Can Process Digitization Improve Firm Innovation Performance? Process Digitization as Job Resources and Demands

: Process digitization as a means to enhance innovation performance has garnered increasing attention from researchers and firms. Digital-driven innovation helps businesses achieve sustainable development. Following the job demands–resources model, we argue that process digitization contributes to job resources and job demands simultaneously, both of which are related to innovation performance. Process digitization offers additional job resources that contribute to enhanced work engagement and foster innovation performance at lower levels, whereas increasing job demands results in exhaustion and hinders innovation performance at higher levels. Therefore, we propose that firms with moderate levels of process digitization are more likely to have higher innovation performance. We further propose that employee training and pay can buffer the negative impact of process digitization on innovation performance. Training serves as an ex ante measure that enables knowledge-based employees to effectively respond to job demands without becoming exhausted. Conversely, pay serves as an ex post measure that compensates for resource depletion caused by excessive job demands, thereby alleviating the adverse effect of exhaustion on innovation performance. Our empirical results based on data from 3823 Chinese firms strongly support our hypotheses.


Introduction
In recent years, a growing number of firms have adopted both established technologies (e.g., cloud computing) and emerging digital technologies (e.g., artificial intelligence) to facilitate competitive positioning and pursue strategic objectives [1].The term process digitization refers to incorporating these digital technologies into a firm's operational processes (e.g., manufacturing process, marketing process, and administration process) [2].Process digitization denotes the transition from manual to digital means of conducting business activities [3], which aligns with the concept of digitization proposed by Slywotzky and Morrison [4] as "a disruptive, creative force that is revolutionizing how people work, play, communicate, buy, sell, and live."Through process digitization, firms can enhance operational efficiency and quality [3], improve process safety [5], and foster innovation performance [6].Furthermore, process digitization can also serve as a beneficial tool for businesses to gain sustainable competitiveness.Unlike capital and labor, emerging technologies represented by digitalization have improved the operational models of enterprises and built entirely new management systems.These effects not only have short-term impacts on innovation but also long-term implications, which makes digitalization a sustainable means of enhancing a firm's competitiveness.
The limited research on the effect of process digitization on innovation performance focuses on the benefits of process digitization in optimizing and automating internal operational processes, enhancing the efficiency of decision making, and consequently improving innovation efficiency and performance within the firm.Andriole [7] argues that process digitization aids firms in effectively and efficiently gathering information and data from diverse sources, facilitating the implementation of pertinent management activities such as monitoring and controlling, thereby optimizing and automating internal operational processes, and consequently, improving innovation performance.
The present study falls within this category, investigating the influence of process digitization on firm innovation performance by proposing an inverted U-shaped relationship between process digitization and firm innovation performance.We adopt the job demands-resources (JD-R) [8] model and posit that process digitization can furnish firm research personnel with additional job resources crucial for fostering innovation.These job resources encompass enhanced information sharing [9], streamlined innovation synergy [10], and refined perception of innovation [11], all of which facilitate job engagement, and consequently, superior innovation performance.However, when process digitization surpasses a certain threshold, it offers managers a powerful tool for monitoring and controlling employees [7], creating an excessive burden on job demands.These heightened job demands can erode employees' sense of autonomy [12] and deplete their psychological resources [13], culminating in exhaustion that hampers innovation performance.
Following the buffering hypothesis of the JD-R model, we further theorize that the effects of process digitization may differ when employees are provided with training and high pay.We argue that training and pay, as job resources, can buffer the negative impact of excessive job demands caused by a high level of process digitization on exhaustion, consequently resulting in a flattened inverted U-shaped curve.
Notably, the moderating mechanisms of training and pay differ.Training primarily serves as an ex ante measure, enhancing employees' innovative abilities and enabling them to effectively address heightened innovation demands, thereby preventing exhaustion even when facing high job demands.Conversely, pay predominantly serves as an ex post measure, compensating employees for depleted work resources and mitigating potential innovative performance declines following exhaustion.
Using data on 16,321 observations from 3823 publicly traded Chinese firms from 2014 to 2020, Our study makes several contributions to the literature, and our empirical results convincingly support our predictions.First, process digitization has emerged as a crucial research topic in the digitalization era.Despite firms' increasing adoption of digital technologies to enhance innovation performance, the actual effects of process digitization on innovation performance are more complicated than prior studies have suggested.We contribute to the literature by developing a conceptual model that supports the dual role of process digitization with regard to both job resources and demands.Specifically, we argue that an excessively low or high level of process digitization can impede innovation performance.Second, this study elucidates how ex ante (training) and ex post (pay) measures may buffer against the adverse impact of process digitization on innovation performance.

Job Demands-Resources (JD-R) Model
The JD-R model has two main hypotheses: the dual pathways hypothesis and the buffering hypothesis.The dual pathways hypothesis argues that job demands and resources drive two distinct processes: impairment and motivation [8].Impairment occurs when job demands persistently exceed job resources, resulting in a constant depletion of employees' energy during work processes.This can eventually lead to exhaustion or burnout, which can have detrimental effects on both employee well-being (e.g., health problems) and organizational outcomes (e.g., job performance).Conversely, motivation occurs when abundant job resources exist (rather than low job demands), and will improve employee work engagement, resulting in beneficial effects such as high job performance and low turnover intention.The buffering hypothesis argues that various job resources can buffer against the impact of job demands on strain [8,14,15].For example, a high level of autonomy or social support can attenuate the positive relationship between job demands and exhaustion [14].

Digitization and Innovation Performance
In the digitalization era, firms face increasingly intense competition and must continuously innovate to enhance their core competitiveness.Only through persistent improvement in innovation performance can firms survive and achieve sustainable development [16].Innovation performance can be categorized into three dimensions: organizational, group, and individual.Among these dimensions, individual innovation performance is widely recognized as the primary driver of overall innovation performance because employees are the main source of innovation [17].Indeed, every innovation originating from an organization or a group is essentially the result of individual efforts.Improving and optimizing individual innovation capabilities can consequently enhance the overall improvement of team and organizational innovation performance [17].Therefore, we have chosen individuals (i.e., knowledge-based employees) as the focal point of our research to delve into the intricacies of firm innovation performance from a micro-level perspective.This approach allows us to study the micro-foundation of corporate innovation performance and investigate why and how an individual's intrinsic psychological factors affect the overall innovation performance of the firm [17].In the context of this study, we define innovation performance as outcome-oriented, in that we focus on the tangible and quantifiable results that arise from employees' novel ideas [18].
Following the JD-R model, we argue that firm process digitization can provide knowledge-based employees with job resources that are indispensable for fostering innovation performance.First, process digitization enhances information sharing between knowledge-based employees.Specifically, process digitization empowers knowledgebased employees to utilize digital technologies for standardized and modular encoding and decoding of information, thereby facilitating the transmission, sharing, reception, and comprehension of information [9,19].Second, process digitization facilitates innovation synergy by enabling managers to effectively utilize digital technologies for comprehensive oversight and management of innovation projects, as well as rational task allocation among research and development personnel [10].Third, process digitization refines the perceptions of innovation among knowledge-based employees.Process digitization empowers knowledge-based employees to efficiently access innovative resources, such as knowledge and information, utilizing digital technology and intelligent hardware and software devices [20].Simultaneously, digital tools enable knowledge-based employees to overcome temporal and spatial constraints in their work [21], thereby providing them additional opportunities to invest in innovative research.
However, when process digitization surpasses a certain threshold, excessive job demands (e.g., managerial monitoring and control) may lead to employee exhaustion, thereby hampering innovation performance.First, process digitalization can lead to increased job demands due to the standardization of work methods and quantification of work outcomes, especially in the digital era when external factors such as customer preferences often dominate the quantitative benchmarks for work [22].This may lead to a phenomenon known as "digital Taylorism" and consequently escalate job demands [23].Second, under monitoring, whether employee performance meets the established standards becomes evident [12,24].Consequently, knowledge-based employees may need to pretend to be actively engaged in their work [12].This can lead to increased consumption of work resources (psychological resources), resulting in job burnout and reduced innovation performance.Third, the utilization of digital tools in management practices may result in centralization, thereby diminishing employees' autonomy and eroding their sense of control [13].This can significantly undermine employees' sense of belonging, instill job and future insecurity, and engender psychological resistance that hampers innovation performance [23].
In summary, two competing forces exist between firm process digitization and innovation performance.Process digitization is positively associated with innovation performance (i.e., the motivational process caused by additional tangible and intangible job resources provided by process digitization); however, it is also negatively associated with innovation performance (i.e., the impairment process caused by excessive job demands resulting from process digitization).Therefore, we propose an inverted U-shaped relationship between process digitization and innovation performance: innovation performance first increases as process digitization increases but then declines as process digitization increases further.
Hypothesis 1.The relationship between firm process digitization and innovation performance has an inverted U-shaped.

Moderating Effect of Pay
Financial incentives, particularly direct monetary incentives such as fixed or variable pay, are prevalent in organizations worldwide as a means to allure, retain, and incentivize employees [25,26].In the JD-R model, pay is posited to be a prospective determinant of work engagement and job resources "located at the level of the organization" [27,28].Although process digitization has increased job demands for knowledge-based employees, we expect this negative impact could be alleviated when employees perceive an increase in pay.Consequently, their work motivation, job satisfaction, and outcome expectancy will become enhanced.
First, higher pay will increase employees' work motivation and mitigate the fatigue associated with standardized processes.Providing a higher salary incentivizes individuals to willingly exert effort in their work tasks, thereby buffering the adverse effects of excessive job demands resulting from process digitization.Second, higher pay allows knowledgebased employees to use alternative means to compensate for the job resources consumed due to excessive job demands.For example, they can improve their quality of life or afford a better education for their children.Third, financial rewards engender a positive outcome expectancy when employees perceive better prospects.This assuages feelings of insecurity arising from the potential loss of work autonomy coming with process digitization.Thus, the following hypothesis is proposed: Hypothesis 2. Pay level flattens the curvilinear relationship (an inverted U-shape) between firm process digitization and innovation performance.

Moderating Effect of Training
The effects of training practices on organizational innovative performance have been widely discussed.Innovation scholars have highlighted the role of training in creating a climate for sharing ideas and information, promoting more communication among employees, and improving individual ability, which relieves the threatening impact of process digitization on employees and enhances firm innovation performance.
First, by providing training for employees, firms can create a climate of continuous learning, thereby amplifying the enhancement of information sharing resulting from process digitization.Training promotes effective employee adaption to new digital work environments and employee willingness to share information with other knowledge-based workers [29].Second, training programs provide opportunities for employees from various departments to collaborate, thereby enhancing their psychological identification and sense of belonging to the firm [30] and fostering a heightened level of innovation synergy following process digitization.Third, training equips knowledge-based employees with the necessary skills and knowledge to foster innovation, thereby mitigating the risk of job demands leading to exhaustion and impeding innovative performance.Training motivates employees to improve the status quo instead of passively accepting it [31].Therefore, knowledge-based employees feel an enhanced sense of control, which reduces the likelihood of exhaustion resulting from additional job demands arising from process digitization.Thus, the following hypothesis is proposed:

Hypothesis 3. Training flattens the curvilinear relationship (an inverted U-shape) between firm process digitization and innovation performance.
A conceptual model of this study is illustrated in Figure 1.
motivates employees to improve the status quo instead of passively accepting it [31].Therefore, knowledge-based employees feel an enhanced sense of control, which reduces the likelihood of exhaustion resulting from additional job demands arising from process digitization.Thus, the following hypothesis is proposed:

Hypothesis 3. Training flattens the curvilinear relationship (an inverted U-shape) between firm process digitization and innovation performance.
A conceptual model of this study is illustrated in Figure 1.

Data Description
We test our hypotheses by examining panel data for 3823 listed firms with 16,321 observations from 2014 to 2020 in China.Following previous procedures, first, we exclude the specially treated firm samples with business sustainability problems.Second, we delete samples with missing data.Third, for continuous variables, we winsorize both ends at the 1% level to help eliminate the influence of outliers on the regression analysis.
The rationale behind choosing listed firms in China for this study is that the application of process digitization is widely proposed and implemented to help employees improve work efficiency in Chinese firms and the success or failure depends on employee fit within the digitized work environment.For example, the most popular communication and management software used in Chinese listed firms, Ding Talk (latest version number: 5.1.11),has 400 million users, with a market share of nearly 50% in China.Users' opinions on Ding Talk are polarizing, as the average rating for Ding Talk across 2.75 million users is only 2.3 out of 5.
We use multiple data sources to avoid bias.The data on the dependent variable, innovation performance, are collected from the CNRDS.Data on the independent variable, process digitization, and moderator variables (i.The dataset used to estimate firm process digitization is the EDT Digital Application Scoring Database.It has an advantage over its predecessors, as EDT provides textual analysis information by examining all words that occur in announcements made by listed firms.As Loughran and McDonald [32,33] explain, textual analysis enables researchers to convert qualitative information into quantitative measures.Hence, EDT provides measures of real digitization applications in listed firms.

Model and Measurements
We use a fixed-effects model to estimate the influence of process digitization on innovation performance.Furthermore, we incorporate a one-year lag on the independent

Methodology 3.1. Data Description
We test our hypotheses by examining panel data for 3823 listed firms with 16,321 observations from 2014 to 2020 in China.Following previous procedures, first, we exclude the specially treated firm samples with business sustainability problems.Second, we delete samples with missing data.Third, for continuous variables, we winsorize both ends at the 1% level to help eliminate the influence of outliers on the regression analysis.
The rationale behind choosing listed firms in China for this study is that the application of process digitization is widely proposed and implemented to help employees improve work efficiency in Chinese firms and the success or failure depends on employee fit within the digitized work environment.For example, the most popular communication and management software used in Chinese listed firms, Ding Talk (latest version number: 5.1.11),has 400 million users, with a market share of nearly 50% in China.Users' opinions on Ding Talk are polarizing, as the average rating for Ding Talk across 2.75 million users is only 2.3 out of 5.
We use multiple data sources to avoid bias.The data on the dependent variable, innovation performance, are collected from the CNRDS.Data on the independent variable, process digitization, and moderator variables (i.The dataset used to estimate firm process digitization is the EDT Digital Application Scoring Database.It has an advantage over its predecessors, as EDT provides textual analysis information by examining all words that occur in announcements made by listed firms.As Loughran and McDonald [32,33] explain, textual analysis enables researchers to convert qualitative information into quantitative measures.Hence, EDT provides measures of real digitization applications in listed firms.

Model and Measurements
We use a fixed-effects model to estimate the influence of process digitization on innovation performance.Furthermore, we incorporate a one-year lag on the independent and moderator variables to address potential reverse causation.We centered all continuous independent variables using the grand mean before creating the interaction terms to reduce multicollinearity concerns.To reduce the potential influence of extreme values, the continuous variables are winsorized at the 1st and 99th percentiles.

Dependent Variable
We select a firm's annual number of new patent applications [34] to represent innovation performance, owing to the strict legal protection that prevents others from using a patented invention without permission for a limited period.Following previous studies, patent counts not only enable scholars to compare innovative performance between firms in terms of new technologies, but they also exhibit a strong correlation with other indicators of innovation, such as new businesses and new products.Furthermore, we use the number of total patents as alternative measures, which include utility model, appearance, and invention patents.This is discussed in the supplementary analyses in result part.

Independent Variable
To measure the degree of process digitization for a firm, following previous research [35], we use textual analysis to count the frequency of the application of process digitization keywords in the annual reports of listed companies.The keyword list constructed by CSMAR includes all forms of business process digitization: production process digitization (i.e., intelligent manufacturing, automated factories), marketing process digitization (i.e., intelligent customer service, intelligent marketing, digital marketing), sales and service process digitization (i.e., mobile payments, third-party payments, NFC payments), and communication and administration process digitization (i.e., human-computer interaction, social networks).We also centralize the process digitization variable and calculate its quadratic term.

Moderating Variables
Our first moderator is employee pay level.In prior research, data on pay levels were collected from personnel records of the host firm, which were measured by self-report categories since the firms refused to provide detailed salary data [36].For the self-report categories measure, even if employee participants report their pay accurately, if the host firm has several types of salary plans, self-reported pay could provide a misleading picture of the overall firm pay level.Considering this, we calculate employee pay levels as the salaries payable to employees published in corporate financial statements, divided by the total number of employees.
The second moderator variable is employee training.Following the resource-based approach, we assess employee training based on the financial data disclosed in firms' annual reports, including employee education fees, which are the direct costs of training.Moreover, we add the labor union expenditure as the extent to which the firm provided additional opportunities for employee training since labor unions regularly organize activities through which employees can share their work experience.

Control Variables
We control for a range of variables at both the top management team (TMT) and firm levels to mitigate potential alternative explanations.The TMT influences firm innovation activities [37]; therefore, we control for firm TMT characteristics, such as board size, TMT overseas experience, TMT average age, and TMT female ratio.In our study, we define the TMT as managers above the vice president level and their subordinate managers within a firm.We coded it as 1 if the firm has a TMT member with overseas experience and 0 otherwise.Managers' overseas experience reflects overseas educational or working experience.We also consider TMT average age and TMT female ratio, which are captured by the average age of managers in the TMT and the ratio of female managers in the TMT.We also control for board size, measured as the number of board members.
At the firm level, we control for firm size, measured as the logarithm of the number of employees; firm age, measured as the logarithmic number of years since a firm's initial public offering; and return on assets (ROA).We measure leverage using the proportion of short-term debt to total assets; and fixed assets, measured as the ratio of tangible assets to total assets.We control for occupational health protection as additional employee work environment-related variables, such as well-being and health, directly affect work efficiency and organizational performance [38].

Empirical Results
Table 1 reports the basic descriptive statistics and variable correlations.Firms have 70.86 patents on average, with a maximum of 1277 patents.The average number of invention patents is 30.19 and the maximum is 579, indicating that nearly half of the firms' patents are invention patents.The variable correlation results suggest that both process digitization and its square term are significantly and positively correlated to firm innovation performance.Furthermore, employee pay level and employee training are also significantly and negatively correlated with firm innovation performance.Moreover, the maximum variance inflation factor (VIF) is 7.80 and the average VIF index is 2.57, which is well below the threshold value of 10.This indicates that this study is not affected by multicollinearity.Table 2 reports the model results.Model 1 serves as the baseline and includes only control variables.In Model 2, we include the linear term of the independent variable and the two moderators.Model 3 adds the square term of process digitization to test the inverted U-shaped relationship between process digitization and innovation performance.Model 4 tests the moderating effect of employee pay level.Model 5 tests the moderating effect of employee training.Model 6 represents the full model.H1 proposes the existence of an inverted U-shaped relationship between process digitization and innovation performance.As observed in Model 3, the coefficient for process digitization is positive and significant (b = 16.234,p < 0.01), whereas that for its squared term is negative and significant (b = −12.274,p < 0.01).The results on the inverted U-shaped relationship are consistent across all models.To further validate the marginal effect of relatedness, we follow the established methodology employed in previous studies [39] and conduct a Utest using STATA.(version number: STATA 7)Specifically, we examine the steepness of the slope at both ends of the data range for relatedness (lower bound: slope = 0.164, p = 0.00; upper bound: slope = −0.435,p = 0.0234; Table 3).In addition, we check the location of the turning point of the inverted U-shaped relationship, finding that the results are all within the relatedness data range.The above results are shown in Table 3, and Figure 2 provides a graphical analysis of the inverted U-shaped relation.Thus, H1 is supported.
H2 predicts that the average employee pay level flattens the curvilinear relationship between process digitization and innovation performance.In Model 4, the coefficient of the interaction term squared process digitization ×employee pay level is significantly positive (b = 0.432, p < 0.01), which partially supports H2.Furthermore, we use figure analysis to examine the slopes of the curve shape.Based on the results in Table 2 for Model 4, we plot the relationship between process digitization and innovation performance for firms with high and low employee pay levels, respectively (Figure 3).As Figure 3 shows, the line exhibits a pronounced flattening trend at higher pay levels compared to lower ones, which supports H2.We then replicate this analysis to analyze H3.In Model 4, the coefficient of the interaction term squared process digitization × employee training is found to be significantly positive (b = 0.008, p < 0.01).The findings shown in Figure 4 further validate the flattening effect of employee training, as evidenced by a significantly flatter line observed at higher levels of employee training compared with that at lower levels.Thus, H3 is supported.

Robustness Tests
To ensure the robustness of our findings and mitigate potential alternative explanations, we conduct an instrumental variable test as a robustness check.We also use alternative measures to re-run our tests as additional analyses.
First, we instrument firms' process digitization degree using the Baidu index of digitization at the province level to eliminate firm-level unobserved heterogeneity and re-

Robustness Tests
To ensure the robustness of our findings and mitigate potential alternative explanations, we conduct an instrumental variable test as a robustness check.We also use alternative measures to re-run our tests as additional analyses.
First, we instrument firms' process digitization degree using the Baidu index of digitization at the province level to eliminate firm-level unobserved heterogeneity and re-

Robustness Tests
To ensure the robustness of our findings and mitigate potential alternative explanations, we conduct an instrumental variable test as a robustness check.We also use alternative measures to re-run our tests as additional analyses.
First, we instrument firms' process digitization degree using the Baidu index of digitization at the province level to eliminate firm-level unobserved heterogeneity and re-

Robustness Tests
To ensure the robustness of our findings and mitigate potential alternative explanations, we conduct an instrumental variable test as a robustness check.We also use alternative measures to re-run our tests as additional analyses.
First, we instrument firms' process digitization degree using the Baidu index of digitization at the province level to eliminate firm-level unobserved heterogeneity and reverse causality between process digitization and innovation performance.Following prior research [40], we use the province-level search volume of feature words related to the digital economy such as "digital economy" and "digitalization" in the Baidu index (Baidu index: Baidu is the world's largest Chinese-language search engine, one of the 10 largest global websites.The Baidu Index is a data analysis platform based on Baidu's massive netizen behavior data and is one of the most important statistical analysis platforms currently on the Internet.The Baidu Index can reflect keyword search volume, and keywords related to an amount of the retrieval keywords will also appear in our search index.It is available online: https://index.baidu.com/v2/index.html#/,accessed on 24 May 2024), crawled using R language from 2014 to 2020 as province-level instrumental variables.Firm process digitization will pique the interest of local Internet users; thus, the search index of local Internet users will be positively correlated with firm process digitization.However, Internet user searches have little impact on innovation activities within firms, which fulfills the exclusivity requirement of instrumental variables.In the first-stage regression, we observe that the coefficients on the instrument variable Baidu index are significantly different from 0 at the 1% level, and are further validated by an F-test.The second-stage results from our instrumental variable regressions, as presented in Table 4, align with our baseline estimates and alleviate concerns regarding endogeneity potentially influencing or distorting our findings.Second, we supplement the stability of the results by replacing the dependent variable with total patent counts.In the main analyses, we use the number of invention patents as the measurement of innovation performance to guarantee the competitive advantages of new patents.However, some utility model patents may be closely related to firm process digitization.For example, firms can apply for new utility model patents, which reflect a digital access control system to facilitate the employee attendance assessment process.As invention patents do not cover all possible cases, we use the total patent counts for supplementary analyses to provide a more complete picture of firms' innovation capability and performance.We conduct all analyses again using this new sample, and the results remain consistent with all significant effects unchanged (Table 5).

Conclusions
The study uses fixed-effects regression analysis to investigate the impact of process digitization on innovation performance with a panel dataset comprising 16,321 observations from 3823 Chinese publicly traded firms from 2014 to 2020.Furthermore, this research explores the moderating roles of average pay levels and employee education and training expenditure in the relationship between process digitization and innovation performance.Our key findings are as follows.First, the relationship between process digitization and innovation performance is U-shaped.Second, higher employee pay levels help mitigate the inverted U-shaped relationship between process digitalization and innovative performance.Third, increased expenditure on employee training also alleviates this inverted U-shaped relationship.

Discussion and Contribution
First, this study contributes to an understanding of the effects of process digitization on the innovation performance of firms.Drawing on the JD-R model theory, this study investigates the curvilinear relationship between process digitization and innovation performance from an employee perspective.The existing literature on digital transformation generally suggests a positive association between digitization and innovation performance, emphasizing its role in enhancing efficiency.However, this study expands on prior research by proposing that process digitization not only provides employees with valuable work resources but also imposes increased demands on them.Consequently, an inverted U-shaped relationship exists between process digitization and innovation performance.This study takes a micro individual-level perspective and reveals that the enhancement of innovation performance is not solely propelled by organizational-level factors such as external incentives or innovation environment, but also hinges on the intrinsic psychological motivation of employees.Indeed, as innovation is driven by individuals, the innovative performance of distinct organizations is inevitably subject to variation.Our research points out a new angle, suggesting that future studies on corporate innovation performance should comprehensively account for the micro-level contextual factors of the organization to ensure the external validity of their findings.
Second, this study further examines the influence of different employee work environment characteristics on the nonlinear relationship between process digitization and innovation performance while considering average employee pay levels and training expenditures as moderating factors.The findings reveal that higher average employee salaries and increased expenditure in employee training can complement and alleviate job demands arising from process digitization within organizations.
These results offer valuable insights for managers to understand employees' resource requirements following process digitization implementation.In practice, managers should be mindful that process digitization not only facilitates the innovation of knowledgebased employees, but also intensifies their pressure.Therefore, a higher level of process digitization is not always preferable, and managers need to determine the optimal degree of process digitization that aligns with the firm's actual status.In addition, managers ought to actively monitor the psychological well-being of knowledge-based employees and offer support as needed to enhance work engagement and prevent exhaustion.Lastly, managers must possess preventive (e.g., training) and remedial (e.g., pay) measures in order to strive for improved innovation performance of the firm.

Limitations and Further Research
First, this study lacks discussion on the underlying mechanisms through which process digitalization influences innovation performance.Previous research has examined crucial factors that affect innovation performance from an employee perspective, such as employee knowledge [41] and employee involvement [42].Future studies can develop a mediated model to reveal the internal mechanisms underlying the effects of process digitalization on innovation performance.
Second, further explorations are needed that apply diverse theories and more perspectives on process digitalization.This study adopts the JD-R model theory to investigate the impact of process digitalization from an employee point of view, that is, considering average salary level and training expenditure as moderating variables.Prior research has examined the influence of digitization transformation from several theoretical aspects, including organizational learning [43,44] and knowledge-based views [45].
Third, in addition to process digitalization, the impact of technological digitization and business digitization on firms' innovative behavior warrants further discussion.Firms' digitization applications can be classified into technological digitization (e.g., artificial intelligence, facial recognition) and business digitization (e.g., digital finance, virtual economy).This study only considers the influence of process digitalization on firms'
e., employee pay level and employee training expenditure) are from the most recent information in the China Listed Firm's Digital Transformation Research Database (EDT since 2010) and Enterprises' Contribution to Common Prosperity Research Database (CPE since 2011).The EDT and CPE databases have recently been cleaned and validated by CSMAR and are now available for public access on CSMAR's website.Data on control variables are collected from WIND.
e., employee pay level and employee training expenditure) are from the most recent information in the China Listed Firm's Digital Transformation Research Database (EDT since 2010) and Enterprises' Contribution to Common Prosperity Research Database (CPE since 2011).The EDT and CPE databases have recently been cleaned and validated by CSMAR and are now available for public access on CSMAR's website.Data on control variables are collected from WIND.

Figure 2 .
Figure 2. Inverted U-shaped relationship between process digitization and innovation performance.

Figure 3 .
Figure 3. Process digitization and innovation performance: the moderating effect of employee pay level.

Figure 4 .
Figure 4. Process digitization and innovation performance: the moderating effect of training expenditure.

Figure 2 .
Figure 2. Inverted U-shaped relationship between process digitization and innovation performance.

Figure 2 .
Figure 2. Inverted U-shaped relationship between process digitization and innovation performance.

Figure 3 .
Figure 3. Process digitization and innovation performance: the moderating effect of employee pay level.

Figure 4 .
Figure 4. Process digitization and innovation performance: the moderating effect of training expenditure.

Figure 3 .
Figure 3. Process digitization and innovation performance: the moderating effect of employee pay level.

Figure 2 .
Figure 2. Inverted U-shaped relationship between process digitization and innovation performance.

Figure 3 .
Figure 3. Process digitization and innovation performance: the moderating effect of employee pay level.

Figure 4 .
Figure 4. Process digitization and innovation performance: the moderating effect of training expenditure.

Figure 4 .
Figure 4. Process digitization and innovation performance: the moderating effect of training expenditure.

Table 1 .
Descriptive statistics and correlations.

Table 4 .
Instrumental variables test results.

Table 5 .
Alternative measure test results.