It might seem that following people over time provides the best indication
of how people change with age, Sample attrition can undermine that assumpti
on. This study describes the impact of health impairment, and depression on
attrition in the Notional Survey of Families and Households. It analyzes t
he impact of that attrition on estimates of the age-specific changes in dep
ression over a six-year period In doing so, it illustrates methods for asse
ssing and perhaps correcting the effects of attrition. Results show that th
e cross-sectional relationship of baseline depression to age differs sharpl
y for those who later drop out compared with those who stay in Much Of the
difference, but not all, vanishes with adjustment for health and impairment
. The probability of dropping out increases with poor health, impairment an
d depression at baseline. The impact of impairment and depression on attrit
ion increases with age. Panel models that ignore the attrition imply that d
epression decreases in old age. Models that adjust for the hazard of attrit
ion imply that depression riser by an amount that increases with age.