on to include the body count of Vietnam and the numbers of nuclear warheads and intercontinental ballistic missiles cited as the measure of national security.
The computer scientist, the mathematician, the statistician, and the accountant all deal with “reality,” while the poet, the writer, the wordsmith deal with, well, just words. You may find, however, that the world of numbers is not as accurate as you think it is, especially the world of the public opinion poll.
If you believe in public opinion polls, I’ve got a bridge you might like to buy. Depending upon which poll you believed just before the New Hampshire primary in February 1988, you would have known that Robert Dole would beat George Bush 35 percent to 27 (Gallup); or Dole would win 32 percent to Bush’s 28 percent (Boston Globe); or that Dole and Bush were even at 32 percent each (ABC-Washington Post); or Bush would win 32 percent to Dole’s 30 (WBZ-TV); or Bush would win 34 percent to Dole’s 30 percent (CBS-New York Times). Of course, George Bush won the actual vote 38 percent to 29 percent.
Things weren’t much better on the Democratic side, either. While most primary polls were correct in identifying Michael Dukakis as the winner, the margin of victory varied from 47 percent to 38 percent. Dukakis won with 36 percent of the vote. For second place, though, the polls really missed the call. Two had Paul Simon ahead of Richard Gephardt for second place, while a third had the two tied and the others had Simon behind by a thin margin. In the actual vote, Simon finished third with 17 percent of the vote, while Gephardt finished second with 20 percent. No one predicted Gephardt’s 20 percent of the vote, not even the surveys of voters leaving the polling places after they had voted. This last point should not be overlooked, for it reminds us that no poll is worth anything unless people tell the pollster the truth. Since no pollster can ever know whether or not people are telling the truth, how can we ever be sure of any poll?
Things didn’t improve during the presidential campaign either. In August, 1988, before the Republican National Convention, seven polls gave seven different answers to the question of who was ahead. The CBS-New York Times poll had Dukakis leading Bush 50 percent to 33 percent, while a poll taken by KRC Communications/Research had Dukakis ahead only 45 percent to 44 percent. When the ABC News poll came out with Bush ahead 49 percent to 46 percent, many people in the polling business discounted the results. ABC promptly took another poll three days later which showed Dukakis ahead 55 percent to 40 percent. That was more like it, said the other professional poll takers.
Even as presented, such polls are deceptive. Any poll has a margin of error inherent in it, but pollsters don’t discuss that margin very much. They like their polls to have an air of precision and certainty about them. The KRC polls just mentioned had a margin of error of plus or minus 4 percent. This means that, in the first poll KRC took Dukakis really had anywhere from 49 to 41 percent, while Bush had anywhere from 48 to 40 percent. In other words, Dukakis could have been ahead 49 to 40 percent, or Bush could have been ahead 48 to 41 percent. The poll didn’t tell you anything.
Polls have become important commodities to be sold. Television news programs and newspapers use polls to show that they have the inside information, thus boosting their ratings and their circulation. Also, the more dramatic or unexpected the results of a poll, the better the chances the poll will be featured prominently on the evening news program. In addition to all this hype and use of polls as news, politicians, corporations, special- interest groups, and others have vested interests in the results of particular polls. Such people and groups have been known to design and conduct polls that will produce the results they want. In other words, polls can be and are a source of a lot of doublespeak.
How do you read a poll? Actually, it’s not all that hard, but the problem is that most poll results don’t give you enough information to tell whether the poll is worth anything. In order to evaluate the results of a poll, you need to know the wording of the question or questions asked by the poll taker, when the poll was taken, how many people responded, how the poll was conducted, who was polled, how many people were polled, and how they were selected. That’s a lot of information, and rarely does a poll ever give you more than just the results.
In 1967, two members of Congress asked their constituents the following question: “Do you approve of the recent decision to extend bombing raids in North Vietnam aimed at the strategic supply depots around Hanoi and Haiphong?” Sixty-five percent said yes. When asked, “Do you believe the U.S. should bomb Hanoi and Haiphong?” however, only 14 percent said yes. In 1973, when Congress was considering articles of impeachment against President Nixon, a Gallup poll asked the question, “Do you think President Nixon should be impeached and compelled to leave the Presidency, or not?” Only 30 percent said yes to this question. They were then asked, “Do you think the President should be tried and removed from office if found guilty?” To this, 57 percent said yes.
The most popular form of polling these days is the telephone poll, where a few hundred people are called on the telephone and asked a couple of questions. The results are then broadcast the next day. The two ABC polls mentioned earlier were based on telephoning 384 and 382 people, respectively. Just remember that the U.S. population is over 245 million.
According to Dennis Haack, president of Statistical Consultants, a statistical research company in Lexington, Kentucky,
most national surveys are not very accurate measures of public opinion. Opinion polls are no more accurate than indicated by their inability to predict Reagan’s landslide in 1980 or Truman’s win in 1948. The polls were wrong then and they have been wrong many other times when they tried to measure public opinion. The difference is that with elections we find out for sure if the polls were wrong; but for nonelection opinion polls there is no day of reckoning. We never know for sure how well surveys measure opinion when elections are not involved. I don’t have much confidence in nonelection opinion surveys.
The Doublespeak of Graphs
Just as polls seem to present concrete, specific evidence, so do graphs and charts present information visually in a way that appears unambiguous and dramatically clear. But, just as polls leave a lot of necessary information out, so can graphs and charts, resulting in doublespeak. You have to ask a lot of questions if you really want to understand a graph or chart.
In 1981 President Reagan went on television to argue that citizens would be paying a lot more in taxes under a Democratic bill than under his bill. To prove his point, he used a chart that appeared to show a dramatic and very big difference between the results of each bill (see Figure 1). But the president’s chart was doublespeak, because it was deliberately designed to be misleading. Pointing to his chart, President Reagan said, “This red space between the two lines is the tax money that will remain in your pockets if our bill passes, and it’s the amount that will leave your pockets if their bill is passed. On the one hand, you see a genuine and lasting commitment to the future of working Americans. On the other, just another empty promise.” That was a pretty dramatic statement, considering that the maximum difference between the two bills, after five years, would have been $217.
Figure 1
President Reagan’s misleading and biased chart, compared with a neutral presentation regarding the same tax proposals.
The president’s chart showed a deceptively dramatic difference because his chart had no figures on the dollar scale and no numbers for years except 1982 and 1986. The difference in tax payments was exaggerated in the president’s chart by “squashing” or tightening the time scale as much as possible, while stretching the dollar scale, starting with an oddly unrounded $2,150 and winding up at $2,400. Thus, the chart had no perspective. Using the proper method for constructing a chart would have meant starting at $0 and going up to the first
