This paper investigates the relationship between the views expressed in the minutes of the meetings of the Central Bank of Brazil's Monetary Policy Committee (COPOM) and the real economy. It applies various linguistic machine learning algorithms to construct different measures of the uncertainty contained in the minutes of the COPOM. To achieve this, we first infer the content of the paragraphs of the minutes with Latent Dirichlet Allocation (LDA). Secondly, we build an uncertainty index for the minutes with Word Embedding and K-Means. Thirdly, we create two topic-uncertainty indices. The first topic-uncertainty index is constructed from paragraphs with a higher probability of topics related to general economic conditions. The second topic-uncertainty index is built from paragraphs with a higher probability of topics related to inflation and the monetary policy decision. Then, via a Structural VAR, we explore the lasting effects of these uncertainty indices on some Brazilian macroeconomic variables. Our results show that an unexpected increase in the minutes' uncertainty leads to a depreciation of the exchange rate and a decline in industrial production and retail trade. Moreover, we show that a positive shock to the general economic conditions topic-uncertainty index leads to higher inflation, whereas a positive shock to the inflation and monetary policy decision topic-uncertainty index leads to lower inflation.
Read full abstract