Two-Tier Earnings Structures and labour Market Reform

ABSTRACT

Two-tier earnings structures have recently received renewed attention in the context of labour market reform, but relevant evidence for emerging economies is lacking. Relying on four waves of repeated cross-section individual-level micro data and Oaxaca-type earnings decompositions, this study investigates the interplay between the tightening of employment protection legislation in China through the introduction of the Labour Contract Law in 2008 and the earnings gap between covered and uncovered urban workers. It shows that in urban China average monthly earnings are significantly higher for workers with as compared to those without an employment contract and documents that labour market reform went hand in hand with a doubling in the log earnings gap. As there were no newly emerging differences in observable characteristics between the two groups of workers, this doubling was entirely due to reinforced impacts of differences in workers’ observable characteristics on earnings. Results are consistent with prominent theoretical models of a two-tier labour market, complement existing results for Europe and are robust to a wide range of alternative specifications, including those that addresses selection into labour force participation.

KEYWORDS:

• Two-tier earnings structures
• employment protection legislation
• labour market reform

JEL CLASSIFICATION:

• J31
• J38
• O17

Acknowledgments

I thank Amanina Abdur Rahman, Yang Huang, Dewen Wang, participants of a workshop at the World Bank office in Beijing, an anonymous reviewer and David A. Peel (the co-editor) for helpful discussions, comments and suggestions, and Yayun Pan and Bong Sun Seo for outstanding research assistance. Findings, interpretations and conclusions expressed in this paper are entirely my own and do not necessarily represent the views of the World Bank, its affiliated organizations, its Executive Directors or the Governments these represent.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ In CHIP, information on the main variables of interest is usually elicited regarding the year prior to the collection of data. Thus, survey data that was collected in 2014 is usually referred to as CHIP 2013.

² Results for categorical variables depend on which category is omitted. To arrive at results that are independent of such a choice, this study first estimates the group models with a standard dummy coding. Then, the coefficient vectors are transformed so that they represent deviations from the grand mean and the coefficient for the base category is added.

³ The CHIP data do not result from purely random sampling and the descriptive evidence should not be interpreted as representative for China’s labour market. Nevertheless, estimates from Sections 3.2 are expected to be asymptomatically consistent.

⁴ According to Mulligan and Rubinstein (2008, 1084), ‘the selection bias disappears for groups with [observed] characteristics […] such that practically all of them work.’ In practical terms, logit regressions with a binary variable for wage work as the dependent variable are estimated separately for the different groups of workers and waves of the CHIP data. Then, the sample used for the two-step procedure is reduced to individuals with a predicted probability of wage work that exceeds 80%.