The rapid integration of industrial robots across various sectors in China, particularly within the context of the Fourth Industrial Revolution and intelligent manufacturing, has profoundly reshaped the labor market. China's position as the world's largest industrial robot market, with installation rates increasing at an average annual growth rate exceeding 28%, necessitates a comprehensive understanding of the implications for labor distribution and migration patterns. The study focuses on the impact of industrial robot application (IRA) on the individual migration decisions (IMD) of China's substantial floating population (376 million in 2020). This research aims to empirically analyze the complex relationship between IRA and IMD, considering both direct and indirect effects, and accounting for individual and regional heterogeneity. The study leverages the China Migrants Dynamic Survey Database (CMDS) and employs a conditional logit model to examine the effects of city-level IRA on individual migration choices. Understanding this interaction is critical for effective policymaking in managing the socio-economic transformations caused by technological advancements and urban migration in China, and potentially offering a model adaptable to other global contexts.

Existing literature on IRA's impact on the labor market focuses on the substitution and creation effects. The substitution effect highlights the displacement of low-skill, repetitive jobs by robots, potentially leading to technological unemployment. Conversely, the creation effect emphasizes the generation of new, higher-skilled jobs due to increased productivity and economic growth. Studies show varying outcomes, with some suggesting net job losses in the short term and others arguing for net positive employment effects in the long run. Regarding IMD, the literature highlights various factors influencing individual utility maximization, such as economic factors (wage levels, employment opportunities), political factors (government policies, governance), and environmental factors (environmental quality, infrastructure). The existing literature, however, largely lacks a comprehensive examination of the direct relationship between IRA and IMD at the individual level, a gap this study aims to address.

This study uses a conditional logit model to analyze the impact of IRA on IMD. The model is based on the principle of individual utility maximization, where migrants choose the city offering the highest utility considering economic factors, individual characteristics, and city-specific attributes. The data employed are drawn from three sources: 1) City-level data on industrial robot installations, derived from the International Federation of Robotics (IFR) data and China's Industrial Statistical Yearbooks. Bartik-style measures are utilized to construct city-level IRA variables, reflecting the exposure to robots in each city. 2) City-level data on economic characteristics and environmental factors collected from the China Statistical Yearbook for Regional Economy and China City Statistical Yearbook. Variables include per capita GDP, industrial enterprise density, government fiscal expenditure, average wage, population size, education level, medical services, industrial structure, fixed asset investment, PM2.5 concentration, and average housing price. 3) Individual-level data on the floating population are obtained from the China Migrants Dynamic Survey (CMDS) 2018. Individual characteristics include age, gender, education, health, marital status, household registration type, employment unit type, whether involved in individual businesses, and whether performing routine tasks. The study matches the city-level IRA data with both city-level characteristics and individual-level data from CMDS. The data structure for the conditional logit model creates a dataset where each migrant has a series of alternative cities with associated dummy variables representing the choice of destination. Potential endogeneity is addressed using a two-stage least squares (2SLS) approach, with US IRA as an instrumental variable. Robustness checks include alternative variable definitions, lagged effects, dynamic processes, and Poisson regression. Heterogeneity analysis examines the impact of IRA across diverse individual characteristics and city-level attributes using interaction terms within the model.

The baseline regression results show a significant negative relationship between IRA and the probability of the floating population choosing a city, even after controlling for economic and individual characteristics. For every 1% increase in IRA, the probability decreases by 0.58%. This remains robust to endogeneity concerns addressed through the 2SLS analysis. Poisson regression further supports this finding. Heterogeneity analysis reveals significant differences in the impact of IRA across various groups. The negative effect of IRA on IMD is weaker for: highly educated individuals, those in good health, younger individuals, unmarried women, individuals with agricultural household registrations, employees of state-owned enterprises, and those performing non-routine tasks. Conversely, the impact is stronger for older, male, and married migrants. City-level heterogeneity shows that the negative effect is stronger in coastal areas and areas with higher marketization levels. Higher wages, however, attract migrants to areas with high IRA, and this is reflected in both city-level and individual-level data. Mechanism analysis using city and individual-level panel regressions indicates that IRA increases city-level wages and housing prices, while increasing individual-level wages, working hours, and the difficulty in finding jobs and purchasing houses.

The findings support the hypothesis that IRA negatively impacts IMD, primarily due to a stronger substitution effect than creation effect. This is consistent with the observation that the rapid increase in robot density in China correlates with decreased employment in the manufacturing sector. However, this effect is not uniform, highlighting the complex interaction between technological advancements and individual migration decisions. The heterogeneity results show that migrants weigh numerous factors, not just IRA, when making migration decisions; these factors include wages, housing prices, job security, and environmental quality, reflecting individual utility maximization. The mechanism analysis confirms that IRA influences IMD indirectly through its impact on wages, housing costs, and job market competition. These findings demonstrate the intricate and context-specific nature of the IRA-IMD relationship, making simplistic conclusions about automation and migration inappropriate. This study’s insights enrich the understanding of labor market dynamics under automation in a rapidly developing economy.

This study provides a nuanced analysis of the complex relationship between industrial robot application (IRA) and individual migration decisions (IMD) in China. The negative association between IRA and migration probability is robust to various econometric techniques and holds across numerous individual and city-level characteristics. Policy implications emphasize the need for proactive measures to mitigate job displacement, enhance worker skills, improve social security, and address the impact of automation on housing costs and environmental quality. Future research could focus on refining the measurement of IRA at the firm or job level, expanding the geographical scope to other countries, and exploring the long-term dynamics of automation's impact on urban development and regional inequalities.

This study has some limitations. The use of industry-level IRA data matched to individual-level data introduces potential aggregation bias. Additionally, the focus on China might limit the generalizability of findings to other contexts. Future research would benefit from more granular data on IRA and broader international comparisons. Furthermore, the study primarily focuses on the short-term impacts of IRA, and longer-term analysis would enhance understanding.