Sunday, 9 May 2010

So, is this constituency size thing for real?

Updated 10th May 08:00
This is an updated post from the one I first published last night because it takes into account comments on the main site about re-working the data on total constituency size rather than total votes cast.

They say a day's a long time in politics. Well I'm just catching up on '24' on Sky whilst fiddling around with the turnout figures from 6th May 2010 General Election.

I just wanted to test whether it was really true that the Tory Constituencies are larger than the Labour ones.

We're being told that Labour Seats are about 8% smaller than Tory ones meaning that the value of a Labour-vote is, in aggregate, worth about 8% more than a Tory one.

If proved, that's clearly wrong and explains why a 7-point lead for the Tories denies them power whereas a 7-point lead for Labour gives them a comfortable majority.

These stories can grow with the telling so I fiddled around on Access and Excel to generate some crosstabs to test the relationship between total votes cast and the winning party.

The picture below speaks for itself. [click for a hi-res version]

And here's the Full Raw Data incl Minor Parties

And here's the link to the Full 2010 General election Results Spreadsheet derived from The Guardian's helpful tables.

It's clear that the Tory seats really do have much higher votes cast. If all constituencies were of an equal size you'd see the red/blue/yellow bars all muddled up. But they aren't. The Tory Blue ones are clustered around the larger constituencies. And the Labour wins in the smaller ones.

But following publishing the post, WorriedVoter and others asked whether there would be a different result if I used the total electorate rather than votes cast to neutralise the effect of differential turnout. I didn't have the total electorate size but WorriedVoter explained it could be calculated from the Total Votes and %Turnout figures, which I did have. Doh! Why didn't I think of that.

If you have the figure for total votes cast and the turnout in each constituency then you can compute the constituency size. Can you not do this and then put the results through your crosstab again to see if a different picture emerges?
by Worried Voter May 9th, 2010 at 11:14 pm

In truth the picture isn't quite so clear and that's because it looks like turnout was higher in the seats that voted Tory but it's still significant and you can see visually that Tory MPs are disproportionately returned in the larger constituencies and Labour MPs in the smaller ones, inferring that it's easier to elect a Labour member with fewer votes.. There's no clear pattern for the LibDems, which are evenly distributed.
So, when you consider the effect of total-votes-cast in a constituency, the Tories are clearly clustered in those constituencies with larger turnout. When you consider the total electorate, the picture isn't quite so clear but still shows a preponderance of Tory seats in the larger constituencies and Labour in the smaller ones. There doesn't seem to be a pattern with the LibDems, which indicates that constituency size isn't as important, which in turn suggests that they have nothing to fear from constituency equalisation.

But whatever you feel about this analysis, it has to be right in a democracy that each vote carries equal weight. That isn't the case at the moment and that's something that has to be fixed.

Bunnco - Your Man on The Spot


DavidL said...

But Bunnco do votes cast really answer the question? My suspicion is that a major factor in it taking more votes to elect a tory is that there is higher turnout in tory seats. Basically tories still vote in safe seats and turnout in safe labour seats is often spectacularly low.
Average seat size is surely by population. I think the biggest problem is that the Boundaries commission wimped out of doing its job properly and failed to reflect the increasing percentage of the population in the south east in particular and the south in general. Anyone looking at the maps from this election can see the consequences of this.

Cerrig said...

Very interesting Bunco. But would it not be more appropriate to use electorate, rather than votes cast? The chart could just show that Tory voting constituencies have higher turnouts?

Cerrig said...

DavidL - snap

bunnco said...

Of course it would be better to show the total electorate but I don't have those figures to-hand and, whilst turnout was variable across the country, it clustered around 60-70% and the chart shows a greater variability than that.

It't not a perfect analysis I grant you but it's a good approximation and significant over 649 results.

It certainly doesn't disprove the constituency size notion.

Robert Barnes said...

Very illuminating; thanks.