Big data application and firm markups: evidence from China - Scientific Reports

This study investigates the relationship between big data applications and firms' price markups. By constructing a heterogeneous firm model with variable markups, we analyze the mechanisms through which big data applications influence firms' price markups and conduct empirical tests using micro-level firm data. The results indicate that big data applications significantly enhance firms' price markups. Mechanism analysis reveals that promoting product innovation and improving production efficiency are two key channels through which big data applications contribute to higher markups. Furthermore, the positive effect of big data applications on firms' markups exhibits heterogeneity across organizational, technological, and environmental dimensions. These findings suggest that while big data applications positively influence firms' markups, the realization of this effect depends on the synergistic support of various complementary resources. The research uncovers the intrinsic mechanisms through which big data applications shape firms' competitive advantages and market power, providing valuable insights for policy formulation. In the era of the digital economy, data have become a critical factor of production. A growing number of corporate managers are gradually moving away from decision-making based on intuition and adopting data-driven analytical methods to support more scientific decision-making processes. However, as a byproduct of corporate production activities, data often remain in their raw stored state without effective governance, intelligent analysis, and in-depth mining, making it difficult to transform them into high-value information or structured knowledge that can support decision-making. With the increasing prevalence of general-purpose technologies such as big data, enterprises are now able to integrate and analyze large-scale unstructured data from multiple sources, gaining deeper business insights and thereby building sustainable competitive advantages in the market. Despite the broad application prospects of big data, existing literature primarily focuses on analyzing its impact on short-term productivity and performance, while research on how enterprises can derive sustained competitive advantages and market power from such investments remains relatively scarce. The evolution and evaluation of corporate competitive advantages have always been a focal point in academic research, with indicators such as productivity and sales revenue receiving widespread attention. However, in the context of the digital economy, digital platform enterprises, leveraging the network effects, economies of scale, and flywheel effects of data, are more likely to drive market structures toward monopoly or oligopoly, exhibiting a pronounced "winner-takes-all" characteristic. Moreover, although many enterprises continue to increase their investments in information technology (IT), the full realization of their potential often depends on long-term and systematic investments in intangible assets, including new business processes, business models, and skills training. This makes traditional indicators such as productivity and sales revenue inadequate for timely and comprehensive reflection of the actual transformations, innovations, and indirect benefits occurring within enterprises, leading to the so-called "IT productivity paradox." In this context, overemphasizing productivity and similar metrics as standards for evaluating corporate competitive advantages may instead trigger more intense price competition, trapping enterprises in an "efficiency paradox" where increased production does not translate into increased profits, while also failing to accurately measure the true benefits brought by IT investments. In contrast, the firm price markup -- the deviation between price and marginal cost -- provides a more comprehensive reflection of a firm's competitive advantage. This indicator not only captures a firm's performance in both "cost reduction" (lowering marginal costs) and "quality improvement" (raising product pricing) but also offers a more integrated measure of the firm's ability to transform internal technological investments into sustained competitive advantages. Existing literature has explored the determinants of firm markups from various perspectives, including market competition and industrial concentration, international trade, and supply and demand shocks. Furthermore, Crouzet & Eberly and De Ridder suggest that intangible assets, particularly software-related assets, have a positive impact on markups. This raises an important question: As another form of intangible asset, can big data also enhance firm markups? However, few studies have incorporated big data and firm markups into a unified analytical framework for in-depth investigation. As a new factor of production in the digital age, data is characterized by non-rivalry, reproducibility, and timeliness. Over the past decade, data has experienced explosive growth in scale, variety, and generation speed, often referred to as "big data." Faced with increasingly vast and complex data resources, how to effectively utilize this data has become a key focus for enterprises. Big data application refers to the process by which enterprises leverage technologies such as data collection, storage, cleaning, analysis, and mining to process massive, multi-source, and rapidly growing data. Through this process, valuable information is extracted and applied to real-world business operations to optimize decision-making and create value. Conceptually, "big data" emphasizes the characteristics of the data itself, while "big data application" focuses more on the practical process of extracting value, supporting decisions, and driving innovation based on data. Through big data applications, firms can conduct in-depth mining of multi-source, massive, and unstructured data to obtain actionable business insights, thereby driving business transformation and building competitive advantages in the market. Existing research has explored the economic consequences of big data adoption by firms, accumulating extensive empirical evidence, particularly regarding firm productivity and performance (e.g). Unfortunately, aside from Eeckhout and Veldkamp, who examined the impact of data on markups from a macroeconomic perspective, research on big data applications and firm markups is nearly absent. Moreover, their study focuses on macroeconomic theoretical reasoning and lacks sufficient micro-level empirical support. Differing from the existing literature, this study provides an in-depth micro-level analysis of the relationship between big data application and firm markups. Similar to general-purpose technologies such as artificial intelligence, systematic measurement of big data applications at the enterprise level remains relatively scarce, which has become a primary challenge in accurately understanding the economic impact of big data. Existing studies primarily rely on core variable methods and questionnaire surveys to construct measurement indicators for enterprise big data applications. The former often uses metrics such as the number of data analysts as key proxy variables (e.g). However, big data applications require the synergistic coordination of multiple complementary inputs, making it difficult to capture their full framework using a single indicator alone. The latter mainly depends on structured survey tools (e.g), which still struggle to completely avoid subjective biases at the methodological level. Corporate annual reports, with their broad coverage of listed companies, high authority, and comprehensive content, serve as an ideal data source for constructing big data application metrics. However, these reports often contain substantial noise, making effective identification and processing of noise crucial for improving measurement accuracy. Large language models (LLMs), as cutting-edge tools in artificial intelligence, demonstrate significant potential in identifying textual noise and extracting unstructured information. For instance, Li et al. utilized LLMs to measure corporate culture and its economic consequences from listed companies' annual reports. Fang et al. extracted unstructured information, such as policy objectives, target industries, policy tools, and implementation mechanisms, from 3 million Chinese industrial policy texts. This study innovatively employs LLMs to deeply mine corporate annual report texts, aiming to construct a more objective and precise measurement system for enterprise big data applications. Based on this, this paper examines the impact of enterprise big data applications on price markups and their underlying mechanisms, using Chinese A-share listed companies from 2002 to 2023 as the research sample. Compared to existing studies, the contributions of this paper are mainly reflected in the following three aspects: First, from a micro-enterprise perspective, it integrates big data applications and enterprise price markups into a unified analytical framework, systematically investigating the impact of big data applications on price markups. This extends the research of Eeckhout and Veldkamp while broadening the theoretical boundaries of big data's influence in the microeconomic domain. Second, it adopts an innovative measurement method based on large language models to scientifically and accurately gauge the level of big data application at the enterprise level, significantly enhancing the objectivity and precision of indicator construction. This provides a reliable data foundation for empirical analysis and offers a referential measurement tool for subsequent identification of big data applications at the micro level. Third, building on the theoretical analytical framework of Antoniades, it constructs a heterogeneous firm variable markup model to theoretically elucidate how big data applications positively influence enterprise price markups through two key mechanisms: promoting product innovation and enhancing production efficiency. This further deepens the understanding of the mechanisms through which data empowers competitive advantages for enterprises. The structure of the remaining parts of this paper is as follows: Section "Analytical Framework and Theoretical Model" presents the analytical framework and theoretical model; Section "Research Design" outlines the research design; Section "Empirical Results and Analysis" displays the regression results; Section "Mechanism Analysis" conducts mechanism tests; Section "Heterogeneity Analysis" performs heterogeneity analysis; and the final section concludes with implications.