Core inflation is widely used to guide monetary policy, yet most existing measures were not designed to forecast future inflation and often perform poorly in this task. This paper constructs a forward-looking core inflation measure for Brazil using disaggregated consumer price data and generalized non-negative ridge regressions that estimate the weights of price subcomponents so that the resulting index is maximally predictive of future headline inflation. Using monthly data from 1999 to 2025, we evaluate its out-of-sample forecasting performance and compare it with the official core measures used by the Central Bank of Brazil. The results show improved forecasts across different inflation regimes, generally matching or outperforming existing indicators.
Comissão Organizadora
Comissão Científica