Reply to post:

Lies, damn lies and election polls: Why GE2015 pundits fluffed the numbers so badly

tom dial Silver badge

Extremely careful and much costlier pollsters (mostly academic) ask each respondent a number of questions, some and occasionally all of them different from "for whom do you plan to vote?" From the answers, and a theoretical model they then use to make predictions (although the primary goal is to improve the behavioral model and better understand the sources of voting and other political behavior. Along with this they also use a process (somewhat disrespectfully called "farmerizing" when I was in graduate school long ago) to word the questions for clear comprehension by those with no more than about an eighth grade (US) education level. Polls like that are less likely to be gamed (we thought at the time) and might be more likely to give valid predictions The procedure also may allow them to detect and better discount responses that are untrue or ambiguous. It has been noted, however, that projecting the outcome based solely on responses of a sample to the main question is not much less accurate (if it actually is), especially if adjusted based on even quite simple demographic models.

However, with legislatures selected from single member districts with a plurality requirement for election, a national sample is very likely to be inadequate in some cases, including what seems to have occurred in the UK last week, where a narrowly based regional party - the SNP - did major damage to a national party - Labor - that depended a lot on the region. In such circumstances, polling organizations probably would do better by using district polls to project the outcome in each district, and combine those results to project the national legislature composition. This article and others I have seen do not suggest that they did this, perhaps because of the associated cost.

Single member plurality-win districts have a strong tendency to suppress national parties such as the Liberal Democrats and UKIP with relatively small memberships, while allowing strong regional parties like the SNP and sometimes relatively strong national parties like the Conservatives or Labor to elect disproportionate numbers. That may be good or bad, and depend on ones party allegiance and the details.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon