Abstract Loyalty shifts within state security forces critically impact dissent campaign outcomes and have frequently precipitated regime breakdowns. Existing datasets are overly aggregated, impeding our understanding of the micro-dynamics underpinning individual decisions and coordinated efforts to abandon the regime. To address this gap, we introduce the Disloyalty During Campaigns (DDC) dataset (1970–2013), the first global disaggregated dataset documenting acts of disloyalty by security forces during nonviolent and violent campaigns. DDC records the type of disloyalty, number of participants, ranks, branches, and whether disloyalty originated at upper or lower levels of the security forces. We show that disloyalty occurred in only 13% of campaign-years but in 45% of campaigns, with widespread shifts often originating from individual acts. We also find that disloyalty is positively correlated with campaign success but becomes statistically significant only when participation exceeds 1,000. Coups significantly increase success on their own, while other forms are only effective in combination.
Dahl et al. (Thu,) studied this question.