Can we infer the "laws" of finance from big data?

Abstract

A distinction often made between the physical sciences and social sciences is that the latter are not as amenable as the former to controlled experiments for rigorously verifying predictions made from theory. However, astronomy is a hard science in which it is impossible to do experiments and is almost exclusively based on observations. Just as collection and analysis of high-quality data from the period of Tycho Brahe led to the formulation of empirical laws by Kepler and was later followed by the theoretical groundwork of Newton that precisely explains the motion of planetary bodies, it is possible that the use of big-data, especially from financial markets, will eventually lead to a well-established set of "laws" of economic activity. In this talk we will explore the first steps in this direction, focusing on how analysis of price and transaction data from financial markets (including bitcoin as well as more traditional assets such as currencies and equities) suggests the existence of empirical regularities ("stylized facts") that may be universal across space, time and asset classes. In particular, we shall discuss the heavy-tailed distribution of asset price fluctuations (the "inverse cubic/square law"), application of Wishart random matrix spectral statistics to infer cross-correlations in the fluctuations of different assets and the possibility of using graph-theoretic measures such as structural balance as indicators of systemic crises.