NILT: Does it tell us anything beyond it may still have sampling issues?

20% of nationalists aren’t nationalist

29% of people that support SF/SDLP don’t think of themselves as nationalist

42% of SF/SDLP supporters don’t support Irish unity

55% more people support the SDLP than SF

41% of Catholics support the SDLP. Only 27% support SF.

Anyone want to add their spots on survey details that, yet again, seem to clearly conflict with reality?

And this bullshit survey is yet again being treated as a serious barometer of political attitudes in the north?

  • I wonder why SF or the SDLP or the Irish News or the ATN… or even a combination of all 4 haven’t commissioned their own independent opinion polll before now.

  • Mark McGregor

    oneill,

    Why would they? We have regular accurate polls of the people that count – they are called elections. They always seem to be at odds with the NILT attempt to survey attitudes. Funny that.

  • Why would they?

    To prove the NILT wrong with more impressive methodology.

    Only 55% of the electorate vote now, meaning 45 % have no opinion then on, for example, our constitutional future?

    They don’t do it for exactly the same reason they don’t employ PWC or some other beancounters to produce an independent feasibility study into the economic implications of a UI.

  • Mark McGregor

    I love the way people seem to think some academics have managed to survey the non-registered and non-voting and made them into a demographic.

    Great the way QUB managed, yet again, to engage these people more than numerous elections!

    (of course none of the NILT returns that continually conflict with reality would ever be reflections of the authors’ hopes/expectations?)

  • alan56

    I suppose everyone likes surveys when the result suits them. When it doesnt then its flawed. Politico’s do it all the time. Serious analysts should be a little more thoughtful in their reactions.

  • Great the way QUB managed, yet again, to engage these people more than numerous elections

    If you go out of your way to ask me my opinion on who is the best candidate standing in Castlereagh East, I’ll more than likely give it; not sure that I’d be that motivated one way or the other to make my way to the polling station to confirm it where it matters.

    But you haven’t said why you think eg the Irish News hasn’t employed a polling company to do the same job as the NILT?

  • Zig70

    The survey maybe needs to answer it’s critics and add in a question on ‘did you vote and who for’. Data is king. To me it highlights that the gift of a UI is mainly in the southern government lining up ducks, but it hasn’t given the issue a moments notice since the early seventies, from what I can see. You don’t get the feeling off being wanted. That for me is a big issue.

  • Chris Donnelly

    oneill

    The last time I can personally recall the Irish News quoting ‘independent’ polling figures was when an article in the paper proclaimed on the front page that Brid Rodgers was 3,000 votes ahead of Pat Doherty in the race to beat Willie Thompson in the 2001 West Tyrone Westminster contest.

    Methinks they’ve learnt their lesson since then…….

    Mark’s right. There is only one public opinion contest that counts.

    These types of surveys are interesting but until unionism can reliably claim to have a catholic electoral base then this is a nod to the unicorn-believing mentality oft-quoted by a Slugger regular not used to being approvingly cited by myself…

  • IJP

    Mark

    I don’t think it’s fair to imply those carrying out the survey “wanted” a particular outcome. The point is that, year after year, the survey seems to be biased disproportionately towards those who identify with more moderate parties. That has inevitable implications.

    Zig70

    You’re right in theory, but you’d be amazed how difficult that is in practice. Even during our exit poll, a lot of people were unsure how they had voted only last year, and some genuinely struggled to name the candidate they had just given their first preference to!

    The key would be a bit of sensible weighting.

  • John Ó Néill

    On one hand, the ability to suspend disbelief and promote a survey whose measure of political sentiment is unrecognisable when compared with actual elections in 2010 and 2011, is impressive. But then, we’d a statistics office who decided that people who didn’t indicate a religion in a census (hardly an insignificant act here) should then be assigned to a religion for statistical purposes.

    Did the survey mention hallucenogenics?

  • I think that there are actually two Elections for every one in the North.
    There is something called a real Election where real votes are cast.
    This mightily depresses unionists.
    Then the Belfast Telegraph or some other body will commission another poll or election.
    The sole purpose of the second poll is to bring consolation to unionists. And they get to feel a whole lot better.
    They can content themselves that the “pretend” election is actually the real election.
    Nationalists are supposed to feel annoyed but funnily enough I have never met anyone who actually cared.

    Mr Donnelly is of course right to point up the West Tyrone situation. But I seem to recall an opinion poll in West Belfast around the time that Gerry Adams and Joe Hendron were in contention. I believe it was conducted at the corner of Falls Road and Clonard Street.
    It predicted a landslide for Adams of PDR Korean proportions.

  • AntrimObserver

    The NI Life and Times survey has consistently underestimated the true level of support for SF amongst the nationalist population.

    NILT interviewers conduct the survey on a face-to-face basis in respondents’ homes – so the anonymity and privacy of the polling booth is lost. Think about it, a stranger (whose religion or affiliation you do not know) comes into your house and asks if you support a extreme, republican party. People naturally opt for the ‘safe’ answer Is it any wonder the SDLP has fared so well in the NILT figures yet so badly in the private, polling booth?

  • RyanAdams

    There can be no doubt about it; 1200 is just too small a sample to be anywhere near representative of NI.

    However, We have seen great change in this country since 2007. Its difficult for me to say it, but a united ireland as opposed to direct rule in years preceding 2007 was much more attractive then Peter Hain, Maria Eagle and co running the place. I can see why before devolution so many catholics were in favour of a United Ireland.

    Now our own local politicians are back in government, and unionists can’t do anything with out nationalists consent in the assembly and vice-versa, growing numbers of catholics are accepting the status-quo. Given the dropping assembly election turnout, I think it would be fair to say fewer people want to change that status-quo.

  • orly

    How come the figures for “prods” seem to be a lot closer to the money if the sampling is so naff?

    Do “catholics” just lie more for the laugh?
    Are they ashamed of something?
    Do they enjoy setting Barry McElduff up for a meltdown – again for the laugh?
    Or is voting for SF and the SDLP just the “tribal” thing to do even though many seem to harbour reservations about the whole UI project?

  • Mick Fealty

    I applaud Chris’s acceptance of the Baker doctrine. I do, however think we ought not to continue shooting the messenger just because we don’t like the message.

    If the fault is with the sampling method let’s try to identify itrather than identify outcomes which appear to contradit common knowledge/sense.

    You put surveys like this in the field precisely so you can dig beneath common knowledge. There are some tough and portentious questions being asked of nationalism here. You can choose to that as a threat or an opportunity.

  • AntrimObserver

    Orly wrote:

    “How come the figures for “prods” seem to be a lot closer to the money if the sampling is so naff?”

    Because, historically, openly admitting support for, say, the DUP, did not carry with it the same implications in terms of personal security, job prospects etc as admitting to being a Shinner would have done. People just don’t like admitting to some stranger from a government funded agency (e.g. NISRA) that they support SF.

    Nationalist’s, and especially republicans, long adhered to the ‘whatever you say, say nothing’ rule. It’s no coincidence that it was the republican community that in the past fiercely opposed the Census.

    No, the NILT sampling procedures are fine both in terms of the size of the sample (1200 is more than adequate for our population) and in terms of the randomness of the household selection procedure (addresses are selected from the Land and valuation Agency’s database to ensure a wide geographical dispersion).

    The problem lies in the way the Life and Times survey is conducted: face-to-face and in the respondent’s own home. People clam up. They give a safe answer. And in these days of identity theft (and governments losing data on train seats) is it any wonder?!

  • Naughton

    I debated whether to post on this thread at all, as Slugger is once again slipping into that phase where the usual suspects tout the usual lines and those who aren’t sufficiently true to their nationalist/unionist roots slope off and shake their heads ‘not cross, just disappointed’

    But for what its worth…..

    Sampling – NILT sample is a little over 2000 randomly selected across NI, the response of about 1200 is par for the course (638 refusals is also par for the course). The results of such a survey are a report purely on those who responded – its not a census its a sample – so there is the potential for statistical bias. That may be the case, but year on year sampling showing a trend would need hard evidence to refute, not opinion based on “My friends don’t think like that”.

    I listened with interest to Barry McElduff’s incredulity and conspiracy theories on Talkback and dispaired that, once again, ignorance of how statistics work and can be interpreted becomes another whataboutery special. Barry (and others on this forum) cannot compute how election votes can be reconciled with the survey results, but the two are very different exercises and ask different questions.

    Firstly the Election is a record of the votes of the 55% of those who have registered to vote – i.e. those who haven’t registered aren’t counted and those who didn’t vote aren’t either. Without some evidence of preferences and policy priorities of the non voting block, it is dangerous to make assumptions about them, but too often posters on Slugger assume a straight catholic=nat, protestant= unionist.

    btw Barry’s assertion that he knows exactly what SF/SDLP voters think when they mark a ‘1’ on a ballot paper was breathtaking in its ignorance and arrogance.

    Secondly, those voting may (and many do) use their vote tactically. I voted for Conall McD 1, Anna Lo 2 as I thought he would keep a balance in the S Belf constituency that suited this mixed marriage, catholic, Alliance supporter. My wife (a former Ulster Unionist voter), voted Alliance 1 and SDLP 2 as she can’t stand Michael McGimpsey. So, based on a 100% sample of my house and Barry’s insight into voting we must be in more in favour of a United Ireland than not, but that’s just rubbish!

    Thirdly the 600+ who refused to engage with NILT may be an interesting group – some will be too busy, some will never fill in surveys, and I suspect, but can’t prove, some will on political grounds not want to help a ‘Northern Ireland’ survey of any sort – which may skew the results.

    And that is a valid question for ARK to answer – can they provide a further breakdown – perhaps geographical, religion etc and over a number of surveys? That would help identify if there is a statistical bias creeping in.

    My personal experience and that of many I work with is that a binary approach to national identity is lazy and often inaccurate. I’m Northern Irish, Irish and British in that order, my wife would swap the British and Irish, but would still see herself as all three. Some of our friends would drop the Irish element and more woudl drop the British, but all would have Northern Ireland in there with at least one other facet of identity ( German in one case!).

    Mick asked in the previous thread whether this survey is picking up someting about a dislocation between the parties stilled wrapped in their flags and an electorate which is increasingly disengaged. It would need much more evidence to be sure, but I suspect that even backed with the most robust evidence possible, many will dispute it or dismiss any suggestion that the question which has had so much blood spilled over it is increasingly irrelevant.

    The sad truth is a binary ‘us vs them’ ‘UI vs UK’ view is much easier and comfortable than a genuine shared identity.

  • Chris Donnelly

    Mick
    Unless Pete Baker has morphed into Turgon I think you’ve the wrong man for the doctrine there…..

  • pippakin

    Why are ‘republicans’ so aggressive/defensive about the results of a poll? after all the years of violence and hatred it can hardly be surprising to find that some, perhaps most, people are conflicted about nationality. The real change where people can make ‘clean’ decisions will take time and a new generation unaffected by the past.

  • “the two are very different exercises and ask different questions.”

    Agreed, Naughton. For example, I decided this year to allow my reaction to political responses to the Rathlin ferry scandal and related issues to influence my voting pattern. As a consequence some individuals and parties didn’t get a number in their box this time around for the Stormont and local elections. I felt no obligation to ‘reward’ those individuals/parties who ‘facilitated’ misgovernance.

    As for the constitutional question, I predict that my particular shared sovereignty option won’t appear in a referendum – or in an NILT survey.

  • Mick Fealty

    Ah.. I redact Chris!! Shouldn’t dabble in mindreading! My bad!!!

    Indeed Naughtons. I dont think some unionist pols are being too wise hardening it their way either. Tho pols will be pols!

    A smart nationalist response would be to probe this reality harder and see if it can yield more fruitful results.

    The ‘it’s not really real’ response reminds me of nothing so much as Scottish Labour before the deluge of the ‘Celtic spring’.

  • ayeYerMa

    Pippakin, I don’t think most people are conflicted about “nationality” at all. In NI we have the identity of British (UK + British Isles), Irish (island + 26 county passporters), and Northern Irish/Ulster (regional identity). You can hold all 3 without any sort of conflict.

    Being a Unionist, the only conflict that I don’t understand is why many (especially southern commentators) seem to think that a 26 county passport is required to feel Irish. It’s merely playing on the semantics where the 26 counties has hijacked the terms “Irish” and “Ireland”, and a throw-back to the now defunct territorial claim. It reminds me of a story when I was in the US and was about to talk to a hotel concierge when the first thing he said to me was “So you’re Irish?”. I said “Yes, how’d you know that?!” He said “Your passport”. He’d simply seen the last word of the “United Kingdom of Great Britain and Northern IRELAND” on my passport!

    PS: this thread is hilarious to see the hysterics and clutching at straws from the usual suspects!

  • Intelligence Insider

    “1200 is just too small a sample to be anywhere near representative of NI.”

    It’s the equivalent of polling almost 50,000 people in a UK wide survey!

  • AntrimObserver

    Correct, Int.Insider.

    The NILT sample size is not the problem. 1200 is more than enough for the entire population of Britain i.e. if the same randomisation procedures are followed.

    A few years back, when the conflict in Afghanistan was really intense, Coalition-backed polling organisations were trumpeting findings such as ‘Most Afghans support the American presence’ etc. Surprise, surprise these surveys had been conducted face-to-face and in the respondents’ homes. Your average Kabul resident is hardly going to openly admit to being a raving Jihadi supporter to some govt rep now is he?

  • slappymcgroundout

    “That may be the case, but year on year sampling showing a trend would need hard evidence to refute…”

    What is the trend? How about, the trend is the same failed methodology repeating itself over time. And if you’re going to make the argument that non-voters account for the bizarre voting preference result, then the burden of proof is on you and not the critic and the mere assertion does not carry the burden of proof. 11 to 17 versus 29 to 14. Explain that with something more than mere assertion that “the survey includes those who don’t vote”.

    But before you do that, pay your euros or pounds and sign up for this fine piece:

    http://www.jstor.org/pss/2683617

    One would tend to think that the speed of light would be more accurately measured since light doesn’t have a mind that processes lord knows how many variables in making this, that and the other statement or decision, but yet…

    Now back to where I was, the obvious response to the pathetic claim of non-voter accounting for SF 11 when real was 28 and DUP 18 when real was 38 is simply this:

    SDLP 17 to 14
    UUP 16 to 16.

    That’s a difference of 3%, which cannot possibly account for the DUP diff of 20% and SF diff of 18%. Whoever at the Bel-Tel who offered that lame and pathetic excuse ought to resign her/his position and hang the head in shame. The numbers indicate that the UUP sustained no loss between preference and vote while the SDLP only lost 3% from preference to vote. So in Deity’s Name is anyone offering “stayed at home” as excuse when 3% cannot possibly account for 38%.

    Do some of you not know basic math? Too much kiddie lit and art history and not enough math there at university?

    Almost forgot, but the logical fallacy in the intro quote from the one soul? The a priori assumption that the data is valid and so the trend is valid. If the data is flawed so is the trend. So go back and prove that the data is valid. Show me how a 3% drop in combined SDLP/UUP support accounts for a combined rise of 38% for DUP/SF. I won’t be holding my breath.

  • Nordie Northsider

    Not suprisingly, the BT’s stable-mate, The Sunday Indo, went for a story on this today, under the subtle headline ‘Northern Catholics ‘Don’t Want’ United Ireland’: http://www.independent.ie/national-news/northern-catholics-dont-want-united-ireland-2679148.html

    Music to the Indo’s ears, of course. No mention of the methodolgy problem other than: ”SDLP and Sinn Féin cast doubts on the accuracy of the findings contained in the survey’. This is what passes for news. Mark asked: ‘And this bullshit survey is yet again being treated as a serious barometer of political attitudes in the north?’ Sadly yes. I’ve had it quoted at me three times already.

  • slappymcgroundout

    Almost forgot, but for the only way out the box, try looking at declined to answer. That’s the only hope. Has to be some of that, since 11 + 17 = 28 while 14 + 29 = 43, so perhaps an ungodly number of SF voters declined to state preference. Ditto the DUP also, since 16 + 18 = 34 while 16+ 38 = 54. So if you want to save the survey, look at the declined to answer and then claim that all of them were SF/DUP and miracle of miracles, just so happened that they all voted and the break or split at the election was 20:18 DUP:SF (18+20=38 and 11+18=39, so you’ll need 20% declined to answers who voted DUP and 18% declined to answers who voted SF, or else some combination of that and preferences for as yet unnamed parties that rather exceeded their actual vote, ie., instead of “decline to answer” said Green Party instead, something like that)(again, can’t say that they said, SDLP or UUP, since the preference numbers are rather too close to actual vote numbers).

  • Reader

    slappymcgroundout: Almost forgot, but for the only way out the box, try looking at declined to answer.
    You’re only half way there – there is the 4th zone in the Venn diagram – answered the poll but didn’t vote (for instance – the wishy-washy). There’s a lot of slack there too.
    For anyone who can be bothered, the two overlapping circles are “voters” and “poll respondents”

  • ayeYerMa

    Any one-off poll should be taken with a pinch of salt, but given that this isn’t a one-off poll (the trend has been going this way for some time http://commons.wikimedia.org/wiki/Category:Demographics_of_Northern_Ireland) and that even considering a reasonable margin of error, this simply seems like sour grapes from Republicans.

    Remember that many of us could equally try to discredit the validity of election results too: http://sluggerotoole.com/2007/01/17/in-passing-mairtin-o-muilleoir-admits-voter-fraud/

  • ayeYerMa

    Remember also that the 2009 poll pretty much gave the same picture that this one does, and it had a 15% party support figure for Sinn Fein. This isn’t so far off their 16% of electorate. SF supporters are a zealous bunch, and not even everyone is on the electoral register.

  • PaulT

    ayeYerMa, your looking at trend totally wrong, here’s a link off your link

    http://commons.wikimedia.org/wiki/File:NILT_NIRELND2.png

    Nationalists take great heart from this, and any unionist with half a brain would be hitting the panic button!

    Roll on 2016, 0r 2017 or 2018….., the clock is ticking

  • slappymcgroundout

    “You’re only half way there – there is the 4th zone in the Venn diagram – answered the poll but didn’t vote (for instance – the wishy-washy). There’s a lot of slack there too.
    For anyone who can be bothered, the two overlapping circles are “voters” and “poll respondents””

    No. You posit theory. We have numbers.

    Survey:
    SDLP: 17
    UUP: 16

    Vote:
    SDLP: 14
    UUP: 16

    That’s a 3% difference, combined. Not even of those who expressed an opinion in the poll failed to vote, at least if the poll claims to be representative.

    That’s why Mick is entirely wrong when he babbles on about looking deeper. The abstract of the piece that I linked:

    Abstract
    Bias has different sources. Measurement errors create “bad” data and biased estimates. But selection biases occur even with “good” data and can be both subtle and large in magnitude. Selection biases are not easily detected by internal examination of the data. Detection is more likely by comparison with external data sources.

    The external data source is the vote. The SDLP and UUP are what one might expect, with 3% of those expressing a survey preference for SDLP not showing up to vote while the UUP managed to pull a 16 and 16. But SF is 11 then 29 while the DUP is 18 then 38. Again, not enough difference between SDLP and UUP survey/vote results to account for the dramatic rise of SF/DUP. It is either the most extraordinary swing on declined to respond or the survey is horribly flawed. I’m going with the latter.

    Lastly, re the notion of “trend” once again:

    A biased sample causes problems because any statistic computed from that sample has the potential to be consistently erroneous.

    So if they’re “guilty” of the same sample bias over time, consistently erroneous. See SF at 9% in their 05 poll when SF topped the SDLP by something around 50K votes in the Westminster elections.

  • IJP

    I really don’t think the “Shy SF voter” line stands up to scrutiny these days. Our aforementioned exit poll found people quite content to declare they had voted SF – more declared for SF than SDLP overall (ie the tally was roughly what you would expect).

    It seems to me to be a sampling error.

  • Reader

    slappymcgroundout: The external data source is the vote.
    Nope – it’s just a different measurement – not external. Voters are a biased sample of the population too. Can we agree that what we hope to find is the opinion of the general population on the national question; and that our data sets are:
    1) A self-selected sample of 55% of the electorate with their votes for different parties.
    2) A separate self-selected sample of 60% of the people approached; with their expressed party preferences and their opinions on the national question.
    We don’t know whether voters are more or less likely than everyone else to acquiesce to polling. I suggest less likely because:
    1) People with extreme preferences are more likely to turn out and vote – and
    2) People with moderate preferences are more likely to be willing to discuss their preferences with a stranger.

  • Reader

    IJP: Our aforementioned exit poll found people quite content to declare they had voted SF – more declared for SF than SDLP overall (ie the tally was roughly what you would expect).
    Depends on where you were standing, doesn’t it? SF may have under-reported in Kilcooley and over-reported on the Falls Rd. And the issue with the survey is that both the DUP and SF reported well under the vote. “Shy” or “scared” might not cut it – “embarrassed” might well do so. Quite right too.

  • IJP

    We were “standing” in nine constituencies!

    One of the problems with NI politics is that too much of it is based on allegation and conjecture.

    Like I say, there was no evidence at all of “shy SF” or “shy DUP”. Can you provide some?!

  • Irish Aussie

    Mick would like to know whats wrong with the sample,Ive had 5 goes at trying to get this up, i can only do this at work, so here goes, apologies for spelling and grammer.
    The key word for all polls is random random random
    The telephone with its anonimity(important everywhere), ease of contact and most importantly ability to generate random numders, makes it a brilliant tool for market research, so why have Queens gone to some property register to create a sample? as opposed to the tried and tested method.
    I,ve personally never heard of such a thing and would take some convincing that it is some how a good thing.
    The more contact you have with the punters the less random your sample becomes, an informal vetting process starts from both sides and this is a convoluted process.
    The more you ask punters to do, fill this in, respond to that, ring here, the less random your sample becomes, also think about what sort of people will involve themselves in what is a bit of an effort
    Face to Face verses anonimity
    The Australian franchise of the Gallup organisation, the people who basically started opinion polling in the 1950’s made the big statement a few years back that he’d discovered the secret to polling, face to face interviews, end result rubbish data, reputation and company in shambles.
    Anominity is important everywhere, people fear being judged
    In summary
    your poll sample needs to be as random as possible
    contact with punters as short and susscinct as possible
    as anonymous as possible
    and above all it needs to be random.

    PS nothing wrong with sample size, if anything its overkill, you can get good data from 200, if your sample is right.

    Often wondered myself why no polls about the big question, it may be nothing more sinister than cost, a 1 off 400 person survey 5years ago cost 3-5000 in oz, and to get real value it needs to be done regularly, which is probably to much for a small public sector dominated economy
    sorry if bit rushed,probably no one reading anymore anyway

  • ayeYerMa

    Another thing that hasn’t been mentioned that validates the sample used is to look at this graph of “Unionist vs. Nationalist”:

    http://upload.wikimedia.org/wikipedia/commons/c/cc/NILT_UNINATID.png

    The ratio of “Unionist” vs. “Nationalist” is almost exactly the same as the ratio of our elected political parties. Consider the large number of “neithers” and also consider that people obviously don’t vote for parties purely on one issue, then everything in these surveys makes total sense.

    On politics.ie, there is a rather eye-opening discussion on these labels of “Unionist” and “Nationalist”:
    http://www.politics.ie/northern-ireland/162795-2010-northern-ireland-life-times-survey-results-published-today-35.html

    Essentially there are zero Protestants embracing the label of “Nationalist” (http://upload.wikimedia.org/wikipedia/commons/f/fb/NILT_UNINATID_PROTESTANTS.png) and zero Catholics embracing the label of “Unionist” (http://upload.wikimedia.org/wikipedia/commons/5/54/NILT_UNINATID_CATHOLICS.png).

    This implies that the labels of “Unionist” and “Nationalist” are actually more accurately religious labels of “non-Catholic” and “non-Protestant” respectively, and have much more to do with religion than opinions on the border. Some interesting points also on how these labels of “Unionist” and “Nationalist” would be more honestly described as merely religious euphemisms that remain as relics of the “peace process” to keep the snouts of our current political parties in the trough. I think this has some interesting implications for our political system – why is its core based on such meaningless and dishonest religious terminology, and why does our media continue with the usage of such sloppy terminology? Would a 50:50 policy based on religion similar to the PSNI have been a more honest model of our assembly?

    So fascinating that I think this would deserve a Slugger thread of its own.

  • The local media have been plugging this poll like there was no tomorrow and yet a salient fact kept being ignored in the coverage, that of the sick state of the economy in the State.
    I wonder why they would give that distortion so little comment?

  • AntrimObserver

    IJP wrote,

    “I really don’t think the “Shy SF voter” line stands up to scrutiny these days. Our aforementioned exit poll found people quite content to declare they had voted SF – more declared for SF than SDLP overall (ie the tally was roughly what you would expect).

    It seems to me to be a sampling error.”

    ————————————————————————-

    Totally wrong, IJP.

    Your exit poll was conducted on the street, yes? So, there’s nothing to link an individual who states they voted for SF with their actual address, yes?

    NILT surveys, on the other hand, are conducted in the respondent’s OWN HOME.

    So, whilst names are not recorded, the respondent knows that the stranger (the interviewer) will be aware that SF supporters live at the property. Thus, they opt for the ‘safe’ rsponse e.g. SDLP support.

    There’s a world of difference between stopping a stranger in the street and asking who he voted for (the exit poll) and going to someone’s house (NILT) and asking for their political allegiance.

    NILT’s sampling is fine. It’s the research method that is flawed.

  • between the bridges

    NILT: Does it tell us anything beyond it may still have sampling issues?

    Yes it tells us that socialist republican’s sh it themselves at the thought that not all CNR may want a glorious 32 county Marxist banana republic. It also confirms what I have always believed and that is that the CNR community has more in common with the PUL community than they have with the 26 county splitters

  • Barnshee

    I say chaps lets have a “border poll” and settle the matter for a while at least

  • becky980

    This thread has drawn me from lurker to poster, in case my (fairly rudimentary) knowledge of surveys and sampling can shed any further light.

    Mick would like to know whats wrong with the sample,Ive had 5 goes at trying to get this up, i can only do this at work, so here goes, apologies for spelling and grammer.
    The key word for all polls is random random random
    The telephone with its anonimity(important everywhere), ease of contact and most importantly ability to generate random numders, makes it a brilliant tool for market research, so why have Queens gone to some property register to create a sample? as opposed to the tried and tested method.
    I,ve personally never heard of such a thing and would take some convincing that it is some how a good thing.

    There are significant shortcomings with random digit dialling (RDD), particularly as people move away from landlines to relying solely on mobile phones. While you can achieve a decent sample with landlines by isolating numbers beginning with particular area codes in the geographical area of interest, this isn’t as possible with mobile telephones. You can thus only target the population who have landline telephones, and you’re likely to miss younger people in particular if you rely on this.

    The property register provides a better sampling frame (from which to draw the sample) because it provides as full a list as possible of all places where the target population – that is, everyone over 18 living in Northern Ireland – is dwelling. From a sampling frame as extensive as this, you are more likely to be able to generate an appropriate random sample than you would by RDD, not least because you have a better idea where your entire population is and how best to reach them. Most major government surveys in the UK use property registers, postcodes, etc, and multi-stage sampling to ensure it’s as robust as possible.

    Face-to-face interviewing has a number of advantages, not least in ensuring that the respondent understands the question, which can be more difficult in other forms of survey administration. It also means the interviewer can use aids (holding up flags, words, signs) for clarity, as the NILT does.

    Where issues might arise is in ‘interviewer effects’, as others have said, where the respondent feels they should give what they perceive to be more socially acceptable answers in the presence of another person. Many face-to-face interviews include a self-completion questionnaire for more sensitive questions, and I think the NILT survey does for some aspects, but not, if I remember, community relations and political affiliation. In Northern Ireland interviewer-pleasing can be a particular issue because very basic information about the interviewer – such as their surname – might cause the interviewee to place them on one side or the other, and thus to tailor their responses according to what they consider most appropriate in those circumstances.

    PS nothing wrong with sample size, if anything its overkill, you can get good data from 200, if your sample is right.

    A sample of 200 of a population of 1.8 million will generally give you around a 7% margin of error (e.g. the ‘true’ value could be plus or minus 7 the given percentage); by comparison, a sample of 1200 gives you less than 3% margin of error (both at 95% confidence level), so it’s much more robust. Twelve hundred, as I understand it, is a decent sample. Once you get beyond that kind of number, you need substantially more respondents to reduce the margin of error by much (e.g. another 1000 to reduce the error to 2%), and the associated cost means that’s not really worthwhile.

  • RyanAdams

    Heres reality:

    22.5% of the population voted for nationalists in the May Assembly election.

    The rest voted unionist, other, and of course the real winner of the election was apathy.

  • JR

    Ryan,
    “The rest voted unionist” – no they didn’t
    24.7% of the electorate voted for unionist parties, the rest didn’t vote