Taylor runs into the Training manager in the hallway. “Hey, Taylor! Glad I caught you. I’d like to update our training survey. We have a five-point scale with the midpoint ‘Neutral.’ Personally, I want respondents to say one way or the other if they agree or disagree. Let’s talk about this in the staff meeting. Can you bring some guidelines about when to use or not use a midpoint in the Likert scale to the meeting?”
Taylor looks at the clock and realizes it’s less than three hours until the staff meeting. The research hasn’t turned up any decent answers so far…
Have you been in a similar situation where you wondered about the effectiveness of your response scale design and when to include or exclude a midpoint? You aren’t alone.
LET’S TALK ABOUT THE LIKERT SCALE
Rensis Likert, an American social psychologist, is forever associated with the five-point survey response scale he developed 85 years ago. The response scale we commonly see today (Strongly Disagree, Disagree, Neutral, Agree, Strongly Agree) is a close version of Likert’s original scale (Strongly Disapprove, Disapprove, Undecided, Approve, Strongly Approve). When we offer a “Neutral” option, what are we really getting in response?
Regardless of how you label the scale, some respondents may use a midpoint as a “dumping ground.” Common interpretations of the midpoint include:+
- Neutral
- Middle point
- It depends
- Don’t care
- Unsure
- Confused
- Need more
- Other
SHOULD YOU REMOVE IT?
By excluding a midpoint from the Likert scale, the four-point Likert scale becomes a forced-choice scale. This causes different problems. Respondents who would have selected a midpoint now are forced to choose either an agreement or disagreement, which threatens the validity of the data.
A forced-choice Likert scale is an ordinal scale, rank-ordered from Strongly Agree to Agree, to Disagree to Strongly Disagree (or vice versa). You can’t apply certain common statistical analyses to data obtained from ordinal scales.
TIPS FOR MIDPOINT USE
When including a midpoint in the Likert scale, the distance between Strongly Disagree, Disagree, and Neutral may be as equidistant from each other as the numbers 1, 2, and 3. When the midpoint is clearly labeled as Neutral, and respondents are familiar with their topic enough to be able to give a neutral answer, this is good news for potential use of parametric statistics from the survey responses.
Instead of simply removing a midpoint from the Likert scale, survey designers should ask themselves when to omit the midpoint and how to minimize the potential survey bias when excluding a midpoint.
Based on our review of more than two dozen research reports, we recommend the following guidelines (a detailed report can be found in “Evidence- based survey design: The use of a midpoint on the Likert scale,” Chyung, S. Y.; Roberts, K.; Swanson; I.; and Hankinson, A., Performance Improvement Journal, 58(1), pp. 15-23, 2017):
Include a midpoint when:
- You intend to use the data as interval data to generate average scores and parametric statistics.
- Respondents are familiar with the topic.
Exclude a midpoint when you expect that some respondents may choose the midpoint because they:
- Are unfamiliar with the topic or have not yet formed their opinion.
- Are under pressure to please an audience and do not want to select negative answers.
In either case, to prevent the midpoint being used as a dumping ground (or worse):
- Improve the clarity of your survey items.
- Offer options off the scale (i.e., Not Applicable) .
Seung Youn (Yonnie) Chyung, Ed.D., is a professor of the Department of Organizational Performance and Workplace Learning in the College of Engineering at Boise State University. The other three authors all earned degrees at Boise State University. Katherine Roberts, MS, is a senior instructional designer with WellMed.
Andrea Hankinson, MA, MS, is manager of Executive Development at the Public Service Commission, a department of the Government of Alberta, in Edmonton, Alberta, Canada. Ieva Swanson, MS, is a Retail Learning and Development specialist at TD Ameritrade.