Conferences

Survey research or electoral studies have repeatedly shown that political attitudes and behaviour are influenced by the collective economic and social situation, rather than the personal situation of individual citizens. This has lead to an increased interest in time-series or comparative analyses of the relationship between collective experiences and public opinion. In this field, multilevel analyses are now widely used as a privileged tool for studying the effect of contextual variables on individual survey responses. The most frequent procedure is to use contextual indicators at the level of entire nations, based on either official statistics, or ad hoc aggregations from micro-level survey. However, these mainstream procedures convey a series of limitations, on both practical and conceptual grounds. The most obvious practical limitation is that the necessarily limited number of higher-level cases (i.e. nations) potentially undermines efforts to develop sensitive, unbiased, and precise statistical models. More problematic even, these traditional approaches imply that procedures for computing new survey-based contextual-level indicators are of particularly little cost-effectiveness, because they require a representative sample for each higher-level case. As a consequence, potential micro-level data sources for creating new contextual-level indicators are limited to a very small number of existing international surveys. This creates a sizeable and, in our view, partially unnecessary obstacle for creative approaches to contextual data generation. The main conceptual flaws of these procedures derive form the fact that they introduce some rather arbitrary assumptions regarding the relationship between the communitarian backgrounds of collective experiences, and social representations. Why would the national community be the only relevant frame of influence for public opinion? And why would social and economic circumstances have a homogeneous impact on collective experiences within the national community, independently of the particular location of concrete events?

In order to over-come part of these shortcomings, we have started to develop an innovative procedure for generating contextual variables for multi-level analyses, which are spatially weighted. The starting point of our approach is close to geographically weighted regression analysis, which assumes that the effect of an explanatory variable on an outcome variable is a function of the geographical location of the cases (which can be individuals or areas). Extending this logic from physical distance/proximity between survey cases to multifaceted social interdependencies and influences, we suggest a range of continuous proximity functions for defining open-ended social contexts, focusing either on the territorial, symbolic, or practical foundations of social communities. This way, theoretical hypotheses regarding the social dimensions underlying collective experiences can be empirically tested. The conceptual assumptions and analytical procedures of this approach will be introduced. Then, a substantive research application will be presented. On the basis on a large-scale comparative survey across all countries of the former Yugoslavia, we are modelling the effects of collective experiences to violent conflict and massive social exclusion on public opinion within transitional societies. Research outcomes open new avenues for conceptualising social dynamics that shape public opinion at the meso-level of social reality, which are generally inconspicuous in multi-level models relying exclusively on individual-level and country-level indicators.