579.2
Estimating Societal Trends from Heterogeneous Cross-Sectional Time Series Surveys – Some Challenges Demonstrated on the Example of Church Attendance Trends in Europe

Thursday, July 17, 2014: 3:45 PM
Room: 416
Oral Presentation
Markus QUANDT , Data Archive for the Social Sciences, Leibniz Institute Social Sciences, Koeln, Germany
Ferruccio BIOLCATI RINALDI , University of Milan, Italy
Cristiano VEZZONI , University of Trento, Italy
Sample-based social science surveys have initially been a tool to collect data for analyses with a limited time-horizon. With the advent of long-standing survey programmes such as Eurobarometer, EVS/WVS, ESS, or ISSP, the option of deriving society level trend information from cross-sectional data collected over different time points has come into the world. It has often been demonstrated that larger trends can be observed from repeated cross-sectional surveys within the same survey program, when some methodological homogeneity can often be taken for granted. It is however much less clear whether data from different survey programs, with different methodological details, are sufficiently homogeneous to be cumulated into a common source database for building even longer, denser, and geographically more complete trends.

The present study uses a database of responses to church attendance questions in European surveys, compiled from the survey programmes named above. This database is analysed with respect to possible problems arising from the requirements of harmonisation across time and countries, given a variety of languages, question and questionnaire formats, and other design properties of the individual surveys. The database presently comprises more than 800 time/country samples from 32 European countries or regions and covers almost all years from 1986 to 2010. Analyses aim to establish the presence or absence or particular national and supra-national trends, identify aberrations of samples from trend patterns and investigate possible methodological factors behind such aberrations. Further, the analysis may eventually contribute to answering such questions as what the effects of different response formats, sampling design, administration modes etc. on reported average attendance levels are.