HUMAN CAPITAL SPILLOVERS, LABOR MIGRATION
AND REGIONAL DEVELOPMENT IN CHINA
Yuming Fu
National University of Singapore
4 Architectural Drive, Singapore, 117566
email: rstfuym@nus.edu.sg
Stuart A. Gabriel
University of California, Los Angeles
110 Westwood Plaza, Los Angeles, CA 90095-1481
email: stuart.gabriel@anderson.ucla.edu
June 8, 2009
Human Capital Spillovers, Labor Migration and Regional Development in China
HUMAN CAPITAL SPILLOVERS, LABOR MIGRATION
AND REGIONAL DEVELOPMENT IN CHINA
Abstract
This study applies unique data from the 1990s period of economic liberalization in China to
evaluate the effects of human capital spillovers on urbanization and regional agglomeration of
human capital. We examine these effects via a utility maximizing directional migration model,
which accounts for heterogeneous migration costs and benefits among population strata. We use
model estimates to decompose and evaluate human capital spillover effects as derive from three
distinct sources, including productivity effects (social returns to schooling), skill premia (skill
complementarity in production), and non-wage benefits (quality of life and learning
opportunities). In contrast to extant literature emphasizing skill complementarity, we find
significantly stronger non-wage than wage effects in the determination of regional human capital
agglomeration. However, among low-skill migrants, non-wage benefits are substantially
reduced—due likely to urban segregation that deprives low-skill migrants of social
externalities. This finding suggests limited human capital spillovers among low-skill migrants
and hence dampened long-run growth benefits to Chinese urbanization. Finally, we find that
urban concentration of skilled workers was more important than foreign direct investment, the
prominent source of technology transfer in China during the 1990s, in attracting skilled workers.
Keywords: human capital spillovers, place-to-place migration, urbanization, regional
development, human capital agglomeration, China.
JEL codes: J24, J31, J61, O15, O18, R23
1
Human Capital Spillovers, Labor Migration and Regional Development in China
I.
Introduction
Human capital spillovers play a central role in modern theories of economic growth and
development (e.g. Romer, 1986; Lucas, 1988 & 2004). While recent literature has focused on the
link between human capital spillovers and urban economic growth (e.g. Glaeser, Scheinkman, &
Shleifer, 1995; Glaeser & Shapiro, 2003; Rodriguez-Pose & Vilalta-Bufi, 2005; Shapiro, 2006),
few studies have directly evaluated spillover effects via the prism of labor migration. In the
context of spatial equilibrium, urban growth is closely linked to interregional labor migration
(Glaeser, 2008). Regional variability in human capital spillovers may give rise to migration of
labor due to (1) productivity effects associated with social returns to education—in that local
concentration of human capital may raise the productivity and hence the wage rates of all local
workers (e.g. Rauch, 1993; Moretti, 2004a & 2004b; Fu 2007); (2) skill premia associated with
skill complementarities—as local concentrations of human capital may enhance the productivity
of high skill workers (e.g. Giannetti, 2003; Berry & Glaeser, 2005); and (3) non-wage benefits—
as local human capital concentration may enhance quality of life (e.g. Shapiro, 2006) and social
opportunities conducive to learning and human capital accumulation (e.g. Eaton & Eckstein,
1997; Glaeser, 1999; Glaeser & Mare, 2001). 1 Furthermore, the migratory response to these
human-capital-spillover related incentives may differ across skill-based population strata so as to
exacerbate regional disparities in human capital concentration and income (e.g. Giannetti, 2003;
Berry & Glaeser, 2005; Moretti, 2008; World Bank, 2008).
This study contributes to the growing literature on human capital spillovers and regional
growth in two respects. First, we seek to provide a more accurate account of the different effects
of human capital spillovers on the migratory responses of skill-based population strata as
2
Human Capital Spillovers, Labor Migration and Regional Development in China
well as evaluate their influence on the evolution of regional human capital agglomeration. We do
so by applying a utility-maximizing directional migration model to directly examine the
influence of spillover effects on interregional migration. Such a model, widely used in the placeto-place migration literature (e.g. Gabriel, Shack-Marquez & Wascher, 1993; Gabriel, Mattey &
Wascher 1995; Davies, Greenwood, & Li, 2001; Hunt & Mueller, 2004), allows for competing
migration incentives as well as heterogeneous migration costs and preferences in the
determination of mobility outcomes among population strata. We use model estimates to
decompose and identify the three distinct human capital spillover effects, associated respectively
with labor productivity, skill premia, and non-wage benefits, and to assess the influence of those
effects on both the magnitude and the skill mix of directional labor migration. Prior urban growth
studies relying on cross-city regressions typically do not control for heterogeneous migration
costs and preferences nor do they distinguish among the three different spillover effects. While
migration analyses in the extant literature largely focus on skill selectivity a la Roy (1951) and
related labor market efficiency in moving human capital to regions with relative short labor
supply (e.g. Borjas, 1987; Borjas, Bronars & Trejo, 1992; Chiswick, 1999; Dahl, 2002; Hunt &
Mueller, 2004; and Chiquiar & Hanson 2005), those studies largely overlook the influence of
human capital spillovers on migratory response as well as their related contribution to the
regional divergence in human capital concentration.2
Our second contribution is to fill the knowledge gap surrounding the influence of human
capital spillovers on labor migration and urbanization in a major emerging market economy.
New insights regarding the relative importance of the three different effects of human capital
spillovers on labor migration incentives may have important implications for economic
development. According to Lucas (2004), the mechanism of urbanization and economic
3
Human Capital Spillovers, Labor Migration and Regional Development in China
development in a developing economy hinges on the type of human capital spillovers prevalent
in its cities. In the presence of human capital spillovers in learning, cities become increasingly
attractive to new low-skill rural immigrants because of the declining cost of human capital
accumulation, a non-wage benefit associated with a growing human capital concentration in
cities. Consequently, urbanization facilitates the development goal of transition from a dual-skill
traditional economy to a high-skill modern economy. In marked contrast, when human capital
spillovers are manifest only in the productivity effect, urbanization continues over time but
without human capital accumulation by new low-skill immigrants. Accordingly, by identifying
the respective contributions of wage vs. non-wage effects of local human capital concentration to
the migration incentives of skill-based population strata, our research sheds light on the relative
importance of different types of human capital spillovers to China’s urbanization process. The
research findings also enable us to evaluate the extent to which urbanization confers income
growth opportunities on people of different skill strata.3
Our analysis of internal labor migration in China builds off of a unique dataset of stratified
place-to-place population flows derived from the 1995 one-percent national population survey in
China. The data reflect the migration choices of population groups stratified by education and
age between 1990 and 1995, a dynamic period during which regulations governing residential
location were eased and Chinese cities began to open both to foreign direct investment (FDI) and
to rural migrant workers. FDI was the primary source of technology transfer in the wake of the
surge in China’s manufacturing exports during the 1990s.
Results of the analysis suggest that both urban human capital spillovers and FDI generate
positive productivity, skill premium and non-wage benefit effects and these effects are stronger
4
Human Capital Spillovers, Labor Migration and Regional Development in China
for higher-skill migrants even when differences in migration costs are taken into account. 4
Accordingly, both urban human capital spillovers and FDI contribute to regional human capital
agglomeration; however, these two influences are not spatially correlated. Research findings
indicate a greater contribution of human capital spillovers to regional development divergence.
Furthermore, the non-wage benefit effects of urban human capital concentration appear smaller
than the total wage effect (the sum of the general wage effect and the skill premium effect),
especially for low-skill migrants, suggesting a relatively weak role of human capital spillovers in
learning in motivating the urbanization of low-skill workers during our sample period. This
result appears consistent with the observation that low-skill migrants in Chinese cities typically
lacked access to formal employment contracts and urban social services, such as education,
housing, and health services, which served to suppress both the opportunities and the incentives
for them to accumulate human capital in cities (Wang & Zuo 1999).5
The paper proceeds as follows. In Section II we provide additional background on Chinese
internal labor migration and regional development. Section III presents the empirical place-toplace migration odds model and the identification of the wage and non-wage effects associated
with urban human capital concentration and FDI. Data description and variables are contained in
Section IV. Section V reports on estimates of the directional migration model, whereas Section
VI evaluates the wage and non-wage effects of urban human capital concentration and FDI on
migration incentives and the evolution of regional human capital agglomeration. Section VII
provides concluding remarks.
5
Human Capital Spillovers, Labor Migration and Regional Development in China
II.
Chinese Internal Labor Migration Background and Literature
In the three decades of central planning prior to 1980, labor migration in China was directed
by national economic development plans (see World Bank, 2008, p154). A large wave of
migration to cities occurred in the 1950s during China’s initial phase of industrialization. In the
1960s and 1970s, labor migration was dominated by relocation of coastal industries to interior
provinces (as a part of the national defense strategy) and by assignment of educated urban youth
to rural farms (in the context of China’s Cultural Revolution). This second wave of labor
migration sought to more evenly allocate human capital across regions. During the 1980s and in
the wake of rural economic reforms, substantial numbers of farmers migrated to industrializing
rural towns. At the same time, some of the previously relocated coastal skilled workers and
urban youth were allowed to return to their original home cities. But a restrictive rural-to-urban
migration policy, via the household residential registration (Hukou) system, remained in place
and was regarded as a major impediment to efficient urban agglomeration in China (Au &
Henderson, 2006).
The 1990s was a decade of economic liberalization and elevated population mobility. The
privatization of state-owned enterprises and the inflow of FDI created strong growth in privatesector employment in Chinese cities. Also, the liberalization of the land market (Fu & Somerville,
2001) and the privatization of state housing (Fu, Tse & Nan, 2000) allowed considerable
expansion of private-sector housing opportunities. Both reforms resulted in elevated labor
migration to cities. Li (2004) estimates that inter-provincial migration totaled about 11 million
people during the first half of the 1990s; also, twice that number moved within provinces. Zhang
and Song (2003) estimate that about 70 percent of China’s urban population growth during the
6
Human Capital Spillovers, Labor Migration and Regional Development in China
1990s derived from net migration. Accordingly, the level of urbanization in China increased by
about 1 percentage point a year from 28 percent in 1990 to 33 percent in 1995 (Shen, 2005).6
The vast magnitude of China’s rural-to-urban population flow has been noted in the economics
literature and numerous studies have focused on the migration propensities of rural populations
(e.g., Johnson, 2003; Zhao, 1999 & 2003). Other studies, including Liang and White (1997), Wu
and Yao (2003) and Poncet (2006), have examined inter-provincial migration during the 1980s
and 1990s and have indicated increased responsiveness of the migration flows to regional
disparities in employment opportunities and earnings. Studies of regional development in China
also have documented increased income inequality between the coastal and interior regions
during the 1990s and have typically attributed such divergence to economic policies and
globalization (FDI) that favored coastal regions (e.g. Fujita & Hu, 2001; Démurger et. al. 2002).
Unlike prior studies, however, this study relies on unique directional migration matrices by
population strata to examine the effects of skill-based migrant self-selection in response to both
wage and non-wage incentives in the determination of directional migration flows and evolution
of regional development disparities. While extant welfare analysis of rural-urban migration
largely focuses on the impact of migration on consumption and investment in rural origins (e.g.
Zhao, 2002; De Brauw & Rozelle, 2008; De Brauw & Giles, 2008), migrant prospects in
destination cities have received much less attention in the literature. It is to those issues that we
now turn.
III. A Directional Migration Odds Model and the Identification of Wage and Nonwage Effects
Our empirical analysis involves two steps. In the first step, we specify and estimate a
7
Human Capital Spillovers, Labor Migration and Regional Development in China
directional migration model to determine the marginal effects on place-to-place migration of
regional differentials in wage rates, skill premia, urban human capital concentrations, foreign
direct investment (FDI), amenities, cost of living, and migration costs. Holding constant regional
disparities in wage rates and skill premia, the estimated effects of urban human capital
concentration and FDI represent non-wage benefits associated with interregional migration. In
the second step, the total influence of urban human capital concentration and FDI on migration
odds and skill mix of migration flows are decomposed into indirect effects associated with urban
wage-rate differentials (the productivity effect), returns-to-schooling differentials (the skill
premium effect), and direct (non-wage benefit) effects
We use a utility-maximizing framework to describe individual place-to-place migration
choice. Let a resident of type k in region i derive a utility Uk,ij from migration to region j. We
assume that the utility is a linear function of relevant economic and amenity conditions in the
origin and destination regions, denoted by a vector zij; thus,
U k ,ij = zij βk + ωk ,ij ,
(1)
where βk is a conforming vector of utility coefficients, which may vary depending on the type of
resident indexed by k, and ωk,ij is a random disturbance. Assume N alternative destination regions.
The probability that this individual migrates to region j (including j=i), denoted by πk,ij, is
π k ,ij = Prob(U k ,ij > U k ,is )
for all s ≠ j .
(2)
McFadden (1973) has shown that when the N disturbances are independent and identically
distributed with Weibull distribution, 7 the probability in Equation (2) is a conditional logit
function:
8
Human Capital Spillovers, Labor Migration and Regional Development in China
π k ,ij =
exp( z ij β k )
.
N
∑ exp(z β
j =1
ij
k
(3)
)
Direct estimation of the conditional logit function of πk,ij, as in Davies, Greenwood and Li
(2001), is complex because πk,ij depends on the vector zij for all potential destinations. A simpler
approach, as suggested in Gabriel, Shack-Marquez and Wascher (1993) and Gabriel, Mattey and
Wascher (1995), is to estimate the function of the migration odds ratio πk,ij /πk,ii, which describes
the probability of an individual in region i moving to region j, relative to that of staying put:
π k ,ij
= exp(Zij β k ) .
π k ,ii
(4)
In Equation (4) Zij ≡ zij− zii measures the relevant origin and destination conditions and Zijβk
represents the net benefit of migration for type-k residents. Similar to prior studies of place-toplace migration, we include in Zij (i) differential economic conditions between the destination
and the origin, denoted by Ej−Ei, (ii) migration “push” factors in the origin regions Bi, (iii)
migration “pull” factors in the destination regions Aj, (iv) the expected cost of migration between
the origin and the destination Xij, and (v) origin fixed effects DOi and the destination fixed
effects DDj. Thus the net benefit from migrating from region i to region j can be expressed as:
Zijβ k = ( E j − Ei )β kE + Biβ kB + A jβ kA + X ij β kD + DOi + DD j , for all j ≠ i, (5)
where βkE, βkB, βkA, and βkD are related coefficient vectors.
We use data on place-to-place population flows to measure the migration odds ratio. Let mk,ij
be the number of type-k residents in region i who migrated to region j over a given period and
mk,ii be the number of type-k residents remaining in region i. With a prediction error εk,ij, the
9
Human Capital Spillovers, Labor Migration and Regional Development in China
empirical counterpart of Equation (4) becomes:
rk ,ij ≡
mk ,ij
mk ,ii
= exp(Zij β k ) + ε k ,ij .
(6)
The error term in Equation (6), however, is not normally distributed, since, for any give origin,
we typically observe a significantly positive odds ratio only for a few destinations but a small or
zero odds ratio for most of the destinations. Extant studies that estimate the migration odds ratio
equation often use a log linear transformation (e.g. Gabriel, Justman & Levy, 1987; Gabriel,
Shack-Marquez & Wascher, 1993; Poncet, 2006). But, in cases involving many potential
destinations and hence a large number of near-zero directional migration odds ratios, the log
transformation is problematic. Accordingly, we instead use a Box-Cox transformation of
Equation (6):
(r )
k ,ij
λ
= exp(λ Zij β k ) + ε k ,ij ,
(7)
where the constant λ is chosen such that the residuals have an approximate normal distribution.8
We employ a unique dataset of place-to-place population flows fully stratified by education
attainment and age. Such stratification enables an examination of how migration motives and
costs vary across different population groups. We use a parsimonious specification and assume
that the education and age effects are additive; that is, for education group e and age group a, the
migration odds ratio is given by:
(r )
a ,e ,ij
λ
= exp(λ Zij βa ) × exp(λ Zij βe ) + ε a ,e,ij .
(8)
Our analysis is similar to Hunt and Mueller (2004) in that we seek to account for both migrant
selectivity as well as the tradeoff between wage and non-wage motives. Hunt and Mueller (2004),
however, employ a nested logit specification where the skill and demographic attributes
10
Human Capital Spillovers, Labor Migration and Regional Development in China
of the individuals are assumed to affect the upper-level choice of whether or not to migrate and
the location attributes are assumed to influence the lower-level choice of migration destinations.
The nested logit model applies to individual choices, where each choice is influenced by the
attributes of all potential destinations. The computational demands of such a model often limit
the sample size and the number of potential destinations that can be analyzed. Our stratified odds
ratio specification of Equation (8) provides a much simpler approach to analyzing the
determinants of the size and skill composition of place-to-place migration flows covering large
population samples and a large number of potential destinations.
To evaluate the different effects of a regional growth determinant q, such as urban human
capital concentration and FDI, on migration incentives, let Zij ≡ {qij , wij , sij , Z% ij } , where qij, wij and
sij denote, respectively, the origin-destination differentials in growth determinant q, urban wage
rate, and skill premium and Z% ij denote the rest of the variables in Zij. Further, let
f k ,ij ≡ exp(Zij β k ) = exp ( β w,k wij + β s ,k sij + β q ,k qij ) ⋅ f%k ,ij
(
f%k ,ij ≡ exp Z% ij β% k
)
(9)
(10)
where β% ij is a subset of βij corresponding to Z% ij , be respectively the baseline predicted directional
migration odds as in Equation (6) and the alternative predicted migration odds at the mean values
of qij, wij and sij, which are zero. To isolate the productivity and skill premium effects of qij on fk,ij,
we allow wij and sij to be endogenous to qij but hold Z% ij constant, so that:
d ( ln f k ,ij )
dqij
= β w,k
∂wij
∂qij
+ β s ,k
∂sij
∂qij
+ β q ,k
(11)
= β w , k δ w, q + β s ,k δ s , q + β q , k ,
11
Human Capital Spillovers, Labor Migration and Regional Development in China
where δw,q and δs,q are the coefficient estimates obtained from the regression of wij and sij
respectively on their determinants including qij. Equation (11) provides the decomposition of the
influence of q on log odds ratio into the productivity effect, skill premium effect, and non-wage
benefit effect, holding other determinants of the place-to-place migration odds constant.
According to Equation (11), the component effects on migration odds fk,ij are proportional to fk,ij;
hence, the impact of urban human capital concentration or FDI on urban growth, which works
through interregional migration, depends not only on their effects on productivity and non-wage
benefits but also on the migration costs between the city in question and other regions, which
affect the baseline migration odds. Failure to account for migration costs in many extant studies
of urban growth likely results in inaccurate estimates of the effects of human capital
concentration and FDI.
Furthermore, to evaluate the effects of urban human capital concentration and FDI on the
evolution of regional human capital agglomeration, we calculate the baseline mean years of
schooling of directional migration flows hij as
hij = ∑ he ( f young ,e ,ij m young ,e,ii + f old ,e,ij mold ,e,ii )
e
∑( f
young ,e ,ij
m young ,e,ii + f old ,e,ij mold ,e,ii )
(12)
e
where he is assumed to be 5 for e = “primary school or below”, 8, for “middle school”, and 12,
for “high school or above”. Note that hij is not a linear function of the place-to-place migration
odds. Hence the effects of qij on hij cannot be simply expressed as a linear combination of
β w,k δ w,q , β s ,k δ s ,q and β q ,k . For decomposition, we compute the mean years of schooling under
alternative predicted migration odds given by Equation (10):
(
h%ij = ∑ he f%young ,e,ij m young ,e,ii + f%old ,e,ij mold ,e,ii
e
) ∑ ( f%
young ,e ,ij
)
m young ,e,ii + f%old ,e,ij mold ,e,ii .
(13)
e
12
Human Capital Spillovers, Labor Migration and Regional Development in China
We then regress the counterfactual change Δhij ≡ hij − h%ij on wij, sij and qij to obtain estimates γw, γs
and γq, so that the total effect of q on hij, holding Z% ij constant, can be expressed as
dhij
dqij
= γ wδ w , q + γ s δ s , q + γ q .
(14)
The decomposition according to Equation (11) and Equation (14) allows the effects of urban
human capital concentration to be compared with those of FDI.
IV. Data and Variables
Our place-to-place population flow statistics are derived from the one-percent Population
Survey conducted in 1995 (see National Population Survey Office, 1997). The survey covers all
30 provincial-level jurisdictions in China in 1995 (including 22 provinces, 5 autonomous regions
and 3 provincial level cities); in this study, we refer to these 30 jurisdictions as provinces. Within
each province, the survey randomly sampled one-half to one-third of the county-level
jurisdictions. Altogether, more than 12 million people were sampled in this first comprehensive
source of information on unregulated migration choices among Chinese households. We use the
population survey data to compute province-to-province population-flow matrices for sample
strata sorted by gender, age and schooling. For each stratum, the population-flow statistics mk,ij,
i,j = 1,…,29, are defined as the number of people residing in region j in 1995 (for no less than 6
months) whose regular residence 5 years earlier (in 1990) was in region i (the origin region).
Accordingly, our analysis is based on the first comprehensive and statistically-based estimates of
directional migration by socio-economic strata available for modern China.
13
Human Capital Spillovers, Labor Migration and Regional Development in China
The 30 provinces vary considerably in population size. Tibet, located in the southwest high
plateau, was the least populated region with just 2.2 million people in 1990, whereas Sichuan,
located in the fertile upper-Yangzhi-River basin, was the most populous of China’s provinces
with over 100 million people. Per-capita national income also varied considerably, from RMB
654 Yuan in the southwest province of Guizhou to RMB 4,822 Yuan in Shanghai, the emerging
economic powerhouse at the mouth of Yangzhi River in the east.9 These provinces, except Tibet,
form the set of alternative origins and destinations in our place-to-place migration analysis. We
exclude Tibet from our analysis as it was largely isolated from the rest of the Chinese economy
during our sample period due to the lack of transportation links.
Appendix A lists the regional economic indicators that we employ to explain the observed
migration odds ratios, including average urban wage rates, urban labor force size, cost-of-living
measured by per capita consumer spending, urban housing affordability measured by housing
space per person, foreign direct investment (FDI) as a share of provincial fixed investment, urban
human capital concentration measured by high-school graduate share of urban population, and
urban skill premium measured by returns to schooling. In addition, we include an indicator of
summer and winter temperature extremes to reflect upon provincial climatic amenities. The 29
provinces are grouped into seven geographic regions so as to control for regional fixed effects
due to cultural, geographical and policy differences not accounted for by the other variables. In
general, per-capita income is lower in the western interior and rises with proximity to the east
coast.
The last two columns in Appendix A show the rate of inter-provincial migration in our
sample of economically active population (aged 15 to 65). Note that the data reveal limited
14
Human Capital Spillovers, Labor Migration and Regional Development in China
provincial level variability in rates of out-migration, relative to a preponderance of in-migration
to a few fast-growing provinces. The four biggest winners in terms of population gain as a
percentage of their 1990 population were Beijing (the national capital in the northern coastal
region), Shanghai (the emerging commercial center of China in the southern coastal region),
Guangdong (a leading area of economic liberalization in the coastal south), and Xinjiang (a far
northwestern province with a rich resource base). Overall, about 1.14% of the population
migrated beyond their original province during the 5 year period. Zhang et. al. (1998) show that
household migration during this period was predominately rural-to-urban; in that regard, urbanto-rural migration (from both urban districts and county-level cities) accounted for only 3.6% of
all migrants, whereas rural-to-urban migration accounted for nearly 60% of all population moves
(see Table 1). Overall, 78% of migrants originated from rural areas. Rural-rural migration was
more likely within provinces, whereas urban destinations were more dominant among interprovincial moves.
*** Insert Table 1 about here ***
Table 2 shows the distribution of the sample population by age and educational attainment.
Our sample of economically active persons (aged 15 to 65) in the one-percent National
Population Survey consists of about 8.4 million people, of which 52% are below age 35. The
majority of the population did not complete high school (about 49% had at most a primary
school education and whereas 36% had completed middle school). Note as well that the younger
age group (below age 35) had a higher education attainment level than the older group.
Furthermore, Table 2 shows that the younger and more educated groups were more mobile (had
higher average migration odds ratios). The correlation coefficients of the migration odds ratios
15
Human Capital Spillovers, Labor Migration and Regional Development in China
indicate substantial variability in migration choice patterns across the age and education groups.
Those differences are most pronounced between the low-education groups (those with at most
primary school education) and the more educated groups. Similar to Hunt and Mueller (2004),
we find relatively small differences in migration choice by gender. Accordingly, we focus our
analysis on age and education based population strata as shown in Table 2.
*** Insert Table 2 about here ***
Table 3 provides the sample statistics of the explanatory variables used in Zij, where origindestination (o-d) differences are calculated as the destination value minus the origin value of the
variables. We report the mean absolute value and the standard deviation of each variable (the
mean values of o-d difference variables are always zero). Since the majority of inter-regional
migrants are of rural origin, we include in Zij measures of both rural farming conditions (arable
land per rural resident) and urban-rural consumption disparities (urban-to-rural ratio of per capita
consumer spending) to account for the push incentives of rural residents to migrate. A larger
allocation of arable land per rural resident would contribute to a higher farming income and
hence reduced incentives for rural residents to leave their home. The origin urban-rural ratio of
consumer spending reflects the urban-rural gap in income and cost of living, which often reflects
a lack of trade openness in the region (Wei & Wu 2001). We hypothesize that a larger gap would
motivate more highly skilled workers to migrate from the region but deter unskilled workers
from leaving their rural home.
*** Insert Table 3 about here ***
16
Human Capital Spillovers, Labor Migration and Regional Development in China
The remaining independent variables in Zij are intended to account for destination choices.
They are divided into three groups: (i) real wage incentives a la the Roy hypothesis; (ii) nonwage incentives; and (iii) costs of migration. Among the first group are the log values of both
origin-destination differentials in urban wage rates and in living expenses. We further include the
origin-destination differential urban returns to schooling, which allows for specification and test
of the Roy Hypothesis in the context of China’s sizable labor migration. 10 In addition, we
include a control for urban workforce size that indicates the agglomeration of urban employment
opportunities and hence further proxies differential employment opportunities between the origin
and destination provinces. A larger urban workforce may offer an additional labor-market
pooling benefit for the more educated workers. The average urban housing space per person is
also included to further account for the real wage differences between alternative destinations.
The non-wage migration incentives are accounted for by three measures. The first one
pertains to origin-destination differences in natural amenities. Here we compute a temperature
discomfort index, defined as the square root of the sum of the lowest temperature squared and
the highest temperature squared. A high value of this index indicates that the province has a
more extreme temperature either in the winter or in the summer or in both. We would expect
provinces with a temperate climate (a low discomfort index) to be more attractive to migrants.
The other two measures pertain to social interaction opportunities that enhance learning: the
level of FDI and the high-school graduate share of urban population. As suggested earlier, FDI
introduces new production technologies and provides access to foreign markets, the knowledge
of which is valuable in China’s emerging market economy. The share of high-school graduates
in the city population indicates the strength of learning externalities as hypothesized in Lucas
(2004).11 Although these last two measures would also affect the local demand for skills,
17
Human Capital Spillovers, Labor Migration and Regional Development in China
such effects would have been accounted for by the skill-specific wage differentials across
regions.
Finally, the analysis includes controls for migration costs and for provincial-level fixed
effects. A bell-shaped distance function is included to capture the variable cost of migration,
which is assumed to increase with distance at an increasing rate for short distances but at a
decreasing rate for long distances. The variable cost not only reflects the pecuniary transport cost
of relocation but also the cultural and information gaps that tend to be significantly higher once
an individual moves beyond the adjacent provinces. Specifically, the distance-specific
disincentive of migration is computed as exp(-(dij/d0)2), where dij is the direct distance between
the capitals of the origin and the destination provinces as measured on a map and d0 is chosen to
be 9, or about ¾ of the median dij, to maximize the explanatory power the distance function.
Finally a constant is included to allow for a fixed cost of migration which must be overcome for
the migration to be profitable.
Table 3 also shows the correlation between the variables in Zij. The two origin variables are
positively correlated; places more abundant in farm land per rural resident also tend to have
higher urban per capita purchasing power relative to the consumption level in rural areas. More
temperate provinces (hence a lower temperature discomfort index) tend to offer higher expected
urban wage rates but a relatively lower value for other variables. Provinces with a relatively
higher concentration of high-school graduates in cities offer somewhat higher returns to
schooling, possibly reflecting the skill complementarity in employment suggested by Giannetti
(2003) and Berry and Glaeser (2005). We further note that destination levels of FDI and human
capital concentration raise urban cost of living somewhat faster than urban wage rates—an
18
Human Capital Spillovers, Labor Migration and Regional Development in China
indication of positive non-wage benefits from FDI and human capital-related spillovers
according to Roback (1982).
In addition, Table 3 shows the correlation of migration flow statistics with Zij. We observe
that migration odds are somewhat positively correlated with o-d differential wage rates and
housing space per person but slightly negatively correlated with o-d differential high-school
graduate share of urban population. We observe some degree of skill-based selectivity in
migration choices with respect to migration distance and o-d differentials in urban workforce
size, FDI level, and return to schooling. Further, the mean years of schooling of migration flows
is somewhat larger to regions with a relatively lower temperature index (more favorable climate),
to relatively more distance regions, and, interestingly, to regions with a lower high-school
graduate share of urban population.
V.
Estimates of the Directional Migration Odds Model
The system of six directional migration odds equations, one for each of the three education
groups (primary-school education or below, middle-school education, and high-school education
or above) by two economically actively age groups (aged between 15-34 and 35-65), as given by
Equation (8), are estimated using equation-weighted least squares (WLS) methods. The WLS
estimator accounts for heterogeneity in the residual error across the equations. We choose a λ of
0.25 to obtain an approximately normal distribution in the residual errors. In order to limit the
number of coefficients to be estimated and to improve the robustness of coefficient estimates, we
assume that the regional fixed effects are common to all education and age groups. We thus have
three sets of βe estimates pertaining to the three education groups and one set of βa reflecting the
19
Human Capital Spillovers, Labor Migration and Regional Development in China
differential migration incentives of the older relative to the younger age groups; these estimates
are reported in Table 4.
*** Insert Table 4 about here ***
An interesting picture of skill-based selectivity in labor migration emerges from Table 4. The
availability of farm land in origin rural areas significantly affects the incentive to migrate. For
example, more limited availability of farm land results in elevated propensities to migrate among
the more highly educated groups, whereas the older group shows damped propensities to migrate.
As expected, disparities in urban-rural per-capita consumption spending, reflecting a lack of
urban-rural economic integration in the origin region, exerts varying effects on out-migration
across population education strata; it depresses migration odds of low-skill rural workers, who
might be more financially constrained to migrate, but spurs migration among the more highly
educated and older groups who are able to seek more distant opportunities.
The next set of measures proxy for real-wage related incentives to migrate. We find the urban
wage differential between destination and origin provinces to be positive and highly significant
in the determination of the propensity to migrate; the effect is somewhat smaller in magnitude
for low-skill migrants as well as for older migrants. Older migrants have a shorter time frame
over which to discount pecuniary returns to a move, as do low-skill migrants whose chance of
establishing a career in cities is often limited. Further, our estimates support the Roy (1951)
hypothesis in the context of a major emerging market economy. Migration to places with
relatively higher returns-to-schooling is damped among migrants with only a primary school
education, with cost of living held constant; in marked contrast, migration to those same
provinces by more highly educated migrants is significantly elevated. Older migrants are also
20
Human Capital Spillovers, Labor Migration and Regional Development in China
attracted to provinces offering greater returns-to-schooling, perhaps because these places also
offer greater returns to experience. The destination-origin region differential in the size of the
urban workforce, as a measure of labor market opportunities associated with the relative scale of
the provincial urban job markets, also exerts significant positive effects on place-to-place
migration. The estimated effects vary significantly across educational strata, however, and are
considerably more pronounced for individuals with at least a middle school education relative to
those with not greater than a primary school education, indicating the importance of the labormarket pooling benefits for skilled workers.
As expected, differences in living expenses per capita between destination and origin regions
work to significantly dampen migration. However, those effects are less pronounced in size than
the pull of the wage differentials. Moreover, higher levels of destination average living space
per person, as a measure of housing availability and affordability, serve to significantly enhance
migration to those areas. Those effects appear somewhat less important among higher human
capital migrants, for whom housing affordability may be less of an issue.
With respect to non-wage incentives to migrate, the presence of temperature extremes,
indicative of a less amenable climate, appears to discourage movement by more educated and
older groups, but has little effect on moves by low skill migrants; this result is similar to the
findings in Hunt and Mueller (2004) and reflects a greater willingness to pay for climate
amenities by higher skilled population strata.
We further investigate the benefits derived from the presence of FDI and human capital,
holding constant regional variations in skill-based compensation. We find that both urban
concentration of high school graduates and FDI share of provincial fixed investment offer
21
Human Capital Spillovers, Labor Migration and Regional Development in China
positive non-wage benefits across education strata; however, migrants in the top educational
stratum appear to attach notably greater significance to these benefits than do those in the lowest
educational stratum. As discussed above, an important benefit of local concentration of FDI and
human capital is the learning opportunities they provide. As such, one might expect low-skill
migrants to attach greater importance to these benefits than do high-skill migrants. Indeed,
Glaeser (1999) and Lucas (2004) assume the benefit of learning opportunities accrues primarily
to the less skilled and Duleep and Regets (1999) show that migrants having a greater skill gap at
their destinations invest more in learning. However, to take the full advantage of the learning
opportunities associated with local human capital spillovers, one would need complementary
financial resources to finance the investment in human capital accumulation as well as access to
formal employment, housing and education that provide opportunities for social interaction.
However, financially-constrained low-skill migrants in Chinese cities by and large lack such
resources and access, due to a multitude of institutional barriers (Wang & Zuo, 1999). Hence, the
finding of diminished non-wage benefits to the low-skill migrants associated with destination
FDI and human capital concentration is not surprising. Of course, the non-wage benefits of
destination FDI and human capital concentration could derive from enhanced consumer
amenities (Shapiro, 2006), for which the willingness to pay would increase with individuals’
education attainment. If such were the case, we should expect older migrants to evidence a
greater willingness to pay, as they do with respect to the climate amenity. If, on the other hand,
non-wage benefits derive from human capital spillovers that enhance learning, which helps to
raise future labor income, the older migrant group would place less value on such benefits. We
observe almost no difference in the evaluation of these non-wage benefits across the age strata.
As such, it is plausible that both learning opportunities and consumer amenities play a role in
22
Human Capital Spillovers, Labor Migration and Regional Development in China
migrant assessment of non-wage benefits associated with local concentration of FDI and human
capital. Regardless, the low-skill migrants stand to benefit much less than do high-skill migrants
from the non-wage-related human capital spillovers in cities.
The results in Table 4 also include estimates of the costs of migration. The propensity to
migrate declines with distance between origin and destination regions.
While this finding
conforms to the literature more generally, note that we here specify the relationship to take a
bell-shaped form, which provides better explanatory power than a negative exponential or
quadratic form. Moreover, as would be expected, for higher human capital migrants for whom
the expected economic return on migration is elevated, the adverse effect of distance on
migratory propensities is substantially damped, relative to coefficients estimated for lower
educational attainment strata.
However, for older migrants the adverse effects of distance
between origin and destination region are relatively high, due perhaps to social ties and familial
responsibilities that serve to reduce the propensity to migrate over longer distances among older
population groups. The fixed costs of migration, as indicated by the constant terms, appear
smaller in magnitude for the higher human-capital migrants, who probably find it less costly to
be informed of distant opportunities. However, those same fixed costs are greater for older
migrants. At the mean values of the independent variables Zij, the net migration benefit measured
by Zijβk rises monotonically from -9.51, to -8.21 among population strata ranked from low to
high educational attainment. Similarly, a net migration benefit estimate of -1.64 is computed for
the older (relative to younger) migrant group, suggesting higher fixed costs of migration among
the older group, ceteris paribus.
23
Human Capital Spillovers, Labor Migration and Regional Development in China
Finally, the origin and destination region fixed effects often have significant impacts on
provincial rates of in- and out-migration. The Northeast area, for example, characterized by a
cold harsh climate and declining production in heavy industry, sends more migrants but receives
fewer migrants than would be predicted by the model’s independent variables. Migration is
damped to the Central and Southwestern regions, areas characterized by poorer infrastructure
and less open economies.
VI. Decomposition of the Effects of Human Capital Concentration and FDI on
Skill-based Migration Flows.
In order to evaluate and decompose the effects of urban human capital concentration (q1) and
FDI share of fixed investment (q2) on migration odds, we first estimate the effects of these two
determinants on urban wage rates (productivity) and returns to schooling (skill premium)
respectively. In so doing, we control for other determinants of urban wage rates and returns to
schooling, including climate (temperature discomfort index), share of state-owned enterprises
(SOE) in urban total employment, and provincial urban workforce size.12 The OLS estimates,
reported in Table 5, show that (1) a percentage point increase in high-school graduate share (q1)
raises urban wage rates by about 1.68% (δw,q1) and returns to schooling by 0.056 percentage
point (δs,q1); (2) a percentage point increase in FDI share (q2) raises urban wage rate by about
0.33% (δw,q2) and return to schooling by 0.025 percentage point (δs,q2). These results indicate the
presence of strong productivity spillovers associated with both urban human capital
concentration and FDI. In addition, the positive skill premia associated with urban high-school
graduate share and FDI share suggest the presence of positive skill complementarities in
production.
24
Human Capital Spillovers, Labor Migration and Regional Development in China
*** Insert Table 5 and Table 6 about here ***
The estimates of the productivity, skill premium, and non-wage benefit effects of urban
human capital concentration and FDI on migration odds and mean years of schooling, computed
according to Equation (11) and Equation (14), are reported in Table 6. In the case of the lowest
education stratum, a percentage point increase in o-d differential high-school graduate share (q1)
raises the migration odds by about 5.553% through the productivity effect, by -0.325% through
the skill premium effect, and by 2.981% through the non-wage effect, whereas a percentage
point increase in o-d differential FDI share (q2) raises the migration odds by about
1.092%, -0.147% and 1.538%, respectively, through the three different effects. Across all
education strata, both the productivity effect and the skill premium effect associated with highschool graduate share appear considerably greater than those attributable to the FDI share,
although the FDI share is somewhat more variable than the high-school graduate share across the
provinces as shown in Table 3.
The total wage effect (productivity effect plus skill premium effect) associated with a
percentage increase in high-school graduate share is 5.228%, 6.267% and 6.945%, respectively,
for the three education strata; in comparison, the non-wage benefit effect is 2.981%, 4.409% and
6.378%, respectively. The results indicate notably smaller non-wage benefits of urban human
capital spillovers for the low-skill migrants in comparison both with the productivity benefits and
with the non-wage benefits for higher-skill migrants. With respect to FDI share, the non-wage
effects are greater than the total wage effects for all education strata, although both effects are
notably smaller for the bottom education stratum. The substantial weaker effects of both urban
human capital concentration and FDI on the migration odds by low-skill population stratum,
25
Human Capital Spillovers, Labor Migration and Regional Development in China
even after controlling for differences in migration costs, may be due to segregation in
employment and housing that deprives low-skill migrants of social opportunities in destination
cities.
The skill-based selectivity in migratory choices revealed by the foregoing analysis may result
in rising regional disparities in human capital agglomeration. In order to evaluate the influence
of urban human capital concentration and FDI on the evolution of regional human capital
agglomeration, we decompose the counterfactual changes in the mean years of schooling of
directional migration flows Δhij ≡ hij − h%ij , computed according to Equations (12) and (13). Two
versions of h%ij are computed, one using the baseline age effect on migration odds
(
)
(
)
% β%
f%old ,e,ij ≡ exp ( Z ij β a ) exp Z
ij e
(
)
and
the
other
using
an
alternative
age
effect
% β% exp Z
% β% . As reported in the last two columns in Table 6, the negative effect
f%old ,e,ij ≡ exp Z
ij a
ij e
of the differential urban wage rate on the migration odds of the older population, who are
typically less schooled than the young population, and the positive effect of differential return to
schooling largely offset each other, so that the total wage effect on the mean years of schooling is
little affected by using the baseline age effect vis-a-vis the alternative age effect.
The decomposition results show that both urban concentration of human capital and level of
foreign direct investment contribute to human capital agglomeration. The impact of the former,
however, appears much stronger than that of the latter: a percentage point increase in high-school
graduate share, on the one hand, raises the mean years of schooling by 0.012 year and 0.025 year,
respectively, through the wage and the non-wage effects; a percentage point increase in FDI
share, on the other hand, raises the mean years of schooling by about 0.004 year and 0.016 year,
26
Human Capital Spillovers, Labor Migration and Regional Development in China
respectively. If human capital is central to urban growth, as many authors maintain (e.g. Lucas,
1988; Glaeser, Scheinkman & Shleifer, 1995; Black & Henderson 1999), the predominant role
that urban human capital spillovers play in generating regional divergences in human capital
agglomeration, as suggested by our results, is in contrast to earlier studies that attribute Chinese
regional disparities in economic development mainly to economic policies and globalization,
both of which were closely linked to FDI concentration (e.g., Fujita & Hu, 2001; Démurger et. al.
2002). Furthermore, we note that the wage (productivity and skill complementarity) effects of
both high-school graduate share and FDI share on the mean years of schooling of migration
flows and hence on regional human capital agglomeration are not nearly as strong as the nonwage effects. This finding appears to contrast with findings reported in the literature that recent
increases in human capital agglomeration in developed economies are largely due to positive
skill complementarities in production (Giannetti, 2003; Berry & Glaeser, 2005).
VII. Conclusions
This study applies unique data from the 1990s period of economic liberalization in China to
evaluate the influence of human capital spillovers on urbanization and regional human capital
agglomeration. We evaluate these effects via a utility maximizing directional migration model,
which accounts for heterogeneous migration costs and willingness to pay for migration benefits
among population strata. The modeling framework identifies human capital spillovers that derive
from three distinct sources, including productivity effects (social returns to schooling), skill
premia (skill complementarity in production), and non-wage benefits (quality of life and learning
opportunities). Results show that non-wage benefits of urban human capital spillovers are
almost as large as wage effects among educated migrants. However, among low-skill migrants,
27
Human Capital Spillovers, Labor Migration and Regional Development in China
the non-wage benefits are substantially reduced—due likely to segregation in employment and
housing that deprives low-skill migrants of social opportunities in cities.13 The analysis further
evaluates the estimated magnitudes of spillover effects attributable to foreign direct investment
(FDI), given the prominent role of FDI as a source of technology transfer in China during the
1990s. Results indicate that spillover effects associated with FDI are considerably smaller than
those associated with urban human capital agglomeration—both in terms of wage and non-wage
effects and across skill-based population strata. These findings stand in marked contrast to prior
studies indicating the importance of FDI to disparate regional development trajectories in China
(e.g., Fujita & Hu, 2001; Démurger et. al. 2002). Furthermore, we find significantly stronger
non-wage than wage effects on regional human capital agglomeration, suggesting the limitations
of skill-premia based explanations of regional human capital agglomeration (Giannetti, 2003;
Berry & Glaeser 2005). These results demonstrate the usefulness of the utility-maximizing
directional migration framework in evaluation of structural determinants of population mobility
and regional development.
Research findings help to resolve conflicting policy implications often found in the extant
empirical rural-urban migration literature (Lall, Selod & Shalizi, 2006).
Our analysis, for
example, suggests that, in order to reap elevated economic and social benefits of urbanization,
greater policy attention must be paid to the removal of the institutional barriers that segregate
low-skill migrants in cities and diminish their human capital investment incentives. In the long
run, it is the creation of opportunities for the accumulation of human capital among low-skill
migrants, as in Lucas (2004), which allows urbanization to generate sustainable and equitable
income growth opportunities in emerging economies. Improved employment and social welfare
outcomes would also encourage low-skill migrants to take advantage of more distant
28
Human Capital Spillovers, Labor Migration and Regional Development in China
economic opportunities associated with uneven regional development and human capital
agglomerations (World Bank 2008). Further, our findings suggest that the benefits of social
interactions among educated workers have been more important than foreign direct investment to
disparate regional trajectories in China’s economic development. Accordingly, regional policies
that both attract human capital and foster human capital accumulation may succeed in generating
more sustained economic growth.
Acknowledgement: The authors are grateful for helpful comments from Bob Edelstein, John
Quigley, Daniel Felsenstein and seminar participants at the Technion-Tel Aviv University-USC
Israel Symposium on Real Estate and Urban Economics, USC Lusk Center Asia-Pacific Real
Estate Research Symposium, and the economics workshop at Fudan University. We thank Siqi
Zheng for providing the estimates of China’s regional returns to schooling.
References
Au, Chun-Chung, and J. Vernon Henderson, "How Migration Restrictions Limit Agglomeration
and Productivity in China," Journal of Development Economics 80:2 (2006), 350-388.
Berry, Christopher R., and Edward L. Glaeser, "The Divergence of Human Capital Levels
Across Cities," Papers in Regional Science 84:3 (2005), 407-444.
Black, Duncan, and J. Vernon Henderson, "A Theory of Urban Growth," Journal of Political
Economy 107:2 (1999), 252-84.
Borjas, George J., “Self-Selection and the Earnings of Immigrants,” American Economic Review
77:4 (1987), 531-553.
29
Human Capital Spillovers, Labor Migration and Regional Development in China
Borjas, George J., "Does Immigration Grease the Wheels of the Labor Market?" Brookings
Papers on Economic Activity (2001), 69-133.
Borjas, George J., Stephen G. Bronars, and Stephen J. Trejo. "Self-Selection and Internal
Migration in the United States," Journal of Urban Economics 32:2 (1992), 159-185.
Chiquiar, Daniel, and Gordon H. Hanson, "International Migration, Self-Selection, and the
Distribution of Wages: Evidence from Mexico and the United States," Journal of
Political Economy 113:2 (2005), 239-281.
Chiswick, Berry, "Are Immigrants Favorably Self-Selected?" American Economic Review 89:2
(1999), 181-185.
Dahl, Gordon B. "Mobility and the Return to education: Testing a Roy Model with Multiple
Markets," Econometrica 70:6 (2002), 2367-2420.
Davies, Paul S., Michael J. Greenwood, and Haizheng Li, "A Conditional Logit Approach to US
State-to-State Migration," Journal of Regional Science 41:2 (2001), 337-360.
De Brauw, Alan and Giles, John, "Migrant Labor Markets and the Welfare of Rural Households
in the Developing World: Evidence from China," World Bank Policy Research Working
Paper no. 4585 (2008).
De Brauw, Alan and Rozelle, Scott, "Migration and Household Investment in Rural China,"
China Economic Review 19:2 (2008), 320-35.
30
Human Capital Spillovers, Labor Migration and Regional Development in China
Démurger, Sylvie, Jeffrey D. Sachs, Wing Thye Woo, Shuming Bao, Gene Change, and Andrew
Mellinger, "Geography, Economic Policy, and Regional Development in China," NBER
Working Paper no. 8897 (2002).
Duleep, Harriet Orcutt, and Mark C. Regets, “Immigrants and Human-Capital Investment,”
American Economic Review 89:2 (1999), 186-191.
Eaton, Jonathan, and Zvi Eckstein, "Cities and Growth: Theory and Evidence from France and
Japan." Regional Science and Urban Economics 27 (1997), 443-474.
Fan, C. Simon and Oded Stark, “Rural-to-Urban Migration, Human Capital, and
Agglomeration,” Journal of Economic Behavior and Organization 68:1 (2008), 234-247.
Fujita, Masahisa, and Dapeng Hu, "Regional Disparities in China 1985-1994: The Effects of
Globalization and Economic Liberalization," Annals of Regional Science 35:1 (2001), 337.
Fu, Shihe, "Smart Cafe Cities: Testing Human Capital Externalities in the Boston Metropolitan
Area," Journal of Urban Economics 61:1 (2007), 86-111.
Fu, Yuming, David K. Tse, and Nan Zhou, "Housing Choice Behavior of Urban Workers in
China's Transition to a Housing Market," Journal of Urban Economics 47:1 (2000), 6187.
Fu, Yuming, and C. Tsuriel Somerville, "Site Density Restrictions: Measurement and Empirical
Analysis." Journal of Urban Economics 49:2 (2001), 404-423.
31
Human Capital Spillovers, Labor Migration and Regional Development in China
Gabriel, Stuart A., M. Justman, and A. Levy, "Place-to-Place Migration in Israel: Estimates of a
Logistic Model," Regional Science and Urban Economics 17 (1987), 595-606.
Gabriel, Stuart A., J. Shack-Marquez, and William L. Wascher, "Does Migration Arbitrage
Regional Labor Market Differentials?" Regional Science and Urban Economics 23
(1993), 211-223.
Gabriel, Stuart A., Joe P. Mattey, and William L. Wascher, "The Demise of California
Reconsidered: Interstate Migration Over the Economic Cycle," Economic Review
(Federal Reserve Bank of San Francisco) 2 (1995), 30-45.
Giannetti, Mariassunta, "On the Mechanics of Migration Decisions: Skill Complementarities and
Endogenous Price Differentials," Journal of Development Economics 71 (2003), 329-349.
Glaeser, Edward L., "Learning in Cities," Journal of Urban Economics 46:2 (1999), 254-77.
Glaeser, Edward L., Cities, Agglomeration and Spatial Equilibrium (New York: Oxford
University Press, 2008).
Glaeser, Edward L., and David C. Mare, "Cities and Skills," Journal of Labor Economics 19:2
(2001), 316-342.
Glaeser, Edward L. and Jesse M. Shapiro, "Urban Growth in the 1990s: Is City Living Back?"
Journal of Regional Science 43:1 (2003), 139-65.
Glaeser, Edward L., Jose A. Scheinkman, and Andrei Shleifer, "Economic-Growth in a CrossSection of Cities," Journal of Monetary Economics 36:1 (1995), 117-143.
32
Human Capital Spillovers, Labor Migration and Regional Development in China
Hunt, Gary L., and Richard E. Mueller, "North American Migration: Returns to Skill, Border
Effects, and Mobility Costs," Review of Economics and Statistics 86:4 (2004), 988-1007.
Johnson, D. Gale, "Provincial Migration in China in the 1990s," China Economic Review 14
(2003), 22-31.
Lall, Somik V., Harris Selod, and Zmarak Shalizi, “Rural-urban migration in developing
countries: a survey of theoretical predictions and empirical findings,” World Bank Policy
Research Working Paper no. 3915 (2006).
Li, Si-Ming, "Population Migration and Urbanization in China: A Comparitive Analysis of the
1990 Census and the 1995 National One Percent Sample Population Survey,"
International Migration Review 38:2 (2004), 655-685.
Liang, Zai, and Michael J. White, "Market Transition, Government Policies, and Interprovincial
Migration in China: 1983-1988," Economic Development and Cultural Change 45 (1997)
321-339.
Liu, Zhiqiang, "The External Returns to Education: Evidence from Chinese Cities," Journal of
Urban Economics 61:3 (2007), 542-64.
Lucas, Robert E., Jr., "On the Mechanics of Economic Development," Journal of Monetary
Economics 22 (1988), 3-42.
Lucas, Robert E., Jr., "Life Earnings and Rural-Urban Migration," Journal of Political Economy
112:1 (2004), S29-S59.
33
Human Capital Spillovers, Labor Migration and Regional Development in China
Moretti, Enrico, "Estimating the Social Return to Higher Education: Evidence from Longitudinal
and Repeated Cross-Sectional Data," Journal of Econometrics 121:1-2 (2004a), 175-212.
Moretti, Enrico, "Workers' Education, Spillovers and Productivity: Evidence from Plant-Level
Production Functions," American Economic Review 93:3 (2004b), 656-690.
Moretti, Enrico, "Real Wage Inequality," IZA Discussion Paper no. 3706 (2008).
National Population Survey Office, 1995 National One-Percent Population Survey Report (in
Chinese) (Beijing: China Statistics Publishing House, 1997).
Poncet, Sandra, "Provincial Migration Dynamics in China: Borders, Costs and Economic
Motivations," Regional Science and Urban Economics 36:3 (2006), 385-398.
Rauch, James E., "Productivity Gains from Geographic Concentration of Human Capital:
Evidence from the Cities," Journal of Urban Economics 34:3 (1993), 380-400.
Roback, Jennifer, “Wages, Rents, and the Quality of Life,” Journal of Political Economy 90:6
(1982), 1257-1278.
Rodriguez-Pose, Andres, and Montserrat Vilalta-Bufi, "Education, Migration, and Job
Satisfaction: The Regional Returns of Human Capital in the EU," Journal of Economic
Geography 5:5 (2005), 545-566.
Romer, Paul M., “Increasing Returns in Long-Run Growth,” Journal of Political Economy 94:5
(1986), 1002–37.
34
Human Capital Spillovers, Labor Migration and Regional Development in China
Roy, Andrew D., “Some Thoughts on the Distribution of Earnings,” Oxford Economic Papers 3
(1951), 135-146.
Shapiro, Jesse M., "Smart Cities: Quality of Life, Productivity, and the Growth Effects of Human
Capital," Review of Economics and Statistics 88:2 (2006), 324-35.
Shen, Jianfa, "Counting Urban Population in Chinese Censuses 1953-2000: Changing
Definitions, Problems and Solutions," Population Space and Place 11:5 (2005), 381-400.
Wang, Feng, and Xuejin Zuo, "Inside China's Cities: Institutional Barriers and Opportunities for
Urban Migrants," American Economic Review 89:2 (1999), 276-280.
Wei, Shang-Jin, and Yi Wu, “Globalization and Inequality: Evidence from Within China,”
NBER Working Paper no. 8611 (2001).
World Bank, World Development Report 2009: Reshaping Economic Geography (Washington,
DC: World Bank, 2008).
Wu, Zhongmin, and Shujie Yao, "Intermigration and Intramigration in China: A Theoretical and
Empirical Analysis," China Economic Review 14 (2003), 371-385.
Zhang, Kevin H., and Shunfeng Song, "Rural-Urban Migration and Urbanization in China:
Evidence from Time-Series and Cross-Section Analyses," China Economic Review 14:4
(2003): 386-400.
Zhang, W.M., X.R. Li, L.Q. Ye, G. Xie, and Y. Hu, “An Analysis of China’s Population
Migration,” Economic Research Intelligence (in Chinese, Jingji Yanjiu Cankao) 57
(1998), 2-18.
35
Human Capital Spillovers, Labor Migration and Regional Development in China
Zhang, Junsen, Yaohui Zhao, Albert Park, and Xiaoqing Song, "Economic returns to schooling
in urban China, 1988 to 2001," Journal of Comparative Economics 33:4 (2005), 730-752.
Zhao, Yaohui, "Leaving the Countryside: Rural-to-Urban Migration Decisions in China."
American Economic Review 89:2 (1999), 281-286.
Zhao, Yaohui, "Causes and consequences of return migration: Recent evidence from China."
Journal of Comparative Economics 30 (2002), 376-394.
Zhao, Yaohui, "The Role of Migrant Networks in Labor Migration: The Case of China."
Contemporary Economic Policy 21:4 (2003), 500-511.
36
Human Capital Spillovers, Labor Migration and Regional Development in China
Table 1. Distribution of migration flows between rural and urban areas
(1995 Population Survey, including migration both within and across provinces)
Origin
(1990 residing place)
Destination (1995 residing place)
Urban district
County-level cities
Rural counties
Total
Urban district
1.8%
1.0%
1.7%
4.5%
County-level cities
5.3%
10.3%
1.9%
17.5%
Rural counties
41.8%
17.6%
18.7%
78.0%
Total
48.8%
28.9%
22.3%
100%
Source: Zhang et. al. (1998), Table 1.
37
Human Capital Spillovers, Labor Migration and Regional Development in China
Table 2. Sample size distribution and sample statistics of odds ratio by population strata
Education
level
Age
group
Primary
sch. or
below
15-34
1,495,76
4
35-65
15-34
35-65
15-34
2,612,208
2,117,577
918,346
774,493
35-65
481,854
Middle
school
High
school or
above
Sample
size
% total
sample
17.8%
31.1%
25.2%
10.9%
9.2%
Odds ratio
Mean
0.056
%
0.016%
0.074%
0.034%
0.108%
Std dev.
Correlation in migration odds ratio
High
Primary school Middle school
school
0.164
age 15-34
%
35-65
15-34
35-65
0.061%
0.310%
0.164%
0.402%
0.55
0.42
0.18
0.26
0.33
0.37
0.34
0.79
0.92
0.82
5.7% 0.061% 0.401%
0.07
0.17
0.82
0.90
15-34
0.87
Source: Authors’ calculation.
38
Human Capital Spillovers, Labor Migration and Regional Development in China
Table 3. Sample statistics of explanatory variables
(29 origin by 28 destination provinces, 812 observations)
Mean
absolute
value
0.337
0.626
2.723
0.152
0.356
0.771
0.105
0.018
0.069
2.937
0.267
8.767
0.030
0.052
0.072
Variables
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
log origin urban-rural differential per capita consumer spending
log origin farm land per rural resident
o-d differential temperature discomfort index
o-d log differential urban wage rate
o-d log differential per capita urban consumer spending
o-d log differential urban workforce size
o-d differential FDI share of fixed investment
o-d differential urban return to schooling
o-d differential high school graduates share of urban population
log destination housing space per person
negative exponential o-d distance squared, exp(-dij2/81)
Mean years of schooling of o-d migration flow
o-d migration odds ratio, primary school or below (%)
o-d migration odds ratio, middle school (%)
o-d migration odds ratio, high school or above (%)
Correlation coefficients
Variables
2
3
4
5
6
7
8
1
0.569
3
-0.230
4
0.166 -0.149
5
0.059
0.376
0.633
6
0.082
0.273 -0.182
0.180
7
0.375
0.147
0.336
0.371
0.035
8
0.040
0.210
0.108
0.289
0.156
0.162
9
-0.165
0.186
0.419
0.587 -0.323
0.044
0.288
10
0.021
0.058 -0.016
0.038
0.322
0.097 -0.024
11
-0.118
0.000
0.000
0.000
0.000
0.000
0.000
12
0.166 -0.171 -0.067 -0.167
0.001
0.065 -0.014
13
0.016
0.044
0.124
0.083
0.170
0.093 -0.008
14
-0.017
0.030
0.176
0.135
0.183
0.161
0.064
15
0.027
0.038
0.176
0.166
0.273
0.192
0.159
9
10
-0.274
0.000
-0.160
-0.051
-0.025
-0.026
0.128
-0.116
0.122
0.152
0.178
Standard
Deviation
0.161
0.584
3.740
0.193
0.461
1.003
0.150
0.023
0.090
0.199
0.287
1.573
0.083
0.145
0.159
11
-0.219
0.324
0.271
0.267
Note: origin-destination (o-d) differences of the variables are calculated as the destination value minus the
origin value.
39
Human Capital Spillovers, Labor Migration and Regional Development in China
Table 4. Estimates of the directional migration odds model
Estimation method
Variables
log origin farm land per rural resident
log origin urban-rural log differential per
capita consumer spending
o-d log differential urban wage rate (βw,e)
o-d differential urban return to schooling
(βs,e)
o-d log differential urban workforce size
o-d log differential per capita urban
consumer spending
log destination housing space per person
Iterative Equation Weighted Least Square
Primary school Middle school High school or
Age
or below
above
35-65
-0.423 (2.1)** -0.905 (5.1)*** -1.153 (7.1)*** 0.301 (2.2)**
-0.445 (0.7)
1.979 (3.9)***
4.607 (11)*** 1.630 (3.2)***
3.275 (4.9)***
3.502 (6.4)***
3.787 (7.9)*** -0.698 (1.3)
5.871 (2.1)**
9.354 (4.1)*** 13.31 (4.5)***
0.451 (5.9)***
0.503 (7.7)*** 0.181 (2.2)**
-1.428 (4.0)*** -1.294 (4.4)***
-1.518 (6.0)*** 0.588 (2.0)**
-5.813 (1.6)
0.218 (2.3)**
2.961 (7.8)***
o-d differential temperature discomfort
index
o-d differential FDI share of fixed
investment (βq2,e)
o-d differential high-school graduate
share of urban population (βq1,e)
negative exponential o-d distance
squared, exp(-dij2/81)
Constant
0.046 (1.6)
Regional fixed effects
3.110 (9.8)***
-0.054 (2.2)**
2.888 (10)*** 0.429 (1.4)
-0.081 (3.8)*** -0.047 (1.9)*
1.538 (2.5)**
2.700 (5.1)***
3.593 (7.4)*** 0.340 (0.7)
2.981 (2.1)**
4.409 (4.1)***
6.378 (7.2)***
0.455 (0.4)
4.314 (22)***
3.438 (22)***
2.777 (21)***
0.372 (2.1)**
-19.06 (16)***
-19.18 (19)***
-18.39 (21)***
-3.74 (3.8)***
Origin
Destination
Coastal South
***
0.908 (5.9)
-0.424 (3.7)***
Northeast
0.778 (5.8)***
-0.831 (6.8)***
Central North
0.046 (0.4)
-1.225 (10.7)***
Central South
1.064 (7.2)***
-0.808 (6.3)***
Southwest
0.947 (5.0)***
-0.889 (5.5)***
Northwest
1.049 (4.9)***
0.356 (2.2)**
Age 15-34
Age 35-65
0.368
0.334
Age 15-34
Age 35-65
0.296
0.283
Age 15-34
Age 35-65
3.272
3.402
R-squared
0.435
0.343
Skewness
0.021
-0.031
Kurtosis
3.952
2.888
0.501
0.387
-0.632
-0.312
4.057
2.825
Note: The regression equation is ( ra ,e ,ij ) = exp(λ Z ij β a ) × exp(λ Z ij β e ) + ε a , e ,ij , where the dependent variable
λ
is migration odds ratio for each of the three education groups and two age groups and λ=0.25. βa for age 15-34
is set to be zero. The number of observations is 812 (29 origin provinces by 28 destination provinces). tstatistics are in parentheses and ***, **, * denote statistical significance at 1%, 5% and 10% level respectively.
40
Human Capital Spillovers, Labor Migration and Regional Development in China
Table 5 OLS estimates of urban wage rate and return-to-schooling determinants
Dependent variable
Determinants
o-d log
differential urban
wage rate
o-d differential
return to schooling
δw,q
δs,q
***
0.056 (4.9)***
o-d differential FDI share of fixed investment
0.331 (8.4)***
0.025 (5.3)***
o-d differential temperature discomfort index
-0.029 (18.6)***
0.001 (4.4)***
o-d differential SOE share of urban employment
-1.295 (18.1)***
0.060 (5.1)***
o-d log differential urban workforce size
-0.050 (7.8)***
0.008 (7.4)***
o-d differential (ave. years of schooling × return to schooling)
0.012 (0.4)
o-d differential high-school graduate share of urban population
R squared
1.684 (22.6)
0.514
Note: The number of observations is 812. t-statistics are in parentheses and
significance at 1% level.
0.199
***
denotes statistical
41
Human Capital Spillovers, Labor Migration and Regional Development in China
Table 6. Decomposition of the effects of urban concentration of human capital and FDI
on migration odds and regional human capital agglomeration
Migration statistics
Effects
Productivity effect (w)
High-school graduate share
of urban population (q1)
FDI share of fixed
investment (q2)
Skill premium effect (s)
Place-to-place migration odds by education
strata (%)
Primary
High
Middle
school or
school or
school
below
above
3.275
3.502
3.787
βw,e
Counterfactual change in mean
years of schooling (year)
Baseline Alternative
age effect age effect
γw
0.357
0.411
βw,eδw,q1
5.516
5.898
6.378
γwδw,q1
0.600
0.692
βw,eδw,q2
1.085
1.161
1.255
γwδw,q2
0.118
0.136
βs,e
-5.813
5.871
9.354
γs
10.815
9.463
q1
βs,eδs,q1
-0.325
0.328
0.523
γsδs,q1
0.605
0.529
q2
βs,eδs,q2
-0.147
0.148
0.236
γsδs,q2
0.273
0.239
5.191
6.226
6.901
1.205
1.221
0.939
1.309
0.391
0.375
Total wage effect
βw,eδw,q1 +
βs,eδs,q1
βw,eδw,q2 +
βs,eδs,q2
q1
q2
γwδw,q1
+ γsδs,q1
γwδw,q2
1.491
+ γsδs,q2
Direct (non-wage benefit) effect
q1
βq1,e
2.981
4.409
6.378
γq1
2.505
2.486
q2
βq2,e
1.538
2.700
3.593
γq2
1.648
1.630
Note:
Decomposition
of
migration
odds
is
given
by
Equation
(11):
d ln f k ,ij dqij = βw,k δ w,q + βs ,k δ s ,q + βq,k . Estimates of βw,e, βs,e, βq1,e and βq2,e are obtained from Table 4 and
(
)
those of δw,q and δs,q are from Table 5. Decomposition of counterfactual change in mean years of schooling of
migration flow is given by Equation (14): dhij dqij = γ wδ w,q + γ sδ s ,q + γ q . γw, γs, γq1, and γq2 are jointly estimated
in OLS regression of the counterfactual changes of mean years of schooling; these estimates are all statistically
significant at 1% level.
42
Human Capital Spillovers, Labor Migration and Regional Development in China
Appendix A. Provincial economic indicators
Region* Average
urban wage
rate, 1990
(yuan
/month)1
CN
Beijing
2,653
Tianjing
CN
2,438
Hebei
CN
2,019
Liaoning
CN
2,180
Shandong
CN
2,149
Shanghai
CS
2,917
Jiangshu
CS
2,129
Zhejian
CS
2,220
Fujian
CS
2,162
Guangdong
CS
2,929
Hainan
CS
1,982
Inner Mongolia NE
1,846
Jilin
NE
1,888
Heilongjian
NE
1,850
Shanxi
NC
2,111
Henan
NC
1,825
Shannxi
NC
2,042
Anhui
SC
1,827
Jiangxi
SC
1,729
Hubei
SC
1,903
Hunan
SC
2,038
Guangxi
SW
2,049
Sichuan
SW
2,011
Guizhou
SW
1,947
Yunnan
SW
2,130
NW
Gansu
2,407
Qinghai
NW
2,632
Ningxia
NW
2,252
Xinjiang
NW
2,289
Coefficient of variation
16.3%
Province
Urban p.c.
consumer
spending,
1990 (yuan
/month.)1
Urban-torural p.c.
consumer
expenditure
ratio, 19901
Rural p.c.
arable
land
1990
(mu)1
Urban
workforce
1990
(10,000)1
FDI share of
fixed
investment
(1990-93
cumulative)1
Housing
space per
person
(sq.m.)2
Lowest /
High-school Urban
return to highest
graduate
schooling 4 temperature
share of
(Celsius)1
urban
3
population
Outflow,
% of 1990
population2
Inflow,
% of 1990
population2
1,548
1.61
1.02
454.9
11.4%
20.8
31.0%
8.58% -15.0 /37.6
1.32%
7.16%
1,310
1.79
1.68
284.3
11.5%
16.0
22.8% 8.77 % -13.0 /36.9
1.00%
3.03%
664
1.43
2.16
652.7
3.9%
20.4
13.9% 5.16 % -12.2 /38.5
0.80%
1.01%
1,074
1.68
2.7 1,012.2
11.1%
16.8
14.9% 6.50 % -24.3 /32.0
0.56%
1.23%
681
1.32
1.48
767.5
11.3%
20.7
10.9% 5.86 % -13.5 /37.5
0.52%
0.72%
1,908
1.51
1.12
508.1
21.4%
22.0
28.2% 6.70 % -5.2 /39.6
1.06%
6.72%
841
1.07
1.38
879.9
13.9%
27.0
11.4% 6.23 % -6.5 /40.0
0.80%
1.77%
912
1.02
0.91
476.0
6.2%
32.3
9.0% 5.64 % -4.1 /40.3
1.43%
1.38%
837
1.18
0.95
310.9
40.2%
22.9
10.0% 4.37 %
4.0 /41.7
0.91%
1.54%
972
1.14
1.04
785.5
33.3%
17.8
11.8% 5.09 %
4.1 /37.8
0.49%
4.42%
708
1.39
1.17
105.9
34.1%
15.1
24.7% 7.85 %
7.4 /38.6
1.86%
1.99%
703
1.43
6.76
369.7
1.6%
13.6
16.6% 4.60 % -25.3 /33.7
1.27%
1.41%
882
1.51
4.82
517.3
5.1%
15.9
17.0% 5.82 % -27.3 /33.0
1.23%
0.62%
918
1.71
7.46
856.2
2.9%
15.3
18.1% 6.19 % -25.9 /34.2
1.75%
0.65%
607
1.34
3.07
438.7
1.8%
18.1
15.3% 5.97 % -20.2 /33.4
0.56%
0.67%
499
1.26
1.58
692.6
2.6%
19.6
12.0% 6.62 % -9.4 /36.8
1.02%
0.38%
614
1.49
2.6
379.2
5.0%
18.4
17.5% 8.66 % -9.0 /39.3
0.93%
0.60%
588
1.18
1.58
484.8
3.6%
19.1
10.2% 4.19 % -6.3 /40.3
1.63%
0.34%
652
1.30
1.33
386.1
5.5%
20.8
12.9% 5.61 % -2.1 /39.5
1.76%
0.42%
759
1.38
1.45
698.5
7.0%
24.3
17.4% 4.97 % -2.5 /39.6
0.89%
0.59%
666
1.32
1.16
551.0
5.8%
26.1
11.7% 5.18 % -4.5 /40.6
1.58%
0.43%
576
1.41
1.16
311.8
15.7%
16.0
14.8% 6.38 % -0.8 /37.0
1.73%
0.36%
616
1.32
1.13
936.1
3.9%
24.5
8.1% 6.15 % 0.55 /37.6
1.64%
0.44%
445
1.10
1.21
225.5
2.2%
15.8
13.0% 3.43 % -3.4 /33.5
1.62%
0.56%
628
1.39
1.54
291.9
1.8%
19.0
12.3% 4.51 % -0.9 /30.3
0.81%
0.69%
552
1.76
3.17
231.9
0.4%
15.4
14.7% 7.71 % -12.8 /34.3
1.27%
0.72%
786
1.86
2.71
66.4
0.3%
10.6
20.2% 1.53 % -20.0 /30.7
2.82%
1.33%
634
1.41
3.48
67.4
0.8%
15.5
17.7% 6.42 % -21.2 /34.4
1.28%
1.19%
904
1.89
4.06
300.6
0.8%
18.5
33.2% 4.58 % -22.8 /34.1
1.19%
4.24%
38.7%
19.0% 74.2%
57.6%
116.1%
26.7%
38.8%
44.7%
157%
27.3%
Sources: 1. China Statistics Yearbook; 2. 1% Population Survey, 1995; 3. Population Census 1990; 4. 1998 Urban Household Survey.* Notation for
regions: CN=coastal north, CS=coastal south, NE=northeast, NC=north central, SC=south central, SW=southwest, and NW=northwest.
43
Human Capital Spillovers, Labor Migration and Regional Development in China
Appendix B. Estimates of returns to schooling by provinces
Variable
Coeff (t-stat)
***
Variable
Coeff (t-stat)
Variable
Coeff (t-stat)
MALE
0.159 (19)
Zhejian
-0.029 (3.5)
Hubei
-0.036 (10)***
EXPYEAR
0.013 (4.7)***
Fujian
-0.042 (4.5)***
Hunan
-0.034 (4.5)***
EXPYEAR2
0.000 (0.3)
Guangdong
-0.035 (4.6)***
Guangxi
-0.022 (2.8)***
YSCH
0.086 (25)***
Hainan
-0.007 (0.5)
Sichuan
-0.024 (3.9)***
Constant
7.820 (156)***
Inner Mongolia
-0.040 (10)***
Guizhou
-0.051 (14)***
Provincial dummy × YSCH
Jilin
-0.028 (7.3)***
Yunnan
-0.041 (14)***
Tianjing
Heilongjian
-0.024 (3.0)***
Gansu
-0.009 (0.9)
0.002 (0.2)
***
Hebei
***
-0.034 (4.2)
Shanxi
-0.026 (3.3)
Qinghai
-0.070 (3.1)***
Liaoning
-0.021 (3.1)***
Henan
-0.020 (2.1)**
Ningxia
-0.022 (1.3)
Xinjiang
-0.040 (11)***
***
***
Shandong
-0.027 (3.2)
Shannxi
0.001 (0.1)
Shanghai
-0.019 (1.9)*
Anhui
-0.044 (5.0)***
Jiangshu
***
-0.023 (2.9)
Jiangxi
***
-0.030 (3.1)
89 city fixed effects
R-squared
0.386
Note: The dependent variable is ln(Employment Income). The sample is from 1998 Urban Household Survey and includes 117664 individuals from 90
cities. across 29 provinces. MALE is a dummy variable equal to unity for males, EXPYEAR is the number of years of working experience, and YSCH is the
number of years of schooling of the individual. t-statistics are in parentheses and ***, **, * denote statistical significance at 1%, 5% and 10% level
respectively.
44
Human Capital Spillovers, Labor Migration and Regional Development in China
Footnotes
1
Shapiro (2006), for example, shows that college graduate concentration contributes to
metropolitan employment growth by enhancing both local productivity and quality of life, with
the productivity enhancement accounting for about 80 percent of the local employment growth.
2
The cited studies provide evidence that interregional differences in returns to skills are
important determinants of both the magnitude and the skill composition of interregional
migration flows. Such migration flows are found to contribute to the convergence in regional
gaps between skill demand and skill supply (see Borjas 2001).
3
In contrast, in models of rural-urban migration without the benefit of learning in cities,
migration can generate perverse results of welfare deterioration, as in Fan and Stark (2008).
4
We proxy human capital spillovers and FDI via, respectively, the share of educated workers
in the local urban population in 1990 and the share of FDI in local fixed investment over 19901993 period.
5
The lack of human capital accumulation by low-skill immigrants is reflected also in
widening income inequality in Chinese cities since the early 1990s, as the Lucas (2004) model
would predict. The income Gini coefficient in China, according to World Bank (2008), rose from
33.5% in 1990 to 46.9% in 2004.
6
The accuracy of official urban population statistics is impaired by the exclusion of new rural
migrants to cities. Shen (2005)’s estimation of urban population adjusts for such undercounting.
7
The Weibull distribution has a cumulative distribution function F(ω ) = exp( −e −ω ) .
45
Human Capital Spillovers, Labor Migration and Regional Development in China
8
The Box-Cox transformation of x is (xλ−1)/λ, which becomes ln(x) when λ approaches zero
in limit.
9
In 1995, 1 US dollar was equal to about 8 RMB at the official exchange rate.
10
The provincial-level return to schooling is estimated using a sample of residents in 90 cities
across the provinces derived from 1998 Urban Household Survey (see Appendix B). Earlier
samples would be desirable but are, unfortunately, unavailable to support the estimation of return
to schooling across the provinces. Returns to schooling have been rising across Chinese cities
since the 1980s, in the wake of economic reforms and China’s integration into the global
economy (see, e.g. Zhang et. al. 2005). If migrants are forward looking, the 1998 measures of
returns to schooling could be a more appropriate indication of the migrations incentives during
1990-1995 than a backward-looking measure.
11
The share of college graduates in population was too small in 1990.
12
We also include in the wage rate regression differences in average years of schooling
multiplied by urban return to schooling to account for wage rate difference due to differential
average years of schooling.
46
Human Capital Spillovers, Labor Migration and Regional Development in China
13
Most recently, anecdotal evidence suggests substantial reverse migration by unskilled
urban workers to rural villages in the wake of the 2008 global economic crisis and related
slowdown in Chinese economic growth.
Such outflows may similarly reflect the low
opportunity costs of return migration, owing in part to limited accrual of human capital by lowskill migrants to Chinese cities.
47