Interpreting institutional investment activity as a markov process: A stock recommender

Article type: Research Article

Authors: Nayanar, Nikhil

Affiliations: School of Industrial Engineering, Purdue University, West Lafayette, IN 47907, USA | E-mail: nnayanar@alumni.purdue.edu

Correspondence: [*] Corresponding author: School of Industrial Engineering, Purdue University, West Lafayette, IN 47907, USA. E-mail: nnayanar@alumni.purdue.edu.

Abstract: Institutional investors like hedge funds are particularly resilient in times of volatility, primarily due to their access to a slew of complex trading strategies and extensive research capabilities; both of which are generally outside the reach of common investors. Fortunately, the U.S. Securities and Exchange Commission has mandated that large institutional investors publicly disclose the positions held by them at the end of a quarter, within the following 45 days. We then ask the question “Given our access to snapshots of positions held by large investors, can we extract alpha by approximating their aggregate behaviour?” In this paper, we introduce a stock recommendation model driven by the aggregate behaviour of institutional investors. We interpret stocks as the states of a Markov chain. The corresponding state transition matrix is defined based on the aggregate behaviour of investors. By designing the transition matrix to hold for aperiodicity and irreducibility, the steady state distribution of this chain is treated as a ranked list of stocks. We build a long only, equally weighted portfolio of stocks to trade using these recommendations. We observe that the returns on this portfolio beat existing models in the literature, and standard hedge fund indices both in terms of annualised returns and Sharpe ratio.

Keywords: Stochastic modelling, econophysics, institutional investors, algorithmic trading, markov process

DOI: 10.3233/IDT-220264

Journal: Intelligent Decision Technologies, vol. 17, no. 3, pp. 673-685, 2023