https://orcid.org/0000-0001-7752-2155
Figures
Abstract
Data-driven decision-making (DDDM) has become integral to managerial and organizational processes in the era of digitalization and internationalization. This study explores the impact of DDDM on international firm performance. Leveraging AI language models, specifically BERT and ChatGLM2-6B, to quantify DDDM, we find that DDDM positively impacts international firm performance. To uncover the mechanisms underlying this correlation, we develop a framework explaining how DDDM creates sustainable value for firms, thereby enhancing international firm performance across four dimensions: pollution prevention (current internal), green innovation (future internal), sustainability information disclosure (current external), and sustainability vision co-creation (future external). Additionally, this study reveals that the positive impact of DDDM on international firm performance is amplified by higher market competition, greater foreign shareholding, and state ownership.
Citation: Xu M, Lu B (2026) The data-driven decision-making, sustainable value creation, and international firm performance: Micro-level evidence based on AI language models. PLoS One 21(2): e0340731. https://doi.org/10.1371/journal.pone.0340731
Editor: Chaohai Shen, East China Normal University, CHINA
Received: November 20, 2024; Accepted: December 24, 2025; Published: February 11, 2026
Copyright: © 2026 Xu, Lu. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data for this study are publicly available from the Zenodo repository (https://zenodo.org/records/15612235).
Funding: This study was supported by (1) the Major Humanities and Social Sciences Research Project in Zhejiang Higher Education Institutions (Grant No. 2024QN097); and (2) the General Scientific Research Project of the Department of Education of Zhejiang Province (Grant No. Y202455624). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
1. Introduction
In recent years, the digital economy has emerged as a pivotal catalyst for economic and industrial transformations worldwide, reshaping how businesses operate and compete [1]. Digital technologies, such as the Internet of Things, big data, cloud computing, and artificial intelligence, are transforming firm performance in global markets by redefining cross-border business processes and reshaping profit distribution across regions. These transformative changes underscore the importance of understanding how the application of digital technologies (usually termed as digitalization) impacts international business.
As a key capability enabled by digitalization, data-driven decision-making (DDDM) has attracted increasing attention for its potential to enhance decision accuracy through systematic data analysis [2]. However, debates persist regarding how to effectively measure DDDM and evaluate its actual impact on firm performance. While some studies use case-based approaches to explore the theoretical implications of DDDM [3], others rely on survey-based indicators [2]. These methods often suffer from limited sample sizes and potential subjectivity, highlighting the need for more robust and scalable approaches to measuring DDDM.
This study addresses this challenge by introducing a novel methodology that leverages advanced AI language models, including BERT and ChatGLM2-6B, to construct a firm-level DDDM index based on text data from the Management Discussion and Analysis (MD&A) sections of firms’ annual reports. Using this index, we analyze the relationship between DDDM and international firm performance within the theoretical framework of sustainable value creation [4]. This framework emphasizes the creation of internal and external sustainable value, both in the short and long term, as a source of firm-specific advantages in international markets.
Empirically, this study examines a dataset of 2,873 Chinese listed firms from 2007 to 2022, offering insights into how DDDM adoption influences international firm performance. China, as the world’s second-largest economy, provides a compelling context for exploring the broader implications of DDDM, given its prioritization of the digital economy as a national strategic priority [5]. Our findings confirm that DDDM enhances firms’ sustainable value and international competitiveness, contributing to the ongoing discourse on the strategic role of digitalization in the global economy.
By integrating DDDM into the sustainable value framework, this study not only advances the understanding of DDDM’s role in international business but also provides a scalable tool for assessing its adoption and impacts. The remainder of this paper is organized as follows. Section 2 reviews the relevant literature. Section 3 theoretically analyzes the relationship between DDDM and international firm performance, as well as the underlying mechanisms. Section 4 introduces the data, variable measurement, methodology and model specifications. Section 5 presents and discusses the empirical results, while Section 6 concludes this study.
2. Literature review
2.1. Digitalization and international firm performance
Various theories have been widely applied to explain how digital technologies improve firm performance in international markets. According to the transaction cost theory, digital platforms and tools help reduce information asymmetry, lower production and coordination costs, and facilitate cross-border operations [6–10]. Resource-based theory emphasizes digital capabilities such as data analytics, knowledge sharing, and virtual integration as sources of firm-specific advantages and dynamic capabilities [11–13]. These perspectives offer useful insights into the mechanisms through which digitalization contributes to international competitiveness. Recent studies further extend these perspectives by highlighting the distinct characteristics of digital platform firms and data-driven practices in international business. Ojala et al. (2018) [14] argue that the internationalization process of digital ventures differs significantly from traditional firms due to novel business models and platform architectures. Zeng et al. (2019) [15] emphasize that internal resources alone are insufficient for platform firms in foreign markets, highlighting the need for leveraging external networks and real-time data. These developments underscore the importance of examining DDDM in international business.
2.2. Sustainable value and international firm performance
In the context of an increasingly pronounced global pursuit of sustainable development, sustainability practices have been confirmed as crucial determinants for firms engaging in international business [16,17]. Several important reasons help explain why sustainable value is associated with international firm performance. First, from the perspective of international competition, the increase in sustainable value enables firms to gain competitive advantages in international markets [18–20]. Firms that hesitate to adopt sustainable practices inevitably face pressure the competitive pressures exerted by sustainability-oriented competitors [21]. Second, from the perspective of investors and shareholders, firms with high sustainable value are considered relatively reliable [22], and are able to access external financial resources (e.g., debt) at relatively low cost [19]. Third, from the perspective of host countries, the increase in sustainable value helps firms establish good relationships with host governments, thereby gaining continuous financial and political supports [23]. For instance, environmental protection taxes implemented by host countries allow firms that adopt sustainable development strategies to reduce tax burden [24]. Additionally, the implementation of sustainable strategies enhances internal controls, reducing the potential litigation risks in overseas markets [25], thereby promoting a steady growth.
It is worth noting that implementing sustainable strategies is not easy, as it involves significant implementation costs [26]. Research has found that in practice, some “greenwashing” or “window-dressing” firms, despite poor environmental performance, create a superficial strategic image that misleads outsiders into believing they are genuinely practicing sustainable development strategies. This approach aims to gain the benefits of sustainable strategies without bearing the costs of implementing them [27,28].
Nevertheless, relevant discussion on how digitalization or DDDM affects international firm performance remains insufficient. Firstly, the methods used to measure digitalization or DDDM exhibit some limitations. Specifically, the keyword text-searching method fails to capture the semantics of sentences and paragraphs, often leading to misjudgments and omissions. For example, when examining annual reports of listed firms, we found that a significant portion of digital-related keywords appeared in sections on industry background or future development plans, which do not necessarily reflect digitalization practices. Consequently, we highlight the limitations of directly extracting keywords without considering their specific semantics and propose addressing this issue with more nuanced approaches. Secondly, there is limited empirical evidence regarding the influence of DDDM on international firm performance due to data availability constraints, as DDDM is difficult to measure accurately. Moreover, understanding the potential mechanisms by which DDDM affects international firm performance is crucial. Existing studies remain fragmented and lack a theoretical framework to systematically summarize these mechanisms. In this study, we adopt a quantitative method to measure the extent of firms’ DDDM practices, using AI language models. We further highlight the ways in which DDDM enables value creation and provide empirical evidence from Chinese listed firms.
3. Theoretical framework and hypotheses
3.1. The effects of DDDM on international firm performance
DDDM extends beyond the mere utilization of digital technologies; it involves the strategic use of data and advanced analytics to guide managerial decisions and organizational behaviors [29]. DDDM provides firms with rational decision-making solutions across stages, including R&D, production, management and marketing [30], which could benefit firms’ competencies, such as entrepreneurial orientation, international knowledge, learning capability, network, marketing capability, and innovation capability [31]. Recent developments in generative AI have further amplified the potential of DDDM by enhancing firms’ ability to extract insights, model future scenarios, and explore non-traditional strategies [32]. Furthermore, it allows organizations to transcend the limitations of conventional algorithms and human intuition by supporting the exploration of non-traditional, innovative solutions, while also reinforcing predictive analytics and strategic planning processes [33]. In operational domains, such as supply chain management and benchmarking operations, generative AI contributes to more informed decision-making, ultimately improving firm competitive advantages [34]. Additionally, generative AI empowers entrepreneurs to innovate business models by enabling decisions that were previously difficult to make [35]. Thus, DDDM offers a novel perspective to understanding how digitalization translates into international firm performance.
DDDM facilitates firms in accessing and expanding overseas markets. Using predictive algorithms to evaluate a firm’s current status and forecast its market potential, DDDM can optimize decision-making and accelerate internationalization [36]. By using digital platforms and ecosystems, DDDM serves as an approach to internationalization, accumulating knowledge and relationships, and creating value for global customers [37]. AI readiness, as a form of cognitive technology supporting DDDM, helps organizations optimize investment decisions and select diversified growth paths in resource-constrained contexts, thereby boosting market performance for small and medium-sized firms (SMEs) [38]. It is worth noting that DDDM is particularly important for SMEs, as it helps reduce barriers, such as time, resources, information and network constraints, thereby promoting their participation in the international market [38–40]. DDDM can help female-led SMEs improve the efficiency of market intelligence in the internationalization process, allowing entrepreneurs to better understand customer needs, market trends, and competitor dynamics, thereby making wiser decisions [41]. Advanced data-mining techniques, such as machine learning and statistical approaches, support firms in strategically modeling and interpreting data, and underpin the market-oriented strategies of most SMEs through accessible information sources [39]. Therefore, we propose the following hypothesis:
Hypothesis 1. DDDM has a positive effect on international firm performance.
3.2. The mechanism of sustainable value creation
We argue that sustainable value creation can be a crucial mechanism for explaining how DDDM promotes international firm performance. We analyze how DDDM creates sustainable value across four dimensions: current internal, future internal, current external, and future external. Fig 1 illustrates the sustainable value creation framework of DDDM, which is based on the model proposed by Hart and Milstein (2003) [4].
Firstly, from the current internal dimension, DDDM enables firms to minimize waste emissions during production and operations, effectively preventing pollution issues [1], thus generating short-term internal sustainable value [4]. Specifically, DDDM, as an emerging management practice, can identify and resolve inefficiencies in production and operation processes, optimize energy consumption patterns, and promote energy conservation and pollution reduction [42,43]. With the widespread adoption of DDDM in management processes, inefficient and high-energy-consuming production procedures are gradually eliminated, reducing overall energy consumption [44], thereby improving firms’ green total factor productivity [45]. Meanwhile, DDDM contributes to the advancement of renewable energy [46], promotes the development of clean energy sources [47], and optimizes the energy structure [48]. DDDM improves the efficiency of resource utilization [49], and reduces emissions during production, operation, and service processes [1]. Therefore, we propose the following hypothesis:
Hypothesis 2a. DDDM increases firms’ sustainable value from the current internal dimension through pollution prevention, thereby improving international firm performance.
Secondly, from the future internal dimension, DDDM promotes the development of green technologies [1], driving green technology innovation and enabling firms to gain sustainable value internally in the long run [4]. Green technology innovation primarily refers to innovative activities related to energy conservation, emission reduction, pollution control, waste recycling, and green management [50]. At the firm level, DDDM significantly promotes green innovation [51], as firms adopting DDDM possess greater organizational flexibility and tend to integrate more proactively into global innovation networks, thereby more effectively keeping pace with international trends in cutting-edge green innovation [52]. The driving force behind green innovation comes not only from technological advancements of individual firms but also from the spillovers of green innovation among firms. For example, Industry 4.0 technologies, which leverage advanced digital technologies and big data analytics to drive data-based decisions, enhance the dissemination of information and accumulation of knowledge among firms, encouraging them to embrace open innovation practices, consequently fostering advancements in green technology innovation [53]. At the industry level, green technology innovation has positive spillover effects. When individual firms develop new green technologies, their competitors often adopt similar green technologies to maintain their market share and competitive position. Such green technology spillover spreads along the industry chain, benefiting firms at various stages and improving the overall level of green innovation in the industry [54]. Therefore, we propose the following hypothesis:
Hypothesis 2b. DDDM increases firms’ sustainable value from the future internal dimension through green innovation, thereby improving international firm performance.
Third, from the current external dimension, DDDM, as a kind of objective and relatively open decision-making approach, can increase information disclosure, thereby reducing information asymmetry between firms and external parties, enabling firms to gain sustainable value externally in the current period [4]. Information disclosure systems evolve accordingly and intrinsically with the advancement of digitalization, driving down communication costs and enhancing the transparency of information exchange [55]. This serves to extend interaction and reduce information asymmetry between firms and external parties, such as customers, suppliers, social organizations, governments, and the media [4]. As a result, firms leveraging DDDM accumulate increased external confidence from the public, enhancing corporate reputation. In the interaction between firms and external parties, disclosing the firms’ own sustainable development-related practices can integrate the needs of stakeholders, which in turn further encourages firms to adopt more sustainable practices to maintain ongoing external confidence and reputation [4]. Therefore, we propose the following hypothesis:
Hypothesis 2c. DDDM increases firms’ sustainable value from the current external dimension through sustainability information disclosure, thereby improving international firm performance.
Finally, from the future external dimension, DDDM enables firms to embrace their social commitments [56] and to co-create sustainability visions with external parties, thereby enhancing firms’ external sustainable value [4]. DDDM strengthens firms’ collective inclinations, thereby fostering the drive to meet their social responsibilities [57], mainly because the application of digital technologies in the process of DDDM changes how firms participate in collective practices [58]. Meanwhile, the openness and inclusiveness of DDDM encourage external parties to participate in firms’ decision-making, promoting socially responsible practices [59]. Research has found that some firms are leveraging digital platforms to co-create social responsibility initiatives (e.g., donations) with consumers, serving as an emerging form of digital marketing [60]. Some firms are leveraging digital training technologies (e.g., VR or AR) to improve employees’ job skills, which can be regarded as a socially responsible practice [57]. Crucially, firms adopting DDDM can swiftly address the values of social stakeholders during the collaborative co-creation of collective practices [57]. Therefore, we propose the following hypothesis:
Hypothesis 2d. DDDM increases firms’ sustainable value from the future external dimension through sustainability vision co-creation, thereby improving international firm performance.
4. Empirical model design
4.1. Regression models
Our empirical analysis explores how international firm performance responds to DDDM using the following baseline specification:
where i indexes firms, and t indexes years. The dependent variable is lnsales and the independent variable is DDDM. X is a set of control variables at the firm level to account for other factors that may affect international firm performance. ,
,
, and
represent firm, year, three-digit industry and province fixed effects, respectively. The key coefficient of interest is
, estimated on DDDM, which indicates the relationship between DDDM and international firm performance. The standard errors are clustered at the firm level to alleviate unobserved heterogeneity.
We investigate mechanisms using the specification shown in Eq (2), which is designed to examine the relationship between DDDM and mechanism variables (Mechanism).
where Mechanism denotes mechanism variables. The key coefficient of interest is , estimated for DDDM, which captures the relationship between DDDM and mechanism variables. All of the control variables and fixed effects in the specification adhere to those used in the baseline specification.
This method of mechanism analysis is derived from the idea of mediation effect analysis. Although the idea of mediation effects has been widely adopted, many scholars have raised concerns about its potential endogeneity issues. In response, various approaches have been proposed to improve the reliability of mediation analysis. A common practice [61,62] is to propose one or several mediator variables whose relationship with the dependent variable is theoretically intuitive and temporally and logically proximate. Under these conditions, formal causal inference methods are unnecessary to assess the causal link from mediator variables to the outcome.
Furthermore, to investigate the moderating effects of firm characteristics, we select several independent variables, and add interaction terms between these variables and DDDM into the baseline specification, as shown in Eq (3).
where Moder denotes firm characteristic variables. All other variables and fixed effects are the same as the baseline specification. The key coefficient of interest is , estimated for the interaction term, which indicates whether the examined firm characteristics moderate the relationship between DDDM and firm performance.
4.2. Variable definitions and explanations
4.2.1. The independent variable.
The independent variable is the DDDM index (DDDM), which measures the degree to which firms adopt DDDM. Existing research on DDDM demonstrates multidisciplinary characteristics, yet quantitative measurements primarily rely on questionnaire surveys [2,63,64]. To address the limitations of the questionnaire method, such as subjectivity and limited sample size, this study adopts a text classification approach based on AI language models and unstructured text data to quantify DDDM.
The DDDM index was constructed using 50,000 annual reports of Chinese listed firms (2007−2022). First, Python was employed to extract the Management Discussion and Analysis (MD&A) sections via localization algorithms, filter irrelevant content (e.g., financial tables) using keyword matching, and split overlength paragraphs into 1.13M coherent units. Second, 50,000 paragraphs were randomly sampled for manual annotation based on predefined criteria for identifying DDDM practices, yielding 3,000 positive samples (labeled 1) and 47,000 negative samples (labeled 0). To mitigate insufficient learning of positive sample features by LLMs, 3,000 negative samples were resampled with all positive samples to form a balanced 6,000-sample training set, wherein 1,000 pieces were allocated as the development set. Additionally, 5,000 units were randomly sampled as the test set from the total sample. Third, A cascading strategy for text classification was implemented: Stage 1 used BERT to screen 1.13 million paragraphs (98% recall on the test set), generating 70,000 class-1 candidate texts; Stage 2 employed ChatGLM2-6B—retrained with tailored prompts—to classify these texts, achieving 96.4% overall accuracy. This strategy demonstrated superior performance over a single model in both efficiency and precision. Finally, following academic conventions, the DDDM index was calculated as the natural logarithm of (1 + the number of class-1 paragraphs in each firm’s annual report).
4.2.2. The dependent variables.
As for the dependent variables, we use the natural logarithm of sales revenue in overseas markets (lnsales) to proxy for international firm performance in the baseline regressions. Although firm performance is multidimensional, sales growth has been widely accepted as one of the proxy variables for firm performance [65]. Similar to prior work [66], we also use the natural logarithm of the number of related firms in overseas markets (lnaffiliates) as another proxy for international firm performance in the robustness test.
4.2.3. The control variables.
As for the control variables, following prior work [67], we select a group of firm-level variables, including a firm’s net return on assets (ROA), the natural logarithm of the capital-labor ratio measured by dividing the total fixed assets by the number of employees (lncapital), total liabilities scaled by total assets (liability), shareholders’ equity divided by the market value of equity (Book_Market), the shareholding ratio of the five largest shareholders (top5), and the percentage change in domestic sales revenue (Home_sales). All continuous variables are winsorized at 1%.
4.3. Data and sample
Data used to measure DDDM index are sourced from annual reports of Chinese listed firms and these reports are collected from CNINF.COM. Data on Chinese listed firms are collected from the China Stock Market and Accounting Research Database (CSMAR) and Chinese Research Data Services Platform (CNRDS). Our sample comprises 2,873 A-share listed firms in China from 2007 to 2022. Listed firms that satisfied one of the following conditions were excluded: (1) were in finance-related industries; (2) were listed as ST and *ST; (3) had missing financial data; or (4) were established in the observed year.
Table 1 provides a comprehensive summary of variables in this study, including their labels, definitions, explanations and sources.
4.4. Statistics and correlations
Table 2 shows descriptive statistics of the full sample of 22,117 observations. The independent variable, DDDM, is ranging from 0 to 2.9440, with a median of 0.6930, while the dependent variable, lnsales, ranges from 3.3420 to 14.8900 with a median of 10.1900. This indicates DDDM and international firm performance exhibit substantial variation in our sample.
Table 3 presents the Pearson correlation coefficients among the variables. We also test the variance inflation factor (VIF) to ensure the low multicollinearity among variables. The mean VIF is 1.23.
5. Empirical results
5.1. Baseline results
Following Eq (2), we employ multi-dimensional fixed effects to estimate the impact of DDDM on international firm performance. Table 4 reports the baseline results. Column (1) shows the results estimated under two-way fixed effects without control variables. The coefficient on DDDM is 0.0408, statistically significant at the 5% level. Column (2) shows the results in the presence of the aforementioned control variables with firm and year fixed effects, and standard errors clustered at the firm level. The coefficient on DDDM is 0.0399, statistically significant at the 5% level. Column (3) includes industry and province fixed effects to control for industry and geographical factors that may influence international firm performance. The coefficient on DDDM is 0.0365, statistically significant at the 5% level. These results indicate that DDDM significantly improves international firm performance, supporting Hypothesis 1.
5.2. Endogeneity and robustness
This section conducts several tests to ensure the robustness of the baseline results. Tables 5 and 6 present the results of these tests.
5.2.1. IV estimations.
In this subsection, we apply the Instrumental Variable (IV) method to address concerns about possible endogeneity, particularly omitted variable bias and reverse causality. Following the approach of the Bartik instrument [68], we first calculate the province-industry-year average of DDDM (except itself) in the previous term (Avg_DDDM). This Avg_DDDM is expected to correlate with a firm’s DDDM, but it will not directly affect a firm’s decision-making. Therefore, the potentially endogenous component of DDDM can be partially removed, reducing concerns about endogeneity. Next, we compute the national average growth rate of DDDM (Delta_DDDM), which is plausibly exogenous to individual firm behavior. Then, we multiply Avg_DDDM and Delta_DDDM to obtain the instrumental variable, IV_DDDM. We use a two-stage least squares approach (2SLS) to estimate Eq (2) with this instrument.
The results in the first and second stages of IV (2SLS) are reported in the first and second columns of Table 5. More specifically, the coefficient on IV_DDDM is significantly positive at the 1% level in the first stage, and the coefficient on DDDM is significantly positive at the 5% level in the second stage. This IV passes both the weak identification and under-identification tests: the Cragg-Donald Wald F-statistic (17.515) exceeds the Stock-Yogo critical value at the 10% level (16.38), and the p-value of Kleibergen-Paap rk LM statistic is statistically significant at the 1% level. The IV (2SLS) results support the baseline findings that DDDM is positively related to international firm performance.
5.2.2. Alternative proxies of variables.
We conduct robustness tests on the selection of the proxy for the dependent variable to alleviate concerns that our baseline analysis may yield biased findings due to reliance on a single measure. Following the extant literature, we adopt two alternative proxies for DDDM measured by text-searching and sourced from CNRDS. This database has compiled over 100 digitization-related keywords to construct a firm-level digitization index. One alternative proxy for DDDM is the natural logarithm of the number of searched keywords (lnword), and the other is the natural logarithm of the number of paragraphs that contain these keywords (lnpara). Columns (1) and (2) of Table 6 report the results of regressions using these alternative measures of DDDM. We find that the coefficients on DDDM are significant at the 1% level. These results confirm that our baseline finding is robust to alternative proxies for DDDM.
Similarly, we consider the proxy for the independent variable. Following extant literature, we use the natural logarithm of the number of related overseas firms (lnaffiliates) as an alternative proxy for international firm performance. Column (3) of Table 6 reports the results of regressions where international firm performance is measured by the number of related overseas firms. We find that the coefficient on DDDM is still significant at the 5% level. Accordingly, the baseline findings hold with an alternative proxy for the dependent variable.
5.2.3. Sample selection.
Next, we conduct robustness tests on sample selection to reduce concerns that the baseline results might depend on a specific sample period. In Column (4) of Table 6, we set 2009 as the starting year, given that 2008 was the year of the global financial crisis. The results are generally consistent with baseline findings, alleviating concerns that the baseline results might be driven by a particular sample period.
5.2.4. Alternative method of regression.
Furthermore, we conduct robustness tests using an alternative estimation method to reduce concerns that the baseline results might depend on the specific regression method. We adopt the Poisson pseudo maximum likelihood (PPML) to re-estimate Eq (2). PPML is a variant of Poisson regression, suitable for non-negative dependent variables that include zero values [69], which is precisely the case in this study. Column (5) of Table 6 reports that the coefficient on DDDM is significant. Accordingly, the baseline findings hold under the alternative regression model.
5.3. Mechanism analysis
In this section, we test Hypotheses 2a to 2d and examine why DDDM is positively related to international firm performance.
5.3.1. The mechanism of pollution prevention.
First, from the current internal dimension, we investigate whether DDDM can reduce pollution emissions and increase firms’ short-term sustainable value internally. Regression results are shown in Table 7. Column (1) reports the regression results with the mechanism variable, Monitor, which equals1 if the firm is monitored as a polluter and 0 otherwise. The results in Column (1) show that the coefficient on DDDM is negative and significant at the 1% level, which indicates that DDDM reduces the probability of being monitored as a polluter. Column (2) reports the regression results using Accidents as the mechanism variable, which equals 1 if the firm experiences a major environmental pollution incident and 0 otherwise. The results in Column (2) show that the coefficient on DDDM is negative and significant at the 5% level, which indicates that DDDM reduces the probability of environmental pollution incidents. Column (3) reports the regression results using Illegal as the mechanism variable, which equals 1 if the firm experiences illegal events related to environment protection and 0 otherwise. The results in Column (5) show that the coefficient on DDDM is negative and statistically significant at the 5% level, indicating that DDDM reduces the likelihood of environmental illegal events. Columns (4) and (5) report the regression results using ISO9001 and ISO14001 as the mechanism variables, which equal 1 if the firm has obtained the ISO9001 or the ISO14001 respective certification, and 0 otherwise. These results show that the coefficients on DDDM are both positive and significant at the 1% level, which indicates that DDDM facilitates firms’ attainment the environmental certifications.
The results in Table 7 indicate that the positive impact of DDDM on international firm performance is associated with its effect on pollution prevention. When the adoption of DDDM increases, firms tend to reduce waste and emissions, lower the incidence of environmental pollution events, and are more likely to obtain environmental certifications. Therefore, Hypothesis 2a is supported.
5.3.2. The mechanism of green innovation.
Secondly, from the future internal dimension, we investigate whether DDDM can promote green innovation and increase firms’ long-term sustainable value internally. The mechanism of green innovation can be captured by the number of green invention patents independently obtained (Green_Invention), the number of green utility model patents independently obtained (Green_Utility), and the sum of these two patents (Green_sum). The results of regressions are presented in Table 8. The coefficients on DDDM are significantly positive in all three columns. DDDM not only promotes growth in firms’ green invention patents but also boosts the development of green utility model patents. The improvement in green innovation performance increases firms’ long-term sustainable value, strengthens their international competitiveness, and consequently enhances international firm performance. Therefore, Hypothesis 2b is supported.
5.3.3. The mechanism of sustainability information disclosure.
Thirdly, from the current external dimension, we investigate whether DDDM can encourage sustainability information disclosure and increase externally generated short-term sustainable value. The mechanism of sustainability information disclosure can be captured by ESG scores, which are typically assessed by external agencies that collect and analyze relevant data, and evaluate firms based on environmental, social and governance (ESG) criteria. Specifically, we select the following four variables as proxies: (1) environmental performance score (E); (2) social performance score (S); (3) governance performance score (G); and (4) overall ESG score (ESG). These variables are individually substituted as the mechanism variable in Eq (2). The regression results, reported in columns (1) to (4) of Table 9, show that the coefficients on DDDM are all significantly positive. This confirms that DDDM promotes the disclosure of sustainability information, thereby increasing the current sustainable value obtained externally by firms, aiding them in participating in international competition, and improving their international firm performance. Therefore, Hypothesis 2c is supported.
5.3.4. The mechanism of sustainability vision co-creation.
Finally, from the future external dimension, we explore whether DDDM encourages sustainability vision co-creation and increases externally generated long-term sustainable value. To measure firms’ interaction with external stakeholders [70], we use the following three variables as proxies: (1) analyst attention (analyst), which counts the number of analysts (teams) tracking and analyzing the firm, reflecting the degree of interaction between the firm and investment analysts; (2) research report attention (reported), which counts the number of research reports tracking and analyzing the firm, reflecting the degree of interaction between the firm and research institutions; (3) sentiment of forum posts (positive), which measures the proportion of positive posts in the firm’s stock forum, reflecting the degree of interaction between the firm and retail investors. These variables are individually substituted as the Mechanism in Eq (2). The regression results, reported in columns (1) to (3) of Table 10, show that the coefficients on DDDM are significantly positive. This indicates that DDDM enhances firms’ interactions with analysts, research institutions, and the public, forming a virtuous cycle that increases future external sustainable value and promotes firms’ performance. Accordingly, Hypothesis 2d is verified.
5.4. Moderating effects of firm heterogeneity
Recent studies suggest that DDDM can be further contextualized within institutional theory, which highlights how decision-making practices are shaped by broader regulatory and normative environments [71,72]. In light of this, we examine the Herfindahl Index (HHI), the ratio of shares owned by foreigners (foreign_share), and the nature of firm ownership (SOE) as three proxies for institutional environments, which serve as the heterogeneity variables in this study. Table 11 presents the results of these regressions.
First, the relationship between DDDM and international firm performance can be affected by the level of market competition (HHI). DDDM enables firms to maintain competitive advantages, which helps firms expand overseas markets at a relatively low cost. As expected, when the competition in a market becomes intensifies, the influence of DDDM is expected to be stronger. To examine whether the positive impact of DDDM on international firm performance is strengthened as market competition increases, we add interaction terms between DDDM and HHI into the baseline specification. We choose two Herfindahl indexes (HHI_A and HHI_B) as proxies for the level of competition in a market, and report the results in the first and second columns of Table 11. As expected, the results show that the estimated coefficients of the interaction terms are negative and significant. In other words, as the Herfindahl index decreases and the market competition becomes more intense, the positive relationship between DDDM and international firm performance is reinforced.
Second, the role of the ratio of shares owned by foreigners (foreign_share) should be particularly considered in a study relevant to international firm performance. The interaction term between DDDM and foreign_share is introduced into the baseline specification and the result is shown in Column (3) of Table 11. The results show that the estimated coefficient of the interaction term is significantly positive, indicating that as the proportion of foreign shareholding increases, the impact of DDDM on international firm performance also increases. This suggests that firms with a higher proportion of foreign shareholding are better able to leverage sustainable value, thereby improving their international firm performance.
Third, the nature of firm ownership (SOE) is especially relevant to the research of the aforementioned mechanism, which could be captured by the dummy variable of SOE that equals 1 if the firm is state-owned and 0 otherwise. The interaction term between DDDM and SOE is introduced into the baseline specification and the result is shown in Column (4) of Table 11. The coefficient on the interacted term is negative and significant, which evidences that the fact that firms are state-owned can weaken the impact of DDDM on international firm performance is weaker when firms are state-owned, compared to non-state-owned firms. One possible explanation is that state-owned firms are relatively less affected by external environments or social shareholders, thereby reducing the necessity of increasing sustainable value and diminishing the impact of DDDM on international firm performance. Therefore, DDDM is more conducive to overcoming the challenges of international firm performance faced by non-state-owned firms.
6. Conclusion and discussion
6.1. Conclusion
This study introduces AI language models to quantify DDDM, laying an important foundation for subsequent research on digital management and digital economy. By leveraging two AI language models, BERT and ChatGLM2-6B, to conduct the text classification tasks, we effectively identify DDDM practices in annual reports of listed firms, and develop a DDDM index.
Furthermore, this study attempts to complement the current literature by highlighting the importance of DDDM to international firm performance. We explore the relationship between DDDM and international firm performance, using firm-level panel data. The sample includes about 2,873 listed firms in China and spans the period of 2007–2022. We start our empirical analysis by investigating international firm performance as a function of DDDM and other firm-level variables with the time, firm, province, and industry fixed effects. We find that DDDM is positively associated with international firm performance. Next, we attempt to answer why DDDM affects international firm performance by exploiting the mechanisms. Our theoretical framework, inspired by Hart and Milstein (2003) [4], underscores how DDDM creates sustainable value and promotes international firm performance in four dimensions. Specifically, DDDM has pollution prevention effects, green innovation effects, sustainability information disclosure effects, and sustainability vision co-creation effects in the dimensions of current internal, future internal, current external, and future external, respectively. This enhances firms’ competitive advantages in international markets, thereby improving international firm performance. Finally, we examine how firms’ characteristics moderate the effects of DDDM on international firm performance. The positive impacts of DDDM on international firm performance is enhanced when the level of market competition is higher, the proportion of foreign shareholding is higher, and the firm is state-owned.
6.2. Limitations and future research
This study primarily focuses on Chinese listed firms, which may not fully capture the economic impacts of DDDM on firm performance. The exclusion of small and medium-sized enterprises from the research sample is a limitation. Future research should focus on the full range of firms to provide a more general understanding of the effects of DDDM on firm performance across firm sizes. In addition, recent development of AI technology is likely to increases the application scope of DDDM; however, despite all these promising developments, caution remains that human intelligence should not be overlooked. Future research should focus on the comparison of AI-based and human-based decision-making to find out how to optimize the decision-making in practice. Future studies could also attempt preliminary experiments that integrate the proposed framework with emerging AI advances to further enrich the understanding of DDDM in practice.
Acknowledgments
We sincerely thank Professor Ma Shuqin for her guidance on the research direction of this study. We also extend our sincere thanks to Dr. Zhang Chenyu, AI engineer Ling Yun, Dr. Cheng Yang, Dr. Jin Yanyi, and Ms. Xu Yi for their assistance in constructing the DDDM index.
References
- 1. Zhu Q, Ma D, He X. Digital transformation and firms’ pollution emissions. Technological Forecasting and Social Change. 2023;197:122910.
- 2. Brynjolfsson E, McElheran K. The Rapid Adoption of Data-Driven Decision-Making. American Economic Review. 2016;106(5):133–9.
- 3. Zaitsava M, Marku E, Di Guardo MC. Is data-driven decision-making driven only by data? When cognition meets data. European Management Journal. 2022;40(5):656–70.
- 4. Hart SL, Milstein MB. Creating sustainable value. AMP. 2003;17(2):56–67.
- 5. Wen H, Zhong Q, Lee C-C. Digitalization, competition strategy and corporate innovation: Evidence from Chinese manufacturing listed companies. International Review of Financial Analysis. 2022;82:102166.
- 6. Feliciano-Cestero MM, Ameen N, Kotabe M, Paul J, Signoret M. Is digital transformation threatened? A systematic literature review of the factors influencing firms’ digital transformation and internationalization. Journal of Business Research. 2023;157:113546.
- 7. Lehdonvirta V, Kässi O, Hjorth I, Barnard H, Graham M. The Global Platform Economy: A New Offshoring Institution Enabling Emerging-Economy Microproviders. J Manage. 2019;45(2):567–99. pmid:30774168
- 8. Rangan S, Sengul M. Information technology and transnational integration: Theory and evidence on the evolution of the modern multinational enterprise. J Int Bus Stud. 2009;40(9):1496–514.
- 9. Chen W, Kamal F. The impact of information and communication technology adoption on multinational firm boundary decisions. J Int Bus Stud. 2016;47(5):563–76.
- 10. Jean R-JB, Kim D, Cavusgil E. Antecedents and outcomes of digital platform risk for international new ventures’ internationalization. Journal of World Business. 2020;55(1):101021.
- 11. Chang K-H, Chen Y, Huang H-F. Information technology and partnership dynamic capabilities in international subcontracting relationships. International Business Review. 2015;24(2):276–86.
- 12. Kim D, Jean R-JB, Sinkovics RR. Drivers of Virtual Interfirm Integration and Its Impact on Performance in International Customer–Supplier Relationships. Manag Int Rev. 2018;58(3):495–522.
- 13. Parente RC, Geleilate J-MG, Rong K. The Sharing Economy Globalization Phenomenon: A Research Agenda. Journal of International Management. 2018;24(1):52–64.
- 14. Ojala A, Evers N, Rialp A. Extending the international new venture phenomenon to digital platform providers: A longitudinal case study. Journal of World Business. 2018;53(5):725–39.
- 15. Zeng J, Khan Z, De Silva M. The emergence of multi-sided platform MNEs: Internalization theory and networks. International Business Review. 2019;28(6):101598.
- 16. Kolk A. The social responsibility of international business: From ethics and the environment to CSR and sustainable development. Journal of World Business. 2016;51(1):23–34.
- 17. Aguilera-Caracuel J, Guerrero-Villegas J, García-Sánchez E. Reputation of multinational companies. EJMBE. 2017;26(3):329–46.
- 18. Kim J, Kim M, Im S, Choi D. Competitiveness of E Commerce Firms through ESG Logistics. Sustainability. 2021;13(20):11548.
- 19. Malik N, Kashiramka S. “Impact of ESG disclosure on firm performance and cost of debt: Empirical evidence from India”. Journal of Cleaner Production. 2024;448:141582.
- 20. Behl A, Kumari PSR, Makhija H, Sharma D. Exploring the relationship of ESG score and firm value using cross-lagged panel analyses: case of the Indian energy sector. Ann Oper Res. 2021;313(1):231–56.
- 21. Zhang D. Environmental regulation, green innovation, and export product quality: What is the role of greenwashing? International Review of Financial Analysis. 2022;83:102311.
- 22. Bansal M, Samad TA, Bashir HA. The sustainability reporting-firm performance nexus: evidence from a threshold model. JGR. 2021;12(4):491–512.
- 23. Liu Y, Quan B, Xu Q, Forrest JY-L. Corporate social responsibility and decision analysis in a supply chain through government subsidy. Journal of Cleaner Production. 2019;208:436–47.
- 24. Wang X, Ye Y. Environmental protection tax and firms’ ESG investment: Evidence from China. Economic Modelling. 2024;131:106621.
- 25. Zhang H, Zhang H, Tian L, Yuan S, Tu Y. ESG performance and litigation risk. Finance Research Letters. 2024;63:105311.
- 26. Palmer K, Oates WE, Portney PR. Tightening Environmental Standards: The Benefit-Cost or the No-Cost Paradigm? Journal of Economic Perspectives. 1995;9(4):119–32.
- 27. Delmas MA, Burbano VC. The Drivers of Greenwashing. California Management Review. 2011;54(1):64–87.
- 28. Sajid M, Zakkariya KA, Suki NM, Islam JU. When going green goes wrong: The effects of greenwashing on brand avoidance and negative word-of-mouth. Journal of Retailing and Consumer Services. 2024;78:103773.
- 29. Doshi AR, Bell JJ, Mirzayev E, Vanneste BS. Generative artificial intelligence and evaluating strategic decisions. Strategic Management Journal. 2024;46(3):583–610.
- 30. Brynjolfsson E, McElheran K. Data in action: data-driven decision making and predictive analytics in US manufacturing. In: Rotman School of Management Working Paper, 2019.
- 31. Lee Y-Y, Falahat M, Sia B-K. Impact of Digitalization on the Speed of Internationalization. IBR. 2019;12(4):1.
- 32. Fosso Wamba S, Guthrie C, Queiroz MM, Minner S. ChatGPT and generative artificial intelligence: an exploratory study of key benefits and challenges in operations and supply chain management. International Journal of Production Research. 2023;62(16):5676–96.
- 33. Singh K, Chatterjee S, Mariani M. Applications of generative AI and future organizational performance: The mediating role of explorative and exploitative innovation and the moderating role of ethical dilemmas and environmental dynamism. Technovation. 2024;133:103021.
- 34. Dubey R, Gunasekaran A, Papadopoulos T. Benchmarking operations and supply chain management practices using Generative AI: Towards a theoretical framework. Transportation Research Part E: Logistics and Transportation Review. 2024;189:103689.
- 35. Shore A, Tiwari M, Tandon P, Foropon C. Building entrepreneurial resilience during crisis using generative AI: An empirical study on SMEs. Technovation. 2024;135:103063.
- 36. Neubert M. The Impact of Digitalization on the Speed of Internationalization of Lean Global Startups. TIM Review. 2018;8(5):44–54.
- 37. Nambisan S, Zahra SA, Luo Y. Global platforms and ecosystems: Implications for international business theories. J Int Bus Stud. 2019;50(9):1464–86.
- 38. Denicolai S, Zucchella A, Magnani G. Internationalization, digitalization, and sustainability: Are SMEs ready? A survey on synergies and substituting effects among growth paths. Technological Forecasting and Social Change. 2021;166:120650.
- 39. Hervé A, Schmitt C, Baldegger R. Internationalization and Digitalization: Applying digital technologies to the internationalization process of small and medium-sized enterprises. TIM Review. 2020;10(7):28–40.
- 40. Jean R-JB, Kim D. Internet and SMEs’ internationalization: The role of platform and website. Journal of International Management. 2020;26(1):100690.
- 41. Pergelova A, Manolova T, Simeonova-Ganeva R, Yordanova D. Democratizing Entrepreneurship? Digital Technologies and the Internationalization of Female-Led SMEs. Journal of Small Business Management. 2018;57(1):14–39.
- 42. Cai S, Chen X, Bose I. Exploring the role of IT for environmental sustainability in China: An empirical analysis. International Journal of Production Economics. 2013;146(2):491–500.
- 43. May G, Stahl B, Taisch M, Kiritsis D. Energy management in manufacturing: From literature review to a conceptual framework. Journal of Cleaner Production. 2017;167:1464–89.
- 44. Ishida H. The effect of ICT development on economic growth and energy consumption in Japan. Telematics and Informatics. 2015;32(1):79–88.
- 45. Wang J, Liu Y, Wang W, Wu H. How does digital transformation drive green total factor productivity? Evidence from Chinese listed enterprises. Journal of Cleaner Production. 2023;406:136954.
- 46. Zheng J, Wang X. Can mobile information communication technologies (ICTs) promote the development of renewables?-evidence from seven countries. Energy Policy. 2021;149:112041.
- 47. Tang C, Xu Y, Hao Y, Wu H, Xue Y. What is the role of telecommunications infrastructure construction in green technology innovation? A firm-level analysis for China. Energy Economics. 2021;103:105576.
- 48. Xu Q, Zhong M, Li X. How does digitalization affect energy? International evidence. Energy Economics. 2022;107:105879.
- 49. Leng J, Ruan G, Jiang P, Xu K, Liu Q, Zhou X, et al. Blockchain-empowered sustainable manufacturing and product lifecycle management in industry 4.0: A survey. Renewable and Sustainable Energy Reviews. 2020;132:110112.
- 50. Wang Q, Qu J, Wang B, Wang P, Yang T. Green technology innovation development in China in 1990-2015. Sci Total Environ. 2019;696:134008. pmid:31465930
- 51. Zheng Y, Zhang Q. Digital transformation, corporate social responsibility and green technology innovation- based on empirical evidence of listed companies in China. Journal of Cleaner Production. 2023;424:138805.
- 52. Srinivasan R, Swink M. An Investigation of Visibility and Flexibility as Complements to Supply Chain Analytics: An Organizational Information Processing Theory Perspective. Production and Operations Management. 2018;27(10):1849–67.
- 53. Mubarak MF, Tiwari S, Petraite M, Mubarik M, Raja Mohd Rasi RZ. How Industry 4.0 technologies and open innovation can improve green innovation performance? MEQ. 2021;32(5):1007–22.
- 54. Li J, Lian G, Xu A. How do ESG affect the spillover of green innovation among peer firms? Mechanism discussion and performance study. Journal of Business Research. 2023;158:113648.
- 55. Heim H. Digital Fashion Revolutions: Supply chain transparency, digitalization and the non-disclosure paradox. Fashion Practice. 2022;14(3):329–51.
- 56. Truant E, Borlatto E, Crocco E, Bhatia M. ESG performance and technological change: Current state-of-the-art, development and future directions. Journal of Cleaner Production. 2023;429:139493.
- 57. Jiang W, Wu J, Yang X. Does digitization drive corporate social responsibility?. International Review of Economics & Finance. 2023;88:14–26.
- 58. Young A, Selander L, Vaast E. Digital organizing for social impact: Current insights and future research avenues on collective action, social movements, and digital technologies. Information and Organization. 2019;29(3):100257.
- 59. Adams CA, Frost GR. The internet and change in corporate stakeholder engagement and communication strategies on social and environmental performance. Journal of Accounting & Organizational Change. 2006;2(3):281–303.
- 60. Kull AJ, Heath TB. You decide, we donate: Strengthening consumer–brand relationships through digitally co-created social responsibility. International Journal of Research in Marketing. 2016;33(1):78–92.
- 61. Zheng J, Jiang Y, Cui Y, Shen Y. Green bond issuance and corporate ESG performance: Steps toward green and low-carbon development. Research in International Business and Finance. 2023;66:102007.
- 62. The Persistent Effects of Peru’s MiningMita. Econometrica. 2010;78(6):1863–903.
- 63. Gul R, Leong K, Mubashar A, Al-Faryan MAS, Sung A. The Empirical Nexus between Data-Driven Decision-Making and Productivity: Evidence from Pakistan’s Banking Sector. Cogent Business & Management. 2023;10(1).
- 64. Bar-Gill S, Brynjolfsson E, Hak N. Helping Small Businesses Become More Data-Driven: A Field Experiment on eBay. Management Science. 2024;70(11):7345–72.
- 65. Delmar F, Davidsson P, Gartner WB. Arriving at the high-growth firm. Journal of Business Venturing. 2003;18(2):189–216.
- 66. Contractor FJ, Kundu SK, Hsu C-C. A three-stage theory of international expansion: the link between multinationality and performance in the service sector. J Int Bus Stud. 2002;34(1):5–18.
- 67. Bhandari KR, Zámborský P, Ranta M, Salo J. Digitalization, internationalization, and firm performance: A resource-orchestration perspective on new OLI advantages. International Business Review. 2023;32(4):102135.
- 68. Goldsmith-Pinkham P, Sorkin I, Swift H. Bartik Instruments: What, When, Why, and How. American Economic Review. 2020;110(8):2586–624.
- 69. Correia S, Guimarães P, Zylkin T. Fast Poisson estimation with high-dimensional fixed effects. The Stata Journal: Promoting communications on statistics and Stata. 2020;20(1):95–115.
- 70. Cui C, He G, Xie D. Information externalities, analyst research resource allocation, and stock pricing efficiency. China Journal of Accounting Studies. 2022;10(1):1–25.
- 71. de Curtò J, de Zarzà I, Fervier LS, Sanagustín-Fons V, Calafate CT. An Institutional Theory Framework for Leveraging Large Language Models for Policy Analysis and Intervention Design. Future Internet. 2025;17(3):96.
- 72. Safaei M, Longo J. The End of the Policy Analyst? Testing the Capability of Artificial Intelligence to Generate Plausible, Persuasive, and Useful Policy Analysis. Digit Gov: Res Pract. 2024;5(1):1–35.