Understanding Political Polls

The two tables above summarize three recent polls conducted in South Carolina. (CLICK ON THE TABLE TO MAKE IT LARGER)

The top table shows a "point estimate" of the percentage of South Carolina voters planning to vote for each candidate. Is this the true percentage? Maybe but probably not. The truth may be less than or greater than the point estimate.

To get a better estimate of the true percentage, the poll includes a margin of error. By applying the margin of error to the point estimate, we strengthen the estimate by adding a lower and upper bound to the truth. The second table applies each poll's margin of error to each candidate's point estimate. For example, in the Clemson poll, the estimated percentage of voters planning to cast a ballot for John McCain was 29%. Applying the margin of error (4.6%), we can now say the estimated percentage of voters planning to cast a ballot for John McCain is at least 24.4% and at most 33.6%.

Don't just compare the point estimates of two candidates. Instead, compare the upper and lower bounds. If the ranges (of two candidates) do not overlap, there is statistically significant separation between the two candidates.

The margin of error is based on two main factors: Sample size and confidence level. Obviously, the larger the sample size, the better the estimate. A larger sample size will decrease the margin of error. The confidence level is a measure of strength for the estimate. Most results of political polls are stated with 95% confidence. This means there is a 95% probability that the upper and lower boundaries contain the true percentage. There is a 5% probability that the true percentage lies outside the range.

I hope this helps.