Urban agglomeration and job search behavior: evidence from China

Abstract

Relying on data from 2007 to 2008 CHIP, we use the duration model to analyze the relationship between urban agglomeration and job search behavior. We find that urban agglomeration can improve the quality of the labor market; the duration of unemployment is shorter in big cities. Furthermore, we find that job search intensity of migrants is stronger but urban dwellers can find jobs more easily. Our study enriches the existing research on urban agglomeration and job search behavior in the labor market.

Keywords:

• Urban agglomeration
• unemployment duration
• search behavior
• matching quality

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This work was supported by Key Technologies Research and Development Program.

Notes on contributors

Jiafeng Zong, Nankai University

More than 30 papers were published in CSSCI journals such as the Quarterly Journal of Economics, the Journal of Urban Planning and the Journal of Public Administration. He has over 2 projects of the National Natural Science Foundation of China. He is Chief Chinese expert (University of Nankai-University of Glasgow, UK) in Natural Science Foundation (NSFC) and the Economic and Social Research Council (ESRC) of the United Kingdom.

Zongjian Lin, Nankai University

Paper: The Impact of the New Rural Pension Scheme on Retirement Sustainability in China: Evidence of Regional Differences in Formal and Informal Labor, sustainability, 2018.

Man Guo, Harbin Institute of Technology (Shenzhen)

Book: Performativity, Corporate Behavior, Institutional Change and Rituals in China. Schaker Verlag, 2014. Paper: Entrepreneurs and Ritual in China’s Economic Culture, Journal of Institutional Economics, 2019.

Liang Zhou, PICC China

Notes

1 Because there is a lack of valid data, people often use the survival function or hazard function to characterize the distribution of sample survival time and analyze the impact of each co-variable on the hazard rate. Cox (1972) suggests that the Cox PH model does not require the benchmark risk function to be known in advance and it belongs to the semi-parametric model.

2 $\text{exp}\hspace{0.17em}\beta$ is the coefficient of influence of urban agglomeration on the hazard rate. $\beta$ is greater than 0, and the risk rate increase and the survival time decrease.

3 Where, the “total city” include urban area, suburb and country. “districts under city” include urban area and suburb. This paper uses districts under city data because urban agglomeration exists mainly in the districts under city.

4 This variable is the living expenses for a families in 2007, which reflects data about township household sub-database from CHIP for 2007, and this variable is represented by monthly rent or monthly debt.

5 Networking has a significant impact on the ease that migrants find a job at their destination (Zenou, 2009). The survey in CHIP 2008 shows whether migrants live with other migrants from the same home town, if yes we assume that they have certain networks.

6 The population of local city (country) includes agricultural and non-agricultural accounts, and out-land population includes agricultural and non-agricultural accounts too.

7 The urban level co-variables are reduced to the price in 2008.

8 This paper exchanges the real foreign investment to RMB according to the exchange rate in 2008 in 《China urban statistical yearbook (2010)》.

9 The caliber in this national census is the population of municipal districts, however, the administrative unit is prefecture-level and country level cities in 2007-2008 CHIP because elements can be freely circulated among cities (Liu B, lI l, 2012), we combine municipal districts together and use the population of municipal districts of the fifth national census to measure the urban population. This paper chooses census data as a measure of the total population of the city mainly because the data includes the non-household population, however ‘China Urban Statistical Yearbook’ only consists urban population with household registration.

10 Since the dependent variable is the hazard rate that is taken logarithm, so coefficients are taken index and the following calculation are the same.

11 There are ‘arrive city’ for migrants in New York and Rio de Janeiro. Glaeser and David (2001) analyzed the probability of finding jobs for poor migrants by observing their gathering behaviors.

12 High-skilled, medium-skilled and low-skilled labor accounted for 50.18%, 37.35% and 12.64% respectively,

13 This paper sets control variables according to the location information in CHIP 2007 because there are no location information of interviewees and there is no migration in urban. The location information in this survey consists five parts: main-district location, urban-rural location, town center, town-rural location and special area. We set main-district location and urban-rural location as control variables and the remaining three types are 0.

14 There is a considerable proportion of population without citizenship in the Chinese resident population and the proportion increases constantly during urbanization. According to the sixth census in 2010, 666 million people live in towns and account for 49.68% of the total population, however, ‘2010 National Economic and Social Development Bulletin’ shows there are 246 million migrant workers and it accounts for 35.34% of town population.

15 There are main five kinds of places for migrants to work: urban, suburban, urban-rural, rural and other in 2008 CHIP. We set location in urban as 1, ($\text{city}\mathit{cenrer}\_\mathit{location}$), location in suburban as 1 ($\text{sub}\mathit{urban}\_\mathit{location}$) and set location in urban rural as 1($\text{urbanrural}\_\mathit{location}$).

16 We reclassify the skill level according to the year of education, and 1 to 9 years are low-skilled labor force, 9 to 12 years are medium-skilled labor force, more than 12 years for high-skilled labor force (Lu et al. 2012; Zhao Wei, 2009).

17 There are 16 industries and 12 ownership in the survey and we eliminate the e rural private enterprises and rural individual enterprises. The occupational category of workers is divided into 11 categories and we excludes private business owners (managers), individual owners and farmers.

18 Wang (2013) analyzed the non-agricultural population from 1949 to 2008 and showed that the city size distribution in the national level is more even in the long run.

19 Table 7-8 are IV Probit coefficients. According to the Newy's two-stage estimation method, the first stage conduct OLS regression about endogenous variables and the tool variables, and put the endogenous variables and the residual in the first stage into the original model, the result of which is the endogenous variable with the two-stage IV Probit regression.