politicalbetting.com » Blog Archive » The leading academic forecasters gather for the biggest event before May 7th
It’s an early start for me this morning to get to Westminster for what looks set to be the most important gathering of academic forecasters before the big day.
My gut feeling tells me that YouGov's panel is bust. ICM seems to far towards the Tories, so I suspect that it's on the upper end, but the ComRes poll in the SW (admittedly small and constituency so I am naturally sceptical) is supportive evidence.
Either the online or the phone polls are going to be horribly wrong. I'd back ICM over YouGov.
It seems safe to say nothing much is moving, but the problem is knowing what the underlying position is that it isn't moving from. Between UKIP and the LibDems in coalition there's plenty of unprecedented stuff to trip up the pollsters, and there's at least a few points variation between how different competent pollsters see it.
Not that enamoured by YG's methodology. As I understand it, what we get in effect is a self selecting sample of a self selecting fixed sample. Their pool is a tiny section of the population all of whom are self selected into the pool, and we have to take it more or less on trust that the pool distribution represents the population distribution. There is a secondary concern that the sample from the pool represents the pool because its is also self-selected, but even assuming that is controlled for, if the pools views are not representative (computer literate people with an internet connection prepared to answer VI questions and 10-15 minutes of questions about their toothpaste preferences) all we will see is a fair representation of the pools views, not the populations view. It might work, but for my taste it assumes too much.
For sure, there are a lot of assumptions in the YG polling, but what they do have is a large number of polls so we get a lot of data points and there has been a lot of movement in it over the last 5 years. i don't think we can discount it, it's within the same broad range as the other polls. The online vs phone gap isn't all that much.
My personal view is that we have a small tory lead of around 0.5% - 1% and that will be reflected on the day, what we don't have is momentum. In 21 days time it'll be over, I don't see any game changers. The collapse of the labour vote in Scotland is hiding it's strength in England. Whatever comes out of Scotland will be anti tory so it can be discounted.
In order to get a situation where Con + LD > Lab + SNP, I think we need a tory lead of ~5%. I can't see that happening at present, therefore Ed is PM.
Surely participants in YouGov polls must suffer from severe polling fatigue? Do they still need to go through the list of marketing questions these days, or are the polls stand-alone?
For sure, there are a lot of assumptions in the YG polling, but what they do have is a large number of polls so we get a lot of data points and there has been a lot of movement in it over the last 5 years. i don't think we can discount it, it's within the same broad range as the other polls. The online vs phone gap isn't all that much.
My personal view is that we have a small tory lead of around 0.5% - 1% and that will be reflected on the day, what we don't have is momentum. In 21 days time it'll be over, I don't see any game changers. The collapse of the labour vote in Scotland is hiding it's strength in England. Whatever comes out of Scotland will be anti tory so it can be discounted.
In order to get a situation where Con + LD > Lab + SNP, I think we need a tory lead of ~5%. I can't see that happening at present, therefore Ed is PM.
I still think we really don't have a scooby what will happen
We have no ideas how many marginal will change hands because of a leak of 2-3000 votes to either or both of the greens or the kippers. Not enough to give either them a first three place in the local vote, but enough to change the winner. There are 50 or so seats with less than 2000 majority.
We are mostly speculating where the LD collapse vote will go, it might split to Labour and the Conservatives, or it might stay at home. Ditto the receding kipper vote. We assume a SNP riot in Scotland, but it doesn't take very many points of shift to LAB at the last minute to make it a much more modest affair given the huge majorities that need to be overturned.
We have no real idea how differential turnout, especially on the back of individual voter registration is going to effect the result, it might well show a marked dip in younger voters, and hence reduce the left leaning turnout considerably.
We talk endlessly about needing an 11%+ swing for a Tory majority, but that is UNS, and UNS is looking increasingly unreliable in a multi party election. If Labour pours on votes in the inner cities and bleeds votes to the kippers in the suburbs it might need a lot less, if the Tories pile on votes in the shires and loose seats to Labour in the Midlands they might need more.
Even if I wasn't out the of country until several months after the election, my money would be staying in my pocket for this election with the exception of Scotland.
Surely participants in YouGov polls must suffer from severe polling fatigue? Do they still need to go through the list of marketing questions these days, or are the polls stand-alone?
I did a political poll the other day. And another one about television programmes that I never watch. So your answer is yes.
Not that enamoured by YG's methodology. As I understand it, what we get in effect is a self selecting sample of a self selecting fixed sample. Their pool is a tiny section of the population all of whom are self selected into the pool, and we have to take it more or less on trust that the pool distribution represents the population distribution. There is a secondary concern that the sample from the pool represents the pool because its is also self-selected, but even assuming that is controlled for, if the pools views are not representative (computer literate people with an internet connection prepared to answer VI questions and 10-15 minutes of questions about their toothpaste preferences) all we will see is a fair representation of the pools views, not the populations view. It might work, but for my taste it assumes too much.
Frankly i find all polling samples a little bit suspicious considering the frequency with which some people claim to be polled. And that includes the phone companies. It gives the impression that if you are a willing participant in polls then you go on a database and you are more likely to be called back in the future. Although maybe this doesn't actually undermine the results.
Somewhat like the debate this evening where the biggest name is missing PBers will note that my ARSE, in the history of mankind the greatest name in election forecasting, is absent from the Political Studies Association shindig.
Clearly I wish these pale imitatiions well as they fumble around in the dark and then strangely begin to ease grudgingly around my ARSE.
I just wonder, given the way FPTP works whether we will in fact return to a two party situation in England and Sctotland. Lab vs Con in England, SNP vs Lab in Scotland. Only in Wales will we still have a significant third party presence.
Customers who are applying to work, study or settle in the UK may also need an Immigration Health Surcharge reference number to access health services whilst in the UK, and may need to make an additional payment as part of your application. To gain a health surcharge reference and check whether you are required to make a payment, please log onto https://www.immigration-health-surcharge.service.gov.uk/payment/start
Latest ARSE with added APLOMB 2015 General Election & "JackW Dozen" Projection Countdown :
50 hours 50 minutes 50 seconds
I see that you have resisted the temptation of entering the contest so as to give us small fry a chance. Very noble!
Well quite .... I don't enter PB competitions anymore.
Apart from being very noble Mrs JackW was getting a little tired of offloading OGH's competition prizes of various cases of hair tonics and restoratives down to local church bazaars.
Mind you the pews around North Hertfordshire are now full of some interesting elderly men with "winning here" hair styles.
Just 3 weeks to go and no announcement as yet concerning any all-night PB election event at a suitable hostelry. Presumably this means that none is planned.
A few days ago Ashcroft polled South Ribble: the result was a dead heat between Con and Lab on 39%. The Greens were on 2%, but it turns out this is one of the few seats where the Greens aren't standing. Could prove crucial:
Just 3 weeks to go and no announcement as yet concerning any all-night PB election event at a suitable hostelry. Presumably this means that none is planned.
Latest ARSE with added APLOMB 2015 General Election & "JackW Dozen" Projection Countdown :
50 hours 50 minutes 50 seconds
I see that you have resisted the temptation of entering the contest so as to give us small fry a chance. Very noble!
Well quite .... I don't enter PB competitions anymore.
Apart from being very noble Mrs JackW was getting a little tired of offloading OGH's competition prizes of various cases of hair tonics and restoratives down to local church bazaars.
Mind you the pews around North Hertfordshire are now full of some interesting elderly men with "winning here" hair styles.
Though Shadsy did, with Labour on 300 and Tories on 254. Interesting, but perhaps realising that the best chance with theses contests is not being in the centre of the pack.
I think that the Ipsos Mori poll today is one of the more important polls for some time. It is a slightly more volatile than average (because of their methodology as I understand it) phone poll.
If that were to show a Tory lead of, say, 5% then the view that the Yougov panel model is broken would gain a lot of strength. If it were to show something similar to Yougov then conversely the ICM would really look like a fairly extreme outlier.
At the moment in the competition I have gone for a modest Labour plurality largely based on the majority of polling. I may well change my mind and get somewhat closer to Jack's ARSE if Ipsos Mori showed a significant Tory lead.
The big event of the day could be very entertaining indeed
Ed Miliband will use tonight’s opposition leaders’ debate to launch a “forensic” attack on Nicola Sturgeon’s policies and her record in government, senior Labour sources have revealed.
That sounds great, but he might have a slight problem...
new analysis disclosed that Ed Miliband voted in the same way as SNP MPs more than 90 per cent of the time in the last five years, including 27 out of 28 votes on welfare and 62 out of 65 votes on the economy.
Latest ARSE with added APLOMB 2015 General Election & "JackW Dozen" Projection Countdown :
50 hours 50 minutes 50 seconds
I see that you have resisted the temptation of entering the contest so as to give us small fry a chance. Very noble!
Well quite .... I don't enter PB competitions anymore.
Apart from being very noble Mrs JackW was getting a little tired of offloading OGH's competition prizes of various cases of hair tonics and restoratives down to local church bazaars.
Mind you the pews around North Hertfordshire are now full of some interesting elderly men with "winning here" hair styles.
Though Shadsy did, with Labour on 300 and Tories on 254. Interesting, but perhaps realising that the best chance with theses contests is not being in the centre of the pack.
In those inimitable words of "Yes Prime Minister" Shadsy has delivered a "very brave" entry.
Given around 25 losses to the SNP, Labour on 300 seats implies approx. 60+ gains from the Tories .... Hhhmmm.
At the moment in the competition I have gone for a modest Labour plurality largely based on the majority of polling. I may well change my mind and get somewhat closer to Jack's ARSE if Ipsos Mori showed a significant Tory lead.
Like night follows day and Nigel declaring UKIP's manifesto as "drivel" so the great and the good of PB begin to warm to my ARSE.
The big event of the day could be very entertaining indeed
Ed Miliband will use tonight’s opposition leaders’ debate to launch a “forensic” attack on Nicola Sturgeon’s policies and her record in government, senior Labour sources have revealed.
What can I say? That sounds absolutely fascinating in the extreme. Can't wait.
Latest ARSE with added APLOMB 2015 General Election & "JackW Dozen" Projection Countdown :
50 hours 50 minutes 50 seconds
I see that you have resisted the temptation of entering the contest so as to give us small fry a chance. Very noble!
Well quite .... I don't enter PB competitions anymore.
Apart from being very noble Mrs JackW was getting a little tired of offloading OGH's competition prizes of various cases of hair tonics and restoratives down to local church bazaars.
Mind you the pews around North Hertfordshire are now full of some interesting elderly men with "winning here" hair styles.
Though Shadsy did, with Labour on 300 and Tories on 254. Interesting, but perhaps realising that the best chance with theses contests is not being in the centre of the pack.
In those inimitable words of "Yes Prime Minister" Shadsy has delivered a "very brave" entry.
Given around 25 losses to the SNP, Labour on 300 seats implies approx. 60+ gains from the Tories .... Hhhmmm.
Latest ARSE with added APLOMB 2015 General Election & "JackW Dozen" Projection Countdown :
50 hours 50 minutes 50 seconds
I see that you have resisted the temptation of entering the contest so as to give us small fry a chance. Very noble!
Well quite .... I don't enter PB competitions anymore.
Apart from being very noble Mrs JackW was getting a little tired of offloading OGH's competition prizes of various cases of hair tonics and restoratives down to local church bazaars.
Mind you the pews around North Hertfordshire are now full of some interesting elderly men with "winning here" hair styles.
Though Shadsy did, with Labour on 300 and Tories on 254. Interesting, but perhaps realising that the best chance with theses contests is not being in the centre of the pack.
In those inimitable words of "Yes Prime Minister" Shadsy has delivered a "very brave" entry.
Given around 25 losses to the SNP, Labour on 300 seats implies approx. 60+ gains from the Tories .... Hhhmmm.
If Yougov is right that is a dawdle. If.
Interestingly which is correct - A projection based on YouGov or the projection of Peter Kellner of YouGov ?
Not that enamoured by YG's methodology. As I understand it, what we get in effect is a self selecting sample of a self selecting fixed sample. Their pool is a tiny section of the population all of whom are self selected into the pool, and we have to take it more or less on trust that the pool distribution represents the population distribution. There is a secondary concern that the sample from the pool represents the pool because its is also self-selected, but even assuming that is controlled for, if the pools views are not representative (computer literate people with an internet connection prepared to answer VI questions and 10-15 minutes of questions about their toothpaste preferences) all we will see is a fair representation of the pools views, not the populations view. It might work, but for my taste it assumes too much.
People have made exactly the same arguments in the past re: YouGov and ended up with egg on their faces. I remember all the Labour supporters refusing to believe the YouGov polls showing that Boris was going to win the Mayoralty in 2008.
I think that the Ipsos Mori poll today is one of the more important polls for some time. It is a slightly more volatile than average (because of their methodology as I understand it) phone poll.
If that were to show a Tory lead of, say, 5% then the view that the Yougov panel model is broken would gain a lot of strength. If it were to show something similar to Yougov then conversely the ICM would really look like a fairly extreme outlier.
At the moment in the competition I have gone for a modest Labour plurality largely based on the majority of polling. I may well change my mind and get somewhat closer to Jack's ARSE if Ipsos Mori showed a significant Tory lead.
Anthony Wells did a piece on phone vs Internet polls the other day.
It's not just YG showing what is essentially a draw, it's all the polls except ICM. As someone who expects a big Tory seat lead at a minimum (50 plus agead of Labour), I am more inclined to believe ICM have it broadly right (though 39% is surely too high), but there's currently not much supporting evidence. I also think YG will be more aware of potential issues with its methodology than anyone else given the reputational and business issues it has at stake. It's hard to believe the company does not have things under constant review. But that's the same for all of them I guess.
The big event of the day could be very entertaining indeed
Ed Miliband will use tonight’s opposition leaders’ debate to launch a “forensic” attack on Nicola Sturgeon’s policies and her record in government, senior Labour sources have revealed.
That sounds great, but he might have a slight problem...
new analysis disclosed that Ed Miliband voted in the same way as SNP MPs more than 90 per cent of the time in the last five years, including 27 out of 28 votes on welfare and 62 out of 65 votes on the economy.
Latest ARSE with added APLOMB 2015 General Election & "JackW Dozen" Projection Countdown :
50 hours 50 minutes 50 seconds
I see that you have resisted the temptation of entering the contest so as to give us small fry a chance. Very noble!
Well quite .... I don't enter PB competitions anymore.
Apart from being very noble Mrs JackW was getting a little tired of offloading OGH's competition prizes of various cases of hair tonics and restoratives down to local church bazaars.
Mind you the pews around North Hertfordshire are now full of some interesting elderly men with "winning here" hair styles.
Though Shadsy did, with Labour on 300 and Tories on 254. Interesting, but perhaps realising that the best chance with theses contests is not being in the centre of the pack.
In those inimitable words of "Yes Prime Minister" Shadsy has delivered a "very brave" entry.
Given around 25 losses to the SNP, Labour on 300 seats implies approx. 60+ gains from the Tories .... Hhhmmm.
If Yougov is right that is a dawdle. If.
Interestingly which is correct - A projection based on YouGov or the projection of Peter Kellner of YouGov ?
Kellner's projection is that things are going to change. Polling, ICM apart, says otherwise. Ipsos Mori has been the one phone pollster to consistently favour Labour. If it is different then the panels are concealing any underlying movement rather than revealing it. Given the nature of the panels and those who can be bothered to be on them this sounds plausible to me on the basis that those on the panels are more committed and less likely to change their minds but plausibility is not evidence.
@SMukesh The divergence between phone and internet polls is interesting if only because at some point in the future phone polls will be choosing from a diminishing pool of responders. This could be mitigated by increasing the percentage of calls to "mobiles", but is unlikely to be fully effective. (people on mobiles will not generally respond to polls unless they are in their homes at the time)
Latest ARSE with added APLOMB 2015 General Election & "JackW Dozen" Projection Countdown :
50 hours 50 minutes 50 seconds
I see that you have resisted the temptation of entering the contest so as to give us small fry a chance. Very noble!
Well quite .... I don't enter PB competitions anymore.
Apart from being very noble Mrs JackW was getting a little tired of offloading OGH's competition prizes of various cases of hair tonics and restoratives down to local church bazaars.
Mind you the pews around North Hertfordshire are now full of some interesting elderly men with "winning here" hair styles.
Though Shadsy did, with Labour on 300 and Tories on 254. Interesting, but perhaps realising that the best chance with theses contests is not being in the centre of the pack.
In those inimitable words of "Yes Prime Minister" Shadsy has delivered a "very brave" entry.
Given around 25 losses to the SNP, Labour on 300 seats implies approx. 60+ gains from the Tories .... Hhhmmm.
That certainly makes Betfair's 8.6 for their Labour 301-325 seats band look like great value, especially when compared with his own firm's 6.0 for the same bet!
Labour has scrapped all plans to run billboard posters of David Cameron during the general election campaign in what it says it is a deliberate attempt to avoid “negative personalised adverts” and raise the tone of debate.
Maybe i'm missing something - but as I understand it, the internet pollsters have a big panel of, say, 100,000 people - and, each time a poll is commissioned, *invite* ~1000 people to fit a demographic profile.
Surely it would make more sense to only poll specific individuals in marginal constituencies (who are on the electoral roll) and then repoll exactly the same people to see if they change their mind. Repoll these people immediately after they've voted - rinse & repeat every month for the next 5 years.
Then, instead of weighting the responses, converting to vote shares and baxtering - we would just get direct seat projections.
I guess my question boils down to - why use random sampling when measuring changes in opinion over time, when individual opinion changes can be captured cheaply, accurately and regularly - and are probably more useful in determining the end result (which seats actually change hands)?
Labour has scrapped all plans to run billboard posters of David Cameron during the general election campaign in what it says it is a deliberate attempt to avoid “negative personalised adverts” and raise the tone of debate.
Um, right...
Not sure why that is not consistent with this 'video'.
Maybe i'm missing something - but as I understand it, the internet pollsters have a big panel of, say, 100,000 people - and, each time a poll is commissioned, *invite* ~1000 people to fit a demographic profile.
Surely it would make more sense to only poll specific individuals in marginal constituencies (who are on the electoral roll) and then repoll exactly the same people to see if they change their mind. Repoll these people immediately after they've voted - rinse & repeat every month for the next 5 years.
Then, instead of weighting the responses, converting to vote shares and baxtering - we would just get direct seat projections.
I guess my question boils down to - why use random sampling when measuring changes in opinion over time, when individual opinion changes can be captured cheaply, accurately and regularly - and are probably more useful in determining the end result (which seats actually change hands)?
You wouldn't know how representative of the population in that constituency your sample was in terms of their likelihood to change their vote. At the extreme end you might have a sample of activists who are never going to change their vote come what may. That in a nutshell is my concern with the current YG methodology as well, if it tends to sample politically engaged people, then any effects will be damped as it would take something very major to make them switch allegiance.
@SMukesh The divergence between phone and internet polls is interesting if only because at some point in the future phone polls will be choosing from a diminishing pool of responders. This could be mitigated by increasing the percentage of calls to "mobiles", but is unlikely to be fully effective. (people on mobiles will not generally respond to polls unless they are in their homes at the time)
But "being In your home" doesn't obviously introduce a bias, it's a fairly universal experience!
There must be some better method which we just haven't thought of yet though. Use Facebook or Twitter in some way, or back to approaching people in the street?
Unless you are value trading in the market then betting on Labour and Ed Miliband at that seat level shows a level of faith that hasn't been seen since "Stuart Truth" was prognosticating on PB over two years back.
@Ishmael_X Possibly combining the two? The point about mobile users being more reluctant to respond to polls while on the move is down to how the users tend to perceive calls to them in different situations.
Latest ARSE with added APLOMB 2015 General Election & "JackW Dozen" Projection Countdown :
50 hours 50 minutes 50 seconds
I see that you have resisted the temptation of entering the contest so as to give us small fry a chance. Very noble!
Well quite .... I don't enter PB competitions anymore.
Apart from being very noble Mrs JackW was getting a little tired of offloading OGH's competition prizes of various cases of hair tonics and restoratives down to local church bazaars.
Mind you the pews around North Hertfordshire are now full of some interesting elderly men with "winning here" hair styles.
Though Shadsy did, with Labour on 300 and Tories on 254. Interesting, but perhaps realising that the best chance with theses contests is not being in the centre of the pack.
In those inimitable words of "Yes Prime Minister" Shadsy has delivered a "very brave" entry.
Given around 25 losses to the SNP, Labour on 300 seats implies approx. 60+ gains from the Tories .... Hhhmmm.
That certainly makes Betfair's 8.6 for their Labour 301-325 seats band look like great value, especially when compared with his own firm's 6.0 for the same bet!
Latest ARSE with added APLOMB 2015 General Election & "JackW Dozen" Projection Countdown :
50 hours 50 minutes 50 seconds
I see that you have resisted the temptation of entering the contest so as to give us small fry a chance. Very noble!
Well quite .... I don't enter PB competitions anymore.
Apart from being very noble Mrs JackW was getting a little tired of offloading OGH's competition prizes of various cases of hair tonics and restoratives down to local church bazaars.
Mind you the pews around North Hertfordshire are now full of some interesting elderly men with "winning here" hair styles.
Though Shadsy did, with Labour on 300 and Tories on 254. Interesting, but perhaps realising that the best chance with theses contests is not being in the centre of the pack.
In those inimitable words of "Yes Prime Minister" Shadsy has delivered a "very brave" entry.
Given around 25 losses to the SNP, Labour on 300 seats implies approx. 60+ gains from the Tories .... Hhhmmm.
If Yougov is right that is a dawdle. If.
Interestingly which is correct - A projection based on YouGov or the projection of Peter Kellner of YouGov ?
Kellner's projection is that things are going to change. Polling, ICM apart, says otherwise. Ipsos Mori has been the one phone pollster to consistently favour Labour. If it is different then the panels are concealing any underlying movement rather than revealing it. Given the nature of the panels and those who can be bothered to be on them this sounds plausible to me on the basis that those on the panels are more committed and less likely to change their minds but plausibility is not evidence.
Maybe i'm missing something - but as I understand it, the internet pollsters have a big panel of, say, 100,000 people - and, each time a poll is commissioned, *invite* ~1000 people to fit a demographic profile.
Surely it would make more sense to only poll specific individuals in marginal constituencies (who are on the electoral roll) and then repoll exactly the same people to see if they change their mind. Repoll these people immediately after they've voted - rinse & repeat every month for the next 5 years.
Then, instead of weighting the responses, converting to vote shares and baxtering - we would just get direct seat projections.
I guess my question boils down to - why use random sampling when measuring changes in opinion over time, when individual opinion changes can be captured cheaply, accurately and regularly - and are probably more useful in determining the end result (which seats actually change hands)?
You wouldn't know how representative of the population in that constituency your sample was in terms of their likelihood to change their vote. At the extreme end you might have a sample of activists who are never going to change their vote come what may. That in a nutshell is my concern with the current YG methodology as well, if it tends to sample politically engaged people, then any effects will be damped as it would take something very major to make them switch allegiance.
Sure, there are going to be problems - but I'm saying there could potentially be fewer problems than with the classic random sample method. In a way, you can benefit (in accuracy terms) from the sample not being representative - you don't want your sample to include people who aren't going to be on the electoral roll on polling day. You don't want people who think they voted last time, but didn't actually vote. You probably don't want people who are signed up members of political parties. These are the types of things that can be captured from your polling panel - so you only end up with the type of people who will swing the vote.
I'm no psephologist - obviously - but it does strike me that dismissing internet polls is a bit silly. With all the extra data that can be captured with technology, surely it's only a matter of time before the methodology of internet polls becomes more accurate in achieving the ultimate objective (the constituency tallies) than the old tried & trusted methods. We might not be there yet, I'll grant you
I must say I can't really understand what the point of having professional election forecasters is. It's good fun for amateurs, but why pay people when you could just wait for the votes to be counted?
Unless you are value trading in the market then betting on Labour and Ed Miliband at that seat level shows a level of faith that hasn't been seen since "Stuart Truth" was prognosticating on PB over two years back.
That ended well.
Well yes, a trading bet on 301-325 should produce a healthy profit if Labour were indeed seen to be heading for around the 300 seat mark. At some stage it should prove possible to lay this off at around even money or indeed less, bearing in mind that the Red Team invariably builds a large initial lead from the early declaring city constituencies before being pegged back.
Morning all and it would be fun to be a fly on the wall today as all the left-wing so called academic experts try to come to terms with the fact their beloved Labour party is heading down the plughole. John Curtice, SLAB's chief cheerleader in particular would be fun.
OGH is clearly devastated that a Tory poster has appeared in his area of Bedford for the 1st time.
In fact, the big UKIP demographics are the ones generally that are unlikely to have everyday internet access. So, I am not sure why phone pollsters would find more elderly Tory voters, but fewer elderly UKIP ones.
Morning all and it would be fun to be a fly on the wall today as all the left-wing so called academic experts try to come to terms with the fact their beloved Labour party is heading down the plughole. John Curtice, SLAB's chief cheerleader in particular would be fun.
OGH is clearly devastated that a Tory poster has appeared in his area of Bedford for the 1st time.
Curtice isn't a cheerleader for SLAB methinks. He is one of the most bearish academics on their predicament.
Labour has scrapped all plans to run billboard posters of David Cameron during the general election campaign in what it says it is a deliberate attempt to avoid “negative personalised adverts” and raise the tone of debate.
Um, right...
Does that amount to an admission that Labour did at some stage actually have such plans?
Unless you are value trading in the market then betting on Labour and Ed Miliband at that seat level shows a level of faith that hasn't been seen since "Stuart Truth" was prognosticating on PB over two years back.
That ended well.
Well yes, a trading bet on 301-325 should produce a healthy profit if Labour were indeed seen to be heading for around the 300 seat mark. At some stage it should prove possible to lay this off at around even money or indeed less, bearing in mind that the Red Team invariably builds a large initial lead from the early declaring city constituencies before being pegged back.
Unless there is some consistency in the early declarations the big driver will be the exit poll especially as the 2010 offering showed a level of accuracy that many thought was a rip off from my ARSE.
I must say I can't really understand what the point of having professional election forecasters is. It's good fun for amateurs, but why pay people when you could just wait for the votes to be counted?
One imagines some serious money is being moved around by traders, companies and billionaires based on probable election outcomes and the likely legislative program that follows.
If EIC is looking like a winner one imagines the non-doms will be contacting their estate agents.
They have definitely got the balance much better in this election than they did in 2010. Unfortunately it was in 2010 that a majority was there for the taking.
In fact, the big UKIP demographics are the ones generally that are unlikely to have everyday internet access. So, I am not sure why phone pollsters would find more elderly Tory voters, but fewer elderly UKIP ones.
Perhaps "county" ladies and WI members answer the phone to unlisted numbers and Victor Meldrew's either ignore it or tell them to p155 off
Unless you are value trading in the market then betting on Labour and Ed Miliband at that seat level shows a level of faith that hasn't been seen since "Stuart Truth" was prognosticating on PB over two years back.
That ended well.
Well yes, a trading bet on 301-325 should produce a healthy profit if Labour were indeed seen to be heading for around the 300 seat mark. At some stage it should prove possible to lay this off at around even money or indeed less, bearing in mind that the Red Team invariably builds a large initial lead from the early declaring city constituencies before being pegged back.
Unless there is some consistency in the early declarations the big driver will be the exit poll especially as the 2010 offering showed a level of accuracy that many thought was a rip off from my ARSE.
I couldn't possibly comment.
My one worry with the exit poll is that Scottish voters may not admit to voting for Jim Netenyahu Murphy !
In fact, the big UKIP demographics are the ones generally that are unlikely to have everyday internet access. So, I am not sure why phone pollsters would find more elderly Tory voters, but fewer elderly UKIP ones.
My guess is that internet panels tend to pick up the most politically committed people, which is why they tend to have high UKIP scores.
You wouldn't know how representative of the population in that constituency your sample was in terms of their likelihood to change their vote. At the extreme end you might have a sample of activists who are never going to change their vote come what may. That in a nutshell is my concern with the current YG methodology as well, if it tends to sample politically engaged people, then any effects will be damped as it would take something very major to make them switch allegiance.
Sure, there are going to be problems - but I'm saying there could potentially be fewer problems than with the classic random sample method. In a way, you can benefit (in accuracy terms) from the sample not being representative - you don't want your sample to include people who aren't going to be on the electoral roll on polling day. You don't want people who think they voted last time, but didn't actually vote. You probably don't want people who are signed up members of political parties.
I'm no psephologist - obviously - but it does strike me that dismissing internet polls is a bit silly. With all the extra data that can be captured with technology, surely it's only a matter of time before the methodology of internet polls becomes more accurate in achieving the ultimate objective (the constituency tallies) than the old tried & trusted methods. We might not be there yet, I'll grant you
The problem is self-selection.
If we had a well-behaved, spam-free internet you'd just email people. We don't, and the only way round that is an opt-in. People self-select by opting in, and then again by not dropping out when they realise what a ballsachingly tedious exercise in futility answering an online poll actually is. (It is very much more than a simple "How you gonna vote?") This may not matter because you get equal proportions of oddballs on each side, but that isn't obvious. There is also the following purely theoretical danger: if one major party said to each new activist recruit, by word of mouth, that they were expected to sign up to internet polling panels, and the other party didn't, that would distort the results. There is no evidence AT ALL of this happening but on the other hand I can see no way of effectively guarding against it.
It's sort of like electing the Pope: just as wanting to be Pope is a very undesirable quality in a potential Pope, you don't want to poll people who are positively eager to be polled.
I am also becoming suspicious about the way YouGov polls do not appear to be reacting to events. That said, the longer moving average better demonstrates the recent fall off in support for UKIP and The Greens in favour of the 'traditional' parties. It still has Labour ahead, but to my mind suggests that could still be a lot to play for. We possibly also need to bear in mind that great chunks of the country (and the public schools) are still enjoying their half-term holiday. Click to enlarge...
I must say I can't really understand what the point of having professional election forecasters is. It's good fun for amateurs, but why pay people when you could just wait for the votes to be counted?
These people are really just academics whose area of research overlaps with election forecasting, rather than professional forecasters. I would guess the true professional forecasters are employed in the City, helping inform financiers' speculation.
Firstly I declare an interest as a member of the YouGov panel. YouGov does indeed poll the same small % of the electorate over and over again. I have completed no fewer than 3 polls this week which included voting intention questions. Most of the surveys are frankly just turgid and repetitive. I only do them because over the course of a year I get £100-150 in payments from YouGov.
If you look at the last 2 weeks of YouGov polls, the Tories have been 33-37 and Labour 33-36 and almost all the leads either way have been caused by +1 or -1 for one side or the other. The LibDems, UKIP and Greens have barely shifted, again maybe +1 or -1.
On Monday voter registration closes. I hope Andy JS does one of his excellent spreadsheets on the size of constituency electorates and the difference from 2010. It will be interesting to see if in marginal seats, particularly those with large transient communities like students, there has been a substantial fall in registrations.
I remain convinced we will see the final votes at something like Tory 35-37, Labour 29-31, LibDem 12-15 and UKIP 10-12. In Scotland I expect the SNP to possibly pick up 25-30 Labour seats. If they pick up more then Ed is looking like ending up with fewer MPs than Gordon Brown did in 2010.
Anecdote alert: My eldest cousin lives in Jim Murphy's seat. We were discussing how it looks for him last night. She said she and her neighbours like Murphy but wouldn't dream of voting for him. Her view is they wouldn't be surprised of Dr David Monty scrapes through the middle of a dogfight between Murphy and the SNP. I wouldn't go that far but would be thrilled if it comes to pass.
Mr. Mark, I saw Peston last night (on the TV, I hasten to add). I kept waiting for him to point out the IMF had criticised Osborne's plans, only to end up praising the UK economy as the model to follow.
Maybe i'm missing something - but as I understand it, the internet pollsters have a big panel of, say, 100,000 people - and, each time a poll is commissioned, *invite* ~1000 people to fit a demographic profile.
Surely it would make more sense to only poll specific individuals in marginal constituencies (who are on the electoral roll) and then repoll exactly the same people to see if they change their mind. Repoll these people immediately after they've voted - rinse & repeat every month for the next 5 years.
Then, instead of weighting the responses, converting to vote shares and baxtering - we would just get direct seat projections.
I guess my question boils down to - why use random sampling when measuring changes in opinion over time, when individual opinion changes can be captured cheaply, accurately and regularly - and are probably more useful in determining the end result (which seats actually change hands)?
If you got an unrepresentative sample to begin with, you'd get repeatedly wrong results. That's why pollsters go for randomness.
"The suggestion that you can laugh at everything, except certain aspects of Islam, because Muslims are much more prickly that the rest of the population - what is that, if not discrimination?"
If we had a well-behaved, spam-free internet you'd just email people. We don't, and the only way round that is an opt-in. People self-select by opting in, and then again by not dropping out when they realise what a ballsachingly tedious exercise in futility answering an online poll actually is. (It is very much more than a simple "How you gonna vote?") This may not matter because you get equal proportions of oddballs on each side, but that isn't obvious. There is also the following purely theoretical danger: if one major party said to each new activist recruit, by word of mouth, that they were expected to sign up to internet polling panels, and the other party didn't, that would distort the results. There is no evidence AT ALL of this happening but on the other hand I can see no way of effectively guarding against it.
Very true, but it actually worse than that. Even if each party got all their activists to sign up, and even if each got the same number of activists to sign up, its still nothing like a representative sample, because activists are 100% certain to vote and the population isn't and activists are 99% unlikely to change their vote, and the population isn't. So you could be in a position where three quarters of the sample are politically engaged activists whose votes are set in stone, and one quarter regular punters, which would show some vote movement (due to the regular punters) but it might move about a completely unrepresentative mean (owing to the activists biasing the result) or it might move in a very damped fashion due to political events not changing the perceptions of the bulk of the sample. Its all speculation, but the problem with a biased sample, is you don't know how biased it is until you get to compare it with the population (ie at the election) which is too late to be useful.
In fact, the big UKIP demographics are the ones generally that are unlikely to have everyday internet access. So, I am not sure why phone pollsters would find more elderly Tory voters, but fewer elderly UKIP ones.
My guess is that internet panels tend to pick up the most politically committed people, which is why they tend to have high UKIP scores.
Yep - that sounds reasonable.
Generally, the Labour online and telephone poll scores are pretty much the same. The big variances seem to be the UKIP and Tory scores. That's one of the reasons I see a big Tory seat lead. When push comes to shove a lot of UKIP-inclined voters are going to go for the Tories. It could be that ICM has found them further down the line to making that final decision. On a really bad night for Labour, there could also be Tory tactical voting for UKIP and the LDs, which would see Labour come in much lower than 250 seats - especially if it is an almost total wipe-out in Scotland.
Labour has scrapped all plans to run billboard posters of David Cameron during the general election campaign in what it says it is a deliberate attempt to avoid “negative personalised adverts” and raise the tone of debate.
Um, right...
Pointing out the record of a party in govt isn't negative personal campaigning
If the conservatives made an advert highlighting Miliband announcing. Leftist policies that wouldn't be an attack ad, when they tweet pics of him pulling funny faces etc that is
I must say I can't really understand what the point of having professional election forecasters is. It's good fun for amateurs, but why pay people when you could just wait for the votes to be counted?
I imagine bookies for one will be interested in their techniques and accuracy! Likewise, while it does fit more into your 'wait and see' argument, the exit polling can seriously damage the reputation of the companies that commissioned and produced it (1992) or can give them great credit (2010). I guess those doing it would rather the latter.
But on a more general note, understanding how the dynamics of public opinion, voting actions and election results all fit together is simply something that's useful to know. To what extent does swingback occur, and why? How potent is media endorsement? What effect does a good or bad performance in a debate have? Do marginal behave differently from safe seats, and if so, how? What of seats where minor parties are in contention? How do you model Scotland? These are all legitimate academic questions which have a relevance that goes beyond predicting the results for May 7 to what the nature of political campaigning is - and to wider political questions still, about power itself.
I forgot to add that something which has always concerned me about pollsters is they don't ask as the first question: are you registered to vote? They just assume people are.
I forgot to add that something which has always concerned me about pollsters is they don't ask as the first question: are you registered to vote? They just assume people are.
Yeah, you teach them sitting in your Scottish hideout ! A whole industry are just a bunch of fools compared to an almost extinct specie in Scotland.
Pointing out the record of a party in govt isn't negative personal campaigning
The tag line of the advert is "David Cameron is too scared to debate his record"
That is negative personal campaigning
Do you understand the difference?
No it isn't. You are mistaken
They are saying he is scared of defending his record in govt... That is a fair comment.
If Cameron was scared of heights and the tag line was 'David Cameron is scared of heights' that would be irrelevant and unfair and you would them have a point if you pointed it out
They have definitely got the balance much better in this election than they did in 2010. Unfortunately it was in 2010 that a majority was there for the taking.
I forgot to add that something which has always concerned me about pollsters is they don't ask as the first question: are you registered to vote? They just assume people are.
Yeah, you teach them sitting in your Scottish hideout ! A whole industry are just a bunch of fools compared to an almost extinct specie in Scotland.
Thank you for that valuable contribution to the discussion on polling accuracy, I am sure we were all enlightened.
Comments
Looks pretty fixed to me.
My gut feeling tells me that YouGov's panel is bust. ICM seems to far towards the Tories, so I suspect that it's on the upper end, but the ComRes poll in the SW (admittedly small and constituency so I am naturally sceptical) is supportive evidence.
Either the online or the phone polls are going to be horribly wrong. I'd back ICM over YouGov.
My personal view is that we have a small tory lead of around 0.5% - 1% and that will be reflected on the day, what we don't have is momentum. In 21 days time it'll be over, I don't see any game changers. The collapse of the labour vote in Scotland is hiding it's strength in England. Whatever comes out of Scotland will be anti tory so it can be discounted.
In order to get a situation where Con + LD > Lab + SNP, I think we need a tory lead of ~5%. I can't see that happening at present, therefore Ed is PM.
We have no ideas how many marginal will change hands because of a leak of 2-3000 votes to either or both of the greens or the kippers. Not enough to give either them a first three place in the local vote, but enough to change the winner. There are 50 or so seats with less than 2000 majority.
We are mostly speculating where the LD collapse vote will go, it might split to Labour and the Conservatives, or it might stay at home. Ditto the receding kipper vote. We assume a SNP riot in Scotland, but it doesn't take very many points of shift to LAB at the last minute to make it a much more modest affair given the huge majorities that need to be overturned.
We have no real idea how differential turnout, especially on the back of individual voter registration is going to effect the result, it might well show a marked dip in younger voters, and hence reduce the left leaning turnout considerably.
We talk endlessly about needing an 11%+ swing for a Tory majority, but that is UNS, and UNS is looking increasingly unreliable in a multi party election. If Labour pours on votes in the inner cities and bleeds votes to the kippers in the suburbs it might need a lot less, if the Tories pile on votes in the shires and loose seats to Labour in the Midlands they might need more.
Even if I wasn't out the of country until several months after the election, my money would be staying in my pocket for this election with the exception of Scotland.
50 hours 50 minutes 50 seconds
Clearly I wish these pale imitatiions well as they fumble around in the dark and then strangely begin to ease grudgingly around my ARSE.
Apart from being very noble Mrs JackW was getting a little tired of offloading OGH's competition prizes of various cases of hair tonics and restoratives down to local church bazaars.
Mind you the pews around North Hertfordshire are now full of some interesting elderly men with "winning here" hair styles.
http://lordashcroftpolls.com/2015/04/south-ribble/
If that were to show a Tory lead of, say, 5% then the view that the Yougov panel model is broken would gain a lot of strength. If it were to show something similar to Yougov then conversely the ICM would really look like a fairly extreme outlier.
At the moment in the competition I have gone for a modest Labour plurality largely based on the majority of polling. I may well change my mind and get somewhat closer to Jack's ARSE if Ipsos Mori showed a significant Tory lead.
That sounds great, but he might have a slight problem... http://www.telegraph.co.uk/news/politics/david-cameron/11540433/David-Cameron-SNP-will-be-chain-to-Labours-wrecking-ball-if-Ed-Miliband-wins.html
Twa cheeks...
Given around 25 losses to the SNP, Labour on 300 seats implies approx. 60+ gains from the Tories .... Hhhmmm.
http://ukpollingreport.co.uk/blog/archives/9350
Phone polls seem to favour the Conservatives. However this could be an effect of internet polls seeming to favour UKIP/Greens.
That sounds great, but he might have a slight problem... http://www.telegraph.co.uk/news/politics/david-cameron/11540433/David-Cameron-SNP-will-be-chain-to-Labours-wrecking-ball-if-Ed-Miliband-wins.html
Twa cheeks...
Isn't Nicola Sturgeon already backing down wanting Barnett formula rather than financial independence?
@Claire_Phipps: Labour gives @BuzzFeedUK first look at its latest anti-Cameron ad http://t.co/oV3IoWnkSA #ELECTION2015
Besides it has potency.
The divergence between phone and internet polls is interesting if only because at some point in the future phone polls will be choosing from a diminishing pool of responders.
This could be mitigated by increasing the percentage of calls to "mobiles", but is unlikely to be fully effective.
(people on mobiles will not generally respond to polls unless they are in their homes at the time)
Maybe i'm missing something - but as I understand it, the internet pollsters have a big panel of, say, 100,000 people - and, each time a poll is commissioned, *invite* ~1000 people to fit a demographic profile.
Surely it would make more sense to only poll specific individuals in marginal constituencies (who are on the electoral roll) and then repoll exactly the same people to see if they change their mind. Repoll these people immediately after they've voted - rinse & repeat every month for the next 5 years.
Then, instead of weighting the responses, converting to vote shares and baxtering - we would just get direct seat projections.
I guess my question boils down to - why use random sampling when measuring changes in opinion over time, when individual opinion changes can be captured cheaply, accurately and regularly - and are probably more useful in determining the end result (which seats actually change hands)?
@Douglas4Paisley: Watching this, you can see why David Cameron is too scared to debate tonight. https://t.co/n6Opwwv2L6
Are you sure UKIP's voters are mainly elderly, as opposed to say middle aged?
Not sure why that is not consistent with this 'video'.
There must be some better method which we just haven't thought of yet though. Use Facebook or Twitter in some way, or back to approaching people in the street?
Unless you are value trading in the market then betting on Labour and Ed Miliband at that seat level shows a level of faith that hasn't been seen since "Stuart Truth" was prognosticating on PB over two years back.
That ended well.
Umm, not very, Bob...
@afneil: IMF says tax revenues will be 36.1% GDP by 2020 instead of OBR's 36.3%. Be sceptical of such decimal-point forecasts so far into distance.
Possibly combining the two? The point about mobile users being more reluctant to respond to polls while on the move is down to how the users tend to perceive calls to them in different situations.
He knows my ARSE from the YouGov ELBOW.
The net pollsters had 'gap' inaccuracies going up to 5.5%
Which neatly summed it up!
I'm no psephologist - obviously - but it does strike me that dismissing internet polls is a bit silly. With all the extra data that can be captured with technology, surely it's only a matter of time before the methodology of internet polls becomes more accurate in achieving the ultimate objective (the constituency tallies) than the old tried & trusted methods. We might not be there yet, I'll grant you
OGH is clearly devastated that a Tory poster has appeared in his area of Bedford for the 1st time.
http://www.huffingtonpost.co.uk/laurence-stellings/ukip-poll-voters_b_6631026.html
In fact, the big UKIP demographics are the ones generally that are unlikely to have everyday internet access. So, I am not sure why phone pollsters would find more elderly Tory voters, but fewer elderly UKIP ones.
Does that amount to an admission that Labour did at some stage actually have such plans?
I couldn't possibly comment.
If EIC is looking like a winner one imagines the non-doms will be contacting their estate agents.
They have definitely got the balance much better in this election than they did in 2010. Unfortunately it was in 2010 that a majority was there for the taking.
If we had a well-behaved, spam-free internet you'd just email people. We don't, and the only way round that is an opt-in. People self-select by opting in, and then again by not dropping out when they realise what a ballsachingly tedious exercise in futility answering an online poll actually is. (It is very much more than a simple "How you gonna vote?") This may not matter because you get equal proportions of oddballs on each side, but that isn't obvious. There is also the following purely theoretical danger: if one major party said to each new activist recruit, by word of mouth, that they were expected to sign up to internet polling panels, and the other party didn't, that would distort the results. There is no evidence AT ALL of this happening but on the other hand I can see no way of effectively guarding against it.
It's sort of like electing the Pope: just as wanting to be Pope is a very undesirable quality in a potential Pope, you don't want to poll people who are positively eager to be polled.
114.4 (pre-method change) implied vote= 33.98% (last six samples)
120.2 (post-method change) implied vote= 35.69%
Conservative 2015:2010 vote ratio - Yougov
94.5 (pre-method change) implied vote = 34.93% (last six samples)
89.6 (post-method change) implied vote = 33.09%
If you look at the last 2 weeks of YouGov polls, the Tories have been 33-37 and Labour 33-36 and almost all the leads either way have been caused by +1 or -1 for one side or the other. The LibDems, UKIP and Greens have barely shifted, again maybe +1 or -1.
On Monday voter registration closes. I hope Andy JS does one of his excellent spreadsheets on the size of constituency electorates and the difference from 2010. It will be interesting to see if in marginal seats, particularly those with large transient communities like students, there has been a substantial fall in registrations.
I remain convinced we will see the final votes at something like Tory 35-37, Labour 29-31, LibDem 12-15 and UKIP 10-12. In Scotland I expect the SNP to possibly pick up 25-30 Labour seats. If they pick up more then Ed is looking like ending up with fewer MPs than Gordon Brown did in 2010.
Anecdote alert: My eldest cousin lives in Jim Murphy's seat. We were discussing how it looks for him last night. She said she and her neighbours like Murphy but wouldn't dream of voting for him. Her view is they wouldn't be surprised of Dr David Monty scrapes through the middle of a dogfight between Murphy and the SNP. I wouldn't go that far but would be thrilled if it comes to pass.
Mr. Mark, I saw Peston last night (on the TV, I hasten to add). I kept waiting for him to point out the IMF had criticised Osborne's plans, only to end up praising the UK economy as the model to follow.
Only it seemed to have slipped his mind.
http://www.bbc.co.uk/news/world-europe-32325884
"The suggestion that you can laugh at everything, except certain aspects of Islam, because Muslims are much more prickly that the rest of the population - what is that, if not discrimination?"
Generally, the Labour online and telephone poll scores are pretty much the same. The big variances seem to be the UKIP and Tory scores. That's one of the reasons I see a big Tory seat lead. When push comes to shove a lot of UKIP-inclined voters are going to go for the Tories. It could be that ICM has found them further down the line to making that final decision. On a really bad night for Labour, there could also be Tory tactical voting for UKIP and the LDs, which would see Labour come in much lower than 250 seats - especially if it is an almost total wipe-out in Scotland.
Pointing out the record of a party in govt isn't negative personal campaigning
If the conservatives made an advert highlighting Miliband announcing. Leftist policies that wouldn't be an attack ad, when they tweet pics of him pulling funny faces etc that is
Do you understand the difference?
But on a more general note, understanding how the dynamics of public opinion, voting actions and election results all fit together is simply something that's useful to know. To what extent does swingback occur, and why? How potent is media endorsement? What effect does a good or bad performance in a debate have? Do marginal behave differently from safe seats, and if so, how? What of seats where minor parties are in contention? How do you model Scotland? These are all legitimate academic questions which have a relevance that goes beyond predicting the results for May 7 to what the nature of political campaigning is - and to wider political questions still, about power itself.
That is negative personal campaigning
Do you understand the difference?
They are saying he is scared of defending his record in govt... That is a fair comment.
If Cameron was scared of heights and the tag line was 'David Cameron is scared of heights' that would be irrelevant and unfair and you would them have a point if you pointed it out
Do we know who their ad agency is?