Handling attrition and non-response in the 1970 British Cohort Study

With Richard Wiggins

CLS Working paper 2014/2.

The 1970 British Cohort Study (BCS70) is a continuing multi-purpose, multidisciplinary longitudinal study based on a sample of over 17,000 babies born in England, Wales and Scotland in 1970. The study has collected detailed information from the cohort members on various aspects of their lives, including their family circumstances at birth, education, employment, housing and partnership histories. There have been nine sweeps of data collection so far: at birth and at ages 5, 10, 16, 26, 30, 34, 38 and most recently age 42 (2012). This paper studies the extent of attrition in BCS70 and how it affects sample composition over time. We examine the determinants of response then construct inverse probability weights. In the last section, we use a simulation study to illustrate the effectiveness of weights and imputations in dealing with unit non-response and item missingness respectively. Our findings show that when the predictive power of the response models is weak, the efficacy of non-response weights is undermined. Further, multiple imputations are effective in reducing the bias resulting from item missingness when the magnitude of the bias is high and the imputation models are well specified.

Download paper in PDF