Looking at local newspaper feeds, twitter feeds of reporters on the ground and so on here is what I can make out so far…

WILTSHIRE and DYFED-POWYS have both declared and were both won by the Conservatives.

BEDFORDSHIRE has gone to a second round between the Conservatives and Labour, pretty much neck and neck and the moment. Ind, Lib Dem and British Freedom second preferences to be reallocated.

CLEVELAND has gone to a second round between the Conservatives and Labour. Green and Independent 2nd preferences to be reallocated.

DORSET is reported to have an independent in first place so far.

ESSEX has gone to a second round between the Conservatives and an Independent. The Conservatives have a lead of 10,000 or so votes, but there are lots to be redistributed, including Labour, a second independent and UKIP.

GWENT’s first round was won by an Independent, Ian Johnston, and has gone to a second round

KENT is still counting, but the former Independent chair of the police authority Ann Barnes is ahead on the first round and looks set to win

MERSEYSIDE should be ready to declare shortly, Labour have won easily on first preferences.

NORTH WALES has gone to a second round between Labour and an Independent

NORTHUMBRIA hasn’t quite declared yet, but has been easily won by Labour. They have declared now and Vera Baird has indeed won comfortably on the first round.

SOUTH YORKSHIRE has apparently been won by Labour on the first round, but no figures yet

SUFFOLK has gone to a second round with Labour and the Conservatives absolutely neck and neck, there are votes from UKIP and an Independent to be redistributed.

WEST MIDLANDS has gone to a second round, but Labour’s first round lead looks unassailable.

UPDATE: Labour have won Corby with a 22% lead, so the Populus(?) poll by Lord Ashcroft got the lead correct. It is a swing of around about 12.7%, so significantly better than the national polling position, which is currently showing a swing of around about 8.5%.

There won’t be many results tonight – only the two safe Labour by-elections, Manchester Central and Cardiff South and Penarth, and one of the Police Commissioner elections, Wiltshire, are counting. Everybody else is starting their counts on Friday. However, feel free to discuss results here as the come in (or, more likely, when you wake up in the morning!)

UPDATE: The results in the two safe Labour seat were much as expected – very comfortable Labour holds. Labour held Manchester Central with a towering 69% of the vote with the Lib Dems in second place, a swing of 17% from the Lib Dems to Labour. Everyone else lost their deposit, with the Conservatives only narrowing beating UKIP into third place. Turnout was 18%.

Cardiff South and Penarth has a higher turnout but more modest swings. Labour held the seat easily with 47% of the vote with the Conservatives in second place. There was a swing of 8.4% from the Conservatives to Labour, pretty much in line with national polling (it would be the equivalent of a ten point Labour lead in the national polls).

Finally the only overnight Police result was Wiltshire, which returned a Conservative Commissioner after redistributing second preferences, Labour came in second place (note that in 2010 general election votes in Wiltshire Labour were a distant third behind the Lib Dems,… although we don’t yet know to what extent police votes will reflect general election votes). The two independent candidates came third and last, but had 21% of the vote between them suggesting there may be potential for some Independent victories later on today


Ipsos MORI’s monthly political monitor is out with topline figures of CON 32%(-1), LAB 46%(+3), LDEM 9%(nc). Changes are from MORI’s October monitor.

The 14 point Labour lead is the largest MORI have shown this Parliament and 46% the highest Labour have been, but before anyone gets too excited it is almost certainly due to a particularly odd sample, rather than a genuine shift in support.

As regular readers will know, most pollsters these days use some form of political weighting – in most cases weighting how people say their voted in the 2010 election to what the actual result was, adjusted slightly to account for people’s known shortcomings in accurately reporting their vote (known as “false recall”). MORI are one of the few companies who don’t do this – MORI’s reasoning is that they believe that degrees of false recall can change quickly, and weighting by it therefore risks weighting out genuine changes in support. In contrast companies like ICM and Populus believe that false recall exists, but changes only slowly over time, meaning that recalled vote is suitable for weighting as long as false recall is accounted for.

In practice the phone companies are assuming quite a low level of false recall at the moment, so most polls have past vote weighted very close to the actual results of the 2010 election:

  • Populus’s samples normally have 23% people who voted Conservative at the last election, 19% people who voted Labour, 15% who voted Lib Dem, 42% who voted for other parties, didn’t vote or won’t say. This equates to CON 37%, LAB 30%, LDEM 23%.
  • ICM’s samples normally have 23% people who voted Conservative at the last election, 19% people who voted Labour, 14% who voted Lib Dem, 44% who voted for other parties, didn’t vote or won’t say. This, again, equates to CON 37%, LAB 30%, LDEM 23%.
  • ComRes’s samples normally have 23% people who voted Conservative at the last election, 19% people who voted Labour, 14% who voted Lib Dem, 44% who voted for other parties, didn’t vote or won’t say. This, again, equates to CON 37%, LAB 30%, LDEM 23%.
  • This month’s MORI sample has 22% people who voted Conservative at the last election, 27% people who voted Labour, 14% who voted Lib Dem, 37% who voted for other parties, didn’t vote or won’t say. Roughly speaking (as we don’t have the figure for others), this equates to CON 32%, LAB 40%, LDEM 21%.

So this month’s MORI sample has significantly more people who voted Labour in 2010 than people who voted Conservative, and has about a third more 2010 Labour voters in its sample than other telephone polling companies. Unusual sample and, consequently, an unusual result.

It is important to note that this sample is NOT typical of MORI, so don’t go away with the idea that MORI consistently have vastly more 2010 Labour voters in their samples than 2010 Conservative voters – they don’t. Most of the time MORI’s samples are far more in line with other companies and their results are also normally very much in line. It’s just this month’s sample that’s wacky.

Gay marriage… again

Guido and the Spectator have a story up about a letter from Andrew Hawkins at ComRes accusing David Cameron of misrepresenting polling data over gay marriage. It seems to be an issue that keeps coming back, so I thought I may as well revisit the issue once again and look at what the polls actually say on gay marriage.

Now, to start at the beginning with Andrew’s letter. In Cameron’s original letter he wrote that 10% of Tory voters said legalising gay marriage would make them less likely to vote Conservative, compared to 7% of people who said it would make them more likely. Cameron’s letter claimed the difference between these figures was so close as to not really be significant. The specific factual criticisms Andrew makes about these claims are correct. What Cameron appears to have done is quote figures from page 5 of this poll, which refer to the Labour party, rather than the figures from page 2 of the poll which refer to the Conservative party – the page 2 figures would suggest a rather bigger gap!

However, looking only at the opinions of current Conservative voters seems a rather bizarre thing to do in the first place. The Conservative party, after all, would presumably be interested not only in holding onto current support, but in gaining support from people who are not already supporting them. If you look at the figures overall, 10% of people said gay marriage made them more likely to vote Conservative, as opposed to 13% who made it less likely… the same 3 point gap as in the figures Cameron wrongly quoted.

Gay marriage is a swings and roundabouts issue – it no doubt alienates some people with socially conservative or religious views, it appeals to other people with more socially liberal views. Lord Ashcroft’s polling on the same subject has somewhat more useful crossbreaks. Its overall findings are almost identical to those of ComRes – Ashcroft found 10% said they were more likely to vote Tory as a result of gay marriage, 12% said they were less likely to vote Tory as a result of gay marriage. Both polls found former Con voters saying it made them much less likely to vote Tory. However Ashcroft also provided cross-breaks for people who didn’t support the Conservatives in 2010 or now, but said they might consider doing so. These people said the policy would make them more likely to support a party – but it was a smaller group and by a much smaller margin. So, the polling is consistent in showing the potential loss is bigger than the potential gain… but not by a significant amount.

As it happens, I think “would X make you more or less likely to vote Y” are often useless and misleading questions anyway, as I have ranted about before. They give an issue false prominence, when actually people’s votes will more likely be driven by bigger issues like the economy and perceptions of party competence, secondly people responding to polls are not stupid and tend to use them to register their support or opposition to a policy regardless of whether it affects their vote, third people are poor judges of what actually drives their voting intention, so the bigger impact of policies may be on broader perceptions of party image that are not picked up in polls like this.

In short, what drives voting intention is an extremely complicated question. Political scientists spend whole careers studying it, doing complicated analysis of data sets from British election studies and trying to model it. It would be very nice if we could understand it in a single polling question and I can understand people’s attraction to questions that look as though they do… but it really doesn’t work like that.

Moving on to the second part of the letter, whether polls consistently show people support gay marriage. Cameron says in his letter that “All of the published polls have found that more voters support equal civil marriage – however described – than oppose it”. This isn’t really true either.

Below are a list of polls that essentially asked if people supported gay marriage and gave them a straight yes/no or agree/disagree question:

Ashcroft/Populus? – May 2012. Support 42%, Oppose 31%
ICM/Sunday Telegraph – Mar 2012. Support 45%, Oppose 36%
Populus/Times – Mar 2012. Support 65%, Oppose 27%
ComRes/Independent – Oct 2011. Support 51%, Oppose 34%.
YouGov/Sunday Times – Nov 2012. Support 51%, Oppose 38%
YouGov/Sunday Times – May 2012. Support 51%, Oppose 35%

As you can see, they all show more support for gay marriage than opposition to it, though there is some variation in results and in wording (for example, the Populus one talked about “equal rights” in the question, the Ashcroft one gave a “don’t mind either way” option).

So far Cameron’s statement looks true… but now look instead at the questions below from Angus Reid and YouGov, they gave a three way option – asking if people supported gay marriage, supported civil partnerships but not gay marriage, or opposed both. This gives a slightly different picture.

YouGov/Sunday Times – Mar 2012. Support gay marriage 43%, support civil part only 32%, oppose both 15%
YouGov – Sep 2011. Support gay marriage 46%, support civil part only 28%, oppose both 17%
Angus Reid – July 2011. Support gay marriage 43%, support civil part only 34%, oppose both 15%

There is still a plurality in favour of gay marriage, but if you add together those who support civil partnership but not gay marriage and those who oppose both of them then in some cases there are more people opposed to gay marriage than support it – so when David Cameron says all polls show more people support gay marriage than oppose it however you ask the question, it isn’t really true. If you ask it as a three way question, it suggests that the two way question isn’t necessarily giving the whole picture, and that some of the people who say they support gay marriage would actually be happier with just civil partnerships. I suspect some of this is people not wanting to look bigoted – they don’t want to look homophobic so they say they support gay marriage even though they’d rather just have civil partnerships. Alternatively it may just be simple confusion between gay marriage and civil partnerships.

The broader picture, therefore, is that there is a very large majority of people who support legal recognition of gay relationships, and if people are forced to choose more people support gay marriage than oppose it… but if you don’t force people into that artificial yes-no there is a substantial minority of people who think the half-way house of civil partnership is enough.

Andrew Hawkins’ letter refers to the ComRes/Coalition For Marriage poll that found 70% of people agreed with the statement “Marriage should continue to be defined as a life-long exclusive commitment between a man and a woman”. I have written about this before here, and my view remains the same. I think it has been rather over interpreted. Perhaps the logical inference is that anyone agreeing to this statement must be opposed to the laws on marriage being changed so that same sex-couples can marry, but in this case we don’t have to infer, we have lots of other questions in other polls that actually asked directly about gay marriage, so we can be fairly certain that 70% of people are not interpreting the sentence that way.

Andrew speculates that online or offline fieldwork could explain the difference between different polls on gay marriage. In theory I agree with him – people may be embarrassed to admit to unfashionable views in a live telephone interview and may be more willing to openly admit them in an online poll. It would be good to see some directly comparable online and offline questions on gay marriage to see if it’s true in this case. However, almost all the polls I have mentioned above were conducted online anyway, including those by the traditionally telephone based researchers like ICM and Populus, so in this case the difference in results does seem to be mostly down to question wording.

So in summary,

(1) Polls asking if supporting gay marriage would make you more or less likely to vote Conservative show marginally more people saying less likely than more likely, but there are good reasons to be very cautious about questions asked in this format. People are not good at reporting what actually drives their voting behaviour. There is also the effect on the wider image of the Conservative party, whether it is seen as tolerant and in touch with modern Britain, which effects voters but is harder to pick up in polls, not to mention perceptions of Cameron and his leadership (as Lord Ashcroft wrote here, if Cameron did change his mind it wouldn’t just be the policy effect, but whether he would look weak, or like he was flip-flopping). I would be extremely, extremely cautious about drawing any conclusions about whether the net benefit of backing a policy is positive or negative.

(2) Taking the polling as a whole, all polls that ask a direct yes-no or agree/disagree question on gay marriage show that more people support it than oppose it. However, if you give people the option of gay marriage, just civil partnerships or neither, then in some cases marginally more people oppose gay marriage than support it. The only polls that show a majority opposed to gay marriage that I am aware of are those commissioned by groups campaigning against it, which tend to ask slightly different questions about whether marriage should be redefined, rather than asking direct questions about same-sex marriage.

(3) David Cameron wrongly quoted the ComRes poll in his letter, and it isn’t true to say that all polls however asked show that more people support gay marriage than oppose it. However, the broad thrust of his letter is basically right – the polls suggest that the British public are, generally speaking, pretty positive towards gay marriage.

Tomorrow we have some actual proper elections to look at – three by-elections, the new Police and Crime Commissioners across Wales and most of England, plus the Bristol mayoral election and the normal local government by-elections.

The three Parliamentary by-elections are the two caused by the resignation of veteran Labour MPs to contest the police elections (Alun Michael in Cardiff South and Penarth and Tony Lloyd in Manchester Central) and one in Corby caused by the resignation of Louise Mensch to go and live in the USA. The first two are safe Labour seats that promise little of interest – there is no obvious prospect of any upset and both will be easily held by Labour. The Corby by-election is more interesting, but just as easily predictable – the Conservative majority at the last election was only 3.6%, so with current national opinion polls showing a swing to Labour of around 9% they will win Corby at a stroll. If Corby behaves in line with national polling Labour should win with a majority of around 15%, in fact the polling of the seat commissioned by Lord Ashcroft suggests they will do even better than that, showing a Labour lead of 22 points.

The police elections are far less predictable. The elections take place in all English and Welsh police forces outside London, with each force electing a Police and Crime Commissioner using the Supplemental vote system (the same system used in the London mayoral election – you get a first and a second preference vote, if no candidate gets 50% on the first round all but the top two candidates are eliminated and the second preferences of people who voted for eliminated candidates are redistributed).

As I write there has not been any substantial polling of voting intentions in the police elections – there was a MORI poll that briefly asked about it, but given only around 150 people in the sample said they were certain to vote it won’t tell us much. Under normal circumstances we should expect Labour to do very well in the elections – it is a mid-term vote for positions that people don’t really understand, a perfect opportunity for a protest vote. On Thursday we have several unknowns – first is the extent to which people vote on the issue of crime, where the Conservatives have a traditional advantage. Second is the impact of the Liberal Democrats only contesting some seats. Third is the impact of independent candidates – you regularly get polls showing people like the idea of independent candidates, but in Parliamentary elections they invariably don’t vote for them. We shall see if people do end up voting in substantial numbers for non-party candidates.

Finally, there is the issue of turnout – both how it affects the results, and on the turnout figure itself. A lot of the media discussion in advance of the elections has been about how low turnout will be, whether it will be lower than the 23% recorded in the 1999 European elections, whether it will be as low as the 18% the Electoral Reform Society predicted. There have been a couple of polls asking whether people are likely to vote which have shown between 15% and 28% of people saying they are certain to vote, but don’t pay too much attention to that: turnout is remarkably difficult to predict from opinion polls (partly because the registers they use to work out turnout are not accurate in themselves, mostly because people tend to grossly overestimate their likelihood to vote – responses to the British Election Study are cross-checked against the marked electoral register to see if people actually did vote, and even amongst those people who say they are 10/10 certain to vote a good twenty percent don’t seem to actually do so.

Only Wiltshire police commissioner and the two safe Labour by-elections are counting overnight. Corby and all the other police elections are counting during the day on Friday, with the first results expected to turn up around lunchtime.