Social networks and internal migration in China: A spatial autoregressive model

• Published date: 01 May 2022
• Author: Chuanmin Zhao,Xi Qu
• Date: 01 May 2022
• DOI: http://doi.org/10.1111/rode.12843

1132 

|

     wileyonlinelibrary.com/journal/rode  Rev Dev Econ. 2022;26:1132–1163.

© 2021 John Wiley & Sons Ltd

Received: 2 January 2020 

|

  Revised: 28 July 2021 

|

  Accepted: 6 October 2021

DOI: 10.1111/rode.12843  

REGULAR ARTICLE

Social networks and internal migration in

China: A spatial autoregressive model

ChuanminZhao

|

XiQu

Antai College of Economics & 

Management, Shanghai Jiao Tong 

University, Shanghai, China

Correspondence

Xi Qu, Antai College of Economics 

& Management, Shanghai Jiao Tong 

University, Shanghai 200052, China.

Email: xiqu@sjtu.edu.cn

Funding information

National Natural Science Foundation of 

China, Grant/Award Number: 71973097, 

71601115 and 72031006

Abstract

Individual decisions such as migration can be influenced 

by different types of social networks. This paper examines 

strong and weak network effects on internal migration in 

China by distinguishing between first- time migrants and 

experienced migrants.  Strong networks  are constructed 

by  household  members,  while  weak  networks  are  de-

termined by  the village in  which the individual  resides. 

By employing  spatial econometrics  tools, we  show that 

strong networks have  a significant positive effect on the 

first- time migration decision, while weak network effects 

vary across home provinces and are significant and posi-

tive only for some provinces. We also classify the sample 

to explore the network effects  within the household and 

find that the effects are larger  among siblings, especially 

among siblings  of  the same  gender. In  contrast,  strong 

networks have a smaller impact, and weak networks ap-

pear to encourage migration after people have migration 

experience. Similar  findings also  apply to the  choice of 

destination, months  worked  away from  home, and  mi-

gration earnings.

KEYWORDS

experienced migrants, first- time migrants, strong networks, weak 

networks

JEL CLASSIFICATION

O15; R23; J61

|

1133

ZHAO and QU



|

INTRODUCTION

The migration  of labor  from a low- productivity  agricultural sector to  a highly productive  non-

agricultural  sector  is  one  of  the  important  mechanisms  for  economic  growth  (Carrington 

etal.,1996; Lewis, 1954), especially for China, the largest  transition economy. Since the reform 

and opening- uppolicy releasedin  1978, the loosening of the household registration system and 

the difference between the urban and rural labor markets have brought about large - scale immi-

grations (Cai et al.,2001). In 2018,  China's floating population was 241  million, accounting for 

17.3% of the total population (The 2018 Statistical Communique on National Economic and Social

Development1). The migrant population has become an important driving force for China's rapid 

economic growth and urbanization (Becker & Elfstrom,2010). Meanwhile, migration geograph-

ically separated many adults from their aging parents and their young children (Antman,2013). 

Although workers migrate to earn more to improve the economic and social well- being of  their 

entire family, a  large number of empirical  studies (e.g., Antman, 2010; Chang et al.,2011;  Du 

etal.,2005; Lei etal.,2018; Zhang et al., 2014) have shown that the migration of household mem-

bers has negative impacts on the rural left- behind children and elderly parents, including mental 

and physical  health; well- being;  social behavior; cognitive  development; and achievement  and 

interpersonal relationship in school, family, and life.

Classical theories  argue that  an individual will  choose to  migrate if  the discounted present 

value of  earnings exceeds the migration  cost (Sjaastad,1962)  and migration increases with  the 

wage differential in simple static  models (Dustmann, 2003). These models  emphasize the  role 

of  the  individual  and  family  characteristics  (e.g.,  age,  education,  and  household  land  area), 

job  features  (e.g.,  distance  and  income),  and  market factors  (e.g.,  wage  gains  and  land  own-

ership).  Recent  literature  has  increasingly explored  the  role  of   social  networks  in  migration 

(Bertoli & Moraga,2015; Giulietti et al.,2018; Massey etal.,1993; McKenzie  & Rapoport,2007; 

Munshi,2003). In  theory, social networks  could generate a snowball  effect in the dynamic  pat-

tern of migration (Carrington etal.,1996), thus providing a useful perspective for predicting how 

social networks influence migration decision isofhigh importance.

The implications of the distinction among  different types of migration flows are shown to be 

important (Vanderkamp,1971). However, past  literature rarely distinguishes between the  first- 

time migrants  (people with no  experience working out)  and the  experienced migrants (people 

who report  having migrated in  the past).  From the  perspective of  the first- time  migrants who 

have less contact with  the city, social networks play an important  role in the process of  provid-

ing assistance or financial help (Dolfin & Genicot,2010), transmitting labor market information 

(Baltagi  et al., 2017;  Ioannides  &  Loury,2004),  and  lowering  the  risks  of  movement  (Palloni 

etal., 2001). It  is much  more convenient to  ask relatives  and friends for  help when  they start 

jobsearching. In addition, experienced  migrants are quantitatively important, which constitute 

a great  proportion, and  the number of  experienced migrants  is likely  to vary with  the level  of 

economic activity. Although the experienced migrants may have lower costs of adjustment  con-

nected with a move based  on accumulating experience, social networks also have important  ef-

fects. Chen etal.(2010) find that co- villagers can use a social network to obtain job information 

and reduce  moving costs at  the destination.  In spite of the easier  job- searching process,  social 

interactions are also possible  because of peer  pressure from their siblings or  co- villagers with a 

similar background and general equilibrium effects in the original village.

Given the varied types of networks that play different roles in encouraging migration, different 

network effects  have recently attracted growing  attention in empirical  literature. Granovetter’s 

(1973)  “strength- of- weak- ties”  argument  shows  that  individuals  are  matched  to  jobs  more 

1134

|

ZHAO and QU

frequently or more effectively through weak ties than through strong ties.  Hiwatari (2016) com-

pares four  types of  networks separately and finds  that nonkinship  networks or  customary as-

sociation networks seem  to have significant effects when  making migration decisions. Giulietti 

etal.(2018) estimate the role of weak ties and strong ties in the migration decision and find that 

the former usually provides information about jobs at the destination, whereas the latter usually 

provides information and concrete support to migrate. However, for China's labor market, some 

scholars have put forward different views. Zhang and Guo (2011) claim that “strength- of- strong- 

ties” is more applicable to China's labor market, especially in the low- end labor market. China's 

social structure  is a kind  of “differential pattern”; that  is, social relations  are gradually pushed 

outward with the individual as the center, indicating how close they are to other people. Kinship 

is one  of the  most important and  stable social networks  in rural China,  which is based  on the 

facts of  birth and marriage  (Fei,2001). Bian (1997)  believes that in  Chinese society, which has 

long been known for its emphasis on interpersonal relationships (guanxi), jobs can be channeled 

through strong ties  more easily than through  weak ties since the  trust and obligation of  strong 

guanxi networks can be used to influence employers to assign a job for someone.

Based on this, we try to identify the networks existing among individuals and determine how 

different types of networks affect the migration decisions. We employ the Longitudinal Survey on 

Rural Urban Migration in China (RUMiC) from 2008 to 2009 to identify network effects. The de-

tailed individual- level information on mutual interactions in a village provides us an opportunity 

to better  identify how strong  and weak  networks influence individual decisions. In  particular, 

this paper applies a high- order spatial autoregressive (SAR) model with two networks to empiri-

cally examine the relative effects of social networks on migration decision in China: a strong net-

work based on kinship relationships2 and a weak network based on the village which household 

members belong to. This description of  the social networks is based on two assumptions: one is 

that the villages are independent and the other is that people in the villages know each other. For 

the first assumption,we control the share of  previous migrants from the five closest contacts3 of  

the household head as a  proxyvariable to depictthe impact of  migrants with good peer- to- peer 

relationships on the individual's subsequent migration decision. To relax the second assumption, 

we use clan networks based on  the surname (reported on Chinese characters) of  the household 

in the same village to  narrow down the weak network. Data on social  networks as well as basic 

individual information were collected from RUMiC 2008, which is  1year before migration. We 

also  classify  the  sample  to  explore the  network  effects  within  a  household.  Furthermore,  we 

investigate network effects for experienced migrants to analyze whether there are differences in 

how strong and weak networks influence different types of migrants (first- time migrants versus. 

experienced migrants). This classification that can explore the different mechanisms fills the gap 

in researching the relationship between network and migration.

Our results present strong  evidence that the importance  of social networks varies  according 

to the type  of migration flows. Kinship  (strong) network has a significant positive effect on the 

first- time  migration decision.  The results  are robust in the  benchmark SAR  linear probability 

model and  SAR logit model as  well as in  subsample regressions, and the  effects become larger 

for siblings of  the same gender. Additional evidence is provided to show that the weak  network 

effects vary  across  home provinces  and are  significant and positive  only for  Henan Province. 

Furthermore, the weak  network plays a dominant role  in the nonfirst- time migration decision. 

These results are robust to the sensitivity exercise after considering the clan relations, employing 

the reported village  migration rate and the  household migration rate last year, and eliminating 

the sample of undergraduates and are also confirmed by the full  sample regression, proving sta-

bility and reliability.