Abstract
Large financial transaction datasets are increasingly used to estimate carbon emissions associated with individual spending. However, to effectively target high-emission spending areas and implement successful carbon reduction strategies, policymakers and financial institutions need to understand individual consumer spending behaviour. In this study, we describe an approach to identify spending patterns in large financial transaction datasets, using stochastic block modelling for community detection on a bipartite network. This is an effective method to form communities of consumers who share similar spending patterns across merchant categories, allowing us to identify the categories causing high carbon emissions for each group of consumers. We also introduce a modification to the weights of the bipartite network which allows us to keep the average community spending constant across different categories. The impact and applications of this study are twofold. First, it highlights the importance of transaction datasets and stochastic block modelling in providing insights for financial institutions in their efforts to decarbonise by identifying areas for targeted behavioural strategies for carbon reduction. Second, it provides researchers with a framework to examine how different factors, such as consumer spending patterns, energy usage, or transportation habits, interact with one another. This is done while keeping overall spending levels consistent across various communities, allowing for a controlled analysis of behavioural and economic impacts on carbon reduction efforts.