This repository hosts the Faces of Leaders dataset, which provides facial width-to-height ratio (FWHR) measurements for heads of government from every sovereign country between 1945 and 2020.
This page provides downloads, documentation, methodological details, and version history.
Version 0.1 of the dataset is available in the following formats. The archive (.zip) file contains all output photos.
FWHR was measured using the following protocol:
1. Leader identification
Leader names and tenures use "Identifying Ideologues: A Global Dataset on Political Leaders, 1945–2020". Heads of government are included. Herre B. Identifying Ideologues: A Global Dataset on Political Leaders, 1945–2020. British Journal of Political Science. 2023;53(2):740-748. doi:10.1017/S0007123422000217.
2. Image collection
Standardized frontal photographs were hand selected to fit requirements. If no photographs were suitable, data was omitted. Images were accepted only if the face was unobstructed, approximately frontal (<15° rotation), and of sufficient resolution. Future tasks could include collecting multiple photos for a more robust dataset.
Method based on, though not directly replicated, Ross A Miller, Facial Metrics, Aggression, and the Use of Military Force, Foreign Policy Analysis, Volume 18, Issue 4, October 2022, orac023, https://doi.org/10.1093/fpa/orac023.
3. FWHR measurement
FWHR is defined as bizygomatic width divided by upper facial height (distance from brow line to upper lip). Calculations made using Python and the face_recognition package, using interaction methods listed by Ties de Kok, former Assistant Professor at the University of Washington, and exported to a .csv file using Claude AI model Sonnet 4.6 to automate the process.
v0.1 - added the following countries:
United States
United Kingdom
France
Israel
Griffin Meeks