How to lie with graphics

Abstract

Data visualization is a key element for informing a wide range of users of Official Statistics, from policymakers and media to analysts and statisticians, and to the general public. In this field of communication, National Statistical Offices (NSOs), which compete with other public and private information channels, often distinguish themselves through the quality, accuracy, and efficiency of the information they disseminate, particularly via statistically based graphics, graphical interfaces, and infographics. As with any statistic, these graphics are designed to support the understanding of complex phenomena, enable fair comparisons, and communicate unambiguous facts. While the essence of statistical graphics is to provide visual statistical tests revealing clear patterns and data-based “truths,” some fail to achieve this objective. This paper aims to highlight commonly encountered “visual lies”, defined by Tufte as graphics in which what is represented (or perceived) does not accurately reflect the underlying data. Such representations can distort interpretation and alter the statistical messages conveyed. Our conceptual approach is based on nine rules that illustrate how the understanding of information can be substantially modified through graphical design choices. Our objective is not to promote cynicism or suggest that all misleading graphics are intentional, but rather to encourage a critical perspective on data visualization. Examining how visualizations can mislead also provides insight into how they function, how they are interpreted, and how they can be improved to support trustworthy communication of Official Statistics.