Two Multilevel Modeling Techniques for Analyzing Comparative Longitudinal Survey Datasets

Abstract

Increasing numbers of comparative survey datasets span multiple waves. Moving beyond purely cross-sectional analyses, multilevel longitudinal analyses of such datasets should generate substantively important insights into the political, social and economic correlates of many individual-level outcomes of interest (attitudes, behaviors, etc.). This article describes two simple techniques for extracting such insights, which allow change over time in y to be a function of change over time in x and/or of a time-invariant x. The article presents results from simulation studies that assess the techniques in the presence of complications that are likely to arise with real-world data, and concludes with applications to the issues of generalized social trust and postmaterialist values, using data from World/European Values Surveys.

Educational expansion and the education gradient in health: A hierarchical age-period-cohort analysis

“…As a solution, we examine the different periods as clustered within countries. This method of clustering was proposed by Fairbrother (2014) to create a sufficient number of second-level units. Given that not every country participated in every ESS round, 147 different country-periods are calculated.…”

The mental health consequences of the economic crisis in Europe among the employed, the unemployed, and the non-employed

“…Since the ESS collected information regarding depressive symptoms in only two rounds, we did not have a large enough number of units to include the period effect as a separate level of analysis The advantage of such national-level time-series cross-sectional data is that we were able to simultaneously model the cross-sectional effect, which explains differences 5 With regard to depression, the category of respondents with missing data on income did not significantly differ from the reference category: those with a high income. between countries, and longitudinal effects, which explain differences within countries over time (Fairbrother, 2014;Van der Bracht & Van de Putte, 2014). Applied to our model, this meant our main change variable, measured as the difference between national unemployment rate before the crisis (2006) and during the crisis (2012), was introduced in the model at this period level per country-year, while national pre-crisis unemployment rate was located at the highest level, the country level.…”

Homonegativity among first and second generation migrants in Europe: The interplay of time trends, origin, destination and religion

“…To include simultaneously both a longitudinal and a cross-sectional effect of native disapproval at the destination level, the longitudinal effect at the period level is group mean-centered, as described in Fairbrother (2013). In this way, the longitudinal effect of native disapproval is orthogonal to the cross-sectional effect.…”

“…Given the cross-national nature of the ESS, however, there is a possible solution to overcome this estimation problems for the period effect: considering the different waves as clustered within countries. Such national-level time-series cross-sectional data have the advantage that they enable us to simultaneously model cross-sectional effects, which explain differences between countries, and longitudinal effects, which explain differences within countries over time (Fairbrother 2013). A disadvantage, however, is that these models presuppose that social change happens over time within countries: time trends are each time nested within each survey country.…”