Measurement error contaminates many responses recorded in microdata. It usually increases dispersion and tends to lead to upward-biased inequality and poverty measures. It can cause apparent attenuation of behavioural relationships and distort the view of nonlinearity that may be present in those relationships. Policy interventions aim at altering the underlying variables whose relationships with outcomes are distorted by measurement error. The research effort in this area focuses on developing methods for understanding and reducing the impact of measurement error on inference about social and economic processes.

The impact of measurement error on the measurement of poverty and inequality is studied in cemmap working paper CWP 03/01 (joint with Christian Schluter (Southampton)) published in revised form in the Review of Economic Studies in 2002. Small parameter approximations as set out in 'The effect of measurement error', Andrew Chesher, Biometrika, 1991, are informative about the generic impact of measurement error on the understanding of structural features obtained from measurement error contaminated data. cemmap working paper CWP 02/01uses these methods to study the impact of measurement error on quantile regression estimation and in joint work with Erich Battistin these methods are used to study the impact of measurement error on matching and other programme evaluation procedures.