Over the last decades, substantial progresses have been made by NSIs worldwide in developing new price collection techniques and implementing alternative sampling procedures for compiling Consumer Price Indices (CPIs). Moreover, the availability of transaction data, providing itemized information on quantities and sales at individual product level, has opened new perspectives to the use of alternative index calculation methods, such as weighted bilateral indices and multilateral index methods. Since the elementary aggregates form the building blocks of the CPIs, the definition of elementary aggregates and the choice of the most appropriate formula at this level can have a remarkable impact on the overall index. Although various multilateral methods have been suggested in literature with the common characteristic that price indices are constructed simultaneously for the entire sample period, there is generally no consensus as to which of these methods is the most appropriate for the measurement of price change. To contribute to the discussion on this topic, our paper addresses the issue of choosing the index method at the elementary aggregate level by analysing the performance of the GEKS method, based on Törnqvist indices, and in particular its weighted counterpart, the weighted GEKS. This analysis is conducted against the strong changes in the spending patterns of households occurred in the pandemic period (most part of 2020 and 2021) and given the consequences of the recent rise of inflation mainly due to the sharp increase of the prices of energy products started in the second part of 2021. As a matter of fact, the consumer behavior captured by scanner data provides the opportunity to analyse the performance of the weighted GEKS method as compared to the unweighted GEKS on a strongly perturbed real-world scenario. Therefore, we focus on a generalization of the standard GEKS method as it allows us to consider the reliability of the various bilateral comparisons by using different weights for different months in the reference period. Given what happened in the last three years, the choice of the window length is a crucial issue when deciding to apply multilateral methods for the possible consequence on the inflation profile. The longer the time window, the more data from the past will impact the current-month compilations thus leading to a loss of “characteristicity” that may affect temporal comparison using multilateral indices. Different alternative systems of weights are considered for introducing information on the reliability of the underlying binaries for the weighting matrix. To consider their performance in a period of economic turbulence, as mentioned before, we consider Italian scanner data from December 2018 to December 2022 when consumer behaviors were affected by the recent Covid-19 pandemic and the sharp rise of prices of energy products also related to Russia’s invasion of the Ukraine.
