Why so much variation over the Lib Dem number?
Whenever I go anywhere, like yesterday during a very enjoyable visit to Westminster, virtually everybody I speak to asks me the same question – “Which pollster do you think is best?”.
My answer, as always, is ICM because the record since the mid-90s shows that the firm’s methodology is more likely to pick up the most accurate Lib Dem share. This matters because it has a big impact on the Labour share. The reason YouGov is currently showing the smallest Tory lead is because in my view it is understating the Lib Dems.
In a previous article I argued that the wording of the ICM question is more likely to tease out a higher Lib Dem figure than all the other firms.
Anthony Wells of UK Polling Report sums it up well: “A more significant difference is in ICMâ€™s wording which specifies â€œin your areaâ€. This is very interesting since we know that people give very different answers if they are asked to say specifically how they would vote in their own constituency – people suddenly start taking into account tactical voting or what they think of their MP, and this normally increases Lib Dem support.”
Underlining the importance of the wording point both ComRes and Populus usually use ICM to do the fieldwork for their telephone polls. So it’s possible that in the same month the same interviewers can be working for all three firms – only the questions are worded differently.
In addition, of course, a lot of the variation between the pollsters is down to the weightings. Thus YouGov applies a “party identifier” formula which explains why its Lib Dem figures are generally lower. This is: CON 34.7% LAB 45.8% LD 15.3%.
These numbers seem odd given that that at the 2005 general election for every two people who voted for the Lib Dems just over three supported Labour. The online pollster is weighting at a ratio of 1 to 3.
If I was advising Nick Clegg then I would say that the poll to take most notice of is ICM and if you want reassurance just check out the record from the firmâ€™s numbers over a couple of decades here. No other pollster can point to such an historical performance.