 The consequences of biased polls are well recorded by history. The first famous example of a clearly flawed public opinion poll was taken before the election of 1936 by Literary Digest. The magazine queried a database of its readers and of registered telephone company customers and automobile owners. The poll showed incumbent Franklin D. Roosevelt losing by a sound margin to the Republican Governor of Kansas, Alf Landon. History, of course, proved otherwise. Two things contributed to the enormous error of this result. First, significant difference existed in the organization of the electorate in 1936 compared to previous years. It marked the beginning of a trend that continues today — the electorate became fairly polarized along racial and socioeconomic lines, with the vast majority of blacks, religious and ethnic minorities, and the economic underclass shifting to the Democratic Party. This is the origin of FDR’s “New Deal Coalition.” Second, the poll was heavily biased towards well-off, since the height of the Great Depression was no time for the vast majority of the population to enjoy automobiles or telephones. However, the weighting towards the wealthy would not have mattered had it not been for the (since then indelible) socioeconomic and racial polarization of the electorate. The Literary Digest poll had not failed in earlier elections, when the electorate had not been so polarized.
Another unfortunate error was made in the election of 1948, when pollsters forecast that Thomas Dewey would win in a landslide victory over incumbent Harry S. Truman. However, the now-famous picture of Truman triumphantly holding a large newspaper, with the headline “Dewey defeats Truman,” stands as a second major testament to the fallibility of incorrectly run opinion polls. The error here was even more avoidable; pollsters merely stopped taking polls too early and failed to detect the success of Truman’s whistle-stop campaign tour in the closing days before the election which drew many Democrats back to the party banner.
 If anything can be said about the polling performed before the 2000 election, it certainly did not end too early. An editorial in the 8/21/2000 issue of the Omaha World-Herald pointed out the problematic nature of modern polling, which rushes to constantly reflect the day-to-day changes in public opinion by churning out fast-paced polls with ever-smaller sample sizes to “provide the most current information available.” This is, of course, the same polling procedure that attributed a much wider margin of victory to President Clinton in 1996 than he actually had, and which provided both presidential candidates in 2000 with almost no relevant information (except for the obvious leads each held in specific demographic groups).
Moreover, in the modern day and age, there is a new inherent bias in the polls. Just as there was a time when people outside white suburbia did not have access to telephones, which had made white suburbanites the logical target group for easy polling, there exists another situation today in which white suburbanites tend to be the most unresponsive to telephone polling. In fact, statistics show that the people who refuse to be polled on the phone are typically of a higher income bracket than those that do not. Regardless of the specific demographics of those who respond versus those who do not, it is foolish to think the two groups think in the same way. The margins of error reported on most conventional polls, therefore, can scarcely be trusted for anything.
Additionally, there is the further problem of which demographic groups are being polled. Often, it’s been shown that low-income groups and groups that are numerically weak but politically empowered are under-polled. Such groups basically include all ethnic groups other than whites. Some of these groups are tiny minorities across the whole of American society but are becoming more prominent in the traditional power centers such as the socioeconomic upper middle class. They do not share the political opinions of the standard suburbanite, and should be polled independently to reflect the nature of the electorate.
Low-income groups that are regarded as being unlikely voters are also under-polled. It was demonstrated in this election that “unlikely voters” may vote in great numbers. They were drawn in this year by a great “Get Out The Vote” effort, like the one mounted by the African-American community on behalf of Vice-President Gore. This made the election amazingly tight in many areas, and indeed, tipped the national popular vote sharply in favor of Gore. However, this was not reflected in the polls, because the poorer African-American community was thought to be comprised of “unlikely voters.”
Polls have been designed to have “disenfranchisement” factored in. In this day and age when a fair amount of public policy may or may not be influenced by the results of public opinion polls, though, it only makes sense to have accurate polls. Elected officials should consult their constituents on important matters, but they should not take actions based solely on public opinion polls, much less on ones that possess so many inherent biases as to misrepresent the mood of the public.
|
 |
