Funded by the Italian Minister of University and Research (MUR) under Research Projects of National Relevance (PRIN), code 2020SKJSTF
In this project we propose a novel data-driven modelling approach to evaluate economic policies, with an empirical focus on Italy and with an application to climate change mitigation policies. It builds upon the idea of combining the flexible, realistic, and disaggregated structure of agent-based models with the coherent and consistent data framework provided by the Social Accounting Matrix. To do that we need to address several challenges both on the modelling and on the data representation side. Concerning modelling, we need to face the challenge of letting the model dynamically interact with data. On the data side, we have to integrate and harmonize several data sources in a framework that can serve and guide the development of the model. We identify the key for succeeding in a continuous and bidirectional feedback between data analysis and model building. In particular, we propose a novel approach in structuring data, called Multi-Agent Social Accounting Matrix (MASAM). It consists in a new data framework where the national social accounting matrix is linked with a set of granular information on firms and households, in terms of both stocks and flows, and where financial and environmental aspects are described. This detailed data structure dynamically interacts with the model via advanced estimation, data assimilation, and validation techniques drawn from the statistical learning domain. The main test-bed for evaluating the performance of the model will be its out-of-sample forecasting accuracy of macroeconomic variables of interest. Indeed, the predictive performance of the model will be continuously assessed. In this way, we plan to promptly intervene, spotting and correcting the inconsistencies that arise. A large body of theoretical investigation will be devoted to understanding the proper level of disaggregation that the model needs to provide accurate forecasts. Once a correctly calibrated model with high predictive accuracy is available, we can perform stress-tests on the resilience of the Italian economy to various shocks, providing accurate quantitative predictions on the outcomes of different sets of policies in the short-medium run. At the same time, we will be able to provide realistic scenario analyses, simulating the behavior of the economy over longer time horizons. Given the relevance and the global concern on the topic, we apply our framework to economic policies aimed at tackling climate change. In particular, we will investigate the consequences of fiscal and monetary policy reforms, aimed at mitigating the environmental effects of economic activity on the Italian economy. We believe that our effort toward the development of a consistent data-driven framework for agent-based modelling may constitute an important contribution to both scientific debate and policy making, especially considering the fundamental challenges the world is facing.
WP1 focuses on the construction of the MASAM database, inspired by the Social Accounting Matrix.
WP2 is devoted to construct and calibrate the model, guided by the MASAM.
WP3 explores the impact of heterogeneity and interactions on the economic dynamics.
WP4 is devoted to perform computational experiments on climate change mitigation policies.
Conference 'Multi-Agent Data-driven Modelling in Economics' (MADME 2025, 29-30 May 2025, Venice). Further details about the conference can be found HERE.