Singular spectrum analysis para estimativa de núcleo de inflação no Brasil

Abstract

Inflation measures based on cost-of-living indices may not be the most appropriate for the use of monetary policy, as they may have a high degree of noise in their composition. A proposal to reduce the noise present in consumer prices indexes is the measure of core inflation, which aims to separate the noise from the estimation of the long-term inflation trend. This work evaluated whether the core inflation measures obtained by the Singular Spectrum Analysis (SSA) method are feasible for use by the monetary authority in Brazil. The measures estimated by SSA from the Broad Consumer Price Index (IPCA) data between 1999 and 2021 are compared to the inflation cores used by the Central Bank of Brazil, through the evaluation of the properties of unbiadesness, short-term adjustment dynamics and the predictive ability. The results indicated that the measures estimated by SSA through graphical analysis and grouping by frequency ranges fulfill the tested properties and demonstrate a greater predictive capacity than the measures used by the Brazilian monetary authority. Moreover, the Central Bank cores should be used sparingly, as some of the measures do not have satisfactory results regarding the unbiasedness and predictive ability. Therefore, SSA method grouping by graphical analysis and frequency ranges can provide better inflation core measures for the Central Bank of Brazil.