Forecasting election results accurately can be incredibly hard. But it doesn’t stop campaign managers and political analysts from conducting and analyzing election surveys to gauge election outcomes.

The primary goal of election surveys and polls is to predict which candidate or political party will win, but there are other benefits to it.

Pre-election polls reveal public sentiment toward candidates and political parties.  This helps campaign managers in developing effective campaign strategies that will help the candidate or political party gain public support.

In this article, we’ll be discussing what election polls and surveys are, and how to prevent common errors to accurately predict elections.

What Is a Political Poll?

Although most people use the terms interchangeably, political polls and surveys are not the same. A political poll solicits people’s opinions by asking a single question in a simple, clear, and concise manner, without collecting demographic information or other details to categorize participants into groups.

The main reasons for conducting political polls are to collect voter opinions, measure voter engagement with candidates, and predict election winners. Political polls also help campaign managers understand voters’ concerns and tailor their campaign strategy to position their candidate favorably with voters.

Political polls are typically created by campaign managers for either the political party or the candidates. They are designed for a specific purpose, such as collecting public opinion, identifying the more likable candidate, or forecasting the election outcome.

Campaign managers also use poll results to develop a campaign strategy that resonates with their candidate’s constituency.

Here’s what a political poll looks like:

Who will you vote for in the upcoming Mayor election?

• Brad Greene
• Eva Blanc
• Don’t Know 
• Not Voting

What is a Political Survey?

Unlike political polls, political surveys ask multiple questions that allow you to understand the respondents’ perspectives. It also allows you to review the survey results by demographic groups.

Political surveys have a significant advantage over political polls because, while polls can tell you who people want to vote for, political surveys can tell you why they want to vote for that person.

Here are some common political survey questions:

Are you voting in the next Mayor elections?

• Yes
• No
• Unsure

If yes, who are you voting for?

• Eva Blanc
• Brad Greene
• Other

If not, why are not voting?

• I don’t like the candidates
• I’m not a registered voter
• I’m not yet legal to vote
• Other (kindly specify)

Why are you not voting for the other candidate?

• Personality
• Political party
• Experience
• Other (Kindly specify)

Who do you expect to win the upcoming elections?

– Brad Greene

– Eva Blanc

– I’m not sure

Conducting Surveys Before an Election

Pre-election polls enable political parties and campaign managers to gather voter feedback on candidates’ personalities, ideologies, and likeability. Voters also gain a better understanding of the agendas, profiles, and plans of electoral candidates during political surveys, which influences their voting decisions.

Previously, most political polls were conducted over the phone to ensure the integrity of the results, but phone surveys are becoming way too expensive and time-consuming. 

According to a Harvard Business Review report, a pollster must have at least 800 votes to forecast elections with reasonable accuracy. Obtaining 800 phone responses is a daunting task, especially in the U.S where phone calls can’t be automated.

The more viable option is online surveys that allow instantaneous data collection from respondents. Online surveys are more flexible, less expensive, and faster than telephone surveys, but of course, they have some drawbacks.

However, by adhering to survey guidelines and best practices, you can prevent respondents from giving biased responses or dishonest answers.

Watch Out for Election Survey Sampling Bias

The most common mistake in election surveys is sampling bias. So, pollsters adopt a standard margin of error is +/- 3%, with a 95% accuracy.

In election surveys, sampling bias can lead to inaccurate forecasting and ineffective campaign strategies amongst other problems. It’s a nightmare that every pollster tries to avoid, especially in election polls where the stakes are so high.

Here are some major situations in which sampling bias was a significant issue in US election predictions and why:

Examples of Election Sampling Bias

• Donald Trump vs. Hillary Clinton, 2016

While there are many speculations as to why election polls were off in this election, the sampling bias theory is a reasonable one that is backed by evidence. 

The majority of Democratic-leaning respondents were younger people with a stronger online presence. While the majority of Republican voters were mostly older people who did not participate in the online election survey, so the data gathered did not accurately reflect their opinions.

• Barack Obama vs. Mitt Romney, 2012

The polls were looking pretty good for Republicans in the 2012 elections, especially after Romney performed so well in the presidential debate.

Everyone expected him to win, especially since there were so many favorable poll results. What happened, however, was that Republicans dominated the survey respondents, not that Romney had more support.

Many Republicans confidently supported Romney in the surveys because of his win in the presidential debate, while Democrats were reluctant to answer questions. However, Obama won despite the Democrats’ low response rate in election polls.

• Harry S. Truman vs. Thomas E. Dewey, 1948

The election survey suggested such a wide gap between the candidate that it was ridiculous to believe Dewey would lose. This inaccurate prediction was so notable that the winner, Truman, posed for a photograph with the printed newspaper that predicted his defeat.

The predictors’ results were strong enough that the Chicago Tribune confidently printed the headline “Dewey Defeats Truman” before the final election results were announced.

To obtain the election survey result, there was a nationwide telephone survey asking people about their opinion of the candidates, and who they would vote for. However, because most middle and lower-class families could not afford home phones, the majority of respondents were upper-class.

So, while the survey results heavily favored Dewey, the voter population was not accurately reflected in the survey, which lead to the wrong prediction.

Sampling Techniques for Credible Election Polls

Random-Digit Dialing (RDD)

Although online surveys are becoming more popular, telephone polls remain one of the most personalized ways to collect voter feedback. 

This Random-Digit Dialing technique is also known as probability sampling, it allows researchers to generate phone numbers to contact for polls. It also enables researchers to reach a representative sample of the population who has access to a telephone.

Pros of Random Digit Dialing

• It is a quick and inexpensive way to achieve a significant sample size.
• It’s a fairly common method with plenty of research to back up its effectiveness.
• It enables pollsters who do not have a rich database to reach out to the general population via phone numbers.

Cons of Random Digit Dialing

• It is difficult to reach a specific demographic. RDD provides access to phone numbers but does not include their personality, political affiliations, occupation, or other characteristics that indicate participants’ demographic.
• In countries that don’t have specified dialing codes by location, it’s hard to know the location of respondents.
• Also, there’s no guarantee that all numbers generated will be valid.

Registration-Based Sampling (RBS)

This method selects participants from a pool of registered voters and cross-references their phone numbers to establish the sample size for the survey. It’s an efficient sampling method because it avoids calling numbers that may not belong to registered voters, unlike the RDD sampling method where numbers are randomly generated.

It also boosts the likelihood of the sample size responding to the survey and getting fairly accurate data from them because the participants are inclined to vote.

However, the RBS technique is not bulletproof. Although the sample is drawn from the voter registration list, some people may have moved and changed their phone numbers.

Self-Selected Samples (SSS)

Respondents in this survey method are self-selected volunteers who choose to complete the survey on their own. When using this method, researchers must exercise caution because there is no guarantee that participants accurately represent the population being surveyed.

Dial-in polls and internet polls are the most common applications of this sampling method.

Before sending out a survey to respondents, most online polling and survey platforms filter the respondents to ensure they are representative of the population under study.

Without this filtering method, as is common with most dial-in polls, the poll results may not represent the entire population being surveyed. 

The outcome may be biased because a significant population of people who choose not to participate in the survey may have very different opinions than those who volunteered to participate in the survey.

It’s also difficult to dispute the survey results because there are only responses from people who want to participate. As a result, there is no data to compare the SSS survey results to assess their accuracy.

However, this method optimizes resources because survey participants who have opted in are the ones who are interested in it, and pollsters do not have to spend money on promotions for people who will decline.

Samples from Internet Panels

This is a variation of the self-selected sample method. The sample is selected randomly from people who signed up for part of the survey online. Although the sampling method is random and intended to eliminate bias, this is not always the case.

The sample was drawn from a population that volunteered to be a part of the internet panel, so the respondent pool could still cause selection bias.

Common Errors in Election Survey

Although sampling bias is the most common cause of incorrect election predictions, other errors exist.

Coverage Error

This is mostly largely attributable to geographic or demographic error. For example, certain regions may not be covered in a dial-in survey because the region’s network is spotty, making it difficult to reach people in this region.

It could also be a demographic error, where a particular demographic doesn’t get covered in the survey. Tor example, people in certain professions, do not answer dial-in surveys because they are not available during the hours when the survey is being answered.

Measurement Error

Election surveys aren’t quantitative surveys; they are qualitative. The primary cause of this is poorly structured survey questions for online surveys and inexperienced interviewers for telephone and physical surveys.

Conclusion

Election polls and surveys provide valuable information for forecasting election outcomes accurately, but not always.  Election predictions can be wrong for several reasons, but the most common is sampling bias.

As a result, researchers have developed techniques to avoid sampling bias and ensure that election polls accurately reflect public opinion.