Topic Classification of Central Bank Monetary Policy Statements: Evidence from Latent Dirichlet Allocation in Lesotho

Abstract

Abstract This article develops a baseline on how to analyse the statements of monetary policy from Lesotho’s Central Bank using a method of topic classification that utilizes a machine learning algorithm known as Latent Dirichlet Allocation. To evaluate the changes in the policy distribution, the classification of topics is performed on a sample of policy statements spanning from February 2017 to January 2021. The three-topic Latent Dirichlet Allocation model extracted topics that remained prominent throughout the sample period and were most closely reflective of the functions of the Central Bank of Lesotho Monetary Policy Committee. The topics identified are: (i) International Monetary and Financial Market Conditions; (ii) Monetary Policy Committee and International Reserves; (iii) Regional and International Economic Policy Conditions. The three-topic Latent Dirichlet Allocation model was determined as the most appropriate model through which a consistent analysis of topic evolution in Central Bank of Lesotho Monetary Policy Statements can be performed.