Margin Of Error: Pollsters And The Manipulation Of Canadian Politics|
by Claire Hoy
Post Your Opinion
|Ruin Of The Country
by I. M. Owen
PUBLIC-OPINION polling arrived in this country in November 1941, when George Gallup set up the Canadian Institute of Public Opinion in Toronto. This made Mackenzie King very nervous. He was five months away from his plebiscite on conscription, and was afraid that the Institute would publish a poll on the subject that would reveal a deep division in the country along linguistic lines. (The plebiscite itself, of course, would do just that; King's thought processes were sometimes hard to follow.) It happened that a budding diplomat, Saul Rae, had collaborated with George Gallup in writing a book published the year before, and he was sent to Toronto to talk the Institute out of it. He was successful; and a consequence of this meeting was that the Gallup organization incorporated the Canadian Opinion Company to conduct confidential polls for government agencies, especially the section of the Wartime Information Board run by the social psychologist J. D. Ketchum.
I worked under Ketchum for the last two years of the war, though not directly on polls -- my contribution to victory was a weekly survey of editorial opinion. But I read the reports with interest, and learned from Dave Ketchum something about their limitations as well as their strengths. For instance, the interviewers inevitably put questions into people's heads, so that the poll results didn't necessarily reflect what was actually on their minds. Therefore Ketchum supplemented polls with letters from volunteers across the country whose work brought them into contact with a lot of people.
In the winter of 1943-4 we learned that Tommy Douglas, who was soon to leave the House of Commons to lead the CCF in the Saskatchewan election, had got wind of the WIB polls, suspected that the government was feeding information from them to the Saskatchewan Liberals, and planned to ask a question in the House. As I was known to have illicit contacts with the CCF National Office and M. J. Coldwell's office, I was dispatched to Parliament Hill to head Tommy off at the pass. I was able to tell him truthfully that the subject- matter of our polls contained little or nothing relevant to Saskatchewan affairs and, more importantly, that the total sample was only about two thousand people. As Saskatchewan had less than eight per cent of the population, this meant that there would be at most 160 respondents there, much too small a sample to reveal anything reliably. The question was never asked.
I cite this recollection from the early days of opinion polling because it contains the germs of some of the issues that Claire Hoy raises very effectively in this book: the possible effects of published polls on public opinion (King's anxiety about the plebiscite); the potential usefulness of unpublished polls to political parties (interesting that Douglas may have been the first politician to think of this); and the hopeless skewing you'll get if you try to break national poll results down into small regions -- all the more nowadays when the total sample is usually about a thousand people out of 26 million compared with two thousand out of 11 million. I suspect that in a country with what seem to be increasingly sharp regional differences, the small sample may produce increasingly misleading national results.
The systematic use of polls by political parties didn't begin until 1960, in the U.S. presidential election. It came into this country after Keith Davey read Theodore White's The Making of the President 1960, and persuaded Walter Gordon to hire Kennedy's pollster, Lou Harris, to conduct polls for the Liberal Party. From there it grew into the obsession we see today. In the 1988 election even the NDP, which had been late getting into the game in spite of Douglas's early insight, spent $300,000 on polls, the Liberals more than twice as much, and the Conservatives (hold your breath) $5 million. And, as Hoy points out, "one of the more scandalous oversights in the Canada Elections Act is that it doesn't count party polls as a campaign expense."
Of course, it's not just a campaign activity. The process of polling and interpreting goes on all the time. This, rather than their own conviction, largely determines the policies of the parties, which partly explains the strange anomalies in the 1988 election, when the old protectionist party passionately preached free trade with the United States while the leader of the old free-trade party reminded me irresistibly of Lord Tottenham, the mad protectionist in E. Nesbit's The Treasure Seekers, who crossed the Heath every day talking to himself, saying "Ruin of the country, sit! Fatal error, fatal error!"
Now, in a way it's the essence of democracy that the parties seek to find out from polls what the people want and shape their policies accordingly. But the essence of representative government by party is that each party should be bound together by a set of distinct principles. Poll results, which are quite often approximately right (perhaps even 19 times out of 20, as they say), could be very useful in showing the parties how to convert voters to their principles. instead, they allow the poll results to convert them -- at least for the duration of a campaign.
With the enormous proliferation of polling in the 1980s, it was time for someone to write a serious study of the subject; Claire Hoy has done that and done it very well. Of course it's written in slapdash journalese which the publishers didn't see fit to have translated into English, but it's an excellent book all the same. I've never read Hoy before, but know his reputation as a master of unrestrained insult. He has a chapter on each of the major pollsters, and sure enough he records every error, every self-contradiction, every hypocrisy -- and makes it convincing. Most valuably, perhaps, he exposes the incompetence of his own profession in reporting polls. As I write this, I have just read in the Toronto Star: "The federal Progressive Conservatives are in third place in party popularity with support from only 23 per cent of decided voters, a poll released today says." Well, it doesn't. It says that 23 per cent of 70 per cent of 1,501 respondents support the Conservatives, and within a stated margin of error this is very likely representative of the electorate as a whole. News reports should never say more than that the poll indicates a probability.
Hoy points too to the growing intimacy between the media and the polls. Being no respecter of persons, he attacks his own employer's purchase of 60 per cent of the Angus Reid firm: "Now that Southam actually has a major stake in the Reid poll, we can expect to see even more direct involvement .... the questionable practice of meshing the journalist's and the pollster's function as if they were the same and required the same skills."
Polls, especially the party-preference ones, are well worth following with interest. But don't look at another until you've read this book.