Nowcasting CPI using online retail prices: forecasting combination of dynamic factor models


  • Igor da Silva Carvalho


06/08/2020 - 10:00


Via Zoom


In this work we use daily online retail prices and financial data to construct nowcasts of consumer price index (CPI) for the current month, which we refer simply as inflation. We first adapt the dynamic factor model in Gianonone et al. (2008) to our setting and integrate it into a nowcasting combination algorithm. We also propose an extension to the nowcasting equation to take into account previous inflation level and nowcasting error.We compare our results with a base model and Focus research. The base model is constructed from a simple factor regression using individual categories monthly price change calculated in Cavallo (2013), whereas the Focus research aggregate daily inflation forecasts provided by a large number of institutions to the Brazilian Central Bank. We consider the Focus research output as the golden standard. Our results indicate that (1) retail prices are not enough to capture the whole dynamics of CPI; and (2) the factor regression has to be augmented with recent inflation information to be competitive.

*Texto enviado pelo aluno. 

Membros da banca: 

  • Eduardo Fonseca Mendes (orientador) - FGV EMAp
  • Pedro Guilherme Costa Ferreira - FGV IBRE
  • Marcelo Cunha Medeiros - PUC