You are on page 1of 22

Review of Economic Dynamics 15 (2012) 72–93

Contents lists available at SciVerse ScienceDirect

Review of Economic Dynamics


www.elsevier.com/locate/red

International trade, exhaustible-resource abundance and economic


growth ✩
Beatriz Gaitan a , Terry L. Roe b,∗
a
Free University of Berlin, Garystrasse 21, 14195 Berlin, Germany
b
University of Minnesota, 337h Ruttan Hall, 1994 Buford Av., St. Paul, MN 55108, USA

a r t i c l e i n f o a b s t r a c t

Article history: Countries with oil and other natural resources have grown less rapidly than those countries
Received 25 April 2006 without. This phenomenon is known as the “natural resource curse”. We develop an
Revised 12 July 2011 infinite-horizon, two-country model of trade in which countries are identical, except that
Available online 30 August 2011
one country is endowed with deposits of an exhaustible resource and the other is not.
JEL classification:
Within the context of the model, we show that this phenomenon can be explained in part
F11 by an inelastic demand for the exhaustible resource that increases growth in trade revenues
F43 and induces the resource-abundant country to invest relatively less than the country
O13 lacking in exhaustible resources. These results are derived analytically and illustrated by
O41 an empirical analysis based on plausible parameters obtained from data.
Q32 © 2011 Elsevier Inc. All rights reserved.

Keywords:
Growth
Exhaustible resources
Depletable resources
International trade
History dependent equilibria

1. Introduction

The economics of exhaustible resources received increased attention following Sachs and Warner’s (1995, 2001) sug-
gestion that countries with ample natural resources tend to grow less rapidly than natural resource-scarce countries. This
apparent paradox led to a number of papers with conflicting findings. Sala-i-Martin (1997) and Doppelhofer et al. (2000)
find the export share of primary products to be negatively correlated with economic growth. They also find, however, that
the fraction of GDP in mining and GDP growth are positively correlated. Papyrakis and Gerlagh (2007) find economic growth
among U.S. states to be negatively affected by their natural resource abundance. Empirical studies by Lederman and Maloney
(2003), Stijns (2005) and Brunnschweiler (2008) either come to opposite conclusions or find that natural resource abundance
does not affect growth. Brunnschweiler and Bulte (2008) conclude from their statistical analysis that the apparent paradox
may be a red herring.1
Given this puzzling empirical evidence, the challenge is whether a theoretical model that posits a specific yet plausible
economic structure can help us understand whether and how resource abundance affects economic growth, thus alleviating
to some degree the red herring suggestion. Instead of focusing on the broader concept of natural resource abundance,


We thank Timo Trimborn, Cees Withagen, Harris Dellas and anonymous referees for their helpful comments.
* Corresponding author.
E-mail addresses: beatriz.gaitan@fu-berlin.de (B. Gaitan), troe@apec.umn.edu (T.L. Roe).
1
They find that resource abundance, constitutions and institutions determine resource dependence.

1094-2025/$ – see front matter © 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.red.2011.08.002
B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93 73

we focus on analyzing the effects of exhaustible resource abundance on economic growth.2 Since exhaustible resources
are exported to countries deficient in these resources (referred to henceforth as resource-poor countries), we argue that
resource-poor countries likely play a role in the growth performance of the endowed countries (referred to henceforth as
resource-abundant countries). Further, the abundance of these resources must also influence the economic performance of
resource-poor countries.3 The main contribution of this paper is to show that country interdependence through trade can
give rise to the resource curse under plausible conditions. This interrelationship is the key focus of this paper.
We consider a continuous-time, infinite-horizon two-country model of trade and analyze the conditions which lead a
resource-abundant country to grow less rapidly than a resource-poor country—we call this the resource curse (henceforth
RC). One country is endowed with deposits of an exhaustible resource (such as petroleum). The countries produce an
identical final good with the same technology using a flow of the exhaustible resource and human and physical capital.
At each instant of time, the resource-abundant country extracts and employs some of the exhaustible resource to produce
the final good and exports the remainder to the resource-poor country. International borrowing is not allowed and trade
between the two countries is balanced at each instant of time. Each country has a representative household that maximizes
discounted instant utility from consuming the final good subject to a budget constraint. Consumers across countries have
identical preferences. Thus, countries are identical and differ only with regard to their initial endowments of the exhaustible
resource and possibly capital stocks.
Since, after controlling for corruption, Sachs and Warner (1995) find that resource abundance negatively affects economic
growth, we first consider a setting in which capital stocks across countries are equal as a way to abstract from corruption
and institutional effects. The model suggests that even under these circumstances, the RC can occur. We analytically char-
acterize the conditions causing the rate of growth of the resource-abundant country’s gross domestic product (GDP) to be
transitionally smaller than that of the other country. This “growth gap” solely depends on technological and preference
parameters. Specifically, the size of the elasticity of intertemporal substitution and the exhaustible resource demand’s own
price elasticity play a crucial role in determining whether such a gap occurs. In particular, given an elasticity of intertempo-
ral substitution equal to or less than unity, the more own price inelastic is the derived demand for the exhaustible resource,
the larger the growth gap, that is, the stronger the RC.
This result may seem counterintuitive. In the process of growth, an inelastic demand for a flow resource mined from an
exhaustible resource tends to generate a growing income stream to the resource-abundant country which, one may incor-
rectly conclude favors the country’s relative income growth. An intuitive explanation is the following. An inelastic own price
elasticity of demand for the resource is associated with a factor of production that accounts for a relatively small share in the
value of total output, as is the case with petroleum. In the process of economic growth, the stock of the resource is depleted,
the price of the flow resource increases in transition, and the inelastic own price elasticity causes the total revenue remu-
nerated to the resource-abundant country to increase. Households in the resource-poor country obtain a higher discounted
value of utility by saving to increase their human and physical capital stocks (and income) to remunerate the increasing
cost of the flow resource, thus causing trade to be balanced. The growing income stream from exports of the exhaustible
resource induces households in the resource-abundant country to invest relatively less than those in the other country, and
thus, at the expense of future income growth. This effect is reinforced the stronger incentives are to smooth consump-
tion over time. Our empirical exercise helps to confirm our analytical results and shows that our model can explain about
one-fourth of the average GDP growth gap between resource-poor and resource-abundant countries observed in the data.
Several hypotheses have been put forward to explain why resource-abundant economies grow less rapidly than resource-
poor countries. Tornell and Lane (1995) argue that it may be explained by the struggle of groups attempting to extract
natural resource rents. Sachs and Warner (1995) argue in favor of Dutch Disease effects (see Corden, 1984 for a survey
on this literature). Rodríguez and Sachs (1999, p. 278) argue that “resource-rich countries may grow more slowly because
they are likely to be living beyond their means . . . ”. We argue that the interaction of resource-abundant and resource-poor
countries through international trade plays a role in the growth performance of resource-abundant countries.
The structure of our model draws upon early well-known literature on exhaustible resources.4 While many studies focus
on single economies (e.g. Rodríguez and Sachs, 1999, and Kemp and Long, 1982), only few address international trade issues
associated with exhaustible resources. Among these are Asheim (1986) and Hartwick (1995), who study a two-country
world and investigate whether constant consumption paths are achievable when rents accruing to exhaustible resources are
invested in new capital. Kemp and Long (1980) analyze monopsonistic behavior in a two-country model with exhaustible
resources, absent of physical capital. Brander and Taylor (1998) and Jinji (2007) investigate the effect of international trade
on welfare using Ricardian models of trade with renewable resources.

2
Sachs and Warner’s (1995) initial paper measures natural resource abundance as the ratio of agricultural, mining and fuel exports to GDP. Their follow-
up paper (Sachs and Warner, 2001) shows that excluding agricultural commodities from their measure of resource abundance does not alter their earlier
findings.
3
For example, there is a strand in the economic literature that associates the productivity slowdown in the U.S., which started in the 1970s, to oil price
shocks (see, for example Rotemberg and Woodford, 1996).
4
Hotelling (1931) characterizes the behavior of the prices of exhaustible resources. Dasgupta and Heal (1974), Hartwick (1977), Solow (1974) and Stiglitz
(1974) provide the foundation for modeling exhaustible resources employed in final good production where agents optimize over an infinite horizon.
They address issues of sustainability and inter-generational equity, which are also embedded in our model. Chiarella (1980a) and Cigno (1981) introduce
endogenous technological progress and endogenous population growth, respectively, to the model developed by Stiglitz, and focus on stability properties.
Pezzey and Withagen (1998) analyze the dynamic behavior of consumption in the model developed by Dasgupta and Heal.
74 B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93

Our work relates more closely to that of Chiarella (1980b) and Geldrop and Withagen (1993). Chiarella uses a two-
country model. One country supplies an exhaustible resource, whereas the other produces a final good which it exports
in exchange for the exhaustible resource. Chiarella studies the dynamic behavior and stability properties of consumption
shares. In contrast to this work, we allow both countries to be engaged in the production of the final good. In this way,
the model can transitionally display different cross-country GDP growth rates. Geldrop and Withagen introduce n number
of trading partners and consider neoclassical production functions and extraction costs. They study the dynamic behavior
of consumption. They abstract from any engine of growth, and thus consumption approaches zero in the long run. We
consider two economies and do not consider extraction costs. However, we allow for technological change, which enables
positive levels of consumption and human and physical capital stocks to be achieved in the long run. Similarly to Geldrop
and Withagen, we investigate the dynamic behavior of consumption. In addition, however, we show analytically, for near
steady states, the dynamic behavior of the model’s variables. With the use of an empirical model, we subsequently test
whether the model is able to transitionally generate the differences in GDP growth rates observed in the data.
The paper is organized as follows. In Section 2, we introduce the model and characterize equilibrium. We then proceed in
Section 3 to analyze the convergence properties of the model. In Section 4 we analytically analyze the dynamic behavior of
variables in near steady states and derive conditions under which the resource-abundant country’s income grows less (more)
rapidly than that of the resource-poor country. In Section 5 we simulate the model under plausible parameter values.

2. The model

2.1. The set-up

We consider a two-country world. There is an exhaustible resource-abundant ( A) country (in situ) that is also endowed
with stocks of human capital and physical capital. The other is an exhaustible resourceless country (referred to as the
resource-poor ( P ) country), which is only endowed with human capital and physical capital. For brevity, we use “resource”
in reference to “exhaustible resource”. The A subscript is used to denote variables of the resource-abundant country and the
P subscript is used for those of the resource-poor country. Both countries produce an identical non-perishable final good
using identical technologies and employing human capital, physical capital and a flow of the resource in production. In
each country there are firms producing the final good. In country A , there is also a firm that extracts and sells flows of the
exhaustible resource domestically and internationally. The final good can be consumed, invested or exported. While we allow
for international trade in goods, we abstract from international borrowing and lending. In each country, a representative
household seeks to maximize the discounted utility of consumption, subject to a budget constraint. All households own the
stock of their respective country’s endowments. We abstract from population growth and assume that the same number of
people live in each country and normalize population per country to one.
The final good sector in country i for i = A , P rents physical capital 
K i (t ) and human capital 
H i (t ) and buys a flow of
the exhaustible resource  N i (t ) in competitive markets to produce output 
Y i so as to maximize profits using the following
technology5
 φ  φ  φ1  χ  χ1

Y i = Λ α
V i + (1 − α ) e ηt 
Ni V i = B β
where  K i + (1 − β)
χ
Hi (1)

and parameter restrictions: Λ > 0, B > 0, 0 < α < 1, 0 < β < 1, η  0, φ < 1 and χ < 1.  V i is a composite input of human6

and physical capital. Elasticities of factor substitution are σ ≡ 1/(1 − φ) and μ ≡ 1/(1 − χ ). We denote the rental rate of
human capital in country i (the wage rate per efficiency unit of labor) at instant t by r̃ H i (t ) for i = A , P and, similarly, the
rental rate of physical capital by r̃ K i (t ). In the absence of transportation costs and under perfect competition within and
between countries, the price of the exhaustible resource, q̃(t ), and the price of the final good, which we treat as numéraire,
are common to both countries.

2.2. The extracting sector

The extracting sector of country A finds the optimal path of resource extraction  N (t ) that maximizes the present value
of profits, subject to the constraint that cumulative extractions do not exceed the initial stock of the exhaustible resource

S (0). We presume that extractions are costless. Let r̃ i denote country-i’s return on assets for i = A , P . The extracting sector
solves
 ∞
t  ∞
 
max q̃(t )
N (t )e − 0 r̃ A (τ ) dτ
dt  
N (t ) dt  S (0) . (2)
0 0

Let 
S (t ) denote the stock of the exhaustible resource at time t. Since 
N (t ) is a flow, the resource constraint can also be
written as

5
Setting the profit maximization problem of the final good sector as a static or as a dynamic problem leads to the same results.
 β 1−β
6
Y i = Λ
V iα (e ηt 
N i )(1−α ) for φ → 0 and 
V i = B
Ki 
Hi for χ → 0.
B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93 75

S˙ (t ) = −
 N (t ). (3)
The necessary condition for a maximum and the transversality condition are, respectively, given by (the t notation is omitted
to avoid clutter)
t
q̃˙ /q̃ = r̃ A , lim q̃(t )
N (t )e − 0 r̃ A (τ ) dτ
= 0. (4)
t →∞

As the exhaustible resource is an asset, q̃˙ /q̃ = r̃ A is a no-arbitrage condition equalizing the returns between the exhaustible
resource and other assets.

2.3. Households’ optimization problem

Let 
X H A (t ) and 
X K A (t ) denote the investment in human capital and physical capital in country A. The representative
household of country A solves the problem of maximizing the discounted instant utility U (c̃ A (t )) of consumption c̃ A (t )
∞
     
max e −ρ t U c̃ A (t ) dt where U c̃ A (t ) ≡ c̃ A (t )1−θ − 1 /(1 − θ) (5)
0

subject to the intertemporal constraints

W ˙ (t ) = r̃ (t )

W A (t ) − c̃ A (t ), (6a)
A A
˙K (t ) = 
 K A (t ), 
X K A (t ) − δ K  K A (t )  0, (6b)
A


H (t ) = X (t ) − δ H (t ), 
  H (t )  0, (6c)
A HA H A A

W A (t ) = 

K A (t ) + 
H A (t ) + q̃(t )
S (t ), (6d)
K A (0), 
 H A (0), 
S (0) given (6e)
where ρ > 0 is the rate of time preference and θ −1 > 0 is the elasticity of intertemporal substitution. The depreciation rates 7

of human and physical capital are denoted by δ H and δ K , respectively.


W A (t ) denotes the total wealth of the representative
household of country A and equals the sum of  K A (t ), 
H A (t ) and the value of the stock of the resource q̃(t ) S (t ) at instant t.
Following Krebs (2003), the budget constraint disregards the households’ decision to assign time to different activities and
does not impose a non-negativity constraint on human capital investment ( Z A (t ) ≡ 
X H A  0). Let  K A (t ) +  H A (t ). Using
r̃ A = q̃˙ /q̃ (from 4), the budget constraint (6a)–(6e) can be rewritten as

Z˙ A (t ) = r̃ A (t )
 Z A (t ) + q̃(t )
N (t ) − c̃ A (t ), 
Z A (t )  0, 
Z A (0) given (7)
where q̃(t )
N (t ) is the revenue of the extracting sector of country A at instant t. Defining variables in an analogous manner,
a similar problem also holds for country P , except that wealth is
W P (t ) = 
Z P (t ) = 
K P (t ) + 
H P (t ). The budget constraint of
country P thus equals

Z˙ P (t ) = r̃ P (t )
 Z P (t ) − c̃ P (t ), 
Z P (t )  0, 
Z P (0) given. (8)
The Euler and transversality conditions of the consumer’s problem of country i equal
 
c̃˙ i /c̃ i = (r̃ i − ρ )/θ, lim 
Z i (t )/ e ρ t c̃ i (t )θ = 0 for i = A , P . (9)
t →∞

2.4. Factor price equalization

From the cost-minimization problem and zero-profit condition of the final good firm in country i, we obtain
 σ φσ  φ σ − φ1σ −1
1= α /r̃ V i + (1 − α )σ / q̃/eηt Λ for φ = 0 (10)
where the right-hand term is the unit cost of producing the final good, and
 χ μ − χμ
1
B −1
χμ
r̃ V i = R V (r̃ K i , r̃ H i ) ≡ β μ /r̃ K i + (1 − β)μ /r̃ H i for χ = 0 (11)

is the rental rate of composite input 


V i .8 Since the resource is sold at the same price in both countries, it follows from (10)
that the rental rate of the composite input r̃ V i is also equal across countries r̃ V A = r̃ V P . Denote r̃ V A = r̃ V P by r̃ V . Since r̃ i is
the return on assets in country-i, no-arbitrage possibilities imply

7
For θ → 1, U (c̃ A (t )) = log c̃ A (t ).
1−β
q̃1−α
β
r̃ α r̃ K i r̃ H i
8
For φ → 0, Eq. (10) becomes 1 = Vi
, and for χ → 0, Eq. (11) becomes r̃ V i = R V (r̃ K i , r̃ H i ) ≡ β β (1−β)1−β B
.
α α (1−α )1−α e η(1−α )t Λ
76 B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93

r̃ i = r̃ K i − δ K = r̃ H i − δ H ⇒ r̃ K i = r̃ i + δ K and r̃ H i = r̃ i + δ H . (12)
Substituting for r̃ K i and r̃ H i from (12) into (11), we have

r̃ V = R V (r̃ i + δ K , r̃ i + δ H ). (13)
Substituting (13) into (10) implies that r̃ A = r̃ P , and therefore, r̃ H A = r̃ H P and r̃ K A = r̃ K P . Thus, even in the absence of
international mobility of either type of capital, factor price equalization occurs. We define r̃ ≡ r̃ i , r̃ H ≡ r H i and r̃ K ≡ r̃ K i for
i = A, P .
Substituting (13) into (10) for r̃ V implicitly defines r̃ in terms of q̃/e ηt ; we denote this relation as ψ(q̃/e ηt ). Setting
r̃ = ψ(q̃/e ηt ) in (4), we obtain a differential equation describing the motion of the resource price q̃:
 
q̃˙ /q̃ = r̃ = ψ q̃/e ηt . (14)
Since (14) is a key feature of the model, we solve this equation numerically for given q̃(0) and contrast the result with
others. In Fig. 1, we plot the ratio q̃/r̃ H for q̃(0)/r̃ H (0) = 16.5 with parameter values indicated in Table 4.

Fig. 1. Exhaustible resource price–wage rate per efficiency unit of labor ratio.

Fig. 1 conflicts with Krautkraemer’s (1998) empirical findings and Nordhaus’ observation (see Nordhaus, 1992) of de-
clining relative resource prices.9 Others (Slade, 1985 and Pindyck, 1999), however, find that the real price of exhaustible
resources is U-shaped over time consistent with Pindyck’s (1978) theoretical findings. For illustration purposes, Fig. 2 plots
the natural log of the ratio of the price of crude oil to U.S. manufacturing wage as well as that of bituminous coal and
natural gas.10 The ratio of crude oil prices to manufacturing wages and the ratio of natural gas prices to manufacturing
wages appear to be U-shaped.
In our model, the ratio q̃/r̃ H can transitionally decline depending on the size of q̃(0), which is unknown and needs to
be computed. Its level is likely determined by each country’s human and physical capital stocks and the resource stock
available at time zero. Below, we examine how endowments of capital and the resource, as well as unexpected resource
discoveries, affect the dynamic behavior of q̃.
Factor price equalization and Eq. (9) imply that c̃ A and c̃ P grow at identical rates. Substituting r̃ = q̃˙ /q̃ into (9) yields
  1/θ
c̃˙ i /c̃ i = (q̃˙ /q̃ − ρ )θ −1 ⇒ c̃ i (t ) = q̃(t )/ q̃(0)e ρ t c̃ i (0) for i = A , P . (15)
Thus, c̃ i follows a similar pattern to that of q̃(t ), albeit influenced by the rate of time preference and the elasticity of
intertemporal substitution. Using (15), the transversality condition of country i can be rewritten as

lim 
Z i (t )/q̃(t ) = 0 for i = A , P . (16)
t →∞

2.5. Equilibrium

Definition. An equilibrium are paths of quantities c̃ i (t ),


W i (t ), 
H i (t ), 
K i (t ), 
Y i (t ), 
N i (t ), 
N (t ), 
S (t ) and prices q̃(t ), r̃ i (t ), r̃ H i (t )
and r̃ K i (t ) for i = A, P such that taking prices as given: c̃ i (t ),
W i (t ),  H i (t ) and  K i (t ) solve the consumer’s optimization
problem of country i;  Y i (t ), 
H i (t ), 
K i (t ) and 
N i (t ) solve the optimization problem of the final good sector of country
i = A, P ;  N (t ) and 
S (t ) solve the maximization problem of the extracting sector of country A;

9
See Slade and Thille (2009) for a review of this literature.
10
The plots presented are similar to those by Nordhaus (1992), which include data for the 1870–1989 period.
B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93 77

– the exhaustible resource market clears, i.e. 


N A (t ) + 
N P (t ) = 
N (t ); and
– the final good market clears, i.e.

Y A (t ) + 
 Y P (t ) = c̃ A (t ) + c̃ P (t ) + 
X K A (t ) + 
X K P (t ) + 
X H A (t ) + 
X H P (t ). (17)

An equilibrium is characterized by Eqs. (3), (4), (7)–(12), (17) and 


N A (t ) + 
N P (t ) = 
N (t ).

Fig. 2. Ratio of various exhaustible resource prices to U.S. manufacturing wage. (a) U.S. Manufacturing wage: 1890–1919 NBER (2009a); 1920–1938 NBER
(2009b); 1939–2008 Bureau of Labor Statistics (2009). (b) Coal price: 1890–1948 Potter and Christy (P-C 1962, p. 328); 1949–2008 EIA (2009a, p. 221).
(c) Oil price: 1890–1913 P-C, p. 319; 1914–1945 U.S. Census Bureau (1947, p. 752 Value/Quantity); 1946–1948 U.S. Census Bureau (1952, p. 695); 1949–
2008 EIA (2009a, p. 169). (d) Natural gas price: 1919–1921 P-C, p. 326; 1920–2008 EIA (2009b). The wage rate is the per hourly wage; the price of coal is
per short ton; the price of oil is per barrel of oil and the price of natural gas is per thousand cubic feet.

Proposition 1. The ratio c̃ A (t )/c̃ P (t ) is constant for all t and equals the ratio of total wealth of country A to total wealth of country P
((
Z A (t ) + q̃(t )
S (t ))/
Z P (t )).

Proof. See Appendix A. 2

This implies

c̃ A (t ) c̃ A (0) 
Z A (0) + q̃(0)
S (0)
= = (18)
c̃ P (t ) c̃ P (0) 
Z P (0)
and indicates that relative consumption (including its long-run value) is determined forever by each country’s initial wealth.

3. Long-run growth and convergence

In this section, we investigate the long-run and convergence properties of the model.

3.1. Properties of long-run growth

We begin by defining the model’s constant growth path.11

Definition. A constant growth path (or steady state) is an equilibrium c̃ i (t ),


W i (t ), 
K i (t ), 
H i (t ), 
Y i (t ), 
N i (t ), 
N (t ), 
S (t ), q̃(t ),
r̃ i (t ), r̃ H i (t ), r̃ K i (t ) for i = A, P for some initial conditions of the state variables  K i (0),  H i (0),  S (0) for i = A, P such that
c̃ i (t ),

W i (t ),  K i (t ),  H i (t ), 
Y i (t ), 
N i (t ), 
N (t ), 
S (t ), q̃(t ), r̃ i (t ), r̃ H i (t ), r̃ K i (t ) grow at constant rates.

Proposition 2. For σ < (>) 1 and R V (η + δ K , η + δ H ) < (>)Λα 1/φ or σ = 1 (Cobb–Douglas case), a long-run equilibrium (if it
exists) is characterized by a constant growth path, where c̃ i (t ),

W i (t ), 
K i (t ),  Y i (t ) grow at rate v ; 
H i (t ),  N (t ), 
N i (t ),  S (t ) grow at rate
S ; q̃ grows at rate v q̃ , where
v

11
Since the constant elasticity of substitution (CES) technology violates an Inada condition (e.g. r̃ V is bounded from above (below) by Λα 1/φ for σ <
(>)1), long-run equilibria can exist in which the marginal product of factors of production asymptotically approach their bound. One type of boundary
equilibrium corresponds to the case when a high elasticity of substitution between composite input  V i and flow resource 
N i generates endogenous growth
(see Dasgupta and Heal, 1974). We do not consider this possibility since evidence suggests that the elasticity of substitution between value added and, for
example, crude oil is less than one (see Rotemberg and Woodford, 1996). More generally, our calibration, presented in section five, indicates that boundary
types of equilibria are unlikely and are not pursued further to economize on space.
78 B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93

 
v ≡ (η − ρ )θ −1 , S ≡−
v ρ + η(θ − 1) θ −1 , v q̃ = η (19)
and r̃ i (t ), r̃ H i (t ) and r̃ K i (t ) are constant for i = A , P .

Proof. See Appendix B. 2

Thus, as Stiglitz (1974) finds, a sufficiently large rate of growth of resource-saving technological change η can lead to
long-run non-negative consumption growth rates. The more impatient consumers are (larger ρ ), the “more” negative the
long-run growth rate of resource extraction, thus speeding up its depletion. Provided v > 0, a larger θ increases the rate of
resource extraction (that is, the larger θ , the “more” negative v S̃ becomes).
Following the approach by Dasgupta and Heal (1974), it can be verified that  N (t ) > 0 for all t, and  S (t ) asymptotically
approaches zero as t approaches infinity. The transversality condition (16) holds if v S < 0, see Appendix B. We impose this
restriction throughout the paper. As shown in Appendix B, the long-run growth rate of resource extraction v S also equals
v − σ ( v q̃ − φ η) = v − η . Since v
S must be negative, this result implies that the long-run growth rate of consumption is
bounded from above by the rate of technological change η . This result is in contrast to the typical model employing only
capital and labor, where long-run output and consumption grow at the rate of labor-saving technological change. Unlike the
typical model, we now have an input that can be depleted, in particular, since a large ρ speeds up the long-run rate of
resource extraction so does the growth rate of long-run consumption. While a larger θ makes v S “more” negative, a larger
θ decreases the steady-state consumption growth rate (provided v > 0).
To provide insights into other long-run properties of the model, we proceed in the typical fashion by normalizing vari-
ables so they remain constant in the long run. Let r (t ) ≡ r̃ (t ), r V (t ) ≡ r̃ V (t ), r H (t ) ≡ r̃ H (t ), r K (t ) ≡ r̃ K (t ) and

K i (t ) ≡ 
K i (t )/e vt , H i (t ) ≡ 
H i (t )/e vt , V i (t ) ≡ 
V i (t )/e vt , Y i (t ) ≡ 
Y i (t )/e vt ,
W i (t ) ≡

W i (t )/e vt , c i (t ) ≡ c̃ i (t )/e vt , Z i (t ) ≡ 
Z i (t )/e vt , q(t ) ≡ q̃(t )/e v q̃ t ,
N i (t ) ≡ 
N i (t )/e vS t , N (t ) ≡ 
N (t )/e vS t , S (t ) ≡ 
S (t )/e vS t , for i = A , P . (20)
The final good firm’s profit maximization problem implies12
σ σ
1 − α rV 1 − α rV
Ni = V i = Ω(r ) Z i (21)
α q α q
where

1 (r + δ H )(r + δ K ) 1
Ω(r ) = μ . (22)
β (r + δ H )μ + (1 − β)μ (r + δ K )μ R V (r + δ K , r + δ H ) Bχμ
Substituting for N i from (21) into Ż A (from (7)), an equilibrium in terms of normalized variables is characterized by the
following equations of motion

q̇ = (r − v q̃ )q, (23a)
ċ i = (r − ρ − θ v )c i /θ = (q̇/q)c i /θ for i = A , P , (23b)
Ż P = (r − v ) Z P − c P , (23c)

1 − α rV
Ż A = (r − v ) Z A + q Ω(r )( Z A + Z P ) − c A , (23d)
α q

1 − α rV
Ṡ = − Ω(r )( Z P + Z A ) − vS S (23e)
α q
where r = ψ(q) and r V = R V (r + δ K , r + δ H ) is defined in (11).
Let a “bar” denote steady-state values. Setting q̇ from (23a) equal to zero and setting q̃/e ηt = q in Eq. (10), the steady-
state return on assets and normalized resource price equal
1  φ σ − φ1σ
r̄ = η, q̄ = (1 − α ) φ 1/Λφ σ − α σ /r V for φ = 0 (24)

where r̄ V = R (r̄ K , r̄ H ), r̄ K = r̄ + δ K and r̄ H = r̄ + δ H .


V 13
Setting (23e) equal to zero, it follows that N = − v S̃ S and the
steady-state levels of Z A , Z P and S satisfy

μ
1−β r K βμrH
12
The final good firm’s profit maximization problem implies H i = ( β )μ K i . Since Z i = H i + K i = ( 1−β rK
)μ K i + K i , we have K i = μ Z
β μ r H +(1−β)μ r K i
rH β rH μ
μ
(1−β)μ r K χ χ 1
and H i = μ Z . Using V i = B (β K i + (1 − β) H i ) χ , (11) and (12), we obtain V i = Ω(r ) Z i .
β μ r H +(1−β)μ r K i
μ
1
13
For φ → 0, we have q̄ = (α α (1 − α )1−α Λ/r̄ α
V)
1−α .
B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93 79


ZA + ZP − vS α q̄ σ
=Γ ≡ . (25)
S Ω(r̄ ) 1 − α r̄ V
Setting (23c) and (23d) equal to zero, normalized steady-state consumption levels in countries A and P , respectively, equal

c A = − v
S ( Z A + qS ) and c P = − v
S ZP. (26)

Setting (23b) equal to zero leads to r̄ = η, which is already known. 14

Thus, any combination of ( Z P , Z A , S ) satisfying (25) represents a steady-state equilibrium. Let


  Z +Z 
 A P
T = (Z P , Z A, S)  =Γ (27)
S
denote the set of triplets ( Z P , Z A , S ) satisfying (25). We next demonstrate that T is saddle path stable.

3.2. Convergence

To study the convergence properties of the model, it is convenient to reduce the dimensionality of the system. Inte-
grate (23b) to obtain

c i (t ) = γi q(t )1/θ for γi = c i (0)/q(0)1/θ and i = A , P (28)


and substitute this result into (23c) and (23d). The system of dynamic equations (23a)–(23e) can now be transformed into
a system of lower dimension expressed as
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
q̇ Fq (x) (r − v q̃ )q
⎜ Ż ⎟ ⎜ F Z P (x) ⎟ ⎜ (r − v ) Z P − γ P q1/θ ⎟
⎜ P⎟ ⎜ ⎟ ⎜



⎜ ⎟ = ⎜ ZA ⎟ = ⎜ (29)
⎝ Ż A ⎠ ⎝ F (x) ⎠ ⎝ (r − v ) Z A + q( 1−αα qV )σ Ω(r )( Z A + Z P ) − γ A q1/θ ⎟
r

Ṡ F S (x) −( 1−αα rqV )σ Ω(r )( Z P + Z A ) − vS S
where x ≡ (q, Z P , Z A , S ), r = ψ(q) and r V = R V (r + δ K , r + δ H ) is defined in (11). We know from Li et al. (2003) that system
(29) preserves the dynamic properties of (23a)–(23e).

Proposition 3 (Convergence). Given positive initial conditions ( Z P (0), Z A (0), S (0)) in the neighborhood of T, the state variables
( Z P (t ), Z A (t ), S (t )) converge to a point on the plane of steady states (27). That is, T is saddle path stable.

Proof. The eigenvalues of the Jacobian matrix of the differential system (29) evaluated at a steady state equal (see Ap-
pendix C)

1 − α r̄ V
ε1 = −Ω(r̄ ) , ε2 = r̄ − v = − vS , ε3 = −ε1 − vS , ε4 = − vS . (30)
α q̄φ

Since v
S < 0, then ε2 , ε3 and ε4 are positive while ε1 is negative.15 2

This leads to the following claim.

Claim 1.

(a) Different initial conditions Z P (0), Z A (0) and S (0) can asymptotically lead to a different steady state; and
(b) for each steady state ( Z P , Z A , S  ) there is a neighborhood containing a one-dimensional manifold convergent to a point on the
plane of steady states T at ( Z P , Z A , S  ) such that, for initial conditions ( Z P (0), Z A (0), S  (0)) of this one-dimensional manifold,
the equilibrium path converges to ( Z P , Z A , S  ).

The shaded plane in Fig. 3 portrays set T. We display different initial conditions leading to a different steady state
(Claim 1(a)). On the other hand, initial conditions lying in a given trajectory lead to the same steady state (Claim 1(b)).

14 (β r̄ H )μ ((1−β)r̄ K )μ
Steady-state K i and H i equal μ Z and μ Z , respectively.
β μ r H +(1−β)μ r̄ K i β μ r̄ H +(1−β)μ r̄ K i
μ μ
15
Note that the Jacobian matrix of system (23a)–(23e) evaluated at a steady state has two zero eigenvalues. We know from Li et al. (2003) that the
reduced system (29), a system of lower dimension, keeps the eigenvalues of the Jacobian of system (23a)–(23e) at any equilibrium whose generalized
eigenspaces are not mapped to the null space by the Jacobian of the reduced system. Moreover, the reduced system maintains “the local stability properties
of a dynamical system at an equilibrium” (Li et al., 2003, p. 1626).
80 B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93

Fig. 3. Illustration of Claim 1 (in normalized variables).

Since the set of steady states is a plane of dimension two and the Jacobian matrix of the reduced system (29) has a
single negative eigenvalue, it follows from Li et al. that Claim 1 holds. Provided initial conditions are in the neighborhood of
a steady state, it also follows that for each steady state ( Z P , Z A , S  ) there is a unique path with different initial conditions
converging to it. Note that a steady state is not stable in the sense that disturbing a steady state by providing more physical
or human capital to either economy, or more exhaustible resources to country A, countries may converge to a different
steady state.

4. Dynamic behavior

Since we are interested in providing analytical insights into the conditions causing a resource-abundant country to grow
more slowly than a resource-poor country, we now study the near steady-state variables’ dynamic behavior under the
simplifying assumption that δ ≡ δ K = δ H . This assumption is relaxed in our numerical simulations. With δ K = δ H , it follows
that H i (t ) = (1 − β)μ /(β μ + (1 − β)μ ) Z i (t ), K i (t ) = β μ /(β μ + (1 − β)μ ) Z (t )i and r H = r K . Denote normalized and non-
normalized world aggregate capital by Z and  Z , respectively (i.e. Z = Z P + Z A and  Z = ZP + Z A ).

Proposition 4. If ( Z P (0), Z A (0), S (0)) is in the neighborhood of a steady state and ( Z P (0) + Z A (0))/ S (0) = Z (0)/ S (0) < {>} Γ (as
defined in (25)), then Z , H P , K P , q, c A and c P converge (transitionally increase {decrease}) to Z , H P , K P , q̄, c A and c P , respectively,
where

Z (0) < {>} Z , H P (0) < {>} H P , K P (0) < {>} K P , q(0) < {>} q̄ and c i (0) < {>} c i .

Proof. Let J denote the Jacobian matrix of system (29) evaluated at a steady state. The negative eigenvalue of J equals ε1 ,
which is defined in (30). Let the eigenvector of matrix J associated with ε1 be denoted by u  = (u q , u Z P , u Z A , u S ) so that u
satisfies (J − ε1 I)u = 0. Here, I is a 4 × 4 identity matrix. Note that, in terms of normalized aggregate world capital, the set
of steady states is a ray from the origin given by Z / S = Γ , which is illustrated in Fig. 4. In addition, in ( S , Z ) space and
near steady states, the ratio u Z /u S ≡ (u Z P + u Z A )/u S equals the slope of the ( S , Z ) trajectory towards a steady state which
equals (using (D.1) from Appendix D)
φ σ σ
uZ 1 θ Λα 1/φ q̄ α − vS − ε1
=− D (31)
uS M σ r̄ V r̄ V 1 − α Ω
where
φ σ
1 ΩΛα 1/φ
M =θ − for φ = 0, and M =θ −α for φ → 0. (32)
σ δ+η
1
Also, Ω ≡ Ω(r ) = (β μ + (1 − β)μ ) χμ B for μ = 1 and Ω ≡ β β (1 − β)1−β B for μ → 1 (Ω(r ) is constant for all t when
δ H = δ K ). Finally, D = (σ ( Λαr V1/φ )φ σ v ε+1ε + θ1 ) > 0 and − vS − ε1 > 0. Similarly for q and Z P , near steady states, the slope of
 1
S
the ( Z P , q) trajectory towards a steady state equals (see (D.1) from Appendix D)

uq ε1 + vS q̄
= −1
> 0. (33)
uZP ε1 + v
Sθ ZP
Since v S̃ < 0 and ε1 < 0, uq /u Z P is positive, and the sign of u Z /u S depends on the sign of M.
B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93 81

Fig. 4. Dynamic behavior of Z and S.

Assume the initial values of S and Z are near the ray of steady states where Z (0)/ S (0) < Γ, as illustrated in Fig. 4.
When M > 0, the slope of the ( S , Z ) trajectory towards a steady state is negative (u Z /u S < 0), thus Z and S must move in
opposite directions. For the system to converge to a point on the ray, Z must increase while S must decline. Using (D.1) and
(31), it can be verified that the slope of the ( Z , q) trajectory towards a steady state is positive. Thus, since Z increases, q
increases. Since u q /u Z P > 0 and c i for i = A , P follow the dynamic behavior of q, then H P , K P , c P and c A increase toward
their steady-state levels. The case for Z (0)/ S (0) > Γ can be proven in the same manner. The cases for M  0 are provided
in Appendix D. 2

Claim 2. Suppose the two economies have either reached steady state ( Z old old
P ,ZA , S
old
) or are in transition towards ( Z old old
P ,ZA ,S
old
)
with Z (t )/ S (t ) < Γ and resources are unexpectedly discovered {lost} (refer to ( Z old old
P ,ZA ,S
old
) as the old steady state). Then, in near
steady states

(i) country P reaches a new steady-state level Z P that is larger {smaller} than Z old
P ;
(ii) the new steady-state level c P is larger {smaller} than the old steady-state value;
(iii) c A and c P transitionally increase {decrease} towards a new steady state;
(iv) the normalized price of the resource instantaneously declines {increases} and then transitionally increases {decreases} towards its
steady-state value.

Note that if an unexpected resource discovery is made, the aggregate capital–resource ratio Z / S is less than Γ . Thus
Claim 2 follows directly from Proposition 4.
Claim 2 indicates that, from a steady state, a resource discovery causes Z P to increase. The intuition for this result is
as follows. Following a resource discovery, the price of the resource instantaneously declines. To maintain a least cost-input
combination, the resource-poor country increases its demand of the exhaustible resource. Resource depletion induces the
price of the resource to transitionally increase, causing Z P to transitionally increase.
With δ H = δ K , country- P ’s normalized factor income at the old steady state equals r̄ K H old
P + r K K P = r̄ K Z P , since from
old old

a steady state, a resource discovery causes Z P to transitionally increase, this implies that country P converges to a new
steady-state level of normalized income r̄ K Z P that is larger than its old steady-state level (r̄ K Z P > r̄ K Z old
P ). In addition,
since the steady-state value of normalized imports of country- P equals q̄N P = q̄( 1− α r̄ V σ
α q̄ ) Ω Z P , the value of normalized
exports of country A converges to a new steady state larger than the old steady-state level.
We now focus on the dynamic behavior of Z A . Let ω V and ω N denote the shares of output paid to composite input

V i and flow resource  V A /
N i , respectively. That is, ω V ≡ r̃ V  V P /
Y A = r̃ V  Y P and ω N ≡ (1 − ω V ). At the steady state, ω V =
1/φ
( ΩΛα φσ
δ+η ) . Let

 2  
1 θ −1
λ = − 1 + E
NN (−ε1 − vS ) − ε1 (34)
ωV θ
∂
N q̃
where ε1 and vS are negative (see (30)). E N N ≡ |
N r =r̄
∂ q̃ 
= −σ ω V is the partial own price elasticity of demand of the
16
exhaustible resource evaluated at the steady state. With elasticities of substitution σ smaller than one, E N N is greater
than −1.

16
To obtain E 
N

N , derive the conditional demand for N from the cost minimization problem of the final good sector, and take its partial derivative with
respect to q̃.
82 B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93

Proposition 5. If the ratio Z (0)/ S (0) < Γ , then near steady states H A and K A

(i) increase if λ  0 and


(ii) decline (increase) if λ < 0, and the ratio Z P (0)/ Z A (0) of aggregate capital of country P to the aggregate capital of country A is
sufficiently large (small).

Proof. In near steady states, the slope of the ( Z A , q) trajectory equals (see Appendix D)

σ 
uq (−2ε1 − vS )ε3 q̄ α Z−
A
1
= (35)
uZA Ω r̄ V 1 − α λ Z P / Z A + D ε3 ε3 /(−ε1 )

where ε1 < 0 and ε3 > 0 are eigenvalues defined in (30). Given Proposition 4, when Z (0)/ S (0) < Γ , the price of the
resource transitionally increases. If λ is non-negative, u q /u Z A is positive; thus the aggregate capital of country A ( Z A )
transitionally increases towards a new steady state. If λ is negative and Z P (0)/ Z A (0) ≈ Z P / Z A is sufficiently large, u q /u Z A
is negative, since q transitionally increases Z A transitionally declines. 2

Claim 3. Given a steady-state equilibrium, an unexpected resource discovery causes the capital of the resource-abundant country to

(i) increase near steady states if λ  0 and


(ii) decline (increase) near steady states if λ < 0, and the ratio of aggregate capital of country P to aggregate capital of country A
( Z P (0)/ Z A (0) ≈ Z P / Z A ) is sufficiently large (small).

This result follows directly from Proposition 5. The sign of λ depends upon the steady-state partial own price elasticity
of demand of the exhaustible resource (E  N ), the steady-state share of output paid to the composite input (ω V ), and the
N
inverse of the elasticity of intertemporal substitution θ . An inelastic resource demand (E  N approaching zero) will tend to
N
make λ negative. For θ = 1, the sign of λ is determined by E  NN and ω V .
The intuition for Claim 3 is as follows. A resource discovery causes an instantaneous decrease in its price, and thus
country P increases its imports of the resource. The increase in the quantity imported is partially determined by elasticity
ENN = −σ ω V . As the resource is depleted, q increases, causing H P and K P to transitionally increase and converge to
a steady-state level larger than the steady-state level achieved in the absence of the discovery. Despite the transitional
increase in q, depending upon the factor substitution possibilities between  N P and composite input  V P , country P imports
relatively large amounts of the resource. Accordingly, the value of exports of country A transitionally increase, warranting a
future source of income for this country, thereby decreasing the incentive to save relative to country P . If consumers prefer
a smooth consumption path (measured by θ ), a secured future source of income in country A fosters consumption, relative
to that of country P , at the expense of own investments.
If the ratio Z P (0)/ Z A (0) ≈ Z P / Z A is sufficiently small, a positive resource shock causes Z A to transitionally increase near
steady states (even if λ is negative). This outcome is obtained because the aggregate capital stock of country P , Z P , is not
large enough to generate large gains from trade. Thus, given incentives to smooth consumption, the households of country
A increase H A and K A in the short run at the expense of consumption (numerical simulations in the next section show the
effects of equal and different initial endowments of Z P and Z A ).
We have now established the near steady-state conditions whereby a positive resource shock will cause a decline of the
aggregate capital stock of country A. We have shown that declines in H A and K A due to a resource discovery are associated
with a relatively inelastic resource demand. The partial own price elasticity of demand for  N equals −σ ω V (see Allen, 1964,
p. 373).17 As pointed out by Rotemberg and Woodford (1996, p. 557), the elasticity of substitution of energy and value
added must be less than unity in order to be consistent with the observation that the share of energy as a fraction of total
cost increases with a rise in energy prices. The cost share of the resource in our model equals ω N ≡ q̃ Y A = q̃
N A / N P /
YP =
(Λ(1 − α )1/φ e ηt /q̃)φ σ . Thus, ω N is increasing in q̃ only when φ < 0 ⇒ 0 < σ < 1. Therefore, for our model to lead to an
increase in the share of output paid to  N from a rise in its price q̃, as observed in the data (see Fig. 5), σ must be less than
one, and thus σ ω V < 1. This in turn implies that E  NN = −σ ω V > −1.
Table 1 presents estimates of short-run and long-run demand elasticities of raw inputs, various energy resources, elec-
tricity and aggregate energy. Aggregate energy includes crude oil, natural gas, coal, petroleum products and electricity.
These estimates indicate that the larger the level of aggregation (energy) or processing level (electricity), the larger
(in absolute terms) the own price elasticity of demand. Since we think of exhaustible resources as a raw commodity, the
elasticities of demand of crude oil or natural gas seem to be better suited for our analysis. In this case, elasticities of demand
of raw commodities tend to be larger than −1.0, which helps to support the possibility of a negative λ.
B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93 83

Fig. 5. U.S. share of GDP paid to crude oil and real price of crude oil. (a) The share of GDP paid to crude oil equals the (nominal) value of U.S. crude oil
production plus crude oil net imports (EIA, 2009a, pp. 71 and 85) on (nominal) GDP (Council of Economic Advisors, 2010). (b) EIA (2009a, p. 169).

Table 1
Own price elasticity of demand of various energy sources.

Input Short-run Long-run


Gately et al. (1986, Tb.a 1) M-Cb Crude oil −0.05 to −0.22 −0.34 to −1.09
Griffin (1992, p. 34) M-C Crude oil −0.371
Hamilton (2008, Tb. 3) M-C Crude oil −0.05 to −0.07 −0.21 to −0.30
Bernstein and Griffin (2005, Tabs. D.8–D.9) U.S. Nat. gas −0.08 to −0.34 −0.13 to −0.63
Hamilton (2008, Tb. 3) M-C Gasoline −0.25 to −0.34 −0.58 to −0.86
Griffin (1992, p. 39) M-C Petr. gds. −0.23 −1.04
Bernstein and Griffin (2005, Tabs. D.6–D.7) U.S. Electricity −0.21 to −0.76 −0.50 to −3.11
Berndt and Wood (1975, Tb. 5) U.S. Energy −0.47
Pindyck and Rotemberg (1983, Tb. 2) U.S. Energy −0.36 −0.99
a
Tb. stands for Table.
b
M-C stands for Multi-country.

4.1. Income growth performance

We now focus on the relative GDP growth performance of these two economies. Denote the rate of growth of country-i’s
factor income (GDP i ) by g i ≡ GḊP i /GDP i .

Proposition 6. Let

π = (1 − σ )(ε1 + vS )(2ε1 + vS )/(Ω r̄ V ) + λ (36)


where (ε1 + v
S )(2ε1 + v
S ) > 0. If Z (0)/ S (0) < Γ, in near steady states, g P > (<){=} g A if and only if π < (>){=}0.

Proof. The difference in GDP growth rates, g P − g A , equals


T 
GDP T GḊP P GḊP
gP − gA = − (37)
GDP A GDP P GDP T
where GDP T ≡ GDP P + GDP A equals world GDP. Totally differentiating the right-hand side with respect to q using (D.1) and
simplifying gives the near steady-state slope of the (q, g P − g A ) trajectory18

d( g P − g A ) πε12 (Λα 1/φ /r̄ V )φ σ Z P / Z A
= 1+ . (38)
dq q̄(ε1 + v
S )(2ε1 + v
S) 1 + Λφ σ q̄1−σ (1 − α )σ Z P / Z A
At a steady state, g P − g A = 0. Since Z (0)/ S (0) < Γ , then q transitionally increases towards q̄ (refer to Proposition 4). Near
d( g P − g A )
steady states, dq
< 0 if π is negative. In this case, g P − g A transitionally declines and converges towards zero from
above (g P > g A ). The other cases can be proven in the same manner. 2

17
Note that E 
NN is a partial elasticity only. We are able to derive what we think of as the “total” own price elasticity of demand for N near steady states.
dN q̄ ε3
It is given by dq N
= (2ε1 + vS )ε1 [λ + (1 + E 
NN /ω V )(−2ε1 − v
2
S )].
T
18
To obtain this result, set GDP P = r K ( H P + K P ) = Ω r V Z P and GDP T = ( αrΛV φ )σ Ω Z . Next compute G ḊP P
and GḊP
to obtain g P − g A . Make appropriate
GDP P GDP T
substitutions from (29) into g P − g A and finally take the total derivative of g P − g A with respect to q.
84 B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93

Proposition 6 exploits knowledge about the dynamic behavior of q to determine the behavior of g P − g A , indicating
which country’s factor income grows faster than the other. A key parameter determining the sign of π is the own price
elasticity of demand for N, which determines the sign of λ, (see (34)).
If π < 0 holds, we claim the resource curse condition is satisfied.

Claim 4. If the resource curse condition is satisfied, in near steady states a larger ratio Z P (0)/ Z A (0) ≈ Z P / Z A generates a larger gap
in growth rates ( g P − g A ).

For π < 0, this result is obtained because the right-hand term in (38) increases with an increase in the ratio Z P / Z A .
Note that, near steady states, ratio Z P (0)/ Z A (0) ≈ Z P / Z A negatively influences capital accumulation in country A (refer to
Claim 3) and hence, it dampens the rate of growth of GDP A .

Claim 5. Given a steady state, if the resource curse condition is (is not) satisfied, then a positive shock to the exhaustible resource S
causes GDP P to grow faster (slower) than GDP A .

Claims 4 and 5 follow directly from the proof of Proposition 6.


Thus, a positive shock to S causes an instantaneous decline in its price q, which fosters capital accumulation in coun-
try P , but slows capital accumulation in country A, thus, negatively affecting the GDP growth rate of country A.
The resource curse condition depends upon several parameters. The key parameters are the partial own price elasticity
of demand for N, the share of output paid to composite  V i and the willingness to smooth consumption, θ , all of which are
embodied in λ. While initial conditions do not influence the resource curse condition, the ratio Z P (0)/ Z A (0) influences the
magnitude of the difference in growth rates ( g P − g A ).
There is a large body of economic literature linking the RC to rent-seeking activities and institutions (see Tornell and
Lane, 1995) and capital endowments to differences in the “take-off” time of country growth experiences. Differences in ini-
tial capital endowments almost surely reflect (at least to some extent) differences in policies or institutions that affect both
human and physical capital accumulation (see Acemoglu and Dell, 2009).19 Thus, our model predicts that policies which
discouraged capital accumulation prior to the time of discovery may increase the relative GDP growth underperformance
of resource-abundant countries. This interpretation is an oversimplification since it presumes that the presence of resource
rents does not affect the institutional environment or does not enhance the level of rent-seeking activities.
In our simulations, we study the case in which Z P (0) and Z A (0) are equal and study the effects of different initial
endowments of the resource. Thus, in this model, the sign of π describes the qualitative behavior of relative GDP growth
( g P − g A ). Note that λ < 0 does not necessarily imply that Z A declines in transition. We know from Proposition 6 that Z A
increases in transition for λ < 0 and a sufficiently small Z P (0)/ Z A (0) ratio. For λ < 0, therefore, the capital of country A
can increase in transition (as well as its GDP), but for π < 0, country- A’s income grows less rapidly than that of country P .
Our results also suggest that resource endowments, including discoveries (Claim 5), are important in explaining the GDP
growth gap between resource-poor and resource-abundant countries, however this is often omitted in growth regressions.
The model also suggests that growth regressions should control for country elasticities of intertemporal substitution and
measures of exhaustible resources’ demand elasticities. In addition, the model predicts that the economic conditions at the
point in time when resource discoveries take place may play a role in explaining the intensity of the RC. These conditions
could be measured by capital stocks at the point when resource discoveries occur. If these conditions are not controlled for
or not identified in regression models, it is not surprising that conflicting results are obtained.
In the next section, we investigate whether this simple model can generate the differences in GDP growth rates across
countries that we observe in the data.

5. Quantitative analysis

We next draw upon Proposition 6 to investigate empirically the value of π (defined in (36)) under reasonable parameter
values (Table 4). Some of these values are based upon the literature others are calibrated.
Following the business cycles literature, we set the elasticity of factor substitution between human and physical capital
μ equal to unity, and set the degree to which the economy is physical capital intensive to β = 0.35. The scale parameter of
the composite input (refer to Eq. (1)) is set to unity, B = 1.
We estimate the share of output paid to the resource (ω N ) as the ratio of mining and quarrying intermediate expenditure
of all industries (IMQ) on value added plus IMQ using the input–output tables of various countries reported in Table 2.
Using OECD data (1995, 2002, 2006) ω N ranges from 0.015 for France in 1995 to 0.085 for China in 2000. Using OECD data
(1995), ω N for the U.S. equals 0.065 in 1982 (a year for which the U.S. energy expenditure to GDP ratio is relatively large,
see Atkeson and Kehoe, 1999), while BEA (2008) data suggests a lower estimate of 0.041 for 2006. To generate the cost
shares of exhaustible resources we observe in the data, our base line calibration sets the steady-state value of ω N equal to
ω N = 0.085, and thus ω V = 1 − ω N = 0.915. We test the sensitivity of π to values of ω N ranging from 0.04 to 0.15.

19
We thank an anonymous referee for pointing out this result.
B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93 85

Table 2
Ratio of mining and quarrying intermediate expenditure (IMQ) to value added plus IMQ for selected.a

1990 1992 1995 1996 1997 1998 2000 2002 2006


Brazil 0.021 0.038
Canada 0.048 0.045
China 0.083 0.085 0.080
Czech Rep. 0.080
Denmark 0.018 0.019
Finland 0.031
France 0.034 0.015
Germany 0.027 0.020
Greece 0.024c
Hungary 0.040
Italy 0.051c 0.020
Japan 0.022 0.019
Korea 0.042
Netherlands 0.070c 0.038
Norway 0.031
Poland 0.068
Spain 0.025
US 0.065c 0.041b

Authors’ calculations using: a OECD (1995, 2002, 2006) data, and b BEA (2008) data. c The ratios concerning Italy, the Netherlands, the US and Greece are for
the years 1985, 1986, 1982, and 1994, respectively.

We set ρ = 0.03 a parameter value often used elsewhere. We set the resource-saving technological change growth rate
η equal to 0.05, and test the sensitivity of π to this value.20 Since the long-run return on assets r̄ equals η, we have
r = 0.05.21
To compute π , we set δ ≡ δ H = δ K = 0.03. This value takes into consideration that the depreciation rate of human capital
(δ H ) may be smaller than that of physical capital (δ K ). We relax this assumption in our numerical simulations, and test the
β 1−β
sensitivity of π to different values of δ . Given the mentioned parameter values, r̄ V = r̄ K r̄ H /(β β (1 − β)1−β B ) = 0.16 and
r̄ K = r̄ H = 0.08.
We think of exhaustible resources as a “raw” commodity (e.g. crude oil) and not as a processed commodity (e.g. elec-
tricity or energy). Based on the elasticities listed on Table 1 and in view of Bodenstein et al. findings (2007, pp. 48–49)
we set the partial own price elasticity of demand of the exhaustible resource at the steady state equal to −0.5. Setting
E N = −σ ω V = −0.5, we obtain σ = 0.5/0.915 = 0.55 and hence φ = −0.83.
N
The first-order conditions of the final good sector of country i imply r̃ V = α Λφ · ( Y i /
V i )1−φ , thus Λα 1/φ =
V i /
r̃ V (r̃ V 
1/(φ σ ) 1/ σ
Y i )1/(φ σ ) = r̃ V ω V and Λ = r̃ V (ω V
/α )1/φ . The degree to which output is resource intensive (1 − α ) is set
1/(φ σ ) 1/ σ
to 0.085. We thus set Λα = r̄ V ω
1/φ
= 0.19 and Λ = r̄ V (ω V /α )1/φ = 0.17.
V
In Fig. 6 we plot π for different parameter values and for different values of θ on the horizontal axes. The dashed
line uses the benchmark parameters given in Table 4, except that we set δ H = δ K = 0.03. The remaining plots show the
sensitivity of π , in (θ, π ) space, to various values of ω N , (1 − α ), δ, ρ , η and to price elasticity E 
NN as indicated. Changes
in these parameters affect the values of others parameters, as indicated by the above calculations. Fig. 6.1 through Fig. 6.5
show that the resource-curse condition (π < 0) holds over a range of inter-temporal preferences θ  1, for plausible values
of: the steady-state share of output paid to the resource ω N , the degree to which output is resource intensive (1 − α ),
the rate of time preference ρ , the depreciation rate δ and the rate of factor productivity growth η of N . As shown by
Proposition 5, the existence of the RC is sensitive to elasticity E NN . Fig. 6.6 shows that the more inelastic the price elasticity
is, the “stronger” the curse. With an elasticity E NN = −0.95, for π to be negative it requires that θ is greater than 3.25. In
this case, however, we have σ = 1.04, which contradicts the evidence that the share of output paid to exhaustible resources
increases when resource prices rise.
Thus, the condition consistent with the GDP of an exhaustible resource-poor country growing faster than that of an
exhaustible resource-abundant country seems plausible. The next question is: can the model display the differences in GDP
growth rates that we observe in the data?

20
Note that if the relation between human capital, physical capital and the exhaustible resource were Cobb–Douglas (μ = 1 and σ = 1), then we have
 β 1−β α ηt  1−α
Y i = Λ B α (
Ki 
Hi ) (e N i ) where η(1 − α ) equals the rate of growth of total factor productivity (TFP). Estimates of TFP growth vary substantially over
time and across countries. Jorgenson et al. (2000) include the effects of energy and materials in their estimates of TFP in addition to labor and capital. They
report a TFP growth for the U.S. over the 1958–1996 period of 0.0048. In our model with μ = σ = 1, setting η(1 − α ) = 0.0048 and ω N = 1 − α = 0.085
gives η = 0.0048/0.085 = 0.056.
21
Mulligan (2002) uses U.S. national account data from 1931 to 1997 to estimate the capital rental rate. He finds the after-tax capital rental rate net of
depreciation to be approximately 0.05.
86 B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93

Fig. 6. π (resource–curse condition π < 0).

We now relax the assumption of identical depreciation rates of capital. We set the depreciation rate of physical capital
δ K equal to 0.05 (a parameter often used elsewhere). The depreciation rate of human capital δ H is set equal to 0.025. As a
compromise, and in view of Guvenen’s (2006) findings, we set θ = 1.22
We use the Maddison (2008) data set to compute average GDP per-capita growth rates over a 56-year period for resource-
abundant and resource-scarce countries. A country is considered resource-abundant if it was a net exporter of fuel, ores and
metals in 2000, computed using the World Bank’s World Development Indicators (2008). In Table 3, we list the 20 countries
with the largest and the smallest per-capita-net exports of fuel, ores and metals for which we also have data on GDP growth
from the Maddison (2008) data set.23
Qatar, the United Arab Emirates and Norway are the three countries with the largest net exports of exhaustible resources
in per-capita terms, while Singapore, Belgium and South Korea bring up the rear. In level terms, the U.S. has the largest trade
deficit in exhaustible resources, while in per-capita terms it ranks tenth. We consider 141 countries from the Maddison
data set for which per-capita GDP data are available for the years 1950–2008.24 Dividing the set of countries into two
groups leaves 42 countries with an exhaustible-resource trade surplus and 99 countries with an exhaustible-resource trade
deficit. The average GDP growth rate among countries with an exhaustible-resource trade surplus from 1950 to 2008 is
approximately 1.84 percent, while for those with a trade deficit the average rate is 2.17 percent, a difference of approx. 0.3
percentage points. The Maddison data set for all the countries and years considered leads to an average yearly per-capita

22
Estimates of the elasticity of intertemporal substitution (EIS) vary considerably from study to study. Using U.S. postwar consumption quarterly data and
the real after-tax yield on U.S. treasury bills, Hall (1988) finds that the EIS is unlikely to be greater than 0.1 (θ = 10). Vissing-Jørgensen (2002) differentiates
between asset holders and non-asset holders and estimates that the EIS ranges from 0.3 to 0.4 for stockholders and from 0.8 to 1 for bondholders. Mulligan
(2002) uses tax policy changes to estimate the EIS. Using U.S. national accounts data from 1931 to 1997, he reports estimates of EIS ranging from 0.49 to
2.05.
23
In Table 3, Iraq and Libya are omitted due to a lack of data availability.
24
We consider the former USSR and Czechoslovakia instead of their successor Republics in order to extend the data set over 58 years instead of only 18
years.
B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93 87

Table 3
Per-capita-net exports of fuel, ores and metals in 2000 (US$).

Qatar 17,004 Singapore −867


United Arab Emirates 14,524 Belgium −814
Norway 8,678 South Korea −766
Kuwait 8,170 Japan −712
Bahrain 5,174 Hong Kong −644
Oman 3,828 Israel −631
Saudi Arabia 3,392 Ireland −577
Gabon 1,828 Finland −575
Trinidad & Tobago 1,279 Switzerland −502
Venezuela 1,190 United States −483
Canada 986 France −469
Australia 958 Germany −467
Algeria 696 Italy −450
Iran 379 Sweden −442
Chile 348 Spain −429
USSR 217 Portugal −403
Syria 204 Austria −394
Yemen 201 Lebanon −314
Ecuador 171 Czechoslovakia −272
Malaysia 169 Greece −260
Source: Authors’ calculations using World Bank (2008) data.

Table 4
Baseline parameter values.

Rate of physical capital depreciation δK 0.05


Rate of human capital depreciation δH 0.025
Long-run growth rate of capital and income v 0.02
Share of physical capital in composite  Vi β 0.35
Composite capital input  V i shift parameter B 1.00
Elasticity of substitution between human capital—physical capital μ 1.00
Long-run share of output paid to human and physical ωV 0.915
Human and physical capital intensity α 0.915
Long-run partial own price elasticity of resource demand E
NN −0.50
Elasticity of substitution between resources and composite capital σ 0.55
Output shift parameter Λ 0.17
Resource-saving technological change growth rate η 0.05
Rate of time preference ρ 0.03
Elasticity of intertemporal substitution θ −1 1.00

Table 5
Simulations.

Simulation 1 vs simulation 2 Simulation 2 vs simulation 3 Simulation 4


S 2 (0) > S 1 (0) S 3 (0) > S 2 S 4 (0) = S 2 (0)
Z 1A (0) = Z 1P (0) Z 3A (0) = Z 2A Z 2A (0) = Z 2P (0)
Z 2A (0) = Z 2P (0) Z 3P (0) = Z 2P Z 4A (0) < Z 4P (0)

GDP growth rate of 2 percent. The values of η , ρ and θ reported in Table 4 are consistent with this rate, v = (η − ρ )/θ =
0.02.
Fig. 7 shows simulations using parameter values reported in Table 4. The base simulation is denoted simulation 1. Super-
scribed variables refer to a particular simulation. For example, S 1 (0) denotes the initial stock of the exhaustible resource of
the base simulation. Four simulations are performed for various initial stocks of capital and exhaustible resource. Differences
across simulations are given in Table 5.
The first row of charts in Fig. 7 shows the evolution of ratio Z A / Z P for each of the four simulations (as described in
Table 5). The solid line in chart 7.1 denotes simulation 1 for which Z 1A (0)/ Z 1P (0) = 1. The dashed line in chart 7.1 shows
the effect of increasing the stock of the exhaustible resource in the initial period from S 1 (0) to S 2 (0). Chart 7.2 “starts” the
model from the steady state of simulation 2 by a once and for all increase in the near steady-state level of the remaining
exhaustible resource stock to level S 3 (0) and illustrates the effect on the ratio Z A / Z P . Chart 7.3 contrasts the evolution
of the ratio Z A / Z P from simulation 2 (the dashed line), where Z 2A (0) = Z 2P (0), to that of simulation 4 where country- P ’s
initial stock of capital exceeds that of country A , i.e. Z 4A (0) < Z 4P (0), as shown by the dotted line.
The second row of charts shows the evolution of the aggregate capital stock of country A ( Z A ) for each of the simulations
described above. In simulations 1–3, Z A increases in transition towards its respective steady-state value (charts 7.4–7.5).
Thus, an increase in the initial endowment of the resource promotes the growth of Z A . However, since the ratio Z A / Z P
88 B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93

Z A /
Fig. 7. Simulations 1–4,  Z P = Z A / Z P , Z A and g P − g A .

decreases in transition (charts 7.1–7.2), Z A must be increasing at a lower rate than Z P . In simulation 4, with Z 4A (0) < Z 4P (0)
but with an initial endowment of the exhaustible resource equal to that of simulation 2 (where Z 2A (0) = Z 2P (0)), the long-run
level of Z A is only a fraction of that of simulation 2 (chart 7.6).
Charts 7.1–7.6 confirm Claim 1, that is, different initial conditions lead to a different steady state. Furthermore, Claims 2
and 3 are verified: given a steady-state equilibrium, an unexpected resource discovery causes Z P and Z A to increase (charts
7.2 and 7.5). However, Z P increases at a higher rate than Z A . Simulation 4 (chart 7.6) also verifies Claim 3 that a large
ratio Z P (0)/ Z A (0) negatively affects the accumulation of Z A .
The final row of charts in Fig. 7 shows the difference in GDP growth rates associated with each simulation (g P − g A ), and
lends support to Proposition 6, and Claims 4 and 5. Given the same level of aggregate capital stocks ( Z A (0) = Z P (0)), gap
( g P − g A ) is positive and it becomes larger with a rise in the initial stock of the exhaustible resource, chart 7.7. The growth
gap ( g P − g A ) associated with simulation 3 is larger than that of simulation 2, due to an unexpected and positive shock to
the exhaustible resource stock from S 2 to S 3 (0) (Claim 5, chart 7.8). Chart 7.9 illustrates that the gap ( g P − g A ) found in
simulation 4 is larger than the gap found in simulation 2, due to differences in initial capital stocks (Claim 4). Effectively,
country A never catches up with country P , and converges to a lower steady-state value of GDP, although their long-run
growth rates converge. Simulations 1, 2 and 3 yield a short-run difference in GDP growth rates ( g P − g A ) of at most 0.07%.
This difference is smaller than the average difference revealed by the data (0.3%).
Fig. 8 shows the consumption of the resource-abundant country (c A ), the share of output paid to the resource (ω N )
and resource export revenues (qN P ) profiles for the same simulations shown in Fig. 7. The values shown within the charts
are steady-state values. For brevity, we do not comment on all of the results, but merely show them for the sake of
completeness.
As shown in Fig. 8, the simulations confirm our analytical results that, starting with initial conditions, ( Z P (0) +
Z A (0))/ S (0) < Γ , c A transitionally grows towards a steady state. The effect on c A of increasing the stock of the resource
B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93 89

Fig. 8. Simulations 1–4, c A , qN /(Y A + Y P ) and qN P .

in the initial period (chart 8.1) causes c 2A to be below its base solution values c 1A in the short run (for around 20 years).
The increase in the resource endowment causes an instantaneous decline in q and qN P (chart 8.7) in the short run, but not
in the long run. The productivity of composite capital increases relative to the base solution, which provides incentives for
both countries to increase savings. Due to an own inelastic price effect, the decline in q and the increase in N lead initially
to a decline in the share of output paid to the flow resource (chart 8.4), but converging to the share of simulation 1 in the
long run. Share ω N ranges from 0.01 to 0.085, chart 8.4, which is consistent with the shares observed in the data. As the
resource is depleted, q transitionally increases towards its steady-state value. Since the long-run aggregate capital stock of
country P ( Z P ) in simulation 2 exceeds the level of simulation 1, remuneration qN P to country A eventually exceeds base
solution levels as well (chart 8.7). Relative to country A, the discounted utility of country- P ’ households is maximized by
foregoing consumption, and increasing capital stocks, which result in an increase in output needed to remunerate country A
for imports qN P , thus causing trade to be balanced. Country- A households’ incentive to save is “dampened” by the increased
flow of income qN P that continues to grow into the long run.

6. Conclusions

The apparent paradox that the income of countries with ample natural resources grows less rapidly than that of natural
resource-scarce countries has led to a number of papers with conflicting empirical findings, suggesting that more theoretical
work may determine whether plausible structural features can help explain this paradox. We develop a continuous-time,
infinite-horizon two-country model of trade in which countries are identical, except that one country is endowed with de-
posits of an exhaustible resource (such as petroleum). We find a condition not considered in previous work that may cause
an exhaustible resource-abundant country’s GDP to grow more slowly than that of an exhaustible resource-scarce country.
We call this condition the resource curse condition. This condition depends purely on preference and technological param-
eters. The key preference parameter is the willingness to smooth consumption, measured by the intertemporal elasticity
90 B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93

of substitution; the key technological parameter is the degree of inelasticity of the derived demand for the flow resource
mined from an exhaustible resource. Other parameters matter, but they do not reverse the condition.
The more inelastic the demand for exhaustible resources, the greater the resource curse. An inelastic demand for the flow
resource generates a growing revenue stream which, one may think, favors GDP growth in the resource-abundant country.
Our model predicts the opposite. An intuitive explanation is the following. In the process of economic growth, the price
of the flow resource increases, and an inelastic exhaustible resource demand causes the total revenue remunerated to the
resource-abundant country to increase. The growing income stream from exports of the resource induces households in the
resource-abundant country to invest relatively less than those of the other country, at the expense of future income growth.
In addition, the greater consumers’ preferences to smooth consumption are, the greater their disincentives to invest, thus
exacerbating the curse.
Our computations suggest that the resource curse condition holds for plausible parameter values. We also simulate the
model to test whether it is capable of replicating the gap in GDP growth rates that we observe in the data. Our benchmark
simulation assumes that both countries have the same initial endowment of human and physical capital as a way to abstract
from corruption and institutional effects. The numerical exercises confirm our analytical results and show that the model
can account for about one fourth of the average GDP growth gap between resource-poor and resource-abundant countries
observed in the data.
With equal initial capital stocks across countries, a larger initial exhaustible resource stock induces a transitionally larger
GDP growth gap between the resource-poor country and the resource-abundant country. We explain this phenomenon as
follows. A larger initial exhaustible resource stock leads to a larger world aggregate (human and physical) capital stock in the
long run and therefore, to larger long-run revenues from sales of the exhaustible resource. These larger revenues discourage
investment in the resource-abundant country to a greater extent than the case where the resource endowment is smaller.
The GDP growth gap between the two countries is therefore transitionally larger.
Initial capital stocks do not influence the resource curse condition. However, if the resource curse condition holds, the
larger the ratio of aggregate capital of the resource-abundant country to that of the resource-poor country the smaller the
GDP growth gap between the two countries. This result may explain why some resource-rich countries experience a smaller
negative effect of resource abundance, they had a relatively large stock of capital when their resource discoveries were
made. When allowing for different initial endowments of capital, the model can generate about two thirds of the gap in
GDP growth rates observed in the data.

Appendix A. Proof of Proposition 1

Using (15) and r̃ = q̃˙ /q̃, the budget constraint of the household of country P can be rewritten as
1/θ
Z˙ P (t )
 q̃˙ (t ) 
Z P (t ) q̃(t ) c̃ P (0) −ρ /θ t
− =− e . (A.1)
q̃(t ) q̃(t ) q̃(t ) q̃(0) q̃(t )

Note that the left-hand side of expression (A.1) equals the time derivative d(
Z P /q̃)/dt . Integrating (A.1), we obtain

  ∞
Z P (t ) Z P (τ ) c̃ P (0)
lim − =− q̃(t )(1−θ )/θ e −ρ /θ t dt . (A.2)
t →∞ q̃(t ) q̃(τ ) q̃(0)1/θ
τ
Employing the transversality condition (16) implies

∞
c̃ P (0)

Z P (τ ) = q̃(τ ) q̃(t )(1−θ )/θ e −ρ /θ t dt . (A.3)
q̃(0)1/θ
τ

Since the ratio c̃ A (t )/c̃ P (t ) is constant for all t , let c̃ A (t ) = ς c̃ P (t ), where ς is a positive constant. Using r̃ = q̃˙ /q̃, the
budget constraint of the household of country A can be rewritten as

Z˙ A
 q̃˙ 
ZA c̃ P
− +ς =
N. (A.4)
q̃ q̃ q̃ q̃

Using  S˙ , setting c̃ P (t ) = (q̃(t )/q̃(0))1/θ c̃ P (0)e −ρ /θ t , and integrating, we obtain


N = −

∞ 
˙  ∞
Z A (t ) q̃˙ (t ) 
Z A (t ) c̃ P (0) q̃(t )(1−θ )/θ
− +ς dt = N (t ) dt = 
 S (τ ) − lim 
S (t ). (A.5)
q̃(t ) q̃(t ) q̃(t ) q̃(0)1/θ e ρ /θ t t →∞
τ τ

Since depletion is costless, limt →∞ 


S (t ) = 0. Therefore (A.5) can be simplified to
B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93 91

  ∞
Z A (t ) Z A (τ ) c̃ P (0)
lim − +ς q̃(t )(1−θ )/θ e −(ρ /θ )t dt = 
S (τ ). (A.6)
t →∞ q̃(t ) q̃(τ ) q̃(0)1/θ
τ
Using the transversality condition (Eq. (16)) and rearranging, we obtain

∞
c̃ P (0)
ς q̃(τ ) q̃(t )(1−θ )/θ e −(ρ /θ )t dt = q̃(τ )
S (τ ) + 
Z A (τ ). (A.7)
q̃(0)1/θ
τ

Using (A.3), we yield ς = (q̃(τ )


S (τ ) + 
Z A (τ ))/
Z P (τ ) for all τ . 2

Appendix B. Proof of Proposition 2

Taking the log-time derivative of (10) and (11) and using (14), r̃ K = r̃ + δ K and r̃ H = r̃ + δ H , we obtain
 φ σ 
1 1 r̃ V
r̃˙ = βμ (1−β)μ χμ (η − r̃ ) −1 . (B.1)
+ r̃ V B χ μ Λα 1/φ
(r̃ +δ K )μ (r̃ +δ H )μ

Setting r̃ V = R V (r̃ + δ K , r̃ + δ H ), this is a differential equation in r̃ only. It follows from (B.1) that, in the long run, r̃ either
equals η or it equals the r̃ that solves R V (r̃ + δ K , r̃ + δ H ) = Λα 1/φ . The latter case corresponds to the boundary case, which
we do not analyze. Thus, the long-run value of r̃ equals η .25 It follows from (4) that the long-run growth rate of q̃ equals
v q̃ = r̄ = η and the per-country growth rate of consumption equals v = (r̄ − ρ )/θ = (η − ρ )/θ .
S˙ = −
Next, note that  S equals the ratio 
N. Thus the growth rate of  S˙ /
S = −
N /
S. The transversality condition im-
plies that, as time approaches infinity, both 
N and  S˙ (t )/
S approach zero. Using L’Hôpital’s rule, we obtain limt →∞  S (t ) =
limt →∞ −N (t )/ ˙ (t )/
S (t ) = limt →∞ −
N ˙ (t )/
S˙ (t ) = limt →∞ 
N N (t ), implying that as t approaches infinity, 
S and 
N grow at
equal rates vS (this result was previously derived by Dasgupta and Heal).
The first-order conditions of the final good sector in each country and the market clearing condition for the resource

Z A +
extracted 
N = N P imply 
NA +  N = ( 1−
α
α r̃ V
e φ ηt )σ · (
VA +
V P ) = ( 1−
α
α r̃ V
e φ ηt )σ · ( B1χ r̃ H r̃ K
)μ · μ
ZP
μ , thus
q̃ q̃ r̃ V β μ r̃ H +(1−β)μ r̃ K

 σ σ σ −μ
N q̃ 1−α r̃ V ( r̃ HB χr̃ K )μ
= (B.2)

Z e φ ηt α μ
β μ r̃ + (1 − β)μ r̃
μ
H K

where Z = Z A +Z P . Denoting the long-run growth rates of Z and  N by v and v S , respectively, and since r̃ is constant in the
long run, Eq. (B.2) indicates that the long-run growth rates of  Z and N must satisfy v S − v + σ ( v q̃ − φ η ) = v
S − v + η = 0.
This implies v 
S = v − η = −(ρ − η (1 − θ))/θ . Eqs. (7) and (8) imply that, in the long run, Z i grows at the same rate as c̃ i .
Since 
Z = ZA +  Z P , it is straightforward to show that both  Z A and Z P grow at equal rates. Finally, for the transversality
condition (16) to hold it must also be the case that v − v q̃ = v S < 0. 2

Appendix C. Proof of Proposition 3

The Jacobian matrix of the system of differential equations (29) evaluated at a steady state equals
⎛ ε 0 0 0⎞
1
⎜ Z
F q P ( x)ε2 0 0⎟
⎜ ⎟
J=⎜ ZA ZA ⎟ (C.1)
⎝ F q ( x) F Z ( x) ε3 0⎠
P

F qS (x) F ZS P (x) F ZS A (x) ε4


where 1 = −Ω(r̄ )( 1−
ε α r̄ V σ
α q̄φ ) , 2 = r − v = − v
S, ε ε3 = −ε1 − vS and ε4 = − vS . For j , b = q, Z P , Z A , S ; F j is defined in
j j j j
∂F
(29), F b ≡ ∂ b and Fb (x) equals Fb evaluated at x ≡ (q̄, Z P , Z A , S ). The eigenvalues of matrix J (a diagonal matrix) equal
the elements of its diagonal. 2

Appendix D. Proof of Proposition 4 (continuation)

j
With δ H = δ K , we have that Fb (x) for j , b = q, Z P , Z A , S in (C.1) equal

1
25
One can verify that for σ < (>)1 and R V (η + δ K , η + δ H ) < (>)Λα φ , r̃ converges to η.
92 B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93

Z
F q P (x̄) = (ε1 Z P − c̄ P /θ)/q̄,
 φ σ   
Z r̄ V σ ε1
F q A (x̄) = ε1 σ ZA − + (1 − σ ) Z P − c̄ A /θ /q̄,
Λα 1/φ Ω r̄ V
Z
F Z PA (x̄) = −ε1 ,
 
F qS (x̄) = σ ε1 ε1 /(Ω r̄ V ) − 1 ( Z P + Z A ) /q̄,
F ZS P (x̄) = F ZS A (x̄) = ε1 /q̄.
Thus, the eigenvector of matrix J associated with the negative eigenvalue ε1 equals:
φ σ 
ε1 + v S̃ θ −1 Z P σ r̄ V Q
(u q , u Z P , u Z A , u S ) = n , n, u Z A , − M n (D.1)
ε1 + vS q̄ θ Λα 1/φ ΥD
for n = 0, where

λ Z P / Z A + D ε3 ε3 /(−ε1 )
uZA = n,
Υ Z̄ −
A
1

q̄(−2ε1 − v S )ε3
Υ≡ > 0,
−ε1
φ σ
r̄ V ε1 1
D ≡σ + > 0,
Λα 1/φ vS + ε1 θ
φ σ 
r̄ V
Q ≡ −ε1 σ Z + (c A + c P )/θ /q̄ > 0,
Λα 1/φ
ε1 < 0 and ε3 > 0 are defined in (30), and vS from (19) is negative.
The case M = 0, illustrated in Fig. 4, can be proven in the same manner as case M > 0. Case M < 0 is illustrated in Fig. 4.
We now make use of u q /u Z > 0. Note that points below the ray of steady states can be understood as if the world economy
were endowed with too little aggregate capital Z . Since small amounts of Z reduce the marginal product of the resource, q
must be smaller than its steady-state value. Thus, transitionally q must increase, while Z , Z P , c P and c A increase. The case
Z (0)/ S (0) > Γ can be proven in the same manner. 2

References

Acemoglu, D., Dell, M., 2009. Productivity differences between and within countries. NBER Working Paper 15155.
Allen, R.G.D., 1964. Mathematical Analysis for Economists. St. Martin’s Press, New York.
Asheim, G.B., 1986. Hartwick’s rule in open economies. The Canadian Journal of Economics 19, 395–402.
Atkeson, A., Kehoe, P.J., 1999. Models of energy use: Putty–putty versus putty–clay. American Economic Review 89, 1028–1043.
BEA, 2008. Industry Economic Accounts, The Use of Commodities by Industry after Redefinitions (1987–2006). Bureau of Economic Analysis.
Berndt, E.R., Wood, D.O., 1975. Technology, prices and the derived demand for energy. Review of Economics and Statistics 57, 259–268.
Bernstein, M.A., Griffin, J., 2005. Regional Differences in the Price-Elasticity of Demand for Energy. Rand Corporation, Santa Monica, CA.
Bodenstein, M., Erceg, C.J., Guerrieri, L., 2007. Oil shocks and external adjustment. Board of Governors of the Federal Reserve System, International Finance
Discussion Papers No. 897.
Brander, J.A., Taylor, M.S., 1998. Open access renewable resources: Trade and trade policy in a two-country model. Journal of International Economics 44,
181–209.
Brunnschweiler, C.N., 2008. Cursing the blessings? Natural resource abundance, institutions, and economic growth. World Development 36, 399–419.
Brunnschweiler, C.N., Bulte, E.H., 2008. The resource curse revisited and revised: A tale of paradoxes and red herrings. Journal of Environmental Economics
and Management 55, 248–264.
Bureau of Labor Statistics, 2009. Employment, hours, and earnings, manufacturing average hourly earnings of production workers. US Department of Labor,
http://www.bls.gov/webapps/legacy/cesbtab3.htm.
Chiarella, C., 1980a. Optimal depletion of a nonrenewable resource when technological progress is endogenous. In: Kemp, M.C., Long, N.V. (Eds.), Exhaustible
Resources Optimality and Trade. North-Holland Publishing Co., Amsterdam, pp. 81–91.
Chiarella, C., 1980b. Trade between resource-poor and resource-rich economies as a differential game. In: Kemp, M.C., Long, N.V. (Eds.), Exhaustible Re-
sources Optimality and Trade. North-Holland Publishing Co., Amsterdam, pp. 219–246.
Cigno, A., 1981. Growth with exhaustible resources and endogenous population. Review of Economic Studies 48, 281–287.
Corden, W.M., 1984. Boom sector and Dutch disease economics: Survey and consolidation. Oxford Economic Papers 36, 359–380.
Council of Economic Advisors, 2010. Economic Report of the President, U.S. Government Printing Office, Washington, DC.
Dasgupta, P., Heal, G., 1974. The optimal depletion of exhaustible resources. Review of Economic Studies (Symposium on the Economics of Exhaustible
Resources) 41, 3–28.
Doppelhofer, Miller, Sala-i-Martin, X., 2000. Determinants of long-term growth. A Bayesian averaging of classical estimates (BACE) approach. NBER Working
Paper 7750.
EIA, 2009a. Annual Energy Review 2008. Energy Information Administration, US Department of Energy, Washington, DC.
EIA, 2009b. U.S. natural gas wellhead price. Energy Information Administration, US Department of Energy, http://tonto.eia.doe.gov/dnav/ng/ng_pri_sum_dcu_
nus_a.htm.
Gately, D., Adelman, M.A., Griffin, J.M., 1986. Lessons from the 1986 oil price collapse. Brookings Papers on Economic Activity, 237-84.
B. Gaitan, T.L. Roe / Review of Economic Dynamics 15 (2012) 72–93 93

Geldrop, J.H., Withagen, C.A., 1993. General equilibrium and international trade with exhaustible resources. Journal of International Economics 34, 341–357.
Griffin, J.M., 1992. OPEC and world oil prices: Is the genie back in the bottle? Energy Studies Review, 27–39.
Guvenen, F., 2006. Reconciling conflicting evidence on the elasticity of intertemporal substitution: A macroeconomic perspective. Journal of Monetary
Economics 53, 1451–1472.
Hall, R.E., 1988. Intertemporal substitution in consumption. Journal of Political Economy 96, 339–357.
Hamilton, J.D., 2008. Understanding crude oil prices. NBER Working Paper 14492.
Hartwick, J.M., 1977. Intergenerational equity and the investing of rents from exhaustible resources. American Economic Review 67, 972–974.
Hartwick, J.M., 1995. Constant consumption paths in open economies with exhaustible resources. Review of International Economics 3, 275–283.
Hotelling, H., 1931. The economics of exhaustible resources. Journal of Political Economy 37, 137–175.
Jinji, N., 2007. International trade and renewable resources under asymmetries of resource abundance and resource management. Environmental and Re-
source Economics 37, 621–642.
Jorgenson, D.W., Stiroh, K.J., Gordon, R.J., Sichel, D.E., 2000. Raising the speed limit: U.S. economic growth in the information age. Brookings Papers on
Economic Activity 1, 125–235.
Kemp, M.C., Long, N.V., 1980. The interaction of resource-rich and resource-poor economies. In: Kemp, M.C., Long, N.V. (Eds.), Exhaustible Resources,
Optimality, and Trade. North-Holland Publishing Co., Amsterdam, pp. 197–209.
Kemp, M.C., Long, N.V., 1982. Rybczynski’s theorem in a context of exhaustible resources: the case of time-contingent prices. International Economic
Review 23, 699–710.
Krautkraemer, J.A., 1998. Nonrenewable resource scarcity. Journal of Economic Literature 36 (4), 2065–2107.
Krebs, T., 2003. Human capital risk and economic growth. Quarterly Journal of Economics 118, 709–744.
Lederman, D., Maloney, W.F., 2003. Trade structure and growth. World Bank Policy Research Working Paper 30251.
Li, W., Rychlik, M., Szidarovszky, F., Chiarella, C., 2003. On the attractivity of a class of homogeneous dynamic economic systems. Nonlinear Analysis 52,
1617–1636.
Maddison, A., 2008. Historical statistics for the world economy: 1-2006 ADv. http://www.ggdc.net/maddison/.
Mulligan, C.B., 2002. Capital, interest, and aggregate intertemporal substitution. NBER Working Paper 9373.
NBER, 2009a. U.S. average hourly money earnings in payroll manufacturing industries, Macrohistory database. Income and Employment.
NBER, 2009b. Income and employment, U.S. average hourly earnings, twenty-five manufacturing industries. National Industrial Conference Board 01/1920-
07/1948, Macrohistory database.
Nordhaus, W.D., 1992. Lethal Model 2: The limits to growth revisited. Brookings Papers on Economic Activity 2, 1–43.
OECD, 1995. The OECD input–output database. OECD, Paris.
OECD, 2002. Input–output tables. OECD, Paris.
OECD, 2006. Input–output tables. OECD, Paris.
Papyrakis, E., Gerlagh, R., 2007. Resource abundance and economic growth in the United States. European Economic Review 51, 1011–1039.
Pezzey, J., Withagen, C.A., 1998. The rise, fall and sustainability of capital-resource economies. Scandinavian Journal of Economics 100, 513–527.
Pindyck, R.S., 1978. The optimal exploration and production of nonrenewable resources. Journal Political Economy 86, 841–861.
Pindyck, R.S., 1999. The long-run evolution of energy prices. Energy Journal 20, 1–27.
Pindyck, R.S., Rotemberg, J.J., 1983. Dynamic factor demands and the effects of energy price shocks. American Economic Review 73, 1066–1079.
Potter, N., Christy, F.T., 1962. Trends in Natural Resource Commodities. Johns Hopkins Press, Baltimore.
Rodríguez, F., Sachs, J.D., 1999. Why do resource-abundant economies grow more slowly? Journal of Economic Growth 4, 277–303.
Rotemberg, J.J., Woodford, M., 1996. Imperfect Competition and the effects of energy price increases on economic activity. Journal of Money, Credit, and
Banking 28, 549–577.
Sachs, J.D., Warner, A.M., 1995. Natural resource abundance and economic growth. NBER Working Paper 5398.
Sachs, J.D., Warner, A.M., 2001. Natural resource abundance and economic development. The curse of natural resources. European Economic Review 45,
827–838.
Sala-i-Martin, X., 1997. I just ran four million regressions. NBER Working Paper 6252.
Slade, M.E., 1985. Noninformative trends in natural resource commodity prices: U-shaped price paths exonerated. Journal of Environmental Economics and
Management 12, 181–192.
Slade, M.E., Thille, H., 2009. Whither hotelling: Tests of the theory of exhaustible resources. Annual Review of Resource Economics 1, 239–260.
Solow, R.M., 1974. Intergenerational equity and exhaustible resources. Review of Economic Studies Symposium on the Economics of Exhaustible Re-
sources 41, 29–45.
Stiglitz, J.E., 1974. Growth with exhaustible natural resources: Efficient and optimal growth paths. Review of Economic Studies Symposium on the Economics
of Exhaustible Resources 41, 123–137.
Stijns, J.-P.C., 2005. Natural resource abundance and economic growth revisited. Resources Policy 30, 107–130.
Tornell, A., Lane, P.R., 1995. The voracity effect. American Economic Review 89, 22–46.
U.S. Census Bureau, 1947. Statistical Abstract of the United States. US Census Bureau, Washington, DC.
U.S. Census Bureau, 1952. Statistical Abstract of the United States. US Census Bureau, Washington, DC.
Vissing-Jørgensen, A., 2002. Limited asset market participation and the elasticity of intertemporal substitution. Journal of Political Economy 110, 825–853.
World Bank, 2008. World Development Indicators CD-ROM. International Bank for Reconstruction and Development/The World Bank, Washington, DC, USA.