How do firms transact? Guesstimation and validation of financial transaction networks with satisfiability
Knowledge of monetary flow between firms can give a significant advantage both from a profit or research point of view. So-called firm-to-firm transaction networks are valuable in analyzing a market or an economy. However, such detailed and complete data is seldom available.
In this work, we aim at supporting economists by reusing available financial information from different sources at different levels of detail and completeness. With our technique, experts’ domain knowledge can be fused together with publicly available information to extract a representative, coherent instance of the transaction network. Supporting underspecification is important, as experts may develop partial econometric models. Our technique fills such blanks by systematically guesstimating missing information. Our approach builds upon formal foundations of satisfiability modulo theories and thus obtained transaction networks respect constraints imposed by domain knowledge and input data sources. We outline a taxonomy of general data types in the domain, and we programmatically construct formal predicates describing them. We demonstrate both guestimation of missing information of a transaction network and validation of external, expert-provided models. Finally, we investigate feasibility and performance of the advocated technique over a fragment of the Austrian economy.
C. Tsigkanos, A. Arleo, J. Sorger, S. Dustdar, How do firms transact? Guesstimation and validation of financial transaction networks with satisfiability, in: 2019 IEEE 20th International Conference on Information Reuse and Integration for Data Science (IRI) (2019) 15–22