Designing (Out) Democracy: Legal design choices of citizen-initiated popular vote processes and how they affect their use

Bottom-up direct democratic instruments such as citizen initiatives and veto initiatives are often described as mechanisms that expand citizen influence (Papadopoulos, 1995; Altman, 2018). In practice, however, their use varies widely (Altman, 2010). Out of 27 countries globally which have a citizens' initiative, 13 have never seen a popular vote on one. Existing research attributes this variation to individual institutional attributes—signature thresholds, quorum rules, or initiation mode. Yet processes sharing the same formal type often behave very differently. Drawing on Moeckli's (2025) data on citizen initiatives across 24 European countries, the proportion that successfully navigate signature collection and admissibility requirements ranges from 0.8% to 97.0%. This variation cannot be explained by the presence or absence of individual design features alone. This paper argues that frequency of use reflects not individual attributes but the underlying process architecture: the configuration of institutional conditions across sequential stages. Existing typologies classify popular vote processes by their endpoints (initiation and legal effect) while treating the institutional pathway between those endpoints as a black box. A process-oriented framework opens this box by specifying five analytical stages (initiation, admissibility, activation, popular vote, post-decisional effect) and identifying the institutional conditions that govern progression at each point. Where these conditions are located shapes usage patterns in ways that endpoint-focused classification cannot capture. The analysis relies on a new panel dataset covering 22 bottom-up processes in 15 countries from 1980 to 2023, capturing annual changes in design elements including signature thresholds, collection periods, admissibility requirements, activation rules, quorum requirements, and counterproposal provisions. Unlike existing studies that treat legal design as fixed national traits (Altman, 2010; Breuer, 2011), this research captures how legal configurations evolve over time and how reforms recalibrate accessibility. A central contribution is the systematic inclusion of registered initiative attempts, not only those that reached a vote. Most registration data come from the LIDD project (Moeckli, 2025); historical coding and direct contact with registration bodies supplement these. The paper models two outcomes: initiatives registered per year, and initiatives resulting in a popular vote. Modelling these separately distinguishes rules that discourage initiation from those that impede progression. Zero-inflated negative binomial models account for the high frequency of years with no initiatives. The regression includes variables reflecting the framework's stage-level logic: admissibility requirements (Stage 2), mandatory activation (Stage 3), and binding outcomes (Stage 5), alongside signature thresholds and collection periods (Stage 1) and quorum requirements (Stage 4). Initial results reveal that quorum requirements strongly discourage initiative registrations, not just approvals. Signature thresholds show limited explanatory power cross-nationally. Longer collection periods correlate with more registrations. Counterproposal provisions reduce initiative votes, suggesting elite responsiveness preempts escalation. Legal reforms produce short-term increases in activity. Critically, where blocking conditions are located matters: admissibility requirements at Stage 2 affect progression differently than quorum requirements at Stage 4, even when both create comparable barriers. The paper demonstrates that formal availability is insufficient for meaningful use and that understanding variation in direct-democratic practice requires attention to process architecture rather than endpoint properties alone.