The Ukraine invasion, of course, hit energy prices which didn't help. Again, the government bears no blame for this, their response was woeful and inadequate.
And, the inadequate response is itself in part a consequence of prior decisions by the government since 2010. The decisions to effectively prevent development of renewable generation in England and Wales meant that electricity generation was more dependent upon gas than it needed to be, with greater impacts on electricity prices. The decisions to drop the requirements for thermal efficiency for new homes from 2015 left hundreds of thousands of people who had bought a new build home needing to burn a lot of gas to keep warm. Failures to invest in insulation of existing homes did much the same. Measures that could have been taken to improve public transport had left more people than should have been the case dependent upon cars, who then took a hit with petrol price rises. Volatility in fossil fuel prices and the dependence of Europe on Russian gas weren't suddenly discovered the day Russian tanks rolled further into Ukraine, these were well known factors that governments world wide should have already been taking steps to address - the UK government isn't unique in failing to do so and getting caught with their proverbial pants down, but not being the only government to fail their people over this isn't an excuse.
So well put @Alan Cresswell . I found Shapps more than-the-usual-level of irritating today. No one is saying that external events didn't happen, nor that they had no ill effects. But to claim they had good responses when they clearly didn't. And the chutzpah of this when they've spent the past decade and a half blaming Labour for the consequences of an external event that was heroically well-managed.
So well put @Alan Cresswell . I found Shapps more than-the-usual-level of irritating today. No one is saying that external events didn't happen, nor that they had no ill effects. But to claim they had good responses when they clearly didn't. And the chutzpah of this when they've spent the past decade and a half blaming Labour for the consequences of an external event that was heroically well-managed.
Not long now, please God...
AFZ
Three weeks, or thereabouts.
Some commentators, at least, are hoping for a sane government, which seems not unreasonable, however much Sid & Doris Bonkers and their Reform ilk might disagree.
Even if it was true, it's a problematic position. In 2010 - and more importantly, ever since - he and his party have weaponised a once-in-two-generation, international financial crash against the Labour Party. If we simply hold them to their own standards, then everyone is entitled to weaponise the pandemic against them.
I was just looking up which seat is the "Tipping point" for Labour - i.e. if they win this seat then assuming anything close to uniform swing, then Labour will secure at least 326 seats and therefore have an overall majority.
Annoyingly, it's not a straight-forward question. Suspensions and resignation from the party and boundary changes makes it not a straight-forward question. In theory, Labour had 205 MPs when Parliament was dissolved and so would need to gain 121 seats. This Guardian article, allowing for boundary changes, suggests the number is 126. It may be slightly higher as there are one or two seats in which they are vulnerable. They may lose a Bristol seat to the Green Party, for example.
Anyway, if you order the target seats by size of majority, as done here, then the answer is Colchester. Electoral Calculus currently has Labour with a 84% chance of winning in Colchester.
Or to put it another way,
These seats are the ones to look for on election day. If Labour are winning these seats then a majority is very likely:
York Outer
Rossendale and Darwen
Wolverhampton North East
Macclesfield
Blackpool North and Fleetwood
Monmouthshire
Glasgow West
Scarborough and Whitby
Dunfermline and Dollar
Welwyn Hatfield
Hitchin
Dunbartonshire West
Bolton West
Scunthorpe
Carlisle
Erewash
Bournemouth West
Edinburgh North and Leith
Earley and Woodley
Glenrothes and Mid Fife Colchester Stockton West
Edinburgh East and Musselburgh
Hexham
Ossett and Denby Dale
Middlesbrough South and Cleveland East
Dwyfor Meirionnydd
Doncaster East and the Isle of Axholme
Great Grimsby and Cleethorpes
Pendle and Clitheroe
Paisley and Renfrewshire South
Basingstoke
Bathgate and Linlithgow
Dover and Deal
Penrith and Solway
Cumbernauld and Kirkintilloch
Telford
Paisley and Renfrewshire North
East Kilbride and Strathaven
Livingston
That's a good thing in itself but doesn't necessarily indicate an economy on the up - falling inflation often indicates an economic slowdown or recession (I believe the latest figures indicate flat GDP).
On the subject of the NHS, this is a King's Fund (independent healthcare research group) report.
Multi-year funding increases and a series of reforms resulted in major improvements in NHS performance between 2000 and 2010, but performance has declined since 2010 as a result of much lower funding increases, limited funds for capital investment, and neglect of workforce planning. Constraints on social care spending has also resulted in fewer people receiving publicly funded social care and a repeated cycle of governments promising to reform social care but failing to do so.
On the subject of the NHS, this is a King's Fund (independent healthcare research group) report.
Multi-year funding increases and a series of reforms resulted in major improvements in NHS performance between 2000 and 2010, but performance has declined since 2010 as a result of much lower funding increases, limited funds for capital investment, and neglect of workforce planning. Constraints on social care spending has also resulted in fewer people receiving publicly funded social care and a repeated cycle of governments promising to reform social care but failing to do so.
The worrying thing there is that it's still 'lower funding increases' rather than 'no funding increases' - ie performance has (understandably) dropped off as we've not shovelled increases of cash at the problem to the same extent as was done 2000-2010. And health's one of the protected departments.
There is now a workforce plan, but social care as noted remains a skip fire.
There really needs to be an honest conversation about what we want to fund in this country, where the money's going to come from, how much it's going to cost, and how things are going to be done differently.
On the subject of the NHS, this is a King's Fund (independent healthcare research group) report.
Multi-year funding increases and a series of reforms resulted in major improvements in NHS performance between 2000 and 2010, but performance has declined since 2010 as a result of much lower funding increases, limited funds for capital investment, and neglect of workforce planning. Constraints on social care spending has also resulted in fewer people receiving publicly funded social care and a repeated cycle of governments promising to reform social care but failing to do so.
The worrying thing there is that it's still 'lower funding increases' rather than 'no funding increases' - ie performance has (understandably) dropped off as we've not shovelled increases of cash at the problem to the same extent as was done 2000-2010. And health's one of the protected departments.
There is now a workforce plan, but social care as noted remains a skip fire.
There really needs to be an honest conversation about what we want to fund in this country, where the money's going to come from, how much it's going to cost, and how things are going to be done differently.
Indeed.
With an aging population and new treatments etc. healthcare costs will go up in the medium term. Healthcare inflation is also higher than general inflation. These things are all known by economists.
The point is that, as a country, we will spend more on healthcare in the next decade. The question is how we spend it. Fixing the NHS is the cheapest AND the most equitable way to do it. Not to mention the politically most palatable. Hence, for me, it's a no-brainer.
Of course part of how you fix the NHS is to fix social care. I've seen serious estimates that this will cost £5-£10Bn. But you get a lot back in NHS savings so net is quite a bit less. We've still got to find a few billion though...
On the subject of the NHS, this is a King's Fund (independent healthcare research group) report.
Multi-year funding increases and a series of reforms resulted in major improvements in NHS performance between 2000 and 2010, but performance has declined since 2010 as a result of much lower funding increases, limited funds for capital investment, and neglect of workforce planning. Constraints on social care spending has also resulted in fewer people receiving publicly funded social care and a repeated cycle of governments promising to reform social care but failing to do so.
The worrying thing there is that it's still 'lower funding increases' rather than 'no funding increases' - ie performance has (understandably) dropped off as we've not shovelled increases of cash at the problem to the same extent as was done 2000-2010. And health's one of the protected departments.
There is now a workforce plan, but social care as noted remains a skip fire.
There really needs to be an honest conversation about what we want to fund in this country, where the money's going to come from, how much it's going to cost, and how things are going to be done differently.
Pesky researchers finding new treatments that cost money and keep people alive so they can go on to need new expensive treatments.
I think that healthcare is inevitably going to get more expensive.
It can be partly ameliorate by alleviating poverty which has known correlations to ill health, and similarly drug and alcohol services. Reducing the hospital revolving door problem by sorting out our care system. All will require considerable investment for long term benefit.
It's a shame the 15 minute city has become mired in conspiracy theory and diverting money from road to encouraging active travel and subsidising public transport is electorially a hard sell as enabling people to not find themselves having to drive everywhere would have a positive long term health impact too.
To illustrate my point, think of Roy Jenkins and Harriet Harman, and the radical social policy (for the time) that they espoused.
Although nominally 'centrists', on these issues they were far, far, far to the left of a typical Labour voter in a pub in Salford or Scunthorpe. This is something many working-class voters have cottoned on to, and now claim Labour does not represent them. The question is, who does? The answer is, honestly, no one. Certainly not the tinpot fascists of Reform.
Er... you might have to elucidate a bit @Sighthound as I am only just coming up on 50 so was not very politically aware when "Woy" was in his prime... and even on Harriet Harman it would be helpful to have it pointed out exactly what her "radical social policy" was...
To be honest, the social changes - the end of the death penalty, legislation of homosexuality and general equality legislation - are so mainstream now that I believe they are accepted by everyone bar a few extreme-right reactionaries. I would be the last to decry them. Jenkins and Harmon and their ilk were right.
However, at the time, this social liberalism - for want of a better term - was not universally popular among the working-class community.
I know it is dangerous to generalise about groups of people, but even now there are a fair few working-class people who are very socially conservative - they like 'traditional families' and would like children to be caned in school and perhaps even conscripted into the Army. They tend not to like immigration or multi-culturalism, except for their mates Errol and Abdul who are decent chaps. However, these very same people are left-wing in economic terms and approve heartily in public services like the NHS and the government spending capital on infrastructure programmes. These sort of voters - often also flag wavers and admirers of the monarchy - do not fit any party's template. They were attracted to 'Boris' because his con nodded towards them. And now you can see how some of this group are attracted to Reform.
To illustrate my point, think of Roy Jenkins and Harriet Harman, and the radical social policy (for the time) that they espoused.
Although nominally 'centrists', on these issues they were far, far, far to the left of a typical Labour voter in a pub in Salford or Scunthorpe. This is something many working-class voters have cottoned on to, and now claim Labour does not represent them. The question is, who does? The answer is, honestly, no one. Certainly not the tinpot fascists of Reform.
Er... you might have to elucidate a bit @Sighthound as I am only just coming up on 50 so was not very politically aware when "Woy" was in his prime... and even on Harriet Harman it would be helpful to have it pointed out exactly what her "radical social policy" was...
To be honest, the social changes - the end of the death penalty, legislation of homosexuality and general equality legislation - are so mainstream now that I believe they are accepted by everyone bar a few extreme-right reactionaries. I would be the last to decry them. Jenkins and Harmon and their ilk were right.
However, at the time, this social liberalism - for want of a better term - was not universally popular among the working-class community.
I know it is dangerous to generalise about groups of people, but even now there are a fair few working-class people who are very socially conservative - they like 'traditional families' and would like children to be caned in school and perhaps even conscripted into the Army. They tend not to like immigration or multi-culturalism, except for their mates Errol and Abdul who are decent chaps. However, these very same people are left-wing in economic terms and approve heartily in public services like the NHS and the government spending capital on infrastructure programmes. These sort of voters - often also flag wavers and admirers of the monarchy - do not fit any party's template. They were attracted to 'Boris' because his con nodded towards them. And now you can see how some of this group are attracted to Reform.
I think a lot of those have gone back to Labour. It depends which is more important to them - less immigration and what they consider "woke", or functioning services.
However, at the time, this social liberalism - for want of a better term - was not universally popular among the working-class community.
I know it is dangerous to generalise about groups of people, but even now there are a fair few working-class people who are very socially conservative - they like 'traditional families' and would like children to be caned in school and perhaps even conscripted into the Army.
Yes, and there was always a section of the working class who were traditionally Tory voters on this basis, but as things like the British Social Attitudes survey has chosen this group has generally become older and the country as a whole - including the working class - have become much more liberal on social issues over that time.
They tend not to like immigration or multi-culturalism,
Concerns around immigration are shared more widely than - say - concerns over the traditional family/conscription - but that this is varied quite a bit over time demonstrates that attitudes here are nowhere near as sticky as sometimes made out.
I know it is dangerous to generalise about groups of people, but even now there are a fair few working-class people who are very socially conservative - they like 'traditional families' and would like children to be caned in school and perhaps even conscripted into the Army. They tend not to like immigration or multi-culturalism, except for their mates Errol and Abdul who are decent chaps. However, these very same people are left-wing in economic terms and approve heartily in public services like the NHS and the government spending capital on infrastructure programmes. These sort of voters - often also flag wavers and admirers of the monarchy - do not fit any party's template. They were attracted to 'Boris' because his con nodded towards them. And now you can see how some of this group are attracted to Reform.
Concerns around immigration are shared more widely than - say - concerns over the traditional family/conscription - but that this is varied quite a bit over time demonstrates that attitudes here are nowhere near as sticky as sometimes made out.
I would add that in many Western countries, many working- and lower-middle-class social conservatives (on issues of sexuality, gender, the traditional notion of marriage and family, and reproduction), are themselves from immigrant backgrounds (of any race or religion), non-white backgrounds, or non-Christian backgrounds (of any race).
You're right that the working class of all cultural backgrounds is often supportive of government interventions to help the less fortunate, provide better public services, rein in the influence of the super-wealthy, etc., but a lot people from traditional societies (many of whom are immigrants, non-white, and/or non-Christian) living in the West feel also agree with statements like "hard work pays off", "laziness results in failure", "families need to live together and take care of their children, elderly, and sick/disabled", "religious and traditional cultural associations should play a large role in providing aid and conflict resolution in communities," etc.
And a lot of immigrants come from countries where the government is oppressive, deeply corrupt, or both, and are very distrusting of any politician promising big changes in society. A few of them a particularly distrustful of anything with the traditional language or iconography of socialism (even if they support many left-to-center-left policies) because of associations with politicians or parties from which they or their families immigrated.
Progressives should try to include in their image of social conservatives some, but not all (it is a different form of bigotry to assume all members of these groups are the same) African Pentecostals, South Asian Muslims and Hindus, Polish Catholics, Ultra-Orthodox Jews, etc.
And a lot of immigrants come from countries where the government is oppressive, deeply corrupt, or both, and are very distrusting of any politician promising big changes in society. A few of them a particularly distrustful of anything with the traditional language or iconography of socialism (even if they support many left-to-center-left policies) because of associations with politicians or parties from which they or their families immigrated.
My emphasis, what this analysis misses is that there are equally a reasonable number of first and second immigrants who - precisely because of their background - are much more literate regarding socialist/social-democratic policies, and who support left policies from a position of knowledge.
And a lot of immigrants come from countries where the government is oppressive, deeply corrupt, or both, and are very distrusting of any politician promising big changes in society. A few of them a particularly distrustful of anything with the traditional language or iconography of socialism (even if they support many left-to-center-left policies) because of associations with politicians or parties from which they or their families immigrated.
My emphasis, what this analysis misses is that there are equally a reasonable number of first and second immigrants who - precisely because of their background - are much more literate regarding socialist/social-democratic policies, and who support left policies from a position of knowledge.
Maybe this is truer in the UK where more immigrants come from countries in the EU with more social democratic policies than it is in the US? Or maybe it is easier for people who come from more socialist countries in the “global South” (I’m never comfortable using that term) and who aren’t political dissidents or wealthier people who are strongly anti socialist to immigrate to the UK than it is to the US (it was very hard for people from India or China to immigrate to the US even before Trump)?
I should also amend when I said the beliefs “hard work pays off” and “laziness results in failure” I should have said “hard work should pay off” and “laziness should result in failure”. I didn’t mean to suggest that most immigrants believe that everyone can pull themselves up by their own bootstraps, as some right wing politicians suggest.
Or maybe it is easier for people who come from more socialist countries in the “global South” (I’m never comfortable using that term) and who aren’t political dissidents or wealthier people who are strongly anti socialist to immigrate to the UK than it is to the US (it was very hard for people from India or China to immigrate to the US even before Trump)?
There was a lot of immigration of all classes of people from Commonwealth countries into the UK (which I presume is germane given the topic of the thread).
Or maybe it is easier for people who come from more socialist countries in the “global South” (I’m never comfortable using that term) and who aren’t political dissidents or wealthier people who are strongly anti socialist to immigrate to the UK than it is to the US (it was very hard for people from India or China to immigrate to the US even before Trump)?
There was a lot of immigration of all classes of people from Commonwealth countries into the UK (which I presume is germane given the topic of the thread).
Something I've been meaning to ask is whether any news sources attempt to analyze the voting patterns of different demographic groups (race, ethnicity, gender, income, level of education, religious identity/level of religious practice, etc.) I see some general discussion along the lines of "this constituency has become more competitive for Labour as it has become more multiethnic" or "this constituency is full of affluent Remainers who used to vote Tory but now might vote LibDem", etc., but I never seem to see opinion polling or exit polling that says "this percent of...black voters/single women/voters without a university degree/voters who pray daily or attend religious services at least once a week...intend to vote for or say they voted for this party". The news in the US is full of this kind of opinion polling. I know it's harder to do at the constituency level, which is where is matters for a parliamentary election, but it just feels like a lot of the political analysis is only based on the party vote distribution from the last election and opinion polling of the general public in a constituency, rather than on breaking voters down into demographics. Is this true? Or am I just not looking at the news sources that have this kind of analysis?
You have to remember UK demographics are very different. It can be done by ethnicity (and is) but not hugely routinely simply because (as your post suggests) it tends to swing only certain seats. The UK remains a place (though things are changing) where overall nearly 8/10 people are white.
You can do voting intention of the 2/10 but you need a big sample and typically within a nationally representative 2000 or so (MRP is a different kettle of fish) the numbers will be necessarily small. Which pushes to a specific minority poll. Which someone has to pay for - and people do, obviously - but not generally for run of the mill ‘state of the polls today’ polling.
If you look at the polling tables (which I do occasionally*), the polling companies collect all this data.
Religious affiliation does not get discussed very much in the media, in these terms though. I suspect mostly because, socioeconomic status and level of education is a better predictor of voting intention.
Occasionally, you might here reference to religious groups. I know that most Muslims vote Labour. Members of the Sikh religion do as well, and there were targettes campaigns - in 2016 - to persuade them to vote against Kahn for London mayor.
In the Bexit refendum, level of education had a strong correlation with voting intention.
AFZ
*I am sad enough to be interested in methodology and data. I am not quite up to speed on current methods but all sorts of things have been tried. In a bit I will write a brief summary.
There are well known (though not necessarily accurate!) demographic trends in voting intention. It's generally thought that students are more likely to vote for left of centre parties, more economically deprived people are less likely to vote at all etc. Polling at constituency level is almost always too small a sample to be able to produce any level of accuracy of how things are going to go. But, combining regional polling data with the constituency demographics does allow the larger samples of voting intentions to be projected onto specific locations (I've not seen this done at ward level for local elections, but it's certainly done for constituencies in parliamentary elections). Political parties, who may have their own polls but will all be looking at published polls, will be doing their own analyses of those polls to decide where to concentrate and plan their campaigning activity, and those analyses will include demographics and local knowledge - though parties won't be sharing that information.
I apologise if no one other than me is interested, but this is a basic explanation of polling and how to interpret and understand them, including the pitfalls and problems.
George Gallup, the pioneer of polling allegedly said:
You only need a single spoon to sample the soup, provided it is well stirred.
This scientific and statistical concept is used all the time in all sorts of contexts. If sampling is representative, then it gives you the answer. An opinion poll is no exception. The idea is to sample enough people to understand how the whole population will behave. Indeed, if a sample is representative, it will do so.
1. Why 1000 people? There is some clever maths behind this and it comes down to something called Central Limit Theorem which is really useful for lots of things. Leaving aside the theoretic maths, the key bit is this: If you sample 100 people your estimate of voting intention with have ~10% margin of error. If you sample 1000 then it's ~3% and to get to ~1%, you need 10,000 people in your sample.
The vast majority of polls use a sample size of around 1000. This reflects a balance between getting as accurate an answer as possible and the costs and other limitations of sample size. Contacting 1000 people can be expensive. Contacting 10,000 will cost 10x as much. Moreover, it takes more time. Polls are a snapshot and if they take too long to collect then they will be potentially distorted by changes over time.
If you look at any poll they will state a margin of error. So when it says that 45% of people will vote for Candidate X what they actually mean, in precise, statistical terms, is that we are 95% confident that the true answer lies between 42% and 48%.
This is really important when polls are close. For the most part when people say "polls are wrong" they mean "I don't understand how margin of error works." The problem for political polls is that the result is binary. In a straight-fight between two presidential candidates - like the French second round - then the candidate with 50.00001% wins and the candidate with 49.99999% loses. It would be expected - in fact, almost inevitable - in that situation that some of the polls would have predicted the loser to win.
2. Representative Sampling?
Political polls face a specific challenge around making sure the sample is representative. This is easy to understand - if you took a survey in Mayfair, you would get a very different answer to one taken in Tower Hamlets. The populations are different. So how do you correct for this, to make your sample representative?
This is where polling companies differ. There are lots of factors you might use. There is a problem in that the more factors you use, the more you compromise the statistical power of the survey (making your margin or error effectively higher). What polling companies are always looking for is a magic formula that gives them the answer. Part of the challenge is that this changes over time.
Forty years ago, most polls were conducted by phone, using landline numbers, randomly selected. Now, in 2024, that is a very narrow, very specific population, you end up sampling. Modern polls are almost exclusively drawn from online panels with demographic factors already known. In the UK, demographics are not great correlators with voting intention so it's still not perfect.
So, one of the things you can use is previous voting behaviour. This one is quite nice because we know those numbers from the previous election. Here's a simplistic, worked example.
Poll of 1000 people; If there was an election tomorrow, how would you vote? Red Party: 800
Blue Party: 200
So, 80% / 20%. Job done. But what if, our sample is not representative and we'll over-sampling people who normally vote Red?
So we ask the question, who did you vote for last time? and compare that to the result of the election. In the previous election the vote was Blue 60%, Red 40%. So we know that a truly representative sample would include 60% who voted Blue and 40% who voted Red last time.
In our survey, the people we sampled voted Blue 30% and Red 70%. Ahhh, we may have a bias here! So we need to build a contingency table and do some correction. (Just a table of numbers).
Voting intention……Voted last time: red….. Voted last time: blue
Red……800………………………………680………………………………120
Blue…..200………………………………20………………………………180
(total)..1000……………………………700………………………………300
These numbers suggest we really have over-sampled a group and this will make the result wrong. So we correct that by some simple multiplication. Those that voted blue last time need to be double counted (60% / 30%) and the effect of those that voted red last time needs to be reduced.
so our corrected table looks like this:
Voting intention….. Voted last time: red….. Voted last time: blue
Red……629………………40 / 70 x 688 = 389…………60 / 30 x 120 = 240
Blue……371………………40 / 70 x 20 = 11.....………… 60 / 30 x 180 = 360
(total) ..1000…………………………400…………………..……………600
So are headline figures become:
Red 63% and Blue 37% (From the original 80/20).
That is how corrections work. Different polling companies, use different methods. Continued/
3. Likelihood of Voting
We know that likelihood of voting is one of the biggest distortions in polls because exit polls are generally very accurate. Most (all?) polling companies ask respondents to rate their likelihood of voting on a scale of 1-10. Some will then discard the lower scores and only include the higher ones. Other will simply weight the answers - so someone who says they will 10 out of 10, definitely vote for a candidate will be counted 10 times more than someone who says 1 out of 10 they will vote for that candidate.
4. "Don't knows" / undecideds
What to do with undecideds is another subject of variation. Some companies simply allocate them in proportion - i.e. they make no difference to the result. Some allocate them according to how they voted last time. In a big swing election like this one, this choice makes a big difference to the headline figures - which explains why one poll will have Labour 17% ahead of the Conservatives and another 23% ahead. The gap between the polls has narrowed as we get closer to the election as the number of undecided voters reduces.
Time will tell which method is correct. There is a lot of data that shows people who are undecided mid-term, most often return to voting the same way as they did previously. This close to an election that effect is smaller.
4.Polling question This does not really apply to voting intention as the "if there was an election tomorrow" question is an established, reliable question. However for issues, how the question is phrased makes a measurable difference to the result. One of my favourite ones came up in my Facebook memories this week. I think it was YouGov but it could have been any of the major companies. They found that 94% of people thought that "pointless, non-jobs should be abolished." I really worry about the 6%! Of course, no one should be paid for doing a non-job. The issue here is actually what is a 'non-job' and this is a form of propaganda is trying to argue that certain roles are unimportant. Whether that is true or not, is not engaged with at all. This is one of the classic ways that politicians and other activists use polling to push a point. One of the most important examples of this is around immigration. "Do you think the UK needs lower net migration to reduced the burden on public services?" Most people answer yes to this despite the fact that the data shows that migration is a net contributer to public services by means of filling vacancies and tax revenue. But it's polls like this that are used to demonstate that 'most people want lower immigration.' Beware of the question!
5. MRP
Finally MRP. MRP is Multi-level Regression and Post-stratification.
Our electoral system is very distorting. How a vote share translates into seats in the House of Commons is very variable. For decades now, uniform swing has been the method of choice which is reasonably accurate but not perfect. In essence, uniform swing ranks all of the 650 seats by how much swing from one party to another is needed to make it switch parties at an election. A close, marginal may need much less than 1% swing whilst a safe seat would need a 25% swing to change.
This modelling takes the headline polls: I.e. Labour 40% / Conservative 35% compared to Labour 36% / Conservative 42% at the previous election, for example. So that's an estimated 8% swing to Labour. (Some of the voters have gone elsewhere). So all the seats that require less than an 8% swing are predicted to change hands and the ones that are safer than that to remain the same. With a muti-party system, this is a little more complex but not much. Electoral Calculus does this and you can play with the numbers yourself on their website.
MRP is a different approach, aimed at predicting the seats more accurately. An MRP poll takes a big sample of voting intention nationwide. The reason why they use big samples is that they want to also collect lots of other demographic data and characterise the electorate by multiple factors.
Conceptually they may find from their polling that 95% of white, middle-aged male millionaires will vote Conservative. 70% of Asian, female, public sector workers will vote Labour etc. etc. Which factors they use varies between polling companies. Then for each of the UK constituencies, they use ONS data to understand the likely voting intention. That's just simple multiplication. If this constituency is 90% female Asian, public sector workers and 10% white middle aged millionaires then we can work out (assuming only 2 candidates standing) the voting intention of this constituency: 70% x 90% + 10% x 5% = number voting Labour and 95% x 10% + 30% x 90% = number voting Conservative. So the prediction becomes Labour 63.5% and Conservative 36.5%.
In reality, it doesn't come down to individual voters in that way, as different demographic factors like age, gender, socioeconomic status might pull in different ways but the mathematics of regression means that you can combine these factors and end up with an estimate of the votes for each party in each constituency. They then decide what sort of thresholds to use for certainty. I.e. If one party is 10% ahead, then the prediction is that they will most likely win this seat, whereas a party 3% ahead may be considered leaning that way and 1% would be 'too close to call' Then it's just a matter of adding up all 650 seats and seeing what the totals are. Most experts think that MRP is – or will become – the Gold Standard for polling. Results over the past few years suggest they lead to very accurate predictions.
There you go: An Alien's Guide to Earth Polling. It is definitely simplistic in places and I have left various things out but hopefully that explains where we are. The last important caveat is that comparing one poll to another is a mistake - and you see people do this a lot. They say things like "Reform was at 6% last week and now they're at 15% - wow, they's gone up 9%!!!" Well, actually the 6% poll was Company A and uses a different method to the 15% one from Company B. Company A actually shows Reform now on 10%. Similarly the previous poll from Company B had Reform at 11%, so in both cases the increase is actually 4%.
I will now take your questions and criticisms.
(or more likely you'll all too sensible to read this!)
AFZ
TL:DR Polling is complicated but careful interpretation - based on understanding how they are done - is usually informative.
I have been known to do it for a living (commercially rather than politically). That’s all broadly right but 2k/2.5k for comfort over 1k.
Thank you.
Yeah - quite a few polls do that. It doesn't make sense statistically - 2k doesn't give you much more than 1k. Most political polls published in the UK though will state ~1000-1100 respondents.
I have been known to do it for a living (commercially rather than politically). That’s all broadly right but 2k/2.5k for comfort over 1k.
Thank you.
Yeah - quite a few polls do that. It doesn't make sense statistically - 2k doesn't give you much more than 1k. Most political polls published in the UK though will state ~1000-1100 respondents.
Comes down to what you’re trying to find out and what your cross breaks are. By the time you’re looking for (hypothetically for illustration) one legged Black Caribbean women who live in villages then you might only have 4 in your thousand so the result won’t be statistically significant. And you might have picked up 4 outliers.
Whereas you’ve got more confidence if it’s 8-10. The percentages start to matter less and the raw numbers more.
Here are 7 parties, what do people think? 1000 will do you. The overall answer doesn’t change much on that if you boost it to 2 or 3k but the confidence with which you can do granularity within it does.
I have been known to do it for a living (commercially rather than politically). That’s all broadly right but 2k/2.5k for comfort over 1k.
Thank you.
Yeah - quite a few polls do that. It doesn't make sense statistically - 2k doesn't give you much more than 1k. Most political polls published in the UK though will state ~1000-1100 respondents.
Comes down to what you’re trying to find out and what your cross breaks are. By the time you’re looking for (hypothetically for illustration) one legged Black Caribbean women who live in villages then you might only have 4 in your thousand so the result won’t be statistically significant. And you might have picked up 4 outliers.
Whereas you’ve got more confidence if it’s 8-10. The percentages start to matter less and the raw numbers more.
Here are 7 parties, what do people think? 1000 will do you. The overall answer doesn’t change much on that if you boost it to 2 or 3k but the confidence with which you can do granularity within it does.
Ahh yeah, that makes sense. It protects you from statistical error when correcting for multiple factors.
Liam Fox, Tory candidate for our constituency, has just put a leaflet through our letterbox in person. I recon he must be worried that he won't retain what used to be considered a very safe Tory seat.
On a brief trip out into the countryside today, I noticed one poster for our Tory candidate*, and one for the Labour candidate* in the next constituency, both roadside, and well away from any houses.
Next Village to Arkland was (AFAICS) void of posters for anyone...
*neither had any clearly visible indication of which party they represent, although the Labour poster was largely red, and included part of the Union Jack - make of that what you will!
Liam Fox, Tory candidate for our constituency, has just put a leaflet through our letterbox in person. I recon he must be worried that he won't retain what used to be considered a very safe Tory seat.
That's unusual for a candidate. Usually they'd be chapping doors hoping to talk to people, which is what has potential to get people to vote for them who might otherwise not consider it. Even one good conversation in an hour will be more votes than will be gained by the hundred+ leaflets delivered in the same time. To be just pushing leaflets through doors is a sign of not expecting anyone to want to talk to him - either with everyone giving a polite "I don't have time" or else expecting angry people berating him for the damage done over the last 14 years. It's not a serious attempt to win votes.
Liam Fox, Tory candidate for our constituency, has just put a leaflet through our letterbox in person. I recon he must be worried that he won't retain what used to be considered a very safe Tory seat.
That's unusual for a candidate. Usually they'd be chapping doors hoping to talk to people, which is what has potential to get people to vote for them who might otherwise not consider it. Even one good conversation in an hour will be more votes than will be gained by the hundred+ leaflets delivered in the same time. To be just pushing leaflets through doors is a sign of not expecting anyone to want to talk to him - either with everyone giving a polite "I don't have time" or else expecting angry people berating him for the damage done over the last 14 years. It's not a serious attempt to win votes.
Random tangent - I currently know 4 people, all Scots by birth, education, upbringing and residency, canvassing in Scotland for Labour and the SNP. All say ‘knocking’ - is ‘chapping’ a regional thing even within Scotland? My friends are in Edinburgh, Inverness-shire, and Paisley.
It might be a Glasgow thing. I don't really know, it's just the word used around here. That may just be the few people I interact with regularly.
It's not that "knocking" isn't used. A couple of years back at the council elections we were for the first time in a position where checking on supporters to make sure they had voted on election day was worth while. We were surprised that the descriptions of how to go about doing this (identifying who to chap the doors of, how to talk to supporters, what time to be trying this) used the phrase "knocking up". We decided we were "getting out the vote", and not knocking up anyone.
The polling error of 2016 in the US has been attributed in part to not obtaining representative samples (or weighting to improve the representativeness of a sample) on the basis of level of education, and the error in 2020 has been attributed in part to sample representativeness on the basis of social engagement. White voters without a college degree supported Trump in 2016 much more than white voters overall. Trump made unexpected (by pollsters) gains (not a majority but still enough to affect state level results) in 2020 among black and Latino voters without a college degree and black and Latino men - and that trend seems to be getting stronger.
In 2020, pollsters started identifying white voters without a college degree and, if there weren’t a representative number of them in a sample, weighting their survey responses, but the results were skewed because Trump supporters were more likely than Biden supporters to not answer surveys at all. This wasn’t so much because of the “shy Republican” effect of people who don’t want to tell people they support a controversial candidate or even necessarily because they thought polls were all left-wing and biased. Rather, one explanation was that people who have smaller circles of friends and are less involved in community activities like volunteering, political activism (right or left), public meetings (even religious ones), etc., were more likely to support Trump and also less likely to answer a poll.
It’s also true that during the pandemic, people who were able because of their professions to work from home and therefore more likely to be home if the phone rang with a pollster calling disproportionately supported Biden, and also people who practiced a stricter form of self-imposed lockdown and therefore were more likely to be at home to answer the phone, also disproportionately supported Biden (which might be a bit confusing if you think of the less socially engaged supporting Trump more - this might have been an effect exclusive to the pandemic).
Do any of these demographic effects (education level and community engagement) seem to matter much in UK elections? And have they affected the accuracy of polls, or have pollsters there been taking them into account?
Tonight I winessed Sir Keir being formally interviewed on GB News. It was a very friendly interview and he was allowed to explain himself without interuption.
I am now seeing Sir Keir as a man who will do whatever it takes to win. He has convinced himself that he is the right man to be the PM of this country and that whatever he does to achieve this should be OK
Regarding sampling errors, I recently read a book which cited research showing that even if you ask people how they voted in the previous election, a surprisingly large proportion get it wrong, either because they've genuinely forgotten or for psychological reasons eg. not wanting to admit to themselves that they've changed their mind. I don't know if polling companies allow for this effect in their calculations or not.
In the US, it’s generally public information (or at least information available to any party or group doing voter outreach) what each registered voter’s party registration is (or their lack of any party affiliation), their address, and which elections they have voted in or not. You can’t find out how they voted without asking them though!
If you have been on the polling agency’s roll for a while (I’ve been on YouGov for years) they will have a record of all your previous voting responses. We are regularly asked about what we voted for last time and they do track people changing their mind about previous voting.
If you have been on the polling agency’s roll for a while (I’ve been on YouGov for years) they will have a record of all your previous voting responses. We are regularly asked about what we voted for last time and they do track people changing their mind about previous voting.
But do the local parties know when they send volunteers to distribute literature and knock on doors who has voted recently and who hasn’t?
If you have been on the polling agency’s roll for a while (I’ve been on YouGov for years) they will have a record of all your previous voting responses. We are regularly asked about what we voted for last time and they do track people changing their mind about previous voting.
But do the local parties know when they send volunteers to distribute literature and knock on doors who has voted recently and who hasn’t?
Yes, that information is freely available. They don’t know who they voted for though
Generally, if it's just leaflet delivery the volunteers doing that would know nothing about the households they're going to - they'd simply be given a street name and deliver to all houses there (even though that means delivery to some homes without registered voters). We would respect "no junk mail" notices, not all parties would, but otherwise the literature goes to everyone.
For canvassing the marked register isn't all that useful (IMO and IME). It only says whether or not someone voted in the last election, which isn't necessarily an indicator of likelihood to vote in whatever election is coming - probably OK this time round, someone who votes in council elections is very likely to vote in a general election, but approaching a council election after a general election that relationship definitely doesn't hold. Plus, canvassing has several aims - first to identify those will or are likely to vote for you, information that can be used to target material (maybe a later visit by the candidate close to polling day) to increase chances that they do actually vote for you; second to talk to people who are likely to be in general agreement with you but normally voting for another party, those conversions are what's needed to get a swing towards your party; but, it's also good to talk to someone and convince them to vote even if they're not a regular voter, which has certainly happened for me on polling day chapping the door of a couple who were likely to vote for us but getting their young adult son who hadn't voted in previous elections and wasn't going to vote this time round because he didn't see the point, and a few minutes chat and he was leaving the house to the voting place.
When canvassing the only additional information we take note of is whether someone has a postal ballot. There's usually no point chapping those doors in the week of the election, though even a week after postal ballots have gone out many of those voters are often still thinking about who to vote for and haven't yet voted - that may be particularly true this coming week, because "I haven't decided yet" is a much more common response on the door step than I've observed in previous elections.
Comments
It's like having Daily Mail headlines bandied about as though they were evidence based conclusions.
Not long now, please God...
AFZ
Three weeks, or thereabouts.
Some commentators, at least, are hoping for a sane government, which seems not unreasonable, however much Sid & Doris Bonkers and their Reform ilk might disagree.
https://x.com/ChrisO_wiki/status/1802737457273671786?t=b3hvW6z2Fh-MBtee78B9LA&s=19
As I was saying:
https://x.com/Conservatives/status/1802960272966959243?t=RjMWUOO87GMUramWhQEV2A&s=19
Annoyingly, it's not a straight-forward question. Suspensions and resignation from the party and boundary changes makes it not a straight-forward question. In theory, Labour had 205 MPs when Parliament was dissolved and so would need to gain 121 seats. This Guardian article, allowing for boundary changes, suggests the number is 126. It may be slightly higher as there are one or two seats in which they are vulnerable. They may lose a Bristol seat to the Green Party, for example.
Anyway, if you order the target seats by size of majority, as done here, then the answer is Colchester. Electoral Calculus currently has Labour with a 84% chance of winning in Colchester.
Or to put it another way,
These seats are the ones to look for on election day. If Labour are winning these seats then a majority is very likely:
York Outer
Rossendale and Darwen
Wolverhampton North East
Macclesfield
Blackpool North and Fleetwood
Monmouthshire
Glasgow West
Scarborough and Whitby
Dunfermline and Dollar
Welwyn Hatfield
Hitchin
Dunbartonshire West
Bolton West
Scunthorpe
Carlisle
Erewash
Bournemouth West
Edinburgh North and Leith
Earley and Woodley
Glenrothes and Mid Fife
Colchester
Stockton West
Edinburgh East and Musselburgh
Hexham
Ossett and Denby Dale
Middlesbrough South and Cleveland East
Dwyfor Meirionnydd
Doncaster East and the Isle of Axholme
Great Grimsby and Cleethorpes
Pendle and Clitheroe
Paisley and Renfrewshire South
Basingstoke
Bathgate and Linlithgow
Dover and Deal
Penrith and Solway
Cumbernauld and Kirkintilloch
Telford
Paisley and Renfrewshire North
East Kilbride and Strathaven
Livingston
AFZ
They can say what they like in their contract because they will not have to deliver
Inflation back down to normal levels for starters
That's a good thing in itself but doesn't necessarily indicate an economy on the up - falling inflation often indicates an economic slowdown or recession (I believe the latest figures indicate flat GDP).
Core inflation is still high. A fall in energy prices is hiding a bigger problem.
https://assets.kingsfund.org.uk/f/256914/x/0ab966500b/rise_decline_nhs_england_2000-20_2023.pdf
AFZ
The worrying thing there is that it's still 'lower funding increases' rather than 'no funding increases' - ie performance has (understandably) dropped off as we've not shovelled increases of cash at the problem to the same extent as was done 2000-2010. And health's one of the protected departments.
There is now a workforce plan, but social care as noted remains a skip fire.
There really needs to be an honest conversation about what we want to fund in this country, where the money's going to come from, how much it's going to cost, and how things are going to be done differently.
Indeed.
With an aging population and new treatments etc. healthcare costs will go up in the medium term. Healthcare inflation is also higher than general inflation. These things are all known by economists.
The point is that, as a country, we will spend more on healthcare in the next decade. The question is how we spend it. Fixing the NHS is the cheapest AND the most equitable way to do it. Not to mention the politically most palatable. Hence, for me, it's a no-brainer.
Of course part of how you fix the NHS is to fix social care. I've seen serious estimates that this will cost £5-£10Bn. But you get a lot back in NHS savings so net is quite a bit less. We've still got to find a few billion though...
AFZ
Pesky researchers finding new treatments that cost money and keep people alive so they can go on to need new expensive treatments.
I think that healthcare is inevitably going to get more expensive.
It can be partly ameliorate by alleviating poverty which has known correlations to ill health, and similarly drug and alcohol services. Reducing the hospital revolving door problem by sorting out our care system. All will require considerable investment for long term benefit.
It's a shame the 15 minute city has become mired in conspiracy theory and diverting money from road to encouraging active travel and subsidising public transport is electorially a hard sell as enabling people to not find themselves having to drive everywhere would have a positive long term health impact too.
To be honest, the social changes - the end of the death penalty, legislation of homosexuality and general equality legislation - are so mainstream now that I believe they are accepted by everyone bar a few extreme-right reactionaries. I would be the last to decry them. Jenkins and Harmon and their ilk were right.
However, at the time, this social liberalism - for want of a better term - was not universally popular among the working-class community.
I know it is dangerous to generalise about groups of people, but even now there are a fair few working-class people who are very socially conservative - they like 'traditional families' and would like children to be caned in school and perhaps even conscripted into the Army. They tend not to like immigration or multi-culturalism, except for their mates Errol and Abdul who are decent chaps. However, these very same people are left-wing in economic terms and approve heartily in public services like the NHS and the government spending capital on infrastructure programmes. These sort of voters - often also flag wavers and admirers of the monarchy - do not fit any party's template. They were attracted to 'Boris' because his con nodded towards them. And now you can see how some of this group are attracted to Reform.
I think a lot of those have gone back to Labour. It depends which is more important to them - less immigration and what they consider "woke", or functioning services.
Yes, and there was always a section of the working class who were traditionally Tory voters on this basis, but as things like the British Social Attitudes survey has chosen this group has generally become older and the country as a whole - including the working class - have become much more liberal on social issues over that time.
Concerns around immigration are shared more widely than - say - concerns over the traditional family/conscription - but that this is varied quite a bit over time demonstrates that attitudes here are nowhere near as sticky as sometimes made out.
I would add that in many Western countries, many working- and lower-middle-class social conservatives (on issues of sexuality, gender, the traditional notion of marriage and family, and reproduction), are themselves from immigrant backgrounds (of any race or religion), non-white backgrounds, or non-Christian backgrounds (of any race).
You're right that the working class of all cultural backgrounds is often supportive of government interventions to help the less fortunate, provide better public services, rein in the influence of the super-wealthy, etc., but a lot people from traditional societies (many of whom are immigrants, non-white, and/or non-Christian) living in the West feel also agree with statements like "hard work pays off", "laziness results in failure", "families need to live together and take care of their children, elderly, and sick/disabled", "religious and traditional cultural associations should play a large role in providing aid and conflict resolution in communities," etc.
And a lot of immigrants come from countries where the government is oppressive, deeply corrupt, or both, and are very distrusting of any politician promising big changes in society. A few of them a particularly distrustful of anything with the traditional language or iconography of socialism (even if they support many left-to-center-left policies) because of associations with politicians or parties from which they or their families immigrated.
Progressives should try to include in their image of social conservatives some, but not all (it is a different form of bigotry to assume all members of these groups are the same) African Pentecostals, South Asian Muslims and Hindus, Polish Catholics, Ultra-Orthodox Jews, etc.
My emphasis, what this analysis misses is that there are equally a reasonable number of first and second immigrants who - precisely because of their background - are much more literate regarding socialist/social-democratic policies, and who support left policies from a position of knowledge.
Maybe this is truer in the UK where more immigrants come from countries in the EU with more social democratic policies than it is in the US? Or maybe it is easier for people who come from more socialist countries in the “global South” (I’m never comfortable using that term) and who aren’t political dissidents or wealthier people who are strongly anti socialist to immigrate to the UK than it is to the US (it was very hard for people from India or China to immigrate to the US even before Trump)?
I should also amend when I said the beliefs “hard work pays off” and “laziness results in failure” I should have said “hard work should pay off” and “laziness should result in failure”. I didn’t mean to suggest that most immigrants believe that everyone can pull themselves up by their own bootstraps, as some right wing politicians suggest.
There was a lot of immigration of all classes of people from Commonwealth countries into the UK (which I presume is germane given the topic of the thread).
Something I've been meaning to ask is whether any news sources attempt to analyze the voting patterns of different demographic groups (race, ethnicity, gender, income, level of education, religious identity/level of religious practice, etc.) I see some general discussion along the lines of "this constituency has become more competitive for Labour as it has become more multiethnic" or "this constituency is full of affluent Remainers who used to vote Tory but now might vote LibDem", etc., but I never seem to see opinion polling or exit polling that says "this percent of...black voters/single women/voters without a university degree/voters who pray daily or attend religious services at least once a week...intend to vote for or say they voted for this party". The news in the US is full of this kind of opinion polling. I know it's harder to do at the constituency level, which is where is matters for a parliamentary election, but it just feels like a lot of the political analysis is only based on the party vote distribution from the last election and opinion polling of the general public in a constituency, rather than on breaking voters down into demographics. Is this true? Or am I just not looking at the news sources that have this kind of analysis?
You can do voting intention of the 2/10 but you need a big sample and typically within a nationally representative 2000 or so (MRP is a different kettle of fish) the numbers will be necessarily small. Which pushes to a specific minority poll. Which someone has to pay for - and people do, obviously - but not generally for run of the mill ‘state of the polls today’ polling.
Does that make sense? I’m half asleep…
Religious affiliation does not get discussed very much in the media, in these terms though. I suspect mostly because, socioeconomic status and level of education is a better predictor of voting intention.
Occasionally, you might here reference to religious groups. I know that most Muslims vote Labour. Members of the Sikh religion do as well, and there were targettes campaigns - in 2016 - to persuade them to vote against Kahn for London mayor.
In the Bexit refendum, level of education had a strong correlation with voting intention.
AFZ
*I am sad enough to be interested in methodology and data. I am not quite up to speed on current methods but all sorts of things have been tried. In a bit I will write a brief summary.
I apologise if no one other than me is interested, but this is a basic explanation of polling and how to interpret and understand them, including the pitfalls and problems.
George Gallup, the pioneer of polling allegedly said:
This scientific and statistical concept is used all the time in all sorts of contexts. If sampling is representative, then it gives you the answer. An opinion poll is no exception. The idea is to sample enough people to understand how the whole population will behave. Indeed, if a sample is representative, it will do so.
1. Why 1000 people?
There is some clever maths behind this and it comes down to something called Central Limit Theorem which is really useful for lots of things. Leaving aside the theoretic maths, the key bit is this: If you sample 100 people your estimate of voting intention with have ~10% margin of error. If you sample 1000 then it's ~3% and to get to ~1%, you need 10,000 people in your sample.
The vast majority of polls use a sample size of around 1000. This reflects a balance between getting as accurate an answer as possible and the costs and other limitations of sample size. Contacting 1000 people can be expensive. Contacting 10,000 will cost 10x as much. Moreover, it takes more time. Polls are a snapshot and if they take too long to collect then they will be potentially distorted by changes over time.
If you look at any poll they will state a margin of error. So when it says that 45% of people will vote for Candidate X what they actually mean, in precise, statistical terms, is that we are 95% confident that the true answer lies between 42% and 48%.
This is really important when polls are close. For the most part when people say "polls are wrong" they mean "I don't understand how margin of error works." The problem for political polls is that the result is binary. In a straight-fight between two presidential candidates - like the French second round - then the candidate with 50.00001% wins and the candidate with 49.99999% loses. It would be expected - in fact, almost inevitable - in that situation that some of the polls would have predicted the loser to win.
2. Representative Sampling?
Political polls face a specific challenge around making sure the sample is representative. This is easy to understand - if you took a survey in Mayfair, you would get a very different answer to one taken in Tower Hamlets. The populations are different. So how do you correct for this, to make your sample representative?
This is where polling companies differ. There are lots of factors you might use. There is a problem in that the more factors you use, the more you compromise the statistical power of the survey (making your margin or error effectively higher). What polling companies are always looking for is a magic formula that gives them the answer. Part of the challenge is that this changes over time.
Forty years ago, most polls were conducted by phone, using landline numbers, randomly selected. Now, in 2024, that is a very narrow, very specific population, you end up sampling. Modern polls are almost exclusively drawn from online panels with demographic factors already known. In the UK, demographics are not great correlators with voting intention so it's still not perfect.
So, one of the things you can use is previous voting behaviour. This one is quite nice because we know those numbers from the previous election. Here's a simplistic, worked example.
Poll of 1000 people; If there was an election tomorrow, how would you vote?
Red Party: 800
Blue Party: 200
So, 80% / 20%. Job done. But what if, our sample is not representative and we'll over-sampling people who normally vote Red?
So we ask the question, who did you vote for last time? and compare that to the result of the election. In the previous election the vote was Blue 60%, Red 40%. So we know that a truly representative sample would include 60% who voted Blue and 40% who voted Red last time.
In our survey, the people we sampled voted Blue 30% and Red 70%. Ahhh, we may have a bias here! So we need to build a contingency table and do some correction. (Just a table of numbers).
Voting intention……Voted last time: red….. Voted last time: blue
Red……800………………………………680………………………………120
Blue…..200………………………………20………………………………180
(total)..1000……………………………700………………………………300
These numbers suggest we really have over-sampled a group and this will make the result wrong. So we correct that by some simple multiplication. Those that voted blue last time need to be double counted (60% / 30%) and the effect of those that voted red last time needs to be reduced.
so our corrected table looks like this:
Voting intention….. Voted last time: red….. Voted last time: blue
Red……629………………40 / 70 x 688 = 389…………60 / 30 x 120 = 240
Blue……371………………40 / 70 x 20 = 11.....………… 60 / 30 x 180 = 360
(total) ..1000…………………………400…………………..……………600
So are headline figures become:
Red 63% and Blue 37% (From the original 80/20).
That is how corrections work. Different polling companies, use different methods.
Continued/
3. Likelihood of Voting
We know that likelihood of voting is one of the biggest distortions in polls because exit polls are generally very accurate. Most (all?) polling companies ask respondents to rate their likelihood of voting on a scale of 1-10. Some will then discard the lower scores and only include the higher ones. Other will simply weight the answers - so someone who says they will 10 out of 10, definitely vote for a candidate will be counted 10 times more than someone who says 1 out of 10 they will vote for that candidate.
4. "Don't knows" / undecideds
What to do with undecideds is another subject of variation. Some companies simply allocate them in proportion - i.e. they make no difference to the result. Some allocate them according to how they voted last time. In a big swing election like this one, this choice makes a big difference to the headline figures - which explains why one poll will have Labour 17% ahead of the Conservatives and another 23% ahead. The gap between the polls has narrowed as we get closer to the election as the number of undecided voters reduces.
Time will tell which method is correct. There is a lot of data that shows people who are undecided mid-term, most often return to voting the same way as they did previously. This close to an election that effect is smaller.
4.Polling question
This does not really apply to voting intention as the "if there was an election tomorrow" question is an established, reliable question. However for issues, how the question is phrased makes a measurable difference to the result. One of my favourite ones came up in my Facebook memories this week. I think it was YouGov but it could have been any of the major companies. They found that 94% of people thought that "pointless, non-jobs should be abolished." I really worry about the 6%! Of course, no one should be paid for doing a non-job. The issue here is actually what is a 'non-job' and this is a form of propaganda is trying to argue that certain roles are unimportant. Whether that is true or not, is not engaged with at all. This is one of the classic ways that politicians and other activists use polling to push a point. One of the most important examples of this is around immigration. "Do you think the UK needs lower net migration to reduced the burden on public services?" Most people answer yes to this despite the fact that the data shows that migration is a net contributer to public services by means of filling vacancies and tax revenue. But it's polls like this that are used to demonstate that 'most people want lower immigration.' Beware of the question!
5. MRP
Finally MRP. MRP is Multi-level Regression and Post-stratification.
Our electoral system is very distorting. How a vote share translates into seats in the House of Commons is very variable. For decades now, uniform swing has been the method of choice which is reasonably accurate but not perfect. In essence, uniform swing ranks all of the 650 seats by how much swing from one party to another is needed to make it switch parties at an election. A close, marginal may need much less than 1% swing whilst a safe seat would need a 25% swing to change.
This modelling takes the headline polls: I.e. Labour 40% / Conservative 35% compared to Labour 36% / Conservative 42% at the previous election, for example. So that's an estimated 8% swing to Labour. (Some of the voters have gone elsewhere). So all the seats that require less than an 8% swing are predicted to change hands and the ones that are safer than that to remain the same. With a muti-party system, this is a little more complex but not much. Electoral Calculus does this and you can play with the numbers yourself on their website.
MRP is a different approach, aimed at predicting the seats more accurately. An MRP poll takes a big sample of voting intention nationwide. The reason why they use big samples is that they want to also collect lots of other demographic data and characterise the electorate by multiple factors.
Conceptually they may find from their polling that 95% of white, middle-aged male millionaires will vote Conservative. 70% of Asian, female, public sector workers will vote Labour etc. etc. Which factors they use varies between polling companies. Then for each of the UK constituencies, they use ONS data to understand the likely voting intention. That's just simple multiplication. If this constituency is 90% female Asian, public sector workers and 10% white middle aged millionaires then we can work out (assuming only 2 candidates standing) the voting intention of this constituency: 70% x 90% + 10% x 5% = number voting Labour and 95% x 10% + 30% x 90% = number voting Conservative. So the prediction becomes Labour 63.5% and Conservative 36.5%.
In reality, it doesn't come down to individual voters in that way, as different demographic factors like age, gender, socioeconomic status might pull in different ways but the mathematics of regression means that you can combine these factors and end up with an estimate of the votes for each party in each constituency. They then decide what sort of thresholds to use for certainty. I.e. If one party is 10% ahead, then the prediction is that they will most likely win this seat, whereas a party 3% ahead may be considered leaning that way and 1% would be 'too close to call' Then it's just a matter of adding up all 650 seats and seeing what the totals are. Most experts think that MRP is – or will become – the Gold Standard for polling. Results over the past few years suggest they lead to very accurate predictions.
There you go: An Alien's Guide to Earth Polling. It is definitely simplistic in places and I have left various things out but hopefully that explains where we are. The last important caveat is that comparing one poll to another is a mistake - and you see people do this a lot. They say things like "Reform was at 6% last week and now they're at 15% - wow, they's gone up 9%!!!" Well, actually the 6% poll was Company A and uses a different method to the 15% one from Company B. Company A actually shows Reform now on 10%. Similarly the previous poll from Company B had Reform at 11%, so in both cases the increase is actually 4%.
I will now take your questions and criticisms.
(or more likely you'll all too sensible to read this!)
AFZ
TL:DR Polling is complicated but careful interpretation - based on understanding how they are done - is usually informative.
I’m not disagreeing with your maths but it’s what people buy
Thank you.
Yeah - quite a few polls do that. It doesn't make sense statistically - 2k doesn't give you much more than 1k. Most political polls published in the UK though will state ~1000-1100 respondents.
Comes down to what you’re trying to find out and what your cross breaks are. By the time you’re looking for (hypothetically for illustration) one legged Black Caribbean women who live in villages then you might only have 4 in your thousand so the result won’t be statistically significant. And you might have picked up 4 outliers.
Whereas you’ve got more confidence if it’s 8-10. The percentages start to matter less and the raw numbers more.
Here are 7 parties, what do people think? 1000 will do you. The overall answer doesn’t change much on that if you boost it to 2 or 3k but the confidence with which you can do granularity within it does.
Ahh yeah, that makes sense. It protects you from statistical error when correcting for multiple factors.
Next Village to Arkland was (AFAICS) void of posters for anyone...
*neither had any clearly visible indication of which party they represent, although the Labour poster was largely red, and included part of the Union Jack - make of that what you will!
Random tangent - I currently know 4 people, all Scots by birth, education, upbringing and residency, canvassing in Scotland for Labour and the SNP. All say ‘knocking’ - is ‘chapping’ a regional thing even within Scotland? My friends are in Edinburgh, Inverness-shire, and Paisley.
It's not that "knocking" isn't used. A couple of years back at the council elections we were for the first time in a position where checking on supporters to make sure they had voted on election day was worth while. We were surprised that the descriptions of how to go about doing this (identifying who to chap the doors of, how to talk to supporters, what time to be trying this) used the phrase "knocking up". We decided we were "getting out the vote", and not knocking up anyone.
In 2020, pollsters started identifying white voters without a college degree and, if there weren’t a representative number of them in a sample, weighting their survey responses, but the results were skewed because Trump supporters were more likely than Biden supporters to not answer surveys at all. This wasn’t so much because of the “shy Republican” effect of people who don’t want to tell people they support a controversial candidate or even necessarily because they thought polls were all left-wing and biased. Rather, one explanation was that people who have smaller circles of friends and are less involved in community activities like volunteering, political activism (right or left), public meetings (even religious ones), etc., were more likely to support Trump and also less likely to answer a poll.
It’s also true that during the pandemic, people who were able because of their professions to work from home and therefore more likely to be home if the phone rang with a pollster calling disproportionately supported Biden, and also people who practiced a stricter form of self-imposed lockdown and therefore were more likely to be at home to answer the phone, also disproportionately supported Biden (which might be a bit confusing if you think of the less socially engaged supporting Trump more - this might have been an effect exclusive to the pandemic).
Do any of these demographic effects (education level and community engagement) seem to matter much in UK elections? And have they affected the accuracy of polls, or have pollsters there been taking them into account?
I am now seeing Sir Keir as a man who will do whatever it takes to win. He has convinced himself that he is the right man to be the PM of this country and that whatever he does to achieve this should be OK
But do the local parties know when they send volunteers to distribute literature and knock on doors who has voted recently and who hasn’t?
Yes, that information is freely available. They don’t know who they voted for though
For canvassing the marked register isn't all that useful (IMO and IME). It only says whether or not someone voted in the last election, which isn't necessarily an indicator of likelihood to vote in whatever election is coming - probably OK this time round, someone who votes in council elections is very likely to vote in a general election, but approaching a council election after a general election that relationship definitely doesn't hold. Plus, canvassing has several aims - first to identify those will or are likely to vote for you, information that can be used to target material (maybe a later visit by the candidate close to polling day) to increase chances that they do actually vote for you; second to talk to people who are likely to be in general agreement with you but normally voting for another party, those conversions are what's needed to get a swing towards your party; but, it's also good to talk to someone and convince them to vote even if they're not a regular voter, which has certainly happened for me on polling day chapping the door of a couple who were likely to vote for us but getting their young adult son who hadn't voted in previous elections and wasn't going to vote this time round because he didn't see the point, and a few minutes chat and he was leaving the house to the voting place.
When canvassing the only additional information we take note of is whether someone has a postal ballot. There's usually no point chapping those doors in the week of the election, though even a week after postal ballots have gone out many of those voters are often still thinking about who to vote for and haven't yet voted - that may be particularly true this coming week, because "I haven't decided yet" is a much more common response on the door step than I've observed in previous elections.