Why David Laws was right about the Conservative Party’s poll ratings

Appearing on the BBC earlier today, David Laws made the point that the Conservative Party’s lead in the opinion polls is fairly modest at the moment compared with Labour’s in the run-up to 1997.

This led Mike Smithson to blog:

Why’s the LD schools spokesman getting it so wrong?

You’ll have to indulge me if you think I’ve banged on about this too much – but I have a real “bee in my bonnet” about the phoney invalid polling comparisons that journos, pundits and politicians are rushing to make when they compare the polling position at the moment with what went on in 1996/97…

Mike went on to point out how most pollsters over-estimated Labour’s support before polling day in 1997, adding that:

One pollster was totally out of line – ICM. Then, as now, it mostly had higher ratings for the Lib Dems and much smaller shares for Labour. And guess what happened on election day? Its approach was proved right.

I’d agree with that praise for ICM. But David Laws’s comments are fully justified by the ICM figures.

Since last summer (i.e. August 2008-March 2009), the Conservative lead in the ICM polls has varied between 5% and 15%.

If you look back at August 1995-March 1996*, the Labour lead in the ICM polls varied between 14% and 22%. There is a small overlap between these two ranges, but overall the Labour lead in 1995/6 was much higher than the Conservative lead in 2008/9, averaging 18% for Labour compared to 12% for the Conservatives.

Moreover, due to the way their respective vote shares are distributed across the country, the Conservatives need a larger lead to have an overall majority than Labour – so in fact the relative position for them is even worse than the 18% vs 12% figure suggests.

Sorry Mike, but on this one, I think you’ve got it wrong: the Conservative lead is consistently significantly lower than the comparable Labour lead was.

* This is the comparable period for a 2010 election. But the point also stands for a 2009 general election, for example if you take the period August 1996-March 1997 instead.