Terms of Poll-i-tical Wisdom

A regular prejudice in one direction; in political polling, bias might result from polling only a certain economic or ethnic group, which could have unrepresentative opinions.

Confidence interval
An interval of results that is 95 percent likely to contain the true number; equals reported proportion plus or minus the margin of error. (For an alternative explanation).

A supposition, or statement, used in science, that is precise enough to be tested.

Margin of error
A range of numbers used to determine the confidence interval, usually expressed as "plus or minus and x %". Note that the margin of error is computed on sample size and is only valid if the poll was well designed and executed on a true random sample. A poll with misleading questions will produce misleading results.

Mean, or average
The sum of a measurement, divided by the number of contributors to that measurement.

The sample size, generally the number of people questioned in a poll.

The probability that convincing results would occur due to chance alone; if p is less than .05 (often written "p < .05") the results are accepted as "statistically significant," or likely to be accurate at least 19 times out of 20.

The larger group being studied. In political polls, the population is usually all eligible voters.

An amount so small it's hardly worth measuring (we think.)

A way to choose a sample that represents all the people we want to reach (in political polling, that's usually likely voters). A sample is random if each member of the population has an equal chance of being represented.

Describes whether a measurement gives approximately the same result in repeated tests.

Response rate
The percentage of the originally identified sample that finally responds to the survey.

The number of people who were questioned.

Sampling error
An error derived from a mistake in sampling procedure.

Standard deviation
A statistical number that gauges how far any one measurement is likely to vary from the mean.

Statistically significant
A result which is at least 95 percent likely to be accurate; a result that would be produced by chance no more than 5 percent of the time.

Systematic error
An error or bias introduced by some methodological error.

Tracking poll
A poll conducted each day to assess trends caused by new advertising, public events, etc. Uses a relatively small sample; generally three to four days' worth of results are rolled together.

Validity gauges whether a statistic measures what it is supposed to measure.

Volunteer error
An error caused by the fact that some people had a chance to come forward with their opinions, and that biased the sampling.