My last but one post about Hung Parliaments – which is really just the IoD on hung parliaments – has done so well over a sleepy weekend that I feel a little embarassed:  I really added nothing!   So I feel I should try to add something. My own scatter chart for once.

I am going to beef up my Swingometer to a ludicrously overblown level, so that I can enter results as they come in, work out what they seem to be implying about the final result, speculate a little, that sort of thing.  But there are serious assumptions to tease out first.  For example: how swingy is the swing?  Last election the anti-Labour swing averaged about 6 points.    But its standard deviation was 4.5 points – so only about 66% of results would fall between -1.5 and -10.5.    Worth bearing in mind if a freak result comes  in.

But the second point that had me thinking was this: if you get a result in a seat that ‘counts’ i.e. that is in play – should one expect it to be typical?  A seat with a Labour majority of 10% – did it swing more than one with 30% or 5%?  If so, you might get results far more extreme than under the uniform swing.   Note: this is not (yet) about tactical voting (the lost votes going to the appropriate other party).

The scatter chart below suggests that this did not happen in 2005:

Sorry, a mess of dots.  The basic results are: the swing from Labour was about 2% less in Tory held seats than in Labour held seats – which is what a Tory strategist would want.  But the red dots are downward sloping but not downward sloping enough; there are a lot of ‘swung’ votes that took place in very strongly held Labour seats.

To put it more exactly: the average swing in Labour held seats was 6.8% against Labour; but it averaged 5% in those with a majority in 2001 of less than 10%, and 7.15% in those with higher majorities.

Perhaps back then people voted against Labour more as a protest – when they felt it was unlikely to overturn their MP?  Who knows.  I’ve done 20 minutes on this and have a sore back.  But I doubt conditions are anything like the same now.

Advertisements

10 thoughts on “Quick one: do the electorate respond to anti-Labour campaigning?

  1. Interesting stuff. I think it’d be better to call it ‘Change in Labour vote’ not swing, as isn’t a swing from one party to another?

    Isn’t the question of how the swing plays out is the crux of the debate on whether a uniform swing is uniform in percentage points, ie if in every seat Labour falls by 5% points, or whether it is proportional, ie if 5% is 1/7th of its vote then it loses than in every seat?

    1. Very true. I have a couple of days to work out which refinements are worth doing, and that sounds like one I should consider. I was tempted to add a device that allows the user to enter in results as they come, then work out what would happen if that swing were typical across the whole country – a swingometer I guess! – but then seeing these standard deviations I am rather put off.

  2. “Perhaps back then people voted against Labour more as a protest – when they felt it was unlikely to overturn their MP? Who knows. I’ve done 20 minutes on this and have a sore back. But I doubt conditions are anything like the same now.”

    Indeed: Iraq vs. economy, etc.

    1. I’ve got to earn myself some excel geek time this afternoon. May have a lot there then

  3. “if you get a result in a seat that ‘counts’ i.e. that is in play – should one expect it to be typical?”

    Probably not, but aren’t these the seats that are most worth looking at in terms of predicting the result? Whatever votes might change in Toryshire South or Labourtown Central, it’s not going to affect the seats tally.

    And I completely agree with Matthew on thinking about the swing in terms of a party losing x% of its previous vote, rather than of the vote as a whole. I remember years ago reading an analysis of the 1992 result, which remarked with bafflement that there’d been a bigger Con-Lab swing in safe Con seats than in their marginals, and exactly this thought occurred to me. The bigger they come, the harder they fall.

    1. Darnit, since you are a mathmo I will have to reprogramme the ***ing thing and do it that way. Another 15 minute of my life, wasted on Excel.

      More pertinently, a glance last night at the 1992-7 swings found that the 2nd place party picked up about 100% of the Conservative loss of votes (in terms of %). Since the default assumption ought to be ‘pick up that amount in proportion to your vote share (divided into the non-Tory vote)’, that strikes me as v high tactical voting. Not at all the same in 2005.

  4. And I completely agree with Matthew on thinking about the swing in terms of a party losing x% of its previous vote,

    Except this is the debate (linked to on pollingreport.co.uk) between Nate Silver and another chap, and I thought the other chap made some good arguments about going for a classic Uniform Swing (the terminology is difficult here because both are uniform, as in not having regional variations, but the Uniform Swing is usually taken to mean a linear rather than proportional decline.

  5. Oh dear, I just checked I got his name correct and although I had visualised Nate Silver as a wizened old hand of election statistics, I have just found out he was born in 1978.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s