We’ve been using the terms polls and surveys interchangeably, and that reflects the ambiguity about which term to use. The term “poll” generally refers to a measure of public opinion, while the term “survey” has a more scholarly connotation. For our purposes, we will use the term survey to encompass all types. To be clear, we will use the following definition:
Surveys measure a number of variables for a sample of a population at one point in time
Each of the italicized terms has a specific meaning in this definition:
Variable: something that can be measured and which varies. Opinions, attitudes, intentions, demographic characteristics and similar items constitute the variables in a survey.
Sample: a small group selected to represent the larger group that the survey researcher wants to learn about. Unless the researchers are going to conduct a census (talk to every member of the group in question), they must draw a sample.
Population: the broad group of people with something in common that the researchers want to learn something about. For example, a reporter investigating the popularity of Facebook may define the population for her survey as 18-24 year-olds who have an account on Facebook. For a study to inform the creators of an advertising campaign for Toyota, researchers may want to look at the population of everyone who purchased a new Toyota in the last year in the state of Colorado. In other words, the population can be defined in any way that makes sense given who the researchers want to study for a particular purpose.
One point in time: the survey’s “snapshot” of one particular moment. Because surveys measure things that vary, the results of any one survey are good only for that one point in time. We could ask you your opinion about a controversial issue today and you might tell us how you truly feel about it. But something could happen tomorrow in your personal world or the world at large that could utterly change your opinion about that same topic. For this reason, surveys cannot be used to predict anything – it is scientifically invalid for someone to use the results of any poll or survey to predict the outcome of any activity whatsoever. Despite the fact that you will run across reports all the time that appear to be predicting the outcome of an election or the likelihood of one thing or another based on survey results, there is absolutely no validity in those predictions.