Looking for one or two model papers for a conference paper. Does anyone have a favorite paper on nonresponse bias (unit and/or item) in elderly populations...or papers correlating age with nonresponse? Preferably recent and based on RDD, but that's not essential. Just looking to see if/where there's a gap in this literature. Library and Google Scholar searches aren't turning up much recent work focusing specifically on age, and nothing based on RDD samples specifically.
Thanks in advance!
It is met in most surveys, but there cannot be one single answer. That is, sometimes younger, but sometimes older people respond poorly. Much depends on the survey and how respondents have been tried to contact.
Hello Los Angeles from Boston!
Can you please give more specifics? I am aware of the ongoing evolution in rates of nonresponse to web surveys by age groups. For a possibly unhelpful example, here's a recent article that compares different countries about selection bias with access to the web and doing web surveys in general, but they did not stratify by age (and they did not include the US): http://stats.cwslive.wiley.com/details/journalArticle/2232891/Selection-Bias-in-Web-Surveys.html
Of course, results are going to be different than for phone or in-person surveys, and we still have to look at your specific question, which is by age. Context is important; older people are more likely to answer a survey about an issue affecting older people, for example. Country is important, language is important, subject is important, etc. That probably why you are not finding much in the way of "general" papers on the subject.
Please advise us! Your question is extremely thought-provoking. We might be able to find a more specific paper to guide you with respect to your specific context.
-Monika
Thanks, Monika and Alec, for the comments and for the links to papers. The finding that older respondents will be more likely to respond than younger respondents fits with Groves et al's leverage-saliency theory of unit nonresponse...people are more likely to respond when the topic touches them personally, and when it doesn't, other design features like incentives or contact/recruitment techniques have more power.
See a couple of key papers here:
http://www.jstor.org/stable/3078721
http://poq.oxfordjournals.org/content/68/1/2.short
I found a couple of older papers in my search yesterday, and will share later. Though nothing recent and on RDD specifically. It can also be hard to find papers that look at response rates AND nonresponse bias in general (not just on age).
Monika, since you asked me to clarify my questions, I'll do that...If I were to restate my general question as more specific research questions, they'd look like those below. Let me know if you know of or see anything like that.
RQ1: Are people age 65+ more likely than younger people to be contacted in RDD surveys of the general population?
RQ2: Given contact, are people age 65+ more likely than younger people to complete a phone survey interview in RDD surveys of the general population?
RQ3: If rates of contact and cooperation/response are different in those 65+ v. younger, is there any evidence that those rates produce nonresponse bias in estimates? If the differences in response are large enough, we'd expect that statistics about topics/characteristics that older adults are more likely to endorse will be biased if post-survey adjustments aren't done (e.g,. % retired, % on Medicare, certain illnesses).
RQ4: Are respondents 65+ more or less likely to to not answer survey questions once they begin a phone survey, and is there any evidence that this item nonresponse is related to nonresponse bias?
Thanks in advance for your (or anyone else's) comments/leads.
Thanks Alec...great paper. You're right that Roger's book is a go-to source for "response behavior" (e.g., answering questions and measurement error). My copy is quite annotated. :) I looked in Groves et al "Survey Methodology" but didn't find much on age. Groves "Survey Errors and Survey Costs" has some, but that's out of date.
One of the big open areas in our field (you might know) is getting beyond response rates (who responds and why) to nonresponse bias (what difference does that make in estimates). That's at the unit level. At the measurement/item level, stuff you've mentioned is helpful. Another open area for those pushing the field is to connect more than one error source (e.g., are nonresponse bias and measurement error correlated). I imagine no one's done this looking at older adults yet.
Is aging your specialty? Thanks much for the inputs and leads?
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Applied Statistics
- Survey Design