China Economic Review 13 (2002) 213 – 230

The impacts of income gaps on migration decisions in China Nong ZHU CERDI, 65 Boulevard Franc˛ois Mitterrand, 63000 Clermont-Ferrand, France

Abstract Using survey data from China, this article examines the effects of income gaps on migration decisions and the sources of these gaps. The econometric results support the hypothesis that income gaps significantly influence migration decisions. When income gap reaches a certain level, the reaction of the migration probability to income gap is weaker for men than for women. The relative income of women is less sensitive to an increase in rural income but more sensitive to a decrease in urban income than that of men. Moreover, we find that the urban to rural income gap is larger for women than for men, which suggests that women receive larger monetary return from migration than men do. In decomposing income gaps, we find that the gap for men is largely determined by differences in the attributes of migrants and nonmigrants, whereas for women, the gap is mainly determined by differences in returns to attributes. D 2002 Published by Elsevier Science Inc. JEL classification: E24; J31; O15; R23 Keywords: Internal migration; China; Income gap; Sample selection

1. Introduction China’s economic reform, which began in the late 1970s, introduced the market mechanism into the internal migration process. On the one hand, agricultural reform returned some freedom to farmers, so that they could freely enjoy their time and leave the land. On the other hand, the reforms and opening up increased regional disparities (Bhalla, 1990; Chen & Fleisher, 1996; Fan, 1999; Lyons, 1991). The combination of increased economic freedom (although by no means total as emphasized in the paper by Cai, Wang, & Du, 2001 in this

E-mail address: [email protected] (N. Zhu). 1043-951X/02/$ – see front matter D 2002 Published by Elsevier Science Inc. PII: S 1 0 4 3 - 9 5 1 X ( 0 2 ) 0 0 0 7 4 - 3

214

N. Zhu / China Economic Review 13 (2002) 213–230

issue) and regional disequilibrium has led to massive labor movements, in particular from rural to urban areas. Migration has become an essential part of the development process in China. Research on China’s internal migration has lagged behind that in other fields, mainly due to lack of data (Ma, Liaw, & Zeng, 1997). Under the assumption that internal migration was well controlled by government and, thus, did not need to be studied, authorities did not include any question on migration in the official census before 1990. Furthermore, there was no survey of migration before the mid-1980s. Since then, various surveys have been conducted, and migration research has flourished. In 1990, the variable ‘‘migration’’ was first introduced in the population census. Nevertheless, migration research in China is still at an early stage (Wu & Zhou, 1997, p. 54). This backwardness is manifested in at least three ways. (a) Most researches remain qualitative. (b) The objective of the migration surveys is often subordinate to some governmental policies instead of being designed for scientific research. Consequently, many migration surveys aim only at descriptive or static analysis or are tangential to other surveys, such as ‘‘China 1988 2/1000 Fertility and Birth Control Survey.’’ (c) Studies focusing on the relation between income and migration are mainly devoted to a comparison of income differences among such as the migrants and nonmigrants (Li, 1997). Dynamic interactions between income and labor mobility are rarely examined. This paper fills some gaps in the microeconomic analysis on internal migration in China. The objective is to analyze not only the determinants of rural to urban migration, but also the determinants of migrants’ income. Specifically, we try to study impacts of income gaps on migration decisions and the sources of those gaps. The following section contains background information on migration, focusing primarily on migration in China. Section 3 presents the methodology. Section 4 describes the data used and discusses the empirical results, and the Section 5 concludes.

2. Background Three decades ago, Todaro (1969) and Harris and Todaro (1970) formalized the hypothesis that rural to urban migration in developing countries responds to expected earnings gains. Since then, a flurry of empirical studies has provided evidence supporting this model (Agesa & Agesa, 1999; Levy & Wadycki, 1974; Lucas, 1988; Taylor, 1987; Todaro, 1976). China is a large agricultural country with a dual economy and with major regional disparities. Particularities of the Chinese situation contribute to an environment for labor mobility that is different in some ways from that in both of developed and other developing economies. Until recently, China had tightly restricted rural to urban migration, mainly in response to the devastating famine that occurred between 1959 and 1961. The purpose was to restrict the urban population size, because, among other things, the government was responsible for feeding the urban population. Two fundamental methods were used to achieve this segregation. One was to impose a high opportunity cost for leaving rural areas by tying incomes to participation in daily collective farm work (gongfenzhi). The other was to make it difficult for outsiders to live in urban areas through the denial of urban residence registration (hukou), on which employment, allocation of housing, food and other necessities were

N. Zhu / China Economic Review 13 (2002) 213–230

215

contingent. The prolonged restriction on migration permitted urban to rural income gaps to persist and enlarge greatly. The income gap widened until 1978, declined between 1978 and 1984, and widened again afterwards. The ratio of urban to rural per capita income was 2.57 in 1978, dropping to 1.86 in 1985, and then rising to 2.50 in 1994. From then on, it fluctuated around 2.50 (Department of Comprehensive Statistics of National Bureau of Statistics of China, 1999, p. 22). When the reform began in the late 1970s, the people’s commune came to an end and the household responsibility system (HRS) was propagated throughout rural China. The HRS had two far-reaching and probably unintended effects on the ability of the central government to control migration. First, buying food in urban areas without urban registration status became possible. The HRS increased the food supply dramatically, which led to the availability of food on free market in cities and eventually to the abandonment of food rationing. Second, the HRS returned personal freedom to rural people, including the freedom to choose one’s occupation. Rural workers could freely allocate their time, choose their profession and their mode of production. The large urban to rural income gap encouraged farmers to leave agricultural activities for nonagricultural ones or to migrate to cities. Gradually, these spontaneous movements of the rural population broke through the government’s migration constraints. However, although food shortages are no longer a threat, the government continues to restrict migration by both direct and indirect regulations for three reasons. First, urban residents are unwilling to share their higher living standards with rural people. Second, the government does not want to increase its investment in urban infrastructures to accommodate rural migrants. (In rural areas, it is the local population that bears the burden of infrastructure investment.) Third, due to the state-owned enterprise (SOE) reform, urban unemployment has become a serious problem. The conflict between official policies and farmers’ aspirations have led to two consequences: (a) development of an urban informal sector and aggravation of urban labor-market segmentation (Leila, 1999; Wang & Zuo, 1999) and (b) growth and prospering of the rural nonagricultural sector, which provides job opportunities for rural surplus labor (Aubert, 1995; Byrd & Lin, 1994; Rizwanul & Jin, 1994; Zhao, 1999). One of the prominent features of the Todaro model is the existence of an ‘‘informal sector,’’ in which urban residents not otherwise employed can eke out a subsistence living using their labor power alone. This sector is merely a holding ground for people awaiting entry into the formal sector (Todaro, 1997, p. 271).1 Many researches have confirmed the existence of an informal sector in China’s urban areas. They have identified severe labor market segmentation, in which most of the nonqualified rural migrants arrive in cities to take up marginal jobs characterized by long working hours, poor working conditions, low and unstable pay, and no benefits, such as housing and food subsidies, education for children,

1

According to some studies, in some countries, such as South Korea, the formal and informal sectors are tightly related in urban labor markets: there is mobility between the odd-jobbers of the nonmodern sector and the blue-collar workers of the modern sector (Hashiya, 1996, p. 461).

216

N. Zhu / China Economic Review 13 (2002) 213–230

medical insurance and other social insurance (Wang & Zuo, 1999, p. 277). Moreover, opportunities for the unskilled migrants to enter the formal sector are slim. To these migrants, the modern sector income is in sight but beyond reach. Moreover, it is their registration as migrants, not their economic characteristics, that determines their socioeconomic position. Quite many rural migrants work in modern sectors, but they are engaged in low-level, lowpaying, and temporary jobs such as sanitation workers, porters, etc. Thus, they are not only employed in an ‘‘informal sector,’’ but more seriously, they have access only to an ‘‘informal labor market’’(Zhu, 1998). In spite of these major disadvantages, it can be said that the situation of migrants in China’s ‘‘informal market’’ is better than that of migrants in many other developing countries (Zhu, 1998). Despite the legal political disadvantages they face, farmers’ propensity to leave agricultural sector is high because their opportunity cost is low. In 1999, cultivated land per capita in rural areas was only 0.138 hectares (National Statistics Bureau of China, 2000, p. 382). Given China’s geography and existing technology, there appears to be no short-run solution to the ‘‘surplus labor’’ problem within agriculture, and agriculture is characterized by technical stagnation and low productivity. In other words, it appears to be stuck in the traditional phase as defined by Schultz (1964). Consequently agricultural income hovers around the subsistence level, influenced only by exogenous factors such as agricultural product price adjustment, land-allocation policies, etc.

3. Methodology In this section we develop and test an econometric model of the effect of the income gap on rural to urban migration. Furthermore, the overall urban–rural wage differential may be attributed to three factors (Agesa & Agesa, 1999, pp. 41–42): (a) wage differences in observed characteristics, such as the difference in education level and in age composition of the migrants, (b) wage differences in returns to observed characteristics in the rural and urban areas, in other words, wage discrimination between urban and rural labor market, and (c) wage differences that stem from nonobservable characteristics of urban and rural workers. Based on the estimation of wage equations, we will analyze the sources of income gap and provide some explanations for gender differences in the incidence of rural to urban migration in China. In the econometric work that follows, we estimate the earning equations for rural to urban migrants and nonmigrants in rural areas, and the impact of the income gap on migration decisions. We also analyze the contribution of various sources to the estimated income gaps. 3.1. Effects of the income gap on migration decisions To analyze the impact of income gaps on migration, it is necessary to introduce the difference in urban to rural income into the equation representing the migration decision. However, in our sample, urban income is observed only if an individual has already migrated, which leads to the problem of sample selection bias. We use the method of ‘‘switching

N. Zhu / China Economic Review 13 (2002) 213–230

217

regression and structural probit’’ applied to the studies of wage gap by Van der Gaag and Vijverberg (1988) and Perloff (1991). This estimation method involves three steps: First, we estimate a reduced form probit equation: ð1Þ Pi* ¼ a0 Zi þ b0 Xi þ e0i where P*i represents the migration decision and Zi and Xi represent the independent variables of the selection equation and those of the income equation, respectively. Second, we estimate the urban income equation and the rural income equation by using the two-step procedure proposed by Heckman (1979) to correct for sample selection bias. We introduce the inverse Mills ratio obtained from (1) in the income equations: logWui ¼ bu Xi þ gu lui þ mui

ð2Þ

logWri ¼ br Xi þ gr lri þ mri

ð3Þ

where Wui and Wri represent the migrant’s income and the nonmigrant’s income, respectively. lui and lri are the inverse Mills ratios. ˆ ui, if he or Finally, from Eqs. (2) and (3), we predict for each individual the value of log W ˆ she migrates to the cities, and that of Wri, if he or she stays in the countryside. Then we introduce the income gap into the structural probit equation:2 ˆ ui  logW ˆ ri Þ þ aZi þ ei Pi* ¼ hðlogW

ð4Þ

From this equation, we obtain estimates of the impacts of income factors and non-income factors on migration decisions. 2

Some researchers refer to this procedure as ‘‘switching regression and structural probit.’’ The deduction of 1 – 4 is as follows: Suppose there are two wage regimes (e.g., urban and rural, agricultural and nonagricultural sectors, public and private sectors, etc.): logWui ¼ bu Xi þ mui

ðaÞ

logWri ¼ br Xi þ mui

ðbÞ

The wage gap can be expressed as log Wui  log Wri. Suppose B is something that influences the migration decision other than the wage gap: B ¼ aZi þ ei

ðcÞ

Migration decision may now be expressed as: Pi* ¼ hðlogWui  logWri Þ þ B ¼ hðlogWui  logWri Þ þ aZi þ ei

ðdÞ

Then Pi* ¼ ðbu  br ÞXi þ aZi þ mui  mri þ ei ¼ b0 Xi þ aZi þ e0i ðeÞ where (e) is Eq. (1), the reduced equation in the paper, and (d) is Eq. (4), structural equation, in which the X variables are excluded.

218

N. Zhu / China Economic Review 13 (2002) 213–230

3.2. Income gap decomposition Various reasons, such as individual characteristics, discrimination between urban and rural labor market and other unobservable factors, can lead to the income gap between migrants and nonmigrants. Oaxaca (1973) proposed a decomposition technique for analyzing wage discrimination in labor markets. Later, Reimers (1983) modified this method to correct for sample selection bias. Agesa and Agesa (1999) adopted this method to analyze the income difference between migrants and nonmigrants in Kenya. Based on Eqs. (2) and (3), we estimate the urban income (Eq. (5)) and rural income (Eq. (6)) for each individual: ˆ ui ¼ bˆ u Xi þ gˆ u lui logW

ð5Þ

ˆ ri ¼ bˆ r Xi þ gˆ r lri logW

ð6Þ

So we can obtain (Eq. (7)): ~ ~ logW u  logW r ¼ logWu  logWr ¼ bu X u  bˆ r X r þ gˆ u lu  gˆ r X r

ð7Þ

~u and W ~r represent the geometric average values of the two groups. where W The income gap between the two groups can be decomposed as the following (Reimers, 1983, p. 572): ~ ~ logWu  logWr ¼ ðX u  X r Þ½Dbˆ u þ ðI  DÞbˆ r  þ ½ðI  DÞX u þ DX r ðbˆ u  bˆ r Þ þgˆ u lu  gˆ r lr

ð8Þ

where I is the identity matrix and D is a diagonal matrix of weights. Eq. (8) decomposes the percentage difference between the geometric means of observed wage rates for the two groups into three parts: (a) that due to differences in average characteristics of the groups, including differences in local price levels where the group members live, (b) that due to differences in the parameters of the wage function, caused by labor market discrimination and other omitted factors, (c) that due to differences in selectivity bias. The measure of various sources of the gap depends on the choice of weights in matrix D. If we assume that discrimination penalizes the nonmigrants by preventing them from earning according to the migrants’ wage-offer function, then D equals I. If the discrimination gives the rural to urban migrants an undeserved advantage, and they are paid more than they would get in a nondiscriminatory world, then D equals 0 (Reimers, 1983, p. 573; Oaxaca & Ransom, 1994, p. 8). We suppose that: (a) most of the rural to urban migrants cannot enter the formal sector but only the informal one, where the wage rate is set competitively; (b) a great amount of agricultural surplus labor reduces rural income; and (c) the real income of farmers is artificially lowered by the socialist price system, which overvalued manufactured prod-

N. Zhu / China Economic Review 13 (2002) 213–230

219

ucts—creating high profitability, while ‘‘squeezing’’ agriculture through the ‘‘price scissors’’ (jiandaocha) (Naughton, 1998). Then, we take D = I in the following analysis: ~ ~ logWu  logWr ¼ ðX u  X r Þbˆ u þ X r ðbˆ u  bˆ r Þ þ gˆ u lu  gˆ r lr

ð9Þ

We can define the ratio of urban to rural income, R, and its value R0, corrected for the sample selection bias (Eqs. (10) and (11)): ~ Wu R ¼ ~ ¼ ð1 þ PX Þð1 þ Pb Þð1 þ Pl Þ Wr

ð10Þ

~ Wu ¼ ð1 þ PX Þð1 þ Pb Þ R ¼ ~ Wr ð1 þ Pl Þ

ð11Þ

0

ˆ

ˆ

ˆ

¯

¯

where PX = e( Xu  Xr)b u  1, Pb = eXr(b u  b r)  1 and Pl = egˆ ulu  gˆ rlr  1 represent the respective contributions of the three sources of income gap between migrants and nonmigrants.

4. Empirical results The data we use in this article come from the research project ‘‘Migration and regional development,’’ which is financed by the China National Social Science Fund. 4.1. Sample restrictions and variables We did our survey in Hubei province in March 1993.3 The sample design has three levels. First, 81 cities or counties were classified into three levels: (1) Wuhan, the provincial capital, (2) cities at the prefectural level, and (3) counties or the cities at the county level.4 We chose Wuhan to represent the large cities, Danjiangkou to represent medium cities and four counties located in the East, South, Central and West regions. Second, we chose randomly 9 resident’s committees (6 in streets and 3 in towns) in the two cities, and 1 resident’s committee (in towns) and two villages in each county. Thirdly, all the families in the villages chosen were included in the survey. However, in urban committees, only temporary resident’s families (zanzhuhu) that were formally or informally registered with the selected street committees 3

Hubei province is situated in central China, with a population size of over 53 millions in 1993. Heavy industry, light industry, and agriculture are highly developed there. In terms of demographic characteristics as in terms of socioeconomic conditions, it is a province richer than many others representative of internal China. 4 In the hierarchy of administrative authorities of China, there are six grades: (1) Central Government, (2) province and municipality directly under the Central Government (MCG), (3) prefecture and municipality at prefectural level (MPL), (4) county (xian), municipality at county level (MCL) and district under the jurisdiction of MCG or MPL (qu), (5) town (zhen), township (xiang) and street of municipality (jiedao), (6) rural village and urban resident’s committee (juminweiyuanhui).

220

N. Zhu / China Economic Review 13 (2002) 213–230

were included in the survey. Given the high population density in the city and the relatively low proportion of migrants (generally less than 5%), we systematically sampled the permanent resident’s families (changzhuhu) according to the population of street committees. Our sample includes 2796 households, which are distributed among 21 communities, including 6 resident’s street committees, 7 resident’s committees in town and 8 rural villages. The elimination of observations with incomplete data leaves 2573 valid observations. We used two types of questionnaires: a household questionnaire and an individual questionnaire. In each household, we randomly chose one member of at least 15 years old to answer the household questionnaire and the individual questionnaire. Here, the household was the one registered on the (permanent) residence registration booklet (hukoubu). For temporary residents, the survey was carried out in their destination place of residence. So, information about their original location is reported by the migrants themselves. We define migration as a change of usual residency between towns, townships, or streets. To simplify our study, we consider someone as a migrant if his or her birthplace is different from his or her current residence. Using a life history table, the survey registers migration history from the age of 15 years. However, concerning the other information of the households, only their actual situation is recorded. We divide the migrants into two subgroups: permanent and temporary, according to whether their usual residency place and the hukou place are different (temporary) or the same (permanent). There are two types of hukou: agricultural registration and nonagricultural registration. A permanent rural to urban migration signifies a change from agricultural registration to nonagricultural registration, which is in general difficult for nonqualified agricultural labor. Many researchers have already shown that there exist significant differences between permanent migrants and temporary migrants (Chang, 1996; Fan, 1999; Goldstein & Goldstein, 1993; Ma, 1999; Wang & Zuo, 1999; Wu & Zhou, 1997). Authorities directly control permanent migration; thus, permanent migration depends largely on government regulation. Individual decisions are not the appropriate focus of research in this case. In contrast, temporary migration is properly studied as individual optimizing decisions in the face of market constraints. Permanent migrants are generally highly qualified, integrated in employment programs and government social protection programs. Normally, permanent migrants obtain permanent and stable posts in the urban formal sector and receive the advantages provided by the government. In contrast, temporary migrants are generally without significant labor-market qualifications and they access only to hard manual jobs in the urban informal labor market. Permanent migration is generally a one-way move, and involves severing links with the place of origin. On the other hand, temporary migrants remain closely linked with their places of departure and often keep their plots of land for the sake of security. It is not uncommon for them to return to the countryside and resume their former occupations. Some temporary migrants are seasonal migrants. For the reasons given above, in this paper, we restrict our sample to a subsample including the temporary rural to urban migrants and interviewees who reside in the countryside. Because our survey was executed in Hubei, we can only study the urban and rural labor markets in this province. Hence, the migrants who come from the other provinces are removed from the sample. Also, because their living and employment styles almost match

N. Zhu / China Economic Review 13 (2002) 213–230

221

those of the urban inhabitants, we remove migrants who have lived in their cities for more than 5 years. In other words, we retain only the migrants who arrived in their destination cities or towns after 1988. We consider all those who stay in the countryside, including the rural to rural migrants, as the reference group, and we define them as nonmigrants. There are 1371 individuals in our subsample, including 399 rural to urban migrants and 972 nonmigrants. Previous studies show that migration is strongly selective according to the sex (Agesa & Agesa, 1999; Yang, 1999), so we estimate separate equations for each sex. We have two types of equations: an income equation and a selection equation. In the income equation, the dependent variable is the natural logarithm of monthly income, measured by yuan. In the questionnaire, the corresponding question was ‘‘What is your average monthly income?’’ For interviewees who cannot provide accurate information due to the instability of their job or their wage, their monthly income is calculated from their prior month’s income or their income from several preceding months. For farmers, monthly income is calculated in two parts, one from household income in the last year (1992), weighted by their participation time in household productive activities, and the other from their individual remuneration, such as wage from rural enterprises. In any case, the income is reported by interviewees themselves.5 Regressors in the income equation are age, age squared, education level, and per capita GDP (in 1990) of the actual residency place (district in the cities, town or township in the counties). The dependent variable in the selection equation is binary, taking the value of 1 if the individual has migrated and the value of 0 if the individual has not migrated. The regressors of the selection equation are age, education level, marriage status before migration,6 family size, the number of brothers and sisters (a proxy of the clan size), a dummy variable equal to 1 if the individual is the eldest, and the amount of cultivated land of the household. Table 1 shows the average values of individual and household characteristics of migrants and nonmigrants according to their sexes. Table 1 shows that migrants are younger than nonmigrants; migrants are better educated than nonmigrants and men are better educated than women. Among the male migrants, the average number of schooling years is 7.2, which is equivalent to lower middle-school level. The income of migrants is 70–80% higher than that of nonmigrants. Men’s income is 30– 40% higher than women’s. 4.2. Determinants of the migration decision We first estimate a reduced form probit equation, which includes the independent variables of the migration decision function and those of the income function (see Regressions 5 and 6 in Table 3). Then, we estimate, respectively, the urban income equation and the rural income 5 Individual income is a difficult question to control in the survey. Most of the interviewees reported their income carefully. Even in the cases where interviewees are cooperative, income is always difficult to precisely measure because of its instability, its diversification, and its complexity. In general, we have no alternative but to take the answers of interviewee at face value. 6 In case that migration and marriage took place in the same year, we suppose that migrants were married before migration if the reason of migration is not marriage, while single if it is.

222

N. Zhu / China Economic Review 13 (2002) 213–230

Table 1 Average values of individual and household characteristics of migrants and non – migrants Males Migrants Age Years of schooling Education level (%)a Illiterate or semiliterateb Primary schoolb Junior secondary schoolb Senior secondary schoolb Number of brothers and sisters Eldest status (%)b Married (%) Before migrationb Actualb Monthly income (yuan) Cultivated land size of the household (mu)c Number of observations

Females Nonmigrants

Migrants

Nonmigrants

29.8 7.2

37.1 6.0

28.8 6.5

34.0 5.0

8.4 19.1 48.5 24.0 4.5 30.4

18.6 32.9 32.9 15.6 4.4 32.2

17.3 23.9 45.0 13.8 4.5 25.6

31.0 30.5 25.9 12.6 4.9 27.4

37.3 67.2 255 3.1 198

– 88.2 147 4.5 512

40.5 64.6 187 3.1 186

– 87.1 105 4.3 457

a The four categories of education level above are classified according to the year of schooling: 0 – 3 years, 4 – 6 years, 7 – 9 years, and 10 years or above. b Dummy variable. c One mu is equal to 1/15 hectare.

equation, introducing the inverse Mills ratios from the reduced form probit equation to correct for sample-selection bias. Table 2 shows the estimation results for the income equations. Regression 1 shows the results for the male migrants. The relation between income and age is an inverted U shape, corresponding to the results of prior researches (Agesa & Agesa, 1999; Li, 1997). On one hand, age reflects the accumulation of human capital, which includes the setting up of personnel connections and the accumulation of experience (Li, 1997, pp. 1012, 1020). On the other hand, most of the rural to urban temporary migrants are nonqualified workers, and older workers have obvious disadvantages in unskilled manual labor. The results confirm the positive effects of formal instruction: as the education level increases, income increases. Finally, we find that per capita GDP of the destination place, which is a proxy of the regional development level, has positive effect on the migrants’ income. Regression 3 shows that the age does not significantly influence the income level of female migrants. Only completion of senior secondary school level has a positive and significant effect on migrants’ income. One possible reason lies in that women have more choice in urban labor market. For example, an old woman of low education level may be engaged in family service, such as housekeeper, or run some small business ventures, which usually involve the marking of homemade foodstuffs and handicrafts (Todaro, 1997, p. 273). Notice that, either for men or for women, the effect of education level on income is greater for nonmigrants than for migrants, which implies that human capital return is higher in rural areas than in urban areas. One possible explanation is that because most rural to urban

N. Zhu / China Economic Review 13 (2002) 213–230

223

Table 2 Logarithmic income equation adjusting for sample selectivity Males

Females

Migrants

Non – migrants

Migrants

Non – migrants

Regression 1

Regression 2

Regression 3

Regression 4

. . . ( 0.01)  0.015 ( 0.40)

0.069*** (4.17)  0.081*** (  4.27)

Age 0.039* (1.92) 0.041*** (2.73) Age2 (/100)  0.047* ( 1.77)  0.051*** ( 2.91) Education level Primary school 0.056 (0.30) 0.325*** (2.88) Junior secondary school 0.379** (2.08) 0.720*** (5.67) Senior secondary school 0.630*** (3.26) 1.052*** (7.54) Per capita GDP of actual 0.036*** (5.23) 0.091*** (10.80) residency (/100) Inverse Mills ratio 0.053 (0.56)  0.338* ( 2.33) Constant 3.577*** (8.65) 2.418*** (7.22) R2 .263 .344 Number of observations 198 512

 0.415 * ( 1.91) 0.207 (1.02) 0.381* (1.78) 0.007 (0.84)  0.015 (0.13) 4.931*** (8.74) .156 186

0.088 (0.87) 0.349*** (3.01) 0.644*** (4.64) 0.095*** (11.36) 0.030 (0.23) 1.750*** (4.67) .300 457

Dependent variable: logarithmic income. The t-students are presented in parentheses. ‘‘. . .’’ indicates that the absolute value is less than 0.001. * Indicates coefficient significant at 10% level. ** Indicates coefficient significant at 5% level. *** Indicates coefficient significant at 1% level.

migrants held inferior posts and work as manual workers, it reduces the effect of education level. On the contrary, for farmers, ‘‘a higher education level favors the acquiring and the usage of certain modern factors’’ (Schultz, 1964, p. 176), so it is more probable for them to participate in nonagricultural activities, which may significantly increase their income. This viewpoint is supported by the paper of Rozelle, Zhang, and Huang (2001) in this issue. According to their results, education plays a positive role in allowing farmers to participate in China’s rural offfarm labor market and the rate of return to education to farmers is fairly high. From the income equations, we can predict urban and rural incomes, so we can predict the income gap for a given individual. It enables us to study the impact of this gap on migration decision by the structural probit equation estimation. The last two columns of Table 3 (Regressions 7 and 8) show the estimation results. In Table 3, we see that, either for men or for women, the relations between age and migration probability are inverted U shapes. Education level plays a positive role in the migration decision only for men, but not for women. It seems that migration decision of male migrants is determined by their accumulation of human capital; but that of female migrants depends only on their age. Education level is one of the most important factors that determines net expected profit of migration. On one hand, the higher is the education level, the stronger is the capacity to overcome migration obstacles and the lower is the migration cost. On the other hand, the higher is the education level, other things being equal, the greater is the probability of obtaining an urban job. In general, male migrants occupy the ‘‘primary’’ (Todaro, 1997, p. 279) place in migration, so their education attainment plays a significant

224

N. Zhu / China Economic Review 13 (2002) 213–230

Table 3 Probability of migration Reduced form equation

Structural equation

Males

Females

Males

Females

Regression 5

Regression 6

Regression 7

Regression 8

Age 0.033 (1.00) 0.122*** (3.30) 0.105*** (2.76) 0.109*** (3.06) Age2 (/100)  0.029 (  0.75)  0.138*** (  3.08)  0.109 **  2.49)  0.125*** (  2.90) Education level Primary school 0.035 (0.13)  0.227 (  1.07) 0.183 (0.67) 0.043 (0.22) Junior secondary 0.393 (1.36) 0.028 (0.13) 1.166*** (4.04)  0.012 (  0.06) school Senior secondary 0.515 * (1.70) 0.200 (0.83) 1.179*** (3.87)  0.026 (  0.11) school 0.151*** (9.95) 0.116*** (8.50) Per capita GDP of actual residency (/100) Land size of the  0.161*** (  7.07)  0.116*** (  7.14)  0.034 * (  1.76)  0.051*** (  3.23) household Married  1.907*** (  9.95)  2.024*** (  10.47)  1.544*** (  7.30)  1.542*** (  8.41) Household size 0.070 (1.21) 0.091 * (1.72) 0.072 (1.21) 0.171*** (3.32) Number of brothers 0.070 (0.78)  0.046 (  0.57)  0.101 (  1.09)  0.080 (  1.00) and sisters Eldest status 0.118 (0.76) 0.069 (0.46) 0.057 (0.35) 0.015 (0.10) Income gap 1.376* * * (10.23) 0.609*** (8.25) ˆ ui  log W ˆ ri) (log W Constant  2.277*** (  3.34)  2.795*** (  3.72)  3.383*** (  4.31)  2.236*** (  3.15) Log-likelihood  223.567  252.794  192.098  263.656 85.4 82.0 90.7 84.7 Percentage of correction predictions (%) Number of 717 654 717 654 observations Dependent variable: migrant = 1, nonmigrant = 0. The t-students are presented in parentheses. * Indicates coefficient significant at 10% level. ** Indicates coefficient significant at 5% level. *** Indicates coefficient significant at 1% level.

role: it directly influences their income (see Table 2) and indirectly affects their expected income through migration participation rate. However, for female migrants, since some of them are ‘‘associational’’ migrants, such as the women who migrate to cities to accompany their husband, their migration decision depends more on the ‘‘primary’’ migrant. In addition, as we have mentioned above, women have further advantages in the informal sector. Their job searching depends less on education attainment. Hence, for a given income gap, the effects of education level on migration probability for women are not so important, while those of age are more significant. As expected, lack of land is a force increasing the likelihood of out-migration. Marriage strongly reduces the probability of migration: the setting up of a

N. Zhu / China Economic Review 13 (2002) 213–230

225

family and the birth of children mark the start of a more stable live, which increases the opportunity cost of migration. Household size favors mobility for women but not for men. This fact seems to suggest that women are more hindered by household work. Finally, our results confirm the important role of income gap in migration decision for two sexes: the larger is this gap, the stronger is the migration propensity. 4.3. Simulation of change in probability of migration The results of estimations above can be used to simulate some policy effects. Here, we focus on the response of migration probability to a change of income gap. Following the work of Perloff (1991, p. 679), two methods are used to calculate the effect of income change on migration probability. According to the first method, the migration probability that the model predicts for each individual is averaged across all individuals (using equal weights). In the second method, migrants would be those whose migration probability (according to the structural probit equation) is equal to or higher than 50%. Fig. 1 illustrates the response of the mobility to the percentage change in urban to rural income gap. The curves show that, according to the first method, the average of simulated migration probability of women is higher than that of men; but according to the second method (50% rule), the share of male migrants is in general higher than that of female migrants. These two methods give different results. By examining the variance of simulated migration probabilities and observing their distribution with the help of the graphics, we find

Fig. 1. Response of the mobility to the percentage change in urban to rural income gap.

226

N. Zhu / China Economic Review 13 (2002) 213–230

Table 4 Effect of an increase in the rural income or a decrease in the urban income Male

Initial value Rural income increases 10% Urban income decreases 10%

Female Migration probability

Relative ~u/W ~r income W

Average

1.99 1.56 (  21.5) 1.39 (  30.1)

28.9 24.0 (  17.0) 21.8 (  24.6)

Migration probability

50% Rule

Relative ~u/W ~r income W

Average

50% Rule

25.0 20.9 (  16.4) 18.3 (  26.8)

1.99 1.60 (  16.1) 1.16 (  41.6)

30.3 28.1 (  7.3) 23.3 (  23.1)

22.2 21.6 (  2.7) 17.6 (  20.7)

The percentage changes compared with the initial situation are presented in parentheses. A negative sign means a decrease.

that the simulated migration probability of men is much less uniform than that of women.7 The other finding is that, according to the two methods, the migration propensity of women seems almost as a linear function of income gap. However, the curves that describe the mobility of men, in particular the share of migrants, are concave. It implies that the marginal effect of income gap on mobility progressively decreases. In other words, with the increase of income gap, the reaction of men to the change of income gap becomes less sensitive. The narrowing of urban to rural income gap can result from an increase in the rural income or from a decrease in the urban income. The first results in an increase of the opportunity cost for urban to rural migration. This policy corresponds to the view of Todaro (1997, p. 286), which suggests that rural and agricultural development is the key factor to restrain the economic incentives for rural to urban migration and thus, to solve the urban unemployment problem. The second policy directly reduces the gains of migration. In general, it is not very probable that wage in cities may be reduced. However, in some cities of China, governments levy fees to the farmer immigrants or place some artificial barriers (Wu, 1994; Zhao, 1999), which in fact reduce the income of the rural to urban migrants. We carry out two simulations: 10% increase in the rural income and 10% decrease in the urban income. The first simulation measures the effects of an across-the-board increase in the rural income on migration probability, holding the urban income constant. The second, in contrast, reflects the effects of a decrease in the urban income, holding the rural income constant. The simulations change (increase or decrease) the constant term in the regression on the logarithm of income, which is equivalent to a constant percentage increase in income for all persons. According to the estimations of the last section, the urban to rural income ratio are 1.99 for men as for women, which is considered as the initial situation. Table 4 reports the results of the two simulations. Since urban income is higher than rural income, the impacts on income 7

At the initial point (0), the variance of simulated migration probability of men and that of women are 0.12 and 0.08, respectively. According to scatter diagram, the distribution of the simulated migration probability of women is more uniform, while that of men gathers relatively to two extremes, namely the southwest and the northeast.

N. Zhu / China Economic Review 13 (2002) 213–230

227

Males

Females

207 104 1.99

154 77 2.00

39.3 27.6 12.0

22.1 65.2  1.3

201 113 1.78

155 77 2.02

Table 5 Income gap decomposition Before correction for sample selection bias ~u = eˆbuX¯u + gˆ u¯lu (yuan) Geometric mean of urban income, W ~r = eˆbrX¯r + gˆ r¯lr (yuan) Geometric mean of rural income, W ~u/W ~r Relative income, R = W Contribution of the various components of income gap (%) ¯ ¯ ˆ Difference in attributes, PX = e(Xu  Xr)b u  1 ˆ ¯ ˆ Difference in returns to attributes, Pb = eXr(b u  b r)  1 ˆ u¯lu  gˆ r¯lr g Sample selection bias, Pl = e 1 After correction for sample selection bias ~u/egˆ u¯lu (yuan) Geometric mean of urban income, W ¯ ~ Geometric mean of rural income, Wr/egˆ rlr (yuan) 0 ~ ~ Relative income, R = Wu/(Wr(1 + Pl))

ratio of a 10% decrease in the urban income are greater than a 10% increase in the rural income. For men, if the rural income increases 10%, the income ratio would decrease by 21.5%. Using the first method, the average of simulated migration probability would decrease by 4.9 percentage points; whereas, using the 50% rule, the share of migrants would decrease by 4.1 percentage points. However, according to the second simulation, a 10% decrease in urban income would lead to a 30.1% decrease in the income ratio. The average of simulated migration probability and the share of migrants would decrease by 7.1 and 6.7 percentage points, respectively. We find similar results for women. Compared with men, women’s relative income is less sensitive to an increase in rural income, but more sensitive to a decrease in urban income.8 The reaction of migration probability to the decrease of urban income is not much different for men and for women. However, the reaction of migration probability to the increase of rural income is much weaker for women than for men. 4.4. Income gap decomposition We now discuss decomposition of urban to rural income gap. Table 5 shows that the geometric averages of urban income and rural income are, respectively, 207 yuan and 104 yuan for male migrants and 154 yuan and 77 yuan for female migrants, implying an urban–rural income ratio of 1.99 for men and 2.00 for women. However, when the income averages are corrected for selectivity bias, the ratio becomes 1.78 for men and 2.02 for women. This contrasts with the results reported by Agesa and Agesa (1999, p. 52). Our results suggest that urban to

8

In fact, this finding can also be directly deduced from Table 2. By comparing the results of Regressions 1, 2, 3 and 4, we can see that, in the urban income equations, the constant term is much greater for women than for men; however, in the rural income equations, the constant term is slightly greater for men than for women.

228

N. Zhu / China Economic Review 13 (2002) 213–230

rural income gap is larger for women than for men, providing a greater economic incentive for women to migrate from rural to urban areas. Given the fact that in our data, migration rates for women are about the same as for men, we infer that women are more likely migrate conditional on their potential income gain. The finding that the propensity to migrate is greater for women suggests further research. When the income gaps are decomposed, the distribution between attributes and returns to attributes is quite different for men than for women. For men, the contribution of the difference in attributes (39.3%) is more important than that of the difference in returns to attributes (27.6%). This may explain the evident larger ‘‘propensity to migrate’’ for women. Given their attributes, the potential gain for migrating to the city is greater than it is for men. For various reasons, women have greater opportunities to use the characteristics that contribute to labor productivity in the urban labor force. This is partly because they are less likely to be constrained to household work in urban areas than in their rural homes. Moreover, as Todaro (1997, pp. 272–279) has shown, women have an easier time finding jobs in the city than men do.

5. Conclusions This paper presents a model of migration and income determination and confirms the importance of the urban to rural income gap for migration decisions in China. By simulating the effects of a change in income gap on migration probability, we find that the simulated migration probability of men is much less uniform than that of women. When income gap reaches a certain level, the reaction of the migration probability to income gap is weaker for men than for women. With the increase of income gap, the reaction of men to the change of income gap becomes less and less sensitive. We also simulate the effects of a change in rural or urban income on income gap and on migration probability. The results show that the relative income of women is less sensitive to an increase in rural income but more sensitive to a decrease in urban income than that of men. The reaction of migration probability to the decrease of urban income is not much different for men and for women. However, the reaction of migration probability to the increase of rural income is much weaker for women than for men. We show that when income levels are corrected for selection bias, the urban to rural income gap is larger for women than for men. In the decomposition of income gap, we find that for men, the individual characteristics play a more important role than the gain in returns to attributes; for women, their income gap is mainly determined by the differences in returns to attributes.

Acknowledgments The author gratefully acknowledges the help from Elisabeth Sadoulet and Xubei Luo, and the important comments from the anonymous referees. Naturally, the responsibility of any remaining errors lies with the author.

N. Zhu / China Economic Review 13 (2002) 213–230

229

References Agesa, J., & Agesa, R. U. (1999). Gender differences in the incidence of rural to urban migration: evidence from Kenya. The Journal of Development Studies, 35(6), 36 – 58. Aubert, C. (1995). Exode rural, exode agricole en Chine, la grande mutation? Espace Populations Socie´te´ (Villeneuve d’Ascq), 1995-2, 231 – 245. Bhalla, A. S. (1990). Rural – urban disparities in India and China. World Development, 18(8), 954 – 987. Byrd, W., & Lin, Q. (1994). China’s rural industry: structure, development, and reform. Oxford: Oxford University Press. Cai, F., Wang, D., & Du, Y. (2001). Regional disparity and economic growth in China: the impact of labor market distortions. Paper presented at the International Conference on the Chinese Economy: Has China Become A Market Economy? May 2001, Clermont-Ferrand, France. Chang, S. (1996). The floating population: an informal process of urbanization in China. International Journal of Population Geography, 1996-2, 197 – 214. Chen, J., & Fleisher, B. M. (1996). Regional income inequality and economic growth in China. Journal of Comparative Economics, 22, 141 – 164. Department of Comprehensive Statistics of National Bureau of Statistics of China (1999). Comprehensive statistical data and materials on 50 years of new China. Beijing: China Statistics Press. Fan, C. C. (1999). Migration in a socialist transitional economy: heterogeneity, socioeconomic and spatial characteristics of migrants in China and Guangdong. International Migration Review, 33(4), 954 – 987. Goldstein, A., & Goldstein, S. (1993). Determinants of permanent and temporary migration in Hubei province, PRC. IUSSP Proceedings, 2, 85 – 101. Harris, J., & Todaro, M. P. (1970). Migration, unemployment and development: a two-sectors analysis. American Economic Review, 60(1), 126 – 142. Hashiya, H. (1996, December). Urbanization in the Republic of Korea and Taiwan: a NIEs pattern. The Developing Economies, XXXIV-4, 447 – 469. Heckman, J. (1979). Sample selection bias as a specification error. Econometrica, 47, 153 – 161. Leila, F. (1999). Labor allocation of Chinese rural migrant workers in urban areas: job election or job enforcement? Asian and Pacific Migration, 8(3), 329 – 340. Levy, M. B., & Wadycki, W. J. (1974). What is the opportunity cost of moving? Reconsideration of the effects of distance on migration. Economic Development and Cultural Change, 22(2), 198 – 214. Li, S. (1997). Population migration, regional economic growth and income determination: a comparative study of Dongguan and Meizhou, China. Urban Studies, 34(7), 999 – 1026. Lucas, R. E. B. (1988). Migration from Botswana. Economic Journal, 95, 358 – 382. Lyons, T. (1991). Interprovincial disparities in China: output and consumption, 1952 – 1987. Economic Development and Cultural Change, 39(3), 471 – 506. Ma, Z. (1999). Temporary migration and regional development in China. Environment and Planning A, 31, 783 – 802. Ma, Z., Liaw, K.-L., & Zeng, Y. (1997). Migrations in the urban – rural hierarchy of China: insights from the microdata of the 1987 National Survey. Environment and Planning A, 29, 707 – 730. National Statistics Bureau of China (2000). China statistical yearbook. Beijing: China Statistics Press. Naughton, B. (1998). Provincial economic growth in China: causes and consequences of regional differentiation. Paper presented at the International Conference on Chinese Economy: Openness and Disparities in China, October 1998, Clermont-Ferrand, France. Oaxaca, R. (1973). Male – female wage differentials in urban labor markets. International Economic Review, 14(3), 693 – 709. Oaxaca, R., & Ransom, M. R. (1994). On discrimination and the decomposition of wage differentials. Journal of Econometrics, 61, 5 – 21. Perloff, J. M. (1991). The impact of wage differentials on choosing to work in agriculture. American Journal of Agricultural Economics, 73(3), 671 – 680.

230

N. Zhu / China Economic Review 13 (2002) 213–230

Reimers, C. W. (1983). Labour market discrimination against Hispanic and Black Men. Review of Economics and Statistics, 65(4), 570 – 579. Rizwanul, I., & Jin, H. (1994). Rural industrialization: an engine of prosperity in postreforme rural China. World Development, 22(11), 1643 – 1662. Rozelle, S., Zhang, L., & Huang, J. (2001). Employment, emerging labor markets, and the role of education in rural China. Paper presented at the International Conference on the Chinese Economy ‘‘Has China Become A Market Economy?,’’ May 2001, Clermont-Ferrand, France. Schultz, T. W. (1964). Transforming traditional agriculture. New Haven, CT: Yale University Press. Taylor, J. E. (1987). Undocumented Mexico – U.S. migration and the returns to households in rural Mexico. American Journal of Agricultural Economics, 69(3), 626 – 638. Todaro, M. P. (1969). A model of labour migration and urban unemployment in less developed countries. American Economic Review, 59(1), 138 – 148. Todaro, M. P. (1976). Internal migration in development countries: a review of theory, evidence, methodology and research priority. Geneva: BIT. Todaro, M. P. (1997). Economic development. London: Longman. Van der Gaag, J., & Vijverberg, W. (1988). A switching regression model for wage determinants in the public and private sectors of a developing country. Review of Economics and Statistics, 70(2), 244 – 252. Wang, F., & Zuo, X. (1999). Inside China’s cities: institutional barriers and opportunities for urban migrants. American Economic Review, 89(2), 276 – 280. Wu, H. X. (1994, September). Rural to urban migration in the People’s Republic of China. The China Quarterly, 669 – 698. Wu, H. X., & Zhou, L. (1997). Rural-to-urban migration in China. Asian-Pacific Economic Literature, 10(2), 54 – 67. Yang, X. (1999). Gender differences in determinants of temporary labor migration in China: a multilevel analysis. International Migration Review, 33(4), 929 – 953. Zhao, Y. (1999). Labor migration and earnings differences: the case of rural China. Economic Development and Cultural Change, 47(4), 767 – 782. Zhu, J. (1998). Rural to urban labor transfer in the labor market of threes dimensions of China. PhD thesis, Nankai University, Tianjin.