YouGov poll
21-22 February 2011
|
|
Constituency |
Regional |
|||||||
|
Lead |
SNP % |
Lab % |
Con % |
LD % |
SNP % |
Lab % |
Con % |
LD % |
Oth % |
Weighted |
Lab +9 |
32 |
41 |
15 |
8 |
26 |
40 |
15 |
7 |
12 |
Unweighted |
SNP +13 |
41 |
28 |
18 |
6 |
34 |
26 |
19 |
6 |
16 |
Amongst actual respondents to the YouGov poll SNP were ahead by 13%, but weighting changed this to Labour being 9% ahead.
This is not the first time. The unweighted numbers of the last Scottish YouGov poll, in October, showed a 9% lead for the SNP over Labour. Weighting resulted in this moving to a 6% lead for Labour. I noted at the time that this was in the expected direction but “unusually large”. This is even larger.
Weighting is necessary. Respondents to a poll never resemble the general public. In this case, they were more likely to be male, middle class, and over 24. Adjustments to compensate for each of these would have helped Labour somewhat, but nowhere near enough to bridge the gap in question.
YouGov also weight voters by political party identification and newspaper readership. They believe that as well as matching demographics this is important to ensure the attitudes of their sample match the attitudes of the population (or electorate) as a whole. Other pollsters attempt to do this too, but by different means – such as asking how the person voted at the last election. Based on our latest information, the party ID profile they use for Scotland is:
|
SNP % |
Lab % |
Con % |
LD % |
Oth % |
None % |
Party ID |
16 |
38 |
13 |
10 |
2 |
21 |
This gives Labour a built-in headstart in any voting intention question. We can only pull ahead when we attract substantial shares of people who normally support Labour and an overwhelming lead amongst floating voters.
On request, YouGov provide breakdowns of the party ID of respondents once full tables have been published online. On analysis, this weighting was the main reason for our drop in October between unweighted and weighted results. At that point 24% of respondents to the October poll were classed as SNP identifying voters, but these were reduced to each only being two-thirds of a respondent each. This is likeliest to be the reason that we fell behind in the weighting process.
Both of these methods have come in for criticism. Nick Sparrow (formerly of ICM, the UK’s most accurate pollster):
One potential problem is that many voters, unsurprisingly, do not make a real distinction between present party “support” and the party they “identify with most strongly” and the available data shows that the two measures move together over time. Danger is therefore that movements in party ID may not be a measure of an imbalance in the sample, but a reflection of the changing mood of the electorate. [1]
YouGov has an extremely hard time getting responses from tabloid readers – especially the Mirror/Record stable. Readers of these papers can have their votes counted two or three times to make up for what YouGov consider to be a a shortfall.
A very aggressive weighting process is a sign of a bad way of gathering respondents. National telephone polls do not have to engage in anything like the level of weighting that YouGov now seem to routinely undertake in Scotland.
Comments
You can follow this conversation by subscribing to the comment feed for this post.