A0804
Title: Property of financial adjacency matrix for detecting local groups
Authors: Minseog Oh - KAIST (Korea, South) [presenting]
Donggyu Kim - KAIST (Korea, South)
Abstract: The property of the inverse sample covariance matrix of the assets' returns as the financial adjacency matrix is investigated. Within the framework of the multi-level factor model, employing the covariance matrix as an adjacency matrix is inadequate due to the predominant impact of common factors. To detect the local group structure more effectively, utilizing the inverse of the sample covariance matrix is suggested as an adjacency matrix to reduce the common factor effects and magnify the local factor effects. It is shown that the inverse of a covariance matrix has better properties for being an adjacency matrix to identify local group membership than the original input covariance matrix. The empirical study with the returns of the top 500 trading volume stocks demonstrates that using the inverse covariance matrix as a financial adjacency matrix helps detect local groups.
