社会杂志 ›› 2012, Vol. 32 ›› Issue (4): 68-92.

• 论文 • 上一篇    下一篇

农民工收入与村庄网络 基于多重模型识别策略的因果效应分析

陈云松   

  1. 牛津大学社会学系，纳菲尔德学院
  • 出版日期:2012-07-20 发布日期:2012-07-20
  • 通讯作者: 陈云松 牛津大学社会学系，纳菲尔德学院 E-mail:yunsong2000@gmail.com E-mail:yunsong2000@gmail.com
  • 作者简介: 陈云松 牛津大学社会学系，纳菲尔德学院
  • 基金资助:

    感谢康奈尔大学社会学系Steve Morgan教授、牛津大学社会学系Peter Hedstrm教授、Nan Dirk de Graaf教授、Colin Mills教授、牛津大学统计学系Tom Snijders教授、香港科技大学社会科学部吴晓刚教授、中山大学社会学与社会工作系梁玉成副教授以及伦敦政治经济学院Mark Williams博士的批评和建议

VillageBased Networks and Wages of RuraltoUrban Migrants: Estimating the Causal Effects of Networks Using Combined Identification Strategies

CHEN Yunsong   

  1. Department of Socioilogy & Nuffield College，The University of Oxford
  • Online:2012-07-20 Published:2012-07-20
  • Contact: CHEN Yunsong，Department of Socioilogy & Nuffield College，The University of Oxford E-mail:yunsong2000@gmail.com

PDF

688

摘要：

以往基于家庭网和社交网的实证研究表明，社会网络对农民工的工资收入没有影响。这些结论的得出，很大程度上是由于对农民工个人网的范围界定不准，且对内生性问题解决不够。本文采用22个省份的农户调查数据，针对中国农民工频繁返乡的特点，证实同村打工网的规模与农民工在城市中的收入具有正向因果关系。为解决影响因果判断的内生偏误问题，本文采取赫克曼二阶段法和工具变量方法组合使用的多重模型识别策略，把村庄遭受的自然灾害强度作为工具变量。

关键词: 农民工, 社会网, 内生性, 因果关系

Abstract:

Ruraltourban migration in China since the 1970s represents the largest labour flow ever observed in the world. Despite the proliferation of research seeking to understand its mechanisms and magnitude, little is done to identify the direct causal effects of migrant networks on labour market outcomes at the destination. In addition, some previous studies show that family and social networks do not influence the wages of the ruralto urban migrants. However, this finding may be problematic because the concept of networks has not been correctly defined. Ruraltourban migrants return home frequently within a year. As a result, villages of origin serve as an important intermediary where the migrant workers exchange information. This homecomingandgo pattern implies the pivotal role of the villagelevel outflow of migrants in determining their job wages at the destination. Using data from 22 provinces in China, this paper analyses the effect of villagebased migrant networks on the wages of migrant labourers, with particular attention to the potential endogeneity problem. Heckman’s twostage method is used to correct for the sampling problem. Natural disaster in the village of origin is used as an instrumental variable (IV) to deal with other endogeneity biases. The major innovation of this study is taking the total outflow of migrants at the origins as the focus and having Heckman’s twostage method and the IV approach combined. After controlling for the unobserved factors influencing the migration decision, the model is achieved through instrumenting the outflow size of migrant workers by the endogenous effect of natural disasters. The empirical results show that the size of the migrant network significantly improves the wages of the migrants. The mechanism is straightforward: Villagebased networks transmit jobrelated information in cities among migrants through which migrant workers can get more and better job opportunities. The IV estimate suggests that network effects obtained from conventional Heckit model are downwardly biased. This paper interprets it within the framework of heterogeneous network effects. That is, Heckit estimate applies to all villages/individuals while the IV estimate mainly applies to the subgroup of villages/individuals more affected by natural disasters. With the presence of heterogeneity effects, the IV estimate can be interpreted as a Local Average Treatment Effects (LATE) . One possible mechanism is that less able people (in terms of earning ability at the destination) are more responsive to natural disasters, since they have a relatively lower ability to compensate for losses due to natural disasters. That is, villagers of lower earning ability are more likely to be “pushed” out from the villages by natural hazards. If this is the case, the IV estimate can be interpreted as a weighted average network effect and the weight for the less able migrants is relatively higher. Since less able migrants benefit more from originbased networks, the IV estimate mainly captures the network effects among them and it would then be higher than the Heckit estimate.

Key words: migrant workers, social networks, endogeneity, causality