22
This article was downloaded by: [The University of Manchester Library] On: 05 December 2014, At: 13:22 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Contemporary Record Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/fcbh19 The polls have a lot to answer for! Robert M. Worcester a b c a Chairman of MORI b Visiting Professor of Government , London School of Economics and Political Science c Visiting Professor of Journalism , City University , London Published online: 25 Jun 2008. To cite this article: Robert M. Worcester (1992) The polls have a lot to answer for!, Contemporary Record, 6:2, 356-375 To link to this article: http://dx.doi.org/10.1080/13619469208581217 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities

The polls have a lot to answer for!

  • Upload
    lydat

  • View
    309

  • Download
    13

Embed Size (px)

Citation preview

Page 1: The polls have a lot to answer for!

This article was downloaded by: [The University of Manchester Library]On: 05 December 2014, At: 13:22Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number:1072954 Registered office: Mortimer House, 37-41 Mortimer Street,London W1T 3JH, UK

Contemporary RecordPublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/fcbh19

The polls have a lot toanswer for!Robert M. Worcester a b ca Chairman of MORIb Visiting Professor of Government , LondonSchool of Economics and Political Sciencec Visiting Professor of Journalism , CityUniversity , LondonPublished online: 25 Jun 2008.

To cite this article: Robert M. Worcester (1992) The polls have a lot to answerfor!, Contemporary Record, 6:2, 356-375

To link to this article: http://dx.doi.org/10.1080/13619469208581217

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of allthe information (the “Content”) contained in the publications on ourplatform. However, Taylor & Francis, our agents, and our licensorsmake no representations or warranties whatsoever as to the accuracy,completeness, or suitability for any purpose of the Content. Anyopinions and views expressed in this publication are the opinions andviews of the authors, and are not the views of or endorsed by Taylor& Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information.Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities

Page 2: The polls have a lot to answer for!

whatsoever or howsoever caused arising directly or indirectly inconnection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private studypurposes. Any substantial or systematic reproduction, redistribution,reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of accessand use can be found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 3: The polls have a lot to answer for!

The Polls Have a Lot to Answer For!

ROBERT M. WORCESTER

The polls in the 1992 British General Election have a lot to answer for.For one thing, the most exciting election night in two decades (remem-

ber all those critics who complained that the eve-of-poll polls and the exitpolls took all of the fun out of election night?).

For another, Red Wednesday, the hubris of the Labour Party's rally inSheffield, or, as Norman Tebbit has called it, 'April Fool'. For yetanother, suggesting to the Labour leadership that they were likely to formthe next government. And for putting steel in the backbone of theConservatives' foot soldiers. And tough words in the mouth of the PrimeMinister about Scottish devolution, about electoral reform ('I'll quitrather than accept PR') and about the threat of a vote for the LiberalDemocrats 'putting Labour in'.

Yes, the polls have a lot to answer for at the 1992 General Election,which is now history.

This article examines the record of the polls in the 1992 GeneralElection. It is meant to be complementary to the Enquiry being under-taken by the Market Research Society (a copy of an earlier version wassent to the Committee for their consideration) and to the reviews beingundertaken by the other members of the Association of ProfessionalOpinion Polling Organisations (APOPO) and has been sent to them aswell.

Forgotten now are the headlines on election day: 'Late Surge by Toriescloses gap on Labour in Final Hours of Campaign' was the Times' banner;'Tory Hopes Rise after Late Surge' was the headline over the 'splash' inThe Guardian. In The Daily Telegraph' it was 'Tories Narrow Gap', andin the Financial Times the banner read 'Opinion Polls Indicate Last-Minute Swing from Labour to Tories' while the Daily Express trumpeted'Tory Surge; Polls Show Late Boost for Major'. Today's Political EditorChris Buckland's copy began

Robert M. Worcester is Chairman of MORI and Visiting Professor of Government at theLondon School of Economics and Political Science and also Visiting Professor of Journalismat the City University, London. This article is an edited version of a paper delivered at theLondon School of Economics on 28 April 1992.

Contemporary Record , Vol. 6, No. 2, Autumn 1992, pp. 356-376PUBLISHED BY FRANK CASS, LONDON

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 4: The polls have a lot to answer for!

THE POLLS HAVE A LOT TO ANSWER FOR! 357

Voters were streaming back to the Tory Party last night as theyprepared to cast their verdict in today's real poll of polls - theGeneral Election. A final clutch of opinion surveys showed supportreturning to Mr Major with Conservative waverers catching on tohis vital message that a vote for the Liberal Democrats could putNeil Kinnock into Downing Street.

The City took heart, and both the stock market and sterling shot up.Even the Wall Street Journal pointed out the late surge to the Conserva-tives on the morning of the election. Yet even so, the next day the media'spundits and psephologists and letter writers to the newspapers alike tookup cudgels to attack the pollsters for getting it wrong.

The Polls Got It Wrong!

The pollsters have to regard all of the criticism of their performance asperfectly reasonable, as the polls got it wrong in 1992, more wrong than inany election since their invention over 50 years ago. As I was widelyquoted as saying on the day after the election, it was 'the worst results forthe opinion polls since they were invented.'1 On the day after the electionI said as much on no less than 18 radio and television interviews and inanother half dozen interviews with newspaper and wire service jour-nalists.

What is essential is to have soundly based critical reviews of metho-dology and interpretation in this of all elections. Peer review and criticalcomment are necessary in a democratic society which must trust theintegrity and professionalism of the people who conduct and comment onopinion poll findings. As this election above all has shown, pollsters bearenormous responsibility to do the very best that they can, to strive foraccuracy and equity, to be open about their methods, cooperative to theircritics, to meet comment with courtesy, and, it must be said, suffer fools ifnot gladly, then with patience and good humour.

This must especially be so now, for as never before have the opinionpolls been so wrong in their prediction of the outcome. Over the 13elections between 1945 and 1987, the average error of the final pollspublished on the eve of the election has been 1.4 per cent on the share ofthe parties; in 1992 it was 2.8 per cent. Over 60 per cent of the polls inprevious elections had predicted the Labour and Conservative shares ofthe vote to within two per cent, never mind the usually quoted '3 per centmargin of error'.

In 1992 however, the five major pollsters, Gallup, Harris, ICM, MORI

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 5: The polls have a lot to answer for!

358 CONTEMPORARY RECORD

and NOP averaged a 1.3 per cent Labour lead, the final result was a 7.6per cent lead for the Conservatives. The polls backed the wrong winner,by a wider margin than can be accounted for by sampling error, and werewidely criticised for their inability to perform to expectation.

But it must also be put into perspective. The magnitude of the error onthe lead (the 'gap') was eight per cent. The movement of four voters in ahundred from the Conservative Party to Labour would have put thepollsters 'spot on'. So it is the cause of failing to get a swing of four percent (not eight per cent as one academic psephologist said on thetelevision), that needs to be examined. As one political scientist put it'The failure of the polls to predict the outcome of this election says moreabout the British than it does about the polls'.

Another bit of perspective is provided by the remarkable fact that theConservative's 21-seat majority in the House of Commons could havebeen overturned if the 11 most marginal seats had gone to the party insecond place instead of to the Tories. The most marginal seat was theVale of Glamorgan with a 19 vote plurality; the 11th most marginal seatwas Southampton Test, with 585. If in these 11 seats just 1,241 voters outof the 32,834,188 who voted on 9 April had voted differently, or if 2,499Conservative voters had stayed home on election day, then the hungparliament anticipated by the pollsters would have been the election'sresult.

The Critics

Just a day or two after the election letter writers and critics alike weredismissing any 'late swing lame excuses' from the pollsters.

There have been a number of punditry, psephologists' commentaryand readers' letters in The Times, The Daily Telegraph, The Independent,The Guardian, The Financial Times and elsewhere following the election,expressing various opinions as to why the opinion polls 'got it wrong' on 9April.

Some blamed sample size: one 'director of a company which suppliesmarket research data to the pharmaceutical industry in the Republic ofIreland' pointed out that 'given an electorate of 28 million (sic - it was42.58 million according to the OPCS), the final four opinion pollssampled at best 0.0098 per cent and at worst 0.0062 per cent'. The authorof that particular letter goes on at some length to reveal his remarkablelack of understanding of basic facts of sampling and statistics (given thenature of his employment), but he does not seem to recall that the samesampling methods and sample sizes have in the past three elections been

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 6: The polls have a lot to answer for!

THE POLLS HAVE A LOT TO ANSWER FOR! 359

commendably accurate. Another market research 'expert' rejected thesudden swing to the Tories on election day, saying that 'Each poll ofhardly more than 1,000 meant fewer than two per constituency'. He maybe able to divide, but a statistical expert he is not.

One of the more inventive and apparently well-reasoned letter writersargued that 'any serious [sic] piece of research which relies on a sample ofthe population is conducted over a period of weeks, not one day'. 'Aweek is a long time in polities', I seem to recall a former President of theRoyal Statistical Society saying some years ago. Nor does this chap, longknown for his attacks on the polls and no doubt relishing the opportunityto put the boot in, explain how the polls record in the past is so good iftheir methods so flawed.

And a fellow letter writer speaks up confidently to report that there was'no eleventh-hour swing to the Conservatives', and reports himself to bein a state of profound depression and dismay; no wonder he is depressed,if he is convinced that the British (against all of my experience to thecontrary) are a nation of liars.

Other letter writers also suggest millions of people lying to inter-viewers, no doubt in the case of the three panel studies that wereundertaken and which had comparable results both before and after theelection repeatedly and consistently lying. Possible, but hardly plausible.As one of my colleagues put it in a letter to the Sunday Times Scotland, ifthe British had been massively lying surely they would have been morecreative and we would have the Monster Raving Looney Party at a 50 percent share!

Another claims government by fraud, reporting that one of the firstActs of the 1987 Conservative government was to give the postal vote totwo million expatriate Britons, but ignores the fact that only 34,454 ofthese expats registered to vote. At an average of 53 voters per constitu-ency, they could have made some difference, but not that much!

Others suggest that massive avoidance of the poll tax was the reason forLabour's defeat, yet the Electoral Reform Society reported that in some300 local authorities surveyed, some 230,000 people were involved. Aswe screen out those who say they are not registered or do not know if theyare or not, and this represents some two million people, it can hardly bethe major source of the error.

One of the more simplistic letter writers explains the unreliability of thepublic opinion surveys (and of the media comment on them) is down toignoring the 'don't knows'. He cannot have been paying too closeattention to the many reports on the 'don't knows', the 'floating voters'and the 'undecideds'. This plaintive 'where were the don't knows?' was aquestion put to me by one eminent, if somewhat self-important, televi-

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 7: The polls have a lot to answer for!

360 CONTEMPORARY RECORD

sion interviewer who obviously had not listened to my earlier remarksabout The Times1 Robin Oakley pieces on the 'don't knows' and the 'soft'party supporters. Certainly he must have missed the front page of TheTimes on the day before the election when Oakely's article, the 'splash'that day was headlined 'Parties Make Final Push to Capture FloatingVoters'; 'However, 24 per cent of Liberal Democrat supporters said theymight yet change their vote. So did a fifth of Conservative supporters and16 per cent of Labour backers.2

'Don't wave that paper [The Times] at me', he shouted. Nor did he dohis homework sufficiently well to read Ivan Fallon's excellent articleseach week in The Sunday Times on its MORI panel, or Peter Kellner'sreport of the NOP panel in the Independent on Sunday, each charting theshifts into and out of the 'don't knows', or even watch the BBC's On theRecordMORl panel of 'floating voters' which included both the 'don'tknows' and 'soft' supporters of all parties, or watch the BBC 'Vote Race'which tracked strong party supporters and contrasted their reactions tothose of the 'floaters' to party election broadcasts and politicians'speeches and interviews.

What it seems to me pollsters should not have to put up with are someof the extreme (and personal) remarks from the self-serving analyses ofthe one or two academic psephologists who give their trade a bad name,and who make a tidy living on the side writing articles in the broadsheetsand appearing on television, second guessing pollsters, demanding in-stant and detailed access to data collected that day to pick at it to findwhat in the heat of battle was overlooked so it can be criticised. My adviceto them is to be cool, be correct, dispassionate, have no hidden agenda,political or personal, and if you cannot say anything nice about in-dividuals, do not say anything at all.

Much space in the papers has been given over since the general electionto 'Pollsters Admit Election Mistakes' (30 April) and 'The InvisibleVoters Who Fooled the Pollsters' (1 May) in The Independent as well as tothe various letter writers' expressions of their own prejudices. All goodfun. But is it not time for a little more light and much less heat on thesubject, what Tony Blair has called for, 'considered reflection, not instantwisdom'?

One such analyst, Peter Kellner, one of the best poll-pickers in thebusiness, contended on BBC Newsnight that 'I don't think Labour wasever in the lead throughout the election campaign'.

On Red Wednesday, the night of Labour's Sheffield rally, three pollshad leads of from four per cent to seven per cent for the Labour Party.Kellner contends that there is at most a two per cent error in the eve-of-poll polls not accounted for by late swing, which accounted for about half

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 8: The polls have a lot to answer for!

THE POLLS HAVE A LOT TO ANSWER FOR! 361

of the four per cent error, now agreed by nearly everyone, including theMarket Research Society enquiry team.

He partly bases this contention, denied by pollsters and the generalsecretary of the Labour Party alike, on exit poll evidence of mass non-registration of council tenants and contends that 'what really needs to beaddressed is the way the electoral registers are compiled'. In fact this isbeing done by the Office of Population Censuses & Surveys to update theuseful Todd-Butcher paper done after the 1981 census, and is eagerlyawaited.

His other allegation is that there is a significantly lower turnout ofworking class electors, sufficient to cause a 12 per cent swing in the classprofile of the electorate, and that not 42 per cent, as in the population aredefined as 'middle-class', but 54 per cent of voters. I also find this hard toswallow. The basis for this contention is his belief that the 1992 polls werebased on the 1981 census, but not all of them were, and there was nodiscrepancy between the ones that were and the ones that were not.Further, he ignores the evidence of trade union membership (checkedagainst TUC membership returns) and newspaper readership (checkedagainst ABC sales figures), both of which correlate highly with classdesignation.

The leader in The Independent that same day (30 April) raised otherpoints, alleging that 'Random polls have proved more accurate' (thanquota polls). This is in fact an error. Election polls conducted by randomsampling during the past eight national elections, including the 1975 EECReferendum) were less, not more, accurate, than those conducted byquota, due to poor compilation of registers and, yes, late swing after thebulk of the interviewing was done.

David Butler, also writing in the 30 April Independent raised aninteresting question. He said 'psephologists should come in for at least asmuch blame as the pollsters'. Instead, they are baying for the pollsters'blood. The exit poll results would have forecast around a 16-seat majorityfor the Tories on a national swing if their 10 p.m. results had been usedand their earlier reports not been subject to such statistical jiggerypokerery by the psephologists.

The Record of the Polls

Before looking further at the 1992 election polling, it is worth remember-ing the standard set by our performance in the previous three generalelections. The table below sets the eve of election, predictive, pollsconducted by MORI, against the actual outcome of the elections.

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 9: The polls have a lot to answer for!

362 CONTEMPORARY RECORD

TABLE 1

Cons

Lib

LD

Olh

Lead

1979

Remit

%

45

38

14

3

7

MORI

%

45

37

15

3

8

Dlf

%

0

+1

•1

0

-1

1983

RnuK

%

44

28

26

2

16

MORI

X

44

28

26

2

16

Dlf

%

0

0

0

0

0

1987

Result

%

43

32

23

2

11

MORI

%

44

32

22

2

12

Dll

%

•1

0

+1

0

+1

The Campaign Polls

There were actually fewer mainstream polls published during the 1992election than in 1987. Then there were 54 national, marginals and panelsurveys, this time 50. In all, there were 74,490 interviews taken in thesenational polls, slightly fewer than the 83,083 interviewed in 1987.

In addition, there were another 30 or so other national, regional andconstituency polls conducted by the major polling organisations andother market research companies and others which involved interviewswith another 30,000 electors, making a total of approximately 100,000people interviewed at some point in the 1992 campaign. With an elec-torate of 42.58 million, that is some two people in a thousand that willhave been interviewed. It is no wonder that we still hear the old line 'I'venever been polled and I don't know anybody who has been either'.

What were these polls showing? A pretty steady picture, steady acrosscompany, and, generally, over time. Table 2 shows the detailed pollresults for all of the national polls published by the major pollingorganisations during the campaign.

MORI Panel Studies

But one thing that these snap-shot polls do not show is the movement ofthe electorate between parties, which largely cancelled itself out, but

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 10: The polls have a lot to answer for!

THE POLLS HAVE A LOT TO ANSWER FOR! 363

which reveals the remarkable changeability of the electorate's political al-legiance. The panel surveys provided this information.

The most instructive analyses should come from the two panel studieswhich were recalled after the election. Both happened to be conducted byMORI, one for The Sunday Times, the other for BBC's On the Record.Both were conducted immediately after the election. Both were tele-phone recalls on people interviewed throughout the campaign. In the onedone for The Sunday Times, the interviewing was done by FDS, anindependent market research firm who were sub-contracted to undertakethe recalls, the MORI telephone survey subsidiary, On-Line TelephoneSurveys, being fully stretched with recalls on the panel for the BBC. FDS,on MORI's behalf, contacted 934 panellists on Friday 10 April, between10 a.m. and 9 p.m. This represented a 60 per cent recall of the originalpanel, not bad for a one-day recall, especially when you bear in mind thatsome of the original 1,544 panel were unavailable because they were noton the telephone. The data were of course weighted to both the demo-graphic and political profile of the original panel.

Over one thousand (1,090), of the On the Record panellists wereinterviewed by On-Line, who had the advantage of being able to inter-view not only on Friday, but also on Saturday up until 3 p.m., because ofthe later deadline of the BBC.

The MORllSunday Times panel recall found that only 63 per cent saidthey had made up their minds before the election had been called, downnearly 20 per cent from the more usual 80 per cent that we have measuredin previous elections. And, as noted below, eight per cent said they hadonly made their mind up in the last 24 hours, and 21 per cent during thelast week of the campaign.

As Ivan Fallon reported week after week, the amount of movement inthe electorate - people who switched from one party to another('switchers') or in and out of 'don't know' ('churners') - was higher thanever before, and as reported in the final article, some 11.1 million electorschanged their minds during the campaign out of the 42.58 million in theelectorate. The week before polling day, the panellists indicated theirvoting intention at that time as a Labour lead of one; at week later thecountry voted in the Conservatives by a 7.6 per cent margin, a swing of alittle over four per cent.

The MORI/BBC panel was of a different design, a compromisebetween the BBC client team, their advisor, John Curtice, and MORI. Itconsisted of 'floating voters', and comprised those who after expressing avoting intention said they might change their minds between the timeinterviewed and polling day, plus the full category usually described asthe 'don't knows', which is the ten per cent who said they would not vote,

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 11: The polls have a lot to answer for!

364 CONTEMPORARY RECORD

FieldworkDates

11.311-12.311-12.311-13.311-13.312-13.312-13.3

13.315-16.3

16.317.3

17-18.317-18.318-20.319-20.3

20.319-21.320-21.321-23.322-23.3

23.323-24.3

24.324.3

24-25.325-27.326-27.3

27.327-28.326-29.328-30.329-30.3

30.331.3

31.3-1.431.3-1.431.3-1.4

1-3.42-3.42-3.4

3.43-4.43-4.44-6.44-7.46-7.47-8.47-8.47-8.4

8.4

TABLE 2

Sampling

si,050(54)si,054(53)si,054(100)sl,086(98)s2,186(100)si,544(65)s2,155(133)51,059(54)si,081(98)si,099(54)si,100(54)s984(100)si,262(82)si,257(63)si,096(100)Sl,115(54)si,004(133)sl,085(54)sl,000(99)s2,158(100)sl,109(55)sl,092(100)sl,096(54)sl,105(55)Sl,326(84)El,292(65)S],057(99)sl,136(54)sl,099(54)si,000(133)sl,108(100)•2,152(100)si,080(54)si,126(54)si,095(100)si.302(83)S10,460(330)sl,265(65)sl,043(95)sl,006(133)Si,139(54)Sl,104(54)sl,090(100)sl,093(100)s2,210(100)sl,065(53)sl,731(164)Sl,746(85)s2,748(198)si,106(103)

Company/Publication

NOP(Mail on Sunday)MORI(Times)HARRIS(Observer)HARRIS(0.Express)HARRIS(LWT)MORI(S.Times)(PI)NOP(Indy on Sunday)ICM(S.Express)HARRIS(D.Express)MORI(Times)ICM(Guardian)GALLUP(D.Telegraph)HOP(independent)MORI(S.Times)(P2-T)HARRIS(Observer)ICM(S.Express)NOPJIndy on Sunday)(P-T)IlOP(Mail on Sunday)IIARRIS(D.Express)HARRIS(ITH)MORI(Times)GALLUP(D.Telegraph)ICM(Guardian)HMR(European)HOP(Independent)MORI(S.Times)(P3)HARRIS(Observer)ICM(S.Express)NOP(Mail on Sunday)NOP(Indy on Sunday)(P3-T)HARRIS(D.Express)HARRIS(ITH)MORI(Times)ICH(Guardian)GALLUP(D.Telegraph)HOP(Independent)ICM(Press Association)MORI(S.Times)(P4)GALLUP(S.Telegraph)NOP(Indy on Sunday)ICM(S.Express)NOP(Mail on Sunday)HARRIS(Observer)HARRIS(D.Express)HARRIS(ITN)MORI(YTV)MORI(Times)HOP(Indcpendent)GALLUP(D.Telegraph)ICM(Guardian)

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 12: The polls have a lot to answer for!

THE POLLS HAVE A LOT TO ANSWER FOR!

TABLE 2 (cont.)

365

Con Lab L/Dem

4138403937404039413838

40.53838403739384338384019383938403637394035353738373637

37.53B3735383738373830

38.538

4041434041394140384343

38.54241394241403(14241

4 0.1403942403838414039414241

37.5393939

37.54139414018404039423838

151612161716141617161618171917161516151617

16.5171914201720181617191918

20.5192021221718201721182070172020

er

455555554333324556444344525645454445533464544332

3.54

C-L

-1-3-3-1-41-1-13-5-52-4-31

-5-2-25-4-3

-0.5-1-1-3-22-2-4-11

-6-7-40.5-2-3-20

-3-2-6-2-1-2-3-1-3

0.50

C/USwing

6776

7.5566488

4.57.5

758

6.56.5

37.5

75.75

667

6.54.56.57.5

65

8.59

7.55.256.57

6.55.57

6.58.56.56

6.5767

5.255.5

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 13: The polls have a lot to answer for!

366 CONTEMPORARY RECORD

the four per cent who were 'undecided' and the fewer than one per centwho refused to say how they intended to vote.

In all, the 'floaters' represented some 38 per cent of the base-linesurvey from which the panellists were recruited, so some 62 per cent saidthey were sure of their party loyalty, within a point of the stated firmnessof party commitment in the base-line panel survey for The Sunday Times.

These figures suggest that 1992 saw an electorate who made up theirminds later, and shifted their ground in greater numbers, than we havemeasured before. It is interesting to note just who the 'switchers' were.Examining the voting intentions of the panellists interviewed in weekfour, the week before election day, and reinterviewing those same peoplethe day following the election, we uncovered several very interesting pat-terns:

- Among trade union members there was a 2.5 per cent swing to Labour,against the trend to the Conservatives, and possibly in reaction to themillions of pounds spent by the trade unions on advertising in the finalfew days of the campaign; however, this was countered by . . .

- A bigger, four per cent swing among the eight in ten of the electoratewho are not trade union members; possibly in a backlash against thesame advertising?

- A small, one per cent, swing against the Conservatives among peopleemployed in the private sector, while . . .

- Among people employed in the public sector, civil servants, healthworkers, teachers, policemen and the like, there was a four per centswing between the Liberal Democrats to the Conservatives. The weekbefore, 24 per cent of public sector workers said they intended to voteLib Dem, 42 per cent for the Tories; the day after the election, 20 percent said they had finally voted Lib Dem while 42 per cent said theyvoted Conservative.

- Among women there was a 4.5 per cent swing to the Conservatives,among the over-55s the same, and among council tenants, an 8.5 percent swing to the Tories!

- In the South of England there was a four per cent swing favouring theConservatives.

- Finally, among regular readers of the Daily Express, the swing fromthe week before to the day after the election was 9.5 per cent1.

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 14: The polls have a lot to answer for!

THE POLLS HAVE A LOT TO ANSWER FOR! 367

The Nature of British Elections

There is a tradition among political scientists in Great Britain topatiently explain to visiting American academics, psephologists, poll-sters, politicians and political journalists that the main differencebetween British and American elections is that in British electionspolicies, not people are the deciding factor. Indeed, MORI researchover the past decade has shown a fairly steady 50 per cent/30 percent/20 per cent relationship between voting decision-making mainlyconditioned by policies/leader image/party image.

Yet during this past election one striking finding stands out:throughout the election, from the beginning to the end, John Majorled Neil Kinnock by from nine per cent to 13 per cent as the 'mostcapable' prime ministerial candidate. With hindsight, I believe thatthis question and its replies consistently gave a guide as to theoutcome; we were just not smart enough to see it clearly enough,although The Times'1 Robin Oakley focused an entire article on thisearly in the campaign.3

Another clue was that when asked 'Which two or three issues will bemost important to you in helping you to make up your mind on how tovote?', the NHS, Education, Unemployment - all Labour issues - ledthe field. Yet the one time that a different question was asked, 'Howmuch will you be influenced, if at all, in the way you vote at theGeneral Election by the policies of the political parties on taxation, nofewer than 39 per cent of the electorate said their vote would beinfluenced 'a great deal'. No lies, no prevarication from this four in tenof the electorate. They told us plain: a great deal, they said. Did wehear clearly enough?

Another key finding from the post-election recall survey for theBBC was when voters were asked 'Thinking about the way you voted,which was stronger, your liking the party you voted for or you dislike ofthe other parties?', a majority, 55 per cent, said it was antipathy andonly 37 per cent liking the party they voted for, among the 'floating"voters"' in the panel. Panellists also voted by more than three to one,56 per cent to 18 per cent, that the Conservative Party rather thanLabour could best handle the economy generally. And by four to one,48 per cent to 12 per cent, they regarded John Major, rather than NeilKinnock, as the most capable Prime Minister.

During the Autumn 1991 election boom, ITN had rung us to find outwhat the EOI (Economic Optimism Index) was for September. 'Why',a colleague asked them. 'Because Central Office says they'll call theelection if it goes over 15, as they did in '83 and '87', was the reply. We

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 15: The polls have a lot to answer for!

368 CONTEMPORARY RECORD

checked, and they had! But it was little noticed, except possibly byCentral Office and in an article by Ivor Crewe in The Times* that in theMarch sounding, a sharp reversal had taken place in the nationalmood, and a —2 per cent EOI recorded in February had been replacedby a +15 per cent, with 36 per cent expressing optimism and only 21per cent gloom.

Another clue came late in the campaign, but replicated findings of ayear and more earlier. When asked if they thought a hung Parliamentwould be good or bad for Britain, a majority, 56 per cent, said 'bad',and they included an astonishing 44 per cent of Liberal Democrats. Itwas certainly this group who were worried by the Tories' attack on theLiberal Democrats and the prospect of a hung parliament in which MrAshdown held the balance of power.

Certainly Captains of Industry were in no doubt. Our work for theFinancial Times at the beginning of the campaign and certainly at theend proved that without a doubt. In the first wave of the 'Captains'panel, six in ten of the main board directors of the nation's largestcompanies said that a hung parliament would be bad for their business;by 6-7 April when just such an outcome seemed the most likely, three-quarters said it would be bad.

Sample Considerations

One thing that should now be manifestly apparent to the critics whoclaim that this election proves the inadequacy of the sample sizes of athousand or two, is that sample size is not the key to quality of findings.The first bit of evidence, courtesy of the Press Association, is that theICM poll taken in the third week of the campaign, among a sample of10,460 electors in 330 sampling points and taken over nearly a weekincluding a weekend (31 March-1 April) was almost identical in itsvoting intention to the findings of a 'smaller' poll conducted over thesame period. Interviewing on 31 March-1 April, ICM recorded 36 percent for the Conservatives, 39 per cent for Labour and 20 per cent forthe Liberal Democrats. Also on 31 March-1 April, NOP interviewed1,302 electors in 83 sampling points and found 37 per cent/39 per centand 19 per cent.

A message had been evident long before, but ignored by the samplesize critics. Between 28 September 1991 and 3 October 1991 ICM hadpolled 10,159 electors for PA and found 40 per cent/42 per cent/13 percent; the poll taken before, by ICM as well, but for the Daily Mail,recorded 40 per cent/42 per cent/14 per cent, nearly identical - and thepoll following, by Gallup for the Daily Telegraph 39.5 percent/41.5 percent/15 per cent, again almost identical findings.

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 16: The polls have a lot to answer for!

THE POLLS HAVE A LOT TO ANSWER FOR! 369

But what about the actual constituencies sampled? For each surveythat MORI conducted, we looked back at the 1987 General Electionoutcome of 43 per cent/32 per cent/23 per cent/2 per cent and averagedthe percentage outcome across all seats sampled to see whether ourpolitical 'profile' represented the national picture at that time. For theSunday Times Panel, the profile was Conservative 43 per cent Labour32 per cent, Alliance 23 per cent and Others two per cent. No bias wasfound.

Given the uneven swings that were found in the individual con-stituencies as the results came in on election night, we were concernedto see whether, by some freak chance, we had a sample of seats thathad behaved atypically in 1992. Again, we averaged the outcome inour sample seats and compared our figures with the national outcomeon 9 April. The panel's constituencies had 42 per cent Conservative, 36per cent Labour and 17 per cent Liberal Democrat (compared with thenational picture of 43 per cent/35 per cent/18 per cent). Within a pointfor each party.

The fact that the final polls of the campaign, all taken the day or twobefore polling day, were so consistent, all five had the Tories at 38 percent + / - 1 per cent, Labour at 40 per cent + / - 2 per cent and theLiberal Democrats at 18 per cent + / - 2 per cent, suggests that neithersample size nor sampling 'error' were responsible for the magnitude ofthe difference between the final polls and the actual result.Many other hypotheses have been put forward including mass

economy with the truth, late swing, and media effect. In truth, probablyall played a part.

Polls are snapshots at a point in time, and that point is when thefieldwork was done, not when the results were published. One earlyindicator of the late swing was that MORI's final poll for The Times was inthree independent matched samples, and the Tuesday night sample wasconsiderably more Labour inclined than the interviews done on theWednesday; another was that the Harris exit poll for ITN found asteadilydeclining level of support for Labour - the 3:40 p.m. cut-off figures hadLabour at 39 per cent, the 5:40 p.m. figures were 38 per cent; the 8 p.m.results, the ones that were broadcast - reported Labour at 37 per cent,and their 10 p.m. 'finals' showed Labour at 36 per cent, close to what theyfinally got; another was that the recall interviews done on the dayfollowing the election with two independent samples of electors inter-viewed the week before to see if they had switched, how they hadswitched, and why those that had switched had done so were consistentand accounted for some three of the four per cent error; perhaps the othertwo per cent were lying. Personally, I doubt it.

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 17: The polls have a lot to answer for!

370 CONTEMPORARY RECORD

TABLE 3

EVE-OF-POLL POLLS

Company

Sample

Fleldwork

CON

LAB

LIBDEM

OTHER

LEAD

ERR/LEAD

ERR/SHARE

Harris

2,210

4-7

%

38

40

18

4

•2

9.6

2.55

MORI

1,731

7-8

%

38

39

20

3

-1

8.6

3.0

NOP

1,746

7-8

%

39

42

17

2

-3

10.6

3.5

Gallup

2,748

7-8

%

38.5

38

20

3.5

0.5

7.1

2.25

ICM

2,186

8

%

38

38

20

4

0

7.6

2.4

GE

9

%

42.8

35.2

18.3

3.7

7.6

8.7

Ave

2.74

Ave

The Exit Polls

So much for the eve-of-poll polls failure to read the crystal ball correctly.Surely the exit polls should have been able to read the past, even if pollstaken a day or two before the election failed to foretell the future?

The three exit polls, by ICM for Sky /Today/The Sun, Harris for ITNand NOP for the BBC, when massaged by the psephologists, suggested intheir findings between 298 and 305 seats for the Conservatives, between294 and 307 for Labour, and between 18 and 25 for the Liberal Demo-crats. Pretty consistent, and pretty wrong. Yet their poll results told avery different story. See table below.

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 18: The polls have a lot to answer for!

THE POLLS HAVE A LOT TO ANSWER FOR! 371

TABLE 4

EXIT POLLS - VOTING INTENTION

Con

Lab

LD

Oth

Lead

Error

GE

%

43

35

18

4

8

ICM

%

38

41

18

3

-3

Dlf

%

•5

+6

0

-1

5.5

Harris

%

41

37

18

4

4

Dlf

%

•2

+2

0

0

1

NOP

%

40

36

18

5

4

Dlf

%

•3

+1

0

+1

2

Mar

%

45

37

16

2

8

NOP

%

43

39

16

2

4

Dl

f

%

-2

+

2

0

0

2

GE = General Election actual result (Con 42.8%/Lab 35.2%/Liberal Democrat 18.3%/Others 3.7%. Sample sizes: ICM = c.25,300; Harris National = 4,701 final (41%/36%/18%/4%); Harris 16,164; NOP National = 4,719 reported, final figures based on 5,769;NOP Marginals = 18,747 reported, final figures based on 20,256 (42.6%/38.7%/16.2%/2.1%). Error calculated only on two-party share. NB. Neither NOP national exit pollfigures nor were marginal share vote figures broadcast by the BBC on the night.

The public on the night of the election were clearly misled by the seatprojections done by the broadcasters and their psephologists, not the exitpolls.

The Harris Exit Poll for ITN actually measured 41 per cent for theTories, 36 per cent for Labour, 18 per cent for the Liberal Democrats andfour per cent for other parties, just two per cent out on the share for theConservatives, one for Labour and spot on for the Liberal Democrats andothers, an average error per party's share of less than one per cent. Notbad; and on a uniform swing, Harris's 41 per cent/37 per cent call based on

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 19: The polls have a lot to answer for!

372 CONTEMPORARY RECORD

TABLE 5

EXIT POLLS - SEAT PROJECTIONS - 10 pm

Con

Lab

UbDetn

Other

GE

336

271

20

24

ICM

302

307

18

24

HARRIS

305

294

25

27

NOP

301

298

25

27

their 8 p.m. figures would have given a 12-seat Tory majority on auniform swing and six seats on proportional loss; a most credible predic-tion of within 15 seats at worst. But when they had been massaged at ITN,the Tories were expected to be 21 seats short of a majority, not a majority.I have challenged ITN to come clean on how this happened, on theTuesday night following the election; TV expects, nay demands, instantresponse from us; I am still waiting for the explanation of how theirfigures were arrived at.

NOP was equally commendable. Their 43 per cent/39 per cent/16 percent/2 per cent gave a 0 per cent/+4 per cent/0 per cent/—2 per cent/—2per cent error or +/—2 per cent per party average error. If projected on auniform swing, this share measure would have given an 18 seat Conserva-tive majority, again, most respectable. On a proportional-loss swing, amajority of 14; still neither 'wildly wrong' nor 'grossly misleading' assome of the psephologists have described them.

The ICM poll, which had to close and report at 4 p.m. to meet thedeadlines for their sponsoring newspapers, was the furthest out, despiteinterviewing some 25,300 people - so much for the sample size argument.The fact that 38 per cent voted after 5 p.m., rather more after 4 p.m., andthe late Tory surge (see Harris, above) can explain quite a lot of the ICMerror - but nothing can possibly justify the churlishness of one newspapereditor who despite specific warnings of the risk insisted on publishingfrom data cut off at 4 p.m. and then slagged off ICM and said they wouldnot be paying for 'shoddy goods'.

The NOP Exit Poll also detected the late swing to the Conservatives,

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 20: The polls have a lot to answer for!

THE POLLS HAVE A LOT TO ANSWER FOR! 373

increasing Tory turnout as the evening progressed. At 5:30 p.m. theLabour share was 41 per cent, by 7:30 it had fallen below 40 per cent, by10 p.m. when the polls closed Labour's share was 38.7 per cent.

Why NOP and the BBC chose the risky methodology of marginal theywill no doubt explain; also, the MRS Enquiry will be reporting shortly,and we can look to the excellent choice of the committee personnel toshed light on alternative methodologies and the pros and cons of each.May I just say in defence of any criticism from any quarter that having twoex-NOP directors on the MRS Committee worries me not one jot.

While considering projections, it is interesting to note in passing thatuniform swing, proportional loss and regional weighting all misled thevotes-to-seats projections. The Payne BEIPS model would have forecasta 75-seat Conservative majority on the actual result, the Nomura model a65-seat majority, the Shaw-Harrop model and the MORI model both 64.The majority was 21. Back to the drawing board? No: Labour performedbetter in the marginals, thereby confounding conventional assumptionsabout seats and votes.

Banning Polls

Over the years there have been a number of calls for the banning ofopinion polls in the final seven days of a election campaign. These havenever gone very far either with the public generally or where it counts, inthe House of Commons. It has not been the policy of any of the partiesrepresented in the House to consider such an illiberal act. None the less,one is met all too frequently with the question from radio and TVinterviewers 'Isn't it time polls were banned, like in France?' My answeris usually to the effect that as 'economic' man I would favour it, becausethe parties would continue to conduct their private polls at an even higherlevel, and the City would fall all over themselves to commission, use,leak, and even manufacture polls for their own benefit. As in France now,the pollsters, the psephologists, the pundits, the political writers, thepoliticians and their friends would all know the results of the latest privatepolls; the only ones who would be in the dark are those for whomelections are supposed to be about, the voters.

To ban polls would be an illiberal act, removing from the electorate theone objective and systematic information they are provided by an in-creasingly biased press. Yet calls have already come; letter writers (afew), pundits (one or two), one politician, and some members of thepublic.

As in previous elections, we tested the public mood, on a nationalsurvey conducted among 1,065 adults at 53 constituencies across Great

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 21: The polls have a lot to answer for!

374 CONTEMPORARY RECORD

Britain on 6-7 April 1992. We found some support for banning pollsduring election campaigns - a quarter of the public favoured such a ban -but seven in ten disagreed! Interestingly, the 25 per cent support wasexactly the level for such a ban as was found in 1987.

We also found then, and now, that a quarter of the electorate is infavour of banning party election broadcasts during general elections, onlya few percentage points more than the 22 per cent who would have allcoverage of the election on radio and TV banned, and only double the 12per cent who would (seriously) ban all coverage of the election innewspapers.

Reluctant Choice

A majority, 52 per cent, of electors worried about their belief that theTories would privatise the National Health Service, including 26 per centof their own supporters. At the same time, seven in ten of the public,including nearly half, 47 per cent, of Labour supporters believed thatmost people would pay more taxes under a Labour government.

The Conservatives spent nearly all of their advertising money in thefinal three days (at a weight greater, annualised, than the spending ofProctor & Gamble or Unilever), the Tories levelled all of their guns at theLiberal Democrats' voters 'letting Labour in', and the testimony of theLiberal Democrats' campaign manager was that this did great damage totheir support in the final hours of the campaign. The Tory tabloids did allthey could on their front pages, never mind in their leader columns, toensure the Conservatives were returned to power. Labour's peak cameon 'Red Wednesday', eight days before polling day, with published leadssufficient to give them an overall majority.

Right to the end of the election, the 'floating voters' were higher thanever before. In the end, I must conclude that the reluctant choice camedown to a decision with four or five people in a hundred who, on the day,decided better the devil you know than the devil you don't, and chose tovote Conservative.

Labour's 'Sheffield Rally' proved the beginning of the end for Labourand its leader Neil Kinnock; and the dawning of a fourth term for theConservatives under their newly elected Prime Minister, John Major.

NOTES

1. R. Worcester, 'History of Political Polls in Great Britain', Paper presented to the SilverJubilee Market Research Society Conference, 1982.

2. R. Worcester (ed.), Political Opinion Polling: An International Review (LondonMacmillan; New York: St Martin's Press, 1983).

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14

Page 22: The polls have a lot to answer for!

THE POLLS HAVE A LOT TO ANSWER FOR! 375

3. R. Worcester, British Public Opinion: A Guide to the History and Methodology ofPolitical Polling (Oxford: Basil Blackwell, 1992).

4. B. Chappell, 'Founding Fathers: Henry Durrant', Market Research Society Newslet-ter, Nos. 157-158, April-May 1979.

5. Winston Churchill, The Second World War, Volume VI, Triumph and Tragedy (Lon-don: Cassel, 1954).

6. Lord Moran, Winston Churchill: The Struggle for Survival 1940-1965 (London: Con-stable, 1966).

7. David Butler and Dennis Kavanagh, The British General Election of 1987 (London:Macmillan, 1988).

8. Ivor Crewe, 'Improving, but Could Do Better', in Robert M. Worcester and MartinHarrop, Political Communications: The General Election Campaign of 1979 (London:George Allen & Unwin, 1982).

9. David Butler and Anthony King, The British General Election of 1964 (London:Macmillan, 1965).

10. Conclusions on the Review of the Law Relating to Parliamentary Elections, Cmnd 3717,July 1968.

11. William Whitelaw, MP, Home Secretary, reported in the Daily Mail, 22 March 1983.12. Official Report, 37th Ordinary Session, Parliamentary Assembly, Council of Europe,

30 Sept. 1985.13. Pierre Huet, Testimony to the Council of Europe Parliamentary Assembly hearings,

Strasbourg, 8 Oct. 1984.14. Chris Baur, Testimony to the Council of Europe Parliamentary Assembly hearings,

Edinburgh, 26 Nov. 1984.15. George Foulks, MP, Testimony to the Council of Europe Parliamentary Assembly

hearings, Edinburgh, 26 Nov. 1984.16. Dick Leonard, 'Belgian Leaders Should Read "Areopagitica"', Wall Street Journal, 18

Oct. 1985.17. Leith McGrandle, 'Should We Ban Opinion Polls?', Evening Standard, 2 May 1979.18. R. Worcester, 'Opinion Polls in British General Elections', Paper presented to the 1992

Market Research Society Conference, 19 March 1992.19. Daily Telegraph, 18 March 1983.20. Hugo Young, The Sunday Times, 21 Nov. 1983.21. Bryan Walden, Evening Standard, 31 May 1983.22. Bryan Walden, 'A Second Opinion Can Do No Harm', Sunday Times, 19 May 1987.23. Bernard Levin, The Times, 17 March 1983.24. The Guardian, 10 May 1986.25. Glasgow Herald, 26 Feb. 1987.26. Quoted by Anthony King, 'Why Did They Get It Wrong' in the Daily Telegraph, 11

April 1992.27. 'Parties Make Final Push to Capture Floating Voters', Robin Oakley, The Times, 8

April 1992.28. Times Guide to the House of Commons, 1987 (London: Times Books, 1987).29. Robin Oakley, 'Leadership Gap Still Troubles Labour Despite Lead in Polls', The

Times, 18 March 1992.30. Ivor Crewe, 'One Poll Victory Does Not Make Kinnock's Summer', The Times, 18

March 1992.

Dow

nloa

ded

by [

The

Uni

vers

ity o

f M

anch

este

r L

ibra

ry]

at 1

3:22

05

Dec

embe

r 20

14