[Rasch] Rasch version to Churchill's Paradigm

Thomas Salzberger thomas.salzberger at gmail.com
Fri Nov 5 09:29:17 EST 2010


At 14:19 30.10.2010, Juanito Talili wrote:
>Is there a Rasch version to Churchill's Paradigm for Developing 
>Better Measures

As a marketing researcher I would say yes and no.

But first, I want to point out who Churchill is and what role his 
seminal paper of 1979 has played in marketing, as I assume only few 
will know him outside marketing. Up to the late 1970s measurement in 
marketing was predominantely based on single items or short scales 
which were taken at face value. Some sort of crude content validity 
if you like. During the late 1970s it began to dawn on the scientific 
community in marketing that validity has to be addressed emprically.
Churchill's contribution and some more by others, which were 
published around that time, have had a tremenduous influence on the 
way measurement is carried out in marketing.
Of course, the paper is somehow outdated by now but the basic 
principles arguably haven't changed. I think one can say that 
Churchill made classical test theory the state-of-the-art in marketing.

In fact, there are some reasonable elements in Churchill's 1979 
paradigm, which have been watered down later on unfortunately (even 
by Churchill himself in subsequent versions of the paradigm published 
in text books). In 1979 Churchill made demands for multiple data 
collection during scale development and strongly advised against 
early item deletion based on factor analysis (in order to avoid 
capitalization on chance). He also said each item should have an 
equal amount of the core of the property to be measured. So he 
favoured parallel items. That could easily tested in factor analysis, 
but no one ever does it. He also asked for norms to facilite the 
interpretation of measures.

Today, people routinely apply exploratory factor analysis, delete 
items, then apply confirmatory factor analysis using the same data 
set and at the end of the day a bundle of SEM fit statistics 
demonstrate that everything is okay and measures are valid. And they 
get away with it.

Admittedly, the procedure has been refined, e.g. "advanced versions 
of reliability" are estimated, like composite reliability or average 
variance extracted.
In my mind, this changes very little, as the definition of 
measurement is still the one suggested by Stevens, manifest response 
scales are still treated as linear interval-scaled measures, 
sample-dependent statistics are used to assess precision and validity, etc.

Surprisingly, Rasch was proposed in marketing already in the 1980s (I 
know of two journal papers and one conference presentation). However, 
it sparked little interest at that time.
It is understandable given that Churchill's paper had been published 
only a few years before. These papers were well ahead of the times.

Since about 10 to 12 years ago several Rasch papers have been 
published. So, you may say there is an alternative paradigm around.
At least no one could actually say there would be no alternative.

However, there is a big qualification on that.
If you try to publish something in marketing using Rasch, you still 
encounter very strong resistance. I am sure that sounds familiar to 
researchers from other disciplines.

When you apply Rasch analysis, you always have to explain everything 
from first principles and you have to justify why you use Rasch 
instead of factor analysis.
I think it is really bizarre: use confirmatory factor analysis, use 2 
to 3 items for a latent variable (I recently read an article 
featuring a study comprising 10 latent variables and a total of just 
32 manifest variables - and some of these were very questionable in 
terms of content validity) - and you get away with it. Use a much 
more sophisticated paradigm, and you always have to justify and 
defend yourself.
Ironically, if you do justify and explain everything in detail, then 
reviewers come up with things like "not focused", "I don't believe 
this", "I cannot find the factor loadings", "no contribution" etc.

In a recent submission on gender-based DIF, we applied Rasch 
measurement. We pointed out that Rasch is superior compared to 
confirmatory factor analysis based on theoretical as well pragmatic 
reasons. As space was limited, we couldn't provide a comprehensive 
comparison of Rasch versus CFA; and I do not see the point as this 
has been done before. But simply citing other papers is not enough. 
And on top of that, one reviewer explicitely asked not only for a 
comparison but for a specific result of that comparison: "don't turn 
CFA down". So much to marketing SCIENCE.

I also believe most editors of leading journals actually do not want 
to see critical contributions of this sort.
This becomes obvious by the way reviewers are chosen. Each time, we 
nominated a handful of suitable reviewers, well respected scholars 
knwoledgable of Rasch - but not a single one was ever chosen as a reviewer.
Recently we had three reviewers and each of them explicitely said 
that they do not have a clue about Rasch measurement. This perverts 
the whole review process as you do not get any valuable feedback. 
Just poor nonsense.

Anyhow, I still hope that the situation will change in the 
not-too-distant the future. There are many young researchers who are 
much more open-minded.

On the MBC website www.matildabayclub.net I posted a list of 
publications on Rasch in marketing (follow Downloads and links and 
then choose the item in the menu at the bottom).
Or go directly to (then you do not have the outer frames, though):

If anyone knows of other Rasch papers in the Marketing/Business area, 
I would highly appreciate any hint.

Best wishes,

>A Paradigm for Developing Better Measures of Marketing Constructs
>Gilbert A. Churchill, Jr.
>Journal of Marketing Research, Vol. 16, No. 1. (Feb., 1979), pp. 64-73.
>Stable URL:
>Rasch mailing list
>Rasch at acer.edu.au
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailinglist.acer.edu.au/pipermail/rasch/attachments/20101104/d59fa6f9/attachment.html 

More information about the Rasch mailing list