Rural Migrant Workers’ Welfare and Labor Protection in China under the Labor Contract Law
DuPengcheng(杜鹏程),XuShu(徐舒)andWuMingqin(吴明琴)...............................................................
Abstract:
This paper employs difference-in-differences (DID) approach to evaluate the effects of China’s Labor Contract Law’s implementation on rural migrant workers’ welfare. Our findings suggest that the Labor Contract Law has reduced rural migrant workers’ working hours by 23%, and increased their social insurance coverage by 10% to 26%. This conclusion holds true after removal of sample selection bias and policy expectation factor. Further analysis reveals that Labor Contract Law’s welfare improvement effect was more significant for migrant workers in regions where workers had poor bargaining power. Other economic factors during the same period of time did not affect the law’s labor protection effect. Our findings give a clear answer to controversies over whether the Labor Contract Law can improve labor rights for underprivileged groups, and are of reference value for developing labor protection systems.
Keywords:
labor protection, migrant workers’ welfare, selective bias.
JEL Classification Codes: J53, J61, K12
DOI:1 0.19602/j .chinaeconomist.2019.3.07119602/ j .chinaeconomist.2018.09.02
1. Introduction
In the course of China’s industrialization and urbanization over the past four decades, hundreds of millions of Chinese farmers have migrated to cities in search of jobs and dominated labor markets in cities. According to data from the National Bureau of Statistics (NBS), the number of rural migrant workers in China reached 281.71 million in 2016. Among them, 169.34 million were employed in cities, becoming an important part of industrial workers. They have made great contributions to China’s urban development. However, rural migrant workers generally receive a low degree of labor protection due to poor education and labor skills. Compared with urban workers, rural migrant workers often receive unfair treatment from employers in such aspects as labor benefits. According to the National Rural Migrant Workers Monitoring Survey Report 2015, 77.4% of rural migrant workers received education at or below junior middle school, 39% of them worked for over eight hours a day, and more than 60% of them have signed no labor contract of any form with their employers. Against such background, rural migrant workers’ living conditions and benefits have increasingly aroused public attention.
In 2006, the State Council released the Opinions on Resolving the Problems Facing Rural Migrant
Workers, which calls for protecting and improving migrant workers’ benefits through effective labor systems and regulatory oversight. To further protect workers’ lawful rights and benefits, the Standing Committee of the 10th National People’s Congress (NPC) adopted the Labor Contract Law of the People’s Republic of China (“Labor Contract Law”), which entered into force as of January 1, 2008. Compared with the original Labor Law, major improvements have been made to the Labor Contract Law in the following aspects1: First, it requires employers to sign a labor contract with employees (including those in flexible employment) within one month after employment date. Second, after an employee has worked continuously for 10 years or entered into the second fixed-term contract, an employer must sign a non-fixed-term contract with the said employee. Third, a labor contract must specify working hours, labor compensation, social insurance and other benefits. Obviously, the biggest three improvements of the Labor Contract Law are all intended for flexibly employed persons (such as rural migrant workers), and specify their benefits regarding labor compensation, labor protection, working hours, labor intensity and social insurance.
From formulation to implementation, the Labor Contract Law received extensive attention from academia. To date, many studies have been carried out to investigate the economic effects of China’s Labor Contract Law from the perspectives of corporate cost (Liu, 2008; Liu and Liu, 2014; Shen et al., 2017), corporate employment (Feldmann, 2009; Kaplan, 2009), innovation (Ni and Zhu, 2016) and implementation effects (Gao et al., 2012; Freeman & Li, 2013). However, studies are yet to be carried out to investigate the law’s effects on the protection of rural migrant workers as an underprivileged group. Did the Labor Contract Law improve migrant workers’ welfare? If so, what is the extent of such improvement? Is there any heterogeneous effect on the protection of migrant workers across different regions? In order to answer these questions, which are not addressed in previous studies, this paper uses the implementation of the Labor Contract Law of 2008 as a quasi-natural experiment to evaluate its effects on rural migrant workers’ welfare from the five aspects of working hours, pension insurance, medical insurance, work injury insurance and unemployment insurance, and discuss its heterogeneous effects.
This paper’s marginal contributions are manifested in the following four aspects: First, unlike existing studies concerned with corporate cost and employment, this paper investigates the Labor Contract Law’s effects on various aspects of rural migrant workers’ welfare, thus enriching relevant literature on rural migrant workers. Second, this paper makes a detailed identification of sample selection problems and other possible biases that exist in such studies, and arrives at a precise estimation of improvement in rural migrant workers’ welfare as a result of the Labor Contract Law. Third, this paper creates such indicators as workers’ bargaining power to discuss the law’s heterogeneous effects on migrant workers’ welfare. Fourth, this paper creates a labor demand shock variable to exclude the possible impact of other economic factors during the same period of time, and offers a new test approach for relevant studies using the difference-in-differences (DID) model.
2. Data Explanation and Descriptive Statistics
Data employed in this paper are from the China Household Income Project (CHIP) of 2007 and 2013, which was completed in partnership by the Chinese Academy of Social Sciences (CASS) and the China Institute for Income Distribution of Beijing Normal University. This survey includes three parts: rural household survey, urban household survey and migrant population survey. Among them, urban household survey questionnaire collected information about urban household samples with nonagricultural household registration ( hukou); migrant population survey questionnaire collected
information about samples with agricultural hukou seeking jobs in cities, i.e. information about rural migrant worker samples. Since farmers are primarily engaged in home-based agricultural production without formal employment, their working hours, income and pension insurance are determined by their labor supply decision and weather conditions, and not directly comparable with urban employees’ benefits. For this reason, we have deleted rural household survey samples. The questionnaires collected detailed information such as age, level of education, health status, income, working hours and social insurance. Urban household survey and migrant population survey of 2007 cover 15 cities of nine provinces in China, and each includes 5,000 households. Urban household survey and migrant population survey of 2013 cover 126 cities across 15 provinces in China, and each contains 7,175 urban households and 760 migrant worker household (rural migrant worker) samples.
It needs to be explained that since the imbalance of migrant worker sample size in the two surveys
may lead to bias in estimation result, we restrict samples to the seven cities2 with the same surveys in two years for weighted analysis with sample size as weight. Based on our research questions, we restrict samples to non-self-employed workers aged between 16 and 65, and exclude those with outlier numbers of children, students, working hours and salary income. Based on our research subjects, we divide samples into urban civil-servant (government-affiliated institution) samples, urban non-civil-servant (non-government-affiliated institution) samples, rural-migrant-worker-turned civil servant (governmentaffiliated institution) samples, and non- civil- servant ( non- government- affiliated institution) rural migrant worker samples, and exclude rural-migrant-worker-turned civil servant (government-affiliated institutions) samples to finally arrive at over 7,000 observations.
Table 1 is descriptive statistical result of key variables. As can be seen from the table, rural migrant worker samples account for 39%, and urban household samples represent 61%; civil servant samples from government-affiliated institutions with urban hukou account for 19%, and non-civil-servant samples account for 81%. Workers have relatively long weekly working hours, i.e. 48 hours; 61% of workers have pension insurance, and only around 30% of workers have work injury insurance and unemployment insurance. Average age of workers is 36 years, and average level of education is junior middle school and high school.
Table 2 compares the welfare of workers in different groups before and after the Labor Contract Law’s implementation. After the law’s implementation, workers of all groups saw their welfare level increase by different degrees, but migrant workers’ welfare improved the most. Take working hours for instance, rural migrant workers had to work for an average of 58.43 hours in a week in 2007, which is higher than 40.99 hours of urban civil servants and urban non-civil-servants. In 2013, rural migrant workers had to work for an average of 47.60 hours, but no significant change occurred for the rest two groups. Prior to the law’s implementation, a smaller proportion of rural migrant workers were covered by various types of social insurance. After the law came into effect, rural migrant workers enjoyed the highest increase in social insurance coverage, while almost no significant change occurred for urban civil-servants.
3. Model Specification
3.1 Benchmark Model Specification
This paper’s benchmark empirical model is specified as the different in differences approach under multiple regression framework. In order to examine the Labor Contract Law’s effects on samples of different groups, we define non-civil-servant samples with urban hukou as treatment group 1, rural migrant worker samples as treatment 2, and civil-servant samples with urban hukou as control group. Dummy variables Di1 and Di2 are defined as follows:
When a sample is urban non-civil-servant, the value Di1 is 1; otherwise, it is 0. When a sample is rural migrant worker group, the value of Di2 is 1; otherwise, it is 0. Meanwhile, the time dummy variable T is created, and T= 0 and T= 1 are specified as the periods before and after the Labor Contract Law’s implementation. In this paper, T= 0 means 2007, and T= 1 refers to 2013. Thus, the Labor Contract Law’s effects on rural migrant workers’ welfare can be expressed by equation (1):
Where, fixed to insurance the effect logarithm Yit ( is medica), of the region, level of weekly work of and labor injury εit working is rights unobservable insurance of hours benefits ( injuca) ( lnwktm), factor. of an and Since individual pension unemployment the validity insurance i during of insurance period DID coverage result t, and ( unempca); ( pension), may specifically be subject Zjt medical refers is the to missing level into variable equation (Meyer, (1), which 1995), specifically we introduce includes other control gender, variable age, level Xit of that education, affects workers’ marital welfare status, health status, etc. In addition, we introduce the interaction term between region dummy variable and time dummy variable to control for the effects of regional unobservable characteristics on the result with the passage of time. γ1 and γ2 are parameters to be estimated that this paper is concerned with, and respectively denote the effects of the Labor Contract Law’s implementation on the welfare of urban noncivil-servants and rural migrant workers.
3.2 Sample Selection Problem
In the specification of equation (1), directly conducting OLS estimation will cause selection bias problem. The reason is that we may only observe migrant workers’ income and welfare but cannot observe the wage income and welfare of individuals who remained in countryside. In this manner, the samples we obtain are self- selected samples and cannot satisfy randomness requirement. In addition, sample selection problem will also cause correlation between the Labor Contract Law’s implementation and unobservable factors that affect workers’ welfare, thus resulting in model estimation bias. Specifically, the Labor Contract Law’s implementation may increase farmers’ willingness to seek jobs in cities, so that OLS estimation value will be greater than real value. On the other hand, the implementation of the Labor Contract Law guarantees rural migrant workers’ basic rights, allowing capable migrant workers to make enough money in cities in a short period of time in order to return to their hometowns or start their own businesses. This will result in negative correlation between the law’s implementation and personal competence, and cause OLS estimation value to be smaller than real value and estimation result to be biased.
In order to overcome sample selection bias, we utilize rural samples and migrant worker samples, and adopt Heckman two-step approach. The first step is to estimate the probability of rural individual samples to seek jobs in cities using Probit model, and the specific form is as follows:
Where, Prob(Zi=1) is the probability of whether or not individual farmers choose to seek jobs in cities, and Xi is an exogenous variable that affects such result. The second step is to include inverse Mills ratio
ϕ(α'Xi )/Φ(α'Xi ) into regression equation (1) on the basis of acquiring the estimation result of equation (2) as the correction term of selection bias. Corrected regression equation is shown as follows:
IMRi is inverse Mills ratio, and we may obtain the result that overcomes sample selection bias by estimating equation (3)
4. Empirical Result
4.1 OLS Regression Result
Table 3 provides the OLS regression result for the Labor Contract Law’s effects on workers’ welfare, and explained variables are relevant indicators of workers’ welfare. All columns have controlled workers’ individual heterogeneity characteristic variable, city characteristic variable, city dummy variable, time dummy variable, as well as the dummy variable of interaction between city and time.
Explained variable in Column (1) of Table 3 is the logarithm value of weekly working hours, and the estimation coefficients of variables D1 and D2 are significantly positive. This implies that the working hours of both treatment groups are significantly higher than those of control group. Estimation coefficient of variable D2 is higher than D1, which explains that rural migrant workers’ working hours are longer than those of urban non-civil-servant group. Estimation coefficients of interaction terms T×D1 and T×D2 measure the real effects of the Labor Contract Law on workers’ welfare, and coefficient values are all significantly negative at 1% level. Coefficient value of T×D1 shows that on average, the law reduced working hours for urban non-civil-servants by 6.5%. Estimation coefficient value of T×D2 shows that the Labor Contract Law reduced rural migrant workers’ working hours by 21.4%. Estimation result of control variables is consistent with experience: Male workers’ average working hours are longer compared with women; older individuals have longer working hours, but once a threshold is passed, working hours will diminish with the increase of age. Married individuals have longer working hours compared with those who are unmarried. Workers with poor health conditions have shorter working hours.
Explained variable in column ( 2) is whether a worker is covered by pension insurance, and estimation coefficient of variable D2 is significantly negative, which explains that the probability for migrant workers’ coverage of pension insurance is significantly below that of other workers with urban hukou. Estimation coefficient value of T×D2 shows that the Labor Contract Law increased the percentage of rural migrant workers with pension insurance by 14.2%. Judging by the result of control variables, older workers and those with poorer health conditions are more likely to have pension insurance. Explained variable in Column (3) is dummy variable of medical insurance coverage, and the result of interaction term suggests that the law increased the percentage of urban non-civil-servants with medical insurance by 8.9%, and increased the percentage of rural migrant workers with medical insurance by 12.9% 3.
The Labor Contract Law clearly provides that employers have the obligation to pay social insurance premium for workers during the period of labor contract. Therefore, Columns (4) and (5) investigate the law’s effects on work injury insurance and unemployment insurance benefits of workers. Coefficient of interaction term in Column (4) shows that the law had an insignificant effect on the percentage of urban non-civil-servants with work injury insurance, but increased the percentage of rural migrant workers with work injury insurance by 13.7%. Coefficient of interaction term in Column (5) shows that the Labor Contract Law increased the percentage of urban non-civil-servants with unemployment insurance by 9.2%, and increased the percentage of rural migrant workers with unemployment insurance by 10.3%. Higher income of workers means greater probability of their coverage of social insurance of various types. Regression result of city control variable shows that per capita GDP and urbanization rate are both positively correlated with the percentage of workers with social insurance coverage. Result of table 3 shows that the Labor Contract Law’s implementation offers greater protection for rural migrant workers compared with urban non-civil-servants.
4.2 Regression Result after Correction of Sample Selection Bias
OLS regression result shows that the Labor Contract Law’s implementation significantly reduced migrant workers’ working hours and increased the percentage of migrant workers with insurance coverage. But as mentioned before, sample selection bias is a problem that cannot be neglected in our study. We employ Heckman’s two-step approach to eliminate the impact of selection bias.
First, we estimate probability selection model of whether farmers would seek jobs in cities using rural household samples and migrant worker samples of various years. In selection model, explained variable is the dummy variable of whether migrant workers seek jobs in cities (1 denotes yes and 0 denotes no). Whether migrant workers decide to seek jobs in cities is largely determined by household demographics. Hence, we introduce the total number of family members, percentage of preschool-age children, percentage of students and percentage of elderly persons aged above 65 years as exclusive variables of selection equation. Independent variables of selection equation include all control variables mentioned before, i.e. age, level of education, etc. Regression result of selection equation shows that the number of family members is negatively correlated with decision for migration and the percentage of persons aged above 65 years, and positively correlated with the percentage of preschool-age children and number of students in the household4. Next, we calculate sample selection bias correction term based on selection model, i.e. inverse Mills ratio, to arrive at regression equation (3) after correction for selection bias. Regression result is shown in Table 4.
After controlling individual characteristic variable, city characteristic variable, city dummy variable, time dummy variable and dummy variable of interaction between city and time, inverse Mills ratios of Columns (1), (2),(4) and (5) are significant, which initially verifies the existence of selection bias problem in OLS estimation. After correction of sample selection problem, interaction terms T×D2 of various columns remain significant. Result of Column (1) shows that the Labor Contract Law’s implementation reduced rural migrant workers’ working hours by 23.2%, and results of Columns (2)(5) show that the law increased the probability of social insurance coverage for rural migrant workers by 13% to 26%. Coefficient value of interaction term T×D2 increases by different degrees. For instance, rural migrant workers’ working hours reduced by 23.2%, higher than 21.4% in Table 3. Probability for migrant workers to be covered by pension insurance increased by 25.8%, higher than 14.2% in Table 3. Probability for migrant workers to be covered by unemployment insurance increased by 13%, which is also higher than 10.3% in Table 3. The implication is that sample selection problem led to negative bias in OLS estimation. As mentioned before, a possible reason is that migrant workers who started
businesses in their home towns are the more capable among their peers. Result in Table 4 indicates that after elimination of selection bias, the Labor Contract Law offers stronger protection of rural migrant workers’ welfare compared with urban non-civil-servant group.
4.3 Result of Labor Market Standardization in Controlled Regions
After correction of sample selection bias, Model (3) leads to consistent estimation results. However, the Labor Contract Law’s implementation may not be the only contributing factor to improvement in workers’ welfare. In fact, it may also be subject to the impact of changes in time trend of regional labor market environment (such as labor market standardization). Thus, we employ China’s nationwide 1% population sample survey data of 2005 to create a regional labor market standardization index for samples prior to the law’s implementation. Specifically, this index includes percentage of regional working population without a labor contract, percentage of population without pension insurance, and percentage of population without medical insurance and unemployment insurance. Higher values of these indicators suggest lower levels of labor market standardization. We expand Model (3) into equation (4):
Where, noncontj, nonunempj, nonpesj and nonmedj denote the percentages of working population without labor contract, without unemployment insurance, without pension insurance and without medical insurance, respectively. nonXj×T is the interaction term between the four indicators and the time dummy variable.
Table 5 presents regression results of Model (4). After additionally controlling for the interaction term between regional labor market standardization index and the time dummy variable, regression result in Column (1) shows that the Labor Contract Law reduced rural migrant workers’ working hours by 23.1%, which is consistent with result in Table 4. Coefficients of interaction term between labor market standardization index and time dummy variables nocont×T, nopes×T and nomed×T are mostly significantly positive, which implies that, in regions with lower labor market standardization, workers’
average working hours are longer. Results in Columns (2) and (3) indicate that the Labor Contract Law increased the probabilities for rural migrant workers to be covered by pension and medical insurance coverage by 25.4% and 17.6% respectively, which is consistent with the result in Table 4. Interaction term between labor market standardization index and time dummy variable is mostly significantly negative, which shows that, in regions with higher labor market standardization, it is more likely for workers to be covered by pension and medical insurance. Result in Columns (4) and (5) show that the Labor Contract Law increased the probabilities for rural migrant workers to be covered by work injury and unemployment insurance by 15.7% and 10.5% respectively. Result in Table 5 implies that, after controlling for changes in time trend of regional labor market standardization, the result of the Labor Contract Law’s effects on rural migrant workers’ welfare remains robust.
5. Heterogeneity Analysis
Previous analysis shows that the Labor Contract Law improved workers’ welfare. However, the question is whether the law has any differentiated effects on workers with different bargaining power. To answer this question, we expand Model (4) as follows:
heterogeneous variables of of regions Where, cubic protection China Based interaction variable prior Industrial have on for to the the effects workers’ the Hk bargaining terms same Enterprise is Labor of the definitions Di1×T×Hk the welfare. bargaining Contract power Labor Database Referencing with and of Contract Law’s workers power preceding to Di2×T×Hk calculate implementation. of Law McDonald in workers chapter. different on are the workers coefficients average in & regions, different Solow Table with bargaining the under (1981), 6 different regions, reports law attention power may we bargaining the employ and offer of law’s the for workers different 2002-2007 coefficients measuring effects power. in degrees sample on Other data the the of welfare bargaining to fight for of workers better power welfare. in in a regions region, where and higher workers value had means different stronger bargaining bargaining power. power Variable and greater bapwr is possibility workers’
in regions In Column where (1) workers of Table had 6, coefficient greater bargaining of variable power, bapwr rural is significantly migrant workers’ negative, working which hours shows were that shorter. Coefficient of cubic interaction term T×D2×bapwr is significantly positive, which shows that the Labor Contract Law reduced the working hours of workers in regions where workers had less bargaining power. Coefficient of variable bapwr in Columns (2)-(5) are mostly significantly positive. This suggests that, in regions where workers had greater bargaining power, it was more likely for rural migrant workers to be covered by various types of social insurance. Coefficient of interaction term T×D2×bapwr is significantly negative, i.e. the law increased the probability for workers’ social insurance coverage to a higher extent in regions where workers had less bargaining power. Result in Table 6 indicates the law could better improve workers’ welfare in regions where workers had less bargaining power.
6. Robustness Analysis
the suffered Labor Analysis by Contract different in the foregoing types Law’s of effects workers sections on may defines rural not migrant urban be the civil workers’ same. servants For welfare5. instance, as policy However, the control global economic group financial to examine crisis shocks of 2008 led to a recession and unemployment of rural migrant workers, who returned to the countryside. But employees of government-affiliated institutions still kept their jobs. Raising minimum wage would also have an effect on rural migrant workers’ income, increasing their insurance coverage. But civil servant group was less affected. The implication is that DID quadratic interaction term coefficient reflects the effects of not only the Labor Contract Law, but other economic factors during the same period of time as well. To prove the robustness of estimation result, it is necessary to further control for changes in labor market environment arising from external economic shocks to exclude this factor’s effects. Among them, economic shocks at the regional level can be well controlled for through region dummy variable, time dummy variable and “region×time” dummy variable. This has been manifested in the model specification in the previous chapter. However, shocks to the economic environment at industry level should be controlled for through creation of indicators. While changes in economic environment at industry level are subject to various factors, the shocks of any factor inevitably find expression in changes in labor demand at the industry level (Bound &Holzer, 1993; Autor & Duggan, 2003). Total changes in labor demand for various industries can be deemed as the aggregate effects of economic shocks on such industries during the same period of time. To control for the effect of other economic factors on estimation result, we create the variable of exogenous labor demand shocks suffered by city×industry to control for the effects of economic shocks on rural migrant workers’ welfare.
Referencing Diamond (2016) and Charles et al. (2017), we create an industry demand change indicator based on data of 2007 and 2012 from City Statistical Yearbooks.
Where, φk, j,2007 is the ratio between employment in industry j in city k in 2007 and total employment in the industry j in the same year, i.e. share of workforce in industry j of city k. v- k,j,2013 denotes the aggregate employment of industry j in all other cities in 2012 excluding city k itself. By the same token, v- k,j,2007 denotes the aggregate employment of the industry j of all other cities in 2007. Shkk, j is change in labor demand in industry j of city k in the two years. Since the impact of labor demand of specific industries in specific cities is excluded in the calculation process, this indicator is often used in literature to measure the impact of exogenous labor demand suffered by research subjects. The greater value means greater shocks of labor demand to the industry’s workforce.
Judging by the calculation result, labor demand shocks suffered by various industries are generally consistent with experience6. After calculating the degree of demand shocks of city×industry, we introduce the cubic interaction term between labor demand shocks and DID interaction term to examine whether the Labor Contract Law’s protective effect on workers is subject to labor demand shocks.
Where, Shkk, j is demand shocks of city×industry calculated based on equation (6), and other variables have the same definitions with previous chapter. We additionally control for minimum wage of the same
period of time to reduce the effect of minimum wage on workers’ welfare.
Table 7 reports the result of Model (7), i.e. the Labor Contract Law’s effects on workers’ welfare after controlling the impact of labor demand. Coefficient value of variable Shk in Column (1) shows that increasing labor demand does not have any significant effect on workers’ working hours. A possible reason is that, although higher labor demand will increase working hours to some extent, it will also give workers greater bargaining power and labor rights. Such effects tend to reduce working hours, and the two will offset each other, so that workers’ average working hours will not be affected by changes in industry demand. Coefficients of variable Shk in Columns (2)-(5) show that demand shocks are significantly positively correlated with probabilities for workers to be covered by various types of social insurance, i.e. greater increase in labor demand corresponds to higher probabilities for workers to be covered by various types of social insurance and great improvements in their welfare. This conclusion is consistent with economic intuition. Except for Column (2), coefficients of cubic interaction term T×D2×Shk in various columns are all insignificant, which explains that the Labor Contract Law’s protective effects on migrant workers will not be affected by differences in other economic shocks during the same period of time. Compared with results in Table 5, the coefficient value and significance of quadratic interaction term T×D2 also have little change. These conclusions jointly suggest that, after controlling for the demand shocks caused by other factors to an industry, the Labor Contract Law’s effect on rural migrant workers’ welfare will not change.
7. Concluding Remarks
The Labor Contract Law is an important law for worker protection in China. Amid the controversies over whether the law’s implementation will improve underprivileged workers’ benefits, the academia is yet to answer this question based on a reasonable methodology. Fewer studies have been carried out to uncover changes in workers’ welfare in this process. In the course of China’s urbanization over recent years, the welfare of vulnerable workers and especially rural migrant workers has received
unprecedented attention. Against the backdrop of China’s efforts to build a law-based and harmonious society on all fronts, precise evaluation of China’s Labor Contract Law’s protective effects on workers is of great significance.
Unlike previous literature, this paper selects working hours and probabilities for social insurance coverage as indicators of rural migrant workers’ welfare based on CHIP data, and offers a quantitative evaluation of the Labor Contract Law’s effects on migrant workers’ welfare using difference- indifferences (DID) model. In addition, Heckman two-step approach is employed to eliminate sample selection bias caused by individual farmers’ choice to seek jobs in cities. Our findings suggest that, compared with workers with urban hukou, the Labor Contract Law improves rural migrant workers’ welfare by a greater degree. It reduced migrant workers’ working hours by 23%, and increased the probabilities for their social insurance coverage by 10% to 26%. Furthermore, this paper also discusses workers’ bargaining power, level of education and gender, and finds that the Labor Contract Law protected rural migrant workers to a greater extent in regions where workers had poor bargaining power. After excluding the shocks of economic factors during the same period of time, the Labor Contract Law’s labor protection effects should still exist.
This paper’s conclusions further enrich relevant literature on rural migrant workers’ welfare, and offer a clear answer to the question as whether China’s Labor Contract Law could improve the welfare of underprivileged persons. This paper also offers heterogeneous evidence regarding the effectiveness of the Labor Contract Law’s implementation, which is a meaningful supplement to existing studies.