University of Southern California Dissertations and Theses

The decisions of migration and remittances in rural China

(USC Thesis Other)

The decisions of migration and remittances in rural China

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

THE DECISIONS OF MIGRATION AND REMITTANCES IN RURAL CHINA

by

Qi Qin

____________________________________________________________________

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)

December 2007

Copyright 2007 Qi Qin
ii
Dedication
To my parents,
Shengyang Qin and Taiju Shen,
who support me always with their endless love;
to all my professors,
whose guidance and inspiration are life-long and invaluable assets;
and to the saints in my church as well as my friends,
who have encouraged me and comforted me throughout these years.
iii
Acknowledgements

First, I would like to express my deepest thanks and appreciation to my
academic advisor and dissertation committee chair, Dr. Cheng Hsiao, for his patient
direction and valuable advice throughout the course of this research. Without his
insightful guidance and boundless support, I could not have completed the present
work.
I would like to extend my sincere gratitude to Dr. Yaohui Zhao for
generously providing the data used in this project as well as for her helpful advice on
the topic of the migration in China. The other members of my dissertation
committee, Dr. Yongheng Deng and Dr. Jeff Nugent, gave useful suggestions and
encouragement. I also wish to thank Dr. Guofu Tan, Dr. John Strauss and Dr. Hyeok
Jeong for their help in our discussions. In addition, I have benefited from the
suggestions and encouragement of the professors and doctoral students who have
attended several presentations of this research in the Department of Economics at the
University of Southern California.
Dr. Karen Hennigan at the Center for Research on Crime (formerly SSRI) at
USC has kindly provided me not only various types of material support during my
research, but also generous psychological guidance and warm-hearted help during
my doctoral studies. I also wish to express my earnest thanks to Kathy Kolnick for
her immeasurable support and constant encouragement throughout the whole process
of my research.
iv
I am very thankful to Mary Jeng, Robert Shiao and Yan Shen, whose love
and support have sustained me during my studies. I wish to thank Young Miller, Dr.
Harry DeAngelo and Dr. Linda DeAngelo, as well as all the other professors,
colleagues and staff at USC who have given me kind help and support.
I am grateful to my parents for their love and patience over the years of my
study. Most of all, I would like to express a heart-felt thanks to the Lord for His
unconditional love and empowering supply – “And He has said to me, My grace is
sufficient for you, for My power is perfected in weakness. Most gladly therefore I
will rather boast in my weaknesses that the power of Christ might tabernacle over
me.” (2 Corinthians 12:9)
v
Table of Contents
Dedication ii

Acknowledgements iii

List of Tables vi

Abstract viii

Chapter 1. Introduction 1
Chapter 2. Literature Review 5
2.1. Theory of Migration and Remittances 5
2.2. Empirical Findings 13

Chapter 3. Methods 15
3.1. Individual-level Decision 15
3.2. Household-level Decision 19

Chapter 4. Data 24

Chapter 5. Empirical Results of Migration and Remittances Decisions 33
5.1. Individual-level Tobit Model for Migration and Remittances 33
Decisions
5.2. Household-level Tobit Model for Migration and Remittances 46
Decisions

Chapter 6. The Impact of Migration and Remittances on the Agricultural 52
Productivity
Figure 1. Last shipment of food aid from the World Food 54
Programme (WFP)
6.1. The Single Equation Estimation 60
6.2. The Simultaneous Equation Estimation 66

Chapter 7. Conclusions and Future Research 75

References 77
vi
List of Tables

Table 1. Distribution of individuals regarding to migration and remittances 25
in 1998

Table 2. Distribution of migrants regarding to remittances in 1998 25

Table 3. Characteristics of non-migrants and migrants in our sample 26

Table 4. Characteristics of migrants giving remittances or not in our sample 29

Table 5. Distribution of households regarding to migration and remittances 30
in 1998

Table 6. Distribution of migrant households regarding to remittances 30
in 1998

Table 7. Distribution of the number of migrants in the household with 30
migrants in 1998

Table 8. Characteristics of households with and without migrants in our 31
sample

Table 9. Characteristics of migrant households with and without 32
remittances in our sample

Table 10-1. Estimation of individual-level decisions 35

Table 10-2. Estimation of individual-level decisions comparable to 37
household-level decisions

Table 10-3. Estimation of individual-level decisions (compare three 39
specifications of household income)

Table 11. Remittances’ sensitivity to various explanatory variables (a 43
summary from Rapoport and Docquier (2006))

Table 12. Estimation of household-level decisions 47

Table 13. Average outputs per mu of the five main types of crops 55

Table 14. Agricultural input and output of households with and without 58
migrants
vii
Table 15. Agricultural input and output of households with and without 59
remittances

Table 16-1. Estimation of total agricultural output with the Cobb-Douglas 64
production function

Table 16-2. Estimation of agricultural output per mu with the Cobb-Douglas 65
production function

Table 17-1. Comparison with the result of Rozelle et al. (1999) (1) 70

Table 17-2. Comparison with the result of Rozelle et al. (1999) (2) 73

Table 17-3. Comparison with the result of Rozelle et al. (1999) (3) 74
viii
Abstract
Using recent household survey data from rural China, this dissertation
investigates the decisions of migration and remittances at both individual and
household levels and their impact on agricultural productivity. Previous researchers
wonder if remittance behavior can be predicted by the migrants' characteristics; if
yes, then these two decisions cannot be identified empirically. We find that the study
of remittances is distinct from that of migration, thus deserving separate treatment;
and empirically we are able to identify these two decisions. In the literature on
migration and remittances, some people treat both as individual-level decisions while
taking account of household characteristics; others think that households are the
decision makers instead of individuals. We find that individuals and households
exhibit different behavioral patterns in migration decisions, thus offering stories from
two different angles. Finally, we find that migration with remittances improve the
agricultural productivity in rural China. Remittances from migrants bring in
advanced technology and increasing capital investment to the farm. In addition, the
implicit contract between migrants and their rural households is effective in
improving the laborers’ efforts in agricultural production.

1
Chapter 1. Introduction

At the beginning of their extensive review of the microeconomics of
remittances, Rapoport and Docquier (2006) asked, "Is the study of remittances in
essence distinct from that of migration? More precisely, may remittance behavior,
for the most part, be predicted by the migrants' characteristics, or is there something
beyond that, thus justifying a separated treatment?" If the answer is yes, then it
implies the severity of the identification problem when people try to study these two
decisions jointly in the empirical research. This difficulty in identification results in
little literature in simultaneous-equation study of the decisions of migration and
remittances.
Edward Funkhouser's (1995) comparative study on remittances to the capital
cities of El Salvador and Nicaragua helps to answer this question. The data in this
study revealed that while the number of migrants and the general economic
conditions prevailing in the two countries during the 1980s were quite similar, both
the number of households that received remittances from relatives abroad and the
average amount of those remittances in San Salvador double those for Managua. To
explain this puzzle, two alternative explanations may be suggested: is it that migrants
self-select differently in the two countries? Or is it that, among those who emigrated,
"remitters" self-select differently? The empirical result shows that differences in
remitting behavior could not be accounted for by differences in households' or
migrants' observed characteristics, including the timing of migration. By contrast, the
estimation of remittance functions revealed substantial differences in remitting

2
behavior between the two samples, allowing a conclusion that different motivations
to remit are central to explaining inter-country differences in remittance behavior.
Because the determinants of remittances go beyond the migrant's
characteristics, the decision about remittances needs to be studied separately from
the decision about migration. Since these two decisions are so closely related to each
other, even with the severe identification problem, we should do a simultaneous-
equation study of them.
In the migration literature, some studies have been conducted at the
individual level, others at the household level. We think that individuals and
households have different behavioral patterns. Individuals explicitly participate in
both migration and remitting decisions. A rural laborer first decides whether to
migrate or not; and then if he or she decides to migrate, the individual further decides
whether or not to remit and the level of remittances to send to his or her rural
household. However, households only participate explicitly in the migration
decision. Regarding whether to send out a migrant, the main concern of households
is the possibility that rural households can receive remittances and how much
migrants will send to them. Expected transfer from migrants that rural households
estimate beforehand is the most crucial factor that affects the household's decision
whether or not to invest in a member's migration. Due to the different decision
processes of individuals and households, we will model the decisions at both the
individual level and the household level.

3
The migration of labor from rural areas to urban centers is one of the most
pervasive features of agricultural transformations and economic growth worldwide.
In China, 71 percent of the current population is rural; the migration of rural labor to
urban areas since the mid-1980s has created the largest labor flow in world history.
Despite barriers to labor mobility imposed by China's household registration system
(hukou bu), China has about 50 to 100 million rural-to-urban migrants (Roberts,
1997) annually. Huang and Pieke (2003) report that the number of rural-to-urban
migrants was 45 million in 1997, 55 million in 1998 and 67 million in 1999.
Accompanying this large-scale migration from rural to urban areas,
remittances are sent back by migrants working in urban areas to families in rural
areas. Since financial markets and insurance markets in rural areas of China are at a
primary stage, and in some areas not available at all, remittances from migrants play
an important role in the economic development of rural China.
What are the determinants of China's current migration at such an
unprecedented scale? What motivates migrants to give remittances to their families
in rural areas? What impacts does migration bring to China's rural areas? These are
the questions we will address in this dissertation.
We will review the literature in Chapter 2. Then we suggest our econometric
model of the migration and remittances decisions in Chapter 3. We describe the data
used in this study in Chapter 4 and report our empirical results of the migration and
remittances decisions in Chapter 5. In Chapter 6, we will study the impact of

4
migration and remittances on agricultural productivity. Finally we conclude the
dissertation and discuss future research in Chapter 7.

5
Chapter 2. Literature Review
2.1. Theory of Migration and Remittances
2.1.1. Migration Decision
The causes of internal migration are summarized as follows.
(1) Earnings and employment opportunities
The expected earning gap between rural and urban areas is well established
as the pull factor of migration to urban areas.
(2) Migrant networks
Migrant networks can reduce information costs by providing specific job
information to potential migrants. They can reduce psychological costs by providing
supportive relationships to migrants at destinations. And finally, they can reduce the
probability of unemployment by providing direct job search assistance from fellow
villagers. The existence of a good migrant network in the destination area will
encourage migration.
(3) Distance
Researchers find that migration over short distances is much more common
than migration over long distances. This may be caused by the greater cost of
transportation, the difficulty of obtaining information about more remote
alternatives, and/or less alienation in a nearby setting.

6
(4) Family strategies to contain risks
Unpredictable weather contributes to volatile agricultural income in rural
areas of most developing countries. Given the quasi-absence of credit and insurance
markets, rural households may smooth their consumption by investing in migration
to access outside income source that has little correlation with earnings in home
areas.
(5) Wealth and capital markets
Incomplete or imperfect local capital markets may affect migration in direct
and indirect ways. First, it can directly encourage out-migration through restrictions
on the ability of families to access loans. And it can indirectly push people to migrate
through effects on employment creation.
The opportunity cost of financing costly migrations is more affordable for
wealthier families. This has two important implications. First, given that other
factors are equal, it is more common for relatively better-off households to send out
migrant workers and this in turn may exacerbate inequality in incomes locally.
Second, as a region becomes wealthier, out-migration may actually increase (up to a
point) because the financial constraint of investing in migration is loosened.
(6) Relative economic standing in the community
The relative economic standing in the communities of origin may be a
concern of households in addition to the absolute gain in earnings through migration.
One study of migration from rural Mexico to the United States (not internal
migration) found that individuals with low incomes relative to others in their village

7
were more likely to migrate. However, it would be hasty to generalize from this
evidence.
(7) Availability and quality of amenities.
Improved amenities in a location may imply better future development of
both agricultural and industrial sectors. Thus out-migration may decrease. However,
some forms of improved local amenities could exacerbate net out-migration. For
example, improved rural transport could encourage out-migration.

2.1.2. Remittances Decision
The theories explaining the behavior of remittances are summarized as
follows.
(1) Altruism
The most common motivation to remit is that migrants care about those left
behind. However, this altruistic inclination to remit was more frequently assumed
than tested against competing theories. Funkhouser (1995) proposed a behavioral
model of remittances based on altruism and provided the following testable
implications:
(a) migrants who have higher earnings potential remit more;
(b) low-income households receive more than high-income households;
(c) the strength of family ties between the migrant and the remaining
household members as well as the migrant's intentions to return have positive
effects on remittances;

8
(d) as the number of other migrants from the same household increases,
remittances by a given migrant should decrease; and
(e) the time profile of remittances should depend on a comparison between
the migrants' time-discount factor and their earnings profile outside of home
community.
As the discussion below will show, these predictions are compatible with a
number of other possible motives. Stark (1991, Chapter 1) modeled the altruism
motive and found that the main testable implication of the altruistic model against
other motives is that transfers cannot increase with the income of rural households.
Another interesting prediction of the pure altruism hypothesis is that an increase by
one dollar in the income of the migrant, coupled with a one-dollar drop in the
recipient household's income, should raise the amount transferred exactly by one
dollar.
(2) Exchange
The most natural way to think of pareto-improving exchanges involving
remittances is to assume that remittances simply "buy" various types of services such
as care of the migrant's assets (e.g., land, cattle) or relatives (children, elderly
parents) at home. Such motivations are generally the sign of temporary migration,
and signal the migrants' intention to return.
Cox (1987) modeled the case of non-altruistic agents only and a fixed amount
of service. He shows that the amount transferred increases with the quantity of
service to be offered but reacts ambiguously to an exogenous increase in the

9
recipient's pre-transfer income. Thus in contrast to the altruistic model, the exchange
model predicts that an increase in the recipient's income may raise the amount
transferred.
The first two models above considered migration and remittances as
individual decisions. In a context of imperfect capital and insurance markets in rural
areas of most developing countries, remittances may be part of an implicit migration
contract between the migrant and his or her family, allowing the household to
achieve higher income and/or smooth consumption. This kind of contract will also
raise the problem of informational asymmetry. For instance, migrants have
disadvantages in getting information about the economic conditions at home. Since
remittances provide those left behind with an insurance against bad economic times,
such informational asymmetries may also give rise to moral hazard.
(3) Insurance and moral hazard
Weather is an important input of agricultural production. It is stochastic, and
its realization during the course of production is unpredictable and exogenous. This
implies the volatility of agricultural income. Urban and foreign jobs are generally
subject to risks uncorrelated with those impeding on agricultural activities at home
(e.g., crop failure, cattle disease, etc.). Hence, in the quasi-absence of credit and
insurance markets, rural households will react to this income volatility by allocating
some members to employment outside of agriculture, via urban or foreign migration,
so that migrants can insure the remaining members of the family against drops in
rural incomes. To operate, however, such pareto-improving arrangements must also

10
be self-enforcing. This is generally achieved because of a sufficient degree of
altruism within the family. Furthermore, households can ultimately sanction
opportunistic behavior using a variety of retaliation strategies. Alongside reputation
(loss of prestige), or ostracism, default to remit may also be sanctioned by denying
the migrant rights to inheritance or return to the village for retirement, an option that
most migrants want to keep open. This also implies that, other things equal, better-
off families that can monitor the migrants' behavior through inheritance procedures
would tend to rely on migration more than poor families (Hoddinott, 1994). We
discuss the role of inheritance in more details in (4) below.
Insurance and altruistic motives could result in a different prediction with
respect to the sign of the effects of pre-transfer income on remittances. Although
purely altruistic models predict higher remittances to lower-income households, the
bargaining model of insurance could imply the opposite because in the bargaining-
cum-exchange model, greater familial wealth increases the family's bargaining
power.
In addition, the number of migrants within a given household may also
provide a basis for discrimination between altruism and insurance. As pointed out by
Agarwal and Horowitz (2002), the number of migrants is expected to reduce
altruistic transfers by any particular migrant as all sources of transfers (whether
public or private) are perfect substitutes under the pure altruistic hypothesis. By
contrast, if each migrant individually subscribes to an insurance contract, no such
negative effect is expected.

11
In the discussion above, we assume the exogeneity of the recipients' income.
However, as agents become insured against risks, they may reduce their level of
effort (moral hazard). At an empirical level, some studies provide evidence of
opportunistic behavior on the recipients' side (e.g., Azam and Gubert (2002)).
(4) Family loan arrangements: the investment motive
The same kind of rationale may be used to explain remittances as repayments
of loans on investments in education and/or migration. In this case, the purpose of
migration is to increase household income rather than to reduce the uncertainty of
income.
If investments are the underlying familial motivation for sending migrants
away, this implies that the household will keep on sending migrants as long as the
total family income is thereby increased. Rapoport and Docquier (2006) modeled the
investment motive and they find that the effect of familial wealth on the number of
migrants is theoretically unclear. On one hand, migration incentives would seem to
be greater for members of poor families, since the wage differential is larger for poor
families. On the other hand, given the liquidity constraints of financing costly
migration activities, relatively wealthier families are more likely to take advantage of
such investment opportunities.
(5) Inheritance as an enforcement device
What enforcement device can rural households that invest in migration
implement to make sure that they will receive compensating transfers from migrants
in the future? Two basic mechanisms generally serve as enforcement devices to

12
make family arrangements incentive-compatible: punishment and social norms. At
the household level, the most powerful threat that can be used to secure remittances
is to deprive the migrants of their rights to inheritance and/or return.
Assuming that there is a benchmark, such as a minimum amount of money
that each migrant is expected to remit, Hoddinott (1994) argues that parents can
encourage transfers above this benchmark level by offering a "reward" of the future-
inheritable asset. Recently, de la Brière et al. (2002) summarized the main
predictions of this inheritance motive (which they called the investment hypothesis)
as follows: the amount of remittances increases with (a) the remaining household's
wealth, (b) the migrant's earning potential, (c) the probability of inheriting (which
depends on the age of the parents, the number of siblings, etc.), and decreases with
(d) the degree of risk aversion, given that inheritance is more risky than other
available forms of savings.
(6) Mixed motives
In reality, a combination of different motives lies behind the behavior of
remitting. Researchers agree that discriminative tests between these models are not
always available. First, different individuals may be heterogeneous in their
motivations to remit; and even the same individual can have multiple motivations to
remit. For instance, informal familial arrangements may take account of both
investment and insurance, and the enforcement of such informal contracts may
depend on loyalty and trustworthiness, which is to some degree related to altruism.
Such complex interdependencies have long been recognized in the empirical

13
literature (e.g., Lucas and Stark, 1985) and, more recently, in the theoretical
literature.

2.2. Empirical Findings
Recently there have been some papers that studied the question of what
factors contribute to the decision to migrate in China (Rozelle et al. 1999, Zhao
1999a, b; 2001a, b, Yao 2001). They conclude that:
(1) Young men with a middle school education who come from families with
more laborers and less land are more likely to migrate.
(2) The availability of nonfarm employment in the local area reduces the
probability of migration.
(3) The number of experienced migrants is found to have significant and
positive effects on subsequent migration from a local area.
However, with the exception of Rozelle et al., all of these papers ignore the
importance of remittances in the migration decision. The expected remittance is a
main determinant of migration. There is a big income gap between China's urban and
(relatively poor) rural areas. In addition, people in the rural areas face extremely
immature financial and insurance markets, which obstruct the development and
modernization of their rural areas. So the remittances from the migrants can help
rural households to overcome stringent financial constraints.
There are few papers studying the phenomenon of remittances and the
impacts of migration on Chinese rural areas. Rozelle et al. (1999) studied the

14
relationship between migration, remittances and agricultural productivity in China.
They used data from a survey of 787 farm households from 31 villages in two
provinces in the northeast part of China, conducted by one of the authors (Rozelle) in
the summer of 1995. Their conclusion is that migration has a negative direct
influence on maize production; remittances have positive direct effects on maize
production; migration also affects remittances positively. In total, the net impact of
migration on maize production is negative.
We find that the econometric model they construct is improper. First, they
ignore the discrete character of the decision function of the number of migrants in
the household. Second, remittances that the household received from migrants is zero
when there is no migrant, and remittances can be zero even when the household
sends out migrants, which implies another selection function should be added. We
take account of the selection issue and the censoring character of the data in our
model. In addition, we use new data covering more areas of China and offering more
detailed information about the household and the village than the data used in
Rozelle et al. (1999). The data also includes rich information in the production of all
main crops instead of just maize. Thus we can construct a more precise indicator of
agricultural productivity.

15
Chapter 3. Methods
3.1. Individual-level Decision
The model for individuals includes a migration decision at the beginning; and
then if the individual decides to migrate, he or she further decides how many
remittances to send to his or her rural household.
The two variables of decisions about migration and remittances, are
) , (
* *
i i
R d
.
The underlying response variable of the migration decision,
*
i
d
, and remittances,
*
i
R
,
are defined by the regression relationship
i i i
X d µ ! + =
*
, (1)
i i i
Z R ! " + =
*
, (2)
where X is a vector of factors contributing to the migration decision of the
individual; and Z is a vector of factors determining the amount of remittances.
The error term of the migration equation at the individual level,
µ
, includes
unobserved characteristics of the individual that are related to migration, such as the
degree of altruism (individuals have some concern for increasing the welfare of the
whole rural family in the decision of migration), the ability to pursue agricultural
and/or non-agricultural production in rural areas, the (expected) earning ability in
urban areas, the degree of personal connection/working network in urban areas
(relationship with existing migrants and/or relatives in urban areas), etc.
The error term of remittances equation at the individual level,
!
, includes
unobserved characteristics of the individual that are related to the decision of sending

16
remittances such as the degree of altruism, the degree of the constraining power of
the implicit family contract concerning remittances, etc.
The error terms of migration and remittances equations at the individual level
are correlated at least in the two aspects below:
(1) the motivation of pure altruism;
(2) the various types of abilities of the individual can be observed much
better within the rural household, although they remain unobserved from the
data that we can obtain. Hence, the implicit family contract concerning how
many remittances are expected from the migrant will take into account the
individual’s abilities observed only by the household.
Assume that the joint distribution of the two error terms of the two equations,
i
u
and
i
!
, is bivariate normal distribution with zero mean and covariance matrix
!
!
"
#
$
$
%
&
2 2
2 2
' µ'
µ' µ
( (
( (
.
In the data observed, we only have
) , (
i i
R d
. In practice,
*
i
d
is unobservable.
What we observe is a dummy variable,
i
d
, defined by
1 =
i
d
if
0
*
>
i
d
(migrate)
0 =
i
d
if
0
*
!
i
d
(not migrate)
And the observed remittances,
i
R
, is a censored variable that is either zero or
positive.

17
So we have
*
i i
R R =
if
0
*
>
i
R
(send remittances)
0 =
i
R
if
0
*
!
i
R
(not send remittances)
We use MLE to estimate the model. The probability functions of the three
cases (not migrate, migrate without remittances, and migrate with positive
remittances) are discussed below.

Case 1: not migrate
) ( ) 0 ( Pr
µ
!
"
i
i
X
d ob
#
$ = =
, (3)
where ! is the cumulative function of standard normal distribution.

Case 2: migrate without remittances
! ! ! !
+"
#
#
" #
+"
#
#
" #
= = $ =
%
&
%
&
µ µ ' µ ' ' µ ' µ
X
Z
X
Z
i i
d f d f d d f R d ob ) ( ) | ( ) , ( ) 0 , 1 ( Pr
*
(4)
) , ( ~ ) | (
2
4
2
2
µ
µ!
!
µ
µ!
"
"
" µ
"
"
µ ! # N f
let
2
4
2 2
|
µ
µ!
! µ !
"
"
" " # = ,
) ( ) | (
|
2
µ !
µ
µ!
"
#
µ
#
#
$
! µ !
% %
& =
'
%
( %
Z
d f
Z
(5)

18
Substitute (5) into (4), we have
µ
!
µ
!
!
"
# !
µ $
µ
µ$
µ
!
%
µ
µ
d
Z
e R d ob
X
i i
) (
2
1
) 0 , 1 ( Pr
|
2
)
2
1
(
*
2
2
& &
' = ( =
&
) +
&
*
(6)

Case 3: migrate with remittances
The density function is
) 1 | ( ) 1 ( Pr
|
= ! " =
i i
d Z R f d ob #
µ $
(7)
) (
) 1 ( Pr
) , | ( Pr
) 1 | (
|
!
! " # µ
!
" µ "
Z R f
d ob
Z R X X ob
d Z R f
i
i
$ %
=
$ = $ >
= = $
(8)
Substitute (8) into (7), we have the new density function as
) ( ) , | ( Pr ! ! " # µ
"
Z R f Z R X X ob $ % $ = $ >
(9)
!
+"
#
= # = # >
$
µ % µ & % $ µ
X
d f Z R X X ob ) | ( ) , | ( Pr
(10)
) , ( ~ ) | (
2
4
2
2
!
µ!
µ
!
µ!
"
"
" !
"
"
! µ # N f
let
2
4
2 2
|
!
µ!
µ ! µ
"
"
" " # = ,
µ
! "
# $ % µ
$
"
"
µ
"
%
$ µ
$
µ$
µ
d e Z R X X ob
v
X
] ) (
2
1
[
|
2
2 2
|
2
1
) , | ( Pr
& &
' +
&
(
= & = & >

)
) (
( 1
|
2
! µ
!
µ!
"
!
"
"
# $ $
% $ =
X
(11)

19
Plug (11) into (9), we get the final density function as
)
2
1
(
|
2
|
2
2
2
2
1
)]
) (
( 1 [ ) ( )]
) (
( 1 [
!
"
! ! µ
µ!
!
! µ
!
µ!
!
# " "
"
"
$
%
"
!
"
"
$
&
'
& &
( & = & '
& &
( & e
v X
Z R f
X
v
(12)

The likelihood function for the model of individual-level decisions is
µ
!
µ
!
!
"
# ! !
$
µ %
µ
µ%
µ
!
$
µ µ
µ
d
Z
e
X
L
X
R d d
) (
2
1
) (
|
2
)
2
1
(
0 , 1 1
2
2
*
& &
' (
&
' =
&
) +
&
* = =
+ , ,

)
2
1
(
|
2
0 , 1
2
2
*
2
1
)]
) (
( 1 [
!
"
! ! µ
!
µ!
!
# " "
"
"
$
%
> =
&
% %
' % &
(
e
v X
R d
(13)

3.2. Household-level Decision
For households, they only explicitly participate in the migration decision.
However, expected remittances are the main factor that contributes to the decision of
rural households on whether they invest in migration.
The main difference of the econometric model for households from that for
individuals is that we should estimate a reduced-form model first, while handing
expected remittances; then we estimate the coefficients of the structural-form model
from the coefficients of the reduced-form model.
Again, the underlying response variable of migration decision,
*
i
d
, and
expected remittances,
*
i
R
, are defined by the regression relationship

20
i i i i
R X d ! " # + + =
* *
, (14)
i i i
Z R ! " + =
*
, (15)
where X is a vector of factors contributing to the migration decision of the
household; Z is a vector of the factors that households use to estimate the expected
remittances.
Substituting (15) into (14) yields
i i i
u W d + = !
*
, (16)
where W includes factors that affect households' expectation of remittances and other
determinants of migration; and
! " #
i i i
u + =
. (17)
There are variables that affect both migration and expected remittances. Thus
we first estimate a reduced-form model. To identify the simultaneous-equation
model, we need at least one variable that is not included in both X and Z. The social
norm of remitting in the village is a variable that enters into the equation of expected
remittances but not in the equation of migration directly. So we can identify the
model.
We can rewrite the structural-form model as
i i i i i i i
R X X u W d ! " # # $ + + + = + =
* 2 2 1 1 *

i i i i i i
X Z Z R ! " " ! " + + = + =
2 2 1 1 *

where
1
X represents the variables that explain migration but not expected
remittances;
1
Z includes the variables that explains expected remittances but not

21
migration;
2
X represents the variables that explains both migration and expected
remittances. Then the reduced-form model becomes
i i i i i
X Z X d µ ! " # ! " # + + + + = ) ( ) (
2 2 2 1 1 1 1 *
(18)
i i i i
X Z R ! " " + + =
2 2 1 1 *
(19)
Similar to the error terms at the individual level, the error term of the
migration equation at the household level,
µ
, includes the unobserved
characteristics of the household that is related to migration, such as the expected
degree of altruism among the household members, the capacity of pursuing
agricultural and/or non-agricultural production in rural areas as a household, the
degree of the family connection in urban areas and/or that with other rural
households with migrants, etc.
The error term of the remittances equation at the household level,
!
, includes
unobserved characteristics of the household that help to form the expectation of
remittances such as the expected degree of altruism among the household members,
the degree of the constraining power of the implicit family contract concerning
remittances, etc.
With the reduced form model, the error terms of migration and remittances
equations at the household level are correlated by the definition in Equation (17).
Assume that the joint distribution of the two error terms of the two equations,
i
u
and
i
!
, is bivariate normal distribution with zero mean and covariance matrix

22
!
!
"
#
$
$
%
&
2 2
2 2
' µ'
µ' µ
( (
( (
.
Let
! " #
1 1
=
! " # $
2 2 2
+ =
Then
i i i i i
X Z X d µ ! ! " + + + =
2 2 1 1 1 1 *
(20)
In practice,
*
i
d
is unobservable. What we observe is a dummy variable,
i
d
, defined
by
1 =
i
d
if
0
*
>
i
d
(send migrant(s) to urban area)
0 =
i
d
if
0
*
!
i
d
(not send any migrant to urban area)
We are also unable to get data of expected remittances. We assume that rural
households have rational expectations and then use the actual remittances received
by households to approximate the expected remittances. The observed remittances,
i
R
, is a censored variable that is either zero or positive.
So we have
*
i i
R R =
if
0
*
>
i
R
(expect/receive positive remittances from migrants)
0 =
i
R
if
0
*
!
i
R
(expect/receive no remittances)

23
We can estimate the reduced-form model (Equations (20) and (19)) by MLE.
The likelihood function for the model of household-level decisions is similar to that
of individual-level decisions, except that we should replace
! X
by
! W
.
We then estimate the coefficients of the structural-form model
2
! and !
from the coefficients of the reduced-form model by applying OLS to the following
equation:
!
"
#
$
%
&
+
!
"
#
$
%
&
!
!
"
#
$
$
%
&
=
!
!
"
#
$
$
%
&
'
'
'
'
2
1
2
2
1
2
1
0
(
(
)
*
+
+
,
,
I
(21)
where ) , , , (
2 1 2 1
! ! ! !
" " # # are MLE estimators of ) , , , (
2 1 2 1
! ! " " ; I is an identity matrix;
) , (
2 1
! ! are error terms.

24
Chapter 4. Data
Data used in this study came from a Chinese Ministry of Agriculture rural
household survey conducted in August and September of 1999 in six provinces:
Hebei, Shannxi, Annui, Hunan, Sichuan and Zhejiang. Two counties in each
province, one township in each county, one administrative village in each township
and three natural villages in each village were chosen for the survey. In each of these
36 villages, 20 to 30 households were randomly chosen for interviews. A total of 824
households were interviewed.
Household interviews collected data at both the individual and the household
levels.
1
Individual information includes basic personal characteristics such as gender,
age, education, etc. At the household level, information was collected on household
income, wealth, inputs and outputs of agricultural production, and many other
aspects of household economic activities. A separate village survey was conducted in
sample villages to gather information on community characteristics.
Migrants were identified from the household survey as household members
who had urban work experience during 1998. Since the range of their age is from 16
to 66, we limit our individual-level study on those whose age falls in this range.
From Table 1, we find that 13 percent of individuals worked outside the rural area
surveyed in 1998. Table 2 shows that about two out of three migrants sent
remittances home in 1998. Table 3 summarizes individual characteristics of all

1
Interviews were conducted in the homes of the households. In most cases, the head of the household
answered questions. When the household head was absent, another adult member was interviewed. A
household was usually skipped if there was no adult present.

25
observations, non-migrants, and migrants in our sample. We find that the young
single male is more inclined to migrate. Migrants are more educated than non-
migrants in the whole sample and in their own households. The households of
migrants have more laborers (age 16-59), fewer young dependents (age 0-15), and
slightly fewer elderly dependents (age 60 and above). Migrant households are also
wealthier than non-migrant households, on average. The villages of the migrants'
households are closer to the nearest town and have more total migrants.

Table 1. Distribution of individuals regarding to migration and remittances in
1998

All
No
migrants
Migrants Remittances
Number of individuals
(age 16-66)
2579 2234 345 224
Percentage (%) 100 86.62 13.38 8.69
Source: survey.

Table 2. Distribution of migrants regarding to remittances in 1998

All No Remittances Remittances
Number of individuals (age
16-66)
345 121 224
Percentage (%) 100 35.07 64.93
Source: survey.

26

Table 3. Characteristics of non-migrants and migrants in our sample

Non-
migrants
All Migrants
Individual characteristics
Male (%) 48.93 50.91 63.77
Married (%) 77.62 74.53 54.49
Age (years) 38.8165 37.4416 28.5392
Years of education 5.8975 6.0884 7.3246
Difference from household average years
of education
-0.1534 0 0.9930
Household characteristics
# Laborers 3.3344 3.3870 3.7275
# Young dependents 0.8044 0.7922 0.7130
# Elderly 0.4064 0.4044 0.3913
Total asset 41754.05 42252.43 45479.59
Average asset 9368.11 9475.72 10172.55
Village characteristics
Distance between the village and the
closest town
4.2122 4.1256 3.5652
Village percentage income from remittance 0.1510 0.2060 0.5621
# Total other migrants in the village 44.0546 48.5417 77.5971
Source: survey

27
Table 4 lists characteristics of remitters and non-remitters among migrants.
Married male migrants are more inclined to remit than single men. Remitters are
older and less educated than non-remitters. The households of remitters have fewer
laborers and more young dependents. An interesting feature of the households of
remitters is that not only are they less wealthy than non-remitters, but also their total
and average assets are below the average level for the whole sample that includes
both migrants and non-migrants. Comparing the characteristics of migrants and
remitters, we find that they have different personal and household characters. This
implies that the decisions about migration and remittances do not follow the same
rationale.
After deleting 4 observations with many missing values, we use the
remaining 820 households to do our empirical study at the household level. From
Table 5, we can see that 255 households (31%) sent at least one household member
into the migrant labor force. Table 6 shows that among those migrant households, 72
percent received remittances in 1998. Table 7 describes the distribution of the
number of migrants in households with migrants. About 72 percent of migrant
households sent only one migrant and about 24 percent of migrant households sent
two persons to work outside the rural region. The remaining 4 percent sent three to
five migrants.
Table 8 describes the sample of non-migrant and migrant households. Two
types of descriptive statistics are presented: household and village characteristics.
Looking at household characteristics, we can see that households participating in

28
migration have more laborers, fewer young dependents, and slightly fewer elderly
dependents. They also have higher average years of education and higher years of
education received by the most educated person among household members. They
own more total assets and average assets. There are also some noticeable differences
in village characteristics. The total number of migrants in other households in
villages of the migrants' households, on average, is much greater than that in villages
of the non-migrant' households. The income of villages that migrant households
indwell is more dependent on remittances as a source of total revenue.
Table 9 describes the sample of migrant households with and without
remittances using the same descriptive statistics as Table 4. Households with
remittances in 1998, on average, received 4880.33 yuan from migrants. From
household characteristics, we can see that households receiving remittances actually
send fewer laborers to work outside the village than those migrant households not
receiving remittances. Households receiving remittances are weaker in the conditions
of education and wealth than those not receiving remittances.

29

Table 4. Characteristics of migrants giving remittances or not in our sample

No
Remittances
Remittances All
Individual characteristics
Male (%) 54.92 68.61 50.91
Married (%) 49.18 57.40 74.53
Age (years) 27.2705 29.2332 37.4416
Years of education 7.5984 7.1749 6.0884
Difference from household average
years of education
1.2391 0.8583 0
Household characteristics
# Laborers 4.2049 3.4664 3.3870
# Young dependents 0.6639 0.7399 0.7922
# Elderly 0.4098 0.3812 0.4044
Total asset 64326.25 35168.87 42252.43
Average asset 14273.09 7929.20 9475.72
Village characteristics
Distance between the village and the
closest town
3.2213115 3.7533632 4.1256
Village percentage income from
remittance
0.1592622 0.7825 0.2060
# Total other migrants in the village 98.4754098 66.1749 48.5417
Source: survey

30

Table 5. Distribution of households regarding to migration and remittances in
1998

All No migrants Migrants Remittances
Number of households 820 565 255 184
Percentage (%) 100 68.90 31.10 22.44
Source: survey.

Table 6. Distribution of migrant households regarding to remittances in 1998

All No Remittances Remittances
Number of households 255 72 183
Percentage (%) 100 28.24 71.76
Source: survey.

Table 7. Distribution of the number of migrants in the household with migrants in
1998

All
1
migrant
2
migrants
3
migrants
4
migrants
5
migrants
Number of
households
255 183 60 8 2 2
Percentage (%) 100 71.76 23.53 3.14 0.78 0.78
Source: survey.

31

Table 8. Characteristics of households with and without migrants in our sample

Without
Migrants
All
Households
With
Migrants
Household characteristics
# Migrants 0 0.4207 1.3529
# Laborers 2.7009 2.9561 3.5216
# Young dependents 0.9186 0.8573 0.7216
# Elderly 0.4230 0.4085 0.3765
Average years of education 5.5964 5.7489 6.0866
Maximum years of education
among household members
8.0478 8.3012 8.8627
Total asset 38217.18 39500.14 42342.79
Average asset 9538.22 9541.33 9548.23
Village characteristics
Village percentage income from
remittance
0.0097 0.1812 0.5612
# Total other migrants in the
village
36.3027 46.2720 68.3608
Source: survey

32

Table 9. Characteristics of migrant households with and without remittances in
our sample

No
Remittances
Remittances
All
Households
Household characteristics
Remittances 0 4880.33 1091.22
# Migrants 1.4167 1.3279 0.4207
# Laborers 3.8889 3.3770 2.9561
# Young dependents 0.5972 0.7705 0.8573
# Elderly 0.4167 0.3607 0.4085
Average years of education 6.1660 6.0553 5.7489
Maximum years of education
among household members
9.1806 8.7377 8.3012
Total asset 60942.74 35024.78 39500.14
Average asset 13496.74 7994.72 9541.33
Village characteristics
Village percentage income from
remittance
0.0091 0.7784 0.1812
# Total other migrants in the
village
78.5833 64.3388 46.2720
Source: survey

33
Chapter 5. Empirical Results of Migration and Remittances
Decisions

5.1. Individual-level Tobit Model for Migration and Remittances
Decisions

5.1.1. Migration Decision at the Individual Level
In total, we find that gender, marital status, age and age square, the difference
of average household assets from the village mean, the total number of other
migrants in the village, and the distance between the village and the closest town are
significant determinants of migration (see Tables 10-1, 10-2 and 10-3).
Males are more inclined to migrate than females. This is consistent with the
tradition in rural China that men go out for work and women stay at home taking
care of children and the elderly. More importantly, rural households will more likely
invest more on male children's education. On average, males receive 2.29 more years
of education than females. And the variables measuring education years are
significantly negatively correlated to gender (male = 0, female = 1). Given that
education is a type of human capital, this implies that men will have a better chance
of getting jobs and will earn more in the destination area. This also explains why the
variables measuring education are not significant in our regression and hence are
omitted.
Single persons are more likely to migrate than married people. Not all
married migrants are able to bring their spouses and children to urban areas where

34
they are employed. So the psychological cost of migration for married migrants is
larger.
The age of migrants in our sample ranges from 16 to 66. So our study of the
migration decision focuses on rural people whose age is in this range. We find that
the propensity for migration has an inverse U-shape on age. The maximum point
estimated of all regressions is about age 29. This seems reasonable because most of
the jobs that rural migrants get in urban areas are labor-intensive; young and
energetic rural laborers will more readily be employed.
Migrants are more possibly from relatively wealthier families than from
poorer families. Due to the cost of financing the migrant, better off households are
more able to invest on migration.
“Migrant networks are sets of interpersonal ties that connect migrants, former
migrants, and non-migrants in origin and destination areas through ties of kinship,
friendship, and shared community origin” (Massey et al. 1993, p. 448). Migrant
networks can reduce information costs by providing specific job information to
potential migrants, can reduce psychological costs by providing supportive
relationships to migrants in destinations, and can reduce the possibility of
unemployment by providing direct job search assistance from fellow villagers. Our
results confirm that the migrant network has a significant positive impact on
migration. Households in a village with more migrants will be more inclined to send
family members to work outside the rural area.

35
Table 10-1. Estimation of individual-level decisions

Only variables significant All variables for discriminating motives of remittances All variables
Variable Migration Remittances Migration Remittances Migration Remittances
Intercept
-2.959
***

(-6.131)
-11937.9
***

(-12.349)
-2.950
***

(-5.690)
-10545.9
**

(-6.011)
-2.921
***

(-5.224)
-7922.5
(-1.410)
Gender
-0.424
***

(-5.589)

-0.430
***

(-5.414)

-0.454
***

(-5.577)
-434.572
(-0.682)
Marital
-0.305
**

(-2.589)

-0.3110
**

(-2.569)

-0.290
*

(-2.226)
518.389
(0.526)
Age
0.128
***

(4.749)

0.129
***

(4.747)

0.130
***

(4.212)
-29.740
(-0.094)
Age Square
-0.00223
***

(-6.665)

-0.00224
***

(-6.663)

-0.00227
***

(-5.654)
0.267
(0.058)
Education Years
-0.00209
(-0.138)
-126.205
(-1.245)
-0.00223
(-0.146)
-120.086
(-1.061)
Number of Migrants
-81.691
(-0.047)

-1239.658
(-0.673)
Square of Number of Migrants
-54.739
(-0.127)

286.295
(0.634)
Number of Young Dependents
-0.103
*

(-2.067)

-0.104
*

(-2.060)

-0.108
*

(-2.011)
-148.321
(-0.407)
Number of Elderly Dependents
0.139
*

(2.250)

0.137
*

(2.194)
0.127
*
(1.962)
-344.946
(-0.784)

36
Table 10-1, continued

Only variables significant
All variables for discriminating motives of
remittances
All variables
Variable Migration Remittances Migration Remittances Migration Remittances
Number of Other Laborers
0.0537
†

(1.699)

0.0514
†

(1.596)

0.0413
(1.228)
-379.782
(-1.431)
Total Number of Other Migrants in the
Village
0.00831
***

(9.282)

0.00828
***

(9.195)

0.00832
***

(9.095)

Distance to the Closest Town
-0.0499
**

(-2.885)

-0.0504
**

(-2.900)

-0.0505
**

(-2.892)

Difference of Average Household Asset
from Village Mean
0.00000804
***

(3.460)
-0.130
***

(-8.142)
0.00000799
***

(3.399)
-0.137
***

(-7.809)
0.00000786
***

(3.339)
-0.145
***

(-6.897)
Household Income excluding
Remittances in 1998

-0.0754
***

(-3.312)

-0.0633
*

(-2.553)

-0.0520
(-1.904)
Migration Income in 1998
0.959
***

(82.572)

0.953
***

(80.609)

0.953
***

(71.886)
Village-Average Portion of Remittances
in Total Income

3294.323
***

(11.495)

3273.347
***

(11.482)

3301.539
***

(10.609)
Sigma(1)
4104.306
***

(14.854)

3992.540
***

(15.020)

3911.124
***

(11.193)
Rho(1,2)
0.620
***

(6.305)

0.586
***

(5.263)

0.556
***

(3.283)
Number in parentheses is t-statistics;
***
Coefficient different from zero at 0.001 significance level;
**
Coefficient different from zero at 0.01
significance level;
*
Coefficient different from zero at 0.05 significance level;
†
Coefficient different from zero at 0.1significance level; Note:
Provincial dummies are included but not reported here.

37
Table 10-2. Estimation of individual-level decisions comparable to household-level decisions

Only variables significant All variables for discriminating motives of remittances All variables
Variable Migration Remittances Migration Remittances Migration Remittances
Intercept
-2.978
***

(-6.185)
-12379.4
***

(-13.367)
-2.953
***

(-5.708)
-10737.2
***

(-6.278)
-2.877
***

(-5.161)
-4849.7
(-0.871)
Gender
-0.418
***

(-5.509)

-0.427
***

(-5.382)

-0.456
***

(-5.594)
-516.416
(-0.818)
Marital
-0.306
**

(-2.610)

-0.311
**

(-2.569)

-0.290
*

(-2.229)
788.659
(0.798)
Age
0.127
***

(4.722)

0.128
***

(4.718)

0.128
***

(4.152)
-146.709
(-0.469)
Age Square
-0.00221
***

(-6.619)

-0.00223
***

(-6.615)

-0.00225
***

(-5.594)
1.799
(0.396)
Education Years
-0.00218
(-0.145)
-133.752
(-1.227)
-0.00227
(-0.149)
-126.465
(-1.120)
Number of Migrants
81.665
(0.048)

-1164.945
(-0.645)
Square of Number of Migrants
-137.632
(-0.335)

276.610
(0.633)
Number of Young Dependents
-0.104
*

(-2.071)

-0.106
*

(-2.061)

-0.118
*

(-2.207)
-65.006
(-0.174)
Number of Elderly Dependents
0.138
*

(2.208)

0.134
*

(2.126)

0.129
*

(1.995)
-281.358
(-0.643)

38
Table 10-2, continued

Only variables significant
All variables for discriminating motives of
remittances
All variables
Variable Migration Remittances Migration Remittances Migration Remittances
Number of Other Laborers
0.0635
*

(2.027)

0.0573
†

(1.782)

0.0357
(1.057)
-618.068
*

(-2.455)
Total Number of Other Migrants in the
Village
0.00829
***

(9.249)

0.00825
***

(9.158)

0.00833
***

(9.027)

Distance to the Closest Town
-0.0503
**

(-2.899)

-0.0506
**

(-2.898)

-0.0483
**

(-2.722)

Difference of Average Household Asset
from Village Mean
0.00000811
***

(3.489)
-0.128
***

(-6.901)
0.00000802
***

(3.409)
-0.135
***

(-6.892)
0.00000771
***

(3.326)
-0.133
***

(-5.937)
Nonagricultural Income per Adult
(Excluding Remittances)

-0.230
**

(-2593)

-0.193
*

(-2.165)

-0.246
(-2.579)
Agricultural Income per Adult
-0.327
(-1.032)

-0.331
(-0.970)

-0.431
†

(-1.165)
Migration Income in 1998
0.958
***

(82.048)

0.953
***

(80.055)

0.957
***

(70.869)
Village-Average Portion of Remittances
in Total Income

3326.678
***

(11.480)

3292.987
***

(11.458)

3155.392
***

(10.051)
Sigma(1)
4179.736
***

(14.336)

4016.896
***

(14.649)

3785.218
***

(11.734)
Rho(1,2)
0.642
***

(6.692)

0.596
***

(5.300)

0.515
***

(2.809)
Number in parentheses is t-statistics.
***
Coefficient different from zero at 0.001 significance level;
**
Coefficient different from zero at 0.01 significance level;
*
Coefficient different from zero at 0.05 significance level;
†
Coefficient different from zero at 0.1significance level; Note: Provincial dummies are included but not reported here.

39
Table 10-3. Estimation of individual-level decisions (compare three specifications of household income)

Total Income
Agricultural and Nonagricultural Income
(Excluding Remittances)
Agricultural and Nonagricultural Income per
Adult (Excluding Remittances)
Variable Migration Remittances Migration Remittances Migration Remittances
Intercept
-2.959
***

(-6.131)
-11937.9
***

(-12.349)
-2.966
***

(-6.157)
-12037.3
***

(-12.516)
-2.978
***

(-6.185)
-12379.4
***

(-13.367)
Gender
-0.424
***

(-5.589)

-0.423
***

(-5.557)

-0.418
***

(-5.509)

Marital
-0.305
**

(-2.589)

-0.301
**

(-2.562)

-0.306
**

(-2.610)

Age
0.128
***

(4.749)

0.128
***

(4.796)

0.127
***

(4.722)

Age Square
-
0.00223
***

(-6.665)

-0.00224
***

(-6.727)

-0.00221
***

(-6.619)

Square of Number of
Migrants

Number of Young
Dependents
-0.103
*

(-2.067)

-0.109
*

(-2.161)

-0.104
*

(-2.071)

Number of Elderly
Dependents
0.139
*

(2.250)

0.143
*

(2.313)

0.138
*

(2.208)

Number of Other Laborers
0.0537
†

(1.699)

0.0532
†

(1.677)

0.0635
*

(2.027)

Total Number of Other
Migrants in the Village
0.00831
***

(9.282)

0.00830
***

(9.230)

0.00829
***

(9.249)

40
Table 10-3, continued

Total Income
Agricultural and Nonagricultural
Income (Excluding Remittances)
Agricultural and Nonagricultural Income
per Adult (Excluding Remittances)
Variable Migration Remittances Migration Remittances Migration Remittances
Distance to the Closest Town
-0.0499
**

(-2.885)

-0.0510
**

(-2.895)

-0.0503
**

(-2.899)

Difference of Average
Household Asset from Village
Mean
0.00000804
***

(3.460)
-0.130
***

(-8.142)
0.00000804
***

(3.470)
-0.138
***

(-8.432)
0.00000811
***

(3.489)
-0.128
***

(-6.901)
Nonagricultural Income
(excluding Remittances)

-0.0615
**

(-2.539)

-0.230
**

(-2.593)
Agricultural Income
-0.241
*

(-2.188)

-0.327
(-1.032)
Household Income excluding
Remittances in 1998

-0.0754
***

(-3.312)

Migration Income in 1998
0.959
***

(82.572)

0.958
***

(84.441)

0.958
***

(82.048)
Village-Average Portion of
Remittances in Total Income

3294.323
***

(11.495)

3254.252
***

(10.940)

3326.678
***

(11.480)
Sigma(1)
4104.306
***

(14.854)

4073.660
***

(14.415)

4179.736
***

(14.336)
Rho(1,2)
0.620
***

(6.305)

0.618
***

(6.238)

0.642
***

(6.692)
Number in parentheses is t-statistics.
***
Coefficient different from zero at 0.001 significance level;
**
Coefficient different from zero at 0.01 significance level;
*
Coefficient different from zero at 0.05 significance level;
†
Coefficient different from zero at 0.1 significance level; Note: Provincial dummies are included but not reported here.

41
Greater distance between the village and the closest town significantly
decreases the possibility of migration. In China, rural towns have much better
transportation facilities to urban areas than villages and they are usually the
necessary starting point for migrants to get transportation to urban areas. Longer
distance between the village and the closest town means greater cost of migration.
The number of other laborers (age 16-59) in the household has positive
effects on migration. It means that a household with more laborers is more inclined
to send out family members to work outside the rural area. The relative pain of the
loss of labor is much smaller for households with a more abundant labor supply or
perhaps even redundant labor.
The number of young dependents (age 0-15) has a negative influence on the
migration decision. The greater the number of young children in the household, the
more time and energy will be needed to take care of them and thus the incentive and
possibility to send out migrants is reduced.
The number of elderly dependents (age no less than 60) positively affects the
possibility of migration. Elderly dependents can still help with young children,
housework and other productive work. Hence they help to relieve the burden that the
younger generation would bear for the rural household, leaving the younger ones free
to work in urban areas without worrying about the proper care of their rural homes.

42
5.1.2. Remittances Decision at the Individual Level
From our regression results, the main determinants of remittances are the
migrant's income, household income (including agricultural and nonagricultural
income) other than remittances, the difference of household average asset from the
village mean, and the village-average portion of remittances in total income.
Using the summary of predictions of motives by Rapoport and Docquier
(2006) (see Table 11), we find that our regression result supports the theories of
altruism, investment, and exchange motives in many aspects; we can also reject the
inheritance motive. Due to data constraints, we lack information about adverse short-
run income shocks in home areas. Thus we are unable to say much about the
insurance motive.
The strong evidence we get for altruism is that our result satisfies the specific
predictions listed in Table 11 and does not contradict to any other predictions of
signs of variables. An interesting prediction of the pure altruism hypothesis is that an
increase by one dollar in the income of the migrant, coupled with a one-dollar drop
in the recipient household's income, should raise the amount transferred exactly by
one dollar. Formally, the transfer-income derivatives should satisfy the following
condition:
1 =
!
!
"
!
!
h m
I
T
I
T
,
where T is the remittances from migrants(m) to households(h), and I is the
pre-transfer income.

43

Table 11. Remittances’ sensitivity to various explanatory variables (a summary
from Rapoport and Docquier (2006))

Motives Individual motives Familial arrangements
Explanatory
Variables
Altruism Exchange Inheritance Insurance Investment
Migrant’s
income
>0 >0 >0 nde (*) >0
Migrant’s
education
nde <0(*) nde nde >0 (*)
Distance
from family
!0 nde <0 nde >0
Number of
migrants
<0 nde
Inverse U-
shape effect
nde nde
Income of
rural
household
<0
>0 or <0
(*)
nde (*) nde (*) >0 or <0
Adverse
short run
shocks in
income of
rural
household
>0
>0 or <0
(*)
nde >0 >0
Assets of
rural
household
nde nde >0 (*) nde nde
Specific
predictions
1 =
!
!
"
!
!
h m
I
T
I
T

It is
possible
that
0 >
!
!
h
I
T

Role of
parental
assets and
number of
migrants
(i)
Irregular
basis
(ii) No
effect of
h
I in the
long run
Inverse
Ushaped
effect of
h
I
Note: nde = no direct effect (after controlling for migrants’ and/or recipients’ incomes)
(*) Remarkable prediction

44
Our result shows the difference of the derivatives of migrant's income and
recipient's income is about 1.03, very close to 1.
The migrant's income is the most significant determinant of remittances,
although all other explanatory variables are all significant at 0.001 significant level.
The positive sign of migrant income satisfies the predictions of altruism, exchange,
and investment motives. If other factors are fixed, the migrant will send back about
95 percent of his earnings. It implies the large marginal effect of a migrant's income
on remittances.
Recipient's income is measured by three means: (1) total household income
excluding remittances in Table 10-1; (2) agricultural income and nonagricultural
income excluding remittances in Table 10-3 (Column 2); (3) agricultural income per
adult and nonagricultural Income excluding remittances per adult in Table 10-2. All
estimated results shows that recipient's income has a negative impact on remittances.
This is consistent with the predictions of altruism, exchange, and investment
motives.
The relative ranking of household assets per member in the village negatively
affects the money sent by migrants. This means that migrants will remit less if their
rural households are relatively wealthier within the village. This contradicts the
prediction of the inheritance motive, which implies that wealthy households are more
able to secure remittances by rewarding or depriving the migrants of their rights to
inheritance. Other evidence for supporting the inheritance motive is that the number
of migrants/heirs has an inverse U-shape effect on remittances. We test it by

45
including the number of migrants and its square in our original regressions. The
result is not significant. This again shows that the inheritance motive is not supported
in our case.
The social norm of remitting in the village also plays an important role. The
proportion that remittances contribute to the total household income of other
villagers is a proxy for the local remittance norm. Remittances will increase with this
proportion.
We also include variables describing the characteristics of the individual. But
individual characteristics are not significant. We think that the migrant's income may
reflect the impact of the observed and unobserved characteristics of the individual
because it is significantly correlated with the observed individual characteristics
(gender, marital status, age, and education).

5.1.3. Different Decision Patterns for Migration and Remittances
One question raised by previous researchers for the study of migration and
remittances is whether the study of remittances is distinct from that of migration. In
empirical study, this implies the identification problem on migration and remittances.
Our results of individual-level decisions clearly prove that the decisions for
migration and remittances follow different patterns; we can distinctively estimate the
two decisions. Interesting evidence supporting our statement is that the difference of
average household assets from the village mean positively affects migration but
negatively affects remittances, both at 0.001 significance level. Wealthier households

46
are more able to finance the investment on migration; but migrants from poor
households will remit more, given that the influence of altruism dominates that of the
inheritance motives. This finding is supported by both theoretical study and many
empirical researches.

5.2. Household-level Tobit Model for Migration and Remittances
Decisions

5.2.1. Migration Decision at the Household Level
We can see that the migration decision is related to not only individual
characteristics, but also household characteristics. Thus the whole household does
play an important role in the migration decision. Although the remittances decision is
made by the migrant, the household will take the expected remittances as the most
important factor in deciding whether to support some household member(s) to work
in urban areas or not. Hence, it is also interesting to investigate migration decisions
from the perspective of the whole household, which will expect remittances as the
return on migration.
Our regression result (Table 12) shows that expected remittances, the number
of laborers in the household, the total number of other migrants in the village, the
number of elderly dependents in the household, and the difference of average
household consumption durables from the village mean mainly determine whether to
send any household member to migrate or not.

47

Table 12. Estimation of household-level decisions

With insignificant
variables
Without insignificant
variables
Variable
Migration Remittances Migration Remittances
Intercept
-7628.148
(-1.562)

-2664.453
(-0.697)
Expected Remittances
0.00283
***

(7744.38)

0.00264
***

(6778.18)

Number of Total Laborers
0.351
***

(5.683)

0.351
***

(5.959)

Number of Elderly Dependents
-6.469
***

(-3457.5)
2341.004
(0.951)

Difference from Village Mean
for Average Consumption
Durable
0.000327
(0.20)
-0.116
(-0.978)

Total Number of Other
Migrants in the Village
0.00810
***

(4.371)

0.00829
***

(4.430)

Nonagricultural Income per
Adult (Excluding Remittances)

0.795
*

(2.294)

0.688
*

(2.224)
Agricultural Income per Adult
-0.320
(-0.142)

Village-Average Portion of
Remittances in Total Income

4553.658
*

(2.156)

4540.268
**

(2.676)
Sigma(1)
11705.770
***

(21.106)

11909.223
***

(21.934)
Rho(1,2)
-0.560
***

(-3.299)

-0.517
***

(-3.521)
Number in parentheses is t-statistics.
***
Coefficient different from zero at 0.001 significance level
**
Coefficient different from zero at 0.01 significance level
*
Coefficient different from zero at 0.05 significance level
†
Coefficient different from zero at 0.1 significance level
Note: Provincial dummies are included but not reported here.

48
As we mention above, the main motivation for households to invest in
migration is to gain remittances. One-yuan increase of expected remittances will
cause the possibility of investing in migration to increase 2.83%.
The total number of laborers in the household also has a significant positive
impact on the migration decision at the household level, as it does in the migration
decision at the individual level. The reason is also the same: the negative effect of
the loss of labor is much smaller for households with more laborers.
Unsurprisingly, the total number of other migrants in the village has
significant positive effects in the migration decision at the household level, as it does
in the migration decision at the individual level. Again it shows that good migration
networks encourage the household to invest in migration.
With regard to the variable measuring the relative ranking of wealth of the
household, the result at the household level also confirms the implication from the
result at the individual level - relatively wealthier households are more able to invest
on migration.
However, the variable measuring the number of elderly dependents has an
opposite sign in the household-level decision from that in the individual-level
decision. More elderly dependents in the family will reduce the inclination of the
household to send migrants to urban areas. It seems that households will be more
likely to keep laborers at home areas, when they have more elderly dependents in the
family to look after. It implies that the whole household, as a unit making migration
decisions, has a different utility function from that of one individual considering

49
migration. The family decision makers may put more weight on the utility of elderly
dependents in the total household utility than the individual decision makers.
Comparing the migration estimation at the household level with that at the
individual level, we find that except the expected remittances and the number of
elderly dependents, the estimated coefficients of variables contributing to both
individual-level and household-level migration decisions have the same signs and
comparable magnitudes, implying the effectiveness of our models used in this study.

5.2.2. Remittances Decision at the Household Level
Since expected remittances are crucial for the household to decide whether to
send any laborer to migrate or not, we need to figure out how households form the
expectation. Assuming rational expectations, we can use the realized remittances as a
proxy for the expected remittance. Our study shows that the main factors that
households use to predict remittances are the village-average portion of remittances
in total household income, the nonagricultural income per adult in the household, the
agricultural income per adult in the household, the number of older adults in the
household, and the difference of average household consumption durables from
village mean.
The village-average portion of remittances in total household income is a
proxy for the local remittance norm. We get a positive estimated coefficient of this
variable, and it is statistically significant. Migrants from the village where household

50
income generally depends more on the resource of remittance will send more money
back home.
The local nonagricultural income per adult in the household positively affects
remittances. Households in rural areas usually assume that the local nonagricultural
output is positively correlated with the nonagricultural output in the migrants'
destination area. Then higher local nonagricultural income is a signal that migrants
may also earn more money in urban areas; thus the household expects more
remittances. This is consistent with the predictions of the investment motive.
The local agricultural income per adult in the household has a negative
relationship with remittances. Lower local agricultural income implies higher
remittances. Since we are unable to distinguish the adverse short-run shocks from
low productivity in our data, this may be justified by the altruism, insurance and
investment motives.
The results also show that households with more elderly dependents will
receive more remittances. Due to the lack of social security and health insurance in
rural China, people in rural areas traditionally have to depend on the support of their
children when they become old. This can be explained by altruism, and the
investment motive as repayment of the loans used to finance the migrant's education
etc., and can be partly traced to the exchange motive as long as the old household
members are still able to take care of some housework and local production work.
Comparing the remittances estimation at the household level with that at the
individual level, we find that the village-average portion of remittances in total

51
household income have an important positive influence on both the individual’s
remitting behavior and the household’s expectation of remittances; however,
nonagricultural income (excluding remittances) has exactly opposite impacts. More
nonagricultural income earned in rural areas decreases the individual’s motive of
altruism to remit, but it increases the expected amount of remittances by rural
households with the investment motive of sending migrants.

52
Chapter 6. The Impact of Migration and Remittances on the
Agricultural Productivity

Migration may have many impacts in rural areas. One important issue is how
total agricultural output and/or agricultural productivity are affected by migration.
Since China is a developing country with chronically low agricultural productivity
compared to developed countries, Chinese policymakers wonder whether China's
food needs can still be met with such a huge scale of migration out of rural areas.
Sufficient national food supplies have been achieved only in recent years.
The United Nations' World Food Programme (WFP) had offered a 26-year program
of food aid to China until 2005. Since 1979, the WFP has provided China with aid
valued at almost US$1 billion. However, the WFP phased out assistance at the end of
2005. As the WFP reported, “The arrival of the last shipment of WFP food aid to
China marks a significant watershed in the campaign to end global hunger, the
world's largest humanitarian agency declared today” (Figure 1). Although increasing
national income in the past decades greatly lessens the problem of hunger in China,
the extremely large population always makes "food security" a serious issue for the
Chinese government, as well as for many other developing countries. Hence,
Chinese policymakers are very concerned about agricultural production.
The ratio of rural population to the total Chinese population was more than
70% in the 1990s (China Statistical Yearbook). Thus the problem of hunger in China
is mainly caused by the low aggregate agricultural output. Since the cultivated land

53
area is almost fixed, or even decreasing because of the urbanization progress in
China, the low agricultural output is due to the low agricultural productivity on land.
In China, permanent ownership of all land remains with the state; rural
households are allotted plots to farm. Because of this concern for food security, the
government requires every household that receives some agricultural land to return a
certain quantity of grains to support urban dwellers. Thus, rural households need
laborers at home to farm in order to keep allocated land.
The decrease of labor in rural areas may hurt agricultural productivity. But
the increased capital achieved through remittances will help rural households to
overcome stringent credit and risk constraints so that they can apply more advanced
techniques of production and/or invest more on land with more or higher-quality
fertilizer, insecticides, etc. We want to study what the net impact of migration with
remittances on agricultural productivity will be.

54

Figure 1. Last shipment of food aid from the World Food Programme (WFP)
The MV Blue Dream docks in the southeastern city of Shenzhen in April 2005
carrying the last shipment of WFP food aid to China

Source: www.wfp.org

55
In our data, we have detailed production information about the five main
types of crops planted in these six provinces: wheat, early rice, middle rice, late rice
2

and maize--instead of maize only that Rozelle et al. (1999) studied. With more data
available we can see that the output of any crop alone cannot be a good measure of
aggregate agricultural output. So we should construct an indicator of aggregate
output for all crops that households grow. However, different types of crops have
different distributions of output per mu (see Table 13). For example, the average
output per mu of wheat is 609 jin and the average output per mu of middle rice is
985 jin.
3
To make the total output of these two types of crops comparable to each
other, we should multiply the total output of middle rice by 0.62 ( 62 . 0
985
609
! ) if
using the output of wheat as the base measure.

Table 13. Average outputs per mu of the five main types of crops
Crop 1. Wheat
2. Early
Rice
3. Middle
Rice
4. Late
Rice
5. Maize
Average
Output per
Mu (Jin)
609 680 985 807 682

2
Early rice, middle rice and late rice are types of rice planted during the early, middle and late part of
the year.

3
Mu is a land area measure used in China; 1 hectare equals 15 mu. Jin is a measure of weight; 1 jin is
equal to 0.5 kilogram.

56
We use the relative portions of the average output per mu of crop i
(averprod
i
) (i = 1, 2, 3, 4, 5) compared to that of wheat as the weights of output for
each crop. The weighted total output of all crops (WTP) becomes the measure of
aggregate output of all types of crops. So the WTP of the j
th
household is
i i
ij j
averprod
averprod
output WTP
1
5
1
!
=
" = .
An interesting feature of households with migrants is that they have higher
productivity (output per mu), but they have fewer laborers staying at home and less
total capital than those households without migrants (see Table 14). The same feature
exists when we compare households receiving remittances with those having no
remittances (see Table 15). Moreover, households with migrants have fewer adults at
home but spend much more total labor days and labor days per capita (regardless of
age) on agricultural production. Again, the same feature exists when we compare
households receiving remittances with those having no remittances. More
surprisingly, the average expense of hiring laborers in the migrants’ and/or remitters'
households is lower than that in households without migrant and/or remittances. So
the significant gap of agricultural labor days between these two kinds of households
may be caused mainly by greater efforts of members in the migrants’ and/or
remitters' households. Migrants may work hard even during their visits home, if they
do not work in urban areas for the whole year. The mean value of migration days in
1998 for a migrant is 250 - traditionally migrants go back to their rural home in the
busiest time of agricultural production, such as harvest time, if possible. Moreover,

57
the remaining household members may work harder than those in households
without migrants. This shows that the implicit contract between migrants and those
who remain at home overcomes the moral hazard problem (the remaining members
may reduce their level of effort when they are insured against risk). In total, although
the households with migrants and/or remitters have much less fixed capital
investment and fewer laborers staying home, they still take hold by increasing effort
in agricultural production, and successfully achieve higher output per mu. Thus the
positive impact of migration and remittances on agricultural production overcomes
the negative effect from the loss of agricultural labor.
We study the net impact of migration with remittances in two approaches. In
Section 6.1, we use the approach of the standard Cobb-Douglas production function.
In Section 6.2, we follow the setup of the recursive system in Rozelle et al. (1999)
and compare our estimated results with theirs.

58

Table 14. Agricultural input and output of households with and without
migrants

Without
Migrants
With
Migrants
Weighted Total Output 3592.14 4378.32
Weighted Total Output per mu 607.455 623.4894
# Migrants 0 1.3529
# Return Migrants 0.1522 0.1529
# Laborers 2.7009 3.5216
# Laborers not migrated 2.7009 2.1686
Land Area 6.4245 6.90705
Total Capital 4143.45 3805.54
Variable Capital Investment 831.4203 812.1262
Productive Assets (Fixed Capital Investment) 3312.03 2993.41
Nonproductive Assets 33534.04 39332.93
Cost of Hiring Laborers 77.08 28.7712
Proportion of Hiring Laborers 0.1168 0.1176
Laboring Days 90.7008 121.2415
Laboring Days per capita (Regardless of Age) 23.9440 41.0023
Laboring Days (with Estimated Hiring Labor
Days) per mu 20.5675 21.4611
Laboring Days (with Estimated Hiring Labor
Days) per mu per capita 5.5753 7.2189
Variable Capital Investment per mu 135.6658 121.2968
Productive Assets (Fixed Capital Investment) per
mu 1219.91 505.02
Source: survey

59

Table 15. Agricultural input and output of households with and without
remittances

Without
Remittances
With
Remittances
Weighted Total Output 3614.70 4581.97
Weighted Total Output per mu 563.2128 645.2661
# Migrants 0.1604 1.3207
# Return Migrants 0.1462 0.1739
# Laborers 2.8365 3.3696
# Laborers not migrated 2.6761 2.0489
Land Area 6.4180 7.1009
Total Capital 4303.45 3142.60
Variable Capital Investment 823.7277 831.3355
Productive Assets (Fixed Capital
Investment)
3479.72 2311.26
Nonproductive Assets 36141.11 32559.16
Cost of Hiring Laborers 73.2085 24.6840
Proportion of Hiring Laborers 0.1164 0.1196
Laboring Days 94.3475 119.7587
Laboring Days per capita (Regardless of
Age)
25.5083 41.7764
Laboring Days (with Estimated Hiring
Labor Days) per mu
21.0418 20.1711
Laboring Days (with Estimated Hiring
Labor Days) per mu per capita
5.8211 6.9710
Variable Capital Investment per mu 134.0307 121.7212
Productive Assets (Fixed Capital
Investment) per mu
1202.37 309.7724
Remittances 0 4864.67
Source: survey

60
6.1. The Single Equation Estimation
We use the standard Cobb-Douglas production function to estimate the
impact of remittances. We have two alternative assumptions to incorporate
remittances into the production function:
(1) Remittances affect the technology used in agricultural production only;
(2) Remittances affect the capital input in agricultural production only.

(1) Remittances affect the technology used in agricultural production
We assume that remittances affect the technology used in agricultural
production only. The Cobb-Douglas production function is then written as
µ ! ! !
e N L K A Y
R c
3 2 1
"
= (22)
where Y is the total agricultural output; A is the technology; R is remittances; K is
the total capital input; L is the total land area; N is the total labor input.
We can take log to get the linear regression equation for the total agricultural
output as
µ ! ! ! + + + + " = ) log( ) log( ) log( ) log( ) log(
3 2 1
N L K R A c Y (23)
To study the agricultural productivity – output per mu, we divide both sides
of Equation (22) by the total land area, L, and get
µ ! ! !
e N L K A
L
Y
R c
3 2 1
) 1 ( " #
= (24)
Similarly, we can take log to get the linear regression equation for the agricultural
output per mu as

61
µ ! ! ! + + " + + # = ) log( ) log( ) 1 ( ) log( ) log( ) log(
3 2 1
N L K R A c
L
Y
(25)
We can further distinguish capital input as fixed capital input (
f
K ) and
variable capital input (
v
K ), and assume that the two types of capital inputs make
different contributions to agricultural production. Fixed capital input consists
primarily of production equipment. Variable capital inputs include expenses for
seeds, fertilizers, pesticide, irrigation etc. Then we have
µ ! ! ! ! + + + + + " = ) log( ) log( ) log( ) log( ) log( ) log(
3 2 12 11
N L K K R A c Y
v f
(26)
µ ! ! ! ! + + " + + + # = ) log( ) log( ) 1 ( ) log( ) log( ) log( ) log(
3 2 12 11
N L K K R A c
L
Y
v f
(27)

(2) Remittances affect the capital input in agricultural production
We assume that remittances affect the capital input in agricultural production
only – a proportion of remittances, d , are added to the existing capital, K . The
Cobb-Douglas production function is then written as
µ ! ! !
e N L R d K A Y
3 2 1
) ( " + = (28)
The log linear regression equation is
µ ! ! ! + + + " + + = ) log( ) log( ) log( ) log( ) log(
3 2 1
N L R d K A Y (29)
And
) log( ) 1 log( ) 1 log( ) log( K
K
R
d K
K
R
d R d K + ! + = ! + = ! +

62
Using Taylor expansion,
K
R
d
K
R
d ! " ! + ) 1 log( if 1 < !
K
R
d
Then we get
µ ! ! ! ! + + + " " + + = ) log( ) log( ) log( ) log( ) log(
3 2 1 1
N L
K
R
d K A Y (30)
And the log linear regression equation for the agricultural output per mu is
µ ! ! ! ! + + " + # # + + = ) log( ) log( ) 1 ( ) log( ) log( ) log(
3 2 1 1
N L
K
R
d K A
L
Y
(31)
If we distinguish capital input as fixed capital input and variable capital input,
and assume the current-year remittances affect current variable capital only and does
not change existing productive equipments (fixed capital input) – a proportion of
remittances,
v
d , are added to the existing variable capital,
v
K ., then we have
µ ! ! ! ! ! + + + " " + + + = ) log( ) log( ) log( ) log( ) log( ) log(
3 2 12 12 11
N L
K
R
d K K A Y
v
v v f
(32)
µ ! ! ! ! ! + + " + # # + + + = ) log( ) log( ) 1 ( ) log( ) log( ) log( ) log(
3 2 12 12 11
N L
K
R
d K K A
L
Y
v
v v f
(33)
The estimation results of both the total agricultural output and the agricultural
output per mu are listed in Table 16-1 and Table 16-2, respectively. The estimated
results in Table 16-2 are the same as those in Table 16-1, except that the coefficient
of land area (log) now equals the original coefficient minus 1 (because we divide
both sides of the total agricultural output equations by the land area, L). Despite the
different specifications of estimated equations, we find that the estimated coefficients

63
are similar across specifications. We can calculate the estimated value of d and
v
d
as 0.23 and 0.10, which justifies our assumption for Taylor expansion, because about
94% of ) (
K
R
d! and 96% of ) (
v
v
K
R
d ! falls into 1 according to our data. In all four
specifications, remittances make a significant contribution to agricultural
productivity on land (output per mu), and total agricultural output, correspondingly.
Considering its indirect effect on existing capital inputs (remittances in previous
years help to build up existing capital) and even labor inputs (households can hire
laborers to replace lost migrant laborers), remittances can have a larger positive
impact to production.

64

Table 16-1. Estimation of total agricultural output with the Cobb-Douglas
production function

Variable Absolute Remittances
Ratio of
Remittances to
Total Capital
Ratio of
Remittances to
Variable Capital
Land Area (log)
0.8727
***

(43.33)
0.8433
***

(29.78)
0.8773
***

(43.79)
0.8444
***

(29.92)
Laboring Days of
All Workers (log)
0.05882
***

(3.77)
0.05732
***

(3.68)
0.05680
***

(3.65)
0.05592
***

(3.60)
Total Capital
(log)
0.01679
†

(1.79)

0.01867
*

(1.97)

Variable Capital
(log)

0.03907
†

(1.69)

0.04372
†

(1.89)
Fixed Capital
(log)

0.00785
**

(2.55)

0.00813
**

(2.65)
Absolute
Remittances
0.00000515
***

(2.89)
0.00000506
**

(2.85)

Ratio of
Remittances to
Total Capital

0.00425
**

(2.69)

Ratio of
Remittances to
Variable Capital

0.00448
***

(3.36)
Adjusted R
2
0.8678 0.8687 0.8676 0.8692
Number in parentheses is t-statistics.
**
Coefficient different from zero at 0.01 significance level
*
Coefficient different from zero at 0.05 significance level
†
Coefficient different from zero at 0.1 significance level
Note: Provincial dummies are included but not reported here.

65

Table 16-2. Estimation of agricultural output per mu with the Cobb-Douglas
production function

Variable Absolute Remittances
Ratio of
Remittances to
Total Capital
Ratio of
Remittances to
Variable Capital
Land Area (log)
-0.1273
***

(-6.32)
-0.1567
***

(-5.53)
-0.1227
***

(-6.12)
-0.1556
***

(-5.51)
Laboring Days of
All Workers (log)
0.05882
***

(3.77)
0.05732
***

(3.68)
0.05680
***

(3.65)
0.05592
***

(3.60)
Total Capital
(log)
0.01679
†

(1.79)

0.01867
*

(1.97)

Variable Capital
(log)

0.03907
†

(1.69)

0.04372
†

(1.89)
Fixed Capital
(log)

0.00785
**

(2.55)

0.00813
**

(2.65)
Absolute
Remittances
0.00000515
***

(2.89)
0.00000506
**

(2.85)

Ratio of
Remittances to
Total Capital

0.00425
**

(2.69)

Ratio of
Remittances to
Variable Capital

0.00448
***

(3.36)
Adjusted R
2
0.1299 0.1359 0.1286 0.1395
Number in parentheses is t-statistics.
**
Coefficient different from zero at 0.01significance level
*
Coefficient different from zero at 0.05significance level
†
Coefficient different from zero at 0.1significance level
Note: Provincial dummies are included but not reported here.

66
6.2. The Simultaneous Equation Estimation
Because our research and the research Rozelle et al. (1999) use different data,
we use the regression model of Rozelle et al. (1999) with our data to see if we can
obtain results similar to theirs.
Rozelle et al. (1999) uses the iterative three-stage least-squares to estimate
the simultaneous equations of migration, remittances and agricultural production.
They estimate the recursive system as below:
Y Y
C
Z R M Y ! " " " " + + + + =
3 2 1 0
(34)
R R
Z M R ! " " " + + + =
2 1 0
(35)
M M
Z M ! " " + + =
1 0
(36)
where migration measured by the number of migrants in the household, M, and
Remittances, R, are endogenously determined with yields per mu,
C
Y .
In Table 17-1, the first column of the estimation of each equation is the result
of Rozelle et al. (1999). The second column of the estimation of each equation is the
result when we use all variables that they suggested except those that we do not have
in our data. The third column shows the estimated result when we add some
variables that are important predictors in migration, remittances, and agricultural
production. The estimated coefficients of migration variable in the remittances
equation (Equation 35) and both migration variable and remittances variable in the
yield equation (Equation 34) by Rozelle et al. (1999) are very significant. Thus,

67
Rozelle et al. (1999) can calculate the total direct plus indirect effect of an additional
migrant on yields is -101 jin per mu. The calculation is as below:
101 63 . 818 44 . 0 63 . 461 ! " # + ! =
$
$
#
$
$
+
$
$
M
R
R
Y
M
Y
C C
.
However, our estimated results are very different from Rozelle et al. (1999). In
particular, the estimated coefficients of both the migration variable and remittances
variable in the yield equation (Equation 34) in both our specifications are not
significant. Hence, we cannot calculate the net effect of migration with remittances.
We find that Rozelle et al. (1999) put both the number of migrants in the
household and the household size in the yield equation, ignoring that the household
size is a significant explanatory variable of the migration. So we delete the number
of migrants from the yield equation and find that the estimated result reported in
Table 17-2 is well improved compared to the result in the second column in Table
17-1. With the significant estimation of the remittances in the yield equation, we now
can calculate the net effect of migration with remittances as 22 jin per mu, which
implies that migration with remittances has positive impact on agricultural
productivity.
Compared to the model of Rozelle et al. (1999) that only shows household
size and the number of migrants, our data offers the detailed labor input information
on agricultural production measured by the labor days on the farm. We can estimate
the impact of migration with remittances on the effort level of the household on
agricultural production, instead of only the number of laborers, which actually

68
remains unknown as a function of the number of household members and the number
of migrants in Rozelle et al. (1999). In addition, the agricultural productivity can be
better approached by the actual labor input instead of the number of household
members and the number of migrants in Rozelle et al. (1999). So we can combine the
single-equation model of Cobb-Douglas production function with the simultaneous
equations in Rozelle et al. (1999) as below:
M M
Z M ! " " + + =
1 0
(37)
R R
Z M R ! " " " + + + =
2 1 0
(38)
N N M
Z D N ! " " " + + + =
2 1 0
(39)
µ ! ! ! ! + + " + + + = ) log( ) log( ) 1 ( ) log( ) log( ) log( ) log(
3 2 12 11
N L K K A c
L
Y
v f
C
(40)
where
M
D is a dummy variable of migration that measures the impact of the implicit
migration contract on the effort level of agricultural production; A is measured by
the experience of household head in our model.
The regression result in Table 17-3 shows that migration contributes
positively to the labor input. This implies that China’s rural migration to urban areas
has no negative lost-labor effect on agricultural production; in addition, the implicit
contract between migrants and their rural households provides an incentive for the
entire household to work harder on the farm, and thus increases the agricultural
productivity. The final estimation of the yield function in the simultaneous system is
similar to that in the single-equation model in section 6.1.

69
Given our data description and regression results, we find that migration with
remittances has positive effects on the technology and capital investment of
agricultural production; more surprisingly, the implicit contract of the migrants’
family successfully deals with the moral hazard problem and even increases the labor
input on the farm. The improvement in all three aspects increases the agricultural
productivity of migrants’ and/or remitters’ household. Thus, we do not agree with
the conclusion of Rozelle et al. (1999) that the net impact of migration on
agricultural production is negative.

70
Table 17-1. Comparison with the result of Rozelle et al. (1999) (1)

Independent
Variable
Dependent Variables
(i). Number of Migrants (ii). Remittances (iii). Yield per mu
Rozelle’s
Estimation
My
Estimation1
My
Estimation2
Rozelle’s
Estimation
My
Estimation1
My
Estimation2
Rozelle’s
Estimation
My
Estimation1
My
Estimation2
Migration Effect:
Number of
Migrants

818.63
**

(5.28)
1514.187
*

(1.96)
1540.018
*

(2.01)
-461.63
**

(3.32)
59.2655
†

(1.86)
5.6559
(0.18)
Total Remittances
0.44
**

(3.21)
0.0047
(0.56)
0.0058
(0.66)
Human Capital and Household Characteristics:
Experience of
Head
0.0053
*

(2.47)
-0.0507
(-0.90)
-0.0551
(-0.98)

0.90
(0.52)
40.5453
**

(2.57)
38.2676
*

(2.43)
Education of Head
0.045
**

(4.38)
0.0031
(0.37)
0.0053
(0.62)

18.62
**

(3.03)
-1.1377
(-0.50)
-1.0679
(-0.46)
Household Size
0.27
**

(11.97)
0.2132
**

(10.19)
0.2199
**

(10.44)

-6.86
(0.32)
-19.8366
**

(-3.75)

Young Dependents
-0.32
**

(10.43)
-0.186
**

(-5.93)
-0.2028
**

(-6.35)
-35.12
(0.69)
39.8003
(0.16)
-29.0007
(-0.12)

Elderly
Dependents
-0.11
**

(3.31)
-0.1154
**

(-3.38)
-0.1205
**

(-3.50)
-20.82
(0.51)
757.7507
*

(2.38)
709.2585
*

(2.23)

Land per capita
0.037
**

(3.25)
0.0075
(0.41)
0.0091
(0.49)
13.41
(0.75)
104.5083
(0.58)
110.3986
(0.61)

Nonproductive
Assets
-0.037
**

(3.25)
-1.45E-8
(-0.03)
-2.92E-8
(-0.05)
6.74
(0.45)
0.0144
**

(2.85)
0.0146
**

(2.88)
4.26
(0.52)
0.00062
**

(3.41)
0.000495
**

(2.69)
Productive Assets
-1.21E-6
(-1.16)
-1.37E-6
(-1.29)

-0.0128
(-1.24)
-0.0136
(-1.32)

0.00064
(1.35)

71
Table 17-1, continued

Independent Variable Dependent Variables
(i). Number of Migrants (ii). Remittances (iii). Yield per mu
Rozelle’s
Estimation
My
Estimation1
My
Estimation2
Rozelle’s
Estimation
My
Estimation1
My
Estimation2
Rozelle’s
Estimation
My
Estimation1
My
Estimation2
Village and Plot Characteristics:
Frequency of Land
readjustment
-0.033
*

(2.24)
-0.0078
(-0.50)
-0.0079
(-0.50)
5.48
(0.16)
77.8959
(0.50)
85.708
(0.55)
-24.00
(1.34)
-1.2126
(-0.25)
-3.8782
(-0.80)
Degree of Land
Readjustment
-0.00036
(0.60)

-4.64
**

(3.41)

1.03
(1.08)

Plot Distance from
Household

-13.46
(0.47)

Size of Plot
-12.78
(1.94)
-1.8191
**

(-3.97)
-3.1245
**

(-3.91)
High-quality Land Dummy
73.41
*

(2.4)

Input of Agricultural Production:
Variable Capital
Investment on Land per mu

0.0906
(1.57)
0.0903
(1.35)
Labor Days on Land
0.1614
†

(1.90)
Expense of Hiring Labor
-0.1599
(-0.98)

72
Table 17-1, continued

Independent Variable Dependent Variables
(i). Number of Migrants (ii). Remittances (iii). Yield per mu
Rozelle’s
Estimation
My
Estimation1
My
Estimation2
Rozelle’s
Estimation
My
Estimation1
My
Estimation2
Rozelle’s
Estimation
My
Estimation1
My
Estimation2
Instruments:
Most Educated in
Household (year)
-0.04
**

(3.93)
-0.0035
(-0.37)
-0.0075
(-0.78)

Village Enterprise Dummy
-0.063
(1.38)

Percentage Out-migrants
0.83
**

(2.81)
2.9324
**

(13.34)
2.857
**

(12.86)

Average Summer Crop
Shock

-5.75
(1.74)

Village Percentage Income
from Remittance

-3.31
(0.24)
2209.858
*

(2.46)
2199.484
*

(2.46)

Number in parentheses is t-statistics.
**
Coefficient different from zero at 0.01significance level
*
Coefficient different from zero at 0.05significance level
†
Coefficient different from zero at 0.1significance level
Note: Provincial dummies are included but not reported here.

73
Table 17-2. Comparison with the result of Rozelle et al. (1999) (2)

Independent Variable Dependent Variables
(i). Number of Migrants (ii). Remittances (iii). Yield per mu
My
Estimation1
Delete # of
Migrants
My
Estimation1
Delete # of
Migrants
My
Estimation1
Delete # of
Migrants
Migration Effect:
Number of Migrants
1514.187
*

(1.96)
1872.486
*

(2.49)
59.2655
†

(1.86)

Total Remittances
0.0047
(0.56)
0.0118
**

(2.91)
Human Capital and Household Characteristics:
Experience of Head
-0.0507
(-0.90)
-0.0559
(-0.99)

40.5453
**

(2.57)
39.526
**

(2.57)
Education of Head
0.0031
(0.37)
0.0045
(0.52)

-1.1377
(-0.50)
-1.1368
(-0.55)
Household Size
0.2132
**

(10.19)
0.21985
**

(10.44)

-19.8366
**

(-3.75)
-13.9937
**

(-2.88)
Young Dependents
-0.186
**

(-5.93)
-0.2003
**

(-6.28)
39.8003
(0.16)
71.3464
(0.29)

Elderly Dependents
-0.1154
**

(-3.38)
-0.1215
**

(-3.53)
757.7507
*

(2.38)
686.5689
*

(2.26)

Land per capita
0.0075
(0.41)
0.0089
(0.47)
104.5083
(0.58)
84.6071
(0.49)

Nonproductive Assets
-1.45E-8
(-0.03)
-3.51E-8
(-0.06)
0.0144
**

(2.85)
0.014
**

(2.77)
0.00062
**

(3.41)
0.00052
**

(3.21)
Productive Assets
-1.21E-6
(-1.16)
-1.33E-6
(-1.26)
-0.0128
(-1.24)
-0.011
(-1.10)

Village and Plot Characteristics:
Frequency of Land
readjustment
-0.0078
(-0.50)
-0.0075
(-0.48)
77.8959
(0.50)
90.8924
(0.58)
-1.2126
(-0.25)
-4.3796
(-0.94)
Size of Plot
-1.8191
**

(-3.97)
-1.6652
**

(-3.61)
Instruments:
Most Educated in
Household (year)
-0.0035
(-0.37)
-0.0068
(-0.70)

Percentage Out-
migrants
2.9324
**

(13.34)
2.8839
**

(13.00)

Average Summer Crop
Shock

Village Percentage
Income from
Remittance

2209.858
*

(2.46)
2022.793
*

(2.84)

Number in parentheses is t-statistics.
**
Coefficient different from zero at 0.01significance level
*
Coefficient different from zero at 0.05significance level
†
Coefficient different from zero at 0.1significance level
Note: Provincial dummies are included but not reported here.

74
Table 17-3. Comparison with the result of Rozelle et al. (1999) (3)

Independent Variable Dependent Variables
(i). Number of
Migrants
(ii).
Remittances
(iii). Labor Days per
mu per Capita
(iv). Yield
(Log)
Migration Effect:
Number of Migrants
1576.252
**

(4.26)

Total Remittances
Human Capital and Household Characteristics:
Migration Dummy(=1if have
migrants)
0.12
**

(5.03)

Experience of Head
-0.0755
(-1.34)
0.0443
†

(1.66)
Laborers
0.2187
**

(11.94)

Laborers not Migrated
46.9858

(0.24)
-1.2514
**

(-6.02)

Young Dependents
0.0204
(0.75)
0.8651
(0.00)

Elderly Dependents
0.1034
**

(3.01)
789.505
*

(2.35)

Land per capita
0.005
(0.25)
74.6581
(0.39)

Nonproductive Assets per
capita
1.253E-6
(0.53)
0.0502
*

(2.23)
0.000082
**

(3.17)

Productive Assets per capita
-8.98E-6
†

(-1.76)
-0.039
(-0.78)
-0.00009
†

(-1.65)

Input of Agricultural Production:
Labor Days on Land per mu
(Log)
0.0612
**

(4.07)
Total Land (Log)
-0.069
**

(-4.42)
Productive Assets per mu
(Log)
0.008
*

(2.14)
Investment on Land per mu
(Log)
0.0392
†

(1.71)
Investment on Land per mu
0.0051
*

(2.11)

Village and Plot Characteristics:
Percentage Out-migrants
2.9128
**

(10.32)

Village Percentage Income
from Remittance
2160.125
**

(3.49)

Number in parentheses is t-statistics.
**
Coefficient different from zero at 0.01 significance level;
*
Coefficient different from zero at 0.05
significance level;
†
Coefficient different from zero at 0.1 significance level
Note: Provincial dummies are included but not reported here.

75
Chapter 7. Conclusions and Future Research

In this dissertation, we show that even within the same country, motivations
underlying the behavior of remitting are different from the incentives for migration,
after taking into account the characteristics of the province and the village. The most
convincing evidence, in our estimation, are the opposing signs of the variable
measuring wealth in the migration and remittances equations at the individual level -
wealthier households are more likely to invest in migration, but migrants from
relatively poor households will remit more. In further research, we will collect data
about rainfall and other meteorological information, where available, and use these
exogenous variables to further identify the joint decisions of migration and
remittances and test the motive of insurance in remittances under negative weather
shocks.
In the migration literature, some studies are conducted at the individual level,
others at the household level. We think that individuals and households have
different perspectives while considering migration; both need to be examined for a
more complete picture. Comparing the empirical result at the household level with
that at the individual level, we find that for households, the expected remittances are
the main concern when households consider an investment in migrants; for
individuals, factors affecting their earning potentials, such as human capital and
migration networks, are the priority in their migration decision. Empirically we also
discover the different directions of impacts for the same predictor. Firstly,
nonagricultural income (excluding remittances) has opposite effects on the

76
individual’s remitting behavior and the household’s expectation of remittances.
Secondly, the number of elderly dependents has positive effects in individuals'
migration but negatively affects households' migration. These results imply that the
whole household has a different utility function concerning migration from that of an
individual. But despite the two differences we discuss above, the estimated
coefficients of variables contributing to both individual-level and household-level
migration decisions have the same signs and similar magnitudes, implying the
effectiveness of the models used in this study. In future research, we will build
models to incorporate the individual-level decision into the household-level decision.
Our results on individual remittances decision-making strongly support the
altruism motive, and are consistent with investment and exchange motives. The
higher effort level on agricultural production in the migrants’ and/or remitters'
households implies that the implicit contract between migrants and the remaining
members effectively deals with the moral hazard problem. Higher labor inputs and
greater financial resources of remitters' households result in a significantly higher
agricultural productivity than that of households without remittances. Thus, we do
not agree with Rozelle et al. (1999) who concluded that migration leads to worse
agricultural productivity, and we are optimistic about the "food security" problem in
China. In the future, we will develop a simultaneous model that includes both the
tobit feature of migration and remittances decisions and the impacts of migration
with remittances on agricultural productivity.

77
References
Agarwal, R. & A.W. Horowitz. 2002. “Are International Remittances Altruism or
Insurance? Evidence from Guyana Using Multiple-Migrant Households.”
World Development 30(11): 2033-2044.

Amemiya, Takeshi. 1985. Advanced Econometrics, Cambridge: Harvard University
Press.

Azam, Jean-Paul & Flore Gubert. 2005. “Those in Kayes: The Impact of
Remittances on Their Recipients in Africa.” Revue économique 56: 1331-
1358.

Benjamin, Dwayne & Loren Brandt. 1998. “Administrative Land Allocation,
Nascent Labor Markets, and Farm Efficiency in Rural China.” Working
paper. Toronto: University of Toronto.

Cox, Donald. 1987. “Motives for Private Income Transfers.” Journal of Political
Economy 95(3): 508-546.

de la Brière, B., E. Sadoulet, A. de Janvry & S. Lambert. 2002. “The Role of
Destination, Gender, and Household Composition in Explaining Remittances:
An Analysis for the Dominican Sierra.” Journal of Development Economics
68(2): 309-328.

Funkhouser, Edward. 1995. “Remittances from International Migration: A
Comparison of El Salvador and Nicaragua.” Review of Economics and
Statistics 77(1): 137-146.

Hoddinott, John. 1994. “A Model of Migration and Remittances Applied to Western
Kenya.” Oxford Economic Papers 46: 459-476.

Huang, Ping & Frank N. Pieke. 2003. “China Migration Country Study.” Paper
presented at the Conference on Migration, Development and Pro-Poor Policy
Choices in Asia, Dhaka.

Jamison, Dean T. & Lawrence J. Lau. 1986. Farmer education and farm efficiency.
Baltimore: Johns Hopkins University Press.

Lucas, Robert. 1987. “Emigration to South Africa's Mines.” American Economic
Review 77(3): 313-30.

78
Lucas, Robert. 1998. “Internal Migration and Urbanization: Recent Contributions
and New Evidence.” IED Discussion Paper 91. Boston: Boston University.

Maddala, G.S. 1983. Limited-independent and Qualitative Variables in
Econometrics. New York: Cambridge University Press.

Rapoport, Hillel & Docquier, Frédéric. 2006. “The Economics of Migrants’
Remittances.” In Handbook of the Economics of Giving, Altruism and
Reciprocity, Vol. 2, ed. Serge-Christophe Kolm and Jean Mercier Ythier,
1135-1200. Amsterdam: Elsevier Press.

Roberts, Kenneth D. 1997. “China's ‘Tidal Wave’ of Migrant Labor: What Can We
Learn From Mexican Undocumented Migration to the United States.”
International Migration Review 31( 2): 249-293.

Rozelle, Scott, Taylor, J. Edward & deBrauw, Alan. 1999. “Migration, Remittances,
and Agricultural Productivity in China.” American Economic Review 89(2):
287-291.

Rozelle, Scott, Linxiu Zhang & Jikun Huang. 2002. “Employment, Emerging Labor
Markets, and the Role of Education in Rural China.” Working paper. Davis:
University of California.

Singh, I., Lynn Squire & John Strauss. 1986. “An Overview of Agricultural
Household Models-The Basic Model: Theory, Empirical Results, and Policy
Conclusions.” In Agricultural Household Models, Extensions, Applications,
and Policy, ed. I. Singh, Lynn Squire, and John Strauss, 17-47. Baltimore:
Johns Hopkins University Press.

Stark, Oded. 1991. The Migration of Labor. Cambridge: Blackwell Press.

Stark, Oded & Lucas, Robert. 1988. “Migration, Remittances, and the Family”,
Economic Development and Cultural Change 36(3): 465-81.

Taylor, J. Edward. 1992. “Remittances and Inequality Reconsidered: Direct, Indirect
and Intertemporal Effects.” Journal of Policy Modeling 14(2): 187-208.

Taylor, J. Edward & Philip L. Martin. 1999. “Human Capital: Migration and Rural
Population Change.” In Handbook of Agricultural Economics, ed. Gordon
Rausser and Bruce Gardner, New York: Elsevier Press.

World Food Programme. 2005. www.wfp.org News - Press Release.

79
Yao, Yang. 2001. “Egalitarian Land Distribution and Labor Migration in Rural
China.” Working paper, Beijing: China Center for Economic Research,
Beijing University.

Zhao, Yaohui. 1999a. “Labor Migration and Earnings Differences: the Case of
China,” Economic Development and Cultural Change 47(4): 767-782.

Zhao, Yaohui. 1999b. “Leaving the Countryside: Rural to Urban Migration
Decisions in China.” American Economic Review (Papers and Proceedings),
89, 281-286.

Zhao, Yaohui. 2001a. “Causes and Consequences of Return Migration: Recent
Evidence from China.” Working paper. Beijing: China Center for Economic
Research, Beijing University.

Zhao, Yaohui. 2001b. “The Role of Migrant Networks in Labor Migration: The Case
of China.” Working paper. Beijing: China Center for Economic Research,
Beijing University.

Abstract (if available)

Abstract Using recent household survey data from rural China, this dissertation investigates the decisions of migration and remittances at both individual and household levels and their impact on agricultural productivity. Previous researchers wonder if remittance behavior can be predicted by the migrants' characteristics

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays in the study of institutions and development

PDF

Three essays on real estate risk and return

PDF

Essays on the econometrics of program evaluation

PDF

Housing consumption-based asset pricing and residential mortgage default risk

PDF

Children 's migration and the financial, social, and psychological well-being of older adults in rural China

PDF

Maternal full-time employment and childhood obesity

PDF

The impact of agglomeration policy on CO₂ emissions: an empirical study using China’s manufacturing data

PDF

An econometric evaluation of IMF program effectiveness

PDF

Income inequality and economic growth: A theoretical and empirical analysis

PDF

Intergenerational transfers in rural China: Do children's gender and birth order matter?

PDF

Mechanisms of human skin cell migration and wound healing

PDF

International brides: cross-border marriage migration in China and Japan through a feminist lens

PDF

Assessment of the impact of second-generation antipscyhotics in Medi-Cal patients with bipolar disorder using panel data fixed effect models

PDF

Technomobility in the margins: mobile phones and young rural women in Beijing

PDF

Essays on product trade-ins: Implications for consumer demand and retailer behavior

PDF

An economic model of the security motive in the demand for children: An application to Malaysia

PDF

Insurance mechanisms, forest clearance, and the effect of government policies in rural economies

PDF

The divergence of the economic fortunes of Hindus and Muslims in British India: a comparative institutional analysis

PDF

Getting a new blanket: China’s conceptualization of “security” in the post-Deng Xiaoping era

PDF

Adaptation, assets, and aspiration. Three essays on the economics of subjective well-being

Asset Metadata

Creator Qin, Qi (author)

Core Title The decisions of migration and remittances in rural China

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Publication Date 10/22/2009

Defense Date 05/29/2006

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag agricultural productivity,China,migration,OAI-PMH Harvest,remittances

Place Name China (countries)

Language English

Advisor Hsiao, Cheng (committee chair), Deng, Yongheng (committee member), Nugent, Jeffrey B. (committee member)

Creator Email qinqi11@yahoo.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m878

Unique identifier UC1318547

Identifier etd-Qin-20071022 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-557783 (legacy record id),usctheses-m878 (legacy record id)

Legacy Identifier etd-Qin-20071022.pdf

Dmrecord 557783

Document Type Dissertation

Rights Qin, Qi

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu