.)
In 2008, that would have meant around 6 polls being seriously off. But think how it is almost unheard of for the media coverage of an opinion polls to be, “Well, you know this one looks completely wrong. So don’t really pay any attention until another comes along to give us a clearer idea.” Instead, each poll is pretty much treated by the media as if it must be right.
In practice, therefore, more polls means rogue polls, which in turn means more political reporting that in fact is based on nonsense. Which isn’t so good.
It’s true that it also means more opportunities for stories based on non-rogue polls, but by their very nature (results out of line with previous polls) rogue polls are more likely to generate widespread media coverage than an unexceptional poll. The media’s collective reaction is usually, “ooh, exciting poll – let’s report it” rather than “ooh, dodgy looking poll – let’s wait and see”.
It’s also debatable whether the heavy emphasis in political reporting on the relatively popularity of the parties, rather than what they’re actually doing or saying, is to the good. Certainly it would be an odd form of political reporting that ignored popularity completely, but you can have too much of a good thing. (One example: at the time of writing this piece I tried searching on The Times website for stories with opinion poll gordon brown and with credit crunch gordon brown. The former exceeded the latter by 456 – 445.)
As for what 2009 will bring, continued general election speculation is likely to encourage more polls, whilst financial difficulties on the part of those who commission polls will depress the number. The overall impact? Time will tell…
* The graph shows for each quarter the total number of opinion polls published which included general election voting intentions for Great Britain. Private polls where the voting intention results were subsequently published are included. Exit polls are excluded. Polls are allocated to quarters depending on when the last day of their fieldwork took place. Where some fieldwork was shared between two different polls, the polls are counted separately as long as at least one of them had some fieldwork unique to itself.