Yes my friends we have entered the political silly season. With a month til the US presidential elections more money will be spent on advertising and on public opinion polls than any other time of the four-year political cycle. It is a good time to be in advertising or research. It is a sure bet that most, if not all, of the surveys will have their results promoted as a means of subtly swaying public opinion.

If you look or listen closely these reports all include a simple measure, generally reported at the end of the report. It is common practice to report the margin of error for surveys where the respondents were selected at random. This allows the user of the information to decide for themselves the validity of the data being presented.

There are two components to this measure that are useful for survey researchers and those who make decisions based upon the data we provide. First is the margin of error, also known as the confidence interval. This is the “+/-“ component as in the survey results are valid within plus or minus four points. For example if we use the four-point interval and 55% respondents select a given candidate then we can be assured that if we surveyed the entire relevant population then the value selecting the candidate would range between 51% and 59%.

The second component is the confidence level. This aspect gives us a sense of how confident we can be in our estimation. It is expressed as a percentage (typically, 90%, 95% or 99%). This tells us that if we sampled the relevant population multiple times how certain we can be that the estimate would lie within the confidence interval. If our sample was large enough we might say that 95% of the time we are certain that the estimate falls within the 51% to 59% range.

These values can be used to determine the necessary sample size needed to reach a desired level of confidence. They are not linearly related so doubling the sample size may not have the same level of effect on your confidence. However, all things being equal a larger sample (assuming it is random) will generate a greater sense of confidence in your survey results.

A good tool to bookmark is the sample size calculator from Creative Research Systems.