6 Growth of Small Open-Economies With Capital Accumulation

6 Growth of Small Open-Economies with Capital
Accumulation
It is significant to examine possible effects of trade upon personal income

distribution in a globalizing world economy. It is argued, for instance, that
the rise in inequality between the rich and the poor and the decline of real
wages of the less skilled in the US is closely related to international trade
with low-wage countries.1 Nevertheless, some economists argue that the
role of foreign trade for enlarged inequality is negligible. Keller (2004) ex-
amines the extent of international technology diffusion and channels
through which technology spreads.2 It is shown that productivity differ-
ences explain much of the variation in incomes across countries, and tech-
nology plays a key role in determining productivity. The pattern of world-
wide technical change is determined largely by international technology
diffusion because a few rich countries account for most of the world’s
creation of new technology. Cross-country income convergence turns on
whether technology diffusion is global or local. There is no indication that
international diffusion is inevitable or automatic, but rather, domestic
technology investments are necessary. Winter et al. (2004) examine the
impact of trade policy reform on poverty in developing countries. It is
demonstrated that there is no simple generalizable conclusion about the re-
lationship between trade liberalization and poverty. In the long run and on
average, trade liberalization is likely to be strongly poverty alleviating, and
there is no convincing evidence that it will generally increase overall pov-
erty or vulnerability. But there is evidence that the poor may be less well
placed in the short run to protect themselves against adverse effects and
take advantage of favorable opportunities. In extensively and intensively
connected world markets, workers are confronted with increasing competi-
tion from other countries and capital owners can move their wealth easily
1 See Nahuis (2003: Chap. 2), which surveys some empirical studies on possi-
ble causes for the decrease in the relative wage of low-skilled workers in the US in
the 1980s. For trade and wage inequalities for developing economies, see also
Marjit and Acharyya (2003) and Edmonds and Pavcnik (2006).
2 Other channels of enlarged inequalities are modeled by, for instance, Krug-
man and Venables (1995), Manasse and Turini (2001).

176 6 Growth of Small Open-Economies with Capital Accumulation
to wherever returns appear likely to be the highest. Yet, modern technol-

ogy tends to diminish the demand and therefore the wages for low-skilled
workers, while pushing up the demand for highly-educated specialists. As
globalization is deepening, it is important to provide analytical frameworks
for analyzing global economic interactions. For instance, it is important to
examine how a developing economy like India or China may affect differ-
ent economies and different people in a special economy as its technology
is improved or population is enlarged; or how the global trade patterns
may be affected as technologies are further improved or propensities to
save are increased in developed economies like the US or Japan.
As mentioned in Introduction, classical economists constructed different
trade theories to explain why countries make trade. They argued that coun-
tries make trades due to various reasons under different conditions. They
trade because they are different from each other. These differences may be
either in real terms such as climates, technology and natural resources, or
in monetary variables, such as prices, interest rates and wage rates. Classi-
cal economists proved that it does often benefit a nation to exchange desir-
able things which it cannot produce. Nations may benefit from trading as
each of them may produce things it does relatively well. Nevertheless, we
have so far assumed that total factor supplies are fixed. This chapter intro-
duces capital accumulation into trade models.
As observed by Wong (1995), international factor mobility has received
little attention in the literature of international trade. In most books on in-
ternational trade, international trade is considered nearly synonymous to
international trade in goods. Many trade theorems are obtained when only
goods are allowed to move between countries. Two reasons for the omis-
sion are pointed out. The first is that historically economic relationships
between countries have been determined mainly by the movement of
goods. During most periods of economic history, international factor
movements did not play a significant role in economic evolution. Move-
ments of factors were generally strictly regulated by governments or were
very costly. It happened, though not very frequently, that the gaps between
factor prices in different countries became so large that some significant
movements of factors did occur. The second reason is that international
factor movement can be analyzed by the same framework for studying
trade in goods. The Arrow-Debreu general equilibrium framework can be
extended to examine factor mobility by regarding factors inputs as nega-
tive outputs. Production outputs and factor inputs can thus be analyzed si-
multaneously and symmetrically. In a classical work, Mundell (1957)
shows international trade in goods and factor mobility are substitutes in the
senses that either of them reaches the same world equilibrium and an in-
crease in the volume of one will discourage the volume of the other. This
6.1 The One-Sector Growth (OSG) Model of a Closed Economy 177
result is obtained under static conditions. Factor mobility may not be sub-
stitutive with trade in goods once we perceive economic evolution in a dy-
namic and accumulative way. This becomes evident when we introduce
changeable returns to scale in economic dynamics. It can be seen that in
general, factor mobility and trade in goods cannot be treated in symmetry.
The Mundell substitutability between trade in goods and factor mobility is
generally not valid. This chapter will suggest a dynamic one-commodity
and multiple-country trade model to examine interdependence between
trades and global growth. We analyze trade issues within the framework of
a simple international macroeconomic growth model with perfect capital
mobility.
This chapter studies a few models of small open economies with inter-
national capital mobility. As mentioned before, the main deviation of this
book from traditional approaches in modeling dynamics of international
trade is how to model households’ decision making. Section 6.1 introduces
the one-sector growth (OSG) model of an isolated economy. In the rest of
this book, we use the OSG framework to stand for the one sector growth
model developed in this section and its various extensions. Section 6.2 ex-
amines the Ramsey growth model (which is the most popular approach in
economic growth theory with optimal foundation) also for a closed econ-
omy. As the OSG approach is an alternative approach to the Ramsey ap-
proach, we will also compare the two approaches. Section 6.3 describes
dynamics of a small country economy. An open economy can import
goods and services and borrow resources from the rest of the world or ex-
ports goods and services and lend resources abroad. For convenience of il-
lustration, assume that there is a single good in the world economy and the
price of the goods is unity. Section 6.4 extends the model in Sect. 6.3 to a
multi-regional economy. The model examines economic growth of a
multi-regional small open economy in a perfectly competitive economy.
The national economy consists of multiple regions and each region has one
production sector and one housing sector. Households move freely among
regions, equalizing utility level among regions by choosing housing, goods
and saving. A region’s amenity is endogenous, depending on the region’s
output and population. We explicitly solve the dynamics of the multi-
regional economy. As a concluding remark, Sect. 6.5 discusses the theo-
retical basis for the utility function used in Chap. 6. Section A.6.1 intro-
duces a typical model of a small overlapping-generalizations (OLG) econ-
omy, proposed by Galor. Section A.6.2 studies a small country model
proposed by Ikeda and Gombi to study the equilibrium dynamics of sav-
ings, investment and the current account. Section A.6.3 proves Lemma
6.4.1. Section A.6.4 studies the Keynesian consumption function and ex-
amines its possible relations to the consumption function obtained from the
OSG approach. Section A.6.5 introduces the Solow growth model and ex-
amines its possible relations to the OSG growth model.
6.1 The One-Sector Growth (OSG) Model of a Closed

Economy3
The main deviation of this book from traditional approaches in modeling

dynamics of international trade is how to model households’ decision mak-
ing. Although this approach has been extensively explained by Zhang,4 this
section introduces the one-sector growth (OSG) model of an isolated
economy. In the rest of this book, we use the OSG framework to stand for
the one sector growth model developed in this section and its various ex-
tensions.
6.1.1 The Model
We consider an economy of one production sector and one type of house-

holds. With regard to production, almost all the aspects of the OSG model
are similar to the standard Solow and Ramsey one-sector growth models. It
is assumed that there is only one (durable) good in the economy under
consideration. Households own assets of the economy and distribute their
incomes to consume and save. Production sectors or firms use inputs such
as labor with varied levels of human capital, different kinds of capital,
knowledge and natural resources to produce material goods or services.
Exchanges take place in perfectly competitive markets. Production sectors
sell their product to households or to other sectors and households sell
their labor and assets to production sectors. Factor markets work well. Sav-
ing is undertaken only by households, which implies that all earnings of
firms are distributed in the form of payments to factors of production, la-
bor, managerial skill and capital ownership.
First, we describe behavior of the production sector.5 Time is repre-
sented continuously by a numerical variable which takes on all values from
zero onwards ( t ≥ 0 ). Let K (t ) denote the capital existing at each time t
and N (t ) the flow of labor services used at time t for production. Capital
is malleable in the sense that one need distinguish neither its previous use
3 This section is based on Zhang (2006a: Chaps. 1 and 2).

4 See, for instance, Zhang (2005a, 2006b).
5 The description of behavior of producers and production sectors follows the
traditional approach (e.g., Burmeister and Dobell, 1970, and Azariadis, 1993).
nor the factor productions of its previous use. We use the conventional
production function to describe a relationship between inputs and output.
The function F (t ) defines the flow of production at time t . The produc-
tion process is described by neoclassical production function6:
F (t ) = F (K (t ), N (t ) ). In the rest of this book, we omit time subscripts and
in the subsequent analysis whenever no ambiguity results.
We assume that the production function exhibits constant returns to
scale.7 It is straightforward to check that a linear homogenous production
has the following properties:
(i) The production function can be written in terms of per capita output as
a function of per capita capital
F K (6.1.1)
= F (k , 1) ≡ f (k ), k ≡ .
N N
Output per worker depends only on the amount of capital employed by one
worker. Equation (6.1.1) is called the intensive form of the aggregate pro-
duction function. It is also referred to as the per-worker production func-
tion.
(ii) The slope of f (k ) represents the marginal product of capital, i.e.

∂F ∂ (F / N )
FK = = = f ' (k ) > 0 .
∂K ∂ (K / N )
(iii) The marginal product of labor can be obtained by f (k ) as follows
6 A production function F ( K , N ) is called neoclassical if it satisfies the following

conditions: (1) F (K , N ) is non-negative if K and N are non-negative; (2)
F (0 , 0) = 0 ; (3) marginal products, FK and FN are non-negative; (4) there exist
second partial derivatives of F with respect to K and N ; (5) the function is
homogeneous of degree one: F (λK , λN ) = λF (K , N ), for all non-negative λ ;
(6) the function is strictly quasi-concave.
7 For the production function F (K , N ) we define the homogeneity of degree
n for capital and labor inputs as follows: F (λK , λN ) = λn F (K , N ), where λ is an
arbitrary non-negative number. When n = 1, we say that the production function
has constant returns to scale. It is linearly homogeneous or homogeneous of
degree one.
∂F ∂ (NF / NK ) ∂ ( f (k ) / k )
FN = = = = f (k ) − k f ' (k ) > 0 .
∂N ∂(N / K ) ∂ (1 / k )
(iv) The Euler Theorem holds
KFK + NFN = F .
We assume (identically numerous) one production sector. Its goal of

economic production is to maximize its current profit
π (t ) = p(t ) F (t ) − (r (t ) + δ k ) p(t ) K (t ) − w(t ) p(t ) N (t ) ,
where p (t ) is the price of product, r (t ) is the real rate of interest, w(t ) is
the real wage rate, and δ k is the fixed depreciation rate of capital. We as-
sume that the output good serves as a medium of exchange and is taken as
numeraire. We thus set p (t ) = 1 and measure both wages and rental flows
in units of the output good. The rate of interest and wage rate are deter-
mined by markets. Hence, for any individual firm r and w are given at
each point of time. The production sector chooses the two variables K and
N to maximize its profit. Maximizing π with regards to K and N as
decision variables yields
r + δ k = FK = f ' (k ), w = FN = f (k ) − kf ' (k ), (6.1.2)
in which k (t ) ≡ K (t ) / N (t ). We assume that factor markets work quickly

enough so that our system always displays competitive equilibrium in fac-
tor markets. Since we assumed that the production function is homogenous
of degree one, we have KFK + NFN = F , or
rK + wN + δ k K = F . (6.1.3)
This result means that the total revenue is used up to pay all factors of
the production.
We now describe behavior of consumers. Consumers obtain income
Y = rK + wN , (6.1.4)
from the interest payment rK and the wage payment wN . We call Y the
current income in the sense that it comes from consumers’ daily toils
(payment for human capital) and consumers’ current earnings from owner-
ship of wealth. The current income is equal to the total output as we ne-
glect any taxes at this initial stage. The sum of income that consumers are
using for consuming, saving, or transferring are not necessarily equal to
the temporary income because consumers can sell wealth to pay, for in-
stance, the current consumption if the temporary income is not sufficient
for buying food and touring the country. Retired people may live not only
on the interest payment but also have to spend some of their wealth. The
total value of wealth that consumers can sell to purchase goods and to save
is equal to p(t )K (t ) (with p(t ) = 1 ). Here, we assume that selling and
buying wealth can be conducted instantaneously without any transaction
cost. The disposable income is equal to
Yˆ (t ) = Y (t ) + K (t ). (6.1.5)
It should be noted that in Eq. (6.1.5), like Y (t ), the value of wealth, K (t ),

is a flow variable. The disposable income is used for saving and consump-
tion.
At each point of time, consumers would distribute the disposable in-
come between saving, S (t ), and consumption of goods, C (t ). The budget
constraint is given by
C (t ) + S (t ) = Yˆ (t ). (6.1.6)
In our model, at each point of time, consumers have two variables to de-
cide. A consumer decides how much to consume and to save. Equation
(6.1.6) means that consumption and savings exhaust the consumers’ dis-
posable personal income. The slope of the budget line is equal to − 1, i.e.,
dS / dC = − 1.
We assume that utility level, U (t ), that the consumers obtain is depend-
ent on the consumption level of commodity, C (t ), and the saving, S (t )
U (t ) = U (C (t ), S (t )). (6.1.7)
A typical consumer is to choose his most preferred bundle (c(t ), s (t )) of

consumption and saving under his budget constraint. Here, c ≡ C / N and
s ≡ S / N . The utility maximizing problem at any time is defined by
Max U (c(t ), s(t ))
c,s ≥ 0
s.t.:
c(t ) + s(t ) ≤ yˆ (t ), (6.1.8)
in which yˆ (t ) ≡ Y (t ) / N (t ) + k (t ). The following theorem holds.

Proposition 6.1.1
Let U (c , s ) : R+2 → R1 be a C 1 function that satisfies the monotonicity as-
sumption, which says that ∂U / ∂c > 0 and ∂U / ∂s > 0 for each (c , s )
satisfying the constraint set in problem (6.1.8). Suppose that (c * , s * )
maximizes U on the constraint set. Then, there is a scalar λ * > 0 such that
∂U * *
(c , s ) ≤ λ * , ∂U (c* , s* ) ≤ λ * .
∂c ∂s
We have ∂U / ∂c = λ * if c* ≠ 0 and ∂U / ∂s = λ * if s * ≠ 0 . If both
c* > 0 and s * > 0, then
∂U * *
(c , s ) = λ * , ∂U (c* , s * ) = λ * .
∂c ∂s
Conversely, suppose that U is a C 1 function, which satisfies the
monotonicity assumption and that (c * , s * ) > 0 and the first order condi-
tions. If U is C 2 and if
0 1 1
H = 1 U cc U cs = 2U cs − U cc − U ss > 0 ,
1 U sc U ss
then (c * , s * ) is a strict local solution to the utility maximization problem. If

U is quasiconcave and ∇U (c, s ) for all (c, s ) ≠ (c * , s * ), then (c * , s * ) is a
global solution to the problem.
The proof of this proposition and other general properties of the prob-
lem can be found in standard textbooks of microeconomics or mathemati-
cal economics.8 We require that U is a C 2 function, and satisfies
U c > 0, U s > 0 for any (c , s ) > 0 . It can be shown that that
0 < ds / dyˆ < 1 and 0 < dc / dyˆ < 1 in the case of U sc ≥ 0 under the sec-
ond-order condition of maximization. We denote an optimal solution as
function of the disposable income
(c(t ), s(t )) = (c( yˆ (t )), s( yˆ (t ))).
8 See Chiang (1984), Mas-Colell, et al. (1995), and Simon and Blume (1994).
The vector (c( yˆ (t )), s ( yˆ (t ))) is known as the Walrasian (or ordinary or
market) demand function, when it is single-valued for all positive dispos-
able income.
If the utility function is taken on the Cobb-Douglas function
U (t ) = c ξ (t )s λ (t ), ξ + λ = 1, ξ , λ > 0 , (6.1.9)
where ξ and λ are parameters, it is straightforward to solve the optimal

choice of the consumers as
c(t ) = ξyˆ (t ), s (t ) = λyˆ (t ). (6.1.10)
It appears reasonable to consider population as independent of economic
conditions, as a first approximation. We assume that the population dy-
namics is exogenously determined in the following way
N& (t ) = nN (t ),
where n is a constant growth rate of N .
The change in the households’ wealth is equal to the net savings minus
the wealth sold at time t , i.e.
K& (t ) = s ( yˆ (t ))N (t ) − K (t ). (6.1.11)
Inserting
K& (t )
k&(t ) = − nk (t ),
N (t )
in Eq. (6.1.11) yields
k&(t ) = s ( yˆ (t )) − (1 + n )k (t ),
where
Yˆ (t )
yˆ (k (t )) ≡ = f (k (t )) + k (t ).
N (t )
In a stationary state
s( yˆ ( k ) ) = (1 + n )k .
It can be shown that this equation has a unique solution.
Theorem 6.1.19
Given a neoclassical production function and a utility function that is a C 2
function, and satisfies U c > 0, U s > 0 for any (c(t ), s(t )) > 0 . Let the
bordered Hessian be positive for any nonnegative (c(t ), s (t )). Then the
capital-labor ratio converges monotonically to a unique positive steady
state. The unique stationary state is stable.
The stability guaranteed above is local. We now show that if s( yˆ ) is

concave in ŷ , then the system is globally stable. Because of
d 2 c / dy 2 = − d 2 s / dy 2 by equation 1 = dc / dyˆ + ds / dyˆ , concavity of s
implies convexity of c . From the first-order conditions, it is straightfor-
ward to give that conditions under which s is concave, we omit the ex-
pression because we lack a clear economic interpretation. Asymptotical
stability can be proved by applying Lyapunov’s theorem.10
We illustrate dynamics of capital-labor ratio in Fig. 6.1.1. The dynamic
system has a unique stable equilibrium.11
k&(t )
0 k* k
k& = s ( yˆ ) − (1 + n )k
Fig. 6.1.1. Evolution of capital-labor ratio in the OSG model
9 The theorem is proved in Zhang (2006a, Chap. 2).

10 See Zhang (1991, 2005a).
11 The proof is referred to, for instance, Burmeister and Dobell (1970), or
Zhang (2005a).
6.2 The Ramsey Growth Model and the OSG Approach 185
6.2 The Ramsey Growth Model and the OSG Approach12
Section 6.1 introduces the OSG model. This section examines the Ramsey
growth model (which is the most popular approach in economic growth
theory with optimal foundation). As the OSG approach is an alternative
approach to the Ramsey approach, we will compare the two approaches.
The Ramsey growth model is a neoclassical model of economic growth
based primarily on the work of Frank Ramsey. Keynes considered Ram-
sey’s 1928 classic paper to be
one of the most remarkable contributions to mathematical economics ever

made.13
The influences of Ramsey’s classical contribution are reflected in the

fact that almost all the contemporary dynamic models of national, urban,
interregional, or international economic growth with microeconomic foun-
dation for behavior of households are based on the paper and its variations
(like the overlapping-generations model in discrete version).14
Most aspects of the Ramsey model are similar to the OSG model de-
fined in Sect. 6.1. The variables, F (t ) , K (t ) , N (t ) , k (t ) , w(t ) , r (t ) , in
the Ramsey model have the same meanings as in the OSG model. The
production process, marginal conditions, population growth are the same
as in the OSG model. We now describe households’ behavior. In the Ram-
sey approach, households’ decisions on saving are represented by assum-
ing that consumers maximize the discounted value of their flow of utility,
using a constant rate of impatience. The extended family is assumed to
grow at an exogenously given rate n . Let the number of adults at time 0 be
unity, the family size at time t is N (t ) = e nt . The household’s preferences
are expressed by an instantaneous utility function, u (c(t )), where c(t ) is
the flow of consumption per person, and a discount rate for utility, denoted
by ρ . For simplicity, specify u (c ) as
12 This section is based on Zhang (2005a: Chap. 2).

13 Macroeconomic Dynamics (2006: vol. 10). The early literature on the Ram-
sey model, is referred to Ramsey (1928), Cass (1965), and Koopmans (1965).
14 Although the acceptance of this framework has not been based on any pro-
found reason or empirical test, any valid alternative to the dominant framework
will rationally meet with obstacles.
c1 − θ (t ) − 1 (6.2.1)
u (t ) = , θ > 0,
1−θ
where θ is a parameter, u ' > 0 , u" < 0 , and u satisfies the Inada condi-
tions: u ' → ∞ as c → 0 and u ' → 0 as c → ∞ .
Assume that each household maximizes utility U as given by
∞
U = ∫ u (c(t ))e nt e − ρt dt , c(t ) ≥ 0 , t ≥ 0 .

0
The household makes the decision subject to a lifetime budget con-

straint. We denote the net assets per household by k (t ) which is measured
in units of consumables. The total income at each point of time is equal to
w + rk . The flow budget constraint for the household is
k& = w + rk − c − nk = f (k ) − c − nk . (6.2.2)
The equation means that the change rate of assets per person is equal to
per capita income minus per capita consumption and the term, nk .
The first-order conditions are
∂J
= 0 ⇒ λ = u ' e −( ρ − n ) t ,
∂c
dλ ∂J dλ (6.2.3)
=− ⇒ = − (ρ − n )λ ,
dt ∂k dt
where λ is the present-value shadow price of income. By Eq. (6.2.3), we
can derive
u" c  1 dc  (6.2.4)
r=ρ−  .
u '  c dt 
This equation says that households choose consumption so as to equate
the rate of return r to the rate of time preference ρ plus the rate of de-
crease of the marginal utility of consumption u ' due to growing per capita
consumption c. Inserting Eq. (6.2.1) in Eq. (6.2.4) yields
r−ρ f'− ρ (6.2.5)
c& = c (t ) = c.
θ θ
The trajectory of the economy is determined by Eqs. (6.2.2) and (6.2.5).

The phase diagram in c(t ) and k (t ) is shown by Fig. 6.2.1. Along the ver-
tical line defined by f ' (k * ) = ρ , the change rate of consumption per capita
is equal to zero, that is, c& = 0 . The consumption per capita increases to the
left of the curve and falls to the right. Along the locus defined by
c = f − nk , the change rate of the capital-labor ratio equals zero. The
capital-labor ratio falls above the curve and increases below it. With the
requirement ρ > n (without which the utility becomes unbounded along
feasible paths), the intersection of the two curves determines a unique
steady state, (k * , c * ).
c(t )
c&(t )
c*
k&(t )
k* k (t )
Fig. 6.2.1. The dynamics of the Ramsey model
The two eigenvalues are given by
(ρ − n ) ± (ρ − n )2 − 4 f "c / θ (6.2.6)
φ1, 2 = .
2
The Ramsey model is controlled by a system of two differential equa-
tions. Together with the initial conditions and the transversality condition,
this system determines the path of the two variables. At stationary state,
the per capita variables, k , c and y ( ≡ Y / N ), grow at the rate, 0, and
the level variables, K , C and Y , grow at the rate, n. It can be shown that
the system has a unique steady state. Since the two eigenvalues have the
opposite signs, the system is locally saddle-path stable.15
The dynamical behavior of the Ramsey model is controlled by
k& = w + rk − c − nk = f − c − nk ,
r−ρ f ' (k ) − ρ (6.2.7)

c& = c= c.
θ θ
If we find some equation of preference change in the OSG model to
generate the same behavior as Eq. (6.2.7), then the two systems should ex-
hibit the same behavior in terms of consumption, capital accumulation and
incomes, even though they are built on different assumptions.
We now consider consumption of the OSG model when the utility func-
tion is specified as in (6.1.9). The consumption per capita in the OSG
model is given by
c(t ) = (1 − λ (t ))[ f (k (t )) + k (t )].
Differentiation of this equation with respect to time yields
c& f'+δ & λ& (6.2.8)
= k− .
c f (k ) + δk 1− λ
For Eqs. (6.2.7) and (6.2.8) to be equal, it is sufficient for λ (t ) to evolve

according to
f' +1 & f' − ρ (6.2.9)
λ& = ξk − ξ.
f +k θ
The propensity to own wealth λ tends to rise (fall) when k& rises (falls);
it tends to rise (fall) when r < (>) ρ . We may interpret that the direction of
change in λ is influenced by the direction of change in wealth as well as
whether the rate of return of wealth is larger or smaller than the rate of
time preference. If the wealth is increasing and the rate of time preference
is larger than the rate of return, then the propensity to save will definitely
rise. If the wealth is falling and the rate of time preference is smaller than
the rate of return, the propensity tends to fall. In the other cases, the pro-
pensity may either increase or decrease.
15 We will not provide a complete analysis of the model. Refer to Takayama
(1985) and Romer (1996) in detail.

Under Eq. (6.2.9), the consumption per capita in the OSG model
evolves in the same way as in the Ramsey model. We now examine the
fundamental equation of the OSG, i.e.
k& = λf (k ) − (1 − λ + n )k .
By c = (1 − λ )( f + k ), the above equation can be rewritten as
k& = f (k ) − c − nk . (6.2.10)
Theorem 6.2.1
Let the production sectors be identical in the OSG model and the Ramsey
model. If the propensity to save, λ (t ), evolves according to Eq. (6.2.9),
then the OSG model generates the same dynamics of capital-labor ratio,
k (t ), and per-capita consumption, c(t ), as the Ramsey model does.
This example illustrates how the Ramsey model is related to the OSG
model. We can similarly examine relationships between the two ap-
proaches when utility functions are taken on other forms.
We now explain another difference between the OSG and Ramsey ap-
proaches. The OSG model determines consumption as follows
c& = (1 − λ )( f ' + 1)k& ,
where λ is a constant. The change rate of consumption is positively re-
lated to the rate of interest. The rational household has increasing, station-
ary, or decreasing consumption according to whether the wealth rises, is
stationary, or falls. The consumer adapts consumption level not according
to the difference between the interest rate and discount rate for utility as
the Ramsey model predicts, as shown below. According to the this model,
a Japanese consumer would consume more, irrespective of low interest, if
his wealth increases; he would consume less, irrespective of high interest
rate, when his wealth falls. On the other hand, the Ramsey model predicts
f ' (k (t )) − ρ
c&(t ) = c(t ).
θ
This implies that the difference between r and ρ determines whether
households choose a pattern of per capita consumption that rises, stays
constant or falls over time. The optimizing household has increasing, sta-
tionary, or decreasing consumption according as the current real interest
rate exceeds, equals, or falls short of the utility discount rate. According to
this result, consumption always falls if the interest rate is low and the util-
ity discount rate is high.
It should be noted that in 1937 Paul Samuelson published an article on
discounted utility. Since then, the discounted utility was rapidly adopted as
the framework of choice for intertemporal decisions. It is worthwhile to
cite from Samuelson’s following cautions:
Any connection between utility as discussed here and any welfare concept is
disavowed.16
It is completely arbitrary to assume that the individual behaves so as to maxi-
mize an integral of the form envisaged in [the discounted utility model].17
It is easy to see what Samuelson means by looking at the form of utility

formulation in the Ramsey optimal growth theory
∞
∫ U [C (t )]e
− ρt
dt .
0
The specified form means that the household’s utility at time 0 is a

weighted sum of all future flows of utility. The parameter, ρ (≥ 0), is de-
fined as the rate of time preference. A positive value of ρ means that utili-
ties are valued less the later they are received. There are two assumptions
involved in the Ramsey model. The first is that utility is additional over
time. Although we may add capital over time, it is unrealistic to add utility
over infinite time. Intuitively it is not reasonable to add happiness over
time. It is well known in utility theory that when we use utility function to
describe consumer behavior an arbitrary increasing transformation of the
function would result in identical maximization of the consumer at each
point of time. Obviously, the above formulation will not result in an iden-
tical behavior if U is subjected to arbitrarily different increasing transfor-
mations at different times. The second implication of the above formation
is that the parameter ρ is meaningless if utility is not additional over
times. It should be noted that Ramsey considered the meanings of this pa-
rameter from ethical perspectives. Ramsey interpreted the agent as a social
planner, rather than a household. The planner chose consumption and sav-
ing for the current and future generations. Ramsey assumed ρ = 0 and
16 Samuelson (1937: 161). It is nowadays common to use the utility in modeling

and comparing welfare.
17 Samuelson (1937: 159). This reservation does not affect the dominant role of
the discounted utility. Almost all the papers involved with intertemporal decisions
in well-cited theoretical economic journals use the utility concept.
considered ρ > 0 “ethically indefensible”. If ρ = 0, by equation con-

sumption per capita always grows if the interest rate is positive irrespective
of whether wealth grows or falls.
The Ramsey framework has not been proved to be effective when we
take account of variations in households’ preferences. As empirical studies
have convinced existence of great differences in impatience among house-
holds and the Ramsey approach proves futile for dealing with the issues, it
is necessary to search alternative ways to explain the reality. In fact, as
shown in a recent survey on studies of estimating individuals’ discount
rates by Frederick et al.,18 rates differ dramatically across studies, and
within studies across individuals. There is no convergence toward an
agreed-on rate of impatience. It is estimated, for instance, by Warner and
Pleeter19 that individual discount rates vary between 0 and 70 percent per
year. As observed by Frederick et al.
The [discounted utility] model, which continues to be widely used by econo-

mists, has little empirical support. Even its developers – Samuelson, who origi-
nally proposed the model, and Koopmans, who provided the first axiomatic
derivation – had concerns about its descriptive realism, and it was never em-
pirically validated as the appropriate model for intertemporal choice.
… [D]eveloping descriptively adequate models of intertemporal choice will not
be easy.20
Although the validity of the Ramsey approach has been questioned over
years, it is quite another matter to create a “more effective” alternative.21
Instead of searching for another approach, the main attitude toward the ap-
proach is illustrated by Turnovsky (2000: 273)
Any model as widely employed as the representative agent model begins to

take on a life of its own and to be accepted almost as an axiom. It is therefore
18 Frederick et al. (2002).

19 Warner and Pleeter (2001). The studies, for instance, by Rader (1981) and
Jouini and Napp (2003) also hold that there is no reason to believe that different
consumers have identical time preferences for utility streams. It should be re-
marked that Becker (1992) first observed that if individuals have heterogeneous
constant rates of impatience, the representative agent will not in general use a con-
stant rate to discount the future.
20 Frederick et al. (2002: 393-84).
21 As rational economics predicts, the acceptance of an alternative approach to a
dominant idea like the discounted utility may take years or decades. Keynes
(1936) says: “the difficulty lies, not in the new ideas, but in escaping the old ones,
which ramify … into every corner of our minds.”
useful to remind ourselves periodically of its limitations. Despite the criticisms

that have been made, we feel that the representative agent model provides a
useful framework that offers a good deal of insight, and we shall continue to
develop it further.
This “reformist” attitude of refinement and extensions will not solve es-
sential problems in the approach.
This book will use an alternative utility function not only because the
validity of the discounted utility concept has been questioned from phi-
losophical, psychological or/and empirical aspects,22 but also because
many obviously significant issues, such as growth with heterogeneous
households and growth with interregional dynamics, can hardly be prop-
erly discussed with the concept as having become evident in the history of
theoretical economics in the last 40 years. Since the Ramsey approach has
been so influential in the contemporary monetary growth theory, I will de-
scribe monetary growth models within the approach in the main context.
6.3 A Small Open Economy with Capital Accumulation
This section describes dynamics of a small country economy.23 An open

economy can import goods and services and borrow resources from the
22 A comprehensive survey on time discounting and time preference is given by

Frederick et al. (2002). It is observed that there is a growing list of anomalies for
the discounted utility model – patterns of choice that are inconsistent with the
model’s theoretical predictions. Scholars have also proposed different models to
replace the discounted utility model; nevertheless no one has been successful in
the replacement. As observed by Frederick et al., when this model “eventually be-
came entrenched as the dominant theoretical framework for modeling intertempo-
ral framework for modeling interremporal choice, it was due largely to its simplic-
ity and its resemblance to the familiar compound interest formula, and not as a
result of empirical research demonstrating validity.” This book is based on an al-
ternative utility approach to handle with intertemporal choice problems without
reference to the compound interest formula. Truth should be expressed with sim-
plicity and beauty.
23 Refer to, for instance, Song (1993), Lane (2001), Koolmann (2002), Obstfeld
and Rogoff (1995b), Benigno and Benigno (2003), and Galí and Monacelli (2005),
for the literature of open economies. It can be seen that the model here can be
generalized and extended in different directions. Nevertheless, this book will not
deal with open economies when we study international trade. We try to treat every
economy as a part of the integrated whole. As shown later on, as we can develop a
global economy model of any number of economies, it is not necessary, at least
technically, to be concerned with small open economies.
6.3 A Small Open Economy with Capital Accumulation 193
rest of the world or exports goods and services and lend resources abroad.
For convenience of illustration, assume that there is a single good in the
world economy and the price of the goods is unity.24
6.3.1 The Model with General Production and Utility Functions
The production sector is identical to that in the OSG model for the closed
economy. Let K (t ) denote the capital stocks employed by the economy at
time t and N ( = 1 ) the flow of labor services used at time t for produc-
tion.25 The production function F (t ) defines the flow of production at time
t. We assume that F (K (t ), N ) is neoclassical.26 Let w(t ) stand for the real
wage rate and r * (t ) the real interest rate for borrowing or lending in the
world capital market at time t. For illustration, we fix the interest rate dur-
ing the study period. The marginal conditions are
r * + δ k = f ' (k (t )), w(t ) = f (k (t )) − k (t ) f ' (t ), (6.3.1)
in which k (t ) ≡ K (t ) / N . We assume that factor markets work quickly

enough so that our system always displays competitive equilibrium in fac-
tor markets. As r * is fixed, we see that both k (t ) and w(t ) are functions of
r * as
k = g (r * ), w = h(r * ).
24 We may also classify output into traded and non-traded goods. There are
some models of small open economies with non-traded goods (for instance, Engel
and Kletzer, 1989). We will deal with an open monetary economy with traded and
non-traded goods in Chap. 11. By the way, it is important to introduce tariffs (see,
Sen and Turnovsky, 1989), fixed specific factors (see, for instance, Guilló and
Perez-Sebastian, 2007), and renewable resources (see, for instance, Harris, 1981;
Tawada, 1982; Brander and Taylor, 1997; Hannesson, 2000) into small-open eco-
nomic models. It can be seen that the main ideas of these approaches can be inte-
grated into the framework proposed in this section.
25 We assume a homogeneous population of size 1 and full employment of the
production factors at any point of time.

26 We may introduce changeable returns to scale to open economies either via
knowledge accumulation as or Marshallian externalities. Recent publications on

the issues are numerous.
It is straightforward to show dg / dr * = 1 / f " < 0 and dh / dr * = − g < 0.

That is, as the rate of interest rises, both the capital density and the wage
rate fall.
We now describe behavior of consumers. Denote a(t ) the wealth per
capita at time t. A typical consumer obtains current income
y (t ) = r * a (t ) + w(t ), from the interest payment ra and the wage payment
w. We call Y (t ) ( ≡ y (t )N ) the current income. Introduce B(t ) as the
value of the economy’s net foreign assets at t and define b(t ) ≡ B (t ) / N .
According to the definitions, we have a(t )N = K (t ) + B(t ). That is
a(t ) = k (t ) + b(t ). As
Y (t ) = r * a(t )N + wN = F (t ) − δ k K (t ) + E (t ), (6.3.2)
where E (t ) ≡ r * B(t ) and we use
r * K (t ) + w(t )N = F (t ) − δ k K (t ).
The current income of the households is equal to the sum of the econ-
omy’s net output, F − δ k K , and the country’s interest earned on foreign
assets, r * B . The gross national product (GNP) is measured as the sum of
the value of the net output produced within its borders and net interna-
tional factor payments. The GNP is given by F + E. The output produced
within the country’s geographical borders is called gross domestic product
(GDP), The GDP is given by F . A country’s current balance at time t is
the change in the value of its net claims over the rest of the world – the
change in its net foreign assets. If B& (t ) > 0, the economy as a whole is
lending (in this case we say that the current account balance is in surplus);
if B& (t ) < 0, the economy as a whole is borrowing (the current account bal-
ance is in deficit); and if B& (t ) = 0, the economy as a whole is neither bor-
rowing nor lending (the current account balance is in balance).
The disposable income is given by Y (t ) = Y (t ) + a (t )N . From (6.3.2),
we have
Y (t ) = (1 + r * )a(t )N + wN = w0 N + (1 + r * )B(t ), (6.3.3)
where
w0 ≡ (1 + r * )g + h .
Here, we require Y (t ) > 0, i.e.

a(t )N + F (t ) > − r * B(t ) + δ k K (t ).

The requirement means that the sum of the economy’s total assets and
output is more than the sum of the interest payment to rest world and the
capital depreciation. Otherwise, the economy has nothing to consume after
paying the two parts.
The disposable income is used for saving and consumption. At each
point of time, consumers would distribute the disposable income between
saving, S (t ), and consumption of goods, C (t ). The budget constraint is
given by
C (t ) + S (t ) = Y (t ). (6.3.4)
A typical consumer is to choose his most preferred bundle (c(t ), s (t )) of

consumption and saving under his budget constraint. Here, c ≡ C / N and
s ≡ S / N . The utility maximizing problem at any time is defined by
Max U (c, s )
c,s
s.t.: c(t ) + s (t ) = y (t ). (6.3.5)

We denote an optimal solution as function of the disposable income
(c(t ), s(t ) ) = (c( y (t )), s( y (t ) ).
The change in the households’ wealth is equal to the net savings minus
the wealth sold at time t , i.e.
a& (t ) = s ( y ( k ) ) − a(t ) . (6.3.6)
From y (t ) = w0 (r * ) + (1 + r * )b(t ) and a(t ) = g (r * ) + b(t ), for any fixed

level of r * , we see that the national economic dynamics is determined by
the following motion of per-capital foreign assets
b&(t ) = s(w0 (r * ) + (1 + r * )b(t )) − g (r * ) − b(t ). (6.3.7)
In a stationary state
s (w0 (r * ) + (1 + r * )b ) = g (r * ) + b .
We now show that this equation has a unique solution. Define
Φ(b) ≡ s (w0 (r * ) + (1 + r * )b ) − g (r * ) − b . (6.3.8)
The condition Y ≥ 0 is guaranteed if
~ w (r * )
b≥b ≡− 0 *.
1+ r
We have Φ (b ) = h / (1 + r * ) > 0. As 0 < s ' ( y ) < 1, for sufficiently large b,

~
we may have Φ(b ) < 0. As Φ ' (b) = (1 + r * )s' − 1, we may have Φ ' (b ) < 0
if r * is small and s ' is properly smaller than unity.27 If Φ (b ) < 0 and
~
Φ' (b ) < 0 for b ≥ b , then the dynamic system has a unique equilibrium. Let
b* stand for an equilibrium point. As
∂b&
= Φ' (b* ) = (1 + r * )s ' − 1.
∂b b = b*
Hence, if Φ ' (b ) < 0, the dynamic system has a unique stable equilibrium
point.
In the case of Φ' (b ) < 0 at the equilibrium we can examine effects of
change in the rate of interest. Take derivatives of Φ' (b ) < 0 with respect to
r*
dg 1 − (1 + r * )s' (6.3.9)
Φ'
db
= − (w0 (r *
) + b )s ' + = − bs ' ,
dr * dr * f"
where we use
dh 1 + r *
dw0
= g + (
1 + r )
* dg
+ = < 0.
dr * dr * dr * f"
Using Φ' < 0, 0 < s ' ≤ 0, f " < 0 and 1 > (1 + r * )s ' , we conclude
db / dr * > 0. As the rate of interest increases, the value of the country’s net
foreign assets rises. As
dy 1 + r *
+ b + (1 + r * ) * ,
db
*
=
dr f" dr
27 We will specify production and utility functions to further demonstrate these
properties.
we see that the sign of dy / dr * is ambiguous in general. As the impacts on

c and s are in the same direction as on y , we need additional conditions
in order to explicitly judge the impacts on c and s.
6.3.2 The Dynamics with the Cobb-Douglas Functions
To illustrate the analytical results and to simulate the model, we now ex-
amine the model with the following utility and production functions
F (t ) = AK α (t )N β , α , β > 0, α + β = 1,
U (t ) = c ξ (t )s λ (t ), ξ + λ = 1, ξ , λ > 0. (6.3.10)
From f = Ak α and Eqs. (6.3.1), we solve
k = r01/ β , w = βAr0α / β , (6.3.11)
where
αA
r0 = .
r + δk
*
We have y = w0 + (1 + r * )b, where w0 = (1 + r * )k + w. The optimal

choice of households is given by
c = ξy , s = λy.
From Eq. (6.3.7), s = λy and y = w0 + (1 + r * )b, we have
b& = λw0 − r01 / β − (ξ − λr * )b . (6.3.12)
Assume ξ − λr * > 0. From the definitions of ξ and λ and the rate of

interest should be small, it is reasonable to require ξ > λr * . The solution
of the linear differential equation is given by28
b(t ) = (b(0 ) − b* )e − (ξ − λr )t + b* , (6.3.13)

*
28 It should be noted that in an important paper, Fischer and Frenkel (1972)

demonstrates that an initially capital-scarce country will first acquire debt as it ac-
cumulates capital, then will begin to reduce its foreign liabilities, and may eventu-
ally become a net creditor to the rest of the world. Nevertheless, saving in their
work is not derived from optimizing behavior.
where
λw0 − r01 / β
b* ≡ .
ξ − λr *
As ξ − λr * > 0, we conclude that as t → + ∞, b(t ) → b* . In the long

term, the system approaches its equilibrium position. Figure 6.3.1 depicts
the foreign asset dynamics with different initial conditions. We see that if
the initial foreign asset is above the equilibrium level, it decreases over
time, and vice versa.
b
b(t )
equilibium
b(t )
Fig. 6.3.1. Convergence towards the equilibrium point
Lemma 6.3.1
Assume ξ − λr * > 0. The dynamic system has a unique stable equilibrium
point.
As ξ − λr * > 0 , the sign of b* is the same as that of λw0 − r01/ β . From

w0 = (1 + r * )r01/ β + βAr0α / β , we see that the equilibrium value of b* is posi-
tive if
αξ (6.3.14)
r* > − βδ k ,
λ
where we also use r0 = αA / (r * + δ k ). We conclude that if the propensity
to save is high, it needs higher level of the interest rate to satisfy (6.3.14).
Inequality (6.3.14) implies that in the long term, whether or not the econ-
omy owns wealth employed by other countries is independent of its pro-
ductivity parameter A.
We simulate the model to illustrate how the equilibrium values are af-
fected by parameters. First, we examine the impact of r * on the equilib-
rium values. We specify the other parameters as follows
α = 0.35, A = 1, λ = 0.7, δ k = 0.05. (6.3.15)
1.5 c
b λ
0.02 0.04 0.06 0.08 0.1 0.12 0.14 1.45
-2
1.4
-4
1.35
-6
0.02 0.04 0.060.08 0.1 0.12 0.14 λ
-8 1.25
a) the foreign assets b) the per-capita consumption level
Fig. 6.3.2. The impact of the international interest rate, 0 < r * < 0.15
We now allow λ to vary, with r * = 0.03 and the values of the other pa-
rameters specified as (6.3.15), we have the effects of change in the propen-
sity to save as in Fig. 6.3.3.
α = 0.35, A = 1, λ = 0.7, δ k = 0.05. (6.3.15)
As the propensity to save rises, both the foreign assets and the consump-
tion level increase. It should be noted that in an autarky economy, a rise in
the propensity to reduce per-capita consumption level when the propensity
to save is high. As the economy in trade is faced with a fixed rate of inter-
est, accumulated wealth will not reduce the return rate from wealth as in
the autarky case.29
8 1.7
6
1.6
4
2 1.5
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.4

-2
-4
-6 0.3 0.4 0.5 0.6 0.7 0.8 0.9
a) the foreign assets b) the per-capita consumption level

Fig. 6.3.3. The impact of the propensity to save, 0.1 ≤ λ ≤ 0.9
29 The costs due to depreciation is paid by producers.

6.3.3 Autarky Interest Rates and the Trade Pattern
We are concerned with an open economy where the rate of interest is fixed
in international market. In order to describe impact of trade on the national
economic growth, it is proper to compare two extreme types of economies
– a completely open economy and an isolated economy with the same
preference and technology.30 The key concept for the comparison is the au-
tarky real interest rate, that is, the rate of interest that prevails in an econ-
omy barred from international borrowing and lending.
As the two economies have similar variables, if x is a variable value in
the open economy, then x̂ stands for the corresponding variable value in
the autarky economy. We assume that the two economic systems have the
same preference and technology, which are specified as in (6.3.10). From
Sect. 6.3.3, we know that the dynamics of the open economy are described
by
f (t ) = Ak α (t ), k = r01 / β , w = βAr0α / β , U (t ) = c ξ (t )s λ (t ),
y (t ) = w0 + (1 + r * )b(t ), c(t ) = ξy (t ), s = λy (t ), a (t ) = k (t ) + b(t ),
b(t ) = (b(0 ) − b* )e − (ξ − λr )t + b* , (6.3.16)

*
where r0 = αA / (r * + δ k ) and w0 = (1 + r * )r01/ β + βAr0α / β .

From Sect. 6.3.2, we know that the dynamics of the autarky economy
are described by
fˆ (t ) = Akˆα (t ), rˆ(t ) = αAkˆ − β (t ) − δ k , wˆ (t ) = βAkˆα (t ),
yˆ (t ) = Akˆ α (t ) + δkˆ(t ), cˆ(t ) = ξyˆ (t ), sˆ(t ) = λyˆ (t ),
1/ β
 λA  − βξ k t λA 
kˆ(t ) =  kˆ β (0) − e +  ,
 ξ k  ξ k  (6.3.17)
where ξ k ≡ 1 − λδ > 0. The equilibrium value of kˆ(t ) is given by

kˆ = (λA / ξ ) .
1/ β
k
First, we are interested in comparing the equilibrium values of the two

systems. As
30 As trade will affect an economy’s preference and technology, validity of this
comparison is limited.
αξ k αξ − βλδ k
rˆ = − δk = ,
λ λ
we see that rˆ − r * > (< ) 0 if αξ / λ > (< ) r * + βδ k , that is
α
> (< ) λ.
r + α + βδ k
*
We conclude that for the autarky economy’s equilibrium rate of interest

to be higher than the internationally fixed interest rate (i.e., rˆ > r * ), we
should require that the economy’s propensity to save is “properly” small or
the international interest rate is small. In the remainder of this section, our
discussion is limited to the case of
α > λ (r * + α + βδ k ).
Under this condition, we have rˆ > r * . Under α / (r * + α + βδ k ) > λ , we
have
λw0 − r01 / β ((α + r * + βδ k )λ − α )r01 / β
b* = = < 0.
ξ − λr * α (ξ − λr * )
The equilibrium value of foreign assets is negative because of the low
propensity to save. From rˆ > r * , r = αAk − β − δ k and rˆ = αAkˆ − β − δ k , we
see that the capital intensity of the open economy is higher than the capital
intensity of the autarky economy, i.e., k > k̂ . We have f > fˆ and w > ŵ.
From
y = (1 + r * )r01 / β + βAr0α / β + (1 + r * )
((α + r *
+ βδ k )λ − α )r01 / β
,
α (ξ − λr * )
α /β 1/ β
 λA   λA 
yˆ = A  + δ   ,
 ξk   ξk 
we have
y − yˆ
= (1 + r *
) +
(r * + δ k )β +
r01 / β α
(1 + r ) ((α + r + βδ k )λ − α ) ξ k ~ β (r * + δ k ) ~
*
r ~ r
*
− − δr = − ,
α (ξ − λr )*
λ α (ξ − λr ) λ
* (6.3.18)
r ≡ (λ (r * + δ k )/ αξ k ) < 1 (under α / (r * + α + βδ k ) > λ ). We

in which ~
1/ β
see that in general, we cannot judge the sign of y − ŷ. As

( )
c − cˆ = ξ y − yˆ , the impact of trade on the long-term consumption is
ambiguous.
Lemma 6.3.2
If α / (r * + α + βδ k ) > (< ) λ , then the autarky economy’s equilibrium rate
of interest is higher (lower) than the internationally fixed interest rate. The
capital intensity, per-capita output and wage rate in the trade economy are
higher (lower) than the corresponding variables in the autarky economy.
We now simulate the model to illustrate behavior of the economy. First,

we examine effects of the international interest rate. We specify the other
parameters as in (6.3.15). As r * rises, the difference between the interna-
tional interest rate and the interest rate in the autarky system rises. When
r * reaches near 0.094, the difference is changing its sign from negative to
positive. We see that free trade will always benefit the economy in terms
of (long-term) per-capita consumption.
0.06 0.25
0.04 r * − rˆ 0.2
0.02
0.15
0.02 0.04 0.06 0.08 0.1 0.12 0.14
-0.02
0.1
c − cˆ
-0.04
-0.06 0.05
-0.08
0.02 0.04 0.06 0.08 0.1 0.12 0.14
a) the interest rates b) per-capita consumption levels
Fig. 6.3.4. The differences in the two economies as r * changes, 0 < r * < 0.15
We now allow λ to vary, with r * = 0.05 and the values of the other pa-
rameters specified as (6.3.15), we have the effects of change in the propen-
sity to save as in Fig. 6.3.5. As the propensity to save rises, the difference
6.4 Growth and Agglomeration of a Small-Open Multi-Regional Economy 203
between the international interest rate and the interest rate in the autarky
system rises. When r * reaches near 0.76, the difference is changing its
sign from negative to positive. We see that free trade benefits the economy
in terms of (long-term) per-capita consumption.
0.6
0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.5
-0.2
-0.4
0.4 c − cˆ
0.3
-0.6
r * − rˆ 0.2
-0.8
0.1
-1
0.3 0.4 0.5 0.6 0.7 0.8 0.9
a) the interest rates b) per-capita consumption levels
Fig.6.3.5 The differences in the two economies as τ changes, 0 < τ < 0.9
6.4 Growth and Agglomeration of a Small-Open Multi-

Regional Economy
Extending the model in Sect. 6.3 to a multi-regional economy, we examine

economic growth of a multi-regional small open economy in a perfectly
competitive economy.31 The national economy consists of multiple regions
and each region has one production sector and one housing sector. House-
holds move freely among regions, equalizing utility level among regions
by choosing housing, goods and saving. A region’s amenity is endogenous,
depending on the region’s output and population. We explicitly solve the
dynamics of the multi-regional economy. The system has a unique stable
equilibrium point. We simulate the motion of the model and examine ef-
fects of changes in the rate of interest, the preference, and amenity parame-
ters. We show, for instance, that a productivity improvement in the region
with lowest productivity reduces the GDP and GNP and a rise in the pref-
erence for large cities may accelerate agglomeration of the population and
economic activities into a region with high productivity.
In his Interregional and International Trade published in 1933, Bertil
Ohlin argues that regional economics and international economics should
be studied together as they share the same research objective. Neverthe-
less, regional aspects of economics had been largely neglected in (main-
31 This section is based on Zhang (2008a).

streams of) economics until new economic geography has recently ob-
tained much attention.32 A main purpose of economic geography is to ex-
plain the empirical fact that economic activity and population distribution
are not spatially random. It is important to give a micro-economic founda-
tion for studying clustering of people and firms using a general equilibrium
framework. Although many papers and books have been recently pub-
lished in the field of new economic geography, almost all these works have
been concerned with imperfect competition and have neglected capital ac-
cumulation.33 The purpose of this section is to develop a general equilib-
rium framework with multiple regions and capital accumulation under per-
fect competition. Different from the new economic geography which has
been mainly concerned with monopolistic competition, scale economies,
and transport costs in economic geography, this model studies perfection
competition, amenity, and technological differences (with constant returns
to scale).34 We show how different regions in an open economy interaction
with each other with capital accumulation and differences in amenity, fac-
tor endowments and productivity. Although some attempts have been
made to apply neoclassical growth theory to address spatial growth issues,
these models do not take account of land and regional differences in amen-
ity, which are generally considered as important factors of economic geog-
raphy.35 The objective of this section is to study growth of a small open
economy with economic geography. There are some economic models
which deal with growth and capital accumulation of small open econo-
32 We refer two comprehensive surveys on the literature to Henderson and Thisse
(2004) and Capello and Nijkamp (2004).

33 See Zhang (2003a) and Baldwin and Martin (2004) for the literature on the
topic. Obviously, capital accumulation and capital mobility are important vari-
ables for explaining spatial dynamics. Clustering of people into a single metropoli-
tan area like Tokyo or Shanghai can hardly be explained without taking capital as
endogenous variables. As argued by Zhang (2005a, 2006a), capital accumulation,
which is the key aspect of the neoclassical growth theory and is obviously a key
dimension of modern economic evolution, is largely neglected in new growth the-
ory as well as new economic theory.
34 A comprehensive model of economic geography should take account of not
only scale economies and transport costs (like in Krugman’s models), but also
amenity and factor endowment differences (like in Glaeser’s approach and this
model). In this stage, we are concentrated on perfect competition, amenity, factor
endowments and productivity differences among regions).
35 Extending neoclassical growth theory to spatial economics is made by, for in-
stance, Richardson (1977), Henderson (1985), and Henderson et al. (1995).

mies.36 As pointed out by Baldwin and Martin (2004: 2675-6), “Many of

the most popular economic geography models focus on labor. … These are
unsuited to the study of growth.” Capital accumulation is seldom modeled
with land use pattern and land markets in the literature of geographical
economics. Fujita and Thisse (2002: 389) state the current situations of
spatial economic growth as follows: “Clearly, space and time are intrinsi-
cally mixed in the process of economic development. However, the study
of their interaction is a formidable task. … Not surprisingly, therefore, the
field is still in its infancy, and relevant contributions have been few.” This
section attempts to make a contribution to economic growth with space by
developing an economic growth model with economic geography, basing
on the neoclassical growth theory within the context of growth theory of
small open economies. Moreover, as far as the interregional economic is-
sues are concerned, the model in this section is similar to the GTAP model
which is a multiregion, multisector, computable general equilibrium
model, with perfect competition and constant returns to scale (see, for in-
stance, McDougall, 2002). Attempts have been made to extend the stan-
dard static GTAP model to dynamic model.37 There are many other models
developed by regional scientists. For instance, some models based on the
input-output system or/and the gravity theory are proposed to examine in-
terregional trade patterns.38 Nevertheless, our model differs from these ap-
proaches in that we model behavior of households differently.
6.4.1 The Multi-Region Trade Model with Capital Accumulation
Most aspects of the model are the same as the model in Sect. 6.3, except
that we add housing sector and refine national economy into multiple re-
gions. The system consists of multiple regions, indexed by j = 1, ..., J .
Perfect competition is assumed to prevail in good markets both within each
region and among the regions, and commodities are traded without any
barriers such as transport costs or tariffs. The labor markets are perfectly
competitive within each region and among regions. Let prices be measured
in terms of the commodity and the price of the commodity be unity. We
denote wage and interest rates by w j (t ) and r j (t ) , respectively, in the
36 Refer to, for instance, Obstfeld and Rogoff (1999), Lane (2001), Koolmann
(2001, 2002), Benigno and Benigno (2003), and Galí and Monacelli (2005), for
the literature on economics of open economies.
37 This approach is represented by Ianchovichina and McDougall (2001).
38 See, for instance, Isard (1953), Hewings and Jensen (1986), Batten (1982),
Boomsma and Oosterhaven (1992), and Canning and Wang (2005).

j th region. The interest rate is identical throughout the national economy,

i.e.
rj (t ) = r * ,
where r * is the rate of interest fixed in the international market. We as-

sume a homogenous population. A person is free to choose his residential
location. We assume that any person chooses the same region where he
works and lives. Each region has fixed land. Land quality, climates, and
environment are homogenous within each region, but they vary among re-
gions. We neglect transportation cost of commodities between and within
regions. The assumption of zero transportation cost of commodities im-
plies price equality for the commodity among regions. As amenity and
land are immobile, wage rates and land rent may not be equal among re-
gions.
We introduce
N  the given population of the economy;

L j  the given (residential) area of region j;
~
K (t ) and A(t )  the capital stocks employed by and the total wealth of
the national economy at t ;
F j (t )  the output levels of region j ’s production sector at time t ;
K j (t ) and N j (t )  the levels of capital stocks and labor force employed
by region j ’s production sector;
c j (t ) and s j (t )  the per-capita consumption level of commodity and
saving made by per capita in region j;
l j (t ) and R j (t )  the lot size per household in region j and region j ’s
land rent.
Behavior of producers
We assume that there are only two productive factors, capital, K j (t ) ,
and labor, N j (t ), at each point of time t. The production functions are
given by
F j (K j (t ), N j (t )), j = 1, L, J ,
where F j is the output of region j. Assume F j to be neoclassical. We

have
F j (t ) K j (t ) (6.4.1)
f j (t ) = f j (k j (t )), f j (t ) ≡ , k j (t ) ≡ .
N j (t ) N j (t )
Markets are competitive; thus labor and capital earn their marginal prod-
ucts, and firms earn zero profits. The rate of interest, r * , and wage rates,
w j (t ), are determined by markets. Hence, for any individual firm r * and
w j (t ) are given at each point of time. The production sector chooses the
two variables, K j (t ) and N j (t ), to maximize its profit. The marginal con-
ditions are given by
r * + δ kj = f j' (k j ), w j (t ) = f j (k j ) − k j f j' (k j ), (6.4.2)
where δ kj are the depreciation rate of physical capital in region j . As r *

is fixed, from r * + δ kj = f j' and w j = f j − k j f j' we obtain that both k j (t )
and w j (t ) are functions of r * as
k j = φkj (r * ), w j = φwj (r * ). (6.4.3)
We see that k j and w j are invariant in time. It is straightforward to

show dφkj / dr * = 1 / f j" < 0 and dφwj / dr * = − φkj < 0 . As the rate of in-
terest rises, both the capital densities and the wage rates fall.
Behavior of consumers
Each worker may get income from land ownership, wealth ownership
and wages. In order to define incomes, it is necessary to determine land
ownership structure. It can be seen that land properties may be distributed
in multiple ways under various institutions. To simplify the model, we ac-
cept the assumption of “absentee landownership” which means that the in-
come of land rent is spent outside the economic system. A possible case is
that the land is owned by the government.39 Households rent the land in
39 Another two popular assumptions in the literature of spatial economics are

the equally shared landownership and the public ownership. In the former, land is
owned equally by all households in the system. Zhang (2007a) also studies the dy-
namics for the equally shared landownership. In the latter, for instance as accepted
in Kanemoto (1980), the government rents the land from the landowners at certain
rent and sublets it to households at the market rent, using the net revenue to subsi-
dize city residents equally. Zhang (1998b, 2008b) constructs multi-regional
growth models, which are similar to this model but for a closed economy. It
competitive market and the government uses the income for military or
other public purposes. Consumers make decisions on choice of lot size,
consumption levels of services and commodities as well as on how much
to save.
Let a~ j (t ) stand for the wealth owned by a household in region j. The
household in region j obtains income
y j (t ) = r * a~ j (t ) + w j , j = 1, L, J , (6.4.4)
from the interest payment, r *a~ j (t ), and the wage payment, w j . The dis-
posable income is equal to
yˆ j (t ) = y j (t ) + a~ j (t ). (6.4.5)
At each point of time, a consumer distributes the total available budget

among housing, l j (t ), saving, s j (t ) , consumption of goods, c j (t ) . The
budget constraint is given by
R j (t )l j (t ) + c j (t ) + s j (t ) = yˆ j (t ) = r * a~ j (t ) + w j + a~ j (t ). (6.4.6)
At any time, consumers have three variables to decide. A consumer de-

cides how much to consume housing, to consume goods and to save. Equa-
tion (6.4.6) means that consumption and savings exhaust the consumers’
disposable personal income.
We assume that utility level, U j (t ) , that the consumers obtain is de-
pendent on the consumption level of commodity, c j (t ) , and the saving,
s j (t ) . The utility level of the consumer in region j , U j (t ) , is specified as
follows
U j (t ) = θ j (t )l ηj 0 (t )c ξj 0 (t )s λj 0 (t ), η 0 , ξ 0 , λ0 > 0 , (6.4.7)
in which η 0 , ξ 0 , and λ0 are a typical person’s elasticity of utility with re-

gard to lot size, commodity and savings in region j . We call η 0 , ξ 0 , and
λ0 propensities to consume lot size, to consume goods, and to hold wealth
(save), respectively.
In (6.4.7), θ j (t ) is called region j ’s amenity level. The concept of
amenity measures a region’s attractiveness for households. Amenities are
should be noted that we can refine the model by introducing urban structures
(Zhang, 1996, 2007c).
affected by infrastructures, regional cultures and climates.40 People cluster

together for different reasons. For instance, people like to socially interact
with each other. This implies that a large population of the region is attrac-
tive. A region with large population offers more opportunities for em-
ployment. Good climates make a region attractive as residential location.
As argued by Glaeser et al. (2001), consumption amenities have increas-
ingly played important role in economic geography. Public services, acces-
sibilities, local transportation systems, pollution, and human relations such
as discrimination all involve externalities and affect amenities. We incor-
porate amenity into the consumer location decision by assuming that
amenity is an endogenous variable.41 The regional population dynamics is
influenced by many changing characteristics of environmental quality such
as air quality, levels of noise pollution, open space, and other physical and
social neighborhood qualities at each location. Environmental quality can
be reflected in part by its effect on the location choice of individuals.
Many kinds of externalities may actually exist at any location. Some may
be historically given, such as historical buildings and climate; others such
as noise and cleanness, may be endogenously determined by the location
of residents. Households may prefer a low-density residential area to a
high one, as there tend to have more green, less noise, more cleanness and
more safety in a low-density area. Nevertheless, there are other factors,
such as social interactions, which may make high-density area attractive.
We assume that amenity is affected by production and consumption activi-
ties. We specify θ j as follows
θ j (t ) = θ j F ja (t )N bj (t ), j = 1, L, J ,
where θ j (> 0) , a and b are parameters. We don’t specify signs of a

and b as economic activities and population may have either positive or
negative effects on regional attractiveness. As F j = f j N j , we can rewrite
the above equations as follows
θ j (t ) = θ j f ja (t )N aj + b (t ), j = 1, L, J . (6.4.8)
40 See, for instance, Kanemoto (1980), Diamond and Tolley (1981), Blomquist et
al. (1988), and Andersson et al. (2003).
41 This section does not take account of externalities for producers. Firms often
prefer to locate close to other firms. An explicit introduction of externalities will

make the spatial structure far more complicated. In the literature of spatial eco-
nomics, various externalities have been analyzed (see, Henderson, 1974; Upton,
1981; and Abdel-Rahman, 2004).
Maximizing U j (t ) subject to the budget constraints (6.4.6) yields
l j (t ) R j (t ) = ηyˆ j (t ) , c j (t ) = ξyˆ j (t ) , s j (t ) = λyˆ j (t ) , (6.4.9)
in which
1
η ≡ ρη 0 , ξ ≡ ρη 0 , λ ≡ ρη 0 , ρ ≡ .
η 0 + ξ 0 + λ0
According to the definitions of s j (t ), the wealth accumulation of the
representative person in region j is given by
a~& j (t ) = s j (t ) − a~ j (t ) . (6.4.10)
As households are assumed to be freely mobile among the regions, it is

reasonable to consider that households migrate where utility is higher. Un-
der the assumptions that households can move freely and rapidly, the util-
ity level of people should be equal, irrespective of in which region they
live, i.e.
U j (t ) = U m (t ) , j , m = 1, L, J . (6.4.11)
We neglect possible costs for migration. In reality, even to change a

house in a small town costs. Although it is not difficult to introduce migra-
tion costs into the model, it will become far more difficult to explicitly get
analytical results. Instead of wage equalization (which is often used as the
equilibrium mechanism of population distribution), we assume that con-
sumers obtain the same level of utility in different regions as the equilib-
rium mechanism of population distribution among the regions. Although
the condition of utility equalization is often used in the literature of urban
economics,42 the assumption of utility equalization is rarely used in the lit-
erature of economic dynamics as the temporary equilibrium condition of
population distribution. It is argued that this assumption is more reasonable
than the assumption of wage equalization in interregional analysis.
The total capital stocks employed by the economy is equal to the sum of
the capital stocks employed by all the regions. That is
J J (6.4.12)
K (t ) = ∑j =1
K j (t ) = ∑ k (t )N (t ).
j =1
j j
42 Fujita (1989) provides some models with the assumption of equalizing utility
levels among households in different locations.

The total wealth of the national economy is the sum of the wealth
owned by all the households
~ J
(6.4.13)
A(t ) = ∑ a~ (t )N (t ).
j j
j =1
We introduce B(t ) as the value of the economy’s net foreign assets at t .

The income from the net foreign assets, E (t ), which may be either posi-
tive, zero, or negative, is equal to r * B(t ). The national industrial output is
equal to the national net saving. That is
~ J
(6.4.14)
S (t ) + C (t ) − A(t ) − r * B(t ) + ∑δ kj K j (t ) = F (t ),
j =1
where
J J J
C (t ) ≡ ∑ c j (t )N j (t ), S (t ) ≡ ∑ s j (t )N j (t ), F (t ) ≡ ∑ F j (t ).
j =1 j =1 j =1
The assumption that labor force and land are fully employed is repre-
sented by
J
(6.4.15)
∑ N (t ) = N , l (t )N (t ) = L ,
j =1
j j j j j = 1, L, J .
We have thus built the multi-regional model of a small open economy

with capital accumulation.
We now explain trade balance in the model. According to the definitions
of the national wealth, the capital stocks employed by the economy and the
net foreign assets, we have
~
A(t ) = K (t ) + B(t ).
Similar to the variable, B(t ), we introduce B j (t ) as the value of region

j ’s net external assets at t . We have
B j (t ) = (a~ j (t ) − k j )N j (t ).
We have
J
B (t ) = ∑ B j (t ).
j =1
From Eqs. (6.4.4), we obtain the national current income, Y (t ), as fol-

lows
Y (t ) ≡ ∑ y j (t )N j (t ) = ∑ (r * a~ j (t ) + w j )N j (t ).
J J
j =1 j =1
From Eqs. (6.4.2) and K (t ) = K (t ) + B(t ), we have
Y (t ) ≡ ∑ y j (t )N j (t ) = ∑ (r * a~ j (t ) + w j )N j (t ).
J J
j =1 j =1
where we also use Eqs. (6.4.12), (6.4.13) and (6.4.15). From this equation
and Eq. (6.4.16), we have
J
Y (t ) = F (t ) − ∑δ kj K j (t ) + r * B (t ).
j =1
The national current income is equal to the sum of the economy’s net
output
J
F (t ) − ∑δ kj K j (t ),
j =1
and the country’s interest earned on foreign assets, r * B(t ). The gross na-
tional product (GNP) is measured as the sum of the value of the net output
produced within its borders and net international factor payments. The
GNP is given by F (t ) + E (t ). We introduce
G (t ) ≡ F (t ) + E (t ).
Similarly, region j ’s gross regional product (GRP), G j (t ), is given by
G j (t ) ≡ Fj (t ) + r * B j (t ).
The output produced within the country’s geographical borders is called

gross domestic product (GDP), The GDP is given by F (t ). A country’s
current balance at time t is the change in the value of its net claims over
the rest of the world – the change in its net foreign assets. If B& (t ) > 0 , the
economy as a whole is lending (in this case we say that the current account
balance is in surplus); if B& (t ) < 0 , the economy as a whole is borrowing
(the current account balance is in deficit); and if B& (t ) = 0 , the economy as
a whole is neither borrowing nor lending (the current account balance is in

balance).
Finally, we describe trade among regions. In the rest of this section, we
omit t in expressions, wherever without causing confusion. First, we cal-
culate
F − (R l + c + s − a~ )N − δ K = r * B ,
j j j j j j j kj
(6.4.17)
j j
where we use
R j l j + c j + s j = ra~ j + w j + a~ j , Fj = (r * + δ kj )K j + w j N j .
That is
F j = (R j l j + c j + s j − a~ j )N j + δ kj K j + r * B j .
The regional output, F j , is used for housing, R j l j N j , consumption,

c j N j , net saving, (s j − a~ j )N j , the depreciation of capital, δ kj K j , and the
region’s interest payment for the capital owned by the other regions or for-
eign countries, r * B j . We express Eq. (6.4.17) as follows
(F j − R j l j N j − c j N j − δ kj K j ) + (s j N j − a~ j N j − r * B j ) = 0 ,
where trade balance, F j − R j l j N j − c j N j − δ kj K j , is expressed as net pro-

duction minus consumption. If this term is positive, then the region is an
exporter, which corresponds on the capital account side to a positive,
s j N j − a~ j N j − r * B j , the wealth accumulation minus interest payment re-
ceived.
6.4.2 The Interregional Dynamic Behavior
Multi-regional dynamic system is seemingly complicated. Nevertheless, its

motion is given by a set of (unconnected) linear differential equations. The
following lemma, which is proved in Appendix A.6.3, shows how we can
determine the motion of all the variables in the dynamic system.
Lemma 6.4.1
The variables, k j , w j and f j , are uniquely determined as functions of
r * . The motion of the levels of the per-capita wealth is given by
λw j (6,4,18)
a~ j (t ) = a~ j (0)e −λ t + * , j = 1, L, J ,
*
λ
in which λ* ≡ 1 − λ − λr * . The other variables are uniquely determined by
the following procedure: yˆ j (t ) = (1 + r * )a~ j (t ) + w j , j = 1, ..., J →
g j yˆ ηj (t )N
N j (t ) = , j = 1, ... , J ,
∑ g yˆ η (t )
J
j =1 j j
where
a / (η − a − b ) 1 / (η − a − b )
 fj   θ j Lηj  ξ +λ
g j ≡     , η ≡ ,
 θ Lη  η −a−b
 f1   1 1 
→ F j (t ) = f j N j (t ) → F (t ) = ∑ j F j (t ) → K (t ) = ∑ j k j N j (t ) →
~ ~
A(t ) = ∑ j a~ j (t )N j (t ) → B(t ) = A(t ) − K (t ) → G (t ) = F (t ) + r * B(t ) →
B j (t ) = (a~ j (t ) − k j )N j (t ) → l j (t ) = L j / N j (t ) → R j (t ) = ηyˆ j (t ) / l j (t ) →
c j (t ) and s j (t ) by Eqs. (6.4.9).
As η + ξ + λ = 1, we see that if η + ξ > r *λ , then λ* > 0 . As it is rea-

sonable to assume the rate of interest, r * , to be small, the requirement of
η + ξ > r *λ is generally acceptable. As λ* > 0 , from Eqs. (6.4.18) we
see that
a / (η − a − b ) 1 / (η − a − b ) (6.4.19)
 fj   θ j Lηj  ξ +λ
g j ≡     , η ≡ ,
 θ Lη  η −a−b
 f1   1 1 
as t → ∞ . The dynamic system has a unique equilibrium point and the
equilibrium point is stable. In summary, we have the following theorem.
Theorem 6.4.1
The dynamic system has a unique stable equilibrium point.
From Eqs. (6.4.19), the wealth of a household in region j in equilib-

rium is positively related to the wage rate, w j , the rate of interest, r * , and
propensity to save, λ . The disposable income is also positively related to
w j , r * , and λ . We note that the sign of the parameter, η , is important
for determination of the population distribution. The sign of η is the same

as that of η − a − b . If a ≤ 0 and b ≤ 0 , which implies that both the out-
put level and the region’s population have negative effects on the region’s
amenity level, then we definitely have η > 0 . If η > 0 , then an increase
in the region’s residential area, or/and per-capita output level, or/and the
per-capita wealth attracts more people to the region. In the case of a = 0 ,
the condition becomes η < b , which implies that the amenity level is posi-
tively related to the region’s population and the impact of the population
on the amenity is stronger than the housing condition on the utility level.
We now explain behavior of the model. In the remainder of this section,
we are concerned with a three-region economy, i.e., J = 3 . For illustra-
tion, we specify the production functions as follows
F j (t ) = A j K αj (t )N βj (t ), α + β = 1, α , β > 0 ,
where A j is region j ’s productivity. We have f j = A j k αj . By Eqs.

(6.4.2), we solve
1/ β
 αA 
kj =  * j  , f j = A j k αj , w j = βf j .
r + δ 
 kj 
Region j ’s capital intensity, k j , per-capita output level, f j , and wage
rate, w j , are invariant in time and are dependent on the rate of interest in
the international market and the region’s productivity. As the rate of inter-
est rises, k j , f j and w j fall. According to Lemma 6.4.1, we can deter-
mine k j (t ) and yˆ j (t ). We examine how the population distribution is af-
fected by different factors. To simulate the model, we specify the
parameter values as follows
r * = 0.04 , N = 10 , α = 0.3 , a = 0 , b = − 0.1,
 A1  1.5   θ1   4  η 0   0.07   L1   3 
               
 A2  = 1.3  ,  θ 2  =  3 .8  ,  ξ 0  =  0.10  ,  L2  =  4  ,.
A   1       λ   0.83   L   6
 3    θ 3   4.3   0    3  
 δ k1   0.05  (6.4.20)
   
 δ k 2  =  0.06 
 δ   0.05 
 k3   
The rate of interest is fixed at 4 per cent and the population is 10 . Re-
gion 1 has the highest level of productivity. Region 2 ’s level of productiv-
ity is the second, next to region 1’s. We term region 1 as the coastal re-
gion (CR), region 2 the inner region (IR), and region 3 the hinterland
region (HR). It should be remarked that although the specified values are
not based on empirical observations, the choice does not seem to be unre-
alistic. For instance, some empirical studies on the US economy demon-
strate that the value of the parameter, α , in the Cobb-Douglas production
is approximately equal to 0.3. With regard to the technological parame-
ters, what are important in our interregional study are their relative values.
This is similarly true for the specified differences in land and amenity pa-
rameters among regions.
First, we calculate the time-independent variables
k1 = 9.966 , k 2 = 6.988 , k3 = 5.584 , f1 = 2.990 , f 2 = 2.330 ,
f 3 = 1.625 , w1 = 2.093 , w2 = 1.631, w3 = 1.173.

Region 1’s per-capita capital level, per-capita output level, and wage
rate are higher than the other two regions’ corresponding variables. Figure
6.4.1 plots the motion of the time-dependent variables. The initial condi-
tions are specified as
a~ (0 ) = a~ (0) = a~ (0) = 6 .
1 2 3
The initial per-capita wealth are equal in the three regions. As shown in
Fig. 6.4.1, the variables approach their equilibrium values.
The initial values of the per-capita wealth, k j (0), are smaller than their
equilibrium values. Figure 6.4.1a shows that the wealth levels of the
households in all the regions increase over time. From Fig. 6.4.1b and 1c,
the CR’s population and output level rise and the other two regions’ popu-
lation and output levels fall. The GDP, F , rises over time. The GNP and
the CR’s GRP rise and the other two regions’ GRPs fall. The net foreign
assets, B(t ), rise. The current account balance, B& (t ), is in surplus over
time. The per-capita consumption levels rise. The lot size in region 1 falls
and the land rent rises. The lot sizes in the other two regions rise and the
land rents fall.
12 25
11
k2
6
N1 20 F F1
10 k1 15
4
9
8
N2 10
7
k3 2
5 F2
N3 t
5 10 15 20 25 30t 5 10 15 20 25 30 5F310 15 20 25 30 t
~ (t )
(a) a (b) N j (t ) (c) F (t ) and F j (t )
j
25
G B'
3
20 G1 20 B
'
15 10 B1 2 B 1
10
B2
G2 25B330 t
1
5 5 10 15 20 B2'
-10 t
5G310 15 20 25 30 t B3' 5 10 15 20 25 30
(d) G (t ) and G j (t ) (e) B(t ) and B j (t ) (f) B' (t ) and B j (t )
8 2.5
1.4 c1 2 R1
1.2
c2 6
l3 1.5
4
t 2
l2 1 R2
5 10 15 20 25 30 l1 0.5
c3 20 25 30 t
0.8
5 10 15 20 25 30 t 5 10 15
R3
(g) c j (t ) (h) l j (t ) (l) R j (t )
Fig. 6.4.1. The motion of the national economy
6.4.3 The Rate of Interest and Interregional Dynamics
The previous section plots the motion of the variables. This section exam-
ines how the rate of interest affects the national economy. As we have ex-
plicitly solved the model, it is straightforward to make comparative dy-
namic analysis. First, we examine the case that all the parameters, except
the rate of interest, r * , are the same as in (6.4.20). We study what will
happen to the dynamics of the economic system if the rate of interest is
changed as follows43
43 As we have explicitly solved the dynamics, we can also carry out compara-
tive dynamic analysis by assuming that the rate of interest varies in time, r * (t ).
r * = 0.04 ⇒ 0.03.
The rate falls from 4 % to 3 % in the international market. We intro-
duce a variable, ∆x(t ), to stand for the change rate of the variable, x(t ), in
percentage due to changes in the parameter value. That is
x(t ; a ) − x(t ; a0 )
∆x(t ) ≡ ×100 ,
x(t ; a0 )
where x(t ; a ) stands for the value of the variable, x(t ), with the parameter
value, a , at time t and x(t ; a0 ) stands for the value of the variable, x(t ),
with the parameter value, a0 , at time t .
The effects on k j , f j and w j are given as follows
∆k1 = 18.325 , ∆k 2 = 16.243 , ∆k3 = 18.325 , ∆f1 = 5.177 ,

∆f 2 = 4.619 , ∆f 3 = 5.177 , ∆w1 = 5.177 , ∆w2 = 4.619 ,
∆w3 = 5.177 .
As the rate of interest is reduced in the international market, the capital
intensities, per-capita output levels and wage rates are increased. The capi-
tal cost reduction increases the capital intensities.
Figure 6.4.2 plots the effects of the change in the rate of interest. Figure
6.4.2a shows that the wealth of the households in all the regions rise ini-
tially and then fall. In the long term, the wealth of a typical household in
any region will be reduced. From Eqs. (6.4.18), we see that a reduction in
r * increases the wage rate and increases the parameter, λ* . Initially, the
increases in the wealth due to the wage income increases are larger than
the losses in the wealth due to the reduction in λ* , the net results increase
the wealth. Nevertheless, as time passes, the losses due to the reduction in
λ* become larger. From Fig. 6.4.2b and 2c, both the CR’s population and
output level rise. The other two regions’ population fall but their output
levels are increased. The GDP rises over time. The GNP and the three re-
gions’ GRPs rise. The net foreign assets, B(t ), rise in the long term. The
current account balance, B& (t ), is in surplus over time. The lot size in re-
gion 1 falls and the land rent rises. The lot sizes in the other two regions
rise and their land rents rise.
2 7
∆k1 6 ∆F1
0.5
∆k2 1 ∆ N1 5
t 5 10 15 20 25 30
t 4 ∆F
5 10 15 20 25 30 -1 3
-0.5 -2 ∆N 3 2 ∆F2
-1 -3
1
∆F3
∆k3 ∆N 2 5 10 15 20 25 30 t
(a) ∆k j (t ) (b) ∆N j (t ) (c) ∆F (t ) and ∆F j (t )
6 t 1
0.5 ∆B
5 10 15 20 25 30 '
-2.5
4 ∆G1 -5 ∆B3 ∆B '
2 ∆B2 3
∆G
-7.5 5 10 15 20 25 30 t
-10 -0.5
2 ∆G2 -12.5 ∆B1 ∆B '
∆G3 -15
∆B
-1
5 10 15 20 25 30 t -17.5 -1.5
∆B1'
(d) ∆G (t ) and ∆G j (t ) (e) ∆B (t ) and ∆B j (t ) (f) ∆B' (t ) and ∆B j (t )
4 3
∆c1 3 ∆l2 2 ∆R1
0.5 1
2
5 15 20 25 30 t
10∆c 1 ∆l3 5 10 15 20 25 30
t
2 -1
-0.5
5 10 15 20 25 30t -2 ∆R3
-1
∆c3 -1
-2
∆ l1 -3
-4
∆R2
(g) ∆c j (t ) (h) ∆l j (t ) (l) ∆R j (t )
Fig. 6.4.2. A fall in the rate of interest
6.4.4 Preference Change and Interregional Dynamics
This section examines how changes in the preference affect the national
economy. First, we examine effects of a fall in the propensity to save,
specified as follows: λ0 = 0.83 ⇒ 0.80 . We note that the preference will
not affect the capital intensities, k j , per-capita output levels, f j , and
wage rates, w j . Hence, we will not mention these variables in this section.
The simulation results are plotted in Fig. 6.4.3. Figure 6.4.3a shows that
the wealth of the households in all the regions fall. From Fig. 6.4.3b and
3c, both the CR’s population and output level fall. Some people move
away from the CR and they migrate either to the IR or to the HR. As the
propensity to save falls, the propensities to consume goods and housing are
relatively increased. This results in that the CR becomes less attractive be-
cause of its high land rent. As the two other regions’ populations are in-
creased and the per-capita output levels are not affected by the preference
change, the CR’s output falls and the other two regions’ output levels rise.
The national output level falls as the propensity to save falls. The GNP and
the CR’s GRP fall and the other two regions’ GRPs rise. The net foreign
assets, B(t ), fall. The current account balance, B& (t ), is in deficit. The lot
size in region 1 rises and the land rent rises initially and then falls. The lot
sizes in the other two regions fall and their land rents rise. The per-capita
consumption levels rise initially mainly because the propensity to consume
goods rises and the disposable income does not fall much; but the per-
capita consumption levels falls because the per-capita wealth is reduced as
a consequence in the fall in the propensity to save.
30 t 3
5 10 15 20 25 3 ∆F3
-1 ∆N 3
-2 ∆k1 ≥ ∆k2 ≥ ∆k3
2
∆N 2 2
∆F2
1 1
-3
∆F
-4 5 10 15 20 25 30t 5 10 15 20 25 30t
∆N1 ∆F1
~ (t )
(a) ∆a j (b) ∆N j (t ) (c) ∆F (t ) and ∆F j (t )
∆B3 ∆B53' 10
∆G3 30t
3
5 10 15 20 25 15 20 25 30t
2 -1 ∆B2 -0.1 ∆B2'
∆G2 -2
1
-3 ∆B1 -0.2 ∆B1'
t -0.3
5 10 15 20 25 30 -4
∆B ∆B '
-1
∆G1 ∆G -5
(d) ∆G (t ) and ∆G j (t ) (e) ∆B(t ) and ∆B j (t ) (f) ∆B ' (t ) and ∆B j (t )
3
∆c1 ≥ ∆c2 ≥ ∆c3 ∆ l1 t 3
∆R3
2 5 10 15 20 25 30 2
1
-1
1 ∆R2
-2
∆ l2 t
t 5 10 15 20 25 30
5 10 15 20 25 30 -3
∆l 3 -1 ∆R1
(g) ∆c j (t ) (h) ∆l j (t ) (l) ∆R j (t )
Fig. 6.4.3. A fall in the propensity to save
We now examine effects of a rise in the propensity to consume housing,

specified as follows: η 0 = 0.07 ⇒ 0.1. The variables, k j , f j and w j are
noted affected by the preference change. Hence, we will not mention these
variables in this section. The simulation results are plotted in Fig. 6.4.4.
Figure 6.4.4a shows that the wealth of the households in all the regions
fall. An increase in the propensity to consume housing implies a relative
decline in the propensity to save. The relative decline in the propensity to
save explains the decreases in the per-capita wealth. As housing becomes
relatively more important for residents, people move away from the CR
and migrate to the IR and the HR. The consumption levels in the three re-
gions are reduced. From Fig. 6.4.4b and 4c, both the CR’s population and
output level rise. Some people move away from the CR and they migrate
either to the IR or to the HR. The other variables change as in the case of
the decrease in the propensity to save.
t 50 50
-2.5
-5
5 10 15 20 25 30 40 ∆N 3 40
-7.5 ∆k1 ≥ ∆k2 ≥ ∆k3 30
∆N 2 30 ∆F3
-10
-12.5
20
10
20
10
∆F2
∆F
10 15 20 25 30 t
-15
-17.5 5 10 15 20 25 30t 5
-10
∆N1 -10 ∆F1
~ (t )
∆B3 t ∆5B3' 10' t

40
30
∆G3 -5
5 10 ∆B25
15 20
2
30 -0.25 ∆B2 15 20 25 30
-0.5
20 ∆G2 -10 -0.75
-1 ∆B1'
10
-15 ∆B1 -1.25
5 10 15 20 25 30
t ∆B -1.5
-10 ∆G1 ∆G -20 ∆B '
10 70
-2.5 5 10 15 20 25 30 t ∆ l1 60 ∆R2
-5 5 10 15 20 25 30
t 50 ∆R3
-7.5 ∆c1 ≥ ∆c2 ≥ ∆c3 -10 40
-10
-12.5 ∆l2 30
20
-15
-20
10 ∆R1
-17.5 -30 ∆l3 t
5 10 15 20 25 30
(g) ∆c j (t ) (h) ∆l j (t ) (l) ∆R j (t )
Fig. 6.4.4. A rise in the propensity to consume housing
6.4.5 Amenity and Interregional Dynamics
This section examines how changes in the parameters in the amenity func-
tions, θ j (t ) = θ j N bj . First, we examine effects of a fall in the parameter,
b , as follows
b = − 0.1 ⇒ − 0.05 .
A decrease in the parameter value may result from that, for instance,
people like more to live in large cities. Although cities have disamenities,
cities also offer a great variety of life styles and many work opportunities.
It is significant to examine how this parameter affects regional agglomera-
tion. The variables, k j , f j and w j are not affected by the preference
change.
The simulation results are plotted in Fig. 6.4.5. The per-capita wealth
and consumption levels are not affected by the parameter change. As peo-
ple like more to live with other people in the same region, people immi-
grate to the CR from the IR and HR. As the parameter is increased, people
tend to agglomerate into regions with high productivity. This parameter
may help us understand why people tend to be concentrated in a few met-
ropolitan areas, for instance, in Japan. As b rises, the CR’s output level is
increased mainly due to the increase in the labor force and the output lev-
els in the other two regions fall. As more people move to the region with
the highest productivity, the national output level rises.
1 ∆F1
10 ∆ N1 10
t
0.5 5 10 15 20 25 30t 5 10 15 ∆20
F 25 30
∆k j = 0 t
-10
-20
-10
-20
5 10 15 20 25 30 -30 ∆N 2 -30 ∆F2
-0.5 -40
-50 ∆N 3 -40
-50
-1 -60 -60 ∆F3
~ (t )
∆G1 0.5
10 2 ∆B1' ≥ ∆B '
-10 5 10 15 ∆G25 30t
20 1 ∆B1 5 10 15 20 25 30 t
-20 ∆B -0.5
-30 ∆G2 5 10 15 20 25 30t -1
-40
-50
-1 ∆B3 -1.5 ∆B3' ≥ ∆B2'
-60 ∆G3 -2 ∆B2 -2
(d) ∆G (t ) and ∆G j (t ) (e) ∆B (t ) and ∆B j (t ) (f) ∆B ' (t ) and ∆B j (t )
1
125 10 ∆R1
0.5 100 ∆ l2 5 10 15 20 25 30 t
∆c j = 0 75 ∆l3 -10
-20
5 10 15 20 25 30 t 50 -30 ∆R2
-0.5 25
t
-40
∆R3
10 15 ∆
20l1 25 30
-50
-1 5 -60
(g) ∆c j (t ) (h) ∆l j (t ) (l) ∆R j (t )
Fig. 6.4.5. The preference for living in the large region is increased
We now examine effects of changes in the CR’s amenity parameter, θ1 .

We specify the change as follows: θ1 = 4 ⇒ 4.2 . A change in the pa-
rameter does not affect k j , f j , w j , a~ j and c j . As shown in Fig. 6.4.6,

the effects are similar to those caused by the rise in b as shown in Fig.
6.4.5. Although the effects are similar in the two cases, the changes in the
two parameters may be caused by different mechanisms. For instance,
changes in b may be caused by the population’s attitude towards urban
life; changes in θ1 may be caused by an improvement in the CR’s infra-
structures.
1
10 ∆ N1 10 ∆F1
0.5
∆k j = 0 ∆F
t t
5 10 15 20 25 30 t 5 10 15 20 25 30 5 10 15 20 25 30
-0.5 -10 -10
∆ N 2 = ∆N 3 ∆F2 = ∆F3
-1 -20 -20
(a) ∆a~ j (t ) (b) ∆N j (t ) (c) ∆F (t ) and ∆F j (t )
1
10 ∆G1 ∆B1 ∆B t 0.5
0.4
∆B1' ≥ ∆B '
5 ∆
10G 15 20 25 30
t -1
5 10 15 20
∆B3
25 30 0.3
0.2
-10 ∆B2 0.1
∆G2 = ∆G3 -2 ∆B53' ≥
10 ∆15
B2' 20 25 30
t
-20 -0.1
1
0.5
20 ∆l2 = ∆l3 10 ∆R1
10
t t
5 10∆c
15j =
20 025 30 t 5 10 15 20 25 30
-0.5 5 10 15 20 25 30 -10
-10 ∆R2 = ∆R3
-1 ∆ l1 -20
(g) ∆c j (t ) (h) ∆l j (t ) (l) ∆R j (t )
Fig. 6.4.6. A rise in the CR’s amenity parameter, θ1
6.4.6 Productivity and Interregional Dynamics
This section examines how changes in the regions’ productivities affect the
national economy. First, we examine effects of a rise in the CR’s produc-
tivity, specified as follows: A1 = 1.5 ⇒ 1.7 . The effects on the capital in-
tensities, k j , per-capita output levels, f j , and wage rates, w j , are given
as follows
∆k1 = ∆f1 = ∆w1 = 19.579 , ∆k j = ∆f j = ∆w j = 0 , j = 2 , 3 .
The capital intensity, per-capita output level, and wage rate of the CR
are increased and the corresponding variables in the IR and HR are not af-
fected. The simulation results for the time-dependent variables are plotted
in Fig. 6.4.8. Figure 6.4.8a shows that the wealth of the households in the
CR is increased and the variables in the IR and HR are not affected. From
Fig. 6.4.8b and 8c, both the CR’s population and output level rise. The
populations and output levels in the other two regions are reduced as some
people from these regions immigrate to the CR. The net effect on the na-
tional output is positive. The trade balance is worsened initially but soon
improved. The lot size in region 1 falls and the land rent rises. The lot
sizes in the other two regions rise and their land rents fall.
20 40 ∆F1
15
∆k1 ∆ N1
30t ∆F
20
10 5 10 15 20 25
-20 5 10 15 20 25 30 t
5 -20
∆ k 2 = ∆k 3 = 0 -40
∆N 2 = ∆ N 3 -40 ∆F2 = ∆F3
5 10 15 20 25 30t
~ (t )
∆G1 2
40
5 ∆B 1 ∆B1' ≥ ∆B '
20 ∆B1 ∆B3
∆G t 5 10 15 20 25 30t 5 10 15 20 25 30 t
-20
5 10 15 20 25 30 -5 ∆B2 -1 ∆B3' ≥ ∆B2'
-2
-40
∆G2 = ∆G3 -10
-3
120 40
15 ∆c1 100 20
∆R1
80
10 60 ∆l2 = ∆l3 5 10 15 20 25 30
t
40 -20
5 20
t -40
∆c10
2 = 15
∆c20
3 t -20 5 10 15 20 25 30 ∆R2 = ∆R3
5 25 30
∆ l1
(g) ∆c j (t ) (h) ∆l j (t ) (l) ∆R j (t )
Fig. 6.4.7. An improvement in the CR’s productivity, A1
We now study effects of a rise in the HR’s productivity, specified as fol-

lows: A3 = 1 ⇒ 1.3. The effects on the capital intensities, k j , per-capita
output levels, f j , and wage rates, w j , are given as follows
∆k j = ∆f j = ∆w j = 0, j = 1, 2 , ∆k3 = ∆f 3 = ∆w3 = 45.471.
As in the previous case, only the variables in the HR are affected. Al-
though the effects to similar to those in the previous case, it should be
noted that the national output level is reduced in the long term. As the
HR’s productivity is increased, its wage rate is increased and the region
becomes more attractive. People immigrate to the IR from the CR and IR.
As the HR’s productivity after the change, A3 = 1.3 , is still lower than the
CR’s productivity, A1 = 1.5 , the net result of the reallocation of the popu-
lation reduces the national output level.
40 600 ∆F3 10 ∆F
30 ∆k3 500
t
20
400
300 ∆N 3 -10
5 10 15 20 25 30
200
10 100 -20
∆k1 = ∆k2 = 0 t 5 10
∆15N120= 25
∆N30
2 t -30 ∆F1 = ∆F2
5 10 15 20 25 30
~ (t )
(a) ∆a j (b) ∆N j (t ) and ∆F3 (t ) (c) ∆F (t ), ∆F1 (t ) and ∆F2 (t )
600 5 4
500 2.5 ∆B ∆B3 t
400 ∆G3 5 10 15 20 25 30
3
300 -2.5
-5 ∆B2 2
200
100
-7.5 ∆B1 1
∆B ' ∆B3'
∆G -10
t
= ∆25G230 t
5 10 ∆15 -12.5
G1 20 ∆B52' 10∆B15' 20 25 30
(d) ∆G (t ) and ∆G j (t ) (e) ∆B(t ) and ∆B j (t ) (f) ∆B' (t ) and ∆B j (t )

1
40 40 600
∆c3 20 ∆ l1 = ∆ l 2 500
30
20 -20 5 10 15 20 25 30
400
300
∆R3
-40 200
10
∆c1 = ∆c2 = 0 -60 100
∆R1 = ∆R2 t
5 10 15 20 25 30
t -80 ∆l3 t 5 10 15 20 25 30
(g) ∆c j (t ) (h) ∆l j (t ) (l) ∆R j (t )
Fig. 6.4.8. An improvement in the HR’s productivity, A3
By the way, we can also take other forms of technological changes in

the system. A most common way is to assume that technological changes
are exogenous and time dependent, as specified as follows
A j (t ) = A0 j exp(g j t ), j = 1, 2 , 3 ,
where A0 j are the initial levels of region j ’s productivity and g j is the

growth rate of productivity. We can simulate the model as before. As little
new insights can be obtained for the type of technological change, we omit
further examination.44
6.4.7 Conclusions
This section proposed an economic growth model of a multi-regional small

open economy in a perfectly competitive economy. The national economy
consists of multiple regions and each region has one production sector and
one housing sector. Following the traditional literature of small open
economies, we assume that the rate of interest is fixed in international
market. The production side is the same as in the neoclassical growth the-
ory. Different from, for instance, the Solow growth model and the Ramsey
model, we used a utility function, which determines saving and consump-
tion with utility optimization without leading to a higher dimensional dy-
namic system like by the traditional approaches. Households move freely
among regions, equalizing utility level among regions by choosing hous-
ing, goods and savings. A region’s amenity is endogenous, depending on
the region’s output and population. The dynamics of J -regional national
economy is controlled by a J -dimensional differential equations system.
We explicitly solved the dynamics of the multi-regional economy. The
system has a unique stable equilibrium point. We simulated the motion of
the model and examined effects of changes in the rate of interest, the pref-
erence, and amenity parameters.
The comparative dynamic analysis provides some important insights.
For instance, if productivity improvement occurs in the region with the
highest productivity, the national output rises; but if productivity im-
provement occurs in the region with the lowest productivity, the national
output falls. This hints on the possibility that if the central government im-
proves a technologically less advanced region, the national output may ac-
tually fall. We also show that amenities may strongly affect agglomeration
of people and economic activities. It should be remarked that from the em-
pirical study on wage determination with a large panel of French workers,
Combes et al. (2003) have found that individual skills account for a large
fraction of existing spatial wage disparities with strong evidence of spatial
sorting by skills and endowments only appear to play a small role. As far
44 See, Zhang (2005a), for various sources of knowledge creation, diffusion and
utilization and how different aspects of technological changes can be introduced

into economic growth theory.
6.5 On the Alternative Utility Function 227
as relations between wage disparity and endowments in small open

economies are concerned, our model gives similar results. It remains to
prove whether this property of economic development can be observed
under more general conditions. We may extend and generalize the model
in different ways. We may analyze behavior of the model with other forms
of production or utility functions. There are multiple production sectors
and households are not homogenous. Any extension will cause some ana-
lytical difficulties because of the nature of regional dynamics. Moreover, it
is known that issues related to tax competition among regions has increas-
ingly caused interests in economic geography.45 We can extend the dy-
namic equilibrium framework proposed in this section to deal with these
issues. As mentioned before, the new economic geography has mainly
concerned with monopolistic competition, scale economies, and transport
costs in economic geography.
6.5 On the Alternative Utility Function
As a concluding remark, we discuss the theoretical basis for the utility

function used in this chapter.46 We assume that at any point of time the
consumer has preferences over alternative bundles of commodities, which
can be divided into goods, services, and time distribution of the consumer.
The behavioral rule consists of maximization of these preferences under
budgets restrictions of finance, or time, or human capital, or energy.
A commodity is characterized by its location, date at which it is avail-
able and its price. At each point of time, the consumer is faced with a
commodity bundle consisting of (finite) real numbers
{x (t )}, j = 1, 2 , L, m ,
j
indicating the quality of each commodity. The commodity space consists

of commodity bundles. Here, we omit issues related to spatial location. Let
us denote the price of commodity j by p j (t ). For simplicity, we omit
time index of x and p except in some circumstances. Both commodity
vector x, and price vector p , can be represented by points in Euclidean
45 See, for instance, Andersson and Forslid (2003), Baldwin and Krugman (2004),
Bayindir-Upmann and Ziad (2005), and Borck and Pflüger (2006).
46 This is section is referred to Zhang (2005a: Chap. 2). The discussion on the re-
lationship between the preference structure and utility function is actually based on
Barten and Böhm (1982).
space R m , i.e., x ∈ R m and p ∈ R m . The value of the commodity bundle

at any point of time is given by p(t )x(t ) = ∑ j p j (t )x j (t ) . The consumption
set, denoted by X , consists of all consumption bundles, which are possi-
ble. Let us assume that the consumer’s choices are restricted by the fact
that the value of his consumption should not exceed his income w(t ), at
each point of time. The budget set
β ( p , x , t ) ≡ {x ∈ X px ≤ w},
is the set of possible consumption bundles whose value does not exceed
the income.
The consumer has tastes and desires. They are important in analyzing
why the consumer chooses a bundle from the consumption set. Mathemati-
cally, we represent the preference structure by the consumer’s preference
relation, f t , at each point of time which is a binary relation on X . For
any two bundles, x(t ) and y (t ), x∈X and y ∈ X , x f y means that x is
at least as good as y at time t. Before discussing the relation between the
preference relation and utility functions, we introduce the following axi-
oms.
Axiom 1 (Reflexibility)
For all x ∈ X , x f x , i.e., any bundle is as good as itself.
Axiom 2 (Transitivity)
For any three bundles, x , y , z in X such that x f y and y f z it is
true that x f z .
Axiom 3 (Completeness)
For any two bundles x and y in X , x f y or y f z .
Axiom 4 (Continuity)
For every x ∈ X the upper contour set {y ∈ X y f x} and the lower con-
tour set { y ∈ X | x f y} are closed relative to X .
A preference relation f which satisfies the first three axioms is a com-

plete preordering on X and is called a preference order. A bundle x is
said to be strictly preferred to a bundle y , i.e., x f y iff x f y and not
y f x . A bundle x is said to be indifferent to a bundle y, i.e., x ~ y iff
Appendix 229
x f y and y f x . The indifference relation defines an equivalent relation

on X , i.e., ~ is reflexible, symmetric, and transitive. We always assume
that X includes at least two bundles x' and x" such that x' f y". In order
to solve the problem of the representability of a preference relation by a
numerical function, we introduce the concept of utility function.
Definition 6.5.1
Let X denote a set and f t a binary relation on X at time t . Then a func-
tion u from X into real R is a representation of f , i.e., a utility function
for the preference relation f , if, for any two points x and y,
ut ( x) ≥ ut ( y ) iff x f t y at point of time t .
It seems that Pareto was the first to recognize that arbitrary increasing
transformation of a given function would result in identical maximization
of a consumer. From the above definition we see that for any utility func-
tion ut and any increasing transformation f : R → R the function
vt = f o ut is also a utility function for the same preference relation f .
The following theorem is referred to Debreu (1959) or Rader (1963).
Theorem 6.5.1
Let X denote a topological space with a countable base of open sets and
f a continuous preference order defined on X , i.e., a preference relation
that satisfies Axioms 1-4. Then there exists a continuous function u .
The above theorem shows that under certain conditions the concepts of
utility and of the underlying preferences may be used interchangeably to
determine demand at any point of time.
Appendix
A.6.1 Growth of a Small Overlapping-Generations Economy
This section introduces a typical model of a small overlapping-

generalizations (OLG) economy. The model is proposed by Galor (1994).47
47 It should be noted that Galor also examines the case when there are tariffs.
The model is built for an economy of perfect competition with an infinite

horizon. In every period two, consumption and investment goods, are pro-
duced with two factors, capital and labor in the world. The labor force,
N (t ), of the country grows at a fixed rate, n , that is
N (t + 1) = (1 + n ) N (t ).
For simplicity, assume that the small open economy specializes in pro-
ducing the investment good. The output produced in every period is used,
partly as capital in next period production processes while the rest is ex-
ported and traded freely in return for the consumption good.
Production of the investment good in the small economy occurs within a
period according to the following neoclassical production function:
F (t ) = F (K (t ), N (t )), alternatively
K (t )
f (t ) = f (k (t )), k (t ) ≡ ,
N (t )
where K (t ) and N (t ) are respectively the capital stock employed by the
production sector and the labor force of the economy. Assume that the rate
of capital depreciation is zero. Choose the investment good as the nu-
meraire. The marginal conditions are
r (t ) = f ' (k (t )), w(t ) = f (k (t )) − k (t ) f ' (k (t )), (A.6.1.1)
where r (t ) and w(t ) are the prevailing wage and the interest rates, respec-
tively at t .
Individuals are identical in the system over time and each individual
lives only two periods. In the first period individuals work and earn w(t )
and in the second they are retired. Individuals born at t are characterized
by their intertemporal utility function u (ct (t ), ct (t + 1)), where ct (t ) is the
consumption of period t . Assume that u is twice continuously differenti-
able and strictly quasi-concave, defined over the consumption set R+2 . It is
also required Du >> 0 , ∀ c ∈ R+2+ , and for ct (t ), ct (t + 1) > 0 ,
lim u1 (ct (t ), ct (t + 1)) = ∞ , lim u1 (ct (t ), ct (t + 1)) = 0 .

ct (t )→ 0 ct (t )→∞
Young individuals at time t supply their unit-endowment of labor ine-

lastically and allocate the resulting income w(t ) between first period con-
sumption, ct (t ), and savings St (t ). The budgets over two periods are thus
given by
Appendix 231
p(t )ct (t ) + S (t ) ≤ w(t ),
1 + r (t + 1) S (t ) (A.6.1.2)
ct (t + 1) = .
p(t + 1)
Individuals maximize u (ct (t ), ct (t + 1)) subject to the above constraints.

The problem has a unique solution for strictly positive prices. Let p(t ) be
fixed in the world market and constant over time. Then, we can express the
solution as
 w(t )   w(t )  (A.6.1.3)
ct (t ) = ct  , r (t + 1) , S t (t ) = w(t ) − pct  , r (t + 1) ,
 p   p 
where ∂S / ∂r ≥ 0 and ∂S / ∂w > 0 .
If neither international lending nor borrowing is allowed, the stock of
capital changes according to
K (t + 1) = N (t ) S t (t ) = F (t ) + K (t ) − X (t ), (A.6.1.4)
where X (t ) is the economy’s export at t . The trade balance is guaranteed

by
X (t ) = p[N (t − 1)ct −1 (t ) + N (t )ct (t )]. (A.6.1.5)
We have thus described the model. Under certain conditions the dy-
namic system has a unique (locally) stable equilibrium. We will not illus-
trate the dynamic analysis of the discrete system as almost all the dynamic
models in this book are in continuous time.
A.6.2 Habits and Current Account Dynamics
Many empirical studies have identified the following stylized facts about
current experiences (Ikeda and Gombi, 1998): (i) adverse productivity
shocks improve the current account; (ii) savings and investment display a
positive correction in the short- and long-run; and (iii) temporary increases
in fiscal spending deteriorate the current account whereas permanent ones
exert at most weaker negative effects on it. To explain these findings,
much attention in theoretical literature has been paid to the intertemporal
aspects of savings, investment, and the current account (Sen, 1994;
Obstfeld and Rogoff, 1995). This section introduces a small country model
proposed by Ikeda and Gombi (1998) to analyze the equilibrium dynamics
of savings, investment and the current account and to explain the stylized
facts just mentioned.
The model is constructed for a small open economy populated with infi-
nitely lived identical agents. There is a composite traded good that can be
used for consumption and investment. The good is taken as numeraire.
Given the market wage rate, w(t ), households supply one unity of labor at
each point of time t and hold non-human wealth, a(t ), in the form of
bonds, b(t ), and equities. Bonds can be either purchased or issued freely at
a constant interest rate, r , in the international market. To describe habit
formation of households, introduce z (t ) the time- t consumption habit, i.e.,
the average of past consumption rates defined by
z (t ) = ∫ c(t )exp(− α (t − s)) ds ,

t
−∞
or equivalently
z& (t ) = α (c(t ) − z (t )), (A.6.2.1)
where α is the discount rate for past consumption rates, and c(t ) is the
consumption rate. The consumers’ lifetime utility is specified as
∞
U0 = ∫ [u (c(t ), z(t )) + v(g (t ))]exp(− θt ) dt ,

0
where g (t ) represents government services. It is assumed that v is con-

cave in g (t ) and u satisfies: (1) uc > 0 ; (2) u z ≤ 0 ; (3) uc + u z > 0; (4)
u is concave in (c, z ); (5) lim c →0 uc (c, z ) = ∞ uniformly in z ; and (6)
lim c →0 [uc (c, c ) + u z (c, c )] = ∞ .
Let x(t ) stand for lump-sum tax payments to the government. Given the
initial conditions (a(0 ), z (0 )), consumers choose {c(t ), z (t ), a(t )} to maxi-
mize U 0 subject to the following budget constraint
a& (t ) = ra(t ) + w(t ) − c(t ) − x(t ), (A.6.2.2)
and the non-Ponzi game condition. Together with the transversality condi-
tions for λ (t ) and ξ (t ), the optimal conditions are given by
u z (c, z ) = λ (t ) − αξ (t ),
ξ&(t ) = (θ + α )ξ (t ) − u (c, z ),
z
Appendix 233
λ& (t ) = (θ − r )λ (t ), (A.6.2.3)
where λ (t ) (≥ 0) and ξ (t ) (≤ 0) are respectively the shadow price of sav-

ing and the shadow price of habit formation. Assume θ = r in the rest of
this section (which implies λ& (t ) = 0 ). This assumption will bring the
model the zero root property that the steady state depends on the initial
condition (Sen and Turnovsky, 1990), even though it is frequently ac-
cepted in the literature.
The representative firm maximizes the present value of its future net
cash flows, choosing labor, N (t ), and rate of net investment, I (t ). Ne-
glecting depreciation of physical capital, we have K& (t ) = I (t ), where
K (t ) is the capital stock. Assume that the government levies a tax on equi-
ties at constant rate τ ≥ 0 . The arbitrage-free rate of return on equities is
equal to r + τ . The government maximizes
∞
V0 = max ∫ [AF (K , N ) − wN − I (1 + φ (I / K ))]exp[− (r + τ )t ]dt ,

0
where AF (K , N ) is the production function which is linearly homogene-

ous in N (t ) and K (t ) with productivity parameter A and Iφ (I / K ) de-
notes the capital adjustment cost function satisfying
 I   I   I 
φ (0 ) = 0, φ '   > 0, 2φ '   + φ "  > 0 .
K K K
Together with, the normalization N (t ) = 1, K& (t ) = I (t ), and the trans-
versality condition for K (t ), the optimal conditions are
AFN (K , 1) = w(t ),
I  I I 
1 + φ   + φ '   = q(t ),
K K K
  I   I 
2
q& (t ) = (r + τ )q (t ) −  AFK (K ,1) +   φ '   ,

  K   K 
where q(t ) is the shadow price of investment. From the second equation,
we solve
I (t ) = η (q(t ))K (t ), η (1) = 0, η ' (q ) > 0 .

The dynamics are thus given by
K& (t ) = η (q(t ))K (t ),
[
q& (t ) = (r + τ )q(t ) − AFK (K ,1) + η 2φ ' (η ) . ] (A.6.2.4)
Governments
The government’s budget is assumed to be balanced at any point of
time, i.e.
g (t ) = x(t ) + τV (t ),
where V (t ) is equities at time t .
Steady state and stability

The dynamics of ( z (t ), ξ (t ), K (t ), q(t )) are given by Eqs. (A.6.2.4) and
z& (t ) = α (c(t ) − z (t )),
ξ&(t ) = (θ + α )ξ (t ) − u z (c, z ).
It can be shown that the dynamic system exhibits saddle point stability.
A.6.3 Proving Lemma 6.4.1
We now prove Lemma 6.4.1. First, we note that k j , w j and f j (k j ) are de-
termined as functions of r * , which is fixed in the international market.
Hence, we treat them as constants in the dynamic analysis. Substituting
Eqs. (6.4.8), c j = ξyˆ j and s j = λyˆ j in Eqs. (6.4.9) and l j = L j / N j into
the utility functions, we have
U j = ρ ξ 0 + λ0θ j f ja Lηjξ ξ λλ N aj +b −η yˆ ξj + λ , j = 1, L , J .
Inserting the above equations in Eqs. (6.4.11), we get

η
 yˆ j  (A.6.3.1)
n j = g j   , j = 2, L, J ,
 yˆ1 
Appendix 235
where g j and η are defined in Lemma 6.4.1 and
N j (t )
n j (t ) ≡ .
N1 (t )
From the population constrain in (6.4.15), we have
N
N1 = .
∑
J
1+ j =2
nj
From this equation and Eqs. (A.6.3.1), we determine the population dis-
tribution as functions of ŷ j as follows
g j yˆ ηj N
Nj = , j = 1, ... , J ,
∑
J
g yˆ η
j =1 j j
in which g1 = 1. From Eqs. (6.4.4) and (6.4.5), we obtain
yˆ j = (1 + r * )a~ j + w j , j = 1, L, J . (A.6.3.2)
As r * and w j are independent of t , we see that ŷ j in region j is line-

arly related to a~ j and is independent of any other time-dependent vari-
ables.
From s j = λyˆ j in (6.4.9) and Eqs. (6.4.10), we have
a~& j = λyˆ j − a~ j , j = 1, 2 , L, J . (A.6.3.3)
Substituting Eqs. (A.6.3.2) into Eqs. (A.6.3.3) yields

a~& j = λw j − (1 − λ − λr * )a~& j , j = 1, L, J . (A.6.3.4)
As each equation is unconnected to the rest of the equations, it is

straightforward to show that the solutions of Eqs. (A.6.3.4) are given by
Eqs. (6.4.18).
A.6.4 The Keynesian Consumption Function and the OSG

Approach
The Keynesian theory of consumption is that current real disposable in-

come is the most important determinant of consumption in the short run.
Many attempts were made to correlate aggregate consumption expendi-
tures over time with aggregate disposable income and other variables. The
traditional Keynesian consumption function posits that consumption is de-
termined by current disposable income, i.e.
C (t ) = a + bY (t ), a > 0 , 0 < b < 1,
where a and b are constant, C (t ) is real consumption at time t , and Y (t )
is real disposable income (which is the same as the current income in our
model), which equals GNP minus taxes. It can be seen that if we swap the
real disposable income in the Keynesian model with the disposable per-
sonal income in our model in Sect. 6.1,
The parameter, b , is the marginal propensity to consume, which meas-
ures the increase in consumption in association with per unit increase in
disposable income. The intercept, a , measures consumption at a zero level
of disposable income.
Consumer spending
450 line where income = spending
consumption
spending is assumed to rise but less quickly

than income
Real disposable income

Fig. A.6.4.1. The Keynesian consumption function
Because of the intercept, the Keynesian consumption is not a propor-

tional relationship between consumption and income. The ratio of con-
sumption to income is termed the average propensity to consume (APC),
i.e.
C a
APC = =b+ .
Y Y
Appendix 237
The average propensity to consume declines as income increases. The

average propensity to consume is greater than the marginal propensity to
consume, by the amount of a /Y . The ratio of saving to income is termed
the average propensity to save (APS), i.e.
Y −C a
APS ≡ =1− b − .
Y Y
We have
APC + APS = 1.
As Y increases, the average propensity to save rises.
The OSG model with the Cobb-Douglas utility function determines the
relationship between consumption and disposable personal income
C (t ) = ξYˆ (t ) = ξY (t ) + ξK (t ).
We now connect the two theories by treating a in the Keynesian model
as a wealth-related variable. If we assume that the intercept a is depend-
ent on wealth and marginal propensity to propensity, b , is related to the
propensity to consume, ξ , in the following way
a = ξK (t ), b = ξ ,
then the Keynesian consumption function is identical to the consumption
function in the OSG model. We may call our consumption function as a
generalized Keynesian consumption function. We can define the APC and
APS, denoted by c (t ) and s (t ), respectively, for the OSG model in the
same way as in the Keynesian consumption function. In the OSG model
C ξ (Y + K ) K
c (t ) = APC ≡ = =ξ +ξ ,
Y Y Y
S − K λY − δK K
s (t ) = APS ≡ = =λ −ξ ,
Y Y Y
where we use ξ + λ =1. It should be noted that according to the definition
of the APS
S (t ) − K (t ) K& (t )
s (t ) = APS ≡ = .
Y (t ) Y (t )
The APC in the OSG model rise as wealth increases or as current in-
come declines; The APS in the OSG model rise as wealth falls or as cur-
rent income rises. It should be noted that APC + APS = 1. It is possible

for the APS to be negative in the OSG model.
It should noted that the life cycle hypothesis has also played an impor-
tant role in analyzing households’ The hypothesis was developed by Irving
Fisher and Roy Harrod, before later being extended by Ando and Modi-
gliani (1963).48 It assumes that individuals come a constant percentage of
the present value of their life income. This is dictated by preference and
income. The hypothesis is to explain the empirical work on consumption
function.49 It has been observed that the relationship between consumption
and current income would be nonproportional and the intercept of con-
sumption function is not constant over time. As stated out by Modigliani
(1966), “The point of departure of the life cycle model is the hypothesis
that consumption and saving decisions of households at each point of time
reflect a more or less conscious attempt at achieving the preferred distribu-
tion of consumption over the life cycle, subject to the constraint imposed
by the resources accruing to the household over its lifetime.” Consumption
depends not just on current income but also on long-term expected earn-
ings over their lifetime. Zhang (2005a) shows how the OSG model can be
related to this hypothesis.50
A.6.5 The Solow Growth Model and the OSG Approach
The Solow growth model, often also called the neoclassical growth model,
is a work horse of economic growth theory. Most neoclassical models are
extensions and generalizations of the pioneering works of Solow and Swan
in 1956.51 The behavior of the production sector is the same as in the OSG
model in Sect. 6.1.52 Nevertheless, the Solow model assumes that the
agents regularly set aside some fairly predictable portion ŝ of its output
for the purpose of capital accumulation; hence
48 See also Modigliani and Brumberg (1954) and Modigliani (1986).

49 This section is referred to Froyen (1999: 282-286).
50 Zhang (2005a) also discussed the permanent income hypothesis developed by
Friedman (1957) and possible relations between the permanent income hypothesis
and the OSG approach.
51 Solow (1956) and Swan (1956). The Solow model is often called the Solow-
Swan model because Swan’s work is similar to Solow’s seminal paper.

52 The description of behavior of producers and production sectors follows the
traditional approach (e.g., Burmeister and Dobell, 1970; Azariadis, 1993; and
Zhang, 1999).
Appendix 239
k&(t ) = sˆf (k (t )) − (n + δ k )k (t ).
We see that the differential equation for per-worker-capital accumula-
tion in the Solow model is mathematically identical to the capital accumu-
lation equation in the OSG model defined by Eq. (1.1.12) in Sect. 6.1
k&(t ) = λf (k (t )) − (ξ + n )k (t ),
if we specify U = c ξ s λ . The Solow model and the OSG model have the
same dynamic properties – the system has a unique stable equilibrium. But
the OSG model holds that the saving rate is time-dependent; the Solow
model predetermines the saving rate.
We now show that under certain circumstances the OSG model can ex-
plain what the Solow model forecasts. The OSG model endogenously de-
termines saving and consumption. For simplicity, we let δ k = 0 . The
OSG’s capital accumulation is given
K& (t ) = λF (t ) − ξK (t ). (A.6.5.1)
In the OSG model, the APS is given by

λF (t ) + λK (t ) − K (t )
s (t ) = .
F (t )
We are interested in when s (t ) in the OSG model equals the predeter-
mined saving rate ŝ in the Solow model, i.e.
λF (t ) + λδK (t ) − δK (t ) (A.6.5.2)
= ŝ .
F (t )
If the propensity to save λ is considered as an endogenous variable, the
above equation holds if
K (t ) (1 − sˆ )δ (A.6.5.3)
λ (t ) = sˆ + (1 − sˆ )δ = sˆ + (< 1).
F (t ) + δK (t ) f (k (t )) / k (t ) + δ
As ξ (t ) + λ (t ) = 1, ξ (t ) is also a function of k / f (k ). If the propensity

to save is related to the ratio of capital per capita and output per capita as
in Eq. (A.6.5.7), then s (t ) in the OSG model is constant and is equal to the
saving rate, ŝ , in the Solow model. Inserting Eq. (A.6.5.7) into Eq.
(A.6.5.1) yields
K& (t ) = λ (t )F (t ) − (ξ (t ) + λ (t )δ k )K (t ) =
λ (t )(F (t ) + δK (t )) − K (t ) = sˆF (t ) − δ k K (t ).
We see that under Eq. (A.6.5.7) the evolution of capital in the OSG model
is identical to that in the Solow model.
Theorem A.6.5.1
Let the production sectors be identical in the OSG model and the Solow
model. If the saving rate, ŝ, in the Solow model and the propensity to save
λ (t ) in the OSG model satisfy Eq. (A.6.5.8), then the OSG model is identi-
cal to the Solow model in terms of the saving rate (out of current income), the
consumption rate, the interest rate, the wage rate, output, income, consump-
tion, and saving.

6 Growth of Small Open-Economies With Capital Accumulation

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

6 Growth of Small Open-Economies With Capital Accumulation

Uploaded by

Copyright:

Available Formats

6 Growth of Small Open-Economies with Capital

It is significant to examine possible effects of trade upon personal income

man and Venables (1995), Manasse and Turini (2001).

to wherever returns appear likely to be the highest. Yet, modern technol-

6.1 The One-Sector Growth (OSG) Model of a Closed

The main deviation of this book from traditional approaches in modeling

6.1.1 The Model

We consider an economy of one production sector and one type of house-

3 This section is based on Zhang (2006a: Chaps. 1 and 2).

(ii) The slope of f (k ) represents the marginal product of capital, i.e.

6 A production function F ( K , N ) is called neoclassical if it satisfies the following

We assume (identically numerous) one production sector. Its goal of

in which k (t ) ≡ K (t ) / N (t ). We assume that factor markets work quickly

It should be noted that in Eq. (6.1.5), like Y (t ), the value of wealth, K (t ),

A typical consumer is to choose his most preferred bundle (c(t ), s (t )) of

in which yˆ (t ) ≡ Y (t ) / N (t ) + k (t ). The following theorem holds.

then (c * , s * ) is a strict local solution to the utility maximization problem. If

where ξ and λ are parameters, it is straightforward to solve the optimal

K& (t ) = s ( yˆ (t ))N (t ) − K (t ). (6.1.11)

The stability guaranteed above is local. We now show that if s( yˆ ) is

Fig. 6.1.1. Evolution of capital-labor ratio in the OSG model

9 The theorem is proved in Zhang (2006a, Chap. 2).

6.2 The Ramsey Growth Model and the OSG Approach12

one of the most remarkable contributions to mathematical economics ever

The influences of Ramsey’s classical contribution are reflected in the

12 This section is based on Zhang (2005a: Chap. 2).

U = ∫ u (c(t ))e nt e − ρt dt , c(t ) ≥ 0 , t ≥ 0 .

The household makes the decision subject to a lifetime budget con-

The trajectory of the economy is determined by Eqs. (6.2.2) and (6.2.5).

The two eigenvalues are given by

r−ρ f ' (k ) − ρ (6.2.7)

For Eqs. (6.2.7) and (6.2.8) to be equal, it is sufficient for λ (t ) to evolve

15 We will not provide a complete analysis of the model. Refer to Takayama

(1985) and Romer (1996) in detail.

It is easy to see what Samuelson means by looking at the form of utility

The specified form means that the household’s utility at time 0 is a

16 Samuelson (1937: 161). It is nowadays common to use the utility in modeling

considered ρ > 0 “ethically indefensible”. If ρ = 0, by equation con-

The [discounted utility] model, which continues to be widely used by econo-

Any model as widely employed as the representative agent model begins to

18 Frederick et al. (2002).

useful to remind ourselves periodically of its limitations. Despite the criticisms

6.3 A Small Open Economy with Capital Accumulation

This section describes dynamics of a small country economy.23 An open

22 A comprehensive survey on time discounting and time preference is given by

6.3.1 The Model with General Production and Utility Functions

in which k (t ) ≡ K (t ) / N . We assume that factor markets work quickly

production factors at any point of time.

knowledge accumulation as or Marshallian externalities. Recent publications on

It is straightforward to show dg / dr * = 1 / f " < 0 and dh / dr * = − g < 0.

where E (t ) ≡ r * B(t ) and we use

Here, we require Y (t ) > 0, i.e.

a(t )N + F (t ) > − r * B(t ) + δ k K (t ).

A typical consumer is to choose his most preferred bundle (c(t ), s (t )) of

s.t.: c(t ) + s (t ) = y (t ). (6.3.5)

From y (t ) = w0 (r * ) + (1 + r * )b(t ) and a(t ) = g (r * ) + b(t ), for any fixed

The condition Y ≥ 0 is guaranteed if

We have Φ (b ) = h / (1 + r * ) > 0. As 0 < s ' ( y ) < 1, for sufficiently large b,

27 We will specify production and utility functions to further demonstrate these

we see that the sign of dy / dr * is ambiguous in general. As the impacts on

6.3.2 The Dynamics with the Cobb-Douglas Functions

From f = Ak α and Eqs. (6.3.1), we solve

As η + ξ + λ = 1, we see that if η + ξ > r λ , then λ > 0 . As it is rea-