Strategies for dealing with survey error depend on the source of the error. So you should try to investigate the extent to which each different type of error contributes to any discrepancies between the survey and official data you observe in order to formulate the best strategy to adjust for the discrepancies.

Techniques for tackling survey error are many, varied and often complex (see, for instance, Groves 2004). Here we consider one particularly common form of adjustment: weighting. This is a particularly useful strategy for dealing with unit non-response and relatively little or no information on the nature of the non-respondents. If you know that the distribution of your sample on a key variable (maybe but not necessarily the dependent variable) is different from official (accurate) sources, even if you are unsure of the source of the discrepancy or whether it is in fact an error in the survey, it is usually reasonable to weight your data to match the official marginal distribution for that variable. This could be simply for presentational purposes, but if the use of weights affects your substantive conclusions it is important to report how and why.

The University of Manchester; Mimas; ESRC; RDI

Countries and Citizens: Unit 4 Combining macro and micro data by Steve Fisher, University of Oxford is licensed under a Creative Commons Attribution-Non-Commercial-Share Alike 2.0 UK: England & Wales Licence.