h1

Is a methodology change behind the CR poll shock?

January 31st, 2007


    Best to wait until we’ve seen how they’ve done the calculations

There’s a poll with, apparently, some sensational figures from Communicate Research in the Independent this morning but judging by Andrew Grice’s accompanying report by and the quoted party shares there is something not quite right.

The voting intention shares with changes on December are CON 34% (-2): LAB 29% (-8) LD 21% (+7). So Ming’s party is up half again on the figure it recorded just five weeks ago.

What makes me suspicious that something has changed is this from Grice; “.To help determine how the “don’t knows” might vote, CommunicateResearch asked people: “Generally speaking, do you think of yourself as Conservative, Labour, Liberal Democrat or other?..The results of the “party identification” question show that Labour is struggling to retain the backing of its natural supporters…”

Grice goes on: “Some 82 per cent of those people who regard themselves as natural Tories intend to support the party in a general election – the highest “core vote” rating of the three main parties.”

This party identifier question was not even asked in the December Communicate poll and now it is being suggested that the responses are central to the calculation.

PBC regulars might recall that I have not been happy with the new monthly CR survey for the paper and after raising some issues was told at the end of November that the methodology was “under constant review“.

My guess, and until I have analysed the full data tables this is only a guess, is that CR are using the new question as a form of past vote weighting to ensure it has a politically balanced sample.

    If that is the case then it is very interesting and I applaud the pollster and the paper for testing new approaches.

I have invited the boss of Communicate, Andrew Hawkins, to comment and he is usually very helpful. Clearly, though you cannot make a valid comparison with previous surveys if there has been a change in approach. Hopefully the full dataset will be available during the day and when it is I will update the article.

Whatever the move to the Lib Dems is entirely in line with the recent ICM and YouGov surveys – so the trend is consistent.

UPDATE 1245. The boss of Communicate, Andrew Hawkins, has been in touch, about the changes in the methodology for this month’s poll. This is what he writes: “Yes there has been a change (two in fact). First, we’ve changed slightly the question wording to incorporate the three main party names in the actual questions (both main and squeeze). Second, we reallocated those who said don’t knows or refused according to their party identification, but also applied turnout weights to these.”

There is little doubt that the main beneficiaries from naming the parties are the Lib Dems. CR have had very small numbers and part of the big increase this month might be due to that change. The second element also partially explains the big change on last month’s poll.

The firm has yet to go all the way and adopt past vote weighting.

Mike Smithson






Comments are closed.