Ivan Makarov’s presentation at the All-Moscow Seminar “Expert opinion and data analysis”
V.A. Trapeznikov Institute of Control Sciences of the Russian Academy of Sciences together with HSE University’s International Centre of Decision Choice and Analysis hosted the All-Moscow Seminar “Expert opinion and data analysis”. Ivan Makarov, a Research Assistant at the International Laboratory for Digital Transformation in Public Administration, made a report on the topic “The Problem of Multicollinearity in Public Administration Indicators and Tools for Overcoming It”, within which he presented a part of his PhD dissertation study to specialists in the field of data analysis.
The presented study examines the problem of multicollinearity (strong correlations) between certain indices used for analysis in science and public administration. To give examples, the presence of multicollinearity among the Worldwide Governance Indicators and among several indices describing digital development was shown. The simultaneous use of such indicators can lead to distorted analysis results, as this may mean multiple counting of the same characteristics assessed by different indicators. This study examines the application of principal component analysis to overcome this problem by extracting principal components from correlated indicators and using them as mutually independent "basic" indices. The possibilities and specifics of using "basic" indices for quantitative analysis are explored. In particular, the issue of secondary aggregation of "basic" indices is examined and the possibility of their application for constructing indicative risk corridors is demonstrated. The analytical tools developed in the study are demonstrated by assessing the general levels of risks associated with digital development faced by countries around the world.
During the discussion of the study, a number of issues were touched upon. Among other things, the following were discussed: the reasons for choosing a relatively early time period for analysis; the ability to meaningfully interpret the extracted principal components; the need to analyze the indicators’ distributions to determine the boundaries of risk corridors; and the reasons for assessing countries with particularly rapid digital development as facing significant risks associated with digital technologies. The questions asked and valuable comments made during the discussion helped to further improve the study.
