AUSTRALIA'S `STOLEN CHILDREN': WHICH POLL WOULD A POLL-FOLLOWING PRIME MINISTER HAVE FOLLOWED?
A report into indigenous Australian children taken from their families recommended that the Government apologize. Three polls on the question of whether the Government should apologize produced three quite different results: a `yes', a `no' and one which was more evenly divided. This paper shows why this happened. It relates the results to three quite different understandings of what opinion polls should model: opinion expressed through plebiscites; `real' opinion; and opinion based on some sort of deliberation. And it explores the relationship between what a poll-following Prime Minister might have done and scholarly judgments about `quality' in public opinion polls.
The Government seems reluctant to sound out the electorate on Aboriginal issues. Perhaps the following referendum that I have prepared would help.
The `Stolen Generations' report has revealed the attempt by past governments and other agencies to destroy the culture of indigenous Australians. Children were forcibly removed from their families and many abused and maltreated. The emotional scars are still felt today. The policy was called `cultural genocide', akin to the attempt to destroy the Jews of Europe in the 1940s. As a first step in the reconciliation process, I would like to apologise for what was done in my name by past governments and individuals. Yes  NO  Bob Selinger, Eastwood (Sydney Morning Herald 12 June 1997)
As I read Bob Selinger's survey I was reminded of an episode of Yes, Prime Minister ... An equally emotive statement to the reverse effect could have been substituted for every statement he makes and the average victim of the survey would have been driven towards the opposite response. For example ...
It is a fact that many Aboriginal children were the victims of a culture of heavy drinking and the resulting neglect. Disease and malnutrition were in plague proportions, and children's lives in dreadful and immediate risk. Do you believe the Government had a responsibility to protect the lives of innocent children in this situation? I wonder whether we really want to know what a fair survey would reveal about the opinions of the majority of Australians ... G. Kettlewell, Warrawong (Sydney Morning Herald 20 June 1997)
To suspect that the Howard Government might be `reluctant to sound out the electorate' on Aboriginal issues, or on any other issue for that matter, is about the last thing one would expect a student of Australian politics to lay at the door of the present government. And not only at the door of the present government. Since the Whitlam Government came to office, a quarter of a century ago, all Australian governments have kept a close eye on the published polls; the major parties have supplemented these with their own programs of opinion research; and most commentators have assumed that governments are more likely to follow the polls than to defy them.
Whether prime ministers ought to follow the polls is a separate question. In Australia, poll-following has often been raised by commentators as a criticism or concern: prime ministers are meant to lead not to follow. Governments suspected of following the polls are more often derided for being poll-driven than praised for deferring to public opinion. As Mick Young, a former Labor Party National Secretary and member of Whitlam's shadow cabinet, once lamented: `we are fast becoming a generation of political activists without the courage or conviction to make a decision lest we unleash the wrath of the pollsters' (Australian, December 24, 1977).
Among Americans, by contrast, politicians who trim their sails to the polls are more likely to be seen as true democrats. `No principle, in democratic theory, has been more fundamental', declared one American political scientist fifty years ago, `than the belief that political decisions ought to be made by the people as a whole or in accordance with their desires' (Ranney 1946, p. 534). George Gallup, whose polls were rapidly coming to substitute for `the people as a whole', declared that `Democracy doesn't require that every man [sic] be a political philosopher'; it required `merely that the sum total of individual views add up to something that makes sense' (Gallup 1944, pp. 74-5, emphasis in the original). Nearly fifty years later, Benjamin Page and Robert Shapiro were to take precisely the sort of data that Gallup had helped make commonplace and demonstrate that the collective preferences of the American public did indeed `make sense': that repetitive surveying had proved public opinion in the aggregate to be rational, stable, and predictable. Their landmark report was grounded in the `conviction that ordinary citizens are not to be feared, that governments should respond to their wishes, and that the politically active should learn more about what the public wants' (Page and Shapiro 1992, pp. xi-xii, 29-30).
Those who believe that governments do follow the polls, or at least that they ought to, often assume that the polls measure what they are supposed to measure and that the task of interpreting their results is relatively straightforward. They assume, in other words, that public opinion exists independently of its being polled and that one poll is as likely to give public opinion its voice as any other: `Technology makes it easier and easier to discover the public's policy preferences [via] "instant" media referenda' (Cronin 1989, pp. 158-9). Political journalists almost always assume that if the government's research is telling the Prime Minister one thing, the Opposition's research won't be telling the Leader of the Opposition anything very different.
What happens, however, when two or more differently worded polls point the parties in two or more quite different directions? Under these circumstances, what would--or should--a poll-following Prime Minister do? Choose at random? Take note of the most `balanced' of the polls, ignoring those which offer `a one-sided dose of policy-relevant information' (Page and Shapiro 1992, p. 29)? Or follow the poll which best fits the Government's preferred position?
Polls taken after the release of the `stolen children' report, in May 1997, placed the Prime Minister, Mr Howard, in precisely this dilemma: one suggested widespread support for a government apology to Aboriginal people; one suggested widespread opposition to such a course; and one fell somewhere in between.
This paper describes what the polls showed and offers an explanation for the very considerable differences in the figures they produced; this explanation hinges on context effects, backgrounding and the effects of question wording. It then asks whether one poll might be said to represent a truer picture of public opinion than either of the others and argues that this depends on whether public opinion is to be understood in terms of plebiscites, whether it presupposes knowledge and awareness on the part of the public, or whether it entails public deliberation. If one poll does not reflect public opinion, in any absolute sense, more closely than any of the others, what should a poll-following prime minister do?
POLLED OPINION ON THE `STOLEN CHILDREN'
An inquiry into Indigenous Australian children, `forcibly removed from their families', from the very first days of European occupation to the present, was established by the Keating Government in August 1995. Bringing Them Home, the report prepared by the Human Rights and Equal Opportunity Commission (HREOC), was handed to the Howard Government early in April 1997 (SMH April 5, 1997). It found, among other things, that `between one in three and one in ten Indigenous children were forcibly removed from their families between 1910 and 1970; it followed the United Nation's Convention on Genocide, by labeling the policy that had underpinned these practices `genocide'; and it called on governments, state and federal, to make reparation--starting with `acknowledgment and apology' (HREOC 1997, pp. 4, 27, 34).
But even before the Government tabled the report (i.e. presented it to parliament), on May 26, it was busy, off the record, attacking the inquiry's findings and dissociating itself from its recommendations (SMH May 20, 1997, May 22, 1997). On the day the report was tabled the Prime Minister, addressing a Reconciliation Convention, expressed his `personal sorrow' for the events the inquiry had laid bare (SMH May 27, 1997). Later he made it clear that no apology would be forthcoming from the Government itself since this might be misconstrued as an admission of wrong-doing. `If you're going to speak on behalf of the nation', he added, `then the nation must support it' (SMH May 29, 1997). Behind Howard's thinking, one commentator was quick to remark, was `the hard-headed view that mainstream Australia does not believe a national apology is warranted' (Weekend Australian May 31-June 1, 1997).
Did the public think a national apology was warranted? The weekend before the report was to be tabled in parliament, AGB McNair--a polling organization contracted to the Herald--sprang into action. A week later, the Australian's Newspoll entered the lists. And a week after that, the Morgan Poll ventured forth on behalf of The Bulletin. In their very different ways, each poll sought to establish how the Prime Minister could respond `on behalf of the nation'; or rather, if he responded by apologizing, whether `the nation' would `support it'.
The highest level of support for an apology was recorded by the first of the polls, the one taken by AGB McNair before the report was tabled. In this poll respondents were told of `the forcible removal of Aboriginal children from their families' and of the recommendations of the Human Rights [and Equal Opportunity] Commission, that `all Australian parliaments acknowledge the responsibility of previous governments' for these events and `formally apologise for the suffering caused'. Asked whether they agreed or disagreed `with this [sic] recommendation', 65 percent agreed.
The lowest level of support, 37 percent, was recorded by the last of the polls, the one conducted by Morgan. Here respondents were told that `Aboriginal children' had been `taken from their families between 1910 and 1970 and placed within the white community'; that `the report' (no mention of its being a report of the Human Rights and Equal Opportunity Commission) had been `dubbed' The Stolen Children Report; and that the Prime Minister had decided against the Government's giving `a formal national apology' because to apologize would be to `indicate in some way that the present generations are responsible and can be held accountable for errors, wrongs and misdeeds of earlier generations'.
Registering a level of support lower than that reported by AGB McNair but higher than that suggested by Morgan--indeed, almost exactly bisecting the two--was the survey conducted by Newspoll. Here respondents were informed of neither the recommendation of the Commission nor the reaction to it of the Prime Minister; they were simply told of an inquiry known as the `stolen children' inquiry and asked whether the Government should apologize to the Aboriginal people because of `the events revealed' (Table 1).
TABLE 1 Opinion on whether the Government should apologize for the `stolen children'
Poll Apologize Don't apologize Don't know n (Percent) AGB McNair 65 30 5 (2065) Newspoll 50 40 10 (1200) Morgan 37 57 6 (522)
`Recently a report by the Human Rights Commission has recommended that all Australian Parliaments officially acknowledge the responsibility of previous governments for the laws and policies which led to the forcible removal of Aboriginal children from their families. The Human Rights Commission has recommended that all Australian parliaments formally apologize for the suffering caused by those laws and policies. Do you agree or disagree with this recommendation?' (AGB McNair, May 23-25, 1997).
`Now for a question about the inquiry referred to as the "stolen children" inquiry. Do you think the Federal Government should or should not apologize to the Aboriginal people for the events revealed in the "stolen children" inquiry?' (Newspoll, May 30-June 1, 1997).
`Thinking now about the report into the Government policy which allowed Aboriginal children to be taken from their families between 1910 and 1970 and placed within the white community. The report has been dubbed "The Stolen children" report. The Prime Minister, Mr Howard, has decided against the Government giving a formal national apology. He said that to do so is [sic] to indicate in some way that the present generations are responsible and can be held accountable for errors, wrongs and misdeeds of earlier generations. Overall, do you agree or disagree with Mr Howard's statement?' (Morgan Poll, June 4-5, 1997).
Source: Sydney Morning Herald May 27, 1997, for AGB McNair; The Australian June 3, 1997, for Newspoll; Roy Morgan Research Finding No. 2993 and Computer Report No. 1434, for The Bulletin June 17, 1997.
What do these disparate data mean? Part of the explanation for the very different readings offered by each of these polls may have to do with the circumstances in which each was conducted. The AGB McNair poll was conducted before the Commission's report was tabled in the parliament; Newspoll tested opinion just after the Prime Minister had discounted the possibility of a formal apology; while the Morgan Poll was taken a week after that. It may not be entirely coincidental, therefore, that it was the AGB McNair poll that provided the most favorable response to the report and the Morgan Poll the least favorable. But the time frame--less than two weeks--is surely too short for this to account for any more than a small proportion of the shift, especially when the Leader of the Opposition and much of the media had taken the Prime Minister to task for not offering an apology.
The larger part of the explanation, if not the whole of it, almost certainly lies elsewhere--in the ways in which the issue was framed by the polls. As we have seen, each of the polls varied: in terms of the information about the report offered to (or withheld from) respondents; in the range of reactions to the report brought to respondents' attention; and in the status of the players whose reactions were incorporated as background information by the polls. And where the polls led, however unwittingly, respondents were inclined to follow.
Thus, AGB McNair. boosted support for an apology by the simple expedient of ensuring that the only information it provided, or recommendations it rehearsed, were sourced to the `Human Rights Commission'. Morgan, by contrast, was able to persuade respondents that the Prime Minister should not apologize not only by offering a misleadingly narrow account of the period covered (`between 1910 and 1970') and by failing to identify authorship of the report; but, more importantly, by drawing respondents' attention exclusively to the Prime Minister's reaction and by rehearsing his reasons for rejecting an apology while not offering any countervailing reasons why the report ought to be accepted. Newspoll achieved a more even split by omitting any reference to the source of the report and any direct reference to the nature of its recommendations, thereby avoiding not only arguments in favor of the report but reasons for opposing it. Apart from being told of a report on `stolen children', affecting Aboriginal people, respondents were given no cues as to which way they should jump. The result? A division of opinion weighted slightly towards an apology--an understandable outcome, given the information.
The importance of the cues offered by Morgan is underscored by the response elicited by a second, but almost identical, question asked in the same Morgan Poll. Having just indicated whether they agreed or disagreed with the Prime Minister's decision, Morgan's respondents were asked `whether or not there should be a Federal Government apology'. On this occasion, however, the response options were doubled; instead of a two-fold choice they were offered a four-fold choice: `The Federal Government should formally apologise even if the apology makes it easier to claim compensation' (23 percent chose this option); `the Federal Government should formally apologise only if it does not make it easier to claim compensation' (chosen by 27 percent); `The Federal Government should not apologise because it is enough [sic] individual politicians to apologise' (l0 percent); `The Federal Government should not apologise because the policy was legal and well meaning at the time' (agreed to by 37 percent of those interviewed).
In response to the earlier question--the question which had immediately preceded this--37 percent had said they favored the Government's apologizing; but with the follow-up question the number jumped to 50 percent, provided that any apology did not make it easier (for Aborigines) to claim compensation. In response to the earlier question, 57 percent said they favored Howard's decision not to apologize; now, 47 percent did so. In short, what had been a margin of 20 percentage points in favor of Howard's position suddenly became a margin of three percentage points against. The explanation, almost certainly, lies not only in the responses allowed by the second of the two questions, but in the absence in the second question of any reference to the Prime Minister's position or his reasons for holding it.
MIRROR, MIRROR ON THE WALL ...
On the categorical question of an apology, can one of the three results--AGB McNair's, Newspoll's or the first of Morgan's results--be said to have been more truly representative of public opinion than either of the others? The answer to this question depends on what one understands by public opinion. Depending on where one starts, one can argue for the superiority of the second poll over the first and third, the inadequacy of all three polls, or the virtues of the first and third against the second.
OPINION POLLS AS PLEBISCITES
More than fifty years ago, George Gallup advised that the `most practical solution to the problem of reporting and describing public opinion on a typical issue'--including when a `definite and specific proposal' is advanced `before there has been very much discussion of the problem which underlies the proposal'--involved polling companies taking the following four steps: starting interviews with an information question designed to find out what their respondents knew about the issue; following-up with an open-ended question to elicit the direction of their thoughts; possibly, putting in a question to see how strongly respondents felt about the matter; then asking `one or more categorical questions to discover how each person would vote on specific proposals' (Gallup 1944, p. 36. This was subsequently expanded to include a fifth dimension, `reasons why'; see, Gallup 1947, 1948, pp. 40-9).
However, the polling Gallup conducted through the American Institute of Public Opinion Research generally ignored every dimension of the practice he had enunciated so famously--except the last: the instruction to pollsters that they ask categorical questions to see how people would vote (see, Cantril 1951, pp. 917ff; Gallup 1972). `Most questions asked by polling organisations', he conceded, `are intended to reveal how the public would vote if the same questions were put to voters in a nationwide referendum (Gallup 1948, p. 40). How each person would vote on a specific proposal was the nub of it; not their knowledge, not their own ideas about the issue and not the strength of their convictions. In a democracy, Gallup insisted, `the majority usually registers sound judgment on an issue, even though a good many are ignorant or uninformed' (Gallup 1944, p. 75, 1948, p. 85; also Page and Shapiro 1992, pp. 14-27). Gallup was strongly attracted to plebiscites or referendums as the democratic ideal (Gallup and Rae 1940, pp. 30-1); he constantly thought of his polls as modeling referendums, even if the referendums they were modeling were not to be regarded as binding (Gallup 1944, pp. 7-8; 1948, p. 8). In his manifesto for `The New Science of Public Opinion Measurement', he described polls as `sampling referendums' no less (Gallup 1938, p. 1; for the dating of this pamphlet, see Converse 1987, p. 116); the fact that, in a political system of non-compulsory voting, his sampling was more representative of the citizenry than any referendum was ever likely to be was a strength, he said, not a weakness (Gallup cited in Cronin 1989, p. 180).
Given this account of public opinion, which of the polls on the `stolen children' fits best? The answer, surely, is the poll conducted by Newspoll. Admittedly, Newspoll was no different from the other polls in assuming rather than establishing that most of its respondents had heard of the issue (more precisely, it assumed that those who hadn't heard of, or thought about, it would show up as `don't knows'); in eschewing an open-ended question in favor of a closed one; and it making no attempt to measure the intensity of opinion. Where Newspoll differed, however, was in confining itself to a `categorical question'--a question designed `to discover how each person would vote' (assuming voting to be compulsory) on a `specific proposal'. Newspoll was the only poll to offer respondents no substantive information about the nature of the Commission's inquiry or about the recommendations which came out of it. It rehearsed neither the case for a Government apology nor the case against. It presented respondents with a single question posing the simplest of choices encased by a minimum of information-nothing more but nothing less than the sort of information voters would receive at the ballot box if they were casting their vote on a particular proposition at a plebiscite or referendum. It simply asked whether the Federal Government `should or should not apologise'.
PUBLIC OPINION AS `REAL' OPINION
`Dr Gallup', one his early critics remarked, `takes many of his polls before there can be any public opinion--indeed, before the public can have any information on which to base an opinion, and he asks his questions of all and sundry, most of whom do not yet realize that there is an issue the decision of which may affect them' (Rogers 1949, p. 43). `Even on issues which affect the bulk of the people', argued other democratic theorists, pushing the argument a good deal further, `it is unusual to find as many as twenty per cent of [people] sufficiently concerned to form any definite opinion' (Benn and Peters 1959, p. 337, though the evidence they cited was hardly relevant). John Plamenatz put the matter more bluntly: `Most people have no opinions about most public issues' (Plamenatz 1973, p. III, another opinion offered without any evidence).(1)
For public opinion to be `real', according to Rogers, the public did not need to be particularly well-informed; on the contrary, `much of what we call "opinion" is accepted without question' and the problem with the polls, he argued, was that they asked questions before people had had a chance to take their bearings from party leaders and other `authorities' (Rogers 1949, pp. 28-30). Benn and Peters appear not to have demanded an especially knowledgeable electorate either; the basis on which they judged the public not to hold `definite' opinions--what we have called `real' opinions--was not so much how little the public knew but how little they had done to `spread' their views (Benn and Peters 1959, p. 337). With Plamentaz, too, what damned the polls was not just that the public did not know much but how poorly the polls predicted political action (Plamenatz 1973, p. III).
On this view, then, public opinion in relation to the `stolen children' could only be represented through the polls if the public was aware of the report, had been given an opportunity to hear it discussed, and had formed some `definite' opinion on the matter--even if not a well-informed one. Or to put the point another way: the only people one could count as `the public' on this issue would be those who knew of the report (independently of being told of it by the interviewers), who had had a chance to discuss the issue or at least to listen to those whose views on such matters they trusted, and who had formed clear opinions about an apology--either in favor or against.
Whether respondents had heard of the report, were aware of the reactions to it, or had discussed it among themselves are things that none of the polls attempted to discover. The first poll was conducted after the report had received some media coverage but before the Prime Minister, officially, had shown his hand; the second came hard on the heels of the release of the report and the Prime Minister's declaration that there would be no formal apology; even the third poll, which followed a week later, may have been taken before the issues had had time to sink in. The question of an apology bore directly on the interests of very few of the respondents. Given the susceptibility of the respondents to the manner in which the three polls chose to present what was fundamentally the same proposition, it seems reasonable to infer that even if the respondents had been aware of the report, they probably knew little of its findings, had not given it much thought and certainly had not formed definite opinions about it.
In this sense, public opinion on the `stolen children' might well be said not to have existed. If that is the case, one poll could hardly be said to have produced a truer picture of it than another; on the contrary, each would have to be condemned as misleading.
PUBLIC OPINION AS DELIBERATIVE OPINION
Criticism of the polls based on their attempt to measure opinion before it has had time to take shape has been taken a good deal further by those concerned not just that polled opinion should be `real' but that it should have other qualities as well--that it should be well-informed; that it should be shaped by respondents' direct participation in public debate or at least by their exposure to the arguments of the main protagonists in such a debate; and, as a consequence, that it should be not easily shifted.
It is an ideal of this kind that underwrites a number of contemporary attempts to reconceptualize the position of the public in the democratic process. One such attempt is captured by James Fishkin's idea of a `deliberative opinion poll'--not `an ordinary opinion poll' which `models what the public thinks, given how little it knows', but a poll which `models what the public would think, if it had a more adequate chance to think about the questions at issue' (Fishkin 1991, p. 1; 1995, pp. 205ff for British and some American experiments along these lines; Merkle 1996 and Price and Neijens 1998 for a more critical view). A second is James Farr's paradigm of the jury, a group brought together to hear the evidence, to deliberate on it and to come to a decision (Farr 1993, p. 388; Adonis and Mulgan 1997, pp. 238-40 for a review of the experimental evidence). A third is Daniel Yankelovich's quest for `quality' in public opinion--consistency, firmness, and a willingness to take responsibility for the consequences of one's views--something that might transform `mass opinion' into `public judgment' (Yankelovich 1991, ch. 2).
But the contrast Fishkin draws between an `ordinary opinion poll' and a `deliberative opinion poll' is too sharp. Not all pollsters subscribe to the plebiscitary model of public opinion celebrated by Gallup; and some `ordinary polls' are not as far removed from Fishkin's ideal as his own characterization of them suggests. On the contrary, it is precisely because they have shared an understanding of public opinion very similar to the one championed by Fishkin that some pollsters over the years have attempted to educate, inform, or tutor respondents--to brief them on the pros and cons of an issue--before asking them (in Gallup's terms) to `vote'.(2)
If the extent to which it encourages deliberation is the criterion of a good poll, then both the AGB McNair and the Morgan polls might be said to have modeled public opinion better than the Newspoll. Both, after all, did attempt to give respondents some background on which they might base their judgments; Newspoll did not. This is not to be blind to the fact that, measured against the deliberative model, both polls left a lot to be desired. For one thing, the information offered by the two polls was one-sided--AGB McNair's in favor of the Commission's recommendations, Morgan's against; a `balanced' introduction might have combined elements of argument and counter-argument (see Schuman and Presser 1981, pp. 184-92). For another, respondents (as the word suggests) merely responded; there was no scope for any other form of interaction. Nonetheless, however deficient the execution, deliberation of some form cannot be gainsaid.
WHAT WOULD A POLL-FOLLOWING PRIME MINISTER HAVE DONE?
The Sydney Morning Herald, having commissioned a poll on the `stolen children', was not about to entertain any doubts about its validity. AGB McNair's poll, it argued, showed that the `Australian community' had `no trouble distinguishing between a formal apology and the acceptance of guilt. Nor should Mr Howard' (SMH May 29, 1997). A few days later, however, under the headline `Black apology splits voters', the political editor of the Australian used Newspoll's findings to argue that `Australians are divided on whether the Federal Government should apologise'. Had it been an issue on which the paper felt strongly, however, the 50:40 split might have been reported rather differently; two years earlier, for example, when Newspoll reported that public support for Australia's becoming a republic had reached 50 percent (with 35 percent against) the Australian's headline declared in large type: `Republic yes' (Australian June 22, 1995). The Bulletin, armed with the Morgan Poll, took a very different line. `John Howard's veto of a national stolen generation apology has received firm [sic] public support', it announced (The Bulletin June 17, 1997). There was no reference to the item which had offered respondents a four-fold choice and yielded a result marginally in favor of a Federal Government apology; and there was no mention of either the AGB McNair poll or the Newspoll.
Which of the three results should a poll-following Prime Minister have followed? Howard Schuman (1986), in his presidential address to the American Association of Public Opinion Research, counseled his audience not to confuse survey questions with policy questions: marginal frequencies from single polls should not be used to derive answers to complex questions of policy. Survey results, he emphasized, were sometimes a function of the way questions were worded and the contexts in which questions were asked; polls typically assume that an issue has only two possible solutions, when the reality facing policymakers is usually more complex; and the opinions recorded by polls might be ill-informed and unstable. Marginal frequencies, he insisted, only made sense in comparative terms--between one subgroup and another, across cultures or through time.
In the case of the stolen children, the Prime Minister was confronted by the sort of poll data (including, perhaps, data derived from his party's private polls) which would have alerted him to the importance of question wording: he was certainly not hostage to just one poll. The policy issue--whether or not to make a simple, largely symbolic, but highly significant statement--was not particularly complex, especially if one discounted the Prime Minister's view that there was a clear connection between the making of an apology and the government's legal liability for compensation. And this was one of those issues (of the kind Schuman himself happily acknowledged) on which there were no prior data and where time was of the essence; if the Prime Minister was going to apologize on behalf of the nation, and be seen to be genuine, he needed to do so without too much delay.
What the polls revealed, perhaps because respondents knew little about the issue, was a pattern of opinion which was unstable; this, in turn, pointed to a public with contradictory potentials for political persuasion. The Morgan Poll certainly suggested that if the Prime Minister stuck to the script that `present generations' cannot be `held accountable for errors, wrongs and misdeeds of earlier generations' he might have carried the day. But, on the evidence produced by AGB McNair, had the Prime Minister decided to endorse the Commission's recommendations he may well have carried opinion in that direction, too. Indeed, since any endorsement of the Commission's findings by the government would have been backed by the Opposition, persuading the public of the merits of an apology might have been easier than getting electors to agree that the recommendations were best ignored.
The three strategies outlined by Schuman (1986, pp. 437-8) for coping with the `inherent ambiguity of marginals'--the inclusion of open-ended questions, the asking of a wide variety of questions and the enlisting of representatives from opposing sides to help develop questions--should be recognized as important elements of sound survey practice. But the survey practice to which they apply needs to be spelt out. Schuman's paradigm of public opinion is grounded not in the plebiscitary model (which he attacks) or in the deliberative model (which was not to come into vogue until a few years later); it is based on the quest to uncover what we have called `real' opinion--a quest which lies at the heart of much, but by no means all, public opinion research.
If its value to survey researchers of this kind is large, its utility for politicians is limited. Apart from the fact that political leaders (or their advisers) almost always find themselves having to deal with poll data that fall short of Schuman's ideal, they are often concerned not with how public opinion can be properly measured but with how it might be politically managed. The quest to establish the truth about public opinion is often less pressing than the need to determine how successfully opinion can be `spun' or the public persuaded.
Conflicting poll findings of the kind we have isolated in connection with the `stolen children' are not unusual; polls in Australia on Aboriginal land rights (Goot 1994a), immigration (Goot 1991) and the monarchy (Goot 1994b) testify to that. The differences are not those generated by responses to different but closely related propositions; for example, permitting abortions when the mother's life is in danger versus allowing them when the mother simply doesn't want another child. Rather, they are differences generated by different ways of presenting what is essentially the same proposition, whether it be about feelings of happiness (Turner 1985, pp. 165-75) or attitudes to homosexuality (Smith 1994, p. 198, note 10).
The question on the `stolen children' run by AGB McNair was not the question suggested by Bob Selinger, in his letter to the Herald four days after the poll was published; but it did have a similar sort of bias. The first question asked by the Morgan Poll was not the sort of question dismissed by another Herald correspondent, G. Kettlewell, four days after this poll was published; but it shared the same sort of skew. Newspoll produced the fairest survey; yet a survey which exposed respondents to the sorts of arguments used in both the AGB McNair poll and the Morgan Poll would have produced a result that was not only fairer but better informed than either poll on its own.
What is at stake here is not just that reasonably familiar set of issues to do with which polls are fair, whether the opinions they seek to report are well informed or how far the polls might satisfy any other criterion of `quality' (see Price and Neijens 1997, for a fairly comprehensive list); what is also at stake, though less often noticed, are our understandings of public opinion itself. If, as George Gallup avowed, public opinion polls model plebiscites, then the best of the polls on the `stolen children' was the poll undertaken by Newspoll. It had a binary structure; it did not rehearse the arguments for or against an apology; and, appropriately for a country where voting is compulsory, it made no attempt to filter for awareness or interest. If, as Gallup's critics maintain, public opinion polls should attempt first to establish--following a period of wide discussion and considerable public debate--what respondents know and how much they care before they attempt to establish what positions they have arrived at, then a strong case can be made for saying that none of the polls provided a valid measure of public opinion. But if, as some contemporary theorists have suggested, public opinion should not be thought of in terms of what the public thinks or what public opinion polls normally measure but in terms of what the public would think after being `immersed in intensive deliberative processes' (Fishkin 1991, p. 81) or after being forced to accept the consequences of its own views, in a firm and consistent manner (Yankelovich 1991, p. 38), then some combination of the polls produced by AGB McNair and Morgan might take us some way towards approximating the ideal.
Meanwhile, poll-following prime ministers, of the kind Mr Howard is often said to be, might be placed in the exquisite position of being able to pick whatever poll took their fancy. `Public opinion', as V. O. Key defined it, was `simply ... those opinions held by private persons which governments find it prudent to heed' (Key 1964, p. 14). On this view, whatever poll a prime minister pinned his or her policy on would, ipso facto, constitute public opinion. And while some leaders may want to know `what a fair survey would reveal' (as the Herald's correspondent hoped), most might be more than happy to find a survey which tells them what--given their own preferences--they might be able to get away with.
(1) The evidence most often cited for the absence of real opinions, or the presence of `pseudo' opinions, derives from panel data; see Neumann (1986, ch. 1). The most sophisticated defence of the polls, though one which ultimately rests on collective rather than individual behavior, is in Page and Shapiro (1992, ch. 1); see also Smith (1994). For further discussion, see the chapters by Hochschild, by Kinder and Herzog and by Noon in Marcus and Hanson (1993).
(2) In Australia, the poll conducted by Irving Saulwick & Associates for the Sydney Morning Herald and the Melbourne Age, from the early 1970s to 1995, did this much more often than polls produced by Morgan or Newspoll. One reason why these last two enjoyed better reputations in political and journalistic circles than the Saulwick poll may have been that they fitted in better than Saulwick with plebiscitary understandings of public opinion.
Adonis, A. and Mulgan, G. (1997): `Back to Greece: the Scope for Direct Democracy'. In G. Mulgan (ed.) Life After Politics: New Thinking for the Twenty-First Century, London, Fontana Press.
Benn, S. I. and Peters, R. S. (1959): Social Principles and the Democratic State, London, Allen & Unwin.
Cantril, H. (ed.) (1951): Public Opinion 1935-1946, Princeton, Princeton University Press.
Converse, J. (1987): Survey Research in the United States: Roots and Emergence 1890-1960, Berkeley, University of California Press.
Cronin, T. E. (1989): Direct Democracy: The Politics of Initiative, Referendum and Recall, Cambridge, Harvard University Press.
Farr, J. (1993): `Framing Democratic Discussion'. In G. E. Marcus and R. L. Hanson (eds.) Reconsidering the Democratic Public, University Park, The Pennsylvania University State Press.
Fishkin, J. S. (1991): Democracy and Deliberation, New Haven, Yale University Press.
Fishkin, J. S. (1995): The Voice of the People, 2nd edn, New Haven, Yale University Press.
Gallup, G. (1938): The New Science of Public Opinion Measurement, Princeton, American Institute of Public Opinion.
Gallup, G. (1944): A Guide To Public Opinion Polls, Princeton, Princeton University Press.
Gallup, G. (1947): `The Quintamentional Plan of Question Design', Public Opinion Quarterly, 11, 385-93.
Gallup, G. (1948): A Guide To Public Opinion Polls, end edn, Princeton, Princeton University Press.
Gallup, G. (ed) (1972): The Gallup Poll: Public Opinion 1935-1971, New York, Random House.
Gallup, G. and Rae, S. F. (1940): The Pulse of Democracy, New York, Simon and Schuster.
Goot, M. (1991): `Public Opinion as Paradox: Australian Attitudes to the Rate of Immigration and the Rate of Asian Immigration', International Journal of Public Opinion Research, 3, 277-94.
Goot, M. (1994a): `Polls as Science, Polls as Spin: Mabo and the Miners'. In M. Goot and T. Rowse (eds.), Make A Better Offer: The Politics of Mabo, Pluto Press, Leichhardt.
Goot, M. (1994b): `Contingent Inevitability: Reflections on the Prognosis for Republicanism'. In G. Winterton (ed.), We, The People: Australian Republican Government, St. Leonards, Allen & Unwin.
HREOC (1997): Bringing Them Home: A Guide to the Findings and Recommendations of the National Inquiry into the Separation of Aboriginal and Torres Strait Islander Children flora their Families, East Sydney, Human Rights and Equal Opportunity Commission.
Key, V. O. (1964): Public Opinion and American Democracy, New York, Alfred A. Knopf.
Marcus, G. E. and Hanson. R. L. (eds.) (1993): Reconsidering the Democratic Public, University Park, The Pennsylvania University State Press.
Merkle, D. M. (1996): `The National Issues Convention Deliberative Poll', Public Opinion Quarterly, 60, 588-619.
Neumann, W. R. (1986): The Paradox of Mass Politics, Cambridge, Harvard University Press.
Page, B. I. and Shapiro, R. Y. (1992): The Rational Public: Fifty Years of Trends in Americans' Policy Preferences, Chicago, University of Chicago Press.
Plamenatz, J. (1973): Democracy and Illusion, London, Longman.
Price, V. and Neijens, P. (1997): `Opinion Quality in Public Opinion Research', International Journal of Public Opinion Research, 9, 336-60.
Price, V. and Neijens, P. (1998): `Deliberative Polls: Toward Improved Measures of "Informed" Public Opinion', International Journal of Public Opinion Research, 10, 145-76.
Ranney, J. C. (1946): `Do the Polls Serve Democracy?', Public Opinion Quarterly, 10, 349-60. In R. M. Christenson and R. O. McWilliams (eds.), Price of the People: Readings in Public Opinion and Propaganda, New York, McGraw-Hill, 1962.
Rogers, L. (1949): The Pollsters, New York, Alfred A. Knopf.
Schuman, H. (1986): `Ordinary Questions, Survey Questions, and Policy Questions', Public Opinion Quarterly, 50, 432-42.
Schuman, H. and Presser, S. (1981): Questions and Answers in Attitude Surveys, New York, Academic Press.
Smith, T. (1994): `Is there Real Opinion Change?', International Journal of Public Opinion Research, 6, 187-203.
Turner, C. F. (1985): `Why do Surveys Disagree? Some Preliminary Hypotheses and some Disagreeable Examples'. In C. F. Turner and E. Martin (eds.), Surveying Subjective Phenomena, Vol. 2, New York, Russell Sage Foundation.
Yankelovich, D. (1991): Coming To Public Judgment, Syracuse, Syracuse University Press.
For their comments on an earlier version of this paper I am grateful to Tim Rowse, the Journal's referees and the acting managing editor, Erich Lamp.
Murray Goot holds a personal chair in Politics at Macquarie University, New South Wales, Australia 2109; E-mail: email@example.com. He has co-edited Australia Gulf War (1992), Make A Better Offer: The Politics of Mabo (1994) and Taxing Australia (forthcoming).
|Printer friendly Cite/link Email Feedback|
|Publication:||International Journal of Public Opinion Research|
|Date:||Dec 22, 1998|
|Previous Article:||AGENDA SETTING AND COMMUNITY CONSENSUS: FIRST AND SECOND LEVEL EFFECTS.|
|Next Article:||Political Communication: Politics, Press and Public in America.|