On the excellent Political Betting blog, Mike Smithson makes an important point against what is sometimes regarded as a electoral law. Namely, that:
“There is a notion which seems to be deeply ingrained in the minds of many pundits and Labour figures that there is some immutable law of polling that says that Governments always recover by election day… It might be that the race will narrow as polling day approaches but it certainly has not happened with the two general elections that the Labour government has fought.”
Drawing on data from the 2001 and 2005 election, Smithson argues that the reverse is true – in reality, as the election drew closer, the large lead that Labour enjoyed declined (although in both cases remained large enough to return them to power).
It is certainly sloppy thinking to boldly state that governments always improve their polling position in the run up to the election. However, it is also worth pointing out that this is a case where micro-level data (i.e. opinion polling) might have certain limitations for drawing conclusions. The elections that are used as evidence in this case were abnormal historically, largely because it was very clear which party was going to win them.
The interesting question is how voters shift their allegiance in the run up to an election when faced by the possibility of a change of government. In other words, there could well be macro-level variables at play here to do with the larger political situation. We might theorize, for example, that in a ones-sided contest, voters will feel able to give the government a slap over issues that don’t directly impact their lives (for example, in the case of many people, the Iraq War in 2005). In contrast, a close election might lead to emergence of more conservative (with a small ‘c’) sensibilities, where people who are generally happy with their day-to-day lot revert back to the governing party as decision day approaches.
I am simply working through some ideas here. It would be interesting to look at evidence of this historically, to see how preference patterns change in close versus one-sided election (although to find close elections, we really need to go back to 1992 and then the 70s. This was a period of time when opinion poll technique was far less sophisticated than it is today).
All that said, Smithson’s point about lazy “this always happens” generalizations should certainly stand.