MULTIPLE CONTEXTS AS CROSS-CLASSIFIED MODELS - THE LABOR VOTE IN THE BRITISH GENERAL-ELECTION OF 1992

Voters make their decisions in social and geographical contexts that c an be seen as different levels in an overall data structure. Increasin gly these structures are being analyzed by multilevel models, but this approach has so far been limited to structures that are strictly hier archical. This paper outlines the approach of cross-classified multile vel models in which units at lower levels in the structure can be nest ed in more than one higher-level unit simultaneously. An appropriate m odeling framework is outlined, models are specified, and particular at tention is paid to efficient computation. The approach is illustrated through a cross-classified legit analysis of Labor versus Conservative support for a nationally representative sample of voting behavior for the 1992 British General Election. The data Is structured so that ind ividual voters at level 1 are nested within constituencies at level 2 which are cross-classified by geographical and functional regionalizat ions at level 3. A conclusion discusses the general utility of a cross -classified approach to geographically based contextual research, whil e two technical appendices provide details on model estimation.