(607) 592-0546
Applications: Election Forensics, Electoral Geography, Simulations and Data Methodologies: Automated Data Gathering, Geographic Analysis, Information Fusion, Machine Learning, Statistical Estimation Relevant Projects: Election Forensics.

Walter Mebane

Professor, Political Science


Center for Political Studies, Statistics

My primary project, election forensics, concerns using statistical analysis to try to determine whether election results are accurate.  Election forensics methods use data about voters and votes that are as highly disaggregated as possible.  Typically this means polling station (precinct) data, sometimes ballot box data.  Data can comprises hundreds of thousands or millions of observations.  Geographic information is used, with geographic structure being relevant.  Estimation involves complex statistical models.  Frontiers include:  distinguishing frauds from effects of strategic behavior;  estimating frauds probabilities for individual observations (e.g., polling stations);  adjoining nonvoting data such as from in-person election observations.

Hotspot Analysis, Extreme Fraud Probabilities, South Africa, 2014

Hotspot Analysis, Extreme Fraud Probabilities, South Africa, 2014