Tuesday 16th July
Wednesday 17th July
Thursday 18th July
Friday 19th July
Download the conference book
Multi-level Relationships and Social Mechanisms |
|
| Convenor | Professor Juergen Friedrichs (University of Cologne) |
| Coordinator 1 | Dr Alexandra Nonnenmacher (University of Hanover) |
Explaining cross-national differences between macro level variables often requires to include the micro level (and possibly the meso level) in the analysis, resulting in a macro-micro-macro model comprising contextual hypotheses and aggregation based on to the "Coleman boat". These are not sufficient to describe the much more complex relationship between a country and its residents. In the following, we list major methodological problems associated with "contextual" studies:
* Number of levels: In many contextual analyses, one or more meso levels, either spatial (e. g. regions, cities, neighborhoods) or social (e.g. social networks, school classes, departments), may have to be considered, as well as mediating effects for example of the mass media.
* Which characteristic(s) of the country level to consider? If macro-micro-correlations are not caused by composition effects, often the question still remains whether the country level characteristics have to be seen as a proxy for other country features, which are probably hard to measure. For example, the positive correlation between a country's affluence and its residents' life satisfaction is well established in the literature, but the causal relationship is not: Is it really collective wealth that leads to individual well-being?
* Social mechanisms or effects: How does a given characteristic or condition transpire from the aggregate to the individual level? What exactly is a "social mechanism"- if not a set of propositions?
* Similar Effects? Do we find similar mechanisms, such as "role models" for context effects on different forms of individual behaviour?
The topics listed show that cross-national analysis still raises many methodological problems, if one goes beyond stating correlations between country characteristics and tries to explain such relationships. We invite scholars to present papers which address the above questions from a theoretical, methodological or empirical point of view.
The paper deals with life satisfaction among women depending on their status on the labour market and number of children they have. The author claims that women living in the countries with more liberal labour laws and open labour markets are happier to have children and combine their work with childbearing than those women who have to re-enter labour market in the countries with very rigid labour legislation. The hypothesis is tested with the help of the EVS data for 2008 year, including 28 countries. The results of propensity score matching as well from the multilevel regression analysis proved the tested assumption and showed that females are happier in those countries with family oriented labour legislations and at the same time where it is easy to find a new job.
While Max Weber's (1968:1) demand for an understanding explanation already tried to bridge the gap between 'explaining' natural sciences and 'understanding' humanities, the deductive-nomological explanation (Hempel/Oppenheim 1948) became the dominant paradigm in quantitative social research.
Proponents of mechanism-based explanations (Hedström 2005; Elster 2007) criticize the lack of understanding in merely explanatory accounts and plead for more fine-grained analyses of the underlying causal chains ("opening the black box and showing the cogs and wheels of the internal machinery"). This is particularly important for theoretical justifications of contextual-level effects in order to avoid both contextual fallacy and invalid action-theoretical assumptions on the micro level (Farkas 1974).
The aim of our contribution is first to illuminate the desires, beliefs, and opportunity model proposed by Hedström (2005). The DBO model offers both a parsimonious action theory and an advantageous toolbox that allows to unveil the hidden causal chains in macro-to-micro explanations.
Second, we demonstrate this model's applicability by carving out the social mechanisms supposed to account for 'frog-pond' effects in educational and other settings (Marsh et al. 2008; Rosar et al. 2008; Becker/Birkelbach 2010). Particular attention is paid to disentangling the very distinct mechanisms depending on different outcomes (e.g., academic self-concept vs. teachers' evaluations) and different predictors (e.g., contextual-level achievement vs. schools' social composition) inducing frog-pond effects. These theoretical foundations are supported by empirical implementations in terms of state-of-the-art multilevel analyses.
Most of the research on the link between volunteering and subjective well-being suggest a general positive effect of volunteering on happiness. Yet, volunteering did not always have a significant effect on subjective well-being indicators (e.g., Haller & Hadler, 2006) or it was associated with a high level of life satisfaction in some countries, whereas in others not (Haski-Leventhal, 2009). Perhaps in some countries individuals can derive satisfaction from acting in the benefit of others, while in other countries being prosocial is not an important factor that predicts happiness. This might occur as a result of country differences in prosociality indicators. Using data from World Value Survey we examine what contextual level factors might serve as a mechanism that would make clear under which conditions volunteering is associated with a high level of subjective well-being. We have employed multilevel regression models in order to test the moderating effect of contextual indicators, such as frequency of certain educational goals (e.g., tolerance and respect for people) and of volunteering. Our results indicate that individuals that participate in voluntary activities report lower subjective well-being in countries where tolerance is perceived as an important quality that children should be encouraged to learn at home than in countries where tolerance is not an essential educational goal.
Theoretical models on the interaction between macro and micro levels are often tricky to be empirically tested. For instance, the Coleman "boat" suggests a bi-directional cross-level interaction across time: a predefined context influences the change across time of individual variables (e.g., attitudes), in turn the individual change ("adaptation" in Coleman's terms) feeds back the context and possibly changes it. With the spread of multilevel modelling in social science, the first path (context-to-individuals) has been tested by applying multilevel methods. On the other hand, the most straightforward solution to assess the opposite path (individuals-to-context) is still by using indicators derived from aggregated individual scores (e.g., Inglehart & Welzel, 2005; Morselli & Passini, 2012). However the use of aggregated indicators in surveys is strongly dependent by the scale of the context units and becomes problematic when small contexts are taken into account (e.g., region or municipalities). The smaller the contextual unit in terms of sample size, the larger the standard error and the feebleness of the aggregated indicators. In this paper we present the use of geographical weighting techniques developed in the R-package "spacom" (Junge et al., 2012) to overcome this problem. According to this method, the aggregation is performed taking into account the neighbouring effect of all available contextual units. Thus, the aggregated score for each unit is not only the raw aggregation of individual micro-data, but also the sum of weights of its surrounding units. This method has been shown of building theoretically and statistically more robust indicators which can be used in longitudinal models using cross-sectional cross-national surveys.