peterbirks (![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
peterbirks) wrote2017-04-18 03:45 pm
peterbirks) wrote2017-04-18 03:45 pm
Entry tags:
Some early election thoughts
I've been throwing in some preliminary numbers for the election. If I eliminate Scotland, Wales and NI as different regions requiring a different technique, and just focus on the 533 English constituencies, the latest polls indicate as a baseline C +6%, Lab -7%, UKIP -4% and LD + 5%.
However, applying this in a blanket fashion to all 533 constituencies would obviously be a mistake.
What I planned to do was to take the voting in the 2016 referendum and to "map" it onto the 2017 election (not as easy as it sounds because the constituencies were not precisely the same). For every percentage point that the referendum vote moved away from 52%:48% in the direction of "leave", I would add 0.7% to UKIP and subtract 0.7% from LD. Now, for want of a lack of certainty, I did not plan to map any pro- or anti-Leave skew to Lab or Con's vote in the forthcoming election. That might well be wrong, but I don't think it will be massively so.
It's taken me a couple of hours to put together the "database". I'll update in future on how I see this affecting the predicted result. And, of course, I'm yet to start on Scotland or Wales, where it's such a tough call to design an algorithm that I will probably just analyze it seat by seat.
In case I haven't expressed this sufficiently clearly, let's take a hypothetical "Remain", "Labour Held" seat which in 2015 voted:
20,000 Lab
16,000 Con
12,000 LD
5,000 UKIP
It voted 53:47 for Remain.
So our base result would be:
18,600 Lab
16,960 Con
12,600 LD
4,800 UKIP.
However, there's a 5 percentage point difference in favor of Remain from the national result in 2016. Applying the skew to the base result, we get:
18,600 Lab
16,960 Con
13,041 LD
4,632 UKIP
In this particular case, there would be no change, either as a result of the base swing or as a result of the fact that the constituency was pro-Remain. I'll have to go through each of the 533 seats (well, I'll have to design a function to tell the spreadsheet to do it and to highlight the changes in various colours, then to add up the totals! another couple of hours' work) to see how it might affect the final result.
However, applying this in a blanket fashion to all 533 constituencies would obviously be a mistake.
What I planned to do was to take the voting in the 2016 referendum and to "map" it onto the 2017 election (not as easy as it sounds because the constituencies were not precisely the same). For every percentage point that the referendum vote moved away from 52%:48% in the direction of "leave", I would add 0.7% to UKIP and subtract 0.7% from LD. Now, for want of a lack of certainty, I did not plan to map any pro- or anti-Leave skew to Lab or Con's vote in the forthcoming election. That might well be wrong, but I don't think it will be massively so.
It's taken me a couple of hours to put together the "database". I'll update in future on how I see this affecting the predicted result. And, of course, I'm yet to start on Scotland or Wales, where it's such a tough call to design an algorithm that I will probably just analyze it seat by seat.
In case I haven't expressed this sufficiently clearly, let's take a hypothetical "Remain", "Labour Held" seat which in 2015 voted:
20,000 Lab
16,000 Con
12,000 LD
5,000 UKIP
It voted 53:47 for Remain.
So our base result would be:
18,600 Lab
16,960 Con
12,600 LD
4,800 UKIP.
However, there's a 5 percentage point difference in favor of Remain from the national result in 2016. Applying the skew to the base result, we get:
18,600 Lab
16,960 Con
13,041 LD
4,632 UKIP
In this particular case, there would be no change, either as a result of the base swing or as a result of the fact that the constituency was pro-Remain. I'll have to go through each of the 533 seats (well, I'll have to design a function to tell the spreadsheet to do it and to highlight the changes in various colours, then to add up the totals! another couple of hours' work) to see how it might affect the final result.