Psychiatric Morbidity and Nonignorable Attrition from the British Household Panel Study

In this paper we study the importance of possible non-random attrition on variations in psychiatric morbidity using data from the first eight waves of the British Household Panel Study. To study attrition we use the model of Hausman and Wise (1979) which applies to continuous longitudinal data with nonignorable or informative dropout. The model combines a multivariate linear model for the underlying response with a probit regression model for the dropout process. The latter incorporates dependence of the probability of dropout on unobserved, or missing, observations. Parameters in the model are estimated by using maximum likelihood. It is seen that the assumption of an informative dropout rather than a random dropout process has practical implications for the interpretation of the data. In particular we find evidence of attrition bias among the cohorts of men and women who were over 64 years of age at the time of the first interview in 1991. The model predicts that among men and women over 64 years of age at the time of the first interview those with high levels of psychiatric morbidity are more likely to leave the panel. As a consequence, the study may understate levels of psychiatric morbidity in late adulthood. Email: [email protected] Telephone: 01206-873901 Fax: 01206-873151 JEL code: C33.