Is David Herdson right about Labour and the Lib Dems?

Over on Political Betting, David Herdson has several times made comments such as this:

Since the start of March, there have been sixty national opinion polls published and they have shown a remarkable consistency in the total share identified for Labour plus Lib Dems. When one party rises, the other tends to fall.

I’ve been intrigued by this because David certainly knows his stuff, but that’s not been my impression of the data – which has been that changes in Lib Dem support are more equally balanced between Labour and Tory than the much more lopsided position David suggests.

So I’ve done some number crunching, looking at this question from two angles:

a) How do the changes in party support between polls from the same pollster compare?

I’ve looked at the changes in vote share between consecutive polls from the same pollster for all such pairs of polls since the start of March.

There is a correlation of -0.46 between the Labour and Liberal Democrat changes (i.e. for a 1 point increase in Labour’s share there is typically a 0.46 fall in the Lib Dem share and vice vera).

However, there is also a very similar correlation between the Conservative and Liberal Democrat changes: -0.33. That’s slightly less than the Labour / Lib Dem correlation, but not by much.

Both of these correlations are though higher than that between Conservative and Labour, which comes in at -0.16, suggesting that it is votes flowing to and from the Liberal Democrats which is the most important factor.

b) How do the actual levels of support in the polls compare?

A different way of making the comparison is to look at overall levels of support rather than changes in it. However, this gives a similar picture with the highest correlation being between Labour and Liberal Democrat (-0.27), ahead of Lib Dem/Conservative (-0.14) and with Labour/Conservative the least correlated (-0.04).

So the verdict? It’s true that the Liberal Democrat share of the vote is more closely related to the Labour share than the Conservative share, but the different is not that great. There’s more to what is going on than simply shifts to and from Labour.

Update: as Mike Smithson  points out in the comments, I wrongly attributed the PB story to him initially. Apologies Mike.

Read more by or more about or .
This entry was posted in Polls.


  • Mike Smithson 12th Apr '10 - 3:55am

    I did not write the article to which you refer and have never made such an assertion.

  • You also need to alter the “Mike” at the bottom of the second paragraph

  • I wonder what the trend data is for Undecided/wont say and how it correlates.
    If pundits on the politics show are to be believed it is still very high for four weeks out.

  • I would be interested if you repeated the exercise with “Others”, there was one poll which suggested we could take 40% of the others vote where the favoured Party wasnt standing. If true this could represent an extra 2 or 3% for us & wouldnt show up fully in the polls.

  • David Herdson 12th Apr '10 - 3:27pm

    Interesting analysis.

    You are of course right that there’s more going on than simply centre-left voters switching between Labour and the Lib Dems, which is nigh-on inevitable in a 3+ party system. One factor which is commonly ignored though is one of the more potent is strength of intent to vote at all.

    Thanks for the quote in the article. As I implied both there and elsewhere in the pbc piece, it’s not a direct 1:1 relationship (“When one party rises, the other tends to fall”). There is though a moderately strong relationship as your number-crunching shows, certainly stronger than that of any two other parties. I hadn’t gone into that level of statistical analysis but it was clear from the sample of polls I was looking at that frequently a LD rising share was accompanied by a fall in the Labour one, though rarely of exactly the same magnitude. That the range of shares published by all but one (Angus Reid) of the regular pollsters for the two parties combined was smaller that at least one individually was a strong indicator of a negative correlation.

    In the end, my thinking came down to this: in tactical politics, the voters you most want to go after are those most likely to switch to you, and those you’re most worried about are the flakiest on your own side. As you’ve shown, the Lib Dems and Labour do have a moderately strong collelation are in direct competition in relatively few seats. This made it all the more sensible for Adonis to write his article on the subject he did (though the exact words he chose were poor). I would agree however that it could only ever be one strand to an election campaign.

  • David Herdson 12th Apr '10 - 3:36pm

    One further comment re the LD vs Con correlation.

    I wrote the piece asking about whether Labour was right to target the Lib Dems so openly. As Mark points out, the Lab/LD correlation is a lot greater than the Lab/Con one. Had I been writing the piece following a Lib Dem making similar overtures to Labour it might well have been different but the relationship between the LD/Con shares is far less relevant to Adonis than it would be to a Lib Dem considering doing something similar.

Post a Comment

Lib Dem Voice welcomes comments from everyone but we ask you to be polite, to be on topic and to be who you say you are. You can read our comments policy in full here. Please respect it and all readers of the site.

To have your photo next to your comment please signup your email address with Gravatar.

Your email is never published. Required fields are marked *

Please complete the name of this site, Liberal Democrat ...?


Recent Comments

  • Leekliberal
    @nigel hunter: Hear hear!...
  • James Fowler
    Simon R has it right, as does James Moore. A Corbyn-lite slab of radicals' pet interventionist, illiberal policies won't turn the polling dial a millimetre. The...
  • Denis Loretto
    ....agenda. Buried within these motions is lots of good stuff but when is it publicised? We need to cut these down to crisp statements of the points that really...
  • Richard
    ... tell me aint so. How could one party be in favour of legalising sales of cannabis and at the same time banning sales of tobacco. Surely consistency would re...
  • Denis Loretto
    I reckon the problem is not too little policy but too much. Twice a year we have national conferences which earnestly work their way through worthy but hideousl...