How does digital transformation relieve the employment pressure in China? Empirical evidence from the national smart city pilot policy

The study addresses whether digital transformation—operationalized via China’s national Smart City Pilots (SCP) launched in three batches between 2012 and 2014—alleviates or aggravates employment pressure among urban job seekers. Against the backdrop of elevated unemployment risks due to economic crises and COVID-19, the paper conceptualizes employment pressure as negative emotional states driven by labor market conditions and uncertainty. Given mixed prior evidence on technology’s substitution versus creation effects on employment, the authors set out to evaluate SCP’s net impact on employment pressure and to unpack mechanisms from micro (firm-level selection and technology adoption) to macro (agglomeration, industrial upgrading, and innovation) levels, using a quasi-natural experiment framework.

Prior research shows technology affects labor through both substitution and creation channels. Studies document negative substitution effects reducing employment or polarizing jobs, especially among middle-skill workers (Autor et al., 1998; Autor and Dorn, 2013; Autor and Salomons, 2018), while other work finds complementary effects and new labor demand, including in services and for lower-skilled workers under certain conditions (Acemoglu and Michaels, 2002; Graetz and Michaels, 2018; Lordan and Neumark, 2018; Dauth et al., 2018). The digital economy can mitigate information asymmetry and frictional unemployment, enhance innovation and entrepreneurship, improve human capital quality, and optimize long-term employment structures (Kuhn and Mansour, 2014; Atasoy, 2013; Rageth and Renold, 2020; Wu and Yang, 2022). Smart cities, as key vehicles of digital transformation, have been studied for effects on public services, governance, innovation, emissions, and resident well-being, but fewer studies integrate micro- and macro-mechanisms to link smart city policies directly to employment pressure. This paper fills that gap by theorizing configuration optimization and technology upgrading effects that induce siphoning, factor substitution, and efficiency at micro levels, and agglomeration, industrial restructuring, and regional innovation at macro levels.

Data: Panel data from 2003–2019 at macro and micro levels. City-level data for 284 prefecture-level cities (4,828 observations) from China Statistical Yearbook include total employed workers, household registration population, wages of employed workers, per-capita gross regional product, and total retail sales of consumer goods. Firm-level data for Chinese A-share listed companies in Shanghai and Shenzhen (2,329 firms; 28,707 firm-year observations) from CSMAR include number of employees, operating revenue growth, cost of sales, gearing ratio, return on net assets, and income tax expense. Firm data cleaning excluded ST/*ST firms, serious missing data, short panels; linear interpolation used for limited missingness. SCP pilot lists obtained from MHURDC. Identification: Multi-period difference-in-differences (DID), treating SCP (2012–2014 batches) as a quasi-natural experiment. Treatment equals 1 for cities (and firms located therein) in pilot years; 0 otherwise. Baseline models: City-level employment ln(citylabor_ct) on SCP×Post with city and year fixed effects and controls. Firm-level employment ln(labor_it) analogously specified with firm and year fixed effects and controls. Spatial DID: To capture spatial spillovers, the model includes a spatial lag of the dependent variable using spatial proximity matrix W1 (adjacency) and an inverse economic geography matrix W2, following Feng et al. (2019, 2021). Variables: Employment pressure proxied by total employed urban workers (city-level) and firm employees (firm-level). Explanatory variable SCP×Post indicates pilot implementation status. Controls: city-level (employment, population, wages, per-capita GRP, retail sales) and firm-level (employees, revenue growth, cost of sales, leverage, ROE, tax expense). Spatial weights: W1 adjacency; W2 economic-geographic weights based on inverse distance and per-capita GDP scaling. Validity checks: Event-study (parallel trends) per Jacobson et al. (1993); placebo timing shocks (lead treatments at t−3/−4/−5); random treatment assignment simulations; robustness via tail trimming (1%/5%), controlling for city benchmark factors and trends (provincial capital, SEZ, Hu line), PSM-DID (one-to-two nearest neighbor, caliper 0.05), and joint fixed effects (province×year; industry×year). Mechanism tests: Allocation/efficiency via firm TFP (LP method); technology upgrading via patent counts (applications and grants, total and inventions); siphoning via spatial DID direct/indirect effects; factor substitution via interactions with average wages and asset composition proxies; industrial structure transformation via tertiary share, rationalization/advancement indexes, industrial digitization (employment share in info/computer/software services within tertiary sector), and digital industrialization (Peking University Digital Financial Inclusion Index); regional innovation via city innovation index (China City and Industry Innovation Power Report). Heterogeneity analyses: by city size, region (east/central/west), human capital (tertiary students per 10k), government intervention (budget expenditure/GDP), digital infrastructure (broadband subscribers per capita), firm age (new vs old), ownership (SOE, private, foreign), industry (primary/secondary/tertiary), and factor intensity (capital-/labor-/technology-intensive). Further analyses: whether SCP drives firms’ digital transformation (text-mined digitalization index from annual reports) and effects on average employee pay.

Baseline effects: • City level: DID coefficients 0.1185 (without controls) and 0.0743 (with controls), implying SCP increases employment in pilot cities by about 7.43% relative to non-pilot cities (p<0.01). • Firm level: DID coefficients 0.1841 and 0.1690 with controls, implying about 16.9% higher firm employment in pilot cities (p<0.01). Validity and robustness: • Event-study parallel trend tests show no pre-trends at city and firm levels. • Placebo policy timing (t−3/−4/−5) yields insignificant coefficients. • Random treatment simulations (500 runs) center near zero and differ from the true effect. • Robust after tail trimming (1%/5%), adding benchmark factor interactions, PSM-DID, and joint fixed effects. Mechanisms: • Allocation optimization and efficiency: SCP raises firm TFP overall by 3.40% (0.0340**, LP-based). Effects are stronger for technology-intensive firms (+5.02%, 0.0502**), not significant for capital- or labor-intensive firms. • Technology upgrading: Significant increases in innovation outputs (p<0.01): patents granted (+4.3438), patents applied (+6.9002), invention patents granted (+2.3536), invention patents applied (+4.6585). • Siphoning (spatial spillovers): Spatial DID direct effects are positive and significant (W1: +0.0528***; W2: +0.0757***), while indirect (neighbor) effects are negative (W1: −0.0213***; W2: −0.0290***). Total effects positive (W1: +0.0314***; W2: +0.0467***), indicating employment gains in pilots and outflows from neighbors. • Factor substitution: The triple interaction SCP×Post×average wage is negative (−0.0026***), indicating higher wage contexts see more digital substitution of labor. Asset-based substitution proxies are generally insignificant overall, but technology-intensive firms show significant substitution (0.0662***). • Industrial structure transformation: No significant effects on tertiary share or on rationalization/advancement indexes, but significant gains in industrial digitization (+0.0812***) and digital industrialization (+0.1146***), signaling digital restructuring pathways. • Regional innovation: City innovation index increases overall (+0.2354***), with larger effects in large cities (+0.2886***) than in small/medium cities (+0.2083***). Heterogeneity: • Cities: Larger cities see stronger employment gains (large: +0.0797*** vs medium/small: +0.0324***). By region: western (+0.1826***) > eastern (+0.1164***) > central (+0.0250**). Low-education cities experience somewhat stronger gains (+0.0672***) than high-education ones (+0.0537***). Lower government intervention cities benefit more (+0.0785*** vs +0.0363**). High digital infrastructure cities show stronger effects (+0.1021*** vs +0.0595***). • Firms: Old firms benefit more (+0.1890***) than new (+0.0633**). By ownership: foreign (+0.3953***) > private (+0.1745***) > state-owned (+0.1426***). By industry: tertiary (+0.1532***) ≈ secondary (+0.1535***) > primary (+0.5945***, noting small N). By factor intensity: technology-intensive (+0.1665***) > capital-intensive (+0.1502***) ≥ labor-intensive (+0.1455***). Further analyses: • SCP significantly promotes firms’ digital transformation (overall +2.9510***; strongest in technology-intensive firms). • SCP is associated with reduced average wages: overall −0.1364***; largest reduction in labor-intensive (−0.2008***), then capital-intensive (−0.1288***), least in technology-intensive (−0.0680***). Overall conclusion: New jobs created by SCP exceed jobs lost, so SCP alleviates employment pressure while exhibiting notable spatial siphoning and technology-driven substitution dynamics.

The findings directly address the research question by showing that digital transformation via smart city pilots reduces employment pressure, as evidenced by significant increases in employment at both city and firm levels. Mechanistically, SCP improves allocation efficiency (higher TFP) and accelerates technology upgrading (more patents), thereby enabling firms to expand output and employment. At the same time, smart cities attract labor and capital from neighboring areas (siphoning), and digital technologies substitute for certain tasks (especially in technology-intensive contexts), but the creation effects dominate, yielding net employment gains. The macro outcomes—industrial digitization, digital industrialization, and enhanced regional innovation—further support employment expansion. Heterogeneous effects highlight where policy leverage is strongest (large, western, low-intervention, high-digital-infrastructure cities; old, foreign/private, tertiary and technology-intensive firms) and which worker groups benefit more (highly educated). Notably, average wages decline, especially in labor-intensive sectors, suggesting distributional frictions even as employment rises. These results are relevant for policymakers designing digital and urban transformation strategies to relieve employment pressure while managing spatial and distributional side-effects.

Using multi-period DID on 2003–2019 city- and firm-level panels, the study shows that China’s smart city pilot policy significantly alleviates urban employment pressure: employment rises by about 7.43% at the city level and 16.9% at the firm level in pilot areas relative to non-pilots. Robustness checks (parallel trends, placebo timings, random assignments, PSM-DID, joint fixed effects, and tail trimming) support causal interpretation. Mechanism tests indicate SCP drives allocation optimization (higher TFP), technology upgrading (patenting), spatial reallocation (positive direct and negative neighboring spillovers), digital restructuring of industries, and stronger regional innovation. Heterogeneous gains are larger in large and western cities, in cities with low government intervention and stronger digital infrastructure, and among old, foreign/private, tertiary, and technology-intensive firms; highly educated labor benefits more. Policy recommendations include: strengthen SCP promotion and oversight in small/medium and central-region cities; reduce administrative barriers, broaden financing, and appropriately decentralize to curb excessive government intervention; invest in workforce skills to align with digital economy needs; and mitigate adverse livelihood impacts by supporting displaced workers and providing reasonable compensation.