IS

Regression analysis of country effects using multilevel data: A cautionary tale Non-technical Summary Researchers often wish to compare and explain differences in socio-economic outcomes across countries. They aim to uncover ‘country effects’, i.e. to identify how different policy environments and institutions affect these outcomes and to inform the policy debate about how to improve outcomes. The outcomes considered vary widely, ranging from labour force participation and wages to political and civic participation rates, and social and political attitudes. The most popular quantitative approach to analysis is multivariate regression analysis of data from surveys or registers in multiple countries in which individual outcomes are modelled as a function of both individual-level and country-level characteristics. The popularity of this approach is illustrated by the European Sociological Review between 2005 and 2012. Of the 340 articles published, some 75 exploit multilevel datasets with individual respondents within countries. The statistical method that is most often applie