Conditioning is somewhat similar to

Conditioning is somewhat similar to contamination in its effects on the survey results,
but arises from a different cause. Conditioning may be evidenced in two alternative
ways. On the one hand, respondents may become so used to the questions that are
posed that they become adept at providing answers that are untrue but reflect what
they think the survey researchers want to hear. In general, the issues of affirmation
bias, demand characteristic, and social desirability bias are dealt with in Chapter 8 in
relation to one-time surveys. These issues are magnified in intensity in panel surveys,
in which the repeated nature of the survey can provide a means for respondents to
learn more about the intentions of the survey, and as a result provide more incorrect
responses. In other words, the repetitive nature of a panel survey can result in a learning
process for respondents that, in turn, leads respondents to provide answers that are
increasingly less truthful, and increasingly unrepresentative of the population from
which the sample was originally drawn.
The second aspect of conditioning occurs when the repetitive nature of the survey
actually leads to changes in the behaviour being measured in the survey. In household
travel surveys, survey researchers are all too familiar with elderly people who travel
very little deciding to reschedule their travel to the day on which travel details are
to be completed, so as to have something to report in the survey. Conversely, some
respondents who travel a great deal may put off some travel from the travel day to
another day, in order to have less to report, thereby reducing the burden of the survey.
Such behaviours are not uncommon in these one-off surveys, but may be exaggerated
into actual persistent behaviour change by panel respondents that is unrepresentative
of the general population. For example, a panel survey about household recycling may
induce panel members to be more conscientious in recycling household waste, or even
to start recycling when previously they did not. These behaviour changes are more evident
in panel surveys, in which the repetitive nature of measurement induces behaviour
shifts in the behaviours that are the subject of the panel survey measurement.
When such conditioning effects are likely to occur, the best mechanism to deal with
them is to ensure that no panel member remains in the panel for more than a small
number of waves. It has been noted in some long-running panels that, if no attempt
is made to remove potentially conditioned members, as many as 60 per cent of initial
recruits to the panel may still be in the panel after a decade of annual waves. Such a
high number of continuing panel members is not a problem if the survey is unlikely to
produce conditioning effects, but would result in serious bias in the results for a survey
that could have caused extensive conditioning of the respondents.