You are on page 1of 28

Urbanization, informal sector, and development


Kazuhiro Yuki

Faculty of Economics, Kyoto University, Yoshida-honmachi, Sakyo-ku, Kyoto, 606-8501, Japan
Received 19 July 2004; received in revised form 30 June 2005; accepted 17 September 2006
Abstract
Urbanization is a universal phenomenon in the postwar world, but its qualitative nature is varied across
countries. Some nations have experienced urbanization that accompanied skill upgrading, industrialization,
and the expansion of the urban formal (modern) sector, but others urbanized without such modernization
and underwent the expansion of the urban informal (traditional) sector simultaneously. The question that
naturally arises is what are the underlying causes for the differences. The purpose of the paper is to tackle
this question analytically based on a dynamic model of urbanization and development.
© 2006 Elsevier B.V. All rights reserved.
JEL classification: J62; O11; O15; O17
Keywords: Urbanization; Informal sector; Development; Income distribution
1. Introduction
Urbanization proceeded universally in the post-war world. The degree of urban concentration
increases with a nation's income level, while the speed of urbanization shows an inverted U
pattern: the speed of urbanization of middle-income nations is highest and that of developed
nations is lowest (Mazumdar, 1987). Since middle-income nations are, on average, those that are
undergoing highest economic growth and the most drastic changes in their economic structures,
the evidence suggests that urbanization is related to the income growth, industrialization, and
modernization of an economy.
However, the speed of urbanization seems not to be explained solely by economic growth,
because many of nations that urbanized most rapidly, i.e. African nations among low-income
Journal of Development Economics 84 (2007) 76–103
www.elsevier.com/locate/econbase

Insightful comments and suggestions from two anonymous referees and an editor are gratefully appreciated. All
remaining errors are the author's own.

Tel.: +81 75 753 3532.
E-mail address: yuki@econ.kyoto-u.ac.jp.
0304-3878/$ - see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jdeveco.2006.09.004
nations and Latin American and Caribbean nations among middle-income nations, grew
relatively slowly (Mazumdar, 1987; Fay and Opal, 2000).
Behind the changes in the aggregate variable are migration decisions of rural residents, which
are mainly driven by economic factors, that is, better job opportunities in urban areas. However,
immigrants do not necessarily enjoy higher welfare in urban areas, due to the dualistic nature of
the urban economy in a typical developing country.
The urban economy of a typical underdeveloped nation can be classified into the formal or
modern sector and the informal or traditional sector.
1
The informal (traditional) sector is
characterized by small firm sizes, unskilled jobs, low wages, and the loose enforcement of laws
and regulations, and it incorporates sectors such as petty trading, domestic services, repair
services, and basic manufacturing. The formal (modern) sector, by contrast, possesses contrasting
features and includes modern manufacturing and skill-intensive services such as finance and
health care. According to Ranis and Stewart (1999), in most developing countries, the size of the
urban informal sector is substantial, in many cases accounting for over half the urban workforce.
Further, earnings of workers in the informal sector are close to or sometimes lower than those in
rural areas after adjustment for the cost of living.
2
The large share of the informal sector in the
urban economy implies that many immigrants do not attain noticeable improvements of their
economic conditions.
It seems that urbanization is associated with sizable job creation in the formal sector and thus
large welfare gains by immigrants occur, only if an economy achieves industrialization. For
example, the case study of Thailand and the Philippines by Ranis and Stewart (1999) shows that,
during the 1960s and the 1970s, both nations experienced large-scale urbanization accompanied
by rapid industrial growth, resulting in the expansion of the formal sector and the contraction of
the informal sector. After the 1980s, Thailand kept following this trend, while the Philippines
encountered urbanization with stagnant (almost zero) industrial growth and had to deal with the
rapid growth of the informal sector.
The question is why some economies attain urbanization accompanying the expansion of the
formal sector and why others urbanize without such modernization but with the expansion of the
informal sector. This paper presents a dynamic model of urbanization and development in order to
answer the question.
The model economy is inhabited by a continuum of individuals and is composed of the rural
agriculture sector, the urban informal (traditional) sector, and the urban formal (modern) sector.
The rural sector and the urban informal sector employ unskilled workers, whereas the urban
formal sector employs skilled workers and physical capital. The rural sector produces an
agricultural good, and the urban formal sector produces a manufacturing good, both of which are
tradables. The urban informal sector produces a nontradable good that would correspond to
services such as petty trading and domestic services in real economy.
3
The rural agriculture sector
1
Strictly speaking, the modern/traditional classification is based on production technologies, while the formal/informal
classification is based on establishment size or social security coverage, so they are distinct. In this paper, however, the
two taxonomies are used interchangeably, because the focus is on differences in skill requirements between the sectors, as
detailed below.
2
According to Jamal and Weeks (1988), for example, rural–urban income differentials are substantial in many sub-
Saharan African nations but the income differentials for unskilled workers are very small and, in some nations, their
urban incomes are actually lower than rural incomes.
3
Ranis and Stewart (1999) find that about 70% of the labor force in the informal sector in Manila in 1986–87 are
engaged in trading and only about 13% are in manufacturing. In Bangkok, the proportion of workers in the informal
sector engaged in manufacturing is larger but still only about 17%. Similar numbers are found for many Latin American
economies as well (Marcouiller et al., 1997).
77 K. Yuki / Journal of Development Economics 84 (2007) 76–103
exhibits diminishing returns to scale, reflecting the scarcity of land, and output is shared equally
among workers, while the urban sectors have constant returns to scale technologies and wages are
determined competitively. As different skill requirements of the formal and informal sectors
indicate, this paper focuses on the role of human capital accumulation in development. The
accumulation of physical capital is thus given only a minor role by assuming free international
capital mobility.
4
An individual in the model lives for two periods, the first period as a child and the second period
as an adult. In childhood, he or she receives a transfer from the parent to invest in assets or
education. Education is needed to become a skilled worker, is provided by teachers (skilled
workers), and must be financed by the received transfer. That is, there are no credit markets for such
investment in this economy. In adulthood, depending on human capital accumulation in the
previous period, the individual becomes a skilled or unskilled worker, and makes occupational and
associated locational choices. There are no barriers to interregional migration. Then, he or she
receives labor and capital incomes, which are spent on the consumption of the three goods and a
transfer to his or her sole child. The utility function is assumed to be Cobb–Douglas, hence the
share of spending on each itemis constant over time. Generations go by in this fashion. Individuals
of the same generation are heterogeneous in terms of received transfers, and this heterogeneity
translates into diverse choices of investments, occupations, and locations, and brings interesting
dynamics.
In the model economy there exist two types of steady states. The first type called the equal
opportunity steady state has features of a typical developed economy: many people are educated
and skilled; most people live in urban areas and are engaged in non-agricultural activities; and
inequalities between skilled and unskilled workers and between urban and rural residents are
small. (In the model, distributions of net wage and wealth are equalized.) The second type called
the unequal opportunity steady states, by contrast, characterizes a typical stagnant developing
economy: many people live in rural areas and are engaged in agriculture; many of the urban
residents are not skilled and remain in the informal sector; and the inequalities are substantial.
Given an initial distribution of transfers over the population, an economy approaches one of the
steady states in the long run. If an economy starting with limited asset accumulation is on a path to
the equal opportunity steady state, it experiences urbanization that accompanies skill upgrading,
the expansion of the formal sector, the shrinkage of the informal sector, and falling inequality.
Instead, if it is on a track to an unequal opportunity steady state, it urbanizes without skill
upgrading or modernization but with the expansion of the informal sector. Thus the model exhibits
two distinctive patterns of urbanization observed in real economy. The critical determinant of the
long-run success of an economy is the fraction of the population who can afford education in the
initial period, which reflects the degree of equality of initial asset distribution unless aggregate
asset accumulation is very low. This is consistent with empirical findings, for example, by
Deininger and Squire (1998) and Deininger and Olinto (2000), which show that an economy's
growth rate is affected negatively by initial land inequality (a proxy for initial asset inequality).
These results are obtained with fixed productivities. The paper also examines effects of one-
time productivity changes in the rural agriculture sector and the urban formal sector on the
urbanization and development of an economy. Both a decrease in the agricultural productivity and
an increase in the formal sector's productivity promote urbanization but make it more likely for an
4
The model becomes analytically intractable once the closed capital market is assumed. In considering the urbanization
of contemporary developing nations, this simplifying assumption would be defended, since the link between domestic
savings and investment is becoming weaker with growing international capital flows.
78 K. Yuki / Journal of Development Economics 84 (2007) 76–103
economy starting with an unfavorable wealth distribution to end up in an unequal opportunity
steady state. The result is suggestive since many sub-Saharan African nations actually expe-
rienced falling agricultural productivity,
5
rapid urbanization associated with the expansion of the
informal sector, and stagnant growth for many years. The effects of the productivity increase in
the formal sector suggest that skill-biased productivity growth is not necessarily beneficial for the
long-run development of a poor economy.
This paper is related to the literature that explores the mechanisms of urbanization and the
evolution of the informal sector. Many papers in this literature, including Fields (1975), Gupta
(1997), and Ranis and Stewart (1999), are some variants of the Harris and Todaro (1970) model in
the sense that the formal sector is characterized as the sector that pays higher wages to identical
workers than the informal sector does, due to exogenous institutional factors, such as entry
restrictions to the sector, compliance with minimum wage regulation, and the presence of trade
unions. By contrast, the present paper distinguishes the two sectors in terms of skill requirements
to workers. Empirical evidence supports this characterization: for example, Marcouiller et al.
(1997) found that workers in the formal sector are significantly more educated than workers in the
informal sector. Further, the conventional classification is at odds with some empirical findings:
according to Funkhouser (1997), while there exist unexplained sectoral wage differentials after
adjustment for workers' characteristics, the differentials seem not to be explained largely by the
presence of labor regulations or unions.
6
Of course, each characterization would capture a part of
reality and the present paper's skill-based classification reflects its emphasis on human capital as
the determinants of development and the regional and sectoral distributions of workers.
In this literature, Bencivenga and Smith (1997) and Banerjee and Newman (1998) are
particularly worth mentioning. Bencivenga and Smith (1997) constructs a two-period OLG model
with the rural, urban formal, and urban informal sectors and analyzes how migration,
unemployment (employment in the informal sector), physical capital accumulation, and economic
development are related. In their model, individuals are innately skilled or unskilled, but their
types are private information and an adverse selection problem arises. Unlike a conventional
Harris–Todaro model, this adverse selection, not exogenous institutional factors, gives rise to the
dual structure of the urban economy. In equilibria, only skilled workers migrate to the urban region,
while unskilled workers stay in the rural sector. Depending on parameter values, there exist
multiple steady state equilibria, where the high-income steady state is more urbanized and
industrialized but has a greater share of its urban workforce in the informal sector.
Banerjee and Newman (1998) examine implications of differences in technological and
institutional conditions between urban modern and rural traditional sectors in development. In the
model economy, individuals can move freely between the sectors and earn higher wages in the
more productive modern sector. This does not lead to full urbanization, however, since the rural
sector has one advantage: loans for fulfilling opportunities to consume indivisible goods are
available more easily, because the degree of information asymmetry between lenders and
borrowers is lower in the close-knit community. Hence, there exists a tradeoff between the better
access to credit in the traditional sector and the higher productivity in the modern sector, and thus
some of them stay in the rural sector. Under this setting, the degree of urbanization could be too
5
Fulginiti and Perri (1997) examine changes in agricultural productivity in 18 developing countries over the period
1961–1985 and find that at least half of these countries have experienced productivity declines in agriculture including all
the sub-Saharan African nations in the sample.
6
Also, Magnac (1991) constructs a Roy model to test the hypothesis of competitive labor market against that of
segmented market for married women in the urban areas of Columbia and does not reject the competitive hypothesis.
79 K. Yuki / Journal of Development Economics 84 (2007) 76–103
low: if some of those staying in the rural sector and receiving loans are forced to migrate, loans
available to the urban sector increase and the interest rate falls, which in turn promotes migration
and raises total output.
The present paper is also related to the literature that investigates the interplay between income
distribution and growth through human capital accumulation, which includes Galor and Zeira
(1993), Ljungqvist (1993), Benabou (1996a,b), Durlauf (1996), Galor and Tsiddon (1997), and
Yuki (2003). Closely related are Galor and Zeira (1993), Ljungqvist (1993), and Yuki (2003).
Galor and Zeira (1993) considers a small-open OLG model, where, as in the present paper,
intergenerational transfers motivated by impure altruism (utility from making transfers) constrain
educational decisions, while capital is freely mobile internationally. There are two sectors, one
using skilled labor and physical capital and the other employing unskilled labor, to produce a single
final good. They examine how initial wealth distribution affects the dynamics of the model
economy. Ljungqvist (1993) examines a dynamic model with a single production sector where
individuals are infinitely lived (or have perfect altruism), the final good is produced by skilled
labor, unskilled labor, and physical capital, and education is required to remain skilled for a fixed
units of time. He shows that there exists a continuum of steady states that are different in terms of
the proportion of skilled workers and wage inequality. In these papers, wages are determined by
supply factors, whereas, in the present paper, unskilled wage is affected by demand factors as well.
Banerjee and Newman (1993) examine the interaction between occupational choices and the
distribution of wealth in the process of development. There exist three kinds of occupations –
self-employed workers, entrepreneurs, and wage earners – in the economy. Due to credit market
imperfections, certain amounts of wealth are required to become an entrepreneur or a self-
employed worker, and thus occupational patterns are affected by the distribution of wealth.
Occupational patterns in turn affect future wealth distributions and their interaction determines the
fate of the economy.
Yuki (2003) constructs a similar model with the tradable modern sector and the nontradable
traditional sector. However, its focus is on the mechanism by which initial distribution affects the
long-run outcome of an economy and the effect of the sectoral shift of consumption associated
with income growth (Engel's law) on development.
This paper is organized as follows. Section 2 presents the model economy and derives
equilibrium conditions. Section 3 analyzes the model's dynamics and Section 4 presents and
interprets the results. Section 5 concludes the paper.
2. Model
The model economy is of a discrete-time OLG variety. The economy is composed of a
continuum of individuals who live for two periods, the first period as children and the second
period as adults. There are two regions, urban and rural, and, as detailed below, each region has
different production sectors and available jobs. This is the only difference between the regions in
the model: there are free mobility of goods and factors of production; hence prices are equalized
between the regions. There is no uncertainty in the model.
2.1. Lifetime of an individual
2.1.1. Childhood
In childhood, an individual receives a transfer from his or her parent. Then he or she
allocates the transfer for two investment options, assets and education, in order to maximize
80 K. Yuki / Journal of Development Economics 84 (2007) 76–103
future income.
7
The educational investment is required to become a skilled worker and enjoy
higher earnings in adulthood. The investment is a discrete choice, making education optional
and incurring a fixed cost. Consider an individual who was born into lineage i in period t −1.
His or her generation is called generation t. Then, his or her education costs e
t
, and its gross
return is w
h,t
−w
l,t
in the next period, where w
h,t
and w
l,t
are skilled and unskilled wages in
period t, respectively. Assume that the education cost is the cost of hiring current skilled
workers as teachers and it is proportional to w
h,t −1
:e
t
=s
e
w
h,t −1
.
8
The investment must be
self-financed because loan markets for such investment are not available: the child's future
income is not considered a valid collateral in the financially underdeveloped economy. The
other option, the investment in assets, is a continuous choice, and brings a gross rate of
return of 1+r
t
. It is easily shown that, in an equilibrium, the return from the investment in
education becomes at least as high as the return from the investment in assets, i.e. w
h,t
−w
l,t

(1+r
t
)e
t
.
Suppose that the individual has received b
t
i
units of income as a transfer from the parent. He
or she allocates the transfer between the investments in assets a
t
i
and in education e
t
i
in order to
maximize future income. If the return from the educational investment is strictly higher than
the one from the asset investment, optimal investment choices are given by the following
equations:
9
a
i
t
¼ b
i
t
; e
i
t
¼ 0; if b
i
t
be
t
; ð1Þ
a
i
t
¼ b
i
t
−e
t
; e
i
t
¼ e
t
; if b
i
t
ze
t
: ð2Þ
Since innate abilities of individuals are identical, transfers solely determine the investment and
resulting occupational choices.
2.1.2. Adulthood
At the beginning of adulthood, an individual makes occupational and associated locational
choices. Available jobs are constrained by the human capital investment in the previous period.
Then, the individual obtains income from assets and labor supply and spends the income on
consumption and a transfer to his or her child. Each adult is assumed to have a single child. There
are three different consumption goods, goods uh, ul, and rl in the economy. Characteristics of
these goods are described later in this section. The utility function of agent i in generation
7
Alternatively, one can suppose that the investment decisions are carried out by the parent in order to maximize the
child's future income. Note that the transfer in the model corresponds to total intergenerational transfers including
bequests, education, and other inter-vivos transfers in real life. The decision that the child has to make is the allocation of
the whole transfers between education and assets.
8
World Bank (1983) reports that about 95% of current expenses in primary school systems of low income countries are
teacher salaries.
9
Actually the relative return from education is determined as a result of people's investment decisions, since it depends
on the numbers of skilled and unskilled workers in the economy. More formal analysis of the investment decision is
described in Section 3.1.
81 K. Yuki / Journal of Development Economics 84 (2007) 76–103
t depends on the consumption of the three goods c
j,t
i
, j =uh, ul, and rl, and the transfer to his or her
child b
t+1
i
:
u
i
t
¼ j
j
ðc
i
j;t
Þ
g
j
ðb
i
tþ1
Þ
g
k
;

j
g
j
þ g
k
¼ 1: ð3Þ
The individual maximizes the utility function subject to the budget constraint,

j
p
j;t
c
i
j;t
þ b
i
tþ1
¼ w
i
t
þ ð1 þ r
t
Þa
i
t
; ð4Þ
where p
j,t
is the price of good j, w
t
i
is the wage of the individual, r
t
is the interest rate, and a
t
i
is his
or her asset holding in period t. Solving the maximization problem gives the following
consumption and transfer rules:
p
j;t
c
i
j;t
¼ g
j
fw
i
t
þ ð1 þ r
t
Þa
i
t
g; ð5Þ
b
i
tþ1
¼ g
k
fw
i
t
þ ð1 þ r
t
Þa
i
t
g: ð6Þ
2.1.3. Generational change
At the beginning of period t +1, current adults pass away, current children become adults, and
new children are born into the economy. Since each adult has one child, the population is constant
over time. The population of each generation is normalized to be one.
2.2. Production
The urban region has two production sectors, sector uh and sector ul. Sector uh employs skilled
workers and physical capital to produce good uh, and sector ul employs unskilled workers to produce
good ul. The former sector may be considered as the urban formal sector such as modern manufac-
turing and the latter sector as the urban informal sector such as petty trading, personal services, and
repair services. Good uh serves as a capital good as well. In contrast, the rural region has sector rl that
employs unskilled workers to produce good rl. This sector corresponds to agriculture in real economy.
The production functions for sectors uh, ul, and rl in period t are given by
Y
uh;t
¼ ðA
uh
H
u;t
Þ
a
uh
ðK
t
Þ
1−a
uh
; ð7Þ
Y
ul;t
¼ A
ul
L
u;t
; ð8Þ
Y
rl;t
¼ A
rl
ðL
r;t
Þ
a
rl
; ð9Þ
0ba
uh
b1; 0Va
rl
b1; ð10Þ
where A
j
( j =uh, ul, and rl) is the productivity in sector j; H
u,t
is the number of skilled workers
employed in sector uh;
10
L
u,t
is the number of urban unskilled workers (employed in sector ul); L
r,t
is the number of rural unskilled workers (employed in sector rl); and K
t
is the amount of physical
capital employed in sector uh. Sector rl exhibits decreasing returns to its only input L
u,t
, which
intends to capture the fact that agricultural productivity is constrained by limited arable land.
10
The rest of skilled workers is employed in the education sector. It is assumed that the sector is located in the urban area.
82 K. Yuki / Journal of Development Economics 84 (2007) 76–103
2.3. Equilibrium
Assume that goods uh and rl are tradables and their prices are determined in international
markets. Remember that goods uh and rl correspond roughly to manufacturing goods and
agricultural goods, respectively in real economy. Normalize the price of good uh to be 1, and
denote the price of good rl by p
rl
. By contrast, good ul, which may be considered as service
goods produced with technologies intensive in unskilled labor, is assumed to be non-tradable,
thus its price is determined in the domestic market.
As for factors of production, capital is assumed to be freely mobile internationally. The
assumption would be more realistic than the other extreme of the closed market, and enables the
paper to focus on human capital accumulation rather than physical capital accumulation as the
primary source of development. Denote the exogenously determined interest rate by r. Wage rates
are determined competitively in urban sectors uh and ul, and in the education sector. By contrast,
in the rural sector rl, where the production technology exhibits decreasing returns, labor income is
determined so that the product is equally shared among workers, as is assumed in Lewis (1954).
11
Thus, the average productivity of labor is higher than the marginal productivity in this sector, i.e.
there exists disguised unemployment. Since the wage rate of the sector, p
rl
A
rl
(L
r,t
)
α
rl
− 1
,
approaches infinity as L
r,t
goes to 0, good rl is always produced in the economy.
The wage of skilled workers is obtained by solving the profit maximization problem of a firm
in sector uh:
12
w
h
¼ a
uh
A
uh
1−a
uh
r
_ _
1−a
uh
a
uh
: ð11Þ
The wage rate w
h
is exogenously determined. Since the cost of education e
t
is the cost of hiring
skilled workers as teachers, it is exogenous as well, e
t
=e(=s
e
w
h
).
13
The wage rate of unskilled
workers must satisfy the following equations:
w
l;t
¼ p
ul;t
A
ul
ð12Þ
¼ p
rl
A
rl
ðL
r;t
Þ
a
rl
−1
: ð13Þ
11
This assumption reflects the fact that traditional agriculture in developing economies is typically engaged by
collectives or family farms. After a minor modification, the assumption is also compatible with share-cropping systems
widely observed in Asian countries.
12
The profit maximization problem gives the following first order conditions:
r ¼ ð1−a
uh
ÞðA
uh
H
u;t
Þ
a
uh
ðK
t
Þ
−a
uh
;
w
h;t
¼
a
uh
H
u;t
A
uh
H
u;t
Þ
_
a
uh
K
t
Þ ð
1−a
uh
;
Solving the first equation for K
t
and substituting it into the second equation gives Eq. (11).
13
The market clearing condition of the education sector is given as follows:
w
h
H
e;t
¼ eH
tþ1
;
or H
e;t
¼ s
e
H
tþ1
;
where H
e,t
is the number of ‘teachers’ in the education sector and H
t+1
is the number of ‘students’ in period t. Assume
that the education technology is reasonably efficient, that is, s
e
is not very high. Then the equation is satisfied without
rationing; that is, all children who want to receive and can afford education are able to find their teachers.
83 K. Yuki / Journal of Development Economics 84 (2007) 76–103
The wage rates in the two regions are equalized, because there is free labor mobility between the
regions, and there are no mobility costs of goods or factors of production. The unskilled wage is,
in general, not constant over time.
The price of non-tradable ul, p
ul,t
, is determined domestically. The market clearing condition is
14
p
ul;t
A
ul
L
u;t
¼ g
ul
w
h
H
t
þ w
l;t
ð1−H
t
Þ þ ð1 þ rÞ

i
a
i
t
_ _
: ð14Þ
In this equation, H
t
is the total number of skilled workers and ∑
i
a
t
i
is the aggregate asset.
15
The right-
hand side of the equation is obtained by aggregating the consumption rule for good ul (Eq. (5)) over
the population. Let B
t
be aggregate intergenerational transfers. Then ∑
i
a
t
i
=B
t
−eH
t
is satisfied, since
currently skilled workers have paid the cost of education e out of transfers received from their
parents. Substituting ∑
i
a
t
i
=B
t
−eH
t
and Eq. (12) into the above equation and solving for p
ul,t
A
ul
, one
can obtain
p
ul;t
A
ul
¼
g
ul
L
u;t
−g
ul
ð1−H
t
Þ
fw
h
H
t
þ ð1 þ rÞðB
t
−eH
t
Þg: ð15Þ
The price level p
ul,t
is increasing in H
t
and B
t
, and decreasing in L
u,t
. Further, fromthe substitution of
Eq. (13) and L
u,t
=1−H
t
−L
r,t
into the above equation, the number of unskilled workers in the rural
region, L
r,t
, is determined:
p
rl
A
rl
ðL
r;t
Þ
a
rl
−1
¼
g
ul
f½w
h
−ð1 þ rÞeŠH
t
þ ð1 þ rÞB
t
g
ð1−g
ul
Þð1−H
t
Þ−L
r;t
: ð16Þ
The left-hand side of the above equation is decreasing in L
r,t
, and the right-hand side is increasing in
L
r,t
. Further, LHSNRHS, as L
r,t
→0, and LHSbRHS, as L
r,t
→(1−γ
ul
)(1−H
t
). Hence, the existence
and uniqueness of L
r,t
as the solution for the above equation is assured.
16
In order to indicate the
14
It is assumed that the urban sector ul supplies good ul to the both regions, which might seem unnatural considering the
fact that good ul includes services such as petty trading and personal services in real economy. The assumption is made just
for simplicity: all the results in later sections except those on sectoral allocations of workers remain intact with the
alternative assumption that good ul consumed by individuals in each region is supplied by the nontradable sector located in
the region, although the analysis becomes significantly more complicated. As for sectoral allocations of workers, as
detailed in footnotes 19, 26, and 29 in Section 4, most of the qualitative results go through under the alternative assumption.
15
Due to free international capital mobility, a large portion of the aggregate asset may have been invested abroad, if
there do not exist enough investment opportunities within the economy.
16
If unskilled wage in the rural region is determined competitively, i.e. w
l,t

rl
p
rl
A
rl
(L
r,t
)
α
rl
−1
, L
r,t
is the solution to the
following equation:
p
rl
A
rl
ðL
r;t
Þ
a
rl
−1
¼
1
a
rl
g
ul
f½w
h
−ð1 þ rÞeŠH
t
þ ð1 þ rÞB
t
g
ð1−g
ul
Þð1−H
t
Þ− 1 þ
1−a
rl
a
rl
g
ul
_ _
L
r;t
: ð17Þ
By comparing the above equation with Eq. (16), it is clear that L
r,t
is smaller under the competitive rural labor market.
This implies that, in terms of static output maximization, there are too many farmers (too few urban workers) under the
income sharing. However, the competitive labor market does not necessarily lead to higher aggregate output in the longer
term, since, as is detailed in later sections, the long-run development of the economy crucially depends on income levels
of unskilled workers. If the profit share in the rural agriculture sector, 1−α
rl
, is high or skilled workers take most of
profits from the sector, unskilled workers might have lower incomes in the competitive market and thus long-run
aggregate output could be lower.
84 K. Yuki / Journal of Development Economics 84 (2007) 76–103
dependence of L
r,t
on H
t
and B
t
, let L
r,t
=L
r
(H
t
,B
t
), which is a decreasing function of H
t
and B
t
. Then,
from Eq. (13), w
l,t
=w
l
(H
t
,B
t
) is an increasing function of H
t
and B
t
; and from Eq. (12),
p
ul,t
=p
ul
(H
t
,B
t
) is also an increasing function of H
t
and B
t
. These results are summarized in
the following lemma.
Lemma 1.
ðiÞ
AL
r
ðH
t
; B
t
Þ
AH
t
b0;
AL
r
ðH
t
; B
t
Þ
AB
t
b0: ð18Þ
ii ð Þ
Ap
ul
ðH
t
; B
t
Þ
AH
t
N0;
Ap
ul
ðH
t
; B
t
Þ
AB
t
N0: ð19Þ
iii ð Þ
Aw
l
ðH
t
; B
t
Þ
AH
t
N0;
Aw
l
ðH
t
; B
t
Þ
AB
t
N0: ð20Þ
3. Dynamics
The previous section has presented the model and derived the equilibrium conditions. In the
model economy, individuals live only for two periods and take part in each market for one period
alone, thus each market is populated by individuals of a single generation every period. Hence,
the model economy can be considered as a sequence of static economies.
What connects these static economies across periods are intergenerational transfers. Because
of the credit constraint, transfers directly affect individuals' investment and occupational choices,
and consequently consumption and transfer decisions. Further, the distribution of transfers over
the population determines a proportion of individuals who can afford education, and thus it affects
the relative return from education and investment decisions. The distribution also affects supplies
and demands of goods and their prices and allocations. Hence, in general, the time evolution of
the distribution of transfers must be examined in order to understand how the structure of the
economy, such as regional and sectoral distributions of workers, production and employment
shares of each sector, and distributions of earnings and assets, changes over time.
This section first derives the dynamic equation relating the current transfer to the next
period's transfer within a lineage (individual dynamics). The dynamics depend on the time
evolution of two aggregate variables that in turn are determined by the dynamics of the
distribution of transfers. However, it turns out that directions of motion of the two aggregate
variables, which can be obtained without knowledge on the distributional dynamics, is enough
to derive the model's implications. Hence the dynamics of the aggregate variables are
characterized next. Although the two dynamics interact, for exposition, initially the dynamics
of each variable are analyzed fixing the other, then the both dynamics are analyzed together by
introducing a phase diagram.
3.1. Individual dynamics
Consider an individual born into lineage i in period t −1, who belongs to generation t. He or
she allocates the transfer b
t
i
between investments in assets a
t
i
and in education e
t
i
so as to
maximize future income. If the transfer is less than the cost of education, i.e. b
t
i
be, the transfer is
spent only on assets and he or she becomes an unskilled worker, as described above. On the other
hand, if the transfer is at least as large as the cost of education, i.e. b
t
i
≥e, the investment decision
85 K. Yuki / Journal of Development Economics 84 (2007) 76–103
is more complicated. Since investment decisions of others affect unskilled wage w
l
(H
t
,B
t
) and the
relative return from education, the individual has to take into account their actions in making the
decision. The key variable affecting the decision is the fraction of individuals in generation t who
have received transfers b
t
i
larger than e, Fr
t
. In short, when only a small number of individuals
can afford education, all of them take education and become skilled, and when many individuals
have access to education, not all of them become skilled workers.
Unequal opportunity case: More precisely, when the proportion of individuals who can afford
education is small, the return from education is higher than the return from assets, even if all of
them actually take education, i.e. w
h
−(1+r)eNw
l
(Fr
t
,B
t
). In this case, the cut-off level of transfer
b
t
i
=e determines whether an individual becomes a skilled worker or an unskilled worker, and
H
t
=Fr
t
is satisfied. This case is named the unequal opportunity case since access to the better
investment opportunity, education, is constrained by received transfers. For a currently skilled
worker, i.e. those who have received b
t
i
≥e, the dynamic equation linking the received transfer b
t
i
to the transfer given to the next generation b
t+1
i
is
b
i
tþ1
¼ b
s
ðb
i
t
Þug
k
fw
h
þ ð1 þ rÞðb
i
t
−eÞg: ð21Þ
The equation is derived by substituting w
t
i
=w
h
, a
t
i
=b
t
i
−e, and r
t
=r into the transfer rule (6). The
assumption γ
k
(1+ r) b1 is made so that the fixed point for the equation, ðb
s
Þ*u
g
k
1−g
k
ð1þrÞ
½w
h
−ð1 þ rÞeŠ, exists.
Similarly, for a currently unskilled worker, i.e. those who have received b
t
i
be, the corre-
sponding dynamic equation is
b
i
tþ1
¼ b
u
ðb
i
t
; Fr
t
; B
t
Þug
k
fw
l
ðFr
t
; B
t
Þ þ ð1 þ rÞb
i
t
g: ð22Þ
The equation is obtained from the substitution of w
t
i
=w
l
(Fr
t
,B
t
), a
t
i
=b
t
i
, and r
t
=r into
Eq. (6). The fixed point of the equation for given Fr
t
and B
t
is denoted by b
*
u
ðFr
t
; B
t
Þu
g
k
1−g
k
ð1þrÞ
w
l
ðFr
t
; B
t
Þ.
17
Equal opportunity case: In contrast, when the return from education fails to be higher than the
return from assets, if all of them invest in education, i.e. w
h
−(1+r)e≤w
l
(Fr
t
,B
t
), the number of
skilled workers H
t
is determined at the point where the two returns are equated, that is, H
t
is the
solution of w
h
−(1+r)e=w
l
(H
t
,B
t
). Now not all of financially eligible individuals take education
and become skilled workers, i.e. H
t
≤Fr
t
. The dynamics of transfers of both skilled and unskilled
workers are described by the single equation b
t+1
i
=b
s
(b
t
i
) (Eq. (21)).
The economy belongs to one of the two cases depending on levels of two aggregate variables,
Fr
t
and B
t
. The dividing line separating the two cases on the (Fr
t
,B
t
) space is derived in the
following lemma.
Lemma 2. The dividing line between the unequal opportunity case and the equal opportunity
case on the (Fr
t
, B
t
) space is
B
t
¼ DðFr
t
Þu
1
g
ul
ð1 þ rÞ
1−g
ul
−Fr
t

p
rl
A
rl
w
h
−ð1 þ rÞe
_ _ 1
1−a
rl
_ _
½w
h
−ð1 þ rÞeŠ: ð23Þ
17
In general, this is not the long-run transfer level of a lineage of a currently unskilled worker, since his or her
descendants may become skilled workers and Fr
t
and B
t
may change over time. Although this fixed point may appear to
have no economic importance, it turns out that the level of b
u

(Fr
t
,B
t
) is crucial for aggregate dynamics (detailed later).
86 K. Yuki / Journal of Development Economics 84 (2007) 76–103
Proof. In the Appendix. □
It is assumed 1−g
ul

P
rl
A
rl
w
h
−ð1þrÞe
_ _
1
1−a
rl
N0, which ensures that, for some combinations of positive
Fr
t
and B
t
,w
h
−(1+r)e Nw
l
(Fr
t
,B
t
) is satisfied and thus individuals with b
t
i
≥e become skilled
workers and all the goods are produced. The economy belongs to the equal opportunity case
when Fr
t
or B
t
is large enough that B
t
≥D(Fr
t
) is satisfied. Based on the above lemma, the
number of urban unskilled workers and the number of skilled workers in the equal opportunity
case are obtained.
Lemma 3. In the equal opportunity case, i.e. B
t
≥D(Fr
t
),
L
r;t
¼
p
rl
A
rl
w
h
−ð1 þ rÞe
_ _ 1
1−a
rl
; ð24Þ
H
t
¼ D
−1
ðB
t
Þ ¼ 1−g
ul

p
rl
A
rl
w
h
−ð1 þ rÞe
_ _ 1
1−a
rl

g
ul
ð1 þ rÞB
t
w
h
−ð1 þ rÞe
: ð25Þ
Proof. In the Appendix. □
3.2. Aggregate dynamics
The previous subsection has shown that the individual dynamics depend on the dynamics of
two aggregate variables, aggregate transfers B
t
and the fraction of individuals who have
received transfers b
t
i
≥e, Fr
t
. These dynamics in turn depend on the time evolution of the
distribution of transfers, which cannot be characterized without imposing strong assumptions
on the distribution. However, in order to derive the model's implications, the information on
directions of change of the two aggregate variables is sufficient. Thus this subsection analyzes
the dynamics of the two aggregate variables qualitatively. For exposition, initially each of them
is examined separately fixing the other variable, then their interaction is taken into account at
the end.
3.2.1. Dynamics of B
t
The dynamic equation linking current aggregate transfers B
t
to transfers in the next period B
t+1
is derived by aggregating the individual dynamics of skilled workers (Eq. (21)) and of unskilled
workers (Eq. (22)) over the population:
B
tþ1
¼ g
k
½w
h
H
t
þ w
l
ðH
t
; B
t
Þð1−H
t
Þ þ ð1 þ rÞðB
t
−eH
t
ފ: ð26Þ
Remember that in the unequal opportunity case, i.e. w
h
−(1+r)eNw
l
(Fr
t
,B
t
), H
t
can be replaced
by Fr
t
in the above equation. By contrast, in the equal opportunity case, w
h
−(1+r)e=w
l
(H
t
,B
t
) is
satisfied. Hence the dynamic equations are expressed as,
B
tþ1
¼ g
k
f½w
h
−ð1 þ rÞeŠFr
t
þ w
l
ðFr
t
; B
t
Þð1−Fr
t
Þ þ ð1 þ rÞB
t
g; for B
t
bDðFr
t
Þ; ð27Þ
B
tþ1
¼ g
k
½w
h
−ð1 þ rÞe þ ð1 þ rÞB
t
Š; for B
t
zDðFr
t
Þ: ð28Þ
87 K. Yuki / Journal of Development Economics 84 (2007) 76–103
Aggregate transfers, when the other aggregate Fr
t
is fixed, are shown to converge to particular
levels of aggregate transfers, as is illustrated graphically in Fig. 1:
Lemma 4. Assume
g
k
ð1þrÞ
1−g
ul
b1 and let Fr
t
be fixed.
(i) If Fr
t
bH
**
u1−
g
ul
1−g
k
ð1þrÞ

P
rl
A
rl
w
h
−ð1þrÞe
_ _ 1
1−a
rl
, aggregate transfers converge to B*(Fr
t
), which is a
solution to the following equation:
B ¼ g
k
f½w
h
−ð1 þ rÞeŠFr
t
þ w
l
ðFr
t
; BÞð1−Fr
t
Þ þ ð1 þ rÞBg: ð29Þ
Further,
AB*ðFr
t
Þ
AFr
t
N0 is satisfied.
(ii) If Fr
t
≥H**, aggregate transfers converge to
B
**
u
g
k
1−g
k
ð1 þ rÞ
½w
h
−ð1 þ rÞeŠ: ð30Þ
Proof. In the Appendix. □
3.2.2. Dynamics of Fr
t
The dynamics of Fr
t
, the fraction of the adult population who had received enough wealth to
take education, depend on the dynamics of intergenerational transfers of skilled and unskilled
workers. For skilled workers, the following result holds.
Lemma 5. Assume γ
k
Ns
e
. Then, descendants of currently skilled workers can afford education.
Proof. It is enough to show that the child of a currently skilled worker can afford education. From
Eq. (21), b
t+1
i
≥γ
k
w
h
Ns
e
w
h
=e. □
The assumption implies that a parent has reasonably strong altruism towards his or her child
and the education technology is not very inefficient. From the lemma, Fr
t
is non-decreasing over
Fig. 1. Dynamics of aggregate transfers for fixed Fr.
88 K. Yuki / Journal of Development Economics 84 (2007) 76–103
time. The question is whether Fr
t
remains constant or increases over time, and the answer
depends on the level of transfers at the fixed point of the dynamic equation of unskilled workers,
b

u
ðFr
t
; B
t
Þu
g
k
1−g
k
ð1þrÞ
w
l
ðFr
t
; B
t
Þ, relative to the cost of education e:
Lemma 6.
If b

u
ðFr
t
; B
t
Þu
g
k
1−g
k
ð1 þ rÞ
w
l
ðFr
t
; B
t
ÞVe;
() B
t
V
½1−g
k
ð1 þ rފe
g
k
g
ul
ð1 þ rÞ
 ð1−g
ul
Þð1−Fr
t
Þ−
p
rl
A
rl
e
g
k
1−g
k
ð1 þ rÞ
_ _
1
1−a
rl
_ _
ð31Þ

1
1 þ r
½w
h
−ð1 þ rÞeŠFr
t
; ð32Þ
then Fr
t+1
=Fr
t
.
Proof. In the Appendix. □
From the lemma, Fr
t+1
NFr
t
happens only when b
u

(Fr
t
,B
t
) Ne, although Fr
t+1
=Fr
t
is possible
if the distribution of transfers {b
t
i
} is such that there do not exist unskilled workers satisfying b
t
+1
i
≥e. However, if b
u

(Fr
t
,B
t
) Ne continues to hold for many periods, lineages of current unskilled
workers accumulate wealth gradually and Fr
t
starts to increase eventually (see Fig. 2).
18
3.2.3. Joint dynamics of B
t
and Fr
t
Now the dynamics of the two aggregate variables are considered together by adding directions
of motion of Fr
t
to Fig. 1 that illustrates the dynamics of B
t
for given Fr
t
. To do so, one more result
regarding the relative location of the two critical loci, B
t
=D(Fr
t
) and b
u

(Fr
t
,B
t
) =e, is needed.
18
To be more accurate, the position of b
u

(Fr
t
,B
t
) on the figure moves as Fr
t
and B
t
change.
Fig. 2. Necessary condition for Fr
t +1
NFr
t
.
89 K. Yuki / Journal of Development Economics 84 (2007) 76–103
Lemma 7. b
u

(Fr
t
, B
t
) =e is located below B
t
=D(Fr
t
).
Proof. In the Appendix. □
Fig. 3 is the resulting phase diagram describing the joint dynamics of B
t
and Fr
t
. In the figure,
feasible combinations of Fr and B must be in the area bound by Fr =0, Fr =1, and B=eFr. The
economy must satisfy B≥eFr because Fr is defined as the fraction of adult individuals who had
received transfers b
i
larger than the cost of education e in their childhoods. Remember that the
region below the locus B=D(Fr) is the unequal opportunity case and H=Fr is satisfied, and the
region on or above the locus is the equal opportunity case and H=D
−1
(B) ≤Fr is true. Directions
of change of B and Fr are expressed with vertical arrows and horizontal arrows, respectively. In
the unequal opportunity case, the direction of motion of B is determined by the position of current
(Fr,B) relative to B=B

(Fr), and in the equal opportunity case, it is determined by the current
position relative to B=B
⁎⁎
. As for the direction of change of Fr, it is determined by the current
location of (Fr,B) relative to b
u

(Fr,B) =e. Denote Fr at the intersection of b
u

(Fr,B) =e and B=
B

(Fr) by Fr

. In the next section, main results of the model are presented with the help of this
diagram.
4. Analysis
4.1. Steady states and transitional dynamics
The inspection of Fig. 3 reveals that there exist two types of steady states in this economy,
(Fr
ss
,B
ss
) =(1,B
⁎⁎
) and (Fr
ss
,B
ss
) =(Fr,B

(Fr)), where Fr ≤Fr

is satisfied. In the former steady
Fig. 3. Phase diagram.
90 K. Yuki / Journal of Development Economics 84 (2007) 76–103
state, the number of skilled workers is H
⁎⁎
=D
−1
(B
⁎⁎
), skilled wage (net of the cost of education)
is equal to unskilled wage, and all the individuals hold the same level of wealth. By contrast, in
the latter type of steady states, the number of skilled workers is Fr (≤Fr

bH
⁎⁎
), the net wage
and wealth of skilled workers are higher than the corresponding ones of unskilled workers.
Further, as for the distribution of the population over the rural and urban areas in the steady states,
the following lemma is obtained.
Lemma 8. (i) In the steady state (Fr
ss
, B
ss
) =(1,B
⁎⁎
), the number of skilled workers and the
numbers of rural and urban unskilled workers are
H
⁎⁎
¼ 1−
g
ul
1−g
k
ð1 þ rÞ

p
rl
A
rl
w
h
−ð1 þ rÞe
_ _ 1
1−a
rl
; ð33Þ
L
r
ðH
⁎⁎
; B
⁎⁎
Þ ¼
p
rl
A
rl
w
h
−ð1 þ rÞe
_ _ 1
1−a
rl
; ð34Þ
L
u
ðH
⁎⁎
; B
⁎⁎
Þ ¼
g
ul
1−g
k
ð1 þ rÞ
: ð35Þ
(ii) In the steady state (Fr
ss
,B
ss
) =(Fr,B

(Fr)), Fr ≤Fr

, the number of unskilled workers in the
urban area is
L
u
ðFr; B

ðFrÞÞ ¼
g
ul
1−g
k
ð1 þ rÞ
1 þ
½w
h
−ð1 þ rÞeŠ−w
l
ðFr; B

ðFrÞÞ
w
l
ðFr; B

ðFrÞÞ
Fr
_ _
: ð36Þ
Proof. In the Appendix. □
Based on this lemma, the following proposition compares the different steady states in terms of
the distributions of the population over the two areas and the sectoral compositions of production.
19
Proposition 1. (i) The equal-opportunity steady state, (Fr
ss
,B
ss
) =(1,B
⁎⁎
), is more urbanized and
has a smaller informal sector in the urban area than the unequal opportunity steady states, (Fr
ss
,
B
ss
) =(Fr,B*(Fr)), Fr ≤Fr

:
L
r
ðH
⁎⁎
; B
⁎⁎
ÞbL
r
ðFr; B

ðFrÞÞ; ð37Þ
and L
u
ðH
⁎⁎
; B
⁎⁎
ÞbL
u
ðFr; B

ðFrÞÞ; for Fr VFr
y
: ð38Þ
(ii) The outputs of goods rl and ul are lower and the output of good uh is higher in the equal-
opportunity steady state in comparison to the unequal-opportunity steady states.
(iii) Among the unequal opportunity steady states, one with higher Fr is more urbanized (has
smaller L
r
(Fr,B

(Fr))) and has a smaller informal sector relative to the formal sector in the urban
region (
L
u
L
u
þFr
is lower).
19
The results remain unchanged in the modified model with two regional nontradable sectors supplying good ul
(footnote 14) by defining L
u
to be the total number of unskilled workers in the two nontradable sectors and L
r
to be the
number of unskilled workers in the rural tradable sector. (This is also true for Propositions 3 and 4 in the next subsection.)
Further, it can be shown that the number of unskilled workers in each nontradable sector is smallest at the equal
opportunity steady state and, among unequal opportunity steady states, the number of total rural workers and the
proportion of informal workers in the urban region decrease with Fr. That is, the qualitative results of the original model
are obtained in the modified model as well.
91 K. Yuki / Journal of Development Economics 84 (2007) 76–103
Proof. In the Appendix. □
The equal-opportunity steady state has features of a typical developed economy: many people are
educated and skilled; most of the people live in urban areas and are engaged in non-agricultural
activities; and inequality between skilled and unskilled workers is small (in the model, the
distributions of net wage and wealth are equalized). In contrast, the unequal-opportunity steady
states characterize a typical stagnant developing economy: many people live in rural areas and are
engaged in agriculture; many of the urban residents are not skilled and in the informal sector; and
inequality between skilled and unskilled workers is large. Among such stagnant economies, one with
a greater portion of skilled workers is more urbanized, less dependent on agriculture, has a smaller
informal sector relative to the formal sector in the urban region, accumulates larger wealth, and
attains lower inequality between skilled and unskilled workers because of higher unskilled wage.
20
Starting fromthe initial distribution of transfers over the population, the economy approaches one
of the steady states in the long run. FromFig. 3, it can be seen that the critical determinant of success
and failure of the economy is Fr
0
, the initial proportion of individuals who can afford education:
Proposition 2. Assume B
0
bB

(Fr
0
).
(i) If Fr
0
≤Fr

, the economy converges to the steady state (Fr
ss
,B
ss
) =(Fr
0
,B

(Fr
0
)) in the long
run. On the transitional path, Fr =H remains constant and B increases, thus L
r
(H,B)
decreases and L
u
(H,B) and w
l
(H,B) increase over time.
(ii) If Fr
0
NFr

, the economy converges to the steady state (Fr
ss
,B
ss
) =(1,B
⁎⁎
) in the long run.
On the transitional path, (a) in the region b
u

(Fr,B) ≤e, the results in (i) hold; ( b) in the
region where b
u

(Fr,B) Ne and BbD(Fr) are satisfied, both Fr and B increase, thus L
r
(H,B)
decreases and H and w
l
(H,B) increase over time; and (c)in the region B≥D(Fr), both Fr
and B increase, but this time L
r
(H,B) and w
l
(H,B) remain the same, while H decreases and
L
u
(H,B) increases over time.
Intuitively, the reason why Fr
0
is critical for the long-run performance of an economy can be
explained as follows.
21
In this model, the number of skilled workers increases, if a portion of
unskilled workers accumulates wealth sufficient for their children to take education and get skilled
jobs. Unskilled wage increases as the degree of urbanization rises through the migration of
unskilled workers from ‘over-populated’ agriculture (sector rl) to the urban informal sector, which
is possible if the price of the non-tradable good produced in the urban informal sector (good ul)
goes up. The price of the good, which is determined jointly with the number of rural unskilled
workers, depends positively on the number of skilled workers and aggregate assets: a greater
number of skilled workers implies the higher demand and lower supply of the good, and larger
asset accumulation leads to the higher demand. Thus, for the economy to initiate ‘modernization’,
the proportion of people who have enough wealth to take education and aggregate assets must be
above certain levels, i.e. b
u

(Fr,B) Ne.
If the initial distribution of wealth is such that a relatively large portion of the population can
afford education and becomes skilled workers, the price of good ul is higher due to the higher
20
Formally, from Lemmas 1(iii) and 4(i),
dw
l
ðFr
t
; B

ðFr
t
ÞÞ
dFr
t
¼
Aw
l
ðFr
t
; B

ðFr
t
ÞÞ
AFr
t
þ
Aw
l
ðFr
t
; B

ðFr
t
ÞÞ
AB
t
dB

ðFr
t
Þ
dFr
t
N0:
21
A more detailed explanation of the mechanics of a related model is found in Yuki (2003).
92 K. Yuki / Journal of Development Economics 84 (2007) 76–103
demand and the lower supply of the good, and the number of rural unskilled workers is smaller,
resulting in higher unskilled wage. If the price level is above the critical level, i.e. b
u

(Fr
0
,B
0
) Ne,
a richer portion of unskilled workers can send their children to school and the expansion of the
modern sector starts immediately. Even when the initial price level is not high enough for the
‘take-off’ due to limited asset accumulation, the large pool of skilled workers makes rapid asset
accumulation possible, hence the price level rises to the critical level eventually and the
modernization process begins. Once the structural change is initiated, it continues autonomously.
An increase in the number of skilled workers and asset accumulation lead to a further increase in
the price of good ul, stimulate the migration of unskilled workers to the urban region, and boost
unskilled wage. This allows children of less affluent unskilled workers to access education and
thus increases the number of skilled workers further, which, in turn, inflates the price and unskilled
wage even more. As long as skilled wage (net of the cost of education) is higher than unskilled
wage, this process continues. In the long run, the economy reaches the state in which the return
from education is equated with the one from assets, thereby attaining a state of equal opportunity.
In contrast, if the economy starts with relatively small Fr
0
, either because wealth is concentrated
in the few rich or because limited wealth is dispersed among the many poor, skilled labor is scarce,
the price of nontradable ul is low, and many of unskilled workers remain in agriculture. Children of
unskilled workers are not financially able to obtain education and the number of skilled workers
does not increase. Still, if initial asset accumulation is low, i.e. B
0
bB

(Fr
0
), further asset
accumulation increases the relative price, promotes the migration of unskilled workers, and raises
unskilled wages over time. However, asset accumulation alone is not enough to raise the relative
price to the critical level necessary for structural change. Since skilled labor remains limited,
inequality between skilled and unskilled workers does not disappear and the investment choices are
affected by family income even in the long run.
22,23
As for the issue of internal migration, the proposition shows that, as long as an economy starts
with small asset accumulation, i.e. B
0
bB

(Fr
0
), it always experiences urbanization along its
transition to a steady state, which is consistent with the fact that urbanization occurred universally
for most of the post-war period. However, the economy experiences different kinds of urbanization
depending on its development path.
When an economy fails to initiate industrialization and ends up in the unequal-opportunity
steady state, i.e. (Fr
ss
,B
ss
) =(Fr,B

(Fr)), Fr ≤Fr

, urbanization takes place without an increase in
well-paid jobs for skilled workers, hence the proportion of ‘informal’ unskilled jobs increases
22
Using panel data of wider coverage and of higher quality than those of earlier studies, Deininger and Squire (1998) and
Deininger and Olinto (2000) discover that an economy's growth rate is affected negatively by initial land inequality (a proxy
for initial asset inequality) and positively by its mean years of schooling per working person (a proxy for human capital).
Further, they find that the average educational attainment is negatively affected by initial land inequality; the effect of human
capital is larger for a lower-income economy; and initial land and income inequalities affect negatively the income growth of
the poor, but not of the rich. Also, Birdsall and Londono (1998) find that initial inequalities of land and education have strong
negative effects on subsequent income growth, especially that of the poor. Their findings are consistent with predictions of
the model, as long as low Fr
0
reflects high initial asset inequality, that is, unless asset accumulation is very low.
23
The model may be used to understand growth experiences of two contrasting economies, Brazil and Korea. At the
beginning of the 1960s, the income and wealth distributions of Brazil were far less equal than Korea (Deininger and
Olinto, 2000). Both Brazil and Korea experienced rapid growth in the 1970s, but Brazil attained only a modest increase in
the proportion of workers with secondary education from a much smaller starting value, resulting in much smaller
decreases in returns to education (Birdsall et al., 1995). According to Williamson (1993), a large portion of differences in
enrollment rates of secondary education between the two economies in the early 1970s can be explained by the greater
income inequality in Brazil, not GNP per adult, the relative price of teachers, or school age population ratio. And Brazil
stagnated after the 1980s, while Korea continued to grow fast. In terms of the model, Korea may be on a path to the equal
opportunity steady state and Brazil on a path to an unequal opportunity steady state.
93 K. Yuki / Journal of Development Economics 84 (2007) 76–103
over time in the urban region. This result accords with growing urban slum dwellers observed in a
typical stagnant developing economy. Wealth accumulation raises the demand for nontradable ul,
which corresponds to services such as trading and housework in real economy, increases urban
unskilled jobs supplying the good, and attracts rural unskilled workers into the urban region.
By contrast, when an economy is on a track to the equal-opportunity steady state, i.e. (Fr
ss
,B
ss
)
=(1,B
⁎⁎
), it goes through three stages of urbanization. At the initial stage when an economy grows
without an increase in skilled workers, i.e. b
u

(Fr,B) ≤e, it urbanizes without an increase in skilled
jobs, thus the share of informal jobs in the urban region increases. Once the economy ‘takes off’,
i.e. b
u

(Fr,B) Ne, urbanization accompanies an increase in skilled jobs. At this stage, urbanization
is driven not only by the increasing demand for non-tradable ul, but also by an increase in the
number of educated people searching for skilled jobs. An increase in the number of skilled workers
and asset accumulation lower wage differentials between skilled and unskilled workers over time,
and urbanization stops once the economy attains equal opportunity. At this point the wage
differential between the regions disappears as well.
24
4.2. Roles of productivities
So far, the analysis has been performed for given levels of productivities. This subsection
examines how changes in the productivities of sector rl (agricultural productivity) and of sector
uh (manufacturing productivity) affect the distribution of the population between the regions,
sectoral compositions of production, and the dependency of the initial distribution on long-run
steady states. First, effects of a decrease in the productivity of sector rl, A
rl
, are investigated.
25
Proposition 3.
26
Suppose that an economy experiences a one-time and permanent decrease in
A
rl
at the beginning of a period that is expected in the previous period. Then,
(i) Fr

increases. That is, the economy is more likely to end up in an unequal opportunity
steady state, (Fr
ss
,B
ss
) =(Fr,B

(Fr)), Fr ≤Fr

.
(ii) Instantaneous effects on sectoral allocations of workers and wages are:
24
Larson and Mundlak (1997) find that the ratio of average labor productivity of agricultural workers to that of non-
agricultural workers converges to one as an economy develops. Some evidence indicates that the size of the informal
sector and the wage differential between the informal and formal sectors decrease with development through human
capital accumulation (Marcouiller et al., 1997; Ranis and Stewart, 1999). Marcouiller et al. (1997) find sizable wage
differences between formal and informal sectors after controlling for individual characteristics and industry-specific
factors in El Salvador and Peru but not in Mexico in the late 1980s. Consistent with the model's implications, the
Mexico's income level is higher than the others and the share of the informal sector is smaller. Higher economic growth
of Mexico may be partly due to its more equal initial asset distribution: at the beginning of the 1960s, the distribution of
land holdings in Mexico was much more equal than the other nations (Deininger and Olinto, 2000).
25
Results are presented for the productivity decline rather than the more frequent productivity increase, since, as
summarized in the proposition, the former promotes urbanization. As for manufacturing, for the same reason, the
productivity increase is examined. In this way, one can see clearly how the productivity changes in the two sectors, both
promoting urbanization, have different implications for development and inequality.
26
In the modified model with two regional nontradable sectors, the following results are additionally obtained. As for
the instantaneous effects in (ii), in the unequal opportunity case, the fraction of workers in the rural tradable (agriculture)
sector declines and the number of workers in the urban informal sector rises; when the economy belongs to the equal
opportunity case both before and after the productivity change, the fraction of workers in the informal sector falls in the
urban region, while the number of workers in the rural nontradable sector decreases and that of the urban informal sector
increases by the same amount. Regarding the long-run effects in (iii), in all the steady states, the number of workers in the
rural nontradable sector falls and that of the urban informal sector rises.
94 K. Yuki / Journal of Development Economics 84 (2007) 76–103
(a) (Unequal opportunity case) L
r,t
and w
l,t
decrease, L
u,t
increases, and H
t
=Fr
t
does not
change.
(b) (Equal opportunity case) If the economy belongs to the unequal opportunity case
after the productivity decline, H
t
and L
u,t
increase, and L
r,t
and w
l,t
decrease. If it
remains in the equal opportunity case, H
t
increases, L
r,t
decreases, and L
u,t
does
not change.
(iii) Long-run effects on sectoral allocations of workers and wages are:
(a) (Unequal opportunity steady states) At the steady state with given Fr, B

(Fr), L
r
, and w
l
decline, and L
u
rises.
(b) (Equal opportunity steady state) H
⁎⁎
increases, L
r
decreases, while B
⁎⁎
and L
u
do not
change.
Proof. In the Appendix. □
A fall in agricultural productivity A
rl
decreases unskilled wage directly and makes it more
difficult for unskilled workers to afford their children education (b
u

(Fr,B) =e shifts outward on
the (Fr,B) space). The lower unskilled wage translates into the slowdown of asset accumulation
and the reduction of long-run aggregate transfers (B=B

(Fr) shifts downward), which reduces the
demand for non-tradable ul and unskilled wage. As a result, the minimum size of Fr
0
necessary
for successful development increases (Proposition 3(i)). That is, as agricultural productivity is
lower, a disadvantageous asset distribution in the initial period becomes a more serious hindrance
to development. Further, the productivity decline has negative effects on the unequal opportunity
steady states: unskilled wage falls and wage and wealth inequalities worsen (Proposition 3(iii)
(a)). In contrast, it does not affect the wage and inequalities in the equal opportunity steady state
(Proposition 3(iii)(b)).
27
Proposition 3(ii) and (iii) show that urbanization is not necessarily associated with income
growth. A fall of agricultural productivity depresses farmers' incomes and forces them to leave
their lands and migrate to the urban area immediately. They move to the urban informal sector
in the unequal opportunity case and to the modern sector in the equal opportunity case.
Moreover, the decline of unskilled wage and the migration of rural workers are persistent. In
particular, when Fr
0
is low and the economy remains in the unequal opportunity case in the
long run, the long-run effect on unskilled wage is stronger and the one on migration is weaker
than the short-run effects: the lower income due to the productivity decline slows down asset
accumulation, and has a negative effect on the demand for good ul and thus unskilled wage. As
can be shown easily, a fall of the price of the agricultural good rl has the same qualitative
effects. Empirical findings suggest that agricultural productivity and prices of agricultural
goods are important determinants of internal migration in many stagnant developing countries,
especially in sub-Saharan African nations.
28
27
The divergent development and urbanization experiences of Thailand and the Philippines after the 1980s, which is
mentioned in the introductory section, could be due to differences in initial conditions of the two economies. That is,
Thailand started with the more equal distribution of land and higher agricultural productivity (more abundant fertile land)
(Inral and Power, 1991; Siamwalla et al., 1993), both of which are conductive to urbanization with development according
to the model. Although pro-industry policies, such as export taxes on agricultural products, overvaluation of the exchange
rate, and various subsidies to industry, lead to industrialization in the Philippines in the 1970s, they were not sustainable.
28
Becker and Morrison (1988) find that per capita caloric intake in rural areas is the most important determinant of
urbanization in sub-Sahara African economies for the years 1970–80. Jaeger (1992), based on survey data from Ghana,
shows that policy reforms that shift rural–urban terms of trade in favor of agriculture resulted in a significant reverse
migration from urban to rural regions in the late 1980s.
95 K. Yuki / Journal of Development Economics 84 (2007) 76–103
The next proposition summarizes effects of a permanent increase in the productivity of
sector uh.
Proposition 4.
29
Suppose that an economy experiences a one-time and permanent increase in
A
uh
at the beginning of a period that is anticipated in the previous period. Then,
(i) Fr

increases. That is, the economy is more likely to end up in an unequal opportunity
steady state, (Fr
ss
,B
ss
) =(Fr,B

(Fr)), Fr ≤Fr

.
(ii) Instantaneous effects on sectoral allocations of workers and wages are:
(a) (Unequal opportunity case) L
u,t
, w
h,t,
and w
l,t
increase, L
r,t
decreases, and H
t
=Fr
t
does not change.
(b) (Equal opportunity case) If the economy belongs to the unequal opportunity case after
the productivity increase, H
t
, w
h
, and w
l,t
increase, and L
r,t
decreases. The effect on L
u,t
is ambiguous. If it remains in the equal opportunity case, H
t
, L
u,t
, w
h
, and w
l,t
increase,
while L
r,t
decreases.
(iii) Long-run effects on sectoral allocations of workers and wages are:
(a) (Unequal opportunity steady states) At the steady state with given Fr, B

(Fr), L
u
, w
h
,
and w
l
increase, and L
r
decreases.
(b) (Equal opportunity steady state) H
⁎⁎
, B
⁎⁎
, w
h
, and w
l
increase, L
r
decreases, and L
u
does not change.
Proof. In the Appendix. □
A rise in manufacturing productivity raises the cost of education, which is proportional to
skilled wage, more than unskilled wage and makes it harder for children of unskilled workers to
accomplish upward mobility (b
u

(Fr,B) =e shifts outward on the (Fr,B) space). It also raises total
income and long-run aggregate transfers (B=B

(Fr) shifts upward), which raises the demand for
good ul and unskilled wage. It turns out that the former effect dominates, thus the critical level of
Fr for modernization, Fr

, goes up and the economy is more likely to end up in the unequal
opportunity steady state (Proposition 4(i), see Fig. 3). In particular, if the initial condition satisfies
Fr
0
∈(Fr
old

, Fr
new

], the productivity increase drastically changes the economy's steady state
from (Fr
ss
,B
ss
) =(1,B
⁎⁎
) to (Fr
ss
,B
ss
) =(Fr
0
,B

(Fr
0
)). The result suggests that an uneven
productivity growth of the more advanced manufacturing sector, that is, a skill-biased techno-
logical progress, is not always beneficial to the long-run development of a low-income nation.
For an economy starting with relatively large Fr
0
(Fr
0
NFr
new

), however, the productivity
increase is certainly advantageous. In such an economy, the rise of the cost of education does not
become an impediment to convergence to the equal-opportunity steady state, although it slows down
the convergence process somewhat. Further, the productivity growth raises the number of skilled
jobs, income, and wealth in the equal-opportunity steady state (Proposition 4(iii)(b)).
30
The increase
in manufacturing productivity also promotes urbanization. In the short run, the productivity increase
29
In the modified model with two regional nontradable sectors, the following results are additionally obtained. As for
the instantaneous effects in (ii), when the economy remains to be in the same (unequal or equal opportunity) case after the
productivity change, the number of workers in the rural nontradable sector declines and that of the urban informal sector
rises; otherwise, the number of workers in the rural nontradable sector decreases and the fraction of workers in the
informal sector falls in the urban region. Regarding the long-run effects in (iii), in all the steady states, the number of
workers in the rural nontradable sector falls and that of the urban informal sector rises.
30
On the other hand, when the initial condition satisfies Fr
0
≤Fr
old
†, the positive productivity shock yields mixed
results: while it raises wages of both skilled and unskilled workers (Proposition 4(iii)(a)), it worsens wage inequality
between them, as can been seen from Eq. (36) in Lemma 8.
96 K. Yuki / Journal of Development Economics 84 (2007) 76–103
raises skilled wage, total income, and the demand for good ul and encourages the migration of rural
workers to the urban informal sector in the unequal opportunity case (Proposition 4(ii)(a)), while, in
the equal opportunity case, it promotes migration to the formal sector as well (Proposition 4(ii)(b)).
Since the increase in total income stimulates asset accumulation and raises the demand for good ul
further in the longer term, the effects are persistent (Proposition 4(iii)).
4.3. Discussion
Some of the model's assumptions are made for the purpose of simplifying analyses rather than
for capturing reality. As for the assumption of the single nontradable sector ul, it has been shown
in footnotes 14, 19, 26, and 29 that main results are not affected even if regionally separated
nontradable sectors are allowed. In this subsection, the robustness of main results with respect to
modifications of several other assumptions and the model's implications for international
migration are briefly discussed.
4.3.1. Urban formal sector's production function
When the model is modified so that sector uh (the urban formal sector) employs unskilled
workers as well as skilled workers, skilled wage changes with the ratio of skilled to unskilled
workers in the sector. With this modification, the model becomes intractable with the present
setting that skilled workers are employed as teachers in the education sector, hence the cost of
education is assumed to be the cost of purchasing a fixed amount of good uh. Then, although the
analysis becomes more complex, the results of the original model remain intact qualitatively for
the most part. The only major difference is that now an increase in Fr =H raises unskilled wage
through the higher proportion of skilled workers in sector uh, as well as the higher demand for
good ul and the lower number of unskilled workers, in the unequal opportunity case.
4.3.2. Utility function
In the model, the utility function is Cobb–Douglas and thus the proportion of spending on each
itemis constant and independent of income and price levels. In real economy, the proportion of the
consumption of agricultural goods decreases with development (the Engel's law). This feature can
be captured by a slightly modified utility function, u
i
t
¼ ðc
i
rl;t

¯
c
rl
Þ
g
rl
P
j p rl
ðc
i
j;t
Þ
g
j
ðb
i
tþ1
Þ
g
k
. That is,
an individual must consume at least c
rl
¯ units of good rl irrespective of his or her income.
Qualitatively, the results remain intact with positive c
rl
¯ , but quantitatively they are affected. In the
unequal opportunity case, positive c
rl
¯ implies smaller demand for good ul, the lower degree of
urbanization, and lower unskilled wage for given Fr and B, resulting in smaller asset accumulation
and higher Fr

, i.e. low Fr becomes a more serious obstacle to development. In contrast, in the
equal opportunity case, H
⁎⁎
increases and L
u
decreases due to the lower demand for good ul.
4.3.3. Rural economy
When the production function of sector rl (the agriculture sector) exhibits constant returns to
scale, the wage rate equals the marginal product of labor, i.e. there is no disguised unemployment.
In this setting, when Fr and B are low, both sectors rl and ul exist, while when Fr and B are high,
only sector ul operates, i.e. the economy is completely urbanized. The dynamics are also affected.
If the agricultural productivity is high enough that
g
k
1−g
k
ð1þrÞ
p
rl
A
rl
Ne is satisfied, irrespective of
initial conditions, the economy reaches the equal opportunity steady state in the long run. By
contrast, when the productivity is low, the long-run outcome is determined by initial conditions as
in the original model. In a transition to the equal opportunity steady state, as long as sector rl is in
97 K. Yuki / Journal of Development Economics 84 (2007) 76–103
operation, urbanization proceeds without changes in unskilled wage, and once urbanization is
completed, unskilled wage starts to increase.
If the production function of the rural sector is identical to the original model but the wage is
determined competitively, given asset distribution, the number of workers in the sector is smaller
and aggregate output is higher compared to the original setting, as is explained in footnote 16.
This does not imply that unskilled wage and income are higher, since a portion of the output is
distributed as profits. Main results would be qualitatively unchanged, but whether the economy
goes through urbanization accompanying modernization or not now depends on the distribution
of profits over the population as well.
4.3.4. International migration
The model has some implications for effects of international migration on the urbanization and
development of lowincome nations. Migration is motivated economically by returns frommigration
and migration costs. Returns frommigration would be greater for unskilled workers, since the return
to education and wage inequality are lower in developed nations, while migration costs would inhibit
migration of unskilled workers with limited wealth. When Fr and B are very low in the absence of
international migration, unskilled wage is low, many of unskilled workers cannot cover migration
costs, and thus skilled workers would be overrepresented in immigrants. If this is the case, the
presence of international migration tends to decrease Fr, exacerbate wage inequality, lower the
degree of urbanization, and raise the proportion of the informal sector in the urban region. On the
other hand, when Fr and B are not very low, immigrants are disproportionately unskilled workers,
hence international immigration would raise Fr and have the opposite effects to the previous
situation. In particular, if Fr is only slightly less than Fr

without international migration, increased
migration allows the economy to attain equal opportunity in the long run. Further, remittances from
immigrants may contribute to wealth accumulation by unskilled workers and lower Fr

, further
easing the burden of low Fr in development.
5. Conclusion
Urbanization is a universal phenomenon in the postwar world, but its qualitative nature is
varied across countries. Some nations have experienced urbanization that accompanied skill
upgrading, industrialization, and the expansion of the urban formal (modern) sector, but others
urbanized without such modernization and underwent the expansion of the urban informal
(traditional) sector simultaneously. The question that naturally arises is what are underlying
causes for the differences. The purpose of the present paper has been to tackle this question
analytically based on a dynamic model of urbanization and development.
The main findings of the paper can be summarized as follows. (1) In the model economy there
exist two types of steady states. The equal opportunity steady state has features of a typical
developed economy and the unequal opportunity steady states have features of a typical stagnant
developing economy in terms of the accumulation of human capital, the sectoral and regional
distributions of the population, and sectoral and regional income inequalities. (2) Given an initial
distribution of transfers (wealth) over the population, an economy approaches one of the steady
states in the long run. If an economy starting with limited asset accumulation is on a path to the
equal opportunity steady state, it experiences urbanization that accompanies skill upgrading, the
expansion of the formal sector, the shrinkage of the informal sector, and falling inequality.
Instead, if it is on a track to an unequal opportunity steady state, it urbanizes without such
modernization but with the expansion of the informal sector. Thus the model exhibits two
98 K. Yuki / Journal of Development Economics 84 (2007) 76–103
distinctive patterns of urbanization observed in real economy. The critical determinant of the
long-run outcome of an economy is the initial fraction of the population who is sufficiently
wealthy to obtain education. (3) Both a decrease in agricultural productivity and an increase in the
formal sector's productivity promote urbanization but make it more difficult for an economy
starting with an unfavorable wealth distribution to reach the equal opportunity steady state in the
long run.
Appendix A. Proofs of lemmas and propositions
Proof of Lemma 2. From the definition of the two cases, the dividing line is equivalent to
w
h
−(1+r)e=w
l
(Fr
t
,B
t
). By substituting p
rl
A
rl
(L
r,t
)
α
rl
−1
=w
h
−(1+r)e into Eq. (16) and solving for L
r,t
,
L
r;t
¼ 1−g
ul
−H
t

g
ul
ð1 þ rÞB
t
w
h
−ð1 þ rÞe
: ð40Þ
By plugging H
t
=Fr
t
into the above equation and substituting the resulting equation into
w
h
−(1+r)e=p
rl
A
rl
(L
r,t
)
α
rl
−1
,
w
h
−ð1 þ rÞe ¼ p
rl
A
rl
− 1−g
ul
−Fr
t
g
ul
ð1 þ rÞB
t
w
h
−ð1 þ rÞe
_ _
a
rl
−1
: ð41Þ
Eq. (23) is obtained by solving the above equation for B
t
. □
Proof of Lemma 3. L
r,t
is obtained from the condition w
l,t
=p
rl
A
rl
(L
r,t
)
α
rl
−1
=w
h
−(1+r)e. As for
H
t
, w
h
−(1+r)e=w
l
(H
t
,B
t
) must hold by the definition of the equal opportunity case. From the
proof of the previous lemma, it can be seen that w
h
−(1+r)e=w
l
(H
t
,B
t
) is equivalent to B
t
=D
(H
t
). □
Proof of Lemma 4. First, the existence and uniqueness of the solution to Eq. (29) for given Fr
t
is
proved.
It is enough to prove that (a)
ARHS
AB
t
b1 and (b) RHSN0 when B
t
=0 are satisfied. From Eq. (13),
Aw
l
ðFr
t
; B
t
Þ
AB
t
¼ −ð1−a
rl
Þp
rl
A
rl
ðL
r;t
Þ
a
rl
−2
AL
r;t
AB
t
: ð42Þ
From Eq. (16),
AL
r;t
AB
t
¼
−g
ul
ð1 þ rÞ
p
rl
A
rl
ðL
r;t
Þ
a
rl
−2
½ð1−g
ul
Þð1−a
rl
Þð1−Fr
t
Þ þ a
rl
L
r;t
Š
: ð43Þ
Substituting this equation into Eq. (42),
Aw
l
ðFr
t
; B
t
Þ
AB
t
¼
ð1−a
rl
Þg
ul
ð1 þ rÞ
ð1−g
ul
Þð1−a
rl
Þð1−Fr
t
Þ þ a
rl
L
r;t
: ð44Þ
Hence,
ARHS
AB
t
¼ g
k
ð1 þ rÞ þ g
k
ð1−Fr
t
Þ
Aw
l
ðFr
t
; B
t
Þ
AB
t
;
¼ g
k
ð1 þ rÞ
ð1−a
rl
Þð1−Fr
t
Þ þ a
rl
L
r;t
ð1−g
ul
Þð1−a
rl
Þð1−Fr
t
Þ þ a
rl
L
r;t
V
g
k
ð1 þ rÞ
1−g
ul
b1:
99 K. Yuki / Journal of Development Economics 84 (2007) 76–103
(b) is obviously satisfied. Thus, the solution to Eq. (29), B

(Fr
t
), exists and is unique for given
Fr
t
. If the dynamics of aggregate transfers are always described by Eq. (27), global convergence
to B

(Fr
t
) is assured by this result as well. Similarly, if the dynamics are always determined by
Eq. (28), global convergence to B
⁎⁎
is obvious. The actual dynamics are given by Eq. (27) in the
unequal opportunity case, i.e. B
t
bD(Fr
t
), and by Eq. (28) in the equal opportunity case, i.e.
B
t
≥D(Fr
t
). If B

(Fr
t
) bD(Fr
t
) is satisfied, the fixed point of Eq. (27) is below the critical level of
B
t
dividing the two cases on the (Fr
t
,B
t
) space, hence B
t
converges to B

(Fr
t
) globally (see Fig. 1).
Otherwise, B
t
converges to B
⁎\⁎
globally (see Fig. 1). The critical level of Fr
t
satisfying B

(Fr
t
) =
D(Fr
t
) is H
⁎⁎
, which is obtained by substituting B
t+1
=B
t
=D(Fr
t
), Eq. (23), into Eq. (28) and solving
for Fr
t
. Finally,
AB⁎ðFr
t
Þ
AFr
t
N0 is true, since the RHS of Eq. (29) is increasing in Fr
t
in the unequal
opportunity case. □
Proof of Lemma 6. Fr
t+1
=Fr
t
is satisfied iff children of currently unskilled workers cannot
afford education, i.e. for any lineage i satisfying b
t
i
be, b
t+1
i

k
{w
l
(Fr
t
,B
t
) +(1+r)b
t
i
}be (from
Eq. (22)) must be true. The inequality (31) is derived from this condition. Since w
l
(Fr
t
,B
t
) is equal
to the right-hand side of Eq. (16) when H
t
=Fr
t
, the inequality becomes
g
k
1−g
k
ð1 þ rÞ
g
ul
f½w
h
−ð1 þ rÞeŠFr
t
þ ð1 þ rÞB
t
g
ð1−g
ul
Þð1−Fr
t
Þ−L
r;t
Ve; ð45Þ
fL
r;t
Vð1−g
ul
Þð1−Fr
t
Þ−
g
k
1−g
k
ð1 þ rÞ
g
ul
e
f½w
h
−ð1 þ rÞeŠFr
t
þ ð1 þ rÞB
t
g: ð46Þ
Substituting Eq. (13) into the same inequality (31) and solving it for L
r,t
, one can also obtain
L
r;t
z
p
rl
A
rl
e
g
k
1−g
k
ð1 þ rÞ
_ _ 1
1−a
rl
: ð47Þ
Then, Eq. (32) is obtained from these two inequalities (46) and (47). □
Proof of Lemma 7. From Eqs. (23) and (32) the statement is true iff
g
k
1−g
ul
−Fr
t

p
rl
A
rl
w
h
−ð1 þ rÞe
_ _ 1
1−a
rl
_ _
½w
h
−ð1 þ rÞeŠN½1−g
k
ð1 þ rފ
 e ð1−g
ul
Þð1−Fr
t
Þ−
p
rl
A
rl
e
g
k
1−g
k
ð1 þ rÞ
_ _ 1
1−a
rl
_ _
−g
k
g
ul
½w
h
−ð1 þ rÞeŠFr
t
;
ffg
k
½w
h
−ð1 þ rÞeŠ−½1−g
k
ð1 þ rފegð1−g
ul
Þð1−Fr
t
ÞNðp
rl
A
rl
Þ
1
1−a
rl
 g
k
½w
h
−ð1 þ rÞeŠ
−a
rl
1−a
rl
−g
1
1−a
rl
k
f½1−g
k
ð1 þ rފeg
−a
rl
1−a
rl
_ _
:
Since γ
k
[w
h
−(1+r)e] N[1−γ
k
(1+r)]e is satisfied by the assumption γ
k
w
h
Ns
e
w
h
=e, the left-
hand side of the above equation is positive, while the right-hand side is negative, proving the
inequality. □
100 K. Yuki / Journal of Development Economics 84 (2007) 76–103
Proof of Lemma 8. (i)H
⁎⁎
is obtained from the substitution of B
t
=B
⁎⁎
, Eq. (30), into B
t
=
D(H
t
), Eq. (23). L
r
(H
⁎⁎
,B
⁎⁎
) is from Lemma 3. Finally L
u
(H
⁎⁎
,B
⁎⁎
) =1−H
⁎⁎
−L
r
(H
⁎⁎
,B
⁎⁎
).
(ii) By substituting H
t
=Fr, B
t
=B

(Fr), and p
ul,t
A
ul
=w
l
(Fr,B) into Eq. (36),
w
l
ðFr; B*ðFrÞÞ ¼
g
ul
L
u
−g
ul
ð1−FrÞ
fw
h
Fr þ ð1 þ rÞðB*ðFrÞ−eFrÞg: ð48Þ
From Eq. (29),
B

ðFrÞ ¼
g
k
1−g
k
ð1 þ rÞ
f½w
h
−ð1 þ rÞeŠFr þ w
l
ðFr; B*ðFrÞÞð1−FrÞg: ð49Þ
By plugging the above equation into Eq. (48) and solving for L
u
, Eq. (36) is obtained. □
Proof of Proposition 1. (i) The first relation is derived from the facts that L
r
(H,B) is a decreasing
function of its arguments (Lemma 1(i)), and H
⁎⁎
NFr, B
⁎⁎
NB

(Fr) are satisfied for Fr ≤Fr

. As
for the second relation
L
u
ðH
⁎⁎
; B
⁎⁎
ÞbL
u
ðFr; B*ðFrÞÞ; ð50Þ
f1b1 þ
½w
h
−ð1 þ rÞeŠ−w
l
ðFr; B

ðFrÞÞ
w
l
ðFr; B

ðFrÞÞ
Fr; ðfrom Lemma 8Þ; ð51Þ
f½w
h
−ð1 þ rÞeŠ−w
l
ðFr; B

ðFrÞÞN0: ð52Þ
(ii) is directly obtained from (i). The first result of (iii) is straightforward from Lemma 1(i) and
the result
AB

ðFr
t
Þ
AFr
t
N0 of Lemma 4(i), and the second result is from Eq. (36) of Lemma 8, Lemma 1
(iii), and
AB

ðFr
t
Þ
AFr
t
N0. □
Proof of Proposition 3. (i) Recall that Fr

is the intersection of B=B

(Fr) and b
u

(Fr,B) =e.
From Eqs. (29) and (32), the exact form of Fr

is
Fr
y
¼
½1−g
k
ð1 þ rފe ½1−g
k
ð1 þ rފ 1−g
ul

p
rl
A
rl
w
h
−ð1þrÞe
_ _
1
1−a
rl
_ _
−g
k
g
ul
ð1 þ rÞ
_ _
g
k
g
ul
½w
h
−ð1 þ rÞeŠ þ ½1−g
k
ð1 þ rފ½1−g
ul
−g
k
ð1 þ rފe
: ð53Þ
(ii)(a) H
t
=Fr
t
is not affected, since the educational investment of currently skilled workers has
been made in the previous period. From Eqs. (16) and (13), L
r
(H
t
,B
t
) and w
l
(H
t
,B
t
) decline with
the fall of A
rl
, hence the results follow. (b) Since the productivity decline shifts B
t
=D(Fr
t
), Eq.
(23), outward, there is a region that used to be in the equal opportunity case but nowbelongs to the
unequal opportunity case. For such region, w
l,t
obviously drops. H
t
increases since H
t
=Fr
t
rather
than H
t
≤Fr
t
is satisfied now. Then, from Lemma 1, L
r,t
decreases. Regarding L
u,t
, the substi-
tution of p
ul,t
A
ul
=w
l,t
into Eq. (15) gives
w
l;t
ðL
u;t
−g
ul
Þ ¼ g
ul
f½w
h
−ð1 þ rÞe−w
l;t
ŠH
t
þ ð1 þ rÞB
t
g: ð54Þ
Since H
t
rises and w
l,t
falls, the right-hand side of the equation shifts upward, while the left-
hand side shifts downward. Hence, L
u,t
increases. In contrast, if the economy remains in the equal
opportunity case, the results follow from Lemma 3.
101 K. Yuki / Journal of Development Economics 84 (2007) 76–103
(iii)(a) Since the decrease of w
l
(H,B) lowers the right-hand side of Eq. (29), B*(Fr) and thus
w
l
decline. Then, the increase of L
u
is straightforward from Eq. (36). (b) Straightforward from
Lemmas 4(ii) and 3. □
Proof of Proposition 4. (i) Remember that w
h
and e=s
e
w
h
are proportional to A
uh
. Then, the
result is straightforward from Eq. (53).
(ii) (a) From Eq. (16), L
r
(H
t
,B
t
) declines with the increase of w
h
−(1+r)e, hence from Eq. (13),
w
l
(H
t
,B
t
) rises. (b) If the economy belongs to the unequal opportunity case after the productivity
increase shifts B
t
=D(Fr
t
), Eq. (23), outward, H
t
increases since H
t
=Fr
t
rather than H
t
≤Fr
t
is
satisfied now. Then, from Eq. (16), L
r,t
declines, and from Eq. (13), w
l,t
rises. From Eq. (15), the
effect on L
u,t
is ambiguous. On the other hand, if the economy remains in the equal opportunity
state, the results follow from Lemma 3 and Eq. (13).
(iii) (a) Since the increases of w
h
−(1+r)e and w
l
(H,B) raise the right-hand side of Eq. (29),
B*(Fr) and w
l
(Fr,B*(Fr)) increase. Then, from Eq. (36), L
u
rises. (b) Straightforward from
Lemma 4(ii), Lemma 8(i), and Eq. (13). □
References
Banerjee, A., Newman, A., 1993. Occupational choice and the process of development. Journal of Political Economy 101
(2), 274–298.
Banerjee, A., Newman, A., 1998. Information, the dual economy, and development. Review of Economic Studies 65 (4),
631–653.
Becker, C.M., Morrison, A.R., 1988. The determinants of urban population growth in sub-Saharan Africa. Economic
Development and Cultural Change 36 (2), 259–278.
Benabou, R., 1996a. Heterogeneity, stratification, and growth: macroeconomic implications of community structure and
school finance. American Economic Review 86 (3), 584–609.
Benabou, R., 1996b. Inequality and growth. In: Bernanke, B., Rotemberg, J. (Eds.), NBER Macroeconomic Annual 1996.
MIT Press, Cambridge, MA, pp. 11–92.
Bencivenga, V., Smith, B., 1997. Unemployment, migration, and growth. Journal of Political Economy 105 (3), 582–608.
Birdsall, N., Ross, D., Sabot, R., 1995. Inequality as a constraint on growth in Latin America. In: Turnham, D.,
Foy, C., Larrain, G. (Eds.), Social Tensions, Job Creation and Economic Policy in Latin America. OECD, Paris,
pp. 175–207.
Birdsall, N., Londono, J.L., 1998. No tradeoff: efficient growth via more equal human capital accumulation. In: Birdsall,
N., Graham, C., Sabot, R. (Eds.), Beyond Tradeoffs: Market Reforms and Equitable Growth in Latin America. Inter-
American Development Bank, New York, pp. 111–145.
Deininger, K., Squire, L., 1998. New ways of looking at old issues: inequality and growth. Journal of Development
Economics 57 (2), 259–287.
Deininger, K., Olinto, P., 2000. Asset Distribution, Inequality, and Growth. World Bank, mimeo.
Durlauf, S., 1996. Theory of persistent income inequality. Journal of Economic Growth 1 (1), 75–94.
Fay, M., Opal, C., 2000. Urbanization Without Growth: A Not So Uncommon Phenomenon. World Bank, mimeo.
Fields, G.S., 1975. Rural–urban migration, urban unemployment and underemployment, and job-search activity in LDCs.
Journal of Development Economics 2 (2), 165–187.
Fulginiti, L.E., Perri, R.K., 1997. LDC agriculture: nonparametric Malmquist productivity indexes. Journal of
Development Economics 53 (2), 373–390.
Funkhouser, E., 1997. Mobility and labor market segmentation: the urban labor market in El Salvador. Economic
Development and Cultural Change 46 (1), 123–153.
Galor, O., Zeira, J., 1993. Income distribution and macroeconomics. Review of Economic Studies 60 (1), 35–52.
Galor, O., Tsiddon, D., 1997. The distribution of human capital and economic growth. Journal of Economic Growth 2 (1),
93–124.
Gupta, M.R., 1997. Informal sector and informal capital market in a small open less-developed economy. Journal of
Development Economics 52 (2), 409–428.
Harris, J.R., Todaro, M.P., 1970. Migration, unemployment and development: a two-sector analysis. American Economic
Review 60 (1), 126–142.
102 K. Yuki / Journal of Development Economics 84 (2007) 76–103
Inral, P., Power, J.H., 1991. The Philippines. In: Krueger, A.O., Schiff, M., Valdes, A. (Eds.), The Political Economy of
Agricultural Pricing Policy. Johns Hopkins University Press, Baltimore, pp. 149–194.
Jaeger, W.K., 1992. The causes of Africa's food crisis. World Development 20 (11), 1631–1645.
Jamal, V., Weeks, J., 1988. The vanishing rural–urban gap in sub-Saharan Africa. International Labor Review 127 (3),
271–292.
Larson, D., Mundlak, Y., 1997. On the intersectoral migration of agricultural labor. Economic Development and Cultural
Change 45 (2), 295–319.
Lewis, A.W., 1954. Economic development with unlimited supplies of labour. Manchester School of Economic and Social
Studies 22 (2), 139–191.
Ljungqvist, L., 1993. Economic underdevelopment: the case of missing market for human capital. Journal of Development
Economics 40 (2), 219–239.
Magnac, T., 1991. Segmented or competitive labor markets? Econometrica 59 (1), 165–187.
Marcouiller, D., de Castilla, V.R., Woodruff, C., 1997. Formal measures of the informal-sector wage gap in Mexico, El
Salvador, and Peru. Economic Development and Cultural Change 45 (2), 367–392.
Mazumdar, D., 1987. Rural–urban migration in developing countries. In: Mills, E.S. (Ed.), Handbook of Regional and
Urban Economics, vol. 2. Elsevier, Amsterdam, pp. 1097–1128.
Ranis, G., Stewart, F., 1999. V-goods and the role of the urban informal sector in development. Economic Development
and Cultural Change 47 (2), 259–288.
Siamwalla, A., Setboonsarng, S., Patamasiriwat, D., 1993. Agriculture. In: Warr, P.G. (Ed.), The Thai Economy in
Transition. Cambridge University Press, Cambridge, pp. 1–82.
Williamson, J.G., 1993. Human capital deepening, inequality, and demographic events along the Asia-Pacific rim. In:
Ogawa, N., Jones, G.W., Williamson, J.G. (Eds.), Human Resources in Development along the Asia-Pacific Rim.
Oxford University Press, Singapore, pp. 129–158.
World Bank, 1983. World Development Report 1983. Oxford University Press, New York.
Yuki, K., 2003. Sectoral Shift, Wealth Distribution, and Development. Kyoto University, mimeo.
103 K. Yuki / Journal of Development Economics 84 (2007) 76–103

K. Yuki / Journal of Development Economics 84 (2007) 76–103

77

nations and Latin American and Caribbean nations among middle-income nations, grew relatively slowly (Mazumdar, 1987; Fay and Opal, 2000). Behind the changes in the aggregate variable are migration decisions of rural residents, which are mainly driven by economic factors, that is, better job opportunities in urban areas. However, immigrants do not necessarily enjoy higher welfare in urban areas, due to the dualistic nature of the urban economy in a typical developing country. The urban economy of a typical underdeveloped nation can be classified into the formal or modern sector and the informal or traditional sector.1 The informal (traditional) sector is characterized by small firm sizes, unskilled jobs, low wages, and the loose enforcement of laws and regulations, and it incorporates sectors such as petty trading, domestic services, repair services, and basic manufacturing. The formal (modern) sector, by contrast, possesses contrasting features and includes modern manufacturing and skill-intensive services such as finance and health care. According to Ranis and Stewart (1999), in most developing countries, the size of the urban informal sector is substantial, in many cases accounting for over half the urban workforce. Further, earnings of workers in the informal sector are close to or sometimes lower than those in rural areas after adjustment for the cost of living.2 The large share of the informal sector in the urban economy implies that many immigrants do not attain noticeable improvements of their economic conditions. It seems that urbanization is associated with sizable job creation in the formal sector and thus large welfare gains by immigrants occur, only if an economy achieves industrialization. For example, the case study of Thailand and the Philippines by Ranis and Stewart (1999) shows that, during the 1960s and the 1970s, both nations experienced large-scale urbanization accompanied by rapid industrial growth, resulting in the expansion of the formal sector and the contraction of the informal sector. After the 1980s, Thailand kept following this trend, while the Philippines encountered urbanization with stagnant (almost zero) industrial growth and had to deal with the rapid growth of the informal sector. The question is why some economies attain urbanization accompanying the expansion of the formal sector and why others urbanize without such modernization but with the expansion of the informal sector. This paper presents a dynamic model of urbanization and development in order to answer the question. The model economy is inhabited by a continuum of individuals and is composed of the rural agriculture sector, the urban informal (traditional) sector, and the urban formal (modern) sector. The rural sector and the urban informal sector employ unskilled workers, whereas the urban formal sector employs skilled workers and physical capital. The rural sector produces an agricultural good, and the urban formal sector produces a manufacturing good, both of which are tradables. The urban informal sector produces a nontradable good that would correspond to services such as petty trading and domestic services in real economy.3 The rural agriculture sector
1 Strictly speaking, the modern/traditional classification is based on production technologies, while the formal/informal classification is based on establishment size or social security coverage, so they are distinct. In this paper, however, the two taxonomies are used interchangeably, because the focus is on differences in skill requirements between the sectors, as detailed below. 2 According to Jamal and Weeks (1988), for example, rural–urban income differentials are substantial in many subSaharan African nations but the income differentials for unskilled workers are very small and, in some nations, their urban incomes are actually lower than rural incomes. 3 Ranis and Stewart (1999) find that about 70% of the labor force in the informal sector in Manila in 1986–87 are engaged in trading and only about 13% are in manufacturing. In Bangkok, the proportion of workers in the informal sector engaged in manufacturing is larger but still only about 17%. Similar numbers are found for many Latin American economies as well (Marcouiller et al., 1997).

Generations go by in this fashion. it urbanizes without skill upgrading or modernization but with the expansion of the informal sector. this simplifying assumption would be defended. if it is on a track to an unequal opportunity steady state. . There are no barriers to interregional migration. Then. occupations. If an economy starting with limited asset accumulation is on a path to the equal opportunity steady state. is provided by teachers (skilled workers). it experiences urbanization that accompanies skill upgrading.4 An individual in the model lives for two periods. In childhood. many of the urban residents are not skilled and remain in the informal sector. and the inequalities are substantial. Given an initial distribution of transfers over the population. hence the share of spending on each item is constant over time. for example. distributions of net wage and wealth are equalized. Thus the model exhibits two distinctive patterns of urbanization observed in real economy. while the urban sectors have constant returns to scale technologies and wages are determined competitively. which show that an economy's growth rate is affected negatively by initial land inequality (a proxy for initial asset inequality). by contrast. Instead. characterizes a typical stagnant developing economy: many people live in rural areas and are engaged in agriculture. since the link between domestic savings and investment is becoming weaker with growing international capital flows.) The second type called the unequal opportunity steady states. an economy approaches one of the steady states in the long run. Both a decrease in the agricultural productivity and an increase in the formal sector's productivity promote urbanization but make it more likely for an 4 The model becomes analytically intractable once the closed capital market is assumed. and must be financed by the received transfer. and falling inequality. most people live in urban areas and are engaged in non-agricultural activities. This is consistent with empirical findings. and output is shared equally among workers. The first type called the equal opportunity steady state has features of a typical developed economy: many people are educated and skilled. depending on human capital accumulation in the previous period. he or she receives a transfer from the parent to invest in assets or education. The utility function is assumed to be Cobb–Douglas. and this heterogeneity translates into diverse choices of investments. and locations. and brings interesting dynamics. by Deininger and Squire (1998) and Deininger and Olinto (2000). which reflects the degree of equality of initial asset distribution unless aggregate asset accumulation is very low. reflecting the scarcity of land. In considering the urbanization of contemporary developing nations. Yuki / Journal of Development Economics 84 (2007) 76–103 exhibits diminishing returns to scale. this paper focuses on the role of human capital accumulation in development. In the model economy there exist two types of steady states. The critical determinant of the long-run success of an economy is the fraction of the population who can afford education in the initial period.78 K. the expansion of the formal sector. In adulthood. As different skill requirements of the formal and informal sectors indicate. and makes occupational and associated locational choices. he or she receives labor and capital incomes. and inequalities between skilled and unskilled workers and between urban and rural residents are small. The accumulation of physical capital is thus given only a minor role by assuming free international capital mobility. These results are obtained with fixed productivities. the individual becomes a skilled or unskilled worker. The paper also examines effects of onetime productivity changes in the rural agriculture sector and the urban formal sector on the urbanization and development of an economy. which are spent on the consumption of the three goods and a transfer to his or her sole child. Individuals of the same generation are heterogeneous in terms of received transfers. there are no credit markets for such investment in this economy. Education is needed to become a skilled worker. That is. (In the model. the first period as a child and the second period as an adult. the shrinkage of the informal sector.

Unlike a conventional Harris–Todaro model. there exists a tradeoff between the better access to credit in the traditional sector and the higher productivity in the modern sector. not exogenous institutional factors. and stagnant growth for many years. this adverse selection. In the model economy. unemployment (employment in the informal sector). The effects of the productivity increase in the formal sector suggest that skill-biased productivity growth is not necessarily beneficial for the long-run development of a poor economy. where the high-income steady state is more urbanized and industrialized but has a greater share of its urban workforce in the informal sector. Yuki / Journal of Development Economics 84 (2007) 76–103 79 economy starting with an unfavorable wealth distribution to end up in an unequal opportunity steady state. and the presence of trade unions.K. the conventional classification is at odds with some empirical findings: according to Funkhouser (1997). such as entry restrictions to the sector. Gupta (1997). . however. (1997) found that workers in the formal sector are significantly more educated than workers in the informal sector. and urban informal sectors and analyzes how migration. This paper is related to the literature that explores the mechanisms of urbanization and the evolution of the informal sector.6 Of course. gives rise to the dual structure of the urban economy. there exist multiple steady state equilibria. while unskilled workers stay in the rural sector. In equilibria. Hence. the differentials seem not to be explained largely by the presence of labor regulations or unions. Under this setting. individuals can move freely between the sectors and earn higher wages in the more productive modern sector. Magnac (1991) constructs a Roy model to test the hypothesis of competitive labor market against that of segmented market for married women in the urban areas of Columbia and does not reject the competitive hypothesis. Bencivenga and Smith (1997) constructs a two-period OLG model with the rural. Marcouiller et al. The result is suggestive since many sub-Saharan African nations actually experienced falling agricultural productivity. because the degree of information asymmetry between lenders and borrowers is lower in the close-knit community. compliance with minimum wage regulation. since the rural sector has one advantage: loans for fulfilling opportunities to consume indivisible goods are available more easily. Bencivenga and Smith (1997) and Banerjee and Newman (1998) are particularly worth mentioning. the degree of urbanization could be too 5 Fulginiti and Perri (1997) examine changes in agricultural productivity in 18 developing countries over the period 1961–1985 and find that at least half of these countries have experienced productivity declines in agriculture including all the sub-Saharan African nations in the sample. urban formal. and economic development are related. including Fields (1975). Depending on parameter values. each characterization would capture a part of reality and the present paper's skill-based classification reflects its emphasis on human capital as the determinants of development and the regional and sectoral distributions of workers. individuals are innately skilled or unskilled. and thus some of them stay in the rural sector. the present paper distinguishes the two sectors in terms of skill requirements to workers. due to exogenous institutional factors. By contrast. Empirical evidence supports this characterization: for example. only skilled workers migrate to the urban region. but their types are private information and an adverse selection problem arises. and Ranis and Stewart (1999). Further. are some variants of the Harris and Todaro (1970) model in the sense that the formal sector is characterized as the sector that pays higher wages to identical workers than the informal sector does. physical capital accumulation. This does not lead to full urbanization. Banerjee and Newman (1998) examine implications of differences in technological and institutional conditions between urban modern and rural traditional sectors in development. In their model. Many papers in this literature.5 rapid urbanization associated with the expansion of the informal sector. In this literature. 6 Also. while there exist unexplained sectoral wage differentials after adjustment for workers' characteristics.

and education is required to remain skilled for a fixed units of time. They examine how initial wealth distribution affects the dynamics of the model economy. The present paper is also related to the literature that investigates the interplay between income distribution and growth through human capital accumulation. and. loans available to the urban sector increase and the interest rate falls. He shows that there exists a continuum of steady states that are different in terms of the proportion of skilled workers and wage inequality. and Yuki (2003). its focus is on the mechanism by which initial distribution affects the long-run outcome of an economy and the effect of the sectoral shift of consumption associated with income growth (Engel's law) on development. the final good is produced by skilled labor. Benabou (1996a. urban and rural. an individual receives a transfer from his or her parent. Galor and Zeira (1993) considers a small-open OLG model. Ljungqvist (1993).1. assets and education. and Yuki (2003). as detailed below. unskilled labor. each region has different production sectors and available jobs.1. Then he or she allocates the transfer for two investment options. Durlauf (1996). Childhood In childhood. Galor and Tsiddon (1997). 2. as in the present paper. to produce a single final good. intergenerational transfers motivated by impure altruism (utility from making transfers) constrain educational decisions. whereas. certain amounts of wealth are required to become an entrepreneur or a selfemployed worker. one using skilled labor and physical capital and the other employing unskilled labor. Due to credit market imperfections. Yuki (2003) constructs a similar model with the tradable modern sector and the nontradable traditional sector.80 K. wages are determined by supply factors. In these papers. This is the only difference between the regions in the model: there are free mobility of goods and factors of production. and thus occupational patterns are affected by the distribution of wealth. There are two sectors. entrepreneurs. in the present paper. unskilled wage is affected by demand factors as well. where. Section 5 concludes the paper. while capital is freely mobile internationally. the first period as children and the second period as adults. The economy is composed of a continuum of individuals who live for two periods. Yuki / Journal of Development Economics 84 (2007) 76–103 low: if some of those staying in the rural sector and receiving loans are forced to migrate. and wage earners – in the economy.b). Section 2 presents the model economy and derives equilibrium conditions. This paper is organized as follows. 2. and physical capital.1. Lifetime of an individual 2. Model The model economy is of a discrete-time OLG variety. There are two regions. Section 3 analyzes the model's dynamics and Section 4 presents and interprets the results. Occupational patterns in turn affect future wealth distributions and their interaction determines the fate of the economy. However. There exist three kinds of occupations – self-employed workers. in order to maximize . There is no uncertainty in the model. Ljungqvist (1993) examines a dynamic model with a single production sector where individuals are infinitely lived (or have perfect altruism). Ljungqvist (1993). which includes Galor and Zeira (1993). which in turn promotes migration and raises total output. Closely related are Galor and Zeira (1993). Banerjee and Newman (1993) examine the interaction between occupational choices and the distribution of wealth in the process of development. hence prices are equalized between the regions.

Then.t in the next period. The investment is a discrete choice. His or her generation is called generation t. . and rl in the economy. 9 Actually the relative return from education is determined as a result of people's investment decisions. It is easily shown that. The decision that the child has to make is the allocation of the whole transfers between education and assets. making education optional and incurring a fixed cost. the individual obtains income from assets and labor supply and spends the income on consumption and a transfer to his or her child. Suppose that the individual has received bti units of income as a transfer from the parent. ul. The other option. where wh.t are skilled and unskilled wages in period t. Characteristics of these goods are described later in this section. education. i. and other inter-vivos transfers in real life.t and wl. Then. eit ¼ 0. wh.1.t − wl. since it depends on the numbers of skilled and unskilled workers in the economy. Adulthood At the beginning of adulthood. an individual makes occupational and associated locational choices. the investment in assets. in an equilibrium. Each adult is assumed to have a single child.K. 2. is a continuous choice. and its gross return is wh. the return from the investment in education becomes at least as high as the return from the investment in assets. Note that the transfer in the model corresponds to total intergenerational transfers including bequests.t ≥ (1 + rt)et. Consider an individual who was born into lineage i in period t − 1. He or she allocates the transfer between the investments in assets ati and in education eti in order to maximize future income. The utility function of agent i in generation 7 Alternatively.t − wl. Available jobs are constrained by the human capital investment in the previous period.2. ð1Þ if ait ¼ bit −et . respectively. eit ¼ et . 8 The investment must be self-financed because loan markets for such investment are not available: the child's future income is not considered a valid collateral in the financially underdeveloped economy.1. one can suppose that the investment decisions are carried out by the parent in order to maximize the child's future income.e. There are three different consumption goods. goods uh. 7 The educational investment is required to become a skilled worker and enjoy higher earnings in adulthood. if bit zet : ð2Þ Since innate abilities of individuals are identical. If the return from the educational investment is strictly higher than the one from the asset investment.t − 1:et = sewh. transfers solely determine the investment and resulting occupational choices. optimal investment choices are given by the following equations:9 ait ¼ bit . More formal analysis of the investment decision is described in Section 3. his or her education costs et. and brings a gross rate of return of 1 + rt.t − 1. bit bet . 8 World Bank (1983) reports that about 95% of current expenses in primary school systems of low income countries are teacher salaries. Assume that the education cost is the cost of hiring current skilled workers as teachers and it is proportional to wh. Yuki / Journal of Development Economics 84 (2007) 76–103 81 future income.

t Þgj ðbitþ1 Þgk . The population of each generation is normalized to be one.t ¼ Aul Lu. Sector rl exhibits decreasing returns to its only input Lu. j ð4Þ where pj. Production The urban region has two production sectors. Generational change At the beginning of period t + 1.t þ bitþ1 ¼ wit þ ð1 þ rt Þait .10 Lu. and new children are born into the economy.t ¼ gj fwit þ ð1 þ rt Þait g. sector uh and sector ul. It is assumed that the sector is located in the urban area. ul. ul. and Kt is the amount of physical capital employed in sector uh. personal services. and the transfer to his or her i child bt+1: X uit ¼ jðcij. gj þ gk ¼ 1: ð3Þ j j The individual maximizes the utility function subject to the budget constraint. the rural region has sector rl that employs unskilled workers to produce good rl.t cij. which intends to capture the fact that agricultural productivity is constrained by limited arable land. In contrast. The former sector may be considered as the urban formal sector such as modern manufacturing and the latter sector as the urban informal sector such as petty trading.2. 10 The rest of skilled workers is employed in the education sector.t is the number of urban unskilled workers (employed in sector ul).t ¼ Arl ðLr. 0V arl b1. Hu. current adults pass away.t cij. and rl in period t are given by Yuh. and ati is his or her asset holding in period t. The production functions for sectors uh. 0bauh b1. Sector uh employs skilled workers and physical capital to produce good uh. Since each adult has one child. Yuki / Journal of Development Economics 84 (2007) 76–103 i t depends on the consumption of the three goods cj. bitþ1 ¼ gk fwit þ ð1 þ rt Þait g: ð5Þ ð6Þ 2.3. This sector corresponds to agriculture in real economy. and sector ul employs unskilled workers to produce good ul.1. rt is the interest rate. and rl.t . current children become adults. .t Þarl .t ¼ ðAuh Hu. Good uh serves as a capital good as well.t Þauh ðKt Þ1−auh . and rl) is the productivity in sector j. j = uh. Yrl. Solving the maximization problem gives the following consumption and transfer rules: pj. 2. wti is the wage of the individual.t is the number of rural unskilled workers (employed in sector rl).t. ul.82 K. Yul.t is the number of skilled workers employed in sector uh.t is the price of good j. X pj. Lr. the population is constant over time. and repair services.t. ð7Þ ð8Þ ð9Þ ð10Þ where Aj( j = uh.

The wage of skilled workers is obtained by solving the profit maximization problem of a firm in sector uh:12 ! 1−auh 1−auh auh wh ¼ auh Auh : ð11Þ r The wage rate wh is exogenously determined. the average productivity of labor is higher than the marginal productivity in this sector. which may be considered as service goods produced with technologies intensive in unskilled labor.t ¼ auh À a Auh Hu. As for factors of production. The assumption would be more realistic than the other extreme of the closed market. all children who want to receive and can afford education are able to find their teachers. 13 The market clearing condition of the education sector is given as follows: wh He.e. Wage rates are determined competitively in urban sectors uh and ul. 12 The profit maximization problem gives the following first order conditions: r ¼ ð1−auh ÞðAuh Hu. After a minor modification. Equilibrium Assume that goods uh and rl are tradables and their prices are determined in international markets.t goes to 0. is assumed to be non-tradable. respectively in real economy.K. Hu. Since the wage rate of the sector.t is the number of ‘teachers’ in the education sector and Ht+1 is the number of ‘students’ in period t. labor income is determined so that the product is equally shared among workers. capital is assumed to be freely mobile internationally.13 The wage rate of unskilled workers must satisfy the following equations: wl. approaches infinity as Lr.t Þarl −1 : ð12Þ ð13Þ This assumption reflects the fact that traditional agriculture in developing economies is typically engaged by collectives or family farms. that is. Assume that the education technology is reasonably efficient.t 11 Solving the first equation for Kt and substituting it into the second equation gives Eq.t ¼ pul. Remember that goods uh and rl correspond roughly to manufacturing goods and agricultural goods. By contrast. (11). there exists disguised unemployment. et = e(= sewh).t ¼ eHtþ1 . Yuki / Journal of Development Economics 84 (2007) 76–103 83 2. Normalize the price of good uh to be 1. the assumption is also compatible with share-cropping systems widely observed in Asian countries. in the rural sector rl. Then the equation is satisfied without rationing. . Since the cost of education et is the cost of hiring skilled workers as teachers. good rl is always produced in the economy. and enables the paper to focus on human capital accumulation rather than physical capital accumulation as the primary source of development. se is not very high.t)αrl − 1. ¼ se Htþ1 .t Þauh ðKt Þ−auh . good ul. i.t Þ uh ðKt Þ1−auh . where He. Denote the exogenously determined interest rate by r. thus its price is determined in the domestic market.3. and in the education sector. where the production technology exhibits decreasing returns.11 Thus. it is exogenous as well. and denote the price of good rl by prl. wh. By contrast. that is.t Aul ¼ prl Arl ðLr.t or He. prlArl(Lr. as is assumed in Lewis (1954).

one can obtain pul. wl. from the substitution of Eq. is determined: prl Arl ðLr. unskilled workers might have lower incomes in the competitive market and thus long-run aggregate output could be lower.t.t = αrlprlArl(Lr. because there is free labor mobility between the regions. Hence. The market clearing condition is14 ( ) X pul. and there are no mobility costs of goods or factors of production. not constant over time. as is detailed in later sections.t. is high or skilled workers take most of profits from the sector. Lr.t Þarl −1 ¼ gul f½wh −ð1 þ rÞeŠHt þ ð1 þ rÞBt g : ð1−gul Þð1−Ht Þ−Lr. the existence and uniqueness of Lr.t. LHS N RHS. Further. (13) and Lu. This implies that. Yuki / Journal of Development Economics 84 (2007) 76–103 The wage rates in the two regions are equalized.t ul arl ul ð17Þ By comparing the above equation with Eq. i. although the analysis becomes significantly more complicated.t is the solution to the following equation: prl Arl ðLr. The unskilled wage is.t Þarl −1 ¼ 1 gul f½wh −ð1 þ rÞeŠHt þ ð1 þ rÞBt g   : arl ð1−g Þð1−Ht Þ− 1 þ 1−arl g Lr.t into the above equation.t ¼ gul wh Ht þ wl.e.t → (1− γul)(1 − Ht). which might seem unnatural considering the fact that good ul includes services such as petty trading and personal services in real economy. The price of non-tradable ul. 26. Further.84 K. Lr. a large portion of the aggregate asset may have been invested abroad. .t Aul ¼ gul fwh Ht þ ð1 þ rÞðBt −eHt Þg: Lu.t is smaller under the competitive rural labor market.t ð1−Ht Þ þ ð1 þ rÞ ait : ð14Þ i In this equation. the competitive labor market does not necessarily lead to higher aggregate output in the longer term.t is increasing in Ht and Bt.t. (16). if there do not exist enough investment opportunities within the economy.t Aul Lu. in general. the long-run development of the economy crucially depends on income levels of unskilled workers. most of the qualitative results go through under the alternative assumption. Let Bt be aggregate intergenerational transfers. it is clear that Lr. 16 If unskilled wage in the rural region is determined competitively. there are too many farmers (too few urban workers) under the income sharing.t Aul. (12) into the above equation and solving for pul. and 29 in Section 4. The assumption is made just for simplicity: all the results in later sections except those on sectoral allocations of workers remain intact with the alternative assumption that good ul consumed by individuals in each region is supplied by the nontradable sector located in the region. pul. (5)) over the population. and decreasing in Lu. as Lr.t → 0.t ð16Þ The left-hand side of the above equation is decreasing in Lr. since. the number of unskilled workers in the rural region. since currently skilled workers have paid the cost of education e out of transfers received from their parents.t)αrl − 1. If the profit share in the rural agriculture sector. and the right-hand side is increasing in Lr. 1 − αrl. Ht is the total number of skilled workers and ∑iati is the aggregate asset.t −gul ð1−Ht Þ ð15Þ The price level pul.t = 1 − Ht − Lr. and LHS b RHS. As for sectoral allocations of workers.16 In order to indicate the 14 It is assumed that the urban sector ul supplies good ul to the both regions. Substituting ∑iati = Bt − eHt and Eq. However.t as the solution for the above equation is assured. in terms of static output maximization. Then ∑iati = Bt − eHt is satisfied. as Lr.15 The righthand side of the equation is obtained by aggregating the consumption rule for good ul (Eq.t. as detailed in footnotes 19. 15 Due to free international capital mobility. is determined domestically.

1. as described above. the time evolution of the distribution of transfers must be examined in order to understand how the structure of the economy. initially the dynamics of each variable are analyzed fixing the other.Bt) is an increasing function of Ht and Bt. i. wl. the investment decision .Bt) is also an increasing function of Ht and Bt. ðiÞ ðiiÞ ðiiiÞ ALr ðHt . Bt Þ N0. (12). These results are summarized in the following lemma. from Eq. Hence the dynamics of the aggregate variables are characterized next. is enough to derive the model's implications. if the transfer is at least as large as the cost of education. such as regional and sectoral distributions of workers. Yuki / Journal of Development Economics 84 (2007) 76–103 85 dependence of Lr. He or she allocates the transfer bti between investments in assets ati and in education eti so as to maximize future income. who belongs to generation t. individuals live only for two periods and take part in each market for one period alone. Bt Þ ALr ðHt . On the other hand. and from Eq. Hence. for exposition. then the both dynamics are analyzed together by introducing a phase diagram. Bt Þ Apul ðHt . the transfer is spent only on assets and he or she becomes an unskilled worker. If the transfer is less than the cost of education. and thus it affects the relative return from education and investment decisions. Dynamics The previous section has presented the model and derived the equilibrium conditions.t = wl(Ht. in general. it turns out that directions of motion of the two aggregate variables. Individual dynamics Consider an individual born into lineage i in period t − 1. N0: AHt ABt Awl ðHt . The distribution also affects supplies and demands of goods and their prices and allocations. and distributions of earnings and assets. and consequently consumption and transfer decisions. However. the distribution of transfers over the population determines a proportion of individuals who can afford education. i.e. N0: AHt ABt ð18Þ ð19Þ ð20Þ 3. (13).t = Lr (Ht.t = pul(Ht. The dynamics depend on the time evolution of two aggregate variables that in turn are determined by the dynamics of the distribution of transfers.e. Then. which is a decreasing function of Ht and Bt. which can be obtained without knowledge on the distributional dynamics. the model economy can be considered as a sequence of static economies. What connects these static economies across periods are intergenerational transfers. changes over time. In the model economy. Bt Þ Awl ðHt . This section first derives the dynamic equation relating the current transfer to the next period's transfer within a lineage (individual dynamics). Because of the credit constraint.Bt). transfers directly affect individuals' investment and occupational choices. bti b e. Bt Þ N0. pul. let Lr. 3. production and employment shares of each sector. b0: AHt ABt Apul ðHt .t on Ht and Bt.K. Hence. Although the two dynamics interact. Bt Þ b0. Lemma 1. bti ≥ e. thus each market is populated by individuals of a single generation every period. Further.

even if all of them actually take education. The dynamics of transfers of both skilled and unskilled i workers are described by the single equation bt+1 = bs(bti) (Eq. Yuki / Journal of Development Economics 84 (2007) 76–103 is more complicated. Bt Þ þ ð1 þ rÞbit g: ð22Þ The equation is obtained from the substitution of wti = wl(Frt. this is not the long-run transfer level of a lineage of a currently unskilled worker. all of them take education and become skilled. ati = bti.Bt). i. ðbs Þ*u gk 1−gk ð1þrÞ ½wh −ð1 þ rÞeŠ.Bt) space is derived in the following lemma. The key variable affecting the decision is the fraction of individuals in generation t who have received transfers bti larger than e.Bt) is crucial for aggregate dynamics (detailed later). The economy belongs to one of the two cases depending on levels of two aggregate variables. and rt = r into * Eq. i. since his or her descendants may become skilled workers and Frt and Bt may change over time. Equal opportunity case: In contrast.e. Although this fixed point may appear to ⁎ have no economic importance. Since investment decisions of others affect unskilled wage wl(Ht.Bt). Ht ≤ Frt. Similarly. Frt . Frt and Bt. the dynamic equation linking the received transfer bti i to the transfer given to the next generation bt+1 is bitþ1 ¼ bs ðbit Þugk fwh þ ð1 þ rÞðbit −eÞg: ð21Þ The equation is derived by substituting wti = wh.e. Ht is the solution of wh − (1 + r)e = wl(Ht.Bt).e. not all of them become skilled workers. This case is named the unequal opportunity case since access to the better investment opportunity. when the return from education fails to be higher than the return from assets. the cut-off level of transfer bti = e determines whether an individual becomes a skilled worker or an unskilled worker. (6).Bt). those who have received bti b e. The assumption γ k (1 + r) b 1 is made so that the fixed point for the equation. ati = bti − e. exists. it turns out that the level of bu (Frt. In short. Bt Þu gk 17 1−gk ð1þrÞ wl ðFrt . wh − (1 + r)e N wl(Frt. and rt = r into the transfer rule (6).e. i. Bt Þ. for a currently unskilled worker. Lemma 2. those who have received bti ≥ e. The dividing line separating the two cases on the (Frt. The dividing line between the unequal opportunity case and the equal opportunity case on the (Frt. that is. the corresponding dynamic equation is bitþ1 ¼ bu ðbit . Bt Þugk fwl ðFrt . Frt. and when many individuals have access to education. In this case. and Ht = Frt is satisfied. the number of skilled workers Ht is determined at the point where the two returns are equated. Unequal opportunity case: More precisely. if all of them invest in education.Bt) and the relative return from education. Bt ) space is ( ! 1 ) 1 prl Arl 1−arl 1−gul −Frt − Bt ¼ DðFrt Þu ½wh −ð1 þ rÞeŠ: gul ð1 þ rÞ wh −ð1 þ rÞe ð23Þ 17 In general. i. . wh − (1 + r)e ≤ wl(Frt. Now not all of financially eligible individuals take education and become skilled workers. i. the individual has to take into account their actions in making the decision. For a currently skilled worker.e. when only a small number of individuals can afford education.86 K. The fixed point of the equation for given Frt and Bt is denoted by bu ðFrt . is constrained by received transfers. the return from education is higher than the return from assets. (21)). when the proportion of individuals who can afford education is small. education.

Bt Þð1−Frt Þ þ ð1 þ rÞBt g. □ h i 1 Prl Arl It is assumed 1−gul − wh −ð1þrÞe 1−arl N0.t ¼ prl Arl wh −ð1 þ rÞe  1 1−arl . aggregate transfers Bt and the fraction of individuals who have received transfers bti ≥ e. which cannot be characterized without imposing strong assumptions on the distribution. 3. in the equal opportunity case. These dynamics in turn depend on the time evolution of the distribution of transfers.Bt). In the Appendix. ! 1 1−arl ð24Þ prl Arl Ht ¼ D ðBt Þ ¼ 1−gul − wh −ð1 þ rÞe −1 − gul ð1 þ rÞBt : wh −ð1 þ rÞe ð25Þ Proof.Bt) is satisfied. In the Appendix. Bt ≥ D(Frt ).e. Btþ1 ¼ gk f½wh −ð1 þ rÞeŠFrt þ wl ðFrt .2. wh − (1 + r)e = wl(Ht. i. (21)) and of unskilled workers (Eq.2. For exposition. then their interaction is taken into account at the end.e. 3. which ensures that. Aggregate dynamics □ The previous subsection has shown that the individual dynamics depend on the dynamics of two aggregate variables. Btþ1 ¼ gk ½wh −ð1 þ rÞe þ ð1 þ rÞBt Š.Bt) is satisfied and thus individuals with bti ≥ e become skilled workers and all the goods are produced. i. the information on directions of change of the two aggregate variables is sufficient. Ht can be replaced by Frt in the above equation.wh − (1 + r)e N wl(Frt. in order to derive the model's implications. Yuki / Journal of Development Economics 84 (2007) 76–103 87 Proof. By contrast.  Lr. Hence the dynamic equations are expressed as.K. Dynamics of Bt The dynamic equation linking current aggregate transfers Bt to transfers in the next period Bt+1 is derived by aggregating the individual dynamics of skilled workers (Eq. The economy belongs to the equal opportunity case when Frt or Bt is large enough that Bt ≥ D(Frt) is satisfied. Based on the above lemma. However. In the equal opportunity case. for Bt zDðFrt Þ: ð27Þ ð28Þ . Thus this subsection analyzes the dynamics of the two aggregate variables qualitatively. the number of urban unskilled workers and the number of skilled workers in the equal opportunity case are obtained.1. Frt. wh − (1 + r)e N wl(Frt. for Bt bDðFrt Þ. Bt Þð1−Ht Þ þ ð1 þ rÞðBt −eHt ފ: ð26Þ Remember that in the unequal opportunity case. (22)) over the population: Btþ1 ¼ gk ½wh Ht þ wl ðHt . initially each of them is examined separately fixing the other variable. Lemma 3. for some combinations of positive Frt and Bt.

BÞð1−Frt Þ þ ð1 þ rÞBg: * ðFr Further. From the lemma. Yuki / Journal of Development Economics 84 (2007) 76–103 Fig. Assume γk N se. 1: Lemma 4. From i Eq. are shown to converge to particular levels of aggregate transfers. ABAFrt t Þ N0 is satisfied. Proof. the fraction of the adult population who had received enough wealth to take education. Aggregate transfers. Then. 3. Assume gk ð1þrÞ b1 and let Frt 1 fixed. aggregate transfers converge to B*(Frt ). Lemma 5. when the other aggregate Frt is fixed. It is enough to show that the child of a currently skilled worker can afford education. (ii) If Frt ≥ H**. which is a k ð1þrÞ solution to the following equation: rl rl h B ¼ gk f½wh −ð1 þ rÞeŠFrt þ wl ðFrt . depend on the dynamics of intergenerational transfers of skilled and unskilled workers. For skilled workers. □ The assumption implies that a parent has reasonably strong altruism towards his or her child and the education technology is not very inefficient. (21). Dynamics of Frt The dynamics of Frt.2.2.88 K. In the Appendix. 1−gul   be P A (i) If Frt bH ** u1− 1−g gul − w −ð1þrÞe 1−arl . the following result holds. 1. as is illustrated graphically in Fig. aggregate transfers converge to B** u gk ½wh −ð1 þ rÞeŠ: 1−gk ð1 þ rÞ □ ð29Þ ð30Þ Proof. Dynamics of aggregate transfers for fixed Fr. Frt is non-decreasing over . bt+1 ≥ γkwhNsewh = e. descendants of currently skilled workers can afford education.

Frt+1 N Frt happens only when bu (Frt.18 3. In the Appendix. Bt ÞV e. 18 ⁎ To be more accurate. Joint dynamics of Bt and Frt Now the dynamics of the two aggregate variables are considered together by adding directions of motion of Frt to Fig. To do so. 1−gk ð1 þ rÞ ( 1  1−arl ) ½1−gk ð1 þ rފe prl Arl gk  ð1−gul Þð1−Frt Þ− ( Bt V ) gk gul ð1 þ rÞ e 1−gk ð1 þ rÞ 1 ½wh −ð1 þ rÞeŠFrt . Bt Þu 1−g gð1þrÞ wl ðFrt .Bt) N e continues to hold for many periods. Yuki / Journal of Development Economics 84 (2007) 76–103 89 Fig. ⁎ If bu ðFrt . . ⁎ k bu ðFrt .Bt) N e. Bt = D(Frt) and bu (Frt. The question is whether Frt remains constant or increases over time. 1 that illustrates the dynamics of Bt for given Frt. lineages of current unskilled workers accumulate wealth gradually and Frt starts to increase eventually (see Fig.Bt) on the figure moves as Frt and Bt change.2. Necessary condition for Frt + 1 N Frt.3. one more result ⁎ regarding the relative location of the two critical loci. 2.K. and the answer depends on the level of transfers at the fixed point of the dynamic equation of unskilled workers. 1þr ð31Þ − ð32Þ then Frt+1=Frt. although Frt+1 = Frt is possible if the distribution of transfers {bti} is such that there do not exist unskilled workers satisfying bt i ⁎ +1 ≥ e. Proof. However. Bt Þ. is needed. Bt Þu gk wl ðFrt . the position of bu (Frt. 2). relative to the cost of education e: k Lemma 6. time.Bt) = e. □ ⁎ From the lemma. if bu (Frt.

Analysis 4. ⁎ Lemma 7. where Fr ≤ Fr† is satisfied. feasible combinations of Fr and B must be in the area bound by Fr = 0.90 K. Remember that the region below the locus B = D(Fr) is the unequal opportunity case and H = Fr is satisfied. bu (Frt . (Frss. respectively. it is determined by the current position relative to B = B⁎⁎.B) = e and B = ⁎(Fr) by Fr†. 3 is the resulting phase diagram describing the joint dynamics of Bt and Frt. and in the equal opportunity case.B⁎⁎) and (Frss. The economy must satisfy B ≥ eFr because Fr is defined as the fraction of adult individuals who had received transfers bi larger than the cost of education e in their childhoods. and the region on or above the locus is the equal opportunity case and H = D−1(B) ≤ Fr is true. Steady states and transitional dynamics The inspection of Fig. and B = eFr. In the former steady . the direction of motion of B is determined by the position of current (Fr. 4. Directions of change of B and Fr are expressed with vertical arrows and horizontal arrows. Proof. In the next section. As for the direction of change of Fr. In the figure. 3 reveals that there exist two types of steady states in this economy. Fr = 1. In the unequal opportunity case.B⁎(Fr)). main results of the model are presented with the help of this B diagram.Bss) = (1. Phase diagram.B) relative to B = B⁎(Fr). it is determined by the current ⁎ ⁎ location of (Fr.B) = e. Denote Fr at the intersection of bu (Fr. In the Appendix. Bt ) = e is located below Bt = D(Frt ).B) relative to bu (Fr.1. 3. □ Fig.Bss) = (Fr. Yuki / Journal of Development Economics 84 (2007) 76–103 Fig.

ð34Þ ð35Þ gul : 1−gk ð1 þ rÞ (ii) In the steady state (Frss. and all the individuals hold the same level of wealth. the net wage and wealth of skilled workers are higher than the corresponding ones of unskilled workers.B⁎⁎). as for the distribution of the population over the rural and urban areas in the steady states. (i) In the steady state (Frss. skilled wage (net of the cost of education) is equal to unskilled wage. B⁎⁎ ÞbLr ðFr. B⁎ ðFrÞÞ. . Fr ≤ Fr†. (This is also true for Propositions 3 and 4 in the next subsection. is more urbanized and has a smaller informal sector in the urban area than the unequal opportunity steady states. 1−gk ð1 þ rÞ wh −ð1 þ rÞe ð33Þ Lr ðH ⁎⁎ . B⁎ ðFrÞÞ 1−gk ð1 þ rÞ Proof. By contrast.19 Proposition 1.B*(Fr)). In the Appendix. the number of skilled workers is Fr (≤ Fr† b H⁎⁎). B⁎ ðFrÞÞ 1þ Fr : ð36Þ Lu ðFr. among unequal opportunity steady states. Bss ) = (1.B⁎(Fr)). the following proposition compares the different steady states in terms of the distributions of the population over the two areas and the sectoral compositions of production. B⁎⁎ Þ ¼  prl Arl wh −ð1 þ rÞe 1 1−a rl . it can be shown that the number of unskilled workers in each nontradable sector is smallest at the equal opportunity steady state and. B⁎⁎ Þ ¼ Lu ðH ⁎⁎ . (iii) Among the unequal opportunity steady states.Bss ) = (Fr. That is. (i) The equal-opportunity steady state. one with higher Fr is more urbanized (has smaller Lr(Fr. B⁎ ðFrÞÞ ¼ wl ðFr. the number of unskilled workers in the urban area is & ' gul ½wh −ð1 þ rÞeŠ−wl ðFr. (Frss. the number of skilled workers and the numbers of rural and urban unskilled workers are H ⁎⁎ ¼ 1− 1  1−a rl gul prl Arl − . the qualitative results of the original model are obtained in the modified model as well. in the latter type of steady states.B⁎⁎).) Further. Further. the following lemma is obtained.B⁎(Fr))) and has a smaller informal sector relative to the formal sector in the urban region ( LuLu is lower).Bss ) = (1. Fr ≤ Fr†: Lr ðH ⁎⁎ . Yuki / Journal of Development Economics 84 (2007) 76–103 91 state. the number of total rural workers and the proportion of informal workers in the urban region decrease with Fr. B⁎⁎ ÞbLu ðFr. Bss ) = (Fr. for Fr V Fry : ð37Þ ð38Þ (ii) The outputs of goods rl and ul are lower and the output of good uh is higher in the equalopportunity steady state in comparison to the unequal-opportunity steady states. □ Based on this lemma. þFr 19 The results remain unchanged in the modified model with two regional nontradable sectors supplying good ul (footnote 14) by defining Lu to be the total number of unskilled workers in the two nontradable sectors and Lr to be the number of unskilled workers in the rural tradable sector.K. (Frss. the number of skilled workers is H⁎⁎ = D− 1(B⁎⁎). Lemma 8. B⁎ ðFrÞÞ. and Lu ðH ⁎⁎ .

but this time Lr(H.B⁎⁎) in the long run.B⁎(Fr0 )) in the long run. In contrast. Fr = H remains constant and B increases. while H decreases and Lu(H. (ii) If Fr0 N Fr†. On the transitional path. less dependent on agriculture. the reason why Fr0 is critical for the long-run performance of an economy can be explained as follows. (a) in the region bu (Fr. the distributions of net wage and wealth are equalized). and inequality between skilled and unskilled workers is small (in the model. Yuki / Journal of Development Economics 84 (2007) 76–103 Proof.92 K. both Fr and B increase. many of the urban residents are not skilled and in the informal sector. 3. which is possible if the price of the non-tradable good produced in the urban informal sector (good ul) goes up. and attains lower inequality between skilled and unskilled workers because of higher unskilled wage. thus Lr(H. (b) in the ⁎(Fr. the price of good ul is higher due to the higher 20 Formally. the initial proportion of individuals who can afford education: Proposition 2. (i) If Fr0 ≤ Fr†. □ The equal-opportunity steady state has features of a typical developed economy: many people are educated and skilled. thus Lr(H. From Fig.B) and wl(H. one with a greater portion of skilled workers is more urbanized. has a smaller informal sector relative to the formal sector in the urban region.Bss ) = (Fr0. for the economy to initiate ‘modernization’. The price of the good. the results in (i) hold. the economy approaches one of the steady states in the long run.B) and wl(H. the proportion of people who have enough wealth to take education and aggregate assets must be ⁎ above certain levels.B) region where bu decreases and H and wl(H. . dwl ðFrt .e.B) ≤ e. it can be seen that the critical determinant of success and failure of the economy is Fr0. the number of skilled workers increases. Among such stagnant economies. most of the people live in urban areas and are engaged in non-agricultural activities.B) N e and B b D(Fr) are satisfied. both Fr and B increase. and (c)in the region B ≥ D(Fr). Unskilled wage increases as the degree of urbanization rises through the migration of unskilled workers from ‘over-populated’ agriculture (sector rl) to the urban informal sector.B) N e. and inequality between skilled and unskilled workers is large. which is determined jointly with the number of rural unskilled workers. B⁎ ðFrt ÞÞ Awl ðFrt .21 In this model. accumulates larger wealth. B⁎ ðFrt ÞÞ Awl ðFrt .B) remain the same.B) increase over time. Intuitively.B) increase over time. If the initial distribution of wealth is such that a relatively large portion of the population can afford education and becomes skilled workers. Thus. from Lemmas 1(iii) and 4(i). and larger asset accumulation leads to the higher demand.B) decreases and Lu(H. ⁎ On the transitional path. depends positively on the number of skilled workers and aggregate assets: a greater number of skilled workers implies the higher demand and lower supply of the good. i. In the Appendix.20 Starting from the initial distribution of transfers over the population.Bss ) = (1. the unequal-opportunity steady states characterize a typical stagnant developing economy: many people live in rural areas and are engaged in agriculture. the economy converges to the steady state (Frss. bu (Fr. B⁎ ðFrt ÞÞ dB⁎ ðFrt Þ ¼ þ N0: dFrt AFrt ABt dFrt 21 A more detailed explanation of the mechanics of a related model is found in Yuki (2003). Assume B0 b B⁎(Fr0 ). the economy converges to the steady state (Frss. if a portion of unskilled workers accumulates wealth sufficient for their children to take education and get skilled jobs.B) increases over time.

e. hence the proportion of ‘informal’ unskilled jobs increases 22 Using panel data of wider coverage and of higher quality than those of earlier studies. Deininger and Squire (1998) and Deininger and Olinto (2000) discover that an economy's growth rate is affected negatively by initial land inequality (a proxy for initial asset inequality) and positively by its mean years of schooling per working person (a proxy for human capital). i. Fr ≤ Fr†. skilled labor is scarce. in turn. if the economy starts with relatively small Fr0.B⁎(Fr)). Further. the economy reaches the state in which the return from education is equated with the one from assets. the large pool of skilled workers makes rapid asset accumulation possible. the economy experiences different kinds of urbanization depending on its development path. not GNP per adult. When an economy fails to initiate industrialization and ends up in the unequal-opportunity steady state.Bss) = (Fr. However. and initial land and income inequalities affect negatively the income growth of the poor. An increase in the number of skilled workers and asset accumulation lead to a further increase in the price of good ul. Since skilled labor remains limited. This allows children of less affluent unskilled workers to access education and thus increases the number of skilled workers further. Still. promotes the migration of unskilled workers.. it continues autonomously. 23 The model may be used to understand growth experiences of two contrasting economies. a large portion of differences in enrollment rates of secondary education between the two economies in the early 1970s can be explained by the greater income inequality in Brazil.B0) N e. further asset accumulation increases the relative price. asset accumulation alone is not enough to raise the relative price to the critical level necessary for structural change. i. while Korea continued to grow fast.e. 1995). In the long run. In terms of the model. and the number of rural unskilled workers is smaller. Even when the initial price level is not high enough for the ‘take-off’ due to limited asset accumulation. Both Brazil and Korea experienced rapid growth in the 1970s. stimulate the migration of unskilled workers to the urban region. or school age population ratio.23 As for the issue of internal migration. Korea may be on a path to the equal opportunity steady state and Brazil on a path to an unequal opportunity steady state.K. And Brazil stagnated after the 1980s. 2000). this process continues.e. i. Birdsall and Londono (1998) find that initial inequalities of land and education have strong negative effects on subsequent income growth. the effect of human capital is larger for a lower-income economy. either because wealth is concentrated in the few rich or because limited wealth is dispersed among the many poor. Yuki / Journal of Development Economics 84 (2007) 76–103 93 demand and the lower supply of the good. Once the structural change is initiated. inflates the price and unskilled wage even more. they find that the average educational attainment is negatively affected by initial land inequality. B0 b B⁎(Fr0). unless asset accumulation is very low. Children of unskilled workers are not financially able to obtain education and the number of skilled workers does not increase. At the beginning of the 1960s. which. resulting in much smaller decreases in returns to education (Birdsall et al. As long as skilled wage (net of the cost of education) is higher than unskilled wage. and raises unskilled wages over time. but not of the rich. Their findings are consistent with predictions of the model. (Frss. However. i.22. Also. a richer portion of unskilled workers can send their children to school and the expansion of the modern sector starts immediately. inequality between skilled and unskilled workers does not disappear and the investment choices are affected by family income even in the long run. the income and wealth distributions of Brazil were far less equal than Korea (Deininger and Olinto. thereby attaining a state of equal opportunity. and many of unskilled workers remain in agriculture. B0 b B⁎(Fr0). which is consistent with the fact that urbanization occurred universally for most of the post-war period. According to Williamson (1993). but Brazil attained only a modest increase in the proportion of workers with secondary education from a much smaller starting value. ⁎ resulting in higher unskilled wage. as long as low Fr0 reflects high initial asset inequality. especially that of the poor. hence the price level rises to the critical level eventually and the modernization process begins.e. and boost unskilled wage. In contrast. that is. bu (Fr0. the price of nontradable ul is low. urbanization takes place without an increase in well-paid jobs for skilled workers. the relative price of teachers. if initial asset accumulation is low. . Brazil and Korea. it always experiences urbanization along its transition to a steady state. as long as an economy starts with small asset accumulation. If the price level is above the critical level. the proposition shows that.

Yuki / Journal of Development Economics 84 (2007) 76–103 over time in the urban region. Ranis and Stewart. As for manufacturing. Some evidence indicates that the size of the informal sector and the wage differential between the informal and formal sectors decrease with development through human capital accumulation (Marcouiller et al. the analysis has been performed for given levels of productivities. (ii) Instantaneous effects on sectoral allocations of workers and wages are: 24 Larson and Mundlak (1997) find that the ratio of average labor productivity of agricultural workers to that of nonagricultural workers converges to one as an economy develops. At the initial stage when an economy grows ⁎ without an increase in skilled workers. the productivity increase is examined. (Frss.24 4.B) ≤ e. Consistent with the model's implications. it goes through three stages of urbanization. Once the economy ‘takes off’. As for the instantaneous effects in (ii).25 Proposition 3. which corresponds to services such as trading and housework in real economy.e. Roles of productivities So far. 1997. i.. Then.26 Suppose that an economy experiences a one-time and permanent decrease in Arl at the beginning of a period that is expected in the previous period. (1997) find sizable wage differences between formal and informal sectors after controlling for individual characteristics and industry-specific factors in El Salvador and Peru but not in Mexico in the late 1980s. when an economy is on a track to the equal-opportunity steady state. Marcouiller et al. and urbanization stops once the economy attains equal opportunity. bu (Fr.Bss) = (1. effects of a decrease in the productivity of sector rl. That is. one can see clearly how the productivity changes in the two sectors. and attracts rural unskilled workers into the urban region. thus the share of informal jobs in the urban region increases. At this stage. This subsection examines how changes in the productivities of sector rl (agricultural productivity) and of sector uh (manufacturing productivity) affect the distribution of the population between the regions. By contrast. In this way.e. bu (Fr. but also by an increase in the number of educated people searching for skilled jobs. while the number of workers in the rural nontradable sector decreases and that of the urban informal sector increases by the same amount. ⁎ i. for the same reason. First. Fr ≤ Fr†. 2000). (i) Fr† increases. it urbanizes without an increase in skilled jobs.94 K. i. Wealth accumulation raises the demand for nontradable ul. Regarding the long-run effects in (iii). the following results are additionally obtained. in the unequal opportunity case. the fraction of workers in the informal sector falls in the urban region. urbanization accompanies an increase in skilled jobs. and the dependency of the initial distribution on long-run steady states. the Mexico's income level is higher than the others and the share of the informal sector is smaller. the distribution of land holdings in Mexico was much more equal than the other nations (Deininger and Olinto. the economy is more likely to end up in an unequal opportunity steady state.B⁎(Fr)). increases urban unskilled jobs supplying the good.Bss) = (Fr. .B) N e. as summarized in the proposition.e.2. since. 26 In the modified model with two regional nontradable sectors. the number of workers in the rural nontradable sector falls and that of the urban informal sector rises. when the economy belongs to the equal opportunity case both before and after the productivity change. 1999). An increase in the number of skilled workers and asset accumulation lower wage differentials between skilled and unskilled workers over time. both promoting urbanization. Arl. are investigated. 25 Results are presented for the productivity decline rather than the more frequent productivity increase. in all the steady states. the former promotes urbanization. This result accords with growing urban slum dwellers observed in a typical stagnant developing economy. the fraction of workers in the rural tradable (agriculture) sector declines and the number of workers in the urban informal sector rises. have different implications for development and inequality. At this point the wage differential between the regions disappears as well. urbanization is driven not only by the increasing demand for non-tradable ul. sectoral compositions of production. Higher economic growth of Mexico may be partly due to its more equal initial asset distribution: at the beginning of the 1960s. (Frss.B⁎⁎).

and Lu. the minimum size of Fr0 necessary for successful development increases (Proposition 3(i)). In the Appendix.t increases. they were not sustainable. the decline of unskilled wage and the migration of rural workers are persistent. and has a negative effect on the demand for good ul and thus unskilled wage. while B⁎⁎ and Lu do not change. Although pro-industry policies. 1993). As can be shown easily.B) space). That is. The lower unskilled wage translates into the slowdown of asset accumulation and the reduction of long-run aggregate transfers (B = B⁎(Fr) shifts downward). based on survey data from Ghana.27 Proposition 3(ii) and (iii) show that urbanization is not necessarily associated with income growth. A fall of agricultural productivity depresses farmers' incomes and forces them to leave their lands and migrate to the urban area immediately. a disadvantageous asset distribution in the initial period becomes a more serious hindrance to development. Lu. Further.t and wl. it does not affect the wage and inequalities in the equal opportunity steady state (Proposition 3(iii)(b)). and wl decline. lead to industrialization in the Philippines in the 1970s. Moreover.t and wl. They move to the urban informal sector in the unequal opportunity case and to the modern sector in the equal opportunity case. as agricultural productivity is lower. 1991. 28 Becker and Morrison (1988) find that per capita caloric intake in rural areas is the most important determinant of urbanization in sub-Sahara African economies for the years 1970–80. In contrast. especially in sub-Saharan African nations. Empirical findings suggest that agricultural productivity and prices of agricultural goods are important determinants of internal migration in many stagnant developing countries.28 The divergent development and urbanization experiences of Thailand and the Philippines after the 1980s. As a result. (b) (Equal opportunity case) If the economy belongs to the unequal opportunity case after the productivity decline. □ A fall in agricultural productivity Arl decreases unskilled wage directly and makes it more ⁎ difficult for unskilled workers to afford their children education (bu (Fr. Siamwalla et al. and Ht = Frt does not change.t decreases. Ht and Lu. Thailand started with the more equal distribution of land and higher agricultural productivity (more abundant fertile land) (Inral and Power.t increase. (iii) Long-run effects on sectoral allocations of workers and wages are: (a) (Unequal opportunity steady states) At the steady state with given Fr.t decrease. both of which are conductive to urbanization with development according to the model. Jaeger (1992). shows that policy reforms that shift rural–urban terms of trade in favor of agriculture resulted in a significant reverse migration from urban to rural regions in the late 1980s. the long-run effect on unskilled wage is stronger and the one on migration is weaker than the short-run effects: the lower income due to the productivity decline slows down asset accumulation. 27 .t decrease. which is mentioned in the introductory section. Proof. and various subsidies to industry. Lr. B⁎(Fr). could be due to differences in initial conditions of the two economies. (b) (Equal opportunity steady state) H⁎⁎ increases. overvaluation of the exchange rate. and Lu rises. That is.. Ht increases. Yuki / Journal of Development Economics 84 (2007) 76–103 95 (a) (Unequal opportunity case) Lr. Lr decreases. and Lr. the productivity decline has negative effects on the unequal opportunity steady states: unskilled wage falls and wage and wealth inequalities worsen (Proposition 3(iii) (a)). In particular. when Fr0 is low and the economy remains in the unequal opportunity case in the long run. which reduces the demand for non-tradable ul and unskilled wage.K.B) = e shifts outward on the (Fr. a fall of the price of the agricultural good rl has the same qualitative effects. Lr.t does not change. If it remains in the equal opportunity case. such as export taxes on agricultural products.

in all the steady states.t. the positive productivity shock yields mixed results: while it raises wages of both skilled and unskilled workers (Proposition 4(iii)(a)). when the economy remains to be in the same (unequal or equal opportunity) case after the productivity change. (Frss. see Fig. Lu. As for the instantaneous effects in (ii). and Lr. a skill-biased technological progress. however. (iii) Long-run effects on sectoral allocations of workers and wages are: (a) (Unequal opportunity steady states) At the steady state with given Fr. otherwise. In the short run.B) = e shifts outward on the (Fr. as can been seen from Eq. The effect on Lu. In the Appendix. B⁎⁎.t . wh. while Lr. and wealth in the equal-opportunity steady state (Proposition 4(iii)(b)). Lr decreases. It turns out that the former effect dominates. That is.t decreases. (b) (Equal opportunity steady state) H⁎⁎.Bss) = (1.Bss ) = (Fr. Frnew].B⁎(Fr0)). the economy is more likely to end up in an unequal opportunity steady state. the productivity increase 29 In the modified model with two regional nontradable sectors. and wl. more than unskilled wage and makes it harder for children of unskilled workers to ⁎ accomplish upward mobility (bu (Fr.B) space). although it slows down the convergence process somewhat. which is proportional to skilled wage. wh . and wl. The result suggests that an uneven productivity growth of the more advanced manufacturing sector. the following results are additionally obtained.t decreases. Then. the productivity increase drastically changes the economy's steady state from (Frss. Lr. the productivity growth raises the number of skilled jobs. Fr ≤ Fr†. wh. (36) in Lemma 8. . the number of workers in the rural nontradable sector decreases and the fraction of workers in the informal sector falls in the urban region. 3).t . and wl increase. It also raises total income and long-run aggregate transfers (B = B⁎(Fr) shifts upward). Regarding the long-run effects in (iii). and Lr decreases. □ A rise in manufacturing productivity raises the cost of education. Yuki / Journal of Development Economics 84 (2007) 76–103 The next proposition summarizes effects of a permanent increase in the productivity of sector uh.t increase.t decreases. In particular. Ht. In such an economy. which raises the demand for good ul and unskilled wage. the number of workers in the rural nontradable sector falls and that of the urban informal sector rises. goes up and the economy is more likely to end up in the unequal opportunity steady state (Proposition 4(i). the productivity increase is certainly advantageous. it worsens wage inequality between them.Bss) = (Fr0. † For an economy starting with relatively large Fr0 (Fr0 N Frnew). and Lu does not change. and Ht = Frt does not change.B⁎(Fr)). that is. (ii) Instantaneous effects on sectoral allocations of workers and wages are: (a) (Unequal opportunity case) Lu. 30 On the other hand.t increase.29 Suppose that an economy experiences a one-time and permanent increase in Auh at the beginning of a period that is anticipated in the previous period. Proof. if the initial condition satisfies † † Fr0 ∈ (Frold. and wl. and wl increase. Fr†. Further. thus the critical level of Fr for modernization. wh.30 The increase in manufacturing productivity also promotes urbanization.t is ambiguous. the number of workers in the rural nontradable sector declines and that of the urban informal sector rises. If it remains in the equal opportunity case.t increase. wh.B⁎⁎) to (Frss. Lu.96 K. Ht. (b) (Equal opportunity case) If the economy belongs to the unequal opportunity case after the productivity increase. (i) Fr† increases. Proposition 4. when the initial condition satisfies Fr0 ≤ Frold†. income. is not always beneficial to the long-run development of a low-income nation. B⁎(Fr). the rise of the cost of education does not become an impediment to convergence to the equal-opportunity steady state.

irrespective of k initial conditions. Discussion Some of the model's assumptions are made for the purpose of simplifying analyses rather than for capturing reality. As for the assumption of the single nontradable sector ul.t −crl Þgrl Pj p rl ðcij. 4. In the ¯ unequal opportunity case. 4. skilled wage changes with the ratio of skilled to unskilled workers in the sector. only sector ul operates. as long as sector rl is in .e.e. there is no disguised unemployment. although the analysis becomes more complex. it has been shown in footnotes 14. the economy reaches the equal opportunity steady state in the long run.3. while when Fr and B are high. uit ¼ ðcirl. Urban formal sector's production function When the model is modified so that sector uh (the urban formal sector) employs unskilled workers as well as skilled workers.3. That is. i. In this setting. 4.3. In contrast. The dynamics are also affected. Yuki / Journal of Development Economics 84 (2007) 76–103 97 raises skilled wage. Since the increase in total income stimulates asset accumulation and raises the demand for good ul further in the longer term. i. in the equal opportunity case. when the productivity is low. i. With this modification. resulting in smaller asset accumulation and higher Fr†. 26. the lower degree of ¯ urbanization. hence the cost of education is assumed to be the cost of purchasing a fixed amount of good uh. k If the agricultural productivity is high enough that 1−g gð1þrÞ prl Arl Ne is satisfied. total income. the effects are persistent (Proposition 4(iii)).2. but quantitatively they are affected. 19. the model becomes intractable with the present setting that skilled workers are employed as teachers in the education sector. the results of the original model remain intact qualitatively for the most part. Rural economy When the production function of sector rl (the agriculture sector) exhibits constant returns to scale. while. ¯ Qualitatively. the economy is completely urbanized. as well as the higher demand for good ul and the lower number of unskilled workers. an individual must consume at least crl units of good rl irrespective of his or her income. In a transition to the equal opportunity steady state. and 29 that main results are not affected even if regionally separated nontradable sectors are allowed. The only major difference is that now an increase in Fr = H raises unskilled wage through the higher proportion of skilled workers in sector uh. In real economy. Then. in the unequal opportunity case. the results remain intact with positive crl .K.t Þgj ðbitþ1 Þgk . 4.1. Utility function In the model. the wage rate equals the marginal product of labor.3. the utility function is Cobb–Douglas and thus the proportion of spending on each item is constant and independent of income and price levels. In this subsection. This feature can ¯ be captured by a slightly modified utility function. positive crl implies smaller demand for good ul. the robustness of main results with respect to modifications of several other assumptions and the model's implications for international migration are briefly discussed. both sectors rl and ul exist. and the demand for good ul and encourages the migration of rural workers to the urban informal sector in the unequal opportunity case (Proposition 4(ii)(a)). in the equal opportunity case.e. and lower unskilled wage for given Fr and B.3. H⁎⁎ increases and Lu decreases due to the lower demand for good ul. By contrast. the long-run outcome is determined by initial conditions as in the original model. low Fr becomes a more serious obstacle to development. it promotes migration to the formal sector as well (Proposition 4(ii)(b)). when Fr and B are low. the proportion of the consumption of agricultural goods decreases with development (the Engel's law).

immigrants are disproportionately unskilled workers. exacerbate wage inequality. industrialization. the shrinkage of the informal sector. urbanization proceeds without changes in unskilled wage. many of unskilled workers cannot cover migration costs. (1) In the model economy there exist two types of steady states. but others urbanized without such modernization and underwent the expansion of the urban informal (traditional) sector simultaneously. it urbanizes without such modernization but with the expansion of the informal sector. If an economy starting with limited asset accumulation is on a path to the equal opportunity steady state. 4. and raise the proportion of the informal sector in the urban region. if Fr is only slightly less than Fr† without international migration. This does not imply that unskilled wage and income are higher. the number of workers in the sector is smaller and aggregate output is higher compared to the original setting. The main findings of the paper can be summarized as follows. further easing the burden of low Fr in development. but whether the economy goes through urbanization accompanying modernization or not now depends on the distribution of profits over the population as well. hence international immigration would raise Fr and have the opposite effects to the previous situation. and thus skilled workers would be overrepresented in immigrants. The purpose of the present paper has been to tackle this question analytically based on a dynamic model of urbanization and development. the sectoral and regional distributions of the population. When Fr and B are very low in the absence of international migration.4. If the production function of the rural sector is identical to the original model but the wage is determined competitively. Yuki / Journal of Development Economics 84 (2007) 76–103 operation.98 K. 5. Returns from migration would be greater for unskilled workers. (2) Given an initial distribution of transfers (wealth) over the population. since a portion of the output is distributed as profits. Conclusion Urbanization is a universal phenomenon in the postwar world. The question that naturally arises is what are underlying causes for the differences. Instead. remittances from immigrants may contribute to wealth accumulation by unskilled workers and lower Fr†. the expansion of the formal sector. Some nations have experienced urbanization that accompanied skill upgrading. lower the degree of urbanization. while migration costs would inhibit migration of unskilled workers with limited wealth. and falling inequality. if it is on a track to an unequal opportunity steady state. as is explained in footnote 16. International migration The model has some implications for effects of international migration on the urbanization and development of low income nations. and the expansion of the urban formal (modern) sector. unskilled wage starts to increase. an economy approaches one of the steady states in the long run. The equal opportunity steady state has features of a typical developed economy and the unequal opportunity steady states have features of a typical stagnant developing economy in terms of the accumulation of human capital. and once urbanization is completed. Main results would be qualitatively unchanged. unskilled wage is low. given asset distribution. Migration is motivated economically by returns from migration and migration costs. In particular. when Fr and B are not very low. since the return to education and wage inequality are lower in developed nations. Further. and sectoral and regional income inequalities. but its qualitative nature is varied across countries. Thus the model exhibits two . it experiences urbanization that accompanies skill upgrading. If this is the case.3. On the other hand. the presence of international migration tends to decrease Fr. increased migration allows the economy to attain equal opportunity in the long run.

t −gul ð1 þ rÞ ¼ : arl −2 ABt ½ð1−gul Þð1−arl Þð1−Frt Þ þ arl Lr.Bt). ALr. Bt Þ ALr. ABt ABt ð1−arl Þð1−Frt Þ þ arl Lr. (16). As for Ht.t ¼ −ð1−arl Þprl Arl ðLr.t)αrl − 1 =wh − (1+r)e into Eq. First. Lr. Appendix A. ABt Awl ðFrt . From the definition of the two cases. (42). Bt Þ ¼ gk ð1 þ rÞ þ gk ð1−Frt Þ . Awl ðFrt . It is enough to prove that (a) ARHS b1 and (b) RHS N 0 when Bt = 0 are satisfied. Proofs of lemmas and propositions Proof of Lemma 2. (3) Both a decrease in agricultural productivity and an increase in the formal sector's productivity promote urbanization but make it more difficult for an economy starting with an unfavorable wealth distribution to reach the equal opportunity steady state in the long run. From the proof of the previous lemma. (16) and solving for Lr. ! g ð1 þ rÞBt arl −1 : ð41Þ wh −ð1 þ rÞe ¼ prl Arl − 1−gul −Frt ul wh −ð1 þ rÞe Eq.t ¼ 1−gul −Ht − gul ð1 þ rÞBt : wh −ð1 þ rÞe ð40Þ By plugging Ht = Frt into the above equation and substituting the resulting equation into wh − (1 + r)e = prlArl(Lr. ARHS Awl ðFrt .t Š prl Arl ðLr. (23) is obtained by solving the above equation for Bt. Bt Þ ð1−arl Þgul ð1 þ rÞ ¼ : ABt ð1−gul Þð1−arl Þð1−Frt Þ þ arl Lr. the existence and uniqueness of the solution to Eq. (13).Bt) must hold by the definition of the equal opportunity case.t)αrl − 1 = wh − (1 +r)e. Lr.Bt) is equivalent to Bt = D (Ht). the dividing line is equivalent to wh − (1 +r)e =wl(Frt.t.t = prlArl(Lr.t)αrl − 1.t Hence.K. By substituting prl Arl(Lr. Yuki / Journal of Development Economics 84 (2007) 76–103 99 distinctive patterns of urbanization observed in real economy.t Þ Substituting this equation into Eq. □ Proof of Lemma 3. (29) for given Frt is proved.t Þarl −2 : ABt ABt From Eq. □ Proof of Lemma 4.t ð44Þ ð43Þ ð42Þ . From Eq.t g ð1 þ rÞ V k b1: ¼ gk ð1 þ rÞ 1−gul ð1−gul Þð1−arl Þð1−Frt Þ þ arl Lr. wh − (1 + r)e = wl(Ht. it can be seen that wh − (1 + r)e = wl(Ht. The critical determinant of the long-run outcome of an economy is the initial fraction of the population who is sufficiently wealthy to obtain education.t is obtained from the condition wl.

(28). if the dynamics are always determined by Eq.Bt) space. Otherwise. the fixed point of Eq. B⁎(Frt). ð1−gul Þð1−Frt Þ−Lr. (27) is below the critical level of Bt dividing the two cases on the (Frt. From Eqs. □ Proof of Lemma 6. (28) in the equal opportunity case. Proof of Lemma 7. global convergence to B⁎⁎ is obvious. which is obtained by substituting Bt+1 = Bt = D(Frt).e. (13) into the same inequality (31) and solving it for Lr.t Vð1−gul Þð1−Frt Þ− gk gul f½wh −ð1 þ rÞeŠFrt þ ð1 þ rÞBt g: 1−gk ð1 þ rÞ e ð45Þ ð46Þ Substituting Eq. Bt b D(Frt). i. Yuki / Journal of Development Economics 84 (2007) 76–103 (b) is obviously satisfied. Similarly. The critical level of Frt satisfying B⁎(Frt) = D(Frt) is H⁎⁎. since the RHS of Eq. (23) and (32) the statement is true iff  gk 1−gul −Frt − "  1 # 1−arl prl Arl ½wh −ð1 þ rÞeŠN½1−gk ð1 þ rފ wh −ð1 þ rÞe (   1 ) 1−arl prl Arl gk  e ð1−gul Þð1−Frt Þ− −gk gul ½wh −ð1 þ rÞeŠFrt . If B⁎(Frt) b D(Frt) is satisfied. Finally. ABAFrt t Þ N0 is true. (16) when Ht = Frt. □ . for any lineage i satisfying bti b e.100 K. (22)) must be true. i. Bt converges to B⁎\⁎ globally (see Fig. one can also obtain Lr. proving the inequality.Bt) + (1 + r)bti} b e (from Eq.e. hence Bt converges to B⁎(Frt) globally (see Fig. Thus. 1). (23). (27) in the unequal opportunity case. and by Eq. Eq. Eq. 1). If the dynamics of aggregate transfers are always described by Eq. (28) and solving ⁎ ðFr for Frt. exists and is unique for given Frt. i. The actual dynamics are given by Eq.t z 1  1−a rl prl Arl gk : e 1−gk ð1 þ rÞ ð47Þ □ Then.t 1−gk ð1 þ rÞ f Lr. (29) is increasing in Frt in the unequal opportunity case.e. (27). e 1−gk ð1 þ rÞ 1 1−arl f fgk ½wh −ð1 þ rÞeŠ−½1−gk ð1 þ rފegð1−gul Þð1−Frt ÞNðprl Arl Þ  −arl  −arl 1 1−arl 1−a 1−a  gk ½wh −ð1 þ rÞeŠ rl −gk rl f½1−gk ð1 þ rފeg : Since γk[wh − (1 + r)e] N [1 − γk (1 + r)]e is satisfied by the assumption γkwh N sewh = e.t. Since wl(Frt. while the right-hand side is negative. (29). Bt ≥ D(Frt). bt+1 = γk{wl(Frt. the inequality becomes gk gul f½wh −ð1 þ rÞeŠFrt þ ð1 þ rÞBt g V e. global convergence to B⁎(Frt) is assured by this result as well. (32) is obtained from these two inequalities (46) and (47). Frt+1 = Frt is satisfied iff children of currently unskilled workers cannot i afford education. the solution to Eq. into Eq.Bt) is equal to the right-hand side of Eq. the lefthand side of the above equation is positive. The inequality (31) is derived from this condition.

(29). Then. . As for the second relation Lu ðH ⁎⁎ .t −gul Þ ¼ gul f½wh −ð1 þ rÞe−wl. since the educational investment of currently skilled workers has been made in the previous period. Bt = B⁎(Fr).B⁎⁎). wl. (b) Since the productivity decline shifts Bt = D(Frt).t ŠHt þ ð1 þ rÞBt g: ð54Þ Since Ht rises and wl. (23).t ðLu. Yuki / Journal of Development Economics 84 (2007) 76–103 101 Proof of Lemma 8. wl ðFr. B⁎ ðFrÞÞ Fr. Hence.t into Eq. ðfrom Lemma 8Þ. □ ⁎ Proof of Proposition 3. if the economy remains in the equal opportunity case. B*ðFrÞÞ ¼ gul fwh Fr þ ð1 þ rÞðB*ðFrÞ−eFrÞg: Lu −gul ð1−FrÞ ð48Þ From Eq. the substitution of pul. (30). and pul.B) into Eq. and H⁎⁎ N Fr.tAul = wl.K.t. hence the results follow.Bt) decline with the fall of Arl. B*ðFrÞÞð1−FrÞg: 1−gk ð1 þ rÞ ð49Þ □ By plugging the above equation into Eq. B⁎ ðFrÞÞN0: (ii) is directly obtained from (i). and the second result is from Eq. Proof of Proposition 1. B*ðFrÞÞ. B⁎⁎ N B⁎(Fr) are satisfied for Fr ≤ Fr†. Eq. (ii) By substituting Ht = Fr. outward. From Eqs. Eq. Eq.t Aul = wl(Fr. Regarding Lu. (i) Recall that Fr† is the intersection of B = B⁎(Fr) and bu (Fr. Finally Lu(H⁎⁎.t obviously drops.B⁎⁎) = 1 − H⁎⁎ − Lr (H⁎⁎.t falls. Lu. For such region. the exact form of Fr† is n o h   1 i prl Arl ½1−gk ð1 þ rފe ½1−gk ð1 þ rފ 1−gul − wh −ð1þrÞe 1−arl −gk gul ð1 þ rÞ : ð53Þ Fry ¼ gk gul ½wh −ð1 þ rÞeŠ þ ½1−gk ð1 þ rފ½1−gul −gk ð1 þ rފe (ii)(a) Ht = Frt is not affected. the right-hand side of the equation shifts upward. Ht increases since Ht = Frt rather than Ht ≤ Frt is satisfied now. there is a region that used to be in the equal opportunity case but now belongs to the unequal opportunity case. while the lefthand side shifts downward. The first result of (iii) is straightforward from Lemma 1(i) and ⁎ ðFr the result ABAFrt t Þ N0 of Lemma 4(i). From Eqs.t increases. from Lemma 1. wl ðFr. In contrast. B⁎ ðFrÞÞ ð50Þ ð51Þ ð52Þ f ½wh −ð1 þ rÞeŠ−wl ðFr. Lr(H⁎⁎.Bt) and wl(Ht. into Bt = D(Ht). (29) and (32). (36) of Lemma 8. (23). Lr.B) = e.B) is a decreasing function of its arguments (Lemma 1(i)). B⁎⁎ ÞbLu ðFr. (i) The first relation is derived from the facts that Lr(H. (i)H⁎⁎ is obtained from the substitution of Bt = B⁎⁎. and ABAFrt t Þ N0. (15) gives wl.B⁎⁎) is from Lemma 3. f 1b1 þ ½wh −ð1 þ rÞeŠ−wl ðFr. (36). Lemma 1 ⁎ ðFr (iii). (48) and solving for Lu. (16) and (13). B⁎ ðFrÞ ¼ gk f½wh −ð1 þ rÞeŠFr þ wl ðFr.t decreases. (36) is obtained. Eq. Lr(Ht. the results follow from Lemma 3.

B*(Fr)) increase. 165–187. K. if the economy remains in the equal opportunity state. 631–653. Journal of Development Economics 53 (2).Bt) declines with the increase of wh − (1 + r)e. and growth: macroeconomic implications of community structure and school finance.. pp.. 1998.... 1997. Mobility and labor market segmentation: the urban labor market in El Salvador. the dual economy. the effect on Lu. Newman. S. Inequality as a constraint on growth in Latin America. Lu rises. Then. urban unemployment and underemployment. No tradeoff: efficient growth via more equal human capital accumulation.P. In: Turnham. From Eq. mimeo. and growth. Review of Economic Studies 65 (4). (16). Urbanization Without Growth: A Not So Uncommon Phenomenon. LDC agriculture: nonparametric Malmquist productivity indexes. 582–608.. L. Journal of Political Economy 105 (3). Deininger. hence from Eq.. Job Creation and Economic Policy in Latin America. 1997. Squire. 1998. NBER Macroeconomic Annual 1996. Rural–urban migration. (23). 175–207. J. 2000.).B) lowers the right-hand side of Eq. Then. Morrison. migration. V. unemployment and development: a two-sector analysis.. 1970.. Fay. Theory of persistent income inequality.. Banerjee. R.. and development. Economic Development and Cultural Change 46 (1).R. L. 126–142. A. Tsiddon.. (13). and from Eq. B*(Fr) and thus wl decline. R. Funkhouser. (13). B. R. D. Lemma 8(i). Economic Development and Cultural Change 36 (2). In: Birdsall. 1996a. Fulginiti. Ht increases since Ht = Frt rather than Ht ≤ Frt is satisfied now.... 409–428. 1975. wl. D. OECD. (16). D. E. Journal of Development Economics 2 (2). (36). □ Proof of Proposition 4. 1997. Review of Economic Studies 60 (1). World Bank.K. A. C. 2000. MIT Press. Yuki / Journal of Development Economics 84 (2007) 76–103 (iii)(a) Since the decrease of wl (H. Bencivenga. 11–92.. wl (Ht. Occupational choice and the process of development. (b) If the economy belongs to the unequal opportunity case after the productivity increase shifts Bt = D(Frt).. Deininger. Paris. pp. J. Fields.. C.. 274–298. (13). (29). Durlauf. N. 1996b. Graham. Asset Distribution. N.t declines. A. (iii) (a) Since the increases of wh − (1 + r)e and wl (H.. Foy. Inequality. Then. M.. World Bank. Galor... In: Bernanke.. Journal of Development Economics 52 (2).B) raise the right-hand side of Eq. 111–145. Londono. B. K. 93–124.. Then. G.. 259–278. American Economic Review 86 (3).M.. 1996. New ways of looking at old issues: inequality and growth.102 K. 259–287. mimeo. Journal of Economic Growth 2 (1). (13). Lr(Ht. and job-search activity in LDCs. Gupta. (29). M. Zeira. A. 75–94. Benabou. The distribution of human capital and economic growth. Harris. Journal of Development Economics 57 (2). Benabou. 123–153. Heterogeneity. the result is straightforward from Eq. InterAmerican Development Bank. 1988.. M.. Sabot. (ii) (a) From Eq.). the results follow from Lemma 3 and Eq. Income distribution and macroeconomics. from Eq. (b) Straightforward from Lemmas 4(ii) and 3. (53). G. The determinants of urban population growth in sub-Saharan Africa.). the increase of Lu is straightforward from Eq.R. Opal. Eq. stratification. outward. (36).. and Growth. Perri.S. On the other hand. American Economic Review 60 (1). Inequality and growth. □ References Banerjee. Ross. Birdsall. 584–609. and Eq. 373–390... Cambridge. pp. Smith.E.. O. 1993. 1997. (b) Straightforward from Lemma 4(ii). 35–52. N. B*(Fr) and wl (Fr..t rises. 1993. . MA. A. Todaro.. Lr. Galor. from Eq. 1998. Journal of Political Economy 101 (2). Birdsall. Social Tensions. Beyond Tradeoffs: Market Reforms and Equitable Growth in Latin America. Sabot. C. Information. R. (Eds. Migration.. R. New York. (Eds. Informal sector and informal capital market in a small open less-developed economy. J.R. Journal of Economic Growth 1 (1). Olinto. P.t is ambiguous.Bt) rises. O. Rotemberg. Becker. (Eds. Newman. Unemployment.L. 1997.. C. J. (i) Remember that wh and e = sewh are proportional to Auh. (15). Larrain.. 1995.

The vanishing rural–urban gap in sub-Saharan Africa. In: Ogawa. On the intersectoral migration of agricultural labor.. Mazumdar. Human Resources in Development along the Asia-Pacific Rim. 1954.. 165–187.. (Eds. V. Economic Development and Cultural Change 45 (2). 1–82. and demographic events along the Asia-Pacific rim. World Development 20 (11).. El Salvador.. J. 1988. 1991. 1993.).. Singapore. Economic Development and Cultural Change 45 (2). J. inequality.). C. Wealth Distribution.G. Y. 1991. de Castilla. Jones. International Labor Review 127 (3). G.. pp.. World Bank. Economic development with unlimited supplies of labour. Magnac. V. 139–191. Patamasiriwat. Baltimore.. Rural–urban migration in developing countries. 1993. In: Krueger. Amsterdam. P.. Segmented or competitive labor markets? Econometrica 59 (1). vol. 259–288. Elsevier. World Development Report 1983. A.W... N.G. mimeo. Schiff. In: Mills. (Ed. Mundlak. The Philippines. (Ed. Handbook of Regional and Urban Economics. Formal measures of the informal-sector wage gap in Mexico. A. 367–392. 2. 1997. and Peru. 1992. Economic underdevelopment: the case of missing market for human capital.. L. D.R. V-goods and the role of the urban informal sector in development. Williamson. Weeks. Journal of Development Economics 40 (2). Yuki / Journal of Development Economics 84 (2007) 76–103 103 Inral. 1999.. and Development. Woodruff. J..S.K. pp. Ranis. Jamal. 149–194. D. W. E.. 295–319.. A. Power. Cambridge University Press. Human capital deepening. Williamson. S. M. J. 1997. 2003..H. Oxford University Press. Marcouiller. 1097–1128. Setboonsarng. The Thai Economy in Transition.K. Siamwalla. Valdes. 219–239. F.. 129–158. Economic Development and Cultural Change 47 (2). pp. 1987... P. In: Warr.O.). Yuki. Sectoral Shift. 271–292. Manchester School of Economic and Social Studies 22 (2).. New York. Larson. . D. T.. 1631–1645. Johns Hopkins University Press.). pp.. Agriculture. 1993. D. G. Kyoto University.G. K. The Political Economy of Agricultural Pricing Policy. Cambridge. Oxford University Press. Lewis.W. The causes of Africa's food crisis. Stewart. (Eds.. 1983. Ljungqvist. Jaeger. A.