How the choice of reference group matters: economic integration of rural-to-urban migrants in China

ABSTRACT

This paper investigates the economic integration of rural-to-urban migrants in China from a subject-centered perspective, placing migrants’ choice of reference group at the centre of the analysis. I empirically demonstrate that the choice of reference group significantly shapes migrants’ success frames and their aspirations to attain comparable economic status as urban residents (i.e. aspirations for economic integration). To study the causal effect of reference group choice on economic integration and to address the endogeneity of reference group choice, I implement instrumental variables estimation using exogenous variation in home ties. The results show that choosing urban residents as the reference group shapes both particular frames of success, which entails not only income but also possession of consumer durable goods, and aspirations for economic integration, which together, in turn, stimulate migrants to realise greater economic integration. After disentangling self-motivated work hours from total work hours by fixing the causal pathways of reference group choice, I find evidence that migrants attain a higher degree of economic integration by working longer hours. These findings suggest that migrants' success frames and aspirations for economic integration, embedded in their choice reference group, are essential for understanding what shapes their pathways to economic integration.

KEYWORDS:

• Reference group
• economic integration
• rural-to-urban migrants
• subject-centered approach

Acknowledgements

I am extremely grateful to Min Zhou and Lauren Duquette-Rury for their guidance and support. I also want to thank Jeffrey Guhin, Ruben Hernandez-Leon, Kayuet Liu, Xiaoting Sun, Roger Waldinger, and two anonymous reviewers for their valuable comments and feedback. Any remaining errors are the sole responsibility of the author.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 A positive impact of total working hours on integration can be expected because motivated working hours are part of it. However, when the causal effect of having urban residents as the reference group choice is controlled, motivated working hours will be controlled as well, and the impact of total working hours is composed of regular working hours only, which will be weaker. In this way, I can disentangle self-motivated from regular working hours.

2 For the few cases where initial income is larger than the mainstream income, meaning individuals are fully economically integrated at the beginning (i.e. reverse of Figure 2), the following rules applied: (i) When current income is larger than mainstream income, the economic integration degree is 100, despite the initial income level; and (ii) when current income is less than mainstream income, the economic integration degree is 0.

3 Note that some migrants manage to attain urban hukous.

4 Annual total working hours are the product of average hours worked per day, average days worked per week, and months worked in 2013.

5 I use years of work because it is more accurate than years of migration in capturing the integration process within an occupation in an urban destination during migration.

6 Although my endogenous variable is a dummy, I can still proceed with the general approach to get consistent results (Heckman and Robb 1985).

7 If years since migrants first worked is $a$ and years since migrants chose urban residents as the reference group is $b$. We will always have $a>b$ because the impact of the reference group choice on economic integration will not start until after they start work. In other words, $\beta =(a/b)\hat{\beta }$ with $(a/b)\ge 1$. Therefore, $\hat{\beta }$ will always be the lower bound.

8 The results of the Hausman tests show that ‘having urban residents as the reference group’ is endogenous. The results of underidentification tests show that I can reject the null hypothesis and that the instruments are correlated with the endogenous regressor – having urban residents as the reference group. The results on weak identification tests show that Cragg-Donald Wald F statistic are 24.936 and 22.272 which are larger than Stock–Yogo weak ID test critical value (19.93) with 10% maximal IV size and indicates that the instruments are not weakly correlated with the endogenous regressor. The results on overidentification tests show that I fail to reject the joint null hypothesis that the instruments are exogenous.