Charlotte political scientist Fred Batista offers primer on election polling

During election season, news organizations report an abundance of polling results. Depending on an individual’s news source, polls may indicate favorable data for one candidate while another outlet’s polling information may favor another. How do we know what to trust?

Fred Batista, associate professor of political science and public administration, shared his expertise to explain the numerous questions about polling.

How do polls work?

Polls are surveys designed to measure and quantify public opinion on political preferences, attitudes about issues as well as their views on products. Pollsters draw a sample of the population of interest, devise a strategy to reach the selected respondents (face-to-face, by telephone or via internet, for example) and ask questions using a structured questionnaire.

Why are political polls important for an election?

Polls are important for elections in many ways. First, they improve democracy and representation by informing the candidates about the preferences of the citizens they intend to represent in office. Second, polls inform campaigns about the main issues voters want to see discussed in public debates as well as about which subgroups or the electorate are undecided or willing to change their vote upon receiving new information from the campaigns. Finally, voters can use the information publicized by polls to understand how other voters see the process, and potentially adjust their own preferences based on that.

How do campaigns use polling information?

Campaigns use polling information to understand which voters are more open to persuasion by their messages, which voters they should focus on retaining, and what issues those subgroups of voters view as important.

Do polls actually give an accurate representation of the populace?

High-quality polls rely on accurate data (from the Census and other sources) about the characteristics of the population of voters and attempt to sample a subgroup of voters that represent that general population. However, given not all eligible voters will turn out to vote in the election, it is usually the case that the population of actual voters differs from the population of eligible voters. This poses a major challenge for polls. In general, most polling companies use models that identify the most likely voters based on their profile and adjust their estimates after taking that into account.

Who is chosen to participate in the polls?

All eligible voters are included in what pollsters and statisticians call sampling frames. Sampling frames are the lists comprising all units of the population of interest (voters), along with additional information about their location and demographic characteristics. Once polling companies have that information, they will design different strategies to randomly select individuals from those lists and will make sure that the final sample has similar aggregate characteristics as the population of voters. Given that the sample will inevitably display some degree of random variation in comparison to the population, poll results will also include estimates with their margins of error, precisely to account for the fact one cannot interview the totality of voters in a short period of time.

How do people know if a poll is trustworthy?

In general, high-quality polls tend to have a few common characteristics. First, they tend to be transparent about the main features of their methodology. Notably, due to market pressures, polling companies cannot share all details in their methodologies, but key elements related to sample composition and weighting as well as how they reach and interview respondents are explained in poll reports. Second, more credible polls tend to be transparent about who funds their data collection, when that is not self-funded. Polls are expensive and it is a good practice to inform consumers about the entities that paid to receive that information about the electorate. Third, we can also look at polling companies’ track record in previous elections and how they respond when some of their results diverge from vote tallies, as polling error can happen and may be explained by factors outside the control of pollsters, such as late swings. All in all, transparency of information about the service is key.

Why do polls from different sources offer conflicting data?

Polls often provide conflicting data due to the use of different methodologies. Different polling companies may rely on different lists of voters to create samples and on different weighting methods to generate corrections when their samples diverge from the population of interest. It is also widely known that the mode of collection – face-to-face, telephone, online – may affect the outcomes. For the most part, different techniques are chosen based on considerations of cost, speed of data collection, and often come from specialists having sincere disagreements on specific technical issues. Also, polls can often differ simply because they are conducted days apart from one another, especially in very dynamic campaign environments.