OSR News!!!

News

As I announced a few weeks ago, I am going to use this space to blog about scales issues. After twenty years of reviewing scales, I have a unique vantage point in the field of marketing. In tweet form these will be called Bruner’s Scale-Related Pet Peeves. (To follow my tweets at Twitter, look for marketingscales.) You won’t be able to get the details of this blog by following me at Twitter but you’ll at least get the main point and can then read it all here if you are interested.)

My second topic in this series of pet peeves involves a question that many scale users ask:

HOW MANY POINTS SHOULD BE ON A RESPONSE SCALE?

Let me begin by answering a simpler question. If the question is how many response points are marketing scholars using, there is no doubt that they are over-whelmingly using either 5 or 7 point response formats. Without a lot of tedious analysis I don't know which is more prominent but my hunch is that it is 7. Interestingly, for some reason, Likert-type scales tend to have 5 points and semantic-differentials tend to have 7 points. Of the three decades of literature I have reviewed, hardly any scales have used less than five points. A few have used more, such as 9 and 11. Even numbered response points are extremely rare.

The more difficult question to ask is why a particular number of points is used. It is speculation on my part since I am not aware of anyone studying marketing researchers and asking them why they have chosen to use a certain number of points for their scales. No doubt, part of it is tradition. We have learned about scales from more experienced researchers and they have tended to use either 5 or 7 points. But, it isn't just tradition; there is a logic for the choice of 5 and 7 response points. Those numbers of points (5 and 7) usually allow respondents sufficient flexibility to express themselves. Fewer would overly constrain them. Using more and more points eventually loses meaning since, for example, it is less clear what a 10 on an 11 point scale means compared to a 4 on a 5 point scale. This becomes clearer when you try to provide verbal anchors for the points. A 4 on a Likert-type scale would be labeled ”agree” and 5 would be “strongly agree.” On an 11 point scale you would label the 11 “strongly agree” but, how would you label the 10, the 9 and the 8? While it is true that labels are not necessary, I think the difficulty researchers encounter in labeling them indicates that respondents may not be able to reliably differentiate between them either.

Another way to approach the answer is to ask about the number of items composing the scale. One easy rule is that the fewer the number of items on a scale, the more the number of points it should have (within reason). For example, if you only have 3 items then I would definitely use 7 point scales. If you only have a 1 item scale you could consider something longer like 9 or 11 points. At the other extreme, if you have a lot of items then something shorter is fine, e.g., 5. The point is that there is an relationship between the number of items on a scale and the number of points. A goal is to find a mix that sufficiently detects variation. If our measures are not sensitive enough then relationships may exist but our scales and analyses are not measuring them. Yet, too many items and points can lead to diminishing returns and may, in fact, lead to greater measurement errors. The explanation and discussion can get a lot more complicated but I think you get the point. (See the article referenced below for more details and citations.)

The bottom line is that the number of points on a psychometric scale depends upon how many items are being used and the level of sensitivity needed. Having said that, 5 and 7 are the hands-down favorite for numbers of points used by marketing scholars for their psychometric measures.

BTW, these points I've made are most relevant for measures of attitudes, opinions, intentions, affect, motivations, and personality traits. They may not have much bearing on measures of facts such as demographics (e.g., level of education) and actual behavior (e.g., number of hours online a week). I am not aware of rules for them but I'm sure some could be developed.

Bruner II, Gordon C. and Paul J. Hensel (1993), "Multi-Item Scale Usage in Marketing Journals: 1980 to 1989," Journal of the Academy of Marketing Science, 21 (Fall), 339-344.

.

Comments: Dr. Gordon C. Bruner II

Office of Scale Research