Anchoring Bias in Runoff Polls: Evidence from the 2025 Ecuadorian Election

The study examines temporal anchoring bias in recall-weighted runoff polls using three empirical strategies: (1) a meta-analysis of 12 publicly available polls, (2) a provincial-level counterfactual simulation model, and (3) microdata analysis of a face-to-face survey (CIEES). Repository Contents Meta-analysis dataset (publicly compiled data) Structured database of 12 pre-election runoff polls (February–April 2025). Variables: polling firm, fieldwork dates, mode, sample size, vote intention (valid votes), disclosed weighting procedures. Python scripts implementing fixed- and random-effects meta-analysis, heterogeneity tests (Q, I²), and pooled estimates. Counterfactual simulation materials Official provincial-level first- and second-round vote shares from Ecuador’s National Electoral Council (CNE). Synthetic provincial microdata (27 districts × 1,000 simulated cases). Python scripts implementing geographic weighting, first-round recall calibration, and migration sensitivity scenarios. Python replication code Fully documented scripts for reproducing all tables and figures in the article. Meta-analysis aggregation procedures. Counterfactual estimations and sensitivity analysis. Weighting comparisons and vote loyalty matrix generation. Effective sample size and design effect calculations. Data Availability Statement The CIEES data used in this study derive from a face-to-face survey conducted in Ecuador in 2025 and were shared with the author for academic research purposes. The author does not hold ownership rights over the dataset. Due to third-party data ownership and the protection of human subjects, the individual-level data cannot be made publicly available. Replication materials, including full Python code, documentation, and all publicly available datasets used in the analysis, are available from the author upon reasonable request.