How can big data enhance the timeliness of official statistics?: The case of the U.S. consumer price index

Abstract

The daily consumer price index (CPI) produced by the Billion Prices Project (BPP CPI) offers a glimpse of the direction taken by consumer price inflation in real time. This is in contrast to the official U.S. CPI, which is compiled monthly and released with an average of a three-week delay following the end of the reference month. A recent body of research contended that the movements of online prices are representative of those of offline retail prices, making the BPP CPI a natural candidate for accurately improving the timeliness of the official CPI. We assess the predictive content of the BPP CPI using a variety of MIDAS models that accommodate data sampled at different frequencies. These models generate estimates that remain robust to the variety of time periods considered and, by the standard of the existing literature, contribute to a significant upgrade in the forecast accuracy of official consumer price inflation figures. The paper then sketches the broad implications of BPP CPI for the consumer price statistics maintained by national statistics offices and discusses how the proposed improvement in the timeliness of the official CPI fits in this perspective.