Polls on party or candidate preferences are usually accurate. But what about polls on controversial issues?

Public opinion polls have shown substantial support for a government-run health insurance system, among other things. However, deception influences such results, although it is not the poll-takers who are being deceptive.
 
While polls on candidate and party preferences are usually quite accurate, polls on controversial issues are often highly inaccurate. Fred Barnes, former senior editor of the New Republic, discussed polls on state Equal Rights Amendments:

. . . In 1975, polls showed that a majority favored state ERA's in New York and New Jersey. But 57 percent voted against an amendment in New York and 51 percent voted no in New Jersey. Three years later in Florida, polls showed an equal rights amendment would win voter approval by two-to-one. It lost two-to-one. In 1980 in Iowa, a pre-election poll found that a state ERA was favored by 48 to 23 percent. On election day, it failed by 55 to 45 percent. In Maine in 1984, support for a state ERA was pegged at 62 percent in a poll taken one month before the election. But it turned out that 63 percent voted against the amendment.

And in 1986, polls showed that an ERA in Vermont was all but certain to be ratified by the electorate. It lost. Barnes concluded: "polls about equal rights amendments lie. And they've been lying for years."

That's not all. A New York Times poll published about two weeks before the November 1993 election showed liberal New Jersey Governor James Florio leading his Republican opponent by 15%. Three polls taken during the first half of October showed him ahead by 12 to 14%; two of them gave him over half the vote. Florio lost. Conservative George F. Allen ran a controversial 1993 Virginia governorship race. In May he was down by 29%. Even the final survey taken the weekend before the November election showed him trailing by 19%. He won handily, 58% to 41%. Why?

To understand, we must know how the political situation can affect opinion polls. In Nicaragua, many of the polls taken before the spring 1990 election that ousted the Marxist Sandinistas showed them ahead. No wonder: respondents couldn't be sure that the poll-takers weren't Sandinista agents who'd report them if they responded incorrectly.

Intimidation plays a role in polling even here. Respondents sometimes react to their perceptions about the poll-takers and the viewpoints they appear to represent. The results show not what people think, but what they are willing to admit to inquisitive strangers.

The January 29, 1990 Time cover story was about the National Rifle Association, and included a telephone poll of over 600 gun owners. According to the poll, blacks were about 12% of the U.S. population, but only 6% of the gun owners. Time presented these results with no assessment of their validity. But anyone familiar with this issue could see that for obvious reasons, very few African-Americans would be foolish enough to admit — to an authoritative middle-class voice on the telephone — that they own firearms. Was Time so obtuse that it believed these absurd results — or was it simply unwilling to admit that such polls are unreliable because many respondents have good reasons to lie?

If "only plutocrats or racists oppose Cause X," many citizens will anticipate a disapproving reaction and won't admit that they too oppose it. The more controversial the subject, the more likely such a reaction is perceived to be. A choice between parties or candidates is simply a preference, but a stance on a controversial issue reveals something about the respondent. Eager to avoid the possibility of disapproval, many respondents will tell the pollsters what it appears they want to hear, even if it is the opposite of what they believe.

Some voters deal with this by claiming to be "undecided" — so poll-takers claim that their polls were accurate when taken, but went awry because many "swing" voters changed their minds. However, the more controversial the issue, the more likely voters are to have deep feelings about it, and the less likely they are to "change their minds." The survey taken just prior to the 1993 Virginia governorship race showed the conservative trailing his liberal opponent by 19%, but he won by 17%. Did so many voters "change their minds" at the last minute?

Another classic example was California's Proposition 187 — a 1994 ballot initiative that proposed to bar illegal immigrants from receiving various public services. On September 27, the San Francisco Chronicle reported that Proposition 187 continued to lead in the polls by about two to one. During the period of October 8-11, a Los Angeles Times poll showed 187 still ahead by almost two to one. However, opponents had been labeling it racist. A few days after the Times poll appeared, they began to air radio spots claiming: "white supremacists are behind 187." The Chronicle article reporting this (October 13) admitted that the "link" (to a laboratory that supported genetics research) was "at most tenuous," but the article's headline was, "White Supremacist Link Trips 187."

The poll results were dramatic. A Field Institute poll conducted between October 21 and October 25 showed that Proposition 187's lead had shrunk to 14 points. Another Field Poll conducted during the same time period claimed that 187's lead was down to 12 points (52% to 40%). Two pollsters declared that this decline indicated a "very close" race with the possibility of 187 being defeated. One article cited a poll that had been announced four days before the election, showing 187 dead-even, with 11% undecided — a huge drop from its 20-point lead in mid-October. (This apparent trend, according to the article, revealed "how far 187 has fallen.") Another article on election day stated that 187 was leading "very slightly" in the polls.

Proposition 187 won easily, 59% to 41%. The poll that showed it "tied" four days before the election was a whopping 18% off, which means that the huge "drop" from its 20-point lead in mid-October never actually occurred. Many pro-187 voters had heard the allegations of racism, and were intimidated into telling poll-takers that they would vote against it, fully intending to vote for it.

Political poll results are affected by the sociopolitical climate at the time. The proffered answer to a question can also vary according to how it is asked, who asks it, the general impressions imparted by the media before the poll is taken, and the anticipated response to a given answer. 

In Great Britain, polls have shown a vast majority approving of the National Health Service. But the BBC routinely defends it, presenting interviews with its defenders. They criticize the U.S. system, implying that anyone who supports it must be a nasty bloke who rejoices that the poor lack medical coverage. NHS problems — long waiting lists for major operations and a dire shortage of beds and facilities — are treated as bad-luck flukes, not flaws of the system. In such cases, the only poll that could accurately reflect public opinion is one in which the respondents were a select group of people who had never in their lives read a newspaper, listened to the radio, or watched television.