Comparison of alternative methods to infer the financial network: How companies are influential in a financial market?

Details

Supervisors

Béreau, Sophie

Faculty

Louvain School of Management

Degree label

Master [120] en sciences de gestion

Abstract

Along this thesis the analysis is based on a sample of daily closing prices of 40 different companies representing the Belgian financial market. To do so, several methods has been used to reveal the different clusters that are present in the financial network. After that some measures of centrality has also been computed to detect which company is the most influential one among the others. These tools are the Minimal Spanning Tree, the hierarchical tree and the Granger causality network. Among empirical finance literature there are many ways to analyze the relationships between the institutions in play. However this paper is based on some specific tools introduced by (Mantegna, 1999; Mantegna & Stanley, 1999) for the Minimal Spanning Tree or (Granger, 1969) for the Granger causality test. These types of graphs allow dealing with a great scale of data and complex links between the agents in a financial network. They also bring some specific indicators that offer the possibility to answer the question about which companies tend to be clustered together and why they are close among the financial market where they take place. The first part of the paper exposes a literature review that focuses in the first place on the presentation of the market under analysis here and how the selection has been made. Then it explains the different methods used in the empirical part and how they work. Indeed, it is important to understand the complexity of the models and how they produce their results. These models are the Minimum Spanning Tree, the hierarchical tree and the Granger causality network. After that comes the second part of the paper with the empirical study. This one will first present the methodology used to collect the data needed. Then I expose how I treated all the collected data into the specific software R for the different methods used. After that I analyzed the results obtained on a yearly basis of a time frame of 15 years. Then a more global analysis will be done to highlight the specific results and tendencies. To finish some tracks to go further will also be exposed as well as a conclusion.