Using multilateral hedonic methods to capture product relaunches, Belgium

Statistics Belgium currently uses the GEKS-Törnqvist/CCDI multilateral method for supermarket scanner data. For consumer electronics scanner data, we use the Imputation Törnqvist GEKS (or ITGEKS) to capture quality differences between new and discontinued products. In supermarket scanner data, there are also products that leave the market and others that enter the market. Although quality changes are less pronounced for traditional supermarket products, the relaunch of products in supermarkets may coincide with a price level change, especially in the context of shrinkflation. In shrinkflation, the price of a product tends to remain the same, but the quantity received by consumers decreases, effectively increasing the price faced by consumers. Failure to take account of these relaunches could potentially lead to a bias in the CPI. Traditional (multilateral) methods use unique product identifiers to match the same product over time. In the case of relaunches, a new product may be given a new product identifier and any potential price change is missed when using these matched model methods. We show that multilateral index methods that use hedonic quality adjustments (e.g. ITGEKS or Time Dummy Hedonic method) can be used with supermarket scanner data and that they solve or provide insight into the relaunch problem. Under hedonics, products with similar characteristics are considered as similar products and relaunches or shrinkflation are effectively captured.