A proposal for constructing and evaluating core inflation measures

There is no unifying framework for evaluating core inflation measures, so we propose a methodological framework to close this gap. It allows us to construct, evaluate, and rank core inflation measures by applying it to countries and regions with different characteristics, such as Chile, Colombia, Peru, the euro area, and the United States. Our methodology uses highly disaggregated data of consumer price indexes, and hinges on a standard quadratic loss function. We show that the usual indicator that excludes food and energy, which is the most widespread measure of core inflation among central banks, performs poorly across the five countries analyzed, due to substantial bias, low persistence, high volatility, and low forecasting power. Therefore, our recommendation is to revise its use. By optimally selecting the CPI components to be excluded, the properties of core inflation measures can be significantly improved. Finally, we argue that when there is a preference regarding the use of fixed exclusion measures, nothing is lost and much can be gained by optimally selecting the excluded items, instead of sticking with the usual ad hoc criteria of excluding food and energy. Results remain robust to changes in the sample and methodology.