The survey results are based on 2,421 telephone interviews conducted under the direction of Princeton Survey Research Associates from a nationwide sample of the voting age population, from Thursday afternoon, Oct. 31, through Sunday morning, Nov. 3, 1996. For results based on the total sample, there is 95% confidence that the error attributed to sampling and other random effects is +/- 2 percentage points. For results based on registered voters 1,875, the sampling error is +/- 3 percentage points, and for results based on likely voters 1,211, the sampling error is +/- 3 points.

The sample for this survey is a random digit sample of telephone numbers selected from telephone exchanges in the continental United States. The exchanges were selected with probabilities proportional to their size, with the first eight digits of the numbers selected to be proportionally stratified by county and by exchange within county. Estimates of the number of telephone households within each county are derived from the 1990 Census Bureau data on residential telephone incidence, updated.

At least four attempts were made to complete an interview at every sampled number, with calls staggered to maximize the chances of reaching a potential respondent. Non-response in telephone interview surveys produces some known biases in survey- derived estimates. To compensate, the sample data are weighted in analysis using parameters derived from a special analysis of the most recently available Census Current Population Survey (March 1994).

A detailed description of the survey methodology will be contained in the full report to be published tomorrow (Nov. 2).