Does the digital divide reduce the probability of formal employment for workers？ --Empirical evidence based on CHFS microdata

doi:10.21203/rs.3.rs-3331031/v1

Download PDF

Article

Does the digital divide reduce the probability of formal employment for workers？ --Empirical evidence based on CHFS microdata

https://doi.org/10.21203/rs.3.rs-3331031/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The digital economy, as an emergent economic construct, has progressively become a major driver for employment promotion. Nevertheless, as the digital economy evolves, the "digital divide" has subtly begun to sway individuals' employment decisions. Utilizing China Household Financial Survey (CHFS) data and the Probit model, this paper embarks on an empirical exploration of the digital divide's impact on workers' employment decision-making behavior. The mechanism of the digital divide is scrutinized by incorporating aspects of information accessibility and risk tolerance. The study unveils that: (1) the digital divide markedly diminishes the likelihood of workers securing formal employment; (2) the digital divide can deter workers from venturing into formal employment channels by curbing their access to information and risk tolerance; (3) this deterrent effect is more pronounced among workers in central and western regions, non-metropolitan cities, rural areas, and those burdened with high household indebtedness; (4) the digital divide also significantly undermines workers' subjective well-being. The findings bear significant theoretical and practical implications for labor market improvement and the enhancement of workers' employment quality.

Physical sciences/Energy science and technology/Energy infrastructure

Physical sciences/Energy science and technology

Physical sciences/Engineering

Physical sciences/Materials science

Digital divide

formal employment

behavioral economy

According to the China Internet Development Report 2022, China's digital economy was projected to reach 45.5 trillion yuan in 2021, contributing to 39.8% of the nation's GDP. This illustrates the digital economy's pivotal role as one of the primary catalysts propelling China's economic growth. The digital economy is swiftly permeating diverse sectors, including industry, agriculture, healthcare, trade, and retail. Traditional industries are leveraging digital transformation to realize high-quality development, with digital advancements persistently refining while multiple nodes of digital accomplishments are emerging.

While the digital economy fortifies digital industries and unveils the dividends of technological innovation, it simultaneously presents numerous challenges. Certain workers, due to their superior Internet usage skills, have managed to amass human capital, thus gaining entry into formal employment channels and reaping higher wage returns. Conversely, workers necessitating more advanced Internet-using skills confront not only difficulties in accessing efficient and diversified employment information but may also potentially face a shrinking income share owing to labor market restructuring. Consequently, the evolution of the digital economy subtly shapes people's employment behavior[1].

Scholars have pointed out that the digital economy may lead to structural polarization of the workforce[2]. The negative effects of the digital divide are becoming increasingly evident[3], with a primary digital divide characterized by the digital infrastructure "access gap" and a secondary digital divide characterized by digital technology The primary digital divide, characterized by the digital infrastructure "access gap" and the secondary digital divide, characterized by the digital technology "usage gap", is triggering a tertiary digital divide, mainly characterized by the digital "income gap"[4]. The digital divide also significantly inhibits entrepreneurship and credit [5]. Existing studies have more often explored the relationship between the digital economy and workers' employment and income from a macro perspective, but with the rapid development of the digital economy, the "digital divide" deserves further study. If so, what are the ways in which it works? Answering these questions will help improve the labor market theory and provide empirical support to promote the integration of the digital economy and the real economy, enhance high-quality employment, and promote high-quality economic development.

The key contributions of this study are as follows: Firstly, leveraging the newly released 2019 China Household Finance Survey data, this study scrutinizes the influence of the digital divide on workers' formal employment from a micro perspective. By focusing on individuals as the units of analysis, the study enriches the existing body of knowledge within the labor market theory. Secondly, the research delves into and validates the instrumental roles of information access and risk tolerance, providing empirical evidence that could guide future efforts to enhance workers' digital literacy. Thirdly, by centering on the digital divide, the study probes into the factors that shape the nature of workers' employment, offering crucial practical implications for bridging the digital divide and fostering high-quality employment.

A Review of Research on the Digital Divide, Formal Employment

The notion of the "digital divide" first surfaced in the 1990s, defined as the disparity in information opportunities between individuals who have access to new information technologies and those who do not[6] As the 20th century drew to a close, digital inequality—initially between countries and communities, and subsequently among individuals—captured significant political and academic attention[7] The OECD (2001) characterizes the digital divide as "the disparities in the accessibility of information and communication technologies and their utilization across individuals, households, businesses, and regions of varying socioeconomic statuses." [8]

Concerning the root causes of the digital divide, some scholars posit that the divide in the physical-level internet access conditions among different groups constitutes the first digital divide, or the access divide. In contrast, the variance in digital skills and technology usage forms the second digital divide, or the use divide[9].Also some studies suggest that the disparities in internet accessibility and usage among varying societal groups contribute to the digital divide.[10]

Labor market segmentation theory delineates between primary and secondary labor markets, with the latter typically characterized by inferior working conditions and relatively diminished benefit packages and wage levels. Workers in informal employment, who are commonly classified under secondary labor markets, continue to constitute a substantial proportion in developing nations. These individuals are often relegated to the periphery of the core workforce, possessing minimal bargaining power with employers, limited opportunities for further training, and are prone to recurrent unemployment. In the digital era, the digital divide may bifurcate the workforce into two broad categories: "Integrated Workers," who are well-qualified for high-productivity, information-centric roles and may eventually transition into formal employment, and "Excluded Workers," who, due to persistent unemployment, lose out on steady employment. This group may find themselves relegated to informal employment, losing their regular positions due to repeated unemployment, being compensated poorly, and shuttling between precarious jobs. At the present juncture, labor market research is centered around the concept of full employment. However, as fintech penetrates various industries, the question arises as to whether workers' employment decision-making will undergo corresponding changes.

Theoretical assumptions

The digital economy continues to overturn people's perceptions of traditional employment, and the digital divide is triggering changes in the labor market structure. On the one hand, digital technology has broadened people's access to employment information, and for high-skilled digital workers, job search costs are significantly reduced. In contrast, the scarcity of big data talents has increased the demand for high-skilled workers[11], which ultimately helps to increase the probability of formal labor force employment[12]. The "skill premium" in the context of the digital economy also helps promote higher wages for high-skilled digital workers [13]. On the other hand, digital skills have a significant human capital accumulation effect [14]. Workers with a small digital divide learn new knowledge and skills independently through modern digital media such as the Internet and continuously improve their human capital. The knowledge-learning effect they generate will help improve their job performance and strengthen their comparative advantage in employment. However, workers with a large digital divide gap may reduce their human capital investment due to their inability to reach the expected income [15] and are eventually restricted to the informal employment sector. Based on the above analysis, this paper proposes the hypothesis

H1: The digital divide will increase the probability of people entering formal employment channels.

Based on existing research, the article argues that the digital divide affects workers' employment decision-making behavior through two main channels.

First information acquisition ability. Digital technology is continuously widening the boundary of workers' information access, workers can search for intended jobs through the Internet channel [16], and they can learn rich knowledge through various online platforms to promote human capital accumulation, improve their vocational skills and employability, and enhance formal employment opportunities. Especially for highly skilled digital workers, their job search channels through the Internet are more diverse, and their communication with enterprises is more efficient, which not only helps to increase the probability of workers' job search success but also helps to improve workers' subsequent job wages[17]. However, for workers with poor digital skills or no exposure to digital technology, it will not be easy to enhance their comparative advantage in employment with the help of digital technology, and they cannot enter the formal employment channel.

Second risk tolerance. Digital technology helps to enhance people's ability to analyze, cognize, and grasp risks, improving risk tolerance[18]. Highly skilled digital workers with higher risk tolerance prefer to enter formal employment channels through occupational mobility and obtain better social security and higher levels of wage income. Movers with poor digital skills have lower risk tolerance and are unwilling to take the unpredictable risks of changing jobs[19]. Based on the above analysis, this paper proposes the following hypothesis.

H2a: The digital divide reduces workers' access to information, which in turn inhibits workers from entering formal employment channels.

H2b: Digital literacy reduces workers' risk tolerance, which in turn inhibits workers from entering formal employment channels.

We are examining the impact of the digital divide on the nature of workers' work. The empirical model is set up in this paper as follows. (1) Based on the Probit model, we analyze whether digital literacy affects workers' employment decision-making behavior.

Econometric model construction

$$Prob\left(Mainwor{k}_{i}=1\right)={\beta }_{0}+{\beta }_{1}\ast Di{g}_{i}+{\beta }_{2}\ast Control{s}_{i}+{\epsilon }_{i}$$

where$Mainwork$ is the explanatory variable, representing whether or not formal employment, if yes, it takes 1, otherwise, it takes 0.$Dig$ is the core explanatory variable, representing the individual digital divide index.$Controls$ are all control variables, which will be explained in the next subsection with focus.$\epsilon$ represents the equation residuals. The subscripts$i$ represents the different observed individuals.

Data sources and variable selection

Data source

The data utilized in this study originate from the fourth iteration of the China Household Finance Questionnaire Survey (CHFS), conducted nationwide in 2019. The CHFS sample encompasses 29 provinces (autonomous regions and municipalities directly under the central government), 170 cities, 345 districts and counties, and 1360 village (residential) committees, excluding Tibet, Xinjiang, Hong Kong, Macao, and Taiwan. With a sample size of 34,643 households, the data offers nationally and provincially representative insights.

The China Household Financial Survey gathers data on household demographic characteristics, assets and liabilities, insurance and social security, income and expenditures, among other facets. Several measures were deployed to control for sampling and non-sampling errors, ensuring a high quality of data. Post data cleaning, a total of 11,965 samples were retained for analysis.

Variable description

Explanatory variable: formal employment.The nature of work is generally classified as formal and informal. The specific classification is based on whether a labor contract is signed and the type of labor contract [20], the duration of the labor contract[21], and whether they enjoy social insurance benefits [22] as the division criteria. This paper defines temporary work (with a workplace but without a formal labor contract, such as odd jobs) and freelance work as informal employment and taken as 0. Being employed by another person or unit (with a formal labor contract) is defined as formal employment and taken as 1, thus serving as a proxy variable for whether formal employment is available.

Explanatory variables: digital divide.In this paper, based on the approach of Yin Zhichao[23], the digital divide is defined in terms of two dimensions: accessibility and usage. Two indicators are selected to represent "accessibility": whether to use a smartphone; or whether to open a third-party payment account. Three indicators are selected to represent "usage", namely, Internet access cost, online shopping cost, and online shopping frequency in the past year. These five sub-indicators are factored to generate the digital divide index (Dig) and transformed into 0-100, and the formula is as follows:

$$Dig=100\times \left[max\left(dd\right)－dd\right]/\left[max\left(dd\right)－min\left(dd\right)\right]$$

Where dd is the value obtained by factor analysis of the above five indicators for each individual, max(dd) is the maximum value, and min(dd) is the minimum value. In other words, the digital divide index measures the percentage of the gap between dd and the maximum value of an individual in the maximum gap. The results of the KMO test indicate that the sample is suitable for factor analysis.The descriptive statistics are presented as shown in Table 1, which details urban and rural statistics.

Table 1

Urban and rural statistics.
Category		Mean	SD	Min	Max
Rural	Digital Divide	40.68	12.52	12.26	96.26
Rural	Formal Employment	0.412	0.492	0	1
City	Digital Divide	34.34	11.27	3.01	85.55
City	Formal Employment	0.684	0.465	0	1

Control variables. The control variables in this paper mainly include age (Age), gender (Sex), education (Edu), health (Health), whether married (Marri), number of family members (Familynumb), whether the household owns its own home (House), and household asset and liability ratio (ROA). The variables are defined in Table 2.

Mediating variables. One is information acquisition ability. In this paper, "How much do you usually pay attention to economic and financial information?" is used as a proxy variable for information acquisition ability. as a proxy variable of information acquisition ability[27]. The second is risk tolerance. In this paper, "If you have a sum of money to invest, which investment project would you be most willing to choose?" as a proxy variable for risk tolerance[24]. The variables are defined in Table 2.

Instrumental variables. Considering the possible omission of variables in the model and the problem of endogeneity due to the mutual causality between the digital divide and the nature of work, this paper selects the community-level digital divide (in addition to itself) and the provincial-level Internet penetration rate as the instrumental variables of the core explanatory variables in this paper for testing. On the one hand, the faster the provincial Internet penetration rate and the smaller the community-level digital divide, the higher the probability of workers using the Internet, satisfying the correlation hypothesis; on the other hand, the above two variables belong to macro-level data, satisfying the homogeneity hypothesis, so the above two instrumental variables selected are reasonable. The variables are defined in Table 2.

Table 2

Classification, name and definition of variables.
Variable classification	Variable Name	Variable Definition
Explained variables	Mainwork	Dummy variable. If the respondent is formally employed, take 1; otherwise, take 0.
Core explanatory variables	Dig	Respondents' digital divide, calculated according to Eq. (1).
Digital Divide Indicator	Phone	Dummy variable. If the respondent uses a smartphone, take 1; otherwise, take 0.
	Eaccount	Dummy variable. If the respondent opens an electronic payment account, take 1; otherwise, take 0.
	Internetfee	Respondents' average monthly communication network fee expenditure last year
	Onlinefee	Respondents spent on online shopping last year
	Onlinefre	Respondents' frequency of online shopping last year
Control variables	Age	Retention of observation data for respondents aged 18–60
	Sex	If the respondent is female, take 1, otherwise, take 0
	Edu	Respondents' education level. The values are no school, elementary school, junior high school, high school, junior college/vocational high school, college/high school, undergraduate, master's degree, and doctoral degree, respectively, and the higher the number, the higher the education level.
	Health	Respondents' health degree. Very good, good, fair, bad, and very bad, respectively, with values from 1 to 5, with larger numbers representing poorer health
	Marri	Respondent's marital status, married, take 1; otherwise, take 0
	Familynumb	The number of household members is the actual number of household members, i.e., those who share income and expenses with the respondent
	House	Respondent takes 1 if he/she owns a house; otherwise, takes 0
	ROA	Total household liabilities and total household assets are added by 1 and logarithmically divided to obtain
Mechanism Variables	Infor	The degree of respondents' usual attention to economic and financial information is used as a proxy variable for information acquisition ability, including "never pay attention, rarely pay attention, average, very much pay attention, and very much pay attention", with values from 1 to 5, with larger numbers representing stronger information acquisition ability.
Mechanism Variables	Risk	Respondents' investment preference is used as a proxy variable for risk tolerance, including "not willing to take any risk, slightly low risk and slightly low return items, average risk and average return items, slightly high risk and slightly high return items, high risk, and high return items," with values from 1 to 5, with higher numbers representing risk tolerance. The higher the risk tolerance.
Tool Variables	Internet	Number of Internet users by province divided by regional resident population
Tool Variables	Cdig	According to the respondent's community, remove their own digital divide index, and then calculate the average of the digital divide in their community

Descriptive statistics

In this study, a descriptive statistical analysis was performed on the explanatory variables, core explanatory variables, factor indicator system, control variables, mediating variables, and instrumental variables, as illustrated in Table 3. Our analysis revealed that the numerical divide exhibited a minimum value of 3.01, a maximum value of 96.26, and an average value of 35.53. These findings suggest a significant degree of numerical divide within the sample data, characterized by a broad range of values and a relatively low mean.

Table 3

Descriptive statistical analysis.
Variable Name	N	Mean	SD	Min	Max
Mainwork	11965	0.633	0.482	0	1
Dig	11965	35.53	11.78	3.01	96.26
Phone	11965	0.954	0.211	0	1
Eaccount	11965	0.925	0.263	0	1
Internetfee	11965	5.346	0.788	0	8.517
Onlinefee	11965	7.876	1.375	2.303	12.79
Onlinefre	11965	6.828	1.832	1	10
Age	11965	39.29	9.866	18	60
Sex	11965	0.572	0.495	0	1
Edu	11965	4.77	1.784	1	9
Health	11965	2.235	0.821	1	5
Marri	11965	0.825	0.38	0	1
Familynumb	11965	3.91	1.459	1	15
House	11965	0.925	0.263	0	1
ROA	11965	0.431	0.439	0	13.29
Infor	11951	2.134	0.999	1	5
Risk	10941	2.157	1.117	1	5
Internet	11965	0.325	0.051	0.232	0.436
Cdig	11965	36.36	6.553	21.91	81.5

Baseline regression and endogeneity test

Benchmark regression.Columns (1) and (2) of Table 4 present the estimated results of the digital divide's impact on the nature of employment, respectively. Upon introducing control variables in column (2) of Table 4, the digital divide's marginal effect on formal employment is approximately − 0.156%, significant at the 1% level. This suggests that the digital divide diminishes the likelihood of workers transitioning into formal employment channels, thereby validating the Hypothesis H1 proposed in this study.

In addition, the coefficient of the age of the household head is positive and significant at the 1% level, indicating that the probability of formal employment increases with the age of the worker, which may be because the social experience of the worker increases with age, which will help the worker to be properly employed; gender does not affect the nature of individual employment; the higher the level of education, the higher the probability of choosing formal employment, which is because education tends to be closely related to human capital is closely related to human capital; health is also an important human capital[25], and the regression results show that the coefficient of health is significantly negative, indicating that the healthier the worker is, the higher the probability of formal employment; being married makes people more inclined to look for stable jobs, so the probability of formal employment is higher for married workers. Household size significantly reduces the probability of formal employment, which may be since household size includes young children and elderly parents, a group that requires more flexible time for caregiving. Workers choose informal employment to balance work and family. The higher the household gearing ratio, the lower the probability of workers entering formal employment.

Table 4

Basis regression and endogeneity test.
Variables	(1)	(2)	(3)
Variables	Probit	Probit	IVProbit
Dig	-0.030***	-0.016***	-0.038***
	(0.001)	(0.001)	(15.915)
Age		0.013***	0.013***
		(0.002)	(8.049)
Sex		-0.023	0.001
		(0.027)	(0.022)
Edu		0.396***	0.334***
		(0.009)	(29.674)
Health		-0.052***	-0.032*
		(0.016)	(1.958)
Marri		0.209***	0.095**
		(0.041)	(2.292)
Familynumb		-0.095***	-0.090***
		(0.010)	(9.276)
House		0.187***	0.206***
		(0.050)	(4.281)
ROA		-0.083***	-0.135***
		(0.030)	(4.415)
Constant	1.411***	-1.190***	-0.102
	(0.040)	(0.118)	(0.645)
Obs	11965	11965	11,965
Pseudo R²	0.054	0.227	0.231
Phase I F-value			513.72
Wald/DWH test			106.22
Note: *, , * indicate significant at the 1%, 5%, and 10% levels, respectively, and standard errors are in parentheses.

Endogeneity test. The regression results from the benchmark model above might be subject to endogeneity issues arising from reciprocal causality and omitted variables. First, concerning reciprocal causality: on one hand, enhanced digital skills are crucial for workers to better integrate into contemporary society and navigate technological innovations that characterize modern societal development, thus preventing loss of formal employment opportunities. Conversely, formal employment also necessitates workers to possess superior digital skills. As modern society progresses rapidly, digital technology becomes increasingly pervasive. Workers consistently augment their knowledge of digital information technology, such as the Internet, whereas for those in informal employment groups, the digital divide continues to widen.Secondly, there's the issue of omitted variables. Although this study controls for attributes like age and gender, workers' employment decisions might still be influenced by unobservable variables, such as personality traits and upbringing.

For this reason, this paper adopts an instrumental variables approach to address endogeneity. Existing literature on the digital divide at the individual level is sparse, and individual behavioral decisions are studied based on individual digital divide gaps, whether or not they use the Internet, and whether or not they use mobile payments, etc. The literature addresses endogeneity using instrumental variables such as provincial Internet penetration[26], county-level digital divide, community-level monthly household postal and telecommunications costs[27],and smartphones[28].

Considering that the digital divide indicator system in this paper already contains indicators such as "smartphone," "Internet access cost," and "electronic payment account," this paper, based on the combination of scholars' studies, The provincial Internet penetration rate and the community-level digital divide (in addition to the workers themselves) are selected as the instrumental variables. On the one hand, the faster the Internet penetration rate, the higher the probability of workers using the Internet, and the higher the digital divide gap among workers may be, while the individual-level digital divide among workers is easily influenced by the surrounding community, thus satisfying the correlation hypothesis. On the other hand, the Internet penetration rate uses macro-level data, while the community-level digital divide level excludes workers' influencing factors, thus satisfying the homogeneity hypothesis. The paper concludes that the selected instrumental variables are theoretically feasible.

The results from the first-stage regression (Table 3, Column (3)) indicate that the F-statistic is significantly higher than ten. The p-values of CLR, K-J, AR, and Wald are all statistically significant at the 1% level, allowing us to reject the null hypothesis that "weak instrumental variables exist"[29] This suggests that the instrumental variables chosen for this study are strongly correlated with the endogenous variables.The second-stage regression results demonstrate a significantly positive coefficient of digital literacy (Table 3, Column (3)), substantially higher than the coefficient of the baseline regression (Table 3, Column (2)). This implies that the negative impact of the digital divide on formal employment is likely to be underestimated if endogeneity is not taken into account. Furthermore, the DWH endogeneity test results indicate that we can reject the hypothesis of "all variables being exogenous" at the 1% statistical significance level.In conclusion, the selected instrumental variables are appropriate, and the study's findings remain robust.

Robustness test

In this section, variables are designed from different perspectives for robustness analysis to test the findings' reliability further. This paper replaces the core explanatory variables with a digital divide dummy (Dig_dummy) and digital divide type (Dig_type).

Firstly, if a household head has a negative option in the above factor model (i.e., no smartphone, no third-party payment account, did not incur Internet costs last year, did not make online purchases last year), then the digital divide dummy variable (Dig_dummy) is defined as 1; otherwise, it is 0. Secondly, the above factor model indicators are set as dummy variables and summed. If the household head fully enjoys the digital dividend (i.e., has a smartphone, has opened a third-party payment account, has incurred Internet costs last year, has made online purchases last year), then the digital divide type (Dig_type) is 0. If the household head does not enjoy the digital dividend at all (i.e., does not have a smartphone, has not opened a third-party payment account, has not incurred Internet costs last year, has not made online purchases last year), then the digital divide type (Dig_type) is 5. The digital divide type from 1 to 5 represents the degree of severity of the digital divide faced by households. Finally, this paper matches the sample data with the Digital Financial Inclusion Index of Peking University (Dig_PK), which is limited to the range of [0,100], takes negative numbers for processing, and conducts the digital divide robustness test at the provincial level. The regression results are shown in Table 5. After a series of robustness tests, we find that the digital divide reduces the probability of formal employment of workers, and the conclusions of this paper still hold.

Table 5

Robustness tests.
Variables	Mainwork
Variables	(1)	(2)	(3)
Dig_dummy	-0.208***
	(0.042)
Dig_type		-0.160***
		(0.034)
Dig_PK			-0.075***
			(0.006)
Controls	YES	YES	YES
Obs.	11965	11965	11965
Pseudo R²	0.217	0.217	0.226

Mechanism analysis

Drawing on the steps of mediating effect analysis by J. Fang et al. (2017) [30] and C. L. Wen et al. (2014) [31], the following mediating effect analysis is further done for information accessibility and risk tolerance, respectively, and the models are set as in (3) and (4).

$$E{M}_{i}={{\theta }}_{1}+{{\theta }}_{2}Di{g}_{i}+{{\theta }}_{3}Control{s}_{i}+{{\xi }}_{i}$$

$${Mainwork}_{i}={{\lambda }}_{1}+{{\lambda }}_{2}Di{g}_{i}+{{\lambda }}_{3}E{M}_{i}+{{\lambda }}_{4}Control{s}_{i}+{\text{ϵ}}_{i}$$

where $E{M}_{i}$ represents the mediating variable. First, the relationship between the digital divide and the mediating variables is tested, and then whether there is still a significant relationship between the digital divide and formal employment by introducing mediating variables based on the original model (1).

Information accessibility

As shown in column (1) of Table 6, the digital divide significantly reduces people's information accessibility, and column (2) indicates that when the model controls for information accessibility, the digital divide still has a significant inhibitory effect on formal employment, with this mediating effect of 5.46%. The regression results confirm hypothesis H2a.

Table 6

Analysis of the mediating effect of information access.
	(1)	(2)
Variable	Infor	Mainwork
Dig	-0.013***	-0.015***
	(0.001)	(0.001)
Infor		0.063***
		(0.014)
Controls	YES	YES
Obs.	11951	11951
Pseudo R²	0.038	0.228
Note: *, , * indicate significant at 1%, 5%, 10% level respectively, standard error in parentheses

Risk tolerance.

As shown in column (1) of Table 7, the digital divide significantly reduces people's risk tolerance, and column (2) indicates that when controlling for risk tolerance, the digital divide still has a significant inhibitory effect on the propensity to formal employment, with a mediating effect of 5.87%. The regression results confirm hypothesis H2b.

Table 7

Analysis of mediating effects of risk preferences.
	(1)	(2)
Variable	Risk	Mainwork
Dig	-0.020***	-0.015***
	(0.001)	(0.001)
Risk		0.044***
		(0.013)
Controls	YES	YES
Obs.	10941	10941
Pseudo R²	0.027	0.222
Note: *, , * indicate significant at 1%, 5%, 10% level respectively, standard error in parentheses

Group test

This section designs the following model (5) to examine the differential effect of the digital divide on the nature of workers' employment.

$${Mainwork}_{i}={{\lambda }}_{1}+{{\lambda }}_{2}Di{g}_{i}+{{\lambda }}_{3}{DDummy}_{i}+{{\lambda }}_{4}Di{g}_{i}\ast {DDummy}_{i}+{{\lambda }}_{5}Control{s}_{i}+{\text{ϵ}}_{i}$$

$DDummy$ Respondents are grouped according to their different characteristics, representing whether they are from rural areas (take 1 if they are from rural areas (Rural); otherwise, take 0), whether they are from mid-west areas (set two dummy variables respectively, take 1 if they are from mid/west areas (Mid/West); otherwise, take 0), whether they are from first-tier cities (take 1 if they are from non-first-tier cities ( Notdeveloped), whether the household asset-liability ratio is high (grouped according to the median of the sample, if the respondent's household asset-liability ratio is greater than the median, it is defined as a high asset-liability ratio household (High_ROA), take 1; otherwise, take 0).

The regression results in Table 8 demonstrate the differential effect of the digital divide on the nature of workers' employment. Column (1) shows that the digital divide reduces the probability of rural workers entering formal employment channels (-0.008, t = 0.003) and is significant at the 1% level. Column (2) shows that the digital divide has a greater inhibitory effect on the entry of workers into formal employment channels in the central and western regions compared to the eastern regions (0.007,t = 0.003; 0.005,t = 0.003) and is significant at least at the 10% level. Column (3) shows that the inhibitory effect of the digital divide on the entry of non-Tier 1 cities into formal employment channels is even greater (-0.004,t = 0.003) and significant at the 1% level. This may be due to the fact that cities, although absorbing part of the employment group, appear to be lacking in social security and therefore are not fully open to workers[32]. The coefficient of column (4) is not significant (-0.001,t = 0.002), but the negative sign indicates, to some extent, that for workers with poorer economic status may enter informal employment channels. The above regression results show that the digital divide has more obvious characteristics of urban-rural differences, regional differences, urban differences, and family differences, and the negative effects of the digital divide are more significant for workers with poorer family economic status and living in underdeveloped areas.

Table 8

Heterogeneity analysis.
Variable	Mainwork
Variable	(1)	(2)	(3)	(4)
Dig	-0.017***	-0.018***	-0.011***	-0.015***
	(0.001)	(0.002)	(0.002)	(0.002)
DigRural*	-0.008***
	(0.003)
Rural	-0.513***
	(0.112)
DigMid*		0.007**
		(0.003)
DigWest*		0.005*
		(0.003)
Mid		-0.414***
		(0.115)
West		-0.353***
		(0.106)
Dig Notdeveloped*			-0.004*
			(0.003)
Notdeveloped			-0.238**
			(0.095)
DigHgih_ROA*				-0.001
				(0.002)
Hgih_ROA				0.195
				(0.131)
Controls	YES	YES	YES	YES
Obs.	11965	11965	11965	11965
Pseudo R²	0.229	0.229	0.238	0.227
Note: *, , * indicate significant at 1%, 5%, 10% level respectively, standard error in parentheses

The subjective impact of the digital divide

The aforementioned analysis probes into the objective impacts of the digital divide on workers' employment behavior. Nonetheless, even when workers are formally employed and enjoy excellent social security and superior welfare benefits, the digital divide may still incite shifts in their subjective attitudes. To address this, the current study investigates the correlation between the digital divide and workers' subjective attitudes (Subatt), utilizing the resident happiness question from the questionnaire (responses ranging from 'very unhappy' to 'very happy', assigned numerical values from 1 to 5, respectively, to denote increasing subjective happiness).

The regression results illustrate in Table 9, the digital divide significantly diminishes workers' subjective well-being. In summary, the digital divide lowers the likelihood of workers' formal employment, leading to adverse objective consequences. Simultaneously, it incites a negative subjective effect by decreasing workers' subjective well-being.

Table 9

Subjective impact test of digital divide.
	(1)
	Subatt
Dig	-0.007***
	(0.001)
Controls	YES
Obs.	4625
R ²	0.031
Standard errors are in parenthesis, * p < 0.01, p < 0.05, * p < 0.1

Research findings

The advent of the digital economy stands as a pivotal characteristic of the Fourth Industrial Revolution, with digital technology representing both its core essence and a critical instrument fostering balanced economic growth. The adept mastery and utilization of this 'tool' are vital parameters determining whether various regions and demographics can partake in the benefits of economic development.Drawing upon the principles of labor market segmentation theory and employing data from the 2019 CHFS, this study conducts an empirical exploration into the manner in which a pronounced digital divide impairs individuals' access to conventional employment channels. The digital divide exerts influence on individuals' employment decision-making processes by impacting their accessibility to information and risk tolerance levels.

The study's heterogeneity analysis illuminates a more considerable inhibitory effect of the digital divide on access to formal employment channels among workers located in mid-western, non-metropolitan, and rural areas as well as those with high levels of household indebtedness. These findings highlight notable disparities in the urban-rural, regional, and household realms. The study further discerns that, from an objective standpoint, the digital divide diminishes workers' likelihood of securing formal employment. Concurrently, from a subjective standpoint, it also affects their perception of well-being.

Realistic revelation

Firstly, in light of the robust growth of the digital economy, data and information associated with workers are steadily evolving into key resource components. However, there has been no concurrent enhancement in workers' social welfare benefits. Therefore, while acknowledging the various positive impacts of the digital economy, it is equally important to address issues such as workers' social security and income. Simultaneously, we need to consider the adverse effects engendered by the digital divide, striving to enhance workers' living standards and promote high-quality employment in China from a more equitable standpoint.

Secondly, the depth of the digital divide is frequently attributed to the requirement for advanced skills and application know-how. Consequently, alongside fortifying infrastructural development and boosting internet penetration, it is imperative to augment workers' internet usage skills. This could be accomplished by designing more user-friendly applications and minimizing the barriers to use for demographics like the older "silver-haired" group.

Thirdly, to steadily bridge the digital divide and enhance human capital investment in the digital era, the strategy needs to encompass not only the elevation of digital literacy but also the application of targeted measures considering the factors influencing human capital development. This may include bolstering legal safeguards for workers, thereby enabling secure access to information, more realistic risk expectations, and a wider array of social capital accumulation methods. All these steps will contribute to fostering a healthy labor market.

Lastly, amidst the vigorous development of the digital economy, efforts should be concentrated on diminishing familial, urban-rural, urban, and regional disparities. The establishment of robust social security mechanisms should be prioritized, with online education and government-led training being deployed as key instruments to continually improve human capital, thereby augmenting the benefits for "disadvantaged" members of society.

Data source

Author Contributions

Conceptualization: Qiankun Wang.

Data curation: Chunhua Zhang, Guanyu Hou.

Formal analysis: Qiankun Wang.

Methodology: Chunhua Zhang,Qiankun Wang

Software: Qiankun Wang,Guanyu Hou.

Supervision: Guanyu Hou..

Validation: Qiankun Wang,Chunhua Zhang, Guanyu Hou.

Visualization: Guanyu Hou.

Writing – original draft: Qiankun Wang,Chunhua Zhang.

Writing – review & editing: Guanyu Hou.

Acemoglu D, Restrepo P.,“Robots and jobs: Evidence from US labor markets”. Journal of Political Economy, 2020, 128(6): 2188–2244.
Autor D H, Levy F, Murnane R J.,“The skill content of recent technological change: an empirical exploration”.The Quarterly journal of economics, 2003, 118(4): 1279–1333.
Forman C, Goldfarb A, Greenstein S.,“The Internet and local wages: a puzzle”. American Economic Review, 2012, 102(1): 556 − 75.
Li Yi, Ke Jiesheng., “Three-level digital divide:income growth and income distribution effects of rural digital economy”. Agricultural Technology Economics, 2021(8):14.
Chen Xiaojie,He Guangwen,Chen Yang. “Digital divide and farmers' digital credit behavior - based on data from the 2019 Rural Financial Inclusion Survey in less developed areas”. Finance and Economics Series,2022(1):46–56.
Ghobadi S, Ghobadi Z.“How access gaps interact and shape digital divide: a cognitive investigation“. Behaviour & Information Technology, 2015, 34(4): 330–340.
Castells M.༌“The information age: economy, society and culture (3 volumes)”. Blackwell, Oxford, 1996, 1997: 1998.
OECD.༌“Understanding the Digital Divide.OECD Digital Economy Papers 49, OECD Publishing, 2001.
Attewell P.,“ Comment: The first and second digital divides”. Sociology of education, 2001, 74(3): 252–259.
Riccardini F, Fazio M.༌“Measuring the digital divide[C//IAOS Conference on Official Statistics and the New Economy. 2002, 27: 29.
Kong, Gao-Wen, Liu, Sasha, Kong, Dong-Min.༌“Robotics and employment-an exploratory analysis based on industry and regional heterogeneity”. China Industrial Economics,2020(8):80–98.
Song, Lin, He, Yang.༌“The impact of Internet use on employment choices of rural laborers in China”. China Population Science,2020(3):61–74 + 127.
Cao J, Zhou YL.༌“Research progress on the impact of artificial intelligence on economy”. Dynamics of Economics, 2018(1):13.
DiMaggio P, Bonikowski B.༌“Make money surfing the web? The impact of Internet use on the earnings of US workers”. American Sociological Review, 2008, 73(02): 227–250.
Xu Linqing.༌“Analysis of the impact of labor market segmentation on rural labor supply behavior”. Economic System Reform,2008(3):36–39.
CHEN Guifu,HAN Jing,HAN Kai-Ming.༌“ Urban digital economy development, skill-biased technological progress and labor underemployment”. China Industrial Economics,2022,No.413(08):118–136.
Stevenson B.,“The impact of the Internet on worker flows”. The Wharton School, University of Pennsylvania, 2006, 53(11): 1–24.
Zhang, Wanting.,“Digital divide and farm household entrepreneurship”. Journal of Shanxi University of Finance and Economics,2022(2):103–114.
Yin C-Chao,Gong Xue,Guo Peiyao.,“The impact of mobile payment on entrepreneurship-micro evidence from the China Household Finance Survey”. China Industrial Economics,2019(3):119–137.
Du Yao,Wang Zhong,Huang Jianye.༌“Did the minimum wage reduce informal employment? -- An empirical study based on mobile population data”. Southern Population,2019(3):69–80.
Li S.E.,Zhao Z.༌“The evolution of employment relationship in urban labor market and its influencing factors”. Economic Research,2012(9):85–98.
Zhao Jian.༌“A review of research on temporary employment ”. Economic Dynamics,2011(7):131–136.
Yin C-Chao,Jiang Jia-Ling,Yan Yu.“Does the digital divide affect household income”. Finance and Trade Economics,2021(9):66–82.
Li Xuan,Yang Zhen,Zhang Yaqian.,“Why do investors' subjective and objective risk preferences differ - micro evidence from CHFS ”.. Southern Economics,2015(11):16–35.
Grossman M. 1.,“ On the Concept of Health Capital and the Demand for Health”.. Columbia University Press, 2017: 6–41.
Yin C-Chao,Jiang Jia-Ling,Yan Yu.,“Does the digital divide affect household income”.Finance and Trade Economics,2021(9):66–82.
Zhang, Yao Yao.,“ Digital divide and farm household entrepreneurship”. Journal of Shanxi University of Finance and Economics,2022(2):103–114.
Yin C-Chao,Gong Xue,Guo Peiyao.༌“ The impact of mobile payment on entrepreneurship–micro evidence from the Chinese Household Finance Survey”. China Industrial Economics,2019(3):119–137.
Stock J H, Wright J H, Yogo M.༌“A Survey of Weak Instruments and Weak Identification in Generalized Method of Moments”. Economic Statistics, 2002,20(4):518–529
J. Fang,Z.L. Wen,M.Q. Zhang.༌“Analysis of mediating effects of categorical variables”. Psychological Science,2017(2):471–477.
Wen Zhonglin,Ye Baojuan.༌“ Mediated effects analysis:Methods and model development”. Advances in Psychological Science,2014(5):731–745.
Wang Chunguang.༌“The impact of China's urban employment system on the survival and development of rural migrants moving to cities”. Journal of Zhejiang University (Humanities and Social Sciences Edition), 2006(5):5–15.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Does the digital divide reduce the probability of formal employment for workers？ --Empirical evidence based on CHFS microdata

Status:

Version 1

Abstract

Literature Review and Prior Research Analysis

Theory and hypotheses development

A Review of Research on the Digital Divide, Formal Employment

Theoretical assumptions

Econometric model construction and model selection

Econometric model construction

Data sources and variable selection

Data source

Variable description

Empirical Analysis

Descriptive statistics

Baseline regression and endogeneity test

Robustness test

Mechanism analysis

Further analysis

Group test

The subjective impact of the digital divide

Research Conclusions and Implications

Research findings

Realistic revelation

Data source

Declarations

References

Additional Declarations

Status:

Version 1