In general, countries should note that differences in survey methodologies can lead to Inconsistencies in output. All countries should therefore aim to reduce sampling and non-sampling error (‘bias’) as much as possible by:
• Using a population frame which accurately reflects the target population (therefore which is up-to-date and representative).
• Using well-designed samples which are of sufficient size to produce reliable data (that is having
low standard errors for the aggregates suggested in this paper).
• Careful design and testing of questions, definitions and question sequences.
• Reducing unit and item non-response rates as far as possible (by, for example, using well designed questionnaires and following up outstanding responses); and
• Minimizing errors arising from data entry, editing and other data processing (by appropriate staff training and documentation).