Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

The other UK losers: pollsters, Liberal Democrats, and 50% of Scots voters

May 9th, 2015, 9:34am by Sam Wang

Reanalyzed data from Lord Ashcroft's poll. Click to see original data

(reanalyzed data from Lord Ashcroft’s post-election poll)

Contrary to all pre-election polls, the Conservatives won an outright majority, with 329 out of 650 seats. The final popular vote was 37% for Conservatives, 31% for Labour, where a near-tie was predicted. But the Labour Party was not the only loser on Thursday. There were three others:

1. Pollsters. Why were opinion polls so far off? It is unlikely to have been a problem with telephone sampling, since pollsters who used cellphone and other sampling methods made similar errors. A probable culprit is bad likely-voter modeling: turnout was 66% of the electorate, the highest since 1997.

2. Liberal Democrats. A second reason is a last-minute fade by minor parties, specifically the Liberal Democrats, who won a paltry 8% of the vote, 2% below final opinion polls and an amazing 15% behind their previous election performance. Aligning with the Conservatives did not pay off for them. They end up with 8 seats, way down from 55 seats. That ties them with the Democratic Unionist Party, which most US readers have not heard of.

LibDem voters from 2010 scattered toward all parties. I have re-analyzed Lord Ashcroft’s post-election poll to see where they went. See above – only 28% of them voted LibDem again in 2015. That is a remarkable crash.

Turnout error and minor-party fade are known errors. Generally, multi-party elections are tough; there are increased possibilities for strategic voting. Still, these factors can catch analysts and consumers unawares. Here is one preliminary postmortem from YouGov.

3. 50% of Scottish voters. Speaking of strategic voting, a big winner was the Scottish Nationalist Party (SNP), which won 56 out of 59 seats in Scotland. Yet they won exactly 50% of the vote there, very similar to the 45% “Yes” vote on last year’s Scotland separation referendum. And this is a cohesive group: of those who voted SNP in 2010, 95% came back in 2015. That is amazing loyalty.

For 50% of voters to capture 95% of the seats is an enormous deviation from proportionality. Under these conditions, about half of Scots voters are likely to be shut out of UK politics for the foreseeable future. This will sound radical, but it would be a strategic move for the Conservatives and Liberal Democrats to drop out of Scotland politics, to make way for Labour candidates, a less-bad option from their point of view.

Another possibility is to do away with “first-past-the-post” voting. However, I imagine people like having their M.P., just as we like having our own Congressman over here.

Tags: United Kingdom

14 Comments so far ↓

  • WDR

    Why would the conservatives want to drop out? Labour winning Scottish seats would be worse for them than SNP, because SNP can never challenge them at the national level.

    • Sam Wang

      I agree that it would not be an easy choice. The alternative is a voting bloc that will always push for separation from the U.K.

  • O.T. Ford

    The Tories are governing alone despite winning 37% of the vote. Aren’t 63% of all Brits losers, then? Why single out the SNP?

  • Amitabh Lath

    Yes of course it is the turnout (or likely voter) model. It has been clear for a while now that the level of systematic uncertainty associated with the LV models is approaching or exceeding the statistical uncertainty.

    Like any systematic uncertainty, this one can be attacked methodically. First it has to be understood (maybe polls specific to voting behavior?) and dissected. But so far polling firms have treated LV models as some sort of secret sauce, not even willing to discuss what goes into them, much less assign any uncertainty to them. Occasionally one sees an artifact (such as the US midterms in 2014, around the 3rd week of September) which may be due to polling firms switching from a “loose” to “tight” LV model. Hardly the model of transparency one would hope for.

  • pechmerle

    Note, however, that Amitabh is not consistently applying his own excellent adiabatic – nonadiabatic elections distinction from prior related topic.

    Looking at the numbers, we readily see that Labour’s 2010 % hold is actually higher than Tories. The difference, as Amitabh correctly noted the other day is, in a word, Scotland! Labour lost to a highly concentrated party (“naturally” gerrymandered), while Tories lost votes to parties, esp. UKIP, widely dispersed around Engliand and thus did not lose those seats, as Labour did across smaller Scotland.

    If we set Scotland aside, Labour gained about the same number of seats from other parties as Tories. But in Scotland, Labour was wiped out by SNP, whereas Tories had for years had only small support in Scotland anyway. So the result is non-adiabatic — the external factor of a huge surge in Scottish nationalism occurring since the independence referendum was put on the table several years ago.

    This, to me, calls into question Amitabh’s earlier comment that conservative parties outperform their polling figures. Since this was a non-adiabatic election, it neither supports nor refutes that as a systematic proposition.

    Also, for those here who may not be aware of it, a move away from first-past-the-post system was put to the U.K. voters in a referendum in 2011. The result was 2-1 (67% – 33%) against! The voters aren’t comfortable with system-change, but in exchange they are going to get dysfunctional governance.

    And, a further implication of the results of this election will be pressure on the SNP from its voters: you almost got independence in 2014; you now have the 3rd largest bloc of seats at Westminster. What are you going to get for us in return for that remarkable loyal support we have shown you?

  • Amitabh Lath

    You are right, I guess the only way we know for sure we are in a non-adiabatic situation is for historians to do a post-mortem. Looking back at the decades after WW2 we can say that was a particularly stable time and voting behavior did not change, until after the civil rights movement when the Confederate south changed from D to R.

    It may well be the SNP is a flash in the pan, they will do a whole lot of not much, their voters will realize no way is Scotland turning into Scandinavia with kilts, and a suitably chastened Labor will be back. Might take a couple of cycles.

  • Amitabh Lath

    The dominant theory in the 1990’s was that opposing parties would start to resemble each other as the losing party in any election inched towards the winner.

    Francis Fukuyama and the “end of history” ideas were popular among the types that listen to NPR and read the Atlantic Monthly.

    It’s like finding a minimum of a function.
    You can start at the right or left, but slowly you get to the same point. Bill Clinton was a good example of inching to the right.

    But after 9/11 I guess we got a function with two minima. There is the “keep the foreigners out + spend less” point vs. the “immigration is good + help the poor”.

    The Republicans in the US, Likud in Israel, BJP in India, Tories in the UK are all examples of the former.

    Note that other than the US, the lefty opposition is in tatters. In Israel they don’t even have a coherent party (was Labor, then Kadima, now Zionist…). In India the Congress party rolled the dynastic dice and best they had was the loser great-grandson of Nehru, and in the UK Ed beat David in the battle of Millibands and drove Labor over a cliff.

    But recall all these lefties were polling much better right up to the election.

    In the US, if it were not for a couple of historic and charismatic executive personalities (Obama, now H. Clinton) I suspect we would be going the same way.

    • Todd Horowitz

      “In the US, if it were not for a couple of historic and charismatic executive personalities (Obama, now H. Clinton) I suspect we would be going the same way.

      That’s not a very satisfying theory. Once you start thinking about “historic and charismatic personalities” as exceptions, then history becomes a series of exceptions. After all, the conservative shift in the US in the 80s was driven by the enormous charisma of Ronald Reagan, while Obama’s election was aided by the fact that Bush jr. made such a historic hash out of US domestic and foreign policy.

      The different outcomes for the US on the one hand, and Israel and the UK on the other, can be put down to some less idiosyncratic factors. One is the difference between parliamentary and presidential democracy. The historic balance between the Republicans and Democrats have shifted in the US, with the Democrats becoming the Presidential party, and the Republicans the Congressional party. If we had a parliamentary democracy, Obama might have been ousted in 2010.
      Also, in the UK, the center-left was in power for the Iraq/Afghan debacle AND the financial collapse, whereas in the US those disasters fell on Bush jr.
      Then there’s the whole multiparty factor. A US President who received less than 37% of the vote would be considered a historic failure, but in the UK that’s a landslide victory.

    • Amitabh Lath

      Yeah, executive charisma is a fairly lame parameter to hang this on. I was trying to figure out why the pollsters have been so far off, and the idea occurred of a function with two minima, spaced far enough apart that if your estimator falls into one minimum, you can fool yourself into thinking you have fairly precise handle on things. You can jiggle parameters (age, race, rural/urban) by small amounts and things behave linearly according to expectation.

      But being stuck in a false minimum means if you (or reality) picks a voting population with just a tiny bit more rural voters, or older voters, you don’t get a result that is just a tiny bit more conservative, but hugely more so.

      So in this picture, the minimum corresponding to conservatives is deeper by default, but some one-shot variable (like say, an executive candidate with charisma, maybe something else) can pull the liberal well deeper.

      As for first past the post, multiparty etc. that should all be baked into the form of the function, no?

    • Matt McIrvin

      Politics in the US has become significantly divided along ethnic/racial lines. Through most of the 20th century, partisanship was tempered by the race/civil-rights question cutting across party lines. But now it mostly doesn’t, and the Democrats have become the overwhelming favorite of everyone who isn’t white/non-Hispanic, while the Republicans get a substantial majority of white people.

      And the non-white/Hispanic vote gets larger and larger… but is still largely AWOL in anything other than presidential elections.

      I think everything in the current US is dependent on that. The American left, which is not very left to begin with, pretty much always loses big among white people, but the right loses catastrophically among everyone else, and they keep doing things that put them further and further into that hole. If they didn’t, they could probably keep dominating.

      I don’t have a sufficient feel for how this maps onto the situation in other countries.

    • Todd S. Horowitz


      Yeah, maybe there are two minima. I guess maybe my point is that the rightward tilt of the UK and Israeli elections doesn’t mean that the conservative minimum is deeper, just that the UK and Israel are stuck on one side, while the US is in the other… but of course you need different models for the US congressional and presidential electorates.

  • Mike Martin

    The general take is that the Labour party was a big loser. However if you take away the SNP losses, it looks like they gain seats. I understand that they lost seats, but the circumstances were unusual. They took 7% of the 2010 Tory vote, while the Tories took 6% of theirs. They also took a greater percentage than the Tories of every other party, except the UKIP. They also held a greater percentage of their 2010 vote than did the Tories. A loss is a loss, but this is not the end of the world for the Labour party.

  • Anthony Dunn

    Todd S Horowitz May 12, 2015 @ 11:01

    Some of us would beg to take issue with your assertion that the UK has shown a rightward tilt with the most recent election.

    Despite the temptation to lump UKIP into the right wing vote share, the party defies such easy and convenient categorisation. Look at what it stands for: a mix of populism that straddles left and right whilst being firmly in the authoritarian (as opposed to liberal) end of the spectrum. As such it has drawn at least as much and possibly rather more support from Labour than from the Tories.

    Look at the groups from whence UKIP draws the bulk of its support: the more elderly, white, less educated, low skilled, lower income, working class (you would call them “middle class”…!) who feel that the world has left them behind. Which is why many would like to turn the clock back to the 1950s when the UK was still a top dog, we still had an Empire, immigration was relatively low and they were almost assured full employment with a decent wage. Much of that has gone – and with it their attachment to the Labour party.

    The Tory party, dominated by so-called “Toffs” from public school and overwhelmingly funded by the Hedge Fund industry, hardly offers an alternative home and, with the LibDems in Coalition for the past five years (but NEVER again!), the latter’s repository as a “protest vote” option went down the plughole.

    The issue over here is how soon the Tories lose their majority and what then? The Democratic Unionists from N’orn Ireland to prop them up?

    The other thing that has not gone away, despite the attempts of the Tories and their cheerleaders in the print meedjah to suppress it, is electoral reform. When the Tories obtain one MP per 36,000 votes whereas the LibDems have one per 399,000 votes and UKIP one MP for four and a half million votes, there can be no denying that the UK has a broken voting system. When you add in an unreformed House of Lords (it still retains an hereditary element in 2015!), an unreformed system of political party financing, unreformed meedjah ownership, unreformed meedjah conduct and an unreformed lobbying industry then you have a democratic deficit as wide as the Grand Canyon.

Leave a Reply to WDR (Cancel)