Education, Human Capital, and Regional Economic Differentials YAO Xianguo & ZHANG Haifeng Zhejiang University, China Dec. 17, 2006.

Education, Human Capital, and

Regional Economic Differentials

YAO Xianguo & ZHANG Haifeng

Zhejiang University, China

Dec. 17, 2006

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Introduction

Regional differentials in China have been one of key issues.

In 1978, GDP per capita are 175RMB and 680RMB in poorest

and richest provinces, respectively. In 2004, GDP per capita in

Shanghai, the richest region, is 55306RMB, while the poorest

Guizhou is about 4125RMB.

As Human Development Report (UNDP, 2005) noted, “If they a

re countries, Guizhou would rank just above Namibia and Shan

ghai alongside Portugal”.

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Introduction

Why does output per capita vary so much across regions in

China? And what policies can be implemented to reduce the

growing gap?

A number of researchers focus on these questions and have

done many helpful investigations on China’s regional

differentials. And education is one of the important factor.

Different from the previous literature, we collect provincial

matched-pair panel data, and attempts to provide some new

empirical evidence for these questions.

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Literature Review

Lin et al. (1998) argue that regional disparities are reflected in d

ifferent development opportunities after analyzing the changes

of regional disparities during the period from 1978 to 1995.

In Démurger’s paper (2001), he notes that differences in infrast

ructure, such as traffic network, electronic power supply and tel

ecommunication equipment, can explain regional disparities in

part.

Several studies focus on the relationship between FDI and regi

onal economic growth (e.g. Wei, 2002; Wang et al., 2002)

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Literature Review

Bao et al. (2002) examine the relationship between geographic locati

on and regional economic growth during China’s transition period.

Like many cross-countries literature, many studies discuss regional e

conomic convergence and its conditions within China.

Although in Lucas’ (and many others) view, human capital is the key

factor of international income differences, there is little consensus on

the magnitude of human capital.

Particularly in China’s case, human capital is always absent from stu

dies until recent years.

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Econometric Strategies

The basic idea is simple and it’s originally from the method which is o

ften used to estimate the return to schooling using twins sample. E.g.,

Ashenfleter & Krueger (1994).

In our case, “twins” are urban and rural sub-regions within the same

province. Compared with the differences between two provinces, the

two sub-economies have many homogeneous background charac-te

ristics, government policies, cultures, geographic location (coastal or

inland), resource endowment, marketization process, and so forth.

So, the two sub-economies, urban and rural, can be considered as a

matched-pair.

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Econometric Strategies

Following Topel (1999, p.2959), consider logarithm of earning e

quation,

where i denotes regional fixed effect (at province level). And ln

hit can usually be rewritten as a linear function of all kinds of hu

man capital.

So, the key econometric issue is whether unobservable i corre

lates with our interested variable.

itithitkiit uhky lnlnln

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Econometric Strategies

As Hall & Jones (1999) noted, differences in institutions and

government policies (he calls them “social infrastructure”) is the

key determinant because these infrastructure determines the

differences in capital accumulation and productivity.

In other words, omitted variable errors may arise if these

regional fixed effects are excluded in simple cross-sectional

regressions.

One method to resolve these biased estimations is using fixed

or random effect estimator.

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Econometric Strategies

Different from existing literatures, we estimate the following equ

ation after first-order difference between urban and rural within

the region.

where Variabledenotes Variableur-Variableru in the same year a

nd region.

Meanwhile, we also estimate the traditional fixed effect estimat

or with interactions to allow parameters to vary between urban

and rural.

ititjjhitshitkit vXSky lnln

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Data summary

Data used in this paper is from China Statistical Yearbook, China Population

Statistical Yearbook, and Fixed Assets Investment Statistical Book: 1950-

2000.

All nominal economic variables are deflated to comparable real variables by

price index (CPI of each sub-regions ).

Like many other studies, average schooling and illiteracy are our main

proxies for human capital.

Unfortunately, other kind of human capital, such as overall life expectancy

and skill level, can not be controlled for in empirical estimations due to data

limitation, thus our estimates may be biased upward.

Other variables are presented in text paper.

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Empirical Results

Table 2 presents the results of first-order difference estimates.

The coefficients of average schooling are about 5%, and

coefficients of illiteracy are -0.4 or so. After controlling for year

effect and excluding municipalities, all these coefficients have a

little rise.

To investigate a comparative analysis, regressions including

interactions are implemented with a comparison of OLS, FE

and RE (see table 3). It shows that education has more

influences on average income in rural areas than in urban

areas.

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Empirical Results

The results in table 4 come from regressions under the assumption

that capital-output ratio is regional fixed effect. (similar to Topel (199

9)) And we get the coefficients of average schooling about 12% in r

ural areas, while 6% in urban areas.

Totally, three conclusions may hold:

human capital measured by average schooling is not as importa

nt as expected to improve average income level.

compared with 1980s, education has increasing impacts on aver

age income level since later 1990s.

education has more influence in rural areas than in urban areas.

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Explanations of Results

Both human capital theory and new growth theory give much

emphasis to the impact of human capital, especially education, on

economic development. If so, our estimation results are

somewhat surprising.

Do our conclusions mean human capital is not important for

China’s economic growth? Otherwise, how to explain it?

We think that the background of transition in China should be

considered together. In fact, some micro-studies also indicate that

returns to education in China is much lower than that in developed

economies, but fortunately, it has an increasing trend with the

reform deepening. (see e.g. Li et al., 2003; Zhang et al., 2005)

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Explanations of Results

The relative lag of factor market reform distorts resource

allocation and incentive mechanism. That is, in Zhou’s view,

owners of human capital may close his/her potential human

capital without any incentive.

Education is the main form of human capital, but not the whole.

In Schultz’s definition, ability to deal with disequilibria may be

the most important especially during China’s transition. For

example, human capital measured by education in Zhejiang

province are lower than many other regions, but there are

numbers of entrepreneurs in Zhejiang.

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Explanations of Results

Human capital externalities are considered as one original drive

factor of economic development, such as in Lucas(1988). And

some studies show there is a threshold effect of education exte

rnalities. That is, the private and social return to education are f

airly low unless average schooling reaches certain level. (e.g. Ir

anzo & Peri, 2006)

As we know, there are yet large numbers of illiteracies and sem

i-illiteracies (about 8.33 ％ of overall illiteracy rate), and overall

average schooling level of Chinese people is approximately 8 y

ears. (Central People’s Government)

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Explanations of Results

It is compatible with characteristics of China’s labor market in terms

of different magnitudes of education on urban and rural income.

In rural labor market, abundant labors’ choices for migration or non-f

arm activities, both of which increase their income, mainly depend on

their own qualities (e.g. education).

While in urban labor market, institutional protection is the main deter

minant of their income, while education has little effects.

Additionally, there are a good many imitation and learning behaviors

in rural area. These behaviors often take place when there are some

able-minded or educated people within the villages.

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Conclusions

Like many other studies, our estimations find that education

has positive effects on regional average income level.

We also find evidence different from previous researches that

the impacts of education decrease a lot after controlling for

other fixed factors.

Particularly, a comparative analysis shows that different

influences of education on average income level, larger

magnitude of effects in rural area than in urban area.

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Thank you very much!

Please send suggestions or criticism to

[email protected]

or

[email protected]

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Empirical Results: Table 2

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Empirical Results : Table 3

Edu

cati

on, H

uman

Cap

ital

, and

Reg

iona

l Eco

nom

ic D

iffe

rent

ials

Empirical Results : Table 4