How to Run a Poll: Avoiding Sample Bias

Avoiding Sample BiasSample Bias Danger Zones

Online polling:
Sample bias is more likely in online polls and surveys posted on websites. There is no way to control for bias of any type in online polls.

This “passive” polling is there any time you’d like to take the poll, and is more for entertainment than scientific study.

Email polling:
Sample bias would also be more likely in polls sent to specific email addresses.

Although the pollster could select a particular group of those who provided their email addresses, the respondent can forward the email to others not accounted for demographically. This makes a truly random poll impossible.

Have questions about polls or surveys? Contact us today 1.877.655.7837

Traditional polling:
Traditional or “live agent” polling can also have significant sample bias created by the inflection in the agent’s voice, the agent’s accent, changes in the way a question is worded or where the agent’s voice happens to pause.

Example: a Caucasian agent who has negative feelings/beliefs about Hispanics and hears a heavy Hispanic accent on the other end of the line, may ask the questions in a vocally cold, flat, even hostile way. The poll results could then be skewed depending on the number of Hispanics this agent polled.

A random sample could then be biased by these vocal or vocabulary changes to inaccurately reflect the opinions of Hispanics. The poll may reflect the views of non-Hispanics without this factor ever being identified.

Solving the Sample Bias Dilemma
This is one area where Pollster system really shines. You can deliver the questions the same way, every single time, eliminating the problems of traditional “live agent” polling.

You can also gather specific demographic information for each poll completion, an area where fickle email and internet responses fall short.