TNS-BMRB published a poll on 9th April, which gives us hope for more seats than the media pundits are claiming. The headline UKIP figure was shown as 19%, which led me to believe that it was a more “honest” poll than the likes of those we see from YouGov that have us bobbling around on 12-14%.
Digging into it, it seems they DID prompt for UKIP, but still redistributed the “Don’t Knows and Refused to Answers” in the proportion of the 2010 result, which is unfair to UKIP to the tune of around 4-5%, i.e. the headline figure would be 23-24% if they didn’t do that.
So, I went digging into the poll. For a start, they did a slightly better regional breakdown than most polls do. For example, many of them just have a “South” area that encompasses a variety of sins, mainly London, where UKIP support is markedly lower. This poll split London out, but it did combine 8 of the regions into 4 super-regions, the worst discrepancies being achieved by lumping the South West and Wales together.
It was also a small poll, just 1144 respondents, which means that the leading party in each super-region only had a sample of around 50-70, but that will do if you are only looking for the potential winners.
The real gold dust in this poll was the table on Page 37, buried away amongst 73 pages. This asked the following question:
Still thinking about the next UK general election, which party, if any, will you vote for in your own constituency?
This is significant, because it forced the voter to detach themselves from their general party allegiance, think about the candidates presented to them, the local issues, and also perhaps to consider tactical voting. The base of voters used was those who would probably vote in 2015 General Election. Also, and this was the key to the table, it showed the “Don’t Knows/Refused” respondents separately, so it was not corrupted by weighting to 2010 results. Promptly, I re-distributed them pro-rata to the opinion poll result. I also split out the super-regional poll values to the two regions making each up, using the variance at the Euro elections.
Then I cranked the handle on my numbers machine, a whopping great Excel workbook that contains all the Euro results by council from 2014, and a prediction engine that maps the untidy and non-matching boundaries between councils and constituencies, and allows for a proportion of voters “returning home” to their more natural party. I then applied to that initial post-May 2014 prediction the variance between the opinion poll and the Euro result. This then came up with a results prediction, calculating vote percentages for the 5 “major” parties. To simplify matters, I only did England, to avoid having to deal with the numeric complexities of the SNP and Plaid Cymru in just Scotland and Wales.
Now, at this point, I would guess you are hoping I’m going to give you a list of winning seats? Wrong! This would be extremely divisive. As I said, the numbers were relatively small, and there could be errors when we are getting down to smaller numbers of respondents per region.
The objective was to see what the effect of regional variations in the headline poll figure would have, based on the most “honest” numbers in the poll, and then factored by the actual results in each constituency in May 2014. If the regional variations fluctuate in later polls, then one region might lose some seats in the prediction, but another region would gain them.
We are profiting from the same benefit that the Liberal Democrats have – we have “hot spots” and it is those hot spots that are going to return UKIP MPs. The end result of all this is:
! 42 SEATS !
The answer to Life, The Universe and Everything. I shall be applying the same technique (now encompassed in a workbook) to other more hopeful polls as we approach the election.