Taming the Tail Risks in Markets with Data-Driven Methods

Large empirical literature in economics and finance tries to explore various asymmetries in data ann especially their implications for the decision making. At the same time, the digital age is revolutionizing decision making in an economy in ways that we only now are beginning to understand. Growing interconnections of agents, companies, and computers form new dynamic networks that have changed the nature of financial interactions. Increasing role of algorithms and computers in financial decisions, availability of huge and permanent information streams affecting individuals, firms, countries, and the global trading system change dramatically the way financial sector is impacting society. Data itself are becoming valuable economic resource, large data sets generate risks and rewards increasingly central to economic decision-making and pose challenges for systems of data governance and ownership. 

The debate clearly indicates that the standard assumptions leading to classical asset pricing models do not correspond with reality. In the series of projects, and publications, we aim to show that models being able to learn the patterns from data can be built for more informative and precise models in finance that better characterize the heterogeneous behavior of investors. We aim to show that to understand the formation of expected returns, one has to look deeper into the features of asset returns that are crucial in terms of the preferences of a representative investor. We argue that two important, risk related features are tail events and frequency-specific (spectral) risk capturing behavior at different investment horizons, and we suggest how to infer such risks from data using modern machine learning tools.

Traditional finance has lost touch with many of these developments and needs to reorient its research. Example of types of newly identified risks: