We expect selective daily mobility to be a much more powerful
source of bias than selective residential migration. Indeed,
even if food, physical activity, and transportation preferences
influence neighborhood choice when moving to a new residence,
other criteria likely play a significant and perhaps larger role in
the choice of the residential neighborhood and dwelling (Kestens
et al., 2010b). Obviously, nutritional preferences have a stronger
and more overwhelming influence on the daily choice of the
physical activity facilities and food outlets visited or not. Accordingly,
sorting participants by types of facilities effectively visited
in daily life provides much more information on their nutritional
preferences than sorting them according to the characteristics of
their residential neighborhoods. Hence, a stronger confounding
bias is expected from selective daily mobility than from selective
residential migration.
Therefore it would be problematic to conclude that an increase
in the strength of environment–behavior associations, when
switching from a classical residential neighborhood study to a
GPS mobility study (both analyzed at the individual level), is
attributable to a better assessment of the causal effect of the
environment. Even if GPS tracking is a promising strategy to
48 B. Chaix et al. / Health & Place 21 (2013) 46–51
improve environmental exposure assessment, our concern is that
the appraisal of causal environmental effects moves one step
backward rather than one step forward with such studies, if
carelessly implemented.