The element of surprise: the diverse impact of slightly alternative specifications of context on opinion formation in multi-level modeling

Abstract

The rise of multi-level modeling in social sciences brings new challenges. Multi-level modeling, although used in a great variety of ways, aims at simultaneously assessing the impact of individual-level and context-level characteristics on a dependent variable that is measured at the individual level. However, comprehending how public opinion is affected by context and how people experience contextual changes is a challenge. Little scholarly attention has been paid to the way context is incorporated in these models. The current study compares the use of the same independent variable (GNI) measured in three different ways to predict attitudes regarding EU enlargement, trust in the EU, and European Unification in 25 EU countries, using data from the European Social Survey. We introduce an alternative way to measure change, taking a larger time span into consideration, and find that this is a good way to measure contextual surprise. Our findings show that the way a variable is measured greatly influences the size and even direction of the effect. Hence, it is crucial to assess both conceptually and methodologically the best way to measure context before one simply implements a variable and interprets empty results.