Replication Materials for “From Gatekeepers to Enablers: Party-System Personalization and Executive Constraint Erosion” This replication package contains the compiled dataset and documentation necessary to reproduce all quantitative results reported in the manuscript. The dataset integrates country-year and party-year information from the Varieties of Democracy (V-Dem v15) dataset and the V-Party dataset. All constructed variables, coding decisions, thresholds, and aggregation procedures are described below. The unit of analysis for the quantitative models is the country-year. The dataset combines the following original variables from V-Dem v15: v2x_jucon (judicial constraints on the executive), v2xlg_legcon (legislative constraints on the executive), and v2x_polyarchy (electoral democracy index). From the V-Party dataset, the following variables are used: v2xpa_antiplural (party anti-pluralism index), v2paind (personalization of party leadership), and v2paseatshare (party seat share in the national legislature). All original variables retain their coding scales as defined in the respective codebooks. The analytical sample is restricted to country-years between 1974 and 2019. The sample includes only country-years with a V-Dem polyarchy index between 0.3 and 0.8, operationalized as vulnerable electoral democracies. Country-years with missing party-system variables are excluded. The final analytical sample contains 733 country-year observations across 128 unique countries. Under the primary erosion threshold, 28 erosion events are identified. Executive Constraints is constructed as the mean of judicial and legislative constraints on the executive: Constraints_it = (v2x_jucon + v2xlg_legcon) / 2. Delta Constraints is defined as the year-over-year change in executive constraints: DeltaConstraints_it = Constraints_it − Constraints_it−1. The primary erosion event variable is coded as 1 when DeltaConstraints_it < −0.05 and 0 otherwise. Two additional robustness thresholds are included: a loose threshold of −0.03 and a strict threshold of −0.07. Seat Weight is constructed at the party level as SeatWeight = v2paseatshare / 100. Party-System Anti-Pluralism is computed as a seat-weighted aggregate: PS_AntiPlural_it = Σ (SeatWeight × v2xpa_antiplural). Party-System Personalization is similarly computed as PS_Personalization_it = Σ (SeatWeight × v2paind). The interaction term used in the main models is constructed as Interaction_AP_Pers_it = PS_AntiPlural_it × PS_Personalization_it. The dataset includes both the original country-year V-Dem variables and the aggregated party-system variables. It also includes the interaction term, erosion event indicators under all three thresholds, and sample restriction flags. Case-selection markers identifying country-years in the high anti-pluralism and high personalization cells are included for transparency and replication of the qualitative case selection logic. All variable transformations, seat-weighted aggregations, and erosion thresholds were constructed in Microsoft Excel and cross-validated using R. Linear probability models and logistic robustness models were estimated in RStudio. Predicted probabilities were calculated using model coefficients and mean-centered values as reported in the manuscript. Personalization is interpreted in the manuscript as an observable organizational correlate of weakened intra-party veto capacity. The seat-weighted aggregation captures the distribution of organizational authority across the legislative party system. Executive constraint erosion captures discrete annual declines in judicial and legislative oversight institutions as measured by V-Dem. The replication file allows reproduction of all main models reported in Table 1, robustness models reported in the appendix, predicted probability calculations, and identification of facilitation and blocked case-study country-years. No additional data cleaning beyond what is documented here is required to reproduce the quantitative results. All original data are publicly available through the V-Dem and V-Party projects. The compiled dataset provided in this replication package contains only derived variables and publicly available data transformations. Any errors in compilation are the responsibility of the author.