Context and Scope

Agent-Based Computational Economics

One of the most challenging tasks in macroeconomic models is to describe the macro-level effects from the collective behavior of meso- or micro-level actors. Wheras in 1759, Adam Smith was still making use of the concept of an ‘invisible hand’ ensuring market stability and economic welfare, a more and more popular approach is to make the ‘invisible’ visible and to accurately model each actor individually by defining its behavioral rules and myopic knowledge domain. In agent-based computational economics (ACE), economic actors correspond to dynamically interacting entities (also called agents) who live inside a computer program. For many research topics, it is useful to combine ABM with the stock-flow consistent (SFC) paradigm. SFC-ABM models, however, are often intransparent and rely on very peculiar, custom-built data structures. A tedious task is to generate, maintain and distribute code for agent-based models (ABMs), as well as to check for the inner consistency and logic of such models.

Agent-based computational economics is a modeling approach where independent myopic units, called agents, interact. The outcome of this interaction (called emergence) can be a self-organized pattern much more complicated than the individual agents’ behavioral rules [Gatti et al., 2008, Gaffeo et al., 2008]. In Agent-Based Computational Economics, the agents are the types economic actors, such as individual firms or plant operators, performing certain operations such as investment or bidding at markets. There is, however, a large plurality of different agent-based economic models. Agent-based macroeconomic models are typically constructed of households, firms, banks, the government and a central bank, who are either aggregate entities or interact in a bottom-up fashion [Turrell, 2016]. A more detailed elaboration on agent-based comutational eonomics can be found, for example, in the two books: [Tesfatsion and Judd, 2006, Gallegati et al., 2017].

Stock-Flow Consistency

The appealing framework of sfctools builds around the Stock-Flow-Consistency (SFC) principle. The latter is a main feature of state-of-the-art macroeconomic (agent-based) models [Zezza and Zezza, 2019, Caiani et al., 2016, Reissl, 2021]. It originates from the pioneering work of Copeland [Copeland, 1949] and later Dos Santos [Dos Santos and Zezza, 2008] and has since then found application in many modeling exercises. The principle of stock-flow consistency applies to a wide spectrum of economic streams, ranging from Hydraulic Keynesianism [Andresen, 1998] to Post-Keynesianism [Caverzasi and Godin, 2015] and other schools of thinking. 1 It is therefore a versatile and widely applicable approach. The main idea is that the stocks (changes within an agent) and the flows (between agents) have to be consistent, just as stated in the Laws of Thermodynamics. The concept is therefore especially suitable for ecological macroeconomic models [Dafermos et al., 2014].

1: For an essay on economic modeing streams, have a look at [Dosi and Roventini, 2019].

What can I do wih sfctools?

Agent-based modeling is a powerful approach utilized in economic simulations to generate complex dynamics, endogenous business cycles and market disequilibria. sfctools is an ABM-SFC modeling suite, which i) relies on transparent and robust data structures for economic agents, ii) comes along with a simple descriptive modeling language, iii) provides an easy project builder for Python, making the software runnable and accessible on a large number of platforms, and iv) is manageable from a graphical user interface for ABM-SFC modeling, shipped as part of the suite, assuring analytical SFC-check and double accounting consistency. The package is shipped in the form of an open-source project. Unlike more generic frameworks like mesa or AgentPy, it concentrates on agents in economics. sfctools was designed to be used by both engineering-oriented and economics-oriented scholars who have basic education in both fields. It can be used by a single developer or by a small development team, splitting the work of model creation in terms of consistency and economic logic from the actual programming and technical implementation. This allows software solutions from rapid prototyping up to sophisticated, small-to-medium-sized ABMs. It is therefore a versatile virtual laboratory for agent-based economics.

The framework sfctools will accompany your modeling work along the whole model design process. Typically modelers will start with constructing their agents, i.e. the transactions between agents and their behavioral parameters. Sfctools supports modelers with a basic Agent class. All agents which inherit from this class will automatically be equipped with datastructures, the most important being the balance sheet. A flow matrix sheet collects all cash flows between agents, as well as changes in stocks. Structural model parameters can be read from a simple yaml file to avoid hard-coding. Finally, sfctools will take care about timing your simulation periods and executing batch simulation runs.

Literature References

1: Domenico Delli Gatti, Edoardo Gaffeo, Mauro Gallegati, Gianfranco Giulioni, and Antonio Palestrini. Emergent Macroeconomics An Agent-Based Approach to Business Fluctuations. Springer, 2008.
2: Edoardo Gaffeo, Domenico Delli Gatti, Saul Desiderio, and Mauro Gallegati. Adaptive microfoundations for emergent macroeconomics. Eastern Economic Journal, 34(4):441–463, 2008.
3: Arthur Turrell. Agent-based models: understanding the economy from the bottom up. Bank of England Quarterly Bulletin, pages Q4, 2016.
4: Leigh Tesfatsion and Kenneth L Judd. Handbook of computational economics: agent-based computational economics. Elsevier, 2006.
5: Mauro Gallegati, Antonio Palestrini, and Alberto Russo. Introduction to agent-based economics. Academic Press, 2017.
6: Gennaro Zezza and Francesco Zezza. On the design of empirical stock–flow consistent models. European Journal of Economics and Economic Policies: Intervention, 16(1):134–158, 2019.
7: Alessandro Caiani, Antoine Godin, Eugenio Caverzasi, Mauro Gallegati, Stephen Kinsella, and Joseph E Stiglitz. Agent based-stock flow consistent macroeconomics: towards a benchmark model. Journal of Economic Dynamics and Control, 69:375–408, 2016.
8: Severin Reissl. Heterogeneous expectations, forecasting behaviour and policy experiments in a hybrid agent-based stock-flow-consistent model. Journal of Evolutionary Economics, 31(1):251–299, 2021.
9: Morris A Copeland. Social accounting for moneyflows. The Accounting Review, 24(3):254–264, 1949.
10: Claudio H Dos Santos and Gennaro Zezza. A simplified, benchmark, stock-flow consistent post-keynesian growth model. Metroeconomica, 59(3):441–478, 2008.
11: Trond Andresen. The macroeconomy as a network of money-flow transfer functions. 1998.
12: Eugenio Caverzasi and Antoine Godin. Post-keynesian stock-flow-consistent modelling: a survey. Cambridge Journal of Economics, 39(1):157–187, 2015.
13: Yannis Dafermos, Giorgos Galanis, and Maria Nikolaidi. An ecological stock-flow-fund modelling framework. In 18th FMM Conference. 2014.
14: Giovanni Dosi and Andrea Roventini. More is different... and complex! the case for agent-based macroeconomics. Journal of Evolutionary Economics, 29(1):1–37, 2019.