Abstract

Philippe Blanchard, "Sequence Analysis for Political Science", October 2011

Several areas of political research deal with sequences, that is, successions of standard categorical states or events: political sociology, evolution of regimes, analysis of speeches, geopolitics, comparative studies, or elections. At least three kinds of longitudinal methods, popular in political science, may attempt at treating political longitudinal objects: regression models, event history analysis and time series analysis. Yet, none can unfold the three dimensions of categorical time series, that is, the nature of the states/events composing the sequences, their order and length. Sequence analysis, with the optimal matching algorithm as a core tool, was specifically designed to this task. It is now commonly used in sociology and demography, and more and more in geography and history. This pragmatic, state-by-state comparison of sequences does not make any assumption about an underlying process that would generate sequences. The paper first defines sequences and their empirical applications. Then it details the principles of sequence analysis and its canonical steps. It shows how sequence analysis connects to and/or competes with other multivariate methods, before giving an overview of advanced issues and available software. To illustrate how fruitful this approach can be for political science, I apply it to a retrospective survey conducted among members of the main French activist organization mobilizing against AIDS.

back

top