The same point keeps coming up in comments – if there is some great surge of young voters backing the Liberal Democrats, would it be picked up in the polls?

The short answer is that it should be. Pollsters in the UK do not weight their samples to match the demographic profile of voters, they weight them to match the demographic profiles of UK adults, so most of those young people who wouldn’t normally have voted should have been represented in the samples anyway. Only Harris specifically ask respondents if they are registered to vote, so they would not have been filtered out of other companies’ samples.

Another point I’ve seen raised is how well pollsters cope with young people who are predominantly online or have only mobile phones. Firstly, online really isn’t a problem, since we have plenty of online pollsters. Secondly, all polls weight by age, so this cannot result in an under-representation of young people. The problem would be if mobile-only young people were significantly different to young people with land lines. If landline penetration continues to fall I suspect that this will eventually be a problem that phone pollsters need to tackle, though there is always the option of including mobile phones in samples.

Perhaps the most common question I’ve seen is whether a change would upset pollsters “assumptions”. Pollsters do not assume particular groups are more or less likely to vote. Instead the majority of pollsters factor in likelihood to vote by asking people how likely they are to vote, and then weighting or filtering appropriately. The correct proportion of young people are already represented in pollsters samples, so if those people told pollsters they had become more likely to vote, it would be picked up.

The bottom line on a lot of questions of whether pollsters would pick up a new trend is that pollsters don’t actually make many presumptions up front about how people will behave. With a few minor and well evidenced exceptions (such as Populus and ICM’s reallocation of don’t knows on the assumption they are likely to vote how they did last time), voting intention figures are based on what people tell pollsters, not the pollsters’ preconceived assumptions.

Things can go always go wrong of course, and I expect it’s more likely to happen at an election where there has been a large shift in support, but I can’t see any particular reason to expect the polls to get it wrong this time.


104 Responses to “Young people and the Lib Dems”

1 2 3
  1. This is a very interesting point regarding home telephones, which are on their way to being obselete among young people. Equally if YouGov only poll on the internet, then those people without internet (one presumes to be the older generation!) perhaps might not be included correctly?!

    At a guess I would say that there is probably no voting preference amongst people without the internet, but people without landlines will tend to be the younger generation which might meant that the Lib Dem vote is not represented fully in telephone polls.

  2. Geoff, have you not read Anthony’s article?! The point is that however under- or over-represented any group is by polling methods (or chance) it does not affect the results, as the pollsters know what proportion of those groups should be included and weight the data accordingly.

  3. Anthony,

    Yes, I am inclined to agree entirely with that.

  4. @GEOFF: Geoff, I believe that what Anthony Wells is trying to say is that the fact that many young people don’t have land lines is not a problem (because pollsters weight their samples by age) UNLESS young people without land lines are SYSTEMATICALLY DIFFERENT from young people with land lines. So, unless there is such a systematic difference (which is admittedly possible), the Lib Dem vote should be adequately represented in the WEIGHTED results.

    With that said, a poll is of course only a poll. If a poll were completely accurate, we could do away with the actual election! But it won’t be, and so we can’t. The same is true of a prediction market.

  5. Sorry I see that now. I’m too embarrassed to ever comment on here again.

  6. So no under 35s surprise then. That’s a shame. With respect to the pollster’s reputations, surprise results are fun. I can still remember Ian Hislop on Have I Got News For You making fun of high profile Tories who lost their seats in 97.

  7. @ Anthony

    Thank you – I have been patiently explaining this to my son & his friends since “the surge”.

    i.e. There is not a big pool of uncounted young voters missing from YouGov polls.

  8. Geoff – don’t be! (embarrassed or sorry). Hope I didn’t sound sharp.

  9. Sue:

    Thanks for your earlier comments: I have tried, extremely hard, to be objective in my posts whilst at the same time acknowledging an extremely strong anti-Conservative bias and a longing for a voting system that gives us truly representative government in the future.

    If you read the earlier exchanges you may have noticed that I didn’t receive a reply from Eoin, as to why he chose the last 15, versus previous 15 polls, to illustrate a miniscule downward movement in the Lib Dem position. A cynical person might have observed that fifteen took us back, exactly, to just after three very high Lib Dem ratings.

    I asked “why not 18, for example?”

    As I didn’t receive a reply, here are the LD polling figures for the last two groups of eighteen polls [to be quite clear, this as an equally arbitrary figure, but I’m sure you’ll get my point}

    Most recent: 29.5%

    Previous: 27 %

  10. @Éoin

    To carry a subject over from the last thread, I wouldn’t say that 1983 shows that UNS holds up in a three-party race. In that instance Thatcher was miles ahead, and the Alliance was so evenly spread as to render it’s poll boost ineffective. Analysis of the regional Yougov polls has indicated that Labour would lose about 30 more seats than UNS is predicting, giving 20 to the Tories and 10 to the Lib Dems. I think we can pretty safely assume that come election day similar results will be seen.

    As far as predicting the outcome based on the polls goes, Yougov in particular has called it right in recent elections, but there weren’t so many unknowns in those. The ICM figures showing the Lib Dem support both as the most volatile, but also as potentially the largest, are probably the most interesting we’ve had in several days. All of these newly registered and young voters are, almost by their very definition, soft. That said, they lean heavily to the Lib Dems, and it’s the turnout of this group which will determine the result of the election, one way or the other. I recognise that the pollsters are very careful with their weighting, but that produces numbers that are in the middle of a distribution. The nature of the surge voters and the unusually large size of that distribution means, IMO, that the final result for the Lib Dems is going to be much harder to predict than that for the other two parties.

  11. @jones,

    With the greatest of respect I have fielded about 100 similar posts in the last fortnight since the rise of the Liberal Ds. They ahve got steadily nicer in the way they are addressed to me and yours in certainly in a respectful tone. I am just all out on repeating myself. I reccomend you read the following article.

    Forecasting the 1983 British General Election
    Philip Brown and Clive Payne
    Journal of the Royal Statistical Society. Series D (The Statistician), Vol. 33, No. 2 (Jun., 1984)

  12. for the geeks among us interested in academic studies into various elements of polling/voting i also reccomend

    Picturing the 1992 British General Election
    Graham J. G. Upton
    Journal of the Royal Statistical Society. Series A (Statistics in Society), Vol. 157, No. 2 (1994),…

    Attitudes and Votes at the 1983 General Election: Explorations of the Geography of ‘Deviations’
    R. J. Johnston and C. J. Pattie
    Area, Vol. 20, No. 2 (Jun., 1988), pp. 111-119

    Perceptions of Economic Performance and Voting Behavior in the 1983 General Election in Britain
    Paul Whiteley
    Political Behavior, Vol. 6, No. 4 (1984), pp. 395-410

    Inter-Constituency Migration and Turnout at the British General Election of 1983
    David Denver and Keith Halfacree
    British Journal of Political Science, Vol. 22, No. 2 (Apr., 1992), pp. 248-254

  13. @Jones,

    The last minute surge in registration of 100,000 voters added c.150 voters per constituency.

  14. What if the demographics have changed, e.g. if the surge of late registrations is disproportionately younger people (as has been reported) could they now form a larger share of the electorate than in the pollsters’ model?

    “All of these newly registered and young voters are, almost by their very definition, soft.”

    Are they? If you’ve gone to the trouble of registering for a specific election might you not be more likely to vote than the average?

  15. @Paul Croft

    I did reply

    it is on an earleir thread

  16. @Paul C,

    you asked me “Why 15?”

    My answer: There has been 31 polls since the first debate. I just split them in two. As simple as that

  17. @Paul C,

    The downward trend was not miniscule it was c.2%.

  18. @RogerH
    “What if the demographics have changed, e.g. if the surge of late registrations is disproportionately younger people (as has been reported) could they now form a larger share of the electorate than in the pollsters’ model?”

    Not just late registrations, but those already registered but unlikely to vote pre-LD surge. If turnout is 70-75% with many younger voters, the model will be compromised. I also agree that they is likely to be hard not soft.
    Add to that the voter’s change of mood on hung parliaments and a GB outright majority is unlikely.

  19. > The last minute surge in registration of 100,000 voters added c.150 voters per constituency.

    The Electoral Commission reported registration downloads up from 50,000 to 300,000 (compared to 2005) on the Friday, four days before the deadline. And, of course, it won’t be evenly distributed, as you’ve said yourself. One probably shouldn’t read too much into it but it’s another variable to throw into the pot.

  20. RogerH – the demographics haven’t changed, remember these are people who are newly registered. They are not new people.

    Pollsters weight to the profile of the adult population, not the profile of the registered adult population. They were always included in pollsters samples.

  21. @ Anthony

    Great post. Sets a lot of things straight.

  22. @Éoin

    Entirely understandable. I accept that you’re much better read on the subject than I, and probably have a far better understanding of polling methods. I don’t think that UNS will be wildly inaccurate, only that it would be reasonable, based on the recent regional polling and current state of affairs, to say that it may be off by perhaps 10% in terms of seat numbers. Not enough to majorly transform the result of the election, but perhaps enough to mildly reduce the Labour bias in FPTP.

    My second point wasn’t necessarily directed at you, but I would say that contemporary polling methods have yet to deal with an election with as many variables as this one. Indeed, given the current three-way-tie this is the trickiest election to predict since the inter-war period. I don’t think that the unpredictability of the Lib Dem vote is more likely to result in their gaining a massive boost (though I’d be only too happy if it did). Nor do I subscribe to the idea that polling is failing to reflect Lib-Dem voters. Indeed, they may do worse than the polls are suggesting. My feeling is only that they won’t have it to the nearest percent this time around.

  23. @RogerH

    Yes that is right- i was talking abou tlast minute surge just which from March 15 =100,000

    the total added to each constit of overall registrations is c.500 per seat. ( from 2005)

  24. The Independent reports that the proportion of postal votes in some marginals is as high as 30%.

    25 constituencies won’t count their votes until Friday. Could they keep us on tenterhooks?

  25. The thing about this election is the ‘cat is out of the bag’ and unlike the last election where the Tories won more votes than labour in England,and it was ignored, this time England will see Labour 3rd in % of votes if these polls are to believed and 2nd or 3rd in seats in England, and only their Welsh and Scottish votes/ seats will keep them competitive.
    And with the clear view that the electroral system needs changes from those polled, no matter what the outcome, even an unlikely Lab largest seats scenario, will mean any inconsistancies will be highlighted to the maximum especially by the Lib Dems who are obviously going to be the party with the momentum after the elction no matter what the outcome.

  26. @jones,

    I could give you many examples I think will buck it. And I would happily hear some of yours. Dumfries I suspect will buck it. Finchely may well buck it. Dundee East may buck it. i a m sure every poster has their own idea of what seats might buck the trend. i would be interested if you had more.

    weston super Mare might buck it… Norwich (Clarke’s seat might buck it) really the list could go on.

    I am no fan of UNS, I am simply too well brought up to cast someone else’s hard work in devising the model completely to one side. I feel it would be arrogant to do so.

  27. @ Anthony

    Anthony once again I find that my calculations from your polling table behind the “…More” line for the UKPR Polling Average produces a different result than you have.

    The total of the weighted products of Conservatives is 332.54, divided by 9.90, the sum of the weights, gives 33.6 or 34 rounded. Similarly for LibDems their weighted total is 287.17 divided by 9.90 gives 29.

    So instead of 33 / 27 / 30 (Labour short by 63) shouldn’t you have 34 / 27 / 28 (Con short by 64) – I think? I’m faithfully using your weights and I’ve posted notes to you like this before so please let me know if I’m not using your table correctly.

  28. @ CLIFF

    Lib Dems who are obviously going to be the party with the momentum after the elction no matter what the outcome.
    ————————————————–
    Do you think voters’ interest in politics will survive for long after the election? I rather doubt it.

  29. @ Anthony

    “Pollsters weight to the profile of the adult population, not the profile of the registered adult population. They were always included in pollsters samples.”

    I ask out of ignorance but isn’t what matters not the demographic of the adult population but the demographic of the voters (for which I imagine pollsters adjust their figures)? If the voter demographic has changed because of a higher ‘young’ vote won’t that compromise the accuracy of the pollsters’ weightings?

  30. @ Anthony

    “shouldn’t you have 34 / 27 / 28”

    Sorry that should have been 34 / 27 / 29, finger trouble :)

  31. @MIkeP,

    Do you have LD on 28.52? ;)

  32. @ Anthony

    Actually I should have read your original commentary more closely as it answers my questions!

  33. I have read a couple of times that Asian communities are less likely to vote in this election than the last. The LibDems received a strong anti Iraq vote last time. Does anyone have a view on whether this may be resulting in possible overstatement of LibDem support now (I am assuming the weighting methodologies for turnout don’t differentiate different communities propensities to vote, but may be wrong here.

  34. @RogerH

    The same point came to my mind. I think that AW was saying that from the answers that people polled eg Young people saying they were undecided or they were not intending to vote , this adjustment was being relied upon to then adjust to the demographics of the voting population.
    I don’t know about you but the method doesn’t seem entirely satisfactory to me?

  35. MikeP – all the weightings there are rounded, so you won’t get the same figures doing it by hand.

    RogerH – since there are no accurate figures for the make up of the voter population, pollsters can’t weight to it anyway. In practice it shouldn’t really matter – the adult population contains within it the voter population and as long as unregistered people are relatively accurate in stating their likelihood to vote, it will produce an accurate result.

  36. @ Anthony

    “Do you have LD on 28.52?”

    Nope, I have 29.01, by dividing 287.17 by 9.90

  37. It’s interesting that only Harris filters for registration & consistently produces a lower Labour vote percentage. It may be that other factors are at work, but that could make them more accurate. I can see that registration may be unimportant when elections are months or years away, but it must be important with 10 days to go.

  38. @Éoin

    I don’t think we should discard UNS. It may not be perfect, but it’s the probably the best we can do with limited polling information. Anyway, it’s a subject that’s been raked over a lot recently, so I’ll shut up about it now.

    @RogerH

    “Are they? If you’ve gone to the trouble of registering for a specific election might you not be more likely to vote than the average?”

    _

    True, but these registrations happened in the light of a very particular mood, and that has receded slightly in the intervening period. I was also referring as much to their decisions about who they vote for as much as whether they turn out or not. These voters lack a long-term allegiance to any one party, and would probably be more likely to change their minds once they’ve started to pay more attention to the election than a voter who’s been engaged for longer.

  39. @ Geoff – don’t be embarrassed! We’ve all said stupid things here from time to time, some of us more than once – eg me, and I’m generally AWESOME (according to polls conducted by me) so if I can say something stupid, it can happen to anyone ;-)

  40. @ Anthony

    “all the weightings there are rounded, so you won’t get the same figures doing it by hand”

    Oh shame, ah well I’ll stick to your answers then – unless you could add a place or two to the weightings you display? ;)

  41. @ Amber Star

    I think the turnout n this election will be lower than last time at about 55% -60%, but the people who do vote are the interested ones, and they will keep the PR agenda.

    I also think a lot of people are going to vote for ‘others’ but in a more calculating way this time, so Yes the interested ones will keep it going, and it will have support because its in every parties interest except Lab who want STV & Con who want FTPN to bring it about, and this time they will have the numbers to make it a serious issue.

    And just to put the cat among the pigeons, Scotland could be independent by Christmas so all bets could be off

  42. Jones,

    I would be interested in the seats that you think will buck the trend…….. if you have time to do so of course….

  43. Read with interest views, ratios, trends, % and history as and when the regular polls appear.

    For comparison to views on this site i have visited Electoral Calculus.. Being interested in comparison they state sampling 19135 people over period 19th/24th April they predict at the moment::-
    Con 35.3 297 seats
    Lab 27.02 227 seats
    Lib 27.43 94 seats
    Others 10.24 33 seats

    Cons 29 seats short of majority.

    Interesting also that the political betting market are stating the following as the possible outcome:-

    Cons 250/299
    Lab 200/224
    Lab 100 +

  44. It seems that you could compare predicting polls with forecasting movements on the Stock Exchange. There are two methods-one to look at past trends , so called chartists or technical analysis. The other method is to look and see if one, say, share is underpriced or overpriced compared with others. This is called fundamental analysis and is more successful in predicting share price movements.
    With regard to opinion polls and parties’ standing in them, you would have to eliminate all consideration of past trends and consider which party is over-rated or under-rated or fairly rated, and then get out of the market as soon as possible before it changes again.

  45. @ John Brown
    @ Eoin

    “Cons 29 seats short of majority”

    Wow, that’s quite different from what we’re seeing on UKPR.

    Eoin, didn’t you post something a while back about a fundamental difference in Electoral Calculus methods ?

  46. @ Jones

    “These voters lack a long-term allegiance to any one party”

    Isn’t that true of many more voters? By definition the allegiance of existing voters can’t be greater than the respective party votes at the 2005 GE. The loyal voters won’t decide the result.

    @ Cliff

    “I think the turnout n this election will be lower than last time at about 55% -60%”

    Most commentators seem confident it will be higher than 2005.

  47. @ EOIN

    To be fair with LibDem voters, whilst i subscribe to most of your assumptions, as they clearly are validated with figures, facts and historic trends, I don not agree that the Lib Dem vote will implode, the numbers are simply not there to show this.

    true, over the last week LibDem votes have dropped compared to the week before. It does in fact drop from 34 to 29 (based on YouGov numbers), but trend wise there is a drop, albeit small which i doubt will go under 27-26% by 6th May

  48. Xiby

    “implode”

    Not me sir!

    :)

    Lib scored 22.7% in 2005. I more or less see them matching that. If they have got better at targetting their vote, which it strikes me they become continually efficient at, then I anticipate about 10 gains from red.
    how many from blue I don’t think I could be able to say.

  49. This might be bilge, but I have a hunch that the type of young people who respond to landline calls will fall disproportionately into the following categories. Those who still live at home, and are thus less adventurous types and more likely to vote like their parents. Those who have chosen a more settled and conventional or old-fashioned lifestyle that they have ordered and installed a landline.
    Alternatively, those who rely upon mobiles are likely to be the more adventurous freedom loving types who often have gone away to study or live independently from their parents, and are therefore potentially disproportionately likely to be LibDem this time.
    So, indeed, it may be skewing the polls because of the type of young person reached?
    I admit there is no evidence for this theory in the various polls themselves – but might I have a valid point? If not, please shoot me down in flames!!

  50. I would say Nick Clegg probably has the most brand recognition amongst voters in general .He’s a fitting successor to David Lloyd George and JeremyThorpe.

1 2 3