The possibilities for biasing of mailed
questionnaire results due to only partial returns are all too obvious.
Nonreturners may well have their own peculiar views toward questionnaire
content in contrast to their more cooperative corecipients.
Thus it is
strange that very few published accounts of questionnaire-based research
report any attempt to deal with the problem. Some do not even acknowledge
it.
There are ways of at least partially accounting for the effects of
nonreturns after the usual follow-up procedures, such as postcard reminders.
To the extent that responders are asked to report personal characteristics,
those of returners may be compared to known population parameters.
For
example, the proportion of younger returners might be much smaller than the
population proportion for people in this age group.
Then results should be
applied only cautiously with respect to younger individuals. Anonymous
responses may be categorized according to postal origin (if mailed).
Then
results should be applied more cautiously with respect to underrepresented
areas.
Usually, the best way to account for nonresponders is to select a random
sample of them and obtain responses even at substantial cost.
This is
possible even with anonymous questionnaires, though, in this case, it is
necessary to contact recipients at random and first inquire as to whether
they returned the questionnaire.
Telephone interviews are often satisfactory
for obtaining the desired information from nonresponders, but it is almost
always necessary to track down some nonresponders in person.
In either case,
it may not be necessary to obtain responses to all questionnaire items.
Prior analyses may reveal that only a few specific questions provide a key
to a responder's opinion(s).
