Previous PageTable Of ContentsNext Page

Répartition égalitaire des terres et migration de la main-d'œuvre dans les zones rurales de Chine

Le présent article examine les liens existants entre le système actuel de propriété foncière en Chine et la migration de la main-d'œuvre. Trois hypothèses pouvant être testées de façon empirique sont présentées pour montrer à la fois l'effet de richesse et l'effet de substitution de la propriété foncière et examiner ces implications au niveau de la migration de la main-d'œuvre, compte tenu des différents risques liés à une telle migration. On a ensuite recours à une estimation économétrique pour tester ces hypothèses. La principale conclusion est que la répartition plus égalitaire des terres dans le cadre du système actuel de propriété foncière appliqué en Chine encourage la migration de la main-d'œuvre. Cette conclusion résiste à plusieurs tests de sensibilité.

Distribución igualitaria de la tierra y emigración de la mano de obra en la China rural

El presente artículo explora la relación entre el sistema actual de tenencia de tierras de China y la emigración de la mano de obra. En él se proponen tres hipótesis empíricamente comprobables para mostrar tanto el efecto riqueza como el de sustitución de la tenencia de tierras y para estudiar sus repercusiones en la emigración de la mano de obra, teniendo en cuenta los diversos riesgos inherentes a esta última. A continuación se utiliza la estimación econométrica para comprobar dichas hipótesis. El hallazgo más importante es que la distribución más igualitaria de la tierra, derivada del sistema de tenencia actual de China, fomenta la emigración de la mano de obra. Se demuestra que esta conclusión resiste a varias pruebas de sensibilidad.

Egalitarian land distribution and labour migration in rural China

Jing Li and Yang Yao1

China Center for Economic Research, Beijing University, China

This article explores the relationship between China's current land tenure system and labour migration. Three empirically testable hypotheses are put forward to show both the wealth effect and the substitution effect of landholding and to investigate their implications for labour migration, taking into account the various risks inherent in migration. Econometric estimation is then used to test these hypotheses. The most important finding is that the more egalitarian land distribution resulting from the current land tenure system in China encourages labour migration. This finding is shown to withstand several sensitivity tests.

EGALITARIAN LAND DISTRIBUTION AND LABOUR MIGRATION IN RURAL CHINA

Labour migration is an important component of the transformation of Chinese agriculture. However, it is not clear whether the current land tenure in rural China - characterized by egalitarian land distribution - deters labour migration.2 There are several reasons why this might be so. First, possessing more land is a substitute for migration in an economy characterized by land market imperfections, so that egalitarian land distribution reduces the willingness to migrate of individuals who would otherwise have less land and therefore prefer to migrate. Second, the lack of a land sales market deprives farmers of the opportunity to sell off their land in order to finance their settlement in the city (Yang, 1997). Third, there is a risk that a person who is absent owing to migration might end up by losing land in the next land reallocation (Giles, 2000). While the second reason has some theoretical merit, empirically it is less conclusive. With the deterioration of the terms of trade for agricultural goods, agricultural land prices in rural China would be quite low even if land sales were allowed. The proceeds from selling about one mu, or one-fifteenth of a hectare of land (the average per capita landholding in China), would be insufficient for a peasant wishing to settle permanently in a city.3 The first and third reasons may stand and will be explored in this article.

In the literature, some authors characterize China's current land tenure as a response to market imperfections. Dong (1996) argues that the current two-tier land ownership characterized by collective legal ownership and individual use rights is a rational choice amid the imperfections frequently observed in Chinese agricultural markets. Kung (1994) argues that current Chinese land tenure is a device for the peasants to pool their income risk in order to meet the subsistence constraint. Burgess (1998) uses empirical data to show that egalitarian land distribution has contributed significantly to Chinese farmers' higher nutritional intake compared with that of their counterparts in India.

This article follows the above studies in linking land's insurance functions with labour migration. Migration is characterized by a fair number of risks. In particular, three major risks are involved. First, a large number of migrants do not know before they move whether they will be able to find a job.4 Therefore, there is a risk of not finding a job after a migrant reaches the city, or of losing the job after finding one. In 1998, when the impact of the Asian financial crisis began to hit China, the number of migrants was reduced by 187.3 million, 4.1 percent of the number of migrants in 1997 (MOLSS, 1999). Third, there is a high incidence of wage arrears and non-payment for migrants (Liu, Wang and Yao, forthcoming), which means that a migrant might end up with no income to bring back home at the end of the year. Added to these risks are the substantial fixed costs involved in migration, such as the costs of transportation, job search and obtaining proper education and training.5

Land plays two roles in relation to migration. First, it represents a source of wealth that can provide insurance against the risks of migration. Second, it is also a productive input that complements a person's labour input to generate income. The first role can be characterized as the income effect, and the second role can be characterized as the substitution effect.6 Although selling off land will not help a migrant to settle permanently in the city, the accumulation of several years' income generated from the land will be sufficient to finance the fixed costs for his temporary (although long-term) migration. Because the wages earned by migrants are much higher than agricultural earnings (in the case of poor areas, several times higher), the investment is worthwhile.

The role of land in generating income is still significant in China. In 1999, agricultural income still accounted for 39.9 percent of farm household net income at the national level, and for western provinces the figure reached 67.8 percent (SSB, 2000). In addition, the rural credit market is very thin, particularly in poor regions where even informal borrowing among relatives and friends is limited because everyone is financially constrained. As a result, land becomes the most important source of income for financing migration. On the other hand, the role of land as a productive input complementary to labour also holds back migration, because more land means a higher marginal return to labour. However, this substitution effect exists only when the land rental market is not perfect because, otherwise, a person could rent out his land and obtain its full market value. In reality, the land rental market is always characterized by some form of imperfection, a result frequently found in empirical studies (e.g. Carter and Yao, 2000) and leading to the study of the separability problem in agricultural household models.

The relationship between landholding and migration is therefore shaped by the interaction between the land's income and substitution effects. As the income effect tends to be more significant for households with small landholdings and the substitution effect tends to be more significant for households with large landholdings, it is probable that it is people with medium-sized landholdings who are more likely to migrate. To the extent that it tends to draw the size of everyone's landholding closer to the village mean, egalitarian land distribution encourages migration rather than hinders it.

In the following section, three empirically testable hypotheses are proposed regarding the relationship between landholding and the propensity to migrate. The subsequent section (Empirical results) implements an econometric test of the three hypotheses using data from a recent household survey conducted on 814 farms in six provinces. The most important finding is that people in a village with more equal land distribution are more likely to engage in migration. The final section then discusses the policy relevance of this finding.

TESTABLE HYPOTHESES

Consider a two-period model in which a person endowed with W amount of initial monetary wealth and T0 amount of land can be either a full-time farmer or a full-time migrant. In the first period, he chooses between these two options. However, in order to become a migrant, he has to prepare a monetary payment, M, to pay for the fixed costs discussed above. At the end of the first period, there is the probability T that the land will be reallocated (the probability of no reallocation is 1 - T ), a decision made by the village. At the beginning of the second period, the reallocation decision is revealed. If reallocation happens, the person gets the village average per capita land,

In the second period, the person becomes a full-time farmer if he did not make the migration investment. There is a local land rental market, but it is not perfect in the sense that there is an extra cost on top of the rent for each unit of land rental. If he invested in migration in the first period, the person becomes a full-time migrant in the second period. However, there is a risk that his land would be taken back if a land reallocation happens in the second period. As Giles (2000) shows empirically, this risk dampens farmers' incentive to engage in migration. The wage for the migration job is subject to risks.

The person determines whether to make the fixed investment M and become a migrant by comparing his levels of expected utility of investing and not investing. Let EUM denote his expected utility when he does make the investment, and EUA denote his expected utility when he does not. Then we can define his option value of migration by V = EUM - EUA. The person makes the investment and becomes a full-time migrant if V is greater than zero; he does not make the investment and instead becomes a full-time farmer if V is non-positive.

The effect of landholding

The first issue that we are interested in is how the optional value changes with a change in the landholding T0. There are two effects to be considered. First, land is a kind of wealth so it serves to mitigate the risk faced by the household (see, for example, Binswanger, 1981; Chavas and Holt, 1996). We call this effect the gross income effect. Because of this effect, a person with more land is more likely to migrate. Second, because the land rental market is imperfect, more land holds back a person from migrating. We call this effect the gross substitution effect. This effect arises because a person cannot rent out as much land as he would wish and has to spend more time in farming. As a result, the relationship between landholding and the willingness to migrate exhibits an inverse-U relationship. This is summarized in the following hypothesis (H1):

H1: The relationship between per capita landholding and migration exhibits an inverse-U curve.

Egalitarian land distribution and migration

H1 means that people with medium-level landholdings are more likely to engage in migration. What constitutes "medium-level" is not revealed in the hypothesis, however. Nevertheless, to the extent that the medium landholding tends to be the local average landholding, a corollary to H1 is the following:

H2: If the inverse-U relationship is found, people in a village with more egalitarian land distribution will be more likely to migrate.

This result, if proven, would constitute a piece of strong evidence supporting the role of China's current egalitarian land distribution in providing an insurance device for labour migration. While financing permanent migration through land sales is not a practical choice for most rural residents in China, financing long-term migration through income accumulation based on land is possible, because one year's harvest provides enough income to support the travel and initial accommodation costs for a person's migration from inland provinces to the coastal regions.7 Therefore, the insurance function of land is real and important, especially for poorer farmers with low levels of savings.

However, H1 may be caused by a reverse-causality: more egalitarian land distribution is the result of a village's response to more out-migration. To qualify this reverse-causality story, though, we need to require that the village not only reallocate land but also take land from land-rich households and give it to land-poor households. However, the only reason that a village should conduct a land reallocation because of out-migration is that those migrating households have relatively too much land (as indicated in the introduction, this is an argument against the current land tenure arrangement), yet there is no theoretical (as our result just showed) or empirical evidence to suggest that migrating households should have more land. As a result, the reverse-causality story is not likely to hold.

Another potential problem is that both land distribution and migration are determined by other village factors. For example, poorer villages could be more inclined to have egalitarian land distribution because, for them, collective survival is a keener problem, yet people from such villages are more likely to migrate out because of low income.8 This critique suggests that land distribution is endogenous. For this reason, we will run a test of endogeneity of land distribution against labour migration in the empirical section of this article. In fact, this test, if it rejects endogeneity, is also sufficient to reject the reverse-causality story because this story inevitably implies that land distribution is endogenous.

The effect of land reallocation

Another topic we are interested in is how the probability of a land reallocation affects a person's decision regarding migration. This is examined by studying how the option value of migration, V, changes with the probability of reallocation, _. In this regard, two effects can be found. The first is the conventional negative effect shown by Giles (2000), i.e. facing the probability that he would lose his land if he were to migrate, a person will be less likely to engage in migration. Regarding the second, to the extent that land reallocation draws households closer to the mean landholding in the village and that H1 guarantees that a person with an average landholding is at the height of the inverse-U curve, we know that for both a person with a smaller-than-average landholding and a person with a larger-than-average landholding, a larger _ implies a larger propensity to migrate. This result is intuitive: for a person with a smaller-than-average landholding, the income effect functions; for a person with a larger-than-average landholding, the substitution effect functions. Therefore, the above analysis suggests that the prosperity resulting from land reallocation may have two countervailing effects on migration. As a result, we have the following hypothesis:

H3:. The effect of land reallocation on labour migration is undetermined.

This hypothesis can be contrasted with H2. The latter specifies labour migration as a function of land distribution and is a direct corollary to H1. H3 shows that the prosperity of a land reallocation is quite different from land distribution that can be altered by a reallocation. It is important to note that egalitarian land distribution does not need to come from periodical land reallocations; there are other ways to achieve equal landholding.9

EMPIRICAL RESULTS

In this section, the three hypotheses proposed in the last section are tested empirically. The data is drawn from a village and household survey on 824 households of six provinces in 1999.10 The households were located in 36 villages, most of which were in purely rural areas. The survey was designed to study the impacts of China's current land tenure on land productivity and labour mobility and thus provides useful information ranging from village land tenure arrangements to household labour allocation. Most data are for the year of 1998, but retrospective information was also collected on land tenure arrangements. In the sections below we first describe the variables used in the test, and then present the test strategy and empirical results.

Variables

Consistent with the theoretical model, we use the probit model to estimate individual migration behaviour in 1998. The dependant variable is a 0-1 dummy indicating whether a person is engaged in long-term migration in 1998. By long-term migration, we mean employment outside the individual's own county for more than three months. Among the 2 050 people in the employment age interval (16-60 years old), 296 or 14 percent were engaged in long-term migration in 1998.

For the explanatory variables, we first discuss those that are critical for testing the three hypotheses. The first is land endowment. This is measured by the per capita landholding (expressed in mu) of the family that a person belonged to in 1998. It only includes land allocated by the village. H1 indicates that the relationship between per capita land and migration exhibits an inverse-U curve, so in the regressions per capita landholding squared is added to accommodate this relationship.

For H2, we need a variable with which to measure the dispersion of village land distribution. For this we use the coefficient of variation (CV) of household landholding in a village in 1998.

Finally, in order to test H3, we need a variable to represent the probability of a land reallocation in a village. We have chosen two variables to represent that probability: the number of reallocations that have taken place in a village since the Household Responsibility System11 was established and a household's expectation of land reallocation in the contract period. Since there is a risk that a larger number of reallocations in the past would imply a smaller probability of an immediate reallocation, the first variable alone is not enough to capture the probability of reallocation. The second variable is assigned a value of 1 if the household thought there would be a reallocation in the future, and a value of 0 if otherwise. We assign the same value to people from the same household. This variable complements the first variable by accounting for a person's expectation of future reallocations.

In addition to the key variables discussed above, we include four groups of control variables. The first group consists of personal characteristics. They are age, sex (male = 1, female = 0), marital status (married = 1, single = 0), and number of years of schooling.12

The second group includes five household variables: wealth per capita (in thousand yuan), population, children/labour ratio, the ratio of the amount of dry land and the perception of the difficulty of leasing out land.

Wealth comprises the sum of the monetary value of houses, equipment, durables and bank savings at the beginning of 1998. Although greater wealth might be a result of more migration, which serves to accumulate wealth over time, the causality cannot run from migration to wealth because the wealth figure was for the beginning of 1998, but the migration decision was for 1998. By experience and intuition, we expect that a person from a household with a larger population will be more likely to migrate, whereas a person from a household with more dependant children will be less likely to migrate. Similarly, we would expect a person from a household with more dry land to be more likely to migrate because his land is likely to be inferior in quality. The perception of the difficulty of leasing out land is included as a variable because when a person migrates, he may need to rent out his land, and if land rental is difficult, then his willingness to migrate will be dampened. The variable representing this perception is constructed from the answers to two questions: Does a household need to obtain approval from the village if it wants to rent out land a) within the village and b) to households in other villages? A value of 1 is assigned to the variable if the answer to either of these two questions is yes; otherwise it is assigned a value of 0.

The third group contains seven village variables. Two indicate the village's transportation conditions: distance from the county site (kilometres) and the difficulty involved in reaching a bus route (rated by the values 1-5, with 1 = very convenient, 2 = convenient, 3 = average, 4 = less convenient, 5 = very inconvenient). Two more indicate the village's geographic conditions. These are two dummy variables indicating whether the village is in a plain or a hilly area. The reference group is villages in a mountainous area. In addition, per capita income of the village (in thousand yuan), local wage rate (in yuan) and the tax burden per unit of land (expressed as yuan/mu) are added in order to control for general local economic conditions. Higher local income and wage rates imply a stronger local economy, and, particularly, better local non-farming opportunities, so they may act against out-migration. On the other hand, higher tax may push migration because it implies local return from agriculture. Here, "tax" includes the agricultural tax and the so called santi wutong (local surcharges), whose major purposes are to support local governments (at the township and village levels), education, basic health care and village accumulation. Although the charge base of the surcharges is not necessarily land, most villages distribute the burden according to size of landholding.

Finally, five provincial dummies are added to the regressions (Zhejiang is the reference province). Table 1 presents the means and standard errors of the variables.

TABLE 1

Descriptive statistics of variables (2 050 people)

Variable

Mean

Standard deviation

Main variables

 

 

Household per capita land endowment (mu)

1.13

1.09

CV of land distribution in village1

0.54

0.51

Number of reallocations in village

1.12

1.90

Household expectation of land reallocation

0.68

0.47

Personal variables

 

 

Age

37.66

11.81

Sex

0.51

0.5

Marital status

0.81

0.39

Number of years' schooling

6.05

3.11

Household variables

 

 

Per capita wealth (thousand yuan)

8.64

11.16

Population

4.57

1.44

Children/labour ratio

0.35

0.37

Dry land ratio

0.45

0.43

Perception of difficulty of leasing out land

0.48

0.50

Village variables

 

 

Distance to county site (km)

30.63

21.17

Difficulty of reaching a bus route

1.79

1.04

Located in plain

0.58

0.49

Located in hilly area

0.39

0.49

Per capita income (thousand yuan)

2.23

1.65

Local wage rate (yuan)

16.22

11.35

Tax (yuan/mu)

107.77

65.99

1 CV = coefficient of variation.

Results

We ran six probit regressions to test the three hypotheses based on individual migration decisions, the results of which are presented in Table 2. In R1, we ran an explorative regression with only land endowment together with the control variables. It is shown that land endowment has a negative but highly insignificant effect on the decision to migrate. This result shows that neither the income effect nor the substitution effect dominates over the whole sample.

Before testing the three hypotheses, we first discuss the estimates of the control variables. They are basically the same across the regressions so we will base our discussion on the results of R1. The estimates generally make sense. In the group of personal characteristics, age significantly reduces a person's propensity to migrate. At the margin, being one year older implies a 0.3 percent reduction in the individual's probability of migration. It is also shown that a man is significantly more likely to migrate than a woman. The marginal effect of being a man regarding the probability of migration is 13.3 per cent. In contrast, an individual's marital status is not a significant factor in determining his or her probability of migration. Lastly, education helps a person to migrate. Calculation shows that an additional year of schooling increases an individual's probability of migration by 1.1 percent. All these results are consistent with the findings of other studies (e.g., Zhao, 1997, 1999).

In the group of household characteristics, the effect of per capita wealth is significantly positive although its economic significance is low: 1 000 yuan of wealth increases a person's probability of migration by only 0.4 percent. To the extent that decreasing absolute risk aversion is frequently found in empirical studies, however, this result reinforces the notion that migration is a risky endeavour, a premise invoked in the theoretical model. As one can envision, the result strongly shows that a person in a family with a larger population is more likely to migrate, but a person in a family with a larger children/labour ratio is less likely to migrate. At the margin, one additional household member increases a person's propensity to migrate by 2.5 percent, and one more child relative to labour reduces his propensity by 15.1 percent. The result also shows that land quality is not a significant factor in predicting a person's propensity to migrate, but the perception of the difficulty of leasing out land does reduce a person's willingness to migrate. If village approval is required for a land lease, a person's probability of migration is reduced by 6.6 percent.

TABLE 2

Regression results (2 050 cases)

Variables

R1

R2

R3

R4

R5

R6

Intercept

0.069

-0.081

0.151

0.447

0.161

0.492

 

(0.551)

(0.555)

(0.554)

(0.569)

(0.560)

(0.573)

Main variables

Personal land endowment

-0.058

0.417**

 

 

 

 

 

(0.051)

(0.199)

 

 

 

 

Personal land endowment squared

 

-0.156**

 

 

 

 

 

 

(0.069)

 

 

 

 

Personal-village land endowment difference squared

 

 

-0.150**

 

 

 

 

 

 

(0.075)

 

 

 

CV of land distribution in village

 

 

 

-0.348***

 

-0.328**

 

 

 

 

(0.125)

 

(0.130)

Number of reallocations in village

 

 

 

 

0.057

0.021

 

 

 

 

 

(0.045)

(0.046)

Household expectation of land reallocation

 

 

 

 

-0.114

-0.112

 

 

 

 

 

(0.091)

(0.091)

Personal variables

Age

-0.050***

-0.050***

-0.050***

-0.050***

-0.050*

-0.050***

 

(0.005)

(0.005)

(0.005)

(0.005)

(0.005)

( 0.005)

Sex

0.345***

0.344***

0.342***

0.340**

0.345***

0.340

 

(0.082)

(0.082)

(0.082)

(0.082)

(0.082)

(0.082)

Marital status

-0.072

-0.063

-0.057

-0.062

-0.072

-0.059

 

(0.125)

(0.126)

(0.126)

(0.126)

(0.126)

(0.126)

Number of years' schooling

0.027*

0.027*

0.029*

0.027*

0.026

 

 

(0.016)

(0.016)

(0.016)

(0.016)

(0.016)

 

Household variables

Per capita wealth (thousand yuan)

0.008**

0.008**

0.008**

0.008**

0.0084**

0.008**

 

(0.004)

(0.004)

(0.004)

(0.004)

(0.004)

(0.004 )

Population

0.077***

0.076**

0.075***

0.071**

0.073***

0.067*

 

(0.028)

(0.028)

(0.138)

(0.028)

(0.028)

(0.028)

Children/labour ratio

-0.401***

-0.405***

-0.409***

-0.400***

-0.372***

-0.382***

 

(0.138)

(0.138)

(0.138)

(0.138)

(0.138)

(0.138)

Dry land ratio

-0.028

-0.077

-0.079

-0.206

-0.08

-0.194

 

(0.173)

(0.179)

(0.170)

(0.178)

(0.173)

(0.179)

Perception of difficulty of leasing out land

-0.172*

-0.180*

-0.184*

-0.173*

-0.167

-0.167

 

(0.098)

(0.098)

(0.098)

(0.098)

(0.098)

(0.098)

Village variables

Distance to county site (km)

0.005**

0.005*

0.005**

0.004*

0.005**

0.005

 

(0.002)

(0.002)

(0.002)

(0.002)

(0.002)

(0.002)

Difficulty of reaching a bus route

0.056

0.053

0.053

0.050

0.061

0.056

 

(0.055)

(0.055)

(0.055)

(0.055)

(0.055)

(0.055)

Located in plain

0.559

0.593

0.532

0.531

0.528

0.549

 

(0.365)

(0.366 )

(0.368)

(0.365)

(0.375)

(0.374)

Located in hilly area

0.224

0.232

0.182

0.182

0.154

0.160

 

(0.384)

(0.386)

(0.387)

(0.383)

(0.393)

(0.391)

Per capita income (thousand yuan)

-0.167***

-0.174**

-0.172***

-0.151***

-0.184***

-0.162**

 

(0.049)

(0.050)

(0.049)

(0.050)

(0.051)

(0.052)

Local wage (yuan)

0.006

0.007*

0.006

0.007

0.007

0.007

 

(0.004)

(0.004)

(0.004)

(0.004)

(0.004)

(0.004)

Tax (yuan/mu)

-0.000

-0.000

0.000

-0.0001

-0.0002

-0.0002

 

(0.001)

(0.001)

(0.001)

(0.0008)

(0.001)

(0.001)

Provincial dummies

Heibei

-0.209

-0.257

-0.221

-0.345

-0.445

-0.419

 

(0.227)

(0.231)

(0.228)

(0.232)

(0.290)

(0.290)

Shannxi

-0.961***

-1.094***

-1.010***

-1.099***

-1.093***

-1.127***

 

(0.267)

( 0.273 )

(0.267)

(0.271)

(0.299)

(0.298)

Anhui

-0.596***

-0.683**

-0.641***

-0.731***

-0.599***

-0.706***

 

(0.208)

(0.212)

(0.207)

(0.209)

(0.207)

(0.210)

Hunan

-0.080

-0.216

-0.151

-0.295

-0.093

-0.275

 

(0.182)

(0.190)

(0.182)

(0.192)

(0.181)

(0.193)

Sichuan

-0.416**

-0.531***

-0.461***

-0.592***

-0.383**

-0.549***

 

(0.182)

(0.187)

(0.182)

(0.189)

(0.183)

(0.192)

Note: Standard errors are in parentheses; *, **, and *** indicate significance at the 10 percent, 5 percent and 1 percent significance levels.

In the group of village characteristics, the two variables measuring a village's transportation condition are not significant, neither are the two geographic variables. The local wage rate and tax burden are also not significant. The only significant variable is village per capita income, which reduces a person' probability of migration by 6 percent per thousand yuan at the margin. The insignificant influence of transportation and geographic conditions may simply show that transportation is not a big problem in rural China because most villages have good access to roads. The insignificance of the tax burden contradicts the casual impression that people are driven out of the countryside by a high tax burden. On the other hand, the insignificance of the local wage rate may be because it is its correlation with village income that shows a strong effect.

Finally, compared with Zhejiang, the reference province in the regressions, all five other provinces except Hunan have fewer people engaged in migration (Hunan is not significantly different from Zhejiang). It is interesting to note that this result is obtained despite the fact that Zhejiang is the richest province among the six - largely because Zhejiang has a tradition of migration and has developed extensive migration networks. The "Zhejiang Village" in Beijing is a good example (Ma and Xiang, 1998).

With the discussion of the control variables settled, we turn to the tests of the three hypotheses. In R2 we test H1, which predicts that the relationship between landholding and labour migration exhibits an inverse-U curve. This hypothesis is confirmed because the coefficients for land endowment and land endowment squared have the expected signs and are both significant at the 5 percent significance level. Calculation shows that the maximum of the inverse-U curve is 1.34 mu, very close to the mean per capita land in our sample, which is 1.13 mu.

However, this inverse-U relationship may be caused by other, uncontrolled, village characteristics. For example, villages with small land endowments may also be those with abundant off-farm opportunities so people there tend not to migrate; on the other hand, people in villages with relatively abundant land endowments would also tend not to migrate. In this case, the inverse-U relationship is caused by uncontrolled village variations, not by land endowment per se.13 To further confirm the relationship, we need to show that it holds within a village. To do that, we replace personal land endowment and its square with the square of the difference between personal land endowment and per capita village land endowment and run regression R3. If the inverse-U relationship holds within a village, we expect that this new variable will take a negative sign in the regression. This is indeed what the results of R3 show: the coefficient of this new variable is significantly negative at the 5 percent significance level. Therefore, the inverse-U relationship holds both across villages and within a village.

After H1 has been tested, we test H2, which predicts that people in a village with more equal land distribution are more likely to migrate.14 This test is carried out by R4. The control variables are the same as before; the only difference is that the two land endowment variables are substituted by the CV of village land distribution. As the regression result shows, H2 is strongly verified because the estimate for the CV passes the significance test at the 1 percent significance level.

To answer the question of reverse causality and endogeneity of the village CV, we conduct an endogeneity test suggested by Hausman (see Greene, 1997: 442-444). To conduct this test, we first regress the village CV on all the other variables appearing in R4 as well as village per capita land endowment and village leaders' opinion on whether land distribution should be adjusted to household population change (this creates a dummy variable with a value of 1 representing a positive answer to the question and a value of 0 representing a negative answer). These two variables serve as the instrumental variables for the village CV. We will not report the results of the regression here, but simply wish to emphasize that the two instrumental variables are both significant. Then, as suggested by the Hausman method, we take the regression residual and rerun R3 with all the old variables (including the CV) and the residual together. If the CV is exogenous, then the coefficient of the residual should be insignificant. This is exactly the result we obtain. We can therefore conclude that the reverse-causality and endogeneity story does not hold.

To test H3, we substitute the number of land reallocations in a village and household's expectation of a reallocation for the village CV and run regression R5. The result shows that the estimates for the two new variables are both highly insignificant, which means that neither of the two countervailing forces - one encouraging and one discouraging - contained in H3 dominates a person's decision whether or not to migrate.

To further confirm our results, we combine R4 and R5 by putting the village land distribution CV, the number of reallocations and the expectation of reallocation into one regression, R6. However, the results of R6 virtually repeat those of R4 and R5: the village CV is still significantly negative (albeit now at the 5 percent significance level) and the number of reallocations and expectation of reallocation are still insignificant.

To summarize the results, we have found strong evidence supporting H1 and H2 and have shown that neither of the two countervailing forces behind H3 dominates the other. The most important result is that more equal land distribution increases the chance of migration for people in a village. In addition, greater family wealth also helps a person to engage in migration. Finally, our results on personal characteristics are consistent with the findings of other studies.

CONCLUSIONS

The empirical results in this article have strong policy implications. Legal experts in China are trying to propose a specific law and a chapter in the Property Law to govern rural land tenure. The major aim put forward by the experts is to individualize land ownership (Liang, 1998). To the extent that the current land tenure dampens farmers' incentive to invest in land (e.g. Li, Rozelle and Brandt, 1998; Yao, 1998), this aim is justified. However, if we extend our discussion beyond agriculture and take into account the transformation of peasants into industrial workers, the justification is blurred. Indeed, the only way to solve the Chinese rural problem is to transform most of the 800 million rural residents into urban residents. With this consideration in mind, our finding that more egalitarian land distribution promotes labour migration is remarkable and merits serious consideration. As mandatory social security is still not available in rural areas, perhaps a better choice would be to concentrate on a land tenure system that accommodates both the efficiency and insurance considerations.15

This article also adds to the literature on equality and economic development. Given that labour migration moves workers from agriculture to industry, which has much higher productivity, so it contributes to economic growth. From this perspective, egalitarian land distribution has a positive effect on economic development. If the loss resulting from reduced agricultural investment is smaller than the benefit brought by labour migration, there is a net gain in more equal land distribution. An integrated analysis of the economy-wide effects of land distribution thus is warranted.

BIBLIOGRAPHY

Binswanger, H. 1981. Attitudes toward risk: theoretical implications of an experiment in rural India. Economic Journal, 91(364): 867-890.

Burgess, R. 1998. Market incompleteness and nutritional status in rural China. Paper delivered at the International Conference on Land Tenure and Agricultural Performance in Rural China, Beijing. May.

Carter, M. & Yao, Yang. 1999. Specialization without regret - transfer rights, agricultural productivity, and investment in an industrializing economy. World Bank Policy Research Working Paper 2202, October. Washington, DC.

Carter, M. & Yao, Yang. 2000. Global versus local test of separability in household models. Department of Agricultural and Applied Economics, University of Wisconsin, Madison, USA.

Chavas, J.-P. & Holt, M. 1996. Economic behavior under uncertainty: a joint analysis of risk preferences and technology. Review of Economics and Statistics, 78(2): 329-335.

Dong, Xiaoyuan. 1996. Two-tier land tenure system and sustained economic growth in post-1978 rural China. World Development, 24(5): 915-928.

Fu, Chen. 1998. Land reform in rural China since the mid-1980s. Land Reform: Land Settlement and Cooperatives, 1998/2: 122-137.

Giles, J. 2000. Risk, shock and weak-property rights in the labor allocation decisions of rural Chinese households. Department of Economics, Michigan State University, USA. (Mimeo)

Greene, W. 1997. Econometric Analysis. Upper Saddle River, New Jersey, USA, Prentice Hall.

Kung, J.K. 1994. Egalitarianism, subsistence provision and work incentives in China's agricultural collectives. World Development, 22(2): 175-188.

Kung, J.K. 2001. Property rights, specialization, and the development of land rental markets in rural China. Paper presented at the Workshop on Opportunities and Constraints in China's Rural Transition: A Critical Appraisal of Two Decades of Reform and Development, 9-10 June 2001. Hong Kong University of Science and Technology, Hong Kong SAR, China.

Liang, Huixing. 1998. A study on Chinese Property Law. Beijing, The Legal Press.

Li, Guo, Rozelle, S. & Brandt, L. 1998. Tenure, land rights, and farmer investment incentives in China. Agricultural Economics, 19(1-2): 63-71.

Liu, Shouying, Carter, M. & Yao, Yang. 1998. Dimensions and diversity of property rights in rural China: dilemmas on the road to further reform. World Development, 26(10): 1789-1806.

Liu, Yigao, Wang, Xiaoyi & Yao, Yang. Forthcoming. Migrants and the economic and social dynamics of the Chinese village. Shijiazhuang, China, Hebei Renmin Press.

Ma, L. & Xiang, Biao. 1998. Native place, migration and the emergence of peasant enclaves in Beijing. China Quarterly, 155: 546-81.

MOLSS. 1999. China labor statistical yearbook 1998. Ministry of Labor and Social Security. Beijing, China Statistical Press.

Pratt, J. 1964. Risk aversion in the small and in the large. Econometrica, 32: 122-136.

Song, L. 2001. Process of rural-urban labour migration in China: information flows and networks. Paper presented at the Workshop on Opportunities and Constraints in China's Rural Transition: A Critical Appraisal of Two Decades of Reform and Development, 9-10 June 2001. Hong Kong University of Science and Technology, Hong Kong SAR, China.

SSB. 2000. China statistical yearbook: 1999. State Statistical Bureau. Beijing, China Statistical Press.

World Bank. 1997. Staff Appraisal Report, Qinba Mountains Poverty Reduction Project. 15 May 1997. East Asia Department, World Bank, Washington, DC.

Yao, Yang. 1998. Land tenure arrangements and agricultural performance in China. Zhongguo Nongcun Guancha (Chinese Rural Survey), 6: 1-10.

Yao, Yang. 2000. Chinese land tenure and rural social security. Chinese Social Sciences Quarterly (Hong Kong), Fall 2000: 19-26.

Yang, D. 1997. China's land arrangements and rural labor mobility. China Economic Review, 8(2): 101-115.

Zhao, Yaohui. 1997. Labor migration and returns to rural education in China. American Journal of Agricultural Economics, 79(4): 1278-1287.

Zhao, Yaohui. 1999. Labor migration and earnings differences: the case of rural China. Economic Development and Cultural Change, 47(4): 767-782.


1 Master student and Associate Professor, China Center for Economic Research, Beijing University, Beijing 100871, China. E-mail: yyao@ccer.pku.edu.cn. We wish to thank participants of the workshop on "Opportunities and Constraints in China's Rural Transition: a Critical Appraisal of Two Decades of Reform and Development", held at Hong Kong University of Science and Technology from 9 to 10 June 2001, for their helpful comments.

2 For a description of China's land tenure, see, for example, Liu, Carter and Yao (1998) and Fu (1998).

3 The annual net income per mu of land does not usually exceed 300 yuan in most parts of China. This will lead to a land sale price of between 3 750 and 6 000 yuan if the discount rate is between 5 and 8 percent. This sum would purchase only 1 square metre of an apartment in Beijing, where the average apartment price is about 5 000 yuan per square metre. Prices would be lower in other cities, but not much.

4 Song (2001) found in a survey undertaken in Handan city that about 20 percent of the migrants did not have job-specific information when they left their villages. Liu, Wang and Yao (forthcoming) found, in a study of four migrant recipient villages, that about 30 percent of migrants found their jobs by random search.

5 Song (2001) found that the costs involved in the job search period ranged from 28 to 75 percent of a migrant's monthly wage, depending on where the migrant eventually found his job. A large portion of this cost was the deposit paid to the employer. A World Bank poverty alleviation project implemented in southwestern China provides a loan of up to 3 000 yuan to migrants for their training, transportation and initial living costs after they reach the city (World Bank, 1997). This represents a substantial loan given that the annual cash income for a typical farmer in the project region is less than 500 yuan.

6 More precisely, they should be called the gross income effect and gross substitution effect, as we will see later in the text.

7 The average landholding in our sample is 1.13 mu, from which the gross cash value of a year's rice harvest would be between 400 and 600 yuan, depending in which region the land is located. This income is enough for one person's train ticket (100-150 yuan for a single trip from an inland province to the coast) and about one month's meal costs. Migrants usually live in employer-provided dormitories if they find a job immediately after they reach the city, or live with acquaintances if they do not find a job upon their arrival, so their living costs are minimal.

8 We owe a referee for this argument.

9 One method found in rural China is using the set-aside land - land that is reserved for land adjustments and is not distributed to individual households. With a certain portion of the set-aside land, the village can get by for a long period without a land reallocation.

10 The survey was conducted by the Rural Development Research Centre, Ministry of Agriculture. The six provinces covered

11 The land tenure system adopted in China since the reform took place in the early 1980s.

12 The individual's earning potential from migration is also an important factor determining that person's willingness to migrate. However, earnings are likely to be determined simultaneously with the person's migration decision. We therefore need to find an instrument for earnings. The instrumental variables are likely to be the variables already appearing in the migration equation as well as the characteristics of the destination cities, which we do not have in the data set. Therefore, our approach of not including the earnings explicitly can be seen as a reduced-form approach. Since migration earnings are not likely to be correlated with a person's land endowment, this approach does not bias the estimates related to land.were Zhejiang, Anhui, Hunan, Hebei, Shannxi and Sichuan.

13 We thank an anonymous referee for pointing out this possibility.

14 Since H2 is implied by H1, we do not put these two hypotheses in one regression.

15 For a proposal of such a land tenure system, see Yao (2000).

Previous PageTable Of ContentsNext Page