Sectoral co-movements in the Indian stock market: A mesoscopic network analysis

TL;DR

This paper reviews techniques for analyzing stock market data and applies them to the Indian stock market to visualize sectoral structures and co-movements using network analysis methods.

Contribution

It introduces a mesoscopic network approach based on sectoral indices to reveal sectoral relationships and co-movements in the Indian stock market.

Findings

Minimum spanning tree effectively separates related sectors.

Sectoral co-movements reflect actual production relationships.

Visualization techniques reveal sectoral structures.

Abstract

In this article we review several techniques to extract information from stock market data. We discuss recurrence analysis of time series, decomposition of aggregate correlation matrices to study co-movements in financial data, stock level partial correlations with market indices, multidimensional scaling and minimum spanning tree. We apply these techniques to daily return time series from the Indian stock market. The analysis allows us to construct networks based on correlation matrices of individual stocks in one hand and on the other, we discuss dynamics of market indices. Thus both micro level and macro level dynamics can be analyzed using such tools. We use the multi-dimensional scaling methods to visualize the sectoral structure of the stock market, and analyze the comovements among the sectoral stocks. Finally, we construct a mesoscopic network based on sectoral indices. Minimum spanning tree technique is seen to be extremely useful in order to separate technologically related sectors and the mapping corresponds to actual production relationship to a reasonable extent.