There is an ongoing debate in the literature about the veracity of a relationship between economic growth and environmental quality.  The debate stems largely from a paper by Gene Grossman and Alan Krueger called "The Environmental Impacts of a North American Free Trade Agreement" (1991)  The idea that economic growth leads to environmental quality improvements warrants a closer examination as it has a number of potentially important policy consequences:

1. Suggests that environmental quality will improve "spontaneously" so special care need not be taken of the environment at earlier stages of growth.

• impairment of a country’s ecosystem or depletion of natural resources in the present could preclude that country’s potential for a sustainable future.
• unable to meet its own basic needs, it could become a ward of the international system,
• engage in resource related conflicts internally or with neighbors.

Edward Barbier (1994) argues that: The preservation of natural resources in less developed countries (LDCs) is important because it is rare for capital obtained through depleting natural resources to be reinvested in a sustainable way. Usually this capital is diffused somehow or remains in the hand of a few wealthy citizens.

2. With the increase in global trade, wealthy countries might be better able to shift environmental degradation associated with consumption to poorer countries.

A related concern is that perhaps the types of environmental degradation performed by wealthy countries can be shifted to the future while environmental problems of poorer nations are more evident in the present, again hiding world-wide consequences.

3 Perhaps most importantly, if economic growth is seen to be the best means for attaining common social goals, policies and institutions will be structured to promote growth above all else.

• Yet, growth may not be the best means for reaching broadly desired social goals, and may actually be taking us further from these goals. Economic growth is often antithetical to improvements in environmental quality.
• In the long run, it may well be that continued economic growth is not even possible.

History of debate:  The debate regarding the merits of economic growth for society and for environmental quality has been evident since the late 1960s.

• In 1967, E.J. Mishan published a book entitled "The Costs of Economic Growth."
• In 1971, Nicholas Georgescu-Roegen published "The Entropy Law and the Economic Process".
• In 1972, Donella Meadows et al. published "The Limits to Growth".
• Herman Daly's 1977 book "Steady-State Economics".
• Other essays were also published during this period that focus on the limits to growth.

These books argue that economic growth is not sustainable.

Although these works were broadly read, they had little impact on government policy or global institutions such as The World Bank or the International Monetary Fund (IMF).

Then, in 1987, the United Nations published "Our Common Future". Rather than argue that continued growth is not possible, it argues for the importance of including environmental concerns in economic decision making. The overall message is that growth is fine, as long as it is done in a sustainable fashion.

This report did stimulate change in international economic institutions.

• Sustainable development became a catchword, and
• Social and natural scientists were hired to evaluate economic development programs to ensure that they did not proceed with unwanted environmental consequences leading to unsustainability.
• The report undercut the more serious concerns of earlier authors regarding the sustainability of growth itself.

As the world financial community was adjusting to the idea that unfettered economic growth is not sustainable, Grossman and Krueger published their 1991 paper which concludes that economic growth alone can be good for environmental quality.

So, Grossman and Krueger (1991) not only received a great deal of attention

• Contribution to the debate regarding the merits of increased trade;
• It turned the limit to growth debate on its head;
• Sustainability is no longer an issue because it will follow from growth once a certain income level is reached.

A cottage industry of academics working on the EKC has evolved.

The air pollution data they use are from the Global Environmental Monitoring System (GEMS), a cooperative effort of the World Health Organization (WHO) and the United Nations Environment Program (UNEP).

• The GEMS data are given in annual mean and median values for the 50^th and 95^th percentile of daily observations at each site for SO2, SPM (smaller than 10 mg) and Smoke (larger than 10MG). EXPLAIN
• Data were collected in a total of 42 countries from the years 1977 to 1991.
• The number of sites where it was collected varied over time.
• The average number of sites per city is three.

Given that the data they use are panel data, readings from multiple locations over multiple time periods, a simple ordinary least squares (OLS) model can not be used to analyze the data. Instead, Grossman and Krueger use two types of random effects models to examine the relationship between GDP per capita and urban air pollution concentrations.

The random effects model is used to take into account components of the error term that are common to a given year's measurements at different sites in the same city.

The first random effects model includes coefficients identified as having probable impacts on pollution concentrations. This model is designed to expose the relationship between GDP per capita and urban air pollution across countries.

The second model they use is the same as the first except that fixed site effects are taken into consideration by including separate dummy variables for each site in the sample rather than dummies for single attributes of the location such as residential or coastal. With the fixed effect model, if some site has specific attributes, such as being in a particularly windy spot or being near a factory, this will be picked up by the dummy variable for that site.

This model is designed to expose the relationship between GDP per capita and pollution concentrations over time holding the site constant.They use Cubic form

Regressions were performed for both the 50^th percentile and 95^th percentile concentrations of SO_2, SPM, and smoke at each site during the measured year.

Twelve regressions were performed: three pollutants, 50^th and 95^th percentile measurements, and two different models.

Of the twelve regressions, the EKC is supported by the results from four of these regressions: the 50^th percentile SO₂ data using both the RE and FE models; both the 50^th and 95^th percentile smoke using the RE model.

Only one component of EQ. not necessarily good proxy for EQ

Effect of increased income of diff types of EQ may depend on current income level of the country.

Perhaps an increase in urban air pollution corresponds with a decrease in indoor air pollution, with a corresponding decrease in human health risks.

Quality of data.

There are a number of problems with using the GEMS database that are significant but practically unavoidable because of the limitations with international environmental data.

• Placement of monitors
• Functioning of monitors
• Collection methods

Study comparing the results from the different laboratories found that only 75% of the results obtained were acceptable.

GDP as proxy for income.

They had little choice but to use GDP since few other measures are available.

• Poor measure of wealth
  • does not take into account loss of resources
  • miss certain economic activities,
  • while measuring some economic activities that do not truly reflect positive growth.
• Poor measure of quality of life: relevant to claims about the demand for environmental quality being related to income.
• National measure: Urban income levels are generally higher that rural income levels.

Just because a rich country is cleaner than a poor country does not mean that the poor country will also get cleaner as is becomes wealthier.

The current development paths for poorer countries now are different than they were for the developed countries for a number of reasons.

Cross-country analysis may also miss the facts that wealthy countries may be cleaner because they have superior environmental endowments in the first place, or that they are exporting pollution industries to poorer countries.

A fixed effects model's results do not depend on cross-country comparison, and so are a more reliable indicator of change over time within each country.

Can not separate direct from indirect effects of growth.

It could be that there are intervening factors that lead to the apparent relationship between GDP and air pollution. Promotion of these other factors could lead to decreases in air pollution irrelevant of changes in GDP.

Regression analysis

An additional concern with the Grossman and Krueger study is the quality of their data analysis.

• They fit the data to a cubic form.
• R² is small in many cases.
• Not corrected for autocorrelation.
• Do not test for the acceptability of using a random effects model.

Data Analysis

I obtained data - tested only SO₂

With small changes to the data I was given, I was able to closely, though not exactly, replicate G&K's results.

Hausman Test - the random effects model can not by relied upon to give consistent results.Difference in R².

I ran the same regression using a dummy for each site. It yields the same regression coefficients, yet the reported R² is 0.71, rather than 0.07.

Tests of Robustness

Data Subsets 1. I created a subset by removing China from the data set. There are 152 observations for China, more than for any other country, so I hypothesized that this one country could have an important impact on the regression results.

My hypothesis was correct.

• FE regression results of cubic income become insignificant if China is dropped from the data base.
• RE results are still significant without China, however, once again, the Hausman test rejects the RE model in favor of the FE model.

2. I dropped the data that reported zero concentrations of SO_2. The zero measurements seem implausible given that other sites in the same city have high concentration measurements for the same years. Due to different measuring and analysis techniques discussed above, the zero measurements may well be erroneous. I found again that the regression results are insignificant if reported zero concentrations of SO₂ are dropped.3. A fixed effects regression can also be performed with the data split into two groups:

• Countries with GDP per capita less than $4500 and more then $4500;
• At income level less than and greater than $7000;
• Alphabetically.

By splitting the data in various ways, I was generally not able to replicate the results obtained with the entire data set, which demonstrates the limited robustness of Grossman and Krueger's results.

Annual Mean Value SO₂ Concentrations

Additionally, although Grossman and Krueger looked at 50^th and 95^th percentile daily means of measured SO₂ levels, the GEMS data also provides the annual mean measurement for each site.

The RE regression's findings for annual mean are similar to those for 50^th and 95^th percentile, however the Hausman test once again rejected this model at the .0001 level. The FE results are not significant at the .10 level.National Averaged pollution level

I did this for two reasons:

1. The GEMS data for air pollution concentrations are unbalanced

2. The second reason is that GDP is a national level variable

I also calculated population-weighted national average pollution levels.The only significant results were with 95^th% SO2. I found a U-shaped relationship rather than an inverted U. The trough is at approximately $6000 per capita per year. The peak is not until approximately $350,000 per capita per year.

The results of a quadratic model FE regression using 50^th percentile SO_2, 95^th percentile SO₂ and mean concentrations are not significant. i.e. the data does not hold up to an alternative model form that is equally viable given the proposed relationship. Summary – looks like they were data mining.

Data Analysis with the AIRS Data SetI also tested the relationship between EQ and EG using an expanded and updated GEMS data set called the Aerometric Information Retrieval System (AIRS).

The variables provided by AIRS are slightly different than those provided by the GEMS database.

• AIRS data does not give 50^th and 95^th percentile measurements for each site each year. It just gives the mean annual measurement, no method of measurement

The AIRS database expands upon the GEMS data by adding the more recent year of data, It also revises some of the data for the years 1977 to 1988.

• Smoke results support an inverted U-shaped relationship between income per capita and the pollutant studied, as well as a turning point within the range found by Grossman and Krueger. However, at approximately $10,000 per capita per year, smoke concentrations start to rise again.

• For SO₂, the results suggest a U-shaped relationship, rather than an inverted U-shaped relationship, with a trough preceding the peak pollution level. The peak of approximately $17500 is at an income level far above most countries per capita income, and at a level that many countries may never reach.

• For SPM, the results of the cubic form are not significant.

Given the insignificant of the regression coefficients using a cubic form GDP for SPM, I tried a regression using squared GDP

• The results of the FE model were significant for both coefficients at the .05 level.
• They suggest an inverted U with a turning point at $16,920.

• The broadly cited "evidence" of an EKC found by Grossman and Krueger in their 1991 paper is based primarily on their findings for smoke and SO₂ concentrations.
• Further examination of these results has shown that the RE model results are inconsistent and therefore only the 50^th percentile SO₂ results support an EKC.
• However, the low R² of these results suggest that the relationship between SO₂ and income is weak at best.
• Tests for robustness of these results by splitting the data, using alternative measures for SO₂ concentrations, using national-averaged data, and using an alternative model form did not support the EKC hypothesis.
• Regression analysis with an expanded data set did not support an EKC.
• Other empirical studies have found mixed results.

Conclusion: It is difficult to say anything definitive about the relationship between environmental quality and income level.

Given the weak empirical evidence of an EKC, Why Has Link Between Income and Environmental Quality Been So Readily Accepted?

The mechanism most commonly cited as a link between income level and environmental quality is changing demand.

Although it may be true that wealthier people care more about some aspects of environmental quality, such as endangered species and parkland, many aspects of environmental quality are issues of livelihood and health. Certainly poor people care about loss of fisheries or their children dying from diarrhea or asthma. For the poor, environmental problems can be related to survival. Here, the willingness to pay in infinite, the problem is the capacity to pay.

Additionally, since the poor are less likely to be able to escape environmental degradation through purchasing bottled water or spending weekends in a country home for example, there are reasons to expect they might care more about environmental quality than wealthy people.

Just because a poor person may be too preoccupied with feeding his or her family to take actions to protect environmental quality does not mean that this person does not care about environmental quality.

A number of surveys have been conduced over the years that examine attitudes towards environmental quality in countries around the world. One of the better know surveys is The World Value Survey, which examines attitudes towards various social issues including environmental issues. These find that the Rank of Envir Concern not same as GDP rank.

A 1992 Gallop poll finds that citizens of low to middle income countries were generally more concerned about environmental problems than citizens of advanced industrial counties.

A 1989 Harris poll finds that, not only do citizens of poorer countries care as much or more about environmental quality, citizens from poorer countries as compared to richer countries are willing to pay almost as much to reduce environmental problems and are willing to put in more volunteer hours towards reducing environmental problems than citizens of wealthier countries.

Surveys are notoriously faulty, however, these three surveys bring into question the changing demand thesis.Case Studies

Based on Philippine case studies, Robin Broad asserts that degrading activities by the poor are largely dependent on whether or not they are in a position to control environmental degradation, even if they do care about protecting the environment.Broad writes: "Grossman and Krueger hypothesized that enrichment brings environmental values and motivations. While under certain circumstances this may be true, our research suggests that looking at growth and poverty alleviation as a means to instill environmentalism misses a key point: in the Philippines environmentalism was a demand of the poor, not the rich."

Structural and technological change

With economic growth, the structure of a nation's economy can change in ways that have repercussions for environmental quality.

SCALE EFFECT.

The scale effect is associated with a negative relationship between economic growth and environmental quality at all income levels.

COMPOSITION EFFECT.

Changing composition of output. Increased wealth is associated in the poorest countries with a shift in composition of GDP from agriculture and fishing to industries, and in wealthier countries from industry to services. This mechanism suggests that countries will have different environmental problems at different levels of development.

The EKC theory suggests that this leads countries to have better environmental quality at the extremes of the income scale. The changing composition of output effect could lead to an EKC for certain environmental problems, such as industrial pollution.

A third way is that as countries grow wealthier they may also alter their industrial composition to focus less on polluting industries like steel manufacturing and more on cleaner industries such as computer industries. This would also lead to an EKC for industrial pollution if above a certain income level countries switched to cleaner industries.

TECHNIQUE EFECT

This mechanism creates a positive correlation between economic growth and environmental quality.

References