Design Effects and Misspecification Effects for Cross-National Comparisons Where Sample Design Varies Between Countries

• Activity Number: 567
• Type: Topic Contributed
• Date/Time: Wednesday, August 4, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Olena Kaminska*+ and Peter Lynn
• Companies: University of Essex and University of Essex
• Keywords: Design effects ; Misspecification effects ; Complex surveys ; Cross-national ; Sample clustering ; EU-SILC

Abstract

In cross-national surveys, sample design often differs substantially between countries with respect to clustering, stratification and selection probabilities. Estimation of standard errors of between-country differences typically fails to take this into account. Common approaches are, a) to estimate standard errors for country means but then make informal comparisons, or b) to test differences formally but under an assumption (usually unrealistic) that the sample design is common across countries. Using data from the EU SILC survey, we develop design-based estimates of standard errors of between-country differences in means. We compare these estimates with estimates of type b) above, where only design weights are taken into account. Resultant misspecification effects are presented to demonstrate the importance of correctly estimating standard errors for between-country differences.