Economics Economic Growth and The Environment Article Summary

Description

1 attachmentsSlide 1 of 1attachment_1attachment_1.slider-slide > img { width: 100%; display: block; }
.slider-slide > img:focus { margin: auto; }

Unformatted Attachment Preview

ECONOMIC GROWTH AND THE ENVIRONMENT
Theodore Panayotou
Harvard University and Cyprus International Institute of Management
I. Introduction
Will the world be able to sustain economic growth indefinitely without running
into resource constraints or despoiling the environment beyond repair? What is the
relationship between a steady increase in incomes and environmental quality? Are there
trade-offs between the goals of achieving high and sustainable rates of economic growth
and attaining high standards of environmental quality. For some social and physical
scientists such as Georgescu-Roegen (1971), Meadows et al. (1972), growing economic
activity (production and consumption) requires larger inputs of energy and material, and
generates larger quantities of waste byproducts. Increased extraction of natural resources,
accumulation of waste, and concentration of pollutants would overwhelm the carrying
capacity of the biosphere and result in the degradation of environmental quality and a
decline in human welfare, despite rising incomes (Daly 1977). Furthermore, it is argued
that degradation of the resource base would eventually put economic activity itself at risk.
To save the environment and even economic activity from itself, economic growth must
cease and the world must make a transition to a steady-state economy.
——————————————————————————————————————————-Paper prepared for and presented at the Spring Seminar of the United Nations Economic Commission for
Europe, Geneva, March 3, 2003.
1
At the other extreme, are those who argue that the fastest road to environmental
improvement is along the path of economic growth: with higher incomes comes increased
demand for goods and services that are less material-intensive, as well as demand for
improved environmental quality that leads to the adoption of environmental protection
measures. As Beckerman (1992) puts it, “The strong correlation between incomes, and
the extent to which environmental protection measures are adopted, demonstrates that in
the longer run, the surest way to improve your environment is to become rich,” (quoted
by Rothman 1998, pp. 178). Some went as far as claiming that environmental regulation,
by reducing economic growth, may actually reduce environmental quality (Barlett 1994).
Yet, others (e.g., Shafik and Bandyopadhyay (1992), Panayotou (1993),
Grossman and Krueger (1993) and Selden and Song (1994)) have hypothesized that the
relationship between economic growth and environmental quality, whether positive or
negative, is not fixed along a country’s development path; indeed it may change sign
from positive to negative as a country reaches a level of income at which people demand
and afford more efficient infrastructure and a cleaner environment. The implied invertedU relationship between environmental degradation and economic growth came to be
known as the “Environmental Kuznets Curve,” by analogy with the income-inequality
relationship postulated by Kuznets (1965, 1966). At low levels of development, both the
quantity and the intensity of environmental degradation are limited to the impacts of
subsistence economic activity on the resource base and to limited quantities of
biodegradable wastes.
As agriculture and resource extraction intensify and
industrialization takes off, both resource depletion and waste generation accelerate. At
higher levels of development, structural change towards information-based industries and
2
services, more efficient technologies, and increased demand for environmental quality
result in leveling-off and a steady decline of environmental degradation (Panayotou
1993), as seen in the Figure 1 below:
Environmental
degradation
Pre-industrial
economies
Industrial
economies
Post-industrial
economies
(service economy)
Stages of economic development
Figure 1: The Environmental Kuznets Curve: a development-environment relationship
The issue of whether environmental degradation (a) increases monotonically, (b)
decreases monotonically, or (c) first increases and then declines along a country’s
development path, has critical implications for policy.
A monotonic increase of
environmental degradation with economic growth calls for strict environmental
regulations and even limits on economic growth to ensure a sustainable scale of economic
activity within the ecological life-support system (Arrow et al. 1995) A monotonic
decrease of environmental degradation along a country’s development path suggests that
policies that accelerate economic growth lead also to rapid environmental improvements
and no explicit environmental policies are needed; indeed, they may be counterproductive
if they slow down economic growth and thereby delay environmental improvement.
3
Finally, if the Environmental Kuznets Curve hypothesis is supported by evidence,
development policies have the potential of being environmentally benign over the long
run, (at high incomes), but they are also capable of significant environmental damage in
the short-to-medium run (at low-to-medium-level incomes). In this case, several issues
arise: (1) at what level of per capita income is the turning point? (2) How much damage
would have taken place, and how can they be avoided? (3) Would any ecological
thresholds be violated and irreversible damages take place before environmental
degradation turns down, and how can they be avoided? (4) Is environmental improvement
at higher income levels automatic, or does it require conscious institutional and policy
reforms? and (5) how to accelerate the development process so that developing
economies and economies in transition can experience the same improved economic and
environmental conditions enjoyed by developed market economies?
The objective of this paper is to examine the empirical relationship between
economic growth and the environment in different stages of economic development and
explore how economic growth might be decoupled from environmental pressures.
Particular attention is paid to the role of structural change, technological change and
economic and environmental policies in the process of decoupling and the reconciliation
of economic and environmental objectives. We then examine the experience of the ECE
region in fostering environmentally friendly growth. Whether and how it has been
possible to decouple economic growth from environmental pressures in the ECE region.
What has been the role of structural change, technological change and policy instruments
in this decoupling for the two major groups of countries that constitute the ECE region,
the developed market economies and the economies in transition.
4
II. Empirical Models of Environment and Growth
The environment-growth debate in the empirical literature has centered on the
following five questions.
relationship
between
First, does the often-hypothesized inverted-U shaped
income
and
environmental
degradation,
known
as
the
Environmental Kuznets Curve actually exist, and how robust and general is it? Second,
what is the role other factors, such as population growth, income distribution,
international trade and time-and-space-dependent (rather than income-dependent)
variables? Third, how relevant is a statistical relationship estimated from cross-country
or panel data to an individual country’s environmental trajectory and to the likely path of
present day developing countries and transition economies.
Fourth, what are the
implications of ecological thresholds and irreversible damages for the inverted-U shaped
relationship between environmental degradation and economic growth? Can a static
statistical relationship be interpreted in terms of carrying capacity, ecosystem resilience
and sustainability? Finally, what is the role of environmental policy both in explaining
the shape of the income-environment relationship, and in lowering the environmental
price of economic growth and ensuring more sustainable outcomes?
Empirical models of environment and growth consist usually of reduced form
single-equation specifications relating an environmental impact indicator to a measure of
income per capita. Some models use emissions of a particular pollutant (e.g. SO2, CO2,
or particulates) as dependent variables while others use ambient concentrations of various
pollutants as recorded by monitoring stations; yet other studies employ composite indexes
of environmental degradation. The common independent variable of most models is
income per capita, but some studies use income data converted into purchasing power
5
parity (PPP) while others use incomes at market exchange rates. Different studies control
for different variables, such as population density, openness to trade, income distribution,
geographical and institutional variables. The functional specification is usually quadratic;
log quadratic or cubic in income and environmental degradation. They are estimated
econometrically using cross-section or panel data and many test for country and time
fixed effects. The ad hoc specifications and reduced form of these models turns them
into a “black box” that shrouds the underlying determinants of environmental quality and
circumscribes their usefulness in policy formulation.
There have been some recent
efforts to study the theoretical underpinnings of the environment-income relationship and
some modest attempts to decompose the income-environment relationship into its
constituent scale, composition and abatement effects. However, as Stern (1998) has
concluded, there has been no explicit empirical testing of the theoretical models and still
we do not have a rigorous and systematic decomposition analysis.
We proceed with an overview of the theoretical microfoundations of the empirical
models, followed by a survey of studies whose primary purpose is to estimate the
income-environment relationship. We then survey attempts at decomposition analysis
followed by studies that are focusing on mediating or conditioning variables such
international trade as well as on ecological and sustainability considerations and issues of
political economy and policy.
Finally, we review the experience of the ECE region in terms of the growth and
environment relationship and efforts to decouple the two.
6
III Theoretical Underpinnings of Empirical Models
The characteristics of production and abatement technology, and of preferences
and their evolution with income growth, underlie the shape of the income-environment
relationship. Some authors focus on production technology shifts brought about by
structural changes accompanying economic growth (Grossman and Krueger 1993,
Panayotou 1993). Others have emphasized the characteristics of abatement technology
(Selden and Song 1995, Andreoni and Levinson 1998. And yet others have focused on
the properties of preferences and especially the income elasticity for environmental
quality (McConnell 1997, Kriström and Rivera 1995, Antle and Heidebrink 1995). A few
authors have formulated complete growth models with plausible assumptions about the
properties of both technology and preferences from which they derive Environmental
Kuznets Curves (Lopez 1994, Selden and Song 1995). In this section, we will briefly
review the main theoretical strands of the KC literature.
The model by Lopez (1994) consists of two production sectors, with weak
separability between pollution and other factors of production (labor and capital),
constant returns to scale and technical change and prices that are exogenously
determined. When producers free ride on the environment or pay fixed pollution prices,
growth results inescapably in higher pollution levels. When producers pay the full
marginal social cost of pollution they generate, the pollution-income relationship depends
on the properties of technology and of preferences.
With homothetic preferences
pollution levels still increase monotonically with income but with non-homothetic
preferences, the faster the marginal utility declines with consumption levels and the
higher the elasticity of substitution between pollution and other inputs, the less pollution
7
will increase with output growth. Empirically plausible values for these two parameters
result in an inverted-U-shaped relationship between pollution and income. This tends to
explain why in the case of pollutants such as SO2 and particulates, where the damage is
more evident to consumers and, hence, pollution prices are near their marginal social
costs, turning points have been obtained at relatively low-income levels. In contrast,
turning points are found at much higher income levels, or not at all for pollutants such as
CO2, from which damage is less immediate and less evident to the consumers, and hence
under priced, if priced at all.
Selden and Song (1995), using Forster’s (1973) growth and pollution model with
utility function that is additively separable between consumption and pollution derive an
inverted-U path for pollution and a J-curve for abatement that starts when a given capital
stock is achieved; i.e. expenditure on pollution abatement is zero until “development has
created enough consumption and enough environmental damage to merit expenditures on
abatement” (Selden and Song 1995 p. 164). Two sets of factors contribute to early and
rapid increase in abatement: (a) on the technology side, large direct effects of growth on
pollution and high marginal effectiveness of abatement, and (b) on the demand side,
(preferences) rapidly declining marginal utility of consumption and rapidly rising
marginal concern over mounting pollution levels. To the extent that development reduces
the carrying capacity of the environment, the abatement effort must increase at an
increasing rate to offset the effects of growth on pollution.
A number of empirical EKC models have emphasized the role of the income
elasticity of demand for environmental quality as the theoretical underpinning of
inverted-U shaped relationship between pollution and income (Beckerman 1992, Antle
8
and Heiderbrink 1995, Chadhuri and Pfaf 1996). Arrow et al. (1995) state that because
the inverted-U shaped curve “is consistent with the notion that people spent
proportionately more on environmental quality as their income rises, economists have
conjectured that the curve applies to environmental quality generally” (p. 520).
A
number of earlier studies (Boercherding and Deaton 1972), Bergstrom and Goodman
1973, and Walters 1975) found income elasticities for environmental improvements
greater than one. Kriström (1995) reviewed evidence from CVM studies (Lombrer et al.
1991 and Carson et al. 1994) that found income elasticities for environmental quality
much less than one.
Does the finding of a low-income elasticity of demand for
environmental quality present a problem for EKC models?
McConnell (1997) examines the role of the income elasticity of demand for
environmental quality in EKC models by adapting a static model of an infinitely lived
household in which pollution is generated by consumption and reduced by abatement.
He finds that the higher the income elasticity of demand for environmental quality, the
slower the growth of pollution when positive, and the faster the decline when negative,
but there is no special role assigned to income elasticity equal or greater to one. In fact,
pollution can decline even with zero or negative income elasticity of demand, as when
preferences are non-additive or pollution reduces output (e.g. reduced labor productivity
due to health damages, material damage due to acid rain deposition or loss of crop output
due to agricultural externalities). He concludes that preferences consistent with a positive
income elasticity of demand for environmental quality, while helpful, are neither
necessary nor sufficient for an inverted-U shaped relationship between pollution and
income. McConnell found little microeconomic evidence in non-valuation studies that
9
supports a major role for responsiveness of preferences to income changes in
macroeconomic EKC models.
Kriström (1998, 2000) interpreting the EKC as an equilibrium relationship in
which technology and preference parameters determine its exact shape, proposed a
simple model consisting of: (a) a utility function of a representative consumer increasing
in consumption and decreasing in pollution; and (b) a production function with pollution
and technology parameters as inputs.
Technological progress is assumed to be
exogenous. He interprets the EKC as an expansion path resulting from maximizing
welfare subject to a technology constraint at each point in time; along the optimal path
the marginal willingness to pay (MWTP) for environmental quality equals its marginal
supply costs (in terms of forgone output). Along the expansion path the marginal utility
of consumption, which is initially high, declines and the marginal disutility of pollution
(MWTP for environmental quality) is initially low and rises. Technological progress
makes possible more production at each level of environmental quality, which creates
both substitution and income effects.
The substitution effect is positive for both
consumption and pollution. The substitution effect dominates at low-income levels and
the income effect dominates at high-income levels producing an inverted-U shaped
relationship between pollution and income. Of course, the exact shape of the relationship
and the turning point, if any, depend on the interplay of the technology and preference
parameters, which differ among pollutants and circumstances.
In overlapping generation models by John and Pecchenino (1994,1995), John et
al. (1995) and Jones and Mannelli (1995) pollution is generated by consumption activities
and is only partially internalized as the current generation considers the impact of
10
pollution on its own welfare but not on the welfare of future generations. In these
models, the economy is characterized by declining environmental quality when
consumption levels are low, but given sufficient returns to environmental maintenance,
environmental quality recovers and may even improve absolutely with economic growth.
Andreoni and Levinson (1998) derived inverted-U shaped pollution-income
curves from a simple model with two commodities, one good and one bad, which are
bundled together. Income increases result in increased consumption of the good, which
generates more of the bad. This presents consumers with a trade-off: by sacrificing some
consumption of the good they can spend some of their income on abatement to reduce the
ill effects of the bad. When increasing returns characterize the abatement technology,
high-income individuals (or countries), giving rise to an optimal pollution-income path
that is inverted-U shaped. The abatement technology is characterized by increasing
returns when it requires lumpy investment or when the lower marginal cost technology
required large fixed costs (e.g. scrubbers or treatment plants); poor economies are not
large enough or polluted enough to obtain a worthwhile return on such investments and
end up using low fixed cost, high marginal-cost technologies, while rich economies are
large enough and polluted enough to make effective use of high fixed-cost, low marginalcost technologies.
Different pollutants have different abatement technologies and
correspondingly the income environment relationship may or may not be inverted-U
shaped.
The authors argue that similar results are obtained from other “good-bad”
combinations e.g. driving a vehicle associated with mortality risk which can be abated by
investments in safety equipment: “both the poor who drive very little and the rich, who
invest in safe cars face lower risk from driving than middle-income people”. Indeed,
11
empirically, Khan (1998) found such an inverted-U shaped relationship between
hydrocarbon emissions and household income in California, and Chaudhuri and Pfaf
(1998) between indoor pollution and household income in Pakistan.
IV. The Basic Environmental Kuznets Curve
The 1990s have seen the advent of the Environmental Kuznets Curve (EKC)
hypothesis and an explosion of studies that tested it for a variety of pollutants. In this
section, we review the basic EKC studies that focus on the income-environment
relationship; in subsequent sections we review studies that focus on mediating or
conditioning variables. Appendix Table 1 Summarize the empirical studies of the EKC
hypothesis and their findings and Appendix Figure 1 depicts these findings in a
diagrammatic form. The first set of empirical EKC studies appeared independently in
three working papers by: Grossman and Krueger (1991), in an NBER working paper as
part of a study of the likely environmental impacts of NAFTA; by Shafik and
Bandyobadhyay (1992) for the World Bank’s 1992 World Development Report; and by
Panayotou (1992) in a Development Discussion Paper as part of a study for the
International Labor Office. It is reassuring that these early studies found turning points
for several pollutants (SO2, NOx, and SPM) in a similar income range of $3,000 – $5,000
per capita.
Grossman and Krueger (1993, 1994) estimated EKCs for SO2, dark matter
(smoke) and suspended particles using GEMS (Global Environmental Monitoring
System) data for 52 cities in 32 counties during the period 1977-88, and in per capita
12
GDP data in purchasing power parity (PPP) terms. For SO2 and dark matter, they found
turning points at $4,000-$5,000 per capita; suspended particles continually declined at
even low-income levels. However, at income levels over $10,000-$15,000 all three
pollutants began to increase again, a finding which may be an artifact of the cubic
equation used in the estimation and the limited number of observations at high-income
levels.
Shafik and Bandyopadhyay (1992) estimated EKCs for 10 different indicators for
environmental degradation, including lack of clean water and sanitation, deforestation,
municipal waste, and sulfur oxides and carbon emissions.
Their sample includes
observations for up to 149 countries during 1960-90 and their functional specification
log-linear, log quadratic and logarithmic cubic polynomial functional forms. They found
that lack of clean water and sanitation declined uniformly with increasing incomes and
over time; water pollution, municipal waste and carbon emissions increase; and
deforestation is independent of income levels. In contrast, air pollutants conform to the
EKC hypothesis with turning points at income levels between $300 and $4000.
Panayotou (1992, 1993, and 1995), using cross section data and a translog specification
found similar results for these pollutants, with turning points at income levels ranging
from $3000 to $5000. (The lower figures are due to the use of official exchange rates
rather than PPP rates).
Panayotou also found that deforestation also conforms to the EKC hypothesis,
with a turning point around $800 per capita; controlling for income deforestation is
significantly greater in tropical, and in densely populated countries.
Cropper and
Griffiths (1994), on the other hand, using panel data for 64 countries over a 30-year
13
period, obtained a turning point for deforestation in Africa and Latin America between
$4700 and $5400 (In PPP terms). These turning points are a multiple of those found by
Panayotou and by Shafik and Bandyopadhyay’s studies, a possible consequence of
Cropper and Griffith’s use of panel data. A study by Antle and Heidebrink (1995), which
used cross-section data, found turning points of $1,200 (1985 prices) for national parks
and $2,000 for afforestation. On the other hand, Bhattari and Hammig (2000), who used
panel data on deforestation for 21 countries in Latin America, found EKC with a turning
point of $6,800. Furthermore, while earlier studies have controlled for institutional
factors, such as the level of indebtedness, and found them to have the expected signs,
negative and positive respectively (Bhattarai and Hammig 2000).
Returning to urban environmental quality, the mid-1990s have seen a large
number of studies focusing on airborne pollutants, Selden and Song (1994) estimated
EKCs for SO2, Nox, and SPM and CO using longitudinal data on emissions in mostly
developed countries. They found turning points of $8,700 for SO2, $11,200 for Nox,
$10,300 for SPM, and $5,600 for CO. These are much higher levels than Grossman and
Krueger’s, a discrepancy that which authors explain in terms of reduction of emissions
lagging behind reduction in ambient concentrations. However, this reasoning does not
explain the large difference of their results from those of Panayotou, who also uses
emissions data; the use of longitudinal data versus cross-section may help explain part of
the difference. Cole et al. (1997) estimated income-environment relationships for many
environmental indicators, including total energy use, transport emissions of SO2, SPM
and NO2, nitrates in water, traffic volumes, CFC emissions and methane. They found
inverted-U shaped curves only for local air pollutants and CFCs and concluded that
14
“meaningful EKCs exist only for local air pollutants, while indicators with a more global,
more indirect, environmental impact either increase with income or else have high
turning points with large standard errors” (p.441). This conclusion would lead one to
expect that CO2, the global pollutant par excellance, would increase monotonically with
income, at least within any observable income range since the impacts of global warming
are (totally) externalized to other countries and future generations. Indeed, earlier studies
(e.g. Shafik and Bandopadhyay 1992) obtained such a result. Holtz-Eakin and Selden
(1995) estimated EKCs for CO2 using panel data, and found that CO2 emissions per
capita do not begin to decline until income per capita reaches $35,000, a result that
confirms earlier findings by Shafik (1994).
However, more recent studies, using better data and more sophisticated estimation
techniques, have obtained turning points for CO2 emissions, while higher than those of
local pollutants, still within the range of observable income levels. Schmalnesee et al.
(1998) using a spline regression with ten piece-wise segments and the Holtz-Eakin and
Selden data, have obtained an inverted-U shaped relationship between CO2 emissions and
income per capita in PPP$ (1985). They found negative CO2 emission elasticities with
respect to income per capita at the lowest and highest income splines, and a turning point
in the range of $10, 000 to $17,000 per capita. Galeotti and Lanza (1999a,b) have tested
alternative functional specifications for the CO2-income relationship, including Gamma
and Weibrill functions as well as quadratic and cubic functions. They found turning
points between $15,000 and $22,000 depending on the specification and sample.
Another recent study by Panayotou, Sachs, and Peterson (1999), using a ten
segment piece-wise spline function and panel data for 150 countries during 1960-92, have
15
found results quite similar to those of Schmalensee et al. The income elasticity of
emissions was low at the lowest income spline, and rose to a maximum at around $11,500
per capita (turning point) and turned negative at incomes of about $17,500. Finding an
inverted-U shaped relationship for an invisible pollutant with much delayed effects and
ample scope for a fee-riding behavior, is a bit puzzling but fully explainable by structural
changes accompanying economic growth: from agriculture, to industry, to services, three
sectors with different carbon emissions intensities.
V. Decomposition of the Income-Environment Relationship
The income-environment relationship specified and tested in much of the
literature is a reduced form function that aims to capture the “net effect” of income on the
environment. Income is used as an omnibus variable representing a variety of underlying
influences, whose separate effects are obscured. For this reason, some authors termed the
reduced form specification as a “black box” that hides more than it reveals; “without
explicit consideration of the underlying determinants of environmental quality, the scope
of policy intervention is unduly circumscribed.” (Panayotou 1997, pp. 469). In order to
understand why the observed relationship exists, and how we might influence it, more
analytical and structural models of the income-environment relationships are needed. As
a first step, it must be recognized that the observed environmental quality is the outcome
of the interplay of emissions and abatement within a location specific context, and try to
identify the different effects of economic development on environmental quality
transmitted through the income variables.
16
Panayotou (1997) and Islam, Vincent, and Panayotou (1999) identify three
distinct structural forces that affect the environment: (a) the scale of economic activity,
(b) the composition or structure of economic activity and (c) the effect of income on the
demand and supply of pollution abatement efforts. They name the respective effects on
the environment: the scale or level effect, the structure or composition effect, and the
pure income or abatement effect.
Algebraically:
Kaufman et al. (1998) and Nguyen (1999) have identified analogous effects.
The scale effect on pollution, controlling for the other two effects, is expected to
be a monotonically increasing function of income since the larger the scale of economic
activity per unit of area the higher the level of pollution, all else equal. The structural
change that accompanies economic growth affects environmental quality by changing the
composition of economic activity toward sectors of higher or lower pollution intensity. At
lower levels of income, the dominant shift is from agriculture to industry with a
consequent increase of pollution intensity. At higher incomes, the dominant shift is for
industry to services with a consequent decrease in pollution intensity. Hence, the
changing share of industry in GDP may represent structural change. The composition
effect is then likely to be a non-monotonic (inverted-U) function of GDP, i.e. as the share
of industry first rises and then falls, environmental pollution will first rise and then fall
with income growth, controlling for all other influences transmitted through income.
17
Stripped of its scale and composition effects, the income variable represents
“pure” income effects on the demand and supply of environmental quality. On the
demand side, at low incomes, income increases are directed towards food and shelter, and
have little effect on the demand for environmental quality; while at higher income levels,
income increases lead to higher demand for environmental quality since the latter is a
normal (if not a superior) good. The Engel’s curve for environmental quality translates
into an inverted-J curve between income and environmental degradation (Selden and
Song 1995); that is once the scale and composition effects of income growth are
controlled for, pollution is a non-increasing function of income reflecting the nonnegative elasticity for environmental quality. On the supply side, higher incomes make
available the resources needed for increased private and public expenditures on pollution
abatement, and induce stricter environmental regulations that internalize pollution
externalities. The income variable (stripped of its scale and composition effects) captures
the locus of the equilibrium abatement levels, where demand and supply, both incomedependent, are equal. Hence, the abatement effect is expected to be a monotonically
decreasing function of income. Figure 2 below depicts these three effects based on Islam,
Vincent, and Panayotou (1999):
18
Figure 2. Decomposition of Income Effects on the Environment
Panayotou (1997) specified a cubic functional form for all decomposition effects,
and included variables representing population density, economic growth rate, and a
policy variable (quality of institutions). He tested the model with a panel data set for
thirty countries; he used SO2 data from GEMS and PPP-adjusted GDP figures from
Summers and Heston (1991). The decomposition of the income variable into its
constituent channels improved the overall fit dramatically, compared with the reduced
form equation. The scale of the economy increases SO2 concentrations monotonically,
but at a diminishing rate, and it is particularly strong up to income levels of $3 million
per square kilometer.
The composition effect leads