How Does the Digital Economy Promote a Culture of Business Innovation? A Study Based on Human Capital Allocation Perspective

Abstract

Human capital is the driving force of enterprise innovation. By clarifying the impact of the digital economy on enterprise innovation from the perspective of human capital allocation, we can understand the underlying mechanisms that enable high-quality development dividends on a more nuanced scale. This study incorporated the ‘Broadband China’ strategy to construct a quasi-natural experiment, empirically investigating the impact of digital economy development on micro-level enterprise innovation from a human capital perspective. The findings show that digital economy development can effectively reduce the demand gap and recruitment costs for research and development personnel and significantly improve their efficiency, thereby promoting corporate innovation. Heterogeneity tests show that the micro-level effects of the digital economy are more pronounced for firms with younger entrepreneurs, those registered in eastern Chinese cities, or those that are strongly influenced by Confucian culture. Further analysis shows that by streamlining these channels of human capital, the digital economy can not only boost the overall output of corporate innovation but also significantly improve its quality.

1. Introduction

Currently, developing the digital economy has become key for countries worldwide to achieve economic growth and cultivate new competitive advantages. China’s 19th Party Congress outlined a strategic plan to build a ‘digital China’ and called for the vigorous advancement of the digital economy to promote high-quality economic development. As enterprises are the cells of the national economy and the primary source of innovation and creativity, promoting the digital economy entails promoting enterprise innovation to build a foundation at the micro level, thus enabling high-quality macro-level development [1,2].

Research on the relationship between the digital economy and enterprise innovation is still in its infancy, with some studies focussing on data as a ‘resource’, suggesting that the digital economy is transforming enterprise innovation resources and the way those resources are connected, thereby significantly improving the ability of enterprises to adaptively innovate [3,4,5]. Other studies have analysed the ‘technological’ properties of data, arguing that the digital economy has improved or even reshaped the process frameworks and organisational mechanisms behind corporate innovation, which strongly contribute to firms’ self-growth and modular innovation capabilities [6,7,8]. However, most of these studies have examined the product and organisational domains, and few have examined the micro-level effects of integrating the digital economy and the labour market. This study addresses this gap by analysing and testing the mechanisms behind the impact of digital economy development on firm innovation from the perspective of human capital.

Human capital is an important driver of corporate innovation [9], requiring support from qualified and efficient professionals [10]. For many years, China’s ‘hyper-normal’ labour market frictions have limited the efficiency of human capital allocation, creating a fundamental constraint on business innovation [10] and even impeding steady economic growth. While it is difficult to provide a single definition of the digital economy, there is growing recognition of its essential characteristic of reducing market frictions by digitising economic activities and thereby promoting the efficient allocation and use of resources [1,11]. Can the digital economy also play a positive role in the labour market, and can its dynamic development improve the efficiency of internal and external human capital allocations in firms in order to promote corporate innovation? The answers to these questions can help deepen the research related to the micro-level mechanisms powering the digital economy and provide empirical evidence supporting the optimisation of corporate human resource management and innovation models through the digital economy.

To avoid endogeneity problems caused by reverse causality and omitted variables as much as possible, this study incorporated the ‘Broadband China’ strategy, constructing a quasi-natural experiment to examine the impact of digital economy development on enterprise innovation using a double-difference approach. For context, on 17 August 2013, China’s State Council released the implementation plan for the ‘Broadband China’ strategy, presenting the future development goals and paths for China’s broadband network. The deep development of the digital economy cannot be achieved without the support of broadband networks, and the improvement of network performance and service quality has direct effects on the digital economy. During the implementation of this strategy, 120 cities were selected in three batches in 2014, 2015, and 2016 as demonstration cities for broadband construction that significantly promoted the development of the digital economy. The selection of pilot cities was independent of the development status of local enterprises and, therefore, was relatively neutral to their innovation activities.

In sum, this study empirically examined the digital economy’s impact on enterprise innovation from the perspective of human capital allocation using the ‘Broadband China’ strategy as an exogenous policy shock. The additional potential contributions of this study are stated as follows: first, drawing on the existing literature, this study examines the digital economy’s impact on innovation at the enterprise level, enabling a more nuanced understanding of the mechanisms underlying the digital economy’s output of high-quality development dividends, thus enriching research at the micro level of the digital economy. Second, it offers a new interpretation of human capital characteristics and functions within enterprises in the digital economy context, systematically illustrating the impact of the digital economy’s development on the personnel demand gap and recruitment costs to fill enterprise research and development (R&D) roles. Third, this study introduces human capital into the relationship between the digital economy and enterprise innovation, constructs a model for the mechanism underlying the digital economy’s development and enterprise innovation from the perspective of human capital, and reveals the important role of human capital in this process of digitally empowered enterprise innovation.

The remainder of this paper is structured as follows: Section 2 theoretically analyses the relationship between the digital economy, internal and external human capital allocation of firms, and firm innovation and proposes relevant research hypotheses; Section 3 discusses the research design; Section 4 reports the results of empirical tests of the research hypotheses; Section 5 conducts heterogeneity tests and further analysis; and Section 6 presents the conclusion and recommendations.

2. Theoretical Analysis and Research Hypothesis

2.1. The Digital Economy, External Human Capital Allocation, and Corporate Innovation

The scarcity of human capital in competitive sectors coexisting with human capital redundancies in public and monopolistic sectors in China, along with the inefficient allocation of human capital between industry sectors, has for some time caused total factor productivity losses of up to 20% in the non-agricultural economy, while seriously undermining China’s performance in terms of innovation. Ref. [10] has shown that if China’s factor allocation efficiency could reach US levels, innovation performance could be boosted by 30–50%.

The adverse effects of inefficient external human capital allocation on enterprise innovation activities are mainly related to the supply and demand of knowledge-based human capital in the form of R&D personnel. On the one hand, firms find it difficult to recruit qualified R&D personnel in sufficient numbers, perpetuating the demand gap and hindering knowledge-related innovation efforts [9], thus directly constraining corporate innovation capabilities. On the other hand, this scenario increases recruitment costs for R&D personnel and forces firms to raise wages to attract talent, which may drain innovation budgets. Ref. [12] has shown that incomplete information in the labour market results in final wages, agreed upon between firms and job seekers, based on matching uncertainty and bargaining outcomes, that are generally higher than the ultimate labour productivity.

Technological progress is a driver of efficiency in human capital allocation and can efficiently handle continuous changes in labour supply and demand in terms of labour force structure, new employment trends, and labour mobility across industries [13]. Along with the development of emerging technologies, such as the internet, big data, and artificial intelligence (AI), the scale of China’s digital economy is growing, and the impact of digital economy development on the labour market is becoming increasingly prominent [14]. Specifically, at the micro level of enterprises, the digital economy can reduce the demand gap and recruitment costs for R&D personnel in the following ways: first, the application of digital technology and widespread internet use can reduce incomplete information in the labour market, weaken the spatial and temporal barriers between enterprises and workers, seamlessly link labour supply and demand information, and enable real-time and accurate matching, thereby significantly reducing the cost of recruiting R&D staff and increasing recruitment success rates. Second, the digital economy has made platform-based employment an important model with the availability of more flexible and diverse jobs. Formally employed R&D workers can then engage in related part-time activities, which, to a certain extent, reduces the demand gap for R&D personnel. Thus, the following research hypotheses are proposed from the perspective of external human capital allocation efficiency:

H1: 

The digital economy boosts business innovation by reducing the gap in enterprise demand for R&D staff.

H2: 

The digital economy facilitates business innovation by reducing enterprise recruitment costs for R&D staff.

2.2. The Digital Economy, Internal Human Capital Allocation, and Corporate Innovation

Rapid digital economy development is driving the continuous penetration of digital technologies into industries, thereby helping enterprises deploy generic, integrated, and interactive digital solutions in a planned and progressive manner, while ultimately reorganising and supporting the entire business process, that is, the digital transformation of the enterprise [15]. The impact of the digital economy on human capital allocation within firms mainly depends on the firm’s level of digital maturity. On the one hand, in the digital transformation process, the generation of ideas, new product development, product piloting and manufacturing, logistics, and sales may be improved. Innovation activities also shift from those that are experience-driven to those that are data-driven, thereby significantly increasing R&D staff efficiency [6]. On the other hand, digital technology penetration can substitute for low-skilled labour and increase reliance on high-skilled labour [16]. This optimisation of the internal human capital structure driven by digital transformation improves the ratio of highly qualified R&D personnel and the quality of intellectual capital, thus improving their overall efficiency and innovation capacity.

Furthermore, with the development of the digital economy and the continued digital transformation of enterprises, innovation platforms, such as knowledge-sharing and trading platforms, virtual crowdsourcing spaces, and digital makerspaces, are gradually becoming innovation hubs for many enterprises [17], and digital innovation ecosystems are quietly emerging. Open, collaborative, and cross-firm innovation within these systems has become the norm, not only accelerating knowledge codification and inter-firm exchanges [18] but also enhancing the knowledge-seeking (self-selection) capabilities of R&D personnel, their abilities to resolve conflicts over consistency and flexibility, and their collaboration capabilities [19], ultimately enhancing enterprise innovation. Finally, new technologies have a labour substitution effect [16], where the deep integration of digital technologies with the physical economy increases unemployment risks for low-skilled and low-knowledge employees who may be completely replaced by AI in the future, which in turn stimulates the learning and work motivation of R&D staff. Based on the above-noted analysis, from the perspective of internal human capital allocation efficiency, the following research hypothesis is proposed:

H3: 

The digital economy boosts business innovation by increasing the efficiency of R&D staff.

The theoretical model in this paper is shown in the following diagram (Figure 1).

3. Study Design

3.1. Sample Selection and Data Sources

During the implementation of the ‘Broadband China’ strategy, the relevant departments of the Chinese government selected a total of one-hundred twenty cities in three batches in 2014, 2015, and 2016 as demonstration cities for broadband construction; thus, this paper selects Chinese A-share listed companies from 2010 to 2019 as the initial sample. The list of ‘Broadband China’ demonstration cities used in this paper was obtained from the official website of the Ministry of Industry and Information Technology of China. The company-level technology innovation data were retrieved from the State Intellectual Property Office of China by company name, and the financial data were obtained from the China Stock Market and Accounting Research database. The city-level data were mainly obtained from the China City Statistical Yearbook and the China Regional Economic Statistical Yearbook, from which the city-level indicators of population and employment, as well as national accounting and financial status, were extracted and controlled. Given the study’s needs, the following were excluded: (1) companies in industries with little engagement vis-à-vis technological innovation, such as finance and insurance, residential services, and education, and (2) special treatment (ST) and * ST companies and observations with missing data. To avoid the effects of extreme values, we applied a top and bottom 1% tail shrinkage (Winsorisation) to all continuous variables at the firm level. To control for potential heteroskedasticity and serial correlation problems, the standard errors of all regression coefficients in this study were adjusted for heteroskedasticity and treated with firm-level clusters. These treatments yielded 16,976 firm-year observations.

3.2. Model Design and Variable Definition

The following double-difference model was constructed to test the hypotheses. Considering the temporal differences in the selection of the demonstration cities for ‘Broadband China’, we designed a multi-period staggered quasi-natural experiment; therefore, a two-way fixed effects model was used, controlling for time fixed effects μ_t and firm fixed effects γ_i.

$$ST{P}_{i,t}={\alpha }_0+{\alpha }_1\times BC\_selecte{d}_{i,t}+{\displaystyle \sum Control{s}_{i,t}}+{\mu }_t+{\gamma }_i+{\epsilon }_{i,t}$$

(1)

$$Innovatio{n}_{i,t}={\alpha }_0+{\alpha }_1\times BC\_selecte{d}_{i,t}+{\displaystyle \sum Control{s}_{i,t}}+{\mu }_t+{\gamma }_i+{\epsilon }_{i,t}$$

(2)

$$Innovatio{n}_{i,t}={\alpha }_0+{\alpha }_1\times BC\_selecte{d}_{i,t}+{\alpha }_2\times ST{P}_{i,t}+{\displaystyle \sum Control{s}_{i,t}}+{\mu }_t+{\gamma }_i+{\epsilon }_{i,t}$$

(3)

where μ_t and γ_i represent time and firm fixed effects, respectively; ε_i,t is a random error term; and Controls represents a set of control variables from the firm and city levels. The dependent variable STP includes three variations, namely, the demand gap for R&D personnel (STP_1), the recruitment cost of R&D personnel (STP_2), and the efficiency of R&D personnel (STP_3), which together form a regression model of the number of R&D personnel with firm size, age, nature, and industry type. The residual term reflects the difference between the number of R&D personnel demanded and the actual number of R&D personnel, which is used to measure the demand gap for R&D personnel (STP_1). Higher talent recruitment costs lead to higher wages, thereby draining innovation budgets [20]. Consequently, this study uses the share of R&D personnel compensation in the total R&D expenditure to measure R&D personnel recruitment costs (STP_2). The ratio of the number of patent applications for inventions, utility models, and designs to the total number of R&D personnel is used to measure R&D personnel efficiency (STP_3). Innovation is a dependent corporate innovation variable that reflects the outcome of innovation-related activities and is measured by adding the number of patents granted in the three categories by one and taking the natural logarithm.

BC_selected is an independent dummy variable that takes the value of 1 if the city in which the company is registered is selected as a ‘Broadband China’ demonstration city by the end of the year, and 0 otherwise. BC_selected is used as a proxy variable for the level of digital economy development, and its coefficients are estimated as double-differenced results. According to the direction and significance of α₁ in Model (1), we can determine whether the digital economy reduces the demand gap and recruitment cost of R&D personnel while improving their efficiency. According to the mediation effect test procedure, if the digital economy ultimately promotes enterprise innovation through the abovementioned R&D staffing and efficiency effects, we expect α₁ in Model (2) to be significantly positive and the sign of α₂ in Model (3) to be significant.

As human capital allocation, accumulation, and innovation activities are closely related to corporate capital, this study first considers a firm’s debt and cash flows when selecting the control variables. It also controls for factors closely related to human capital and innovation behaviour in terms of firm characteristics, governance, and external environment. City-level control variables in terms of economic development, population growth, industrial structure, and foreign investment are also selected, and their names and definitions are listed in

Table 1. Definitions of key variables.

Variable Name	Variable Symbols	Variable Definitions
Demand gap for R&D staff	STP_1	Residual term of the regression model of the number of R&D personnel on firm size, age, nature and industry
R&D staff search costs	STP_2	R&D staff remuneration/R&D expenditure
R&D staff productivity	STP_3	Number of annual applications for the three types of patents by enterprises/number of R&D personnel
Corporate Innovation	Innovation	The annual number of patents granted to enterprises in the three categories plus one plus the natural logarithm
Digital Economy	BC_selected	The enterprise’s registered place has been selected as the “Broadband China” demonstration city at the end of the year, the value is 1, otherwise it is 0
Gearing ratio	Lev	Total liabilities/total assets
Operating cash flow	Cfo	Net cash flow from operating activities/total assets
Size of business	Size	Natural logarithm of total assets
Age	Age	Natural logarithm of the number of years the business has been in existence
Characteristic	Soe	The value is 1 if the ultimate controller is the state, 0 otherwise
Growth	Growth	Annual growth rate of revenue from main business
Share of tangible assets	Tang	Net tangible assets/total assets
Concentration of shareholding	Owner	Percentage of shareholding of the largest shareholder
Level of economic development	Ggdp	GDP growth rate of the place of business incorporation
Population growth	Peop	Population growth rate at the place of business registration
Industrial structure	Ind	Value added of tertiary sector as a percentage of GDP where the enterprise is registered
Foreign investment	Fdi	Actual use of foreign direct investment/GDP

.

3.3. Descriptive Statistics

Table 2. Descriptive statistics.

Variables	Volume of Observations	Average Value	Standard Deviation	Median Value	Minimum Value	Maximum Value
STP_1	16,976	7.002	2.708	5.834	−9.519	11.085
STP_2	16,976	0.062	1.156	0.045	0.008	0.667
STP_3	16,976	0.391	2.004	0.329	0.000	0.903
Innovation	16,976	1.207	3.478	0.753	0.000	8.443
BC_selected	16,976	0.358	0.507	0.000	0.000	1.000
Lev	16,976	0.409	0.210	0.400	0.050	1.110
Cfo	16,976	0.043	0.089	0.052	−0.197	0.264
Size	16,976	22.110	1.282	21.892	19.080	28.513
Age	16,976	1.903	0.304	2.837	1.103	4.062
Soe	16,976	0.544	0.481	1.000	0.000	1.000
Growth	16,976	0.181	0.453	0.110	−0.682	4.571
Tang	16,976	0.263	0.156	0.221	0.001	0.773
Owner	16,976	0.357	0.151	0.340	0.000	0.894
Ggdp	16,976	0.129	0.075	0.114	−0.132	0.507
Peop	16,976	0.036	0.081	0.056	−0.124	0.601
Ind	16,976	0.504	2.982	0.510	0.277	0.756
Fdi	16,976	0.048	0.067	0.033	0.000	0.165

reports the results of the descriptive statistics for the main variables, showing the following: there is still an overall shortage of R&D personnel in the sample companies, with an average demand gap (STP_1) of approximately seven people; a proportion of R&D personnel remuneration to R&D expenditure (STP_2) of 6.2%; and an R&D personnel per capita contribution to patent applications (STP_3) of ~0.391. For the standard deviation, maximum, and minimum values for Innovation, there are significant differences in the number of patents applied for by each company. The mean value of BC_selected was 0.358, indicating that 35.8% of the companies were affected by the ‘Broadband China’ strategy.

4. Empirical Results

4.1. Baseline Regression Results

Table 3. Baseline regression results.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	STP_1	STP_2	STP_3	Innovation	Innovation	Innovation	Innovation
BC_selected	−0.091 ** (0.035)	−0.039 ** (0.019)	0.079 *** (0.015)	0.013 *** (0.001)	0.011 ** (0.004)	0.003 ** (0.001)	0.016 *** (0.003)
STP_1					−0.055 ** (0.020)
STP_2						−0.002 * (0.001)
STP_3							0.104 *** (0.028)
Lev	0.031 ** (0.012)	0.018 *** (0.006)	−0.168 * (0.102)	−0.028 ** (0.011)	−0.027 *** (0.007)	−0.023 ** (0.009)	−0.021 ** (0.008)
Cfo	−0.097 *** (0.016)	−0.085 *** (0.014)	0.014 *** (0.004)	0.164 ** (0.066)	0.160 ** (0.063)	0.157 ** (0.061)	0.162 *** (0.054)
Size	−0.315 ** (0.136)	−0.164 ** (0.066)	0.094 *** (0.007)	0.098 *** (0.016)	0.091 * (0.048)	0.097 ** (0.037)	0.088 ** (0.034)
Age	0.002 (0.001)	−0.001 (0.001)	0.007 *** (0.001)	−0.026 (0.052)	0.008 (0.026)	−0.001 (0.001)	0.002 (0.011)
Soe	−0.188 (0.220)	−0.034 ** (0.016)	−0.083 *** (0.017)	−0.096 *** (0.037)	−0.094 *** (0.007)	−0.090 *** (0.147)	0.082 ** (0.032)
Growth	0.263 * (0.140)	0.084 (0.074)	0.090 *** (0.017)	0.073 ** (0.035)	0.141 *** (0.016)	0.315 ** (0.136)	0.364 ** (0.184)
Tang	0.041 * (0.019)	−0.323 (0.215)	0.166 * (0.068)	0.079 *** (0.015)	0.173 *** (0.041)	0.083 ** (0.036)	0.099 *** (0.015)
Owner	0.158 (0.202)	−0.088 ** (0.034)	0.005 * (0.003)	0.360 * (0.194)	0.083 *** (0.016)	0.040 (0.225)	0.821 * (0.419)
Ggdp	−0.086 * (0.046)	−0.343 (0.314)	−0.054 (0.677)	0.315 ** (0.136)	0.129 ** (0.048)	0.108 * (0.054)	0.096 (0.302)
Peop	−0.037 ** (0.016)	−0.097 *** (0.011)	0.147 (0.721)	0.079 (0.369)	0.137 (0.222)	0.068 (0.518)	0.271 (0.204)
Ind	−0.021 (0.147)	0.076 (0.329)	−0.152 (1.638)	−0.005 *** (0.001)	−0.109 *** (0.022)	−0.073 *** (0.004)	−0.031 *** (0.010)
Fdi	0.110 (0.703)	0.001 (0.026)	0.164 ** (0.066)	0.052 (0.701)	0.068 (0.151)	0.032 (0.231)	0.030 (0.991)
Constant	−0.346 ** (0.134)	1.653 *** (0.551)	0.440 *** (0.145)	−9.254 *** (0.029)	3.878 *** (0.127)	3.990 *** (0.135)	−3.594 *** (0.198)
μt	Yes	Yes	Yes	Yes	Yes	Yes	Yes
γi	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Obs	16,976	16,976	16,976	16,976	16,976	16,976	16,976
Adj. R2	0.235	0.194	0.212	0.118	0.264	0.278	0.219
Sobel Test	Z−statistic (5): BC_selected→STP_1→Innovation				−4.452 **
	Z−statistic (6): BC_selected→STP_2→Innovation					−9.807 ***
	Z−statistic (7): BC_selected→STP_3→Innovation						−4.865 **

Note: ***, ** and * denote significance levels of 1%, 5% and 10%, respectively; robust standard errors corrected for heteroskedasticity and adjusted for firm-level clustering are shown in parentheses. Same as below.

reports the results of the baseline regressions for the hypotheses. Columns (1) to (3) show the impact of being selected as a ‘Broadband China’ demonstration city on the demand gap, recruitment cost, and efficiency of R&D staff. The coefficient of BC_selected is significantly negative at the 5% level when the dependent variables are STP_1 and STP_2, indicating that the digital economy can reduce the demand gap and recruitment cost of R&D personnel. Meanwhile, the coefficient of BC_selected is significantly positive at the 1% level when the dependent variable is STP_3, indicating that the digital economy boosts R&D personnel efficiency.

The dependent variable in Column (4) is Innovation, and the coefficient of BC_selected is 0.013 and is still significantly positive at the 1% level, indicating that the digital economy can significantly boost firm innovation. Columns (5) to (7) further incorporate STP_1, STP_2, and STP_3 into the model and show their mediating effects on the digital economy and firm innovation. Unsurprisingly, the coefficient of BC_selected remains significantly positive, at least at the 5% level, and the coefficients of STP_1 and STP_3 are significant at the 5% and 1% levels, respectively, indicating a significant mediating effect. Although the coefficient of STP_2 has weak statistical significance (10% level of significance), as suggested by [21], a Sobel test reports a Z-statistic of −9.807 and significance at the 1% level, again implying a significant mediating effect. In summary, STP_1, STP_2, and STP_3 can significantly mediate the relationship between the digital economy and corporate innovation. This is further explained in conjunction with the direction of their estimated coefficients, that is, the digital economy can promote corporate innovation by reducing the demand gap and recruitment costs of corporate R&D staff and increasing their efficiency. All of the present study’s research hypotheses were thus tested.

4.2. Analysis of Parallel Trends and Dynamic Effects

The parallel trend between the treatment and control groups prior to the exogenous change is an important premise of the double-difference model. Drawing on the treatment offered by [22], the interaction terms between the treatment group (the sample from the ‘Broadband China’ demonstration cities) and the exogenous changes 1–4 years prior (Treat_B4–B1), the year they took place (Treat _D0), and the following 1–5 years (Treat _D1–D5) were constructed and added to the model. This allows a parallel trend to be tested, while the effect of the digital economy is also decomposed into individual years, hence facilitating the examination of its dynamics.

As seen in

Table 4. Analysis of parallel trends and dynamic effects.

Variables	(1)	(2)	(3)	(4)
	STP_1	STP_2	STP_3	Innovation
Treat_B4	−0.101 (0.127)	−0.081 (0.094)	0.129 (0.120)	0.144 (0.161)
Treat_B3	−0.026 (0.075)	−0.025 (0.023)	0.051 (0.049)	0.008 (0.142)
Treat_B2	−0.148 (0.211)	−0.036 (0.105)	0.032 (0.119)	0.050 (0.247)
Treat_B1	−0.003 (0.125)	−0.031 (0.178)	0.086 (0.202)	0.063 (0.368)
Treat_D0	−0.010 * (0.006)	−0.091 * (0.034)	0.131 (0.126)	0.157 (0.457)
Treat_D1	−0.014 *** (0.002)	−0.005 *** (0.001)	0.099 (0.157)	0.116 (0.239)
Treat_D2	−0.056 ** (0.028)	−0.031 *** (0.004)	0.274 * (0.154)	0.002 (0.002)
Treat_D3	−0.181 *** (0.060)	−0.037 *** (0.002)	0.132 ** (0.066)	0.225 * (0.130)
Treat_D4	−0.175 *** (0.066)	−0.159 *** (0.061)	0.174 *** (0.064)	0.282 ** (0.130)
Treat_D5	−0.241 ** (0.117)	−0.332 *** (0.126)	0.205 *** (0.028)	0.467 *** (0.027)
Constant	−9.412 *** (0.626)	−4.528 *** (0.670)	−8.149 *** (0.646)	−8.565 *** (0.671)
Controls	Yes	Yes	Yes	Yes
μt	Yes	Yes	Yes	Yes
γi	Yes	Yes	Yes	Yes
Obs	16,976	16,976	16,976	16,976
Adj. R2	0.272	0.165	0.281	0.296

Note: ***, ** and * denote significance levels of 1%, 5% and 10%, respectively.

, the coefficients of the interaction terms did not reach the 10% significance level prior to implementation of the ‘Broadband China’ strategy, indicating that there was no significant difference between the two groups of sample companies in terms of the demand gap, recruitment costs, work efficiency, and enterprise innovation related to R&D staff, and, therefore, that the sample meets the assumption of a parallel trend. The coefficients of the interaction term began to reach statistical significance in the year that the ‘Broadband China’ strategy was implemented, subsequently showing a significant upward trend. Analysing each year individually, we find that the digital economy has an ‘immediate’ effect on reducing the demand gap and recruitment costs for R&D staff; however, the positive impact on R&D staff productivity and innovation takes approximately three years to emerge, which is consistent with the time required for new technology applications to be translated into endpoints [23].

4.3. Robustness Tests

4.3.1. Placebo Tests

We moved the entry time of each model city forward by three years, reset the value of BC_selected, and then conducted a placebo test using this dummy dataset. As shown in

Table 5. Robustness tests: placebo test.

Variables	(1)	(2)	(3)	(4)
	STP_1	STP_2	STP_3	Innovation
BC_selected	−0.001 (0.035)	−0.085 (0.436)	0.106 (0.240)	0.037 (0.147)
Constant	−9.102 *** (3.224)	0.742 *** (0.210)	0.030 *** (0.006)	1.705 *** (0.144)
Controls	Yes	Yes	Yes	Yes
μt	Yes	Yes	Yes	Yes
γi	Yes	Yes	Yes	Yes
Obs	16,976	16,976	16,976	16,976
Adj. R2	0.193	0.206	0.217	0.179

Note: *** denote significance levels of 1%.

, none of the coefficients for BC_selected are significant, indicating that the inherent differences between the treatment and control groups and the issue of omitted variables have little impact on the basic conclusions of this study.

4.3.2. Replacement of the Measurement Model

To further exclude the influence of other basic firm characteristics in the treatment and control groups, the control group sample was rematched using propensity score matching. The matching variables mainly consisted of basic firm characteristics and governance in the control variables of the double-difference model, and the matching method was a 1:1 radius matching (radius of 0.01). As shown in

Table 6. Robustness tests: re-proportioning control group.

Variables	(1)	(2)	(3)	(4)
	STP_1	STP_2	STP_3	Innovation
BC_selected	−0.038 *** (0.005)	−0.089 *** (0.014)	0.097 *** (0.016)	0.172 *** (0.039)
Constant	10.428 *** (0.224)	9.765 *** (0.212)	−8.052 *** (0.134)	−4.601 *** (0.995)
Controls	Yes	Yes	Yes	Yes
μt	Yes	Yes	Yes	Yes
γi	Yes	Yes	Yes	Yes
Obs	9987	9987	9987	9987
Adj. R2	0.361	0.302	0.255	0.318

Note: *** denote significance levels of 1%.

, the BC_selected regression coefficients based on the matched samples all reached a 5% significance level and were in the same direction as those in Table 3.

4.3.3. Re-Proportioning Control Group

The cities in 2014, 2015, and 2016 were all selected as ‘Broadband China’ demonstration cities, having a staggered distribution over space and time. In this study, the observed samples from 2015 and 2016 were excluded and only those from 2014 were retained to exclude the interference of the staggered quasi-natural context, thus obtaining a double-difference model in a general sense. The regression results in

Table 7. Robustness tests: replacing the econometric model.

Variables	(1)	(2)	(3)	(4)
	STP_1	STP_2	STP_3	Innovation
BC_selected	−0.235 *** (0.029)	−0.048 * (0.030)	0.354 ** (0.138)	0.271 ** (0.104)
Constant	1.941 *** (0.051)	−0.268 (0.295)	−4.347 *** (0.184)	−9.090 *** (0.673)
Controls	Yes	Yes	Yes	Yes
μt	Yes	Yes	Yes	Yes
γi	Yes	Yes	Yes	Yes
Obs	10,073	10,073	10,073	10,073
Adj. R2	0.210	0.272	0.195	0.248

Note: ***, ** and * denote significance levels of 1%, 5% and 10%, respectively.

show general agreement with those in Table 3, indicating that the findings of this study are rather robust.

4.4. Heterogeneity Testing and Further Analysis

4.4.1. Firm-Level Heterogeneity: The Impact of the Age of the Entrepreneur

According to previous theoretical analyses, the digital economy’s impact on companies’ human capital- and innovation-related activities mainly occurs through digital channels that emphasise the application and integration of digital technologies. However, as a major structural shift, the digital transformation of companies carries a dual strategic risk of breaking through organisational inertia and forming new practices, and its effective implementation depends primarily on the willingness of entrepreneurs [24]. The upper echelons theory suggests that the older the entrepreneur, the lower the awareness and understanding of emerging novelties, and the lower the level of corporate risk-taking [25]. Accordingly, the possible heterogeneous impact of the age differences of entrepreneurs on the basic findings of this study was further examined. As shown in

Table 8. Heterogeneity analysis: effect of age of entrepreneur.

Variables	(1)	(2)	(3)	(4)
	STP_1	STP_2	STP_3	Innovation
BC_selected × Youth	−0.014 *** (0.004)	−0.007 ** (0.003)	0.093 *** (0.040)	0.021 *** (0.004)
BC_selected	−0.025 *** (0.006)	−0.178 *** (0.024)	0.154 ** (0.062)	0.126 *** (0.031)
Youth	0.005 (0.012)	−0.082 * (0.039)	0.189 (0.403)	0.102 * (0.052)
Constant	0.694 *** (0.257)	4.467 *** (0.151)	−0.028 (0.032)	−0.493 *** (0.039)
Controls	Yes	Yes	Yes	Yes
μt	Yes	Yes	Yes	Yes
γi	Yes	Yes	Yes	Yes
Obs	16,976	16,976	16,976	16,976
Adj. R2	0.305	0.294	0.311	0.276

Note: ***, ** and * denote significance levels of 1%, 5% and 10%, respectively.

, the coefficients of the cross-product BC_selected × Youth are significant, with Columns (1) and (2) being significantly negative, and Columns (3) and (4) being significantly positive. This indicates that, for firms with young entrepreneurs, the digital economy has a more significant effect on supporting human capital- and innovation-related performance.

4.4.2. Heterogeneity at the City Level: The Impact of Geographical Location

Consistent with the spatial differences in China’s regional economies, eastern cities have the unique advantage of having many network infrastructures in place, with higher internet penetration and digital device usage. The resulting impact of the digital economy on human capital- and innovation-related performance may thus differ depending on geographic location. Theoretically, the digital economy’s contribution to the human capital- and innovation-related performance of firms depends on the interactions between employees and the outside world [26], and the micro-level contribution of the digital economy is likely to be more pronounced in eastern cities with higher internet penetration. In this study, we introduced the moderating variable of city location (1 for a city’s natural location in the east, 0 otherwise) to examine its effect on the digital economy. As shown in

Table 9. Analysis of heterogeneity: the effect of the age of the entrepreneur.

Variables	(1)	(2)	(3)	(4)
	STP_1	STP_2	STP_3	Innovation
BC_selected × Location	−0.032 *** (0.009)	−0.021 *** (0.007)	0.043 ** (0.019)	0.054 *** (0.020)
BC_selected	−0.122 *** (0.032)	−0.049 *** (0.016)	0.105 ** (0.026)	0.063 *** (0.023)
Location	−0.088 ** (0.016)	−0.003 * (0.002)	0.001 * (0.001)	0.282 * (0.144)
Constant	0.063 (0.082)	−0.464 *** (0.142)	0.500 *** (0.030)	1.467 *** (0.071)
Controls	Yes	Yes	Yes	Yes
μt	Yes	Yes	Yes	Yes
γi	Yes	Yes	Yes	Yes
Obs	16,976	16,976	16,976	16,976
Adj. R2	0.281	0.343	0.312	0.295

Note: ***, ** and * denote significance levels of 1%, 5% and 10%, respectively.

, the coefficients of BC_selected × Location are both significant and in the same direction as those of BC_selected, indicating that the digital economy in eastern cities plays a more prominent role in boosting the human capital- and innovation-related performance of enterprises compared with those in central and western cities, in line with theoretical expectations.

4.4.3. Heterogeneity at the Environmental Level: The Influence of Confucianism

The cultural environment has a significant and broad impact on socioeconomic development, especially in emerging economies, where the institutional environment is not yet mature. Confucianism is the most influential philosophy and traditional cultural symbol in China. Ref. [27] shows that the Confucian culture of the informal system and the legal environment of the formal system significantly impact business innovation, for example, by raising the level of corporate human capital and encouraging innovation among employees. Thus, this study expects that the digital economy’s effect on corporate human capital- and innovation-related performance is greater in cities with a strong Confucian culture (Confucianism) (cities with Confucian temples take a value of 1, and 0 otherwise), with the results in

Table 10. Heterogeneity analysis: the influence of Confucianism.

Variables	(1)	(2)	(3)	(4)
	STP_1	STP_2	STP_3	Innovation
BC_selected × Confucian	−0.054 *** (0.019)	−0.049 *** (0.013)	0.111 *** (0.036)	0.078 ** (0.028)
BC_selected	−0.105 ** (0.041)	−0.210 ** (0.084)	0.234 ** (0.107)	0.295 *** (0.111)
Confucian	−0.027 (0.033)	−0.323 (0.345)	0.335 * (0.168)	0.091 ** (0.038)
Constant	1.200 *** (0.465)	1.090 *** (0.325)	10.100 *** (0.424)	6.703 *** (0.181)
Controls	Yes	Yes	Yes	Yes
μt	Yes	Yes	Yes	Yes
γi	Yes	Yes	Yes	Yes
Obs	16,976	16,976	16,976	16,976
Adj. R2	0.111	0.180	0.162	0.209

Note: ***, ** and * denote significance levels of 1%, 5% and 10%, respectively.

supporting this statement.

4.4.4. Further Analysis: Can the Digital Economy Contribute to the Quality of Enterprise Innovation?

China has become the world’s leading country in terms of the number of patent applications and grants. However, despite the increased number of patents, a parallel improvement in innovation quality is not evident. Technological progress and high-quality economic development can only be achieved by fundamentally improving the quality of enterprise innovation. Thus, can digital economy development significantly improve the quality of enterprise innovation? Theoretically, digital capital can significantly increase the frequency and reduce the cost of collaboration and exchanges between R&D professionals inside and outside the enterprise, while facilitating the recruitment of R&D personnel and increasing their efficiency [18].

This study examined the aforementioned question further. Specifically, two options were adopted to replace the Innovation variable in the basic model to measure enterprise innovation quality. First, the number of utility models and design patents granted was excluded based on the above-stated measurement, and only the number of invention patents granted (Ipatent) was retained. The second involved constructing the ‘knowledge width of invention patents’ indicator (Kwidth) from the perspective of the number of cross-classes, based on the cross-class information characteristics of the International Patent Classification number of the invention patents. The regression results of basic model (1) are the same as those of Columns (1)–(3) of Table 3, and

Table 11. Impact of the digital economy on the quality of innovation in firms.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Ipatent				Kwidth
BC_selected	0.267 ** (0.108)	0.024 * (0.013)	0.135 ** (0.058)	0.069 * (0.040)	0.017 *** (0.004)	0.005 * (0.003)	0.102 *** (0.018)	0.108 ** (0.042)
STP_1		−0.075 *** (0.014)				−0.016 ** (0.006)
STP_2			−0.042 *** (0.008)				−0.060 ** (0.025)
STP_3				0.105 ** (0.041)				0.134 *** (0.011)
Constant	3.990 *** (0.135)	0.369 *** (0.087)	0.912 ** (0.394)	−0.893 *** (0.119)	5.537 *** (0.135)	0.002 (0.952)	−0.991 *** (0.384)	0.815 *** (0.138)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
μt	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
γi	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Obs	16,976	16,976	16,976	16,976	16,976	16,976	16,976	16,976
Adj. R2	0.200	0.193	0.181	0.202	0.236	0.241	0.229	0.275
Sobel Test	Z(2)	−6.702 **
	Z(3)		−12.761 ***
	Z(4)			−5.184 **
	Z(6)					5.139 **
	Z(7)						−4.775 **
	Z(8)							9.809 ***

Note: ***, ** and * denote significance levels of 1%, 5% and 10%, respectively.

supplements the regression results of basic models (2) and (3) after the enterprise innovation variables are replaced. Columns (1) and (5) of Table 11 show that the coefficient of BC_selected is still significantly positive, indicating that the development of the digital economy increases the quality of enterprise innovation, in line with theoretical expectations. Columns (2)–(4) and (6)–(8) show that the coefficients of all three STP variables are significant and have the same sign as those in Columns (5)–(7) of Table 3, indicating that they still have significant mediating effects. Collectively, these results show that the digital economy can significantly improve innovation quality while increasing the number of patents by enhancing the recruitment processes for R&D personnel.

5. Discussion of Results

The findings of this study have several policy implications. First, given that the digital economy significantly promotes enterprise innovation, the construction of digital infrastructure must be accelerated and digital applications must be expanded for the dividends of digital economy development to be fully realised. Accelerating the construction of digital intelligence infrastructures, such as 5G base stations and large data centres, and focussing on breakthroughs in key technologies, such as chips and smart sensors, will promote further digital development. Continuously promoting integration with the physical economy, breaking information barriers between labour supply and demand, applying digital technology to interactive learning, staying abreast of tailored and scenario-based consumer needs [28], and continuously improving the quality of enterprise innovation is essential [29]. Building digital innovation platforms and shared co-creation ecosystems covering industry, academia, and research institutions is also recommended to streamline the flow of otherwise ‘fragmented’ innovation resources [30], thus promoting knowledge sharing and helping enterprises sustain innovation. Finally, in terms of industrial policy, while developing preferential measures to promote the digital transformation of traditional industries, it is more important to focus on supporting the development of new digital industries to better promote the development of the digital economy.

Furthermore, the Chinese government should attach importance to the role of human capital as a factor in promoting enterprise innovation through the digital economy and build a policy system for people-oriented science and technology innovation in the context of the digital economy by improving human capital allocation efficiency. Additional measures should be taken by the government to address administrative restrictions on the flow of talent, reduce the institutional transaction costs of human capital, promote the cross-regional flow of labour and capital [31], and recognise the role of human capital as key to promoting enterprise innovation through the digital economy. In response to the demands for highly qualified talent and the elimination of low-skilled jobs arising from digital economy development, local governments and enterprises should consider building a multilevel digital skills training system that is both generalist and professional in nature to enhance the digital skills of workers [32], especially in terms of continued education for low-skilled workers. Thus, these workers can also benefit from the social dividends realised by the digital economy. In parallel, governments at all levels should continue to improve policies to identify and support digital talent, build a people-oriented policy system for scientific and technological innovation in the context of the digital economy, and stimulate the creativity and enthusiasm of science and technology personnel.

Finally, considering the heterogeneous impact of the digital economy on corporate innovation, strategies and paths for digital economy development should be tailored to the needs of enterprises and the local context. A key management-oriented takeaway from this study is that corporate management should be fully aware of the digital economy’s development dividend and the important role of their own digital transformation in view of sustainable corporate development.

Enterprises can overcome tendencies towards short-sighted behaviour in upper management, in terms of innovation investments and organisational changes, through equity incentives and other means. This will allow management to fully appreciate the importance of digital approaches for sustainable and high-quality corporate development and, therefore, craft a scientific and feasible digital transformation strategy that considers their own strengths and resource challenges [17], hence actively responding to the larger objective of building a digital China and smart society. Regions with high levels of digital development should accelerate the development of forward-looking digital industry clusters and innovation platforms to promote original achievements towards a world-class digital economy [33]. Regions with relatively low digital development should accelerate digital infrastructure construction and intellectual property protection, while encouraging key industries and platforms to play a leading role in building digital industry chains and clusters.

6. Conclusions

This study examined the impact of digital economy development on corporate innovation from a human capital perspective. Specifically, the sample enterprises were divided into experimental and control groups based on whether they were ‘Broadband China’ demonstration cities. The findings show that the digital economy effectively reduces the demand gap and recruitment costs for R&D personnel and significantly improves their efficiency, thus boosting corporate innovation. A dynamic effects analysis shows that the digital economy has an ‘immediate’ effect on reducing the demand gap and recruitment costs for R&D staff; however, the positive impacts on R&D staff productivity and firm innovation take time to emerge. Heterogeneity tests show that these micro-level effects of the digital economy are more pronounced for firms with younger entrepreneurs, those domiciled in eastern cities, or those strongly influenced by Confucian culture. Further analysis suggests that the digital economy can significantly contribute to the quantity and quality of enterprise innovation by optimising aspects related to human capital recruitment and performance.

The limitations and future research ideas of this paper include the following. Firstly, to avoid the endogeneity problems caused by reverse causality and omitted variables, this paper adopts a quasi-experimental research method, in which exogenous policy shocks of the ‘Broadband China’ strategy are used as proxy variables of the digital economy. However, it is difficult to guarantee the randomisation of the experimental group using a quasi-experimental research method, which reduces its intrinsic validity to some extent. In the future, we will attempt to construct a direct measurement system for the multilevel and complex concept of the digital economy and use various measurement methods to examine the microeconomic effects of digital economy development. Secondly, to highlight the main line of research, the theoretical model in this paper does not consider the effects of the cost of R&D staff, part-time employment, working style and its effectiveness, and so on. In the future, we will build a systematic theoretical model to analyse the relationship between the digital economy, R&D staffing efficiency, and corporate innovation more comprehensively. Thirdly, as an exploratory study, this paper reveals only one path of digital economy development to promote micro-firm innovation from the labour market allocation level, and a more comprehensive explanation of the micro-action mechanism of the digital economy from other levels (e.g., physical capital allocation level) is needed in the future. Finally, the dependent variable has a standard deviation larger than the average, indicating that many firms have no patents, and a few have many (up to eight patents). These proxies may not fully reflect the digital economy’s underlying effects on labour markets and company innovativeness. Considering that many digital innovations (software) are difficult to patent and that many digital innovations are not only technical but also organisational, future research will look for more comprehensive patent databases and update the way variables are measured to address the measurement implications of the variable measurement problem.