The Impact of Digital Supply Chain Management on Enterprise Total Factor Productivity: Evidence from a Quasi-Natural Experiment in China

Abstract

Digital supply chain management (DSCM) has emerged as a critical driver of enterprise performance in the modern economy, yet empirical evidence on its causal impact on productivity remains limited. This study examines how DSCM adoption affects total factor productivity (TFP) by leveraging China’s supply chain innovation pilot program as a quasi-natural experiment. Using a difference-in-differences approach with propensity score matching, the analysis employs a comprehensive dataset of 2843 Chinese A-share listed companies from 2013 to 2022; the analysis reveals that DSCM adoption leads to an average TFP increase of 14.1%. This positive effect strengthens over time, demonstrating a clear dynamic of organizational learning. Mediation analysis indicates that this productivity enhancement operates through two primary channels: innovation capability enhancement (accounting for approximately 35% of the total effect) and cost efficiency improvement (21%). Crucially, heterogeneity analysis reveals that the positive effects of DSCM are significantly more pronounced in supply-chain-intensive industries, such as manufacturing, and for firms with higher R&D intensity. The findings provide robust causal evidence on the productivity effects of DSCM, offering valuable insights into its underlying mechanisms and key boundary conditions for both enterprise strategy and digital transformation policy.

1. Introduction

The pursuit of enhanced productivity is a cornerstone of sustainable economic development, enabling firms to generate value while optimizing resource consumption. Amid mounting global pressures for environmental accountability and social responsibility, improving total factor productivity (TFP) is no longer just a competitive strategy but a prerequisite for long-term corporate sustainability [1]. Higher TFP allows firms to free up capital for green investments, enhance supply chain resilience, and contribute to stable economic growth. The digital transformation of supply chain management (DSCM) has emerged as a powerful lever for achieving these productivity gains. However, while the operational benefits of specific digital technologies are widely acknowledged [2], a comprehensive understanding of how an integrated DSCM strategy causally impacts firm-level TFP remains elusive. This study addresses this critical gap by examining China’s supply chain innovation and application pilot program as a quasi-natural experiment [3], thereby providing novel causal evidence on the productivity effects of DSCM.

The existing literature has extensively documented the potential of individual digital technologies, such as the blockchain [4], Internet of Things (IoT) [5], and big data analytics [6], to streamline specific supply chain functions. Research has often focused on conceptual frameworks or case studies, highlighting improvements in operational efficiency like reduced lead times or optimized inventory levels. However, two significant gaps persist. Firstly, most studies adopt a technology-centric view, examining technologies in isolation rather than evaluating the synergistic impact of a holistic, firm-wide DSCM strategy. Secondly, much of the empirical work is correlational, making it difficult to disentangle the true causal effect of DSCM from other confounding factors, such as firm size or pre-existing technological capabilities [7]. This study diverges from previous approaches by focusing on a comprehensive DSCM initiative and employing a rigorous research design to isolate its causal impact on TFP.

The context of China offers a compelling setting for this investigation. As Chinese enterprises face intense domestic and international competition, the imperative to boost efficiency while managing costs has accelerated the adoption of digital solutions [8]. Government-led initiatives, such as the pilot program studied here, provide a unique opportunity to observe the effects of large-scale, policy-driven DSCM implementation. This research, therefore, seeks to answer a central question: does the adoption of a comprehensive DSCM strategy lead to a significant increase in enterprise TFP? Furthermore, we explore the underlying mechanisms driving this relationship, offering crucial insights for both managers and policymakers in emerging economies navigating their digital transformation journeys.

To answer these questions, our empirical strategy leverages China’s supply chain innovation pilot program as a quasi-natural experiment. This approach allows us to isolate the causal impact of DSCM on TFP by comparing the performance of “pilot” firms with a carefully matched control group before and after the policy’s implementation, thus mitigating endogeneity concerns that have constrained previous studies. The analysis employs a comprehensive dataset of Chinese A-share listed companies from 2013 to 2022, utilizing a difference-in-differences (DID) methodology to ensure the reliability and validity of our findings.

This study makes several distinctive contributions. Firstly, it provides robust causal evidence on the link between a comprehensive DSCM strategy and TFP growth, moving beyond the study of isolated technologies. Secondly, by identifying dual mediating pathways—innovation enhancement and cost reduction—it unpacks the “black box” of how these productivity gains are achieved. Thirdly, through a novel heterogeneity analysis, it reveals the boundary conditions of DSCM’s effectiveness, showing that its benefits are most pronounced in supply-chain-dependent industries. These findings offer actionable, evidence-based insights for enterprises and policymakers, particularly in emerging economies where guidance on maximizing the returns from digital transformation remains scarce [9].

The structure of this paper is organized as follows: Section 2 provides a comprehensive review of the relevant literature on digital supply chain management and total factor productivity. Section 3 presents the theoretical framework and develops testable hypotheses. Section 4 describes the research methodology, data sources, and analytical approach. Section 5 reports the empirical results and conducts various robustness checks. Section 6 discusses the findings, their implications, and this study’s limitations. Section 7 concludes with policy recommendations and identifies promising directions for future research.

2. Literature Review

2.1. Digital Supply Chain Management and Firm Productivity

The theoretical understanding of digital supply chain management (DSCM) has undergone a profound evolution over the past two decades. Early scholarly work centered on internal tools like enterprise resource planning (ERP) systems, which aimed to enhance operational visibility and coordination across departmental silos [10]. With the advent of technologies such as the Internet of Things (IoT) and blockchain, the academic focus has broadened from the individual firm to the wider external ecosystem [11]. This evolution has given rise to the concept of the “Digital Supply Network” (DSN), which emphasizes the non-linear, dynamic, and highly interconnected nature of modern supply chains. Within the DSN framework, value is not merely transferred linearly but is co-created among a multi-tiered network of partners [12].

Against this backdrop, a substantial body of empirical research has sought to quantify the practical performance impacts of these digital initiatives. The vast majority of studies consistently report a significant positive association between various forms of digitalization and key firm performance indicators, including the core variable of this study, total factor productivity (TFP) [13]. The general consensus is that digital tools systematically enhance operational efficiency by reducing information asymmetries between supply chain partners, optimizing inventory levels, and shortening production and delivery cycles. However, despite these insightful findings, two critical limitations curtail the reliability of the conclusions drawn from this body of literature.

Firstly, many studies adopt a “technology-centric” perspective, focusing on the impact of isolated technologies (e.g., RFID or big data analytics) rather than the synergistic effects of a comprehensive, integrated DSCM strategy. This “seeing the trees but not the forest” approach likely underestimates the network effects and complementary gains that arise when a firm undertakes a systemic digital transformation. Secondly, and more critically, the existing research is overwhelmingly based on correlational analyses of observational data and thus suffers from significant endogeneity concerns. Specifically, more productive firms often possess greater financial resources, superior managerial talent, and keener strategic foresight, which makes them more inclined to invest in and implement DSCM in the first place. The self-selection bias, coupled with potential reverse causality (i.e., high productivity leads to digitalization, not the other way around), makes it exceedingly difficult to infer a true causal relationship from correlational findings alone.

2.2. Mediating Mechanisms and Moderating Factors

To move beyond a simple “input-output” analysis, a second stream of literature has focused on uncovering the intrinsic mechanisms through which DSCM translates into performance advantages. From the perspective of Transaction Cost Economics (TCE), a core transmission channel is cost reduction. Digital platforms automate workflows and enhance transparency, which significantly reduces the friction and transaction costs associated with searching for, coordinating with, and monitoring external partners [14]. In practice, this often manifests as lower inventory holding costs and more efficient procurement processes. Another key channel, understood through the lens of the Knowledge-Based View (KBV), is innovation enhancement. By facilitating seamless data flow and collaboration with suppliers and customers, DSCM can significantly accelerate both product innovation (e.g., shorter new product development cycles) and process innovation (e.g., implementing predictive maintenance and optimizing production techniques) [15].

Furthermore, the literature widely acknowledges that the impact of DSCM is not uniform but is contingent upon various moderating factors. A firm’s dynamic capabilities—its capacity to sense, seize, and reconfigure resources to adapt to environmental changes—play a pivotal role in determining its ability to profit from digital investments [16]. These capabilities are often closely tied to a firm’s internal characteristics, such as its prior R&D intensity, the digital skill level of its workforce, and an organizational culture that encourages change. Concurrently, external environmental factors, such as the intensity of market competition and macroeconomic volatility, also moderate the returns on DSCM investments, with some research suggesting its benefits are more pronounced in highly uncertain market environments [17]. Despite these valuable insights, a systematic empirical test of the core mediating channels within a unified framework, especially in the context of a large-scale, policy-driven DSCM initiative, remains a notable gap in the literature.

2.3. Research Gaps and Contributions

Our comprehensive review of the existing literature, while acknowledging its rich contributions, reveals three critical and interconnected gaps that this study is specifically designed to address.

Firstly, there is a scarcity of robust causal evidence regarding the true effect of a comprehensive, integrated DSCM strategy on TFP. As previously discussed, the pervasive endogeneity issues mean that the policy implications drawn from existing research are tentative at best, as decision makers cannot be certain that an observed positive correlation is a genuine return on investment rather than a reflection of pre-existing firm characteristics. This study addresses this crucial gap by leveraging a quasi-natural experiment created by a national-level pilot policy, which, combined with a rigorous difference-in-differences (DID) methodology, allows for a “cleaner” and more credible identification of the causal effect than has typically been possible.

Secondly, while various potential mediating mechanisms have been proposed, the core pathways through which a holistic DSCM strategy enhances productivity remain unclear. Existing studies often examine a single channel in isolation, lacking a systematic comparison and validation within a unified framework. This study contributes by simultaneously testing two core mediating pathways, “innovation enhancement” and “cost reduction”, within a single model, thereby seeking to open the “black box” between DSCM and TFP and provide a clearer picture of the entire value creation process.

Thirdly, the understanding of the boundary conditions that determine the success of DSCM remains incomplete. Prior research has focused more on firm-level moderators, with less attention paid to macro-level industry characteristics. This study introduces a novel heterogeneity analysis based on industry-level supply chain dependence to explore, in which industrial contexts the benefits of DSCM can be maximized. This not only deepens our understanding of the mechanisms of DSCM but also provides more targeted guidance for the decisions of firms and governments.

In summary, by providing rigorous causal evidence, untangling key transmission mechanisms, and identifying important boundary conditions, this study offers a more complete, nuanced, and actionable analytical framework for understanding how DSCM drives firm productivity in the context of a major emerging economy. Firstly, much of the research focuses on the impact of specific, isolated technologies rather than the synergistic effects of a comprehensive, integrated DSCM strategy. While valuable, this technology-centric approach may underestimate the transformative potential of a holistic digital overhaul. Secondly, many studies rely on correlational data, making it difficult to establish a clear causal relationship due to endogeneity concerns, such as the possibility that more productive firms are simply more likely to adopt advanced digital solutions.

Recent studies have begun to unpack the mechanisms linking DSCM to productivity, identifying mediating factors such as enhanced innovation capabilities, optimized supplier relationships [18], and improved supply chain stability. Heterogeneity in these effects has also been noted, with firm size, industry, and technological readiness acting as important moderators. Despite this progress, a systematic investigation of the causal pathways linking a comprehensive DSCM strategy to TFP, particularly within the institutional context of emerging economies, remains a significant gap.

3. Theoretical Framework and Hypotheses

3.1. The Direct Effect of DSCM on Total Factor Productivity

This study develops a multi-theoretical framework to comprehensively understand the relationship between digital supply chain management (DSCM) and total factor productivity (TFP). By integrating insights from the Resource-Based View (RBV), Transaction Cost Economics (TCE), and the Knowledge-Based View (KBV), alongside concepts from organizational learning and absorptive capacity, we build a robust foundation for our hypotheses. This approach allows us to examine not only the direct impact of DSCM but also its dynamic nature, its underlying mechanisms, and the conditions under which its effects are most pronounced [19]. We posit that DSCM has a direct, positive impact on TFP through two primary theoretical lenses.

Firstly, from the Resource-Based View (RBV), DSCM is not merely a collection of technologies but a strategic capability that is valuable, rare, and difficult to imitate [20]. By integrating real-time data across the supply network, firms develop a unique informational advantage, enabling superior decision making, more efficient resource allocation, and enhanced operational coordination [21]. This “digital capability” becomes an intangible asset that allows firms to outperform competitors who lack such integrated systems. As firms leverage this capability to optimize production, logistics, and inventory, their overall efficiency increases, leading to a direct improvement in TFP.

Secondly, Transaction Cost Economics (TCE) provides a complementary explanation. Supply chains are inherently networks of transactions between independent firms, which incur significant coordination costs (e.g., search, contracting, and monitoring costs) [22]. DSCM systematically lowers these frictions [23]. Digital platforms reduce search costs for partners, smart contracts can automate and secure agreements, and real-time tracking systems minimize monitoring costs. By reducing the marginal cost of coordinating with external suppliers and customers, DSCM allows firms to operate more efficiently within their supply network, directly contributing to higher productivity.

The convergence of these two perspectives—DSCM as a strategic resource and as a transaction cost-reducing mechanism—provides a strong theoretical basis for a positive main effect. Therefore, the study proposes our first hypothesis:

H₁: 

The implementation of digital supply chain management is positively associated with a significant increase in a firm’s total factor productivity.

3.2. The Dynamic Effect of DSCM Implementation

The benefits of a complex strategic initiative like DSCM are unlikely to be fully realized instantaneously. Drawing on organizational learning theory and the dynamic capabilities view, we argue that the impact of DSCM on TFP strengthens over time [24]. The initial implementation phase often involves significant costs, process re-engineering, and a steep learning curve for employees. However, as an organization gains experience, it develops new routines and “dynamic capabilities” that allow it to more effectively leverage the technology. Complementary assets, such as data analytics skills and collaborative relationships with partners, are built over time, leading to accelerating returns. This suggests a dynamic, non-linear relationship where the productivity gains from DSCM become more pronounced with implementation duration [25]. Thus, this study hypothesizes the following:

H₂: 

The positive effect of digital supply chain management on TFP will become stronger over time as firms accumulate experience and develop complementary organizational capabilities.

3.3. The Heterogeneous Effects of DSCM

The extent to which firms can benefit from DSCM is likely contingent on their pre-existing characteristics. The theory of absorptive capacity suggests that a firm’s ability to recognize, assimilate, and apply external knowledge is crucial for leveraging new technologies [26]. In the context of DSCM, firms with a stronger foundation of existing knowledge and capabilities are better positioned to absorb and capitalize on their potential. We argue that a firm’s R&D intensity and its existing digital infrastructure are key indicators of its absorptive capacity. Firms with higher R&D intensity possess superior technical knowledge to understand and adapt DSCM, while those with a solid digital foundation can integrate new systems more seamlessly. Therefore, this study proposes the following:

H₃: 

The positive effect of digital supply chain management on TFP will be more pronounced for firms with higher R&D intensity and a stronger pre-existing digital infrastructure.

3.4. The Mediating Mechanisms of DSCM

To understand how DSCM enhances TFP, we identify two critical mediating pathways: innovation enhancement and cost reduction.

Firstly, from the Knowledge-Based View (KBV), DSCM acts as a catalyst for innovation enhancement [27]. By breaking down information silos between a firm and its partners, digital platforms create a vibrant ecosystem for knowledge sharing and co-creation. Real-time data exchange and collaborative tools facilitate open innovation, allowing firms to more quickly sense market needs, absorb external ideas, and accelerate their product and process innovation cycles. This enhanced innovative capacity is a direct driver of long-term TFP growth.

Secondly, as previously discussed under TCE, DSCM leads to significant cost reduction. This goes beyond just lowering inter-firm transaction costs. From an operations management perspective, real-time visibility and data analytics enable substantial improvements in internal operational efficiency. Firms can achieve more accurate demand forecasting, leading to optimized inventory levels (reducing carrying costs) and more efficient logistics planning (reducing transportation costs). These operational efficiencies directly lower the input requirements for a given level of output, mechanically increasing TFP.

Based on these theoretical arguments, we developed our final two hypotheses:

H₄: 

The positive effect of DSCM on TFP is partially mediated by the enhancement of the firm’s innovation capability.

H₅: 

The positive effect of DSCM on TFP is partially mediated by the reduction in the firm’s operational costs.

4. Data and Methods

4.1. Research Design, Data, and Sample

To identify the causal effect of digital supply chain management (DSCM) on total factor productivity (TFP), this study employs a difference-in-differences (DID) strategy combined with propensity score matching (PSM). This quasi-experimental approach leverages China’s “Supply Chain Innovation and Application Pilot Program” as an exogenous policy shock.

Our dataset comprises Chinese A-share listed companies from 2013 to 2022, sourced from the China Stock Market & Accounting Research (CSMAR) database [28]. After excluding financial firms, special treatment (ST) companies, and firms with missing data for key variables, our final unbalanced panel consists of 2843 unique firms and 18,927 firm-year observations.

The treatment group includes firms officially designated as participants in the pilot program between 2017 and 2020. The control group consists of non-participant firms. A critical aspect of our research design is the justification for using “pilot firm status” as a valid proxy for the implementation of a supply-chain-oriented digital strategy. The official title and objectives of the “Supply Chain Innovation and Application Pilot Program” inherently required applicants to focus on inter-firm collaboration. To be selected, firms had to submit detailed plans demonstrating how they would use digital technologies to enhance coordination, information sharing, and efficiency across their supply chain network, not just within their own organizational boundaries [29]. Therefore, selection into this program serves as a strong indicator that a firm is undertaking a digital transformation with a clear supply chain focus, which directly addresses the core construct of our study.

To mitigate potential selection bias arising from the application process, we use PSM with a 1:1 nearest-neighbor matching and a caliper of 0.05 to construct a comparable control group. The matching is based on a rich set of pre-treatment firm characteristics, including size, profitability, leverage, R&D intensity, and industry classification, ensuring that the treatment and control groups followed a parallel trend before the policy intervention.

4.2. Variable Measurement

Dependent Variable: Total factor productivity (TFP). Following the established literature, we measure TFP using a semi-parametric approach to address endogeneity in production function estimation. Specifically, we employ the Levinsohn–Petrin (LP) method, which uses intermediate inputs (e.g., raw materials) as a proxy for unobserved productivity shocks that are correlated with firms’ input choices. We estimate a Cobb–Douglas production function where the output is value-added, and inputs are capital and labor. The resulting TFP measure, lnTFP, is used as our dependent variable [30].

Core Explanatory Variable: The causal effect of DSCM is captured by the interaction term $trea{t}_i\times pos{t}_t$. $trea{t}_i$ is a dummy variable equal to 1 if firm $i$ is in the treatment group and 0 otherwise. $pos{t}_t$ is a dummy variable equal to 1 for the years during and after the firm’s designation as a pilot firm and 0 for the pre-treatment years.

Mediating Variables: Based on our theoretical framework, we measure two key mediators.

Innovation Capability (Innovation): We use two proxies: (1) the natural logarithm of one plus the number of patent applications filed by the firm in a given year, to capture innovation output quantity; and (2) R&D intensity, measured as R&D expenditure divided by total sales, to capture innovation input.

Cost Efficiency (Cost): Our primary measure is the operating profit margin (operating profit divided by sales). As supplementary measures, we also examine the operating cost ratio (operating costs to sales) and inventory turnover.

Control Variables: We include a standard set of firm-level controls known to affect TFP: firm size (natural logarithm of total assets), leverage (total liabilities to total assets), firm age (natural logarithm of years since establishment), ownership structure (a dummy for state-owned enterprises), and capital intensity (net fixed assets to total assets).

4.3. Empirical Models

Baseline DID Model: To estimate the average treatment effect on the treated (ATT), we use the following two-way fixed effects DID model:

$$ln{TFP}_{it}={\beta }_0+{\beta }_1\left(trea{t}_i\times pos{t}_t\right)+{\gamma }^{\prime }{X}_{it}+{\mu }_i+{\delta }_t+{\epsilon }_{it}$$

(1)

where $ln{TFP}_{it}$ is the TFP of firm $i$ in year $t$. $X_{it}$ is the vector of control variables. ${\mu }_i$ represents firm fixed effects, controlling for all time-invariant firm characteristics, and ${\delta }_t$ represents year fixed effects, controlling for macroeconomic shocks and common time trends. ${\epsilon }_{it}$ is the error term. Our primary interest is in the coefficient ${\beta }_1$.

Mediation Analysis Models: To test hypotheses H₄ and H₅, we employ the widely used three-step causal mediation analysis framework:

$${Mediator}_{it}={\alpha }_0+{\alpha }_1\left(trea{t}_i\times pos{t}_t\right)+{\gamma }^{\prime }{X}_{it}+{\mu }_i+{\delta }_t+{\epsilon }_{it}$$

(2)

$$ln{TFP}_{it}={\theta }_0+{\theta }_1\left(trea{t}_i\times pos{t}_t\right)+{\theta }_2{Mediator}_{it}+{\gamma }^{\prime }{X}_{it}+{\mu }_i+{\delta }_t+{\epsilon }_{it}$$

(3)

In this framework, Equation (1) tests the total effect. Equation (2) tests the effect of the treatment on the mediator. Equation (3) tests the direct effect of the treatment (${\theta }_1$) and the effect of the mediator (${\theta }_2$) on the outcome variable.

$${{ln{TFP}_{it}=\omega }_0+\omega }_1\left(trea{t}_i\times pos{t}_t\right)+{\omega }_2\left(trea{t}_i\times pos{t}_t\times M_i\right)+{\omega }_3{M}_i+{\gamma }^{\prime }{X}_{it}+{\mu }_i+{\delta }_t+{\epsilon }_{it}$$

(4)

Here, $M_i$ is a time-invariant moderating variable (e.g., a dummy for high R&D intensity or for being in a supply-chain-dependent industry). The coefficient of interest is ${\omega }_2$, which captures the differential effect of DSCM based on the moderator [31].

5. Results

5.1. Descriptive Statistics and Parallel Trends Test

5.1.1. Descriptive Statistics

Before proceeding to the regression analysis, we present the summary statistics for the key variables in

Table 1. Descriptive Statistics.

Variable	Obs	Mean	Std. Dev.	Min	Max
TFP (lnTFP)	18,927	9.875	1.234	6.432	13.122
Treat × Post	18,927	0.056	0.230	0	1
Innovation (lnPatents)	18,927	1.892	1.543	0	7.891
Cost Efficiency (OPM)	18,927	0.088	0.121	−0.543	0.654
Firm Size (lnAssets)	18,927	22.543	1.456	19.876	26.432
Leverage	18,927	0.453	0.211	0.054	0.987
Firm Age (lnAge)	18,927	2.876	0.654	1.098	4.123
SOE (Dummy)	18,927	0.345	0.475	0	1
Capital Intensity	18,927	0.312	0.187	0.021	0.876

. The sample covers 18,927 firm-year observations from 2843 unique firms. The data exhibit substantial cross-sectional and time-series variation in our key variables, including TFP, innovation output, and cost efficiency, providing a rich dataset for identifying the effects of DSCM. A preliminary t-test (unreported) shows that firms in the treatment group were, on average, larger and more profitable than those in the control group before the policy, highlighting the necessity of our DID and PSM-DID approaches to control for such pre-existing differences.

5.1.2. Parallel Trends Test

The cornerstone of a credible DID analysis is the parallel trends assumption, which requires that the treatment and control groups would have followed similar trends in the outcome variable had the treatment not occurred. We formally test this assumption by estimating an event study model. The results, visualized in Figure 1, plot the estimated coefficients and their 95% confidence intervals for the years surrounding the policy implementation.

As Figure 1 clearly illustrates, the coefficients for all pre-treatment periods (Pre_3, Pre_2) are statistically insignificant and hover tightly around the zero line. This indicates that prior to the policy intervention, there was no systematic, pre-existing difference in TFP trends between the firms that would later become pilot firms and those that would not. This validation of the parallel trends assumption is crucial, as it provides a solid foundation for attributing any post-intervention divergence in TFP to the causal effect of the DSCM policy itself, rather than to a continuation of historical trends.

5.2. The Main Effect of DSCM on TFP

We begin by estimating the average treatment effect of DSCM on TFP using the two-way fixed effects DID model specified in Equation (1).

Table 2. Baseline DID Regression Results on TFP.

	(1)
Variables	lnTFP
Treat × Post	0.141 ***
	(0.032)
Firm Size	0.057 ***
	(0.015)
Leverage	−0.089 **
	(0.04)
Firm Age	0.021
	(0.018)
SOE	−0.045 *
	(0.025)
Capital Intensity	−0.155 ***
	(0.05)
Constant	2.454 ***
	(0.112)
Observations	18,927
R2	0.677
Firm Fixed Effects	Yes
Year Fixed Effects	Yes

Note: Standard errors clustered at the firm level are in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

presents the baseline regression results.

The coefficient of our core variable, Treat × Post, is positive and highly statistically significant (β = 0.141, p < 0.01). This result indicates that, on average, the implementation of DSCM led to a 14.1% increase in TFP for the pilot firms relative to the control group. This effect is not only statistically significant but also economically meaningful, representing a substantial improvement in organizational efficiency that compares favorably with the impacts of other digital transformation initiatives documented in prior research. This finding provides strong empirical support for Hypothesis 1, confirming that DSCM generates measurable productivity-enhancing effects, consistent with our theoretical arguments based on the Resource-Based View and Transaction Cost Economics.

The control variables also exhibit theoretically consistent relationships with TFP. Firm size is positive and significant, aligning with theories of economies of scale. Leverage shows a negative coefficient, suggesting that financial constraints may hinder productivity-enhancing investments. The overall model explains 67.7% of the variation in TFP, indicating strong explanatory power.

5.3. Robustness of the Baseline Finding

5.3.1. Addressing Selection Bias with PSM-DID

To ensure the credibility and internal validity of our baseline finding, we subject it to an extensive battery of robustness checks, summarized in

Table 3. Robustness Check Results.

	(1)	(2)	(3)
Variables	PSM-DID	Olley-Pakes TFP	Placebo Test
	lnTFP	lnTFP_OP	lnTFP
Treat × Post	0.152 ***	0.138 ***	0.002
	(0.035)	(0.033)	(0.018)
Controls	Yes	Yes	Yes
Firm Fixed Effects	Yes	Yes	Yes
Year Fixed Effects	Yes	Yes	Yes
Observations	12,540	18,927	18,927
R2	0.681	0.655	0.677

Note: Standard errors in parentheses. *** p < 0.01. Column (1) uses a PSM-matched sample. Column (2) uses TFP estimated with the Olley–Pakes method. Column (3) reports the average coefficient from 500 random treatment assignments.

.

A potential concern is that the pilot firms were not randomly selected and might possess unobservable characteristics that make them different. To address this selection bias, we employ a PSM-DID model. After matching firms on a range of pre-treatment observables, the treatment effect, reported in Column (1), remains positive and highly significant (β = 0.152, p < 0.01). The fact that the coefficient is slightly larger suggests that our baseline estimate is conservative and that our finding is not driven by observable selection effects.

5.3.2. Alternative TFP Measurement

As TFP is an estimated variable, our findings could be sensitive to the chosen estimation method. To address this, we re-estimate TFP using the alternative Olley–Pakes (OP) methodology. The result, shown in Column (2), is highly consistent with our baseline (β = 0.138, p < 0.01). This consistency across different advanced TFP estimation techniques significantly strengthens our confidence in the robustness of the core finding.

5.3.3. Placebo Test

To definitively rule out the possibility that our result is due to unobserved confounding factors or pure chance, we conduct a placebo test. We randomly assign treatment status to firms and re-run our baseline regression 500 times. The distribution of these “placebo” coefficients is tightly centered around zero, and our actual coefficient of 0.141 lies far in the right tail. The average placebo coefficient, reported in Column (3), is statistically indistinguishable from zero. This test provides compelling evidence that our result is a genuine treatment effect, not a statistical artifact.

5.4. Unpacking the Mechanisms: Mediation Analysis

To understand how DSCM enhances TFP, we test the two mediating pathways proposed in our theoretical framework: innovation enhancement and cost efficiency.

Table 4. Mediation Analysis Results.

	(1)	(2)	(3)
Variables	Innovation	Cost Efficiency	lnTFP
Treat × Post	0.399 ***	0.016 **	0.061 **
	(0.092)	(0.007)	(0.029)
Innovation	/	/	0.123 ***
	/	/	(0.02)
Cost Efficiency	/	/	1.875 ***
	/	/	(0.45)
Controls and FEs	Yes	Yes	Yes
Observations	18,927	18,927	18,927

Note: Standard errors in parentheses. *** p < 0.01, ** p < 0.05.

presents the results of our causal mediation analysis.

The results provide strong support for both Hypothesis 4 and Hypothesis 5. Column (1) shows that DSCM significantly boosts innovation capability. Column (2) shows that DSCM significantly improves cost efficiency. In the full model Column (3), both mediators are significant predictors of TFP, while the direct effect of DSCM remains significant but is reduced. The innovation channel explains approximately 35% of the total effect, and the cost efficiency channel explains 21%, confirming they are important pathways.

5.5. Heterogeneity Analysis: The Boundary Conditions

5.5.1. The Role of Industry Supply Chain Dependence

This section explores which types of firms DSCM is most effective for. The results are summarized in

Table 5. Heterogeneity Analysis Results.

	(1)	(2)
Variables	High SC Dependence	High R&D Intensity
	lnTFP	lnTFP
Treat × Post	0.095 **	0.081 **
	(0.04)	(0.041)
Treat×Post×Moderator	0.112 **	0.123 ***
	(0.055)	(0.045)
Controls and FEs	Yes	Yes
Observations	18,927	18,927

Note: Standard errors in parentheses. *** p < 0.01, ** p < 0.05.

.

We conduct a new analysis to directly address the reviewer’s concern about the “supply chain” focus of our study. The result in Column (1) is striking: the interaction term Treat×Post×High_SC_Industry is positive and significant (β = 0.112, p < 0.05). This indicates that the TFP gains from DSCM are significantly larger in supply-chain-intensive industries like manufacturing. This powerful finding provides direct empirical evidence that the policy’s mechanism is indeed tied to the optimization of supply chain operations and is not just a generic digitalization effect.

5.5.2. The Role of Firm-Level Absorptive Capacity

Next, we test H₃, which is grounded in absorptive capacity theory. The result in Column (2) shows that the interaction term for firms with high R&D intensity is positive and highly significant (β = 0.123, p < 0.01). Additional untabulated results show similar patterns for large firms and firms with stronger pre-existing digital infrastructure [32,33,34]. This confirms that firms with stronger internal capabilities are better able to absorb and leverage the benefits of DSCM. These findings provide strong support for Hypothesis 3.

5.6. Dynamic Evolution of the Treatment Effect

Finally, we examine how the effect of DSCM evolves over time to test H₂. Figure 2 plots the year-by-year treatment effects from our event study model.

The dual visualization reveals a clear pattern of accelerating returns over time. Figure 2a demonstrates that the effect in the implementation year (Year 0, orange bar) is modest and only marginally significant, which is consistent with the initial costs and disruptions of a major technological overhaul. The effect grows substantially in subsequent years, as shown by the progressively taller bars and deeper red coloring, with statistical significance strengthening over time. Figure 2b further illustrates this trajectory through a smooth upward curve, clearly depicting the accelerating nature of the TFP improvements.

This “S-curve” pattern vividly illustrates the organizational learning curve, peaking in the third year and beyond. The complementary visualizations suggest that firms require time to build complementary capabilities, retrain employees, and re-engineer processes to fully unlock the potential of DSCM. This finding is highly consistent with the dynamic capabilities view and provides robust support for Hypothesis 2, underscoring the importance of a long-term perspective when evaluating digital transformation initiatives.

6. Discussion

6.1. Interpretation of Key Findings

Our study’s primary finding—that DSCM implementation leads to a 14.1% increase in TFP—provides strong, causal evidence of its significant economic value. This result empirically substantiates the core tenets of both the Resource-Based View and Transaction Cost Economics, demonstrating that DSCM functions simultaneously as a strategic, capability-building resource and a powerful mechanism for reducing inter-firm coordination costs.

Perhaps more revealing is the finding that the innovation enhancement channel is a more dominant driver of these productivity gains than cost efficiency. This suggests that the true strategic advantage conferred by DSCM may not lie in immediate operational savings but in its capacity to transform a firm into a more agile, learning-oriented organization. By fostering a collaborative ecosystem for knowledge sharing, DSCM empowers firms to innovate faster and more effectively, which is a more fundamental source of long-term competitive advantage than incremental cost cutting [35,36].

Crucially, our new heterogeneity analysis based on industry supply chain dependence provides a powerful validation for our research focus. The discovery that DSCM’s benefits are significantly amplified in manufacturing and other supply-chain-intensive industries confirms that the policy’s impact is not generic but is specifically tied to the optimization of supply chain operations. This finding not only addresses a critical methodological concern but also adds a new contextual layer to theories like RBV, suggesting that the value of certain digital capabilities is contingent on the operational structure of the industry.

6.2. Theoretical and Practical Implications

Our findings offer several important contributions. Theoretically, this study (1) provides rare causal evidence on the TFP impact of a holistic DSCM strategy, moving beyond the limitations of prior correlational studies; (2) unpacks the “black box” by simultaneously validating dual mediating mechanisms within a unified framework; and (3) identifies industry-level supply chain dependence as a critical boundary condition, enriching our understanding of the contexts in which DSCM thrives.

Practically, our results offer clear guidance. For managers, the key takeaway is that DSCM should be viewed as a strategic investment in innovation and ecosystem capabilities, not just a cost-cutting tool. The findings on heterogeneity suggest that firms should perform a careful assessment of their internal absorptive capacity (e.g., R&D strength) before embarking on large-scale implementation. For policymakers, our research validates the effectiveness of targeted industrial policies that support DSCM [37,38]. However, to maximize returns, such policies should be complemented by investments in regional digital infrastructure and should encourage not just technology adoption but also the organizational changes needed to foster innovation.

6.3. Connecting Productivity to Sustainability Outcomes

While our study’s primary focus is on economic performance, the efficiency gains we document have profound and direct implications for broader sustainability goals. This connection is crucial for understanding the full societal value of DSCM.

From an environmental sustainability perspective, TFP improvements are intrinsically linked to a “greener” footprint. The cost efficiencies we identify are often the result of optimized logistics and better inventory management. Optimized logistics translate into shorter transportation routes and higher vehicle utilization, directly reducing fuel consumption and carbon emissions. Similarly, enhanced demand forecasting capabilities enabled by DSCM lead to less overproduction, which in turn minimizes material waste and energy consumption.

From a social sustainability perspective, DSCM also offers significant benefits. The increased transparency across the supply chain can be leveraged to enhance ethical oversight and ensure compliance with labor standards among suppliers. Furthermore, a more resilient supply chain, as fostered by DSCM, is better equipped to withstand external shocks, which helps ensure the stable supply of essential goods, protect jobs, and contribute to community and economic stability.

7. Conclusions

In conclusion, this study provides robust empirical evidence that the adoption of digital supply chain Management is a significant driver of enterprise productivity. Using a quasi-experimental design, we find that DSCM implementation leads to an average TFP increase of 14.1%, an effect that is primarily channeled through the enhancement of innovation capabilities and the improvement of cost efficiency. Furthermore, we show that this positive impact strengthens over time, highlighting the importance of organizational learning and capability accumulation.

Theoretically, our research contributes by providing strong causal evidence for the value of DSCM, clarifying its underlying mechanisms, and identifying industry-level dependence as a key boundary condition. Practically, our findings underscore the strategic importance of DSCM as an investment in long-term competitiveness and provide actionable insights for both managers and policymakers.

While this study is subject to limitations, such as its focus on listed firms and the inability to capture network spillovers, it opens up critical avenues for future research. The preliminary discussion on sustainability, for instance, calls for future work that directly quantifies the impact of DSCM on environmental and social performance metrics. Moreover, future studies using network data could provide a more complete picture by examining how productivity gains are distributed among supply chain partners. Despite these limitations, the evidence presented here solidifies the understanding of DSCM as a fundamental component of modern industrial strategy, warranting continued attention from all stakeholders committed to building more productive, resilient, and sustainable economies.