In Defence of Polling
Dominic Lawson has today penned another of the staple run up to an election campaign articles - the "should polls be banned" job. To start with, I should probably correct all the inaccuracies. Firstly, pollsters do not make much money out of election campaigns (though to declare my obvious interest, it does pay my wages!). While we are very, very busy, it's the sort of polling that has the lowest margins and brings in the least cash. If Bob Worcester does own an estate on Mustique then I expect he got it from all the commerical and public sector work MORI has done over the decades. He sure didn't buy it from the proceeds of political polls. Political polling for newspapers is a tiny fraction of polling companies' income, and is normally done at knock-down, rock-bottom prices. The reason it is done is not to make money, it is the pollster's shop window. It lets them get publicity and show off their accuracy.
Secondly Lawson questions the accuracy of polls. Yesterday I posted a list of ignorant comments about polls that would invariably pop up during the campaign. Number 8 was "Polls always get it wrong" and it's gratifying to see it turn up so soon. Lawson correctly says that the pollsters got it wrong in 1992, and didn't do that much better in 1997. However, the lack of criticism since 1997 is not because the Labour victories continued to obscure poor performance, it is because the pollsters have stopped being wrong. Most got it pretty right in 2001 (and those that didn't reformed their methods or stopped polling), and the industry got it almost exactly right in 2005. In 2005 all the pollsters were within one or two percent of the right result.
Neither are the adjustments since 1992/1997 additions to help the Conservatives which may not work in other circumstances. The naive characterisation of ICM and Populus's spiral of silence adjustment is that it is just a matter of adding a couple of points to the Conservatives. It is not - ICM and Populus's methodology is not so crude and nor it is blindly "pro-Tory". It is just as capable of accounting for "Bashful Brownites" and the reality is that for at least 6 years it has favoured Labour, not the Conservatives.
Turning to Lawson's main point, do polls influence people's vote? Personally, I think it is very likely they do. Whether that is either avoidable or necessarily a bad thing are different questions. It is inevitable that people are influenced by others. They are influenced by polls, they are influenced by their friends and families, they are influenced by political parties leaflets and door knockers... and for that matter, they are influenced by political columnists. Perhaps we should ban them.
Ask yourself, what would happen if there weren't any polls during the campaign?
Some countries have a ban on polling shortly before elections, Lawson suggests newspapers just not commissioning any. In one sense, this would make very little difference - as I said earlier, pollsters do it for the publicity not the cash, so if newspapers didn't commission opinion polls in the run up to elections, many polling companies would do them anyway. When MORI lost their contract with the Times they continued doing political polling and publishing it themselves. TNS are currently doing political polls with no client. Angus Reid are giving their polls to PoliticalBetting for free to try and get established in the UK market. In an internet age, polls are going to happen and are going to reach the public whether newspapers buy them or not. Last month 150,000 different people visited this site, hardly any of them via mainstream media websites.
Imagine the alternative then, what if there was a legal ban on publishing opinion polls during the election campaign? Firstly, we can still be certain that the polls would take place. There are plenty of people who would still pay for polls even if they couldn't be published - the political parties, bookmakers, bankers. Political journalists wouldn't miss out either, the political parties would all make sure the journalists knew what the polls said.
Already you tend to get columnists and lobby journalists reporting rumours and mumbling about what the party's private polls say. I always advise people to ignore these figures completely - we can't see the actual tables, so readers can't judge for themselves if they are genuine figures or empty spin. In the Lawson scenario, this would still happen - we'd see a nudge and a wink in the newspaper columns as to what the parties private polls say. The only difference would be that the general public would no longer have access to any other data to judge whether it was true or not.
The vacuum of polls would not be filled by staid discussion of policies. People would still want to predict, and the gap would be filled by various inferior attempts to judge attitudes. Unscientific vox pops, rumours and spin from the political parties, BBC have your say comments, bookies odds, the gimmicky news stories you always get about some shop selling red, blue and yellow beer and so on. I am sure the commentariat would still come to their own censensus about the political situation (albeit a less well educated one), would spend just as long speculating about it, and it would probably be just as influential to the public.
Besides, wouldn't it be rather perverse to ban opinion polls during an election campaign? Short of elections and referenda (which are a very crude tool that records only the answer to a single question) the only legitimate measure of what the public think is the opinion poll. Imperfect as they are, opinion polls are the closest we come to finding out the electorate think about education policy, or what should be done on the economy, whether this or that should be banned, etc, etc. The general election is the core of our democratic system - it would be somewhat odd if we allowed political parties, commentators and columnists to pontificate, but froze out our best way of representing the voice of the general public.
