Often I am asked why I use a five point scale in my survey work. After all, a scale with a 'midpoint' makes it easier for respondents to choose the 'neutral' or 'mixed' response, correct? Don't we want to 'force' employees to either be favorable or unfavorable on an issue? The evidence suggests otherwise. Excluding a midpoint has not only been shown to be less preferred by employees, it also has shown to yield very different results when reported as a 'favorable' score, which is the norm for reporting survey results.
At face value, a four point scale is appealing. It has no midpoint, so respondents have to choose a 'favorable' or 'unfavorable' option: For example, consider the following survey item:
My supervisor deals fairly with everyone. Your response options are as follows:In this situation, your opinion will be put into one of two buckets: Favorable (Strongly Agree or Agree) or Unfavorable (Disagree or Strongly Disagree).
Now, consider same item with a midpoint:
My supervisor deals fairly with everyone.
In this situation, your opinion will be put into one of three buckets: Favorable (Strongly Agree or Agree), Mixed (Partly Agree/Partly Disagree) or Unfavorable (Disagree or Strongly Disagree).
Parkes and Langford (2007) showed that, on average, the percent favorable score on a climate survey with a 5-point scale was 58%, which compared to an average of 73% favorable on a four point scale. This 15-point difference has two implications: (1) it makes it is harder to show improvement because there is less variation in the scores, and (2) it overestimates the 'true' favorable had there been a 'mixed' option because with a midpoint, a higher proportion select a favorable response than an unfavorable one to 'give "the benefit of the doubt." In other words, if forced to make an opinion one way or the other, respondents will be more likely to choose the favorable option over the unfavorable option. The study also found that, due to less variation in responses, a four point scale had less predictive power compared to a five point scale as it relates to engagement drivers. Finally, follow-up questions with the respondents revealed the most of those who took the survey on a 4-point scale would have "preferred" to have a midpoint.
Considering that leaders in employee opinion research (e.g., Gallup, Kenexa, Towers Watson) use a five point scale, any normative reports they issue (which are typically reported just as percent favorable), cannot be compared. However, the study did show that there were no differences between 4-point and 5-point scales if results are reported as means. So if you are using a 4-point scale and want to compare it to a 5-point scale, be sure and use means ratjer than "favorable" scores.
In summary, if you are considering a survey, always include a midp-point, as our opinions on many issues at work are not "black and white." Forcing employees to choose a "favorable" or "unfavorable" response can cause bias when reporting results as "percent favorable."
Reference
Parkes, L. P., & Langford P. H. (2007). Including a midpoint option in responses to an organisational climate survey: Does it affect scores, data quality or respondent satisfaction? Proceedings of the 7th Industrial and Organisational Psychology Conference, Adelaide, SA, June 2007.Dave Allen, Ph.D., is the Founder of Allen Analytics, a firm that specializes in employee survey research. click here to contact him.
