What do opinion polls measure?

The question is prompted by the new Irish Times/MRBI poll (print version) the first since their key pre-election poll which registered a 5 percent surge in support for FF following the leaders’ debate. The new one will make grim reading for FF — or rather it would, if there was an election coming up. It also makes for an interesting contrast with the recent SBP poll, which didn’t find a fall in government support.

The headline is a 9 percent drop in FF support since the election, and a 15 percent drop in Bertie’s approval rating. FG and Labour are both up. Now the trouble with opinion polls is that systematic analysis is difficult, since there’ll usually be enough potential explanations around for any outcome. I’ll admit that I was sceptical about the pre-election poll which showed the FF surge, but it had a strong explanatory factor in its favour — the sense that Bertie had done a good job in the debate — and it correctly tracked the actual election outcome.

Yet some of the more triumphal analysis that the poll and the election result itself led to, notably from Brian Cowen, now seems out of place. Cowen was pushing the line that once FF’s early campaign mis-steps, media hostility, and the Bertiegate leaks were overcome, FF’s natural trust advantage with the electorate returned. But where are those people now with FF sitting at 33 percent?

Maybe the poll is a chance to vent disapproval for people who still might vote FF if there was an election at stake. Maybe there’s still a sense that “Bertie will get us out of it” even as dissatisfaction grows over current circumstances. But a cynic has to ask what exactly has been learned about the competence of Bertie’s Cabinet since the election that wasn’t known before. Maybe there was news about the economy: note Cowen’s reference to a “turning point” in our growth prospects (a usage that he obliquely indicated today he regrets).

Part of the explanation would seem to be that election dynamics are special. Even when voters know a mess is coming, an election seems to bring out their risk-averse nature and the incumbent gets the benefit of the doubt that they can deal with it (or perhaps the rough justice vote for them to clean it up). And a good election machine cashes in on this feeling by pushing the risk of changing horses while in motion, which is what FF did (it’s also what Bush did in 2004; his approval ratings turned down very quickly after he won).

So a good opinion poll rating is a good thing and a bad opinion poll rating is a bad thing. But you don’t win elections based on the former and you don’t lose elections based on the latter. Gordon Brown clearly thought so just last month.

UPDATE: See Slugger for more on the SF performance.