Corporate digital transformation and rank and file employee satisfaction

Abstract

Purpose

This paper examines the impact of corporate digital transformation on employee satisfaction. Therefore, this study extends our understanding of the economic consequences of corporate digital transformation from employees’ perspectives.

Design/methodology/approach

The data used to construct our main proxy of employee satisfaction are collected from Kanzhun.com, which provides reviews by rank-and-file employees on their employers. This study uses a large sample of Chinese firms and adopts various empirical methods to examine the impact of digital transformation on employee satisfaction.

Findings

We find a significant positive relationship between corporate digital transformation and employee satisfaction. Moreover, we document that the relationship between corporate digital transformation and employee satisfaction is more pronounced in firms with higher labor intensity and in state-owned enterprises (SOE).

Research limitations/implications

One significant limitation is that corporate digital transformation is constructed based on word frequency analysis. This approach may be influenced by variations in corporate disclosure practices and might not accurately capture the true extent of corporate digital transformation. This limitation is not only present in our research but is also pervasive in many other studies that utilize similar methodologies. Therefore, our results should be interpreted with this caveat in mind.

Practical implications

Our study suggests that corporate digital transformation enhances employee satisfaction, providing direct evidence for managers and regulators to promote corporate digital transformation. Through digital transformation, companies can not only improve operational efficiency but also foster employee satisfaction. This dual benefit underscores the importance of investing in corporate digital transformation for long-term success.

Social implications

Our study suggests that corporate digital transformation enhances employee satisfaction, providing direct evidence for managers and regulators to promote corporate digital transformation. Through digital transformation, companies can not only improve operational efficiency but also foster employee satisfaction. This dual benefit underscores the importance of investing in corporate digital transformation for long-term success.

Originality/value

Our study contributes to the literature on the economic consequences of corporate digital transformation and extends existing research on the determinants of employee satisfaction. Additionally, it provides a novel measurement of employee satisfaction for a large sample of Chinese firms.

Citation

Zhang, B., Wang, S. and Zhou, R. (2024), "Corporate digital transformation and rank and file employee satisfaction", China Accounting and Finance Review, Vol. 26 No. 4, pp. 485-511. https://doi.org/10.1108/CAFR-08-2023-0101

1. Introduction

There is an extremely important element of digital transformation that we don’t talk about enough: the human element of digital transformation. How is the digital transformation of organizations bettering the lives of those who work there? How is it alleviating pain points and freeing up time for more rewarding work? — Mehmed Tiro, the founder and CEO of Muftar Corp

The digital economy has been playing an important role in the national economy (Zhao, Zhang, & Liang, 2020; Bodrožić & Adler, 2022). Since corporations are key components of the market economy, corporate digital transformation is fundamentally tied to the development of the macro-level digital economy (Wu, Hu, Lin, & Ren, 2021). Therefore, the economic consequences of corporate digital transformation have attracted much attention from both practitioners and academic researchers. Previous studies document the positive role of corporate digital transformation in facilitating corporate development, such as enhancing productivity and promoting innovation (Park, 2018; Liu, Yan, Zhang, & Lin, 2021; Yuan, Xiao, Geng, & Sheng, 2021; Niu, Wang, Wen, & Li, 2023). However, few studies have focused on the potential impact of corporate digital transformation on rank-and-file employees, even though this aspect may be one of the most significant benefits of corporate digital transformation. In this paper, we investigate the influence of corporate digital transformation on rank-and-file employee satisfaction.

First, we predict a positive association between corporate digital transformation and employee satisfaction. Corporate digital transformation optimizes firms' production modes and processes (Simsek, Vaara, Paruchuri, Nadkarni, & Shaw, 2019; Qi & Xiao, 2020), reducing unnecessary tasks and enabling employees to focus on more rewarding work. In addition, corporate digital transformation enhances firms' ability to analyze data, simplifies the hierarchy of internal information transmission, and provides more valuable and timely information for employees' daily tasks (Constantinides, Henfridsson, & Parker, 2018; Qi & Xiao, 2020; Nie, Wang, & Pei, 2022). Furthermore, corporate digital transformation enables firms to collect employee work information in real-time and with high precision (Chen, Wang, & Chen, 2020), allowing for a more effective assessment of employee efforts. Overall, we expect that corporate digital transformation supports employees in their daily work and enable companies to collect a large volume of precise information for evaluating employee performance, subsequently leading to improved employee satisfaction.

Second, we predict that the relationship between corporate digital transformation and employee satisfaction is more pronounced in firms with higher levels of labor intensity. This is because corporate digital transformation introduces digital equipment, such as automated production lines and intelligent monitoring systems, influencing labor by reducing repetitive tasks and optimizing workflows (Constantinides et al., 2018; Qi & Xiao, 2020). Consequently, the impact of corporate digital transformation on employee satisfaction is expected to be more significant in labor-intensive firms.

Third, we expect state ownership to affect the association between corporate digital transformation and employee satisfaction by influencing the effectiveness of corporate digital transformation. On one hand, corporate digital transformation has highly uncertain payoffs and a high probability of failure. Non-SOEs have limited access to external resources (Zhu, He, & Chen, 2006; Yu, Wang, & Jin, 2012), making their effectiveness in implementing corporate digital transformation relatively lower compared to SOEs. On the other hand, corporate digital transformation prompts comprehensive changes within companies (Qi & Xiao, 2020). However, compared to non-SOEs, SOEs might face more internal resistance, including difficulties related to organizational structure (Hao & Gong, 2017; Wang, Yan, & Song, 2023). Therefore, the impact of state ownership on the relationship between corporate digital transformation and employee satisfaction is an empirical question that requires examination.

Using a sample of listed Chinese firms from 2014 to 2020, this study examines the influence of corporate digital transformation on employee satisfaction. Our employee satisfaction data is obtained from Kanzhun.com, which provides insights into employees' perceptions of their employers. Utilizing sentiment analysis of the employee reviews from Kanzhun.com, we construct a novel measure of employee satisfaction. Consistent with our first hypothesis, we find a significant positive association between corporate digital transformation and employee satisfaction. Furthermore, consistent with our second hypothesis, we find that the positive association between corporate digital transformation and employee satisfaction is more pronounced for firms with higher labor intensity. We also document that the baseline relationship is stronger for state-owned firms, suggesting that state ownership strengthens the positive impact of corporate digital transformation on employee satisfaction.

In addition, our results remain robust to using several methods to mitigate potential endogeneity problems, including the Heckman two-stage regression method, the instrumental variable two-stage least squares (IV-2SLS) approach, the entropy balancing method (EB), and the difference-in-differences (DID) approach. Furthermore, we also use alternative proxies of corporate digital transformation and alternative measures of employee satisfaction. Besides, we do not find a significant association between corporate digital transformation and either employee turnover rate or departed employee satisfaction. These results help us rule out the alternative explanation that our main findings could be attributed to the departure of employees who were dissatisfied with corporate digital transformation.

This paper contributes to the literature in the following ways. First, it adds to the body of work on the economic consequences of corporate digital transformation from the perspective of rank-and-file employees. Previous studies have focused extensively on the impact of corporate digital transformation on corporate performance (Park, 2018; Forman & Van Zeebroeck, 2019; Wu & Kane, 2021; Liu et al., 2021; Wu et al., 2021; Yuan et al., 2021; Zhao, 2021; Chen, Zhang, Jiang, Meng, & Sun, 2022; Chen & Srinivasan, 2023). However, the impact on rank-and-file employees, who are considered to be one of the most important stakeholders of firms, has not been sufficiently explored. Our study extends this line of research by examining how corporate digital transformation affects employee satisfaction, providing evidence for the employee-related explanation that corporate digital transformation enhances corporate performance, which has not been fully explored in the prior literature.

Second, our study contributes to the stream of research on the determinants of employee satisfaction from the perspective of organizational transformation. Prior literature has explored the factors affecting employee satisfaction from different perspectives (Landsbergis, 1988; Cohen and Spector, 2001; Huang, 2005; Nishii, 2013; Solomon, Nikolaev, & Shepherd, 2022). For example, Cohen and Spector (2001) document that employee satisfaction is improved through fair performance evaluation. Solomon et al. (2022) find that better-educated individuals tend to have a higher degree of work intensity and task pressure, which are often associated with the increased stress, resulting in a negative relationship between education and employee satisfaction. We complement this line of research by documenting that corporate digital transformation promotes employee satisfaction.

Last, this paper contributes to the literature on employee satisfaction by introducing a novel approach to empirically measuring employee satisfaction for Chinese firms. Prior empirical studies mainly rely on data from Glassdoor.com, which collects employee satisfaction ratings and employee reviews of employers in the U.S. (Huang, Li, Meschke, & Guthrie, 2015; Fu, Ji, & Jing, 2020; Dube & Zhu, 2021; Farhadi & Nanda, 2021). However, due to limited data in China, previous studies on this topic are mainly based on theoretical analyses or corporate surveys (Ye, Wang, & Lin, 2005; Huang, 2005; Cui, Zhang, & Qu, 2012; Zhou & Zhang, 2021). To the best of our knowledge, our study is the first one that uses employee reviews extracted from Kanzhun.com and constructs an employee satisfaction measure. This approach overcomes data limitations and serves as a valuable reference for future empirical studies.

The rest of this paper is organized as follows. Section 2 reviews the relevant literature and develops hypotheses. Section 3 describes the data, variables, and research design. Section 4 presents the empirical results. We conclude in Section 5.

2. Literature review and hypotheses development

2.1 Literature review

Employee satisfaction is crucial as it mirrors employees' emotional well-being and contentment concerning their jobs and associated tasks (Ye et al., 2005; Brayfield & Rothe, 1951). High levels of employee satisfaction are associated with reduced employee turnover (Chen, Ployhart, Thomas, Anderson, & Bliese, 2011), increased customer satisfaction (Harter, Schmidt, & Hayes, 2002), enhanced innovation efficiency, and total productivity (Xu, Ni, & Liu, 2021), as well as improved business performance and shareholder value (Edmans, 2011; Faleye & Trahan, 2011). Consequently, previous studies have paid considerable attention to the determinants of employee satisfaction. At the work level, factors include the perceived importance of work tasks (Liu, Mitchell, Lee, Holtom, & Hinkin, 2012), the degree of employee participation in decision-making (Chow, 1983), fairness (Cohen & Spector, 2001), and the experience of work-related stress (Landsbergis, 1988). Factors at the employee level include interpersonal relationships within the workplace (Zhang & Li, 2001), salary and compensation packages (Huang, 2005), employee education (Solomon et al., 2022), and opportunities for learning and growth (Liu & Zhang, 2004). Furthermore, factors at the firm level also significantly influence employee satisfaction, including leadership styles (Li, Tian, & Shi, 2006), workplace atmosphere (Nishii, 2013), and workplace environment (Cui et al., 2012).

Corporate digital transformation represents the essential pathway for firms to navigate the challenges and opportunities arising from the digital economy (Verhoef et al., 2021; Wu et al., 2021). A great body of research has examined the economic consequences of corporate digital transformation. This transformative process profoundly reshapes various aspects of organizations, including their organizational structure, production mode, governance, and management practices (Davenport & Westerman, 2018; Fischer, Imgrund, Janiesch, & Winkelmann, 2020; Qi & Xiao, 2020; Zhang & Chen, 2020; Hanelt, Bohnsack, Marz, & Antunes Marante, 2021; Chen & Hu, 2022). By enhancing the efficiency of internal information flow and resource allocation, corporate digital transformation provides vital information support for managerial decision-making, thereby improving overall production efficiency (Zhang, Lu, & Li, 2021; Minardi, Hornberg, Barbieri, & Solga, 2023) and mitigating cost stickiness (Quan & Li, 2022; Wu & Tian, 2022). Moreover, corporate digital transformation bolsters organizational resilience (Shan, Xu, Zhou, & Zhou, 2021), curbs corporate earnings management (Nie et al., 2022), and enhances corporate stock liquidity (Wu et al., 2021). Furthermore, the application of digital technologies facilitates knowledge integration within the firm (Forman & Van Zeebroeck, 2019), conveniently equips employees with relevant professional knowledge (Wu & Kane, 2021), improves corporate innovation (Park, 2018), and enhances input-output efficiency (Liu et al., 2021) as well as total productivity (Yuan et al., 2021).

Nevertheless, prior literature has paid limited attention to the economic consequences of corporate digital transformation at the employee level. Several studies have explored the impact of corporate digital transformation on the labor market and corporate labor structure resulting from technological advancements like the adoption of robots and artificial intelligence (Acemoglu & Restrepo, 2020; Acemoglu, Autor, Hazell, & Restrepo, 2022). For example, Acemoglu and Restrepo (2020) find that the use of robots has a displacement effect on employment, which decreases the labor demand and wages. However, these studies mainly focus on the influence of the macro-level labor market. Consequently, our study adds to the existing literature by focusing on the firm level of rank-and-file employee satisfaction.

2.2 Hypotheses development

2.2.1 The impact of corporate digital transformation on employee satisfaction

We expect that corporate digital transformation enhances employee satisfaction for the following reasons. First, corporate digital transformation provides support for employee daily work, thereby fostering employee satisfaction. Specifically, the adoption of digital technologies facilitates a more modular and flexible production mode (Qi & Xiao, 2020), enabling firms to effectively monitor and optimize the production process (Simsek et al., 2019). This allows for the timely identification and elimination of redundant work, streamlining the workflow for employees. For example, in the digital era, data can be instantly input into the system as tasks are finalized, removing the redundancy of separate data entry and boosting overall productivity. By minimizing repetitive or redundant work, corporate digital transformation frees employees up to engage in more creative and innovative activities, which provides employees with more fulfillment and promotes their career development (Hackman & Oldham, 1976).

Second, corporate digital transformation facilitates a transition towards a networked information structure (Nie et al., 2022). This transition enhances firms' capacity to collect, translate, analyze, and utilize various types of information, providing valuable data and information support for employee work. Digital platforms emerging from the transformation process enable efficient information transmission (Constantinides et al., 2018). For example, corporate digital transformation enables a direct connection between employees and users, bypassing the traditional intermediary processes between customers and managers (Qi & Xiao, 2020). This direct connection streamlines information flow and feedback loops, allowing employees to receive timely information to support their work. Therefore, corporate digital transformation can improve employee satisfaction by providing them with more support in their daily work (Liu et al., 2012).

Third, corporate digital transformation enables firms to collect large volumes of precise information, both structured and unstructured, which can be used in employee performance evaluation. Therefore, the informativeness of employee performance evaluation is enhanced, increasing employee satisfaction. Specifically, corporate digital transformation significantly improves the corporate ability to collect, integrate, process, and share internal information (Chen, Wang, & Chen, 2020; Yang, Hou, Li, & Wu, 2021; Liu et al., 2021). For example, by using digital technologies such as big data and cloud computing, firms collect more information such as employees' working processes and real-time working statuses, which can be used in employee performance evaluation. This enables firms to obtain informative information used in employee performance evaluation, thus improving employee satisfaction. Based on the above arguments, we present our first hypothesis as follows:

H1.

The level of employee satisfaction is positively associated with the degree of corporate digital transformation.

However, there are arguments that may work against finding a significant positive association between corporate digital transformation and employee satisfaction. Corporate digital transformation may have adverse effects on employee satisfaction because of the concerns about job insecurity and the over-monitored working environment. Prior literature finds that firms' digital advancement often involves the replacement of low-skilled labor with high-tech devices (Zhao, 2021; Menz et al., 2021), reducing the employment opportunities for low-skilled employees (Tabrizi, Lam, Girard, & Irvin, 2019). Besides, corporate digital transformation demands employees to possess relevant digital and analytical capabilities to adapt to the evolving work environment (Chen, Wang, & Chen, 2020). To avoid potential dismissal from the company, most employees are compelled to invest their spare time in learning related technologies. Additionally, while digital technologies enable managers to use monitoring technology to foster self-discipline among employees (Kellogg, 2022), they also create an intense work atmosphere and even a feeling of being infringed upon privacy (Bhave, Teo, & Dalal, 2020). Such monitoring practices can be perceived as a means of oppression (Xie, Zuo, & Hu, 2021), negatively impacting employee satisfaction. Therefore, the impact of corporate digital transformation on employee satisfaction is an empirical question to examine.

2.2.2 Labor intensity

Based on the foregoing discussion, we argue that corporate digital transformation significantly simplifies the workflow of employees, reduces repetitive tasks, and enhances the corporate capability to process and utilize information (Constantinides et al., 2018; Simsek et al., 2019; Qi & Xiao, 2020). Specifically, through automating routine tasks and optimizing workflows, corporate digital transformation enables employees to focus on more rewarding activities, thus potentially increasing employee satisfaction. Additionally, the improved ability to collect and analyze data allows firms to make more informed decisions, providing employees with relevant and timely information to support their work. Compared to non-labor-intensive firms, labor-intensive firms have a stronger dependence on the labor factor (Ni & Zhu, 2016). Therefore, labor-intensive firms are more likely to benefit from corporate digital transformation. Consequently, we expect that the impact of corporate digital transformation on employee satisfaction is more pronounced for labor-intensive firms. The above arguments lead to our second hypothesis:

H2.

The positive association between corporate digital transformation and employee satisfaction is more pronounced for firms with higher levels of labor intensity.

2.2.3 State ownership

On the one hand, corporate digital transformation is a long-term process with high uncertainties that requires substantial resource support (Liu et al., 2021). State-owned firms have resource advantages (Luo & Zhen, 2008; Pan, Xia, & Yu, 2008; Li & Yang, 2015). In other words, state ownership might enhance the effectiveness of digital transformation implementation (Zhao, Wang, & Li, 2021). On the other hand, digital transformation requires firms to reshape organizational structures (Chen, Huang, & Liu, 2020; Qi & Xiao, 2020; Zhang & Chen, 2020). State ownership might impose considerable constraints on the digital transformation process (Shu & Chen, 2024), reducing the effectiveness of corporate digital transformation. Overall, the impact of state ownership on the relationship between corporate digital transformation and employee satisfaction is uncertain. Therefore, we present our third hypothesis as follows:

H3a.

The positive association between corporate digital transformation and employee satisfaction is more pronounced for SOEs.

H3b.

The positive association between corporate digital transformation and employee satisfaction is less pronounced for SOEs.

3. Data, sample selection, and research design

3.1 Data and sample selection

We obtain corporate digital transformation data, corporate governance data, and financial data from the China Stock Market and Accounting Research (CSMAR) database, and employee data from the Wind database. The data used to construct our main proxy of employee satisfaction are collected from Kanzhun.com, which provides reviews by rank-and-file employees on their employers.

Our initial sample covers the period of 2014–2020. Our sample period begins in 2014 because the online date of Kanzhun.com is December 2013. Firms in financial industries are excluded, and we remove observations with missing information for calculating the control variables. The sample selection process yields 6,118 firm-year observations from 2014 to 2020. Table 1 presents our sample selection procedure.

3.2 Empirical model

We examine the impact of corporate digital transformation on employee satisfaction using the following model:

(1)EmpSatisfactioni,t=α0 +α1Digitali,t−1 +γControlsi,t+Year+Industry+ε,

where Emp_Satisfaction_i,t is employee satisfaction, defined as the proportion of positive reviews from Kanzhun.com. Specifically, we use sentiment analysis to identify the emotional tendency of each review, which refers to the emotional attitude expressed within a sentence (Bochkay, Brown, Leone, & Tucker, 2023). In our study, the emotional tendency of each employee review expresses positive or negative emotions towards their employers. We utilize the sentiment analysis technology from the Baidu AI open platform, which has been trained with internet text data accumulated from Baidu in various contexts using deep learning techniques, to detect the emotional tendencies in employee reviews (Sun, Xue, & Cui, 2021; Luo, Wu, Su, & Yu, 2021). This process involves analyzing the text of each review to determine the likelihood of it being positive or negative, respectively. A review is classified as positive (negative) if the probability of being positive (negative) is higher than that of being negative (positive). We calculate the proportion of positive reviews in each firm-year as our main proxy of employee satisfaction (Huang, Teoh, & Zhang, 2014; D'Augusta & DeAngelis, 2020), where a higher proportion of positive reviews indicates a higher level of employee satisfaction.

Our key variable of interest, Digital_i,t−1, is the degree of corporate digital transformation in year t−1. Consistent with prior literature, Digital is defined as the natural logarithm of one plus the total number of word frequencies related to corporate digital transformation (Wu et al., 2021; Zhao, 2021). Word frequency refers to the number of times a specific word appears in a text. Specifically, in our study, we utilize keywords related to “corporate digital transformation” obtained from the CSMAR database. We exclude the cases that there are negation words in the front of the keywords.

Following prior research (Huang et al., 2015; Fu et al., 2020; Dube & Zhu, 2021), we include firm-level characteristics that are likely to affect employee satisfaction, including firm size (Size), defined as the natural logarithm of total assets; firm age (Age), measured by the natural logarithm of one plus the number of years of listing; leverage (LEV), measured by the ratio of total liabilities over total assets; return on assets (ROA), defined as net income scaled by total assets; net cash flow (NCF), measured by the ratio of net cash flow from operations to total assets at the end of the year; property right (SOE), a dummy variable that equals one if a firm is a state-owned enterprise, and zero otherwise; R&D investment (RD), defined as the ratio of total R&D spending to sales revenues.

We also include employee-level factors that may have an impact on employee satisfaction (Zhang & Li, 2001; Cui et al., 2012). Specifically, we control for labor intensity (LaborInt) and average employee salary (AEP). Labor intensity reflects the significance of employees to the firm. Additionally, employee compensation is important to employees, and we expect a positive association between employee satisfaction and employee compensation. Zhou and Tang (2015) find that the degree of market competition affects employee satisfaction through employee training. We thus include the Herfindahl-Hirschman index which captures the degree of market competition (HHI), defined as the sum of the squares of the ratio of each firm sales to the total sales of all the firms in the industry. All the variables are defined in Appendix. In addition, we also include year fixed effects and industry fixed effects in model (1). The stand errors are clustered at the firm level.

4. Empirical results and analysis

4.1 Descriptive statistics

Table 2 reports the descriptive statistics of the variables used in model (1). The mean (median) value of employee satisfaction (Emp_Satisfaction) of the number is 0.547 (0.500), which suggests that employee satisfaction is evenly distributed in our sample. On average, the degree of corporate digital transformation (Digital) is 1.385 with a standard deviation of 1.467 and the median value is 1.099. Regarding the other variables, the average (median) firm size (Size) is 22.522 (22.300). On average, firm age (Age), leverage (LEV), return on assets (ROA) and net cash flow (NCF) are 2.019, 0.428, 0.043, and 0.051, respectively. The mean (median) value of property right (SOE) is 0.316 (0.000), suggesting that seventy percent of firms in our sample are non-state-owned enterprises. The average (median) percentage of R&D investment to operating revenue (RD) is 0.055 (0.040), and the mean values for labor intensity (LaborInt), average employee pay (AEP), and the degree of market competition (HHI) are 0.357, 1.397, and 0.131, respectively.

Table 3 provides Spearman and Pearson correlations between our main variables. Consistent with prior research (Huang et al., 2015; Fu et al., 2020; Dube & Zhu, 2021), the results suggest that employee satisfaction (Emp_Satisfaction) is positively correlated with firm size (Size), return on assets (ROA), and average employee pay (AEP), and is negatively correlated with labor intensity (LaborInt). In addition, Table 3 shows that the degree of corporate digital transformation (Digital) is positively correlated with employee satisfaction (Emp_Satisfaction). These results preliminarily support H1 that corporate digital transformation improves employee satisfaction.

4.2 Main results

4.2.1 The impact of corporate digital transformation on employee satisfaction

Table 4 presents the results from the ordinary least squares estimation of the model (1). Column (1) shows the regression results without control variables but the year and industry fixed effects are included; in Column (2), we add all the control variables. The coefficients for Digital in both columns are positive and significant, supporting H1 that corporate digital transformation improves employee satisfaction. Specifically, on average, a one standard deviation increase in the degree of corporate digital transformation is associated with a 3.2% increase in employee satisfaction. In addition, we also find that firm size (Size), return on assets (ROA), and average employee pay (AEP) are positively correlated with employee satisfaction (Emp_Satisfaction), indicating that firms with larger firm size, higher return on total assets and more average employee may have a higher level of employee satisfaction. The above findings are generally in line with the findings of previous studies (Huang et al., 2015).

4.2.2 Labor intensity

Following prior research (e.g. Pan & Chen, 2017; Dube & Zhu, 2021), we measure the labor intensity of firms (LaborInt) by the ratio of the natural logarithm of the number of employees to the natural logarithm of total assets. If the average labor intensity of an industry is greater than the median of all industries, the industry is considered to be labor-intensive (Ni & Zhu, 2016; Zhu, 2020). Then, we define a dummy variable IFHLB, which equals 1 if the firm belongs to a labor-intensive industry and 0 otherwise.

We divide the sample into two subsamples according to IFHLB and estimate model (1) using the two subsamples, respectively. Column (1) of Table 5 shows that the coefficient of Digital is positive but insignificant in the non-labor-intensive subsample. Column (2) of Table 5 shows that the coefficient of Digital is positive and significant in the labor-intensive subsample. The difference between the coefficient of Digital is significant between the two groups. The results are consistent with H2, suggesting that the positive role of corporate digital transformation on employee satisfaction is stronger for firms with higher labor intensity.

4.2.3 State ownership

To test H3, we define a dummy variable SOE, which equals one if a firm is a state-owned firm, and zero otherwise. We divide the sample into two subsamples (SOEs and non-SOEs) and estimate model (1) using the two subsamples, respectively. Column (1) of Table 6 shows that the coefficient of Digital is positive but insignificant in the non-SOE subsample. Column (2) of Table 6 shows that the coefficient of Digital is positive and significant in the SOE subsample. The coefficient of Digital is significantly different across the two groups. The results show that the positive role of corporate digital transformation on employee satisfaction is stronger for state-owned firms, supporting H3a.

4.3 Endogeneity tests

4.3.1 Heckman two-stage regression

There may be potential sample selection bias in this paper. Our empirical tests are conducted based on a sample consisting of firms having employee reviews on Kanzhun.com. The reviews on employers are self-reported by employees. Thus, there is a concern that firms in this study might be systematically selected, causing potential sample selection bias. Therefore, we use the Heckman two-stage regression method to mitigate potential sample selection bias (Huang et al., 2015; Fu et al., 2020). In the first stage, we estimate a probit regression based on a full sample consisting of all the listed companies to obtain the inverse Mill's ratio (imr). Specifically, the dependent variable of the first stage, IFrev, is a dummy variable that equals one if a firm has employee reviews on Kanzhun.com and zero otherwise. We choose the internet penetration rate (Int_pen) as the IV in the first stage. First, the internet penetration rate reflects the level of internet development in a province. In regions with more developed internet, employees are more likely to use online platforms for communication and share their work experiences (Cai, Xu, & Lu, 2022), meeting the relevance condition. Second, the internet penetration rate does not directly affect employee satisfaction, satisfying the exclusion restriction. Additionally, we select the control variables from the main regression as control variables in the first stage (Lennox, Francis, & Wang, 2012).

In the second stage, we include the inverse Mill's ratio (imr) obtained from the first stage in model (1). Table 7 presents the results of the Heckman two-stage regressions. Column (1) of Table 7 presents the estimation results of the first stage. The results show that employees are more likely to express their views about their employers online when they are located in regions with more developed internet. Column (2) of Table 7 shows the regression results of the second stage. After controlling for the inverse Mill's ratio (imr), the coefficient of corporate digital transformation degree (Digital) is still significantly positive, indicating that our results remain robust after controlling for the potential sample selection bias.

4.3.2 Instrumental variable two-stage least squares (IV-2SLS) approach

There might be potential omitted variable problems or reverse causality problems. For example, employee satisfaction and corporate digital transformation could be simultaneously influenced by factors, causing the omitted variable problem. For reverse causality problems, higher employee satisfaction is often associated with a lower turnover rate (Chen et al., 2011). In this case, corporate operational activities are often more stable, reducing operating risks (Detert & Burris, 2007), providing a suitable environment for transformation and upgrading, and promoting corporate digital transformation. In model (1), the degree of corporate digital transformation is a one-year lagged value, which alleviates the above concern to some extent.

To further mitigate the potential endogenous problems, we adopt the IV-2SLS approach for estimation. Specifically, following previous studies (e.g. Huang, Yu, & Zhang, 2019; Zhao et al., 2020), we use the number of post offices per million people in each prefecture-level city in 1984, in which year the China City Statistical Yearbook began to record the information of post offices, as an instrumental variable (Postal). This IV is likely to positively influence the degree of corporate digital transformation, but it is less likely to affect employee satisfaction directly. The reason is that historical telecommunications infrastructure influences local technological levels and the subsequent development and application of Internet technology. Since Internet technology is essential to facilitate the dissemination and application of digital technology, corporate digital transformation should be positively associated with the historical count of postal services, which meets the relevance condition. Moreover, the number of historical post offices is unlikely to affect employee satisfaction, which satisfies the exclusion restriction. Following prior literature (Nunn & Qian, 2014), we introduce a related time-varying variable to interact with Postal. Thus, the IV is defined as the interaction between the one-year lagged value of the national Internet users (10 million) and the number of post offices owned by every million people in each prefecture-level city in 1984 (Postal*Internet) (Huang et al., 2019; Zhao et al., 2020).

Column (1) of Table 8 reports the first stage regression, where we regress Digital on the IV and all variables included in model (1). The coefficient of Postal*Internet is significantly positive, which is consistent with the above arguments. In addition, the Cragg-Donald Wald F-statistic of the weak instrumental variable test is 41.814, which is higher than 10, indicating that the instrumental variable selected is not a weak instrumental variable.

In the second stage regression, we use the predicted value of the degree of corporate digital transformation (Digital_Hat) derived from the first stage model to replace Digital. The coefficient on Digital_Hat in Column (2) of Table 8 is significant and positive, suggesting that the association between corporate digital transformation and employee satisfaction holds after controlling for the potential omitted variable problem and reverse causality problem.

4.3.3 Entropy balancing method (EB)

Furthermore, we use the entropy balancing method (EB) to alleviate the self-selection bias caused by observable factors. We divide the samples into the treatment group and control group based on the implementation of corporate digital transformation. Next, we balance all the independent variables in model (1) on three moments for continuous separate variables (mean, variance, and skewness) and one moment for dummy separate variables (mean), using the default tolerance of 0.015. Panel A of Table 9 is the balance effect test of the entropy balancing method. There are significant differences between the mean value, variance, and skewness of variables between the treatment and control groups before balancing. After balancing, these differences become insignificant. Panel B of Table 9 shows the results after adopting the entropy balancing method, and the coefficient of the degree of corporate digital transformation (Digital) is positive and significant, indicating that our results remain robust after controlling for the potential endogeneity problem using the entropy balancing method.

4.3.4 Difference-in-differences (DID) approach

In addition, we adopt the difference-in-differences (DID) approach to further mitigate the potential endogeneity problem. As the first digital economy pilot policy in China, the big data pilot zone has contributed to corporate digital transformation by increasing digital subsidies, promoting the development of digital talent, and fostering a supportive digital environment. Therefore, we take this policy as a quasi-natural experiment. Specifically, in September 2015, Guizhou launched the construction of the first big data pilot zone. This initiative was expanded in October 2016 with the announcement of a second batch of zones, including Beijing, Tianjin, Hebei, the Pearl River Delta, Shanghai, Chongqing, Shenyang, Inner Mongolia, and Henan.

(2)Emp_Satisfactioni,t=β0 +β1Treati,t +β2Treati,t *Posti,t+γControlsi,t+Year+Industry+ε

In model (2), Post_i,t is a dummy variable which equals one if the year is after the implementation of the big data pilot zone policy, and zero otherwise. Treat_i,t is a dummy variable which equals one if a firm is located in a big data experimental zone, and zero otherwise. The definitions of other variables are consistent with model (1).

Table 10 presents the estimation results of model (2). The coefficient of the interaction term between Treat and Post is positive and significant, indicating that our results remain robust after using DID regression to mitigate the potential endogeneity problem.

4.4 Other robustness tests

4.4.1 Alternative measures of key variables

First, unlike prior literature that uses employee comprehensive rating scores from Glassdoor.com to measure employee satisfaction (Huang et al., 2015; Fu et al., 2020; Dube & Zhu, 2021; Farhadi & Nanda, 2021), this paper uses the text content of employee reviews from Kanzhun.com to construct the employee satisfaction measure. The reason is that the disclosure of the employee comprehensive rating scores in Kanzhun.com is covered in real-time. We can only get the scores at the data acquisition time (August 2020). To validate the effectiveness of our text-based measure, we use comprehensive rating scores (COMM) for their employers in 2020 as an alternative measure of employee satisfaction and conduct a robustness test based on the subsample of the year 2020. Results are shown in Column (1) of Table 11.

Second, we change the emotional tendency classification of employee reviews from a two-type classification of positive and negative to a three-type classification of positive, neutral, and negative. The alternative measure of employee satisfaction (Emp_Satisfactiontwo) is reconstructed as the proportion of positive reviews in the total number of reviews under this new classification, and the results are shown in Column (2) of Table 11.

Finally, following Zhao (2021), we define a dummy variable, Digital_Dummy, which equals one if a firm engages in corporate digital transformation during the sample period, and zero otherwise. Then, we replace Digital in model (1) with Digital_Dummy. The results are shown in Column (3) of Table 11. The results in Table 11 show that the coefficients of Digital in Columns (1) and (2) and Digital_Dummy in Column (3) are significantly positive, indicating that our results remain robust to using alternative measures of corporate digital transformation and employee satisfaction.

4.4.2 Alternative explanation

There is an alternative explanation for the positive relationship between corporate digital transformation and employee satisfaction. Specifically, employees who are not satisfied with corporate digital transformation may choose to resign voluntarily, resulting in a higher proportion of satisfied employees. In order to rule out this alternative explanation, we conduct the following two tests. First, we examine the relationship between corporate digital transformation and employee turnover rate (Turnover). If our results are driven by this alternative explanation, then corporate digital transformation should be significantly and positively associated with the employee turnover rate. Specifically, Turnover is defined as the absolute value of the percentage change in employee numbers from the previous period to the current period. If the number of employees in the current period is greater than that in the previous period, Turnover is set to 0. A higher value of Turnover indicates a higher employee turnover rate. Column (1) of Table 12 presents the results. The coefficient of Digital is positive but insignificant. Therefore, we do not find a significant relationship between corporate digital transformation and employee turnover rate, which is inconsistent with the expectation of the alternative explanation.

Second, we examine the relationship between corporate digital transformation and the satisfaction of departed employees. If our results are driven by this alternative explanation, corporate digital transformation should be significantly and negatively related to the satisfaction of the departed employees. Using Kanzhun.com, we can determine an employee's status by differentiating between those who have left and those who remain employed with the company. Therefore, using departed employee reviews, we construct a measure to capture departed employee satisfaction (Q_Emp_Satisfaction) and replace Emp_Satisfaction in model (1) with it. As shown in Column (2) of Table 12, the coefficient of the degree of corporate digital transformation is still significantly positive, indicating that departed employees also have higher satisfaction with their employers for firms with a higher degree of corporate digital transformation. This result is also inconsistent with the expectation of the alternative explanation. Taken together, we do not find empirical evidence supporting the alternative explanation and thus this alternative explanation is less likely to drive our results.

5. Conclusion

Using a sample of listed Chinese firms from 2014 to 2020, this study investigates the impact of corporate digital transformation on rank-and-file employee satisfaction. We construct a novel measure for employee satisfaction through sentiment analysis of employee reviews from Kanzhun.com, and find a significant positive relationship between corporate digital transformation and employee satisfaction. In addition, we document that the relationship between corporate digital transformation and employee satisfaction is more pronounced for firms with higher labor intensity and for state-owned firms. Furthermore, our results remain robust when we use several methods to mitigate potential endogeneity problems and conduct other robustness tests.

Our study contributes to the literature in the following important ways. First, our study contributes to the literature on the economic consequences of corporate digital transformation from the perspective of rank-and-file employees. Second, we add to a stream of research examining the determinants of employee satisfaction. Third, this paper contributes to the research on employee satisfaction by providing a novel empirical measure of employee satisfaction. Last, this study also provides practical implications. Our study suggests that corporate digital transformation enhances employee satisfaction, providing direct evidence for managers and regulators to promote corporate digital transformation. Through digital transformation, companies can not only improve operational efficiency but also foster employee satisfaction. This dual benefit underscores the importance of investing in corporate digital transformation for long-term success.

There are also some limitations to our study. One significant limitation is that our key variable of interest, corporate digital transformation, is constructed based on word frequency analysis. This approach may be influenced by variations in corporate disclosure practices and might not accurately capture the true extent of corporate digital transformation. This limitation is not only present in our research but also pervasive in many other studies that utilize similar methodologies. Therefore, our results should be interpreted with this caveat in mind.

Table 2

Descriptive statistics

Variable	N	Mean	Min	P25	P50	P75	Max	SD
Emp_Satisfaction	6,118	0.547	0.000	0.000	0.500	1.000	1.000	0.412
Digital	6,118	1.385	0.000	0.000	1.099	2.485	4.997	1.467
Size	6,118	22.522	20.160	21.505	22.300	23.309	26.610	1.391
Age	6,118	2.019	0.000	1.386	2.197	2.773	3.258	0.891
LEV	6,118	0.428	0.062	0.268	0.417	0.575	0.901	0.201
ROA	6,118	0.043	−0.294	0.017	0.042	0.074	0.209	0.064
NCF	6,118	0.051	−0.139	0.012	0.050	0.091	0.238	0.066
SOE	6,118	0.316	0.000	0.000	0.000	1.000	1.000	0.465
RD	6,118	0.055	0.000	0.018	0.040	0.071	0.282	0.056
LaborInt	6,118	0.357	0.258	0.330	0.358	0.383	0.454	0.040
AEP	6,118	1.397	0.488	0.904	1.199	1.640	4.848	0.751
HHI	6,118	0.131	0.027	0.054	0.089	0.148	0.793	0.135

Note(s): Table 2 provides descriptive statistics for the key variables used in the main tests. All variables are defined in Appendix. All continuous variables are winsorized at 1 and 99% levels

Source(s): Authors’ own work

Table 3

Correlation matrix

	Emp_Satisfaction	Digital	Size	Age	LEV	ROA	NCF	SOE	RD	LaborInt	AEP	HHI
Emp_Satisfaction		0.031**	0.024*	−0.003	−0.019	0.040***	0.011	0.012	−0.014	−0.023*	0.065***	−0.017
Digital	0.040***		−0.036***	0.003	−0.065***	−0.015	−0.061***	−0.143***	0.270***	0.039***	0.140***	−0.073***
Size	0.037***	−0.065***		0.573***	0.579***	−0.138***	0.067***	0.391***	−0.433***	0.544***	0.186***	0.208***
Age	−0.004	0.045***	0.528***		0.389***	−0.298***	−0.039***	0.478***	−0.314***	0.288***	0.115***	0.048***
LEV	−0.019	−0.086***	0.574***	0.377***		−0.455***	−0.193***	0.305***	−0.409***	0.297***	0.075***	0.203***
ROA	0.048***	−0.056***	−0.044***	−0.224***	−0.372***		0.480***	−0.216***	0.102***	0.045***	−0.013	−0.069***
NCF	0.017	−0.058***	0.062***	−0.046***	−0.212***	0.440***		−0.045***	−0.036***	0.217***	−0.014	0.033***
SOE	0.013	−0.143***	0.398***	0.433***	0.310***	−0.124***	−0.055***		−0.279***	0.190***	0.182***	0.093***
RD	0.009	0.323***	−0.364***	−0.234***	−0.382***	−0.024*	−0.042***	−0.230***		−0.203***	0.120***	−0.339***
LaborInt	−0.012	0.030**	0.527***	0.263***	0.274***	0.063***	0.220***	0.176***	−0.195***		−0.252***	0.088***
AEP	0.068***	0.106***	0.248***	0.104***	0.102***	−0.028**	−0.007	0.155***	0.104***	−0.259***		0.081***
HHI	−0.008	−0.059***	0.153***	0.014	0.075***	−0.011	0.061***	0.100***	−0.218***	0.079***	0.088***

Note(s): Table 3 presents correlations for the key variables used in the main tests. All variables are defined in Appendix. The numbers below the diagonal are the Pearson correlations, and numbers above the diagonal are Spearman correlations. *, **, *** indicate statistical significance at the 10 percent, 5 percent, and 1 percent levels, respectively

Source(s): Authors’ own work

Table 4

The association between corporate digital transformation and employee satisfaction

	(1)	(2)
	Emp_Satisfaction	Emp_Satisfaction
Digital	0.011**	0.011**
	(2.37)	(2.35)
Size		0.026***
		(3.49)
Age		−0.012
		(1.47)
LEV		−0.041
		(0.99)
ROA		0.258**
		(2.55)
NCF		0.002
		(0.02)
SOE		0.005
		(0.39)
RD		0.057
		(0.42)
LaborInt		−0.271
		(1.34)
AEP		0.024***
		(2.59)
HHI		−0.003
		(0.03)
constant	0.532***	0.031
	(64.24)	(0.24)
Adj. R2	0.013	0.019
Year effects	Y	Y
Industry effects	Y	Y
N	6,118	6,118

Note(s): Table 4 presents the results of the association between corporate digital transformation and employee satisfaction. All variables are defined in Appendix. Year and industry fixed effects are included. Robust standard errors are clustered at the firm level. T-statistics are in brackets below the coefficient. *, **, *** indicate statistical significance at the 10 percent, 5 percent, and 1 percent levels, respectively

Source(s): Authors’ own work

Table 5

Conditional on labor intensity

	(1)	(2)
	IFHLB = 0	IFHLB = 1
	Emp_Satisfaction	Emp_Satisfaction
Digital	0.005	0.034***
	(0.93)	(3.13)
Size	0.016*	0.047***
	(1.84)	(3.16)
Age	−0.008	−0.017
	(0.84)	(1.12)
LEV	−0.018	−0.073
	(0.37)	(0.90)
ROA	0.216*	0.352
	(1.93)	(1.51)
NCF	0.032	−0.051
	(0.28)	(0.27)
SOE	0.004	0.007
	(0.27)	(0.22)
RD	0.027	0.644
	(0.20)	(1.36)
LaborInt	−0.086	−0.688*
	(0.37)	(1.65)
AEP	0.029***	0.014
	(2.76)	(0.68)
HHI	−0.036	0.081
	(0.24)	(0.38)
constant	0.190	−0.330
	(1.26)	(1.35)
Difference in α1	p < 0.001
Adj. R2	0.019	0.022
Year effects	Y	Y
Industry effects	Y	Y
N	4,397	1,721

Note(s): Table 5 reports the effect of labor intensity on the association between corporate digital transformation and employee satisfaction. Column (1) reports the results of the non-labor-intensive subsample. Column (2) reports the results of the labor-intensive subsample. All variables are defined in Appendix. Year and industry fixed effects are included. Robust standard errors are clustered at the firm level. T-statistics are in brackets below the coefficient. *, **, *** indicate statistical significance at the 10, 5, and 1% levels, respectively

Source(s): Authors’ own work

Table 6

Conditional on state ownership

	(1)	(2)
	SOE = 0	SOE = 1
	Emp_Satisfaction	Emp_Satisfaction
Digital	0.006	0.022**
	(1.12)	(2.48)
Size	0.017*	0.037***
	(1.80)	(2.82)
Age	−0.010	−0.015
	(1.01)	(0.92)
LEV	0.021	−0.125*
	(0.41)	(1.69)
ROA	0.242**	0.264
	(2.08)	(1.17)
NCF	0.078	−0.138
	(0.70)	(0.68)
RD	0.057	−0.033
	(0.36)	(0.12)
LaborInt	−0.323	−0.095
	(1.38)	(0.22)
AEP	0.032***	0.020
	(2.76)	(1.21)
HHI	−0.036	0.031
	(0.24)	(0.14)
constant	0.212	−0.239
	(1.23)	(1.16)
Difference in α1	p < 0.001
Adj. R2	0.019	0.022
Year effects	Y	Y
Industry effects	Y	Y
N	4,180	1,929

Note(s): Table 6 reports the effect of state ownership on the association between corporate digital transformation and employee satisfaction. Column (1) reports the results of the non-SOE subsample. Column (2) reports the results of the SOE subsample. All variables are defined in Appendix. Year and industry fixed effects are included. Robust standard errors are clustered at the firm level. T-statistics are in brackets below the coefficient. *, **, *** indicate statistical significance at the 10, 5, and 1% levels, respectively

Source(s): Authors’ own work

Table 7

Endogeneity tests: Heckman two-stage regression

(1)		(2)
First stage		Second stage
	IFrev		Emp_Satisfaction
Int_pen	0.002*	Digital	0.009**
	(1.77)		(1.98)
Size	0.173***	Size	0.014
	(12.98)		(1.12)
Age	−0.098***	Age	−0.005
	(7.26)		(0.46)
LEV	0.248***	LEV	−0.058
	(3.50)		(1.34)
ROA	0.722***	ROA	0.206*
	(3.76)		(1.85)
NCF	−0.129	NCF	0.009
	(0.73)		(0.09)
SOE	−0.065**	SOE	0.009
	(2.42)		(0.63)
RD	4.340***	RD	−0.213
	(14.59)		(0.78)
LaborInt	6.351***	LaborInt	−0.667
	(16.84)		(1.53)
AEP	0.241***	AEP	0.010
	(12.75)		(0.62)
HHI	−0.063	HHI	0.005
	(0.30)		(0.04)
		imr	−0.103
			(1.06)
constant	−6.993***	constant	0.566
	(23.06)		(1.10)
Pseudo R2	0.1677	Adj. R2	0.019
Year effects	Y	Year effects	Y
Industry effects	Y	Industry effects	Y
N	20,872	N	6,118

Note(s): Table 7 shows the results of Heckman two-stage regression. Column (1) is the estimation results of the first stage; Column (2) is the estimation results of the second stage. All variables are defined in Appendix. Year and industry fixed effects are included. Robust standard errors are clustered at the firm level. T-statistics are in brackets below the coefficient. *, **, *** indicate statistical significance at the 10, 5, and 1% levels, respectively

Source(s): Authors’ own work

Table 8

Endogeneity tests: Instrumental variable two-stage least squares (IV-2SLS) approach

	(1)	(2)
	First stage	Second stage
	Digital	Emp_Satisfaction
Digital_Hat		0.110*
		(1.804)
Postal*Internet	0.0004***
	(3.91)
Size	0.053	0.019**
	(1.61)	(2.143)
Age	0.334***	−0.044**
	(10.76)	(−1.997)
LEV	−0.226	−0.043
	(1.40)	(−0.928)
ROA	−0.708**	0.283**
	(2.09)	(2.383)
NCF	−0.849***	0.117
	(2.66)	(1.001)
SOE	−0.389***	0.045
	(6.13)	(1.504)
RD	2.162***	−0.202
	(3.61)	(−0.967)
LaborInt	2.338**	−0.398
	(2.20)	(−1.471)
AEP	0.045	0.023**
	(1.20)	(2.142)
HHI	0.058	−0.018
	(0.22)	(−0.141)
constant	−2.456	0.310
	(4.65)***	(1.443)
Adj. R2	0.444	0.091
Year effects	Y	Y
Industry effects	Y	Y
N	5,654	5,654
Cragg-Donald Wald F	41.814

Note(s): Table 8 reports the regression results of the IV-2SLS approach. Column (1) presents the first-stage regression and Column (2) presents the results of the second stage. All variables are defined in Appendix. Year and industry fixed effects are included. Robust standard errors are clustered at the firm level. T-statistics are in brackets below the coefficient. *, **, *** indicate statistical significance at the 10, 5, and 1% levels, respectively

Source(s): Authors’ own work

Table 9

Endogeneity tests: entropy balancing method

Panel A: The balance effect test of the entropy balancing method
Variable	Treatment			The control group before the weighted entropy balance method			The control group after the weighted entropy balance method
	Mean	Variance	Skewness	Mean	Variance	Skewness	Mean	Variance	Skewness
Size	22.530	1.832	0.790	22.520	2.092	0.725	22.530	1.832	0.791
Age	2.083	0.569	−0.442	1.923	1.112	−0.593	2.083	0.570	−0.442
LEV	0.424	0.038	0.233	0.434	0.044	0.241	0.424	0.038	0.233
ROA	0.040	0.004	−1.904	0.047	0.004	−1.148	0.040	0.004	−1.904
NCF	0.049	0.004	0.129	0.054	0.005	−0.122	0.049	0.004	0.129
SOE	0.271	0.198	1.032	0.383	0.236	0.481	0.271	0.198	1.032
RD	0.063	0.004	1.629	0.043	0.002	2.256	0.063	0.004	1.629
LaborInt	0.359	0.002	0.016	0.353	0.002	−0.050	0.359	0.002	0.016
AEP	1.420	0.533	1.937	1.362	0.608	2.141	1.420	0.533	1.936
HHI	0.127	0.017	2.871	0.137	0.019	2.547	0.127	0.017	2.871

Panel B: The regression results after adopting the entropy balancing method
	Emp_Satisfaction
Digital	0.011**
	(2.35)
Size	0.026***
	(3.49)
Age	−0.012
	(1.47)
LEV	−0.041
	(0.99)
ROA	0.258**
	(2.55)
NCF	0.002
	(0.02)
SOE	0.005
	(0.39)
RD	0.057
	(0.42)
LaborInt	−0.271
	(1.34)
AEP	0.024***
	(2.59)
HHI	−0.003
	(0.03)
constant	0.031
	(0.24)
Adj. R2	0.019
Year effects	Y
Industry effects	Y
N	6,118

Note(s): Table 9 reports the results of using the entropy balancing method to control for the potential endogeneity concern. Panel A presents the balance effect test of the entropy balancing method. Panel B presents the regression results after adopting the entropy balancing method. All variables are defined in Appendix. Year and industry fixed effects are included. Robust standard errors are clustered at the firm level. T-statistics are in brackets below the coefficient. *, **, *** indicate statistical significance at the 10, 5, and 1% levels, respectively

Source(s): Authors’ own work

Table 10

Endogeneity tests: difference-in-differences model

	Emp_Satisfaction
Treat	−0.012
	(0.61)
Treat*Post	0.040*
	(1.69)
Size	0.027***
	(3.52)
Age	−0.008
	(0.98)
LEV	−0.042
	(1.01)
ROA	0.261***
	(2.58)
NCF	−0.007
	(0.07)
SOE	0.001
	(0.10)
RD	0.076
	(0.57)
LaborInt	−0.266
	(1.30)
AEP	0.022**
	(2.37)
HHI	−0.001
	(0.01)
constant	0.024
	(0.18)
Adj. R2	0.019
Year effects	Y
Industry effects	Y
N	6,118

Note(s): Table 10 shows the results of the DID model. All variables are defined in Appendix. Year and industry fixed effects are included. Robust standard errors are clustered at the firm level. T-statistics are in brackets below the coefficient. *, **, *** indicate statistical significance at the 10, 5, and 1% levels, respectively

Source(s): Authors’ own work

Table 11

Other robustness tests: Alternative measures of key variables

	(1)	(2)	(3)
	COMM	Emp_Satisfactiontwo	Emp_Satisfaction
Digital	0.044*	0.008*
	(1.95)	(1.86)
Digital_Dummy			0.026**
			(2.02)
Size	0.055*	0.024***	0.026***
	(1.86)	(3.33)	(3.48)
Age	0.004	−0.003	−0.011
	(0.12)	(0.35)	(1.40)
LEV	−0.254	−0.074*	−0.041
	(1.39)	(1.88)	(0.99)
ROA	−0.184	0.203**	0.257**
	(0.42)	(2.08)	(2.53)
NCF	0.767*	0.035	−0.003
	(1.70)	(0.36)	(0.03)
SOE	0.102*	−0.031**	0.005
	(1.89)	(2.34)	(0.35)
RD	0.275	−0.038	0.068
	(0.60)	(0.31)	(0.51)
LaborInt	0.695	−0.223	−0.262
	(0.76)	(1.14)	(1.29)
AEP	0.289***	0.018*	0.025***
	(4.42)	(1.82)	(2.66)
HHI		−0.019	−0.002
		(0.16)	(0.02)
constant	−1.777**	−0.065	0.027
	(2.35)	(0.53)	(0.21)
Adj. R2	0.189	0.023	0.019
Year effects	N	Y	Y
Industry effects	Y	Y	Y
N	444	6,118	6,118

Note(s): Table 11 reports the results of using alternative measures of corporate digital transformation and employee satisfaction. Column (1) reports the results using COMM as the alternative measure of employee satisfaction. Column (2) reports the results using Emp_Satisfactiontwo as the alternative measure of employee satisfaction. Column (3) reports the results using Digital_Dummy as the alternative measure of corporate digital transformation. All variables are defined in Appendix. Year and industry fixed effects are included. Robust standard errors are clustered at the firm level. T-statistics are in brackets below the coefficient. *, **, *** indicate statistical significance at the 10, 5, and 1% levels, respectively

Source(s): Authors’ own work

Table 12

Other robustness tests: Alternative explanation

	(1)	(2)
	Turnover	Q_Emp_Satisfaction
Digital	0.001	0.008**
	(1.52)	(2.10)
Size	0.006***	0.019***
	(3.79)	(2.90)
Age	−0.011***	−0.002
	(7.37)	(0.29)
LEV	0.001	−0.023
	(0.11)	(0.65)
ROA	0.329***	0.099
	(10.34)	(1.18)
NCF	−0.037	−0.067
	(1.60)	(0.83)
SOE	0.010***	0.016
	(3.59)	(1.30)
RD	0.096***	−0.072
	(3.27)	(0.61)
LaborInt	0.183***	−0.152
	(4.01)	(0.83)
AEP	−0.019***	0.025*
	(5.81)	(1.74)
HHI	0.021	−0.221**
	(0.98)	(2.08)
constant	0.210***	−0.089
	(7.89)	(0.53)
Adj. R2	0.146	0.010
Year effects	Y	Y
Industry effects	Y	Y
N	6,118	7,531

Note(s): Table 12 reports the results related to the alternative explanation. Column (1) reports the results of the relationship between corporate digital transformation and employee turnover rate. Column (2) reports the results of the relationship between digital transformation and the satisfaction of departed employees. All variables are defined in Appendix. Year and industry fixed effects are included. Robust standard errors are clustered at the firm level. T-statistics are in brackets below the coefficient. *, **, *** indicate statistical significance at the 10, 5, and 1% levels, respectively

Source(s): Authors’ own work

Table A1

Variable definitions

Variable	Variable definitions
Emp_Satisfaction	employee satisfaction, defined as the proportion of positive reviews in each firm-year from Kanzhun.com
Digital	the degree of corporate digital transformation, defined as the natural logarithm of one plus the total number of word frequency related to corporate digital transformation in the annual report one-year lagged
Size	firm size, defined as the natural logarithm of total assets
Age	firm age, defined as the natural logarithm of one plus the number of years of listing
LEV	leverage, defined as the ratio of total liabilities to total assets
ROA	return on assets, defined as the net profits scaled by total assets
NCF	net cash flow, defined as the ratio of net cash flow from operations to total assets at the end of the year
SOE	property right, which equals one if a firm is a state-owned enterprise, and zero otherwise
RD	R&D investment is defined as the ratio of total R&D spending to operating revenue
LaborInt	labor intensity, defined as the ratio of the natural logarithm of the number of employees to the natural logarithm of total assets
AEP	average employee pay, defined as the natural logarithm of (cash paid to and for employees – total executive pay)/(total number of employees – total executive number)
HHI	the degree of market competition, which is measured by the Herfindahl-Hirschman index, is defined as the sum of the squares of the ratio of each firm sales to the total sales of all the firms in the industry
IFrev	whether a company appears on Kanzhun.com, which equals one if a firm has employee reviews in Kanzhun.com, and zero otherwise
Int_pen	the internet penetration rate, defined as the proportion of mobile internet users to the resident population of each province
Postal	number of post offices, defined as the number of post offices per million people in each prefecture-level city in 1984
Internet	Internet users, defined as the one-year lagged value of the number of Internet users in China (10 million)
Digital_Hat	the predicted value of the degree of corporate digital transformation, defined as the predicted value of corporate digital transformation obtained through the first stage of the IV-2SLS approach
Post	a dummy variable, which equals one if the year is after the implementation of the big data pilot zone policy, and zero otherwise
Treat	a dummy variable, which equals one if a firm is located in a big data experimental zone, and zero otherwise
COMM	the comprehensive rating score of employees, defined as the employee comprehensive rating scores for their employers from Kanzhun.com in 2020
Emp_Satisfactiontwo	the alternative measure of employee satisfaction, defined as the proportion of positive reviews in the total number of comments under a three-type classification
Digital_Dummy	whether the company undertakes digital transformation or not, which equals one if the total word frequency related to corporate digital transformation in the annual report one-year lagged is not equal to zero, and zero otherwise
Turnover	employee turnover rate, defined as the absolute value of the percentage change in employee numbers from the previous period to the current period, if the number of employees in the current period is greater than the previous period, Turnover is regarded as 0
Q_Emp_Satisfaction	departed employee satisfaction, defined as the percentage of departed employees' positive reviews about the company to the total number of reviews
IFHLB	whether the firm belongs to a labor-intensive industry, which equals one if a company belongs to such an industry that the average labor intensity is greater than the median of all industries, and zero otherwise

Source(s): Authors’ own work

Appendix

Table 1

Sample selection

Criteria	Firm-year observations
All firm-year observations from 2014 to 2020, excluding firms in financial industries	21,386
Less observations with missing data from CSMAR or Wind for constructing control variables	(410)
Less observations with firms whose names do not appear in Kanzhun.com	(14,835)
Total firm-year observations used in the main tests (total unique firms: 2,120)	6,118

Note(s): Table 1 presents the sample selection procedure of this paper

Source(s): Authors’ own work