**1. Introduction**

Modelling is defined as a mathematical and statistical way of reproducing events and their possible consequences due to policy decisions (Shanahan et al. 2016). In econometric modelling, models formed on a theoretical framework are constructed using one or many exogenous variables, identifying quantitative relationships which generate various responses. Dynamic Stochastic General Equilibrium (DSGE) models have ruled as a tool for policy decisions. Then the 2008 financial crisis became the downfall of this modelling technique because of "no response" to policy problems afterwards. These models cannot predict a crisis or any non-linear event. Meanwhile, the Agent-Based Modelling (ABM) approach emerged as an alternative to DSGE models. ABM emergence property allows foreseeing complex behaviours. The Great Recession made policymakers see the "economy as [a] complex evolving system" consisting of heterogeneous agents and non-equilibrium state continuously change the economy's structure.

Bibliometric analysis is a method to explore the information-rich environment on research activities and findings extracted through data from research publications in academic journals and their citations. Bibliometric indicators help investigate the knowledge structure of a particular field and its scope in the future. In the framework of research developments, questions like "where are we now?" and "where will we be in the future?" are answered by this form of analysis. There are two types of bibliometric analysis techniques: (1) performance analysis and (2) science mapping. In essence, performance analysis considers the contributions of research parts, whereas science mapping considers the relationships among them.

**Citation:** Zehra, Ayesha, and Amena Urooj. 2022. A Bibliometric Analysis of the Developments and Research Frontiers of Agent-Based Modelling in Economics. *Economies* 10: 171. https://doi.org/10.3390/ economies10070171

Academic Editors: Ralf Fendel, Robert Czudaj and Sajid Anwar

Received: 12 April 2022 Accepted: 13 June 2022 Published: 19 July 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

This article is dedicated to exploring the research development of agent-based modelling in economics using bibliometric analysis techniques. This study aims to use bibliometric analytic approaches to investigate the knowledge base of agent-based modelling in economics. We used both performance analysis metrices and scientific mapping methodologies to achieve the study's research aims. Also, ABM in economics is examined in terms of its conceptual and social structure.

#### **2. What ABM Has and What DSGE Lacks**

The primary instrument for generating policy judgments remains DSGE models. These models did not predict the financial crisis of 2008. These models' limitation in addressing many policy concerns is their data-driven approach and lack of macroeconomic data. These models' openness and transparency is a strength, but it also leaves them vulnerable to criticism. It is possible to draw attention to suspicious assumptions. Evidence of inconsistencies are readily apparent. It is possible to identify elements that are not included in the model (Christiano et al. 2018). DSGE modelling is a death tale that has been predicted. Joseph Stiglitz writes, ". . . much of the main parts of the DSGE model are defective—sufficiently seriously wrong that they do not give even a good starting point for creating a good macroeconomic model" (Stiglitz 2018). Vines and Wills want DSGE models to be able to achieve what they desire, which is to allow modellers to get a quick look at crucial issues. Central banks routinely use estimated DSGE models for forecasting and quantitative policy analysis. Estimating these models and interpreting the results to formulate policy are both difficult tasks (Schorfheide 2011). The issues and flaws of DSGE models are similar to those of generalised equilibrium (GE) models. Many researchers have discovered that the agents in DSGE models cannot be constrained so that their uniqueness and stability are preserved.

Furthermore, assuming individual rationality does not imply aggregate rationality. Because reactions to shocks or parameter changes may not resemble in aggregate, the representative agent assumption in these models is not reliable for policy analysis. In DSGE models, solving systems of equations can lead to another difficulty of identification, resulting in skewed estimates of some structural parameters and raising doubts about statistical significance. This modelling technique cannot predict infrequent economic crises, which is not surprising given that fat tail densities are approximated distributions of macroeconomic time series (Fagiolo et al. 2008), and Gaussian distributed shocks are a typical assumption in DSGE models. The assumption that Representative Agents (RAs) are rational prevents these models from addressing distributional issues because it implies that one: agents are fully aware of the economy; two: agents are capable of understanding and solving any problem they encounter without making mistakes; and three: agents are aware that all others follow the same pattern. These issues demonstrate that DSGE models are ill-equipped to solve policy concerns and cannot forecast future crises. The DSGE approach is so enthralled by its internal logic that it confuses the model's precision with the real one. (Caballero 2010).

ABM has evolved rapidly in economics over the previous two decades. Due to the following features of this modelling method;


These models are computer simulations that use a top-down strategy to investigate developing dynamic patterns. Policies and the social behaviours that result from them act like a weather system constantly battered by storms and invasions. The ability to make large-scale modifications and crash systems are inherited. External disturbances throw the equilibrium condition off. ABM allows little effects like herding and fear driving bubbles and crashes to be amplified via feedback processes. Models are non-linear in mathematical terms, implying that the result may not be proportional to the cause. The capacity to represent emergent phenomena resulting from the interaction of each agent is a major advantage of this modelling technique. Emergent phenomena can have traits that are opposed to those of their constituents. Agent-based models are a natural way to describe a system of behavioural elements. ABM can explain that designing a virtual agent with a shopping basket is more natural than describing average effects using a synthetic basket density. The flexibility of these models allows for the addition of new agents and changes in behaviour, learning, evolution, and complexity by altering interaction rules.
