You are currently viewing a new version of our website. To view the old version click .
MDPIMDPI
Search all articles

Review Reports

J. Theor. Appl. Electron. Commer. Res.2025, 20(4), 354;https://doi.org/10.3390/jtaer20040354 
(registering DOI)
by
  • Yu Hu1,
  • Kaiti Zou1,* and
  • Xiaofang Chen1,2

Reviewer 1: Anonymous Reviewer 2: Hiroko Oe

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper aims to investigate how investment in computing power drives new quality productivity within the broader agenda of digital transformation and economic upgrading in China. Therefore, the authors constructed a quasi-natural experiment by establishing national computing power hub nodes and used a panel dataset of high-tech manufacturing firms spanning 2011 to 2021. To estimate the causal effects of computing infrastructure shocks, this paper employed a difference-in-differences (DID) approach, supplemented with robustness checks including placebo tests, PSM-DID, and instrumental variable regressions. The results show that investment in computing power significantly promotes new quality productivity, particularly by enhancing digital empowerment, innovation capacity, and human capital upgrading in high-tech manufacturing enterprises. Additionally, the moderating effects of market competition intensity and firm-level absorptive capacity are both statistically significant and theoretically meaningful. Furthermore, heterogeneous analyses reveal differentiated impacts across firm sizes, ownership types, and regional levels of digital infrastructure, highlighting the uneven distribution of digital dividends in China’s transformation landscape.

To improve the overall readability of this paper and provide greater insight for the readers of this journal, I suggest major revisions. Please check the following suggestions:

  • Clarify the theoretical empirical linkage: While the authors provide a solid empirical framework, the discussion could benefit from a more apparent connection between the concept of “new quality productivity” and existing theories in digital economics or production function upgrading. Anchoring the empirical findings within a more robust theoretical narrative would enhance conceptual depth.
  • Enhance the practical interpretation of mechanisms: The mechanism analysis identifies digital empowerment, innovation capacity, and human capital upgrading as key transmission channels. However, the discussion could be improved by providing more concrete examples of how investments in computing power translate into these mechanisms in actual enterprise operations.
  • Improve the treatment of limitations: The current limitations section is brief. The authors are encouraged to acknowledge potential sources of endogeneity beyond those addressed by the instrumental variable approach, as well as challenges in generalizing findings beyond high-tech sectors or the Chinese context.
  • Refine policy recommendations: The policy implications focus mainly on macro-level strategies such as digital infrastructure planning. It would strengthen this paper's practical contribution by offering more actionable guidance for enterprise-level decision-makers, especially on computing resource allocation and workforce digital training.
  • Improve presentation clarity in some sections: While the overall structure is logical, several sections, particularly the robustness check results, would benefit from additional visualizations or summary tables to improve reader comprehension. A diagram illustrating the mechanism framework may also enhance accessibility.

Author Response

Thank you very much for the insightful and constructive comments provided by Reviewer 1 on our manuscript entitled “The Alchemy of Digital Transformation: How Computing Power Investment Fuels New Quality Productivity.” We have carefully reviewed and considered every point raised. We fully agree that the original version could be strengthened in its theoretical articulation, practical case support, discussion of limitations, and the presentation of several sections. The reviewer’s suggestions have been invaluable in enhancing both the academic contribution and the overall clarity of the paper. Following these recommendations, we have undertaken substantial, point-by-point revisions to improve the manuscript.

Below, we provide our detailed responses to each of Reviewer 1’s comments, together with explanations of the corresponding modifications made in the revised manuscript.

Comments 1: [Clarify the theoretical empirical linkage: While the authors provide a solid empirical framework, the discussion could benefit from a more apparent connection between the concept of “new quality productivity” and existing theories in digital economics or production function upgrading. Anchoring the empirical findings within a more robust theoretical narrative would enhance conceptual depth.]

Response 1: We are grateful for this valuable comment. Following the reviewer’s suggestion, we have substantially strengthened the theoretical–empirical linkage by more explicitly articulating how computing power, digital production theory, and new quality productivity (NQP) are connected within a unified conceptual framework. To address this point, we comprehensively revised Section 2.2 while keeping the overall manuscript structure unchanged. The revision clarifies how digital technologies—especially data, algorithms, and computing power—introduce new intangible inputs that reshape the production function and enable qualitative rather than purely quantitative productivity gains. This reframes computing power as a technological driver of production function upgrading and provides a solid macro-level theoretical foundation for NQP. We also embed NQP within established international research on general-purpose technologies, intangible capital, and computation-augmented production functions, positioning NQP as the natural outcome of digital factor embedding rather than a policy-specific construct.

Furthermore, the revised Section 2.2 integrates the macro perspective of production function upgrading with the micro perspective of dynamic capability theory. At the macro level, computing power induces structural changes in the production system, while at the micro level, dynamic capabilities (sensing, seizing, and reconfiguring) represent the organizational mechanisms through which firms internalize the productivity potential of GPT-type technologies. This integrated logic directly anchors our empirical framework and ensures a smoother narrative flow. Importantly, all enhancements were incorporated within the existing subsection to maintain narrative coherence and avoid structural disruption. We sincerely thank the reviewer for this insightful comment, which has helped improve the conceptual clarity and theoretical rigor of our work.

Comments 2: [Enhance the practical interpretation of mechanisms: The mechanism analysis identifies digital empowerment, innovation capacity, and human capital upgrading as key transmission channels. However, the discussion could be improved by providing more concrete examples of how investments in computing power translate into these mechanisms in actual enterprise operations.]

Response 2: Thank you for this highly constructive suggestion. We agree that the original manuscript focused primarily on theoretical derivation in explaining the mechanisms and lacked vivid, concrete examples from actual enterprise operations, which limited the explanatory power of the analysis.

We have therefore added concrete practical examples for each mechanism in Sections 5.1.1, 5.1.2, and 5.1.3 of the revised manuscript. For sensing capability, we added an example showing how firms use computing power for social media sentiment analysis to predict supply chain risks. For seizing capability, we included an example of how companies, such as those in the pharmaceutical industry, employ computing power to run thousands of molecular simulations (digital twins) to accelerate R&D iterations. For reconfiguring capability, we added an example illustrating how manufacturing firms leverage computing power (e.g., IDC infrastructure) to integrate sales, production, and logistics data and implement automated just-in-time (JIT) inventory management. These additions help make the abstract theoretical mechanisms more vivid and convincing.

Comments 3: [Improve the treatment of limitations: The current limitations section is brief. The authors are encouraged to acknowledge potential sources of endogeneity beyond those addressed by the instrumental variable approach, as well as challenges in generalizing findings beyond high-tech sectors or the Chinese context.]

Response 3: The reviewer’s critique is highly pertinent. We acknowledge that the original conclusion section (Chapter 7) contained almost no discussion of limitations, which falls short of the standards expected for SSCI publications. Although we are confident in the extensive robustness checks conducted in the paper (including IV, PSM, and DML), as the reviewer rightly notes, this does not mean that the study is without limitations.

In the revised manuscript, we have added a new subsection in Chapter 7 titled “7.3 Limitations and Future Research.” In this subsection, we address the limitations of the study as suggested by the reviewer. We explicitly note that using IDC license acquisition as a proxy for computing power investment captures access rather than intensity or quality of use (e.g., GPU vs. CPU), which is the most significant limitation of this research. We also acknowledge that, despite employing multiple causal inference techniques, there may still be unobserved, time-varying firm-level factors (such as a sudden shift toward a more digital-oriented corporate culture) that could simultaneously influence both the likelihood of obtaining an IDC license and productivity improvements.

In addition, we discuss issues of generalizability as recommended by the reviewer. We caution that extending the findings to non–high-tech sectors or contexts outside China should be done carefully, as China’s policy-driven environment (e.g., the “East-to-West Computing” initiative) and the specific characteristics of our manufacturing-sector sample may constrain the external validity of the results.

Comments 4: [Refine policy recommendations: The policy implications focus mainly on macro-level strategies such as digital infrastructure planning. It would strengthen this paper's practical contribution by offering more actionable guidance for enterprise-level decision-makers, especially on computing resource allocation and workforce digital training.]

Response 4: Thank you very much for this valuable and thoughtful suggestion, which has been truly helpful in strengthening the practical contribution of our study. We would also like to clarify that Section 7.2 (Practical Implications and Policy Recommendations) in the original manuscript did include recommendations for enterprise decision-makers in the first paragraph and for policymakers in the second paragraph. We understand that the comparatively longer discussion on policy-level implications may have unintentionally given the impression that the enterprise perspective was less emphasized.

We fully agree with the reviewer that more actionable guidance at the enterprise level would enhance the manuscript’s practical relevance. In response, we have substantially enriched and expanded the enterprise-oriented recommendations in the first paragraph of Section 7.2. Specifically, we incorporated the two important aspects highlighted by the reviewer. First, regarding computing resource allocation, we now advise firms to carefully and dynamically balance private IDC investment (asset-heavy) and the use of public cloud services (asset-light), taking into account their evolving operational needs. Second, on workforce digital training, we emphasize the importance of investing in employees’ digital capabilities to build strong organizational absorptive capacity, ensuring that computing power investments can be effectively transformed into meaningful productivity improvements.

Comments 5: [Improve presentation clarity in some sections: While the overall structure is logical, several sections, particularly the robustness check results, would benefit from additional visualizations or summary tables to improve reader comprehension. A diagram illustrating the mechanism framework may also enhance accessibility.]

Response 5: We sincerely appreciate the reviewer’s thoughtful suggestions on enhancing the clarity and readability of the manuscript. We address this comment in two parts.

Regarding the mechanism framework diagram, we are grateful for the reviewer’s recommendation. We would like to clarify that Figure 1 (“Theoretical Framework”) in the original manuscript already serves as the mechanism framework diagram the reviewer has suggested. It illustrates how computing power investment affects new quality productivity through the three dynamic capabilities—sensing, seizing, and reconfiguring—each further divided into nine sub-paths. In the revision, we have strengthened the in-text references to Figure 1 to ensure that readers can more easily locate and understand this framework.

Regarding the visualization and summarization of robustness checks, we fully acknowledge the reviewer’s concerns. The original Sections 4.2, 4.3, and 4.4 indeed present a large number of robustness checks, making these sections dense and potentially challenging for readers. We agree that improving readability is important. At the same time, because the robustness checks involve a wide variety of methods—event-study plots, placebo tests, sensitivity analyses, IV, PSM, and DML—it is difficult to consolidate them into a single summary table without omitting key information essential for assessing causal validity. For this reason, we chose not to introduce an over-simplified aggregate table, as it might weaken the transparency and rigor of the analysis. Instead, we continue to present each robustness check in detail—through Figures 3, 4, and 5 and Tables 3, 4, and 5—which we believe provides readers with a clearer and more trustworthy understanding of the empirical reliability of our findings.

We would like to once again express our sincere appreciation to Reviewer 1 for the valuable time and constructive feedback provided. We hope that the revisions and responses offered here address the concerns of both the reviewers and the editors, and we look forward to the possibility of our manuscript being accepted for publication in your esteemed journal.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript represents a rigorous and theoretically grounded investigation into how computing power investment drives new quality productivity through dynamic capabilities. The study makes substantial contributions to the digital transformation literature and warrants publication in its current form, subject to minor clarifications outlined below.

Major Strengths:

  1. Methodological Excellence: The authors demonstrate exemplary methodological rigor by employing a comprehensive array of state-of-the-art causal inference techniques, including:
    • Six heterogeneity-robust DID estimators (Gardner 2021, Callaway & Sant'Anna 2021, etc.)
    • Honest DID sensitivity analysis (Rambachan & Roth 2023)
    • Double machine learning
    • Bacon decomposition and negative weight diagnostics
    • Multiple robustness checks (PSM-DID, IV, placebo tests, Oster test)
    This level of methodological sophistication substantially exceeds typical standards and provides compelling evidence for causal claims.
  2. Theoretical Contribution: The conceptualization of computing power as a "contestable strategic digital resource" at the micro-level, integrated with dynamic capability theory's three dimensions (sensing, seizing, reconfiguring), offers a novel and coherent framework. The systematic unpacking of the "resource-capability-performance" black box is particularly valuable.
  3. Comprehensive Mechanism Analysis: The empirical verification of three distinct pathways—risk perception enhancement, innovation opportunity seizure, and data factor reconfiguration—each measured through multiple dimensions and indicators, demonstrates exceptional analytical depth.
  4. Nuanced Heterogeneity Analysis: The examination of boundary conditions across executive cognition, market competition, industrial attributes, organizational capabilities, and regional/digital endowments provides important insights into when and where computing power investments are most effective.
  5. Novel Finding: The identification of strategic information concealment behavior as an unintended consequence represents an important contribution to understanding the complex effects of digital transformation.

Minor Concerns and Suggestions for Improvement:

  1. Generalizability Beyond China: While the Chinese context provides an ideal quasi-natural experiment, the manuscript would benefit from a more explicit discussion of:
    • Which findings are likely generalizable to other institutional contexts
    • How China's unique institutional features (e.g., "East Data, West Computing" strategy, government-driven digitalization) might limit or enhance applicability elsewhere
    • Potential boundary conditions for applying these findings in market economies with different regulatory frameworks
  2. Mechanism Specificity: The authors should clarify whether the three mechanisms operate independently or exhibit complementarities/substitution effects. A brief discussion of potential interaction effects among sensing, seizing, and reconfiguring would strengthen the theoretical contribution.
  3. Long-term Effects: The study period (2011-2022) captures short- to medium-term effects. Adding a brief discussion of potential long-term trajectories (e.g., whether effects plateau, accelerate, or diminish over time) would enhance practical implications.

Minor Technical Points:

  • Page 8, Line 363: Consider providing more detail on the 20% equity threshold rationale
  • Table 1: Adding variance inflation factors (VIFs) would help readers assess multicollinearity concerns
  • Figure 2: Consider adding confidence bands to the heterogeneity-robust DID graphs for easier visual interpretation

 

Author Response

We sincerely appreciate Reviewer 2 for providing insightful comments and constructive suggestions, all of which are highly valuable for enhancing the quality of our manuscript, strengthening its theoretical contribution, and improving its external validity. We have carefully reviewed and fully incorporated these suggestions, and have made the necessary revisions and additions in the corresponding sections. We believe these changes have made the paper more rigorous in its reasoning, clearer in structure, and richer in its discussions.

Below are our point-by-point responses to Reviewer 2’s comments, together with explanations of the related revisions made to the manuscript:

Comments 1:

Generalizability Beyond China: While the Chinese context provides an ideal quasi-natural experiment, the manuscript would benefit from a more explicit discussion of: Which findings are likely generalizable to other institutional contexts? How China’s unique institutional features (e.g., “East Data, West Computing” strategy, government-driven digitalization) might limit or enhance applicability elsewhere? Potential boundary conditions for applying these findings in market economies with different regulatory frameworks.

Response 1:

We thank the reviewer for this highly insightful suggestion. We fully agree that although our quasi-natural experiment in the Chinese context offers clear advantages (e.g., a well-defined exogenous shock), greater attention should indeed be paid to the generalizability of the findings and their potential boundary conditions. We also acknowledge that our original manuscript did not sufficiently address this issue, which constitutes an important limitation.

To address this suggestion—while keeping the conclusion section concise—we have added a more comprehensive discussion in Section 7.3 (Limitations and Future Research). This new content elaborates on the generalizability of our findings, China’s unique institutional characteristics, and potential contextual differences in other market economies. It explicitly defines the boundaries of our research and suggests directions for future cross-country and cross-institutional comparative studies.

We believe that this additional discussion (see Section 7.3 in the revised manuscript) substantially enhances the theoretical rigor and international relevance of the paper.

Comments 2:

Mechanism Specificity: The authors should clarify whether the three mechanisms operate independently or exhibit complementarities/substitution effects. A brief discussion of potential interaction effects among sensing, seizing, and reconfiguring would strengthen the theoretical contribution.

Response 2:

We appreciate the reviewer’s important theoretical question, which prompted us to more deeply examine the interrelationships among the three dimensions of dynamic capabilities (sensing, seizing, and reconfiguring). In the original manuscript, we indeed treated them primarily as parallel pathways and did not fully discuss their potential interactions or complementarities. While incorporating all interaction terms empirically would substantially increase model complexity, clarifying their theoretical relationships is essential for strengthening the contribution of the article.

Accordingly, we added a new theoretical discussion at the end of Section 5.1 (Mechanism Tests), right before Section 5.2. We elaborate on the complementarity and dynamic cyclical relationships among the three mechanisms. Specifically, we argue that the three mechanisms do not operate independently; instead, they form an interlocking cycle: strong sensing capabilities provide the foundation for seizing opportunities; seizing new opportunities (e.g., conducting AI R&D) generates new data that pushes firms to strengthen reconfiguring capabilities for upgrading their data infrastructure; and improved data infrastructure further enhances sensing capabilities. Thus, the computing power investment activates not three isolated pathways, but a mutually reinforcing sensing–seizing–reconfiguring cycle.

Comments 3:

Long-term Effects: The study period (2011–2022) captures short- to medium-term effects. Adding a brief discussion of potential long-term trajectories (e.g., whether effects plateau, accelerate, or diminish over time) would enhance practical implications.

Response 3:

We thank the reviewer for this valuable guidance. We agree that our study period (2011–2022) primarily reflects short- to medium-term effects of computing power investments, while understanding their long-term impact is crucial for a more comprehensive view of digital transformation. Following the reviewer’s suggestion—and in line with our response to Comment 1—we added “long-term effects” as an explicit limitation in Section 7.3 (Limitations and Future Research) and briefly discussed possible long-term trajectories. We emphasize that this is an important direction for future research.

Comments 4(a):

Page 8, Line 363: Consider providing more detail on the 20% equity threshold rationale.

Response 4(a):

We appreciate the reviewer’s attention to this detail. Although the original manuscript mentioned the equity method for long-term equity investments, it did not sufficiently explain why this accounting standard is relevant to our study.

To address this, we added a clarification in Section 3.3.2 (Core Explanatory Variable) regarding the rationale for the 20% threshold. We note that a 20% ownership stake is a widely accepted standard under both international and Chinese accounting regulations (e.g., IAS 28) for determining whether an investor has “significant influence.” In our context, using this threshold ensures that the listed firms in the treatment group have substantial strategic and operational influence over their licensed IDCs, rather than merely serving as passive financial investors.

Comments 4(b):

Table 1: Adding variance inflation factors (VIFs) would help readers assess multicollinearity concerns.

Response 4(b):

We thank the reviewer for this important methodological suggestion. We have conducted VIF tests for all variables in the baseline regression models. The results show that both the average VIF and all individual VIF values are well below conventional thresholds, indicating that multicollinearity is not a concern. To maintain brevity in the main text, we added a concise note summarizing the VIF results at the end of Section 3.3.3 (Control Variables).

Comments 4(c):

Figure 2: Consider adding confidence bands to the heterogeneity-robust DID graphs for easier visual interpretation.

Response 4(c):

We appreciate the reviewer’s careful assessment. We infer that the reviewer is referring to Figure 3 in the original manuscript (“Heterogeneity-Robust DID Estimation Results”) rather than Figure 2 (the parallel trends test). The reviewer is correct that the confidence interval representation across the six subfigures in the original Figure 3 was not visually consistent.

We have followed the suggestion and regenerated Figure 3. In the revised version, all six subfigures consistently display 95% shaded confidence bands, improving visual clarity and comparability.

Once again, we sincerely thank Reviewer 2 for the valuable time and constructive comments. We hope these revisions and responses meet the expectations of both the reviewer and the editor, and we look forward to the possibility of our manuscript being accepted by the journal.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

No more comments.

Author Response

Response:

We appreciate the reviewer’s comments on the overall writing quality and structure of the paper. In response to your indications on the review form that “the English could be improved”, “the methods must be improved”, and “the consistency between conclusions and results can be improved”, we have carried out a relatively comprehensive round of revisions.

First, regarding the English expression, we have systematically polished the language of the entire manuscript. We focused in particular on the introduction, theoretical analysis, research design, results presentation, and conclusion. Long sentences, inappropriate wording, and unclear logical connections have been revised one by one, in order to improve clarity and readability.

Second, regarding the research design and methodological description, we now provide more detailed information on the construction process, data sources, and economic meaning of the main variables, especially the core explanatory variable. At the end of this subsection, we also add a short paragraph explaining the limitations of these variables and clarifying the boundary conditions for interpreting the results. For model selection, we now give a clearer justification for using the staggered difference-in-differences (staggered DID) model, further explaining how this model fits our quasi-natural experimental setting. In addition, we have refined the logical framework figure so that the core role of dynamic capabilities is more clearly highlighted.

Third, regarding the presentation of results and the support for the conclusions, we strengthen the theoretical discussion in the results section. In Section 5.1, we add an integrative paragraph that explains the three mechanism tests through the sensing–seizing–reconfiguring dynamic capability framework and clarifies how computing power investment reshapes firms’ information processing and decision-making in digital business environments. We also connect the empirical insights more explicitly to the theoretical framework and to contemporary issues in electronic commerce, and we state more clearly the practical implications for digital business practitioners, online platforms, and technology-driven firms.

We hope that these revisions address the main concerns raised in your review form and make the manuscript more rigorous and transparent in terms of language expression and structural presentation.