Likert scale: odd or even?

Frequently asked questions about PLS path modeling.
Post Reply
PLS User
Posts: 17
Joined: Wed Apr 24, 2013 7:38 am
Real name and title:

Likert scale: odd or even?

Post by Julen » Wed Sep 18, 2013 3:54 pm

Dear all,

I wonder if it´s better to design an odd or an even Likert scale to be afterwards used in smartPLS.

With even (for example from 1 to 6) likert scales you do not allow the respondants to place themselves in the middle-point, so I guess it might be better in terms of getting more differences between the unswers...

It seems even more usefull if a multigroup analysis is to be conducted...where the sample would be divided into two...

Am I right?

Thank you all.

User avatar
PLS Junior User
Posts: 1
Joined: Sun Oct 24, 2010 11:03 pm
Real name and title:

Post by mgabriel » Mon Dec 09, 2013 1:32 pm


I suggest you to read the seminal article from Rensis Likert: A technique for the measurement of attitudes, published in 1932.

His examples are based on odd scales and I'm a strong supporter of odd scales but you need to build your sentences using extremes (always/never) so the median point will reflect the majority of answers.

Another suggestion (this one came from Dr. Joe Hair) is to use an odd scale ranging from 0 to 10 (eleven points), so the respondent will feel more comfortable to answer since 0 means "not at all" and 10 is something like "100% agree".

Another point you must consider: usually a Likert-type scale is an ordinal scale, and there's no sense to calculate mean and/or standard deviation.

Hope it helps.



PLS User
Posts: 20
Joined: Wed May 08, 2013 1:13 pm
Real name and title:

Post by huracan » Thu Dec 12, 2013 3:50 pm


Yes, it is a good way to follow Gabriel's suggested options.

Frankly speaking, i am not opposed to apply using odd scale such as five or seven point scale in SmartPls.

However, i still have a suspicious point on this issue until now.

It is still an on-going issue among researchers; In my personal view, we can not find the best determined answer yet whether to use odd or even in SmartPls since two types of Likert scale have advantage and disadvantage each other as many scholars addressed.

What i can consider choosing odd or even is that we need to design survey items based upon 'Who is target respondents? What to know in your research? and Can respondents understand questions exactly as you expected?'

I mean, it is more important to think and design in the position of target respondents about whether to use odd or even option because it is the respondents that respond to your survey.

Please let me know if anyone has other opinions on this ongoing issue.



Post Reply