"What's the REAL Tory poll lead?" - UPDATE
In my original thread I outlined why the Con lead may be inflated.
I also invited feedback, and boy, did I get it - some of which has been very insightful, and I am grateful.
Time for updated figures...
1/12
In my original thread I outlined why the Con lead may be inflated.
I also invited feedback, and boy, did I get it - some of which has been very insightful, and I am grateful.
Time for updated figures...
1/12
So, why the need for an update?
Pollsters are semi-transparent about how they model the numbers. Looking at their tables only gets you so far
I did the best with data I had
Since then, some pollsters reached out with new info. I am re-modelling the numbers based on this.
2/12
Pollsters are semi-transparent about how they model the numbers. Looking at their tables only gets you so far
I did the best with data I had
Since then, some pollsters reached out with new info. I am re-modelling the numbers based on this.
2/12
As per the original, we will look at
1) turnout
2) EU vote weighting
3) new voter registration and 2017 Did Not Vote
Turnout is easy.
Most agree that my approach of ironing out the outliers in some polls + applying GE17 turnout by age was fair. No change - Con: -1.6%
3/12
1) turnout
2) EU vote weighting
3) new voter registration and 2017 Did Not Vote
Turnout is easy.
Most agree that my approach of ironing out the outliers in some polls + applying GE17 turnout by age was fair. No change - Con: -1.6%
3/12
For EU vote weighting, it's more challenging.
Different pollsters have revealed they have different approaches to this weighting.
Some polls DO take into account some demographic changes in their EU weighting. But not all polls.
How to remedy this?
4/12
Different pollsters have revealed they have different approaches to this weighting.
Some polls DO take into account some demographic changes in their EU weighting. But not all polls.
How to remedy this?
4/12
On balance, removing the imputed EU VI for people turning 18 in my original analysis is reasonable.
This also avoids a 'double counting' scenario with new voter registrations.
Removing this imputed VI means the reduction in the Con lead is now -1.4% (down from -2.4%).
5/12
This also avoids a 'double counting' scenario with new voter registrations.
Removing this imputed VI means the reduction in the Con lead is now -1.4% (down from -2.4%).
5/12
Now onto the issue of new voter registration and the impact on the Con lead.
We now have partial data on registered voters to more accurately model the impact.
I have outlined this in the thread below.
So what is the impact of these new numbers?
6/12
We now have partial data on registered voters to more accurately model the impact.
I have outlined this in the thread below.
So what is the impact of these new numbers?
6/12
In my original thread I estimated the new voter pool at 4m.
This has been shown to have been too high. It's likely, accounting for deaths, to be more like 3.2m.
This is still an incredible figure.
But not quite as incredible as the 9m 2019 registrations suggested.
7/12
This has been shown to have been too high. It's likely, accounting for deaths, to be more like 3.2m.
This is still an incredible figure.
But not quite as incredible as the 9m 2019 registrations suggested.
7/12
To adjust for the more accurate 3.2m new Lab-leaning voters, I will decrease the weighted impact of them by 20%.
This means that the net impact on reducing the Con lead is -2.53% (down from 2.84%).
However, based on additional feedback, I will go one step further...
8/12
This means that the net impact on reducing the Con lead is -2.53% (down from 2.84%).
However, based on additional feedback, I will go one step further...
8/12
Some pollsters have stated that my non-downweighting for 2017 DNV's does not align with historical patterns of voting.
I knew that.
But I also know that there are record high levels of political engagement.
However, based on feedback, I will downweight 2017 DNV's 25%.
9/12
I knew that.
But I also know that there are record high levels of political engagement.
However, based on feedback, I will downweight 2017 DNV's 25%.
9/12
The result of these adjustments is as follows;
Turnout adjustment: -1.6%
EU voting adjustment: -1.4%
New reg/DNV admustment: -2.05%
Adjustment to Con lead: -5.05%
Since the average Con lead over most recent 10 polls is 9.5%, the 'real' Con lead = 4.45%.
However...
10/12
Turnout adjustment: -1.6%
EU voting adjustment: -1.4%
New reg/DNV admustment: -2.05%
Adjustment to Con lead: -5.05%
Since the average Con lead over most recent 10 polls is 9.5%, the 'real' Con lead = 4.45%.
However...
10/12
We should bear in mind the impact of the Brexit party.
I have analysed this in the thread below.
Because of this, the national Con vote share is inflated by ~3% but this has no impact in the marginals.
Indeed, it damages Con in the Lab marginals
11/12
I have analysed this in the thread below.
Because of this, the national Con vote share is inflated by ~3% but this has no impact in the marginals.
Indeed, it damages Con in the Lab marginals
11/12
Any Con lead can be seen in this context.
~3% of their national vote %, from the perspective of gaining seats, is an illusion.
It's just Con votes piling up in Con seats.
A 4.45% lead on polling day would be a 1% swing to Labour in Lab held seats. And a Lab minority govt
END
~3% of their national vote %, from the perspective of gaining seats, is an illusion.
It's just Con votes piling up in Con seats.
A 4.45% lead on polling day would be a 1% swing to Labour in Lab held seats. And a Lab minority govt
END
