Metrics details This study examines the impact of the Kahramanmaraş earthquake on the BIST100 index through the application of complex network analysis The method begins by examining the correlation distance between the logarithmic returns of companies in the BIST100 index which serves as the foundation for creating filtered weighted networks This technique prioritizes robust financial connections while downplaying weaker ones enabling a thorough examination of the network topological structure and identifying noteworthy financial interactions The study employs a combination of global and local topological metrics to gain insights on the interconnections among markets in terms of public interest and the roles played by various entities It demonstrates how exogenous shocks impact the network structure and reaction Findings indicate initial notable alterations in the network framework which were subsequently alleviated by the implementation of financial rules and market processes This demonstrates the ability of financial markets to recover and maintain stability following a disaster Results offer useful perspectives on the interplay between market dynamics and the resilience of the financial system in the presence of natural disasters This may lead to a decrease in the country’s overall GDP Infrastructure vulnerabilities revealed by earthquakes can make areas more susceptible to additional damage in future disasters potentially exacerbating economic difficulties Rebuilding and recovery efforts often necessitate coordinated responses and investments in disaster resilience and preparedness Given the significant consequences that natural calamities like earthquakes can bring about it is crucial to analyze the repercussions of the Kahramanmaraş earthquake in Turkey This disaster significantly impacted key sectors such as agriculture and potentially had consequences for the Borsa Istanbul 100 index (BIST100) While earlier studies have examined various elements of the effects of earthquakes on economic indicators this study distinguishes itself by using complex network analysis to examine the impact of the Kahramanmaraş earthquake on the topological structure of Borsa Istanbul This approach enhances our understanding of the earthquake’s impact on the complex interconnections between companies listed in the index providing valuable insights into the Turkish stock market resilience and adaptability to significant external disruptions This approach is not only important for scholars and investors but also for policymakers aiming to enhance economic stability and develop effective disaster response strategies This method is particularly powerful because it captures intricate relationships and multifaceted interactions within complex systems The relevance and value of complex network analysis for understanding the impact of the earthquake on the BIST100 index are rooted in its ability to capture the interconnections and systemic effects within financial markets Unlike traditional methods that may focus on isolated financial metrics complex network analysis examines the web of relationships between different entities in the market This approach allows for the identification of key nodes and connections that play crucial roles in the overall market structure we can better understand how shocks like the Kahramanmaraş earthquake propagate through the market this method highlights systemic risks and vulnerabilities that may not be apparent through conventional analysis The ability to visualize and quantify these interdependencies provides a deeper insight into market dynamics enabling more effective risk management and policy-making aimed at enhancing market resilience and stability These networks are constructed by examining the correlations between performances of various financial instruments effectively illustrating their mutual influences A financial correlation network visually depicts nodes representing financial assets and edges representing the correlations between them The value of networks lies in their ability to uncover hidden patterns and structures within financial markets which are not apparent when examining individual assets in isolation investors and analysts can identify groups of assets that exhibit correlated movements and pinpoint key nodes with significant market influence This understanding is crucial in risk management as it helps identify systemic risks and unravel how shocks might propagate through the market these networks are valuable for portfolio diversification as they enable investors to select assets with lower interconnections reducing their portfolio’s exposure to market volatility This work is motivated by the need to understand the substantial but insufficiently investigated influence of natural catastrophes on financial systems particularly in rising markets such as Turkey The Kahramanmaraş earthquake is an exceptional instance that warrants investigation due to its significant economic impact on multiple industries Our research differs from past studies by utilizing sophisticated network analysis to examine the interconnected relationships and systemic impact within the Borsa Istanbul 100 index (BIST100) rather than solely focusing on individual financial measures or broad economic indicators after a disaster The study examines the topological structure of the financial network both before and after the earthquake It offers new insights into how external shocks like earthquakes spread across the market uncovering hidden weaknesses and methods of resilience Our approach not only enhances the current body of knowledge by providing a fresh viewpoint on market dynamics and the effects of disasters but it also assists policymakers and investors in comprehending systemic risks and formulating more efficient strategies for responding to disasters the study is notable for incorporating an advanced analytical framework to examine a crucial and current subject emphasizing the unique characteristics of the Turkish financial market reaction to a significant earthquake it is crucial to highlight the innovative use of complex network analysis in examining the resilience of financial markets specifically in reaction to a natural disaster on a developing market This study provides a distinct perspective by investigating the complex relationships inside the BIST100 index which allows for a comprehensive understanding of how earthquake shocks affect markets Unlike earlier research that primarily considered immediate economic effects using conventional financial measures this study elicits broader systemic implications Utilizing topological metrics and structural entropy allows for a more profound understanding of the network structural modifications that cannot be observed with traditional methods Our research enhances the theoretical comprehension of financial networks during challenging situations and offers practical insights for risk management and policymaking findings uncover concealed weaknesses and resilience mechanisms of the Turkish stock market This unique perspective can provide valuable insights for scholars and practitioners seeking to improve market stability in the event of future crises the study contributes significantly to the present literature on financial market dynamics and disaster impact both in terms of methodology and practical application This study uses complex network analysis to gain a thorough understanding of how the Kahramanmaraş earthquake affects the topological structure of the BIST100 index a matrix is constructed by calculating correlation distances using the logarithmic returns of businesses traded in the BIST100 index The matrix is used as a basis for creating filtered weighted networks that emphasize important financial ties while reducing the impact of less significant links The filtering step is essential because it enables a meticulous and concentrated analysis of the network topological structure revealing the most significant financial interactions the study uses several measurements to examine the network structure Measurements encompass both global and local topological characteristics offering valuable understanding of the overall market interconnection and the unique roles and effects of various entities within it structural entropy is employed to measure the degree of intricacy and diversity inside the network community structure This comprehensive methodology allows for a detailed investigation into the organization and response of the network to external disturbances The analysis integrates both graphical and econometric techniques providing a dual viewpoint on the impact of the earthquake Graphical analysis visually illustrates alterations in the network structure whereas econometric methods measure and evaluate these alterations determining their statistical significance Findings indicate that the earthquake had different impact levels on the BIST100 index effects appeared as noticeable changes in the network topological configuration these effects were alleviated or rectified by the implementation of financial rules and market mechanisms This observation highlights the ability of financial markets to recover and the success of regulatory systems in stabilizing markets after significant external disruptions Our study enhances comprehension of market dynamics and the resilience of financial institutions in the presence of natural calamities The structure of this article is as follows: In Section 2 we provide an extensive analysis of the existing literature about earthquakes and their influence on economies Section 3 outlines the methodology for acquiring a filtered correlation network and it provides details regarding the intricate network metrics employed we provide an overview of the dataset and describe findings The part also includes explanations for graphical findings Section 5 presents the statistical tests conducted and their corresponding interpretations Their research shows that while certain natural disasters cause immediate and noticeable changes in market behavior others have minimal or no significant impact on financial markets His analysis indicates that the long-term economic damage can be substantial with recovery efforts often falling short in restoring pre-disaster economic conditions Findings underscore the importance of comprehensive disaster impact assessments that include both direct and indirect costs as well as the need for effective mitigation strategies to minimize long-term economic disruptions in developing countries but the impact is notably greater in developing countries Despite having lower total damage costs as a percentage of GDP developing countries experience significant economic disruptions underscoring their higher vulnerability and the critical need for disaster risk financing tools This underscores the need for more focused research on the specific impact of natural disasters on regional financial markets like the BIST-100 index to enhance understanding and inform better disaster response strategies The literature also lacks detailed analyses of systemic risks and market resilience using advanced methodologies like complex network analysis the present study utilizes complex network analysis to investigate the ramifications of the Kahramanmaraş earthquake on the Borsa Istanbul 100 index By leveraging this advanced analytical framework the study aims to deliver a more nuanced comprehension of the intricate market dynamics and the resilience mechanisms activated in response to natural disasters Our methodological approach not only deepens the scholarly understanding of how external shocks affect financial systems but also provides critical insights for investors seeking to navigate post-disaster market conditions findings offer practical guidance for policymakers dedicated to bolstering economic stability and formulating robust disaster response strategies By highlighting the specificities of the Turkish stock market reaction to a significant seismic event this research contributes to a more comprehensive body of knowledge that can inform both theoretical models and real-world applications in disaster risk management and economic resilience planning Financial correlation networks provide a complex framework for visualizing and analyzing the intricate interconnections among various financial assets These networks reveal the connections between price movements of different assets showing how the fate of one asset can impact or resemble another By understanding the scale and organization of these connections investors and analysts can improve their capacity to evaluate the inherent risk in a system identify possible paths of contagion during financial crises and develop portfolio strategies that are more robust and long-lasting In the context of a progressively interconnected global financial system correlation networks are crucial for understanding the complex interconnections and hidden vulnerabilities in large sets of financial data they can be transformed into a correlation matrix This matrix can then be filtered or processed further to create a network Equation (1) calculates the difference in the daily closure price of stock \(i\) at time \(t\) To calculate the Pearson correlation coefficient between the stocks \(i\) and \(j\) where \(\left\langle \cdot \right\rangle\) is the temporal average. The Pearson correlation coefficient provided in Eq. (2) quantifies the degree of linear association between two financial time series It measures the degree to which fluctuations in one series can be represented by a linear relationship with the fluctuations in the other series we assign weights to network edges and measure the strength of relationships we consider the complete graph where all nodes are connected by an edge and weighted using the correlation distance function When there is a strong positive-negative correlation the edge weight value will gradually approach zero weak correlation values will cause the edge weight value to approach \(\sqrt{2}\) The weight function is determined by the distance with smaller weights indicating stronger relationships Using this method allows for the utilization of extremely effective filtering techniques allowing for the integration of different association measures without requiring the assumption of multivariate normality in the data we use the TMFG method for the filtration of networks are networks that can be represented on a sphere without any overlapping edges though they are often depicted with crossing edges for simplicity utilizing planar networks to visualize financial correlation networks through the TMFG method provides a well-rounded approach that maintains both structural integrity and visual clarity To quantify structural changes on the BIST correlation network These measurements represent the local or global topological characteristics of the filtered network denoted as \({G}_{F}=({V}_{F},{E}_{F},{d}_{C})\) Global Efficiency (GE) is a crucial metric in comprehending the dynamics and interconnectedness of financial systems especially within the context of financial correlation networks where \({\gamma }_{{ij}}\) represents the shortest path between nodes \(i\) and \(j\) offers a nuanced view of financial networks nodes usually represent different entities like individual stocks and the links show correlation levels between entities Applying GE allows for a precise measurement of the efficiency of information sharing or transmission among entities When a financial network operates with high global efficiency it indicates a market that is closely connected and where information spreads quickly and widely This ultimately results in a greater level of market integration as disruptions in one area of the network can rapidly spread to other areas a decrease in global efficiency could suggest the presence of market segmentation or obstacles to information flow This can result in localized risks and inefficiencies in pricing global efficiency plays a crucial role in examining market stability and the spread of crises fluctuations in GE can indicate the interconnectedness of different sectors and markets This is of utmost importance for regulators and policymakers as it helps in pinpointing possible areas of weakness in financial systems The metric is also useful for constructing diversified investment portfolios Through a thorough analysis of the network global efficiency investors can gain valuable insights into the level of interconnectedness among various assets This knowledge enables them to make well-informed decisions aimed at minimizing their exposure to systemic risk GE can be utilized to analyze the effects of financial regulations and policies initiatives that focus on enhancing transparency and promoting information sharing in markets can enhance global efficiency Examining fluctuations in GE may offer valuable insights into the efficacy of measures for promoting a cohesive and streamlined market When it comes to financial correlation networks it is essential to grasp the local structure and interconnectedness among financial entities The average of the weighted clustering coefficient (\({\bar{C}}_{w}\)) is crucial in this analysis This metric assesses the clustering of nodes by considering the weights of the edges connecting them Weights can be used to indicate the strength or intensity of correlations or financial relationships Calculating \({\bar{C}}_{w}({G}_{F})\) requires taking the average of the weighted clustering coefficients for all nodes in the network the coefficient \({C}_{w}(i)\) is calculated using which requires finding the geometric mean of the subgraph edge weights for neighboring nodes \({d}_{i}\) represents the degree of node \(i\) which reflects the number of direct connections it possesses \({w}_{{ij}}\) denotes the weight of the edge connecting nodes \(i\) and \(j\) indicating the intensity of their financial relationship Having a high mean weighted clustering coefficient in a financial network indicates a substantial level of local interconnectivity Nodes often come together in close-knit groups suggesting that entities within these clusters may exhibit similar market behaviors or risk profiles This phenomenon suggests that the market is resilient as these entities are likely to share information and have an impact on each other’s financial decisions \({\bar{C}}_{w}({G}_{F})\) provides valuable insights into the localized structure of \({G}_{F}\) This concept measures the level of heterogeneity among nodes suggesting that nodes with similar financial attributes or functionalities are more interconnected The selection of algorithm is influenced by network attributes such as size Our main objective is to analyze algorithms that categorize network nodes into separate communities ensuring that each financial entity is linked to only one sector or group This helps to enhance clarity and organization we will focus on a network \({G}_{F}\) in which \(|{V}_{F}|=N\) and designate \({\boldsymbol{A}}\) as the algorithm for community detection By applying \({\boldsymbol{A}}\) to \({G}_{F}\) the partitioning of nodes into communities can be mathematically represented using an \(N\)-dimensional vector \(\vec{\sigma }\) the community to which node \(i\) was assigned is denoted by the \(i\)-th component \({\sigma }_{i}\) \(\vec{\sigma }\) accepts values between 1 (community one) and \(M\) (the total number of detected communities) is computed using the partition \(\vec{\sigma }\) and the community size \(\left|{c}_{i}\right|\) to represent the proportional magnitude of the community in \({G}_{F}\) The value obtained by applying Shannon entropy to the probability vector is subsequently referred to as the structural entropy of the network \({G}_{F}\) (Almog and Shmueli, 2019) Modularity Maximization is a community detection method that revolves around the concept of modularity Modularity \(Q\) of a network partition is a scalar value that measures the density of edges within communities compared to edges between communities where \({w}_{{ij}}\) represents the edge weight between nodes \(i\) and \(j\) \({d}_{i}\) and \({d}_{j}\) are the degrees of nodes \(i\) and \(j\) \(m\) is the total weight of all edges in the network and \(\delta ({c}_{i},{c}_{j})\) is a delta function that equals 1 if nodes \(i\) and \(j\) are in the same community and 0 otherwise When analyzing financial correlation networks it is possible to use Modularity Maximization to detect clusters or communities of entities that have stronger relationships with each other compared to the rest of the network This can be particularly useful when nodes represent financial entities like stocks or market indices and edges represent the correlation between them the network can be efficiently divided into clusters of stocks or sectors that exhibit comparable market behavior or risk profiles Understanding market dynamics and detecting subgroups that may have co-movements in market trends or crises is crucial for portfolio diversification and risk management The Leading Eigenvector method for community detection, developed by Newman (2013) is based on the eigenvector of the largest eigenvalue of the network modularity matrix The modularity matrix \(B\) is the following: where terms are as defined for the Modularity Maximization method The network is divided into communities based on component signs of the leading eigenvector within this matrix Utilizing the Leading Eigenvector method in financial correlation networks is crucial for uncovering hidden structures and relationships that may not be readily apparent This method can provide insights into the segmentation of financial markets highlighting possible areas of market stability or vulnerability Through an analysis of the leading eigenvector financial analysts can uncover valuable information about the primary drivers within the network This includes identifying key stocks or sectors that wield significant influence over the rest of the market These insights are extremely valuable when it comes to developing investment strategies and understanding the complex relationships between various financial entities One of the key reasons for focusing on the Kahramanmaraş earthquake is its timing and the context of the Turkish economy The country has been experiencing various challenges The earthquake added another layer of complexity to an already strained economic environment providing a unique opportunity to analyze how an exogenous shock interacts with existing economic vulnerabilities This context allows for a more nuanced understanding of how financial markets react under compounded stress conditions which is important for developing robust risk management and mitigation strategies The economy of this particular region is characterized by diversity and many prominent sectors such as textiles Kahramanmaraş contributes to 36% of Turkey’s yarn production while Gaziantep serves as a vital center for carpet production Adana and Hatay are notable regions for the cultivation of citrus fruits and the production of iron and steel has also established itself as a significant contributor to the tourism and cuisine sectors The earthquake has raised significant concerns about inflation The increase in demand for different commodities and services during the period of economic recovery is expected to cause prices to rise predictions indicated that inflation would be around 20% by the end of the year the aftermath of the earthquake is anticipated to further increase this rate the surge in demand for rehabilitation and relief efforts may result in volatility in the exchange rate of the Turkish Lira therefore exacerbating economic difficulties A time series is generated to encompass a period of 6 months prior to and 6 months following February 6 we chose 22 August 2022 as the initial date of our time series the BIST100 index was not subjected to trading activities from February 9 to February 14 correlation distance functions assume that the length of time series is identical the analysis omitted companies that commenced trading in the BIST100 index after August 8 examinations are conducted on a grand total of 92 companies To examine the impact of the Kahramanmaraş earthquake on the structure a technique called sliding windows with a length of 5 is used on the time series data \({G}_{F}\) networks are generated using TMFG filtering which is determined by the correlation distances of the time series and consequently a total of 236 filtered network \({G}_{F}\) are acquired The figure on the left is proportionally adjusted in size while the one on the right is in its original size Changes of values of structural entropy obtained by modularity maximization Changes of values of structural entropy obtained by Leading Eigenvector If we assume that a similar trend was observed in the structural entropy obtained through modularity maximization it will strengthen the idea that community structures within the BIST100 network are resistant to shocks or that market dynamics facilitate a quick return to equilibrium even after substantial external disruptions along with the previous examination of structural entropy provides a complete perspective on the robustness of community structures inside the BIST100 index The statement emphasizes the market capacity to sustain equilibrium in its intricate structure as assessed by the dispersion of its components among various communities even when confronted with a significant external disturbance such as the Kahramanmaraş earthquake It can assist in determining whether the earthquake had a significant impact on network measures and in understanding the duration and nature of this impact the NARDL model excels at capturing data asymmetries which is highly significant when considering an earthquake Understanding the impact of such a shock requires considering non-linear and asymmetric effects The market response to negative shocks may differ significantly from its adjustment to positive changes one can reveal intricate connections and gain a deeper understanding of market strengths and weaknesses one can analyze any changes in the time series which is crucial for understanding how the market behaves before and after major events such as the earthquake It can help determine whether the event has significantly impacted market dynamics or if the observed changes are within the normal range of market fluctuations Table 1 presents the NARDL test results for GE. The model coefficients in Table 1 reveal a noteworthy negative correlation for \({Tpr}{e}_{1}\) suggesting that past values of GE have a considerable impact in the opposite direction on current GE changes One way to interpret this is as the network’s inclination to return to a typical level of efficiency while another perspective sees it as a sign of a major structural change caused by the earthquake it appears that the coefficients for \({Tpos}{t}_{p}\) and \({Tpr}{e}_{n}\) do not hold much statistical significance This implies that any changes in GE following an earthquake do not have an immediate or noticeable impact on the network structure in the short term The lack of significance in the results is consistent with the absence of any significant changes in the structural entropy and weighted clustering coefficients shown in the figures provided earlier network structure remained resilient or quickly adapted after the earthquake it seems that there might be some concerns with the distribution of residuals This is supported by the low p-value of the JB test Based on the results of the LM test and the ARCH test there is no evidence of autocorrelation or autoregressive conditional heteroskedasticity in the model The high p-values indicate that these issues are not a concern Based on the results of the cointegration test there is a strong long-term relationship between the pre- and post-earthquake GE measures providing statistical evidence to support this conclusion It can be inferred that even with the occurrence of the earthquake the long-term behavior of the GE is still influenced by its past values Both the short-run and long-run asymmetry tests produce high p-values indicating no significant asymmetrical response in the network GE to the earthquake it appears that the earthquake had a significant impact on the BIST100’s GE in the short term the long-term topological structure of the BIST100 may still be influenced by the same underlying processes as it was before the earthquake occurred Table 2 displays the NARDL test results for \(\bar{C}\) It is evident that the coefficients for \({Tpr}{e}_{1}\) are both significant and negative This suggests that the \(\bar{C}\) prior to the earthquake had a substantial and adverse effect on the subsequent \(\bar{C}\) values It appears that there was a pattern of higher clustering coefficients being followed by lower coefficients which may suggest a tendency towards disintegration in the network clusters or a mean-reverting behavior of the \(\bar{C}\) over time The coefficients for \({Tpos}{t}_{p}\) and \({Tpos}{t}_{n}\) which represent the post-earthquake \(\bar{C}\) values the earthquake did not have a noticeable impact on the overall network structure it seems that there might be some concerns regarding residuals distribution suggested by the low p-value of the JB test (which indicates non-normality) According to the results of the LM test and the ARCH test there is no evidence of autocorrelation or autoregressive conditional heteroskedasticity The F-statistic from the cointegration test indicates that there is a long-run relationship between the pre- and post-earthquake \(\bar{C}\) measures The F-statistic is not above the critical values We infer either that the earthquake did not have a lasting impact or that the correlation is not significant enough to be identified by this analysis Based on the results of the short-run asymmetry test and the long-run asymmetry test it appears that there is no notable short-run asymmetry there is a significant long-run asymmetry in the network \(\bar{C}\) response to the earthquake There appears to be an imbalance in the long term suggesting that the earthquake had a greater impact on the structure of the network over time but only in particular directions or under specific circumstances the NARDL analysis indicates a notable and rapid return to average levels in the network \(\bar{C}\) prior to the earthquake the earthquake did not have a significant short-term impact on the \(\bar{C}\) there are indications of long-term asymmetry suggesting that the earthquake may have had more nuanced and lasting impact on network structure NARDL test results for \(H(\vec{P})\) via modularity maximization are displayed in Table 3 Both the constant term and the \({Tpr}{e}_{1}\) coefficient are highly significant suggesting a robust baseline level for structural entropy and a negative correlation with the entropy of the previous period It is possible that there is a mean-reverting dynamic or a baseline shift in structural entropy after the earthquake It is worth mentioning that coefficients related to the post-earthquake period (\({Tpos}{t}_{p}\) changes in structural entropy immediately after the earthquake were not significantly different from the expected dynamics before the earthquake In line with the visual analyses from previous figures data imply no significant alterations in the network topological structure immediately after the earthquake Diagnostic tests suggest a potential problem with residuals normality the LM test and the ARCH test indicate that there is no noteworthy autocorrelation or conditional heteroskedasticity in the residuals hence the other model assumptions remain valid the F-statistic is significantly higher than the asymptotic critical values indicating a strong long-term relationship between the pre- and post-earthquake measures of structural entropy It is possible to infer that the earthquake impact on the network topological complexity was either absorbed or eventually returned to the long-term trend Both the short-run asymmetry test and the long-run asymmetry test do not provide evidence of significant asymmetrical responses of structural entropy to the earthquake It appears that the earthquake had a consistent impact on the network topological structure regardless of the direction of change in structural entropy the NARDL analysis indicates that the earthquake had a notable initial impact on the structural entropy of the BIST100 index and it is possible that the long-term topological structure of the network remained unchanged The network structural complexity appears to be unaffected by the earthquake as it continues to exhibit its original dynamics in the aftermath Test results for \(H(\vec{P})\) are shown in Table 4 The negative coefficient for \({Tpr}{e}_{1}\) indicates a significant and enduring impact of the structural entropy of the prior period on the current period Data indicate that the network consistently returned to its previous entropy levels demonstrating a pattern of mean reversion in the structural entropy of the BIST100 network over time Coefficients related to the post-earthquake data (along with their lags) do not show any statistical significance it appears that the earthquake did not have a noticeable effect on the structural entropy of the BIST100 index according to the linear framework used in the NARDL model there might be a concern regarding the normality of residuals This is indicated by the JB test statistic although the p-value is not sufficiently low to reject normality at conventional significance levels Based on statistical tests such as the LM test and the ARCH test there is no evidence to suggest any issues with autocorrelation or conditional heteroskedasticity This implies that model errors are independently and identically distributed This suggests a strong indication of a long-term relationship between the pre- and post-earthquake measures of structural entropy It is evident that the earthquake has not only caused immediate effects but also has a lasting impact on the structural network characteristics Historical values play a significant role in determining network entropy Both the short-run and long-run asymmetry tests produce p-values that are quite high suggesting that there is no substantial evidence of significant asymmetrical effects of the earthquake on the network structural entropy in either the short or long term our findings indicate that although the earthquake may have had a notable immediate impact on the BIST100 index its long-term influence on the network structural entropy is not apparent The network structural complexity seems to be resilient to the impact of the earthquake preserving its original dynamics in the long run The earthquake significantly affected the BIST100 network efficiency with a temporary spike in Global Efficiency indicating rapid market response and increased interconnectedness This impact was short-lived as the network quickly stabilized The mean weighted clustering coefficients and structural entropy remained constant showing that local clustering and community structures withstood the disturbance The NARDL model supports these observations indicating no lasting alteration in the BIST100 structure This resilience reflects the market ability to absorb and adjust to unexpected shocks maintaining its fundamental structure while responding dynamically on a larger scale Findings underscore the importance of real-time monitoring systems and risk management techniques for market stability Regulatory agencies should use advanced monitoring to detect abrupt fluctuations and enhance crisis preparedness Promoting investment diversification and supporting resilient industries can further strengthen market robustness The stability of financial markets leads to a positive impact on the real economy due to their ability to withstand and operate effectively even in the face of external shocks This resilience ensures that financial institutions can continue to function smoothly maintaining the flow of capital and liquidity necessary for economic activities By assuring the appropriate allocation of financial resources and effective risk management stable financial markets foster an environment of confidence among investors This confidence encourages investment in businesses and infrastructure the stability of financial markets upholds broader economic stability by preventing severe disruptions that could cascade into other economic sectors This interconnectedness means that when financial markets are robust they provide a solid foundation for sustained economic growth facilitating a virtuous cycle where stable markets support economic activities which in turn contribute to the further stability and growth of financial markets the stability of financial markets becomes a cornerstone for a thriving and resilient real economy Key processes and regulations that influence market responses include circuit breakers and trading halts which temporarily stop trading during extreme volatility to allow time for information dissemination and prevent panic selling Capital adequacy requirements ensure that financial institutions maintain sufficient capital buffers to absorb losses and continue functioning during crises Liquidity provision by central banks during shocks ensures the smooth functioning of financial transactions and prevents liquidity crunches Regular stress testing and contingency planning help identify vulnerabilities and prepare for potential shocks while transparent and timely communication strategies from regulators and market leaders help manage investor expectations and reduce uncertainty Robust risk management frameworks within financial institutions are crucial for identifying These processes and regulations can either prompt or slow down market responses to severe shocks well-designed circuit breakers can quickly stabilize markets while insufficient capital adequacy requirements may hinder recovery efforts policymakers in other markets can design strategies to enhance the resilience and adaptability of their own financial systems Investors need education on market risks and effective communication during crises to prevent panic Ongoing research is vital to understand market dynamics and the impact of external shocks Future studies should explore diverse econometric and network analytic techniques and long-term effects to inform policymaking and enhance financial system resilience the BIST100 response to the earthquake highlights the need for improved market monitoring and continuous research to develop robust financial markets capable of recovering from disruptions thereby supporting a stable and thriving real economy The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request Gaban L (2023) Path-based visibility graph kernel and application for the Borsa Istanbul Stock Network Alharbi FM (2022) How do the remittance outflows react to economic development in Saudi Arabia Shmueli E (2019) Structural entropy: monitoring correlation-based networks over time with application to financial markets Aydinbas G(2023) Sosyoekonomik Boyutuyla Türkiye’de Depremler Üzerine Bir İnceleme: Kahramanmaraş Depremi Örneği (Astudy on earthquakes in Turkey with a socioeconomic dimension: the Kahramanmaraş earthquake example) Sosyal Bilimlere ÇokYönlü Yaklaşımlar: Tarih Nichita A (2022) Coarse graining on financial correlation networks Madkour J (2023) Portfolio optimization using minimum spanning tree model in the Moroccan Stock Exchange Market Hwang DU (2006) Complex networks: structure and dynamics Kirman A (2015) Complex financial networks and systemic risk: a review Dynamic modeling and econometrics in economics and Finance Sanders M (2019) The economic impacts of natural disasters: a review of models and empirical studies Aste T (2022) Dependency structures in cryptocurrency market from high to low frequency Hu Y (2019) Community detection method based on node density Toroczkai Z (2004) Emergence of complexity in financial networks Acevedo L (2022) The impact of natural disasters on economic growth Kwapil TR (2018) Network structure of the Wisconsin Schizotypy Scales–Short Forms: examining psychometric network filtering approaches Dzator M (2021) Climate change and natural disasters: macroeconomic performance and sustainable development Economic effects of natural disasters: theoretical foundations Hric D (2016) Community detection in networks: a user guide Parisi L (2018) Corisk: credit risk contagion with correlation network models Tunçel MB (2023) The impact of 2023 Turkey earthquake on İstanbul Stock Market: evidence from Fourier volatility spillover test Hamurcu C (2023) Vulnerability of stock markets to earthquakes/The 7.7 and 7.6 scale earthquakes in Turkey in 2023: sectors of non-metallic mineral products Basic metals and construction & infrastructure Jaramillo CR (2009) Do natural disasters have long-term effects on growth Lahiani A (2021) Are cryptocurrencies a backstop for the stock market in a COVID-19-led financial crisis Abidin S (2022) Analyzing the asymmetric effect of renewable energy consumption on environment in STIRPAT-Kaya-EKC framework: a NARDL approach for China Joseph IL (2022) The effect of natural disaster on economic growth: evidence from a major earthquake in Haiti Kalayci I (2023) Destroying effects of twin earthquakes on the economy and agriculture in Turkey: lessons of Japan Mamat M (2022) Insurance as an alternative for sustainable economic recovery after natural disasters: a systematic literature review Tol RS (2017) Impact of natural disasters on financial development Liczmanska Kopcewicz K (2021) A dynamic linkage between financial development energy consumption and economic growth: evidence from an asymmetric and nonlinear ARDL model Batool Z (2022) The role of infrastructure and food security to mitigate theloss of natural disasters Valckx K (2014) Natural disasters and economic growth: a meta-analysis van Bergeijk PA (2014) Natural disasters’ impact factors of resilience and development: a meta-analysis of the macroeconomic literature Araujo I (2022) Risk caused by the propagation of earthquake losses through the economy Boto-Garcia D (2023) The effect of natural disasters on hotel demand supply and labour markets: evidence from the La Palma volcano eruption Deng Y (2020) Research on the knowledge association reasoning of financial reports based on a graph network Mantegna RN (1999) Hierarchical structure in financial markets Eur Phys J B-Condens Matter Complex Syst 11:193–197 Aste T (2016) Network filtering for big data: triangulated maximally filtered graph Michis AA (2022) Multiscale partial correlation clustering of stock market returns Rehman A (2023) Economic growth drivers in Romania: evidence from a NARDL analysis Mitchell M (2006) Complex systems: network thinking Girvan M (2004) Finding and evaluating community structure in networks Newman ME (2003) The structure and function of complex networks Newman ME (2013) Spectral methods for community detection and graph partitioning Naoaj MS (2023) From catastrophe to recovery: the impact of natural disasters on economic growth in developed and developing countries Noy I (2009) The macroeconomic consequences of disasters Sen S (2019) Economic impact of natural disasters: an empirical re-examination Roh JJ (2013) Supply chain lessons from the catastrophic natural disaster in Japan Smith RJ (2001) Bounds testing approaches to the analysis of level relationships Bullock S (2017) Hidden structures: using graph theory to explore complex system of systems architectures Proceedings of the Complex Systems Design and Management Deniz D (2023) Field reconnaissance on seismic performance and functionality of Turkish industrial facilities affected by the 2023 Kahramanmaras earthquake sequence Yavuz O (2015) Econometric analysis of natural disasters macroeconomic impacts: an analysis on selected four OECD countries investor sentiments and stock market reactions: evidence from Turkey-Syria earthquakes Fernandes A (2023) Tourism industry at times of crisis: a bibliometric approach and research agenda Selçuk S (2023) Kahramanmaraş Depremlerinin Ekonomiye Etkisi (The effect of Kahramanmaraş earthquakes on the economy) Nam K (2022) Forecasting price movements of global financial indexes using complex quantitative financial networks Shimada G (2022) The impact of climate-change-related disasters on Africa’s economic growth and conflicts: can humanitarian aid and food assistance offset the damage Edrisi Tabriz Y (2022) A mutual information-based network autoregressive model for crude oil price forecasting using open-high-low-close prices Katsavounis S (2021) A survey on the recent advances of deep community detection Strogatz SH (2001) Exploring complex networks Skidmore M (2007) Economic development and the impacts of natural disasters Mantegna RN (2005) A tool for filtering information in complex systems Üzümcüoğlu D (2023) Factors to consider in the reconstruction and design of an earthquake-devastated region: a case study of the Kahramanmaraş earthquake Koumoutsakos P (2022) Multiscale simulations of complex systems by learning their effective dynamics Hoang VN (2002) Impact of natural disasters on the efficiency of agricultural production: an exemplar from rice farming in Sri Lanka Valadkhani A (2004) Measuring the impact of natural disasters on capital markets: an empirical application using intervention analysis Zhang H (2020) Development of stock networks using part mutual information and Australian stock market data Huang C (2019) Dynamic properties of foreign exchange complex network Cheong SA (2021) Understanding changes in the topology and geometry of financial market correlations during a market crash Uddin GS (2023) Analysing and forecasting co-movement between innovative and traditional financial assets based on complex network and machine learning Download references “1 Decembrie 1918” University of Alba Iulia National Institute for Research and Development in Constructions Urbanism and Sustainable Spatial Development “URBAN INCERC” All authors contributed to the writing and revision of the paper The authors declare no competing interests This article does not contain any studies with human participants performed by any of the authors As no studies with human participants have been performed for this article Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations Download citation DOI: https://doi.org/10.1057/s41599-024-04115-w Anyone you share the following link with will be able to read this content: a shareable link is not currently available for this article Please enable JS and disable any ad blocker Euronext Live Markets Comunicati stampa Borsa Italiana non ha responsabilità per il contenuto del sito a cui sta per accedere e non ha responsabilità per le informazioni contenute Borsa Italiana non intende sollecitare acquisti o offerte in alcun paese da parte di nessuno Sarai automaticamente diretto al link in cinque secondi Putin and Lukashenko speak on Union State ties and united historical memory at Volgograd forum Russia proposes joint aviation venture with Belarus as part of expanded industrial cooperation Russian and Belarusian agents accused of plotting attacks on exiled Belarusians in Lithuania Belarus now ‘untouchable’ under Russia’s nuclear umbrella Trump dollar drama sparks global power shift as BRICS countries circle COMMENT: Trump’s tariff drive hits emerging markets as growth and inflation forecasts slide – Oxford Economics Hungarian Commissioner delays EU plan to curb Russian energy dependence COMMENT: Is Russia’s economy headed for hard or soft landing EU allies prepared to back Moldova’s EU accession bid without Ukraine if Hungary’s veto not lifted Post-war order teetering – rights watchdog but playing second fiddle to the Northern Route Think-tank urges EU to consider 'Big Bang' enlargement and sweeping reforms UK insect numbers plummet by two thirds in three years EU swaps China for Russia in procurement of rare earth metals Trump policies won’t harm Emerging Europe badly – and will actively benefit Russia Poland’s manufacturing conditions improve for third consecutive month in April but outlook weakens Hungary and Slovakia warn EU against scrapping unanimity rule over Ukraine membership bid Press freedom declines sharply in several Balkan countries Albania’s 2025 general election to test democratic institutions Western Balkan states increase use of surveillance Bosnia secures €80mn World Bank loan to finance coal phase-out and energy transition Spanish blackout caused by extreme weather and a major renewables failure for the first time Moldova's government hopes for 2% economic growth this year George Simion takes decisive lead in Romanian presidential election first round Serbian president stable after falling ill during US visit COMMENT: Turkish opposition’s efforts to bring down Erdogan regime are just not working Turkey’s inflation dips further to 37.9% in April Turkey’s opposition leader assaulted in Istanbul COMMENT: Embracing “Greater Central Asia” has become a strategic imperative for US Russia and Europe vie for investment in Central Asia Russia removes Taliban from terrorist list acknowledging it as legitimate leadership of Afghanistan Taliban seek $100mn from World Bank to restart work on CASA-1000 electricity transmission project COMMENT: Europe could benefit from partnering with Armenian tech innovation Armenia courts African countries to supply diamond industry Tensions flare in Azerbaijan–Russia ties despite Patriarch’s visit Europe’s quest for Central Asian resources meets reality First commercial flight arrives in Georgia’s separatist Abkhazia region after 33 years “Lost” Ottoman era painting reappears at Sotheby’s London PMI shows ongoing expansion in Kazakh manufacturing sector in April MACRO ADVISORY: Kazakhstan’s high stakes oil gambit Gold mine no cash cow for Kyrgyz government Kyrgyz PM gets polite brush off by US business executives China exports Xi Jinping Thought to Mongolia Boosting Mongolia’s democracy with reform remains an uphill battle China spikes Gazprom gas export plan in Central Asia Turkey’s foreign policy falling apart on all fronts First-ever Uzbek-language startup guidebook series launched to support aspiring entrepreneurs Oil is a major source of income for eight African countries – OWID Sanctioned Russian cargo ship sinks in Mediterranean after explosion Russia's budget oil breakeven price world’s second lowest as oil revenues recover Southeast European countries look to Algeria to diversify energy supplies Sub-Saharan Africa is home to 16% of the global population but 67% of extreme poverty - OWID Russia's expanding pull in Africa is pushing Western influence out EBRD delivers 26% expansion in investments in 2024 Global access to energy starts to fall for the first time in a decade Saudi Arabia hosts kingdom's first Africa summit Putin at 2023 Africa-Russia summit: Wiping debts Botswana throws the diamond industry a lifeline Nelson Mandela worried about natural diamonds Botswana’s 2,492-carat diamond discovery is golden opportunity to replicate legendary Jonker diamond's global legacy Kamikaze marketing: how the natural diamond industry could have reacted to the lab-grown threat Guns and gold: how two coups reshaped Burkina Faso’s mining sector Russia’s Rosatom to support nuclear projects across Africa at AEW2024 Chase and HSBC reportedly unwittingly processed payments for Wagner warlord Prigozhin Burkina Faso the latest African country to enter nuclear power plant construction talks with Russia IMF: China’s slowdown will hit sub-Saharan growth Moscow unlikely to give up Niger toehold as threat of ECOWAS military action looms CAR mercenary becomes first African to die in Ukraine conflict Russia to capitalise on Africa’s growing energy demand Minerals for security: can the US break China’s grip on the DRC BRICS seeks new global alignment at Rio ministerial summit BRICS labour meeting in Brasilia tackles AI impact CONFERENCE CALL: Aurora Women in Tech Global Awards France call for immediate ceasefire in Gaza amid trilateral summit in Cairo Can Eritrea’s mining sector flourish under autocratic rule SDS storms fed by sand and dust equal in weight to 350 Great Pyramids of Giza US Treasury targets Iran's National Oil Co HALLIGAN: Trump's tariffs are either madness or genius – it could go either way Brazil hosts first BRICS Sherpas meeting with expanded membership From oil to minerals: Gabon’s ambitious mining transition Guinea grants final approvals to Rio Tinto for $11.6bn Simandou iron-ore project PANNIER: Prince Karim Aga Khan IV obituary New Aga Khan takes helm of Ismaili Shi'ite Muslims Liberia’s mining resurgence: from conflict to opportunity US adds 17 Liberian-flagged bulk carriers and oil tankers to Russian sanctions-busting blacklist Force majeure at Libya’s Zawiya Refinery threatens exports and oil expansion plans facing loss of Syrian base for Africa operations seen turning to war-torn Sudan or divided Libya Libya’s mineral riches: unlocking a future beyond oil Russia funding war in Ukraine via illegal gold mining in Africa – WGC report Ukraine claims it was behind massacre of Wagner Group mercenaries in Mali Uzbekistan's Startup Garage arrives in North Africa with Casablanca launch EBRD warns of risks for emerging markets pursuing industrial policies Can Morocco's phosphate wealth put it at the centre of the global battery supply chain Foreign agricultural workers flock to South Africa as sector expands Russia's nuclear diplomacy binding emerging markets to the Kremlin Can Niger's military junta seize the country's uranium opportunity Disaster season: heat waves sweep the world – in charts and maps Niger and beyond: Francophone credit delivers coup de grâce From blood diamonds to critical minerals: Sierra Leone’s mining transformation 10 African countries to experience world’s fastest population growth to 2100 Global press freedom indicator slips – Statista SITREP: Middle East rapidly destabilised by a week of missile strikes Gold and guns: how civil war keeps Sudan's mineral riches in the ground Colombian mercenaries trapped in Sudan’s conflict COMMENT: What impact will a raft of protests across the Emerging Markets have Tunisian President Kais Saied re-elected for second term Sri Lanka's LTL Holdings targets African power sector WHO declares "global public health emergency" owing to mpox outbreak in Central Africa Climate crisis-driven global food security deteriorated between 2019 and 2022 and is even affecting the US Reserve Bank of India signals further rate cuts as inflation eases and growth outlook weakens Singapore eases monetary policy again as growth outlook dims and tariff pressures mount Russia and Indonesia work on direct bank settlement framework to bolster trade Bank of Korea expected to lower rates in April despite currency slide IMF: Global debt to exceed pandemic levels soon Chipping point – the climate toll of AI chip-making in East Asia Indonesia’s coal politics: maintaining a strategic market price despite Chinese resistance India moves to cut reliance on Chinese solar modules and PV cells Google backs Taiwan’s geothermal future in landmark Asia-Pacific energy deal mobility infrastructure in western Himalayas India eyes deeper trade ties with trusted economies Record gold prices to hit sales volume of Indian jewellery retailers Beijing VS Washington on rare earths – and why the US is losing Vietnam gold prices hit record highs as global bullion rallies Indonesia to launch national data centre by April 2025 Tehran-based ride-hailing giant sets record with nearly 6mn daily trips Chinese threat to submarine cables emerges in Indo-Pacific BMI: New US tariffs set to reshape economic landscape across Middle East US deploys six B-2 bombers to Indian Ocean ahead of Iran standoff Arab League and UN condemns brutal Israeli airstrikes on Gaza COMMENT: Gulf states court Russia but stop short of strategic shift Fourth round of Iran-US talks to be held in Oman Iran bans skateboards and scooters in public spaces Iraq facing worst drought in history as water resources dwindle and climate crisis intensifies Unemployed Iraqi engineers block Lukoil’s site in Basra Kazakhstan will put national interests over OPEC+ oil limits Israel announces operation Gideon's Chariots to seize control of Gaza if hostage deal deadline missed Israel launches airstrikes on Yemen port city Israel to respond against Yemen for missile on Tel Aviv airport COMMENT: Israel is attempting to destabilise Syria Ryanair and Wizz Air to launch direct flights to Jordan in April Netanyahu to discuss Turkey-Israel relations with Trump at White House on April 7 Syrian foreign ministry urges Kuwait to reopen embassy in Damascus Middle East markets suffer massive losses as Trump tariff effects continue Deadly Israeli air raids heighten risk of renewed Lebanon-Israel conflict Israel launches airstrikes on Beirut's southern suburb amid Lebanon escalation Missile sirens sound in northern Israel amid Lebanon border tensions Is Oman the 'Switzerland of the Middle East' Iran FM to visit Russia ahead of second round of US talks Aid ship to Gaza hit by drone strikes near Malta in international waters COMMENT: Is Israel's Prime Minister committing political suicide Saudi Arabia deploys AI to streamline Hajj Saudi Arabia's Riyadh Air plans first flights for late 2025 Israeli fighter jets strike targets near the Syrian presidential palace Montenegro mega beach deal reignites debate over UAE investments in Western Balkans Protesters against UAE beach development deal target Montenegrin parliament Turkey says it wants no confrontation with Israel in Syria ECLAC cuts Latin America growth forecast amid global uncertainties LatAm set to be world's slowest-growing region as World Bank slashes forecasts Trump’s trade tariffs push Mexico towards recession and weigh on regional growth Latin American leaders blast US tariffs at CELAC summit as China offers economic lifeline Why investors are turning their gaze to Latin America after Trump tariffs Gazprom abandons Bolivia's Azero project after 16 years of unfulfilled promises Brazilian police foil bomb plot targeting Lady Gaga's 2mn people Rio concert IMF suspends Colombia's flexible credit line amid fiscal concerns Colombia’s import surge deepens trade deficit Latin America to miss more than three-quarters of 2030 sustainability targets EU observers reject election fraud allegations in Ecuador Ecuador's Noboa secures presidential win with 55.95% of vote El Salvador plays defiant game with IMF Bitcoin rules Trump and Bukele cement controversial migrant deportation deal Cross-border oil smuggling reveals dark underbelly of US-Mexico trade nexus Mexico rejects bleak IMF recession forecast Panama’s Mulino rules out mining law in bid to restart First Quantum project murder and the collapse of the Peruvian State Trinidad opposition sweeps to power in decisive election victory Venezuela embraces Russian GLONASS in challenge to US satellite dominance Japanese official warns of wider Asian security impact of Trump tariffs ASIA BLOG: Is China preparing to poke the paper tiger in the White House Asian giants thrive in Russia despite sanctions ASIA BLOG: China suddenly recognises need for community Myanmar junta chief to attend regional summit in Bangkok Southeast Asia's leading economies ponder possible US tariffs as 90-day pause takes effect China doubles down on Cambodia ties amid global trade tensions but Myanmar’s celebrations damped by earthquake aftermath Albanese defies ‘incumbency curse’ with a landslide re-election in Australia Pakistan airspace closure impacts around 600 Indian flights war clouds hover over the Indian subcontinent Nusantara: Indonesia’s grand ambition to reshape its future Empty shelves and soaring costs hit Japan’s rice supply China to boost Russian LNG imports in 2025 Myanmar’s recovery stalls as extent of losses still not known North Korea acknowledges deployment of troops to support Russia in Kursk region North Korea is making billions of dollars a year from supplying Russia with weapons India bans Pakistani Youtube channels over provocative content after Kashmir attack US imposes preliminary duties on Southeast Asian solar imports Papua New Guinea tribal conflict leaves 30 dead amid gold mine dispute Singapore election outcome a ‘clear signal of trust South Korea’s PPP narrows presidential field to two finalists China accuses US of sophisticated cyberattacks Thailand's Maha Songkran World Water Festival draws over 558,000 visitors Magnit acquires controlling stake in Azbuka Vkusa German Prosecutors Confirm Termination of Money Laundering Investigation Against Alisher Usmanov Comments by President of the Russian Fertilizers Producers Association Andrey Guryev on bilateral meeting between Indian Prime Minister Narendra Modi and Russian President Vladimir Putin PhosAgro/UNESCO/IUPAC green chemistry research grants awarded for the 8th time to world's best young scientists Download the pdf version Download the pdf version Download the pdf version Download the pdf version Borsa Istanbul has extended the up-tick rule that it introduced on February 24 for a third day, the authority said on February 26 As of January 2, Turkey’s capital markets board (SPK/CMB) lifted a short selling ban for the BIST-50 components The ban was introduced following the catastrophic February 2023 earthquakes that hit the south of the country It was kept in effect for non-BIST-50 stocks Borsa Istanbul’s benchmark BIST-100 index has been stuck around the 10,000-level in Turkish lira terms for almost a year There is also some pressure being felt on the USD/TRY pair The central bank has been burning through some reserves it is not expected that the authority will lose control or burn through a significant sum In its Outlook Turkey 2025, bne IntelliNews noted: “This year kicking in only at the beginning of the Christmas week.” followed by a recovery,” this publication reiterated “May is the month that usually runs according to the principle of 'Sell in May the summer liquidity dry-up then creates a shake-up in August It is followed by a recovery that takes place by November when a shake-up occurs prior to the beginning of the new year rally,” it added When all's said and done, Borsa Istanbul remains in a perfect mess. Given that the government does not allow the lira to appreciate another negative is that currency gains are missing have turned to deposits and money market funds with lira deposit rates and interbank money market rates shifting above the 50% policy rate that prevailed in 2024 With the rate-cutting cycle in progress, 2025 is the year of Turkish government lira bonds. The carry trade is still working out too. And Turkish eurobonds always offer good yields It should be noted that Borsa Istanbul is a highly manipulated market Foreign investors have yet to show real interest in Turkish stocks Widespread manipulative operations regularly take place The trading boards are not deep enough and algorithmic trading via well-known foreign brokerage houses rule the market Given the market situation, even the top BIST-30 index includes many manipulated stocks If you are not an insider privy to those circles doing the manipulating, trading stocks that are out of BIST-30 territory is similar to swimming in a pool full of sharks In 2025, Turkish stocks are supposed to come back into the game. The banking stocks as the first beneficiaries of a rate-cutting cycle The foreign interest seen in shareholder sales held in December may also suggest that automotive and aviation stocks deserve a look Check the box to receive the e-magazine to your inbox every month for free Get notified when there's a new bne IntelliNews Podcasts added Borsa Istanbul and Turkey’s 5-year credit default swaps (CDS) on April 7 fell victim to the global tumult brought about by the unveiling of Donald Trump's "Liberation Day" tariffs Stock exchanges from the US to the Far East have nosedived by 10-30% since the US president's big announcement on April 2 Leading stock exchanges have entered bear market territory. Borsa Istanbul’s benchmark BIST-100 index was down a further 3% d/d during the morning hours of April 7 while Turkey'sn CDS hit the 370s recording the highest levels since November 2023 The Turkish stock exchange fell around 1% on the tariffs announcement day The decline was much smaller than negative outcomes seen on the FTSE and Nasdaq The Turkish stock exchange's losses are still not as stark as the sharp falls seen on other stock markets around the world, as it is controlled by the government to avoid collapses The relatively low 10% baseline tariff that the US has imposed on countries including Turkey could favour Turkish exporters Turkey also stood to benefit from sharp declines seen in oil prices The Brent oil price sank from the $75s per barrel to the $63s in the wake of the Trump tariffs move Turkey imports just about all of its oil and gas needs Turkey can also capitalise on an advantage caused by the sharp increase in the EUR/USD pair The country imports intermediary goods in USD and sells final products to Europe in EUR The EUR/USD rate jumped to the 1.11s from the 1.08s Despite the huge media frenzy and market turmoil that Trump has created the implementation of the sweeping tariffs is not as yet certain Trump previously announced some tariffs on Mexico and Canada that were later lifted and came to be seen as his opening move in trade negotiations has once again done a great job for editors and news programmers thirsty for clicks and ratings Metrics details This study investigates the influence of Environmental and Governance (ESG) scores on the clustering and community formation of companies within various network models Using daily closing prices of 78 companies operating in the Borsa Istanbul Sustainability Index mutual information (both continuous and discrete) and causality (both linear and nonlinear) networks to analyse intercompany relationships We performed community detection using the Leading Eigenvector and Girvan–Newman methods which revealed that companies within the same sector particularly in the financial and manufacturing sectors These intra-sectoral clusters reflect strong market behaviour correlations driven by sector-specific factors mixed-sector communities highlighted the presence of significant inter-sector dependencies To assess the impact of ESG scores on these communities nonparametric tests such as the Kruskal–Wallis significantly influenced community formation companies with strong performance in emission reduction and CSR strategies were found to form more cohesive communities emphasizing the role of sustainability in shaping financial networks Study findings underscore the critical role of ESG factors in financial market dynamics promoting sustainable investment practices by highlighting the importance of integrating ESG considerations into investment decisions These results suggest that sustainability metrics not only affect individual company performance but also contribute to the formation of interconnected communities with shared sustainability practices financial networks can reveal key nodes or hubs which are entities with a high degree of connectivity that play a significant role in the overall stability and functionality of the market Identifying these influential nodes is crucial for understanding systemic risk and potential points of vulnerability within the financial system The analysis of financial networks also enables the study of information flow and influence among market participants By examining how information or shocks propagate through the network analysts can gain insights into the mechanisms of market reactions and the spread of financial contagion This understanding is critical for predicting market behaviours as it allows for the anticipation of how different parts of the market may respond to various stimuli and Governance (ESG) factors have emerged as vital considerations in financial decision-making ESG factors encompass a wide range of criteria that assess a company’s performance beyond traditional financial metrics Environmental factors include a company’s impact on the environment such as carbon emissions and resource usage Social factors evaluate how a company manages relationships with employees Governance factors examine a company’s leadership Companies with high ESG ratings tend to have lower volatility and better long-term returns compared to their peers with poor ESG performance the integration of ESG metrics is recognized for promoting sustainable investment practices investors can identify companies that are committed to sustainability and ethical practices which can lead to positive social and environmental outcomes This approach not only aligns with the growing demand for corporate responsibility but also enhances the resilience of financial markets Sustainable investment practices can help mitigate risks associated with environmental degradation thereby contributing to the stability and health of the financial system Despite the extensive research on financial networks there remains a significant gap in understanding how ESG factors influence the formation and dynamics of networks Traditional financial network studies primarily focus on metrics such as price correlations and return dependencies to analyse market structures and predict behaviours While these studies provide valuable insights into the interconnectedness and risk propagation within financial markets they often overlook the integration of ESG metrics which can offer a more comprehensive understanding of market dynamics Existing studies frequently miss the critical role that ESG factors play in shaping the relationships between market entities ESG metrics encompass a range of non-financial factors that reflect a company’s long-term sustainability and ethical practices current research may fail to capture the full spectrum of influences that drive market behaviours and interdependencies companies with strong environmental practices might be more interconnected due to shared sustainability goals leading to distinct community structures that are not apparent when only financial data are considered The integration of ESG metrics into financial network analysis can provide insights into how sustainability practices affect market interconnectivity and community structures Companies that prioritize ESG criteria may exhibit different connectivity patterns forming clusters based on shared environmental initiatives This can reveal hidden dependencies and relationships that traditional financial metrics might miss such as the tendency for companies with robust ESG practices to be more resilient during economic downturns or crises due to better risk management and ethical governance understanding the impact of ESG factors on financial networks can enhance the predictive power of these models Incorporating ESG metrics can help identify systemic risks and opportunities that are not evident through financial data alone a network analysis that includes ESG factors might reveal that companies with poor environmental practices are more susceptible to regulatory changes or public backlash leading to greater market volatility and risk of contagion the integration of ESG metrics into financial network analysis can inform more sustainable investment strategies Investors increasingly seek to align their portfolios with ethical and sustainable practices and understanding the role of ESG factors in financial networks can guide them in identifying companies and sectors that are not only financially robust but also committed to long-term sustainability This can lead to more informed investment decisions that support both financial returns and positive societal impacts This study aims to bridge the gap in existing research by investigating the impact of ESG scores on clustering and community formation within various financial network models The primary objective is to integrate sustainability metrics into the analysis of financial networks to provide a more comprehensive understanding of market dynamics the study seeks to uncover how environmental and governance considerations influence intercompany relationships and the formation of distinct communities within these networks the study employs several sophisticated network models each offering a unique perspective on intercompany relationships correlation network models will be utilized to identify how company stock prices move in relation to one another edges represent the degree of correlation between the daily closing prices of different companies with higher correlations indicating stronger connectivity This approach will help detect sectoral clusters where companies within the same sector exhibit similar stock price movements due to shared economic factors the study will utilize mutual information network models to collect and analyse both linear and nonlinear connections among stock values of different companies By employing both continuous and discrete mutual information these networks will surpass basic linear correlations to uncover intricate interdependencies and information exchanges among organizations This approach seeks to reveal insights into the impact of information flow and common market behaviours on community structures potentially showing important inter-sectoral linkages causality network models will be utilized to comprehend the directional effects among stock values of different organizations These models will illustrate how fluctuations in the stock price of one firm can forecast those of another firm thus capturing the cause-and-effect interactions within the market We will conduct both linear and nonlinear causality analyses to obtain a full understanding of these interdependencies To assess the influence of ESG metrics on community structures These tests are particularly suitable for this analysis as they do not assume a specific distribution for the data making them robust tools for evaluating relationships between ESG scores and community formation the study aims to provide valuable insights into the role of ESG factors in shaping financial networks By understanding how sustainability metrics influence market connectivity and community structures the findings of this study can inform more sustainable investment practices and contribute to the development of resilient and responsible financial markets Study findings reveal distinct patterns of interconnectivity among companies across different network models Correlation networks highlight strong sectoral linkages particularly within the finance and manufacturing sectors Mutual information networks provide insights into both intra- and inter-sector dependencies while causality networks demonstrate predictive relationships in stock price movements The nonparametric test results indicate that ESG factors such as Emission and Human Rights significantly influence community formation suggesting that companies with strong performance in these areas tend to form more cohesive communities The remainder of this paper unfolds as follows: section “Literature review” embarks on a comprehensive literature review Section “Methodology” delves into the methodology detailing the construction of network models and the application of community detection methods we introduce the dataset and provide an in-depth analysis of the influence of ESG sub-scores on overall ESG performance This section then transitions into a detailed presentation and discussion of network and community results Section “Discussions” explores the nonparametric tests employed to evaluate the impact of ESG scores the section “Conclusions” ends the study with a summary of key insights and thoughtful suggestions for future research directions By integrating ESG factors into financial network analysis this study not only enhances understanding of market dynamics but also promotes sustainable investment practices The insights gained from this research can guide investors and policymakers in fostering more resilient and socially responsible financial markets and Governance (ESG) factors in financial networks is essential for grasping the complexities of market dynamics and sustainable investment practices This literature review synthesizes insights from recent studies on the integration and impact of ESG metrics within financial networks Financial performance was measured through regional GDP growth and productivity showing that better social infrastructure determined higher economic growth suggesting that ESG factors were not fully integrated into performance metrics in these regions These studies collectively emphasize the critical role of ESG factors in financial analysis and emphasize the need for integrating sustainability metrics to foster more responsible financial markets noting its potential to address challenges in ESG disclosure and improve the reliability of ESG ratings These studies collectively emphasize the growing trend toward integrating sophisticated predictive models and machine learning techniques in ESG research to enhance the precision and applicability of ESG assessments The integration of ESG factors into financial network analysis provides valuable insights into market dynamics and promotes sustainable investment practices researchers can enhance the predictive power of financial models and support the development of resilient financial systems particularly those derived from graph theory enhances our understanding of complex interrelated systems nodes represent discrete items (such as stocks or bonds) while edges indicate the connections or affiliations between them where V represents the set of nodes (or vertices) and \(E\subseteq V\times V\) represents the set of edges Weight functions \(\omega :E\to {{\mathbb{R}}}^{+}\) are often used to assign weights to edges in real-world scenarios such as the financial network of the Borsa Istanbul Sustainability Index This generates the formation of a weighted graph This study examines five distinct approaches to network creation each establishing the relationship between nodes that represent companies through different methods for generating edges Weighted financial networks are constructed by assigning weights to the edges based on the selected approach The primary objective of the study is to examine the clusters in financial networks derived from various methodologies specifically focusing on graph communities Within the complex realm of global economics the interconnection of financial networks plays a crucial role in shaping market directions and impacting economic stability composed of intricate systems of institutions are essential in facilitating the movement and distribution of capital globally The interconnectedness of these networks fosters a dynamic ecosystem where economic activity Understanding the interconnections within financial networks is vital for effectively navigating the complexities of contemporary finance as it impacts not only individual institutions but also has extensive ramifications for entire economies and societies and the edge weights measure the correlation between the returns of connected assets To analyse the logarithmic return of the daily closing price of stocks This equation calculates the difference in the daily closing price of stock i at time t To compute the Pearson correlation coefficient between stocks i and j where \(\langle \cdot \rangle\) represents the temporal average The Pearson correlation coefficient is a commonly used statistical measure in finance that quantifies the extent of the linear relationship between two financial time series A coefficient close to 1 indicates a strong positive correlation suggesting that the two time series consistently move together a coefficient close to −1 indicates a significant negative correlation where the variables move in opposite directions A correlation value close to zero indicates a minor or insignificant correlation meaning there is no noticeable linear relationship between time series The weight function derived from the correlation coefficient is given by By starting with a comprehensive initial network setup where each node is intricately connected and assigned weights based on correlation distance we uncover detailed patterns of relationship dynamics the corresponding edge weights tend to decrease towards zero weaker correlations result in weights approaching \(\sqrt{2}\) weights serve as quantifiable measures of relationship intensity with smaller values indicating stronger links The topologically triangulated structure used throughout this paper is denoted as \({G}_{F}=(V,{E}_{F},\omega )\) where \({E}_{F}\subset E\) represents the filtered edge set Implementing this triangulated framework allows for a more detailed understanding of systemic risk and interconnections By using weighted edges derived from correlation distances filtering techniques are applied to enhance the network emphasizing important interactions and reducing the impact of noise the concept of mutual information has become a vital framework for understanding the complex connections and interrelationships within networks Mutual information in financial networks quantifies the degree to which the actions or performances of one entity provide insights into those of another revealing the shared information and interconnections among various financial elements Studying mutual information is crucial for understanding the dynamics that drive market movements and systemic stability in today’s interconnected and complex financial environments This quantitative metric provides valuable insights into the flow and sharing of information among different entities in the financial ecosystem helping one to grasp the relationships that shape the modern financial landscape This study constructs a weighted network by using mutual information associations among time series derived from logarithmic returns of closing prices given in Eq. (1) The use of continuous and discrete bivariate mutual information methodologies allows for the capturing of distinct dynamics Calculating continuous mutual information is essential for uncovering the connections between logarithmic returns of closing prices in financial networks This process involves accurately calculating the continuous mutual information between pairs of organizations offering a quantitative assessment of how interconnected their return behaviours are logarithmic returns are treated as continuous stochastic variables and their joint probability density function is computed using kernel density estimation techniques The integral of the joint probability density function Discrete bivariate mutual information is a crucial step in exploring financial networks using a discrete representation of logarithmic returns to quantify the informational content between pairs of companies the continuous logarithmic returns are discretized into a finite set of bins allowing for the application of discrete mutual information theory The weight assignment schemes used in this study are \(\frac{1}{{I}_{{{c}}}\left(C{l}_{i}{;C}{l}_{j}\right)}\) and \(\frac{1}{{I}_{{{d}}}\left(C{l}_{i}{;C}{l}_{j}\right)}\) for the respective networks Incorporating the inverses of continuous mutual information and discrete bivariate mutual information as edge weights represents an intriguing approach to constructing financial networks When considering continuous mutual information applying edge weights as the reciprocal of estimated mutual information values results in a proportional decrease in the strengths of linkages indicating stronger connections between financial institutions based on logarithmic returns This method amplifies weaker connections by assigning them higher weights resulting in a network representation that emphasizes links that are less connected or less dependent In the context of discrete bivariate mutual information the dynamic weighting technique assigns higher weights to edges with lower mutual information values indicating a focus on relationships with lower information content This reciprocal weighting approach offers a distinct perspective by prioritizing edges representing less statistically significant relationships within discretized financial data The TMFG approach is used as a filtering technique to acquire the sets of EF for both networks generated using continuous and discrete bivariate mutual information It is important to note that in both scenarios the reciprocals of information values are used to determine the appropriate edge weights for TMFG filtering help identify predictive connections between the historical returns of one company and another advanced approaches in nonlinear causality analysis provide a more detailed understanding of the complex relationships that may occur surpassing the limitations of linear models The resulting financial networks encompass not only direct effects but also intricate offering a comprehensive view of how the financial performances of organizations influence one another This extensive investigation into linear and nonlinear causality in financial time series data forms a basis for better understanding the underlying dynamics Consider two time series \(C{l}_{i}\left(t\right)\) and \(C{l}_{j}\left(t\right)\) representing the logarithmic returns of daily closing prices for two distinct companies in the stock market The linear causality between these time series can be formally assessed through the Granger causality test Let \(C{l}_{i}\left(t\right)\) and \(C{l}_{j}\left(t\right)\) be jointly modelled in a lagged vector autoregressive framework as follows: \({\alpha }_{1s}\) and \({\alpha }_{2s}\) are coefficients and \({\epsilon }_{i}\) and \({\epsilon }_{j}\) are white noise error terms The Granger causality test assesses whether past values of \(C{l}_{i}\left(t\right)\) provide significant information in predicting \(C{l}_{j}\left(t\right)\) beyond the information contained in the past values of \(C{l}_{j}\left(t\right)\) itself the null hypothesis \({H}_{0}\) is that \(C{l}_{i}\left(t\right)\) does not Granger-cause \(C{l}_{j}\left(t\right)\) The test involves comparing the F-statistic from the restricted model (excluding lagged values of \(C{l}_{i}\left(t\right)\)) with the F-statistic from the unrestricted model (including lagged values of \(C{l}_{i}\left(t\right)\)) Statistical significance in rejecting \({H}_{0}\) indicates the presence of linear causality from \(C{l}_{i}\left(t\right)\) to \(C{l}_{j}\left(t\right)\) thereby providing insights into the directional influence of one company’s logarithmic returns on another in the stock market are used to investigate the nonlinear causality between time series Let us denote the entropies of \(C{l}_{i}\left(t\right)\) and \(C{l}_{j}(t)\) as \(H\left(C{l}_{i}\left(t\right)\right)\) and \(H\left(C{l}_{j}\left(t\right)\right)\) The conditional entropy of \(C{l}_{i}\left(t\right)\) given \(C{l}_{j}\left(t\right)\) is represented as \(H(C{l}_{i}|C{l}_{j})\) represented as \({TE}(C{l}_{i}\to C{l}_{j})\) measures the decrease in uncertainty regarding the future values of \(C{l}_{i}\left(t\right)\) when the previous values of \(C{l}_{j}\left(t\right)\) are considered surpassing the predictive capability of \(C{l}_{i}\left(t\right)\) based solely on its own previous values represents the Transfer Entropy from \(C{l}_{i}\) to \(C{l}_{j}\) Nonlinear causality is identified when the value of \({TE}(C{l}_{i}\to C{l}_{j})\) is significantly different from zero This method identifies the specific direction in which information flows between the logarithmic returns of the two companies offering valuable insights into nonlinear connections that extend beyond linear links identified by conventional Granger causality analysis Understanding these nonlinear causal connections enhances the description of dynamic interactions in the stock market Assigning weights to edges in networks generated by linear and nonlinear causality analyses using p-values provides a statistical method to determine the importance and reliability of detected causal connections among time series edges can be assigned weights based on the p-values derived from Granger causality tests Smaller p-values indicate greater statistical significance so edges associated with more significant causal relationships are assigned lower weights This methodology ensures that the network represents only statistically robust and reliable causal relationships for networks resulting from nonlinear causality analysis p-values obtained from tests such as those evaluating the significance of Transfer Entropy can be used to assign weights to edges Smaller p-values in the context of nonlinear causality provide stronger evidence for the existence of meaningful causal links By integrating p-values into the weighting of edges these networks not only visually depict causal structure but also allow researchers to focus on the most statistically plausible connections enhancing the reliability of insights derived from network analysis Incorporating p-values in the context of TMFG filtering is highly relevant and enhances the robustness and reliability of network representation The TMFG filtration technique aims to identify the most crucial connections within a network by selecting edges based on their robustness and significance p-values provide an inherent criterion for this objective offering a statistical assessment of the reliability of identified causal connections By assigning lower weights to edges associated with lower p-values the TMFG filtration can be tailored to emphasize connections that are statistically significant This approach ensures that the final graph captures the most significant information flow and highlights edges with a high level of confidence in their causal influence mathematically referred to as subsets of nodes within a graph that exhibit higher connectivity among themselves than with nodes outside the subset are essential structures in network analysis Let \(G=\left(V,\,{E}_{F},\,\omega \right)\) represent a weighted filtered graph A weighted community is then a subset \(C\subseteq V\) where the total edge weight within the community is significantly higher than the expected total edge weight in a null model for a given partition of nodes into weighted communities \({\mathcal{C}}=\left\{{C}_{1},{C}_{2},\ldots ,{C}_{k}\right\}\) the modularity \({Q}_{w}\) for weighted graphs is often employed as an optimization metric The weighted modularity is an extension of the modularity concept for unweighted graphs considering the sum of edge weights within communities where \(w({C}_{i})\) is the total weight of edges within community \({C}_{i}\) and W is the total weight of all edges in the graph Weighted graph communities offer a more nuanced understanding of network structures by capturing varying strengths of connections between nodes which is particularly relevant in applications where the intensity or significance of relationships plays a crucial role The process of identifying and analysing graph communities offers a perspective to comprehend the interrelationships and interdependencies among publicly listed organizations These communities consist of organizations that have comparable financial features they exhibit patterns that contribute to the overall resilience of the market When evaluating various network formation principles graph communities are particularly useful in capturing intricate dynamics Correlation-based networks can reveal the interdependence between enterprises in a certain economic domain by identifying communities that showcase sectors or industries with strongly associated stock price movements networks built on mutual information expose communities that exhibit substantial information exchange suggesting clusters of enterprises that react similarly to market occurrences Introducing causality in network creation adds complexity by revealing how one company’s financial performance causally affects another providing better understanding of cause-and-effect interactions within the financial ecosystem Within the framework of the Borsa Istanbul Sustainability Index graph communities play a crucial role in identifying groups of enterprises that not only share financial connections but also engage in similar sustainability practices Companies in a certain community may have comparable environmental which collectively contribute to the overall sustainability goals of the index Examining these communities enables stakeholders to identify areas of proficiency and prospects for enhancement in relation to sustainability performance Two distinct community detection algorithms are employed in this study, each based on the modularity metric provided in Eq. 9 \({d}_{i}\) and \({d}_{j}\) are the weighted degrees of nodes i and j and m is the total edge weight in the graph Normalizing B by dividing each entry by 2m yields a symmetric matrix The spectral clustering procedure commences by calculating the eigenvectors and eigenvalues of the normalized modularity matrix The components of this eigenvector are subsequently employed as coordinates for the nodes thus mapping them onto a multi-dimensional space a conventional clustering technique is employed to process these coordinates Common options include k-means clustering or hierarchical clustering The partition of nodes that emerges from this process creates separate communities within the graph Nodes that are assigned to the same cluster demonstrate comparable patterns in their connection strengths let \({{\boldsymbol{v}}}_{1}\) be the leading eigenvector corresponding to the largest eigenvalue \({\lambda }_{1}\) of the normalized modularity matrix Form a matrix \({{\boldsymbol{V}}}_{1}\) by stacking the entries of \({{\boldsymbol{v}}}_{1}\) and apply a clustering algorithm to the rows of \({{\boldsymbol{V}}}_{1}\) This clustering process reveals graph communities based on the spectral properties of the modularity matrix The second approach we employ is founded on edge betweenness and is often known as the Girvan–Newman algorithm (Girvan & Newman 2002) the weighted betweenness centrality (\(B(e)\)) of an edge e in a weighted graph is computed as: where \({g}_{{st}}(e)\) represents the number of shortest paths from node s to node t that traverse edge e and \({g}_{{st}}\) is the total number of shortest paths from s to t The weight of an edge is considered in these calculations The Girvan–Newman algorithm then proceeds by iteratively calculating the weighted betweenness centrality for all edges identifying and removing the edge with the highest weighted betweenness The process continues until the desired number of communities is reached or the graph becomes fully disconnected let \({E}_{k}\) represent the set of edges removed during k iterations and \({G}_{k}=\left(V,{E}_{k},{W}_{k}\right)\) be the graph after k iterations where \({W}_{k}\) is the updated weight matrix The iterative removal of edges partitions the graph into disjoint components The modularity is used as an optimization criterion to assess the quality of the resulting communities The Girvan–Newman algorithm for weighted graphs provides a powerful means of community detection considering both the presence and strength of connections This adaptation enhances its applicability to diverse systems where edge weights convey additional information about the underlying network structure This study introduces five methodologies for comparing network communities The initial approach involves comparing two distinct community detection methodologies implemented on the same network The comparison includes the consideration of normalized mutual information When comparing two weighted network communities obtained by different community detection approaches it is essential to employ appropriate evaluation metrics to assess the quality of identified communities One commonly used metric is Normalized Mutual Information (NMI) NMI quantifies the mutual dependence between two sets of communities while considering variations in community sizes where \(I\left({{\mathcal{C}}}_{1};{{\mathcal{C}}}_{2}\right)\) represents the mutual information between the communities \({{\mathcal{C}}}_{1}\) and \({{\mathcal{C}}}_{2}\) and \(H\left({{\mathcal{C}}}_{1}\right)\) and \(H\left({{\mathcal{C}}}_{2}\right)\) are the entropies of \({{\mathcal{C}}}_{1}\) and \({{\mathcal{C}}}_{2}\) with higher values indicating greater similarity between community structures Variation of Information (VOI) is another crucial metric for community comparison It captures both the uncertainty within individual communities and the uncertainty about the relationship between them where \(H\left({{\mathcal{C}}}_{1}\right)\) and \(H\left({{\mathcal{C}}}_{2}\right)\) are the entropies of \({{\mathcal{C}}}_{1}\) and \({{\mathcal{C}}}_{2}\) Lower values of VOI correspond to better community detection results Split Join Distance (SJD) is a metric that measures the dissimilarity between two community structures by considering how communities split or merge This metric is particularly suitable for detecting changes in community memberships where \(s\left({{\mathcal{C}}}_{1},{{\mathcal{C}}}_{2}\right)\) and \(s\left({{\mathcal{C}}}_{2},{{\mathcal{C}}}_{1}\right)\) represent the fraction of nodes split from \({{\mathcal{C}}}_{1}\) to \({{\mathcal{C}}}_{2}\) and from \({{\mathcal{C}}}_{2}\) to \({{\mathcal{C}}}_{1}\) Rand Index (RI) and Adjusted Rand Index (ARI) are widely used metrics for assessing the similarity between two sets of partitions Rand Index quantifies the proportion of pairs of nodes that are correctly classified as either belonging or not belonging to the same community The Adjusted Rand Index adjusts for chance agreement The formulas for Rand Index and Adjusted Rand Index are where \({c}_{1}\) is the number of pairs of nodes in the same community in both \({{\mathcal{C}}}_{1}\) and \({{\mathcal{C}}}_{2}\) \({c}_{2}\) is the number of pairs of nodes in different communities in both \({{\mathcal{C}}}_{1}\) and \({{\mathcal{C}}}_{2}\) and \({\mathbb{E}}\left[{RI}\left({{\mathcal{C}}}_{1},{{\mathcal{C}}}_{2}\right)\right]\) is the expected Rand Index This paper introduces a novel comparison methodology based on random walk that intends to evaluate networks obtained through various methods in addition to the five existing methodologies for comparing network communities In a random walk on a weighted network where lower edge weights signify stronger relationships we can represent the process mathematically using equations that reflect this inverse relationship Let us consider a network represented by \(G=(V,{E}_{F},\omega )\) Each edge \({e}_{{ij}}\in {E}_{F}\) connecting nodes i and j has an associated weight \({w}_{{ij}}\) a lower weight implies a stronger relationship so the transition probability from node i to node j is inversely proportional to the weight \({w}_{{ij}}\) The transition probability \({P}_{{ij}}\) for a walker to move from node i to node j is given by: where \(N(i)\) denotes the set of neighbours of node i and \({w}_{{ik}}\) are the weights of the edges connecting node i to its neighbours \(1/{w}_{{ij}}\) represents the ‘strength’ of the connection and the denominator is the normalization factor ensuring that the sum of probabilities from node i to all its neighbours equals 1 the random walk can be analysed in terms of a Markov Chain whose transition matrix M is defined by the probabilities \({P}_{{ij}}\) The state of the walker after n steps is given by the distribution vector \({{\boldsymbol{v}}}^{\left(n\right)}\) and \({{\boldsymbol{v}}}^{\left(0\right)}\) is the initial state of the walker the system may reach a steady state described by a stationary distribution Π where \({\boldsymbol{\Pi }}{\boldsymbol{M}}={\boldsymbol{\Pi }}\) This stationary distribution depends intricately on the network’s topology and the inverse weighting of the edges providing insights into the most ‘strongly connected’ components of the network This mathematical framework is valuable for understanding various real-world networks where stronger ties have less ‘resistance’ or ‘distance’ and thus are more frequently traversed in the random walk process The Borsa Istanbul Sustainability Index (XUSRD) is an important initiative by Borsa Istanbul aimed at promoting and highlighting sustainable business practices among the companies listed on the exchange aims to assist organizations in formulating policies that consider the potential risks associated with ESG (Environment it seeks to educate investors about the sustainability practices implemented by these companies organizations must meet certain requirements including a cumulative ESG score of 50 or higher individual scores of 40 or higher for each of the ESG pillars and at least 8 category scores of 26 or higher the Borsa Istanbul Sustainability Index includes 80 companies Borsa Istanbul’s adoption of this index reflects its commitment to sustainability and responsible investment aligning with the growing global focus on environmental The index incentivizes organizations to continually enhance their sustainability performance serving as a measurable standard for the creation of various financial instruments such as funds and structured products that adhere to sustainability principles Choosing the Borsa Istanbul Sustainability Index for network analysis with an emphasis on sustainability and corporate responsibility offers several advantages compared to the broader Borsa Istanbul 100 Index The Borsa Istanbul Sustainability Index is a specialized index comprising companies that demonstrate exceptional sustainability performance This focus allows for a detailed examination of how sustainability impacts company networks highlighting the increasing significance of sustainability in investment decisions by evaluating the included companies based on these criteria the index provides a unique perspective on ESG concerns This enables an understanding of the interaction between ESG variables and company policies which is less apparent in more comprehensive indices such as the Borsa Istanbul 100 Index Examining the Borsa Istanbul Sustainability Index can uncover novel and insightful market patterns due to the dynamic nature of sustainability in stock indices companies that prioritize sustainability often exhibit greater resilience to environmental and social risks providing valuable insights into risk management strategies that enhance corporate stability ESG scores are specifically formulated to assess a company’s proficiency in addressing diverse sustainability and ethical obstacles The ratings are divided into various essential sections each emphasizing different aspects of a company’s operations and influence The emissions score assesses a company’s greenhouse gas emissions considering both the overall emissions generated and the intensity of those emissions in relation to company production It discusses the company’s efforts to control and diminish these emissions through diverse programs and adherence to environmental rules The resource use score evaluates the effectiveness and long-term viability of a company’s utilization of natural resources Important indicators include quantifying water and energy consumption and the utilization of sustainable resources companies that successfully execute efficient resource management techniques achieve better scores The innovation score evaluates a company’s allocation of resources towards research and development (R&D) and its capacity to generate novel ideas and solutions that promote sustainability This includes the creation of novel technologies and goods that minimize harm to the environment as well as the acquisition of patents pertaining to sustainable technologies The human rights score focuses on analysing a company’s policies and practices related to safeguarding human rights in its operations and supply chain and initiatives to combat child and coerced labour The product responsibility score assesses a company’s ability to effectively handle the safety and environmental consequences of its products It encompasses factors such as adherence to product safety regulations assessment of the environmental impact throughout the product lifecycle and the incorporation of sustainable materials in the product development process The workforce score evaluates how a company treats its employees and activities related to diversity and inclusion Companies that cultivate a constructive and fair work atmosphere typically achieve higher scores The community score is used to measure a company’s level of involvement and influence in the communities where it operates and efforts to address local concerns and make a positive impact on society The management score domain assesses the calibre and efficiency of a company’s governance structures and processes This includes the board structure and diversity and overall corporate governance regulations The shareholder score evaluates how a company prioritizes and resolves its shareholders’ interests and concerns and the company responsiveness to shareholder resolutions and suggestions The corporate social responsibility (CSR) strategy score evaluates a company’s comprehensive approach to integrating social and environmental factors into its commercial strategy The CSR framework encompasses company’s dedication to CSR values and documentation of CSR actions and results Table A2 details the ESG and sub-scores for each company analysed in this study These scores are based on self-reported data from each company for the fiscal years 2022 or 2023 The ESG scores encompass a range of subcategories Each sub-score provides insight into specific aspects of a company’s environmental offering a comprehensive view of their sustainability performance The goodness-of-fit metrics provided in Table 1 shed light on the effectiveness of the linear regression model in explaining the variability in ESG scores of companies in the Borsa Istanbul Sustainability Index Adjusted R-squared value of 0.9402 suggests that the model can account for approximately 94% of the variability in ESG scores The adjusted R-squared is particularly useful because it adjusts for the number of variables in the model preventing overestimation of the model’s explanatory power A value close to 1 suggests a very good fit The Akaike Information Criterion (AIC) is a measure of the relative quality of statistical models for a given dataset with lower values indicating a better-fitting model an AIC of 377.12 suggests that the model is relatively efficient in balancing goodness-of-fit and complexity The Bayesian Information Criterion (BIC) is another criterion for model selection but with a stricter penalty for models with more parameters The R-squared value of 0.948 indicates that the model explains about 94.8% of the variability in ESG scores It is a measure of how well the regression predictions approximate the real data points A value close to 1 indicates the strong explanatory power of the model the high R-squared and adjusted R-squared values indicate that the model is highly effective in explaining the variability in ESG scores Although AIC and BIC values suggest some model complexity The linear regression model in Table 2 provides useful insights into the impact of many factors on the ESG ratings of companies in the Borsa Istanbul Sustainability Index does not provide meaningful information on its own The coefficient for Emission is positive (0.0279) but lacks statistical significance suggesting that emission levels have a negligible and non-significant influence on ESG ratings there is a strong and positive link between Resource Use and ESG ratings as indicated by a coefficient of 0.117 and a p-value of \(6.7\times {10}^{-6}\) This implies that effective allocation of resources is closely linked to improved ESG performance The impact of Innovation on ESG scores is extremely significant as evidenced by a coefficient of 0.054 and a p-value of \(6.6\times {10}^{-6}\) This shows the crucial role of innovation in attaining higher ESG ratings The analysis indicates that there is a substantial and positive relationship between human rights practices and ESG scores The coefficient of 0.05 and the p-value of 0.039 provide statistical evidence for this relationship It suggests that organizations with stronger human rights practices generally get higher ESG scores The coefficient of 0.089 and the p-value of 0.00014 indicate that there is a strong link between product accountability and higher ESG ratings This means that organizations that accept responsibility for their goods are more likely to have better ESG scores with a coefficient of 0.173 and a p-value of \(6.7\times {10}^{-6}\) exhibits a statistically significant positive relationship with ESG scores implementing effective workforce practices is crucial for attaining high ESG ratings The statistical analysis reveals that community participation has a coefficient of 0.14 and a p-value of \(1.5\times {10}^{-8}\) This emphasizes the crucial role of community involvement in improving ESG performance The analysis reveals a significant and positive relationship between management practices and ESG scores The coefficient of 0.197 indicates a substantial impact while the p-value of \(1.3\times {10}^{-21}\) suggests a highly significant relationship These findings underscore the importance of Management in achieving high ESG scores The impact of Shareholders on ESG ratings is substantial as indicated by a coefficient of 0.041 and a p-value of 0.00117 This suggests that active shareholder participation has a favourable effect on ESG scores the coefficient of CSR Strategy is 0.036 with a p-value of 0.0188 suggesting that a strong CSR strategy is linked to elevated ESG scores ANOVA results in Table 3 provide an in-depth analysis of the impact of various factors on the ESG scores of companies included in the index Each factor is evaluated in terms of its contribution to the overall model and its statistical significance Emission shows an F-statistic of 224.437 with a highly significant p-value of \(4.5\times {10}^{-23}\) indicating a substantial impact on ESG scores Resource Use has an F-statistic of 418.749 and a p-value of \(1.5\times {10}^{-30}\) highlighting its critical importance in the model with an F-statistic of 27.1374 and a p-value of \(1.9\times {10}^{-6}\) Human Rights is another highly significant factor reflected by its F-statistic of 123.313 and a p-value of \(7.7\times {10}^{-17}\) Product Responsibility shows a moderate significance with an F-statistic of 4.46292 and a p-value of 0.038 Workforce management is strongly significant indicated by an F-statistic of 152.156 and a p-value of \(6.6\times {10}^{-19}\) Community involvement also plays a significant role with an F-statistic of 58.7096 and a p-value of \(9.9\times {10}^{-11}\) with an F-statistic of 194.79 and a p-value of \(1.6\times {10}^{-21}\) as shown by an F-statistic of 11.0519 and a p-value of 0.00144 the CSR Strategy demonstrates significance with an F-statistic of 5.79067 and a p-value of 0.01887 These results collectively confirm that factors significantly impact ESG scores reflecting their multifaceted roles in sustainability performance Our study uses the ESG scores of companies in the XUSRD dataset which is highly suitable for examining the impact of various factors on community formation in different network types (correlation networks The relevance of ESG scores lies in their comprehensive measure of a company’s performance across environmental which aligns perfectly with our aim to investigate the factors influencing community formation based on ESG-related criteria The dataset includes a diverse range of variables ensuring a comprehensive analysis of these factors The high adjusted R-squared value demonstrates statistical robustness indicating that the model explains a substantial portion of the variance in ESG scores This provides confidence in the ability of data to reveal meaningful patterns and relationships With data from 78 companies and 249 observations for each time series the dataset is sufficiently large to allow for robust statistical analyses making the findings generalizable and not biased by the small sample size The subsequent part of this section encompasses networks acquired using various techniques as well as the graph communities derived from these networks Such networks are constructed based on distinct methodologies including correlation distance Each method provides a unique perspective on the interactions and dependencies among companies in the Borsa Istanbul Sustainability Index highlighting different aspects of their relationships The figures used in the following subsections incorporate colour-coded indications to represent sectors to which companies belong facilitating a visual understanding of sector-specific clustering and community formation The node colours represent the sectors of each company according to the following rule: red represents commercial This colour-coding scheme allows for an intuitive comparison of sectoral distribution within and across the different network types TMFG with correlation distance and weighted degree distribution Figure 2 shows the clustering results of the correlation network based on company daily closing prices, with clusters identified using the Leading Eigenvector method on the left and the Girvan–Newman method on the right. Emerging filtered correlation network communities In Fig. 2 both methods reveal distinct communities within the network The Leading Eigenvector method identifies several well-defined clusters indicating strong intra-sector correlations particularly within the manufacturing and financial sectors This suggests that companies within these sectors exhibit similar stock price movements likely driven by sector-specific factors and market conditions the Girvan–Newman method also identifies distinct communities though the clustering patterns show slight variations compared to the Leading Eigenvector method The financial and manufacturing sectors again show significant clustering highlighting their strong internal correlations Both methods reveal that companies within the same sector tend to cluster together reflecting shared market influences and sectoral trends clustering results from both methods show how important sectoral affiliation is in determining how daily closing prices are correlated with clear community structures showing up in the manufacturing and financial sectors It indicates that these sectors have more homogenous price movements likely due to common economic drivers and market conditions Table A3 shows the community results that indicate how the Leading Eigenvector and Girvan–Newman algorithms found communities in the correlation network Each community includes companies from diverse sectors This diversity within communities indicates that companies from distinct sectors exhibit significant correlations in daily closing prices likely reflecting interconnected market behaviours and shared external influences suggesting that sectors may be highly correlated in their stock price movements The presence of companies from multiple sectors within each community is consistent with previous research on significant sectoral interdependencies in the Borsa Istanbul market the weighted vertex degree and eigenvector centrality metrics emphasize the roles and significance of specific companies within their respective communities companies in the commercial sector like MGROS have a high-weighted vertex degree of 3.67612 and an eigenvector centrality of 0.0601084 which shows that they have a lot of strong connections and influence in the network ISGYO has a weighted vertex degree of 4.43027 and an eigenvector centrality of 0.0735354 highlighting its critical role in the network structure TOASO from the manufacturing sector has a weighted vertex degree of 3.41767 and an eigenvector centrality of 0.2251145 making it a central player in maintaining network cohesion Figure 3 highlights a network derived from continuous mutual information between company daily closing prices, displayed on the left side. TMFG with continuous mutual information and weighted degree distribution The network topology reveals a high degree of interconnectedness, particularly among the financial and manufacturing sectors, which exhibit higher vertex degrees in Fig. 3 This indicates that sectors share significant informational dependencies reflecting similar responses to market conditions highlights that most companies have a moderate number of connections with a few companies having very high degrees indicating their central role in the network The centrality of these companies implies their status as key players closely monitored and highly influential in stock price movements shows that most connections have relatively high weights indicating strong mutual information between closing prices of companies This suggests that company stock prices contain significant shared information likely driven by sector-specific dynamics and overall market trends Network topology and distributions indicate a robust structure playing pivotal roles in the mutual information network based on daily closing prices This reflects substantial informational linkages within these sectors emphasizing their importance in the market’s overall behaviour Figure 4 displays communities identified in the continuous mutual information network based on company daily closing prices. Emerging filtered continuous mutual information network communities In Fig. 4 derived by using the Leading Eigenvector method Community 1 and Community 4 predominantly consist of companies from the financial and manufacturing sectors with strong intra-sector information sharing Community 2 and Community 3 include a mix of sectors the Girvan–Newman method also identifies four communities While Community 2 centres around a few key financial and commercial companies Community 3 and Community 4 display a blend of manufacturing and technology sectors illustrating sectoral and cross-sectoral informational relationships Both methods highlight that companies within the same sector tend to cluster together due to strong mutual information links reflecting similar market behaviours and shared influences The variations in community structure between methods underscore different perspectives on how these informational relationships manifest within the network clustering results emphasize the significant role of sectoral affiliations in shaping the network topology based on daily closing prices with the financial and manufacturing sectors showing particularly strong internal and external connections In Table A4 presenting continuous mutual information network communities we observe the significance of weighted vertex degree and eigenvector centrality metrics in identifying pivotal companies within their respective communities the company DOAS in the commercial sector exhibits a high-weighted vertex degree of 13.4232 and an eigenvector centrality of 0.0527802 indicating its central role in network connectivity displays a notable weighted vertex degree of 51.0328 and an eigenvector centrality of 0.0634826 highlighting its influence within the network structure These metrics underline the importance of certain companies in maintaining the overall network integrity and their potential impact on market dynamics one can infer that companies with higher degrees and centrality are likely to be influential in driving the behaviour and performance of their communities thereby playing crucial roles in market sustainability and stability companies like AGESA in the financial sector and KORDS in the manufacturing sector also show significant metrics reinforcing their key positions in the network Figure 5 presents the network based on discrete bivariate mutual information between company daily closing prices. TMFG with discrete bivariate mutual information and weighted degree distribution The network topology shows that the manufacturing and financial sectors are closely connected in Fig. 5 which means that they share a lot of information This is probably because of factors that are specific to each sector and the way markets are set up The weighted vertex degree distribution on the top right shows that most companies have a moderate number of connections with a few companies exhibiting high connectivity suggesting their central role in the network Central companies are likely influential in the market as their price movements provide significant informational content for other companies The edge weight distribution on the bottom right shows that most connections have high mutual information values indicating strong relationships between company closing prices The high mutual information suggests that stock price movements share substantial information potentially due to shared economic influences or sector-specific dynamics the network topology and distributions show that sectors like manufacturing and finance are very important with a lot of information flowing within and between sectors This underscores the importance of these sectors in the overall market behaviour as reflected by the discrete bivariate mutual information derived from daily closing prices Figure 6 displays communities identified using discrete bivariate mutual information between daily closing prices, with results from the Leading Eigenvector method on the left and the Girvan–Newman method on the right. Emerging discrete bivariate mutual information network communities The Leading Eigenvector method reveals two major communities in Fig. 6 Community 1 consists primarily of a mix of financial and manufacturing companies indicating strong intra-sector mutual information linkages Community 2 also includes a diverse mix of sectors highlighting the presence of inter-sector dependencies and shared market behaviours The Girvan–Newman method identifies four distinct communities which suggests significant cross-sectoral informational relationships Community 2 and Community 4 display clusters of manufacturing and financial companies reflecting strong sector-specific mutual information linkages The transportation and energy sectors notably centre in Community 3 indicating specialized inter-sectoral connections The methods indicate that companies within the same sector tend to form clusters due to high mutual information values reflecting similar stock price movements driven by common economic factors and sector-specific trends The differences in community structures between the two methods provide complementary views on how informational relationships manifest within the network emphasizing both intra-sector and cross-sector dependencies in the financial market these clustering results highlight the crucial role of sectoral affiliation in forming a mutual information network based on daily closing prices particularly emphasizing the financial and manufacturing sectors In the discrete bivariate mutual information network, community structures reveal different dynamics, as can be seen in Table A5 companies like BIMAS in the commercial sector and TUPRS in manufacturing stand out with high-weighted vertex degrees and significant eigenvector centrality with a weighted vertex degree of 12.8068 and an eigenvector centrality of 0.0327421 indicating strong interconnections and influence This centrality suggests that BIMAS is crucial for the robustness and cohesion of its network cluster TUPRS shows a significant presence with a vertex degree of 8.4654 reflecting its substantial role in community connectivity and information dissemination Figure 7 displays the network, its weighted vertex degree distribution, and edge weights, as well as the network that has been thresholded using the p-values of the linear casualty test. For the linear causality test, we use the order of 1. TMFG with linear causality and weighted degree distribution The network topology in Fig. 7 shows a well-connected structure with several clusters particularly within the financial and manufacturing sectors indicating strong linear causal relationships among companies This suggests that stock price movements in these sectors can predict each other reflecting shared market influences and sector-specific dynamics The weighted vertex degree distribution on the top right reveals that most companies have a moderate number of connections indicating their significant influence within the network These central companies are likely key players whose price movements are pivotal in driving overall market trends The edge weight distribution on the bottom right shows that most connections have high weights indicating strong linear causality between company closing prices This suggests that the price movements of these companies are highly predictive of each other further emphasizing the interdependence within sectors the network topology and distributions suggest a robust structure The presence of strong linear causal relationships highlights the importance of these sectors in the market’s overall behaviour Using the Leading Eigenvector and the Girvan–Newman algorithm, the network communities depicted in Fig. 8 were generated. Emerging filtered linear causality network communities Figure 8 shows the communities identified in the linear causality network based on company daily closing prices using the Leading Eigenvector method on the left and the Girvan–Newman method on the right The Leading Eigenvector method reveals six communities Financial and manufacturing companies primarily compose Community 1 and Community 2 indicating strong intra-sector linear causal relationships suggesting the presence of significant inter-sector dependencies Community 6 shows a smaller cluster with notable interactions within the energy sector The Girvan–Newman method identifies seven communities with a slightly different structure While Community 3 and Community 4 centre around key financial and commercial companies reflecting strong internal causal influences Community 1 and Community 2 also display a mix of sectors Communities 5–7 demonstrate clusters that include companies from various sectors underscoring inter-sectoral causal connections Both methods highlight that companies within the same sector tend to form clusters due to strong linear causal relationships indicating that their stock price movements are predictive of each other Table A6 displays linear causality network communities It shows that financial companies such as AKBNK and GARAN are at the centre of Community 1 which means they play a big part in how the network changes over time based on linear causal relationships derived from daily closing prices Community 2 also features key financial players such as HALKB and ISCTR highlighting the strong interconnections within the financial sector Communities 3 and 4 prominently group manufacturing companies such as TOASO This pattern shows that linear causality is a good way to explain sector-based clustering It shows how network causal influences line up with the boundaries between sectors the Leading Eigenvector and Girvan–Newman community detection methods consistently observe these sectoral clusters demonstrating their robustness in identifying meaningful community structures the weighted vertex degree and eigenvector centrality metrics further emphasize the importance and connectivity of these companies within their respective communities highlighting their pivotal roles in market dynamics AKBNK and GARAN exhibit high vertex degrees and eigenvector centrality indicating their influential positions within the financial sector manufacturing giants like TOASO and VESTL show significant centrality metrics within their clusters underscoring their critical roles in the manufacturing sector’s network dynamics TMFG with nonlinear causality and weighted degree distribution The network topology depicted in Fig. 9 exhibits a complex and interdependent structure with multiple core nodes particularly within the banking and manufacturing sectors This suggests the presence of robust nonlinear causal linkages among these organizations The upper right corner displays the vertex degree distribution which accounts for the weights of the connections It indicates that most companies have a moderate number of connections while certain organizations have a significantly high degree suggesting their central position in the network These pivotal companies are important actors whose unpredictable price swings have a substantial impact on other enterprises in the network reveals that most connections have high weights showing a significant nonlinear causation between company closing prices The closely connected price fluctuations of these companies highlight intricate connections among sectors the network topology and distributions indicate a strong and resilient structure with the financial and manufacturing sectors playing important roles Figure 10 shows the nonlinear network communities that were detected by using the Leading Eigenvector and the Girvan–Newman algorithm. Emerging filtered nonlinear causality network communities In Fig. 10 the Leading Eigenvector method reveals six communities Manufacturing and financial companies predominantly compose Community 1 and Community 4 indicating strong intra-sector nonlinear causal relationships highlighting significant inter-sector dependencies Community 6 shows a cluster primarily centred around the commercial sector with sector-specific nonlinear interactions The Girvan–Newman method identifies nine communities with Community 4 being the largest and most central incorporating a mix of sectors that suggests significant cross-sectoral nonlinear causal relationships Community 1 and Community 2 display a concentration of financial and manufacturing companies indicating strong sector-specific nonlinear influences Communities 5 through 9 demonstrate smaller clusters underscoring the presence of inter-sectoral interactions Both methods stress that companies in the same sector tend to group together because of strong nonlinear causal relationships This is because similar stock price changes are caused by complex market dynamics and factors that are unique to the sector Mixed-sector communities are present in both methods indicating significant cross-sectoral interactions within the market Table A7 illustrates the nonlinear causality network communities showcasing a diverse clustering pattern based on nonlinear relationships between company daily closing prices Community 1 prominently features energy companies like ZOREN and manufacturing companies like SISE and TOASO indicating strong nonlinear causal connections within these sectors Financial firms like GARAN and manufacturing companies like EREGL are central in Community 2 highlighting their significant roles in nonlinear interactions This table confirms that nonlinear causality induces a unique clustering dynamic that intertwines multiple sectors The presence of companies from different sectors within the same community emphasizes the complex multifaceted nature of nonlinear relationships in financial markets This intricacy is evident in companies such as AKBNK and TUPRS which exhibit high eigenvector centrality and influence playing pivotal roles across multiple communities This further underscores the importance of considering nonlinear causality to fully understand the interdependencies and the structural topology of financial networks Figure 11 displays the series that were derived from the edge evaluation using a random walker with a larger transition probability compared to low edge weights. A random walk method with 10,000 steps was chosen. Node visiting values of random walker for different networks In Fig. 11 specific vertices exhibit very high visit frequencies indicating their significant centrality and influence within the network Similar patterns emerge in the continuous and discrete mutual information networks where certain nodes receive more frequent visits indicating their crucial roles in information sharing Linear causality network reveals prominent nodes with high visit frequencies emphasizing their strong predictive relationships Nonlinear causality network displays a more distributed pattern with several vertices showing high centrality we enable the function to consider the impact of previous values on subsequent ones We usually set the quantile threshold for calculating transfer entropy at 0.1 to give more weight to extremes of distributions We use Shannon’s entropy to calculate the transfer entropy We determine the uncertainty of the transfer entropy estimate using 300 bootstrap samples and maintain the stability of the bootstrap estimate by discarding 30 initial samples The resulting transfer entropies in the form of \({G}_{F}1\to {G}_{F}2\) are presented in Table 4 The values in Table 4 show transfer entropies between various types of networks These values come from random walker distribution Transfer entropy quantifies the amount of directed (asymmetric) information transferred between two processes Higher values indicate greater predictability or influence from one network to another we see relatively high transfer entropy values when compared to other networks with the highest influence observed towards the nonlinear causality network (0.028385) This suggests that information flows from the correlation network to the nonlinear causality network is significant indicating that relationships captured by the correlation are relevant to the nonlinear dependencies The linear causality network has the highest transfer entropy for the continuous mutual information network (0.0333629) This means that continuous mutual information is most useful for understanding linear causal relationships This highlights the importance of mutual information in understanding linear dependencies in the network The discrete mutual information network has moderate transfer entropy values across all other networks with the highest influence on the linear causality network (0.0280271) Discrete mutual information captures substantial interactions The nonlinear causality network receives the highest transfer entropy (0.039678) from the linear causality network indicating a strong influence of linear causal relationships on nonlinear dependencies This high value suggests that linear causality encapsulates critical aspects of the dynamics that are further detailed in nonlinear interactions the nonlinear causality network shows significant transfer entropy towards the linear causality network (0.113118) This shows how complicated and two-way this influence is between linear and nonlinear causal networks It also shows how nonlinear dependencies offer a bigger picture to understand linear causal effects transfer entropy values illustrate the interconnectedness and predictive power between different network types Both linear and nonlinear correlation and causality networks have significant values This suggests that these networks are capturing important In addition to looking at the topological structures of networks that were made, it is also important to look at how well communities that were created using two different approaches can cluster. Table 5 displays the outcomes for comparative metrics specified in the preceding sections Community comparison metrics in Table 5 provide a comprehensive evaluation of community detection performance across different network types The correlation network shows a moderate VOI of 1.126 indicating a relatively average level of information loss between true and detected communities Its NMI of 0.588348 suggests a decent alignment of detected communities with true communities and ARI of 0.4722 further confirm that while communities are reasonably accurate there is room for improvement in capturing the intricate structures within the network The continuous mutual information network has the highest VOI of 1.715 indicating significant information loss and less effective community detection reflecting poor correspondence with true communities The high SJD of 66 also points to considerable discrepancies in community structure while the RI (0.6567) and ARI (0.1977) indicate relatively weak community detection performance the discrete bivariate mutual information network performs better with a lower VOI of 1.102 and a moderate NMI of 0.4528 The SJD of 37 suggests a more accurate community structure compared to the continuous counterpart The RI of 0.7126 and ARI of 0.422 indicate an overall moderate performance balancing between information preservation and accurate community detection The linear causality network exhibits the best performance Its NMI of 0.7593 is the highest among all networks showing strong alignment with true communities along with high RI (0.899) and ARI (0.6356) demonstrates superior community detection accuracy the nonlinear causality network has the lowest VOI of 0.573 Its NMI of 0.573 suggests good correspondence with true communities although not as high as the linear causality network The SJD of 53 indicates some discrepancies in community structure but the RI (0.8151) and ARI (0.3471) reflect a strong performance particularly in capturing nonlinear relationships these results highlight that while continuous mutual information struggles to capture community structures effectively linear and nonlinear causality networks provide more accurate and robust community detection reflecting their capability to model complex dependencies in financial networks The correlation network offers balanced performance but improvements can be made by leveraging causality-based methods to capture deeper insights into community structures The next phase of this study involves examining how Environmental and Governance (ESG) scores impact the clustering of companies within these networks By performing nonparametric tests on the ESG scores we aim to determine whether these sustainability metrics significantly affect the formation and structure of communities within the various network models This analysis will provide insights into the role of ESG factors in shaping financial networks and their potential to drive sustainable investment practices We will use nonparametric tests like the Kruskal–Wallis and Gehan to evaluate how ESG scores impact clustering patterns in networks These tests are particularly suitable because they do not assume a specific data distribution making them robust for evaluating the relationships between ESG scores and community formation we can rigorously evaluate whether differences in ESG scores across various clusters are statistically significant providing a deeper understanding of how sustainability practices influence network topology Each of these sub-scores represents a critical aspect of a company sustainability performance and can provide nuanced insights into how different elements of ESG influence network dynamics higher scores in Innovation or Resource Use might be indicative of companies that are more central in certain network communities due to their advanced practices and efficient operations We will treat ESG scores as an independent variable and the community memberships obtained from the various network models as dependent variables By comparing the distribution of ESG scores across different communities we can identify whether certain sustainability practices are more prevalent in specific clusters indicating a correlation between ESG performance and network position This approach will help in understanding whether companies with superior ESG metrics tend to cluster together potentially due to similar strategic goals or operational practices that align with sustainability principles this analysis will explore whether certain ESG dimensions are more influential in determining community structure within specific types of networks we might find that companies with high Human Rights scores cluster together reflecting their commitment to social governance aspects Innovation scores might play a more significant role in community formation highlighting the importance of technological advancements and innovative practices in these clusters Table 6 presents the results of nonparametric tests for ESG scores across correlation network communities highlighting significant findings for various sustainability metrics In Table 6 the Kruskal–Wallis test results for the Leading Eigenvector method indicate that Product Responsibility (\(p=0.061\)) and Human Rights (\(p=0.092\)) are marginally significant suggesting that these factors might influence community formation to some extent shows a more pronounced significance for Resource Use (\(p=0.017\)) indicating a strong impact of resource management practices on community clustering This is further supported by the log-rank test which highlights Product Responsibility (\(p=0.0425\)) and Human Rights (\(p=0.092\)) as notable influencing variables the Kruskal–Wallis test reveals significant results for Innovation (\(p=0.0265\)) and CSR Strategy (\(p=0.008\)) suggesting these elements are key drivers in community formation The Conover test corroborates the significance of Resource Use (\(p=0.0182\)) and Innovation (\(p=0.0748\)) indicating consistent findings across different test methodologies The log-rank test underscores the importance of CSR Strategy (\(p=0.005\)) and Innovation (\(p=0.012\)) further validating the influence of ESG factors on network communities When the findings from the correlation network topologies are considered alongside the results of the nonparametric tests for ESG scores particularly the Conover test for the Leading Eigenvector method indicate that Resource Use significantly affects community formation This suggests that companies with similar resource use practices are more likely to cluster together within the network reinforcing the idea that sustainability strategies are a critical factor in driving stock price correlations The log-rank test further supports this by highlighting the significance of Product Responsibility (\(p=0.0425\)) and Human Rights indicating that these ESG factors also influence clustering the nonparametric tests underscore the importance of Innovation and CSR Strategy suggesting that companies with strong innovation and CSR strategies tend to form distinct communities This aligns with the observed clustering patterns where companies with similar ESG practices are more likely to be grouped together The significance of these ESG factors indicates that sustainable practices are not only important for individual company performance but also play a crucial role in the broader market dynamics and community structures within financial networks the integration of network community analysis and nonparametric test findings highlights the interplay between sectoral affiliation and ESG practices in determining stock price correlations The distinct communities formed within the manufacturing and financial sectors reflect the shared market influences and sectoral trends Table 7 displays the outcomes of nonparametric tests conducted on ESG scores across continuous mutual information network communities The nonparametric test results for ESG scores over continuous mutual information network communities, as presented in Table 7 The tests reveal mixed significance levels across different ESG dimensions and community detection methods shedding light on the nuanced impact of sustainability metrics on network clustering the Kruskal–Wallis test indicates that none of the ESG dimensions show significant differences between communities (all p-values > 0.05) This suggests that ESG scores do not substantially influence the formation of communities when using this method the Conover and log-rank tests do not show any significant results reinforcing the lack of substantial impact of ESG scores on community structure in this context the Girvan–Newman method reveals significant results for several ESG dimensions The Kruskal–Wallis test shows that Emission (\(p=0.032\)) and CSR Strategy (\(p=0.07\)) have lower p-values The Conover test highlights no significant differences while the log-rank test identifies Emission (\(p=0.034\)) and CSR Strategy (\(p=0.065\)) as significantly influencing the clustering of companies These findings suggest that certain ESG scores play a more pronounced role in defining community structures within the Girvan–Newman-derived networks In the continuous mutual information network the Girvan–Newman algorithm reveals that Community 1 predominantly consists of diverse sector companies indicating a broad mix influenced by various ESG factors Community 2 includes key financial and commercial companies which are likely influenced by strong governance and corporate responsibility practices which display a blend of manufacturing and technology sectors suggest that companies within these sectors share similar innovation and resource use strategies Findings highlight how sustainability metrics play a critical role in shaping network topology particularly in the continuous mutual information network Companies with high scores in these ESG areas tend to form distinct clusters emphasizing the importance of shared sustainability practices This clustering not only reflects similar market behaviours but also underscores the systemic impact of ESG performance on financial networks Table 8 displays the outcomes of nonparametric tests conducted on ESG scores across discrete bivariate mutual information network communities The nonparametric test results from Table 8 for ESG scores over discrete bivariate mutual information network communities present several notable insights significant p-values were observed for CSR Strategy in both the Kruskal–Wallis and log-rank tests indicating a substantial impact of CSR Strategy scores on community formation This suggests that companies with similar CSR strategies are more likely to cluster together reflecting shared commitments to corporate social responsibility and potentially similar operational practices Resource Use showed significance in the Conover test pointing towards its role in influencing how companies group within the network only community is significant according to the Conover test reinforcing its importance for community clustering companies with higher resource efficiency and stronger corporate social responsibility practices tend to cluster together reflecting shared sustainability priorities which show significant values in these ESG categories emphasizing their alignment in sustainability practices ESG test results thus complement the community structures observed in the discrete bivariate mutual information network particularly in Resource Use and CSR Strategy underscoring the influence of sustainability practices on their stock price interrelations This alignment suggests that shared ESG goals and operational practices drive the formation of these communities reflecting the interconnectedness of sustainability performance and financial behaviour The significant ESG factors highlight key areas where company sustainability practices are most impactful shaping the network topology and providing insights into drivers of sustainable investment Table 9 presents the results of nonparametric tests for ESG scores across linear causality network communities The nonparametric test results presented in Table 9 for ESG scores over the linear causality network communities reveal some significant findings particularly for the Leading Eigenvector method the Kruskal–Wallis test shows significant results for Human Rights (\(p=0.012\)) indicating that there is a meaningful difference in the Human Rights scores across different communities This is further supported by the Conover test (\(p=0.003\)) and the log-rank test (\(p=0.022\)) suggesting that companies with similar Human Rights scores tend to cluster together more tightly in the network This finding implies that HR practices are a critical factor influencing the formation of communities within the linear causality network the log-rank test reveals significant results for Emission (\(p=0.034\)) and Innovation (\(p=0.02\)) scores in the Girvan–Newman method indicating that these ESG metrics also play a substantial role in how companies cluster This suggests that companies with similar Emission and Innovation scores are more likely to form communities reflecting shared sustainability practices and innovation strategies Results highlight that specific ESG scores significantly impact the clustering of companies within the linear causality network indicate the presence of significant inter-sector dependencies This is reflected in the significant Kruskal–Wallis test results for Human Rights under the Leading Eigenvector method suggesting that companies with similar Human Rights scores tend to cluster together This clustering could be driven by shared practices and policies that transcend sector boundaries indicating that human rights practices are crucial in forming these inter-sectoral links which shows notable interactions within the energy sector aligns with the significant results for Emission scores under the Girvan–Newman method This suggests that companies within the energy sector with similar Emission scores are more likely to form clusters reflecting common environmental practices and regulatory influences The Girvan–Newman method identifies seven communities with a structure that emphasizes both intra- and inter-sector dependencies The significant results for Innovation scores indicate that companies with similar innovation practices are likely to cluster together This suggests that innovation is a key factor in forming causal connections within the network with companies sharing similar Innovation scores showing predictive stock price movements the Conover test results for community scores under the Leading Eigenvector method highlight the importance of community engagement and social responsibility in shaping network communities Companies with strong community engagement practices are likely to be interconnected within the same community driven by shared social goals and responsibilities Table 10 displays the outcomes of nonparametric tests conducted on ESG scores across nonlinear causality network communities The nonparametric test results presented in Table 10 for ESG scores over nonlinear causality network communities provide valuable insights into how specific sustainability metrics influence the clustering of companies several ESG scores show significant impacts on community formation The Kruskal–Wallis test indicates that Innovation (\(p=0.019\)) and Emission (\(p=0.074\)) scores significantly affect community structure These results suggest that companies with similar innovation capabilities and emission policies are more likely to cluster together The Conover test further supports the influence of Human Rights (\(p=0.073\)) and Workforce (\(p=0.17\)) on community formation although it shows slightly different significance levels The log-rank test reveals significant impacts for Emission (\(p=0.056\)) emphasizing the role of environmental and social governance factors in driving network clustering The Girvan–Newman results align with these findings to some extent but also present unique insights The Kruskal–Wallis test highlights the significant impact of Human Rights (\(p=0.057\)) and Shareholders (\(p=0.081\)) suggesting that these factors are crucial in determining how companies group within the network Product Responsibility (\(p=0.013\)) and Workforce (\(p=0.079\)) had a significant impact The log-rank test underscores the importance of Human Rights (\(p=0.023\)) and Shareholders (\(p=0.059\)) suggesting that companies with similar practices in these areas tend to form tighter-knit communities The nonparametric test results for ESG scores over nonlinear causality network communities reveal critical insights into the role of sustainability metrics in influencing company clustering Financial companies such as AKBNK and GARAN highlight the importance of the Emission and Human Rights scores in community formation The log-rank test shows significant results for Emission and Human Rights within the Leading Eigenvector method This suggests that companies with lower emissions and strong human rights policies are more likely to cluster together reflecting shared sustainability practices that influence their stock price movements which features manufacturing giants like TOASO and VESTL shows the impact of Innovation and Human Rights scores on clustering The Kruskal–Wallis test for Innovation and Human Rights indicates that companies excelling in these areas form tighter communities driven by shared innovative practices and robust human rights policies This emphasizes the role of ESG factors in shaping sector-specific causal relationships The diverse sectoral mix in Communities 3–5 suggests significant inter-sector dependencies The Conover test for Management within the Leading Eigenvector method highlights that strong management practices are critical for inter-sectoral clustering This shows that companies with effective management strategies tend to form communities with interconnected stock price movements The energy sector’s notable interactions in Community 6 align with significant log-rank test results for Resource Use and Innovation These findings indicate that companies with efficient resource use and innovative practices within the energy sector form strong causal links reflecting shared operational efficiencies and sustainable practices This study aimed to investigate the clustering behaviour of companies within different network models based on daily closing prices and examine the influence of ESG scores on these clusters By applying various network analysis methods (i.e. continuous and discrete mutual information networks we effectively captured the complex interdependencies and interactions among companies The nonparametric tests performed on ESG scores allowed us to assess the impact of sustainability metrics on community formation fulfilling the primary objective of understanding how ESG factors shape financial networks The comparison of different network models revealed distinct and multifaceted patterns of interconnectivity among companies offering a comprehensive view of market dynamics The correlation networks underscored strong sectoral linkages Continuous mutual information networks provided deeper insights into both intra- and inter-sector dependencies reflecting more nuanced informational relationships that go beyond simple linear correlations These networks highlighted how certain sectors are interconnected through shared information flows revealing a complex web of market interactions Discrete mutual information networks emphasized the importance of specific informational exchanges between companies This model illustrated how discrete events or data points can drive connectivity showcasing the critical role of individual company actions and announcements in shaping market linkages linear causality networks demonstrated how stock price movements are predictive of each other capturing the directional influences and temporal dependencies within the market This approach highlighted the causal pathways through which market shocks or trends propagate across different companies and sectors Nonlinear causality networks expanded this perspective by capturing more complex nonlinear relationships that may not be evident through linear models This method revealed hidden dependencies and intricate feedback loops within the market underscoring the rich and dynamic nature of financial interactions diverse network perspectives underscored the varying degrees of connectivity and influence within the financial market illustrating the complexity and richness of market dynamics we obtained a holistic understanding of how companies are interlinked and the intricate web of causal relationships that drive market behaviour The community detection results across different network models consistently highlighted the significant role of sectoral affiliation in forming clusters The Leading Eigenvector and Girvan–Newman methods both revealed that companies within the same sector This indicates strong intra-sector relationships driven by shared market behaviours and sector-specific factors The robustness of these methods in identifying sectoral clusters demonstrates their effectiveness in capturing the underlying structure of market networks the financial and manufacturing sectors prominently formed distinct communities reflecting the high degree of correlation within these sectors This pattern indicates that companies within these sectors exhibit similar stock price movements likely influenced by common economic drivers and market conditions Sector-specific clusters provide valuable insights into how market participants within the same industry are interconnected The continuous mutual information networks further emphasized the importance of sectoral clustering while also revealing significant inter-sector dependencies Mixed-sector communities identified in these networks highlight how different industries are interlinked through shared informational relationships financial and commercial companies forming a single community suggest that these sectors may be influenced by common external factors Discrete mutual information networks showed specific informational exchanges between companies illustrating how discrete events or data points can drive clustering This approach highlighted the critical role of individual company actions in shaping network structures with sectoral affiliations still playing a key role in forming communities Linear and nonlinear causality networks captured the directional and complex relationships between companies Such networks showed that company stock price movements can predict each other reinforcing the importance of causal influences in market dynamics Clustering results from these models revealed strong internal causal influences within sectors like financial and manufacturing while also showing cross-sectoral causal connections the robust clustering patterns observed across different methods underscore their effectiveness in identifying meaningful community structures These results provide valuable insights into the underlying drivers of market connectivity highlighting the importance of both intra-sector relationships and inter-sector dependencies in shaping the financial network topology we can better grasp the dynamics of market behaviours and factors driving company interconnectivity The nonparametric test results for ESG scores revealed significant insights into the role of sustainability metrics in community formation within financial networks and human rights emerged as significant influence variables in different network models in the Girvan–Newman method applied to nonlinear causality network significant results for Emission and CSR Strategy highlighted their impact on clustering Findings suggest that companies with strong performance in these ESG areas tend to form more cohesive communities reflecting shared sustainability practices and priorities Companies with lower emissions or more robust CSR strategies are more likely to be interconnected within the same community driven by common goals and operational practices related to sustainability This underscores the critical role of ESG factors in shaping financial networks and promoting sustainable investment practices ESG factors such as Human Rights and Innovation showed significant p-values indicating their substantial impact on community structure Companies excelling in human rights practices or innovation tend to cluster together suggesting that these aspects of ESG performance foster strong informational linkages and shared market behaviours the linear causality network revealed significant results for ESG scores like Human Rights and Management highlighting how companies with strong governance and social responsibility practices influence market connectivity The discrete bivariate mutual information network also emphasized the importance of specific ESG factors Significant results for product responsibility and shareholders indicated that these aspects of sustainability play crucial roles in forming community structures Companies with high scores in product responsibility tend to cluster together reflecting their commitment to quality and ethical business practices This clustering based on shared sustainability priorities reinforces the idea that ESG performance drives market behaviours and intercompany relationships findings across different network models and community detection methods demonstrate the multifaceted impact of ESG factors on financial networks They highlight how sustainability metrics not only influence individual company performance but also shape broader market dynamics by fostering interconnected cohesive communities of companies with similar ESG practices This reinforces the importance of integrating ESG considerations into investment decisions to promote sustainable and responsible financial markets the reliance on daily closing prices may not capture all market dynamics and intra-day variations the ESG scores used were limited to publicly available data which may not fully represent the companies’ overall sustainability performance The static nature of network models also does not account for temporal changes in company relationships and market conditions the study focused on a specific set of companies within the Borsa Istanbul Sustainability Index which may limit the generalizability of findings to other markets or indices the nonparametric tests (while robust) may not fully capture the complex interactions between ESG factors and network communities Future research could address these limitations by incorporating intra-day trading data to capture more granular market dynamics and provide a more detailed understanding of company interdependencies Expanding the dataset to include a broader range of companies and multiple stock indices could enhance the generalizability of findings longitudinal studies that track the evolution of network structures and community formations over time would provide valuable insights into temporal changes and the impact of external events on market connectivity Investigating the impact of different types of ESG scores such as those derived from third-party assessments or alternative data sources like social media sentiment could offer a more comprehensive view of sustainability practices integrating machine learning techniques to dynamically adjust network models based on changing market conditions could enhance the robustness and predictive power of the analysis Exploring the interactions between ESG scores and other financial metrics could provide a deeper understanding of how sustainability influences broader market behaviours examining the role of regulatory changes and policy interventions on ESG-driven clustering could inform policymakers and investors about the effectiveness of sustainability regulations Conducting cross-market comparisons to assess how ESG factors influence network structures in different economic contexts could yield valuable insights into the global applicability of the findings developing real-time monitoring tools based on network analysis and ESG metrics could help investors make informed decisions and promote sustainable investment practices this study highlights the intricate relationships between ESG factors and financial network structures demonstrating the significant impact of sustainability metrics on market connectivity The application of various network models and nonparametric tests provided a comprehensive understanding of how companies cluster based on their financial performance and sustainability practices These findings underscore the importance of integrating ESG considerations into financial analysis paving the way for more sustainable and informed investment strategies embracing advanced analytical techniques and expanding the scope of research will further enhance our understanding of the complex interplay between sustainability and market dynamics ultimately contributing to the development of more resilient and sustainable financial systems The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request Akgüller Ö (2019) A threshold method for financial networks and geometric scattering of agents Commun Stat Case Stud Data Anal Appl 5(3):230–242 Lim E-P (2023) On predicting ESG ratings using dynamic company networks Gürbüz AO (2019) ESG and financial performance: impact of environmental and governance issues on corporate performance Yüksel S (eds.) Handbook of research on managerial thinking in global business economics Batrancea LM (2022) Determinants of economic growth across the European union: a panel data analysis on small and medium enterprises Gaban L (2022a) What drives economic growth across European countries Gaban L (2022b) Sources of SMEs financing and their impact on economic growth across the European Union: insights from a panel data study spanning sixteen years Akgüller Ö (2023a) Empirical investigation on how wellbeing-related infrastructure shapes economic growth: evidence from the European Union regions Batrancea I (2021) A panel data analysis of economic growth determinants in 34 African countries Batrancea I (2022c) A panel data analysis on determinants of economic growth in seven non-BCBS countries Tulai H (2023b) Determinants of economic growth for the last half of century: a panel data analysis on 50 countries Zimmermann DJ (2019) RTransferEntropy-quantifying information flow between different time series using effective transfer entropy Makarenko I (2022) Persistence in ESG and conventional stock market indices Lau FC (2010) A network perspective of the stock market Koroleva D (2021) The impact of ESG factors on asset returns: empirical research Fernandez-Izquierdo MA (2017) Integrating multiple ESG investors preferences into sustainable investment: a fuzzy multicriteria methodological approach Ferretti S (2023) On the modeling and simulation of portfolio allocation schemes: an approach based on network community detection Nishikawa L (2019) Foundations of ESG investing: how ESG affects equity valuation Newman ME (2002) Community structure in social and biological networks Zhang C (2024) The enhanced benefits of ESG in portfolios: a multi-factor model perspective based on LightGBM Tian T (2018) Development of stock correlation networks using mutual information and financial big data Kirkulak-Uludag B (2018) Sustainable business practices of Turkish companies listed on the Borsa Istanbul Sustainability Index Vilag RD (2019) The impact of ESG factors on market value of companies from travel and tourism industry Adelina YE (2020) The impact of ESG performance to firm performance and market value Kim S (2023) Learning financial networks with high-frequency trade data Şavlı T (2024a) Modeling the link between environmental and governance disclosures and scores: the case of publicly traded companies in the Borsa Istanbul Sustainability Index Ulussever T (2024b) The relationship between monetary policy and financial asset returns in Türkiye: time Khoojine NS (2023) Analyzing volatility patterns in the Chinese stock market using partial mutual information-based distances Li Z (2021) Understanding the impact of ESG practices in corporate finance Grassberger P (2004) Estimating mutual information Jung HS (2024) Deep-learning-based stock market prediction incorporating ESG sentiment and technical indicators social and governance (ESG) practices and financial performance of shariah-compliant companies in Malaysia Zhang Y (2023) Can firms with higher ESG ratings bear higher bank systemic tail risk spillover Deng M (2024) Can good ESG performance reduce stock price crash risk Corp Soc Responsib Environ Manag 31(3):1469–1492 Lupu R (2022) How is the ESG reflected in European financial stability Ma N (2023) Integration of ESG factors in portfolio management: international trends and practices Silaban A (2023) Exploring the measurement of environmental performance in alignment with environmental Millington T (2022) An investigation into the effects and effectiveness of correlation network filtration methods with financial returns Fain M (2020) Performance measurement of ESG-themed mega-trend investments in global equity markets using pure factor portfolios methodology Newman ME (2006) Finding community structure in networks using the eigenvectors of matrices Aste T (2020) Information network modeling for US banking systemic risk Oh K-S (2022) Integration of ESG information into individual investors’ corporate investment decisions: utilizing the UTAUT framework Lee HS (2023) The roles of finance in ESG management Kartal MT (2024) Assessing the influence of climate risk and technological factors on the ESG market in the European Union Wojcik D (2021) Connectivity and growth: financial centres in investment banking networks Environ Plann A Econ Space 53(7):1789–1809 Kenett DY (2021) Interconnectedness in the global financial market Abbas K (2022) Applying artificial intelligence techniques for predicting the environment and governance (ESG) pillar score based on balance sheet and income statement data: a case of non-financial companies of USA Boginski V (2022) Networks of causal relationships in the US stock market Ziolo M (2019) Inclusion of ESG factors in investments and value addition: a meta-analysis of the relationship In: Effective investments on capital markets: 10th Capital Market Effective Investments Conference (CMEI 2018) Zuev K (2017) Causality networks of financial assets Zuev KM (2020) Unveiling causal interactions in complex systems Shchepeleva M (2022) The impact of ESG-factors on financial stability Giacometti R (2023) Financial contagion in banking networks with community structure Commun Nonlinear Sci Numer Simul 117:106924 Aste T (2022) Simplicial persistence of financial markets: filtering Panigrahi PK (2020) Causal evolution of global crisis in financial networks Jensen HJ (2020) Allometric scaling of mutual information in complex networks: a conceptual framework and empirical approach Chakrabarti AS (2020) Measuring complexity in financial data Tatoğlu E (2020) Does the stock market value inclusion in a sustainability index Cho JH (2024) Corporate social responsibility and financial performance: new evidence from the Korean market Pedrini M (2020) ESG factor integration into private equity Zhang AY, Zhang JH (2023) Renovation in environmental, social and governance (ESG) research: the application of machine learning. Asian Rev Account. https://doi.org/10.1108/ARA-07-2023-0201 Bian D (2023) Detection of illegal transactions of cryptocurrency based on mutual information Razak FA (2022) The causality and uncertainty of the COVID-19 pandemic to Bursa Malaysia financial services index’s constituents Download references Download citation DOI: https://doi.org/10.1057/s41599-024-03527-y Metrics details This research investigates how changes in public discourse on social media (particularly Twitter now X) influence financial performance of companies listed on the e-commerce XTCRT index of Borsa Istanbul amid the COVID-19 pandemic Utilising a unique 4-layer network structure and trading volume data of these companies with Twitter discussions on six categories: culture By using advanced computational methods such as multilayer network analysis nonlinear autoregressive distributed lag analysis we find a high responsiveness of the e-commerce index to social media content particularly in discussions related to health The study underscores that social media discussions significantly shape index composition resulting in noticeable shifts in market dynamics Our study advocates for a more advanced approach to market analysis that incorporates the dynamics of online public discussions Analyses on how these organisations employ technology to enhance productivity expand customer base and generate additional income deepen the understanding of changing dynamics in international trade with particular significance for investors and policymakers operating in an evolving digital landscape control interconnected systems in an increasingly interconnected world This analytical technique can be a useful instrument for policymakers to comprehend and exert influence over public sentiment assess the efficacy of communication tactics and enhance crisis management capabilities Policymakers can gain insights into public discourse by analysing the dissemination of information and attitudes through different social and communication networks This understanding is crucial for influencing public health communication economic strategies and social interventions the analysis can assist in identifying and deterring misinformation and disinformation operations that might harm public health initiatives and incite societal turmoil it is crucial to retain public trust and ensure adherence to health rules Comprehending the emotional tone and prevailing views in online discussions can aid in formulating empathic and impactful public messages which are crucial for efficient administration it can assist in forecasting and alleviating the societal consequences of policy choices by offering insights into public reception and potential responses examining structural characteristics of online discourse networks during significant events such as the COVID-19 pandemic provides a valuable method for monitoring public discourse creating well-informed interventions and promoting a governance approach that is more responsive and adaptable This can ultimately result in enhanced policy outcomes and a more knowledgeable public Our study focuses on companies that are included in the commerce index (i.e. A 4-layer network architecture is constructed by using daily closing prices highest and lowest values and trading volume values of companies throughout the COVID-19 period We investigated the impact of changes in discourse topics on the main structural and topological metrics of the multilayer network The network in question is comprised of individuals who posted various topics on Twitter (now referred to as X) We used deep learning and natural language processing techniques to analyse Twitter topics across six distinct categories: culture Identifying the six categories is crucial for various reasons It facilitates comprehension of the current issues and interests of a society or group This classification also facilitates focused policy formulation marketing tactics and other decision-making procedures that necessitate comprehension of public interest and sentiment across different areas The novelty of our study is that it uses multilayer financial network analysis to offer a broader perspective on the numerous financial relationships and correlations generated by listed companies while two company stocks may move in tandem their reactions to economic indicators or global events might diverge our approach facilitates an understanding of the interconnectedness of different market segments and how fluctuations in one area can ripple through to others By analysing interconnections across various market types financial analysts can identify systemic risks and hidden vulnerabilities that span across multiple market segments It is crucial for developing effective risk mitigation strategies and understanding the complex structure of financial contagion Results showed that public discourse on social media has an impact on short- and long-term company performance public policies should be designed while considering public view and trust in financial markets authorities could support public awareness initiatives for increasing levels of financial literacy and market dynamics among citizens which could yield a knowledgeable public discourse our results suggested that specific conversations about health technology and world events are of interest and concern for citizens authorities are called to implement cross-sectoral policy approaches to better serve public interest and market realities This study is structured as follows: “Literature review” provides an extensive analysis of the impact of social media posts on economies “Methodology” details the methodology employed in the study “Results” presents the outcomes of multiplex analysis and topic classification “Results based on NARDL” focuses on the results using Nonlinear Autoregressive Distributed Lag (NARDL) analysis “Conclusions” provides concluding remarks and future research directions emphasising its impact on the economy and social media social media has significantly changed how information is shared and digested It empowers anyone to create and share content instantly on a worldwide scale social media serves as an expansive virtual platform where users may discuss and exchange ideas These analyses facilitate the understanding of the intricate link between social media activity and commercial enterprises business decisions information shared on social media triggers emotional responses and reactions among investors These emotional responses then impact their choices to buy or sell financial assets (especially in stock markets) and eventually result in observable behaviours the spread of news on social media (regardless of their truthfulness) can change the trends of international or national stock market indicators discourse in news or media outlets and that new content can predict changes in market values it indicates that investors’ perceptions on risk may have a stronger impact on their actions than concrete facts The author highlights the importance of establishing consistent regulations worldwide to prevent market participants from mitigating risk-taking capacity and to ensure the availability of liquidity in global markets we offer a comprehensive examination of how social media influences the financial performance of e-commerce companies during the COVID-19 pandemic By addressing gaps through advanced computational methods and a sector-specific focus our study significantly contributes to the understanding of how social media discourse shapes financial market dynamics particularly for e-commerce companies listed on Borsa Istanbul this approach is vital for understanding intricate networks where companies are linked not just through direct trading ties but also through shared stakeholders market influences and sectoral dependencies retaining significant connections while simplifying overall structure we assume companies are interconnected within a complete correlation distance applying PMFG filtration to enhance the network’s topological structure A multilayer network is defined as a quadruplet \(M=({V}_{M},{E}_{M},V,L)\) where \({V}_{M}\) represents the node-layer combinations and \({E}_{M}\) the filtered edge set The multilayer network enables a more comprehensive view of financial relationships across different layers each representing a distinct aspect of the market is calculated by summing the weights of adjacent edges within a layer The node’s weighted edge overlapped degree \({o}_{i}\) across all layers is given by the multilayer participation coefficient \(P(i)\) measures a node’s involvement across layers with higher values indicating more balanced participation Nodes are categorised based on \(P(i)\): ‘focused’ (\(0 \,<\ 2/3\)) and ‘truly multilayer’ (\(P\left(i\right)\ge 2/3\)) This classification aids in understanding the roles of different nodes within the network The Z-score \(z({o}_{i})\) standardises the overlapping strength we calculate the entropy \(H\) of the multilayer network to measure its complexity where \({p}_{i,\alpha }\) is the probability of a specific state involving node \(i\) in layer \(\alpha\) This metric assesses the unpredictability and information distribution across the multilayer network we gain deeper insights into the interconnected dynamics of Borsa Istanbul financial entities revealing how various layers of interaction contribute to the overall market structure This comprehensive understanding enables analysts and investors to identify critical nodes and connections that may pose systemic risks or offer strategic opportunities it facilitates more informed decision-making allowing for the development of robust investment strategies and resilient financial models that can better withstand market shocks and uncertainties the ability to dissect and analyse these multilayered networks equips stakeholders with the tools needed to anticipate and mitigate potential contagion effects ultimately contributing to a more stable and secure financial environment Zemberek analyses Turkish words based on fundamental components and rectifies incorrectly spelled terms The library is employed for the processes of tokenization Machine learning algorithms often exhibit poorer performance while processing a substantial amount of data points Deep learning approaches are favoured in such instances due to their superior efficiency and improved outcomes An inherent advantage of deep learning compared to classical machine learning is its ability to bypass the requirement for feature extraction Deep learning employs an artificial neural network architecture that relies on word embedding a technique that converts words or phrases into vectors of real numbers every word is shown as a vector with randomly assigned values This study specifically utilises the GloVe (Global Vectors for Word Representation) word embedding paradigm The vector matrices are subsequently incorporated into the embedding layer of the classification model is an unsupervised model that uses co-occurrence probabilities to represent words It evaluates word similarity by analysing the proportion of times they appear together in different contexts The GloVe model employs a weighted least-squares regression equation that places greater emphasis on ratios rather than probabilities to denote associations between words Word frequencies are determined by analysing a co-occurrence matrix and a particular weighting function is utilised in the model Turkish news and tweets are classified using LSTMs Loops interconnection gives the Recurrent Neural Networks (RNNs) their characteristic dynamic temporal patterns and allows them to analyse inputs by drawing on their internal memory RNNs can only hold on to data for so long before it becomes useless which allows it to draw on past learning to generate new findings LSTMs are unique among RNNs since they have four separate layers: input a dense layer and an output layer make up the classification model used in this study Word embedding vector matrices are received by the input layer and sent to the hidden layer which contains LSTM parameters activated with a sigmoid function Overfitting risk can be reduced via dropout The activation function of the Rectified Linear Unit is used by the dense layer to improve learning and guarantee consistent output which produces a probabilistic output for various news categories where \({y}_{t}\) is the dependent variable (e.g. \({x}_{t-j}^{+}\) and \({x}_{t-k}^{-}\) represent the positive and negative components of the independent variable \({x}_{t}\) (e.g. Twitter sentiment or news impact) at time \(t\) decomposed into positive\(\,{x}_{t-j}^{+}\) and negative \({x}_{t-k}^{-}\) changes \({\alpha }_{i}\) captures the autoregressive effect of the dependent variable \({\beta }_{j}^{+}\) and \({\beta }_{k}^{-}\) are the coefficients that measure the impact of positive and negative changes in \({x}_{t}\) on \({y}_{t}\) and \({\epsilon }_{t}\) is the error term This model allows for the distinct measurement of how positive and negative shocks in social media sentiment (or other independent variables) affect the stock market differently capturing potential asymmetries in investor responses enable the analysis of the delayed effects of these shocks crucial in financial markets where responses may not be immediate The NARDL approach is particularly useful for exploring both long-term cointegration and short-term dynamics within financial data long-term relationship between Twitter topic series and multilayer network measurements indicated by the presence of cointegration among variables This capability is essential for understanding the enduring connections between social media trends and financial market behaviour the model effectively captures the immediate adjustments that occur when there are deviations from this long-term equilibrium providing insights into how quickly and to what extent the stock market responds to new information arising from social media discussions This dual focus on both long-term and short-term dynamics makes the NARDL model a powerful tool for analysing the complex interactions within financial systems Given the multilayer nature of the network analysis where different layers might exhibit nonlinear interactions and interdependence the NARDL model can accommodate these complexities by allowing distinct layers to respond differently to changes in the independent variables The BIST TICARET index (XTCRT) is one of several indices listed on the Istanbul Stock Exchange specifically encompassing companies within the commerce sector we focused on companies listed in the XTCRT index that operate in the sphere of electronic commerce The e-commerce sector was chosen due to its rapid growth and significant influence on the modern economy when online shopping and digital transactions surged Analysing this sector provides valuable insights into how digital businesses respond to market conditions and social media discourse which is especially relevant in understanding investor behaviour and market dynamics in the digital age The nine companies selected from the XTCRT index—BIMAS TKNSA—were chosen based on their active involvement in the e-commerce sector These companies represent a cross-section of the industry ranging from retail giants like BIMAS (BIM Birleşik Mağazalar) and MGROS (Migros Ticaret) to diversified holdings with significant e-commerce operations such as DOHOL (Doğan Holding) Each company plays a vital role in the Turkish e-commerce landscape contributing to the sector’s overall market dynamics The selected companies from the XTCRT index represent a diverse range of sectors within the e-commerce landscape BIMAS is a leading supermarket chain known for its expanding online presence BIZIM operates as a wholesaler catering to both retail customers and businesses with a growing emphasis on digital sales channels and has established a notable footprint in e-commerce has a strong digital media presence and is actively pursuing e-commerce initiatives MEPET operates within the fuel retail sector and is developing its digital commerce capabilities has seen significant growth in its e-commerce activities TKNSA is a leading electronics retailer with extensive online sales operations making it a key player in the digital marketplace These companies collectively provide a comprehensive view of the e-commerce sector within the XTCRT index each contributing uniquely to the sector’s overall dynamics each containing the nine selected companies as nodes The choice of a 5-day sliding window to extract financial correlation networks was made due to its alignment with a standard trading week providing an appropriate timeframe for capturing short-term market trends This window balances the need for sufficient data points to establish meaningful correlations while retaining sensitivity to rapid market changes capturing Turkish tweets related to COVID-19 These tweets were identified using keywords such as ‘Covid,’ ‘Corona,’ ‘Kovid’ and ‘Korona,’ reflecting the Turkish adaptations of these terms Despite some interruptions in data collection with regular intervals every 15 min ensuring comprehensive coverage Duplicate tweets were removed to maintain data integrity While the study’s social media discourse data comprises 4,900 posts financial data refer to only nine listed companies the nine entities were carefully chosen for their prominence and representativeness within the Turkish e-commerce sector This selection ensures that the research sample adequately reflects the broader dynamics of the sector allowing for meaningful analysis of the interplay between social media discourse and financial performance The dataset utilised during the deep learning training phase consists of news articles sourced from Turkish websites. It was constructed by selecting six prominent news categories: culture, economy, health, politics, technology and world affairs. There are 700 news entries in each category, for a total of 4900 articles. The number of tweets is presented in Fig. 1. The data set contains the total count of tweets from March 09 to October 31 Missing data gaps were filled using the autoregressive integrated moving average model which generates predictions for the corresponding number of blank sections while considering the days when Borsa Istanbul is closed the time series are synchronised on the same days and the logarithmic differences of the post numbers acquired from Twitter are calculated 161 observations were collected in this case consisting of time series data for both multilayer network measurements and subject classes Logarithmic regressions decrease the time series to 160 observations To assess measurement changes of the multilayer network we additionally employ the 5-length sliding window approach on the time series derived from subject classification of social media postings the median of the number of postings on topics with a length of 5 is calculated The distributions for each window in the topic series exhibit skewness we used the median rather than the mean in multilayer measurements we compared 156-entry series with 5-length sliding windows This section presents the data acquired from the topological measurements of the 4-layer network of the selected businesses Variations of mean multilayer participation coefficients through sliding windows Figure 2 displays a spectrum of average participation coefficients ranging from 0.7 to 1.0 which signifies different degrees of interconnectedness among enterprises Data distribution indicates that there are certain periods where the average participation coefficient is significantly high This demonstrates a robust interconnectedness between the stock actions of companies across all network layers during those specific periods there are substantial declines in the mean coefficient that indicate reduced interconnectedness over specific periods The fluctuations observed may be associated with particular events or market stages presumably influenced by the economic repercussions of the COVID-19 epidemic during the relevant period Lower participation coefficients may indicate periods of market stress or volatility during which stock prices tend to exhibit more segregation and less interconnection larger values may suggest more stable periods characterised by equities moving in a synchronised manner across all parameters under consideration Equation 3 defines the z-score for every node. Again, the mean value is employed at each layer separately. In Fig. 3, we display the average z-scores of the 4-layered network. Variations of mean z-scores through sliding windows Figure 3 indicates that mean z-scores exhibit periodic variations around zero Most data points are concentrated within a narrow range showing that the attributes of nodes are generally near the overall mean of the network there are occasions when average z-scores exhibit noteworthy surges that surpass positive and negative values Positive spikes represent intervals during which node attributes were higher than the mean whereas negative spikes indicate intervals for which attributes were lower Spikes can signify instances of atypical activity or departure from typical behaviour patterns fluctuations in the average z-scores throughout the sliding windows can indicate dynamic shifts in the stock market potentially affected by external occurrences or the intrinsic instability during initial phases of the pandemic Negative z-scores indicate instances when specific stock properties were underperforming relative to the average whereas positive scores correspond to above-average performance Equation 6 provides a definition of the entropy of a multilayer network. Entropy values are shown in Fig. 4, calculated using Eq. 3. Variations of multilayer entropies through sliding windows Figure 4 reveals a range of entropy values spanning from 0.5 to 2.5 Most data points appear to concentrate towards the upper limit of this range indicating a generally elevated degree of complexity or diversity in the interconnections among network levels This suggests a market characterised by numerous influential factors and a significant level of interconnectedness across components of stock performance there are specific time intervals during which entropy decreases to lower levels signifying periods of decreased complexity or heightened predictability in the market These instances may indicate periods when the market exhibited greater homogeneity or when specific layers exerted significant impacts on the entire system The graphs shown in Fig. 5 illustrate changes in the average logarithmic returns of tweet volume for each topic. The average logarithmic returns of tweet volume for each topic The culture graph depicts oscillations within a limited scope indicating that cultural discussions on Twitter were shielded from significant changes caused by market or global occurrences culture-related discussions remain consistent and unaffected by economic fluctuations if there are distinct high or low points in the graph it would be worth exploring whether they coincide with cultural events or releases that could quickly captivate the public’s attention thus offering a temporary diversion from current economic circumstances The economy graph displays substantial changes since economic discourse is typically responsive to market situations If peaks in economic discussions are correlated with falls in the mean multilayer participation coefficients from stock market analysis greater economic discourse could be either a reaction to or an indicator of decreased market stability one should investigate whether increases in the number of economic tweets are associated with notable changes in the market which may be observed through the analysis of multilayer entropy and z-scores it is likely that conversations about health are mostly focused on the COVID-19 outbreak Surges in health-related tweets may align with the timing of lockdown announcements infection rates and other news relevant to the epidemic These peaks may represent moments of elevated market volatility characterised by low participation coefficients or high z-scores as the market reacts aggressively to health-related news An analysis of health data in relation to the entropy graph may indicate a correlation between heightened discourse on health matters and periods of elevated uncertainty in the stock market Political discourse can significantly influence market attitudes The political graph may exhibit relationships with market data wherein noteworthy political events result in heightened market interconnection as indicated by greater participation coefficients as indicated by extreme z-scores or entropy values The technology category is expected to have direct significance for the e-commerce enterprises included in the XTCRT index Surges in technology-related discourse may align with significant advancements in technology introduction of new products and changes in regulations that impact electronic commerce These events may result in greater consistency in the market performance of these enterprises which could be observed as higher average participation coefficients or reduced entropy within specified time periods Global events can significantly influence financial markets particularly when they have economic consequences The global graph can exhibit surges in Twitter engagement in reaction to global occurrences such as elections trade pacts and worldwide economic assessments These spikes may correspond to notable shifts in market behaviour as companies in the XTCRT index respond to global dynamics Besides the graphical representations illustrating the relationships between the series of topics we conduct NARDL analysis to obtain more comprehensive results The NARDL test results for \({T}_{{mp}}\sim {T}_{{culture}}\) are presented in Table 1 \({T}_{{cultur}{e}_{p}}\) (positive changes in cultural tweets) and \({T}_{{cultur}{e}_{n}}\) (negative changes in cultural tweets) are particularly significant in the short term The substantial negative coefficient for \({T}_{m{p}_{1}}\) implies that as the average multilayer participation coefficient increases This suggests a negative correlation between the multilayer network dynamics and the cultural discourse on Twitter The coefficients for \({T}_{{cultur}{e}_{p}}\) and \({T}_{{cultur}{e}_{n}}\) are positive but not statistically significant This suggests that both positive and negative changes in cultural tweets positively correlate with the response variable the relationship is not strong enough to be considered conclusive based on conventional significance levels the coefficients for \({T}_{{cultur}{e}_{p}}\) and \({T}_{{cultur}{e}_{n}}\) are positive suggesting both positive and negative changes in cultural discourse on Twitter may be related to the response variable these coefficients lack statistical significance at conventional thresholds indicating a possible yet inconclusive long-term association The model diagnostics suggest a strong fit with a relatively low residual standard error and a statistically significant F-statistic which supports the general validity of the model The JB test reveals non-normal distribution of residuals Cointegration test indicates that the F-statistic exceeds critical values at conventional significance levels implying the presence of a long-term link between variables The results of the Short Run Asymmetry test and Long Run Asymmetry test are particularly interesting The Short Run Asymmetry test does not demonstrate any substantial asymmetry in the relationship implying that both positive and negative alterations in cultural discourse have an equivalent effect in the immediate period the Long Run Asymmetry test (with a p-value near the standard significance level) suggests the presence of asymmetry in the long-term effects of positive and negative cultural discourse modifications The NARDL analysis indicates an intricate correlation between the multilayer network dynamics of enterprises in the Borsa Istanbul and cultural discourse on Twitter Although there are signs of both temporary and lasting connections they lack sufficient strength to be considered definitive The NARDL test results for \({T}_{{en}}\sim {T}_{{culture}}\) are presented in Table 2 The coefficient for \({T}_{e{n}_{1}}\) exhibits a significant negative value suggesting a robust inverse correlation between past values of multilayer network entropy and the response variable in the near future This implies that an increase in entropy inside the network which may signify greater complexity or unpredictability is linked to a decreasing outcome variable The coefficients for \({T}_{{cultur}{e}_{p}}\) and \({T}_{{cultur}{e}_{n}}\) exhibit negative values but they do not demonstrate statistical significance This suggests that fluctuations in cultural communication on Twitter do not exert a substantial immediate influence on the complexity of the interconnected network The model demonstrates a strong correlation The residual standard error is of a moderate magnitude indicating that the model can capture a significant portion of the variability present in the data such as the JB test for normality of residuals LM test for serial correlation and ARCH test for autoregressive conditional heteroskedasticity indicate that there are no significant problems with the model specification Cointegration indicates that \({T}_{{en}}\) and \({T}_{{culture}}\) exhibit a long-term relationship where any divergence from this relationship is transient and variables will eventually return to their equilibrium state The results of both the Short Run and Long Run Asymmetry tests indicate that there is no substantial asymmetry in the impact of positive and negative changes in cultural discourse on the response variable This suggests that multilayer network entropy is not influenced differently by the direction of change (positive or negative) in cultural tweets the NARDL analysis shows that multilayer network entropy has a strong negative correlation with its previous values do not have a significant effect on the multilayer network entropy in the short or long term The existence of cointegration validates a solid and enduring connection between variables albeit the impact of cultural discussion is not statistically significant These findings indicate that the level of complexity or unpredictability in the network is not significantly affected by the cultural conversations taking place on social media during the investigated period The NARDL test results for \({T}_{{mp}}\sim {T}_{{economy}}\) are presented in Table 3 The presence of a negative coefficient for \({T}_{m{p}_{1}}\) indicates a substantial inverse correlation between preceding values of \({T}_{{mp}}\) and the dependent variable in the near future \({T}_{m{p}_{2}}\) exhibits a positive coefficient The coefficients for \({T}_{{econom}{y}_{p}}\) and its lagged effects exhibit a combination of results and do not demonstrate statistical significance This suggests that temporary fluctuations in economic discussion whether they occur immediately or with a delay do not significantly affect \({T}_{{mp}}\) The model demonstrates a satisfactory level of fit as indicated by an acceptable residual standard error and a statistically significant F-statistic LM test and ARCH test) reveal that there are no significant problems with model specification the JB test indicates the presence of non-normality in the residuals which is a typical issue in time series analysis as the F-statistic exceeds critical values This suggests a stable and consistent relationship between variables over time even though there may not be significant immediate effects The Short Run Asymmetry test does not show any significant asymmetry in the relationship the Long Run Asymmetry test yields a noteworthy outcome indicating that the long-term influence of positive and negative shifts in economic discourse on \({T}_{{mp}}\) may exhibit asymmetry The analysis of the NARDL model indicates that the average multilayer participation coefficients show a significant short-term inverse relationship with previous values the impact of economic discourse is significant Especially the noteworthy outcome of the Long Run Asymmetry test suggests the necessity for additional study into the distinct impacts of positive and negative attitudes in economic discourse on long-term financial market dynamics The NARDL test results for \({T}_{{en}}\sim {T}_{{economy}}\) are presented in Table 4 The substantial negative coefficient for \({T}_{e{n}_{1}}\) signifies a robust inverse correlation between its previous values and the dependent variable in the near future This suggests that an increase in entropy within the network which signifies greater complexity or unpredictability is linked to a reduction in the response variable The coefficients for positive (\({T}_{{econom}{y}_{p}}\)) and negative (\({T}_{{econom}{y}_{n}}\)) changes in economic tweets exhibit both a very small magnitude and a lack of statistical significance This implies that fluctuations in economic discussions on Twitter do not exert a substantial influence on the complexity of the multilayer network in the immediate period the coefficients for \({T}_{{econom}{y}_{p}}\) and \({T}_{{econom}{y}_{n}}\) remain small suggesting that there is no robust and conclusive long-term connection between the economic tweets and \({T}_{{en}}\) as shown by a high R-squared value and a statistically significant overall F-statistic The residual standard error is of considerable magnitude indicating that the model effectively captures a significant portion of the variability present in the data It is observed that the complexity of financial market networks and economic discourse on social media Any deviation from this long-term path is temporary and they will eventually converge back to this equilibrium relationship The results of Short Run and Long Run Asymmetry tests indicate that there is no significant difference in the impact of positive and negative changes in economic discourse on the response variable This conclusion is supported by the high p-values obtained NARDL analysis indicates that there is a strong negative correlation between multilayer network entropy and its previous values changes in economic discourse do not have a significant effect on the entropy of the network This suggests that the level of complexity or unpredictability of the network is not significantly affected by economic debates on social media complexities of market behaviour may not be immediately affected by public economic discussions on platforms such as Twitter The NARDL test results for \({T}_{{mp}}\sim {T}_{{health}}\) are presented in Table 5 The coefficients for \({T}_{{healt}{h}_{p}}\) and \({T}_{{healt}{h}_{n}}\) are both positive but do not have statistical significance alterations in health-related discussions on Twitter do not significantly influence \({T}_{{mp}}\) the coefficients for \({T}_{{healt}{h}_{p}}\) and \({T}_{{healt}{h}_{n}}\) consistently show positive values yet non-significant There is no strong and conclusive long-term association between health-related tweets and \({T}_{{mp}}\) Both Short Run and Long Run Asymmetry tests produce high p-values suggesting that there is no substantial asymmetry in the influence of positive versus negative changes in health discourse on the response variable The analysis of the NARDL model indicates that the average multilayer participation coefficients exhibit a significant negative correlation with their previous values changes in health-related discourse on Twitter do not have a significant impact on \({T}_{{mp}}\) in either the short or long term This suggests that the intricacy or fluctuations of the network as quantified by the participation coefficients are not significantly affected by public health conversations on social media These data emphasise the possible lack of connection or autonomy between commercial trends in the e-commerce industry and public health attitudes especially during the crucial phase of the COVID-19 pandemic The NARDL test results for \({T}_{{en}}\sim {T}_{{health}}\) are presented in Table 6 The coefficients for \({T}_{{healt}{h}_{p}}\) \({T}_{{healt}{h}_{n}}\) and \({T}_{{healt}{h}_{n1}}\) are not statistically significant indicating no significant relationship between positive changes in health tweets Fluctuations in health-related discussions on Twitter do not significantly affect the overall network complexity \({T}_{{healt}{h}_{n}}\) and \({T}_{{healt}{h}_{n1}}\) in the long run remain insignificant there is no clear and lasting connection between the health-related tweets and the entropy of multilayer network The results of Short Run and Long Run Asymmetry tests suggest no significant difference in the impact of positive and negative changes in health discourse on the response variable Multilayer network entropy is primarily influenced by its own past values rather than changes in health-related discourse on Twitter Public health discussions on social media have minimal impact on the complexity or unpredictability of the network within the observed timeframe This might indicate the distinctiveness of health-related discussions which may not directly impact intricacies of financial market behaviours The NARDL test results for \({T}_{{mp}}\sim {T}_{{politics}}\) are presented in Table 7 There is a significant and positive relationship between positive political discourse on Twitter and the response variable the coefficients for lagged effects of positive and negative political tweets do not show statistical significance suggesting that their short-term impact is relatively small the coefficients for \({T}_{{politics}}\) variables show a combination of positive and negative values but they are not statistically significant conclusive long-term relationship between political tweets and \({T}_{{mp}}\) except for a slightly significant coefficient for \({T}_{{politic}{s}_{p}}\) Diagnostic tests suggest that there are no significant problems with model specification The cointegration test reveals that there is evidence of cointegration between \({T}_{{mp}}\) and \({T}_{{politics}}\) This implies a long-term equilibrium connection the Long Run Asymmetry test reveals a significant outcome indicating a potential asymmetry in the long-term effects of positive and negative changes in political discourse on \({T}_{{mp}}\) the average multilayer participation coefficients show a notable short-term inverse relationship with past values the impact of changes in political discourse is more varied The observed correlation between positive tweets and short-term financial market dynamics suggests a possible impact of positive political discourse The significant result in the Long Run Asymmetry test suggests the importance of delving deeper into potential impact of positive and negative sentiments in political discourse on long-term financial market dynamics The findings shed light on the intricate relationship between market behaviour and political discourse particularly in relation to the impact of social media on public perception and market movements The NARDL test results for \({T}_{{en}}\sim {T}_{{politics}}\) are presented in Table 8 The coefficients for \({T}_{{politic}{s}_{p}}\) and \({T}_{{politic}{s}_{n}}\) show positive changes in political tweets but their statistical significance is not established It appears that changes in political discourse on Twitter do not have a substantial immediate effect on the multilayer network entropy The coefficients for \({T}_{{politic}{s}_{p}}\) and \({T}_{{politic}{s}_{n}}\) in the long run are positive but not significant There is no significant long-term relationship suggesting that the discourses of political tweets do not have a substantial impact on the entropy of the multilayer network The residual standard error indicates that the model effectively captures a significant amount of variability in the data Diagnostic tests show that there are no significant problems with model specification residuals do not follow a normal distribution Short Run and Long Run Asymmetry tests show high p-values hence there are no significant difference in the impact of positive and negative changes in political discourse on the response variable The cointegration test indicates the presence of cointegration among variables The F-statistic obtained is significantly higher than critical values providing evidence for the existence of a long-term relationship between variables Multilayer network entropy \({T}_{{en}}\) is greatly impacted by previous values The presence of cointegration indicates a stable long-term relationship among variables even though the impact of political discourse is not statistically significant political discussions on social media have minimal impact on the complexity or unpredictability of financial market networks The NARDL test results for \({T}_{{mp}}\sim {T}_{{technology}}\) are presented in Table 9 The coefficients for \({T}_{{technolog}{y}_{p}}\) and \({T}_{{technolog}{y}_{n}}\) are positive but not significant fluctuations in technology-related discussions on Twitter do not significantly affect \({T}_{{mp}}\) coefficients also show positive non-significant values the sentiment of technology-related tweets does not have an impact on the entropy of the multilayer network in the long run hence we found no significant differences in the impact of positive and negative changes in technology discourse This suggests a long-term balance between variables even though individual long-term coefficients do not show significance Multilayer network entropy is mainly affected by its previous values rather than changes in technology-related discourse on Twitter despite the lack of significant influence from technology sentiment Discussions about technology on social media do not have a significant impact on the complexity or unpredictability of financial market networks Results emphasise the potential disparity between the dynamics of financial market networks and public technology discourse which is crucial for comprehending the wider effects of social media discourse on financial markets The NARDL test results for \({T}_{{en}}\sim {T}_{{technology}}\) are presented in Table 10 The coefficients for \({T}_{{technolog}{y}_{p}}\) and \({T}_{{technolog}{y}_{n}}\) together with their corresponding delayed effects indicate a combination of positive and negative influences It is worth mentioning that \({T}_{{technolog}{y}_{p1}}\) has a substantial positive influence indicating that the delayed effect of optimistic technology-related discussions has a discernible impact on the outcome The long-term coefficients exhibit a comparable trend with \({T}_{{technolog}{y}_{p1}}\) demonstrating a substantial positive influence whilst the remaining coefficients lack statistical significance This suggests a complex and enduring relationship where the delayed favourable discourse about technology has a stronger impact The Short Run Asymmetry test indicates a p-value that is near the significance level suggesting asymmetry in the immediate effects of positive and negative changes in technological discourse suggests the possibility of asymmetry in the long-term effects This implies the existence of a stable relationship between variables in the long run even though certain long-run coefficients may not be significant The NARDL model indicates that multilayer network entropy is affected by previous values and has an intricate association with technology-related discussions on Twitter The notable effect of the delayed favourable technology tweets in both the immediate and extended periods underscores the potential sway of technology discussions on the intricacy or unpredictability of financial market networks Cointegration validates a consistent and enduring connection between variables albeit with varying effects from technology sentiment These findings highlight the subtle and potentially significant influence of technology-related social media debates on altering financial market behaviour The NARDL test results for \({T}_{{mp}}\sim {T}_{{world}}\) are presented in Table 11 The coefficients for \({T}_{{worl}{d}_{p}}\) and \({T}_{{worl}{d}_{n}}\) are positive but not statistically significant changes in global discourse on Twitter do not have a substantial impact on \({T}_{{mp}}\) the coefficients for \({T}_{{worl}{d}_{p}}\) and \({T}_{{worl}{d}_{n}}\) are positive but non-significant world-related tweets do not influence the participation of multilayer network The Short Run and Long Run Asymmetry tests both yield high p-values suggesting no significant asymmetry in the impact of positive versus negative changes in world discourse on the response variable multilayer network means that the participation coefficient is significantly influenced by past values but not by changes in world-related discourse in either short or long term The presence of cointegration confirms a stable long-term relationship among variables although the specific influence of global sentiment is not significant This finding implies that participation coefficients are not strongly affected by global discussions on social media Results underscore the potential independence or disconnect between the dynamics of financial market networks and public global sentiment highlighting the unique characteristics of financial markets in relation to social media discourse The NARDL test results for \({T}_{{en}}\sim {T}_{{world}}\) are presented in Table 12 The statistical analysis indicates that the coefficients for \({T}_{{worl}{d}_{p}}\) \({T}_{{worl}{d}_{n}}\) and \({T}_{{worl}{d}_{n1}}\) are not significant Rapid and delayed shifts in world-related conversation do not influence short-term complexity of multilayer networks The long-term coefficients for \({T}_{{worl}{d}_{p}}\) The absence of a notable enduring correlation suggests that tweets on world affairs do not exert a meaningful influence on the complexity of the multilayer network in the long run Asymmetry tests indicate no significant difference between the impact of positive and negative changes in world discourse on the response variable The cointegration test shows the presence of cointegration among variables a link between variables in the long term may not be statistically significant while multilayer network entropy in financial markets is substantially influenced by prior values of the network it remains unaffected by fluctuations in world-related discourse whether observed over the short term or extended periods Cointegration analysis validates the presence of a consistent and enduring connection between network entropy and historical data including cultural topics discussed on social media does not emerge as statistically significant This underscores that the complexity of financial market networks operates independently of social media conversations revealing a disconnection between the behaviour of these networks and the shifting nature of global mood our findings carry significant implications Despite the proliferation of online discourse on various cultural and societal topics such discussions do not seem to translate into measurable changes in the structure or complexity of financial market networks This points to a fundamental characteristic of financial systems which appear to function autonomously from the influence of social sentiment especially as expressed on global platforms like social media The analysis predominantly focuses on global mood and cultural topics without delving into specific themes or sectors that might have more direct relevance to financial markets or geopolitical events could potentially exhibit a stronger connection to financial network behaviour Future research could benefit from a more granular examination of these themes exploring their distinct impact on market dynamics refining the modelling techniques to capture more subtle variations in global discourse could improve the accuracy and predictive power of such analyses while results might suggest a lack of correlation between global social media conversations and financial network complexity further investigation into other topics and refined methodologies may yield more nuanced insights into the intricate relationship between social discourse and financial markets This study provides valuable insights into the dynamics of financial markets particularly in relation to social media discourse within the specific context of the e-commerce sector in Türkiye we discovered significant short-term and long-term relationships between social media discussions and market behaviour A critical finding is the significant short-term inverse relationship between the average multilayer participation coefficients and their previous values suggesting that market participants in the e-commerce sector may react dynamically and often inversely to recent past performance This reactivity is particularly evident in how economic discourse significantly influences market dynamics underscoring the strong impact of public sentiment on financial behaviours in both the short and long term the analysis reveals a strong negative correlation between multilayer network entropy and its past values indicating that the complexity and unpredictability of network structures are largely shaped by historical data rather than current market fluctuations This finding is crucial for understanding the behaviour patterns specific to the e-commerce sector and predicting future trends within this context the impact on financial markets appears more nuanced We observed that positive political sentiment can have a tangible again emphasising the sensitivity of the e-commerce sector to public discourse Technology-related discussions on Twitter also exhibit a mixed influence While network entropy is predominantly determined by historical values the analysis indicates that technology-related discourse can have a delayed yet significant impact on market behaviour in the e-commerce sector This suggests that technology is increasingly becoming a key factor in shaping market dynamics within this specific sector our NARDL analysis paints a complex picture of financial market dynamics in the e-commerce sector political and technological topics exerts varying degrees of influence Empirical evidence from Borsa Istanbul shows that companies in the XTCRT index are responsive to social media content related to economic and political matters discussions surrounding technology and economic issues impact the topological configuration of the XTCRT index leading to observable shifts in stock prices trading volumes and overall market sentiment Our study’s focus on the e-commerce sector of Borsa Istanbul particularly the nine companies within the XTCRT index provides valuable insights into the specific interactions between social media discourse and financial market performance in this sector The significant interconnectivity among these companies highlights the need for public policies that strengthen market resilience particularly during periods of global challenges like the COVID-19 pandemic Policymakers should consider establishing resilient financial safety nets and contingency plans to manage periods of heightened market volatility and stress specifically within sectors that are highly reactive to public sentiment The variations in average z-scores and network entropy within the e-commerce sector underscore the market’s susceptibility to external disturbances and information transmission To mitigate speculative trading and disinformation policymakers should enhance the regulatory framework related to information transmission and market transparency ensuring that these regulations are tailored to the specific dynamics of the e-commerce sector our study contributes to the literature examining the impact of social media on the stock market particularly in the context of the e-commerce sector during the pandemic period Unlike other studies that may generalise findings across entire stock exchanges our research is specifically targeted at the e-commerce sector revealing how public discourse on certain topics influences market behaviour in both the short and long term it is important to acknowledge the limitations of our study our analysis is based on a sample of nine companies in the e-commerce sector included in the XTCRT index our network analysis focuses on an eight-month period following the onset of the COVID-19 pandemic in Türkiye the social media discourse was retrieved only from one platform (Twitter These limitations suggest that findings are most applicable to the e-commerce sector within Borsa Istanbul and should not be generalised to all companies listed on the Istanbul Stock Exchange the insights gained from this study open several avenues for future research One potential area is exploring the causal links between financial market activities and social media discourse Future studies could use more detailed data and advanced analytical tools such as machine learning algorithms and sentiment analysis to further investigate these relationships Expanding the analysis to include other social media platforms and different forms of digital communication could offer a broader perspective on public opinion and its impact on financial markets while our research is focused on the e-commerce sector within Borsa Istanbul it highlights the broader need for sophisticated analytical techniques and social media monitoring to enhance the stability resilience and responsiveness of financial markets Policymakers and regulators should consider these findings in the context of specific sectors ensuring that regulations and public policies are appropriately tailored to the unique dynamics of each sector they can better manage the complex interplay between public sentiment and financial market performance ultimately contributing to a more stable and resilient economic environment The datasets generated and/or analysed during the current study are available from the corresponding author on reasonable request Ahmed F, Kousar S, Pervaiz A, Ramos-Requena JP (2020) Financial development, institutional quality, and environmental degradation nexus: new evidence from asymmetric ARDL co-integration approach. Sustainability 12(18):7812. https://doi.org/10.3390/su12187812 Ahmed M, Huan W, Ali N, Shafi A, Ehsan M, Abdelrahman K, Khan AA, Abbasi SS, Fnais MS (2023) The effect of energy consumption, income, and population growth on co2 emissions: evidence from NARDL and machine learning models. Sustainability 15(15):11956. https://doi.org/10.3390/su151511956 Ahmed A, Ullah S (2022) Impact of social media on the performance of listed companies. J Bus Socio-Econ Dev 2(3):312–330. https://doi.org/10.1108/JBSED-09-2021-0112 Akgüller Ö, Balcı MA, Batrancea LM, Gaban L (2023) Path-based visibility graph kernel and application for the Borsa Istanbul stock network. Mathematics 11(6):1528. https://doi.org/10.3390/math11061528 Akın AA, Akın MD (2007) Zemberek, an open source NLP framework for Turkic languages. Structure 10(2007):1–5. https://doi.org/10.21105/joss.01241 Alessandretti L, Natera-Orozco LG, Saberi M, Szell M, Battiston F (2023) Multimodal urban mobility and multilayer transport networks. Environ Plan B Urban Anal City Sci 50(8):2038–2070. https://doi.org/10.1177/23998083231164732 Alfaro L, Chari A, Greenland AN, Schott PK (2020) COVID-19 and corporate investment: evidence from past pandemics. National Bureau of Economic Research Working Paper Series, No. 27117. https://doi.org/10.3386/w27117 Aleta A, Moreno Y (2019) Multilayer networks in a nutshell. Annu Rev Condens Matter Phys 10:45–62. https://doi.org/10.1146/annurev-conmatphys-031218-013259 Ali SS, Ramlan H (2020) Stock market forecasting using artificial intelligence during the COVID-19 pandemic. Proceedings of the 3rd International Conference on Communication, Computing, and Digital Systems (C-CODE). https://doi.org/10.1109/ICCCA49749.2020.9259162 Alharbi FM (2022) How do the remittance outflows react to economic development in Saudi Arabia? Evidence from NARDL. Economies 10(5):105. https://doi.org/10.3390/economies10050105 Allen DE, McAleer M (2021) A nonlinear autoregressive distributed lag (NARDL) analysis of the FTSE and S&P500 indexes. Risks 9(11):195. https://doi.org/10.3390/risks9110195 Aramonte S, Avalos F, Xia D (2021) How does COVID-19 revolution affect the financial market? SSRN Electron J. https://doi.org/10.2139/ssrn.3722115 Armanious M, El-Nahas N, Farag A (2020) Using social media data to model stock market movements. Proceedings of the 2020 IEEE International Conference on Industrial Cyber-Physical Systems (ICPS), pp 940–945. https://doi.org/10.1109/ICICS49469.2020.9406630 Baboukardos D, Gaia S, She C (2021) Social performance and social media activity in times of pandemic: evidence from covid-19-related Twitter activity. Corp Gov Int J Bus Soc 21(6):1271–1289. https://doi.org/10.1108/CG-11-2020-0482 Balcı MA, Batrancea LM, Akgüller Ö (2022) Network-induced soft sets and stock market applications. Mathematics 10(21):3964. https://doi.org/10.3390/math10213964 Balcı MA, Batrancea LM, Akgüller Ö, Nichita A (2022) Coarse graining on financial correlation networks. Mathematics 10(12):2118. https://doi.org/10.3390/math10122118 Baker SR, Bloom N, Davis SJ, Terry SJ (2020) COVID-induced economic uncertainty. National Bureau of Economic Research No. w26983. https://doi.org/10.3386/w26983 Basole RC, Russell MG, Huhtamaki J, Rubens N, Still K, Park H (2015) Understanding business ecosystem dynamics: a data-driven approach. ACM Trans Manag Inf Syst 6(2):1–32. https://doi.org/10.1145/2724730 Batrancea I, Balcı MA, Batrancea LM, Akgüller Ö, Tulai H, Rus MI, Masca ES, Morar ID (2023) Topic analysis of social media posts during the COVID-19 pandemic: evidence from tweets in Turkish. J Knowl Econ 1–31. https://doi.org/10.1007/s13132-023-01565-6 Batrancea LM, Balcı MA, Akgüller Ö, Gaban L (2022) What drives economic growth across European countries? A multimodal approach. Mathematics 10(19):3660. https://doi.org/10.3390/math10193660 Batrancea LM, Balcı MA, Chermezan L, Akgüller Ö, Masca ES, Gaban L (2022) Sources of SMEs financing and their impact on economic growth across the European Union: insights from a panel data study spanning sixteen years. Sustainability 14(22):15318. https://doi.org/10.3390/su142215318 Batrancea LM, Nichita A, Balcı MA, Akgüller Ö (2023) Empirical investigation on how wellbeing-related infrastructure shapes economic growth: evidence from the European Union regions. PLoS ONE 18(4):e0283277. https://doi.org/10.1371/journal.pone.0283277 Biswas S, Ghosh A, Chakraborty S, Roy S, Bose R (2020) Scope of sentiment analysis on news articles regarding stock market and GDP in struggling economic condition. Int J 8(7):3594–3609. https://doi.org/10.1007/s40843-020-1545-3 Boccaletti S, Bianconi G, Criado R, Del Genio CI, Gomez-Gardenes J, Romance M, Sendina-Nadal I, Wang Z, Zanin M (2014) The structure and dynamics of multilayer networks. Phys Rep 544(1):1–122. https://doi.org/10.1016/j.physrep.2014.07.001 Bollen J, Mao H, Zeng X (2011) Twitter mood predicts the stock market. J Comput Sci 2(1):1–8. https://doi.org/10.1016/j.jocs.2011.04.001 Boss M, Elsinger H, Summer M, Thurner S (2004) Network topology of the interbank market. Quant Financ 4(6):677–684. https://doi.org/10.1080/14697680400020325 Cavalcante RC, Brasileiro RC, Souza VL, Nobrega JP, Oliveira AL (2016) Computational intelligence and financial markets: a survey and future directions. Expert Syst Appl 55:194–211. https://doi.org/10.1016/j.eswa.2016.02.015 Chang IC, Horng JS, Liu CH, Chou SF, Yu TY (2022) Exploration of topic classification in the tourism field with text mining technology—a case study of the academic journal papers. Sustainability 14(7):4053. https://doi.org/10.3390/su14074053 Chen R, Fan R, Wang D, Yao Q (2023) Exploring the coevolution of residents and recyclers in household solid waste recycling: evolutionary dynamics on a two-layer heterogeneous social network. Waste Manag 157:279–289. https://doi.org/10.1016/j.wasman.2023.06.014 Choi TY, Hofmann E, Templar S, Rogers DS, Leuschner R, Korde RY (2023) The supply chain financing ecosystem: early responses during the Covid-19 crisis. J Purch Supply Manag 100836. https://doi.org/10.1016/j.pursup.2023.100836 Cui Y, Wang Y, Wu J (2020) Multilayer networks: theory and applications. Springer Proceedings in Complexity. https://doi.org/10.1007/978-3-030-24377-3 Çelik S, Gökmen A (2020) The impact of COVID-19 on Borsa Istanbul sector indices: an empirical analysis. SSRN Electron J. https://doi.org/10.2139/ssrn.3616842 Da Z, Engelberg J, Gao P (2011) In search of attention: the impact of online searches on stock market. J Financ 66(5):1461–1499. https://doi.org/10.1111/j.1540-6261.2011.01679.x Das N, Sadhukhan B, Chatterjee T, Chakrabarti S (2022) Effect of public sentiment on stock market movement prediction during the Covid-19 outbreak. Soc Netw Anal Min 12(1):92. https://doi.org/10.1007/s13278-022-00912-2 De Domenico M, Solé-Ribalta A, Omodei E, Gómez S, Arenas A (2015) Multilayer networks: analysis of functional magnetic resonance imaging data. Phys Rev X 5(1):011027. https://doi.org/10.1103/PhysRevX.5.011027 Domalewska D (2021) An analysis of Covid-19 economic measures and attitudes: evidence from social media mining. J Big Data 8(1):42. https://doi.org/10.1186/s40537-021-00449-9 Du WB, Zhou XL, Jusup M, Wang Z (2016) Physics of transportation: towards optimal capacity using the multilayer network framework. Sci Rep 6(1):19059. https://doi.org/10.1038/srep19059 Eren E, Özbekler A (2020) The impact of COVID-19 on stock market volatility: evidence from Borsa Istanbul. SSRN Electronic J. https://doi.org/10.2139/ssrn.3604785 Faccia A, LeRoux CL, Pandey V (2023) Innovation and e-commerce models, the technology catalysts for sustainable development: the emirate of Dubai case study. Sustainability 15(4):3419. https://doi.org/10.3390/su15043419 Falegnami A, Costantino F, DiGravio G, Patriarca R (2020) Unveil key functions in socio-technical systems: mapping from into a multilayer network. Cogn Technol Work 22(4):877–899. https://doi.org/10.1007/s10111-020-00624-x Felch V, Sucky E (2023) In search of a consensus definition of business ecosystems: a qualitative study. J Model Manag 18(6):1834–1857. https://doi.org/10.1108/JM2-09-2021-0232 Fang L, Peress J (2021) Media coverage and the cross-section of stock returns. Rev Financ Stud 34(3):1037–1071. https://doi.org/10.1093/rfs/hhaa014 Fiala V, Kapounek S, Veselý O (2015) Impact of social media on the stock market: evidence from tweets. Eur J Bus Sci Technol 1(1):24–35. https://doi.org/10.11118/ejobsat.v1i1.7 Fuerst S, Sanchez-Dominguez O, Rodriguez-Montes MA (2023) The role of digital technology within the business model of sustainable entrepreneurship. Sustainability 15(14):10923. https://doi.org/10.3390/su151410923 Gani R, Bałdyga J, Biscans B, Brunazzi E, Charpentier JC, Drioli E, Woodley JM (2020) A multi-layered view of chemical and biochemical engineering. Chem Eng Res Des 155:A133–A145. https://doi.org/10.1016/j.cherd.2020.04.014 Gao Q (2022) Systemic risk analysis of multi-layer financial network system based on multiple interconnections between banks, firms, and assets. Entropy 24(9):1252. https://doi.org/10.3390/e24091252 Garcia D (2013) Sentiment during recessions. J Financ 68(3):1267–1300. https://doi.org/10.1016/j.jfineco.2012.07.001 Gasparetto A, Marcuzzo M, Zangari A, Albarelli A (2022) A survey on text classification algorithms: from text to predictions. Information 13(2):83. https://doi.org/10.3390/info13020083 Guan C, Liu W, Cheng JYC (2021) Using social media to predict the stock market crash and rebound amid the pandemic: the digital ‘haves’ and ‘have-mores’. Ann Data Sci 1–27. https://doi.org/10.1007/s40745-021-00332-1 Gunay, S (2022) Social media and financial markets: impact of online discourse on stock volatility. SSRN Electron J. https://doi.org/10.2139/ssrn.3677151 He X, Liu Y (2024) Knowledge evolutionary process of artificial intelligence in e-commerce: main path analysis and science mapping analysis. Expert Syst Appl 238:121801. https://doi.org/10.1016/j.eswa.2023.121801 Hochreiter S, Schmidhuber J (1997) Long short-term memory. Neural Comput 9(8):1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735 Hutchinson MC, Bramon-Mora B, Pilosof S, Barner AK, Kefi S, Thebault E, Jordano P, Stouffer DB (2019) Seeing the forest for the trees: putting multilayer networks to work for community ecology. Funct Ecol 33(2):206–217. https://doi.org/10.1111/1365-2435.13244 Ibrahim AA, Lonardi A, Bacco CD (2021) Optimal transport in multilayer networks for traffic flow optimization. Algorithms 14(7):189. https://doi.org/10.3390/a14070189 Ismail S, Nair RK, Sham R, Wahab SN (2018) Impacts of online social media on investment decision in Malaysia. Indian J Public Health Res Dev 9(11):45–50. https://doi.org/10.5958/0976-5506.2018.01682.2 Jiang Y, Batool Z, Raza SMF, Haseeb M, Ali S, Zain UlAbidin S (2022) Analyzing the asymmetric effect of renewable energy consumption on environment in STIRPAT-Kaya-EKC framework: a NARDL approach for China. Int J Environ Res Public Health 19(12):7100. https://doi.org/10.3390/ijerph19127100 Kadous K, Mercer M, Zhou YD (2017) Undue influence? The effect of social media advice on investment decisions. SSRN Electron J. https://doi.org/10.2139/ssrn.2968407 Kaplan AM, Haenlein M (2010) Users of the world, unite! The challenges and opportunities of social media. Bus Horiz 53(1):59–68. https://doi.org/10.1016/j.bushor.2009.09.003 Kesiku CY, Chaves-Villota A, Garcia-Zapirain B (2022) Natural language processing techniques for text classification of biomedical documents: a systematic review. Information 13(10):499. https://doi.org/10.3390/info13100499 Khadhraoui M, Bellaaj H, Ammar MB, Hamam H, Jmaiel M (2022) Survey of BERT-base models for scientific text classification: covid-19 case study. Appl Sci 12(6):2891. https://doi.org/10.3390/app12062891 Korbel J, Jiang X, Zheng B (2019) Transfer entropy between communities in complex financial networks. Entropy 21(11):1124. https://doi.org/10.3390/e21111124 Korol T (2019) Dynamic bankruptcy prediction models for European enterprises. J Risk Financ Manag 12(4):185. https://doi.org/10.3390/jrfm12040185 Lazzini A, Lazzini S, Balluchi F, Mazza M (2022) Emotions, moods and hyperreality: social media and the stock market during the first phase of covid-19 pandemic. Account Audit Account J 35(1):199–215. https://doi.org/10.1108/AAAJ-07-2020-4654 Li X, Xu X, Liu J, Dong J, Lü J (2022) Multilayer financial complex networks and their applications. IEEE Trans Circuits Syst I: Regul Pap 96(10):4103–4116. https://doi.org/10.1109/TCSI.2022.3203854 Liang Z, Pan D, Deng Y (2020) Research on the knowledge association reasoning of financial reports based on a graph network. Sustainability 12(7):2795. https://doi.org/10.3390/su12072795 Mantegna RN (1999) Hierarchical structure in financial markets. Eur Phys J B-Condens Matter Complex Syst 11:193–197. https://doi.org/10.1007/s100510050929 Menegaki AN (2019) The ARDL method in the energy-growth nexus field: best implementation strategies. Economies 7(4):105. https://doi.org/10.3390/economies7040105 Milano M, Agapito G, Cannataro M (2023) An exploratory application of multilayer networks and pathway analysis in pharmacogenomics. Genes 14(10):191. https://doi.org/10.3390/genes14101991 Milin IA, Buşan G, Ecobici N, Rehman A (2023) Economic growth drivers in Romania: evidence from a NARDL analysis. Sustainability 15(7):5916. https://doi.org/10.3390/su15075916 Naseem S, Mohsin M, Hui W, Liyan G, Penglai K (2021) The investor psychology and stock market behavior during the initial era of Covid-19: a study of China, Japan, and the United States. Front Psychol 12:626934. https://doi.org/10.3389/fpsyg.2021.626934 Nyakuruwa K, Seetharam Y (2023) The evolution of studies on social media sentiment in the stock market: insights from bibliometric analysis. Sci Afr 20:e01596. https://doi.org/10.1016/j.sciaf.2023.e01596 Pennington J, Socher R, Manning CD (2014) Glove: global vectors for word representation. Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP), pp 1532–1543. https://doi.org/10.3115/v1/D14-1162 Pesaran MH, Shin Y, Smith RJ (2001) Bounds testing approaches to the analysis of level relationships. J Appl Econ 16(3):289–326. https://doi.org/10.1002/jae.616 Pianese T, Belfiore P (2021) Exploring the social networks’ use in the health-care industry: a multi-level analysis. Int J Environ Res Public Health 18(14):7295. https://doi.org/10.3390/ijerph18147295 Piñeiro-Chousa J, Vizcaíno-González M, Pérez-Pico AM (2017) Influence of social media over the stock market. Psychol Mark 34(1):101–108. https://doi.org/10.1002/mar.20975 Singh Chauhan S, Srinivasan K, Sharma T (2023) A trans-national comparison of stock market movements and related social media chatter during the Covid-19 pandemic. J Bus Anal 6(3):203–216. https://doi.org/10.1080/2573234X.2023.2228915 Shin Y, Yu B, Greenwood-Nimmo M (2014) Modelling asymmetric cointegration and dynamic multipliers in a nonlinear ARDL framework. Econ Model 42:327–338. https://doi.org/10.1016/j.econmod.2013.12.020 Smales LA (2021) Social media as an information source and its impact on the stock market. Econ Model 95:51–66. https://doi.org/10.1016/j.econmod.2020.12.009 Smith AN, Fischer E, Yongjian C (2012) How does brand-related user-generated content differ across Youtube, Facebook, and Twitter. J Interact Mark 26(2):102–113. https://doi.org/10.1016/j.intmar.2012.01.002 Song J, Ma M (2023) Climate change: linear and nonlinear causality analysis. Stats 6(2):626–642. https://doi.org/10.3390/stats6020034 Stella M, Restocchi V, De Deyne S (2020) #lockdown: network-enhanced emotional profiling in the time of covid-19. Big Data Cogn Comput 4(2):14. https://doi.org/10.3390/bdcc4020014 Sugiki N, Nagao S, Kurauchi F, Mutahari M, Matsuo K (2021) Social dynamics simulation using a multi-layer network. Sustainability 13(24):13744. https://doi.org/10.3390/su132413744 Tumminello M, Aste T, Di Matteo T, Mantegna RN (2005) A tool for filtering information in complex systems. Proc Natl Acad Sci USA 102(30):10421–10426. https://doi.org/10.1073/pnas.0500298102 Ulak MB, Sriram LMK, Kocatepe A, Ozguven EE, Arghandeh R (2021) Resilience characterization for multilayer infrastructure networks. IEEE Intell Transport Syst Mag 14(4):121–132. https://doi.org/10.1109/MITS.2021.3063831 Valle-Cruz D, Fernandez-Cortez V, Lopez-Chau A, Sandoval-Almazan R (2022) Does Twitter affect stock market decisions? Financial sentiment analysis during pandemics: a comparative study of the H1N1 and the COVID-19 periods. Cogn Comput 14:372–387. https://doi.org/10.1007/s12559-022-10044-7 van Dorsser C, Taneja P (2020) An integrated three-layered foresight framework. Foresight 22(2):250–272. https://doi.org/10.1108/FS-01-2020-0001 Wang X, Xiang Z, Xu W, Yuan P (2022) The causal relationship between social media sentiment and stock return: experimental evidence from an online message forum. Econ Lett 216:110598. https://doi.org/10.1016/j.econlet.2022.110598 Wang X, Vergeer M (2024) Effect of social media posts on stock market during COVID-19 infodemic: an agenda diffusion approach. Sage Open 14(1). https://doi.org/10.1177/21582440231171477 Wu X, Wang X, Ma S, Ye Q (2017) The influence of social media on stock volatility. Front Eng Manag 4(2):201–211. https://doi.org/10.3389/fbuil.2018.00021 Wu L, Liu X (2020) Social media predicts fund flows. Rev Financ Stud 33(6):2796–2832. https://doi.org/10.1093/rfs/hhz076 Xie J, Xie B, Zhou K, Li J, Xiao J, Liu C, Zhang X (2023) Multiple probability ecological network and county-scale management. Land 12(8):1600. https://doi.org/10.3390/land12081600 Yan XG, Xie C, Wang GJ (2015) Stock market network’s topological stability: evidence from planar maximally filtered graph and minimal spanning tree. Int J Mod Phys B 29(22):1550161. https://doi.org/10.1142/S0217979215501616 Yılmaz AM (2021) Forecasting stock prices of Borsa Istanbul e-commerce companies using deep learning methods. Turk J Electr Eng Comput Sci 29(6):3297–3310. https://doi.org/10.3906/elk-2008-86 You T, Fiedor P, Holda A (2015) Network analysis of the Shanghai Stock Exchange based on partial mutual information. J Risk Financ Manag 8(2):266–284. https://doi.org/10.3390/jrfm8020266 Zennaro I, Finco S, Calzavara M, Persona A (2022) Implementing e-commerce from logistic perspective: literature review and methodological framework. Sustainability 14(2):911. https://doi.org/10.3390/su14020911 Zhang D, Hu M, Ji Q (2020) The influence of COVID-19 on stock prices and volatility: evidence from a difference-in-differences model. Financ Res Lett 36:101690. https://doi.org/10.1016/j.frl.2020.101690 Zhang H, Chen Y, Rong W, Wang J, Tan J (2022) Effect of social media rumors on stock market volatility: a case of data mining in China. Front Physiol 10:987799. https://doi.org/10.3389/fphy.2022.987799 Download references Download citation DOI: https://doi.org/10.1057/s41599-024-03894-6 Borsa Italiana non ha responsabilità per il contenuto del sito a cui sta per accedere e non ha responsabilità per le informazioni contenute. Accedendo a questo link, Borsa Italiana non intende sollecitare acquisti o offerte in alcun paese da parte di nessuno. Sarai automaticamente diretto al link in cinque secondi. likes to rotate the venue to keep it fresh and while this year’s choice was the solid but unspectacular Royal Horticultural Halls in Westminster you could hardly say that about the wines on show and the setup that ITA designed Over 200 were on show from 41 producers (all but two seeking UK distribution) in 13 key regions - namely Abruzzo and a wonderfully diverse selection of labels to lure the many buyers off-trade representatives et alia who pitched up as ITA London director Giovanni Sacchi puts it and while one can only speculate about the extent of those those who attended were able to delve into many of the scores of indigenous grape varieties Italy has - the most of any country with 377 registered varieties to its name a long way ahead of France in second place on 204.* Perhaps the more obscure varieties are the place to start because two of the three masterclasses were ‘Family Estates: Unveiling lesser-known regions’ (by John Downes MW) and ‘Unknown Italy: white wines from the South’ (by Walter Speller the third was ‘International grapes in Italy’ (by Patrick Schmitt MW) Speller picked out a delightful Pecorino Superiore from the Pesolillo winery in Abruzzo DOC whose joint owner Lorenzo Pesolillo I made a point of seeking out This is precisely the sort of small family-owned set-up that BVI is designed to help when it comes to finding a UK importer Three brothers are involved - Lorenzo (exports) Luca (winemaker) and Marco (admin) - with father Giuseppe the viticulturist (his father having founded the winery in 1961) Organically-farmed vines at 200 metres close to the Adriatic Sea have produced a fresh wine with aromas of white flowers and notes of peach and pear Annual production of around 50,000 bottles (also featuring Montepulciano and Passerina) was taken up entirely by the Italian market until two years ago when the family ventured into export markets that now include Canada “Now we like to get into the UK,” Lorenzo told me Fellow Campania producer, Ocone unfurled a very drinkable 12% abv sparkling Aglianico named Alalunga Vino Spumante di Qualita Sannio 2021 “We make 7,500 bottles of this by the Charmat method but have a traditional method version with 36 months on the lees being released before Christmas,” export manager Antonietta Luongo said Just over a third of the company’s annual production with its premium Vigna del Monaco 2021 label another fine Falanghina from the sought-after Taburno sub-region where their vines are planted up to 600 metres Michele & Andrea Bruno of Boccafolle di Balbia Another of Speller’s shrewd selections was Boccafolle di Balbia in the Calabria town of Mottafollone This intriguing little winery with five hectares under vine only deals in ancient indigenous varietals such as Greco Nero & Bianco was ‘completely unknown’ until last year when it was finally permitted as a varietal Pear and quince are its principal descriptors which complement the herb and lemon notes of the other grape in the blend who happens to be one of Real Madrid’s lawyers were at BVI in the hope of establishing a market in the UK “We’ve made some good contacts with sommeliers and some hotel chains and distributors are showing interest,” Michele declared with 80% going to the United States where I used to live.” The approachable Melara 2022 label (85% Magliocco from 65-year old vines) belied the varietal’s reputation for overt tannins Talking of lower alcohol wines, the Nebbiolo producer, Bosio Family Estates, attracted praise from Beans Boughton MW, buyer for Alliance Wine “These guys are into low alcohol,” Boughton mused although none of their low abv labels were available for tasting “I think their red wines are the best I’ve tasted at 11% because they’re using innovative techniques to get there their Barbaresco and Barolo are the sort of wines that could work for us but we already have a lot of producers who do these but not so many with such good value-to-price ratio as here.” Also offering excellent value for money was Friuli winery, Il Roncal which has 20 hectares of terraced vineyards on Colle Montebello close to the UNESCO world heritage site town of Cividale del Friuli Pinot Grigio and Sauvignon Blanc labels were complemented by some enticing Schioppettino its violet-scented aromas gave way to black pepper and spicy notes on a long finish Another lesser-known red variety, Ciliejolo, whose spiritual home is Tuscany, made up a 50:50 split with Canaiolo in a very quaffable blend made by Chianti producer, Castello del Trebbio which gained biodynamic certification last year mid-market IGT Toscana wine with a very nice balance between acidity and fruit that you can try chilled,” export manager Barbara Ruppel said “Ellis of Richmond imports our other two brands in Maremma and Sardinia but we are looking for distribution here.” Ciliejolo means ‘cherry’ in Italian Reverting to mainstream varieties, Empson & Co a leading exporter of fine Italian wines worldwide is looking for UK distribution for two of their brands: Toscolo in Chianti and Jankara in Sardinia The latter’s Vermentino di Gallura Superiore DOCG 2023 showed particularly well “It is very good Vermentino,which retails in Italy for between €30-35,” she said “It benefits from granitic soils and very fresh winds from the sea.” She revealed the muscular Jankara Cannonau di Sardegna DOC 2020 produced from vines at 700 metres and compared stylistically by Schmitt in his masterclass to Châteauneuf-du-Pape had impressed Ed Fairfax of importer Waud Wines Barbara & Luca Cruciani of Casa Lucciola Talking of Italian white varietals, one small winery that makes nothing but Verdicchio caught the eye - Casa Lucciola in the Marche region Owner-winemaker Luca Cruciani farms four hectares organically at 430m on clay and limestone soils in the Matelica Valley “He lost his mind 25 years ago and planted vines,” wife Barbara joked The couple produce five Verdicchio labels with some very appealing fruit and notable freshness (RRP in Italy €13-21).” It is the grape of our territory and does very well in our terroir “We select the best grapes to produce approximately 12,000 bottles.” By contrast, Muratori produces up to 500,000 bottles per annum of Franciacorta from 50 hectares which is available through Boutique Brands “Muratori has vines in each of the six areas of Franciacorta The Brut NV is the most popular style we have - very user-friendly.” Three other labels were on view including an excellent zero dosage 100% Chardonnay Finally, a quirky SKU to end with. A company named Gloria d’Italia is producing 25cl cans of carbonated spritzer made up of 55% Pinot Grigio (Trentino fruit) Vibrant acidity balances residual sugar of 12g/l “It’s been on the market for six months now - just in Australia where my co-founder Pierro lives and where it’s done well in bottle shops “There is a huge trend for ready-to-drink products and we’re looking for distribution in the US and UK We’re aiming to sell 30,000 cans this year [RRP AU$8 and can increase production to half a million a year * (source: the ‘Wine Grapes’ bible by the prophets Harding The Buyer TVClick below to watch The Buyer's library of online debates, videos and webinars. trade officials and Masters of Wine give their verdict on Borsa Vini Italiani London 2024 Borsa Vini Italiani London 2024 made a strong case for Italy’s viticultural diversity as it offered a snapshot of the country’s wines Featuring 40 producers from 13 different regions the event spanned the length of Italy and the breadth of its production Through the sheer variety of wines offered the event made a compelling case to the buyers in attendance Everything from Franciacorta to rosé to rich reds was available many producers offered sustainably made wines tapping into the increasingly important trend for responsible winemaking That variety is central to the Italian Trade Agency’s message commented: “We have the largest variety of different denominations The fact is that the Italian wine industry is so diversified; we have so many possibilities.” alongside exhibitors and visitors to the event offered his full thoughts on the significance of Borsa Vini Italiani 2024 in the video below three masterclasses also allowed visitors to explore facets of Italian wine in depth Patrick Schmitt MW opened the tastings at Borsa Vini Italiani London 2024 with an exploration of international grapes in Italy offer niche appeal when from an unexpected region John Downes MW then provided a lunchtime tasting of family estates from lesser known regions with family producers from Piemonte to Sicily on offer The masterclass highlighted the centrality of family estates and winemaking traditions in the Italian wine trade The final masterclass saw consultant and writer Walter Speller hone in on the wines of southern Italy a winemaking hotspot that is already much discussed These ranged from well established appellations like Trebbiano d’Abruzzo to undiscovered gems like Terre di Cosenza We are using cookies to give you the best experience on our website You can find out more about which cookies we are using or switch them off in settings This website uses cookies so that we can provide you with the best user experience possible Cookie information is stored in your browser and performs functions such as recognising you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful Strictly Necessary Cookie should be enabled at all times so that we can save your preferences for cookie settings we will not be able to save your preferences This means that every time you visit this website you will need to enable or disable cookies again Essential digital access to quality FT journalism on any device Complete digital access to quality FT journalism with expert analysis from industry leaders Complete digital access to quality analysis and expert insights complemented with our award-winning Weekend Print edition Terms & Conditions apply Discover all the plans currently available in your country See why over a million readers pay to read the Financial Times