It’s not the size of your poll, it’s the relevance of your election data (Part 2 on Election Bias)

On Tuesday night, millions of Americans were surprised that Donald Trump did not win the Iowa Caucus. Based on the polls leading up to the caucus, Donald Trump was the clear frontrunner and looked to be headed to a certain victory. But as you analyze historical trends, it becomes clear that election polls are consistently unreliable at predicting the actual results of elections, and many now believe bias is the result. So what can go wrong so consistently during every Iowa Caucus? Fortunately, QuestionPro may be able to provide some insight on why bias occurs and how it could have affected the Iowa caucus. Below are the four main reasons why surveys yield bias.
Response Bias
Response bias results from problems in the measurement process. This usually happens when surveys are poorly constructed, most prominently containing leading or socially desirable questions. Read more about Response Bias and how to overcome it in a recent post.

Researcher Bias
The viewpoint of the researcher can consciously or subconsciously creep in when they create a poll. This can be very subtle, but all researchers have points of view and very often it can affect the purity of the survey. Writing great questions is an art that requires a great amount of work, practice, and help from others. We have identified some of the common pitfalls in creating a great questionnaire.

Coverage Bias
Coverage bias occurs when the population of a survey sample is incomplete. This is extremely prominent in political polling. People in populated areas are much easier to reach than people in rural areas. Another example is that up until the last election cycle, a significant amount of surveys were conducted using landline phones. Individuals and families without a landline phone, which includes cell phone-only users were not elected members of the population. This leads to a major chunk of the population not being represented. This could be very likely, as many post-election polls show a higher than expected turnout of evangelical voters.

If you want to continue learning, we invite you to read our guide: Sampling Bias

Nonresponse Bias
Non-Response is when people are unwilling to participate in surveys. Americans are becoming harder to reach, and traditional survey methods are becoming less and less effective. When was the last time you responded to anything in the mail? Using modern communication methods like email can improve the effectiveness of responses, but it’s also important to be aware of the survey design. Take a look at some tips on survey design to get more responses from your audience.

What Do You Think?
Why does bias in pre-election surveys occur so regularly? Do you agree with us/ disagree, and if so why? What are the other ways that bias occurs? We want to hear from you, share this post in social media and let’s start a conversation.