Political Affiliate Clustering with Machine Learning in Vietnam Stock Exchange Market

Abstract

This paper aims to cluster politically affiliated groups using machine learning. The sample used in the study is enterprises listed on the stock exchanges of Ho Chi Minh City and Hanoi, research data during the period from 2015 to 2020. Data used in the study include state ownership ratio, the degree of political connection of business leaders and financial indicators in the listed financial statements of enterprises. The author’s study measures political connection by K-means algorithm and then compares the results of the K-means clustering with the traditional method of manual measurement of political connection including two values of 0 and 1, where 0 is no political affiliation and 1 is political affiliation. At the same time, the author runs three clusters to have in-depth insight. The authors conclude that machine learning clustering using the k-means model can replace the traditional method. Politically connected businesses listed on HOSE and HNX with political connections bring many benefits to businesses in investment activities, in accessing resources as well as capital; however, that businesses have a negative impact on business performance. The authors recommend that a moderate degree of political affiliation will help businesses achieve better performance.