How Much Should We Trust Micro-data? A Comparison of the Socio-demographic Profile of Malawian Households Using Census, LSMS and DHS data

摘要

This paper assesses the empirical representativeness of micro-data by comparing the Malawi 2008 census to two representative household surveys – ‘the Living Standard Measurement Survey’ and the ‘Demographic and Health Survey’ – both implemented in Malawi in 2010. The comparison of descriptive statistics – demographics, asset ownership, and living conditions – shows considerable similarities despite statistically identifiable differences due to the large samples. Differences mainly occur when wording, scope, and pre-defined answer categories diverge across surveys. Multivariate analyses are considerably less representative due to loss of observations with composite indicators yielding higher comparability as individual ones. Household-level fixed-effect specifications produce more similar results, yet are not suited for policy conclusions. Comparability of micro-data should not be assumed but checked on a case-by-case basis. Still, micro-data constitute reliable grounds for factually informed conclusions if design and context are appropriately considered.