On 13 September 1998, I request information on the "pros and cons of the
odd-numbered Likert-type scale versus an even-number alternative. The
following are the responses I have received to date, and to all of them, a
BIG thank you. I have for your information opted for an even-numbered
alternative.
Alan Hotte <[log in to unmask]> wrote
"Well, its kind of an old chestnut. I believe I saw a recent evaluation
article which indicated no overall difference in performance. Some folks
like that neutral or centre option and others don't -- is part of the intent
of the question to force folks off the fence... all considerations."
[log in to unmask] <[log in to unmask]> wrote
"I think the issue may be quite simple -- do you want your respondents to be
able to sit on the fence (odd numbered) or indicate either a positive or
negative reaction (even). "
Carol Savill-Smith <[log in to unmask]> wrote
"My only experience is a practical one and, maybe, you have already thought
about it. It is simply if you have an odd number, I have found that people
filling it in often just work their way down the central score, only
deviating one way or another when they feel strongly about something.
However, if you have an even number, they are forced to make a choice one
way or another as they go along - but balanced against this is often that
don't get a "neutral" score if that is what the respondent feels.
The other thing you'll need to think about is how to classify responses
where their choice of answer falls in between 2 of the scales and they tick
or mark it in this way.
No easy answer, but how about an even number, with a little area next to
each question in the scale so people can write something in if they feel
strongly, and a comments box at the end of each section?"
Raphael M. Mutepa <[log in to unmask]> wrote
"Even numbered Likert-type scale forces the respondent to take position with
a certain dose of nuance (agree or strongly agree) whereas odd numbered
Likert-type scale usually allows undecided respondents to adopt a neutral
position by placing themselves in the middle."
Nigel van Zwanenberg <[log in to unmask]> wrote
"Depends a bit on the scale descriptors. If the scale runs from a 'positive'
end such as 'very desirable' to a 'negative' such as 'very undesirable',
then Even forces a choice between the opposite ends. This reduces central
tendency but may annoy respondents who genuinely cannot decide or for whom
the item is irrelevant. On a single scale it probably makes little
difference. Though a separate response category such as 'who gives a damn'
may be helpful as well as/instead of the central position on an odd numbered
scale. in designing biodata forms my colleagues and I have moved more
towards even numbered response categories though these are often not
strictly Likert scales. How about trying piloting the same items with the
odd and even scales and see what the reactions are."
J.Tansey <[log in to unmask]> wrote
"The advantage with an odd numbered Likert scale is that there is then a
middle choice which allows the respondent to express indifference. Otherwise
you force them to take one opinion or another when indifference is a valid
answer. I found that a seven point scale is good but it does depend on the
questions a little."
Dr G.Holt <[log in to unmask]> wrote
"The following reference might be of interest, it deals to some extent with
Likert scales:
Holt, G. D. (1997). Construction research questionnaires and
attitudemeasurement: Relative index or mean? Journal of Construction
Procurement,Vol.. 3, No. 2, pp 88-96. Special edition: Methodological issues
in construction management research. The University of Glamorgan:
International procurement research group."
Clare Tagg <[log in to unmask]> wrote
"Of course even if you have an even likert scale respondents try and defeat
you by either ticking two boxes, ticking the space between boxes or not
ticking anything at all!"
Dr. Peter Stratton <[log in to unmask]> wrote
"One of the main issues is that the mid-point of an odd scale can
incorporate both a genuine midpoint and don't know (or can't be bothered).
If the scale is bipolar so the midpoint is neutral this is especially
troublesome.
Some use a separate 'dont know' box outside the Likert set. But then you
have to decide whether to have substantial missing data, or rescale these
scores as a mid-point anyway.
I am impressed by the evidence that, even if people are sure they do not
know, when forced to guess, there turns out to be useful information in the
guess. So I incline to use an even scale, which forces people to jump to one
side of the fence."
A L Cox <[log in to unmask]> wrote
"As I understand even numbered Likert scales force respondents to choose
between categories such as agree and disagree by allowing them no middle
answer of 'either' or 'neutral'. Oppenheim's book 'Questionnaire Design and
Attitude Measurement' probably describes these in more detail. Generally the
even numbered scales seem more useful for the researcher but in my
experience respondents tend not to like them as we are often asking them to
make difficult choices."
Rosengarten <[log in to unmask]> wrote
"Even is odd! You should always be able to tell whether someone is
indifferent, which is not possible with an even scale.
3, 5 or 7? 5 is best, as 3 looses nuances and 7 confuses."
Dr. Anthony Ingold <[log in to unmask]> wrote
"To some extent I think it depends upon the research situation, but in
general my view is to make respondents make a decision on any questions that
I put to them. Thus I tend to use even numbered scales. When respondents sit
on the fence with a mid point decision, it is not usually very helpful to
the research process.
Some researchers do believe that some respondents genuinely wish to take a
neutral stance. I don't really go with this. Most people will make a
decision when pushed.
--
If there are other perspectives, I would like to hear them, and I will post
a summary of responses towards the end of the year if appropriate. Many
thanks again to all those who have responded.
Regards.
Chris.
--
Christopher Seow
Manchester School of Management
UMIST (UK)
E-mail: [log in to unmask],umist.ac.uk
--
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|