Public digital procurement and corporate total factor productivity

Public procurement refers to the process whereby governments acquire goods or services through tendering procedures using public funds [15]. Owing to its role in allocating public resources, it has long been regarded as a policy tool for increasing economic efficiency. Specialized procurement subtypes (e.g., public green procurement, public innovation procurement) demonstrate greater effectiveness than traditional contracts [16].

Driven by the digital economy wave, public procurement has progressively increased acquisitions of digital products/services, giving rise to PDP. Digital infrastructure has become a critical driver of socioeconomic development globally, especially in China [12]. Traditional supply-side policies exhibit limitations in addressing digital economy challenges [15], prompting governments to intensify digital infrastructure procurement. Against this backdrop, we define PDP as government tenders for digital infrastructure construction, encompassing digital services, products, and integrated applications.

Existing research examines the impacts of PDP on enterprises along three dimensions. First, organizational coordination studies show that digital infrastructure investment significantly reduces communication costs, enhances coordination capabilities, and mitigates information asymmetry [6]. Furthermore, by providing stable demand, PDP strengthens strategic orientation and resource allocation capabilities while alleviating financing constraints, thus driving digital transformation [2]. Second, technological innovation research indicates that public procurement promotes cost reduction and innovation through standardization/modularization [4,5]. Finally, financial security studies demonstrate that PDP creates stable market expectations that incentivize private R&D investment while reducing operating costs and optimizing supply chains [7]. Despite these insights, the literature overlooks the systematic impact of PDP on corporate productivity.

Corporate TFP determines the long-term growth potential of economies and sustainable development foundations. Internally, human capital, R&D investment, and digital technology adoption drive TFP improvement [14,17,18].

However, TFP improvement relies more crucially on external policy synergy than on internal capabilities [19]. Government interventions effectively reduce information asymmetry, lower technology diffusion barriers, and optimize funding chains, directly boosting TFP.

Specifically, PDP’s digitalization enhances transparency/efficiency and diversifies suppliers through streamlined registration. Doing so creates market access for small and medium-sized enterprises (SMEs), especially those in remote areas, enabling equitable competition with large firms via e-platforms [11]. Intensified competition incentivizes operational scaling. Additionally, PDP strengthens enterprise–supply chain–government connectivity, facilitating data flow. E-government platforms empower efficient communication (e.g., via blockchain), improving digital technology application and enabling data-driven decisions. Finally, PDP deepens public‒private symbiosis [15], providing financial support through “relational performance” (mutual value creation via institutional collaboration) and governmental endorsement that reduces financing costs [20].

Therefore, we can formulate the following hypothesis:

Hypothesis 1: PDP has a positive effect on corporate TFP.

Mechanism analysis

Alleviating financial constraint

Corporate competitive advantage stems from unique resources and capabilities, with capital as a crucial resource supporting enterprise development and profitability [23]. Enterprises often resort to external financing to alleviate cost pressures. China’s current financing system is predominantly bank-led. However, enterprises often face unstable early-stage returns and significant profit volatility [8], coupled with information asymmetry in the financing sector, leading to tight financing constraints [24].

According to asymmetric information theory, where the existence of unequal information between market participants can lead to inefficiencies and adverse selection, sellers often have more information about the quality of their products than do buyers, which can lead to potential market failure [25]. Completing technologically advanced and highly competitive PDP projects strengthens a company’s market position, conveys positive signals to external investors about the enterprise’s technological prowess and market potential, and overall operational health, which potentially attracts more external capital inflows and reduces credit costs associated with corporate financing, thereby easing financing constraints [6].

The alleviation of financing constraints provides enterprises with more stable and sustainable funding sources, enabling increased long-term investment in technological R&D, equipment upgrades, process optimization, and human capital. This not only increases resource allocation efficiency and production organization efficiency but also strengthens corporate flexibility and adaptability in response to market fluctuations. Consequently, it reinforces innovation capabilities and operational resilience, thereby effectively driving sustained growth in corporate TFP.

Thus, the current study proposes the following

Hypothesis 3: Financial constraint alleviation mediates the relationship between PDP and corporate TFP.

Enhancing information disclosure

Information asymmetry theory profoundly reveals that incomplete information distribution in capital markets creates significant informational gaps between participants (e.g., investors and firms, owners and managers), potentially triggering adverse selection and moral hazard. Consequently, corporate managers must proactively disclose information to mitigate such asymmetries [25].

PDP increases corporate information disclosure practices and alleviates multidimensional information asymmetries. In the regulatory dimension, government procurement contracts impose compulsory disclosure clauses requiring suppliers to publish core information on their technical capabilities, project execution standards, and digital security compliance, substantially increasing corporate transparency [26]. In the market competition dimension, to gain competitive advantage in PDP tenders, firms possess strong intrinsic incentives to voluntarily augment disclosures. They may release detailed financial reports, technical white papers, project implementation plans, or social responsibility reports to signal comprehensive capabilities in R&D, project management, and regulatory compliance to potential investors, financial institutions, and regulators [27], thereby strengthening their PDP market competitiveness.

Information disclosure drives corporate TFP growth through synergistic capital allocation optimization and increased governance efficacy. From a capital allocation perspective, high-quality disclosures correct investor misvaluation of digital assets and technological potential, reducing equity financing costs and increasing stock liquidity caused by information asymmetry [28]. This redirects capital toward high-efficiency projects, optimizing factor combination efficiency. From a corporate governance perspective, robust disclosure intensifies the oversight of managerial investment decisions, constraining self-serving overinvestment on the basis of informational advantages. It redirects resources to high-return areas such as innovation R&D and compels continuous optimization of internal resource allocation, unlocking the potential of organizational factor reallocation to sustainably increase TFP [29].

Thus, the current study proposes the following

Hypothesis 4: Information disclosure plays a pivotal mediating role in the positive relationship between PDP and corporate TFP.

The enhancing effect of industry–university–research collaboration

IUR cooperation, as a mechanism for promoting innovation and productivity, holds particular significance in the digital era. By pooling resources and expertise from industry, academia, and research institutions, IUR collaboration not only accelerates the R&D and commercialization of technological achievements but also increases corporate technological capabilities and resource allocation efficiency by facilitating knowledge spillovers, technology transfer, and collaborative innovation, thereby driving growth in corporate TFP [30]. In the context of digital public procurement, IUR cooperation can act as a powerful catalyst for increasing TFP. By involving universities and research institutions in public projects, governments not only gain access to a broader range of innovative solutions and expert knowledge but also help bridge the gap between research and commercialization, ensuring that cutting-edge technologies are rapidly and effectively applied in practice [31].

Moreover, IUR cooperation fosters close collaboration between buyers and suppliers, facilitating the transfer of knowledge and technology and further amplifying the positive impact of digital public procurement on TFP. Restricted by several institutional and structural constraints, the commercialization of university-based technology is inefficient in China [32]. This leads to a phenomenon where the output rate of innovation achievements from universities and research institutions is high but the conversion rate to practical applications is low. This issue stems from the lack of adequate policy guidance, incentives, and safeguards in the industrialization and marketization of technological achievements. Securing public procurement bids in the digital realm often necessitates advanced technological capabilities, prompting many enterprises to seek collaborations with academic and research institutions. These partnerships are driven by the high digital technology requirements typically associated with digital service or product-related orders. Compared with enterprises with similar characteristics, equipment suppliers tend to introduce more radical product innovations [33]. This finding underscores the potential for knowledge transfer and innovation stimulation through IUR collaboration. Collaborations with universities have been shown to be instrumental in ensuring the efficient completion of orders.

Therefore, the research hypotheses are as follows:

Hypothesis 5: IUR collaboration has a positive effect on corporate TFP.

Hypothesis 6: IUR collaboration positively enhances the relationship between PDP and corporate TFP.

A conceptual model of the study is proposed in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The mechanism of PDP and corporate TFP.

https://doi.org/10.1371/journal.pone.0330160.g001

Model specification

Baseline model and baseline model.

To identify the impact of PDP on corporate TFP, this paper constructs the following two-way fixed effects models:

(1)

Benchmark mediating effect model.

To analyze the mediating effects of three potential mediators—digital technology innovation, financial constraint alleviation, and information disclosure enhancement—on the relationship between PDP and corporate TFP, we utilize a three-step approach: first, we regress Eq. (1) to verify the influence of PDP on TFP; second, we regress Eq. (2) to test the influence of PDP on each mediator. Finally, Eq. (3) is regressed to further test the simultaneous influence of the independent and mediating variables on the dependent variables.

(2)(3)(4)

Benchmark moderating effect model.

To further estimate the moderating effect of IUR collaboration on the relationship between PDP and corporate TFP, this paper constructs the following two-way fixed effects models:

(5)

Where subscripts i and t represent the enterprise fixed effects and time fixed effects, respectively; is defined as a measure that reflects the efficiency of production and the technological level of an enterprise; is defined as a measure that reflects the scale of PDP contracts obtained by the enterprise in the previous period; and is a mediator variable, where each M represents DT, SA, and KV. is a set of control variables; represents enterprise fixed effects; represents time fixed effects; and is the random error term. In Eq. (1), the coefficient captures the direct effect of PDP on TFP. When is greater than zero, PDP promotes corporate TFP. When is less than zero, PDP inhibits corporate TFP. Eq. (2) tests the effect of PDP on each mediator , with coefficients indicating the strength of this relationship. The final Eq. (3), incorporating the mediators, allows us to assess the indirect effects of PDP on TFP through each mediator. If a significant reduction in the absolute value of compared with , along with a significant , it suggests mediation through . In Eq. (4), the coefficient captures the interaction effect between PDP and IUR on TFP.

Measurement

Explanatory variable.

Referring to the method of [4], we selected PDP as the explanatory variable and employed natural language processing methods to construct the PDP variable. LLM possess robust contextual semantic comprehension capabilities and generalized linguistic knowledge acquired from massive unannotated texts. Through transfer learning and fine-tuning, such models efficiently adapt to domain-specific classification tasks, significantly increasing recognition accuracy for feature-specific terms in complex unstructured texts [34]. Leveraging the scientific rigor and advanced performance of LLM in text representation and classification, we constructed feature lexicons and executed classification via the following procedure.

(i) We used Python to crawl all public procurement contracts from the China Government Procurement Network (2015–2023), extracting key information, including project names, purchasers, suppliers, subject matter descriptions, specifications/service requirements, contract values, and dates. (ii) We subsequently match the supplier names in the procurement orders with the names of A-share listed companies and their subsidiaries, successfully matching 101,373 orders and clearing the data for further processing. (iii) To develop a digital procurement lexicon, we adopted the approach of [4,35], utilizing the Jieba word segmentation tool to tokenize documents such as the “Statistical Classification of the Digital Economy and Its Core Industries”. Adverbs, conjunctions, and other nonsubstantive function words were removed from the segmentation results. The top 200 high-frequency substantive words were selected by descending frequency to form a base lexicon, which was then expanded through LLM-driven semantic association capabilities for synonym generalization and contextual broadening, ultimately yielding a lexicon of over 400 digital economy keywords. (iv) We then match this digital-related keyword library with public procurement text information. If a project contract name, main subject name, specifications, or service requirements contained these keywords, then it was identified as a digital procurement project. To further improve identification accuracy, following [36], we optimized the matching process via LLM fine-tuning: 1,500 data points were randomly selected for manual annotation (two researchers independently labeled records with a kappa coefficient≥0.8), and a pretrained LLM was fine-tuned using masked language modeling and next sentence prediction (the model architecture is shown in Fig 2). Dynamic learning rate scheduling optimized the classification layer parameters, achieving 90% accuracy on the test set. (v) We aggregate the data by firm and year, and the total amount of PDP contracts for each listed company is calculated annually and standardized according to the company’s total assets: PDP = ln(PDP contract amount+1)/ln(corporate total assets). Through this process, 526 observations of PDP data were ultimately obtained.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Pretrained LLM model architecture.

https://doi.org/10.1371/journal.pone.0330160.g002

Explained variable.

In this study, we employed corporate TFP as the dependent variable. TFP is a comprehensive indicator that measures corporate production efficiency given a set of input factors. It reflects the contributions of various factors, such as technological progress and management efficiency improvements, to corporate output [37] Among the methods for estimating TFP, the Olley–Pakes (OP) and Levinsohn–Petrin (LP) approaches are particularly prominent [38,39]. However, the OP method may result in a substantial loss of observations, especially when the sample contains instances of zero investment. To overcome this limitation, we chose the LP method to measure TFP in emerging economies. The fundamental idea of the LP method is to use intermediate inputs as a proxy for TFP. This approach allows researchers to select proxy variables on the basis of available data, thereby minimizing data loss. The baseline Cobb‒Douglas production function for calculating TFP is presented as follows:

Where Y represents output; L denotes labor input; K represents capital input; I denotes intermediate input, measured by cash paid for goods and services; is the error term; and subscripts i and t denote enterprise and time, respectively. In our empirical estimation, these variables are operationalized as follows: total output is represented by operating income, capital input by net fixed assets, labor input by the number of employees, and intermediate input by cash paid for goods and services received. To ensure the robustness of our results, we also employed the OP method to recalculate corporate TFP in the latter part of this study [40].

Mediator variables.

For digital technology, we utilize the digital-related keyword library constructed in the previous sections to identify digital technology innovation patents applied for by enterprises, aggregating the identified patents from the “enterprise-year” dimensions to construct a measurement indicator DT for digital technology innovation at the enterprise level. The specific formula used is DT = ln (total number of DT invention patent applications + 1).

For financial constraints, we calculate the level of financial constraints using firm size and firm age. The specific formula is as follows: SA = –0.737 × size + 0.043 × size² – 0.04 × age. This index generally takes negative values, with a larger absolute value indicating a greater degree of financial constraint.

For information disclosure, we employ the KV index to capture the extent and quality of corporate information disclosure, which reflects the degree of information asymmetry from the perspective of investors. The index is calculated via the following equation:

Where denotes the stock closing price on day t, is the trading volume on day t, and represents the average daily trading volume over the sample period.

Sampling

This study focuses on Chinese A-share listed companies from 2015 to 2023. A pivotal change in 2015 was the mandate to increase the transparency of public procurement information, which facilitated better disclosure of related contract details and enabled us to access data on PDP. Specifically, PDP data were obtained from the Chinese Government Procurement Network; corporate data were extracted from the CSMAR database, annual reports of listed companies, the China National Intellectual Property Administration website, and the China Statistical Yearbook. To ensure sample quality and the reliability of our results, the following sample selection criteria were applied: (i) companies that were subject to special treatment (ST) or other risk warnings (ST) during the study period were excluded, (ii) financial companies were removed, and (iii) observations with missing data for key variables were eliminated. After screening, a panel dataset comprising 26,306 enterprise-year observations was obtained. To mitigate the impact of extreme values, all continuous variables at the 1st and 99th percentiles were winsorized. To control for serial correlation at the enterprise level, cluster-robust standard errors were applied in all regression analyses.

Descriptive statistics

The descriptive statistics associated with the variables are provided in Table 1. For PDP, the mean value is 0.007, with a standard deviation of 0.052. This finding indicates that the overall scale of PDP is relatively small and that there is considerable variation in the extent to which different firms secure public digital orders. For TFP, the mean value is 6.743, with a standard deviation of 0.834. The sample captures sufficient variation to explore the relationship between PDP and corporate TFP. Furthermore, other variables also demonstrate notable variations across firms, attesting to the good distribution of the sample.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Statistical description of the variables.

https://doi.org/10.1371/journal.pone.0330160.t001

Results of the baseline model

Table 2 presents the baseline regression results of our study. Models 1–4 showcase the stepwise regression outcomes, progressively incorporating year fixed effects, firm fixed effects, and finally both year and firm fixed effects. The results of Model 4 reveal that the coefficient of the explanatory variable PDP is 0.128 and significant at the 1% level. The R² value suggests that our model accounts for a substantial portion of the factors influencing corporate TFP. A baseline coefficient of 0.128 implies that a one-unit increase in the standardized PDP index leads to a 0.128 unit increase in the logarithm of TFP. Given that the average TFP in our sample is 6.743 with a standard deviation of 0.834, this effect translates to an approximate calculated percentage, calculated as a 15.35% increase in terms of the standard deviation of TFP. This indicates an economically significant impact.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Baseline regression results.

https://doi.org/10.1371/journal.pone.0330160.t002

To address potential multicollinearity, we calculated the correlation coefficients among the independent variables and obtained the variance inflation factors (VIFs). The mean VIF across all variables is 1.06, with all individual VIF values remaining below 1.5. Consequently, multicollinearity does not pose a significant issue for our regression analysis. Overall, our results support H1, suggesting that PDP significantly encourages enterprises to increase their TFP.

Endogeneity tests

Instrumental variable approach.

To mitigate the potential reverse causality between PDP and corporate TFP, following the research of [4], we employed the city-level annual total number of digital procurement orders as an instrumental variable for enterprises awarded digital procurement. This instrument was chosen on the basis of two considerations. First, a higher initial proportion of PDP orders among listed companies in a city is likely to make it easier for enterprises in that city to obtain such orders, thereby satisfying the relevance requirement. Second, PDP orders received by a city are not directly influenced by the TFP levels of individual enterprises, thereby meeting the exclusion restrictions.

The first-stage regression results are presented in Model 1 of Table 3. The coefficient of IV is 0.024, which is positive and significant at the 1% level, indicating a strong positive correlation between PDP and IV. In the weak instrument test, the F statistic is 2461.56, which far exceeds the critical value, further confirming the validity of the IV. The second-stage regression results are shown in Model 2 of Table 3. The coefficient of PDP is 0.143, which is positive and significant at the 1% level, verifying the robustness of our baseline results.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Endogeneity test results.

https://doi.org/10.1371/journal.pone.0330160.t003

Robustness tests

Alternative measures of digital public procuring.

In this research, we employed two alternative approaches to measure our key explanatory variable: (i) PDP1, defined as the natural logarithm of the PDP contract amount plus one; and (ii) PDP2, a dummy variable assigned a value of 1 if the enterprise received PDP contracts and 0 otherwise. Models 1–2 in Table 4 report the regression results when these alternative measures are used. Upon substituting PDP with PDP1 and PDP2 in the regression analyses, the coefficients obtained were 0.006 (p < 0.01) and 0.042 (p < 0.01), respectively. Regardless of the measurement approach, the impact of PDP on corporate TFP remains significantly positive, which is consistent with our baseline results.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Robustness check results.

https://doi.org/10.1371/journal.pone.0330160.t004

The effect of digital technology innovation

Table 5 provides the results of the regression analysis testing the mediating role of DT in the relationship between PDP and TFP. We conduct hierarchical regression analysis in three steps. In column (1), the results reveal a significantly positive relationship between PDP and DT ( =0.521, p < 0.05). In addition, controlling for PDP, DT must have a significant effect on TFP, and the main effect of PDP should decrease substantially. In column (2), TFP has a significantly positive coefficient for DT ( =0.05, p < 0.01), and the magnitude of the coefficient for PDP ( =0.125, p < 0.01) decreases compared with the coefficient in the baseline regression ( =0.126, p < 0.01). We conduct a Sobel test to assess the magnitude of the mediating effect, resulting in a Z value of 11.82, which is significant at the 1% level. Therefore, H2 is supported.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Mechanistic analysis results.

https://doi.org/10.1371/journal.pone.0330160.t005

Effects of financial constraint alleviation

Table 5 provides the results of the regression analysis testing the mediating role of SA in the relationship between PDP and TFP. In column (3), the results reveal an inverse relationship between PDP =−0.024, p < 0.01). In column (4), TFP has a significantly negative coefficient for SA =−0.303, p < 0.01), and the magnitude of the direct effect of PDP =0.120, p < 0.01) decreases compared with the coefficient in the baseline regression. This suggests that SA partially mediates the relationship between PDP and TFP. We conduct a Sobel test to assess the magnitude of the mediating effect, resulting in a Z value of 2.702, which is significant at the 1% level and affirms the presence of the mediating effect of SA in the relationship between PDP and TFP. Therefore, H3 is supported.

Effects of corporate disclosure enhancement

Table 5 provides the results of the regression analysis testing the mediating role of KV in the relationship between PDP and TFP. In column (5), the results reveal an inverse relationship between PDP and KV ( =0.073, p < 0.01). In column (6), TFP has a significantly negative coefficient for KV ( =0.202, p < 0.01), and the magnitude of the direct effect of PDP on TFP ( =0.112, p < 0.01) decreases compared with the coefficient in the baseline regression. These results indicate that KV partially mediates the relationship between PDP and TFP. We conduct a Sobel test to assess the magnitude of the mediating effect, resulting in a Z value of 5.122 that is significant at the 1% level and confirms the presence of the mediating effect of KV on the relationship between PDP and TFP. Therefore, H4 is supported.

The reinforcing role of industry–university–research collaboration in increasing the impact of public digital procurement on corporate total factor productivity

Table 6 presents the regression results of IUR collaboration on corporate TFP. As evident from Table 6, the regression coefficients of IUR collaboration on corporate TFP remain consistently positive and significant. The results demonstrated in Model 4 reveal that after incorporating a series of control variables and controlling for both year and firm fixed effects, the coefficient of IUR is 0.048 and significant at the 5% level. This finding validates Hypothesis H5, suggesting that IUR collaboration significantly encourages enterprises to increase their TFP.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. The impact of IUR collaboration on corporate TFP.

https://doi.org/10.1371/journal.pone.0330160.t006

Table 7 presents the enhancing effects of IUR collaboration on the relationship between PDP and TFP. As evident from Table 7, the regression coefficients of PDP, IUR, and their interaction term PDPIUR remain consistently positive and significant. The results demonstrated in Model 4 reveal that after incorporating a series of control variables and controlling for both year and firm fixed effects, the coefficient of the interaction term between PDPIUR is 1.385 and significant at the 1% level. To more clearly demonstrate the enhancing effect of IUR collaboration, this paper presents an enhancing effect diagram (see Fig 3). Fig 3 shows that for enterprises with existing IUR collaboration, the positive impact of PDP on corporate TFP is further strengthened. Moreover, this reinforcing effect is even more pronounced at higher levels of PDP. These findings validate Hypothesis H6.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. The enhancing effect of IUR collaboration.

https://doi.org/10.1371/journal.pone.0330160.t007

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. The enhancing effect of IUR collaboration.

https://doi.org/10.1371/journal.pone.0330160.g003

Theoretical contribution

From the perspectives of demand-induced innovation theory and asymmetric information theory, this study employs a comprehensive panel dataset of Chinese listed firms (2015–2023) to investigate the relationship between PDP and corporate TFP. Our findings provide robust evidence that PDP implementation significantly increases corporate TFP, a conclusion that is validated through rigorous endogeneity and robustness tests. This study delves into the underlying mechanisms, pinpointing three pivotal pathways: DT innovation, financing constraint alleviation, and information disclosure. Furthermore, we reveal the positive moderating effect of IUR collaboration on this relationship. These findings enrich existing theoretical frameworks by offering nuanced insights from the perspectives of technological upgrading and resource allocation. Specifically:

1. (i) While the literature has comprehensively examined the impact of digital government policies on corporate development [13], we adopt a public procurement lens to reveal how PDP functions as a policy instrument under such policies, providing a complementary perspective. We innovatively construct a PDP index by integrating LLM techniques with data from Chinese listed firms (2015–2023) and textual information from the China Government Procurement Network, forming panel data for empirical analysis. This approach effectively mitigates the measurement errors inherent in traditional policy metrics [44], offering microlevel evidence for evaluating the economic effects of PDP.
2. (ii) We advance the theoretical frontier of TFP determinants. Prior studies predominantly emphasized supply-side drivers of TFP [10,14], with limited attention paid to demand-side factors. We shift the focus to public procurement—a classical demand‒pull instrument—and rigorously investigate its relationship with TFP. Our findings demonstrate that PDP significantly increases TFP, providing new perspectives and complementary insights into the factors influencing TFP. Moreover, this finding lends support to the theoretical discovery of [45] that, from a supplier sustainability perspective, public procurement can significantly enhance suppliers’ sustainable development capacity and increase corporate TFP.
3. (iii) We advance the microeconomic analysis of PDP. Although prior research has demonstrated positive effects of overall public procurement on corporate productivity [12], it has failed to capture the digital attributes of procurement contracts. In contrast, we innovatively isolate and focus on digital procurement orders, constructing a dedicated PDP index that overcomes the neglect of digital characteristics in traditional studies. This perspective significantly extends the research scope of [46,47] with regard to innovation-oriented and green procurement. Our empirical results indicate that PDP exerts consistently significant positive effects on corporate TFP, in contrast with the U-shaped relationship reported by [12]. We argue that by specifically focusing on high-tech digital procurement (rather than overall government procurement), PDP continuously improves firm efficiency without the initial negative effects potentially present in traditional procurement policies. These findings contribute to a more comprehensive and balanced understanding of the heterogeneous impacts of different types of government procurement.
4. (iv) Furthermore, this study uncovers the intrinsic mechanisms (“black box”) through which PDP enhances corporate TFP. While existing research widely recognizes the significant potential of PDP as an emerging policy tool, its internal mechanisms remain insufficiently explained [4]. We identify and empirically test three critical mediating mechanisms: DT innovation, financing constraint alleviation, and information disclosure. Among these, DT innovation and financing constraint alleviation align with the findings of [2], further validating the robustness and universality of these mechanisms in the PDP context. Moreover, this study adds a novel mechanism, information disclosure, systematically clarifying the joint operation of multiple mechanisms. In doing so, this study effectively extends the understanding of PDP mechanisms in the literature.

Management implications

From a practical perspective, our findings offer significant insights for government digital procurement decisions and corporate development planning.

First, considering that the scale of government digital procurement in China still has substantial room for growth and must address the practical needs of enterprise development, the government should focus on expanding procurement in the digital domain. During implementation, the accurate identification of procurement attributes and enterprise characteristics is essential for enhancing policy effectiveness through differentiated strategies. Simultaneously, accelerating the development of digital government and the deployment of computing infrastructure will promote the efficient flow of key data elements. This will not only significantly improve the TFP of enterprises but also contribute to the inclusive distribution of digital dividends.

Second, the government should construct a policy portfolio toolkit to drive collaborative innovation between digital procurement and enterprise transformation. By designing differentiated policies to support enterprises’ digital transformation—combining direct incentives such as tax reductions and R&D subsidies with indirect support measures such as financing assistance—a synergistic policy force can be formed. This leverages procurement demand to steer enterprises, thus enhancing policy efficacy and fostering corporate growth.

Third, IUR collaboration plays a vital role in the R&D and application processes of government digital procurement. The government must lead in establishing collaborative platforms that integrate research resources, enterprise scenarios, and government procurement needs. These platforms should actively advance PDP projects to break through technological bottlenecks, accelerate the achievement transformation process, and concurrently focus on innovating cooperation mechanisms while optimizing the collaborative environment. This will provide more efficient and convenient resource matching and achievement transformation services for participating enterprises and research institutions.

Limitations and future studies

Notably, our study has several limitations. First, the development of the digital economy has a pronounced long-term and dynamic nature [48]. The data period used in this study (2015–2023) may not fully capture the long-term impact of PDP on TFP. Therefore, future research could further extend the time span, for example, by tracking firm-level microdata (including numerous SMEs) over several decades, to comprehensively capture the long-term dynamic impact of PDP on corporate productivity. On the other hand, although this study innovatively constructs a PDP index via LLM technology, this approach may still suffer from an insufficient capture of data dimensions and policy implementation details. Key factors that influence policy effectiveness—such as the actual implementation efficiency of government procurement contracts and the frequency of firm‒government interactions—remain inadequately examined [49,50]. We recommend that subsequent research collect further interview and survey data from local governments or enterprises, gather and analyze government procurement contract audit documents, and integrate real-time transaction data from government e-procurement platforms. These approaches more comprehensively capture the implementation heterogeneity and granular details of PDP policies, thus increasing the precision and completeness of policy effect evaluations.