SEARCH

SEARCH BY CITATION

Keywords:

  • Aggregate Index Method;
  • Euclidean Distance Method;
  • Partial Least Squares Variable Importance in Projection;
  • United Nations Millennium Development Goals;
  • sustainability;
  • sustainability indicators

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESEARCH METHODOLOGY: THEORETICAL BASIS FOR CALCULATING ENVIRONMENTAL SUSTAINABILITY OF SELECTED OECD AND NON-OECD COUNTRIES
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. ACKNOWLEDGMENTS
  8. LITERATURE CITED
  9. Supporting Information

The objective of the article is to show that indicator aggregation coupled with the use of a multivariate statistical analysis applied to the environmental sustainability indicator data of the UN Millennium Goals (MDG) can allow important inferences to be made. Goal 7 of the UN MDG aims to assess environmental sustainability by ten indicators, the dominant among them being emissions of carbon dioxide equivalent. The data is available over a period of 20 years. We used an aggregate index method to combine the various indicator values into a single index, and studied the trends of countries over the years. We have also compared among two arbitrarily chosen groups of OECD and non-OECD countries, and shown how these countries have performed among themselves and as a combined group over the time period. Using the multivariate statistical method of Partial Least Squares Variable Importance in Projection (PLS-VIP), we showed how to find the key indicators from the larger set and their relative importance over the years. Important results from our research include ranking of countries according to their overall environmental performance in a particular group. The PLS-VIP method showed that certain indicators have shifted in their importance in influencing their overall environmental performance, while others have remained relatively insignificant over the 20-year period. 2014 American Institute of Chemical Engineers Environ Prog, 2014 © 2014 American Institute of Chemical Engineers Environ Prog, 34: 198–206, 2015


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESEARCH METHODOLOGY: THEORETICAL BASIS FOR CALCULATING ENVIRONMENTAL SUSTAINABILITY OF SELECTED OECD AND NON-OECD COUNTRIES
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. ACKNOWLEDGMENTS
  8. LITERATURE CITED
  9. Supporting Information

The Millennium Development Goals (MDG) of the United Nations aims to achieve eight international development targets by the year 2015. The MDG and temporal targets come from the Millennium Declaration, signed by 189 countries, including 147 heads of State and Government, in September 2000 and from a further agreement by member states at the 2005 World Summit Resolution adopted by the UN General Assembly [1]. The agreements represent a partnership between the developed countries and the developing countries “to create an environment—at the national and global levels alike—which is conducive to development and the elimination of poverty.”

Following the conceptual framework of sustainable development as comprising three dimensions, i.e., environmental, economic, and societal, the following eight goals were set for MDG to track progress or regress of the nations in sustainability terms using indicators. These goals were: (1) eradicate extreme poverty and hunger; (2) achieve universal primary education; (3) promote gender equality and empower women; (4) reduce child mortality; (5) improve maternal health; (6) combat HIV/AIDS, malaria, and other diseases; (7) ensure environmental sustainability; and (8) develop a global partnership for development. The website for the MDG Indicators presents the official data, definitions, methodologies, and sources of 80 indicators for monitoring the progress of 234 countries in achieving the MDG goals for more than 20 years [1].

The raw data obtained from the website is large and unwieldy, when considered in the context of making a definitive overall sustainability statement of a chosen country, either at a snapshot in time, or over time. The MDG program has not provided any method for making that assessment, thus allowing differing inferences. The program, however, had listed some priority targets, such as global poverty reduction by 50% by 2012. The records on these priorities tell stories of success or failure, while allowing for learning lessons from the results and fashioning new targets or newer approaches to reach them.

Taking the entire dataset for all countries over all the years would be an onerous task for analyzing the sustainability trajectory of the nations over time, despite the obvious desirability of such an objective. With 25 of the countries, we decided to do a pilot analysis of a smaller set of data using Goal 7, i.e., environmental sustainability, represented by a total of 10 indicators as shown in Table 1. Such a sustainability analysis will set a precedent for extending the application of the method to the larger dataset. We considered the entire 20-year time period from 1990 to 2009 for which complete data are available.

Table 1. Environmental Sustainability goals and targets for MDG, assumptions, and data availability.
Goal 7: Ensure environmental sustainabilityAssumption on higher values of data for sustainabilityAbbreviation used in this workData availability
Target 7.A: Integrate the principles of sustainable development into country policies and programmes and reverse the loss of environmental resources7.1. Proportion of land area covered by forestGood Sparse
7.2. CO2 emissions, total, per capita and per $1 GDP (PPP)BadCO2/$GDP, CO2/Capita, CO2, Thousand Metric TonsGood
7.3 Consumption of ozone-depleting substancesBadODP, Metric TonsGood
Target 7.B: Reduce biodiversity loss, achieving, by 2010, a significant reduction in the rate of loss7.4. Proportion of fish stocks within safe biological limitsGood No data
7.5. Proportion of total water resources usedBad Sparse
7.6. Proportion of terrestrial and marine areas protectedGoodTerr. And Mar. Area Not ProtectedGood
7.7. Proportion of species threatened with extinctionBad No data
Target 7.C: Halve, by 2015, the proportion of people without sustainable access to safe drinking water and basic sanitation7.8. Proportion of population using an improved drinking water sourceGoodPop. Not Using Imprv. Drinking H2OGood
7.9. Proportion of population using an improved sanitation facilityGoodPop. Not Using Imprv. SanitationGood
Target 7.D: By 2020, to have achieved a significant improvement in the lives of at least 100 million slum dwellers7.10. Proportion of urban population living in slumsBad Sparse

Two important objectives are the focus of this article. First, from the indicator data for a country, we proposed to calculate an aggregate index, De for a chosen year. From an intercomparison of the De values, we can infer on the relative environmental sustainability of the countries for a given year. With the De data represented as a function of time, we can conclude whether the chosen countries are moving over time to a better environmental sustainability position with respect to themselves or not. The anticipated outcome of this effort is to enable an analyst in making statements on the trend of environmental sustainability status of the selected countries over time. We detailed the method for De calculation using two simple case studies in a recent publication [2]. Second, we need to ascertain whether there exists a set of indicators that has more significance for making these relative environmental sustainability inferences. In this article, we have analyzed the influence of the different indicators on the aggregate index using a Partial Least Square Variable Importance in Projection (PLS-VIP) method. This approach allowed us to determine the sufficiency of the indicator set in making sustainability assertions and determines the relative rank of the indicators. Additionally, we use the time series data to show the trend for sustainability improvement (or lack thereof) over time. We used this method for a similar purpose for determining relative sustainability of industrial products and processes in a recent publication [3]. That was a static analysis with no temporal domain, unlike this work. A physical observation of the raw dataset is warranted before we progress to computing the De or the VIP scores.

There are methods in the literature for selecting environmental indicators for countries [4]. Collection of environmental data and constructing composite indicators has been done by methods other than the UN MDG effort. For instance, the US Government has an ongoing effort to track the state of the environment through the use of indicators [5, 6]. While the latter does not focus on sustainability, the former focuses on the art of choosing indicators. We take the MDG indicators for an analysis because the data are public, and UN has spent significant resources to collect data for tracking progress. Methods of ranking of countries have been previously critically assessed by Freudenberg [7]. The Environmental Sustainability Index is a well-known comparative index of national-level environmental sustainability [8]. Also, Phillis et al. [9] used the method of Sustainability Assessment by Fuzzy Evaluation (SAFE) to rank the countries for sustainability by several different indicator aggregation formulae. No attempt in these previous works was made to either critically evaluate the statistical necessity of all the indicators used or to determine the relative importance of the indicators in sustainability analyses and conclusions. Our methodology allows the user to (i) provide a sound method to normalize and compute an aggregate index for comparing relative sustainability of competing systems, (ii) rank the given set of indicators using PLS-VIP, according to their contribution to the aggregate index, and allowing to focus on the most needed areas of sustainable development. The MDG environmental sustainability indicator dataset was chosen because of (i) the ready availability of validated open source data, (ii) variety of options for systems (in the form of countries), and (iii) variety of indicators to measure the progress of countries toward fulfilling the MDG goals. The authors are not responsible for the collection and validation of the MDG indicators; it is a set of values already computed by a reputed organization of international importance (UNEP). We are not aware of any effort by the UN to assign weighting factors to their indicators. We therefore decided to base our analysis and draw conclusions on assigning each indicator the same default weighting factor of 1.0, which means no preference for any indicator.

RESEARCH METHODOLOGY: THEORETICAL BASIS FOR CALCULATING ENVIRONMENTAL SUSTAINABILITY OF SELECTED OECD AND NON-OECD COUNTRIES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESEARCH METHODOLOGY: THEORETICAL BASIS FOR CALCULATING ENVIRONMENTAL SUSTAINABILITY OF SELECTED OECD AND NON-OECD COUNTRIES
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. ACKNOWLEDGMENTS
  8. LITERATURE CITED
  9. Supporting Information

MDG Environmental Sustainability Data Analysis

To illustrate the application of our method for aggregate index, we created two arbitrary groups comprising a smaller set of 14 OECD and 11 non-OECD countries from the total of 234 reported in the MDG dataset. The purpose was to observe the comparative overall environmental sustainability performance of the members in each group.

A critical examination of the environmental sustainability indicators as shown in Table 1 would lead one to conclude that the number of suggested indicators is too few, and heavily biased in favor of global warming issues. Based on the availability of data, data quality is denoted as good (available for all the years), sparse (available for 5 or less years), or no data. The indicators of goal 7.2 (CO2 emissions) were calculated using two methods, UNFCC and CDIAC, resulting in a total of six indicators for the measurement of this goal. The CDIAC method was used by all the selected countries in our dataset, and chosen for our analysis. Indicator 7.3 (O3 depletion) was reported by most except the OECD countries belonging to the European Union in our dataset. Based on the complete availability of data, we carried out our analysis with the indicators 7.2, 7.6, 7.8, and 7.9 for OECD countries, and 7.2, 7.3, 7.6, 7.8, and 7.9 for non-OECD countries. For this article, we have abbreviated the indicator names as shown in Table 1.

Table 2. Number of observations out of 20 years having VIP scores above 1 and 0.8.
  CO2/$GDPCO2/CapitaCO2, Thousand Metric TonsODP, Metric TonsPop. Not Using Imprv. Drinking H2OPop. Not Using Imprv. SanitationTerr. And Mar. Area Not Protected
Greater than 1OECD20201409
Non-OECD60201819136
All160020208
Greater than 0.8OECD2018420219
Non-OECD1602020201812
All19019202020

Unidirectional Data

For making sustainability inferences, a country having a smaller De value is considered more sustainable, in accordance with the convention that each indicator is fashioned in a way in which lower numerical values are more desirable than higher ones. To ensure this, an effective method for making all the indicators unidirectional is necessary. This can be done in two ways. The first method is to design an indicator with attributes that make lower values better and higher values worse. In the second method, if an indicator already carries data, an efficient method is established to transform the indicator. For the MDG dataset, the assumptions for data direction are shown in Table 1. We changed the direction of indicators 7.1, 7.6, 7.8, and 7.9 to represent the opposite as better by subtracting the existing values from 100%. If an indicator value is not given in percentage, we can transform it into a new indicator where an inverse of the original value can be used for complying with the convention. In this article, the indicators after transformation contain the word “not” in their description.

Aggregate Index, De

The first step of calculating the aggregate index involves constructing a synthetically chosen reference using the minimum values of the indicators in the dataset. Then, all the indicator values in the dataset are normalized with respect to the synthetic set of indicators and the Euclidean distance computed as shown in Eq. (1) [2].

  • display math(1)

where De, the measure of relative environmental sustainability, is the Euclidean distance of a chosen country Yj, at a point in time obtained after normalizing with respect to a synthetic reference country X0. The idea of the arbitrary reference country in De calculation is to transform the dataset to avoid occurrence of negative data points in the transformed data. The weighting factor cj allows use of weighting preference (usually a societal choice) of any of the indicators in comparison to others. Taking the example of the MDG dataset, the three sets of countries are represented by XOECD (m = 14, n = 6) and Xnon-OECD (m = 11, n = 7) and XAll 25 (m = 25, n = 6) respectively, where m is the number of countries and n is the number of indicators used in the analysis. While considering time series data on countries, as are considered here, the datasets can be represented by a m×n×t matrix, where t is the number of temporal points, and a De value is computed for each country and each year.

Partial Least Square Variable Importance in Projection

PLS-VIP is a multivariate regression method where information from a data space of a larger number of variables is projected into that with a smaller number of variables. We start with the same data matrix, XOECD, Xnon-OECD, and XAll 25 where the indicators are the variables in the data space. PLS-VIP is a supervised model where an overall pattern of the datasets is required. This overall pattern, also known as the response vector (can also be a matrix), is provided by the aggregate index, De. The number of indicators is reduced in a way such that variations in the values of the indicators are most likely to be reflected in the response vector composed of the De data.

An application of PLS-VIP for sustainability assessment has been given by Mukherjee et al. [3]. Variables which are inferred through a mathematical model of the original (observed) data are known as latent variables. In contrast to the original variables, the latent variables are not explicit, nor can be tweaked at will. In this article, the variables are the indicators, and henceforth, we will refer to the original variables as indicators, and the latent variables derived from the original indicators as latent indicators. PLS Regression model is used to decompose the original data matrix into two orthogonal matrices, the loadings (L) and scores (T) of a number of latent indicators, and a residual matrix, E as shown in Eq. (2) [10].

  • display math(2)

The score matrix T is related to the response vector De through a regression matrix b as shown in Eq. (3) [11]. F is the residual vector of De.

  • display math(3)

The indicator set for each country is represented as an option xm. Each option vector xm (m is the total number of countries) can be related to the score vector through weight vectors wj as given in Eq. (4).

  • display math(4)

The VIP for a particular indicator is calculated using the regression coefficient b, weight vector wj, and score vector tj as given in Eq. (5).

  • display math(5)

where wkj is the kth element of the vector.

PLS-VIP is used to identify the importance of each indicator in affecting the aggregate index De. The normalized indicators for calculating De is used in VIP calculation. Indicators with lower VIP scores have little influence on De, and those with the higher VIP scores contribute the most toward De. The average of squared VIP scores equals 1. VIP score greater than 1 is generally used as a criterion for detecting the relative importance of an indicator.

There are different algorithms available to solve PLS regression problems. In our previous work [3], we used an Eigen value decomposition algorithm. In the regression process of the Eigen value decomposition algorithm, the Eigen vector (or principal component) corresponding to the first Eigen value of the matrix XTDeDeTX is used to obtain the first extracted score t1 for the first latent indicator [12]. Once the first latent indicator has been extracted, the residuals from X and De are used to extract the second latent indicator and so on. In the present work, the regression for the PLS is based on PLS1 algorithm [13]. In PLS1 algorithm, the correlation coefficient of X and De is used to obtain the first extracted score t1 for the first latent indicator. After obtaining the first latent indicator, the regression follows by obtaining the second latent indicator from the residuals and so on. This is most appropriate for our problem where response matrix comprises of one column, constituting the aggregated indices, De. Regression coefficient and weight vectors from the first three latent indicator score vector are used for calculating VIP. Both methods produce similar results, PLS1 being more direct.

The code for calculation of De and PLS-VIP scores is written in MATLAB® and available upon request from the authors.

RESULTS AND DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESEARCH METHODOLOGY: THEORETICAL BASIS FOR CALCULATING ENVIRONMENTAL SUSTAINABILITY OF SELECTED OECD AND NON-OECD COUNTRIES
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. ACKNOWLEDGMENTS
  8. LITERATURE CITED
  9. Supporting Information

As per the previous method description, the aggregate index De is computed for XOECD, Xnon-OECD, and XAll 25 sets of countries. The VIP scores computed from the PLS-VIP analysis shows the relative importance of indicators in affecting De.

Aggregate Index Comparison for OECD, Non-OECD, and All 25 Countries

Relative environmental sustainability performance for the time period 1990–2009 of the selected countries in three groups, XOECD, Xnon-OECD, and XAll 25, is shown in Figures 1-3 respectively. Two observations are derived from the information conveyed in these figures. First, the relative environmental sustainability of the countries can be obtained from relative rankings of the countries in terms of De for a given year. Second, the decrease or increase in De over the years conveys the improvement or worsening of environmental sustainability for a particular country. As illustrations, we make the following observations in the following paragraphs. The values in the figures were sorted according to decreasing value of De for the year 2008. It needs to be made clear at the outset that these De numbers are not to be treated as absolute. These numbers will change if more countries are added to group, if more indictors were used for the assessment, also if indicator weighting factors are used. Thus these numbers provide an inter-comparison based on the cohorts we chose and the indicators we used.

image

Figure 1. De for 14 OECD countries calculated with 6 environmental indicators. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Figure 1 shows that among the OECD countries, France and Germany had the lowest De values over the time period, implying their performance was the best in the group in terms of environmental sustainability as derived from the MDG indicators. United States had the highest De, implying poorer environmental sustainability. The numerical values of De remained almost constant for all the countries over the years except for the Republic of Korea, where the environmental sustainability deteriorated, denoted by increase in De, from 1995 to 2000, and improved in subsequent years with steady decrease from 2004 to 2008, and increased again in 2009.

Regarding the position of the United States in the OECD group of countries, two observations are warranted. First, the MDG Goals for environmental sustainability rely heavily on carbon dioxide emissions. United States registered the highest CO2 emissions of any OECD country in the time period, and continues to remain that way to date. Energy being the prime driver of development as expressed in GDP terms, this result is hardly surprising given that the United States is a quarter of the world's GDP and emits roughly a quarter of global carbon dioxide emissions. Second, in recent years because of conservation and energy efficiency gains, the US carbon dioxide emissions have shown modest decrease. Thus even by MDG standards, environmental sustainability of the US is seen to be improving, suggested by the decreasing trend in De values from 2006 to 2009. It now is evident that this trend has continued to date [14]. Since peaking in 2007, by 2012, US CO2 emissions have declined by 24%. This is undoubtedly a result from energy conservation and the revolutionary hydraulic fracturing technologies causing a shift to electricity generation by natural gas in preference to coal. This enables overall CO2 emission reduction, even when the non-fossil energy efforts may suffer as a result.

Figure 2 shows that among the non-OECD countries, Brazil has significantly improved in environmental sustainability, with De decreasing over time. China has the highest De value for all the years, suggesting that it has the worst environmental sustainability, again as deduced from the MDG indicators for environmental sustainability. Another significant finding is the increase of De over time for Thailand and Malaysia, suggesting declines in environmental sustainability in these countries, which can be attributed to their industrial development.

image

Figure 2. De for 11 non-OECD countries calculated with 7 indicators. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

image

Figure 3. De for 14 OECD and 11 non-OECD countries calculated with 6 environmental indicators. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

When all the countries are considered, as shown in Figure 3, the lowest values of De are obtained by Germany, and in recent years, by Brazil. The observation period followed West Germany's merger with previous socialist East Germany with dismal environmental stewardship record. Many inefficient East German factories were shuttered after the merger providing significant emission reductions [15]. Additionally, Germany aggressively invested in non-fossil energy production, primarily in heat and power, rather than transportation for which it relied more on diesel than gasoline, as in the United States. Brazil's superior showing is a result of its heavy reliance of hydropower and a significant reliance on bioethanol from sugarcane as a component of transportation fuels. In a recent study [16], it has been shown that bioethanol from sugarcane, as practiced in Brazil, is arguably the most sustainable of all transportation fuel options, fossil or non-fossil.

The De values for almost all of the countries show a dip in 2002 and 2008, and are most prominent for China. This can be attributed to the peak periods of recession, when the use of fossil fuels dropped significantly resulting in lower carbon dioxide emissions and thus, better environmental sustainability.

Determination of Important Indicators for OECD and Non-OECD Countries

The PLS-VIP method was used to compute the relative importance of indicators for the three sets of countries. Here, we have studied the order in which indicators are important in determining the environmental sustainability of the countries. Observations are made regarding the VIP score above 1 and 0.8 as shown in Table 2, and the increase or decrease of the VIP score of a particular indicator over years. We did not attempt to rank the indicators in any order, because they have changed in relative importance over the years, but such a ranking can be done for a given year if necessary.

Figure 4 shows the VIP scores of indicators for the chosen OECD countries over the 20-year period. For these OECD countries, six indicators were used to compute the De. “CO2/$GDP” was the most important indicator in all the years till 2008. In 2009, “Pop. Not Using Imprv. Drinking H2O” became the most important indicator. From Table 2, we can see that VIP scores above 1 was obtained by “CO2/$GDP” for all the years, “Pop. Not Using Imprv. Drinking H2O” for 14 years, followed by “Terr. And Mar. Area not Protected” for 9 years. However, VIP scores above 0.8 was obtained by “CO2/$GDP” and “Pop. Not Using Imprv. Drinking H2O” in all the years, closely followed by “Terr. And Mar. Area not Protected” and “CO2/Capita.” This shows that “CO2, Thousand Metric Tons” and “Pop. Not Using Imprv. Sanitation” has been relatively unimportant in the years for describing the environmental sustainability of the OECD countries.

image

Figure 4. Important indicators for OECD countries. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Figure 5 shows the VIP scores of the seven indicators used to compute the De for the chosen non-OECD countries. Here, we see that the indicator, “CO2, Thousand Metric Tons,” has been an important indicator with VIP score above 1 for all the years in consideration. “Pop. Not Using Imprv. Drinking H2O” and “ODP, Metric Tons” have also been important indicators with VIP scores above 1 for more than 18 years. “Pop. Not Using Imprv. Sanitation” has been an important indicator with VIP score above 1 till 2002, and its value has declined since then and never reached 1 suggesting the loss in importance of the indicator. A significant increase in importance can be seen for the indicator, “Terr. And Mar. Area not Protected,” where it has consistently attained VIP scores of above 1 since 2004. A VIP score above 0.8 was obtained by all of the indicators for most of the years, except for the indicator, “CO2/Capita.” This indicator has never been relatively important in the indicator set to define the overall sustainability of these countries.

image

Figure 5. Important indicators for non-OECD countries. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Figure 6 shows the VIP scores of the six indicators used to compute the De for all 25 countries. VIP score above 1 is obtained by “Pop. Not Using Imprv. Drinking H2O” and “Pop. Not Using Imprv. Sanitation” in all the years, and these have remained two of the most important indicators among the six. “CO2/$GDP” has VIP score above 1 for 16 years. VIP score above 0.8 is achieved by all the indicators, except “CO2/Capita” suggesting that it is never relatively important in all the years of consideration. The indicator “Pop. Not Using Imprv. Sanitation” has steadily lost its importance since 1998, as more countries move toward achieving this goal.

image

Figure 6. Important indicators for all countries. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

The PLS-VIP method also shows that the relative importance of indicators for groups of countries (OECD and non-OECD) are different, implying that the importance varies with a country's environmental focus. For example, from Table 2, we see that the “CO2, Thousand Metric Tons” is not an important indicator for OECD countries, but an important indicator for non-OECD countries. The indicator has a score of above 0.8 showing its importance when all the countries are compared. “CO2/Capita” is an important indicator for OECD countries when we consider VIP scores above 0.8, but a relatively unimportant indicator when non-OECD and all countries are considered. The indicator “Pop. Not Using Imprv. Sanitation” is relatively unimportant for OECD countries but mostly important for non-OECD countries, and it is one of the most important indicators when all countries are compared.

CONCLUSIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESEARCH METHODOLOGY: THEORETICAL BASIS FOR CALCULATING ENVIRONMENTAL SUSTAINABILITY OF SELECTED OECD AND NON-OECD COUNTRIES
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. ACKNOWLEDGMENTS
  8. LITERATURE CITED
  9. Supporting Information

The PLS-VIP analysis helps us see trends in the indicators, and thus the variation in its performance over time in defining the sustainability of countries. It is important for future sustainability analysis to ascertain when certain indicators have lost or gained their importance. Also, this method allows us to determine the relative importance of indicators, especially when dealing with large set of indicators. Visual representation of VIP scores over the years, with the threshold of scores above 1 (or 0.8) as most important indicators, allow a simple method of analysis of data.

The available MDG data were analyzed for completeness and any missing data points for a 20-year time period was extrapolated and populated before further analysis. Out of the 15 indicators for environmental sustainability, we could use only 6 indicators for OECD countries and 7 indicators for non-OECD countries. We conclude that the indicator sets are too heavily biased with CO2 indicators for any effective environmental sustainability analysis. We understand that the United Nations has agreed upon this set of indicators to represent environmental quality of nations, and our suggestion would be to include further environmental indicators to represent other aspects of environmental sustainability. Some of these new indicators may include acidification potential, smog formation potential, eutrophication potential, fresh water depletion, energy used per unit of GDP, natural resource depletion, etc. [17]. A wider variety of environmental indicators will allow us to have an unbiased analysis when all countries are considered. Also, we have calculated the aggregate index without pre-assigning any weighting factor to any indicator. We understand that this is a possibility for consideration in future research, and the aggregate index has provisions of including the weighting factors for decision makers. However, the weighting factors should be carefully chosen, and the PLS-VIP method can serve as a guideline for making such a choice.

Analyzing the MDG data for the 20-year period, we see that the relative overall environmental sustainability stance has not significantly changed for the countries individually. The exceptions to this observation are Brazil and Singapore in the dataset of all countries considered together, where their De values significantly decreased over the years, suggesting their movement toward sustainability. The rapid industrialization of China and its poorer overall environmental sustainability has been well captured by our analysis using the aggregate index, De. Also, the improved environmental quality for Germany can be seen by the lowest value of De in the analysis with 25 countries. The choice of the cohorts of countries in our analysis has been entirely arbitrary, to illustrate the validity of the method of data analysis. The method is amenable to the inclusion of all countries. The relative rankings of the countries might change when a different set of countries or all the 234 countries are considered together, and if all the indicators have well populated values.

The PLS-VIP method was effective in ranking the indicators in the order of their contribution to De, for all the countries considered together. This is particularly useful for countries to focus on a certain indicator over another, and aim for achieving lower values of that indicator so that the value of De for that country is reduced. The differences that were observed using various sets of developed (OECD) and developing (non-OECD) nations clearly show that when the set group changes, the relative importance of an indicator also changes.

A reduction in the number of indicators may be possible using the analysis of PLS-VIP; however, in the particular case of the MDG indicators, we had to intuitively reduce the dataset and discard certain indicators for lack of completeness of data. It can be stated that the De calculated using the number of important indicators (having a VIP score above 1) gives the true ranking among countries, and should be used if that is a primary goal for a decision maker. This helps in eliminating the indicators which have low contribution to De.

Our analysis establishes that sustainability is a continuous path toward the betterment of a particular nation; at the same time it is highly dependent on the basis against which one analyzes the same. The choice of a reference country having all the least values in all the indicator categories indicate that barring unknown trade-offs, these indicator values are achievable. We have seen when countries progress toward achieving the lowest value of an indicator, the indicator automatically loses its importance. Naturally, when such changes in an indicator value are realized, other indicators gain importance, showing that their contribution toward the De increase. Thus, a continual path toward sustainability is possible when we analyze the combined effect of indicators, and the importance of indicators over time. The use of PLS-VIP together with an overall measure such as De removes the doubt that the use of an aggregate index loses underlying information. As has been seen in this study, no such loss of information exists when all the indicator performances are tracked by the PLS-VIP method.

ACKNOWLEDGMENTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESEARCH METHODOLOGY: THEORETICAL BASIS FOR CALCULATING ENVIRONMENTAL SUSTAINABILITY OF SELECTED OECD AND NON-OECD COUNTRIES
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. ACKNOWLEDGMENTS
  8. LITERATURE CITED
  9. Supporting Information

The work presented here was funded by the EPA Office of Research and Development. This project was supported in part by an appointment to the Research Participation Program at the Office of Research and Development, National Risk Management Research Laboratory, U.S. Environmental Protection Agency, administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the U.S. Department of Energy and EPA. The supplementary information provided with this paper contains an excel spreadsheet where a reader can access the data used to generate the figures in this paper. The Matlab® code used to generate the data is available upon request from the authors.

LITERATURE CITED

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESEARCH METHODOLOGY: THEORETICAL BASIS FOR CALCULATING ENVIRONMENTAL SUSTAINABILITY OF SELECTED OECD AND NON-OECD COUNTRIES
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. ACKNOWLEDGMENTS
  8. LITERATURE CITED
  9. Supporting Information
  • 1
    United Nations. (2013). Millennium Development Goals.; Available from: http://www.un.org/millenniumgoals/. Accessed March 19, 2014.
  • 2
    Sikdar, S., Sengupta, D., & Harten, P. (2012). More on aggregating multiple indicators into a single index for sustainability analyses, Clean Technologies and Environmental Policy, 14, 765773.
  • 3
    Mukherjee, R., Sengupta, D., & Sikdar, S. (2013). Parsimonious use of indicators for evaluating sustainability systems with multivariate statistical analyses, Clean Technologies and Environmental Policy, 15, pp. 669–706.
  • 4
    Niemeijer, D. & de Groot, R.S. (2008). A conceptual framework for selecting environmental indicator sets, Ecological Indicators, 8, 1425.
  • 5
    EPA. (2008). EPA's 2008 Report on the Environment, Washington, DC: National Center for Environmental Assessment, EPA/600/R-07/045F.
  • 6
    Fiksel, J., Eason, T., & Frederickson, H. (2012). A framework for sustainability indicators at EPA, 2012, EPA, EPA/600/R/12/687, http://www.epa.gov/sustainability/docs/framework-for-sustainability-indicators-at-epa.pdf. Accessed March 19, 2014.
  • 7
    Freudenberg, M. (2003). Composite indicators of country performance: A critical assessment. No. 2003/16.OECD Publishing.
  • 8
    Esty, D.C., Marc L., Tanja S., and Alexander De S. (2005). Environmental Sustainability Index: Benchmarking National Environmental Stewardship, New Haven: Yale Center for Environmental Law & Policy p. 47–60.
  • 9
    Phillis, Y.A., Grigoroudis, E., & Kouikoglou, V.S. (2011). Sustainability ranking and improvement of countries, Ecological Economics, 70, 542553.
  • 10
    Cinar, A., Palazoglu, A., & Kayihan, F. (2010). Chemical process performance evaluation, Boca Raton, FL, USA: CRC press.
  • 11
    Chong, I.-G. & Jun, C.-H. (2005). Performance of some variable selection methods when multicollinearity is present, Chemometrics and Intelligent Laboratory Systems, 78, 103112.
  • 12
    Maitra, S. & Yan, J. (2008). Principle component analysis and partial least squares: Two dimension reduction techniques for regression, in Casualty Actuarial Society, Discussion Paper Program.
  • 13
    Martens, H. (1991). Multivariate calibration, Wiley.
  • 14
    Science. (2013). This Week's Section, Science, 341, 1416.
  • 15
    ICIS. (1991). Pollution laws force eastern German closures, ICIS Chemical Business. Available at: http://www.icis.com/resources/news/1991/03/18/12388/pollution-laws-force-eastern-german-closures/, Accessed March 13, 2014.
  • 16
    Mata, T.M., Nídia S.C., Carlos A.V.C., Subhas K.S., and António A.M. (2013). Sustainability analysis of biofuels through the supply chain using indicators, Sustainable Energy Technologies and Assessments, 3, 5360.
  • 17
    Bare, J.C. (2002). Traci, Journal of Industrial Ecology, 6, 4978.

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESEARCH METHODOLOGY: THEORETICAL BASIS FOR CALCULATING ENVIRONMENTAL SUSTAINABILITY OF SELECTED OECD AND NON-OECD COUNTRIES
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. ACKNOWLEDGMENTS
  8. LITERATURE CITED
  9. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
ep11963-sup-0001-suppinfo.docx20KSupporting

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.