*5.3. Network Analysis*

Figure 7 shows the social network maps of the co-occurance matrix, collaboration matrix and coupling.The size of the nodes reflects the frequency of keywords in each cluster. A larger size suggests a stronger citation burst and suggests high significance of the subfield. The conceptual structure captured through keyword co-occurrence indicates diversity within the research sub-fields.

The network indicates three clusters in Keyword Co-occurrence and Author's Coupling. Each cluster represents a theme/field of study in economics research using ABM. Three major themes emerge from the Keyword Co-occurrence network: human behavioral research, climate change and urban development, and the development of agent-based models that may be used to investigate various economic phenomena. The citing authors in the topic area that are mapped are known as author coupling, and these maps can be used to focus on research areas that are shared by many currently active authors. These are split into three groups in this instance. One of the primary cluster authors shares simulation platforms, while the other two work on computational simulations and numerical models together.

The expansion of international collaboration in ABM research was placed in a highly stratified way, resulting in a clear divide between the main contributing countries and many others who collaborated globally on a more occasional basis. The network is made up of a core that is dominated by research outputs from scholars in research-intensive countries, with numerous additional countries gravitating around that core—unsurprisingly, given the emphasis on English-language journals.

#### *5.4. Research HotSpots*

Conceptual structure of agent-based modelling in economics is shown in a thematic map (Figure 8). The map was constructed by using keywords with min word frequency of 250. Minimum cluster frequency per 1000 documents is 5 and number of labels to each cluster are 3. On the x-axis we have centrality, which measures degree of interaction of one network with other networks, while density on the y-axis is measure of internal strength of a network. Themes on the upper right quadrant, i.e., motor themes have well developed internal ties and are important for the structure of the research field. Motor skills have strong centrality and high density. Niche themes are of marginal importance in the research field as they have well developed internal ties but unimportant external ties. Emerging or disappearing themes with low density and low centrality are weakly developed and marginal. Whereas basic themes are of high importance for the field of ABM in economics, but these are not well developed.

**Figure 8.** Thematic Map in Agent-based Modelling Research.

We discovered that research on agent-based modelling in economics can be classified in two ways, the first being the development of an agent-based model, and the second being the use of a developed agent-based model to investigate policy possibilities, based on Keyword Co-occurance analysis. The development of agent-based models employing economic theories, the application of diverse computational methods to calibrate these models, and the simulation of these models are all major topics in ABM currently. Although the largest bubble incorporates numerical models, computer simulations, and computational methods, these motor themes and basic themes both require a significant amount of research to get to a good measure of density and centrality.

#### **6. Research Frontiers**

There are several intriguing options for agent-based modelling research. ABM's versatility in applying to new issues has always been one of its best attributes. While certain classes of models have been established in fields like macroeconomics or financial markets, ABM has always been a transdisciplinary methodology that can be applied to problems involving a variety of rules, interactions, and behavioral phenomena (Steinbacher et al. 2021). ABMs can also be used to investigate issues that arise due to greater AI use, such as the societal impact of ranking algorithms and recommender systems and the potential reinforcement of social inequities and biases. ABMs can be used to build priors for machine learning algorithms in a semi-supervised manner in cases where the given data are noisy or biassed, reducing errors and preventing the amplification of distortions. Artificial agents can also be incorporated into large-scale simulators once they have been built based on the behavior of human subjects (Dosi et al. 2020). Such synergy between ABM and experimental technique is still in its infancy, but it represents an exciting avenue for future research, in our opinion. Further research on the estimation of ABMs is also required, as little is known about the benefits and drawbacks of various techniques. The majority of current models allow for the formulation and stochastic approximation of a likelihood function. As models become more complicated, such approximations will become increasingly difficult. In such instances, Approximation through Bayesian Computation methods and GMM/SMM should be considered (ABC). This approach (Sisson et al. 2007; Toni et al. 2009) employs measurements (moments) of the data other than the likelihood to approximate the posterior distribution of the parameters using a rejection sampling or Markov Chain Monte Carlo technique. While this approach has gained much traction in ecological ABMs (Csilléry et al. 2010), economic applications are still a work in progress.

#### **7. Conclusions**

During the 2008 financial crisis, the discussion began when economists began to investigate the possibility of agent-based modelling techniques for answering policy problems and performing "what-if" scenarios to aid policy decisions. This discussion about the future of agent-based modelling has yielded the desired result, with researchers currently working on developing agent-based models so that simulations based on simple rules can portray the complex economy. The European Central Bank is funding projects to construct agent-based economic models. The science of economics has long been in need of more robust methodologies that do not assume reasonable expectations and do not encourage optimism about the behaviours of agents. The economy appears to be a system, yet individual decisions cause the system's complex nonlinearity. Similarly, agent-based models recognise individual interactions and adapt in response to them.

The objective of the review analysis was to investigate the current developments in economics using this new modelling technique. According to the study's findings, research on the topic of agent-based modelling in economics is growing at a quick pace. Researchers experiment and publish their findings in research papers, books, and conference papers. Researchers are increasingly collaborating in order to improve the quality of their publications. The United States of America (USA), the United Kingdom (UK), Germany, and China are the most productive countries. The most prominent research on this topic has been published in the United States of America (USA), and American-based journals have taken the lead in publishing research in this field. This might be due to the increased formation trend of academic journals in the USA. The most often used keywords by the researchers (e.g., decision making, sustainability, and commerce) indicate the hotspots in ABM research. The main purpose of adopting different bibliometric analysis methodologies was to uncover research trends and the substance of published work. The knowledge structure and research trends were discovered through co-occurrence analysis. According to the findings, several well-established economic themes benefit from ABM techniques, but many require them. Researchers are not introducing central banks in an agent-based economy and conducting more monetary policy experiments. However, monetary policy had a considerable impact during and after the Great Recession. The channels of collaboration among scholars worldwide were uncovered through social network analysis. In the field of agent-based modelling, researchers promote cross-national and intra-national collaboration, which fosters the creation of new ideas. Although there has been much research into using agentbased modelling to comprehend the complexities of economic problems, underdeveloped countries like Pakistan have been slow to adopt this modelling technique. Even after a decade of economic disaster, economists discover agent-based modelling. This policy decision modelling technique is not widely used in economics, especially in monetary policy issues. Agent-based models, in addition to existing ones, could be critical instruments for assessing economic policy.

**Author Contributions:** A.Z. developed the proposed research model by reviewing relevant literature, analysing the data, and writing the paper, as well as revising the manuscript; A.U. contributed to the development of both research and practical implications in the paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Data Availability Statement:** The data for this study was collected from the SCOPUS database and is available upon request.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**

The following abbreviations are used in this manuscript:

ABM Agent-Based Modelling

DSGE Dynamic Stochastic General Equilibrium

#### **References**


Chen, Chaomei. 2017. Science mapping: A systematic review of the literature. *Journal of Data and Information Science* 2: 1–40. [CrossRef]

