Implementation features of large economic models

Agent-based economic models are invaluable tools for understanding complex economic systems and informed political decisions. However, the computational demands of large-scale agent-based models often require significant resources, limiting their accessibility to researchers and policymakers. This study investigates the software implementation details of historical agent-based economic models of Sweden, with a primary focus on the SVERIGE model and additional discussions on SESIM, SUNDSVAL, and MICROHUS. By examining these models and their successful implementations during periods of less capable computer hardware, we derive efficient software development approaches for modern computing systems that do not require supercomputers. These approaches include algorithmic optimizations, memory management techniques, leveraging modern hardware capabilities, and utilizing open-source libraries, frameworks, and cloud computing. Our findings demonstrate that it is possible to create large-scale agent-based economic models, which are both computationally efficient and accessible and have several important implications for the field of agent-based modeling and related disciplines. Addressing the computational bottleneck can help reduce the cost and time required for simulations, making these models more accessible to a wider range of researchers. Enabling the efficient execution of large-scale agent-based economic models can lead to better-informed policy formulation and implementation by better understanding of the potential consequences of these decisions. In addition, our study contributes to the growing movement towards open science and reproducibility in agent-based modeling by emphasizing the importance of efficient software development approaches and promoting open-source libraries and frameworks. A further direction of research in this area is the development of methods and tools for creating economic models for countries with a larger population, such as Russia.