David Maddox: The Numbers Game (8)
It always amazes me how seriously parties take the numbers games and how they crawl over every minute detail from the polls.
My earlier blog on desperation tactics sparked a fairly sharp retort from the Labour offices:
"Polls commissioned by parties do not tell you much - the parties control the questions and crucially the weightings attached to each class of voter (there were 308 people who said they'd vote Labour, but this was "weighted" to 266 in the results). Stick with newspaper polls - they're more reliable!!" It read.
And it was pointed out that while there was a healthy lead for the SNP in its own commissioned poll for Holyrood the Nationalists were behind for Westminster. And if it was compared with the last SNP commissioned poll of August 2008 the SNP were going down which ever way you look at it.
Here they all are along with Westminster seat calculations courtesy as ever of Electoral Calculus:
Westminster (seats won and change in number in brackets):
August 2008: SNP 36% (26 +20) Lab 29% (22 -19) Tories 18% (4 +3) LD 13% (7 -4)
April 2009: SNP 30% (10 +4) Lab 32% (35 -6) Tories 21% (5 +4) LD 13% (9 -2)
Holyrood constituencies:
August 2008: SNP 44% Lab 25% Tories 13% LD 14%
April 2009: SNP 37% Lab 30% Tories 15% LD 13%
What does this prove?
1. Voters are fickle.
2. They vote differently for Holyrood than Westminster.
3. Point 2 suggests that SNP votes are not necessarily for independence.
4. Minor fluctuations in support could lead to dramatic changes in results (see the Westminster seat calculations).
My earlier blog on desperation tactics sparked a fairly sharp retort from the Labour offices:
"Polls commissioned by parties do not tell you much - the parties control the questions and crucially the weightings attached to each class of voter (there were 308 people who said they'd vote Labour, but this was "weighted" to 266 in the results). Stick with newspaper polls - they're more reliable!!" It read.
And it was pointed out that while there was a healthy lead for the SNP in its own commissioned poll for Holyrood the Nationalists were behind for Westminster. And if it was compared with the last SNP commissioned poll of August 2008 the SNP were going down which ever way you look at it.
Here they all are along with Westminster seat calculations courtesy as ever of Electoral Calculus:
Westminster (seats won and change in number in brackets):
August 2008: SNP 36% (26 +20) Lab 29% (22 -19) Tories 18% (4 +3) LD 13% (7 -4)
April 2009: SNP 30% (10 +4) Lab 32% (35 -6) Tories 21% (5 +4) LD 13% (9 -2)
Holyrood constituencies:
August 2008: SNP 44% Lab 25% Tories 13% LD 14%
April 2009: SNP 37% Lab 30% Tories 15% LD 13%
What does this prove?
1. Voters are fickle.
2. They vote differently for Holyrood than Westminster.
3. Point 2 suggests that SNP votes are not necessarily for independence.
4. Minor fluctuations in support could lead to dramatic changes in results (see the Westminster seat calculations).
Labels: David Maddox, elections, Holyrood, Labour, polls, SNP, Westminster
posted by Administrator at
19:07
1 Comments:
Perhaps Labour's press office will want to consult Philip Gould before commenting again?
Weighting is a statistical tool introduced to balance the numbers who actually answer a pollster's questions with the demographics of the population. You may remember polls which showed the Conservatives on course for defeat during the 1980s and early 1990s when they dumbfounded the pollsters and the great search for a way of balancing that. Weighting is controlled by the pollster, not the client.
As for parties controlling the questions, the client will tell the pollster what the client wants to find out about and the pollster will work with the client to frame the questions. Questions and results from reputable polling firms will be published and you can check on what was asked.
If Labour's press office no longer knows the basics about how polling works then they're in real trouble. Not only have they lost sight of politics, they've lost the knowledge of how the tools work.
Post a Comment
<< Home