Macro Sustainability across Countries: Key Sector Environmentally Extended Input-Output Analysis

Abstract

When formulating economic development strategies, the environment and society must be considered to preserve well-being. This paper proposes a comparative sustainability assessment method using environmentally extended input-output analysis and multi-criteria decision aid. Using symmetric input-output tables and sectoral CO₂ emissions and employment data for six countries, linkage coefficients are calculated for 163 sectors in each country. Multi-criteria decision aid tool, ELECTRE III, is used to derive outranking relationships among each country’s sectors using these coefficients as criteria, resulting in a hierarchy of sectors ordered by sustainability. Sectors that frequently appear at the top of the six hierarchies included education, health care, construction, and financial intermediation. China’s results differ significantly because of its concentration of economic activity on the primary/secondary sectors. The results can enable identification of key intervention pathways along which sustainable development could be stimulated. Country-specific recommendations and reflections on economic and sustainability policy initiatives are discussed.

1. Introduction

There is no one-size-fits-all approach to sustainable economic development that can be applied to every country, region, and city in the world. In the face of anthropogenic climate change, depletion of natural resources, dependence on fossil fuels, loss of biodiversity, and worsening socio-economic inequality; every region that intends to address these issues must tailor its sustainable development strategy in a way that considers its own unique set of economic, social, and environmental circumstances. Environmental protection and social well-being are global, not localized, issues that can only be effectively addressed in an international, coordinated manner. Global initiatives have been designed and undertaken to align different nations’ sustainable development efforts towards agreed-upon goals, most notably and recently the 2015 Paris Agreement, which has been signed by 195 countries. The primary aim of the Paris Agreement is to mitigate anthropogenic climate change and ultimately hold global temperatures below 1.5 degrees Celsius above pre-industrial levels; a goal which will require careful consideration of direct and indirect GHG emissions resulting from industrial economic activity [1] (Savaresi, 2016). This paper aims to demonstrate a quantitative framework that can help individual countries most effectively identify how to sustainably manage their economies through key sector analysis, and explain the nature and causes of differences between key sectors across different countries. The six countries analyzed in this study are Austria, China, France, Germany, Sweden, and the USA. These countries were selected due to characteristics pertaining to their energy generation mix, degree of economic development, and social structure, among other factors that are further detailed in Section 2.4.

Understanding the complex internal structure of an economy allows one to gain a clearer picture of the nature and implications of its natural resource throughput as well as its impact on the wider social structure. Input-output analysis (I-O) is a simple yet powerful way of understanding the interdependencies of different components of an economy. EE I-O allows one to quantify the direct and indirect effects of economic activity on environmental and social indicators (CO₂ emissions and employment, respectively, in this study). The EE I-O method of key sector analysis is fundamental to gaining a clearer understanding of an economic structure and is employed to analyze the economies of the six countries. Key sectors can be calculated by different weightings, which in this case are different sectors’ propensity for stimulating economy-wide final demand, CO₂ emissions, and employment as measured by forward and backward linkage coefficients. Using the MCDA method ELECTRE III, the forward and backward linkages are then used as criteria to form an outranking hierarchy, at the top of which is a proxy for the most sustainable sectors in each economy. The implications of these findings for the six countries is explored in a policy discussion, and this methodology could be extended to any number of countries in order to most optimally design their approach to sustainable development.

2. Methodological Literature Review

In this section the theoretical underpinnings, history, and recent applications of I-O, EE I-O and MCDA are explored. The justification for selecting each of the six countries is also described. There are large bodies of existing literature for EE I-O and MCDA, and this study aims to make a valuable contribution to that body of work by performing a comparative macro sustainability analysis using the novel framework which combines both key sector identification using EE I-O and MCDA in this manner.

2.1. Input-Output Analysis

The flow of resources into, within, and out of a region’s economy involves several fundamental components including domestic economic sectors that produce and consume output (intermediate producers/consumers); households and the government that consume output (final consumers); and foreign nations that import and export goods and services into/out of the country. I-O is concerned with the intermediate sectors in an economy and the portions of their output that are then consumed by other intermediate sectors as inputs. The core element of an I-O model is a matrix showing resource flows (usually monetary) between economic sectors over a certain period of time (usually one year) [2] (Leontief, 1986). I-O, as a field of study, was developed by Russian-American economist Wassily Leontief, for which he won the 1973 Nobel Prize in Economics. Leontief was the first to represent intersectoral transactions as a matrix. Although his work resulted in the proliferation of the study and applications of I-O, important work by several others preceded him including Francois Quesnay’s Tableau économique in 1758 [3] (He, 1972), and Léon Walras’s Elements of Pure Economics in 1874 [4] (Walras, 2010). Balanced, symmetric I-O tables are derived from non-symmetric supply and use tables using methods including those outlined by [5] Eurostat (2008). The granularity of sector classification is arbitrary, and I-O tables are generated and used at a variety of sectoral resolutions (for example, the Electricity sector can be included as a whole, or it can be split into electricity generated from different sources such as coal, natural gas, nuclear, and renewables).

A variety of useful insights can be derived from I-O tables that reveal the structure of an economy and the interdependent relationships between its economic sectors. Leontief conceived the Leontief Inverse Matrix L, which is calculated as (I − A)–1 where I is the identity matrix and A is the coefficients matrix (each value z in the I-O matrix as a proportion of its total output) [2] (Leontief, 1986). The Leontief Inverse has served as the basis for extensive applied economic analysis since it was first developed. Leontief’s breakthrough and seminal work on the subject was his 1941 book “The Structure of the American Economy, 1919–1939: An Empirical Application of Equilibrium Analysis” [6](Leontief, 1941). In this work he validated, using 10-sector balanced input-output tables that were assembled using empirical data, that the general coefficient structure of the I-O tables was relatively stable over time, providing encouraging evidence that he had discovered a robust, accurately descriptive economic framework. In 1944, Leontief published “Output, Employment, Consumption, and Investment” [7] (Leontief, 1944), which applied I-O to the economic consequences of WWII. I-O was found to be practically useful to the degree that the United States Bureau of Labor Statistics adopted I-O as its preferred method of employment analysis following the war, and the government constructed a high-resolution I-O table that included 400 sectors [8] (Dorfman, 1973). Numerous other practical applications of I-O have emerged in the decades following Leontief’s pioneering work and it has become a key method used in a variety of economic contexts around the world including tourism impact [9] (Fletcher, 1989), interregional trade analysis [10] (Moses, 1955), and evaluating the structure of income distribution [11] (Miyazawa, 2012), among many others.

A key concept in I-O that was developed as an extension to Leontief’s work is forward and backward inter-industry linkages, which was a concept first introduced by Rasmussen [12] (1956). Forward and backward linkages are coefficients derived from the Leontief Inverse Matrix that measure the ‘downstream’ and ‘upstream’ direct and indirect influences of the activity of a particular sector on other sectors in an economy. To illustrate this in a simplified way, suppose an economy consists of three sectors; Agriculture, Manufacturing, and Transportation. Inputs into Agriculture include services provided by Transportation, which itself requires inputs from the Manufacturing sector, and so on. Therefore, expansion of the Agricultural sector results in indirect upstream expansion of both the Transportation and Manufacturing sectors. The degree to which this is the case is the backward linkage coefficient, and the same principle applies to forward linkages with downstream, consuming sectors. This methodology can be used to identify key sectors, which was an idea first proposed by [13] Hirschman (1958). Key sectors are those which have forward and backward linkage coefficients that are both greater than one, i.e., the sector is more deeply linked on average, in both directions, than all of the other sectors in the economy. Hirschman postulated that economic growth occurs primarily due to these key sectors and their above-average influence on the rest of the economy.

This was the concept underpinning unbalanced growth ([14] Akamatsu, 1961; [15] Hirschman and Lindbolm, 1962; [16] Hansen 1965). The concept of key economic sectors and linkages was further developed in a series of papers by Miyazawa ([17] Miyazawa, 1960; [18] Miyazawa, 1966; [19] Miyazawa, 1968; [20] Miyazawa, 1971) which focused on multi-regional economic systems and income multipliers. Michael Sonis and his colleagues have developed and expanded upon key economic sectors and linkages further, specifically focusing on hierarchies of key linkages/pathways and multiplier product matrices ([21] Sonis and Hewings, 1989; [22] Hewings et al., 1989; [23] Sonis and Hewings, 1992; [24] Sonis et al., 1995; [25] Israilevich et al., 1997; [26] Hewings et al., 1998; [27] Sonis et al., 2000; [28] Okuyama et al., 2004; [29] Guilhoto, Sonis and Hewings, 2005). Contemporary work in input-output analysis has included I-O database construction workflows in virtual laboratories ([30] Geschke and Hadjikakou, 2017; [31] Rahman et al., 2017; [32] Wiedmann, 2017), analyzing temporal changes in input-output networks ([33] Avelino, 2017; [34] Ye et al., 2017), and supply chain analysis ([35] Thekdi and Santos, 2016; [36] Baldwin and Lopez-Gonzalez, 2015)

2.2. Environmentally Extended Input-Output Analysis

Environmentally extended input-output analysis (EE I-O) builds upon I-O and is an analytical technique that aims, broadly, to calculate the hidden, indirect, or embodied environmental and/or social impacts associated with an upstream economic event ([37] Kitzes, 2013). EE I-O incorporates sectoral-level data on resource flows and usage (such as CO₂ emissions and employment) in order to calculate direct and indirect intensity of these flows in response to supply/demand stimuli. It has been conceptualized as a tool to present the complexity of economy-environment interactions by Daly [38] (1968), Ayres & Kneese [39] (1969) and Victor [40] (1971).

EE I-O can be used to determine key paths, sectors, and linkages in an economy for environmental, economic, or social indicators ([41] Lenzen, 2003; [42] Shmelev, 2012; [43] Shmelev, 2019).

Leontief was the first to extend input-output models to account for environmental impact and other resource use/emissions types in his 1970 paper “Environmental Repercussions and the Economic Structure: An Input-Output Approach” ([44] Leontief, 1970). In this paper he calculated technical input-output coefficients in order to analyze how growth in hypothetical economic sectors affected levels of corresponding pollution. Leontief and Ford created the first empirical EE I-O analysis of pollution using a 70-sector input-output table of the United States ([45] Leontief and Ford, 1971). Shortly afterwards, Leontief compared the nature of EE I-O characteristics in developed vs. developing countries ([46] Leontief, 1974). Further environmental extensions included those focused on energy and carbon emissions: [47] Carter (1974), [48] Carter (1976), [49] Herendeen & Tanaka (1976), [50] Proops (1977), [51] Park (1982), [52] Proops (1984), [53] Gay & Proops (1993), [54] Polenske & Lin (1993); [55] Minx et al. (2009), [56] Peters et al. (2011); water pollution: [57] Anderson & Manning (1983), [58] Lenzen and Foran B. (2001), [59] Dietzenbacher & Velázquez (2007), [60] Lenzen (2009); waste and resources: [61] Leontief (1977), [62] Duchin (1990), [63] Duchin (1994), [64] Ferrer and Ayres (2000), [65] Nakamura (1999), [66] Nakamura & Kondo (2002), [67] Hoekstra and ven den Bergh (2002), [68] Suh (2005), [69] Kondo & Nakamura (2005), [70] Dietzenbacher (2005), [71] Nakamura & Kondo (2006).

Subsequent key methodological and interdisciplinary contributions were made by incorporating time-series data [72] (Barker, 1981) and MCDA ([73] Kananen et al., 1990; [74] Luptáčik and Böhm, 1994). Contemporary work in the field of EE I-O has included identifying key points of consumption that drive the emissions of particularly harmful substances such as PM₂.₅ ([75] Nagashima et al., 2017); assessing the sustainability of global supply chains ([76] Acquaye et al., 2017), and evaluating the consequences of food waste ([77] Reutter et al., 2017).

EE I-O is closely related to, and has heavily influenced and informed, the field of industrial ecology, which is the study of material and energy flows through industrial systems ([78] Ehrenfeld and Gertler, 1997; [79] Erkman, 1997; [80] Frosch, 1992; [81] Graedel and Allenby, 2010; [82] Jelinski et al., 1992; [83] Thomas, 1997). It has also informed the development of material flow accounting (MFA), which is the study of material flows on a national or regional scale ([84] Fischer-Kowalski et al., 2011; [85] Haberl et al., 2004; [86] Hinterberger et al., 2003; [87] Moriguchi, 1999); life cycle assessment (LCA), which is used to evaluate the environmental impacts of every stage of a product’s existence (manufacturing, usage, disposal, etc.) ([88] Cabeza et al., 2014; [89] Hellweg and Canals, 2014; [90] Curran, 1996; [91] Guinée, 2001), and urban metabolism, which analyzes the flow of material and energy through cities ([92] Gandy, 2004; [93] Kennedy et al., 2011; [94] Niza et al., 2009; [95] Sahely et al., 2003).

More recent environmentally extended input-output contributions include studies by [96] Lenzen et al. (2012) who found that approximately 30% of global species threats (as represented by the IUCN red list) are due to international trade using EORA database. In their research that linked 25,000 species to more than 15,000 commodities produced in 187 countries and evaluated over 5 bln supply chains from the point of view of their biodiversity impacts, [96] Lenzen et al. (2012) found that USA, Japan, Germany, France, UK, Italy, Spain, South Korea, Canada are the top ‘net importers’ of biodiversity and Indonesia, Madagascar, Papua New Guinea, Malaysia, Philippines, Sri Lanka, Thailand, Russia, Cambodia, Cameroon are the top ‘net exporters’. Such analysis was possible through the links between the threatened species and implicated commodities traded in international markets. The issues of the macroeconomic impacts of individual lifestyles have been addressed in [77,97,98,99].

2.3. Multi-Criteria Decision Analysis

Once a sector’s forward and backward linkage coefficients have been calculated and weighted by final demand, CO₂ emissions, and employment, MCDA can be used to bring that information together into one singular assessment of the most optimal trade-offs between those criteria. MCDA is a broad categorization of decision-making frameworks that allow alternatives to be compared to each other according to differing criteria which are often conflicting and incommensurable. MCDA is highly suitable to this application, where sectors serve as alternatives and their forward and backward linkages serve as criteria with which the alternatives are evaluated. MCDA was initially developed in the 1960s by Bernard Roy. Since then, several branches of methods have emerged over the subsequent decades ([100] Roy and Vanderpooten, 1996).

Types of MCDA include outranking methods such as ELECTRE ([101] Roy, 1978; [102] Roy, 1990), Analytical Hierarchy Process (AHP) ([103] Handfield et al., 2002; [104] Saaty, 2008; [105] Zahedi, 1986), goal programming ([106] Ignizio, 1976; [107] Lee, 1972), weighted sum models ([108] Marler and Arora, 2010; [109] Rowley et al., 2012), and NAIADE ([110] Munda, 1995; [111] Menegolo and Pereira, 1996; [112] Munda, 2005).

The applications of MCDA are numerous, and it has historically been employed extensively in a variety of fields such as health care ([113] Nutt et al., 2010; [114] Baltussen et al., 2010; [115] Diaby et al., 2013), energy ([116] Datta et al., 2011; [117] Shmelev and van den Bergh, 2016; [118] Terrados et al., 2009; [119] Wang et al., 2009), and environmental sustainability ([120] Omann, 2004; [121] De Luca et al., 2017; [122] Fernandes et al., 2017, [123] Shmelev, 2017). MCDA has been used in conjunction with EE I-O in some cases, most notably by [82] Kananen et al. (1990), who evaluated emergency management techniques in response to the propagation of economic and political shocks through the input-output network of the Finnish economy; [74] Luptáčik and Böhm (1994), who employed a multi-objective model which aimed to minimize factor costs to produce Gross National Product (GNP) and also minimize net pollution; [42] Shmelev (2012), who used environmentally-weighted forward and backward linkage coefficients in combination with NAIADE to determine which sectors in the UK economy were the most sustainable; [124] Shmelev and Rodríguez-Labajos (2009), who assessed intertemporal macro sustainability in Austria over a 25-year period; and [125] Shmelev (2011), who used MCDA to assess macro sustainability progress over time in Russia according the UN Sustainability Development Framework of Indicators.

For this study, ELECTRE III was selected. ELECTRE III is an outranking method that was developed by Bernard Roy [101] (Roy, 1978), which incorporates realistic decision-making parameters for different criteria scores, namely an indifference threshold (below which one is indifferent to two alternatives by a given criterion), preference threshold (above which one displays a clear preference for one alternative over another by a given criterion), and a veto threshold (above which, for a certain criterion, an alternative outranks another by all criteria) [126] (Figueira et al., 2005). ELECTRE III was selected for this study because it is an outranking method which clearly displays the preference structure of alternatives, and because its incorporation of the aforementioned thresholds lends to a more realistic modelling of a decision-making process beyond simpler methods such as the weighted sum. It is a robust method that uses a strong sustainability concept (due to its non-compensatory nature), as opposed to the weak sustainability concept used by more rudimentary MCDA methods such as AHP ([127] Cinelli et al., 2014). The concepts of weak/strong sustainability and how they relate to MCDA are explored further in [128] Boggia and Cortina (2010); [129] Myllyviita et al. (2013); and [124] Shmelev and Rodríguez-Labajos (2009).

2.4. Country Selection

Austria, China, France, Germany, Sweden, and the USA were selected for this study because of particular economic, environmental, and social traits of interest that they possess. A comparison of the six countries by two measures each of economic, environmental, and social development indicators (as defined by the UK Office for National Statistics) ([130] Office for National Statistics, 2015) is shown below in Figure 1. These traits affect the structure of their economies in unique ways, and observing how these traits are revealed through descriptive EE I-O analysis could result in useful insights that policymakers can use to adapt sustainable development strategies to each of the countries. The justification for each country’s inclusion in the study is set out below.

Austria was included because of the wide array of accurate data available ([124] Shmelev and Rodrigues-Labajos, 2009); the strong influence of the economy of neighboring Germany; and the importance of its primary sector compared to other similar, highly-developed countries.

China was selected due mainly to its relative concentration of economic activity in the primary and secondary sectors compared to the other countries analyzed. China has grown rapidly over the last (GDP growth was 3003.5% between 1990 to 2016, compared to the global average of 234.6%) ([131] World Bank, 2017); it is a highly CO₂ emissions-intensive economy; and it acts as a point of contrast to the other five countries which are comparatively similar to each other.

France was selected primarily to explore the effect of the country’s high degree of reliance on emissions-free nuclear energy, which made up 76.9% of the country’s electricity generation in 2007 ([132] World Nuclear Association, 2017), the year of the EXIOPOL I-O data used in this study.

Germany is a highly CO₂ emissions-intensive economy with a strong focus on its manufacturing sectors relative to other highly developed western nations; particularly its automotive manufacturing sector. Oil and coal supply 58.4% of the country’s energy demand, compared to the EU average of 51.9% ([133] BP, 2016). Germany has ambitious plans to decarbonize its economy, as outlined in its government’s ‘Climate Action Plan 2050’ ([134] Government of Germany, 2016).

Sweden, in contrast to Germany, relies on hydroelectric power plants to supply 27.0% of its energy demand, compared to the global average of 6.9% ([133] BP, 2016). It generated 46.1% of its electricity from nuclear energy in 2007, in comparison to the aforementioned French figure ([132] World Nuclear Association, 2017). It is one of the least CO₂ emissions-intensive economies in the developed world, with CO₂ emissions per capita at 4.6 metric tons compared to the OECD average of 9.7 metric tons ([131] World Bank, 2017). The importance of the government, health care, and other public services to the economy of Sweden is also interesting to explore and contrast to the other countries.

Figure 1. Comparison of the six countries by select economic, environmental, and social indicators (World Bank, 2016; BP, 2016; CIA, 2016).

Figure 1. Comparison of the six countries by select economic, environmental, and social indicators (World Bank, 2016; BP, 2016; CIA, 2016).

The USA is included because of its size and global significance. The country made up 24.6% of global GDP in 2016 ([131] World Bank, 2017). The USA differs in nature to the other five countries in several key ways, namely that it has much higher GDP per capita and CO₂ emissions per capita (Figure 1) and it has a much larger focus on public administration and defense. Its Gini coefficient is also high relative to the other highly-developed nations, implying that its vast wealth is particularly skewed towards the top of its socio-economic hierarchy.

Exploring differences in key sectors between these six countries will help to shed new light on the nature of their differences, provide insight into how sustainable development strategy should be adapted to meet the unique conditions of each nation, and also strengthen the robustness of environmentally extended input-output analysis as a tool for gaining accurate, data-driven insights into complex economies.

3. Methodology and Data Collection

3.1. Selecting a Data Source

To effectively compare and contrast the six countries, a set of standardized I-O tables with corresponding sectoral final demand, CO₂ emissions, and employment vectors were required. Many national governments compile and publicly release their own supply, use, and input-output tables along with labour statistics, but rarely do two countries’ sector classifications align without some degree of modification or disaggregation. Fortunately, there are a number of sources of aggregated and standardized national input-output tables with corresponding environmental and resource accounts for multiple countries, five of which are compared in Table 1.

Table 1. Comparison of aggregated input-output data sources.

	OECD	EORA	GTAP	WIOD	EXIOBASE
Reference	[135,136,137] OECD, 2015	[138] Lenzen et al., 2013	[139] Aguiar et al., 2016	[140] Timmer et al., 2012	[141] Tukker et al., 2013
Years of data available	1995–2011	1990–2013	2004, 2007 and 2011	1995–2011 (I-O Tables)	2000, 2007
				1995–2009 (Environmental/social accounts)
Number of regions	61	186	147	40	43
Number of sectors	34	26−>1000	57	37	163
Corresponding resource accounts	Air emissions, employment.	Energy use, GHG and other air emissions, land use, water use, employment, crop/livestock use, fertilizer use	Energy, CO2 emissions, labour data	Large range of emissions accounts and socioeconomic indicators.	15 land use types; Employment per three skill levels; 48 types of raw materials; 172 types of water uses
Industry classification consistency and other considerations	Fully consistent across input-output tables. Environmental and social accounts do not match up perfectly.	Inconsistent across regions and years. Uses each country’s native classification systems. Mix of SUT and IOT across commodities, industries, and both. Considerable amount of estimated figures	Consistent across countries and years. Focus is on analysis of international trade	Consistent across countries and years.	Consistent across countries and years.

EXIOPOL was selected as the data source for this study because:

1. It contains standardized, symmetric input-output tables for Austria, China, France, Germany, Sweden, and the USA;

2. It contains complete data for final demand as well as sector-disaggregated CO₂ emissions and employment accounts;

3. It has a very high sectoral resolution (163 sectors) including separate categories for different types of electricity generation and waste management.

A disadvantage to using EXIOPOL is that the most recent data available is from 2007, whereas other sources have more up-to-date data. This fact is outweighed by the benefits listed above and does not affect the ability to illustrate this study’s methodology. EXIOPOL’s wide array of environmental and resource accounts are not based entirely on empirical data because of information availability limitations. The process for calculating and disaggregating the environmental accounts used in EXIOPOL is described in [13] Tukker et al. (2013). For the purposes of this study the data is sufficient to illustrate the usefulness of the method and to make reasonably accurate calculations of the CO₂ emissions and employment forward and backward linkage coefficients.

3.2. Calculating Forward and Backward Linkage Coefficients

To calculate the forward and backward linkage coefficients for each industry weighted by each resource vector, the methodology used in Manfred Lenzen’s 2003 paper ‘Environmentally important paths, linkages and key sectors in the Australian economy’ was employed ([41] Lenzen, 2003). The forward linkage is the resource-weighted row average of the Leontief inverse L, which is then divided by the resource-weighted global average of L. The forward linkage is the resource-weighted column average of the Leontief inverse L, which is then divided by the resource-weighted global average of L.

A linkage coefficient represents how deeply linked a sector is relative to all of the linkages in the economy. For example, a sector with a forward CO₂ linkage coefficient of 1.5 is 50.0% more linked than the average of all sectors in the economy. A key sector is one which has both a forward and backward linkage greater than one, implying that is has an above-average ripple effect on economy-wide final demand, CO₂ emissions, or employment in response to stimulus/contraction in that sector.

3.3. Calculating Sector Sustainability with Multi-Criteria Decision Analysis

After the linkage coefficients were calculated for final demand, CO₂ emissions, and employment, an MCDA using ELECTRE III was used to determine which sectors have the most optimal balance of forward and backward linkage coefficients for the three categories. The intended output of this calculation is a proxy for the most sustainable economic sectors in the country in question. There are established methods in the literature that guide selection of realistic values for the indifference, preference, and threshold levels, which is inherently a highly subjective process. One such guideline is [142] Rogers and Bruen (1998a), which suggests that (a) the veto threshold v should be set closer to the preference threshold as adverse human reaction to the criteria increases and (b) that the preference threshold p and the indifference threshold i should be defined within relatively strict limits.

The threshold settings for this study are shown in Table 2. The thresholds were determined as a proportion of the difference between the highest and lowest linkage coefficients for each criteria. Indifference thresholds were set at 1.0% of the difference, preference thresholds were set at 1.1% (reflecting the first of Rogers and Bruen’s [142] principles) and veto thresholds were set at 80.0% of the difference for final demand and employment, and 70.0% of the difference for CO₂ emissions (conforming to Rogers and Bruen’s [142] second principle). Setting the indifference and preference thresholds so close together follows the concept of strong sustainability because there is a very small degree of compensation among criteria.

Table 2. Criteria thresholds and goal settings for the China ELECTRE III MCDA calculation.

Criteria	Max Linkage Coefficient	Min Linkage Coefficient	Difference (Max–Min)	Indifference Threshold (i)	Preference Threshold (p)	Veto Threshold (v)	Goal
Final demand, forward linkage	18.15	0	18.15	0.18	0.2	14.52	Maximise
Final demand, backward linkage	14.09	0	14.09	0.14	0.16	11.27	Maximise
CO2 emissions, forward linkage	27.36	0	27.36	0.27	0.3	19.15	Minimise
CO2 emissions, backward linkage	9.79	0	9.79	0.1	0.11	6.85	Minimise
Employment, forward linkage	9.05	0	9.05	0.09	0.1	7.24	Maximise
Employment, backward linkage	9.27	0	9.27	0.09	0.1	7.42	Maximise

All six criteria were equally-weighted at 16.67% each. Because of the volume of data produced for the six countries, different weighting schemes were not evaluated in this study due to scope constraints. Utilizing unequal weighting schemes is a direction for future research in this context.

4. Results

4.1. Forward and Backward Linkage Coefficients

Figure 2, Figure 3, Figure 4, Figure 5, Figure 6 and Figure 7 show key sectors and forward/backward linkage coefficients weighted by final demand, CO₂ emissions, and employment. Key sectors (those with forward and backward linkages greater than one) are highlighted in orange. Two countries are compared for each weighting scheme, and the complete coefficient results can be found in Appendix A and Appendix B. The countries in the following comparisons were chosen because of the nature in which their results contrast for different weightings.

Figure 2. Final demand-weighted forward and backward linkages (USA).

Figure 2. Final demand-weighted forward and backward linkages (USA).

Final demand key sectors in the USA are largely centered on the services/tertiary industries, with 18 tertiary sectors out of a total of 25 key sectors. Public administration and defence is deeply linked within the USA economy because of the government’s high degree of military expenditure, which totaled 3.3% of its GDP in 2016 compared to the global average of 2.2% of GDP ([131] World Bank, 2017). Real estate activities, Health and social work, and Construction rounded out the other most deeply linked sectors. Deeply linked secondary sectors in the USA include Construction, which is typical of most countries in this study and Manufacture of motor vehicles, trailers and semi-trailers, due to the USA’ status as the second-largest motor vehicle producer in the world after China.

In contrast to the USA, Germany’s economy has more key sectors in the secondary/manufacturing field (Figure 3). With 14 primary/secondary sectors out of a total of 29 key final demand sectors. Of note is the deep linkage of the Manufacture of motor vehicles, trailers and semi-trailers sector. Automotive manufacturing comprised 20.0% of total industry revenue and employed nearly 800,000 people in Germany in 2015. It is also is the most central sector overall to the German economy from a network analysis standpoint ([143] Blöchl et al., 2011). The most deeply linked final demand sector in Germany is Real estate activities, with Health and social work, Construction, Public administration and defence, and Education also strongly linked. Public administration and defence, although a key sector, is not as deeply linked as it is in the USA. Government military expenditure in Germany is 1.2% compared to 3.3% in the USA ([131] World Bank, 2017).

Figure 3. Final demand-weighted forward and backward linkages (Germany).

Figure 3. Final demand-weighted forward and backward linkages (Germany).

In Austria, coal and natural gas supply 36.0% and 22.4% of total energy consumption, respectively ([133] BP, 2016). This is reflected in Figure 4 that as they are both deeply linked sectors as weighted by CO₂ emissions. Manufacturing industries also have strong linkages throughout the Austrian economy, such as Manufacture of basic iron and steel and of ferro-alloys and first products thereof, Manufacture of gas; distribution of gaseous fuels through mains, and Manufacture of cement, lime and plaster. Of note are sectors with very high backward linkages and forward linkages of zero, which are renewable energy generation sectors (Production of electricity by Geothermal, Production of electricity by solar photovoltaic, Production of electricity by wind, and Production of electricity not elsewhere classified). This phenomenon is unique to Austria among the six countries studied, and is important to note as the true indirect impact of renewables on CO₂ emissions in Austria may be significantly higher than previously thought.

Figure 4. CO₂ emissions-weighted forward and backward linkages (Austria).

Figure 4. CO₂ emissions-weighted forward and backward linkages (Austria).

By comparison, the key CO₂ emissions sectors in France are shown in Figure 5. This structure is quite different than that of Austria. Because France generated 76.9% of its electricity from nuclear energy in 2007 ([132] World Nuclear Association, 2017), electricity generation sectors tend to be less deeply linked in its economy weighted by CO₂ emissions. This contrast is even more striking when compared to coal-focused economies such as Germany, the USA, and China.

The effect of Sea and coastal water transport is apparent compared to landlocked countries such as Austria. Other business activities have an unusually high forward CO₂ emissions-weighted linkage, due mainly to its strong downstream linkages CO₂ emissions-intense sectors such as Sea and coastal water transport; Inland water transport; and Manufacture of cement, lime and plaster.

Figure 5. CO₂ emissions-weighted forward and backward linkages (France).

Figure 5. CO₂ emissions-weighted forward and backward linkages (France).

As can be seen in Figure 6, Sweden’s workforce is largely centered around the tertiary/services sectors, with Health and social work, Education, and Public administration and defence; compulsory social security employing 31.9% of the workforce ([141] Tukker et al., 2013).

Other industries that are deeply linked include Other business activities, Construction, and Wholesale trade. Sweden is the only country where Health and social work has the highest forward and backward employment linkage coefficients. This is likely because of strong state support for healthcare and other public services in Sweden. Health expenditure as a percentage of GDP is 11.9% in Sweden, which is the second-highest out of all OECD countries apart from the USA, which spends 17.1% of GDP on healthcare ([131] World Bank, 2017), although that spending is of a different nature.

Figure 6. Employment-weighted forward and backward linkages (Sweden).

Figure 6. Employment-weighted forward and backward linkages (Sweden).

In stark contrast to Sweden, as well as to the other countries analyzed, China’s labour force is largely focused on the primary and secondary production and manufacturing sectors (Figure 7). China’s primary and secondary sectors employ 67.6% of its workforce, compared to 25.1% in Sweden ([141] Tukker et al., 2013). Of note are the strong backward linkages in meat production and processing in China. Production of meat pigs has a backward employment linkage of 9.27 due primarily to it drawing inputs from employment-heavy sectors such as Pigs farming; Wholesale trade; and Cultivation of vegetables, fruit, nuts. China’s uniqueness among the other nations included in this study is clearly identifiable in these results and will be further elaborated upon in the following sections of this study.

Figure 7. Employment-weighted forward and backward linkages (China).

Figure 7. Employment-weighted forward and backward linkages (China).

Figure 8 and Figure 9 visualize the dispersion of linkage coefficients in France and the USA in three dimensions, with the average CO₂ linkage on the x-axis, the average final demand linkage on the y-axis, and the average employment linkage encoded in the size of the circles. One can start to gain a sense of which sectors are the most sustainable by observing which sectors cluster near the top-left of the charts and have large circles indicative of high average employment linkage coefficients. Less sustainable sectors cluster near the bottom-right of the chart and have small circle sizes/low employment linkages. The remaining linkage coefficient trade-off charts can be found in Appendix C.

Figure 8 and Figure 9 show distinct differences between France and the USA. In France, Real estate activities; Construction; Health and social work; and Public administration and defence cluster towards the top-left of the chart indicating their high degree of overall sustainability. The bottom-right of the chart includes less sustainable sectors such as Manufacture of basic iron; Production of electricity by coal; and Air transport. In the USA, the scale of the x-axis extends past 28.0 to accommodate the extremely high CO₂ emissions linkages from the Production of electricity by coal sector. More sustainable sectors in the USA appear to be Health and social work; Real estate activities, and Construction. Public administration and defence, although it appears to cluster at the top-left of the figure, has very high forward and backward CO₂ linkage coefficients of 4.50 and 2.71, respectively. The USA military is thought to be the largest single institutional emitter of GHGs in the world, and it emitted over 70 million tons CO₂-equivalent in 2016 (excluding foreign bases and operations) ([144] U.S. Department of Energy, 2017).

4.2. Sector Sustainability Rankings

Table 3 below shows the results of the ELECTRE III calculations. The five most and least sustainable sectors in each country according to the six linkage criteria are shown. They are ranked according to how many other sectors in the economy that they dominate with a cutoff proportion of 0.8. The full rankings for all 163 sectors can be found in Appendix D.

Table 3. Five most sustainable sectors by country, by number of other sectors outranked.

Country	Rank	Sector
Austria	1	Other service activities
	2	Health and social work
	3	Education
	3	Public administration and defence; compulsory social security
China	1	Education
	2	Public administration and defence; compulsory social security
	3	Fishing, operating of fish hatcheries and fish farms; service activities incidental to fishing
France	1	Health and social work
	2	Education
	3	Processing of meat pigs
Germany	1	Health and social work
	2	Education
	3	Hotels and restaurants
Sweden	1	Health and social work
	2	Education
	3	Hotels and restaurants
USA	1	Education
	2	Health and social work
	3	Sale, maintenance, repair of motor vehicles, motor vehicles parts, motorcycles, motor cycles parts…

The results of the MCDA show many similarities between the six countries, as well as some prominent differences. Education is in the top three sustainable sectors for all six countries. Health and social work is in the top three for all countries except for China. Financial intermediation appears in the top five for all countries except for China. China’s top five most sustainable sectors are mostly comprised of primary and secondary industries due to the concentration of its final demand and labour force in those portions of the economy. Of note is the peculiar presence of both Mining of coal and lignite; extraction of peat and Processing of meat pigs in the top five most sustainable sectors in France. While these results seem erroneous at first, they are explained by unexpectedly high backward linkage coefficients to key final demand and employment sectors. In the case of Mining of coal and lignite; extraction of peat, the sector has strong backward linkages with (i.e., it draws heavily on for inputs) final-demand and employment heavy industries such as Other business activities; Financial intermediation, except insurance and pension funding; and Computer and related activities. Its forward and backward CO₂ emissions linkages are both practically negligible, therefore it is considered much more sustainable, as calculated using this dataset, than it would seem at first glance. Processing of meat pigs has very strong backward linkages to Other business activities and Wholesale trade, which explain its high ranking in France’s sustainability hierarchy. The EU’s Common Agricultural Policy subsidy program could also play a role in this result. The possibility exists that these unexpected results are due to some degree of empirical inaccuracy in the data; however, the fact that their presence can be explained clearly by linkage coefficients is promising for this method’s capability of eliciting and justifying unexpected and potentially valuable new insights.

The least sustainable sector results also share many similarities across the six countries. Production of electricity by gas and coal appear frequently, as would be expected due to their high CO₂ emissions intensities. China’s results were more similar to those of the other five countries than they were for the top five most sustainable sectors, however the presence of Collection, purification and distribution of water on the list of least sustainable sectors is notable. This is due to it drawing large amounts of input from Production of electricity by coal and Steam and hot water supply, which are both highly CO₂-intensive in China. In France, both Air transport and Sea and coastal water transport are relatively unsustainable due to their very high CO₂ emissions forward and backward linkages and the sectors’ relatively insignificant effects on final demand and employment. In Sweden, Production of electricity by biomass and waste is highly unsustainable due to that same set of characteristics.

Figure 8. Trade-off of average forward and backward linkage coefficients (France).

Figure 8. Trade-off of average forward and backward linkage coefficients (France).

Figure 9. Trade-off of average forward and backward linkage coefficients (USA).

Figure 9. Trade-off of average forward and backward linkage coefficients (USA).

5. Discussion

5.1. Differing Economic Structures of the Six Countries

The results of the above analysis reveal as many similarities between the six countries as they do differences. Many of the countries share key sustainable sectors, and the implications of any differences in economic structure can be drawn by comparing the five highly-developed, western countries (Austria, France, Germany, Sweden, and the USA) to China, which is still rapidly developing and has a greater economic focus on its primary and secondary industries. Sectors appearing at or near the top of the MCDA hierarchies for most of the five western countries include Health and social work; Education; Financial intermediation, except insurance and pension funding; Insurance and pension funding, except compulsory social security; Hotels and restaurants; Computer and related activities; and Other service activities.

These sectors contribute greatly to final demand and therefore economic growth while employing the largest portions of the countries’ workforces. Sectors appearing most frequently at the bottom of the MCDA hierarchies in these countries are generally those associated with electricity production using fossil fuels and extractive industries, particularly mining of metals and ores. Air transport is also quite unsustainable in these countries due to that industry’s enormous usage of fossil fuels which result in high amounts of CO₂ emissions. By comparison, China’s most sustainable sectors contain some of the tertiary industries found at the top of the MCDA in the five western countries, however agricultural and other primary/secondary sectors play more important roles in China, comparatively. The most sustainable sectors in China include Education; Public administration and defence; compulsory social security; Fishing, operating of fish hatcheries and fish farms; service activities incidental to fishing; Cultivation of vegetables, fruit, nuts; Processing of Food products nec; and Manufacture of fish products, many of which are primary and secondary sectors. The tertiary sectors only account for 37.2% of China’s total final demand and 32.4% China’s of total employment. This is about half of the equivalent figure in the other five countries, on average. A breakdown of the broad final demand and employment structures of the six countries can be seen in Figure 10. Of note is China’s heavy employment focus on its primary agricultural/extractive industries, which comprise 42.1% of total employment, although they contribute only 6.0% of final demand.

Figure 10. Industry make-up of final demand and employment across the six countries.

Figure 10. Industry make-up of final demand and employment across the six countries.

5.2. Policy Implications

The insights gained from this study could be highly useful to policymakers who are formulating green economy and investment policies. First, it can offer guidance on where to focus the greening efforts by identifying sectors that are either clearly unsustainable, or are highly economically beneficial but do not have strong linkages with employment and are particularly harmful in their indirect effects on the environment. A list of such ‘pitfall’ industries is shown below in Table 4, which shows the list of sectors that have a strongly positive impact on the economy in terms of final demand, but are not particularly beneficial in terms of employment and have above-average upstream and downstream CO₂ emissions linkages. These are distinguished from the most unsustainable sectors, as many of those tend not to be economically-attractive sectors.

Table 4. Pitfall sectors in the six countries (those with above-average final demand linkages; above-average CO₂ emissions linkages; and below-average employment linkages).

Austria	China	France	Germany	Sweden	USA
Processing of Food products nec	Distribution and trade of electricity	Processing of Food products nec	Petroleum Refinery	-	-
Transmission of Electricity		Chemicals nec
Distribution and trade of electricity
Air transport

Actionable steps that can be taken by policymakers from the results of this study include designing or re-designing economic development and investment plans around the sectors that are the most and least sustainable. The methodology used is flexible and could evolve continually to incorporate new data and shifting objectives, which is important considering sustainable development strategy should be highly dynamic and adaptable ([145] El-Erian and Spence, 2008).

Each of the six countries has ratified the 2015 Paris Agreement. In addition to this shared objective, they also have clearly defined sustainable development strategies, which could benefit to some degree from insight gained using this study’s methodology. The Austrian government has long displayed its commitment to environmental protection through successful pollution reduction initiatives, relatively high environmental spending as a proportion of GDP, and a commitment to future green growth through its ‘Masterplan Green Jobs’ and ‘Growth in Transition’ initiatives ([146] OECD, 2013). Its policy support for renewable energy is strong with 100% of electricity in its largest state, Lower Austria, now being generated from renewables ([147] France-Presse, 2015). China has historically been one of the most environmentally-impactful nations on earth due to severe air pollution resulting from industrial activity ([148] Chan and Yao, 2008; [149] Feng, 1999). However, its government has adopted more environmentally-friendly policies in recent years, with renewable energy near the forefront of that focus ([150] Xiao et al., 2017; [151] Zhao and Luo, 2017). China planned to invest £292 billion into renewable power by 2020 ([152] Reuters, 2017). France has ambitious sustainable development targets including those outlined in its 2015 ‘Energy Transition for Green Growth Act’ and its role as a driving force in the formulation and international adoption of the Paris Agreement ([153] OECD, 2016). However, France still relies heavily on nuclear energy, although that reliance is falling. Its percentage of electricity generated from nuclear come down from 78.1% in 2006 to 72.3% in 2016 ([132] World Nuclear Association, 2017). Germany is highly ambitious in its plans to decarbonize its economy, particularly as it relates to renewable energy policy. Ultimately, it hopes to reduce its total GHG emissions by 80% to 95% by 2050 as per its ‘Climate Action Plan 2050’ ([134] Government of Germany, 2016). At present, Germany’ economy is highly CO₂-intensive and there is much potential for improvement which could be streamlined by identifying and growing key sustainable sectors. Sweden is perhaps the most environmentally successful of the six countries due to its government’s highly ambitious policy support for sustainability, which has resulted in drastic GHG emissions reductions in recent decades ([146] OECD, 2013). Its proximity to the Baltic Sea and the issues surrounding overfishing in that region will continue to be a primary concern for Sweden in the future ([154] Lindegren et al., 2009). The USA currently faces major challenges to its policy support for sustainable development due actions by its current federal government that include proposed funding cuts to its Environmental Protection Agency ([155] Neslen, 2017) and efforts to increase domestic coal production and exports ([144] U.S. Energy Information Administration, 2017). It has also signaled its intention to withdraw from the Paris Agreement ([83] Thomas, 2017).

In spite of this, some individual states are pursuing their own ambitious sustainability and renewable energy policy targets. For example, California’s “2030 Climate Commitment” aims to generate half of the state’s electricity from renewables by 2030 ([156] State of California Energy Commission, 2015). The state of Oregon has also legislated a requirement for 50% of its electricity to be derived from renewables by 2040 ([157] Oregon Department of Energy, 2016).

6. Conclusions

In this study, EE I-O and MCDA have been employed to identify the most sustainable sectors out of a total of 163 sectors in six countries (Austria, China, France, Germany, Sweden, and the USA) according to their economic, environmental, and social impact. Specifically, forward and backward linkages based on the Leontief Inverse Matrices of the I-O tables were weighted by final demand, CO₂ emissions, and number of employees to determine those sectors that had above average upstream and downstream impacts on the economy as a whole. The linkages for each country were then used as criteria in an MCDA calculation using the ELECTRE III method. This process created a hierarchy of dominant relationships between sectors that showed which sectors were most optimal in maximizing final demand and number of employees, whilst reasonably minimizing CO₂ emissions. The results for the six countries reflected the underlying fundamental differences in the structure and nature of these economies, and these differences are contrasted and assessed for their implications and significance to policymakers. Sectors that were most frequently identified as being highly sustainable in most or all of the six countries included Education, Health and social work, and Hotels and restaurants.

The least sustainable sectors included Production of electricity by gas, coal petroleum and other oil derivatives; mining of metals and other ores; Manufacture of cement, lime and plaster; and Air transport. Some sectors that were sustainable in some countries were highly unsustainable in others, and vice versa. Examples of this include Cultivation of crops not elsewhere classified, which is far more unsustainable in France than it is on average for the other countries; Financial intermediation, except insurance and pension funding, which is far more unsustainable in China than it is on average for the other countries; and Mining of aluminum ores and concentrates, which is far more sustainable in France than it is on average for the other countries. These results prove that there is no effective one-size-fits-all policy strategy for sustainable development that applies to all countries. Taking these coefficients into account can help individual nations optimize their pursuit of sustainable development so that collective international targets such as the Paris Climate Agreement can be met in the most efficient and effective way possible.