Randal Marlin

by Propaganda

  affirmatively by roughly 52 per cent). Since the respondent was told to indicate the

  statement that “best represents your opinion,” it becomes apparent that an obvious

  source of the seeming inconsistency lies with the degree of aversion to the contrasting

  statements and that the affirmative claims do not necessarily represent a view with

  which the respondents fully agree. The import of an answer to a survey question can-

  not be fully appreciated without seeing the question in context. To see how mislead-

  ing this form of survey could become, consider the following extreme: “Which best

  represents your opinion? ‘The tooth fairy exists’ or ‘Democratic Party rule will be best

  for the United States.’” People may have aversions to the Democratic Party, but if the

  only alternative is belief in the tooth fairy, most people would find they had to accept

  the second alternative, joking aside. It should not be difficult to construct question

  pairs where the “push” is subtler and not so obviously manipulative.

  Pollster Louis Harris helped improve the perception of President Richard Nixon’s

  standing in public opinion in the early 1970s. Instead of asking whether Nixon should

  be impeached, Harris’s question was whether respondents wanted Nixon “impeached

  and removed from office.” As Michael Wheeler comments, “Impeachment and

  removal, of course, are two quite different things. In essence, Harris was asking not

  whether Nixon should simply be tried, but whether he should be tried and hanged!”49

  Clearly, fewer people would give an affirmative answer to “impeached and removed

  from office” than they would to “impeached” alone.

  An important rule, emphasized by Wheeler, for interpreting the significance of

  poll results, is “to read the questions to see if you yourself would be comfortable giving

  agree/disagree answers. If you would not, then you must discount the results of the

  pol , no matter how conclusive the statistics seem.”50

  Polls have become part of the weaponry of policy justification, so it is not surpris-

  ing that, where politically contested ground is involved, appeals are made to different

  pollsters with their sometimes divergent results. For example, in October 1997, an

  Angus Reid poll said that the majority of Canadians wanted the government to cut

  both taxes and the national debt. However, a poll taken about the same time by Ekos

  said that three-quarters of Canadians were deeply concerned about the growing gap

  between rich and poor and wanted the government to start spending again on social

  programs. As critics pointed out at the time, the discrepancy comes in part from the

  superficiality of the context in which questions are presented. Everyone wants more

  benefits and wants to pay less to get them, so that serious policy can hardly be formu-

  lated on the basis of slapdash responses saying one thing or the other. Frank Graves,

  president of Ekos, was quoted in the same story as saying, “Some polls are being used

  as part of a commando operation for one ideological point of view.”51

  CHAPTER 3: PRoPAgAnDA TECHnIQUE: An AnALySIS 125

  BV-Propaganda-Interior-04.indd 125

  9/5/13 4:52 PM

  Another important rule is to note carefully who is paying for a pol . Canadian public opinion polls about the Kyoto Protocol in 2002 seem to show very different results

  when measured by the Liberal Party and the Government of Alberta, an oil producing

  province. When the latter poll was publicized, the Globe and Mail headlined on the top of its front page in very large type “Support for Kyoto plunges” (November 2, 2002).

  In fact, the story produced no evidence of a shift of opinion. It was simply reporting on

  the results of two different polls, the first of which asked Canadians to choose between

  supporting or opposing ratification. In this pol , conducted by Ekos for the Liberals, 79

  per cent supported ratification of the Kyoto Protocol. The second pol , conducted by

  Ipsos-Reid for the Alberta government, asked a very different question, namely, whether

  respondents supported the Kyoto Protocol or a “made-in-Canada” approach, referring to

  a plan for reducing greenhouse gas emissions by a smaller amount over a longer period.

  As long as the specifications and implications of the second option are not spelled out,

  the patriotic sounding second alternative could sway respondents, who might still, if

  given the alternative between ratifying or not ratifying Kyoto, choose the former.

  Everyone can wish for a slightly improved plan, but political choices rarely give

  everyone exactly what they want. Sometimes the real choice is simply of this kind: do

  you support (Kyoto, or whatever) or don’t you? It is a clever public relations technique

  to present idealized alternatives to the public and then use responses to these alterna-

  tives to sway opinion regarding the real alternatives. A spokesman for the Ipsos-Reid

  poll was reported as saying there was no doubt that support for Kyoto was falling.

  That could well be true as a self-fulfilling prophesy resulting from the widespread

  attention to the second poll results and the Globe and Mail’s reporting of the two

  polls. But it hardly establishes that there had been a change in thinking about whether

  to ratify or not, given these as the only choices.

  5. Lying Respondents. Wheeler notes that a Harris memorandum prepared for inter-

  viewers stated, “It’s been brought to our attention that almost all of our surveys are

  showing the population to be more educated than what the census says it actually

  is ... we feel respondents are exaggerating the amount of schooling they’ve had.”52

  Obviously, if respondents lie, polls cannot be all that reliable.

  6. Dishonesty in Gathering the Information. Some pol -taking involves simply filling in with pencil marks a few pages of coded blank spaces. Pol -takers are also often paid by

  the number of questionnaires filled out. It is easy to cheat by filling in sheets without

  doing interviews, although checkups might make this a risky process. Still, checkups

  are not all that strong a guarantee of integrity in practice. Michael Wheeler gives an

  example:

  In 1968 the New York Times commissioned Gal up to do an intensive survey of atti-

  tudes of Harlem residents. The information was collected, tabulated, and submitted

  126 PROPAGANDA AND THE ETHICS OF PERSUASION

  BV-Propaganda-Interior-04.indd 126

  9/5/13 4:52 PM

  to the Times for publication. An editor was so pleased with the poll that he decided to play it up by sending a reporter and a photographer to get a story about some of

  those who had supposedly been interviewed. At seven of the twenty-three addresses

  Gallup had given them, the newsmen could not even find a dwelling! Moreover,

  five other people who had allegedly been polled could not be traced—the addresses

  existed but apparently the people did not. Not even all the remaining interviews were

  legitimate. In one case the Times reporter learned that the interviewer had talked to four people playing cards and incorporated all their answers into one interview.53

  7. Biased or Incompetent Interpretation of Responses. There are many ways in which

  failure to take account of polling theory and of limitations built into any sampling

  procedure can lead to misinformed conclusions based on polls. There may be a deliber-
r />   ate attempt to deceive, but there may also be errors stemming not from malevolence

  but from wishful thinking or sheer ignorance.

  Katimavik, a youth group in Canada, received public funding under the Liberal

  Trudeau government, but had its funds cut off after the Mulroney Progressive

  Conservatives came to power in 1984. One of the reasons the group got bad press was

  their supposed use of illegal drugs. However, as a letter-writer to the Ottawa Citizen

  pointed out, the interpretation of a poll concerning drug use was very biased:

  One of the questions we had to answer was: “Are you aware of anyone having taken

  drugs during the period as participants in the program?” It is from answers to this

  question that the report concluded that “55 per cent reported taking drugs.”

  The real conclusion of this question, assuming the result is reliable, should have

  been “Fifty-five per cent of the participants are aware of the fact that at least one per-

  son has taken drugs.” When our group responded to this question, we all answered

  Yes as there was, in fact, a participant who did take drugs. He was kicked out in the

  third week of the program.54

  Sometimes a newspaper favouring a certain policy position will give prominence

  to a poll supporting that position, ignoring both methodological weaknesses of that

  poll and other polls supporting a contrary view. The Globe and Mail, for example,

  gave front-page coverage to a poll commissioned by the Canadian Abortion Rights

  Action League (CARAL). The story sought to interpret the results of the poll by

  interviewing Norma Scarborough, president of CARAL, the very group that commis-

  sioned it. It would have been fairer to get opinions from pol ing experts or from those

  not sharing Scarborough’s preconceptions to supplement her views. When I wrote

  to the same newspaper to complain that it appeared to have ignored a 1983 Gallup

  poll indicating that 72 per cent of Canadians are against abortion on demand, I was

  not contradicted.55

  CHAPTER 3: PRoPAgAnDA TECHnIQUE: An AnALySIS 127

  BV-Propaganda-Interior-04.indd 127

  9/5/13 4:52 PM

  Wheeler also reveals the large part that discretion can play in tabulating answers to questions. Suppose the question asks whether there is an energy shortage immediately

  or looming in the future, and the respondent replies, “Well, I think the gas crisis was

  manufactured by the companies.” When the answer is tabulated in the “no energy short-

  age” category, it may be misleading, because recognition that an immediate gas crisis is

  manufactured is not the same as saying that there are no longer term energy shortages.56

  8. Fluctuation of Opinion. It is a common observation that public opinion can be

  greatly affected by events that have a vivid effect on the public imagination. A par-

  ticularly gruesome child murder turned public attitudes against the sex industry

  flourishing on Yonge Street in Toronto in the 1970s. News items, editorial opinion,

  letters, and columns can have a powerful effect as well.57 Pollsters themselves empha-

  size that they are giving only a “snapshot” of opinion at a given time. Politicians who

  rely on favourable polls as a basis for cal ing an election may find out that voters’ opin-

  ions change, particularly if they feel that the election was called earlier than necessary

  merely in order to ride a tide of favourable opinion. This resentment can be aroused,

  fed, and exploited by opposition parties, as former Ontario Premier David Peterson

  found when he lost the election he called in 1990.58

  9. Deliberate Attempts to Manipulate Polls. Since opinion has been shown to be

  affected by prominent events, the idea of manufacturing such events has occurred

  to power-holders and power-seekers. Michael Wheeler describes Charles Colson’s

  successful efforts to improve President Richard Nixon’s standings in the polls, one of

  which included giving a prominent pollster a lucrative contract, whereupon negative

  polls gave way to positive ones from that pollster.

  One direct tactic involved rigging a special newspaper poll on the Vietnam War

  by buying up thousands of papers and flooding the editors with questionnaires filled

  out in favour of Nixon’s policies. Wheeler also reports that during the 1968 campaign,

  Nixon had a source within the Gallup organization who provided advance word on

  when the surveys were going to be taken. “This allowed Nixon to time his activities so

  that they would have the maximum impact on Gal up’s polls.”59

  10. Bogus (Unscientific) Polls. It is possible to produce all manner of unscientific polls, sometimes for amusement. For example, “hamburger polls” judge the popularity of a

  politician, political issue, or some other topic by the number of different kinds of ham-

  burgers purchased. A hamburger will be given a name of one politician, for example,

  while other hamburgers will be named for rivals. They may be essentially the same

  hamburgers with different names. The names under which hamburgers are ordered

  constitutes a “vote” for the politician, and the votes are tabulated to produce a mea-

  sure of the popularity of the different rivals. Unscientific polls can also be carried out

  and publicized with a serious intent to influence public opinion.

  128 PROPAGANDA AND THE ETHICS OF PERSUASION

  BV-Propaganda-Interior-04.indd 128

  9/5/13 4:52 PM

  Conrad Black describes his own publication of a makeshift poll prior to the

  October 1973 Quebec election. His paper, L’Avenir de Sept-Iles, supported the Liberal Party against the PQ. As Black wrote in his autobiography, A Life in Progress:

  Our reporting was fairly balanced for most of the campaign, but we did an editorial

  sandbag on the PQ, complete with publication of a poll indicating a Liberal victory.

  There was no indication of the number of people sampled so the fact that I consulted

  only seven people (including myself ) never came to light.60

  Statistics

  Since 1954, Darrell Huff ’s How to Lie with Statistics 61 has provided guidance to some of the basic forms of deception through the use of statistics. His book, recently

  reissued, is still useful for spotting such techniques today. What follows is a selection

  of his observations; his hypothetical examples are supplemented with a few real-life

  examples. Huff identifies four ways in which statistics can be used deceptively.

  1. The Well-Chosen Average. When we use the word “average,” we can mean different

  things. Perhaps the most common meaning is specifically referred to as the “mean,”

  defined in the following way. Suppose we are thinking of “average” income for workers

  in a plant. We divide the total amount of salaries by the total amount of workers, giv-

  ing us the figure for what one “average” worker makes—the “mean” salary. In practice,

  of course, things are not that simple. Do you include part-time workers and workers

  on disability pay, sick leave, etc.? There are many ways of fiddling figures.

  Huff ’s point is that there are two other thoughts to keep in mind about the

  notion of “average.” One is the idea of “median,” defined as the midpoint in a range of

  things we are considering. In the case of workers’ salaries, we would ask: what is the

  salary of that worker who stands at the midpoint when workers are ranged by
salaries

  from highest to lowest? There is also the notion of “mode,” which in this case refers

  to the income level at which more income earners are grouped than any other level.

  Suppose we draw boxes to represent workers earning up to $10,000, between $10,000

  and $20,000, between $20,000 and $30,000, and so on. The box or category in which

  the greatest number of workers appears would be the “mode.”

  Where there is so-called normal distribution, the mean, median, and mode fall in

  the same place on a graph. It so happens that the height of human beings follows “nor-

  mal” distribution, meaning that we have the greatest number of people in the middle

  of a bell-shaped curve that tapers off gradually at two ends of the graph representing

  the tall and the short. But not every curve is bell-shaped, and it may well happen that

  representation of “average” will vary significantly according to whether we present the

  mean, the median, or the mode. Suppose one person in a company earns $50,000,000

  a year through salary and stock benefits, while the other 99 workers get $50,505.05

  each. The “average” salary could then be represented as $55 million (rounded) divided

  CHAPTER 3: PRoPAgAnDA TECHnIQUE: An AnALySIS 129

  BV-Propaganda-Interior-04.indd 129

  9/5/13 4:52 PM

  by 100, or $550,000. That would conceal the fact that all but one earned a little over $50,000. A false picture is given by using the mean, whereas the median gives a more

  representative “average.” If a corporation is talking to shareholders, the median figure

  might be the one to emphasize, while in labour negotiations management would pre-

  fer the mean figure to show how well-off workers are. Of course, the figures chosen

  here make it obvious how deception might occur, and no one would likely be fooled.

  However, with more complex sets of figures, a similar bias might not be so easily

  detectable.

  When mainstream media accept as news official interpretations of figures without

  critically assessing their truth, the public may be influenced without the media actually

  endorsing the official claims. So when the Bush administration claimed in 2001 that

  proposed tax cuts would benefit 92 mil ion Americans, letting them keep an average