« EPP leaders threaten Cameron | Main | Tory MPs protest Anglicans' Israel boycott »


MORI (or Ipsos-MORI, as we must now call them) are something of a joke at the moment. Just like around the general election, their results fluctuate wildly all over the place. In my opinion the only pollsters worth taking notice of at the moment are ICM and YouGov; NOP having disappeared off the face of the earth since the election. (and I will say that even if they start reporting Labour leads again).

Yougov's findings are in line with what one might expect - the rush of Lib Dems to Labour after the woes of January was only temporary.

To agree with Gary, ICM and Yougov are more reliable than MORI aka the 'random number generator' (you may recall the 9 point lead they gave the Conservatives just two months ago, and indeed Michael Howard's five point lead at the very beginning of the 2005 campaign). MORI bizarrely decide to weight their results neither by past vote, nor by likelihood to vote. A quick glance at their recent history shows wild, inexplicable fluctuations in voting intention figures. I'll pick Yougov and ICM ahead of MORI any day!

The underlying good news of the poll is that people are prepared to listen to the Tories again. This is reflected in the 63% who think Cameron talks a good line.

The bad news is that the substance behind the nice man persona is going to be 18 months away while we conduct a thorough policy review. I have said it before but this takes us dangerously close to the general election.

Agreed. MORI's polling in the run-up to the election showed everything from a 5-point Tory lead to a 10-point Labour one. YouGov, if I recall correctly, were never about 2-3% from the actual result. And let us not let Sir Bob forget his (in)famous call that Kerry has won the Presidency on polling night itself...

The more disturbing statistic is the 63% who believe that Cameron is a flip-flopper. I think that the message is clear. The strategy of seeking to radically alter the party's perception has worked. It need go no further. Now, Cameron needs to adopt genuine Conservative positions and stick to them.

I think MORI need to look at themselves to be honest. Their results are illogical. Why newspapers are still using them, I don't know.


It might be interesting if we had a look at some of the more interesting polls rather than the straightforward "if there was an election tomorrow" kind.

Sometimes looking at the figures from different perspectives and looking at the different questions polled can give us hacks a greater insight into what is actually going on. The Times is always very good at doing this as they have done today. The DT is notable too when it gets Anthony King to comment.

Tracking your polls is sensible but a wider look at what is happening in the minds of voters would give the site and its analysis further gravitas.

"The strategy of seeking to radically alter the party's perception has worked. It need go no further. Now, Cameron needs to adopt genuine Conservative positions and stick to them."

Alex, if only. You must now be informed that to do so would be to truly flip-flop. To show you never truly believed or felt the policy of "change to win". Change is a perpetual process of reinvention and communication with the public in the modern, 21st century blah, blah...

Actually, there's an element of truth in that if we're going with this strategy. I don't think we'll see anything firm until next Christmas, and in the meantime we'll see great realignment. What's more, why put any beef out there before we know who's going to be leading the other two parties? I don't know how many people have seen Ashcroft's more recent polling but it clearly suggests that our realignment in voters' minds is not as profound as you might think.

What's more, though it is highly pleasing to see polls with us leading, especially having been used to getting a monthly kick in the teeth for a decade, nevertheless, we do actually have to do much better. We ought to be another five points up, really, at the expense of the Lib Dems. The reason we're not is the same old story still. It's the easy part that's been done...

Added to which we don't want the same old, same old, lurch to the right nonsense.

I think that the YouGov poll is a good result for The Tories, but it's even better than it looks.
Firstly, once again they posed the question: "Who would make the better prime minister?" putting only Cameron and Brown in the frame. As has been pointed out before, Lib-Dem supporters, without a Lib-Dem to vote for would be more likely to give Brown the nod.
I also take exception to the way they posed a couple of key questions.
Take "Cameron talks a good line, but it's hard to know whether there is any substance behind the words."
What response did they expect to such a loaded question? Any politician would get an affimative answer to that one. Indeed, a similar responce is shown to the statement: "It's hard to know what the government and Labour stand for at the moment.
Interestingly,when suggesting that Cameron "flip-flops", 36% say yes, 26% no and 38% say they don@t know. I think that's
petty good for the Tory leader. Any other politician would have got a similar response.
Perhaps the best news was on economic confidence. From a very poor position, the Tories are now almost level pegging with Labour.

I expect this uncertainly will continue until Cameron clearly states what the Tories believe in. Providing he maintains the popular core positions but puts more emphasis on other issues then the Tories should rise.

"Defence/foreign affairs/terrorism is regarded as the key issue facing the country (34% of the public spontaneously say this) ahead of the NHS (33%), race relations/immigration (30%) and education/schools (28%)."

Clearly immigration is not an issue that Cameron can ignore. However, the NHS and defence should receive more attention and there should be plenty of noise made about education. Not mentioned in the above article is the fact that crime also had a 28% importance rating. Note that issues such as global poverty and the environment barely registered. Maybe Cameron should think twice before emphasising these so much. The Tory Right should also note that the economy didn't come vry high up, suggesting that tax cuts may not be a priority unless they can be linked to public service reform.

I don't think any poll will really mean very much until we know who the runners and riders at the next election will be.

Re methodology - you require a meta-analytic methodology, but we would require something more than the point estimates produced in opinion poll headlines. There is the complication that these estimates are biased by the sampling methodology employed. These things matter so be wary of anyone who tells you just to smooth the estimates (as though they are simple observations) with a moving average type thing.

Note for all those who disagree with the line about 90 days detention and other measures the government may take how high up the list of priorities defence/foreign affairs/terrorism is, (though those who were against Iraq can take some cold comfort at the moment I admit.)

Having talked to various researchers at the HOC I have to say that almost all the letters MPs get on terrorism are generally supportive of the government's 'tough' (exc. NIreland) stance. And I found it hard to stomach when we voted with the LDems and Labour rebels against Blair - it may have pleased those in the Westminister bubble and avid Channel 4 News viewers but it didn't go down well in the country.

Terrorism is one of those 'touchstone' issues - it defines the image of politicians, because it is simple, and it is, to a certain extent, a gut reaction. How far are you prepared to go in balancing some freedoms against others? And this to many people sums up your ability to make reasoned judgements elsewhere.

Editor:It's tempting to over-interpret individual polls and so I'm thinking of starting a ConservativeHome poll of polls to accompany every report of individual polls. If anyone can give me advice on methodology for such an initiative I would be grateful...

A poll of polls? Even better, and which will save even more time for us busy hard-working Tories, how about a handy cut-out-and-keep thread of threads.

New opinion poll shews Conservatives ahead of Labour:
This is a personal triumph for Cameron and proves that he is moving the Party in the right direction; this is pretty much the same as all leaders have achieved in their honeymoon periods; does any one think we would be achieving this magnificent breakthrough with even more right-wing out of touch policies; this is less impressive than Cameron must have been hoping for and what if our traditional supporters stay at home.

New opinion poll shews Conservatives behind Labour:
This proves that Cameron has to work even harder to shed the out-of-touch hard-right image of the Party; this proves that people have seen through the Blairite spin; now is not the time for us to be attacking each other; now is the time for us to unveil revolutionary new policies.

etc etc

On methodology, I agree with Graeme - but isn't there some sort of standard industry method for sampling/weighting? Wouldn't that cut out some of Graeme's concerns, and allow you to go straight to the underlying data, weighting them by age of sample?

Real problem: with these figures you're likely to end up with spreads which overlap, eg Con 35-38; Lab 35-38 etc - which may not give the penetrating analysis we are looking for. Unless, of course, that's the story itself: no one is in the lead?

1AM, when has this Labour Government ever balanced freedoms? They love to crush freedoms wherever they can get away with it. The number of criminal offences introduced since 1997 is truly depressing.

John Major wanted a society "at ease with itself". Blair has created a society which is increasingly intolerant, suspicious, authoritarian and violent.

New Labour's New Britain is ugly and unpleasant.

I don't doubt that it's going to be an uphill struggle to convince Mr and Mrs Voter that we should vote against the govt on matters of liberty over draconian increases in state power. But I think this is one of those touchstone issues absolutely - my gut instinct is that it's never right to lock people up without recourse to trial, whether that's a terrorist subject, or a mental health patient. I think Labour will maintain a lead on this for the same reason we maintained a lead on "asylum" - it's easy to get tabloid endorsement ("kill them! kill them all!"). It's less easy to unpick a mess of overwhelming state power.

We have a lot of debates on this site about basic party direction, a lot of posts come from people who are concerned about what they see as the dilution of our commitment to cut taxes, a change made for what I think are sound electoral reasons. Voting against the govt on the 90 day matter, and the rest, is my own touchstone, and one of the reasons I love the new dispensation. I hated it when we gave off signals that we were quite OK with ID cards, actually, and not too concerned about locking people up endlessly. Everyone has their own party axioms of course, I'm not presuming to tell others what they should be, but mine is that we're a party of liberty or we're nothing. (Not a libertarian party :-0)).

William - I know absolutely nothing about sampling done by pollsters so this is prob irrelevant, but if we had actual random samples then the combination ought to increase the precision not reduce it.

EG if you had a binomial sample of size n where each response could be either conservative or labour (simplification), and a proportion p supported conservatives, where the unbiased estimate of p would be given by x/n, where x is the no. who responded conservative out of the n in the random sample, then the estimated variance of the estimate of p (which gives the "spread") is n*(x/n)*(1-x/n), which is a decreasing function of n, ie if you *could* combine n1 and n2 then the spread would be narrower.

The difficulty I lack is that I don't know how you combine non-random samples, nor even how you estimate the variance within one. A Bayesian estimator is biased by definition, so maybe you could take the first sample as the prior and then update the posterior for the proportion parameters with every new poll? But there must be someone who reads this who knows about sampling methodology!

MORI should be ignored, YouGov is excellent, and its poll results predict a comfortable fourth term for Labour, which is nothing short of disastrous for Cameron.

I give the whole 'Cam Boy Conservative' Project another six months.

'Blair has created a society which is increasingly intolerant, suspicious, authoritarian and violent.'

Surely this has come about, at least in part, as a result of too much freedom for those people who have not learned/grasped what personal responsibility means and confuse 'freedom' with 'free-for-all', and the depressing thing is I don't know how one educates people about personal responsibility. Obviously some schools and some families are aware of the importance of this particular issue, but there are a fair number of families and schools who either don't have a clue what 'personal responsibility' is or don't see it as in their remit to inculcate. And when the young adult leaves school, who is to inform him/her then about his/her responsibility to society, certainly not the media.

"predict a comfortable fourth term for Labour, which is nothing short of disastrous for Cameron"

We're 2 points ahead of Labour when 63% of respondents don't actually know what we stand for... I say that's a good result for all of us. Plus if you run the figures the results shown (with the remaining being for "Other" parties) would turn out a Labour majority of 18. Hardly a comfortable 4th term is it?

I agree that we should see more of what DC and Conservatives stands for but to come up with policies now will just see them stolen by Labour as they've done for the past 8 years.

Cameron cannot be blaimed for why the Tories cannot win. The brand which is highly damaged by Thatcher and Major, and then the ERM will prove difficult to correct.

Also note that the labour government really had not had any "massive" scadals except for Iraq.. The economy has gone suprising well..and people are comfortable with Labour being incharge..

Yougov are always better than the others and I would give no credence to any MORI poll on anything. All the polls however are still Labour biased, at the last election it was about three percent.

The loaded question on what DC stands for was matched by the one for the government - people in very large numbers seem prepared to vote for one of the major parties despite being clear that they don't know what the party stands for. The sustained peformance in the high thirties is creditable and if you look at the local government results things look encouraging for May.

Brown has carefully deliberately and quite wickedly constructed a large client state. Voters in areas where 70% of the income comes from the public sector are naturally worried about the benefits of freedom. This government is all about fear, intimidation and controlling the purse strings and we have to face up to that and overcome it, not by hectoring but by optimism and persuasion.

We won't get past the magic 42% until we can clearly show we can reduce taxes and improve services and provide better opportuntities for waste recycling diversity coordinators - annoying, frustrating and a slow process but could the true believers have a bit of patience, say 18 months worth. If we're not at 43% in a Yougov poll by August 2007 then I will agree that the British Public enjoy wallowing in gentle decline a la 1976. Unitl then consistently strong polling evidence shows we are making slow and steady progress.

Also, the Sun have used a MORI poll because they always like to use unreliable statisitcs if it boosts T Blair; if this will carry over to GB I don't know, and I suspect neither do they.

Jaz are you ACTUALLY a Conservative? You only seem happy when we ditch Conservative values and policies and now you pronounce that our brand was damaged by Margaret Thatcher (you will recall she won 3 elections handsomely and changed Britain for the better)...

"Cameron cannot be blaimed for why the Tories cannot win."

Indeed, the electoral system skewed in favour of Labour can (but I'd choose it over PR anyday). We just need the boundary commission to sort things out.

I am a very strong supporter of Cameron and I campaigned for him to be leader. I think we are right to go for the swing voters in the centre ground....however we are at the beginning of a critical phase now. It is not so much detailed policy we need (which Labour will only steal and which we are still developing) BUT we have to show a golden thread/theme that runs through our new position that defines us simply and clearly so that the public can say "thats what modern Conservatives stand for". This is lacking. It must be put in place or we will get into trouble sooner rather than later.


Just to point out, it doesn't matter what we think - that the MORI poll is not accurate - but what Sun readers see and believe. If the Sun tells their readers that Cameron's Conservatives are behind Labour, the Conservative vote drops off. In this way, opinion polls are not a reflection of opinion, but a means of swaying opinion. It's completely true to say it's the Sun 'wot won it'. The constant drip-drip effect on the readership will affect opinion, and since the Sun makes up a sizeable chunk of the electorate - if the Sun is against Cameron/Conservatives, then it's bad news. Don't underestimate or dismiss the power and influence of the Sun and everything within its covers.

The Sun merely tells its readers what they want to hear - hence why it's the biggest selling newspaper.

The Sun doesn't win it, but it does always back the winner.

Polls will go up, polls will go down, some will be good and some will be bad. We have settled for the right leader and we must stick with him because he is the right leader, now we must think through carefully the platform we want to stand on and stick with that.
The Conservative Party have lacked two things in recent years that as held it back, loyalty and courage. Victory will only come if it truely rediscovers both.

The Sun doesn't win it, but it does always back the winner.

I don't think this is true. While the newspaper will almost certainly declare for the party likely to win the election, and this declaration will not in itself swing the election, its earlier support can be critical in making or breaking a leader and a party, and thus determining will later be in a position to win.

Both Neil Kinnock and William Hague suffered from week in and week out drubbings in The Sun. The credibility of both men was mortally wounded in this way long before they got near a general election.

The Conservative Party have lacked two things in recent years that as held it back...

It has however always been able to use the letter "h".

Indeed, the electoral system skewed in favour of Labour can (but I'd choose it over PR anyday). We just need the boundary commission to sort things out.
The demographic changes behind this are ongoing so inbetween boundary modifications the situation will become increasingly pro Labour until the next set of modifications until the trends behind it end, that said there was a bit of an effect due to Norman Fowler's belief that boundaries were just a diversion whereas Labour and the Liberal Democrats planned nationally their approach to maximise their numbers of seats by arranging locally with Conservative Councils and MP's to campaign for Conservative Wards in many seats to be moved into neighbouring safe Conservative seats hence creating a safer Conservative seat and a potential Labour or Liberal Democrat gain.

Trends change though and if the difference between the 2 parties is the same in an election in 40 years time as it had been in 2005 maybe the Conservative Party would get a majority just as in 1951 the Conservative Party won a majority with fewer votes than Labour got.

As for polls, polls are polls they are not results and people may lie to pollsters and whatever the pollsters might say there is no way of eliminating the chance of this skewing what the polls say.

Graeme Archer: agreed, larger random samples are more reliable. It's not really my field, but I don't think opinion polls are purely random and instead try to satisfy a "representative cross-section", so you don't end up with the views of 400 stock brokers from Surrey, say. However, if you know Group A is a good sample of the whole population, and Group B is also a good sample of the population, couldn't you treat them in the same way (whatever that is?).

My point about spreads wasn't strictly the reliability of the sample, but it's age.
* Opinion Poll A on Monday from 100 people gives Labour 40%;
* Opinion Poll B on Friday from 50 people gives Labour 30%

Does that mean:
(a) Labour have lost 10% in less than a week; or
(b) Labour's "true" support is actually 35% (the arithmetic average of the two); or
(c) do you give greater weight to the one with the bigger sample - so the average for Labour is 36.7%; or
(d) do you give greater weight to the most recent result - ignoring anything over 7 days old - so the average Labour support is, say, 33%; or
(e) a combination of (c) + (d), giving 34.6%.

All you really know is that Labour support is somewhere between 30% and 40%. I can't see how to accommodate this without using a spread estimate (without creating even more spurious precision than exists with opinion polls as it is). Over to you.

I remember opinion polls giving the Labour Party 63% and the Conservatives 19% in the 1990's and yet I doubt that there was ever a time in the last 35 years in which Labour would have got more than 45% of the vote (or less than 20% of the vote) in a General Election and even in 1951 they didn't quite reach 50% of the vote, Labour has never had more than 50% of the vote in any set of mass elections, on the other hand equally the Conservative Party have only once ever got less than 30% of the vote in a General Election and that was in the time of the arguments over the Corn Laws.

William you are right if opinion is very volatile but looking at the monthly polls this doesn't currently appear to been the case (a bit of volatility around LDs recently I accept).

In UK difficulty is that the published results have been amended by past voting statements, likelihood of voting intent & other factors that are not consistent across organisations. Simply averaging polls without knowing the underlying data is therefore not good practice.

But I did it anyway to see what average Tory lead or deficit was monthly since general election:
M -7%, J -8% J -11% A -7% S -7% O -7% N-6% D +2% J 0% F 0%.
Tory share of voting intent (rounded):
M -31%, J 31% J -30% A 32% S 32% O 32% N33% D 37% J 38% F 37%.
So on average we were behind between 7-8% before Cameron and up to 1% ahead after his election and our vote share grew from around 32% to just above 37%.
Not bad - not a massive honeymoon leap but two thirds of the way to stopping a fourth term Labour Government.

I also compared average to YouGov Telegraph monthly poll - very close correalation between average of all and YouGov. YouGov av lead before DC = 7%, YouGov after = 1%

In answer to your question (and thirty years plus on from O level maths) if you know anyone with more recent statisical knowledge I'd sugest a rolling average of last 4 or so polls based on sample size and from that number intending to vote Cons, Lab, LD or other.

So if there were three polls sampling 1000, 1500 & 2000 with 35% voting conservative on first, 40% on second and 37.5% on third you would get total sample of 4500 with total of 1700 voting Tory = 37.8% average. Next poll shows 39% on sample of 1500 so you drop the first poll and new rolling average is 38.7% (1935 of 5000).

Using a rolling average flattens out variability - it might lag a bit if trends change suddenly but with 4 to 6 polls a month that's only really important nearer an election.

Above posts passing the time while uploading large files for sharing on the web - shame DSL isn't as fast uploading as downloading....

Ted: a rolling average on the basis of the last four polls works if they all come out at regular intervals. Probably better to average polls coming out within a fixed period of time - and the sensible choice would be the length of a general election campaign. On the example of 2005, that would be 25 days (proclamation to close of polls).

That still leaves us with the problem of non-comparable samples which Graeme has highlighted - and Ted pointed out himself the latest poll came from a group of people which were 50% Labour voters.

I'm embarrassed to admit that I have a doctorate in mathematical statistics; embarrassed for ever so many reasons, but mainly at the moment because I can't remember anything I was ever taught about sample surveys. I can't even remember why the pollsters DO collect biased samples, they must aim to be increasing variance at the expense of bias (all estimators have a bias-variance trade-off). [BTW William, it's to avoid getting 400 surrey stockbrokers that random samples are most useful; the polling organisations have suboptimal samples (more biased) because (I hope) their estimates have lower sampling variance (so the precision is higher). But they don't use best-least-unbiased estimators, which for a frequentist statistician (which I ain't) are the "best" ie they have zero bias - so as n tends to infinity the estimate of any party's support tends to the true value in the proportion - this is not a property that any of these polls can have, being non-random; but they do have the property that the standard error of the estimated proportion is smaller than the unbiased one, which makes them more useful for between-party comparisons in any one poll (so long as one understands the likely bias in the point estimates supplied; which it would appear we do not!).

Whatever properties individual polls possess, there is one thing that might help, which is that presumably each company uses the same methodology each month, so that we could look at a rolling average of Tory-Labour contrasts per month. So each month, you would have P estimates of Tory-Labour contrasts, where P is the no. of companies, and you could assume that these come from a hyper-distribution - a sort of hierarchical model, where each company supplies an estimate. So what I'm saying in concrete terms is I would start by plotting the Tory-Labour differences for each month, labelled by company. Is company 1 always estimate a positive value? negative? Should become apparent over time (cf the between-company monthly average). If there was a random scatter around c, a constant, you would feel a lot more comfortable about (1) assuming the multilevel model framework and (2) ignoring it and inefficiently estimating the Tory-Labour contrast just by averaging across companies (ie assuming a best estimate of c). If it doesn't look like a random scatter around c, umm..

Could do a bland-altman plot too without invoking any complicated modelling to investigate semi-formally the degree of agreement or systematic difference in both the estimate of the Tory-Labour contrast or the variability in those estimates (ie which are more reliable). I only know how to do this in a pairwise (between company) fashion though.

You're the expert but I would think a simple rolling average would work as methodologies within each remain similiar (William - the point about NOW poll is it isn't weighted - all monthly polls are).

If you look at Anthony Wells site http://www.ukpollingreport.co.uk/blog/?page_id=18 which lists all the polls on voting intention you can see that there are regular polls from YouGov/Telegraph, Ipsos-MORI/the Sun & Observer,ICM/Guardian, Populus/Times, BPIX/Mail on Sunday. Most months there is an ICM, Mori,Populus & YouGov poll. Running a rolling average based on those four would mean replacing the older of the set as each new one comes out.

The comments to this entry are closed.



ConHome on Twitter

    follow me on Twitter

    Conservative blogs

    Today's public spending saving

    New on other blogs

    • Receive our daily email
      Enter your details below:

    • Tracker 2
    • Extreme Tracker