EC O L O G IC A L E C O N O M IC S 6 6 ( 2 0 08 ) 37 1 –3 78

a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m

w w w. e l s e v i e r. c o m / l o c a t e / e c o l e c o n

ANALYSIS

Economic development and environmental quality: A

reassessment in light of nature's self-regeneration capacity ☆

Luisito Bertinellia,⁎, Eric Stroblb , Benteng Zoua,1

a
CREA, Université du Luxembourg, 162A avenue de la Faïencerie, L-1511 Luxembourg
b
Ecole Polytechnique (Paris), F-91128 Palaiseau Cedex, France

AR TIC LE I N FO ABS TR ACT

Article history: We study the relationship between economic development and consumption of natural
Received 1 June 2007 resources using a vintage capital model. Consumption of natural resources is assumed to
Received in revised form generate pollution, part of which will be absorbed by nature's self-regeneration capacity. We
19 September 2007 find that during the transition dynamics, the shape of the pollution output relationship will
Accepted 19 September 2007 depend on the parameter determining nature's self-regeneration capacity. Using footprint
Available online 17 October 2007 and biological capacity data, we show empirically in a repeated cross-section of countries
that the shape of the pollution–output relationship indeed depends on countries' capacity to
Keywords: regenerate part of the resources they consume from nature.
Environmental quality © 2007 Elsevier B.V. All rights reserved.
Footprint
Biocapacity
Vintage capital model
Environmental Kuznets Curve

Jel Classification:
O13; Q01; Q55

1. Introduction
In October 2006 several newspapers announced that the earth
had reached its ecological debt day/overshoot day.1 This concept,
which was first suggested by the Global Footprint Network (www.
footprintnetwork.org) refers to the day in a year when humanity
has consumed what nature can renew. In other words, if more
wood is consumed than trees grow back, or fishery resources are
☆
The third author acknowledges financial support from DFG
(EBIM, under GRK 1134/1).
⁎ Corresponding author. Tel.: +352 46 66 44 6620.
E-mail addresses: Luisito.Bertinelli@uni.lu (L. Bertinelli),
eric.strobl@shs.polytechnique.fr (E. Strobl), Benteng.Zoui@uni.lu
(B. Zou).
1
In 2006, the World Overshoot Day has been estimated to have
occurred on October 9th (see http://www.neweconomics.org).
0921-8009/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.ecolecon.2007.09.013
depleted faster than spawn, then humanity has to dip into its
stocks of natural capital. It is arguably then the tradeoff between
the depletion of resources and the ability to regenerate these that
is at the heart of the sustainability of countries' consumption
habits. In other words, countries should be assessed not only in
terms of the extent of pollution of the environment, but also with
regard to their contribution in terms of rebuilding natural capital.
This could then, for instance, be used to more accurately separate
net debtors to the environment from net creditors, and hence
allocate polluting rights accordingly.2
2
A decade ago, Costanza et al. (1997), for the first time, provided a
monetary figure of the goods and services our earth's ecosystem is
providing. More specifically, they came up with an annual average
value of $33trillion. A considerable amount of comments had
spurred in the wake of this study, and eventually led to an increased
interest in improving and refining the understanding of the quantity
of resources the world can sustainably provide in the long run.
372 EC O LO GIC A L E CO N O M ICS 6 6 ( 2 00 8 ) 3 7 1 –3 78
Unfortunately academic studies generally have tended to
disregard nature's regenerating ability when it comes to
taking account of a country's contribution to polluting the
environment, in part due to data availability. This is obvious,
for instance, in the numerous studies that have tried to relate
economic development to environmental pollution in what
has come to be known as the Environmental Kuznets Curve
(EKC), where pollution is often proxied by emissions of a
particular pollutant, such as carbon dioxide or sulfur. In
contrast, the recent footprint approach,3 has not only taken
important steps to synthesizing many measures of environ-
mental degradation into a single proxy, but also managed to
net out a country's ability to regenerate nature (which is
usually referred to as biological capacity or biocapacity).
From a theoretical perspective, most studies coping with the
issue of the relationship between economic development and
environmental quality are constructed as if pollution were just
another input in the production function, and ignore the
possibility that different countries may be endowed with
varying capacities to absorb pollution; see, for instance, Stokey's
(1998) fundamental contribution to the theoretical modelling of
the EKC. A notable exception in this regard is Aghion and Howitt
(1998, chap. 5), who introduce a potential rate of regeneration in
their environmental quality equation, thus avoiding pollution
just to be the inverse of environmental quality.
In the present paper we use a theoretical framework enabling
us to take account of nature's regeneration capacity in investi-
gating the relationship between economic development and
pollution. More specifically, we first construct a vintage capital
model in which we, as do Aghion and Howitt (1998), incorporate
the hypothesis that environmental quality is bound between
some upper and a lower threshold, the latter one corresponding to
a total collapse of nature. This allows us to endogenously model a
turning point in the pollution–output relationship, as has been
highlighted in the EKC literature. Traditionally, the EKC literature
has attributed the existence of the pollution–output relationship,
and its possible non-linearity, to three factors: a scale effect (i.e., as
economies grow, pollution grows proportionally), a technique
effect (i.e., if environmental quality is a normal good, nations may
use less pollution–intensive techniques of production), and a
composition effect (i.e., pollution may decrease or increase with
income, according to which goods are produced and traded in a
country). Our framework allows us to tackle the two former
explanations of the EKC. As production increases, pollution,
which will depend on the quantity of inputs, will augment (scale
effect). However, as is standard in vintage capital models, obsolete
machines will be replaced by less pollution–intensive techniques
(technique effect). The overall effect will depend on the interplay
of these two factors. The composition effect is not addressed here
since we model a closed economy, producing one type of good.
Vintage capital models are arguably particularly useful to
analyzing issues related to sustainability as they allow to
endogenously integrate the optimal timing of adoption of a
more environmental friendly technology. In using this modelling
strategy, we can then easily show the discrepancies that may
3
Footprint measures synthesize consumption of natural re-
sources, using a common value, i.e., the standardized area of land
needed to regenerate these resources. A detailed account of
footprint is provided in Section 3.
arise when ignoring nature's potential regeneration ability. In
particular, our results show that if there is a bell-shaped
relationship between economic development and environmental
quality, the peak of the curve will be reached for smaller values of
output per capita when one considers some general measure of
pollution net of self-regeneration rather than only pollution.
Using newly constructed cross-country, time-varying foot-
print data that allow the distinction between gross and net
demand on the environment, we then empirically put the
predictions of our theoretical model to the test. Our empirical
results are indeed found to be in line with the theoretical
findings. More specifically, the maximum of the environmental
degradation–economic development relationship is reached for
larger values when one does not take into account nature's self-
regeneration capacity compared to when this aspect is incor-
porated in the measure of degradation. Moreover, we find that
the upward sloping part of the curve is steeper when regener-
ation is neglected. These results arguably have potentially
important policy implications in that they suggest that failing to
account for countries' regeneration capacity may bias the effort
required to maintain a sustainable development. Put differently,
if pollution generation has to be evaluated so as to attribute it a
cost, so should nature's regeneration capacity.
The rest of the paper is organized as follows. Our vintage
capital model is presented in Section 2. Section 3 contains a
description of our data, the empirical framework, and the econ-
ometric results. Concluding remarks are provided in Section 4.
2. The model
In this section, we first present a standard vintage capital model,
where we add an equation of motion representing environmental
quality. Vintage capital models formalize Schumpeter's idea of
“creative destruction”, and have been applied to different issues
related to economic development (see Aghion and Howitt, 1998).
In the present context, our theoretical framework will introduce a
vintage capital structure in the Aghion and Howitt (1998) frame-
work, where the law of motion of environmental quality will
depend on the pollution flow and some upper limit on environ-
mental quality that takes into account the exhaustibility of re-
sources (Bertinelli et al., 2005). Our model is particularly
appropriate to tackle issues related to sustainable economic
development, as it allows the endogenous determination of the
optimal period of adoption of newer, potentially cleaner,
technologies.
2.1. A vintage capital structure
We consider an economy where the labor market is perfectly
competitive, and the production sector produces only one
final good, assigned either to consumption or investment, and
plays the role of the numeraire. Furthermore, the population
level is assumed to be constant.
2.1.1. Production
At time t N 0, output per capita y(t) is assumed to follow a
vintage capital rule
Z t
yðtÞ ¼ iðzÞdz: ð1Þ
tTðtÞ
 EC O L O G IC A L E C O N O M IC S 6 6 ( 2 0 08 ) 37 1 –3 78
where 0 b T(t) b ∞ represents the vintage of the oldest machine
in use, and i(z) is per capita investment in a machine of age z.
2.1.2. Environmental quality
In the present model, households value per capita consump-
tion level c(t) N 0, as well as the quality of the environment.
Following Aghion and Howitt (1998, chap. 5), we measure
environmental quality E(t) as the difference between the
actual quality and an upper limit Ē. Hence, E(t) will always
be negative, and its equation of motion is given by
: Z t
EðtÞ ¼ qEðtÞ  iðzÞegz dz ;
tTðtÞ
|fflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
PðtÞ
where q N 0 is the rate at which the environment can recover,
γ N 0 is the rate at which technology's environmental friend-
liness improves, and P(t) measures the level of pollution. From
Eq. (2) it follows that pollution is a side-product of investment i(z)
in the production sector. Implicit in this equation is the
assumption that newer technologies are less polluting than
older ones, leading thus to reduced pollution per input.
Finally, per capita output y(t) is either consumed, c(t), or
invested in a capital good, i(t) ≥ 0,
yðtÞ ¼ cðtÞ þ iðtÞ:
It is worth pointing towards the fact that in our model,
environmental quality is not simply an inverse measure of
pollution, as we introduce q, the self-regeneration capacity of
nature. In other words, we make a clear distinction between
gross and net pollution, where the latter measure subtracts
the pollution that is absorbed by nature during a predefined
period of time. Pollution absorption, or natural regeneration
can be of different types. One can think, for instance, of waste
assimilation or carbon sequestration by oceans and forests, or
cropland and ocean's abilities to (re)produce biomass.
2.1.3. Central planner
The central planner's objective function will entail per capita
consumption and environmental quality. More particularly,
the planner will choose the paths of consumption and
environmental quality in order to maximize the instanta-
neous utility of the infinitely lived representative household,
Z l Z l
max Uðc; EÞeqt dt ¼ max ½bcðtÞ þ ð1  bÞEðtÞeqt dt;
c 0 c 0
subject to (2), (3), where ρ N 0 is the constant time preference,
0 b β ≤ 1 is a weight parameter between consumption goods and
environmental quality, and i(z), z ≤ 0, T(0) and E(0) are given
functions and initial conditions. Furthermore we assume that
0 b γ ρ b 1, which are necessary and sufficient conditions for the
existence of a balanced growth path in an exogenous growth
model.
2.2. The link between pollution and output
In Bertinelli et al. (2005) we show explicitly that there exists a
finite positive time t ⁎, such that, when t N t ⁎, the optimal
373
scrapping age, T ⁎, can be reached. Furthermore, we derive
explicitly the transition dynamics towards the optimal
scrapping age as well as the interior solutions. Following
these results, we assume that when the optimal scrapping age
is reached, investment grows at a constant rate g, with an
investment level ī, i.e. i(t) = īegt.4 From this one can easily show
that investment, consumption, and output grow at the same
constant rate.
Initially, if T(0) N T ⁎, the economy starts with a relatively low
stock of capital. Then in order to reach an interior solution
(and thus the optimal path if there is one) as quickly as
possible, one possibility is to invest all output. In this case,
ð2Þ
starting from a corner solution, the economy's subsequent
pollution, when 0 b t b t⁎, will be
Z t
PðtÞ ¼ yðzÞegz dz: ð5Þ
tTðtÞ
Indeed, pollution increases with the accumulation of
output y(.) during all periods in which the machine is in use.
Moreover, there is also a delay effect on pollution coming from
output. In the sequel, we will only analyze the impact of
regeneration q through the scrapping age T(.), which will
influence pollution through the lower limit of the integral in
(5). The direct effect of q on environmental quality through the
first expression on the RHS of (2) only acts as a shift parameter.
ð3Þ
We have to distinguish between two cases, i.e., before and
after the optimal scrapping age t⁎ has been reached. If 0 b t b t⁎,
i.e., the optimal scrapping age has not been reached yet and a
country is in its transition dynamics, then the scrapping age is
shown to be (see Bertinelli et al., 2005)
8  
>
> qþq 1 1b 
<t  t  ln egTð0Þ  1  eðqþqÞt ; 0V tb t4;
g g bðq þ qÞ 
TðtÞ ¼ ð6Þ
> T4 ¼  1 ln 1  q þ g  g þ g eqT4 ; t zt4;
>
:
g q q
where T ⁎ is independent of q. This implies that the regener-
ation ability q will affect the scrapping age directly during the
transition, while there is no direct effect of q when the optimal
scrapping rule is reached. However, as the optimal scrapping
rule follows the transition path, there will be an indirect effect
of q on the optimal scrapping age, since it will influence the
position of the scrapping curve.
Straightforward calculation shows that5
 
ATðt; qÞ t 1  b ðq þ qÞt þ 1eðqþqÞt  1
¼ þ b0:
Aq g bgA ðq þ qÞ2
ð4Þ
Therefore T(t; q) is a decreasing function of q for any time
0 b t b t ⁎. Stated differently, a larger regeneration ability (i.e.,
lager q) will shorten the time needed to reach the potential
optimal scrapping age, while a smaller q will postpone the
period when the potential optimal path will be attained.
When t N t⁎, the economy reaches its optimal path, and T ⁎ is
independent of q. However, a scrapping curve, which is above
4
Note that when the optimal scrapping age is reached, a
balanced growth paths (BGP) may potentially, but not necessarily,
appear.
5
It is easy to show that by a Taylor expansion of the
exponential function, (ρ + q)t + 1e−(ρ + q)t b 1.
374 EC O LO GIC A L E CO N O M ICS 6 6 ( 2 00 8 ) 3 7 1 –3 78
another one during the transition (i.e., 0 b t b t⁎), will not cross
this lower curve once the economy has reached its optimal
path (i.e. t N t⁎). In other words, provided all other parameters
and initial conditions are the same, except for environmental
regeneration ability, the time lag introduced due to the weaker
regeneration ability of environment will never vanish.
If t N t⁎, that is, the optimal scrapping age is reached, then
one can show that pollution will decrease and there will be a
bell-shaped EKC. This will be the case if g b γ, i.e., if the rate at
which environmental friendliness of technologies grows is
higher than the growth rate of investment. In this case,
output, investment, and consumption grow at rate g, where g
will be either 0 b g b min {γ, 1 − e− gT⁎}, or g = 0. Environmental
quality will constantly improve and tend to the upper bound
in the long run, but not at a constant rate of growth, and
pollution will always decrease at rate γ − g,
Z regeneration capacity of nature, rather than some gross
i ðggÞt  
P
t P
PðtÞ ¼ i eðggÞz dz ¼ e 1  e1ðggÞT4 ;
tT4 gg
and
PðtÞ g 1  eðqgÞT4
¼ egt :
yðtÞ ðg  gÞð1  egT4 Þ
In this case, pollution is independent of output and time,
but output is increasing with time t.6
Summarizing the results above, we find that during the
transition dynamics the scrapping curve will follow (6), and
thus, (i) the slope of the pollution–output relationship will be
steeper when q = 0 then in the case where q N 0, i.e., when there
is regeneration of nature. Furthermore, (ii) if q N 0, the
maximum of the pollution level will be reached in an earlier
period. Moreover, once the optimal scrapping age has been
attained, P(t) and y(t) become independent of q, and q only
acts as a shift parameter in the case where there is
regeneration, which, in terms of EKC, would translate in a
pollution–output curve with regeneration that should lie
below the curve where there is none.
3. Empirical estimations
Our theoretical model has pinpointed the bias that may arise
when neglecting to take account of nature's self-regeneration
capacity in an Environmental Kuznets Curve (EKC) framework.
In particular, we have demonstrated that the transition
dynamics, as well as the peak of the pollution–income
relationship, may be affected when omitting regeneration in
the equation of motion of environmental quality. In order to
empirically support these theoretical predictions, we will
estimate a standard empirical EKC specification, alternatively
neglecting and incorporating the recovery rate of nature. More
particularly, we estimate:
Pði; tÞ ¼ X½ yði; tÞ þ WðuÞ
6 fuels, and nuclear energy. The footprint for each of these
We should mention that our result is a special case of
Boucekkine et al.'s (1997, p.346) result. These authors prove that
there are an infinite number of solutions in terms of investment
and output replacement, and we pick a particular one.
where the dependent variable Π alternatively proxies total
pollution (as defined below) or total pollution purged from
nature's capacity to absorb pollution, for country i in period t,
and Ω[ y(.)] takes account of the (possibly non-linear) income
per capita level on (net) pollution generation. Finally, the last
term on the RHS of (7), Ψ (u), consists of unobserved factors
that may be common and/or different across countries, as well
as idiosyncratic shocks.
3.1. Data
Our empirical specification resembles the standard EKC
framework, where we need proxies of pollution and the level
of development with per capita income. The major difference
here with respect to most previous studies in this area is our
use of footprint data and its value net of biocapacity, i.e. the
measure of pollution generation. This allows us to explicitly
discriminate between what nations “borrow” from nature and
what they “pay back” in relatively general terms.
3.1.1. Per capita income data
Income per capita is measured in terms of GDP per capita in
thousands of dollars. Data stems from Maddison (2001, 2003)
and is appropriately adjusted for purchasing power parity (and
expressed in 1990 International Geary–Khamis dollars).
3.1.2. Ecological footprint and biological capacity data7
Our proxy for environmental degradation consists of time-
varying, cross-country series of footprint (net of biocapacity)
and is taken from the Global Footprint Network database. The
ecological footprint data measures the requirements of water
and land area that countries need for consumption and waste
absorption, given the prevailing technology. The footprint
data is in a sense a generalized measure of the pollution
variables usually used in EKC studies. Its generalization lies in
the fact that it not only takes account of one specific pollutant,
but accounts for a number of different types of consumption of
resources. In order to standardize these different types of
pollution generating consumption, the data is expressed in
terms of units of biologically productive area, termed global
hectares. Every unit represents an equivalent amount of
biological productivity. Put differently, “one global hectare is
equal to one hectare with productivity equal to the average
productivity of the (…) Earth (…). Thus, one hectare of highly
productive land is equal to more global hectares than one
hectare of less productive land. Global hectares are normal-
ized so that the number of actual hectares of bioproductive
land and sea on this planet is equal to the number of global
hectares on this planet” (p. 9).
In order to get a comprehensive footprint measure, a
consumption approach is adopted. Thus, imports are added,
and exports are deduced from domestic production obtained
ð7Þ from national accounts. The total ecological footprint is
obtained by summing the demands from crop production,
grazing, forestry, fisheries, built-up areas, hydropower, fossil
7
This subsection relies heavily on Monfreda et al. (2004) and
Wackernagel et al. (2005).
 EC O L O G IC A L E C O N O M IC S 6 6 ( 2 0 08 ) 37 1 –3 78 375

Table 1 – Descriptive statistics measured in terms of the sum of its bioproductive areas, also
Mean Standard Min Max expressed in global hectares.
deviation One should note that footprint and biocapacity data are
flow measures. If the former exceeds the latter, this implies
Footprint– − 1.084044 3.588762 −26.98 10.49
that countries will have to consume part of their stock of
biocapacity
Footprint 1.972477 1.642957 0.09 19.9 resources to overcome their excessive consumption.
Biocapacity 3.05652 3.422165 0.03 28.33 Combined data on ecological footprint and biocapacity is
GDP/capita 4695.024 5328.661 334.98 30929.92 available from 1961 to 2002 in our study, and covers 150
countries, of which we retained 116 for which we had
Source: Global Footprint Network (http://www.footprintnetwork.org/).
Footprint and biocapacity are measured in terms of global hectares,
observations on all relevant variables. Units are expressed in
whereas GDP/capita is measured in Geary–Khamis dollars. terms of per capita global hectares. Out of these 116, there are 22
developed countries, while the rest are developing and transi-
tion economies.

activities represents the standardized biological and technical
capacity required for their production. Thus, for instance, the
ecological footprint of cropland includes cereals for human
consumption, cotton, processed oils, and fodder crops for
livestock. For each of these products, their footprint includes
the area demand, expressed in standardized global hectares.
Moreover, in order to take account of energy consumption in
terms of global hectares, it has been proceeded as follows: the
carbon dioxide releases are integrated into the ecological
footprint by measuring the additional biologically productive
area needed to sequester it through afforestation, and taking
account of a fixed fraction of emissions absorbed by oceans.
In contrast, biological capacity or biocapacity, can be
considered as nature's capacity to supply useful biological
resources, and to absorb waste materials generated by mankind,
given prevailing technologies/practices. Biocapacity is an ag-
gregate of the production of various ecosystems within the area,
e.g., arable, pasture, forest, productive sea. The usefulness of
resources is defined according to the needs of the populations.
Thus, the set of what can be considered as useful may change
from year to year. Biocapacity crucially depends on the quantity
of land and water area available to cover consumption, and is
3.2. Descriptive statistics
In Table 1 we provide a number of descriptive statistics of our
variables of interest. Interestingly, the average value of net
footprint, i.e., footprint purged from biocapacity, is negative,
which could be interpreted as that resources consumed by
nations are below what nature can potentially regenerate.
However, and as already mentioned in the introduction, while
this has been the case in earlier periods, it is no longer so.
Moreover, one needs to realise since footprint and biocapacity
are divided by total population in our database, changes in
their measures may be due to population changes.
In Fig. 1 we have plotted the evolution of footprint and
biocapacity per capita over the whole sample of countries in our
database. As one can see, natural regeneration capacity has
drastically fallen over the period under scrutiny. This is, on the
one hand, due to depletion of natural resources (such as forests),
but also due to population growth (i.e. figures are based on per
capita measures). And it has actually been shown when using
total footprint and biocapacity, rather then per capita measures,
that during the 1990s, the former has surpassed the latter
worldwide (see Wackernagel et al., 2005). In Fig. 2, we can see

Fig. 1 – Evolution of footprint and biocapacity (1960–2000).

376 EC O LO GIC A L E CO N O M ICS 6 6 ( 2 00 8 ) 3 7 1 –3 78

Fig. 2 – Evolution of footprint and biocapacity for high and low income countries (1960–2000).

that biocapacity per capita has been dropping most sharply in survey). However, contrary to these previous studies where one
the developing world, mostly since population has grown or a limited number of pollutants were used separately (mostly
fastest there (black dotted line). However, the footprint of carbon dioxide emissions), our footprint measure is more
developed countries, unsurprisingly, by far exceeds developing general in that it entails an aggregated and standardized
countries' footprint, and thus, developed countries contribute measure of several pollution generating activities. So far, only
far more to the unsustainability of the development path a limited number of studies have availed of such a measure in
worldwide.
Finally, one can also see from Table 1 that income per
capita, expressed in 1990 dollars, spans values from around
300 dollars per capita per year in sub-Saharan African Table 2 – Fixed effect estimations
countries to around 30,000 dollars in the United States and
Dependent variable Dependent variable
in petroleum exporting countries.
Footprint Footprint– Footprint Footprint–
3.3. Estimation results biocapacity biocapacity

(1) (2) (3) (4)

In Table 2, estimation results of specification (7) are displayed,
GDP/capita 0.252⁎⁎⁎ 0.082⁎⁎⁎ 0.096⁎⁎⁎ − 0.041
where we allow income per capita to impact non-linearly on
(0.012) (0.028) (0.020) (0.048)
(net) footprint by including its second order term. Also, we take (GDP/ − 0.005⁎⁎⁎ − 0.004⁎⁎⁎ 0.009⁎⁎⁎ 0.008⁎⁎
account of country specific time-invariant unobservables by capita)2 (0.000) (0.001) (0.002) (0.004)
using a fixed effects estimator and for unobservable time (GDP/ −0.0004⁎⁎⁎ −0.0003⁎⁎⁎
effects common to all countries by including a set of year capita)3 (0.00004) (0.00009)
dummies. In column (1) we regressed footprint against income Constant 1.696⁎⁎⁎ 0.019 1.958⁎⁎⁎ 0.227
(0.153) (0.362) (0.154) (0.367)
per capita. As can be seen, the coefficient results point towards
Time Yes Yes Yes Yes
the existence of a bell-shaped curve, i.e., degradation of nature
dummies
increases up to a certain threshold level of development, and No of 4716 4716 4716 4716
decreases thereafter. This is consistent with our theoretical observations
predictions stating that during the transition dynamics the No of 116 116 116 116
pollution–output relationship will always be upward sloping. It countries
is only when the optimal scrapping age is reached that one can F-test 29.45 60.10 31.41 59.08
R-squared 0.22 0.36 0.23 0.36
potentially expect the relationship to revert, where pollution
Max of $24,001 0,749 $21,012 7,674
will decrease with output if the environmental friendliness of
U-shape
newer technologies is sufficiently high.
One should note that the non-linear relationship between Notes: Robust standard errors in parentheses. ⁎⁎⁎, ⁎⁎, and ⁎ indicate 1,
5, and 10% significance levels.
environmental degradation and output suggested by our results
Maximum of U-shaped relationship (last line of the Table) have been
has been already high-lighted in a number of previous studies
obtained using values of the coefficients up to the 7th digit.
on the EKC (see Dasgupta et al., (2002) for a comprehensive
 EC O L O G IC A L E C O N O M IC S 6 6 ( 2 0 08 ) 37 1 –3 78 377

Fig. 3 – Footprint vs. Net footprint, using FE estimation results.

order to assess for the impact of development on pollution (York affecting the pollution are uncorrelated with GDP per capita
et al., 2004; Bagliani et al., 2006). holds true, we can perform a random effect estimation, i.e.,
One can use the estimated coefficients in column (1) to suppose that the error term is composed of the standard
estimate the turning point of the bell-shaped relationship. observation-specific term and a term which varies randomly
More specifically, it is found to be reached at an income per with each cross-sectional observation. Doing so in the first two
capita level of about $24,000 (see last line of the Table). This columns of Table 3, we end nevertheless up with results in
value implies that in 2002 all countries in our sample are terms of sign and magnitude as in columns (1) and (2) of Table 2.
facing an upward sloping EKC, except for the United States and Last, in order to circumvent possible issues of temporal
Norway who have surpassed this threshold value. dependence, we have performed our regression using 4-year
Importantly, our theoretical model has also shown that periods, as is standard for instance in growth convergence
failing to take account of nature's self-regeneration capacity will regressions. Again, our results do not differ qualitatively from
bias the pollution–output relationship. In particular, the value of Table 2 for the first regression, although in the second
income per capita for which the potential turning point of the regression, where we have deducted biocapacity from footprint,
EKC should be reached will be overestimated. Confronting this
theoretical prediction with our dataset, we reestimated the
same specification as in column (1), but deducted biocapacity, as
Table 3 – Robustness checks
a proxy for regeneration, from the footprint measure. Results of
this estimation are presented in column (2) of Table 2. Similarly Random effect 4-year periods of
observation
to the first column, the estimated coefficients point towards the
existence of a bell-shaped curve. There are however two notable Dependent variable Dependent variable
differences, which clearly appear in Fig. 3: (i) the slope of the
Footprint Footprint– Footprint Footprint–
pollution–output relationship before the turning point is biocapacity biocapacity
significantly smaller, and (ii) the threshold value of income per
capita for which the maximum of the relationship is reached is (1) (2) (3) (4)
significantly lower ($10,700) when using our estimated coeffi- GDP/capita 0.293⁎⁎⁎ 0.100⁎⁎⁎ 0.245⁎⁎⁎ 0.105
cients. Both of these aspects are consistent with our theoretical (0.011) (0.028) (0.031) (0.067)
predictions. (GDP/ − 0.006⁎⁎⁎ − 0.004⁎⁎⁎ −0.004⁎⁎⁎ −0.004⁎
In columns (3) and (4) of Table 2, as well as in Table 3, we capita)2 (0.000) (0.001) (0.001) (0.002)
Constant 0.851⁎⁎⁎ − 3.191⁎⁎⁎ 1.683⁎⁎⁎ −0.058
have performed a number of robustness checks. Several
(0.079) (0.316) (0.221) (0.475)
studies analyzing the existence of an EKC have used poly-
Time Yes Yes Yes Yes
nomials of higher order to test for the existence of a bell- dummies
shaped relationship between pollution and output. In col- No of 4716 4716 1158 1158
umns (3) and (4) of Table 2, we use a third degree polynomial of observations
GDP per capita and rerun our estimation. Qualitatively results No of 116 116 116 116
are the same, the only difference being that the maximum of countries
R-squared 0.22 0.36 0.20 0.35
the relationship is reached for a lower level of GDP per capita
Max of $23,418 1,676 $29,671 1,845
when abstracting from biocapacity, and higher level of GDP
U-shape
per capita when taking account of it.
In Table 3, we have performed the same specification as in Notes: Robust standard errors in parentheses. ⁎⁎⁎, ⁎⁎, and ⁎ indicate 1,
5, and 10% significance levels.
columns (1) and (2) of Table 2, using first random rather than
Maximum of U-shaped relationship (last line of the Table) have been
fixed effects, and also using 4-year periods rather then annual
obtained using values of the coefficients up to the 7th digit.
data. If the hypothesis is that the unobservable variables
378 EC O LO GIC A L E CO N O M ICS 6 6 ( 2 00 8 ) 3 7 1 –3 78
the coefficient on the GDP per capita turns out to be marginally
insignificant ( p-value: 0.117).
4. Conclusion
In this study, we have theoretically derived an Environmental
Kuznets Curve (EKC) using a vintage capital model, and
explicitly taking account of the fact that environmental quality
is not just an inverse measure of pollution. Our model
contributes in several ways to the existing literature. First, we
show that a bell-shaped relationship between environmental
degradation and income per capita may appear once the
optimal scrapping age has been reached. Furthermore, we
show that neglecting nature's regeneration capacity has
substantial implications on the shape of the relationship during
the transition dynamics. In particular, before reaching the
optimal scrapping age, the convergence speed will be higher
and thus, the potential maximum of the EKC will be reached for
lower values of income per capita.
Empirically, we substantiate our theoretical findings by
using recently made available panel data on general country
level measures of pollution and biocapacity. This allows us to
explicitly distinguish between the demand and supply of
natural resources, the latter one being usually omitted in
studies of EKC. Adding the biological capacity of nature in our
regressions is demonstrated to significantly change the shape
of the environmental degradation–income per capita relation-
ship. Although we can show the existence of a bell-shaped
curve, its characteristics, i.e., the slope and maximum, depend
substantially on whether self-regeneration of nature is
accounted for.
From a more general perspective, our results imply that
integrating regeneration capacity into and using more general
measures of demand on natural resources in an analysis of
environmental degradation and the level of economic develop-
ment is of prime importance. Indeed, neglecting this aspect
would arguably lead to misvaluing the real contribution of each
country to resource consumption, and provide only a very
restricted picture of environmental depletion. From a policy
perspective, neglecting nature's self-regeneration capacity
induces biased results in terms of consumption of natural
resources compatible with sustainable economic development,
and thus, ignores the value of ecosystems' contribution to
sustainability of our earth and hence may lead to distortionary
policies, such as the misallocation of polluting rights.
REFERENCES
Aghion, P., Howitt, P., 1998. Endogenous growth theory. MIT Press.
Bagliani M., Bravo G., Dalmazzone, S., 2006. A consumption-based
approach to environmental Kuznets curves using the ecological
footprint indicator, University of Torino, WP 01/2006.
Bertinelli, L., Strobl, E., Zou, B., 2005. Sustainable economic
development and the environment: theory and evidence. IMW
working paper, vol. 369.
Boucekkine, R., Germain, M., Licandro, O., 1997. Replacement
echoes in the vintage capital growth model. Journal of
Economic Theory 74, 333–348.
Costanza, R., d'Arge, R., de Groot, R., Farber, S., Grasso, M., Hannon,
B., Limburg, K., Naeem, S., O'Neill, R., Paruelo, J., Raskin, R.,
Sutton, P., van den Belt, M., 1997. The value of the world's
ecosystem services and natural capital. Nature 387 (6630),
253–260.
Dasgupta, S., Laplante, B., Wang, H., Wheeler, D., 2002. Confronting
the environmental Kuznets curve. Journal of Economic
Perspectives 16 (1), 147–168.
Maddison, A., 2001. The World Economy: A Millennial Perspective,
OECD Development Centre. Paris.
Maddison, A., 2003. The World Economy: Historical Statistics,
OECD Development Centre. Paris.
Monfreda, C., Wackernagel, M., Deumling, D., 2004. Establishing
national natural capital accounts based on detailed ecological
footprint and biological capacity accounts. Land Use Policy 21,
231–246.
Stokey, N.L., 1998. Are there limits to growth? International
Economic Review 39, 1–31.
Wackernagel, M., Monfreda, C., Moran, D., Wermer, P., Goldfinger,
S., Deumling, D., Murray, M., 2005. National Footprint and
Biocapacity Accounts 2005: The underlying calculation method.
URL: http://www.footprintnetwork.org/download.php?id=5.
York R., Rosa E., Dietz, T., 2004. The ecological footprint intensity
of national economies. 8(4): 139–154.