**3. Bibliometric Analysis: A Technique of Systematic Literature Review**

One of the most important knowledge discovery methods is synthesising the results of earlier studies. The use of bibliometric analysis is growing in popularity (Zupic and Cater 2015 ˇ ). In a qualitative study of published research papers, journals, and books, the bibliometric technique has been employed (Ellegaard and Wallin 2015). It aids in the identification of frequently referenced authors and institutions, related publications, and the keywords most commonly used in a given study filed (Daim et al. 2006). Furthermore, bibliometric analysis can be used to assess the publication's popularity among specialists and verify the author's reputation (Ball and Tunger 2005). It also aids in literature review by leading the researcher to influential research works or publications, as well as objectively mapping the study field (Zupic and Cater 2015 ˇ ). (Donthu et al. 2021) discused in detail the methodology to conduct bibliometric analysis and concluded that bibliometric analysis can aid knowledge generation not just in business research but also in other sectors, thanks to a better comprehension of science.

Bibliometric approaches are used for a variety of purposes, including performance analysis and science mapping (Cobo et al. 2011). Performance analysis is used to assess individual, institutional, and individual research and publishing performance. A generic approach of domain analysis and visualization is science mapping. A scientific discipline, a field of research, or topic areas related to specific research topics can all be included in the scope of a science mapping study. In other words, an area of scientific knowledge expressed through an aggregated collection of intellectual contributions from members of a scientific community or more clearly defined specialty is the unit of analysis in science mapping (Chen 2017).

Citation analysis, co-citation analysis, bibliographic coupling, co-word analysis, and co-authorship analysis are some of the methodologies utilised in science mapping. Such methodologies are beneficial for illustrating the bibliometric and intellectual structure of a study field when combined with network analysis (Baker et al. 2020; Tunger and Eulerich 2018).

#### **4. Study Design**

The goal of this study was to fulfil two purposes. The first was to determine research trends over time, and the second was to investigate research content in order to assess the application of agent-based modelling techniques in various economic sectors. Hence we combined both bibliometric methodologies; performance analysis and science mapping to reach our objectives.

*Research Objectives*: The analysis is constructed in such a way as to achieve the following goals.


*Research Design*: Many methods are described in Section 3 of the bibliometric analysis. Citation and co-citation analysis (by author and journal), co-word analysis, and network analysis were among the bibliometric methodologies we considered.

*Bibliometric Data Collection*: The Scopus database offered a total of 1568 documents for examination. A sophisticated keyword selection was required for data extraction in order to provide a relevant set of data. The keywords chosen must match the following four criteria: high search volume, relevancy, high conversion value, and low competition. The search terms "agent-based modelling" AND ("DSGE" OR "monetary policy" OR "crisis" OR "central banks") were used to extract bibliometric data.

*Inclusion Criteria*: After filtering the data based on the inclusion and exclusion criteria, a total of 1568 data were gathered. Three inclusion rules were followed: (1) Articles in which one of the keywords appears in the title, abstract, or keywords (2) The publication date ranges from 2000 to 2020. (3) Journal articles, conference papers, and book chapters. If they met all inclusion criteria, English language abstracts were included in the bibliometric review.

*Exclusion Criteria*: All documents with a core subject of agent-based modelling but not relevant to the field of economics were left out of the analysis.

*Methodology and Software*: We used the Scopus dataset in the analysis and the approaches listed above to answer our research questions. R was chosen as software for both visuals and quantitative analysis.

#### **5. Knowledge Base of ABM: Results and Findings**

*5.1. Productivity Assessment*

After the financial crisis of 2008, the agent-based model became a widely researched topic. The crisis was not predicted by DSGE models, and they also did not respond to policy questions. Since the most publications were in 2020, ABM has become a new topic in its development. ABM is also the ideal tool for experimenting with different policy scenarios during a pandemic (Figure 1).

**Figure 1.** Annual scientific output on economics-related ABM: A Scopus database analysis (2000–2020).

Countries such as the United States have played a critical role in the field's continuous progress. The authors are working with researchers from the same country as well as from other ones. The top nations and authors working on the issue of agent-based modelling in economics are shown in Figure 2. According to the findings, American researchers are putting a greater emphasis on this modelling technique and examining its possibilities for solving difficult challenges. Researchers prefer to collaborate with researchers from their own nation rather than researchers from other countries, according to stacked bar charts.

Wilensky, being the most active contributor, also described how to use agent-based simulations to answer complex questions. His writings capture the thrill of re-creating social phenomena in computer simulations to better understand them (Wilensky and Rand 2015) (See Figure 3). In Figure 4, we can see that in which year the authors were most productive and long lived. There are authors from economics who were working on agentbased modelling before the financial crisis took place. But the focus of their research was not directed to answer the policy questions to overcome the aftershocks of crisis.

**Figure 3.** Ranking of authors based on scientific output on economics-related ABM (2000–2020).

**Figure 4.** An overview of the author's output over the years (2000–2020).

Lotka's Law is one of the most fundamental bibliometric rules, and it deals with the frequency with which authors in a specific subject publish. The frequency of publishing by authors in a specific field is described by Lotka's law, presented as follows:

$$f(\mathbf{x}) = \frac{\beta}{\mathbf{x}^a}$$

where *f*(*x*) is the frequency of authors having *x* publications, and *x* is the positive integer, representing number of publications. The parameter estimates of Lotka's Law are *β* = 2.82 and *α* = 0.97. According to the findings, there are 2833 writers with a single ABM publication in economics. There are approximately 70 authors with at least five published works. Over 20 documents were released by only one contributor. Theoretical and observed frequency are depicted graphically in Figure 5.

**Figure 5.** Lotka's law of scientific productivity from 2000–2020 (authors publishing on economicsrelated ABM).

A journal's impact on a specific study topic can be measured by its publications. It can be seen from Table 1 regarding ABM research in economics, Journal of Artificial Societies and Social Simulations (JASSS) is head and shoulders above in the league table. JASSS has the highest number of publications on the theme of agent-based modelling in Scopus index journals followed by Computer environment and urban system on the league table.



The sum of published documents on ABM in the top five journals is around 450 in ten years. As outlined in Table 2, authors are ranked via the Dominance Factor (DF). (Kumar and Kumar 2008) developed the formula as,

$$\text{DF} = \frac{\text{num of multi-autbered problems of an author as first author (Nmf)}}{\text{total num of multi-autbered problems (Nmt)}}.$$

The value of dominance factor indicates collaboration in the field. A value less than 0.5, reflects a good sign for collaboration. Authors who have published nine or more publications on the theme of agent-based modelling are selected and their dominance factor is calculated by using the above formula. Sengupta and Wilensky are top authors with respect to publication number, i.e., 15 and 22, but they rank 6th and 10th, respectively. If the authors' dominance factor values are less than 0.5, this is a good sign of collaboration.

**Table 2.** Dominance factor Ranking.


### *5.2. Importance Assessment*

Table 3 is about author level metrics based on three indices. *H-index* measures the productivity as well as the impact of publication. The H-index is calculated as "author has H publications, and each publication has H or more citations". Whereas *g-index* is one variant of H-index which gives credits for highly cited authors in the data set. (Hirsch 2005), its inventor says: highly cited papers play a key role in the determination of H-index. The selected paper for the top h category are then dropped for the further determination

of the H-index over time. This means that the H-index of the subsequent years are not influenced by the papers of the top category, even if the number of citations increases over time. Value of g-index is always either equal to or greater than H-index.

The *m-index* is another form of H-index. The M index in the Table 3 shows the comparison of authors within the field of agent-based modelling but with very different career lengths. The h-index is constrained by the fact that it is time-based and fieldspecific, and it ignores highly cited works. In the early identification of young researchers, bibliometrics that account for time, such as the m-index, should be evaluated, ideally in conjunction with critical peer review. The m-index has the most potential for identifying early-stage high-potential researchers. An m-index of 1 is normal, 1–2 is above average, and >2 is exceptional, according to a suggested rule of thumb for interpreting the index (Ndwandwe et al. 2021).



We attempted to plot average citations and average total citations (see Figure 6) over time in order to study the most cited literature on the application of ABM in economics. We can readily see that total citations were much higher for work published prior to 2008, with a downward trend after that. On the other side, due to the pandemic, average citations climbed in 2020.

**Figure 6.** Citation analysis of published scientific documents.
