Determination of the income gap between urban residents and rural migrants in China : the role of the Hukou system

(1)

1

Deter i atio of the i o e gap etwee ur a

reside ts a d rural igra ts i Chi a: the role of

the Hukou syste

Ba helor thesis

Ni a Meijer 5 5

Supervisor: Rui Zhuo

February 2017

Abstract

The number of rural migrants to urban areas in China was 132 million in 2006. Between these migrants and local residents inequality exists. This thesis attempts to assess the role of the Hukou system in China on the income gap between urban residents and rural migrants. A difference-in-differences method is used on data on Chinese rural migrants in Sichuan and Hubei province in 2002 and 2007. Wages are compared before and after the Hukou reform in Hubei province in 2004, using Sichuan province as a reference group. I find that the Hukou system has no significant impact on rural

(2)

2 Statement of Originality

This document is written by Student Nina Meijer who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

1. Introduction

The Hukou system is a system that registers individuals and households, implemented in 1951 in China. This registration is made up out of personal facts like age and gender, but it also contains whether you are born in either an urban or a rural area. Ones who are born in rural areas cannot simply move to urban areas. Also, people with rural status cannot simply switch to urban status by migrating to an urban area, while rural migrant is the status they have. Therefore, the Hukou system is nowadays used as a system to control migration (Liu, 2005). Due to Chi a s economic development, factories emerged in urban areas and with that, job opportunities arose. Therefore, an enormous amount of rural individuals and households wanted to move to urban areas. The number of migrants to urban areas was already 132 million per year in 2006. Since this is a population on its own, it is worth being studied. (Démurger et al, 2009).

Rural migrants are not treated the same as urban people in urban areas, in this way inequality arises (Meng & Zhang, 2001). The difference in living standards for urban

residents and rural migrants is made up out of different components. For example, urban residents receive social benefits like medical care, social housing, lifetime jobs, pension etc. Rural migrants do not receive this kind of social welfare. Furthermore, some state

enterprises perform employee discrimination, they prefer to hire urban residents instead of rural migrants. In addition to that, in some cases urban firms that hire rural migrants are being taxed, which causes these firms to stop hiring migrants (Sicular et al, 2007). This makes it even harder for migrants to find jobs in urban areas. The jobs that are left over are mostly the jobs rural migrants have to take. This inequality leads to the research question of this thesis: Is the difference in income between urban residents and rural-urban migrants due to the Hukou system in China?

An answer to this question is relevant for a division of parties. Starting with the government, if there is a significant difference in income for rural migrants before and after a Hukou reform, it is clear to the government that further Hukou reforms need to happen. In addition to that, it is of relevance for the rural-urban migrants themselves. It is important for them to know if they are being unequally treated.

(5)

5 In this thesis, existing literature has been taken into account to form background information about the topic. Determining factors for the rural-urban income gap in China were found to be education, age, working experience and discrimination (Démurger et al (2009); Sicular et al (2007); Meng & Zhang (2001)). Also, I analyzed papers that had different findings regarding the role of the Hukou-system in the determination of the income gap. Liu (2005) found that the Hukou-system is not significant in determining the income gap. However, Zhu (2016) and Afridi et al (2015) found the contrary.

Although research has been done on this topic, this thesis distinguishes itself by focusing on province level. In this thesis the difference-in-differences method is used. On the basis of a Hukou reform in 2004, rural igra ts i o es ha e ee compared before and after the reform. In this method, Hubei province is used as the treatment group, since it has undergone a Hukou reform in 2004, they abolished all Hukou distinctions. Sichuan province is used as the control group, since it has not undergone a Hukou reform ( ‘e e t Chi ese Hukou ‘efor s . The treatment that is used is the Hukou reform. To perform this analysis, data from 2002 and 2007 from the Chinese Household Income Project have been used.

In my empirical analysis, the coefficient of the effect of the Hukou reform is 0.021. Therefore, the Hukou system has a small positive effect on income. However, this variable was not significant. Therefore the answer on the research question, is the difference in income between urban residents and rural-urban migrants due to the Hukou system in China, is no.

This paper is organized as following. First off, different papers are analyzed and compared in the literature review. Second, the method is illustrated in the methodology. Third, empirical results are presented in the results section. Then, the results are being discussed in the discussion. Last, the paper is being summarized and the research question is answered in the conclusion.

2. Literature review

A lot of research has been done on the rural-urban income gap in China. Also, research has been done on the impact of the Hukou system on this gap. In this paper, an amount of papers will be analyzed; these papers either focus on the causes of the rural-urban income gap or on the impact of the Hukou system on the income of rural migrants. Meng & Zhang

(6)

6 were those who researched the rural-urban income gap, they found that the wage gap is for a large part due to discrimination. However, Sicular et al found that the difference in

education is the most significant cause of the income gap. Furthermore, Démurger et al found that the income gap is attributed to the difference in population structure. Also, Zhang & Song found that the income gap was a reason for rural citizens to move to urban areas. Zhu (2002) found that migrants are younger than non-migrants and that migrants are better educated than non-migrants. Last in this section, Akay et al emphasized that the higher income from urban residents is informative for future income for migrants. The first of the next section is Zhu (2016), who found that discrimination against rural migrants has exaggerated between 2002 and 2007. Liu (2005) stated that even without the Hukou system, an income gap could be present, because of differences in experience, local knowledge and social capital. Zhigang & Shunfeng (2006) found that Hukou has a significant impact on the income gap. Similarly, Afridi et al (2015) found that the Hukou syste i flue es i di iduals behavior; giving rural individuals a rural status has influenced them negatively. In contrast, Whalley & Zhang (2007) found that the Hukou system is not the cause of income

differentials.

2.1 The rural-urban income gap

In their research, Sicular et al (2007), Démurger et al (2009) and Zhu (2002) all have

individual characteristics central. Sicular et al (2007) have researched the rural-urban income gap in China. Just like most of the other studies, they used the Oaxaca-Blinder

decomposition analysis. Their data comes from the Chinese Academy of Social Sciences, who performed household income surveys from 1996 and 2003. The decomposition analysis exposes that individual characteristics like age and education, underwrite the income gap. Also, they found that when the share of migrants in an urban area rises, the urban-rural income ratio reduces. In this analysis they included factors which they regard as underlying the rural-urban income gap, and found that differences in education between migrants and urban residents are the most significant cause of the urban-rural income gap in China. But it does not include whether the household has an urban Hukou or not. In this thesis, this particular factor can be added.

(7)

7 Démurger et al (2009) collected their data from the Chinese Household Income Project in 2003. In their research about whether migrants are second class workers in urban China, their finding is that the income gap is attributed to the difference in population structure between migrants and urban residents. They indicate that migrants are mostly younger, less educated and less experienced than urban residents. Therefore, it is harder for them to find jobs, and if they find a job they are usually less paid. They do t regard Hukou as a possible factor for determining the income gap. These findings are in line with the findings from Sicular et al.

I )hu s a alysis o the rural-urban income gap, he found a few possible causes. The first one is the difference in education level and average age between migrants and urban residents. Second, wage discrimination against rural workers, which is also central in this thesis. And last, the gap can be due to unobservable characteristics. Furthermore, he used the method of switching regression and structural probit to analyze the impact of i o e gaps o igratio . He used data fro the resear h proje t Migratio a d regio al de elop e t . His fi di gs ere that igra ts are you ger tha o -migrants in rural areas. Also, he found that migrants are better educated than non-migrants.

Sicular et al (2007), Démurger et al (2009) and Meng & Zhang (2001) all use the decomposition analysis, but Démurger et al & Sicular et al disclaim Meng & Zhangs fi di g on market discrimination and confirm that there is market discrimination. Meng & Zhang (2001) have researched the two-tier labor market in urban China. In their research they used one survey from 1995 and one from 1996. To analyze this data they used a decomposition method, in this method the relative significance of unexplained residuals is compared. They wrote that rural migrants that work in the same jobs as urban residents are being paid less. They found in their paper that although a part of the wage gap between urban residents and rural migrants may be the cause of differences in unobservable characteristics, a large part is due to discrimination. They attribute a large part of the wage gap to discrimination in favor of urban residents. Discrimination is performed in two ways, the first one is the kind in which employers use discrimination while selecting their employees. The second one is when people in the same occupation receive different earnings.

Zhang and Song (2003) used a time-series analysis to analyze rural-urban migration. They used annual data from 1978 till 1999. They used the number of migrants per year as dependent variable and the income gap between rural and urban areas, the growth rate of

(8)

8 real GDP and per capita arable land as independent variables. They found that the rural-urban income gap was a reason for rural citizens to move to rural-urban areas. Therefore, the problem of the rural-urban income gap was a cause of rural-urban migration. The incomes of rural citizens were already lower than urban citizens and that could mean that the Hukou system has nothing to do with the lower wages of rural migrants.

Finally, Akay et al (2012) used a simple regression of subjective well-being method to analyze the income of migrants, using different reference groups. They analyzed the relative concerns or rural-urban migrants in China. With relative concerns, rural igra ts income relative to local urban residents i o e is meant here. They found a negative status effect for rural migrants; negative relative concerns towards other migrants. Also, they emphasized that the higher income from urban residents is informative for future income for migrants. Migrants need to adapt to the new situation and location before their wages increase. However, I suppose that for that to happen, Hukou distinctions need to be abolished. 2.2 The impact of the Hukou system on the income of rural migrants

In this section, opinions are divided in two parts, the ones who support the statement that the Hukou-system determines the income gap, and the ones who oppose this statement. First off, the supporting papers are presented and analyzed. Afterwards, the opposing papers are being taken into account.

2.2.1 Supporting

Zhu (2016) has used the same survey as is used in this thesis, however he did not focus on Hubei and Sichuan province like in this thesis. Zhu has used the Oaxaca-Blinder

decomposition theory. He used data from the China Household Income Project from 2002 and 2007, he chose these years to analyze wage differentials between urban workers and rural migrant workers before and after a Hukou reform. Zhu found that urban residents receive higher average salaries than rural migrants in both 2002 and 2007. A remarkable finding in his paper is that urban residents receive penalties when they work in private firms in both 2002 and 2007. Whereas rural migrants receive the same earnings whether they work for a private firm or in a state-owned enterprise in 2002, but in 2007 they earn less if they work for a private firm. This is remarkable because here it is clear that Hukou

(9)

9 being treated the same way by private firms as urban residents.

Zhigang & Shunfeng (2006) regard Hukou status as a factor causing the rural-urban income gap. They have used a survey from Tianjin in 2003 from October to December. To analyze this data they used a semi logarithmic wage equation. As the dependent variable, they used the natural logarithm of hourly wages. Then, they ran several regressions. In contrast to )hu s (2016) findings, Zhigang and Shunfeng found that urban workers receive higher wages in private firms. However, their findings about migrants are the same, namely there is no significant difference in earnings for migrants in private firms or state-owned enterprises. In addition to these findings, in their research in Tianjin, they found that workers without an urban Hukou or with a temporary urban Hukou receive lower earnings than workers with an urban Hukou.

Afridi et al 5 ha e resear hed rural i di iduals eha ior. To do this they used the experimental design of Hoff and Pandey. They did their own experiment in a classroom a d a alyzed i di iduals eha ior y usi g azes, in this research they discovered the impact of identity. Participants of the games were categorized by rural versus urban via the Hukou system. They e tio ed that the Hukou syste i flue es i di iduals eha ior. Moreover, they stated that discriminating against rural individuals and giving them a rural status has influenced them negatively. Because their performance is worse than urban individuals, they get worse jobs and lower salaries. Just like Zhu (2016), they regard the Hukou system as an important factor causing the income gap.

2.2.2. Opposing

Liu (2005) used an approach in his paper in which he analyzed the age when an individual receives an urban Hukou. To do this he used data from the Chinese Household Income Project in 1995. He used data from various regions and from areas from different sizes. Liu noted that even without the Hukou system, an income gap between migrants and local residents or early migrants could be present. This is because of experience, local knowledge and social capital which urban residents and early migrants control better. However, this gap is easily closed up because migrants catch up with this location specific knowledge. But for this gap to be closed up, one needs to obtain an urban Hukou. Therefore, the Hukou system is an important factor defining this income gap.

(10)

10 Whalley and Zhang (2007) used numerical simulation to analyze impacts of the

Hukou system on income inequality. To do this, they used literature on the global

consequences of immigration restrictions from Hamilton and Whalley. They found that the Hukou system causes movement of labor to be restricted. They state that this helps the Chinese economy to move to a more equal distribution of income. This is because there are less migrants, and therefore less people that are being treated unequally. What they are saying is that the Hukou system is not the cause of income differentials, however it helps the differentials to vanish.

3. Methodology

In this part, the method will be illustrated. First off, the used model is being defined. Afterwards, the data is being described, in this part the origin from the data and the composition of the variables is discussed.

3.1 Model specifications

In this paper, the difference-in-differences method (Imbens & Woolridge, 2007) is being used, for this method, two groups and two time periods are needed. Hubei is the treatment group, Sichuan is the control group, and treatment is the Hukou reform in Hubei in 2004. The difference due to the Hukou reform can be determined by this method, which uses the treatment group to remove trends that are common among the treatment and control group.

The DID framework works as following. Every household is observed in a period before the Hukou reform in Hubei province t=0, and in a period after the reforms t=1 (Abadie, 2005). Also, there are two groups, s=Hubei, Sichuan.

To make a regression out of this, several dummies are created. First, a control group dummy � � which is 1 if a person lives in Hubei and 0 if a person lives in Sichuan. Second, a time dummy � is created, which is 1 if t=1 (in 2007) and 0 if t=0 (in 2002). The regression equation will then be:

(11)

11 In this equation, the dependent variable is log (wage), which is the natural logarithm of hourly wages. The logarithmic value of wage is used to reflect the relative difference in wages. The independent variables are � � , � and �� . in this equation the interaction variable DID consists of � � ∙ � , � is the difference-in-differences estimator. To this equation, control variables are added. Age, schooling and marital status are used as control variables. Unobserved characteristics are among other things working experience, local knowledge and social network, these characteristics are not included in the regression. With the added control variables the regression equations is as following:

�� = � + � � � + � � + � �� + � � �� + � � ��

+ � �� + ��

With using a difference-in-differences method, some assumptions need to be made. First off, all assumptions of the OLS model are applied. Furthermore, a common trend assumption needs to be made. This contains that average change in the control group represents the change in the treatment group if there were no treatment (Abadie, 2005).

To find out if the Hukou system has an impact on income, it needs to be tested if the difference-in-differences estimator is significant. This coefficient is used because � shows the real effect of the treatment, as following:

Post Pre Difference

Treatment � + � + � + � � + � � + �

Control � + � � �

Difference � + � � �

3.2 Data description

The data that is used in this thesis is gathered from the China Institute for Income

Distribution, who created the Chinese Household Income Project. This project was executed in several years, in this thesis the projects in the years 2002 and 2007 will be used. This is

(12)

12 because data before and after the Hukou reform in 2004 needs to be researched. From the 2002 survey the rural-urban migrant individual data will be used. The 2007 survey contains 3 parts, the urban household survey, the rural household survey and the migrant household survey. In this thesis, the migrant household survey is being used. Ideally, data on the years immediately before and after the Hukou reforms are being used. Unfortunately, data is only available from the years 2002 and 2007. Furthermore, since not all provinces have

undergone reforms, comparisons can be done on province level. In this thesis, data on Hubei and Sichuan provinces are being used, since Hubei province has undergone a Hukou reform in 2004 and Sichuan province has not. In this way these two provinces can be compared. The data of 2002 on the rural-urban migrant household data consists of 2000 cases. The data of 2007 on the migrant household survey consists of 5000 households.

To compare wages between 2002 and 2007, 2007 wages are deflated with the consumer price index. In this way, the 2007 wages are measured in 2002 yuan. Only data from individuals aged between 16 and 60 have been used. Furthermore, only positive wage values have been included in the regression. Only observations from people from Hubei and Sichuan are being analyzed. Observations with missing values from the variables used in this paper will be dropped. The variable hourly wages is calculated by using monthly income and working hours per week. The variable schooling is defined by years of schooling.

To the extent that urban residents earn higher wages than rural migrants, I expect that a igra t s i o e ill i rease he he re ei es a ur a Hukou a d e o es a urban resident. If this is true, incomes of migrants in 2007 will be higher than in 2002 in Hubei province. In Sichuan province, outcomes on income in 2002 and 2007 for migrants should be similar. I expect income to be higher for older people, also I expect men to earn more than women. Furthermore, I presume that years of schooling have a positive effect on income. Whether being married has an effect on income is unclear for now. I agree that the difference in income can not totally be devoted to the Hukou system. There can be various other factors instead of the Hukou system that determine the income gap.

In graph 1 the wage in Hubei and Sichuan in 2002 and 2007 is plotted in a wage-time diagram. Just like the expectations, the wages for Hubei citizens rise after the Hukou reform. However, for Sichuan citizens, wages are higher in 2007 as well. From these tables it can be seen that in 2007 average wages for Hubei citizens were higher than for Sichuan citizens. Since the line for Sichuan citizens is steeper than the line for Hubei citizens, wages have

(13)

13 increased more for Sichuan than Hubei citizens between 2002 and 2007. This is not in line with the expectations that Hukou reforms cause wages for rural migrants to rise.

Graph 1 Wage Time diagram (Hourly wage on y-axis)

4. Results

The outcomes of the regression in this thesis are presented following. The DID-estimator coefficient has a positive value. This means that the treatment has a positive effect on hourly wages. This coexists with what is expected. However in this case, the DID-estimator is not significant.

The negative age coefficient is not as expected; apparently it is not certain that income rises with age. That men earn more than women is, however, consistent with my expectations. Also, wages rise with more years of schooling, which coexists with my

expectations. Being married has a positive effect on hourly wages. The R-squared value for this regression is rather low. Also, in this regression some variables have high collinearity with the dependent variable, this can be problematic because the standard deviations will rise.

Table 3

Difference-in-differences regression results

Coefficient SD TIME 0.789*** 0.041 TREATED 0.056 0.051 0 1 2 3 4 5 6 7 2002 2007 Hubei Sichuan

(14)

14 DID 0.021 0.061 Age -0.006*** 0.002 Married 0.261*** 0.042 Schooling 0.058*** 0.006 Constant 0.294*** 0.086 R-squared 0.364 Observations 1763

Note: Data source: China Household Income Project, 2002 and 2007. Dependent variable is the logarithm of hourly wages. *p<0.1;**p<0.05;***p<0.01.

The table shows that wages in 2007 were higher than wages in 2002. Whether it is about a Sichuan citizen or a Hubei citizen, average wages in 2007 were 79.3% higher than in 2002. This result is as was expected in this paper. Furthermore, Hubei citizens received higher wages than Sichuan citizens. For example, Hubei citizens received 6.6% higher wages than Sichuan citizens. However, this result is not significant, therefore it cannot be used. Also, men received higher wages than women, their average income was 22.6% more than

women.

The outcome for this research was insignificant, which was not expected. This unexpected result can be due to several reasons. One reason can be that people need more time to adjust to the new circumstances. In this case, after the reforms it is not said that the migrants instantly change jobs. Perhaps in the long run they will change jobs and receive higher wages. Furthermore, autocorrelation in the standard error can cause problems. Also, there is the problem of heteroscedasticity, which causes the estimators to not be best linear unbiased estimators. As a result, standard errors are uncertain (Stock & Watson, 2003). Table 4

(15)

15 In further research it could be helpful to analyze an individual before and after the treatment, in this way the results become more precise. To do this, real panel data is needed. Also, it can be helpful to take the second time period a bit later after the treatment and the first time period closer to the treatment. Moreover, adding more explanatory variables could improve the research as well. Also, using a larger sample size would improve the model. Last, doing research about the effect of intra-provincial migration would be interesting, these answers could e helpful to a alyze Chi a s o ilizatio . As i this thesis the fo us as o ly on inter-provincial migration.

5. Conclusions

This thesis started by introducing the Hukou system and the situation around the rural-urban income gap in China. This income gap is a popular topic among researchers. Additionally, inequality between rural migrants and local urban residents was highlighted, which leaded to the research question of this thesis: Is the difference in income between urban residents and rural-urban migrants due to the Hukou system in China?

To analyze this subject, different papers were studied and compared. This was to gain more knowledge about the subject and view previous results. Previous findings were that the Hukou system is the cause of the rural-urban income gap (Zhigang & Shunfeng, 2006). However, Whalley and Zhang (2007) stated that the Hukou-system could be the solution for the rural-urban income gap to vanish. Liu (2005) emphasized that even without the Hukou system an income gap would be present. Therefore he stated that the Hukou system is not the defining factor for the income gap.

To develop an answer to the research question, a difference-in-differences method was used. To use this method, a Hukou reform in 2004 in Hubei province was used as the

edu 0.3362 0.2701 0.1539 0.2197 -0.3258 -0.2460 1.0000 married -0.1271 -0.3375 -0.0873 -0.2266 0.6185 1.0000 age -0.1431 -0.1803 -0.0275 -0.1237 1.0000 did 0.3472 0.5453 0.7065 1.0000 treated 0.1325 0.1257 1.0000 time 0.5585 1.0000 logwage 1.0000 logwage time treated did age married edu (obs=1,763)

(16)

16 treatment. Before and after treatment wages in two provinces, one that had Hukou reform and one that did not, were compared. The data that was used came from the China

Household Income Project from 2002 and 2007.

The OLS estimation for the difference-in-differences regression had a few expected outcomes, nonetheless also unexpected outcomes. The DID-estimator turned out to be insignificant, which means that Hukou reforms have no effect on rural igra ts income. Therefore an answer to the research question: Is the difference in income between urban residents and rural-urban migrants due to the Hukou system in China? Is no. In other words, the income gap cannot fully be dedicated to the Hukou-system.

This result can be interpreted as that discrimination against migrants was not already present before the Hukou reform; therefore the reform brings no change. Also, it can be i terpreted that the refor eeds ti e to e fully i ple e ted. Or, it a e that igra ts personal characteristics cause their lower wages. It can also be due to endogeneity effects of the model. Last, as Zhang and Song remarked, was the rural-urban income gap the cause of rural-urban migration, therefore the problem already existed and had nothing to do with the Hukou system.

Although in this paper it was the aim to provide clear empirical results to the research question, this question needs to be researched further. Real panel data is needed to address the question in a superior way.

6. Bibliography

Abadie, A. (2005). Semiparametric difference-in-differences estimators. The Review of

Economic Studies, 72(1), 1-19.

Afridi, F., Li, S. X., & Ren, Y. (2015). Social identity and inequality: The impact of China's hukou system. Journal of Public Economics, 123, 17-29.

Akay, A., Bargain, O., & Zimmermann, K. F. (2012). Relative concerns of rural-to-urban migrants in China. Journal of Economic Behavior & Organization, 81(2), 421-441.

(17)

17 Démurger, S., Gurgand, M., Li, S., & Yue, X. (2009). Migrants as second-class workers in

urban China? A decomposition analysis. Journal of Comparative Economics, 37(4), 610-628.

Imbens & Woolridge. (2007). Difference-in-Differences Estimation. Retrieved from http://www.nber.org/WNE/lect_10_diffindiffs.pdf.

Liu, Z. (2005). Institution and inequality: the hukou system in China. Journal of comparative

economics, 33(1), 133-157.

Meng, X., & Zhang, J. (2001). The two-tier labor market in urban China: occupational segregation and wage differentials between urban residents and rural migrants in Shanghai. Journal of comparative Economics, 29(3), 485-504.

Naughton, B. (2007). The Chinese economy: Transitions and growth. MIT press.

Recent Chinese Hukou Reforms. Retrieved from https://www.cecc.gov/recent-chinese-hukou-reforms

Sicular, T., Ximing, Y., Gustafsson, B., & Shi, L. (2007). The urban–rural income gap and inequality in China. Review of Income and Wealth, 53(1), 93-126.

Stock, J. H., & Watson, M. W. (2003). Introduction to econometrics (Vol. 104). Boston: Addison Wesley.

Whalley, J., & Zhang, S. (2007). A numerical simulation analysis of (Hukou) labour mobility restrictions in China. Journal of Development Economics, 83(2), 392-410.

Zhang, K. H., & Song, S. (2003). Rural–urban migration and urbanization in China:

Evidence from time-series and cross-section analyses. China Economic Review, 14(4), 386-400.

Zhigang, L., & Shunfeng, S. (2006). Rural–urban migration and wage determination: The case of Tianjin, China. China Economic Review, 17(3), 337-345.

Zhu, N. (2002). The impacts of income gaps on migration decisions in China. China Economic

(18)

18 Zhu, R. (2016). Wage differentials between urban residents and rural migrants in urban