1 Introduction

A city is a large and permanent human ecosystem that provides its citizens with a wealth of services and opportunities (Kumar et al. 2020). For decades, rapid urbanization has not only promoted economic and social development, but also brought a series of urban problems such as population expansion, energy shortage, environmental pollution, and traffic congestion (Dong et al. 2019). In order to cope with the above problems and promote sustainable development, a new urban development strategy—smart city initiative emerges at the historic moment and rapidly set off an upsurge in the world. At this stage, more than 1000 smart cities have been launched or built all over the world, especially in Europe, North America, and Asia (Deloitte 2018). Smart city is a new type of urban construction project that aims to facilitate public and social services, optimize the allocation efficiency of factors and resources, promote economic growth, and upgrade the quality of life of citizens (Kumar et al. 2020; Yigitcanlar et al. 2018) by building digital infrastructure and promoting the development of digital technologies.

A large number of studies have found that local construction projects and public policies (such as empowerment zones and enterprise zones in the USA, national high-tech zones in China, infrastructure construction) play a positive role in promoting employment (Czurylo 2023; Garin 2019; Ham et al. 2011; Kim et al. 2012; Kim and Han 2016; King and Shackleton 2020; Majale 2008; Song et al. 2021). Compared with these public actions, smart city pays more attention to the improvement in connotation and is a comprehensive urban development practice integrating urban planning, industrial development, infrastructure upgrading, urban management innovation, public service improvement, and government function transformation. Therefore, the implementation and goal realization of smart city construction projects will have a greater, broader, and more lasting impact on urban employment from multiple aspects. The current global employment situation is still very serious, especially against the background of the impact of the COVID-19 epidemic and the slowdown of global economic growth (International Labour Office 2023). At the same time, the construction of smart cities is still valued by governments. By 2023, global spending on smart city initiatives will reach $189.5 billion.Footnote 1 Whether smart city construction can increase employment opportunities and optimize employment structure and then promote the development of smart cities through fuller employment and better employment structure, has become a problem worthy of attention and discussion.

The construction of smart cities in full swing has successfully attracted the interest of researchers around the world (Mora et al. 2017; Ingwersen and Serrano-López 2018; Zheng et al. 2020). Previous research on smart cities mostly focuses on the following two aspects, delimiting the definition and connotation of smart cities (Giffinger et al. 2007; Caragliu et al. 2011; Yigitcanlar et al. 2018; Zhu et al. 2022), and examining the impact of smart city construction on the economy and society (Caragliu and Del Bo 2018, 2019; Chu et al. 2021; Dong et al. 2021; Guo et al. 2022; Jiang et al. 2021; Wang et al. 2022; Wang and Deng 2022; Xin and Qu 2019; Xu et al. 2022). Only a few studies discuss the relationship between smart city construction and employment. Kim et al. (2016) found the employment creation effect of smart cities when using input–output method to study smart city-related industries in South Korea. Based on individual microscopic data. Luo et al. (2021) discovered that smart city policies can promote employment. However, they discuss the employment effects of smart cities from the perspectives of industries and individuals, which underestimates the impact of smart cities on employment to a certain extent, and they do not further explore the impact mechanism and heterogeneity. Therefore, it is urgent to examine the impact of smart city construction on employment from the perspective of urban employment. Then, the question worth further thinking about is, as a new concept and model of urban development, can smart city construction really promote urban employment? If so, how much does it promote employment? What mechanisms does it function? Is there urban heterogeneity and sector heterogeneity in its impact? For these questions, existing studies do not provide clear answers. The answers to these questions will not only help deepen the academic community’s understanding of smart city development and construction, but also provide policy makers with new insights on the socioeconomic effects of smart city construction.

In 2012, China issued the Notice on Carrying out the National Smart City Pilot Work and established the first batch of smart city pilots. The second and third batches of smart city pilot lists were announced in 2013 and 2015, respectively. So far, nearly 300 cities in China have implemented smart city pilot policies. China’s smart city policies provide a research perspective to answer the above questions. Intelligent infrastructure, convenient public services, industrial transformation and upgrading, refined social governance, and sustainable development are the main goals of China’s smart city construction. The realization of these goals will promote the development of related industries and sectors (e.g., smart city industry chains) and increase the demand for employment. Meanwhile, the smart city construction policy means that resources and elements will continue to gather and spread in pilot cities and more development opportunities, which will increase the attractiveness and talent supply of pilot cities. Moreover, smart cities can use digital technologies such as Internet of things (IoTs), cloud computing, and big data to facilitate the efficiency of urban construction and management, reduce the transaction costs of enterprises, increase the income and productivity of enterprises, and then better play the agglomeration effect and scale effect (Luo et al. 2021), improve the attraction of talents, and ultimately promote employment. It is foreseeable that in the next few years, smart city construction will have a profound impact on urban employment. Therefore, taking the first batch of smart city pilots in China as a quasi-natural experiment and using the DID model, this paper aims to evaluate the impact of smart city construction on urban employment and employment structure, so as to provide empirical evidence and policy reference for smart city construction, employment quality improvement and employment structure optimization in China and other developing economies.

It should be noted in advance that smart city construction involves a wide range of issues, and its impact on employment is predictably complex. There are two mainstreams in the current discussion on how to build a smart city, namely a people-oriented approach and a technology-led approach (Ahvenniemi et al. 2017; Kummitha and Crutzen 2017; Mora et al. 2017, 2019). The former focuses on people’s participation and technology serving people and is considered as the most suitable for smart city development (Mora et al. 2019), while the latter emphasizes the key role of digital technology, especially information communication technology (ICT). Although there is some disagreement, neither of them deny that digital technology represented by ICT has become an important means of shaping smart cities (Luque-Ayala and Marvin 2015; Viitanen and Kingston 2014). In fact, using digital technology to improve urban public services and the quality of life is another important goal of smart city construction. Hence, this paper intends to explore how smart city construction can promote urban employment from the perspective of its impact on digital technology development and public services.

As an attempt to systematically evaluate the effect of smart city construction on employment, the possible contributions of this paper are as follows: First, this paper enriches relevant literature on the implementation effect of smart city construction. When exploring the impact of smart city construction on the economy and society, the existing literature has not conducted an in-depth investigation on the relationship between smart city construction and employment. This paper makes up this gap. Second, from the perspectives of digital economy development and public services, this paper accurately identifies the mechanism of new local policies, smart city construction, affecting employment, and establishes a complete theoretical framework for smart city construction to promote employment, thereby expanding the relevant research on the employment effect of public policy/construction projects. Third, the findings of this paper have strong policy implications. This study discusses the heterogeneous impact of smart city construction on urban employment from the aspects of geographical location, city size, and urban characteristics and further considers its effect on employment structure, which can provide empirical support for the formulation and promulgation of relevant policies.

The rest of this paper is organized as follows. Policy background and research hypothesis are described in Sect. 2. Section 3 outlines methodology, variables, and data in this study. Section 4 discusses the results of empirical analysis, including robustness tests and mechanism analysis. Section 5 explores the results of heterogeneity analysis. Section 6 further examines the impact of smart city construction on urban employment structure. Section 7 concludes the paper and gives policy implications.

2 Background and research hypothesis

In this section, we first introduce the policy background of smart city construction in China. Then, the total impact of smart city construction on employment is discussed and Hypothesis 1 is proposed. Finally, the impact mechanisms of smart city construction on urban employment are analyzed and Hypotheses 2 and 3 are given.

2.1 Smart city policy in China

The idea of a smart city is originated from the note of a smart earth proposed by IBM. The Chinese government put forward the concept of smart city in 2009 and believed that smart city is a new model to strengthen urban planning, construction and management through comprehensive application of modern science and technology, integration of information resources, and coordination of business application systems. In 2012, the Ministry of Housing and Urban–Rural Development officially issued the Notice on Carrying out the National Smart City Pilot Work and the first batch of 90 smart city pilots, and clearly indicated that building smart cities is an important measure to implement innovation-driven development, promote new urbanization, and build a well-off society in an all-round way. The National Smart City Pilot Interim Management Measures and the National Smart City (District, Town) Pilot Index System (Trial) were also issued. In 2013 and 2015, the second and third batches of pilot cities were announced successively, and a total of more than 290 cities (districts, counties and towns) became national pilot smart cities. Subsequently, the National New Urbanization Plan (2014–2020) incorporated smart city construction into the national strategic planning for the first time and pointed out the direction of smart city construction. In 2015, smart city construction was included in the government work report for the first time. The following year, the 13th Five-Year Plan clarified that modern information technology and big data should be fully utilized to build a number of new demonstration smart cities. In 2017, the construction of a “smart society” appeared in the Report of the 19th National Congress. In 2020, the 14th Five-Year Plan further proposed that in line with the new concepts and trends of urban development, improve the level of urban intelligence/smart, and build a livable, innovative, smart, green, humanistic, and resilient city. It can be seen that smart city construction has been highly valued by the central governments in recent years. This paper aims to provide empirical evidence of how China’s smart city construction affects the number and structure of employment across cities.

2.2 The total effect of smart city construction on urban employment

Smart city construction is a huge systematic project, which requires a great deal of both physical capital and human capital (or “smart people”) (Caragliu et al. 2011). In general, smart city construction promotes urban employment from both demand side and supply side. From the demand side, the smart/intelligent construction of cities provides a large number of employment opportunities for the labor force. The construction, maintenance, transformation, and upgrading of infrastructure can promote employment (Majale 2008; Garin 2019; King and Shackleton 2020). Smart city construction can also improve the construction level of new infrastructure, increase the demand for professional, technical, and managerial talents, and further raise the human capital level of cities (Schatz and Johnson 2007; Bakici et al. 2013). From the supply side, the smart/intelligent development of cities can attract labor and promote employment. Smart city construction accelerates the generation and wide application of new technologies and creates conditions for the development of new industries and high-tech industries (Barba-Sanchez et al. 2019; Cao et al. 2019), thus attracting more highly skilled, highly educated, and high creative talents. In addition, smart city construction facilitates the coordination of human capital, financial resources, and information technology (Jiang et al. 2021) and greatly improves urban productivity, which attracts more labor.

The impact of smart city construction on urban employment should be mainly reflected nonfarm sectors, namely the secondary and tertiary industries. On the one hand, although smart city construction increases the demand for agricultural products and facilitates the development of agriculture focusing on agricultural production, the trend of labor transfer from the agricultural sector to the non-agricultural sector with higher productivity and wages has always existed. On the other hand, smart city construction, which regards digital technology and public services as an important grasp, obviously requires more non-agricultural labor. Therefore, the following hypothesis is proposed:

Hypothesis 1

Smart city construction can increase total employment and employment of the secondary and tertiary industries.

2.3 Mechanisms of smart city construction affecting urban employment

2.3.1 Digital technology development

Digital technology, represented by ICT, is the key driving force for cities to address a series of social, economic, and environmental problems in a “smart” way (Batty et al. 2012; Bibri and Krogstie 2017; Bifulco et al. 2016; European Parliament 2014). In order to achieve precise, digital, and dynamic management, smart cities are bound to integrate and calculate information and data from multiple aspects (Guo et al. 2022). Traditional independent processing methods are obviously unable to satisfy the complex requirement in various business scenarios (Yang et al. 2020). Driven by policy support and practical needs, technologies such as big data analysis, deep learning, digital twin, ICT, and IoTs will see significant development. In short, smart city construction is conducive to the innovation and development of digital technology (Luo et al. 2021). There is a virtuous circular interaction between the two, making digital technology one of the most important features of smart cities (Cocchia 2014).

Technological progress/change can increase employment in general (Lv and Hao 2018; Borland and Coelli 2017; Gregory et al. 2021; Autor 2014). Similarly, digital technology development will also have a positive impact on total employment. First, digital technology development satisfies the labor force’s information acquisition needs and reduces the labor force’s information search cost, which promotes entrepreneurship and employment (Appio et al. 2019). Second, the penetration of digital technology into traditional industries will generate new business models and industries (Li et al. 2020; Pan et al. 2020) and increase employment. Third, digital technology development improves the allocation efficiency of capital and labor, thus raising the production efficiency of enterprises and increasing the demand for positions that cannot be automated (Autor and Dorn 2013; Goos et al. 2014; Autor 2015). Fourth, digital technology development will create new jobs, which may be more complex versions of existing jobs, or entirely new jobs (Fossen and Sorgner 2022). Based on the above analysis, the following hypothesis is proposed:

Hypothesis 2

Smart city construction increases total employment by facilitating digital technology development.

It should be noted that the impact of digital technology development on employment in different industries cannot be determined. This is based on the following facts. First, smart city construction promotes the development of both labor-reinstating digital technologies and labor-displacing digital technologies (Fossen and Sorgner 2022), which have different impacts on employment. Second, there are significant differences in labor skill structure, proportion of routine tasks, and degree of labor intensiveness among different industries, resulting in inconsistent responses to digital technology.

2.3.2 Public services

People-oriented is one of the service principles of smart cities (Ahvenniemi et al. 2017) and also the significance and value of smart city construction (Zhu et al. 2022). Lee and Lee (2014) once defined the idea of smart cities as a city where urban residents can obtain smart services regardless of time and place. Smart cities can help improve haze, traffic congestion and health inequality (Chen et al. 2016), enhance urban public service supply capacity (Praharaj et al. 2018; Zanella et al. 2014; Lee et al. 2014; Marsal-Llacuna et al. 2014; Han et al. 2022), and transform traditional cities into a better and more livable place(Cocchia 2014).

On the other hand, good infrastructure and public services can improve the quality of life and urban work experience (Albino et al. 2015; Caragliu et al. 2011; Caragliu and Del Bo 2012) and attract high-quality talents (Wang and Deng 2022). By providing high-quality of life in terms of security, comfort, convenience, and entertainment infrastructure, smart cities become natural environments for creative talents to gather, learn, and share (Jiang et al. 2021). Shapiro (2005) also indicated that the improvement in productivity and quality of life in smart cities was an important reason to boost employment growth by using panel data of US metropolitan areas. In light of the above discussion, the following research hypothesis is put forward:

Hypothesis 3

Smart city construction increases total employment and employment in the secondary and tertiary industries by improving the level of public services.

The research framework of this paper is shown in Fig. 1.

Fig. 1
figure 1

The research framework of this paper

Full size image

3 Methodology, variables, and data

3.1 Sample selection and data specification

China established the first batch of pilot cities for smart city policies in 2012 and expanded the pilot scope in 2013 and 2015. The following approach was used to select the treatment and control groups. First, since the research object of this paper is prefecture-level cities, the prefecture-level cities where the pilot districts and counties are located are eliminated from the sample. If these prefecture-level cities are directly regarded as pilot cities, the impact of smart city construction on urban employment will be underestimated. Second, to improve estimation accuracy and expand the time range of policy estimation, the first batch of pilot prefecture-level cities in 2012 is selected as the treatment group, and the newly added pilot smart cities in 2013 and 2015 are excluded. Finally, the prefecture-level cities with serious data missing or with administrative changes between 2003 and 2019 are removed.

The initial data are mainly obtained from the China City Statistical Yearbook and the statistical yearbooks for each province and city. To achieve accurate results, the raw data are processed as follows: (1) For some missing values, fill in by searching provincial and municipal statistical bulletins and statistical yearbooks and then use interpolation to complete the still missing data. (2) To eliminate the effect of inflation, all nominal variables were at 2003 constant price parity. Therefore, we finally obtained the panel data of 158 prefecture-level cities from 2003 to 2019, including 33 pilot cities and 125 non-pilot cities.

3.2 Model setting

3.2.1 DID

China has officially implemented smart city policy since 2012 and carried out smart city pilots in three batches. A total of more than 290 cities (districts, counties and towns) have been included in smart city pilots. Pilot cities are determined by the Ministry of Housing and Urban–Rural Development according to several standards, which has certain randomness and can be used as an exogenous impact event to design a quasi-natural experiment. DID is the most widely used policy evaluation model. Its basic idea is to examine the performance of policy implementation by effectively separating “time effect” and “policy processing effect”. Therefore, this paper selects DID to evaluate the impact of smart city construction on urban employment. In line with Dong et al. (2021) and Jiang et al. (2021), the following DID model was constructed for the direct conduction mechanism:

$$\begin{aligned} emp_{it}/pri_{it}/sec_{it}/ter_{it}=\beta _0+\beta _1did_{it}+\beta _2Z_{it}+\mu _i+\delta _t+\varepsilon _{it} \end{aligned}$$
(1)

where i represents the city, t represents the year, and \(emp_{it}\), \(pri_{it}\), \(sec_{it}\), and \(ter_{it}\) represent the total employment, primary industry employment, secondary industry employment, and tertiary industry employment, respectively. \(did_{it}\) represents the interaction term between the city dummy variable and the time dummy variable; if a city implements smart city policies in 2012, its did variable in 2012 and subsequent years is set to be 1 and otherwise 0. Next, \(Z_{it}\) represents the control variables, such as urban population, GDP per capita, urbanization rate, and wages, which may affect employment; \(\mu _i\) represents the city fixed effect; \(\delta _t\) represents the time fixed effect; \(\varepsilon _{it}\) represents the random disturbance term. \(\beta _1\) measures the relative impact of smart city construction on employment. If \(\beta _1>0\), this means that smart city construction promotes employment; otherwise, it inhibits the latter.

3.2.2 PSM–DID

In fact, there is a great deal of heterogeneity between cities, and it is difficult to have completely consistent temporal effects. Hence, before using the DID model, it is best to select a group of “non-pilot cities” whose characteristics are as similar as possible to the treatment group as a control group. According to Guo et al. (2022) and Xu et al. (2022), this paper further uses propensity score matching (PSM)-DID to ensure the robustness and scientificity of the conclusion. As the name implies, PSM–DID is a combination of PSM and DID, which can effectively avoid sample selection bias and endogenous problems. The specific steps for PSM–DID are as follows: First, PSM was adopted to find the control groups as similar as possible to the treatment group. Second, the matched treatment and control group are used to re-estimate model (1).

Furthermore, selecting the appropriate covariate is a prerequisite for PSM. In general, variables that affect both employment and smart city pilot policies need to be taken into account in the scope of covariates. Information and communication infrastructure is an important foundation for smart city construction. The government may give priority to cities with higher level of informatization when determining smart city pilots. Therefore, in addition to the original control variables, the scope of covariates should also include urban informatization level (infor). Referring to the stepwise regression analysis used by Xu et al. (2022), covariates ultimately used for matching include control variables (except wag) and infor. infor is the note of informatization level, measured by the number of Internet users per 100 people.

3.2.3 Mediating effect model

According to the previous analysis, this paper examines whether digital technology development and public services are mediating variables between smart city construction and urban employment. Referring to existing literature (Guo et al. 2022; Wang and Deng 2022), the mediating effect models are as follows:

$$\begin{aligned}{} & {} M_{it}=\gamma _0+\gamma _1did_{it}+\gamma _2Z_{it}+\mu _i+\delta _t+\varepsilon _{it} \end{aligned}$$
(2)
$$\begin{aligned}{} & {} emp_{it}/pri_{it}/sec_{it}/ter_{it}=\varphi _0+\varphi _1did_{it}+\varphi _2M_{it}+\varphi _3Z_{it} +\mu _i+\delta _t+\varepsilon _{it} \end{aligned}$$
(3)

where \(M_{it}\) is an intermediary variable, representing digital technology development (\(digital_{it}\)) or public services (\(service_{it}\)). Model (2) is first used to verify the driving effect of smart city construction on digital technology development or public services. Then, the impact of smart city construction and mediating variable \(M_{it}\) (\(digital_{it}\) or \(service_{it}\)) on employment can be determined through model (3). The total effect of smart city construction (\(did_{it}\)) is \(\beta _1\) in model (1), the direct effect is \(\varphi _1\), and the mediating effect of \(M_{it}\) (\(digital_{it}\) or \(service_{it}\)) is \(\gamma _1\varphi _2\). According to Baron and Kenny (1986), the existence of mediating effect can be judged by the significance of \(\beta _1\), \(\gamma _1\), \(\varphi _1\), and \(\varphi _2\) and the magnitude of \(\beta _1\) and \(\varphi _1\).

3.3 Variables

3.3.1 Dependent variables

The dependent variable of this paper is the number of urban employment, including total employment (emp), primary industry employment (pri), secondary industry employment (sec), and tertiary industry employment (ter). Specifically, the total employment is represented by the number of employees in urban units at year end, and the employment in the three industries is represented by the number of employees in the primary, secondary, and tertiary industries, respectively (unit: 10,000 people).

3.3.2 Independent variable

The independent variable (did) is smart city policies. This paper takes the pilot policy of smart city construction as a quasi-natural experiment and determines the independent variable based on the first batch of national smart city pilot lists. If a city implements smart city policies in 2012, its did variable in 2012 and subsequent years is set to be 1 and otherwise 0.

3.3.3 Mediating variables

Digital technology development (digital). Drawing on Huang et al. (2019) and Zhao et al. (2020), this variable includes informatization level (Internet users per 100 people), Internet-related employment (information transmission, computer services and software employment/total employment), Internet-related output (total telecommunications business per capita), and mobile Internet users (mobile phone users per 100 people). Through the method of principal component analysis, the data of the above four indicators are standardized and dimensionally reduced, and the digital technology development index obtained is recorded as digital.

Public services (service). In line with existing literature (Dong and Wang 2021; Xu et al. 2022), this variable includes transportation (the per capita road area in municipal districts), education (education expenditure per capita, the ratio of teachers to students in primary schools, and the ratio of teachers to students in middle schools), medical care (the number of doctors per 10,000 people, the number of hospital beds per 10,000 people, the number of hospitals per 10,000 people), and cultural services (public library holdings per 10,000 people). In the same way as the establishment of digital technology development, the resulting public service index is denoted as service.

The index selection and construction process of mediating variables are shown in Table 1.

Table 1 Index selection and construction of mediating variables
Full size table

3.3.4 Control variables

To analyze the impact of smart city construction on employment more comprehensively, it is also necessary to set control variables that may affect employment. Based on the existing research (Dong 2018; Huang and Wei 2022; Luo et al. 2021; Yang et al. 2020), the control variables are selected as follows: (1) Economic development level (pgdp), expressed by the logarithm of per capita GDP. Cities with higher economic development level can provide more jobs for workers. (2) Population density (pop), expressed by the ratio regional population to regional area. Population density represents the abundance and demand scale of regional labor force, and the greater the population density, the more jobs there are. (3) Urbanization (urb), expressed by the ratio urban population to regional population. (4) Government intervention (gov), expressed by the ratio government fiscal expenditure to GDP. Government intervention is an important force to ensure employment stability and improve employment levels. (5) Financial development level (fin), expressed by the ratio of the total balance of deposits and loans to GDP. Regional financial institutions provide capital guarantee for the development of enterprises and then enhance the demand of enterprises for labor. (6) FDI (fdi), expressed by the ratio actual use of foreign capital to GDP. (7) Average wage (wag), expressed by the logarithm of the average wage of on-the-job employees. Wage is a direct factor affecting labor employment. (8) Human capital (hum), expressed by the number of college students per 100 people. Human capital is an important factor affecting employment. The variable descriptive statistics are shown in Table 2.

Table 2 The statistical description of variables
Full size table

4 Results of empirical analysis

4.1 Benchmark regression results based on DID

The estimated results of model (1) are reported in Table 3, in which columns (1), (3), (5), and (7) do not add control variables. Columns (1) and (2) show that smart city construction (did) has a considerable favorable effect on city’s total employment (emp), with estimated coefficients of 8.627 and 6.968, which are significant at the 1\(\%\) level. In other words, smart city policies have greatly increased employment in pilot cities. After the implementation of the smart city pilot construction, the total employment in pilot cities is 69,680 higher than the cities without the implementation of the policy, on average. Column (3) illustrates that, before adding control variables, the effect of smart city construction on primary industry employment is significant positive. After adding control variables, the estimated coefficient of did in column (4) is positive, and the significance is lower than that in column (3), that is, smart city construction has promoted primary industry employment to a certain extent. According to columns (5) and (6) of Table 3, smart city construction has significantly promoted employment in the secondary industry, with estimated coefficients of 5.373 and 4.622, indicating that smart city construction has increased secondary industry employment in pilot cities. It can be seen from columns (7) and (8) that the estimated coefficients of did are 2.919 and 2.014, respectively, which are significant at the 1\(\%\) level, suggesting that smart city construction has also promoted tertiary industry employment in pilot cities. The above empirical results support Hypothesis 1.

Table 3 The results of the baseline regression
Full size table

4.2 Event study

Before using the DID model, it is necessary to test whether the treatment and control groups meet the common trend. For the purpose of this study, the above empirical results may be overestimated if the treatment and control groups do not satisfied the common trend assumption prior to the implementation of the smart city pilot policy. Therefore, this paper replaces the independent variable on the basis of the benchmark regression model (1) and builds the following model:

$$\begin{aligned} emp_{it}/pri_{it}/sec_{it}/ter_{it}= & {} \alpha _0+\sum \limits _{j=-8,j\ne -1}^{j=6}\beta _jtreat_{i}\nonumber \\{} & {} \quad \times year_j+\alpha _1Z_{it}+\mu _i+\delta _t+\varepsilon _{it} \end{aligned}$$
(4)

where \(treat_{i}\) represents whether city i is a pilot city, \(year_j\in (-8,-2)\cup (0,6)\) represents the dummy variables of policy implementation in the seven years before 2011 and in the six years after 2011, and \(\beta _{-8}\), \(\beta _{-7}\), \(\cdots \), \(\beta _{6}\) (except \(\beta _{-1}\)) represent the policy implementation effect in 2004, 2005, \(\cdots \), 2018 (except 2011), respectively.

The results of the event study are presented in Table 4 and Fig. 2. The left half of Table 4 shows that coefficients of interaction term for the periods prior to pilot policy implementation are close to zero, meaning that the treatment and control groups have met the common trend assumption. The results in Fig. 2 are similar to those in Table 4 and will not be repeated here.

According to Garin (2019), this paper continues to analyze the long-term impact of smart city construction on urban employment using the results of event study. From columns (1), (3), and (4) in the right half of Table 4, it can be seen that the estimated coefficients from \(post\_2\) to \(post\_5\) are significant and gradually increasing and that the estimated coefficients of \(post\_6\) are decreasing and still significant compared to those of \(post\_5\). The results of (a), (c), and (d) in Fig. 2 are more intuitive. Therefore, the above evidence shows that smart city construction has a long-term and lasting impact on total employment and employment in the secondary and tertiary industries.

Table 4 The results of event study
Full size table
Fig. 2
figure 2

Event study

Full size image

4.3 Results based on PSM–DID estimation

To alleviate the problem of possible sample selection bias, this paper adopts PSM to balance the sample. Logit regression is used to calculate propensity scores, and kernel matching is used to match. Table 5 reports the result of the balancing test. After matching, the standardized deviation of each variable is significant reduced, that is, the mean value of the covariates is uniformly distributed, and the balance test requirement is satisfied.

Table 5 Balancing test of PSM (kernel density matching)
Full size table

Model (1) is re-estimated for the matched sample, and Table 6 presents the results of the PSM–DID estimation. It can be seen from Table 6 that the estimated coefficients of did are significantly positive and reduced compared with the benchmark model, indicating that smart city construction still increases total employment and employment in three industries after using the PSM–DID model, which once again verifies Hypothesis 1.

Table 6 The results of PSM–DID estimation
Full size table

4.4 Robustness test

4.4.1 Change the dependent variables

To ensure the reliability of the main findings as much as possible, the robustness test was further conducted by changing the dependent variables. Using the number of employment to evaluate the employment effect of smart city construction may ignore the impact of city size on the estimated results. Therefore, the employment density is used to re-estimate the benchmark model. Employment density is expressed as the number of employees divided by the total area of the city. The results are shown in Table 7. Table 7 shows that smart city construction increases the total employment density and the employment density of the secondary and tertiary industries, indicating that the primary results of this paper are not affected after changing the dependent variables.

Table 7 The results of changing the dependent variables
Full size table

4.4.2 Add new pilot prefecture-level cities

In 2013 and 2015, China added two new batches of smart city pilots. To better estimate the impact of smart city construction on employment, these two batches of pilot prefecture-level cities were added to the original sample as the treatment group, and model (1) was estimated again. Table 8 displays the estimated results of adding new pilot prefecture-level cities. It can be seen that the results in Table 8 remain similar to the benchmark regression results, which once again reveals the robustness of the main findings in this paper.

Table 8 The results of adding new pilot prefecture-level cities
Full size table

4.4.3 Placebo test

In order to avoid the promotion effect of smart city construction on urban employment caused by accident, this paper adopts a random sampling method to conduct a placebo test (Guo et al. 2022). Specifically, a list of virtual smart city pilots was generated by random sampling, and then, model (1) is estimated. This process was repeated 500 times. If the estimated coefficients are mostly distributed around 0, the increase in employment is indeed caused by the smart city pilot policy in 2012. Otherwise, the findings in this paper are biased. The results of the placebo test are shown in Fig. 3. Figure 3b shows that some of the estimated coefficients exceed 0.334 (the probability that the absolute value of the estimated coefficient is greater than 0.334 is 11.6%), and the estimated coefficients do not follow a normal distribution, indicating that the impact of smart city construction on primary industry employment has not passed the placebo test. Furthermore, it can be found that the estimated coefficients are distributed near 0 and approximately normal distribution in Fig. 3a, c and d, implying that the promotion effect of actual smart city construction on total employment and employment in the secondary and tertiary industries is real.

Fig. 3
figure 3

Placebo test

Full size image

4.5 Mechanism analysis

The above conclusions fully demonstrate the positive impact of smart city construction on urban employment. So the question is, does smart city construction improve employment through digital technology development and public services as described in the previous theoretical analysis? To answer this question, the mediation effect model is adopted for further investigation. The results are shown in Table 9. From column (1) of panel A, we can observe that smart city construction promotes the regional digital technology development, and the estimated coefficient is significant at the level of 1\(\%\). Columns (2)–(5) of panel A report the results after adding digital technology development (digital) to the baseline regression model. As can be seen from columns (2) and (5) of panel A, digital technology development has a significant positive impact on total employment and tertiary sector employment. Compared with the benchmark regression results, the estimated coefficients of did are reduced, indicating that smart city construction can promote the total employment and employment of the tertiary industry by facilitating digital technology development. Hypothesis 2 is verified.

However, from column (4) of panel A in Table 9, we notice that the estimated coefficient of digital is negative and insignificant, suggesting that digital technology development is not a mediating variable between smart city construction and employment in the secondary industry. As emphasized by Fossen and Sorgner (2019), new digital technologies have different effects on various occupations due to the existence of labor-displacing and labor-reinstating types of digital technology. Based on individual-level data in Germany from 1994 to 2014, Dauth et al. (2017) discovered that the impact of robots on employment was not significant, because robots reduced employment in manufacturing, while increasing employment in service sector. The results of panel A in Table 9 are similar to Dauth et al. (2017), that is, digital technology compensates for the negative impact on secondary industry employment by increasing employment in the tertiary industry, thus showing a promotion effect on total employment.

In what follows, we discuss the mediating effect of public services. As can be seen from column (1) of panel B in Table 9, smart city construction improves the level of public services in pilot cities, and the estimated coefficient is 0.154 and significant at the level of 1\(\%\). This result is consistent with the conclusions of Dong and Wang (2021) and Xu et al. (2022). Columns (2), (4), and (5) of panel B show that the improvement in public services significantly promotes total employment and employment in the secondary and tertiary industries. It can be found that the estimated coefficients of the core explanatory variables are obviously decreased compared with the benchmark regression results, meaning that public services is a mediating variable by which smart city construction promotes total employment and employment in the secondary and tertiary industries. Hypothesis 3 is confirmed.

Table 9 The results of mediating effect model
Full size table

5 Heterogeneity analysis

5.1 Regional heterogeneity

In China, the eastern and central regions have rapid economic development and relatively complete infrastructure, thus having a first-mover advantage in building smart cities. However, the northeastern and western regions are relatively underdeveloped, and the effect of smart city construction may not be obvious enough. Therefore, the sample cities are divided into four regions of eastern, central, northeastern, and western to examine the regional heterogeneity of smart city construction. The results are shown in Table 10. Panel A in Table 10 displays that smart city construction has significantly boosted employment in the eastern and central regions.

Columns (1)–(3) of panel B in Table 10 demonstrate that in the northeastern region, smart city construction obviously enhances total employment and secondary industry employment, but it has a significantly negative impact on employment in the tertiary industry. Columns (4)–(6) of panel B in Table 10 depict the impact of smart city construction on employment in the western region, and the results illustrate that smart city construction reduces total employment and employment in the secondary and tertiary industries. The above results differ from the benchmark regression results. On the one hand, it shows that the effect of smart city construction in northeastern and western China has not reached the expected level. Tang et al. (2020) found that the development of new smart cities in eastern and central China is relatively good, higher than the national average, while that in northeast and western China is lower than the national average. Such regional differences are most obvious in the sharing, development and utilization of information resources. On the other hand, owing to strong economic and geographical advantages, the “siphon effect” of the eastern and central regions on talents of the northeastern and western regions still exists significantly.

Table 10 The results of regional heterogeneity
Full size table

5.2 Heterogeneity analysis of city size

According to the Notice on Adjusting the Criteria for the Classification of City Scale issued by the State Council of the People’s Republic of China in 2014 and the population of municipal districts of each city in 2012, the sample cities are divided into three groups: small cities, medium cities, and large cities. The heterogeneity results of city size are reported in Table 11. From Table 11, we can observe that for small cities, the impact of smart city construction on total employment and employment in the secondary and tertiary industries is no significant. The smart city construction of medium cities promotes total employment and secondary industry employment. For large cities, smart city construction increases total employment and employment in the secondary and tertiary industries.

It is very interesting that the construction of smart cities in medium cities mainly promotes employment in the secondary industry, while the construction of smart cities in large cities mainly promotes employment in the tertiary industry. This result is not difficult to explain. Compared to large cities, the infrastructure, especially digital infrastructure in medium cities, is relatively poor and imperfect. When building smart cities, the first task of these cities is the construction of new digital infrastructure and the upgrading of existing infrastructure, which will promote the development of the construction industry and equipment manufacturing, thereby increasing the demand for secondary industry labor. Relying on strong financial, technical personnel and infrastructure facilities, large cities can give full play to the effect of resource agglomeration, expand the performance of smart city construction, and create more jobs in the service industry.

Table 11 Heterogeneous impact based on city scale
Full size table

5.3 Heterogeneity in urban characteristics

Human capital, financial development, and informatization level support the implementation of smart city policies (Wang et al. 2022) and can influence the impact of smart city construction on employment. Jiang et al. (2021) also believe that smart city construction is a systematic project that requires the coordination of human capital, finance resource, and information technology. Given this, we further analyzed the heterogeneity in these urban characteristics. According to the research method applied by Jiang et al. (2021), model (1) was re-evaluated by replacing did with interaction terms \(did \times fin\), \(did \times hum\), and \(did \times infor\), respectively. Panels A–C in Table 12 display the results of the heterogeneity test on urban financial development, human capital, and informatization level, respectively.

Panel A indicates that the impact on employment of the smart city with higher financial development level is stronger than that with lower financial development level. An efficient financial system is an important support for the construction of smart cities (Huston et al. 2015). Inadequate financial development affects the ability of smart cities to attract new companies and start-ups as was highlighted by Bakici et al. (2013). A higher level of financial development can ease financing constraints, reduce financing costs, improve capital allocation efficiency, and facilitate the development of enterprises and industries related to smart city construction, thereby expanding employment demand. Therefore, the government can promote the construction of smart cities and urban employment by improving the levels of financial development.

Panel B in Table 12 demonstrates that the higher levels of human capital, the more significant the promotional effect of smart city construction on employment. Human capital is an important driving force of smart city construction. Cities with higher human capital are more innovative and can more quickly grasp and leverage the latest technologies for smart city construction, which will further improve productivity, boost economic development, and ultimately increase employment opportunities.

Panel C in Table 12 shows that smart city construction has a more significant positive impact on employment in cities with higher informatization level. The construction of smart cities relies on the development of information technology and digital economy (Luo et al. 2021). Pilot cities can take advantage of high level of informatization to build urban digital management platforms and information sharing platforms earlier, thus optimizing the allocation efficiency of various elements and resources, improving the income and productivity of enterprises and the efficiency of urban construction and management. These favorable factors will promote more employment. Therefore, the government should focus on improving the urban informatization level when building a smart city, so as to create favorable conditions for increasing employment.

Table 12 Heterogeneous impact based on human capital, finance development, and informatization
Full size table

6 Further discussion: the impact of smart city construction on urban employment structure

The above results show that the different impacts of smart city construction on employment do exist in cities and industries. Does this impact change the employment structure of pilot cities? To this end, this section will examine the impact of smart city construction on employment structure from the perspective of sectoral employment and employment proportion and provide further empirical support for policymakers to formulate a more solid hypothesis.

6.1 Urban sectoral employment

The impact of smart city construction on employment across sectors is investigated in this section. According to the two-digit China Industrial Classification System (CICS), the secondary and tertiary industries are further divided into 18 sectors. The specific results are given in Table 13. It can be found that the estimated coefficients of did in columns (2), (4), (5), (8)–(13), and (15)–(17) are significantly positive, suggesting that smart city construction has increased employment in Manufacturing (manu), Construction (cons), Wholesale and retail (wr), Information transmission, computer services, and software (ics), Finance (fin), Real estate (real), Leasing and business services (lbs), Scientific research and technical service (sci), Water conservancy, environment and public facilities management (wep), Education (edu), Health and social work (hsw), and Culture, sports, and entertainment (cse). Construction industry is related to infrastructure renovation and upgrading, Information transmission, computer services and software and Scientific research and technical service are related to digital technology development, Manufacturing and Finance contribute to the smart construction of cities, and other industries are mostly related to public services. This result means that smart city construction affects the existing sectoral structure of employment by promoting the development of different sectors.

Table 13 The results of urban sector heterogeneity
Full size table

6.2 Employment proportion

In this section, we further discuss the impact of smart city construction on the employment proportion of the three major industries. The employment proportion of the primary/secondary/tertiary industry is expressed as the ratio of employment in the primary/secondary/tertiary industry to total employment. Table 14 reports the estimated result of employment proportion. We find that in Table 14, smart city construction increases the employment proportion of the tertiary industry(terpro), reduces the employment proportion of the secondary industry(secpro), and has no significant impact on the employment proportion of the primary industry (pripro). Therefore, while increasing the employment of the secondary and tertiary industries, smart city construction also promotes the transfer of employment to the tertiary industry and further optimizes the urban employment structure.

Table 14 The estimated result of employment proportion
Full size table

7 Conclusions and inspirations

The concept of smart cities has been proposed for decades, and smart city construction has yielded a positive impact on economic and social development. So, can smart city construction promote urban employment, how much is the employment promotion effect, and what mechanisms does it function? Is there urban heterogeneity and sector heterogeneity in its impact? These doubts have not been taken seriously by relevant scholars.158 cities in China as research objects, this paper empirically examines the impact of smart city construction on urban employment and employment structure by using DID model, PSM–DID model and mediating effect model and further analyzes the transmission pathways and urban heterogeneity.

The main conclusions are as follows: (1) Smart city construction significantly promotes urban employment: compared with non-pilot cities, the total employment, secondary industry employment, and tertiary industry employment in pilot cities increased by 69,680, 46,220, and 20,140, respectively. The conclusion is still robust after using the PSM–DID model, replacing the explained variables, and performing a placebo test. Although smart city construction has a positive impact on employment in the primary industry, it fails to pass the robustness test. (2) The results of the mechanism test show that digital technology development is a mediating variable for smart city construction to promote total employment and tertiary industry employment. Public services mediates the positive impact of smart city construction on total employment and employment of the secondary and tertiary industries. Overall, smart city construction can promote urban employment by promoting digital technology development and improving public services. (3) There was heterogeneity among Chinese cities, with the positive impact of smart city construction on employment mainly reflected in cities located in the eastern and central regions, medium cities, large cities, as well as cities with higher levels of financial development, human capital, and informatization. In addition, the negative impact of smart city construction on employment is mainly reflected in cities in western region. In northeast China, smart city construction increases total employment and secondary sector employment, while reducing tertiary sector employment. (4) Empirical evidence also shows that through different impacts on various sectors, smart city construction has changed the sectoral structure of employment and promoted the transfer of employment to the service industry, that is, optimizing the employment structure of pilot cities.

These conclusions enrich the academic community’s understanding of the development and construction of smart cities, expand the research on the impact of public policy/construction projects on employment, and provide enlightenment and reference for the formulation and promulgation of relevant supporting policies. Based on the above conclusions, the following targeted suggestions are proposed.

First, according to the research, smart city construction can promote urban employment. Therefore, the government can appropriately expand the scope of smart city pilots and support the construction of smart cities in some cities and regions with higher levels of financial development, human capital, and informatization, thereby creating employment opportunities for these regions. Meanwhile, some cities with better infrastructure and higher development level should improve the depth of smart city construction and boost the coordinated development of smart people, smart mobility, smart living, smart governance, smart economy, and smart environment (Giffinger et al. 2007), so as to attract high-end talents to join in urban construction and improve the competitiveness and sustainable development capacity of cities.

Second, in the process of smart city construction, the government should develop digital technology represented by AI, big data, cloud computing, IoTs, etc., to improve the circulation of information, reduce the cost of information acquisition and enhance the production efficiency of enterprises, thus contributing to the promotion of entrepreneurship and employment. Further, to attract and retain talents, the application of digital technology in the fields of education, medical care, transportation, and environment should be accelerated to improve the level of urban public services and the quality of life of residents. Additionally, the government should pay attention to the adverse impact of digital technology on the secondary industry workforce, promptly introduce corresponding employment support policies, and perfect the employment service system.

Third, local governments should adapt to local conditions to build smart cities. In order to expand the positive impact of smart city construction, local governments should focus on improving the level of financial development, human capital, and informatization of cities. Moreover, when formulating smart city construction planning, the central government should avoid the “one-size-fits-all” policies and give certain financial subsidies and preferential policies to cities in underdeveloped areas and small and medium-sized cities to promote balanced development among cities.

Finally, the results of sectoral heterogeneity show that employment in some sectors, especially in the service sector, is positively affected by smart city construction. Therefore, the government should focus on cultivating talents engaged in finance, information technology, scientific research, education, and medical care. Furthermore, the government should give full play to the role of smart city construction in optimizing the urban employment structure, improve the quality of employment and promote full employment, and then facilitate the development and construction of smart cities through fuller and higher-quality employment.