Comparative research using data from the International Survey programs has often focused on Western Europe and North America where it is rather easy to find similar measures of the concepts of interest.
However, research is restricted by the fact that comparison is performed only on similar measures and in a context where there is not much variation between countries.
In this AMCIS seminar, the aim is to compare trust in various institutions in regions where there is much variation between countries and where major changes in governance took place in the recent decades, i.e., in South and Central America, Sub-Saharan Africa, West Asia and North Africa, Asia and Eastern Europe.
In the selected regions, 17 International survey projects comprise measures of institutional trust. The main projects are the Barometers, the World and European Value Surveys, the European Social Survey, Life in transition and the Latin American Public Opinion Project (LAPOP).
The measures of trust are disparate. The answer scales vary from 4-anchor scales to 5, 7, 10 and 11-point scales. More importantly, the institutions and organizations on which trust is assessed – more than 150 institutions and organizations -- vary between projects, between countries and over time.
This is a typical case where multilevel analysis appears the only way to go. Answers to questions on trust for various institutions (level 1) are conceived as nested within respondents (level 2), themselves nested within surveys (the country-year level 3) and within countries (level 4). The data set combines more than 1370 surveys conducted in 143 countries since 1995. It comprises more than 1.8 million respondents and 22 million measures of trust.
This AMCIS seminar shows how the use of a 4-level multilevel longitudinal analysis of repeated measures allows for dealing a posteriori with harmonization issues. In order to deal with the variety of answer scales, Durand harmonizes them on a unique scale and she controls for the length of the original scale at the survey level.
She can keep all the measures of trust in different institutions or organizations by nesting answers within respondents. Since there are too many institutions or organizations, they are grouped by theme a posteriori into 14 large categories of institutions in four different spheres, i.e. political, economical, administration and civil society.
In the end, Durand can compare the level of trust between grouped institutions, between respondents, over time and between regions of the world.
Cross-level interactions allow for testing whether trust in different institutions change similarly over time and in different regions. External indicators of the socio-political and economic situation of the different countries are merged with the data set in order to see whether they can explain some of the variation in trust between countries.
This seminar will present the process used, its pitfalls and advantages, and the results of these analyses. It will conclude on the usefulness of this approach to study other research topics.
Claire Durand is professor at the Department of Sociology at the Université de Montréal, Canada. Furthermore, Durand is past president of the World Association for Public Opinion Research (WAPOR).
Durand's research involves Quantitative Methodology, Survey Research and the performance of electoral surveys. Her current project is 'Pour une analyse historique des données d'enquête /For an historical analysis of survey data'
REC-B3.04 | 15:30 - 16:45