8    Aggregation Bias in ‘Consumption vs. Production Perspective’ Comparisons

Evidence using the Italian and Spanish NAMEAs

Giovanni Marin, Massimiliano Mazzanti and Anna Montini

Introduction

The integration of the National Accounting Matrix including Environmental Accounts (NAMEA) and input–output (I–O) tables (usually referred to as environmentally extended input–output analysis – EE-IOA based on NAMEA data) is a challenging but promising way to analyse the factors behind income-environment relationships in international settings (Cole, 2004; Copeland and Taylor, 2004; Frankel and Rose, 2005). More specifically, it can be used to disentangle production and consumption perspectives through the detailed sector-based information provided by the two frameworks. National and international sources of environmental effects can be ascertained in strict connection with streams of literature such as the ‘ecological footprint’ kind of analysis and decomposition analyses. It is also heavily embedded in the wide realm that deals with sustainable consumption and production (SCP) issues (Harris, 2001), a key pillar of current and future EU policy efforts.

A comparison of the production vs. consumption perspective can have important policy implications. Traditionally, environmental policy focused mainly on production activities as sources of impacts and the actors to be targeted by legislation and regulation. Looking at the role of final consumption for vertically integrated domestic and international impacts can push policy attention towards the possible role of the consumer as an actor of environmental policies, together with the international responsibility for spillover of impacts abroad. A key issue here relates to the modelling of the technology associated with imported goods (produced abroad by the stimulus of domestic consumption), which is tricky in practice given the scarcity of data at that level of detail and at sector level. Given the technology, net trade-embodied pollution arises as a structural phenomenon from a systematic difference in the composition of domestic production compared with the composition of consumption rather than structural level imbalances than cannot be sustainable in time. Systematic differences in turn arise from a production specialization that can be, and usually is, more marked than the consumption specialization of a country.

Current consumption structure is changing slowly in most European countries, although with significant momentum, whereas production specialization is changing faster. In a dynamic setting, consumer behaviour is changing too slowly in terms of embodied environmental efficiency compared with domestic production, thus possibly creating a net demand of pollution abroad through net import. Although consumption structure and behaviour can be less sensitive to environmental policies than production, e.g. due to lacking legal basis to constrain the freedom of choice, there can be room for addressing consumers and their behaviour to contribute to higher efficiency in terms of vertically integrated environmental impacts. The EU strategies on SCP paves the way to this policy direction, and analyses based on EE-IOA, addressing the differences between the two perspectives, can clarify the needs and implications of these policies.

We can affirm that sector-based input–output datasets existing for EU countries offer the possibility of highlighting how emissions are indirectly associated with production. NAMEA-type tables are datasets with coefficients on emission per output that can thus be matched with I–O tables for useful integration. Integration aims at calculating economic-environmental performances by sector by including the role of trade. In other words, it aims to test the hypothesis that given different relative emission efficiency, the structure of imports and exports matters. EE-IOA can provide additional information on environmental implications of economic structure and structural change; its objective is to investigate to what extent changes in final consumption patterns, production technologies and trade patterns (as a result of the decoupling of consumption from production) affect domestic and world-induced air emissions. Moreover, EE-IOA quantifies to what extent the geographical separation between consumption and production activities has occurred and whether it has determined increases or decreases in global environmental pressures.

From a general and methodological point of view, the integration of NAMEA accounting and I–O tables touches upon ecological/environmental economics and industrial ecology frameworks. Due to the striking increase of related works in such realms, the brief survey we provide in the next paragraph aims to give insights into recent developments and offer stimulus for future analyses rather than offering full coverage. It is worth noting that, very recently, there has been increasing interest in these environmental issues in the I–O society. Plenty of environmental papers appeared in the 2010 Sidney Conference of the I–O Society, some of those exploiting NAMEA. A boom of papers on environmentally extended I–O was reached in 2009 that witnessed a peak (Hoekstra, 2010), with a total of 360 papers, from 1969 to 2010. A related field of analyses which has witnessed great relevance in the I–O arena is structural decomposition analysis (SDA), a core technique for analysing factors behind delinking that focuses on sector heterogeneity. Decomposition analysis is one of the most effective and widely applied tools for investigating the mechanisms influencing energy consumption and emissions and their environmental side-effects (Mazzanti and Montini, 2010). SDA has been applied to a wide range of topics, including demand for energy (e.g. Jacobsen, 2000; Kagawa and Inamura, 2004, 2001) and pollutant emissions (e.g. Casler and Rose, 1998; Wier, 1998). Many studies address industry. Nevertheless, services are also relevant: they are less energy intensive but present lower technological contents and can indirectly contribute to strong environmental impacts (we note the NAMEA-based disentangled analyses in Marin and Mazzanti (in press), who present industry vs. services assessments for Italy). Alcántara and Padilla (2009) analyse CO₂ emissions for Spain using I–O (year 2000). They conclude that

Transport activities are the services with the highest level of the direct emissions generated in the production of the sector. These activities are required by the other sectors of the economy to a greater degree than they are for their own final demand. Therefore, the production sold to other sectors causes more emissions than its own final demand. However, in the case of other service activities, direct and indirect emissions related to final demand are much more important, due to the strong pull effect of service activities on other activities of the economy. In this respect, Wholesale and retail trade, Hotels and restaurants, Real estate, renting and business activities, and Public administration services should be highlighted. These services receive scarce attention in the design of policies aimed at reducing emissions, but are notably responsible for the major increase in emissions experienced in recent years.

Trade is the key factor in recent extended I–O and NAMEA works that aim to deal with SCP contents.¹ We recall that the main aim is to assess direct and indirect environmental effects by attributing their relative weights to national consumption and to exports in the explanation of a country’s environmental performance. We present below a brief discussion which centres on recent and comparable works that have touched the core issue of I–O/NAMEA integration with the aim of analysing the sustainability of production and consumption by taking trade into account. Currently, main efforts aim to move away from the domestic technology assumption (DTA) that says that imported goods use the same technology (in terms of structure of intermediate inputs and environmental efficiency) as goods produced domestically.

A very recent example is Arto et al. (2010). They show that Spain is a net emissions exporter and consequently, its consumer responsibility in emissions is higher than its producer responsibility. The difference between both types of responsibility increases by applying the physical DTA. This is substantially due to the fact that the monetary DTA estimates less embodied emissions in imports from non-Annex I countries than the physical DTA.² Linking to NAMEA studies commented on below, we note Watson and Moll (2008). It can be seen that the NAMEAs can be manipulated using EE-IOA to provide the two different perspectives useful for SCP: usual production perspective and an additional consumption perspective. The consumption perspective focuses on the production chains of all final products consumed domestically. This includes goods produced in the home country for domestic consumption and products imported for consumption, but it excludes production chains of exports. We also highlight the recent special issue in Economic Systems Research, (Volume 21, Issue 4, 2009) that deals with applications regarding I–O and carbon footprinting. A study that brings together various frameworks highlighting flexibility of methods and usefulness of integrated use is certainly Moll et al. (2007) (two years, eight countries). The work shows that, according to different sectors and countries, the domestic production patterns and associated direct domestic environmental pressures are rather different. Electricity, gas and hot water production, agriculture and transport and communication services cause the majority of environmental pressures. Direct pressures from private households (mainly for heating and private transport) constitute another important source.

With regard to international factors, it can be seen that a second determinant for cross-country differences in domestic direct pressures is the role of exports. When it comes to consumption and investment patterns, Moll et al. (2007) show that cross-country differences are far less pronounced than production patterns. Analyses focusing on environmental impacts of consumption (by categories) are also found in Huppes et al. (2005): food, heating and transport emerge as core impacting aggregation.³ We also note the extensive IPTS ‘EIPRO’ report (2006). In general, it is the satisfaction and organization of basic needs, i.e. eating, housing and mobility, that is responsible for the majority of production-cyclewide environmental pressures.

Finally, we cited the peak of emphasis that the I–O society gave to environmentally extended I–O analyses in 2010. Among the many works presented, in our view, some of the more intriguing pieces that tried to exploit NAMEA accounting in I–O contexts were by Rueda-Cantuche (2010), Hoekstra et al. (2010), Basina and Sfetsos (2010), Rormose et al. (2010). Three of the authors mentioned appear in this book with some complementary pieces of work.

Methodological and Empirical Literature

Empirical analysis with an extension of the use of the statistical information derived from environmental accounts and the input–output tables requires several considerations to be made. The aim of this chapter is linked to the so-called aggregation bias. As suggested by Lenzen (2011), environmental I–O analyses of environmental issues are often plagued by the fact that environmental and I–O data exist in different classifications.

A recurring problem in EE-IOA is that input–output accounts and environmental statistics are often not compiled by the same statistical agency and therefore often differ with respect to the classification of economic sectors and other definitions. In these cases, analysts have to carry out data collection and harmonization procedures in order to integrate both accounts. What can happen is that: (a) environmentally sensitive sectors are sometimes more aggregated in the economic I–O database than the environmental dataset because monetary I–O tables are compiled with no environmental implications in mind; (b) I–O data are disaggregated into more sectors than environmental satellite data, especially for services sectors (Lenzen, 2011).

There are two basic alternatives for dealing with such a misalignment: either environmental data have to be aggregated into the I–O classification (but some environmental sensitive data will lose their peculiarities) or I–O data have to be disaggregated based on fragmentary information (with several assumptions).

By keeping this in mind, the aggregation bias is likely to severely affect the construction of environmentally extended Multi-Region Input–Output (EE-MRIO) analysis, as recently suggested by Su et al. (2010) and Lenzen (2011), as well as environmentally extended Single Region Input–Output accounts with specific assumption regarding the technology used (embodied in international trade, specifically those in the import data).

As will be explained below, the DTA relies on the consideration that all imported commodities are produced with the same mix of intermediate inputs (in monetary terms and as indicated by the intermediate flows in the input–output table) and with the same environmental efficiency (in terms of emissions per monetary unit of output) as domestic commodities.

Some authors (including Turner et al., 2007; Peters, 2007; Serrano and Dietzenbacher, 2010; Arto et al., 2010) suggest moving away from the DTA because they consider it too simplistic but they recognize that, generally, the DTA produces better estimates than ignoring imports altogether. Ideally, full information on bilateral trade plus corresponding NAMEA data by country is equivalent to analysing trade of impacts at country-by-country differentiated coefficients. However, it requires a wide and often unavailable range of data. A possibility for dealing with the latter is to include only the most important trade partner (and for which required data are available) in terms of emissions embodied in imports and this, as suggested by Andrew et al. (2009), can substantially improve the accuracy of estimates. For the emissions embodied in imports, Andrew et al. (2009) find that the unidirectional trade model gives a good approximation to the full MRIO model when the number of regions in the model is small. Moreover, the assumption that imports are produced with DTA in an MRIO model can introduce significant errors and requires careful validation before results are used.

If we re-examine the issue of aggregation bias, the studies that have analysed the CO₂ emissions embodied in international trade have also been carried out by using an input–output framework at a specific level of sector aggregation. Generally, the choice has been made to a large extent according to economic and energy data availability or, similarly, economic and environmental data availability. A finding for Su et al. (2010) is that levels of around 40 sectors appear to be sufficient to capture the overall share of emissions embodied in a country’s exports.

The issues related to aggregation bias and a possible DTA obviously affect the consumption⁴ and production perspective when looking at the corresponding emissions. As suggested in the introduction, the focus of the EU policy area on SCP forces researchers to consider new tools of analysis and one of them is the EE-IOA based on NAMEA data. The notion of responsibility (either for the consumer or the producer) addresses some considerations.

As suggested by Gallego and Lenzen (2005), there is a sort of domination of producer-centric representation to view the environmental or social impacts of industrial production. When thinking about environmental impacts, crucial questions arise such as who is responsible for what? How is the responsibility to be shared? Should a firm have to improve the eco-friendliness of its products, or it is up to the consumer to buy or not to buy? Questions of this type can be considered when deciding who takes the credit for successful abatement measures that involve producers and consumers. Moreover, the kind of pollutant considered influences policy implications when looking at the ratio between consumptionbased emissions (C) and producer-based emissions (P). If we consider global pollutants, such as CO₂, and C is bigger than P, the country responsibility is bigger than that reported by the official statistics. If we consider local pollutants and C is bigger than P, the country would be displacing environmental costs to other territories.

Gallego and Lenzen (2005) propose a method of re-tracing the flow of past inter-industrial transactions to allocate responsibility for production impacts consistently among all agents such as consumer, producers, workers and investors. According to them, the input–output analysis can be used as a descriptive tool to re-trace the flow of past transactions and examine ex post how, for example, inputs of resources or outputs of pollution were associated with these transactions.

Serrano and Dietzenbacher (2010) define two ways to evaluate the international responsibility of emissions generated by one country – in their analysis they consider Spain in 1995 and 2000 and nine gases – that were shown to be equivalent: the trade emission balance (as the difference between the emissions embodied in a country’s exports and imports) and the responsibility emission balance (as the difference between the responsibility of one country as a producer and its responsibility as a ‘consumer’).

Methodology and Data

In this section, we outline the main features of the domestic technology assumption (DTA henceforth) and we summarize the main issues related to the assessment of the aggregation bias in input–output analysis including NAMEA data.

Domestic Technology Assumption and Aggregation Bias

The Domestic Technology Assumption

The hypothesis behind the DTA is that the imported commodities (either as intermediate inputs or final consumption) are produced with the same mix of intermediate inputs (in monetary terms) and with the same environmental efficiency (in terms of emissions per monetary unit of output) as domestic commodities.

Serrano and Dietzenbacher (2010) formally describe how and under which conditions an environmentally extended MRIO model accounting for worldwide induced emissions could be reduced to a model using only domestic data with an explicit DTA. In addition to assumptions on technology (i.e. the structure of intermediate inputs described by the input–output matrix) and on the vector of emission coefficients, the export of the country on which the analysis is focused should represent a negligible share of world output.

Another requirement, related to the validity of the domestic technology as a proxy of world technology, is that the country produces domestically at least part of all the commodities it consumes as intermediate inputs or final products. For example, this requirement is not fulfilled when a country has no particular raw materials in its soil or subsoil (oil, coal, gas, minerals, metals, etc.) and it is completely dependent on importing these commodities. As a result, the technology for the extracting industries (section C of NACE 1.1) in the input–output tables is biased towards secondary activities within the sector (e.g. basic transformation of raw materials) and it does not describe the main activity (i.e. extraction) properly. This problem is particularly relevant in environmentally extended input–output analyses in which extracting sectors are, in general, among the most polluting industries.

Although the DTA cannot be used to interpret the results as ‘actual worldwide emissions induced by domestic final demand’, it gives information on the potential emissions arising because of domestic final demand if the country has produced domestically the necessary final and intermediate goods (that is, using domestic technology). Estimates using the DTA, if interpreted properly, are therefore a particularly important indicator of consumer responsibility because of its low requirement for data, the possibility of replicating its results and the straightforward and clear hypothesis behind its implementation. For this reason, we claim that estimates based on the DTA should be used as a benchmark in more complex multi-regional environmentally extended input–output analysis aimed at assessing consumer responsibility.

However, the DTA and the overall EE-IOA results might be severely biased when the commodity/sector aggregation is very low and/or when the country which is analysed relies exclusively on import for certain commodities. In the latter case, in fact, either it will not possible to compute any domestic environmental coefficient (because both emissions and output are zero) or, if this sector is aggregated with other sectors, both the technology (the row of the matrix of technical coefficients when considering both imported and domestic intermediate inputs) and the emission coefficient of the aggregated sector could fail to represent technically viable technologies. A possible solution to this problem, although not conclusive, would be to substitute the specific rows of the matrix of technical coefficients and the specific entries of the vector of emission coefficient for these sectors with data of similar countries which have domestic production in these sectors. However, on the one hand, this kind of manipulation is likely to unbalance the whole input–output system and, on the other, the similarity is difficult to check due to the variety of dimensions included in this type of EE-IOA.

Before discussing the way in which aggregation is likely to introduce biases in the estimates of the level of emissions induced by final domestic demand, we will introduce some notation and explain how induced emissions are computed. The notation is summarized in Table 8.1.

Table  8.1  Summary of the relevant notation

Symbol	Dimension	Description
Zd	n × n	Matrix of domestic intermediate inputs
Zm	n × n	Matrix of imported intermediate inputs
	n × 1	Vector of domestic final demand for goods produced domestically
	n × 1	Vector of domestic final demand for goods produced in foreign countries (import of final goods)
	n × 1	Vector of foreign final demand for goods produced domestically (export of final goods)
	n × 1	Vector of foreign final demand for goods produced in foreign countries (re-export)
e	n × 1	Vector of domestic direct air emissions
i	n × 1	Summation vector (column vector of 1s)
I	n × n	Identity matrix
S	m × n	Aggregation matrix
xd	n × 1	Domestic output
xd + m	n × 1	Domestic+imported output
Ad + m	n × n	Matrix of technical coefficients under the domestic technology assumption *
Ld + m	n × n	Leontief inverse under the domestic technology assumption
fd	n × 1	Domestic final demand
b	n × 1	Emission coefficients

Note

*<r> refers to a diagonal matrix with the diagonal composed by the elements of the vector r.

When estimating the emissions induced worldwide by domestic final demand, we need to account for the intermediate inputs induced worldwide (thus using L_d+m as Leontief inverse) and for domestic final demand only (f_d).

Induced emissions (consumption perspective, e_cp) classified by product/industry are given by:

while total induced emissions may be obtained by post-multiplying e_cp by i.⁵

Aggregation Bias

The issue of the choice of the level of aggregation is crucial in any empirical analysis in economics.⁶ Each aggregation results in losses of relevant information and in implicit compensations which are likely to affect the reliability of the results of any empirical analysis. However, aggregation is often unavoidable. First, the most common constraint regards the availability of sufficiently disaggregated raw data. Second, privacy legislation often prevents the diffusion of disaggregated data.⁷ Third, time and computation constraints are likely to induce the researcher to employ readily available and small bases of aggregated data. Finally, when matching various sources of raw data, there is little alternative to aggregation if one or more of the sources is not sufficiently disaggregated, leading to an overall aggregation. This last issue is very common in multiregional input–output models and the general approach involves reducing the overall level of disaggregation to the level of the most aggregated country/ region.⁸

In environmentally extended input–output analysis, aggregation consists of a reduction in n sectors due to data availability constraints. More generally, if either the intermediate input matrices (Z_d or Z_m) or the vector of direct emissions (e) has low disaggregation, it is enough to force the researcher to reduce the level of aggregation of the model to the lowest ‘n’ dimension. This problem is particularly important when using MRIO models because in principle it is sufficient to have data limitations in one dimension of one out of several countries to force the researcher to reduce the overall level of disaggregation. More formally, the way in which we estimate embodied emissions under different aggregations is described by equation (2):

where S is the aggregation matrix. An aggregation matrix is a rectangular matrix (in our case m × n, with m < n) composed by 1s and 0s. The column sum of S will be 1 for each column while the sum of all the entries equals n. Pre-multiplying a column vector by S results in a new vector composed by m rows in which some of the original cells are summed up in a unique entry. When dealing with a square matrix of dimension n, an aggregate square matrix of dimension m can be obtained by pre-multiplying the original matrix by S (m × n) and post-multiplying it by S′ (n × m).

The aggregation in input–output models is related to two main dimensions: the resolution of sector/commodity disaggregation of input–output matrices and related extensions and the level of spatial/geographical aggregation (Miller and Blair, 2009, p. 160).

The issues of sector/commodity aggregation in input–output models and quantification of its bias have been investigated for a long time, the first relevant article explicitly dealing with this issue dating back to 1952 (Hatanaka, 1952). The main concern at that time was related to computational constraints when dealing with big matrices. Aggregation was one way of easing the computation of the Leontief inverse. However, due to tremendous improvements in computational power, the issue of aggregation is currently related to constraints on the availability of or concerns over the quality of disaggregated data. The measurement and decomposition of the bias have been investigated by Morimoto (1970).⁹ Morimoto (1970) defines the aggregation bias as ‘the difference between the outputs which are derived from the aggregated system and those which are derived by aggregating the results of the original system’ (p. 120).¹⁰The main contribution by Morimoto (1970) is related to four theorems which identify the cases in which the aggregation bias does not arise.¹¹ To summarize, the aggregation bias in static input–output models disappears if, alternatively: (a) the sectors/commodities which are aggregated are characterized by the same interindustry structure; (b) the vector of final demand remains unchanged for all aggregated sectors/commodities whereas it changes for all or some of the nonaggregated sectors/commodities.

However, when dealing with extensions (e.g. environmental extensions) either these conditions should be used together or the additional condition of ‘common emissions coefficient among aggregated sectors/commodities’ should be satisfied. A more complete discussion on the theoretical assessment of the aggregation bias in EE-IOA can be found in two recent articles by Su et al. (2010) and Lenzen (2011); as previously indicated Su et al. (2010) focus on a description of the aggregation bias and its generalization and they perform sensitivity analysis in order to identify a minimum level of disaggregation (around 40 sectors) to assure reliable estimates. Lenzen (2011) demonstrates that it is generally desirable to have approximations of disaggregated input–output relations when environmental information is available at a very disaggregated level instead of aggregating environmental information to the level of original actual input–output data.

In our case, the aggregation bias is likely to arise because, when assessing the consumer responsibility, we consider the vector of domestic final demand (thus excluding the vector of export) instead of total final demand. This is equivalent to estimating the effect of a particular impulse (different from the actual vector of final demand) with the risk of obtaining biased results.

The main purpose of the current analysis is to aggregate our original Italian and Spanish data according to relevant aggregations used in other studies and to compare our benchmark estimates (i.e. the estimates arising from the most disaggregated model) with the estimates arising from less detailed aggregations. More specifically, our benchmark consists of a disaggregation of 50 commodities.¹² This benchmark will be compared with the sub-section NACE Rev. 1.1 level (accounting for 30 sectors) and with an aggregation of 16 sectors roughly corresponding to previous studies based on OECD/IEA data sources such as Ahmad and Wyckoff (2003) and Nakano et al. (2009).¹³Table 8.2 summarizes the sectoral detail of each aggregation we tested.

Table  8.2  Sector Aggregation

Aggregation level	Detail
50-sector aggregation	2-digit NACE Rev. 1.1 except 50–52, 65–67 and 70–74
30-sector aggregation	Sub-sections NACE Rev. 1.1 (2-digit capital letters): A (01–02), B (05) CA (10–12), CB (13–14), DA (15–16), DB (17–18), DC (19), DD (20), DE (21–22), DF (23), DG (24), DH (25), DI (26), DJ (27–28), DK (29), DL (30–33), DM (34–35), DN (36–37), E (40–41), F (45), G (50–52), H (55), I (60–64), J (65–67), K (70–74), L (75), M (80), N (85), O (90–93), P (95)
16 sector-aggregation	Agriculture, hunting, forestry and fishing (01–05); Mining and quarrying and petroleum refining (10–14, 23); Food products, beverages and tobacco (15–16); Textiles, apparel and leather (17–19); Wood and wood products (20); Pulp, paper, printing and publishing (21–22); Chemicals (24); Other non-metallic mineral products (26); Iron and steel (271, 2731) and Non-ferrous metals (272, 2732); Fabricated metal products, machinery and equipment (28–32); Motor vehicles, trains, ships, planes (34–35); Plastics, other manufacturing and recycling (25, 33, 36–37); Electricity, gas (40); Construction (45); Transport and storage (60–62); All other services (41, 50–93 excl. 60–62)

Source:  Ahmad and Wyckoff (2003).

Even if several studies acknowledge that their results depend on the choice of the level of aggregation, to our knowledge, just two of them explicitly performed a sensitivity test for aggregation bias. Wyckoff and Roop (1994) found that aggregating their analysis¹⁴ to six sectors (using a disaggregation of 33 sectors as a benchmark) downward biases the carbon embodied in manufacturing imports by about 30 per cent. Su et al. (2010) perform a similar sensitivity analysis on a single country environmentally extended input–output model for China. Compared to their benchmark results obtained with a disaggregation of 122 sectors,¹⁵ the bias in the estimation of carbon emissions embodied in Chinese exports arising from aggregation is positive and around 12 per cent when using a 10-sector aggregation whereas it almost vanishes when using a 42-sector aggregation.

Data Sources

The current analysis relies on input–output tables for Italy and Spain for the years 1995, 2000 and 2005 with a disaggregation of 60 sectors/commodities and on NAMEA sector-level air emissions data with a disaggregation of 50 sectors for the same years and countries. To match the environmental extensions with the input–output table, we reduced the overall level of disaggregation to 50 sectors. In this section, we discuss the features and the limitations of our base data in detail.

Input–Output Tables

A monetary input–output table is a square matrix in which the entries in each column represent the composition (in monetary terms) of the intermediate inputs of a sector whereas the entries in each row represent how the output of a sector is used by the rest of the economic system, either as intermediate inputs or as final consumption.

The Council Regulation (EC) No. 2223/96 of 25 June 1996 on the European system of national and regional accounts in the Community (the so-called ESA 1995) requires each member country to compile and submit supply-and-use tables annually and symmetric (domestic and import) input–output tables every five years to Eurostat. The regulation is very precise as regards the methodology to be used to collect the data and the structure of the published data but allows some flexibility as regards the choice between ‘commodity-by-commodity’ and ‘industry-by-industry’ input–output tables. On the one hand, commodity-by-commodity input–output tables better describe the actual technology in terms of intermediate commodities to produce a specific product whereas industry-by-industry input–output tables describe relationships among sectors regardless of the actual flows of commodities. On the other hand, most of the extensions (e.g. environmental extensions) refer to industries and not to commodities, making the ‘industry-by-industry’ approach more attractive (Eurostat, 2008; Miller and Blair, 2009). Out of the 31 countries which submit their input–output tables to Eurostat (EU27 plus Croatia, Macedonia, Turkey and Norway), ‘industry-by-industry’ tables are only supplied by eight countries (Denmark, Italy, Hungary, Netherlands, Finland, UK, Turkey and Norway).

In our analysis, we use ‘commodity-by-commodity’ input–output tables in order to make the comparison between Italy and Spain possible. The procedure we use to assign ‘industry’ emissions to ‘commodity’ output is based on the hypothesis that direct emissions related to each commodity within a single industry are proportional to the share of the output of each commodity within the industry (Miller and Blair, 2009). Information on the commodity composition of industry output can be found in the make (supply) matrix.

Starting with the make matrix (V) and the vector of total output by industry (x), we compute a matrix which describes the commodity composition of industry output . Each row of the matrix sums to 1 and indicates the relative weight of the different commodities in the total output of the industry (Roca and Serrano, 2007; Miller and Blair, 2009).¹⁶ To obtain the measure of direct emissions generated by the production of a specific commodity (by all of the industries producing that commodity), indicated with e_pp, we multiply the transpose of C by the vector of direct emissions by industry (e_ii):

In Appendix A we compare our results obtained using the commodity-by-commodity approach for Italy with the results we obtain using the industry-by-industry approach.¹⁷ The estimates for total emissions induced by domestic final demand differ by less than 1 per cent in all cases except for CO in 2000 and 2005, thus confirming the validity of the ‘commodity-by-commodity’ approach.

NAMEA Data

The NAMEA approach to identify environmental pressures across production sectors was developed in the late 1980s and 1990s at the Central Bureau of Statistics of the Netherlands (CBS) under the supervision of Steven Keuning (De Boo et al., 1991). As widely presented above,¹⁸ NAMEA data are constituted by a matrix form statistical source where economic (output, value added, final consumption expenditures and full-time job equivalent) and environmental (emissions) indicators can be observed at sector level. In NAMEA, environmentally relevant information is compiled consistently with the way economic activities are represented in national accounts. This framework divides the economy into production sectors and household consumption categories and shows how each industry branch or the household categories contribute to a set of environmental pressures. This allows quite robust analyses on dynamics, correlation, even causation regarding performance and resource productivity indicators. De Haan and Keuning (1996) and Stauvermann (2007), among others, are examples of seminal papers containing long and comprehensive bibliographies of all past works. Furthermore, de Haan (2004) developed and propagated the NAMEA approach in detail and has applied NAMEA for international comparisons.

In the NAMEA tables, environmental pressures (for Italian NAMEA air emissions and virgin material withdrawal) and economic data are assigned to the economic branches of productive resident units or to the household consumption categories (heating, transport and others) directly responsible for environmental and economic phenomena. We focus here on production macro sectors – agriculture and fishing, industry and services – obtained by aggregating the available economic branches at national level to capture the main potential differences in environmental pressure.

Both the Italian, which dates back to 1990 (first published data in 2000), and the Spanish NAMEA include several air pollutants: carbon dioxide (CO₂), nitrogen oxides (NO_x), methane (CH₄), sulphur oxides (SO_x), nitrous oxide (N₂O), ammonia (NH₃), non-methane volatile organic compounds (NMVOC) and carbon monoxide (CO) among others. In the current chapter, we report results for emissions of five different substances (CO₂, NO_x, SO_x, NMVOC, CO)¹⁹ for which NAMEA with the same aggregation of sectors is available both for Italy and Spain.²⁰

Results and Discussion

Overview: Consumption vs. Production Perspective in the Benchmark Case

Before facing the issue of aggregation and its related bias, in this section we briefly discuss the results for Italy and Spain of our benchmark (50 sectors) estimates for the years 1995 and 2005. The 50-sector aggregation level has been obviously considered as the benchmark; as stated by Su et al. (2010), in empirical studies it is logical to take the view that the finer the level of sector disaggregation, the more refined the decomposition results obtained.

Figures 8.1–8.2 and 8.3–8.4 report the contribution of three macro-sectors²¹ to emissions induced by domestic final demand and domestic direct emissions for Italy and Spain respectively.

Figure  8.1  Emissions Induced by Domestic Final Demand by Sector (Italy).

Figure  8.2  Direct Emissions by Sector (Italy).

Figure  8.3  Emissions Induced by Domestic Final Demand by Sector (Spain).

Figure  8.4  Direct Emissions by Sector (Spain).

In Italy, for all emissions except NO_x and CO (in 1995), the contribution of the demand of final products from industry is above 50 per cent. There has been a general shift towards services in the 1995–2005 decade for CO₂, NO_x and SO_x induced emissions. Regarding those pollutants, a weak reduction in environmental pressures caused by industrial activities from 1995 to 2005 appears; efficiency improvements in production processes and product design could be present but a composition effect cannot be excluded.

Agriculture appears almost irrelevant since most of its final products are used as intermediate inputs (the direct emissions by sector are in fact bigger that those induced by domestic final demand).

Table 8.3 (and Figure 8.5) and Table 8.4 (and Figure 8.6) show the comparison between the consumption and production perspective for Italy and Spain respectively. A consumption/production ratio greater than 1 indicates that the emissions arising from the production needed to satisfy the domestic final demand are greater than the emissions directly generated by domestic production sectors. This is equivalent to saying that the amount of emissions embodied in imports is greater than the amount of emissions embodied in export (i.e. the country is a net exporter of emissions).²² The interpretation should be reversed when the consumption/production ratio is smaller than 1.

Table  8.3  Emissions for production and consumption perspective (Italy, 50 sectors; in tons, CO2 in 1,000 tons)

Figure  8.5  Consumption/Production Perspective (Italy, 50 sectors).

Table  8.4  Emissions for production and consumption perspective (Spain, 50 sectors; in tons, CO2 in 1,000 tons)

Figure  8.6  Consumption/production perspective (Spain, 50 sectors).

However, it is important to bear in mind the limitations imposed by the DTA: the estimated amount of emissions embodied in import is, in fact, the amount of emissions that would have been generated by the importing countries if the imported goods were produced domestically (i.e. according to the domestic technology and with the domestic vector of sector emission coefficients).

Though close to 1, the consumption/production ratios for Italy are always below unity except for CO emissions in 2000 and 2005. Furthermore, the average pattern is either stable (CO₂, NMVOC and CO) or even decreasing (NO_x and SO_x). This result, in line with previous analyses such as Moll et al. (2007) but still quite surprising for an OECD country, may have two main explanations. First, Italy maintained industrial specialization in the manufacturing sector, especially in more traditional (and relatively energy intensive) industries, during the considered period. Second, it may be that, within each two-digit industry, there has been a shift from polluting sub-industries (whose products, formerly produced domestically, have been substituted by import) to cleaner sub-industries. This possible shift may lead to a reduction in direct sector emissions in presence of unchanged aggregate monetary domestic output (though with a different sub-industry composition not visible in aggregate monetary data), thus artificially improving the environmental efficiency of the aggregate sector. This hidden structural change worsens the DTA prediction because it affects the sub-industry composition and the real average environmental efficiency of imports. This possible explanation further highlights the importance of using disaggregate data.

The comparison between the patterns of different emissions suggests other somewhat unexpected and interesting results. Local negative externalities generated by NO_x and SO_x (and not by CO₂) emissions, coupled with relatively strict environmental policies controlling these emissions during the considered period,²³ are expected to increase the incentive to move the production of commodities intensive in these emissions abroad (to pollution havens). This should result in an increase of emissions embodied in imports and an increase in the consumption/production ratio. However, we find the opposite which suggests that Italy, due to low stringency of environmental regulation and to lack of enforcement, is behaving as a pollution haven within the EU (Marin and Maz   zanti, in press).

Spain is characterized by the opposite situation and pattern. For all emissions/ years the consumption/production ratio is greater (often far greater) than 1 and the ratio tends to increase in time, reaching the maximum for SO_x in 2000 with 1.395. This means that SO_x emissions induced by domestic final demand are 39.5 per cent greater than SO_x emissions directly generated by Spanish industries. These results are in line with the findings of Arto et al. (2010) and Serrano and Dietzenbacher (2010).

Spain had a very dynamic economy during the 1990s and the early 2000s, with growth mainly driven by the construction and tertiary sectors whereas the share of manufacturing in employment, output and value added has declined steadily.²⁴ This process, coupled with an increased volume of final demand of manufacturing goods (Roca and Serrano, 2007), gave rise to a rapid increase in foreign emissions to produce these goods, thus worsening the balance of emissions embodied in import.

Aggregation Bias

In the following paragraphs, we discuss to what extent the estimates of the consumption perspective change when aggregating our base data, in a smaller subset of sector, with regard to the benchmark level.

Figures 8.7 and 8.8 show the relative magnitude of the bias in the consumption perspective emissions arising from the aggregation of sectors into 30 NACE Rev 1.1 (see also Appendix B and Table 8.A.1) sub-sections and in 16 sectors according to the IEA/OECD studies²⁵ in the Italian case.

Figure  8.7  Aggregation bias %: 30 vs. 50 sectors (Italy).

Figure  8.8  Aggregation bias %: 16 vs. 50 sectors (Italy).

First note that, with few exceptions (CO₂ in 1995 and CO in 1995 and 2000 for the 30-sector aggregation), aggregation tends to overestimate the consumption perspective and this effect is even more evident in the 16-sector aggregation. Moreover, the bias tends to increase in time. The bias tends to be greater for the 16-sector aggregation as opposed to the 30-sector aggregation.²⁶

With regard to the 16-sector aggregation, the magnitude of the bias is particularly evident for SO_x (with a maximum bias of almost 40 per cent in 2005) and it is also relevant for NMVOC, CO₂ and NO_x.

The detailed estimates of the consumption/production perspective ratio for the different levels of aggregation (Table 8.5) show to what extent the aggregation bias is likely to affect our main synthetic indicator, the consumption/production perspective ratio. In all cases (again except CO), moving from the benchmark result (50 sectors) to the result for 16 sectors (to be compared with the set of IEA/OECD multi-regional analyses) artificially makes Italy a net exporter of emissions even within the framework of a pure DTA. Moreover, the relative gap between consumption and production perspectives in the 16-sector case in 2005 becomes quite high for SO_x (+21.6 per cent), NMVOC (+7.9 per cent) and CO₂ (+7.7 per cent),²⁷ suggesting that Italy is a net exporter of emissions. This result is just an artefact because the choice of the sectors to aggregate, although driven by constraints in data availability in IEA/OECD studies, is arbitrary. We also tried different apparently reasonable aggregation and obtained quite volatile results.

Table  8.5  Consumption/production perspective emissions for Italy according to different levels of aggregation

Figures 8.9 and 8.10 report the relative aggregation bias for Spain. Results for Spain are less straightforward than the Italian ones. The bias for the 30-sector aggregation is generally negative (with the only exceptions of very small positive biases for NO_x in 1995) and it is particularly high for NMVOC. No clear trend is found from 1995 to 2005. Moving to the bias for the 16-sector aggregation, it is generally positive (except for NMVOC for which it remains negative, though less important relative to the 30-sector aggregation). Moreover, it tends to decrease in time for CO₂, NO_x and SO_x and to increase for CO.

Figure  8.9  Aggregation bias %: 30 vs. 50 sectors (Spain).

Figure  8.10  Aggregation bias %: 16 vs. 50 sectors (Spain).

Unlike the Italian case, aggregation does not alter the status of Spain as net exporter of emissions for the full set of emissions and years (Table 8.6).

Table  8.6  Consumption/production perspective emissions for Spain according to different levels of aggregation

Comparison with Previous Studies

In the last decade, as previously indicated, some empirical studies have been conducted focusing on carbon or other pollutants embodiments in trade using international-comparable data especially from OECD sources (input–output, CO₂ emissions and bilateral trade) (e.g. Nakano et al., 2009; Ahmad and Wyckoff, 2003), Eurostat sources (e.g. Moll et al., 2007) and single-country sources (e.g. Arto et al., 2010 and Serrano and Dietzenbacher, 2010 for Spain; Su et al., 2010 for China).

Among them, for comparison purposes, we only consider those that include Italy or Spain or both. However, it is important to underline that some of the considered studies were defined with a global and cross-country analysis focus. Ahmad and Wyckoff (2003) in their OECD study consider 24 countries responsible in 1995 for 80 per cent of global emissions and global GDP (in nominal prices); following this study, Nakano et al. (2009) increase the former OECD analysis to 41 countries/regions so that more than 90 per cent of world GDP is covered. The study of Moll et al. (2007) includes eight EU countries²⁸ selected on the basis of data availability and the high coverage purpose of European economic contexts.

A comparison of our CO₂ results with the empirical evidence for the same pollutant found in the recent EE-IOA studies suggests that as far as the Italian case is concerned (Table 8.7), some literature references could be affected by aggregation bias due to a small number of considered sectors. This results in a strong and significant difference among empirical findings with respect to both the consumption and production perspective emissions and the corresponding ratio. In Nakano et al. (2009) and Ahmad and Wyckoff (2003),²⁹ the C/P ratios reported for the Italian case, in 1995 and 2000, are larger than ours and always higher than 1. The Moll et al. (2007) figure is the closest to our 2000 figure for the C/P ratio (0.96); they use a 38-sector aggregation level and if we consider the sensitive results found by Su et al. (2010) and discussed in the second paragraph (levels around 40 sectors appear to be sufficient to capture the overall share of emissions embodied in a country’s export), it may be considered more reliable than other authors’ findings. From a policy point of view, a C/P ratio that ranges from 1.24–1.30 to 0.96–0.97 suggests that, while large studies that involve several countries have to be encouraged because they permit macro area analysis, if they require a low level of sectoral detail to ensure countries’ homogeneity and comparability, their empirical results require caution when they are interpreted.

Table  8.7  CO2 emissions for production and consumption perspective in Italy in different

Notes

1  Author(s), year, DTA or MRIO, aggregation level (#sectors);

2  1992.

Table 8.8 shows a similar comparison for Spain. With regard to this country, the empirical findings reported in the different studies are more homogeneous than the Italian case both for the absolute values of production and consumption perspective CO₂ emissions and the corresponding ratio. This could be interpreted, at least partially, as a confirmation of the reliability of our 50-sector estimates. However, in the light of the Italian results, we could conclude that after a certain degree of aggregation, there is a concrete risk of having biased and volatile results which depend on the specificities of the economic structure of the country and the type of emission considered.

Table  8.8  CO2 emissions for production and consumption perspective in Spain in different studies (Mton CO2)

Notes

1  Author(s), year, DTA or MRIO, aggregation level (#sectors);

2  MtCO₂e;

3  MtCO₂e with Monetary DTA;

4  MtCO₂e with Physical DTA.

Conclusions

The integration of the National Accounting Matrix including Environmental Accounts (NAMEA) and input output (I–O) tables (often referred to as Environmentally extended input–output Analysis – EE-IOA – based on NAMEA data) represents a new way to analyse the determinants of the income-environment relationships in international settings. Moreover, EE-IOA provides analyses of the emissions embodied in domestic consumption and domestic production by considering the structure of intermediate inputs and environmental efficiency in each production sector.

A comparison of a production and consumption perspective may have relevant policy implications. A consumption and production emission ratio greater than 1 denotes a country that is a net exporter of emissions in the sense that it requires an amount of emissions embodied in imports, and thus produced abroad, that is greater than the amount of emissions embodied in export. Usually, the environmental policy points mainly to production activities as responsible actors of impacts to be targeted by legislation and regulation. Looking at the final consumption demand for vertically integrated domestic and international environmental impacts can push policy attention towards the possible role of consumers as actors to be targeted with particular environmental policies, together with the international responsibility for environmental externalities of pollutants’ emissions produced abroad but domestically demanded.

However, similar comparisons require particular assumptions, such as the technology associated with the imported goods, and could be affected by some biases. In this chapter we have analysed and discussed the aggregation bias due to different levels of production sector aggregation for Italy and Spain in 1995, 2000 and 2005. Our empirical findings, for the Italian and the Spanish cases, show that different sectoral aggregation significantly biases the amount of emissions both for the consumption and the production perspective. At the level where we consider only 16 production sectors, the results obtained in both the consumption and production perspective are quite different from those for higher levels of sector disaggregation (e.g. 50 which is our benchmark) both for the amounts of calculated emissions and for the corresponding C/P ratios. With regard to Italy, the 16-sector aggregation level in 2005 shows an emissions amount for CO₂, NO_x and NMVOC which is more than 10 per cent higher than those calculated with the 50-sector aggregation level. Moreover, considering SO_x, the gap between 16-and 50-sector aggregation reaches almost 40 per cent. With regard to Spain, between 16-and 50-sector aggregation levels in 2005, there are differences of below +5 per cent for CO₂, NO_x and SO_x, and almost 5 per cent for CO. NMVOC shows the biggest gap for the Spanish case with an under-estimation of almost –8 per cent compared with the benchmark aggregation level due to the use of a 16-sector aggregation level.

Our results suggest that special attention must be paid when interpreting the EE-IOA of country estimated amounts of embodied emissions, both in domestic final demand and those directly associated with the production sectors when the sectoral aggregation level has a low definition as considered in some recent similar studies.

Appendix A

The methodology we used to employ in a consistent way commodity-by-commodity input–output tables as a proxy of industry-by-industry tables is explained above. While the main analysis relies on results obtained using commodity-by-commodity input–output tables, in this appendix we report the differences between the industry-by-industry approach and the commodity-by-commodity approach as regards the estimation of the emissions induced by domestic demand. This comparison is only possible for Italy because Spain does not publish industry-by-industry input–output tables. The main results are summarized in Table 8.A.1.

Table  8.A.1  Commodity-by-commodity (cc) versus industry-by-industry (ii) approach for Italy (1-ii/cc) (%)

With the only exception of CO emissions, the absolute value of the gap for aggregate consumption perspective emissions is always below 1 per cent. On average, the commodity-by-commodity approach tends to under-estimate the emissions induced by the final demand of agriculture/fishing goods and industrial goods whereas it over-estimates the emissions induced by the final demand of services. Finally, we do not observe relevant changes in the magnitude of the gaps over time.

Appendix B

Table  8.B.1  NACE Rev. 1.1; 2-digit

01	Agriculture, hunting and related service activities
02	Forestry, logging and related service activities
05	Fishing, fish farming and related service activities
10	Mining of coal and lignite; extraction of peat
11	Extraction of crude petroleum and natural gas; service activities incidental to oil and gas extraction, excluding surveying
12	Mining of uranium and thorium ores
13	Mining of metal ores
14	Other mining and quarrying
15	Manufacture of food products and beverages
16	Manufacture of tobacco products
17	Manufacture of textiles
18	Manufacture of wearing apparel; dressing and dyeing of fur
19	Tanning and dressing of leather; manufacture of luggage, handbags, saddlery, harness and footwear
20	Manufacture of wood and of products of wood and cork, except furniture; manufacture of articles of straw and plaiting materials
21	Manufacture of pulp, paper and paper products
22	Publishing, printing and reproduction of recorded media
23	Manufacture of coke, refined petroleum products and nuclear fuel
24	Manufacture of chemicals and chemical products
25	Manufacture of rubber and plastic products
26	Manufacture of other non-metallic mineral products
27	Manufacture of basic metals
28	Manufacture of fabricated metal products, except machinery and equipment
29	Manufacture of machinery and equipment n.e.c.
30	Manufacture of office machinery and computers
31	Manufacture of electrical machinery and apparatus n.e.c.
32	Manufacture of radio, television and communication equipment and apparatus
33	Manufacture of medical, precision and optical instruments, watches and clocks
34	Manufacture of motor vehicles, trailers and semi-trailers
35	Manufacture of other transport equipment
36	Manufacture of furniture; manufacturing n.e.c.
37	Recycling
40	Electricity, gas, steam and hot water supply
41	Collection, purification and distribution of water
45	Construction
50	Sale, maintenance and repair of motor vehicles and motorcycles; retail sale of automotive fuel
51	Wholesale trade and commission trade, except of motor vehicles and motorcycles
52	Retail trade, except of motor vehicles and motorcycles; repair of personal and household goods
55	Hotels and restaurants
60	Land transport; transport via pipelines
61	Water transport
62	Air transport
63	Supporting and auxiliary transport activities; activities of travel agencies
64	Post and telecommunications
65	Financial intermediation, except insurance and pension funding
66	Insurance and pension funding, except compulsory social security
67	Activities auxiliary to financial intermediation
70	Real estate activities
71	Renting of machinery and equipment without operator and of personal and household goods
72	Computer and related activities
73	Research and development
74	Other business activities
75	Public administration and defence; compulsory social security
80	Education
85	Health and social work
90	Sewage and refuse disposal, sanitation and similar activities
91	Activities of membership organizations n.e.c.
92	Recreational, cultural and sporting activities
93	Other service activities

Acknowledgements

The authors thank Roberto Zoboli for its very useful comments and suggestions. Roberto presented a very preliminary version of this work to the conference ‘The Structure of Economic Systems Through Input–Output Applications’ (Accademia dei Lincei, Rome, Italy) in October 2010. The authors also want to acknowledge the Environmental Accounting Unit of ISTAT (Italy) for its very valuable work in regularly providing the NAMEA series. The usual disclaimer applies.

Notes

1  Some main streams of research can be outlined: I–O models accounting for trade and embodied emissions (through energy accounts); global multi-region input–output (MRIO) model; extension for eco-footprint analysis; comparing physical trade balance (PTB) and pollution trade balance (UTB) associated with fossil use; analysing pollution terms of trade, pollution haven tests; analysing I–O tables linked with satellite accounts. For brevity, we refer the reader to the mounting, extensive literature that is also touched on by many contributions in this volume.

2  The physical DTA refers to the use of imports in physical quantities and using, for imports, the same physical environmental coefficient (emissions per kg of import) as domestic physical environmental coefficients (emissions per kg of domestic output). This assumes that, although of different quality (value per physical unit), the emissions content of goods is closely correlated to its weight and less correlated to its value.

3  Automobile driving and related maintenance activities are by far the largest contributing products to total environmental impacts by consumption in the EU25. However, by summing several animal-based foods (meat, meat products, poultry, dairy products), animal food products would become dominant. At the aggregate level of 12 consumption domains, food already emerges as the largest contributor to environmental problems.

4  The consumption-based emissions are computed using production-based emissions minus the emissions embodied in exports plus those embodied in imports.

5  For an exhaustive review on the accounting definitions related to EE-IOA, the reader is referred to Serrano and Dietzenbacher (2010) and Moll et al. (2007).

6  In this section we refer to the aggregation of basic data as opposed to the aggregation of results. The aggregation of results of any empirical analysis in economics is a necessary step when giving an overall picture of the phenomenon under analysis.

7  Due to privacy protection, ISTAT, the Italian National Institute of Statistics, is not allowed to publish data for aggregates with less than three units and it is forced to further aggregate these branches.

8  The aggregation to the minimum common standard is the most widely used approach (Ahmad and Wyckoff, 2003; Nakano et al., 2009). However, a noticeable exception is represented by Huppes et al. (2005) who exploit the very detailed US input–output table and adapt it to the EU economic structure, thus using more disaggregated data relative to publicly available EU input–output tables. Although very interesting, this approach is affected by problems related to differences between US and EU classification structures within each macro-industry.

9  The theoretical results obtained by Morimoto (1970) do not depend on the reason that induces aggregation.

10  In our case, the aggregation bias is defined in terms of emissions induced by domestic demand. However, for the sake of brevity, we will discuss overall aggregation bias only. Results for the assessment of the aggregation bias at sector/commodity level are available upon request.

11  An important point, which often remains implicit, is that the aggregation bias only arises when the vector of final demand is modified relative to the original vector of final demand.

12  This level of disaggregation corresponds roughly to the two-digit NACE Rev. 1.1 classification (see Table 8.B.1 for a description of each sector).

13  OECD/IEA estimates use a disaggregation of 17 sectors. However, both OECD input–output tables and IEA CO₂ emissions from fuel combustion go beyond the two-digit NACE Rev. 1.1 as regards sector 27. This sector is split into ‘Iron and steel’ (271+2731) and ‘Non-ferrous metals’ (272+2732). On the contrary, Italian and Spanish input–output tables and NAMEA do not allow this separation.

14  They employ an environmentally extended MRIO model for six OECD countries (US, Canada, France, Germany, Japan and UK) to estimate the embodiment of carbon in imports of manufacturing products.

15  Note that the benchmark results are obtained by ‘disaggregating’ the original vector of emissions intensities (42 sectors) in order to meet the 122-sector aggregation of the input–output tables. This operation is likely to partly affect the reliability of the estimates for the 122-sector aggregation.

16  Note that when the make matrix is diagonal (that is, when all industries produce only their primary commodity), then the C matrix is an identity matrix.

17  This comparison is not feasible for Spain because the Instituto Nacional de Estadistica (INE) does not produce industry-by-industry input–output tables.

18  For an overview of the methodological issues related to NAMEA, we also refer the reader to Chapters 1 and 2.

19  We also perform all the estimates for 12 additional substances available in the Italian NAMEA only (NH₃, PM₁₀, PM_2.5, As, Cd, Cr, Cu, Hg, Ni, Pb, Se and Zn). Results are available upon request.

20  The Spanish NAMEA used in this chapter is available on the Eurostat website with a 50-sector aggregation and only five pollutants. The INE divulgates a NAMEA with even more pollutants but with only 30 sectors and for this reason is not useful for our purposes.

21  Agriculture and fishing (A-B NACE Rev. 1.1), Industry (C-F NACE Rev. 1.1) and Services (G-O NACE Rev. 1.1). Results at two-digit NACE are available upon request.

22  The equivalence is explained in Serrano and Dietzenbacher (2010).

23  Among others, at EU level, the Council Directive 1980/779/EC substituted by the Council Directive 1999/30/EC of 22 April 1999 ‘relating to limit values for sulphur dioxide, nitrogen dioxide and oxides of nitrogen, particulate matter and lead in ambient air’, the Council Directive 85/203/EEC of 7 March 1985 ‘on air quality standards for nitrogen dioxide’, as last amended by Council Directive 85/580/EEC and the Council Directive 1999/13/EC ‘on the limitation of emissions of volatile organic compounds due to the use of organic solvents in certain activities and installations’.

24  The output share of manufacturing was 32.6 per cent, 31.1 per cent and 26.7 per cent in 1995, 2000 and 2005 respectively.

25  IEA/OECD studies such as Nakano et al. (2009) and Ahmad and Wyckoff (2003) use a disaggregation of 17 sectors which, for sector 27 (Manufacture of basic metals), goes beyond the two-digit detail. IEA/OECD data distinguish between ‘Iron and steel’ (27.1 and 27.31) and ‘Non-ferrous metals’ (27.2 and 27.32). On the contrary, input–output tables and NAMEA published by ISTAT and INE treat sector 27 as a unique sector. This aggregation potentially introduces a bias in our results due to the high emissions intensity of sector 27 and to the heterogeneity in technologies and emissions intensity within sector 27.

26  Note that there is no perfect link between the 16-sector aggregation and the 30-sector aggregation. This fact does not allow the monotonicity of the bias with respect to the number of sectors to be interpreted as a stylized fact. In fact, monotonicity is not found for Spain.

27  The figure for the benchmark case of the 50-sector disaggregation was of –12.9 percent for SO_x, –4.6 percent for NMVOC and –3.6 percent for CO₂.

28  The selected eight economies represent more than two-thirds of EU25’s GDP and more than 60 per cent of EU25’s population. The geographical coverage comprises Spain, UK (1995) and Germany, Denmark, Hungary, Italy, the Netherlands, Sweden (1995 and 2000).

29  A consistent comparison of the absolute levels of CO₂ emissions between IEA/OECD studies and NAMEA-based studies is not possible. In fact, IEA records CO₂ emissions from fuel combustion only and, differently from NAMEA, the principle of recording the emissions generated by resident agents only is not applied in the collection of these data.

References

Ahmad, N. and Wyckoff, A. (2003) ‘Carbon dioxide emissions embodied in international trade of goods’, OECD Science, Technology and Industry Working Papers 2003/15, Paris: OECD.

Alcántara, V. and Padilla, E. (2009) ‘Input–output subsystems and pollution: an application to the service sector and CO₂ emissions in Spain’, Ecological Economics, 69: 905–914.

Andrew, R., Peters, G.P. and Lennox, J. (2009) ‘Approximation and regional aggregation in multi-regional input–output analysis for national carbon footprint accounting’, Economic Systems Research, 21: 311–335.

Arto, I., Roca, J. and Serrano, M. (2010) ‘Accounting for embodied emissions in international trade: physical vs monetary trade flows’, mimeo.

Basina, A. and Sfetsos, A. (2010) ‘Environmental impacts of generating electricity by substituting lignite with photovoltaic technology: an analysis based on a NAMEA-table for the Greek economy’, paper presented at the I–O Society Conference, Sydney, 20–25 June.

Casler, S.D. and Rose, A. (1998) ‘Carbon dioxide emissions in the U.S. economy’, Environmental and Resource Economics, 11: 349–363.

Cole, M.A. (2004) ‘Trade, the pollution haven hypothesis and the environmental Kuznets curve: examining the linkages’, Ecological Economics, 48(1): 71–81.

Copeland, B.R. and Taylor, M.S. (2004) ‘Trade, growth and the environment’, Journal of Economic Literature, 42: 7–71.

De Boo, A., Bosch, P., Gorter, C.N. and Keuning, S.J. (1991) ‘An environmental module and the complete system of national accounts’, Occasional Papers of the CBS, No. NA-046, Voorburg.

de Haan, M. and Keuning, S.J. (1996) ‘Taking the environment into account: the NAMEA approach’, Review of Income and Wealth, 42: 131–148.

de Haan, M. (2004) ‘Accounting for goods and for bads: measuring environmental pressure in a national accounts framework’, PhD thesis, Universiteit Twente.

Eurostat (2008) ‘Eurostat manual of supply, use and input–output tables’, Eurostat Methodologies and Working Papers, Luxembourg: Eurostat.

Frankel, J. and Rose, A. (2005) ‘Is trade good or bad for the environment?’ Review of Economics and Statistics, 87: 85–91.

Gallego, B. and Lenzen, M. (2005) ‘A consistent input–output formulation of shared producer and consumer responsibility’, Economic Systems Research, 17: 365–391.

Harris, R. (2001) ‘Methods for estimating air emissions from the production of goods imported into the UK’, prepared for DG Regional Policy and Eurostat, Working Paper 2/2001/B/5, Eurostat, Brussels: European Commission.

Hatanaka, M. (1952) ‘Note on consolidation within a Leontief system’, Econometrica, 20: 301–303.

Hoekstra, R. (2010) ‘A complete overview of environmental input–output articles’, paper presented at the I–O Society Conference, Sydney, 20–25 June.

Hoekstra, R., Smits, J.P. and van der Helm, R. (2010) ‘Economic growth, structural change and environmental pressures: the case of the Netherlands, 1950–2009’, paper presented at the I–O Society Conference, Sydney, 20–25 June.

Huppes, G., de Koning, A., Suh, S., Heijungs, R., Oers, L., van Nielsen, P. and Guinee, J. (2005) ‘Environmental impacts of consumption in the EU: high resolution input ouput tables with detailed environmental extensions’, Journal of Industrial Ecology, 10: 129–146.

IPTS (2006) ‘Environmental Impacts of Products (EIPRO)’, IPTS/ESTO project IPTS, European Commission technical report series, Seville: IPTS.

Jacobsen, H.K. (2000) ‘Energy demand, structural change and trade: a decomposition analysis of the Danish manufacturing industry’, Economic Systems Research, 12: 319–343.

Kagawa, S. and Inamura, H. (2001) ‘A structural decomposition of energy consumption based on a hybrid rectangular input–output framework: Japan’s case’, Economic Systems Research, 13: 339–363.

Kagawa, S. and Inamura, H. (2004) ‘A spatial structural decomposition analysis of Chinese and Japanese energy demand: 1985–1990’, Economic Systems Research, 16: 279–299.

Lenzen, M. (2011) ‘Aggregation versus disaggregation in input–output analysis of the environment’, Economic Systems Research, 23: 73–89.

Marin, G. and Mazzanti, M. (in press), ‘The evolution of environmental and labor productivity dynamics: sector analyses and decoupling/recoupling trends on a 1990–2007 Italian NAMEA’, Journal of Evolutionary Economics, in press.

Mazzanti, M. and Montini, A. (2010) ‘Embedding emission efficiency at regional level: analyses using NAMEA’, Ecological Economics, 69(12): 2457–2467.

Miller, R.E. and Blair, P.D. (2009) Input–Output Analysis: Foundations and Extensions, 2nd edn, Cambridge: Cambridge University Press.

Moll, S., Vrgoc, M., Watson, D., Femia, A., Pedersen, O.G. and Villanueva, A. (2007) ‘Environmental input–output analyses based on NAMEA data: a comparative European study on environmental pressures arising from consumption and production patterns’, ETC/RWM Working Paper 2/2007, Copenhagen: European Topic Centre – Resource and Waste Management.

Morimoto, Y. (1970) ‘On aggregation problems in input–output analysis’, Review of Economic Studies, 37: 119–126.

Nakano, S., Okamura, A., Sukurai, N., Suzuki, M., Tojo, Y. and Yamano, N. (2009) ‘The measurement of CO₂ embodiments in international trade: evidence from the harmonised input–output and bilateral trade database’, Statistical Analysis of Science, Technology and Industry Working Paper 2009/3, Paris: OECD.

Peters, G. (2007) ‘From production-based to consumption-based national emission inventories’, Ecological Economics, 65: 13–23.

Roca, J. and Serrano, M. (2007) ‘Income growth and atmospheric pollution in Spain: an input–output approach’, Ecological Economics, 63: 230–242.

Rormose, P., Olsen, T. and Hansen, D. (2010) ‘Emissions embodied in Danish import: a unidirectional trade model with 51 countries’, paper presented at the I–O Society Conference, Sydney, 20–25 June.

Rose, A. and Casler, S. (1996) ‘Input–Output structural decomposition analysis: a critical appraisal’, Economic Systems Research, 8: 33–62.

Rueda-Cantuche, J.M. (2010) ‘Underestimation of the performance of the EU carbon dioxide emission reductions via external trade’, paper presented at the I–O Society Conference, Sydney, 20–25 June.

Serrano, M. and Dietzenbacher, E. (2010) ‘Responsibility and trade emission balances: an evaluation of approaches’, Ecological Economics, 69: 2224–2232.

Stauvermann, P.J. (2007) ‘The Regional Accounting Matrix including Environmental Accounts: methodology, application and interpretation – a Dutch pilot study’. Online: www.arpa.emr.it.

Su, B., Huang, H.C., Ang, B.W. and Zhou, P. (2010) ‘Input–Output analysis of CO₂ emissions embodied in trade: the effects of sector aggregation’, Energy Economics, 32: 166–175.

Turner, K., Lenzen, M. and Wiedmann, J.B. (2007) ‘Examining the global environmental impact of regional consumption activities – part 1: a technical note on combining input–output and ecological footprint analysis’, Ecological Economics, 62: 37–44.

Watson, D. and Moll, S. (2008) ‘Environmental benefits and disadvantages of economic specialisation within global markets, and implications for SCP monitoring’, paper presented at the SCORE! conference, Brussels, Belgium, 10–11 March.

Wier, M. (1998) ‘Sources of changes in emissions from energy: a structural decomposition analysis’, Economic Systems Research, 10: 99–112.

Wyckoff, A.W. and Roop, J.M. (1994) ‘The embodiment of carbon in imports of manufactured products: implications for international agreements on greenhouse gases emissions’, Energy Policy, 22: 187–194.