BACK TO TABLE OF CONTENTS

THE UNRELIABILITY OF ELECTION POLLS

It is becoming apparent both in Canada and abroad that polling companies are failing to come up with accurate data on voters' intentions. Most Canadians became aware of this during the June 2004 federal election, when the pollsters repeatedly told us that the election would be very tight, with the conservatives and liberals running neck and neck. This, of course, was not the actual situation, since the Liberals returned with a minority government of 135 seats, with the Conservatives holding 99 seats.

Similarly, in the national election in Australia held in early October, the pollsters predicted a tight race between the Conservative Prime Minister, John Howard, and the socialist leader, Mark Latham. Again, this was highly inaccurate as Mr. Howard was returned with a majority government, with an increased number of seats.

In the November 2, 2004 U.S. Presidential race, each polling company came up with different findings throughout the campaign. There was little consistency in their polling results.

Of course, such varying results hinge on the questions asked, of whom the questions are asked, the margins of error, and the way the data are interpreted.

However, in the last few years, another problem has arisen in regard to polling. Usually polling companies conduct their polls by the random selection of respondents based on national telephone surveys. However, changes in communication technology have thrown a monkey wrench into this process. The telephone has become an imprecise instrument for random probability samples. This is because, at one time, everyone had land telephone lines at their homes or offices. These landlines made it fairly easy to reach the respondents. But now, with call waiting, call display and voice mail, polling by telephone has become much more difficult. Also, more and more individuals no longer bother with landline telephones at all, but rely solely on their cell phones for communication. However, cell phones make many potential respondents unreachable. Even if the pollsters could reach a cell phone owner (U.S. law, for example, forbids polling companies from calling without consent because the respondent pays for the airtime), since cell phones are portable, pollsters never know if a phone's area code actually reflects where the owner resides. As the number of cell phone users grows, problems will increase for the pollsters. Polling companies can make some adjustments to data to take into consideration cell phone users who tend to be younger and more mobile, but this is not a very exact process. Consequently, we should put less and less faith in what the pollsters tell us - about elections or other matters.

BACK TO TABLE OF CONTENTS