The recent availability of high-frequency electronic-point-of-sale ‘‘scanner data’’ together with traditional Consumer Price Index (CPI) data has the potential to significantly change how to compile spatial price indexes. Indeed, one of the main issues when constructing sub-national Purchasing Power Parities (PPPs) is to obtain price data from multiple sources and outlets, which are representative of local consumption patterns and comparable on the basis of a set of price determining characteristics. Within this framework, the aim of this paper is to suggest a new stochastic methodological approach to index numbers based on the Country-Product-Dummy (CPD) method for computing sub-national PPPs in Italy. This approach enables us to use both scanner data and traditional CPI data coming from large-scale retail trade. By using millions of price records concerning food and grocery products collected in supermarkets and hypermarkets of the most important chains of modern distribution located in the 20 regional chief towns in Italy, we provide estimates of Basic Heading (BH) PPPs for the year 2015. By using the CPD based stochastic approach we are able to obtain reliability measures for sub-national PPPs as well as to address methodological issues concerning the choice of the aggregation methods above the BH level.