Abstract
Purpose
This paper aims to investigate how digitalization empowers government auditing with technical power to serve national governance better.
Design/methodology/approach
This study measures the audit digitalization (AD) index by textual analysis method and matches the provincial AD index with the Chinese listed firm’s data from 2011 to 2019. The two-way fixed effect regression is used to explore the nexus of audit digitalization and corporate green innovation.
Findings
The empirical results demonstrate that government audit digitalization stimulates corporate green innovation, especially for substantive green innovation. Research and development expenditures on personnel and capital are influential mediators and are increased by audit digitalization. The heterogeneity analysis indicates the anti-driving effect for fewer audit informatization expenditures or lower environmental audit coverage, and the incentive effect for state-owned enterprises or firms with corporate social responsibility reports.
Originality/value
The incremental contribution lies in recognizing the progress of government audit digitalization and its role from digital to environmental governance, which extends digital capabilities and digital expertise into the government audit view. Based on textual analysis, a reliable dictionary of audit digitalization is built by machine-learning methods. Then, the authors confirm the effectiveness of audit digitalization, especially when other forms of digitalization fail to promote substantive green innovation. This study also attests to the anti-driving and incentive effect from the external governance perspective. The authors’ findings have implications for digital ecological civilization.
Keywords
Citation
Zhang, Z. and Shi, W. (2024), "How does audit digitalization stimulate corporate green innovation? The mediating role of R&D investment", Managerial Auditing Journal, Vol. 39 No. 7, pp. 799-820. https://doi.org/10.1108/MAJ-08-2023-4013
Publisher
:Emerald Publishing Limited
Copyright © 2024, Emerald Publishing Limited
1. Introduction
Strengthening auditing power in green development through digitalization is critical for the world’s digital ecological civilization. Green innovation and digital empowerment safeguard the construction of beautiful China. As a critical component of the national governance system in China, government audits play a pivotal role in developing the digital governance system and have made progress on digitalization over the years. Audit digitalization, a form of digital governance, leverages digital technologies into audit engagements, especially for complex and dynamic environmental auditing. In the government audit view, digital governance enhances the resilience of environmental governance, facilitating the alignment between public and corporate governance.
Nowadays, a consensus is growing to promote green development, and green innovation is an effective force to balance environmental and economic goals (Jiang and Bai, 2022). In two general categories, substantive green innovation is subject to stricter approval requirements than symbolic green innovation (Dang and Motohashi, 2015). However, symbolic green behaviors may muddle through, catering to government and investor supervision with lower compliance costs (Jiang and Bai, 2022). Regardless of the divergence between corporate and public governance, local officials also have the propensity for environmental inaction or even violation, accounting for high negative externalities and uncertainties in green development, which can disturb firm’s environmental attention allocation (Rodrigue et al., 2013). Environmental governance should focus more on the quality of green innovation, evidenced by environmental punishment (Li et al., 2023), environmental regulation (Lian et al., 2022) and environmental auditing (Liu et al., 2022). However, it is complex for performance measurements and dynamic monitoring in the joint audit of economic and environmental responsibilities. A related but unclear question is how to better serve environmental governance, and what is the role of digital governance.
Green innovation literatures validate the digitalization impact, such as firm digitalization (Ning et al., 2023), digital economy (Guo et al., 2023) and digital finance (Kong et al., 2022). Although little, if any, digitalization research indirectly responds to the agency issue (Chen et al., 2024), these forms of digitalization are often controlled by insiders’ orientation within less motivation for environmental governance. From the perspective of external governance, government audits provide a more pertinent setting for digital governance. Yet, there is little empirical evidence on audit digitalization, despite digital audits being proposed ten years ago. Extant auditing research investigates the role of audit committee information technology expertise (Ashraf et al., 2020) and audit partner digitalization expertise (Maghakyan et al., 2023). Following the audit supervision function, government audits have an advantage over other agencies in leveraging digital technologies in environmental governance. But there is a black box of digital governance, and its role in green governance remains unclear.
Audit digitalization captures the role of digital governance on green governance in China’s government audit setting. Along with China’s Golden Auditing Project, digital technologies are leveraged in the audit process to mitigate audit fatigue under limited auditing resources and deal with the complexity of environment auditing, which is the understanding of audit digitalization (Fotoh and Lorentzon, 2023). In theory, audit digitalization fosters digital capabilities and digital expertise, especially for auditors’ digital skills and knowledge diversity, thus facilitating regular environmental auditing with the effective allocation of government audit resources. Despite the ever-increasing attention on environmental policies, digital governance is conducive to grabbing more environmental attention from audits, governments and firms, improving the efficiency of environmental governance and thus shaping the resilience of green development. Therefore, we posit that government audit digitalization stimulates corporate green innovation.
In response to these realistic questions and theoretical propositions, we combined the Chinese listed firms’ data with the provincial-level audit digitalization index. The audit digitalization index is calculated by the dictionary method in the textual analysis, and its keywords dictionary is built on the working reports of local audit institutions by machine-learning methods. This paper empirically examines the impact of audit digitalization on corporate green innovation and its innovation quality, including the underlying mechanisms and heterogenous effects. Our results confirm that audit digitalization significantly increases granted green patents, especially for green invention patents. Then, audit digitalization is more significant than other forms of digitalization to regulate the formalism of green innovation. The research and development (R&D) investments on personnel and capital are significantly increased by audit digitalization, but R&D personnel input is the most effective mechanism for corporate green innovation, and the mediating effect of R&D capital input is only significant for green invention patents. The heterogeneity analysis proves the anti-driving and incentive effects of audit digitalization in diverse settings, such as audit informatization expenditures, environmental auditing coverage, ownership structure and corporate social responsibility (CSR) disclosures.
The incremental contributions have three points: First, drawing upon digital capabilities and knowledge-based theory, audit digitalization captures the digital capabilities and digital expertise of government audits. Despite the government audit literature focusing more on “what is audited,” as we discussed, audit digitalization related to “how to audit” is informative. In textual analysis, our keywords related to audit digitalization are extracted by machine-learning methods to reduce dependency on subjective judgment. This method provides a reasonable evaluation and more valuable evidence on audit digitalization.
Second, starting from the relevance and credibility of digital governance, we attest to the effectiveness of audit digitalization on the question of whether digital governance promotes green governance. Taking government audit function into account, audit digitalization also enriches the digitalization literature from the perspective of external governance. Unlike prior digital orientation, little is known about the status quo of audit digitalization and its role in digital governance. Our empirical results show that audit digitalization has an advantage over other forms of digitalization on digital governance, especially for substantive green innovation (i.e. green invention patents). Additionally, as an influential mediator related to green innovation, R&D investments are also increased by audit digitalization and act as an effective mechanism for the role of digital governance.
Finally, following the attention-based view, our study dissects the nexus of audit digitalization and corporate green innovation from the anti-driving and incentive perspectives, that is the “carrot and stick” strategy related to digital audits. Our anti-driving effects are attributed to the increased probability of being penalized with fewer digital expenditures or lower environmental audit coverage. Digital audits also enrich incentive effects with more effective environmental assessments for state-owned enterprises (SOEs) or firms with CSR reports. In conclusion, our findings extend the theoretical contribution of audit digitalization and provide policy enlightenment for the synergy between digital governance and green governance.
2. Research background and theoretical hypotheses
2.1 Research background
Digital empowerment is highlighted in the national strategies for green development. The National Conference on Ecological and Environmental Protection in July 2023 emphasized the construction of digital ecological civilization and a digital governance system for beautiful China through the application of digital technologies such as artificial intelligence. The evaluation system of Digital Ecological Civilization is being formulated to promote a digital platform and an intelligent perception system for environmental governance in China. That is, digital governance plays a pivotal role in environmental governance. The 14th Five-Year Plan also makes explicit deployments for promoting green development and accelerating digital development.
Audit institutions have made progress on digitalization. The 2009 National Audit Work Conference had previously provided specific guidance on strengthening the construction of audit informatization and exploring digital audit modes. Since then, informatization construction has been a fundamental work of local audit institutions, and the key to informatization lies in digitalization. Since the 11th Five-Year Plan, audit digitalization and its informatization construction are constitutive of the National Auditing Work Plan. The first meeting of the Central Audit Committee in 2018 emphasized that information technology empowers government auditing. As the issuance of the 13th Five-Year National Informatization Plan, audit digitalization is deployed in the provincial Audit Informatization Plan. In recent years, local audit institutions have issued action plans and related deployments on audit digitalization [1].
Based on the third phase of China’s Golden Auditing Project, audit digitalization is aimed at deepening the application of digital technologies, constructing the national-level digital audit platform and sharing comprehensive, real-time, traceable information among audit institutions. In general, the progress of audit digitalization refers to infrastructure, techniques and applications. Digital infrastructures deploy the construction of audit platforms related to data, information, communication and cybersecurity. Digital techniques focus more on innovating audit modes and data management. Digital applications broaden the audit scope from economy to ecology, especially for tracking unstructured data. This intelligent auditing directly connects with audited projects, data collection and error correction. Nowadays, provincial or municipal audit offices officially report the progress of audit digitalization and its role in ecological governance every year [2]. Taken together, digital audits deeply serve the process of ecological governance.
2.2 Theoretical hypotheses
As the digital governance system constructed in China, government auditing has made progress on digitalization beforehand. Drawing on dynamic capabilities and knowledge-based view, audit digitalization penetrates digital governance into ecological governance through digital capabilities and digital expertise (Tang et al., 2022). Digital capabilities are skilled in the collection, analysis and tracking of unstructured data in environmental auditing through geographic information technology (Lu et al., 2020), big data analytics (Li, 2023), cloud platforms (Zhou et al., 2022) and blockchain (Huang et al., 2022). Digital expertise leverages explorative learning and knowledge diversity to deal with the complexity of environmental auditing with dependence on digital infrastructures, intelligent techniques and cross-functional teams (Filatotchev et al., 2023). Building on the integration of digital capabilities and digital expertise, government audits play the role of digital governance with digital skills and knowledge to specialize in audit function, which shapes the resilience of environmental governance (Barr-Pulliam et al., 2022). Following the attention-based view, digital audits enhance the ability and willingness of environmental governance, subsequently facilitating environmental attention allocation for government officials and corporate managers for a long time. Furthermore, audit digitalization may have anti-driving and incentive effects on prompting green innovation.
On the one hand, audit digitalization imposes an anti-driving effect on corporate green innovation by increasing the probability of being audited or penalized. Audit digitalization can build a dynamic and interactive governance through digital capabilities and digital expertise, which precisely discern and rapidly rectify environmental violations and inactions induced by self-interest motivation (Fotoh and Lorentzon, 2023). From the post-audit rectification to pre-audit inspection, digital capabilities can focus more on the governance process than just the violation results to support lifelong environmental accountability. Armed with digital expertise, government audits reinforce their functional advantages over other agencies to monitor funds and resources for green innovation activities. As mentioned above, audit digitalization increases the probability of being audited for environmental inaction and penalized for environmental violations, then impels government propensity for green development and corporate initiative for green innovation (Fotoh and Lorentzon, 2023; Huang et al., 2023).
On the other hand, audit digitalization facilitates incentive alignment between public and corporate governance for green innovations. Environmental performance is more complicated to assess than economic performance (Zhao et al., 2023). In the joint audit setting, digital connectivity breaks through the barriers in auditor expertise and leverages green knowledge diversity, which forms their judgments on the legitimacy and rationality of social-ecological activities (Bianchi, 2018). When digital capabilities alleviate audit fatigue, government audits have higher incentives to leverage explorative learning and knowledge-based expertise in environmental assessments (Sumiyana et al., 2024). Auditors with digital expertise focus more on knowledge-based expertise than rote rules-based accounting to optimize environmental performance measurement and risk assessment, which provides political incentives for officials (Lee, 2024). It follows that managers have a higher motivation to stimulate the input and output of corporate green innovation if their green behaviors can be assessed in a positive feedback loop (Guo et al., 2023). Thereby, digital governance optimizes environmental attention allocation to enhance incentive alignment for green innovation.
Given the anti-driving effect and the incentive effect, the main hypothesis is as follows:
The digitalization of government auditing stimulates green innovation.
Increased R&D investments also respond to the anti-driving and incentive effects of audit digitalization, which facilitates high-quality green innovation. R&D investments, including personnel and capital, are interrelated with green innovation by resource allocation (Koh and Reeb, 2015). Yet, there is a knot in resource misallocation that leads to inefficient green innovation (Yang et al., 2020). Auditors with digital capabilities and expertise have the potential to detect R&D manipulation and its adverse environmental effects. Continuous auditing can investigate misallocation issues and rectify misallocation behaviors concurrently. From post-audit to pre-audit, dynamic monitoring focuses more on the process of resource allocation and green innovation in the long term, so that R&D investments foster more substantive innovation behaviors. In addition, digital audits have more advantages in supervising new factors of green innovation, such as data, technology and talent (Liu et al., 2023). Thereby, sustainable allocation of R&D resources is conducive to promoting green innovation by audit digitalization. Our hypothesis for the mediating effect is as follows:
R&D investments are significant mechanisms for the impact of audit digitalization on green innovation.
3. Research design
3.1 Model setting
The digitalization of government auditing potentially influences green innovation in the theoretical hypotheses above. To analyze the impact of government audit digitalization, we construct the following model (Li and Du, 2021):
Based on the theoretical analysis above, R&D investments are underlying channels of audit digitalization to stimulate green innovation. According to the three-step method (Wang et al., 2022) on the baseline model (1), the next mediating effect models are as follows:
3.2 Textual analysis
3.2.1 Data pre-processing.
In the original document from 2001 to 2019, we extract the 1,701 audit reports about digitalization at the province and city level. Following the rule of Chinese text processing, data cleaning has two steps: the artificial de-noising process and word segmentation (Majumdar and Bose, 2019). In artificial screening, unusual and valuable words about audit digitalization are added to the terms list and then assist in synonym combination (Wang and Wang, 2018), ensuring the word segmentation quality. Simultaneously, the ordinary Chinese stop words list should be augmented by more frequent but useless words, such as audit, bureau and place names (Dyer et al., 2017). Then, word segmentation uses the Jieba package after uploading the terms list and removing the stop words list (Li et al., 2021). Finally, the clean corpus is ready for keywords extraction at the city level and dictionary methods at the province level.
3.2.2 Latent Dirichlet Allocation (LDA) algorithm.
The LDA algorithm generates the topic dimension to connect the word dimension with the document dimension, which is beneficial for identifying valuable keywords through the probability distributions of topics. The Gensim package is commonly used to construct the LDA model, but a thorny issue is the predetermined topic number, which seriously affects the model’s interpretability. In extant research, the perplexity is used to determine the correct number of topics (Blei et al., 2003). Theoretically, the decreased perplexity means increased model interpretability (Dyer et al., 2017), but too many topics generate repetitive categories and lack practical significance (Li et al., 2021). Specifically, Figure 1 depicts that the perplexity is only slightly declined with more topics if the number of topics exceeds 20. Therefore, the proper topic number is 20 to conduct our LDA model. Then, the LDA algorithm outputs the topic distributions of each document and the topic distributions of words in each document, which are crucial statistical indicators to extract valuable keywords.
3.2.3 Keyword vocabulary.
Based on the vectorized and weighted corpus, our keywords are extracted by the similarity of topic distributions of word and document from the LDA model. Before loading the LDA model, the text corpus is vectorized by the bag-of-words model and weighted by the term frequency-inverse document frequency. Given that the low-frequency but representative words cannot be identified from the LDA topics, the cosine similarity of LDA-generated topic distributions is used to extract more keywords (Campr and Ježek, 2015). The similarity compares the word distributions to the topic distributions of each document. In fact, the word distributions are the topic probability distributions of words in each document. The LDA-generated topic distributions represent the topic-related characteristic of words and documents via vectors, which provides the comparability between word and document through topic distributions. Then, digitalization-related words are extracted from the top 10 characteristic words with the highest similarity in each document. In addition, higher top-N values cannot update our keywords in digitalization vocabulary. Finally, we retain 107 keywords after removing low-frequency synonyms from digitalization-related words. These keywords describe the digital capabilities and digital expertise of government audits and are classified into four types: infrastructure, technique, application and characteristic, as shown in Table 1.
3.3 Variable selection
3.3.1 Explained variables.
In the view of green development, corporate green innovation has been widely discussed. Following Popp (2002) and Guo et al. (2023), green innovation (GP) is measured by the number of green patents granted. In extant research, patents represent particularly observable and quantifiable outputs in the innovation process (Huang et al., 2023). Simultaneously, due to the lagged effects of digitalization and patent examination, granted green patents accurately reveals the nexus of audit digitalization and green innovation. Dang and Motohashi (2015) pointed out that the patent examination between patent applications and grants is approximately four years in China. Notably, compared with the patent applications, the patent grants focus on innovation ability instead of willingness. Additionally, based on the China Research Data Service Platform (CNRDS) database, we also compare different patent types between green invention patents (GIP) and green utility model patents (GUP) (Liu et al., 2023), using the citation of green invention patents granted as the proxy for innovation quality (Dang and Motohashi, 2015). Finally, all proxies are taken a natural logarithm to quantify the ability of corporate green innovation.
3.3.2 Explanatory variables.
From the perspective of government audits, the audit digitalization index is measured by the dictionary method (Yu et al., 2023). To improve the index quality, our keywords are extracted by the cosine similarity of topic probability distributions from the LDA model. Then, drawing upon the above digitalization-related dictionary and the clean corpus at the province level, word frequency should be accumulated year by year, because digitalization-related information disclosure on government audits has been intermittent since 2001. In fact, the digitalization projects of government audits are intensively completed in certain years but continuously applied in the following years even with no nominal projects. Therefore, the yearly accumulated word frequency is taken as the natural logarithm to manifest the audit digitalization (AD) index.
3.3.3 Control variables.
We controlled for firm-level and provincial factors related to digital transformation and green innovation. Digitalization-related variables include provincial digital finance (DF) and firm digitalization (Dig). Firm-level variables refer to financial status, market performance and corporate governance in green innovation studies (Liu et al., 2023; Amore and Bennedsen, 2016). In specific, corporate financial factors include firm scale (Size), asset-liability ratio (Lev), profitability (Growth), return on assets (ROA) and corporate market performance (Turn); corporate governance factors include CEO duality (Dual) and ownership concentration (Top1). Other common variables include firm age (Age), economic development (PGDP) and fiscal decentralization (FD). Table 2 shows the definition of the above variables.
3.4 Data and sample
Our sample comprises 31 provinces (excluding Hong Kong, Macao and Taiwan) and 3048 A-share listed firms in China from 2011 to 2019. The firm data is matched to the provincial-level digitalization index for the province in which the firm is located. Next, listed firms with ST or PT status or in financial industries are eliminated. Simultaneously, the observations with missing values for main variables are also excluded. Finally, the 13,991 firm-year observations from A-share listed firms on the Shanghai and Shenzhen Exchanges are used for empirical analysis. We also mitigate the extremum effect by winsorizing continuous variables at the 1% level.
Our panel data include the firm and provincial levels. For firms’ data, the green patent data of listed firms are from the CNRDS database, and the firm digitalization data is from the Guangdong University of Finance (GUF). Other firm data are mainly obtained from the China Stock Market and Accounting Research (CSMAR) database. For the provincial data, the digitalization-related index on government auditing is manually collected from the information construction section of the local audit institution chapter in the China Auditing Yearbook (CAY). Additionally, informatization expenditure of auditing institutions (IE) and provincial coverage of resource and environment audits (NA) are manually collected from the official website or the China Auditing Yearbook. The provincial digital finance index (DF) is from the Digital Finance Research Center of Peking University (PKU). Other provincial data is from the China Statistical Yearbook (CSY).
Table 3 presents descriptive statistics for the main variables. In Panel A, the values of all variables are distributed in appropriate intervals according to existing research (Xiang et al., 2022). In Panel B, the annual averages of our explained and explanatory variables are increased yearly, reflecting the development trend of audit digitalization and green innovation.
4. Empirical analysis
4.1 Benchmark regression
Table 4 shows the impact of audit digitalization on green innovation and its two general categories. As Columns (1) and (2) show, the estimated audit digitalization (AD) coefficient is significantly positive at the 1% level. After controlling other variables, the estimated AD coefficient is consistently significant in Column (2). The results demonstrate that government audit digitalization stimulates the quantity of firms’ green patents granted, which verifies the main hypothesis H1. Clearly, audit digitalization improves the effectiveness of green governance and then prompts green innovation. We also discern this governance effect in two general categories of green innovation, referring to green invention patents in Columns (3)–(4) and green utility patents in Columns (5)–(6). The AD coefficient for green invention patents is significantly positive in Column (4), indicating that audit digitalization stimulates substantive green innovation, not just symbolic green innovation. Compared with other forms of digitalization, audit digitalization (AD) plays a more significant role than digital finance (DF) and firm digitalization (Dig), especially for substantive green innovation in Column (4). For other control variables, firm scales (Size) is significantly positive for green innovation, and profitability (Growth) negatively affects green innovation.
4.2 Influence mechanism
Table 5 verifies the mediation effects of R&D investment in personnel (RD1) and capital (RD2). Columns (1) and (5) show that audit digitalization significantly increases R&D investment in personnel (RD1) and capital (RD2). Columns (1) to (4) show that R&D personnel investment (RD1) is a significant channel for the impact of audit digitalization on green innovation, both green invention patents and green utility patents, whereas Columns (5) to (8) show that the mediation effect of R&D capital investment (RD2) is only significant for green invention patents. In digital audits, R&D investments, especially for R&D personnel investments, facilitate substantive green innovation better than symbolic green innovation, which supports that audit digitalization mitigates resource misallocation and improves green innovation.
For R&D personnel investment (RD1), the audit digitalization (AD) coefficient is 0.007 and significant at the 1% level in Column (1). The RD1 coefficients are all significant for green patents or different patent types in Columns (2) to (4). For R&D capital investment (RD2), the AD coefficient is 0.009 and significant at the 5% level in Column (5). Nevertheless, the RD2 coefficients are only significant at the 1% level for green invention patents in Column (7) and not significant for green utility patents in Column (8). Based on the Sobel test in Table 5, R&D personnel investment and R&D capital investment are efficient mechanisms for substantive green innovation. And, R&D personnel investment is more significant than R&D capital investment for corporate green innovation increased by audit digitalization.
4.3 Robustness checks
Table 6 shows the robustness results in three aspects: variable substitution, model redesign and endogenous problem.
4.3.1 Variable substitution.
Corporate green innovation (GP), as the core variable, is measured by the citations of green invention patents granted in Column (1) of Table 6. In existing research, the invention patents or patent citations are regarded as high-quality innovations (Dang and Motohashi, 2015; Huang et al., 2023). The AD coefficient for the quality of green innovation (GPQ) is significant at the 1% level in Column (1), indicating that government audit digitalization improves the citations of green invention patents granted, even excluding its self-reference. Additionally, the explanatory variable is replaced by the word frequency of the audit digitalization itself without the natural logarithm in Column (2) of Table 6. After replacing the explanatory variables, the audit digitalization (AD2) coefficient is significant at the 1% level.
4.3.2 Model redesign.
The Tobit model is used for re-estimating the baseline model, as the values of core variables are greater than 0. Column (3) of Table 6 shows that the audit digitalization (AD) coefficient on the Tobit model is significantly positive at the 1% level. In addition, the number of green patents and the values of accumulated word frequency of audit digitalization are taken as the explained variable and explanatory variable instead of the natural logarithm, respectively. Given that the above core variables are discrete and greater than 0, we use the Poisson model to re-examine the baseline model in Column (4), and our conclusions are still consistent.
4.3.3 Endogenous problem.
To alleviate the endogeneity caused by reverse causation, the two-stage least squares (2SLS) regression selects the averages of the explanatory variable in other provinces and the one-year lagged explanatory variable as the instrument variable, respectively. In Columns (5) and (6) of Table 6, the 2SLS results all corroborate the significantly positive impact of audit digitalization. Two instrument variables are valid through the Lagrange multiplier (LM) and Wald F statistic. Specifically, the p-value of the Kleibergen-Paap rk LM statistic is 0.000, the Kleibergen-Paap rk Wald F statistics is greater than the Stock-Yogo threshold value of 1% level, and the null hypothesis is rejected (Lee and Moumbark, 2022). Therefore, our results are robust after considering the endogeneity concern.
Focusing on government environmental auditing, the off-office auditing of natural resource assets implemented since 2014 can promote corporate green innovation. Thus, we limit the sample period from 2014 to 2019 so that the concurrent impact of government environmental auditing on green innovation can be mitigated. Column (7) of Table 6 indicates that the audit digitalization (AD) coefficient is consistent with the baseline model.
5. Further discussion
The above findings confirm that government audit digitalization has a governance effect on corporate green innovation. Further, the in-depth question is whether the governance effect of audit digitalization is heterogeneous and how to distinguish between the anti-driving and incentive effects. Therefore, the heterogeneous effects are as follows.
5.1 Anti-driving perspective
Focusing on the anti-driving effect, Table 7 verifies the heterogeneous effects of audit digitalization from the perspective of audit informatization construction and government environmental auditing.
5.1.1 Audit informatization construction.
Informatization construction is a necessity in the schedule of audit institutions. As the issuance of the 13th Five-Year National Informatization Plan, local audit institutions specially deploy informatization construction into the annual Auditing Work Plan. Audit digitalization relies on the information infrastructure of government audits. High-level informatization expenditures enable auditors to enhance digital capabilities and expertise, but there is a narrow room for renewed improvement by audit digitalization. Thereby, the impact of audit digitalization is evident under lower audit informatization expenditure.
In Table 7, Column (1) indicates that audit informatization expenditure (IE) plays a significant moderating role in the audit digitalization - green innovation relationships. The heterogeneity of audit informatization construction is shown in Columns (2) and (3). Diverse groups are divided by annual informatization expenditure in the final departmental statement from the provincial audit authority. The AD coefficient is significant under lower informatization expenditures in Column (2), and not significant in Column (3). As such, the positive effect of audit digitalization is significant for firms in the region with lower informatization expenditure.
5.1.2 Government environmental auditing.
Environmental auditing stimulates regional environmental governance (Wu et al., 2020) and corporate green development (Huang, 2023) in existing research. However, the implementation of environmental auditing can be impeded by limited audit resources and poor environmental expertise. In this sense, audit digitalization reverses the dilemma of environmental governance through digital capabilities and digital expertise, that increases environmental attention allocation. For example, geographic information technologies are skilled in auditing natural resource assets. Thus, the anti-driving effect of audit digitalization is remarkable under lower environmental auditing.
In Table 7, government environment auditing positively moderates the audit digitalization-green innovation relationship at the 1% level in Column (4). The heterogeneous analysis of environmental auditing on digital governance are shown in Columns (5) and (6). Our sample is classified by the median coverage of government environmental auditing since 2014, which is manually calculated by the ratio of prefecture-level cities implementing the off-office audit of natural resource assets. The AD coefficient is positively significant for regions with lower environmental auditing coverage in Column (5), but not significant for firms in regions with higher environmental auditing coverage. The results indicate that audit digitalization responds more significantly under weak environmental auditing.
5.2 Incentive perspective
Firm-level heterogeneity proves the incentive effect of audit digitalization, as evidenced by the heterogeneity of CSR disclosure and ownership nature in Table 8.
5.2.1 Corporate social responsibility disclosure.
Prior studies show that CSR or its disclosure increases green innovation outputs (Kraus et al., 2020). CSR disclosures provide incremental information related to environmental responsibility, which can be acquired by auditors with higher digital expertise (Dhaliwal et al., 2012; Yao et al., 2023). Then, firms with CSR reports are more motivated to narrow the divergence between corporate governance and public governance. We propose that provincial digital governance role significantly exists in firms with CSR reports.
In Column (1) of Table 8, CSR disclosure significantly moderates the connection between audit digitalization and corporate green innovation. Columns (2) and (3) identify the heterogeneity of CSR disclosure in this connection. The moderating or grouping variable is based on whether firms disclose CSR reports. The AD coefficient of firms with CSR reports (CSR) is significantly positive at the 1% level in Column (2), despite its observations being a third of our sample. It is evident that audit digitalization has more substantial incentives in the CSR group. Instead, the NCSR group are unaffected by government audit digitalization, and its coefficient is insignificant in Column (3). Namely, the incentive effect of audit digitalization on green innovation is statistically significant for firms with CSR reports.
5.2.2 Ownership nature.
In China, SOEs are directly supervised by audit institutions and, thus more prone to be directly impacted by audit digitalization. In the regular environment auditing, digital expertise is beneficial for official’s environmental assessments, which provides more incentive to strive for higher environmental performance (Ren et al., 2024). For SOEs, political incentives stimulate corporate green behaviors and more resources for corporate green innovations (Pan et al., 2020; Zhou et al., 2016).
Columns (4) of Table 8 show the significant moderating effect of ownership nature on the audit digitalization-green innovation nexus. Columns (5) and (6) identify the heterogeneity of ownership nature, as a proxy for grouping SOEs (SOE) and non-SOEs (NSOE). It can be observed that the audit digitalization (AD) coefficient is statistically significant for SOE at the 1% level in Column (3) and non-significant for NSOE in Column (4), indicating that more stringent external supervision prompts SOEs’ green innovation to support the incentive effect of audit digitalization.
6. Conclusions and implications
Government audit digitalization prompts sustainable green innovation. In the role of digital governance, the digitalization of government auditing fosters corporate green behaviors, which propels corporate governance to align with public governance and better serve national green governance. We incorporate the provincial audit digitalization index into the firm’s data from 2011 to 2019 and verify the impact of audit digitalization on green innovation and its mechanisms. Based on the machine learning method, keywords related to audit digitalization are extracted to describe the digital capabilities and digital expertise of government audits. Our findings reveal that audit digitalization significantly facilitates corporate green innovation, referring to the growth of not only quantity but also quality. For substantive green innovation, audit digitalization has a significant governance effect than firm- and regional-level digitalization from the perspective of external governance. The influential mechanisms are corporate R&D investments on personnel and capital increased by audit digitalization, thus improving green innovation, especially for substantive green innovation. In theory, the nexus of audit digitalization and green innovation is identified as the anti-driving and incentive effects, which is demonstrated by the heterogeneity analysis. On the one hand, audit digitalization compensates for fewer audit informatization inputs and lower environmental auditing coverage to impose the anti-driving effect on corporate green innovation. On the other hand, audit digitalization provides more incentives to SOEs or firms with CSR reports. In conclusion, audit digitalization attests to the role of digital governance in shaping the resilience of environmental governance.
Our findings have beneficial implications for developing digital ecological civilization. First, audit digitalization is critical for government audits to serve national governance better and construct the national digital governance system. The priority is to further advance the Golden Auditing Project, focusing on the construction of digital infrastructure and the employment of digital technologies. The barriers to audit expertise can be broken on the process of information sharing and knowledge transfer across all levels of audit institutions and connectivity with other government agencies. In the joint efforts of social audits, government audits should provide more incentives for auditors to enhance their digital capabilities and digital expertise in the process of digital governance.
Second, green innovation is associated with not only environmental attention but also governance resilience. On the one hand, the anti-driving role is strengthened to inhibit the formalism of environmental assessments and the window-dressing of firm digitalization. On the other hand, the incentive role means that environmental assessments should be more scientific, not just legitimate, or rather, environmental performance can be concreted by knowledge-based expertise to motivate SOEs or firms with CSR motivations. Digital audits upgrade the “carrot and stick” approach to improve the alignment between corporate and public governance. As a remedy for market failure, this governance alignment is expected to inhibit resource misallocation and R&D manipulation.
Finally, digital governance empowers green governance, particularly through digital audits. Green governance is a complicated project with numerous workloads, but traditional audits have fewer resources and attention allocation. In the role of digital governance, government audits can deal with audit fatigue by digital capabilities and audit complexity by digital expertise. It is imperative for the joint audit of economic and environmental responsibility. To better serve environmental governance, digital audits are aimed at extending environmental auditing from digital to environmental expertise, outcomes-oriented to process-oriented governance and post-audit rectification to pre-audit inspection. The empowerment of digital governance on environmental governance advocates for a beautiful world.
Figures
Keywords related to audit digitalization
Type | Keywords |
---|---|
Infrastructure | Golden Auditing Project (GAP), GAP phase 1, GAP phase 2, GAP phase 3, Auditor office (AO), Office automation (OA), Audit cloud, Audit platform, Cloud platform, Network platform, Analysis platform, Data center, Information center, Web portal, e-Government Affair, Auditing system, Management system, Information system, Data analysis system, Network system, Control system, Office system, Payment system, Conference system, Information security, Network security, Intranet, Extranet, Local area network, IPv6, Optical fiber, Switch, Computer, Storage space, Storage device |
Technique | Computer audit, online audit, intelligent audit, computer assisted audit, real-time audit, digital audit, mobile audit, information systems audit, audit model, analysis model, big data analysis, correlation analysis, dynamic monitoring, data collection, data mining, data transmission, data exchange, data processing, data analysis, data storage, data administration, collection mechanism, system development, software development, computer intermediate examination, analysis team |
Application | Audit techniques, emerging technology, information technology, geographic information technology, geographic information system, remote sensing, satellite, drone, blockchain, artificial intelligence, cloud computing, on the cloud, Information network, Internet, Database, Electronic data, Module, Digital archives, Data chain, Software, Hardware, Soft hardware, Systems software, Copyrighted software |
Characteristic | Sharing, migration, long-distance, interconnection, data mode, big data, digitization, informatization, networking, intellectualization, visualization, automation, paper-free, computerization, electronization, domestication, standardization, integration, overall process, full coverage, cross-department, multi-dimension |
Source: Compiled by the authors
Variable definition
Variable | Descriptions | Data sources |
---|---|---|
GP | The natural logarithm of the number of green patents granted plus one | CNRDS |
GUP | The natural logarithm of the number of green utility patents granted plus one | CNRDS |
GIP | The natural logarithm of the number of green invention patents granted plus one | CNRDS |
AD | The natural logarithm of the yearly accumulated word frequency about government audit digitalization plus one | CAY |
Size | The natural logarithm of total assets | CSMAR |
Lev | The ratio of liability refers to the solvency of total assets to liabilities | CSMAR |
Growth | The growth ratio of operating income | CSMAR |
ROA | Net profit after tax divided by total assets | CSMAR |
MB | The marker values divided by book value | CSMAR |
Dual | If the positions of chairman and general manager are combined, the value is 1, otherwise 0 | CSMAR |
Top1 | Shareholding ratio of the largest shareholder | CSMAR |
Age | The natural logarithm of listed years | CSMAR |
Turn | The natural logarithm of turnover of the circulation stock plus one | CSMAR |
Dig | The natural logarithm of word frequency about firm digitalization plus one | GUF |
DF | The digital Financial Inclusion Index | PKU |
GDP | The natural logarithm of regional per capita GDP | CSY |
FD | The proportion of revenue by expenditure in the general budget | CSY |
Source: Compiled by the authors
Descriptive analysis
Panel A: Descriptive analysis for the whole sample | ||||||||
Variable | Obs. | Mean | SD | Min. | p25 | p50 | p75 | Max. |
GP | 13991 | 0.669 | 0.954 | 0 | 0 | 0 | 1.099 | 3.850 |
GIP | 13991 | 0.250 | 0.569 | 0 | 0 | 0 | 0 | 2.773 |
GUP | 13991 | 0.554 | 0.867 | 0 | 0 | 0 | 0.693 | 3.526 |
AD | 13991 | 4.155 | 0.780 | 0 | 3.761 | 4.344 | 4.754 | 5.215 |
RD1 | 13991 | 0.091 | 0.125 | 0 | 0 | 0.038 | 0.141 | 0.593 |
RD2 | 13991 | 0.044 | 0.132 | 0 | 0 | 0 | 0 | 0.726 |
Size | 13991 | 22.16 | 1.328 | 19.58 | 21.20 | 21.98 | 22.91 | 26.16 |
Lev | 13991 | 0.430 | 0.214 | 0.049 | 0.258 | 0.419 | 0.585 | 0.964 |
Growth | 13991 | 0.181 | 0.432 | −0.577 | −0.016 | 0.113 | 0.269 | 2.802 |
ROA | 13991 | 0.042 | 0.069 | −0.255 | 0.015 | 0.041 | 0.074 | 0.268 |
MB | 13991 | 2.025 | 1.321 | 0.886 | 1.238 | 1.591 | 2.292 | 8.871 |
Dual | 13991 | 0.277 | 0.448 | 0 | 0 | 0 | 1 | 1 |
Top1 | 13991 | 0.347 | 0.151 | 0.089 | 0.230 | 0.323 | 0.448 | 0.757 |
Age | 13991 | 2.015 | 0.942 | 0 | 1.386 | 2.197 | 2.833 | 3.258 |
Turn | 13991 | 6.130 | 0.866 | 3.434 | 5.554 | 6.129 | 6.710 | 7.958 |
Dig | 13991 | 2.466 | 1.496 | 0 | 1.386 | 2.485 | 3.526 | 7.036 |
DF | 13991 | 5.424 | 0.740 | 2.846 | 5.209 | 5.520 | 5.765 | 10.23 |
GDP | 13991 | 11.11 | 0.441 | 9.706 | 10.80 | 11.14 | 11.45 | 12.01 |
FD | 13991 | 0.657 | 0.187 | 0.072 | 0.468 | 0.730 | 0.814 | 0.931 |
Panel B: Sample distribution by year | |||||||||
2011 | 2012 | 2013 | 2014 | 2015 | 2016 | 2017 | 2018 | 2019 | |
AD | 3.756 | 3.812 | 3.879 | 3.966 | 4.055 | 4.151 | 4.325 | 4.465 | 4.532 |
GP | 0.393 | 0.419 | 0.591 | 0.576 | 0.650 | 0.692 | 0.720 | 0.839 | 0.891 |
GIP | 0.137 | 0.169 | 0.215 | 0.201 | 0.256 | 0.307 | 0.306 | 0.248 | 0.321 |
GUP | 0.327 | 0.331 | 0.498 | 0.478 | 0.529 | 0.540 | 0.578 | 0.742 | 0.749 |
Source: Compiled by the authors
Benchmark regression of audit digitalization and green innovation
(1) GP | (2) GP | (3) GIP | (4) GIP | (5) GUP | (6) GUP | |
---|---|---|---|---|---|---|
AD | 0.077*** (3.866) | 0.066*** (3.325) | 0.033** (2.491) | 0.026** (1.984) | 0.066*** (3.361) | 0.057*** (2.934) |
Dig | 0.023*** (3.155) | 0.005 (1.023) | 0.022*** (3.115) | |||
DF | −0.015 (−1.433) | −0.006 (−0.884) | −0.009 (−0.909) | |||
Size | 0.210*** (14.900) | 0.071*** (7.675) | 0.187*** (13.559) | |||
Lev | 0.032 (0.590) | −0.001 (−0.020) | 0.047 (0.891) | |||
Growth | −0.037*** (−2.925) | −0.020** (−2.392) | −0.027** (−2.196) | |||
ROA | −0.031 (−0.294) | −0.127* (−1.817) | 0.066 (0.639) | |||
MB | 0.012* (1.823) | −0.005 (−1.279) | 0.015** (2.350) | |||
Dual | −0.012 (−0.638) | 0.003 (0.232) | −0.016 (−0.912) | |||
Top1 | −0.071 (−0.790) | −0.040 (−0.677) | −0.065 (−0.742) | |||
Age | 0.017 (0.841) | 0.006 (0.450) | 0.007 (0.357) | |||
Turn | −0.002 (−0.257) | 0.003 (0.512) | −0.005 (−0.548) | |||
PGDP | −0.028 (−0.359) | 0.088* (1.699) | −0.086 (−1.116) | |||
FD | 0.172 (0.845) | 0.157 (1.163) | 0.149 (0.746) | |||
Firm FE | Yes | Yes | Yes | Yes | Yes | Yes |
Year FE | Yes | Yes | Yes | Yes | Yes | Yes |
Constant | 0.066 (0.861) | −4.216*** (−4.853) | −0.009 (−0.180) | −2.553*** (−4.453) | 0.048 (0.638) | −3.120*** (−3.676) |
R-squared | 0.108 | 0.132 | 0.045 | 0.054 | 0.091 | 0.112 |
N | 13991 | 13991 | 13991 | 13991 | 13991 | 13991 |
The t-values are reported in parentheses. The statistical significance is denoted by ***(p < 0.01); **(p < 0.05); *(p < 0.1)
Source: Compiled by the authors
Mediating effect of R&D investments in personnel and capital
(1) RD1 | (2) GP | (3) GIP | (4) GUP | (5) RD2 | (6) GP | (7) GIP | (8) GUP | |
---|---|---|---|---|---|---|---|---|
AD | 0.007*** (2.641) | 0.064*** (3.225) | 0.024* (1.825) | 0.056*** (2.888) | 0.009** (2.575) | 0.066*** (3.306) | 0.025* (1.889) | 0.057*** (2.938) |
RD1 | 0.305*** (4.027) | 0.325*** (6.510) | 0.134* (1.806) | |||||
RD2 | 0.038 (0.738) | 0.135*** (3.940) | −0.010 (−0.190) | |||||
Size | 0.016*** (8.912) | 0.206*** (14.518) | 0.066*** (7.114) | 0.185*** (13.360) | 0.019*** (7.513) | 0.210*** (14.810) | 0.069*** (7.381) | 0.187*** (13.538) |
LEV | −0.002 (−0.292) | 0.032 (0.602) | −0.000 (−0.002) | 0.047 (0.896) | −0.006 (−0.616) | 0.032 (0.595) | 0.000 (0.003) | 0.047 (0.890) |
Growth | −0.003** (−2.061) | −0.036*** (−2.847) | −0.019** (−2.269) | −0.027** (−2.161) | −0.001 (−0.498) | −0.037*** (−2.921) | −0.020** (−2.375) | −0.027** (−2.197) |
ROA | −0.036*** (−2.742) | −0.020 (−0.189) | −0.115* (−1.650) | 0.071 (0.685) | −0.040** (−2.080) | −0.030 (−0.279) | −0.122* (−1.740) | 0.066 (0.635) |
MB | 0.005*** (6.092) | 0.010 (1.589) | −0.007* (−1.655) | 0.014** (2.242) | −0.001 (−0.887) | 0.012* (1.829) | −0.005 (−1.247) | 0.015** (2.348) |
Dual | −0.004* (−1.739) | −0.010 (−0.572) | 0.004 (0.339) | −0.016 (−0.883) | 0.004 (1.083) | −0.012 (−0.646) | 0.002 (0.191) | −0.016 (−0.910) |
Top1 | −0.037*** (−3.247) | −0.060 (−0.667) | −0.028 (−0.478) | −0.060 (−0.686) | −0.025 (−1.510) | −0.070 (−0.780) | −0.037 (−0.621) | −0.065 (−0.744) |
Age | 0.034*** (13.600) | 0.006 (0.320) | −0.005 (−0.385) | 0.002 (0.123) | −0.008** (−2.126) | 0.017 (0.856) | 0.007 (0.530) | 0.007 (0.353) |
Turn | 0.004*** (3.828) | −0.004 (−0.403) | 0.002 (0.277) | −0.005 (−0.613) | 0.000 (0.098) | −0.002 (−0.258) | 0.003 (0.509) | −0.005 (−0.548) |
Dig | 0.004*** (4.361) | 0.021*** (2.988) | 0.004 (0.755) | 0.021*** (3.038) | 0.004*** (3.017) | 0.023*** (3.132) | 0.004 (0.911) | 0.022*** (3.119) |
DF | −0.003** (−2.130) | −0.014 (−1.352) | −0.005 (−0.754) | −0.009 (−0.873) | 0.007*** (3.530) | −0.015 (−1.457) | −0.007 (−1.016) | −0.009 (−0.902) |
PGDP | 0.066*** (6.757) | −0.048 (−0.616) | 0.066 (1.282) | −0.095 (−1.229) | −0.013 (−0.924) | −0.028 (−0.352) | 0.090* (1.734) | −0.086 (−1.118) |
FD | −0.142*** (−5.555) | 0.216 (1.057) | 0.203 (1.506) | 0.168 (0.840) | −0.075** (−2.008) | 0.175 (0.859) | 0.167 (1.239) | 0.148 (0.743) |
Firm FE | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
Year FE | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
Constant | −1.008*** (−8.787) | −3.625*** (−3.944) | −2.148*** (−3.545) | −2.712*** (−3.018) | −0.237 (−1.419) | −3.923*** (−4.279) | −2.443*** (−4.041) | −2.849*** (−3.181) |
R-squared | 0.491 | 0.134 | 0.058 | 0.112 | 0.101 | 0.132 | 0.055 | 0.112 |
Observations | 13991 | 13991 | 13991 | 13991 | 13991 | 13991 | 13991 | 13991 |
Sobel-Z | 2.208** | 2.446** | 1.490 | 0.710 | 2.160** | −0.188 |
The t-values are in parentheses. The statistical significance is denoted by ***(p < 0.01); **(p < 0.05); *(p < 0.1)
Source: Compiled by the authors
Robustness test
(1) GPQ | (2) GP | (3) GP | (4) GP2 | (5) GP | (6) GP | (7) GP | |
---|---|---|---|---|---|---|---|
AD | 0.055*** (3.357) | 0.055*** (2.813) | 0.038** (2.105) | 0.101*** (2.609) | 0.073** (2.262) | ||
AD2 | 0.002*** (3.384) | 0.006*** (2.737) | |||||
Size | 0.035*** (3.008) | 0.211*** (14.968) | 0.257*** (24.978) | 0.570*** (9.084) | 0.058*** (3.005) | 0.216*** (10.757) | 0.286*** (11.927) |
Lev | 0.133*** (2.986) | 0.040 (0.744) | 0.114** (2.437) | −0.033 (−0.145) | 0.064 (0.749) | −0.061 (−0.808) | −0.030 (−0.355) |
GROWTH | −0.016 (−1.558) | −0.038*** (−2.957) | −0.039*** (−3.154) | −0.035 (−1.127) | −0.059*** (−5.032) | −0.009 (−0.545) | −0.021 (−1.322) |
ROA | −0.108 (−1.227) | −0.037 (−0.347) | −0.107 (−1.049) | −0.436 (−1.268) | −0.025 (−0.272) | −0.214 (−1.545) | −0.210 (−1.548) |
MB | −0.018*** (−3.398) | 0.012* (1.799) | 0.018*** (3.027) | 0.025 (0.935) | −0.012 (−1.554) | 0.022** (2.549) | 0.031*** (3.576) |
Dual | −0.007 (−0.479) | −0.013 (−0.732) | −0.016 (−0.998) | −0.012 (−0.182) | 0.004 (0.157) | −0.026 (−1.012) | −0.042 (−1.608) |
Top1 | −0.142* (−1.907) | −0.072 (−0.802) | −0.088 (−1.325) | −0.267 (−0.917) | −0.064 (−0.685) | −0.018 (−0.139) | −0.065 (−0.405) |
Age | 0.008 (0.482) | 0.018 (0.909) | −0.035*** (−2.759) | 0.037 (0.488) | 0.122*** (7.952) | 0.025 (0.528) | −0.002 (−0.048) |
Turn | 0.026*** (3.647) | −0.002 (−0.185) | −0.006 (−0.729) | 0.016 (0.605) | 0.014 (1.390) | 0.002 (0.125) | −0.030*** (−2.677) |
Dig | −0.017** (−2.033) | −0.010 (−0.935) | −0.015 (−1.461) | −0.027 (−0.958) | −0.015 (−0.946) | −0.014 (−0.997) | 0.048 (1.643) |
DF | −0.006 (−0.960) | 0.023*** (3.133) | 0.024*** (4.042) | 0.052** (2.221) | 0.005 (0.748) | 0.019* (1.792) | 0.020* (1.798) |
PGDP | 0.301*** (4.627) | −0.004 (−0.055) | −0.050 (−0.650) | −0.416* (−1.677) | 0.109 (1.036) | −0.022 (−0.199) | −0.183* (−1.673) |
FD | −0.088 (−0.522) | 0.138 (0.674) | 0.186 (0.925) | 1.448** (2.241) | 0.082 (0.244) | −0.062 (−0.206) | 0.206 (0.648) |
Fixed effect | Year/Firm | Year/Firm | Year/Area | Year/Firm | Year/Firm | Year/Firm | Year/Firm |
Constant | −3.580*** (−3.070) | −3.598** (−2.479) | −4.483*** (−3.049) | −7.140 (−1.442) | −5.051** (−2.738) | ||
R-squared | 0.202 | 0.132 | 0.070 | 0.106 | 0.077 | ||
N | 13991 | 13991 | 13991 | 10874 | 13991 | 7770 | 8886 |
The t-values are in parentheses. The statistical significance is denoted by ***(p < 0.01); **(p < 0.05); *(p < 0.1)
Source: Compiled by the authors
Heterogeneity of anti-driving perspective
(1) | (2) | (3) | (4) | (5) | (6) | |
---|---|---|---|---|---|---|
DV = GP | Low-IE | High-IE | Low-NA | High-NA | ||
AD | 0.065*** (3.288) | 0.094*** (4.016) | −0.060 (−0.734) | 0.176*** (4.217) | 0.089** (2.446) | 0.085 (0.592) |
AD*IE | 0.670* (1.945) | |||||
IE | −2.925* (−1.887) | |||||
AD*NA | 0.192*** (2.770) | |||||
NA | −0.920*** (−2.797) | |||||
Size | 0.211*** (14.933) | 0.201*** (10.490) | 0.206*** (7.667) | 0.260*** (13.135) | 0.187*** (6.152) | 0.313*** (9.019) |
Lev | 0.035 (0.645) | −0.016 (−0.216) | −0.036 (−0.377) | 0.000 (0.000) | −0.035 (−0.325) | 0.111 (0.862) |
Growth | −0.038*** (−2.961) | −0.046*** (−2.632) | −0.040* (−1.865) | −0.036** (−2.370) | −0.021 (−0.951) | −0.054** (−1.975) |
ROA | −0.029 (−0.274) | 0.008 (0.055) | −0.279 (−1.519) | −0.241* (−1.875) | −0.069 (−0.372) | −0.306 (−1.387) |
MB | 0.012* (1.804) | 0.010 (1.172) | 0.024** (1.979) | 0.030*** (3.753) | 0.020* (1.793) | 0.030** (1.968) |
Dual | −0.012 (−0.671) | −0.027 (−1.078) | −0.001 (−0.035) | −0.015 (−0.662) | 0.009 (0.250) | 0.013 (0.315) |
Top1 | −0.070 (−0.777) | −0.078 (−0.640) | 0.113 (0.689) | 0.006 (0.043) | 0.212 (1.040) | −0.002 (−0.007) |
Age | 0.016 (0.812) | 0.034 (1.253) | −0.010 (−0.255) | 0.004 (0.136) | 0.066 (1.553) | −0.034 (−0.657) |
Turn | −0.002 (−0.265) | 0.011 (0.903) | −0.028** (−2.022) | −0.026** (−2.482) | −0.029* (−1.862) | −0.037** (−2.061) |
Dig | −0.015 (−1.498) | −0.014 (−1.366) | −0.601* (−1.814) | 0.035 (1.221) | 0.023 (0.777) | −0.839 (−0.856) |
DF | 0.023*** (3.131) | 0.024** (2.461) | 0.009 (0.733) | 0.020** (2.075) | 0.028* (1.869) | 0.008 (0.463) |
PGDP | −0.034 (−0.428) | −0.080 (−0.870) | −0.440 (−1.456) | −0.102 (−1.046) | −0.247* (−1.743) | −0.167 (−0.732) |
FD | 0.190 (0.930) | 0.484* (1.717) | 0.271 (0.632) | 0.040 (0.153) | 0.385 (0.793) | −0.061 (−0.120) |
Firm FE | Yes | Yes | Yes | Yes | Yes | Yes |
Year FE | Yes | Yes | Yes | Yes | Yes | Yes |
Constant | −4.174*** (−4.801) | −3.844*** (−3.830) | 5.128 (1.198) | −4.884*** (−4.210) | −1.650 (−1.028) | −0.163 (−0.030) |
R-squared | 0.133 | 0.130 | 0.101 | 0.094 | 0.090 | 0.092 |
Observations | 13991 | 8186 | 5805 | 10220 | 5635 | 4585 |
The t-values are in parentheses. The statistical significance is denoted by ***(p < 0.01); **(p < 0.05); *(p < 0.1)
Source: Compiled by the authors
Heterogeneity of incentive perspective
(1) | (2) | (3) | (4) | (5) | (6) | |
---|---|---|---|---|---|---|
DV = GP | CSR | NCSR | SOE | NSOE | ||
AD | 0.054*** (2.617) | 0.150*** (3.490) | 0.038 (1.630) | 0.012 (0.526) | 0.100***(3.209) | 0.037 (1.386) |
AD* CSR | 0.055** (2.191) | |||||
CSR | −0.187* (−1.740) | |||||
AD* SOE | 0.123*** (4.718) | |||||
SOE | −0.411*** (−3.525) | |||||
Size | 0.209*** (14.606) | 0.225*** (5.840) | 0.216*** (13.222) | 0.214*** (15.138) | 0.223*** (8.949) | 0.228*** (12.468) |
Lev | 0.038 (0.715) | −0.178 (−1.212) | 0.029 (0.494) | 0.040 (0.747) | −0.216** (−2.200) | 0.081 (1.210) |
Growth | −0.036*** (−2.863) | −0.007 (−0.220) | −0.035** (−2.457) | −0.037*** (−2.905) | −0.017 (−0.793) | −0.050*** (−3.102) |
ROA | −0.030 (−0.282) | −0.461* (−1.741) | −0.056 (−0.473) | −0.052 (−0.489) | −0.198 (−0.989) | −0.115 (−0.901) |
MB | 0.012* (1.814) | −0.018 (−1.044) | 0.018** (2.521) | 0.013* (1.930) | 0.004 (0.314) | 0.018** (2.328) |
Dual | −0.011 (−0.626) | 0.002 (0.042) | 0.000 (0.012) | −0.010 (−0.554) | −0.031 (−0.840) | −0.001 (−0.033) |
Top1 | −0.077 (−0.858) | −0.539** (−2.453) | 0.015 (0.149) | −0.086 (−0.954) | −0.151 (−1.035) | 0.049 (0.411) |
Age | 0.022 (1.076) | 0.051 (0.822) | 0.025 (1.116) | 0.035* (1.735) | 0.122** (2.435) | 0.034 (1.390) |
Turn | −0.002 (−0.205) | 0.025 (1.207) | −0.007 (−0.760) | −0.003 (−0.317) | −0.005 (−0.313) | −0.003 (−0.277) |
Dig | −0.015 (−1.436) | −0.023 (−1.078) | −0.010 (−0.864) | −0.015 (−1.511) | −0.016 (−1.027) | −0.015 (−1.083) |
DF | 0.022*** (3.067) | 0.044*** (2.729) | 0.016* (1.949) | 0.023*** (3.249) | 0.052*** (4.212) | 0.011 (1.213) |
PGDP | −0.029 (−0.365) | −0.052 (−0.337) | −0.075 (−0.797) | −0.021 (−0.272) | −0.038 (−0.328) | 0.008 (0.068) |
FD | 0.175 (0.856) | 0.482 (1.151) | 0.227 (0.947) | 0.176 (0.865) | 0.048 (0.146) | 0.290 (1.106) |
Firm FE | Yes | Yes | Yes | Yes | Yes | Yes |
Year FE | Yes | Yes | Yes | Yes | Yes | Yes |
Constant | −3.866*** (−4.215) | −4.317** (−2.314) | −3.521*** (−3.229) | −3.948*** (−4.312) | −4.320*** (−3.133) | −4.762*** (−3.708) |
R-squared | 0.133 | 0.148 | 0.123 | 0.135 | 0.147 | 0.127 |
Observations | 13991 | 3554 | 10437 | 13991 | 5000 | 8991 |
The t-values are in parentheses. The statistical significance is denoted by ***(p < 0.01); **(p < 0.05); *(p < 0.1)
Source: Compiled by the authors
Notes
There are the Three-Year Action Plan for Digital Audit Reform in Chongqing (2023–2025), and Key Tasks for Big Data Audit in 2023 issued by Beijing Municipal Audit Bureau.
The working reports mention that the Ordos Municipal Audit Bureau used satellite and aerial image data to monitor special projects in 2014, the Sichuan Provincial Audit Office used geographic information system (GIS) technology in the 2016 outgoing audit of natural resources assets, the Fujian Provincial Audit Office established audit platform and audit system together with the Provincial Surveying & Mapping Bureau in 2017, the Wuhan Municipal Audit Bureau built a big data platform for data-sharing on natural resource assets and ecological environment in 2018, and the Shandong Provincial Audit Office established an audit GIS to ascertain 60% of environmental cruxes in 2019.
References
Amore, M.D. and Bennedsen, M. (2016), “Corporate governance and green innovation”, Journal of Environmental Economics and Management, Vol. 75, pp. 54-72.
Ashraf, M., Michas, P.N. and Russomanno, D. (2020), “The impact of audit committee information technology expertise on the reliability and timeliness of financial reporting”, The Accounting Review, Vol. 95 No. 5, pp. 23-56.
Barr-Pulliam, D., Brown-Liburd, H.L. and Munoko, I. (2022), “The effects of person-specific, task, and environmental factors on digital transformation and innovation in auditing: a review of the literature”, Journal of International Financial Management and Accounting, Vol. 33 No. 2, pp. 337-374.
Bianchi, P.A. (2018), “Auditors’ joint engagements and audit quality: evidence from Italian private companies”, Contemporary Accounting Research, Vol. 35 No. 3, pp. 1533-1577.
Blei, D.M., Ng, A.Y. and Jordan, M.I. (2003), “Latent Dirichlet allocation”, Journal of Machine Learning Research, Vol. 3 No. 1, pp. 993-1022.
Campr, M. and Ježek, K. (2015), “Comparing semantic models for evaluating automatic document summarization”, in Král, P. and Matoušek, V. (Eds), Text, Speech, and Dialogue, Springer, Cham, pp. 252-260.
Chen, W., Zhu, C., Cheung, Q., Wu, S., Zhang, J. and Cao, J. (2024), “How does digitization enable green innovation? Evidence from Chinese listed companies”, Business Strategy and the Environment, Vol. 33 No. 5, pp. 3832-3854.
Dang, J. and Motohashi, K. (2015), “Patent statistics: a good indicator for innovation in China? Patent subsidy program impacts on patent quality”, China Economic Review, Vol. 35, pp. 137-155.
Dhaliwal, D.S., Radhakrishnan, S., Tsang, A. and Yang, Y.G. (2012), “Nonfinancial disclosure and analyst forecast accuracy: international evidence on corporate social responsibility disclosure”, The Accounting Review, Vol. 87 No. 3, pp. 723-759.
Dyer, T., Lang, M. and Stice-Lawrence, L. (2017), “The evolution of 10-K textual disclosure: Evidence from latent Dirichlet allocation”, Journal of Accounting and Economics, Vol. 64 Nos 2/3, pp. 221-245.
Filatotchev, I., Lanzolla, G. and Syrigos, E. (2023), “Impact of CEO’s digital technology orientation and board characteristics on firm value: a signaling perspective”, Journal of Management, doi: 10.1177/01492063231200819.
Fotoh, L.E. and Lorentzon, J.I. (2023), “Audit digitalization and its consequences on the audit expectation gap: a critical perspective”, Accounting Horizons, Vol. 37 No. 1, pp. 43-69.
Guo, Q., Geng, C. and Yao, N. (2023), “How does green digitalization affect environmental innovation? The moderating role of institutional forces”, Business Strategy and the Environment, Vol. 32 No. 6, pp. 3088-3105.
Huang, L.F., Zhen, L., Wang, J.B. and Zhang, X. (2022), “Blockchain implementation for circular supply chain management: evaluating critical success factors”, Industrial Marketing Management, Vol. 102, pp. 451-464.
Huang, R.B. (2023), “Auditing the environmental accountability of local officials and the corporate green response: evidence from China”, Applied Economics, Vol. 55 No. 34, pp. 3950-3970.
Huang, X., Liu, W., Zhang, Z., Zou, X. and Li, P. (2023), “Quantity or quality: environmental legislation and corporate green innovations”, Ecological Economics, Vol. 204, p. 107684.
Jiang, L. and Bai, Y. (2022), “Strategic or substantive innovation?-the impact of institutional investors’ site visits on green innovation evidence from China”, Technology in Society, Vol. 68, p. 101904.
Koh, P.-S. and Reeb, D.M. (2015), “Missing R&D”, Journal of Accounting and Economics, Vol. 60 No. 1, pp. 73-94.
Kong, T., Sun, R., Sun, G. and Song, Y. (2022), “Effects of digital finance on green innovation considering information asymmetry: an empirical study based on Chinese listed firms”, Emerging Markets Finance and Trade, Vol. 58 No. 15, pp. 4399-4411.
Kraus, S., Rehman, S.U. and García, F.J.S. (2020), “Corporate social responsibility and environmental performance: the mediating role of environmental strategy and green innovation”, Technological Forecasting and Social Change, Vol. 160, p. 120262.
Lee, H. and Moumbark, T. (2022), “Climate change, corruption, and business bribes in South Asia”, Finance Research Letters, Vol. 47, p. 102685.
Lee, T.A. (2024), “A failure of accountancy professionalisation: corporate financial reporting and accounting knowledge”, Accounting, Auditing and Accountability Journal, doi: 10.1108/AAAJ-09-2022-6032.
Li, M. (2023), “Green governance and corporate social responsibility: the role of big data analytics”, Sustainable Development, Vol. 31 No. 2, pp. 773-783.
Li, M. and Du, W. (2021), “Can internet development improve the energy efficiency of firms: empirical evidence from China”, Energy, Vol. 237, p. 121590.
Li, J., Jiao, J., Xu, Y. and Chen, C. (2021), “Impact of the latent topics of policy documents on the promotion of new energy vehicles: empirical evidence from Chinese cities”, Sustainable Production and Consumption, Vol. 28, pp. 637-647.
Li, X., Guo, F., Xu, Q., Wang, S.W. and Huang, H.Y. (2023), “Strategic or substantive innovation? The effect of government environmental punishment on enterprise green technology innovation”, Sustainable Development, Vol. 31 No. 5, pp. 3365-3386.
Lian, G., Xu, A. and Zhu, Y. (2022), “Substantive green innovation or symbolic green innovation? The impact of ER on enterprise green innovation based on the dual moderating effects”, Journal of Innovation and Knowledge, Vol. 7 No. 3, p. 100203.
Liu, Y., She, Y., Liu, S. and Tang, H. (2022), “Can the leading officials’ accountability audit of natural resources policy stimulate Chinese heavy-polluting enterprises’ green behavior?”, Environmental Science and Pollution Research, Vol. 29 No. 31, pp. 47772-47799.
Liu, X., Liu, F. and Ren, X. (2023), “Firms’ digitalization in manufacturing and the structure and direction of green innovation”, Journal of Environmental Management, Vol. 335, p. 117525.
Lu, H., Wei, Y., Yang, S. and Liu, Y. (2020), “Regional spatial patterns and influencing factors of environmental auditing for sustainable development: summaries and illuminations from international experiences”, Environment, Development and Sustainability, Vol. 22 No. 4, pp. 3577-3597.
Maghakyan, A., Jarva, H., Niemi, L. and Sihvonen, J. (2023), “The effect of audit partner digitalization expertise on audit fees”, European Accounting Review, doi: 10.1080/09638180.2023.2298433.
Majumdar, A. and Bose, I. (2019), “Do tweets create value? A multi-period analysis of twitter use and content of tweets for manufacturing firms”, International Journal of Production Economics, Vol. 216, pp. 1-11.
Ning, J., Jiang, X. and Luo, J. (2023), “Relationship between enterprise digitalization and green innovation: a mediated moderation model”, Journal of Innovation and Knowledge, Vol. 8 No. 1, p. 100326.
Pan, X., Chen, X., Sinha, P. and Dong, N. (2020), “Are firms with state ownership greener? An institutional complexity view”, Business Strategy and the Environment, Vol. 29 No. 1, pp. 197-211.
Popp, D. (2002), “Induced innovation and energy prices”, American Economic Review, Vol. 92 No. 1, pp. 160-180.
Ren, S.G., Liu, D.H. and Yan, J. (2024), “How officials’ political incentives influence corporate green innovation”, Journal of Business Ethics, doi: 10.1007/s10551-024-05622-1.
Rodrigue, M., Magnan, M. and Cho, C.H. (2013), “Is environmental governance substantive or symbolic? An empirical investigation”, Journal of Business Ethics, Vol. 114 No. 1, pp. 107-129.
Sumiyana, S., Susanto, E. A A., Rahajeng, D.K.K. and Winardi, R.D. (2024), “Indonesia’s local government internal auditors (LGIAs): reflecting on low motivation in enhancing their dynamic capabilities while being the spearhead of responsible auditing”, Journal of Accounting and Organizational Change, doi: 10.1108/JAOC-10-2022-0159.
Tang, C., Xue, Y., Wu, H., Irfan, M. and Hao, Y. (2022), “How does telecommunications infrastructure affect eco-efficiency? Evidence from a quasi-natural experiment in China”, Technology in Society, Vol. 69, p. 101963.
Wang, B. and Wang, Z. (2018), “Heterogeneity evaluation of China’s provincial energy technology based on large-scale technical text data mining”, Journal of Cleaner Production, Vol. 202, pp. 946-958.
Wang, K.L., Pang, S.Q., Zhang, F.Q., Miao, Z. and Sun, H.P. (2022), “The impact assessment of smart city policy on urban green total-factor productivity: evidence from China”, Environmental Impact Assessment Review, Vol. 94, p. 106756.
Wu, X., Cao, Q., Tan, X. and Li, L. (2020), “The effect of audit of outgoing leading officials’ natural resource accountability on environmental governance: evidence from China”, Managerial Auditing Journal, Vol. 35 No. 9, pp. 1213-1241.
Xiang, X., Liu, C. and Yang, M. (2022), “Who is financing corporate green innovation?”, International Review of Economics and Finance, Vol. 78, pp. 321-337.
Yang, Z., Shao, S., Li, C. and Yang, L. (2020), “Alleviating the misallocation of R&D inputs in China’s manufacturing sector: from the perspectives of factor-biased technological innovation and substitution elasticity”, Technological Forecasting and Social Change, Vol. 151, p. 119878.
Yao, S., Wei, S.Y. and Chen, L.N. (2023), “Do clients’ environmental risks affect audit pricing? Evidence from environmental violations in China”, Managerial Auditing Journal, Vol. 38 No. 5, pp. 634-658.
Yu, F.F., Du, H.Y., Li, X.T. and Cao, J.Y. (2023), “Enterprise digitalization, business strategy and subsidy allocation: evidence of the signaling effect”, Technological Forecasting and Social Change, Vol. 190, p. 122472.
Zhao, X., Jia, M. and Zhang, Z. (2023), “Promotion vs. pollution: city political status and firm pollution”, Technological Forecasting and Social Change, Vol. 187, p. 122209.
Zhou, K.Z., Gao, G.Y. and Zhao, H. (2016), “State ownership and firm innovation in China: an integrated view of institutional and efficiency logics”, Administrative Science Quarterly, Vol. 62 No. 2, pp. 375-404.
Zhou, X., Zhou, D., Zhao, Z. and Wang, Q. (2022), “A framework to analyze carbon impacts of digital economy: the case of China”, Sustainable Production and Consumption, Vol. 31, pp. 357-369.
Acknowledgements
This work was supported by National Key Research Projects of Accounting (No. 2023KJA3-12).