ESRA logo
Tuesday 14th July      Wednesday 15th July      Thursday 16th July      Friday 17th July     




Thursday 16th July, 16:00 - 17:30 Room: L-101


Multilevel survey research, agent based modeling and social mechanisms: towards new frontiers in theory-based empirical research

Convenor Dr Dominik Becker (Heinrich Heine University Düsseldorf )
Coordinator 1Dr Tilo Beckers (Heinrich Heine University Düsseldorf)
Coordinator 2Professor Ulf Tranow (Heinrich Heine University Düsseldorf)

Session Details

This three-part session will lay the spotlight on the link between empirically-oriented theories and empirical research by focusing on the explanatory concept of social mechanisms. When explaining macro-level phenomena such as network structures or social diffusion outcomes, establishing the underlying social mechanisms is a strategy to overcome incomplete explanations which remain restricted on the macro level. Instead, the theoretical and empirical objective is to unveil the meso- or micro-level social mechanisms causing the macro-level explananda. Whether following Coleman’s explanatory macro-micro-macro model, (`wide´) rational action theory or DBO theory (desires, beliefs and opportunities), social scientists address the need for more fine-grained explanatory approaches.

Since about two decades, social mechanism research evolves to be an important paradigm in the social sciences. Yet, though survey data allow for and are often used to study social mechanisms, their methodological potential to do so is only rarely addressed systematically.

In part 1, we invite colleagues establishing social mechanisms as part of their theoretical explanation and actually researching these mechanisms applying different survey research designs. In part 2, we would like to bring together researchers using survey (and/or network) data and linking them to agent-based modeling, an approach which will gain importance to extend and enrich the use of survey data. In part 3, we invite presenters discussing either specific micro- or macro-based mechanisms, i.e. both survey-based and experimental approaches (including mediation and moderation analyses) as well as process tracing. We particularly welcome papers applying multilevel mechanism research, e.g. explicating cross-level interaction effects, or controlling for group-induced selection biases and linking analytical theoretical arguments with their data.

Abstracts should include theoretical references, a specification of the mechanism(s) under study and the method and type of data and analyses.

Paper Details

1. A closer look at the relation between religiosity and formal volunteering. A cross-regional analysis using Swiss data
Ms Elena Damian (University of Cologne)
Professor Elmar Schlüter (Justus-Liebig-University of Giessen)

Past studies revealed that living in a religious society has a significant influence on volunteering for both religious and non-religious individuals (Ruiter & De Graaf, 2006). However, most studies used a cross-national approach and there is little evidence of whether the same relation holds at lower levels of analysis, such as regions and counties. This study seeks to improve upon previous research by conducting cross-regional analyses to test under which conditions and how contextual-level religiosity shapes individual-level formal volunteering. The data used comes from Swiss Volunteering Monitor (2010).


2. Middle-range theories, moderator models and marginal effects: What does sour grapes make taste sweeter?
Dr Dominik Becker (Heinrich Heine University Düsseldorf)
Professor Klaus Birkelbach (University of Duisburg-Essen)

The idea of middle-range theories is that social mechanisms may only hold for a limited domain, (i.e. certain temporal, local, or personal conditions). We assume that statistical moderator models with their capacity to estimate marginal effects of a predictor over meaningful values of the moderator are an adequate technique to 'translate' the idea of middle-range theories into quantitative methods of analysis. Substantially, we use German panel data to analyze whether the impact of biographical 'failures' such as low occupational status and/or income on life satisfaction is moderated by respondents' internal vs. external locus of control.


3. Resilience as a Mechanism for Educational Success Despite Disadvantaged Circumstances
Ms Jennifer Tork (Bielefeld University, CRC 882)

The focus of the study is on youths with a successful educational career who grew up in disadvantaged circumstances of poverty. The construct of resilience is used to explain this, which is a new perspective in the sociological educational research. It could be assumed that children with experience of adversity or disadvantaged circumstances develop well. Using data of the German Socio-Economic Panel (SOEP) logistic regression are estimated. The empirical results show that there are social and personal indicators which promote the development of pupils differently depending on the experience of poverty.


4. Unraveling the paradox of job search via personal contacts and wages: Evidence combining agent based modelling and empirical research
Dr Gerhard Krug (University of Erlangen-Nuremberg)

Researching the effect of personal contacts on wages, two potential social mechanisms are of interest: better job chances and better wages. An agent based model shows that wage regressions using personal contacts as independent variable never produces an unbiased estimate of the true effect size and sometimes the estimate even has the opposite sign of the true causal wage effect of networks. I compare these results to evidence from multilevel survey data and find that in countries where personal contacts are effective in producing job offers, the share of jobs actually found via personal contacts is lower and vice versa.


5. Case study data for validating agent-based models
Professor Sharon Purchase (The University of Western Australia)
Dr Luis Izquierdo (The University of Burgos)
Dr Segismundo Izquierdo (The University of Valladolid)

The paper reports on new ways for developing agent-based simulation models (ABM) from case based data using a suite of qualitative data analysis. Secondary data sources and interviews are analysed to better understand the drivers for innovation in business to business (B2B) networks. Case data provides nuanced and rich descriptions of complex business network processes under-exploited in ABM.
By combining the case data with ABM, we extend and enrich the use of qualitative data in social simulation and provide new ways of validating simulation models when quantitative data is scarce.