[Home Page] [Previous Page]
[Old Roundtable Archives] [New Roundtable]
Posted by Randy on February 04, 2000 at 15:02:36:
In Reply to: Questions, please help posted by karin on February 01, 2000 at 18:12:02:
Here's my tuppins…
According to Dr. Gopala Ganesh at the University of North Texas, in his many, many years of surveying and market research, he has found, on average, that differences between the initial "responders" and the second wave of responders, who were in the original "non-responders" group, are negligible. So the "logical conclusion" that one might make is that these original "non-responders" are no different than the initial "responders".
However, he and I both agree that this does not NECESSARILY mean there still isn't non-response bias, because the second group you surveyed may (or may not) have any specific interest in the subject matter, and therefore may or may not have been influenced by the survey topic. They may have just simply not opened the envelope the first time around. So I think we need to be careful about drawing what appears to be the "logical conclusion", as it may turn out to be the incorrect one. Logical doesn't always mean correct. In this case, intuitive might be more appropriate, which means that we, as researchers, actually have to have EXPERIENCE vs. a nameless, faceless "bumper to bumper" tab package doing the job. As John said, "this is still an art as much as a science", which, especially in the case of non-response, I agree with whole-heartedly.
While I know that re-sampling a portion of the original non-responders is usual and customary, I'm not convinced that just because their responses indicate - or don't indicate - the existence or non-existence of non-response bias, that it does or doesn't actually exist. As I have said before in this forum, there will ALWAYS be non-response bias, at least in this country. And because of that, I have a firm conviction that those who participate will eventually partake of the products and services that were designed, in part, from their input. The others will go hungry.
Here's the deal… on one hand, if these products and services don't match what this specific segment of the population wants, I think it's incumbent upon that segment to take it upon themselves to get involved in order to get what they want. On the other hand, I also think it's incumbent upon us within the research community to attempt to drive higher participation rates from other sectors of the population, which will go a long way toward reducing - or at the very least, understanding - the impact of non-response bias.
I personally think we can never fully quantify non-response bias, and any RIGOROUS attempt to do so is futile. I think it's OK to mention it, understand it, maybe get a "directional" finding on it, but then leave it at that. Because I think it is virtually impossible to know if the people that didn't respond did so because of the topic (and thus might fall under self-selection), or some other esoteric reason like not opening the envelope in the first place. Different reasons for simply electing to open or not open the envelope does not, in my mind, constitute a solid enough platform form for basing non-response bias.
Don't misunderstand me - non-response is a MAJOR issue. But I think we ought to understand how to get at these types of people in other ways than simply assuming a re-send of the same - or similar - survey constitutes that action.
Randy
Subject: