HOW RELIABLE IS THE NEWSNIGHT PREDICTION ?
A week ago, Newsnight produced an analysis of local by-election results by Michael Thrasher and Colin Rallings, which received considerable publicity. Studying this yearâ€™s local by-election results, they suggested that the national equivalent vote shares which the parties would win on May 4th would be CON 33: LD 29%: LAB 28 (compared to 34%, 27%, and 33% respectively in 2002).
On that basis, Labour could expect to lose about 100 seats nationally, the Conservatives about 75, and the Lib Dems would gain over 100.
While it would be embarrassing for Labour to slip to third place, a result on these lines would be truly appalling for the Conservatives. If they were actually to lose seats and votes, compared to the start of Iain Duncan-Smithâ€™s leadership, it is hard to see how David Cameronâ€™s position as party leader would be tenable.
Fortunately, for David Cameron and the Conservatives, there is good reason to believe that their performance, based on this yearâ€™s by-elections, will be rather better than this.
So far this year, there have been 40 by-elections involving all three main parties. 28 of these contests took place in seats last contested in 2003; 7 in seats last contested in 2004, and 5 in seats last contested in 2005.
In 23 of these, the three parties contested the seat on both occasions, making a direct comparison between both elections very easy. However, the absence of by-elections in seats last contested in 2002, (which includes all the seats in London) means any prediction must be treated with considerable caution.
My analysis produces the following results for all seats with the year last contested
2003 CON -1.2%: LAB -3.5%: LD + 4.0%
2004 CON -1.4%: LAB +4.2%: LD -5.2%
2005 CON +6.3%: LAB -7.2%: LD -0.9%
These are the shares for seats where there was an all-party contest on both occasions
2003 CON +3.5%: LAB -3.5%: LD +6.9%
2004 CON -1.9%: LAB +11.2%: LD -5.2%
2005 CON +5.8%: LAB -5.5%: LD -2.2%
The first set of figures implies the following vote shares on May 4th:- CON 35%, LD 29%, LAB 27%. The second set of figures implies CON 38%, LD 31%, LAB 27% (but the Conservative and Lib Dem vote shares are both flattered by a number of rural independents not contesting the seats for a second time).
While this is a small sample, which must be treated with extreme caution, it is nonetheless very hard to work out what Newsnight were basing their prediction on.
There were no by-elections yesterday as it was Maundy Thursday.
Sean is a Tory activist in London and writes a weekly article for the site.