Talking about the [same] revolution? A comparative analysis of main datasets of revolutionary events

Abstract

We conduct a comparative analysis of the main datasets on revolutionary events, including The Nonviolent and Violent Campaigns and Outcomes (NAVCO, versions 1.3 and 2.1) Cross-National Time-Series (CNTS), Revolutionary Episodes Dataset by M. Beissinger, Revolutionary Database by the HSE University (CSRA) and Historical Regime Data (HRD). To compare these sources systematically, we develop several complementary empirical instruments designed to identify and interpret differences across datasets. We show that, although the datasets are not fully equivalent and rely on somewhat different operationalisations, definitions and coding procedures, these differences usually concern event boundaries and the inclusion of borderline cases rather than the results of cross-national quantitative analysis. Across the major sources, we recover similar temporal and spatial distributions of events and broadly the same structural predictors. We therefore argue that, while these datasets are not conceptually identical, they are empirically highly convergent for many comparative purposes.

We conduct a comparative analysis of the main datasets on revolutionary events, including The Nonviolent and Violent Campaigns and Outcomes (NAVCO, versions 1.3 and 2.1) Cross-National Time-Series (CNTS), Revolutionary Episodes Dataset by M. Beissinger, Revolutionary Database by the HSE University (CSRA) and Historical Regime Data (HRD). To compare these sources systematically, we develop several complementary empirical instruments designed to identify and interpret differences across datasets. We show that, although the datasets are not fully equivalent and rely on somewhat different operationalisations, definitions and coding procedures, these differences usually concern event boundaries and the inclusion of borderline cases rather than the results of cross-national quantitative analysis. Across the major sources, we recover similar temporal and spatial distributions of events and broadly the same structural predictors. We therefore argue that, while these datasets are not conceptually identical, they are empirically highly convergent for many comparative purposes.