Introduction

The relationship between economic growth and the environment is, and will always remain, controversial. Some see the emergence of new pollution problems, the lack of success in dealing with global warming and the still rising population in the Third World as proof positive that humans are a short-sighted and rapacious species.

These views are not necessarily inconsistent and growth theory offers us the tools needed to explore the link between environmental problems of today and the likelihood of their improvement tomorrow. It allows us to clarify these conflicting views by use of theory, and when differences still remain, to create useful empirical tests that quantify relative magnitudes.

For many years, the limited natural resource base of the planet was viewed as the source of limits to growth. This was, for example, the focus of the original and subse­quent “Limits to Growth” monograph and the efforts by economists refuting its conclu­sions.^[472] Recently however it has become clear that limits to growth may not only arise from nature’s finite source of raw minerals, but instead from nature’s limited ability to act as a sink for human wastes. It is perhaps natural to think first of the environment as a source of raw materials, oil and valuable minerals. This interpretation of nature’s service to mankind led to a large and still growing theoretical literature on the limits to growth created by natural resource scarcity.

Nature’s other role - its role as a sink for unwanted by-products of economic activ­ity - has typically been given less attention. As a sink, nature dissipates harmful air, water and solid pollutants, is the final resting place for millions of tons of garbage, and is the unfortunate repository for many toxic chemicals. When the environment’s ability to dissipate or absorb wastes is exceeded, environmental quality falls and the policy response to this reduction in quality may in turn limit growth. Growth may be limited because reductions in environmental quality call forth more intensive clean up or abate­ment efforts that lower the return to investment, or more apocalyptically, growth may be limited when humans do such damage to the ecosystem that it deteriorates beyond repair and settles on a new lower, less productive steady state.^[474]

This link between growth and the environment has of course received much more at­tention recently because of the rapidly expanding empirical literature on the relationship between per capita income and pollution. This literature, known as the Environmental Kuznets Curve (EKC) literature has been enormously influential. So to a certain extent, the tables have now turned: there is far less concern over the ultimate exhaustion of oil or magnesium, and far more concern over air quality, global warming, and the emissions of industrial production.

The economics literature examining the link between growth and the environment is huge; it covers, in principle, much of the theory of natural resource extraction, a signif­icant body of theory in the 1960s and 1970s on resource depletion and growth; a large literature in the 1990s investigating the implications of endogenous growth theories; and a new and still burgeoning literature created in the last decade examining the relation­ship between pollution and national income levels.

While no review can settle the perennial debate over the limits to growth, this review hopes to play a positive role in moving the literature forward by identifying important unresolved theoretical questions, reporting on the results of recent empirical work, and providing an integrative assessment of where we stand today.⁵ To do so, we focus on three questions. These are: (1) what is the relationship between economic growth and the environment? (2) how can we escape the limits to growth imposed by environmental constraints? and (3) where should future research focus its efforts?

To answer these questions we start by introducing definitions and providing a pre­liminary result linking the environment and growth. We define the scale, composition and technique effects of growth on the environment, and then use these definitions to prove a useful but negative result on the limits to growth. We show that changes in the composition of national output - as occur when the economy specializes in relatively less pollution intensive services or relatively less natural intensive industries - can at

a primarily theoretical discussion of irreversibilities and hysteresis caused by nonlinearities see the sympo­sium edited by Dasgupta and Maler (2003). For related nonlinear theory see Dechert (2001). For case study evidence from prehistory see Brander and Taylor (1998).

⁴ See the classic book length treatments of renewable and nonrenewable resources by Clark (1990) and Dasgupta and Heal (1979).

⁵ Whether there are serious limits to growth is an unending controversy that reached its peak with the publi­cation of the Limits to Growth by Meadows et al. (1972). See the subsequent contributions by Solow (1973) followed by Meadows, Meadows and Randers (1991) and then Nordhaus (1992). best delay the impact of binding environmental constraints. In the long run, emission intensities must fall towards zero if growth is to be sustainable.

In many models this constraint is met through the substitution of clean inputs for dirty ones, in others via increased abatement, and in still others through some combination of technological progress and the other channels. This result is helpful to us because it allows us to distinguish between empirical regularities that are consistent with a short run growth and environment relationship (along a transition path) from those consis­tent with the long run relationship (along a balanced growth path). It also helps us sort through the literature by focusing on how a given model can generate what we take as our definition of sustainable growth: a balanced growth path with increasing environ­mental quality and ongoing growth in income per capita.^[475]

With our definitions and result in hand we then turn to present some stylized facts on the environment and growth. These facts concern the trend and level of various pollu­tants, and measures of the cost of pollution control. In many cases, the data underlying the construction of these facts is of limited quality; the time periods are sometimes insufficiently long to draw strong conclusions and the relevant magnitudes imprecise relative to their constructs in theory. Nonetheless, they are the best data we have.

Overall these data tell three stories. The first is that by many measures the environ­ment is improving at least in developed countries.

The second feature of the data is that pollution control measures have been both rel­atively successful and relatively cheap. While there are severe difficulties in measuring the full cost of environmental compliance most methods find costs of at most 1-2% of GDP for the U.S. Comparable figures from OECD countries support this finding.^[476]

The last feature of the data is that there is a tendency for the environment to at first worsen at low levels of income but then improve at higher incomes. This is the so-called Environmental Kuznets Curve. We first present raw emission data drawn from the U.S. and then briefly review the empirical literature on the Environmental Kuznets Curve that relies on cross-country comparisons. The raw data from the U.S. are unequivocal, while the cross-country empirical results are far less clear but generally supportive of the finding.

Having reviewed the relevant data and set out definitions we turn to a review of the theory. To do so, we develop a series of 4 simple growth and environment models. The models serve as a vehicle to introduce related theoretical work. For the most part we focus on balanced growth path predictions and eschew formal optimization taking as exogenous savings or depletion rates and sometimes investments in abatement. We do so because in many cases, these rates must be constant along any balanced growth path and hence we identify a set of feasible conditions for sustainable growth. Moreover the resulting simplicity of the models allows us to identify key features of fully developed research contributions already present in the literature.

The 4 models were developed to highlight the different ways we can meet environ­mental constraints in the face of ongoing growth in per capita incomes. In the first, which we dub the Green Solow model, emission reductions arise from exogenous tech­nological progress in the abatement process. Although this model is very simple it provides three key results. First, we show that even with the economy’s abatement intensity fixed, the dynamics of the Solow model together with those of a standard regeneration function are sufficient to produce the Environmental Kuznets Curve. The transition towards any sustainable growth path has environmental quality at first wors­ening with economic growth and then improving as we approach the balanced growth path. This is a surprising result. While numerous explanations for the EKC relationship have been put forward, this explanation is simple, novel, and quite general as it relies only on basic properties of growth functions.

Second, the Green Solow model forms a useful benchmark since this model predicts that a more strict pollution policy has no long run effect on growth. In true Solow tradi­tion, different abatement intensities create level differences in income but have no effect on the economy’s growth rate along the balanced growth path. This result provides par­tialjustification for the current practice of measuring the costs of pollution control to an economy as the sum of current private and public expenditures with no correction for the reduction in growth created. It also points out the stringent conditions needed for a stricter policy to cause no drag whatsoever on economic growth.

Third, the model clearly shows how technological progress in goods production has a very different environmental impact than does technological progress in abatement. Technological progress in goods production creates a scale effect that raises the emis­sions growth rate, technological progress in abatement creates a pure technique effect driving emissions downwards. In this first model both rates are exogenous, and as such they provide especially clean examples of scale and technique effects for us to refer to later. And as we show throughout the review, the presence or absence of technologi­cal progress in abatement is key to whether we can lower emissions, support ongoing growth, and provide reasonable predictions for the costs of pollution control.

The second model, which we dub the Stokey Alternative, was inspired by Nancy Stokey’s (1998) paper on the limits to growth. Here we present a simplified version to highlight the role abatement can play in improving the environment over time. The model we present focuses on balanced growth paths and not the transition paths as emphasized by Stokey, but nevertheless it contains two results worthy of note. The first is simply the observation that once we model abatement as an economic activity that uses scarce resources, increases in the intensity of abatement that are needed to keep pollution in check will have a drag on economic growth. Rising abatement creates a technique effect by lowering emissions per unit output, but also lowers pollution by lowering the growth rate of output.

By rewriting the model along the lines of Copeland and Taylor (1994) so that pollu­tion emissions appear as if they are a factor of production, it is now relatively simple to conduct growth drag exercises for the cost of pollution control in much the same way that others have examined the growth drag of natural resource depletion.^[477] By doing so, the model makes clear the limits to growth brought about by environmental policy.

The second feature we focus on is the model’s prediction of rising abatement inten­sity. In models with falling pollution levels, neoclassical assumptions on abatement, and no abatement specific technological progress, the intensity of abatement must rise continuously through time. For example, in Stokey’s analysis the share of output allo­cated to abatement approaches one in the limit. Since this share represents pollution abatement costs relative to the value of aggregate economic activity, models that rely on abatement alone tend to generate counterfactual predictions of ever-rising abatement costs. This is true even though ongoing economic growth is fueled by technological progress, and hence this result reinforces our earlier remarks about the importance of technological progress in abatement.

Our third model links the source and sink roles of nature by assuming energy use both draws down exhaustible resource stocks and creates pollution emissions that lower environmental quality. This “source and sink” formulation allows us to consider how changes in the energy intensity of production help meet environmental constraints. In this model, the intensity of abatement is taken as constant and there is no technological progress in abatement. Instead the economy lowers its emissions to output ratio over time by adopting an ever cleaner mix of production methods. As such the model focuses on the role of composition effects in meeting environmental constraints. We show that the economy is able to grow while reducing pollution because of continuous changes in the composition of its inputs, but this form of “abatement” has costs. Growth is slowed as less and less of the natural resource can be used in production.

This “source and sink” formulation is important in linking the earlier 1970s and 1980s literature focusing on growth and resource exhaustion with the newer 1990s literature focusing on the link between economic growth and environmental quality. We show that the finiteness of natural resources implies a constraint on per capita income growth that is worsened with higher population growth rates. This constraint is relaxed if the rate of natural resource use is slower as this implies reproducible factors have less of a burden in keeping growth positive. But sustainability also requires falling emissions, and this constraint is most easily met if the economy makes a rapid transition away from natural resource inputs as this reduces the energy and pollution intensity of output.

Putting the constraints from the source and sink side together, we show there exist pa­rameter values for which the twin goals of positive ongoing growth and falling emission levels are no longer compatible. This is not a doomsday prediction. Together with our previous analysis it suggests that abatement or composition shifts alone are unlikely to be responsible for the stylized facts. Technological progress directly targeted to lower­ing abatement costs (i.e. induced innovation) must be playing a key role in determining growth and environment outcomes. Therefore, in the remainder of the paper we turn to a model where technological progress in abatement is set in motion by the onset of active regulation and works to generate sustainable growth paths.

To highlight the importance of technological progress in abatement our final model draws on the analysis of Brock and Taylor (2003) by adopting their Kindergarten Rule model. While the previous models were useful vehicles to discuss the literature and describe possibilities, they were necessarily incomplete because they eschew formal optimization. Optimizing behavior is however important in discussions of the magnitude of drag created by pollution policy, and also important in discussions concerning the timing or onset of active regulation. The Kindergarten model provides two contributions to our discussion.

First, it shows how technological progress in abatement can hold compliance costs down in the face of ongoing growth. In contrast to the Solow model, there are ongo­ing growth drag costs from regulation, but as long as abatement is productive we can generate sustainable growth without skyrocketing compliance costs. By highlighting the important role for progress in abatement, the model points out the need to make endogenous the direction of technological progress as well as its rate.

Second, the model generates a first worsening and then improving environment much like that in Stokey (1998). Incontrasthowevertothe methods employed in the empirical EKC literature, we show that the path for income and pollution will differ systemati­cally across countries. This systematic difference leads to the model’s Environmental Catch-up Hypothesis relating income and pollution paths to countries initial income levels. Poor countries experience the greatest environmental degradation at their peak, but once regulation begins environmental quality across both Rich and Poor converges. Despite this, at any given income level an initially Poor country has worse environmen­tal quality than an initially Rich country. Moreover, since both Rich and Poor economies start with pristine environments, the qualities of their environments at first diverge and then converge over time. In addition to this cross-country prediction, the model also links specific features of the income and pollution profile to characteristics of individual pollutants such as their permanence in the environment, their toxicity, and their instan­taneous disutility. Together these predictions suggest a different empirical methodology than that currently employed, and expand the scope for empirical work in this area con­siderably.

The final section of our review is a summary of the main lessons we have drawn from the literature, offers suggestions for future research and briefly discusses some of the most important topics that we did not discuss elsewhere in the review.

2.