Mr Pack graciously <a href="http://redfellow.wordpress.com/2013/05/03/any-hope-of-realities-here/#comment-3586">commented on my trivial little blog</a>, to enlighten me on what he feels this is all about.
For the record, then:
I agree with Mr Pack that my original remark was misconceived: it assumed that real data was involved. In truth it involves, as he recognises above, unrealsyntheses —
“ie what the vote shares would have been for each party if each ward has been up for election this year and fought by them – which therefore adjusts for the variations in which wards are up in any particular year”.
I’m also leery about “would have beens” and pseudo-statistical evaluation of apples versus oranges. I believe it is provable that the British soldiers at Waterloo had an inch or two of average height above those at the Somme. I wouldn’t suggest that contributed to the comparable outcomes of those events.
Mr Pack, on his blog, now invites us to share John Rentoul’s applause for Mr Pack’s remarkable analysis. To tell the truth, Rentoul’s explication of the “figures” suggests they are even more synthetic that Pack allows:
“The Projected National Share figures are hard to understand: they extrapolate from places that voted to estimate how the whole of Great Britain might have voted if the local elections had been held everywhere, and the main parties had stood candidates in every seat.”
So, we now have a “projection”, a “might” and two “if” concessions. Four levels of manipulation.
A real-world comparison of May 2013 with May 2012 would involve just four County Council Wards which were contested on both occasions: Kendal Strickland and Fell (Cumbria CC), Meriden Tudor (Herts CC), Bixley (Suffolk CC), and Worplesdon (Surrey CC). Even then one of those —Worplesdon — is not a direct comparison, because there was no UKIP participation in 2012. However in Worplesdon, where UKIP ran second in 2013, it was Con minus about 3%, Labour minus about 5% and Lib Dems minus nigh on 16%.
But I’m not drawing nationwide conclusions from that.