___________________________________________________________

The final nationwide telephone poll on the presidential election carried out by Harris finds that the two main candidates are so close that it is impossible to predict a winner. We don't know who will win.

The difficulty of predicting a winner is compounded by the fact that slightly different definitions of who will and who will not vote (i.e. the definition of likely voters) produces three different results:

1. Among those who say they are absolutely certain to vote, Gore has a 3-point lead;

2. Among those who are absolutely or quite certain to vote and say that the election result will make a great deal or quite a lot of difference, Bush has a 2-point lead;

3. Using a third possible definition, people who are absolutely certain to vote, and voted in 1996 or were too young to vote then, shows the two candidates tied.

It seems likely, therefore, that the results will depend on turnout. If the turnout falls below 50% as it did 4 years ago, that will be good for Bush. If, however, the Democrats are successful in turning out a higher proportion of some of their most loyal voters, such as African-Americans, then the turnout might rise to 53 to 54 percent of all adults which would be good for Al Gore. The difference between a high and a low turnout could make the difference between a Gore and a Bush victory.

This Harris Poll was conducted by telephone with a nationwide sample of 1,348 likely voters, interviewed on Friday, Saturday and Sunday, the 3^rd, 4^th and 5^th of November.

Has the race narrowed? And, if so, why?

Ten days ago, we had Bush five points ahead. If our numbers are right, Gore has made some gains in the last ten days of the campaign and Bush has slipped a little. It's impossible to quantify or separate the reasons for this change which is, if it is real, very small, but it may be linked to Governor Bush's misstatement that Social Security is not a federal program or the disclosure that he never admitted publicly his conviction for drunk driving many years ago, or both.

The Electoral College

If this election is as close as our data suggest, then this might be the time when the Electoral College vote and the popular vote break in different ways, unlikely as that is. This can only happen when the two candidates are extraordinarily close, which explains why it did not happen in the 20^th Century and has not happened since Grover Cleveland lost to Benjamin Harrison in 1888.

Millions have already voted

One of the most interesting things in this survey is that 8 percent of all likely voters have already voted by mail. These early voters divided 48% for Bush and 49% for Gore, 2% for Nader, and 1% for Buchanan.

Could the polls get this election wrong and, if so, why?

One day there will be another presidential election like 1948, where the polls will "get it wrong." It is bound to happen. It's only a question of "when." It might even be this year. When it happens it could do so for one or more of any of the following reasons:

1. Late Swing.

Enough people change their minds after they were interviewed by the polls to change the result. This is what seems to have happened in 1948.

2. Defining "likely voters" accurately

Trying to determine who will vote is much harder than determining how they will vote, if they do. Most of the (rather different) methods used by different polling firms to identify likely voters are used because they seemed to work more or less well in the past. But, in my experience, what works well in one election may work much less well in the next. Every election is unique.

Mistaken judgements (that is all they are; this is not pure science) on who are likely voters have probably been the biggest source of error for the major polling firms over the years.

 

3. Non-response bias

One of our biggest nightmares is that the people we don't interview, because they are not at home when we call or because they refuse, will vote differently from those whom we survey. To date we can thank our lucky starts that this has apparently not been a problem. One day it almost surely will be.

4. Flawed polling methodology

No two polling firms use identical methods. They draw their samples of telephone numbers differently; they choose the person to be interviewed differently; they ask different questions in different orders; and they weight the data differently. There is no unique "right way," although there are clearly lots of "wrong ways" to do these things. We try to learn from our mistakes but we always need to keep on learning.

TABLE 1

WHO IS AHEAD AND BY HOW MUCH?

HARRIS TELEPHONE AND ONLINE POLLS

Base: Likely Voters

"If the presidential election were held today, who would you most likely vote for (READ LIST)?"

"Well, if you had to say, would you lean toward (READ LIST)?" Even if you have a very slight preference, we would like to know what it is."

"For the presidential election being held on November 7^th, whom did you vote for (READ LIST)?"

NOTE:

These numbers were calculated after allocating "not sure" and refusals.

The final Harris Online Poll numbers will be released before 7:00 AM on Election Day, November 7^th.

* See Table 2 for definition of likely voters.

TABLE 2

WHO IS AHEAD AND BY HOW MUCH DEPENDS ON YOUR DEFINITION OF LIKELY VOTERS

"If the presidential election were held today, who would you most likely vote for (READ LIST)?"

"Well, if you had to say, would you lean toward (READ LIST)?" Even if you have a very slight preference, we would like to know what it is."

"For the presidential election being held on November 7^th, whom did you vote for (READ LIST)?"

Methodology

This Harris Poll was conducted by telephone within the United States between November 3 – November 5, 2000, among a nationwide cross section of 1,348 likely voters. Figures for age, sex, race, education, number of adults and number of voice/telephone lines in the household were weighted where necessary to bring the total population of all adults into line with their actual proportions in the population.

In theory, one can expect that 95% of surveys with samples of this size would produce results that were within plus or minus 3 percentage points of what they would be if the entire adult population had been polled using the same methods. Unfortunately, there are several other possible sources of error in all polls or surveys that are probably more serious than theoretical calculations of sampling error. They include refusals to be interviewed (non-response), question wording and question order, interviewer bias, weighting by demographic control data and screening (e.g., for likely voters). It is difficult or impossible to quantify the errors that may result from these factors.