PUBLIC OPINION POLLS SIR,—Pharos asks, 'Why were the pollsters, who

fore- cast a much bigger margin for Mr. Bonham Carte r, so far out?'

Which forecasts? Were the polls really 'so far out As far as the Gallup Poll is concerned, we d not attempt to forecast the Torrington result a Pharos will search in vain for evidence that we di In Torrington, as in other by-elections, we were co cerned to : (a) comment upon the campaign as it d veloped; (b) assess the relevance of the by-election

to

the national situation; and, most it lY portantly,

(c) to test our questioning, sampling and analytical techniques. That research considerations are overriding is shown by the fact that the majority of our by- election studies are never published in the News Chronicle, although. they are always available to political students. The decision, on the part of the News Chronicle, to publish some of the findings in the Ipswich and Torrington by-elections was taken because of the exceptional public interest in these contests. But even then research considerations were foremost and the newsworthiness of the results a by product. This can be confirmed by reference to the copy of the 'Technical Notes' which we sent to you at the time of the Ipswich by-election.

If we were going to forecast the result we cou not afford to experiment, nor could we afford operate on the limited scale we employ for b elections. In Torrington, for example, we made on 600 interviews. Moreover, if we were attempting forecast we could not ignore the very importa final four days of the campaign.

If Pharos will check on °lir Torrington relcas he will find three important facts : (a) they stress that the figures relate to the situation four days before polling—no attempt is made to project them on to polling day itself; (b) the percentages relate to the total sample interviewed and do not show a split of votes; (c) very important reservations warn the readers of the dangers of interpreting the figures as a forecast.

The most important of these reservations was that the result would be 'greatly affected by' the ability of Mr. Bonham Carter to establish clearly his party label. We drew attention to the fact that some '10 I

nd

id d.

e-

ld to

IY

a Pt

es THE SPECTATOR, APRIL 11, 1958 Per cent, of the' electorate were still confused about the party labels.' As reference to our figures clearly demonstrated, this one factor of confusion could have swung the result over to Mr. Royle.

Pharos should note that our reference to and estimate of the confusion factor was made before the result was known. The Gallup Poll, I submit, did a very good job 11) Torrington. The split of the Labour/non-Labour vote was anticipated in our weekend figures exactly. We pointed out Mr. Bonham Carter's 10 per cent. confusion handicap which he and his workers would have to overcome.

It was the political commentators, who, after all, were in a far better position than We were to assess the abilities of Mr. Bonham Carter's organisation, Who failed to realise the implications of our findings. We did our best to draw their attention to the paten-. fiat dangers of the Liberal situation. it is now clear that the party-label confusion per- sisted substantially up to polling day itself. The National Liberal tag was worth at least 2,500 votes to Mr, Royle. This last statement is further evidence that re- search considerations dominate our thinking. It would be expedient to stay mute on this point. The National Liberals pose a very serious problem to tC

tion.political forecaster—our role at a general elec- Anything which might help to bring them into more lively participation is definitely not in our interests.—Yours faithfully, • WILLIAM GREGORY Director Social Surveys (Gallup Poll) Limited, 59 Broad Street, W.1 PS.—Pharos raises the old canard that polls in- fluence elections. He backs his hunch both ways— Polls, he states, helped both the Liberal and the Conservative candidates. This contention is as old as polls themselves. So far no one, including Pharos, has produced any evidence to support it. Could this be because there is no truth in it whatsoever?

[Pharos writes: 'By "the pollsters" I was not re- ferring to the Gallup Poll alone : the Mail and the ExpreAs also ran public opinion polls, whose findings tallied reasonably closely with Gallup's; and all three showed the Liberal candidate in the lead by a much wider margin than that which he eventually secured. I agree that these polls arc not in fact forecasts in the meteorological sense; but they are understandably taken as forecasts when they appear in the papers 4,,day or two before polling day. Mr. Gregory denies they have any effect on the actual voting; It would be interesting to know, then, what in fact caused the considerable increase in the Liberal percentage dur- ing the campaign both at Rochdale and at Torring- t°1).'—Editor, Spectator.]