Who Do Sovereign Investors Say They Are? Using Machine Learning Techniques to Build a Taxonomy of Sovereign Investors

Abstract

Sovereign investors are often referred to as long-term investors. But it is still unclear that these investors pertain to the same category, as they vary on many dimensions. Previous research proposed typologies of sovereign investors based on external criteria such as region or size of assets under management. The recent push for sovereign investors to be more transparent led to the availability of large amounts of “internal” qualitative data. Most sovereign investors now have websites and publish annual reports about their missions, investment strategies, and performance. Through these publications, sovereign investors are playing an active role in defining who they are. We use this large corpus of texts to propose a taxonomy of sovereign investors based on their chosen words. We answer the questions: what themes differentiate sovereign investors, and what external criteria are the most informative to categorize them? We use computer-assisted text analysis and structural topic modeling to analyze a corpus of about 300 annual reports for 36 funds over 10 years to answer these questions. This research contributes to defining the boundaries of the category of sovereign investors, identifies themes that supports several identities, and maps out sovereign investors based on both currently used criteria and themes coming from sovereign investors’ own vocabulary. We anticipate this new categorization of investors would help potential partners and organizations assess compatibility in values and identity with these funds. We also wish to present an analytical method new to the field of Engineering Project Organization.

Keywords: Sovereign investors; institutional field emergence; organizational identity; computer-assisted text analysis; structural topic modeling