Harnessing Machine Learning to Address High Levels of Missing Data in Cross-National Studies: From Bias to Precision in Public Service Research

This study addresses the challenge of missing data exceeding 50 per cent in international comparative research, where traditional imputation methods often introduce bias. Machine learning-based imputation captures complex variable relationships, enhancing dataset accuracy and integrity. The paper compares machine learning with traditional methods, validating its superiority through simulations. Findings show that machine learning improves consistency in time trends, uncovers dynamic relationships, and provides reliable tools for policymakers and researchers. Additionally, the study explores the impact of digital technologies on public services across national datasets, demonstrating how machine learning improves data quality and offers valuable insights for decision-making.