Do Spanish energy efficiency actions trigger JEVON’S paradox?

Do Spanish energy efficiency actions trigger JEVON’S paradox? Jos e M. Cansino a , b , * , Rocío Rom an-Collado a , b , Jos e Merch an a a Universidad de Sevilla, Spain b Universidad Aut onoma de Chile, Chile article info Article history: Received 31 August 2018 Received in revised form 7 May 2019 Accepted 28 May 2019 Available online 31 May 2019 Keywords: Jevon's paradox Backfire effect Rebound effect Energy efficiency LMDI abstract This paper explores whether the changes in energy intensity in Spain have led to improvements in the energy consumption levels or to a backfire effect offsetting the expected decrease. Jevon's paradox or backfire effect happens when a rebound higher than 100% causes energy efficiency improvement to raise energy consumption. To test Jevon's paradox or the backfire effect caused by energy efficiency actions, a Logarithmic Mean Divisia Index I (LMDI-I) is used for the Spanish economy. The period under consideration ranges from 2000 to 2015, when three national action plans were implemented. The main methodological novelty of this paper consists on to block the economic activity effect in a second decomposition round to better explore the effect of energy efficiency on energy consumption. As a whole, our results do not support Jevon's paradox for the sectors analyzed. However, they do warn about a possible backfire effect in the industry, transport and service sectors. Major findings follow these same findings when the activity effect is blocked. ©2019 Elsevier Ltd. All rights reserved. 1. Introduction Energy efficiency has become a primary energy policy goal in many countries conditioning their policies towards energyintensive sectors [1e3]. However, Energy Efficiency Improvements (EEI) could lead to changes in the demand for energy services that offset some of the achieved energy savings in the form of rebound effects. Rebound effect has been defined as the additional energy consumption from overall changes in demand as a result of behavioral and other systemic responses to EEIs [4e7]. The idea of the energy rebound effect dates back to 1865; literature also refers to the rebound effect as “Jevons' Paradox” whereby a technology that enhances efficiency does not necessarily lead to less consumption of that resource [8]. The first to study this rebound effect phenomenon in the form of economic literature were Brookes &Khazzoom, thus, the “Jevons’Paradox”is also known as the “KhazzoomeBrookes”Postulate [9,10]. [11,12] identified three types of rebound effects. These are direct rebound–affecting the energy consumption of the energy service affected by the energy efficiency improvement, indirect rebound effect–affecting other goods, services and areas and economy-wide effects which go even further, and affect prices, quantities and produce global economic readjustments. This paper focus on testing Jevon's paradox or KhazzoomeBrookes Postulate for Spain's economy, when this is conceived as a backfire effect caused by energy efficiency actions. The reason why Spain has been chosen is due to its highest energy intensity when is compared to the average of the rest of European Union's Member States. Concretely, in 2008, Spanish energy intensity level was 19% upper the EU-15 average [13,14]. Since then, the Spanish Government has implemented several energy efficiency plans in order to reduce energy intensity. The period under consideration in this research is 2000e2015 and contains three national energy efficiency action plans. The last one includes mandatory actions established by Ref. [15]; which implies a 20% saving on energy consumption levels up to 2020. Most of the measures included in Spain's plans of action are based on technological improvements and are energy intensity oriented. These plans covered the periods 2005e2007, 2008e2012, 2011e2020 (updated after [15] for the period 2014e2020) [16e19]. An in-depth analysis of Spain's Spanish plans of action, including detailed information about measures in force, is due to Rom an-Collado &Sanz [20]. Although many papers do not differentiate between types of rebound effect, whether direct or indirect effects, they are easily recognizable. Following Vivanco et al., [21]; an example of the rebound effect is defined as the way in which fuel efficiency improvements in passenger cars have made driving cheaper, thus *Corresponding author. Universidad de Sevilla, Spain. E-mail address: [email protected] (J.M. Cansino). Contents lists available at ScienceDirect Energy journal homepage: www.elsevier.com/locate/energy https://doi.org/10.1016/j.energy.2019.05.210 0360-5442/©2019 Elsevier Ltd. All rights reserved. Energy 181 (2019) 760e770 resulting in users driving more and buying larger cars (direct effect) and/or spending the remaining savings on other products (indirect effect). As a result, total fuel and energy savings are reduced. In the latter case, a backfire effect is defined as a rebound higher than 100% which implies energy efficiency improvement acting on energy intensity that does not reduce but rather increase energy consumption [9]. To test Jevon's paradox or the backfire effect on energy consumption caused by energy efficiency actions, a Logarithmic Mean Divisia Index I (LMDI-I) is used. Five decomposition factors were considered: population, economic activity, economic structure, energy intensity and energy resource mix. Conducted analysis is a multisectoral one, including the six sectors where Spain's national plans were focused (agriculture, industry, energy, transport, services and residential). Five types of fuels were considered for the energy resource mix (oil, gas, renewable energy sources, electricity and solid fuels). This research uses a valuable paper by Inglesi-Lotz [22] as the start point, to then go further. The aim of this paper is not to estimate the precise rebound effect caused by the implemented energy efficiency measures but to decompose the determinants of energy consumption changes in Spain from the backfire effect perspective. What can be derived from the results herein is whether the changes in energy intensity have led to improvements in the energy consumption levels or to a backfire, the effects of which have managed to offset the expected decrease in energy consumption with improvements in energy intensity. The first contribution of this paper to the literature is from a methodological perspective. Concretely, the methodological approach implemented has allowed us to block the economic activity effect in a second decomposition round to better explore the effect of energy efficiency on energy consumption. The theoretical support for this second-round decomposition is that EEI does not act in a partial equilibrium scenario where all remains equal. Changes in population or in the energy resource matrix happen at the same time as EEI. In fact, the energy rebound effect indicates that technological progress not only improves energy efficiency, but also promotes economic growth, therefore raising the demand for energy; so it is interesting to block the effect of economic growth on energy consumption to observe whether Jevon's paradox occurs or not. The conducted research also contributed to bridge the literature's gap being the first in testing Jevon's paradox (or KhazzoomeBrookes Postulate) for Spanish economy from a multisectoral and expost perspective and through an ex post analysis. The rest of the paper is organized as follows; Section 2offers a review of the various approaches revolving around rebound effect in the literature. Section 3exposes methodological approach developed in the paper. Dataset is detailed in Section 4, while the results obtained are presented in Section 5. Section 6offers the conclusions reached. 2. Literature review There is a broad list of methodological approaches to assess the rebound effect. Computable general equilibrium (CGE) models have been widely used to capture the economy-wide rebound effect providing exante analysis. Turner [23] used a CGE framework to investigate the conditions under which rebound effects may occur in response to increases in energy efficiency in the UK national economy. Allan et al. [24] and Hanley et al. [25] worked with a similar approach. Turner &Hanley [26] used a CGE model of the Scottish economy to consider the factors influencing the impacts of one form of technological changeeimprovements in energy efficiencyeon absolute levels of CO 2 emissions, on the carbon intensity of the economy, and the per capita Environmental Kuznets Curve relationship. The paper by Yu et al. [27], must be also cited. It is also the case of Broberg et al. [28] assessing the economic wide rebound effect for the industry sector in Sweden. More recently, Bye et al. [29] used a multi-sector CGE of the Norwegian economy to explore the cost, emission and energy rebound effects of alternative interpretations of the policy underlying the proposed 2030 energy efficiency goal. Grepperud, &Rasmussen [30] also used a CGE approach for the Norwegian economy. A CGE based analysis for the Spanish case is due to Guerra &Sancho [31]. An econometric approach also shows intensive use in expost analysis. Berkhout et al. [32] might be cited as one of the reference paper. De Borger et al. [33] estimated an energy demand model in first differences to assess the rebound effect's magnitude in YRDUA's industrial sectors using dynamic ordinary least squares and seemingly unrelated regression methods. Llorca &Jamasb [1] analyzed the energy efficiency and rebound effects for road freight transport in 15 European countries during the 1992e2012 period using a stochastic frontier analysis approach and examined the influence of key features of rebound effect in the road freight transport sector. Freire-Gonz alez [34] developed a hybrid methodology that combines econometric estimates, environmental extended input-output analysis and re-spending models that have been developed. As a part of this econometric approach, total factor productivity (TFP) is usually the way to represent the contribution of technological progress. The Cobb-Douglas and CES production function is employed to estimate total factor productivity. Hediger et al. [35] linked econometrics with the emerging topic of Behavioral Economics by researching how households respond to efficiency improvements in their heating system. Not so far from Behavioral Economics, Santarius &Soland [36] exposed the existing rebound discourse to psychological theories. Finally, analysis based on Index Decomposition Analysis (IDA) either in Structural Decomposition Analysis (SDA) as in LMDI version may be found in Alc antara et al. [37]; Cansino et al. [38]efor SDA–Cansino et al. [39]; Lin et al. [40]; Wang et al. [41]; Lin &Liu [42]; Inglesi-Lotz [22,43] for LMDI. Rom an-Collado &Colinet [44]is the most accurate and recent decomposition of energy consumption in Spain from a LMDI approach. Those are expost analysis. 3. Methodology The LMDI-I decomposition model belongs to IDA family. It is a useful tool because of its theoretical support and feasibility. It satisfies relevant properties [45e47]. LMDI-I is complete in decomposition and consistent in aggregation while at the same time excellent in handling zero values [48e50]. Ang &Liu [51] and Ang & Choi [52] analyzed and proposed that the best way to handle it is by substituting zeros for a d value between 10 10 and 10 20 . Ang &Liu [51] also showed that this strategy is robust when an appropriate value is used, and that it provides satisfactory results even for highly extreme cases. Upon comparing various index decomposition analysis methods, Ang [45] concluded that the LMDI method to be the most effective. To analysis whether energy intensity improvements caused the backfired effect on Spain's energy consumption at the sectoral level i, Eq [1] decomposes its energy consumption (E i ) as follows, Ei¼X n j¼1 P, Q P , Qi Q , Ei Qi , Eij Ei ¼X n j¼1 P,A,Si,Ii,Mij (1) where nis the number of final energy sources, E i the energy consumption of sector i,Pthe population, Qis the proxy variable that measures total output of the economy through the Gross Value Added (GVA), Q i is the GVA of sector i. Finally, E ij measures the J.M. Cansino et al. / Energy 181 (2019) 760e770 761 energy consumption of sector ifrom the j-th energy source. LMDI-I model does not present interactions between the variables but rather their relative contributions to the change of the energy consumption over time. From [1], the LMDI-I analysis allows the decomposition of changes in the energy consumption of sector i, between period t and t-1 ( D E i ) for each energy source included in Spain's energy matrix [45,48,53]. In its additive formulation, decomposition is carried out on five factors or effects as shown by the following expression: D Ei¼EiðtÞEiðt1Þ¼ D Piþ D Aiþ D Siþ D Iiþ D Mi(2) When an effect takes positive values, it acts as a driver of energy consumption; when values turn negative, then the effects act as a compensating factor of energy consumption. The definition of each factor is as follows: a) D P i : Population effect. It reflects part of the variation in sectoral energy consumption due to population changes. It acts as a proxy for the country's energy demand. b) D A i : Activity effect. Shows variations in sectoral energy consumption explained by changes in per capita income. c) D S i : Structure effect. It captures variations in sectoral energy consumption due to changes in the relative weight of sectors in production. d) D I i : Intensity effect. It measures variations in energy consumption due to changes in the sectoral energy intensity; that is to say, due to a change in energy consumption per produced unit of each sector between period t and t-1. Being that this is the key decomposition factor for the aim of the paper, in-depth details are provided below. e) D M i : Energy mix effect. Measures the contribution of the penetration of clean energy in the variation of final energy consumption. Each effect can be calculated as follows: D Pi¼X n j¼1 u ij ln PðtÞ Pðt1Þ (3) D Ai¼X n j¼1 u ij ln AðtÞ Aðt1Þ (4) D Si¼X n j¼1 u ijln SiðtÞ Siðt1Þ (5) D Ii¼X n j¼1 u ijln IiðtÞ Iiðt1Þ (6) D Mi¼X n j¼1 u ijln eijðtÞ eijðt1Þ (7) When applying the Mean Value Theorem, u ij corresponds to the weighting factor for the calculation of different effects and is obtained from the logarithmic mean, which considers the variation between two periods of sectoral energy consumption by energy source jaccording to [8]: u ij ¼EijðtÞEijðt1Þ lnEijðtÞlnEijðt1Þ (8) Sign and magnitude of D I i states whether changes in energy intensity have led to improvements in the energy consumption levels or, on the other hand, if the backfire effect has managed to offset the expected decrease in energy consumption with improvements in energy intensity. In other words, it represents if energy intensity decline leads to energy consumption increase or the contrary. As most of the changes in energy intensity specified in Spain's plans come from technological improvements, D I i is the proxy for technological change. However, following [43] and Chertow [54]; caution must be used with this decomposition factor. These authors recognize that there is little social theory to suggest how to specify and measure technological improvements. Chertow [54] states: “Conceptually as well as numerically, P, population, and A, defined as a per capita measure of wealth, consumption, or production, have generally been more accessible to researchers than the technological improvements term.”Without ignoring this limit, from the D I i sign and value, we derive whether the changes in energy intensity provoked backfire effect or not in energy consumption. Going beyond previous research, it is essential to take into account that energy intensity improvements do not all remain equal. In fact, as technological progress not only improves energy efficiency but also promotes economic growth, the activity effect ( D A i ) on energy consumption is partially due to the indirect rebound effect [21]. Therefore, a methodological development of LMDI to block the activity effect on energy consumption would give us finer information about whether the changes in energy intensity have led to improvements in energy consumption levels or to a backfire effect. To explore the, we have proceeded as follows. From [2], a second decomposition of the energy consumption allows us to know the impact of the improvement in the energy intensity once the activity effect is blocked [41,55e57]. To do that, it is assumes that economic growth causes energy consumption. Now D EE i is defined as the difference between changes in energy consumption for sector iand the activity effect D EEi¼ D Ei D Ai¼ D Piþ D Siþ D Iiþ D Mi(9) Rewriting Eq (9) g i INT in Eq (10) lets us calculate whether the changes in energy intensity causes a direct backfire effect on energy consumption between period t and t-1 once indirect backfire effect through economic activity changes is blocked. g i TOTAL ¼ D EEi D Ai¼ D Pi D Ai D Si D Ai D Ii D Ai D Mi D Ai ¼ g i Pþ g i EST þ g i INT þ g i MIX (10) As it is also possible that negative economic growth causes a decrease in energy consumption Diakoulaki &Mandaraka [56] rewrite Eq (10) as follows b i TOTAL ¼ D EEi D Ai D Ai¼ D Piþ D Siþ D Iiþ D Mi D Ai D Ai(11) b i TOTAL ¼ D Pi D Aiþ D Si D Aiþ D Ii D Aiþ D Mi D Ai D Ai D Ai ¼ b i Pþ b i EST þ b i INT þ b i MIX 1 (12) J.M. Cansino et al. / Energy 181 (2019) 760e770762 Eqs (10) and (12) give useful information regarding [15]. The reason is that this legal instrument promotes decoupling between energy consumption and economic growth. Eqs (10) and (12) show the energy consumption changes only attributed to the population, structural and intensity effects, excluding those provoked by the activity effect (where indirect rebound effect is embodied). Both indices measure the efforts that have been made in order to reduce energy consumption when the activity effect is blocked [56,58]. The interpretation of the results for both indices are the same. This let us to test if the changes in energy intensity causes a direct backfire effect once indirect backfire effect is blocked. In fact, g i TOTAL in Eq (10) and b i TOTAL (12) are a finest decoupling index between energy consumption and economic activity while g i P , g i EST , g i INT and g i MIX (also b i P , b i EST , b i INT and b i MIX ) act as partial decomposition index for population, economic structure, energy intensity and energy sources mix effects. Understanding g i TOTAL and b i TOTAL when 1 expresses strong decoupling efforts implies that sectoral energy consumption decreases at the same time that the economy grows. Assuming that g and b have the same meaning, if 0 < g i TOTAL <1 then it expresses weak decoupling efforts, so that the inhibiting effect of sectoral energy consumption is weaker than the effect of economic growth; thus, sectoral energy consumption grows together with the economy. Finally, if g i TOTAL 0, it expresses that there were no decoupling efforts [56,58,59]. According to Roinioti &Koroneos [58] values higher than 0 for g i INT (also for g i P , g i EST and g i MIX ) imply that energy intensity improvements do not causes direct backfire effect on energy consumption. Negative values advise a direct backfire effect. In order to simplify only g notation is used when showing results. 4. Dataset The sectoral energy consumption data measured in kilo tons of oil equivalent (Ktoe) come from the Energy Balances published by Eurostat [60]. The GVA data series, expressed in millions of euros at constant prices, was built up from the statistical series for the sectorial GVA at current prices and the chained volume indexes, published by the National Institute of Statistics (INE for the Spanish acronym). The base year is 2010. The figures for the population for the years 2000e2015 were also obtained from the INE [61,62]. The data for the energy transformation sector does not include the transportation and distribution of energy. The reason for this is that only in the last National Energy Efficiency Plan does the sector include these activities. In the previous plans, it was not included. For the inclusion of the residential sector in the analysis, the activities realized in households have been taken as a proxy. The correspondence of the sectors analyzed, national accounts and energy balances are detailed in Table A1 in annex. The energy sources included in the energy mix factor are solid fuels, oil, gas, renewables and electricity. Offered Table A2 in the annex provides detailed information about its distribution by sectors. 5. Results In Jevon's meaning, rebound effect reflects an individual person or group's behavioral inclination which should be stable for a while. However, obtained results show that the “rebound effect”fluctuates quite frequently-sometime it is negative while soon after it is positive. This is due to Jevons focused in England's consumption of coal after James Watt introduced the Watt steam engine, that means just one (although crucial) energy efficiency improvement. Currently, more than 3 million patent applications were filed worldwide [63]. Given this high number, when patents are incorporated as technological innovations to the productive processes it is expected that not all of them modify the behavior of the agents in the same direction. The index developed in this research captures the average effect on energy consumption for each of the sectors considered and for each of the years under analysis. The fluctuations respond to the high number of technological innovations and to the different reactions of the agents when they are incorporated into the productive activity. 5.1. First decomposition From a sectoral perspective, the results from first decomposition analysis advise against Jevon's paradox or backfire effect in the Fig. 1. Sectorial decomposition analysis. Source: Own elaboration. J.M. Cansino et al. / Energy 181 (2019) 760e770 763 agriculture sector (Fig. 1a). For most of the years under consideration, the energy efficiency effect showed high and negative values. The only exceptions corresponded to sub-period 2000e2005 and 2010e2012. The net effect derived from technological improvements help reduce energy consumption in this sector. On the other hand, the economic activity effect (partly caused by indirect rebound effect) acted as a driver of sectoral energy consumption. Behavior as a driver turned to an inhibitor role only for economic recession period 2007e2013. The results obtained are in line with Arocena et al. [13] using a CGE approach. They found a rebound effect around 30% (far from a backfire effect) for the four types of fuels considered (coal, refined oil, electricity and gas). As we do no develop a partial equilibrium analysis, LMDI's results show that other effects have to be considered when decomposing energy consumption changes. For example, population acted as a driver of energy consumption until demographic changes in 2012 when massive migrated people decided to go back to their countries of origin due to the economic crisis. The mix resource matrix effect also acted as an inhibitor and structure effect, but failed to show a clear pattern. Evidence of a possible backfire effect appeared in the industry sector (Fig. 1b) with the energy efficiency effect showing positive values in nine of the fifteen years under study. However, literature explains that when there is an economic crisis, energy intensity has difficulties in adjusting [38] and 2018). This is due fixed energy consumption levels decoupled from a number of output units (for example, to illuminate factories independently of their activity). Together with this, economic activity and population effects acted as with the agricultural sector while structure effects inhibited energy consumption for most of the years. Our findings are similar than those provided by Guerra and Sancho [31] but for the Spanish economy as a whole. The estimated a rebound effect in a range between 87.4% and 90.8%. The results obtained are also in line with Arocena et al. [13]. They found a rebound effect higher than 70% (close from a backfire effect). A possible backfire effect is rejected upon analyzing the energy sector (Fig. 1c). For the same period, population and economic activity effects acted as in the agriculture and industry sectors. On the contrary, the energy resource matrix effect showed an economy moving slowly to a cleaner energy resource matrix. Structure effect showed an erratic behavior. Arocena et al. [13] found evidence also rejecting possible backfire effect in the energy sector with rebound effect's values around 65%. Transportation was another sector advising of a possible backfire effect, but not for economic crisis years (Fig. 1d). The explanation provided by Vivanco et al. [21] is useful here because these authors wrote that fuel efficiency improvements in passenger cars have made driving cheaper, resulting in users driving more and purchasing larger cars (direct effect) and/or spending the remaining savings on other products (indirect effect). Economic activity and population effects acted as in the previous sectors, while structure effect inhibited energy consumption for most of the years. The last is due to a decrease in logistic activity because of economic recession. For this sector, our results are not similar to those obtained by Arocena et al. [13] but are close to those obtained by Guerra and Sancho [31] for the Spanish economy as a whole. The heterogeneity of the service sector could explain the unclear pattern of energy efficiency; the effect differed from one economic activity to another while and the effects of population acted as usual (Fig.1e). Here, the structure effect drove energy consumption. Finally, the energy intensity effect rejects the backfire effect for the residential sector (Fig. 1f). Higher levels of household automation or smart use of domestic devices could explain this fact (Cansino et al., 2018). The same may be said for population and economic activity, while the energy resource matrix demonstrated a low use of clean energy in residential sector. Here our results are similar to those shown by Freire-Gonzalez [64] for Catalonia's region (in the northeast of Spain) that was a rebound effect between 35% and 49% and by Freire-Gonzalez [65]finding a rebound effect between 64.60% and 74.71% for the Spanish case. 5.2. Second level decomposition The methodological approach developed offers valuable information from results obtained in a second level decomposition analysis when blocking the activity effect. Results for g i INT in Eqs (10) and (12) inform whether the changes in energy intensity have a direct backfire effect on energy consumption once the indirect backfire effect through economic activity changes is blocked. Table 1 includes major results regarding the direct effect on the energy intensity effect from a sectoral point of view. Table A.4 in annex provide detailed results. Regarding the agriculture sector, once the activity effect on energy consumption is blocked, the energy intensity effect takes on positive values that are higher than the one for most of the years. Once again, the economic crisis period is the only exception. In general terms, the agriculture sector seems to move to decouple economic growth from energy consumption. Table 2 lists the coupling/decoupling status for every of the considered sectors. In terms of the industry sector, the direct effect of energy Table 1 Years appearing direct backfire effect caused by energy efficiency. Years Agriculture Industry Energy Transport Services Residential 2000e2001 þþþ 2001e2002 þ 2002e2003 þþ þ 2003e2004 þþ þþþ 2004e2005 þþ þþ 2005e2006 þþþþþ 2006e2007 þ 2007e2008 þ 2008e2009 þþþ 2009e2010 þþþ 2010e2011 þþ þ 2011e2012 þþþ þ 2012e2013 þþþ þ 2013e2014 2014e2015 þþ Source: Own elaboration. J.M. Cansino et al. / Energy 181 (2019) 760e770764 intensity indicates that there will be a backfire effect. This is an important finding regarding Spain's energy efficiency policy because this direct effect worsens when the indirect effect embodied in energy activity is added. As mentioned above, during the economic crisis period, energy efficiency levels failed to react in an easy way to fixed consumption levels, but in this second decomposition analysis, we focus directly on effects derived from technological change on energy intensity. These effects go against decoupling between energy consumption and economic growth. Contrary to what happened in the industry sector, energy intensity did not support Jevon's paradox in the energy sector. Despite this fact, the sector, as a whole, did not show a clear pattern against and with coupling being the population effect as the main barrier for decoupling until changes in the 2012 demographic trend. Exploring possible direct backfire effects in the transport sector motivate these authors to differentiate between two sub periods. Between 2000 and 2006, the energy intensity effect acted against decoupling, thus causing a direct backfire effect on energy consumption. From the last year of the sub period onwards, mainly positive values higher than one were seen. The energy sector seems to slowly move towards a decoupling status. The results seem to support Jevon's paradox when focusing on the service sector. Because it is a less technological intensive option, technical improvements in energy efficiency have little chance. Improvements in processes rather than in technology could be a right road to move ahead of the backfire effect. The service sector, as a whole, shows a strong coupling status between energy consumption and economic growth. Finally, no support of Jevon's paradox appears when analyzing the results for the residential sector, despite of low values for direct energy efficiency effect and an unclear pattern regarding decoupling for the sector. 6. Conclusions Jevon's paradox or the backfire effect on energy consumption caused by energy efficiency improvements was tested for Spain during the period 2000 to 2015. Here, the backfire effect means that there is a rebound higher than 100%; this implies energy efficiency improvement acting on energy intensity that fails to reduce rather than raise energy consumption. To obtain this, a Logarithmic Mean Divisia Index I (LMDI-I) was used in a novel way to better differentiate between direct and indirect effects in a no-partial equilibrium scenario. A multisectoral approach was conducted to focus on the sectors identified by Spain's plans of action as the most energy intensive ones. As a whole, our results do not support Jevon's paradox for the sectors analyze. However, they do indicate a possible backfire effect in the industry, transport and service sectors. Major findings follow along the same lines when the activity effect is blocked and when its results are not worsened by those speaking of a backfire effect causes by a direct effect. Despite the fact that results for the industry sector could be partially explained by fixed levels of energy consumption, it is not easily adjusted during economic crisis years; authorities in Spain might develop strong efforts on this sector together with the transport and service sectors. In the last case, improvements in energy efficiency would derive from innovation in processes rather than technological advances because of low use of the capital factor. Regarding [15] on energy efficiency, the results do not show that Spain's economy is moving towards a decoupling status between economic growth and energy consumption. Of course, energy efficiency improvements are not the only path to reach it but when the rebound effect is reduced, its inhibitor role on energy consumption increases. Finally, as the aim of this paper was not to estimate the precise rebound effect caused by the energy efficiency improvements but to test whether such improvements caused a backfire effect offsetting the expected decrease in energy consumption, further research is needed to measure the size of the various types of rebound effects. Acknowledgements The first and second authors acknowledge the funding received from the SEJ 132 project of the Andalusian Regional Government, the “C atedra de Economía de la Energía y del Medio Ambiente, Red El ectrica de Espa~ na-Universidad de Sevilla. The first and second authors also acknowledge the funding provided by the Universidad Aut onoma de Chile (Chile). The standard disclaimer applies. Table 2 Results of decoupling index. Second level decomposition. Years Agriculture Industry Energy Transport Services Residential 2000e2001 SDE NDE SDE NDE NDE NDE 2001e2002 SDE WDE SDE WDE NDE NDE 2002e2003 NDE NDE NDE NDE SDE NDE 2003e2004 NDE NDE NDE NDE NDE NDE 2004e2005 SDE NDE NDE NDE NDE NDE 2005e2006 SDE SDE NDE NDE NDE NDE 2006e2007 NDE NDE SDE NDE SDE WDE 2007e2008 SDE SDE NDE SDE NDE NDE 2008e2009 WDE SDE NDE NDE NDE NDE 2009e2010 SDE NDE NDE SDE NDE NDE 2010e2011 NDE WDE SDE SDE NDE SDE 2011e2012 NDE NDE NDE WDE NDE NDE 2012e2013 NDE NDE WDE SDE SDE SDE 2013e2014 SDE SDE NDE WDE SDE SDE 2014e2015 SDE SDE NDE NDE NDE WDE No Decoupling Efforts (NDE), Weak Decoupling Efforts (WDE), Strong Decoupling Efforts (SDE) Source: Own elaboration. J.M. Cansino et al. / Energy 181 (2019) 760e770 765 Annexes Table A.1 Correspondence of productive sectors between the Energy Balance and the GVA Aggregate Sectors Energy Balance Gross Value Added Agriculture Agriculture/Forestry Agriculture, forestry and fishing Fishing Industry Mining and Quarrying Extractive industries Food and Tobacco Food industries, beverage manufacturing and tobacco industry Textile and Leather Textile industry, garment making and the leather and footwear industry Wood and Wood Products Wood and cork industry Manufacture of furniture, other manufacturing Paper, Pulp and Print Paper industry Graphic arts and reproduction of recorded media Chemical and Petrochemical industry Chemical industry Manufacture of pharmaceutical products Non-metallic Minerals (Glass, pottery &building mat. Industry) Manufacture of rubber and plastic products Non-ferrous metal industry Manufacture of other non-metallic mineral products Iron &steel industry Metallurgy, manufacture of iron, steel and ferroalloy products Manufacture of metal products, except machinery and equipment Machinery Manufacture of machinery and equipment n.c.o.p. Repair and installation of machinery and equipment Transport Equipment Manufacture of motor vehicles, trailers and semi-trailers Manufacture of other transport material Construction Construction Non-specified (Industry) Manufacture of computer, electronic and optical products Manufacture of electrical material and material Energy Consumption of the energy branch. Includes Oil refineries (Petroleum Refineries) Supply of electric power, gas, steam and air conditioning Coke and oil refining Transport ( a ) Transport Transport Services Services Water supply; sanitation activities, waste management and decontamination Wholesale and Retail; repair of motor vehicles and motorcycles Post and postal activities Hostelry Information and communications Financial and insurance activities Real estate activities Professional, scientific and technical activities; administrative activities and auxiliary services Public administration and defense; compulsory social security; education; health and social services activities Artistic, recreational and entertainment activities Other services Residential Residential Activities realized in households ( a )Includes: land and pipeline transport, maritime, air; as well as, storage and activities attached to transport. Source: Own elaboration. Table A.2 Type of energy sources used for decomposition at sector level Sectors Oil Gas Renewables Electricity Solid fuels Agriculture x x x x Industry x x x x x Energy x x x x x Transport x x x x Services x x x x Residential x x x x x Source: Own elaboration. J.M. Cansino et al. / Energy 181 (2019) 760e770766 Table A.3 Additive LMDI-I decomposition by sectors and effects (kilo tons of oil equivalent) Sectors Effects 2000e2001 2001e2002 2002e2003 2003e2004 2004e2005 2005e2006 2006e2007 2007e2008 2008e2009 2009e2010 2010e2011 2011e2012 2012e2013 2013e2014 2014e2015 D P 11,953 21,553 49,523 53,497 56,171 48,330 50,188 55,056 31,054 12,246 8863 8247 5372 12,995 3423 D M0,247 0002 11,752 0,150 0,330 0,201 0012 0,093 2828 0,202 2306 0,267 0013 0,034 9414 Agriculture D I136,349 57,110 588,156 477,916 39,492 468,454 67,789 170,183 240,000 167,732 59,550 572,091 216,302 48,002 199,650 D S136,878 45,671 92,935 168,760 380,152 46,666 80,358 114,686 8798 46,907 112,595 185,600 393,364 68,699 140,858 D A 87,140 44,826 23,136 34,947 54,524 74,910 68,310 17,755 114,887 10,546 22,237 82,355 34,271 47,783 82,088 D Total 174,381 36,400 579,631 397,750 230,295 298,748 131,079 247,659 335,459 119,327 161,078 312,649 137,432 81,948 271,257 D P 126,436 247,214 544,865 510,187 536,244 474,720 483,961 537,759 301,852 120,291 87,795 70,878 41,640 98,398 27,091 D M 0,110 0016 0,414 0148 0,136 4582 0,028 0,118 0,980 0116 0,470 0,954 0093 0,202 2331 Industry D I 458,003 273,412 1.512,908 551,005 168,849 4.854,929 1.039,091 1.588,609 1.967,654 2.188,941 847,819 508,417 915,144 1.119,123 2.365,192 D S 258,223 153,613 278,658 596,064 375,413 445,555 842,730 567,458 1.665,154 1.440,108 1.247,552 754,131 488,686 130,451 208,021 D A 921,727 514,154 254,555 333,282 520,523 735,810 658,715 173,419 1.116,736 103,596 220,267 707,820 265,656 361,801 649,739 D Total 1.764,498 334,360 2.034,083 798,557 850,339 4.094,535 1.339,066 1.791,846 4.448,672 765,644 532,674 883,610 119,256 725,471 1.536,854 D P 28,94 53,12 113,88 105,92 112,49 119,43 148,21 162,66 101,46 43,07 30,23 26,58 16,95 40,71 12,29 D M0,01 0,09 0,23 0,00 0,96 66,26 0,06 0,37 0,75 0,02 1,40 5,15 1,66 0,05 9,43 Energy D I316,83 119,76 162,15 64,31 9,36 1.784,81 529,91 317,17 127,34 1.361,48 772,66 1.830,95 1.626,30 4,04 837,64 D S105,47 131,00 249,14 136,98 60,88 169,60 15,80 118,61 63,50 1.312,75 357,93 232,71 1.739,47 63,41 1.268,96 D A 210,98 110,48 53,20 69,19 109,20 185,12 201,72 52,45 375,35 37,09 75,85 265,47 108,14 149,70 294,87 D Total 182,383 87,250 254,299 247,779 274,171 1.986,028 195,830 88,731 210,805 42,777 461,737 1.364,503 239,921 168,410 723,321 D P 163,31 315,64 689,12 644,45 685,43 661,71 727,45 809,71 486,26 200,28 141,88 112,00 62,73 147,52 44,15 D M 0,60 0,08 0,07 0,01 0,07 1,03 0,47 0,50 1,79 0,20 0,32 1,24 3,81 0,01 1,44 Transport D I 663,50 884,85 2.270,96 1.281,14 2.200,01 705,75 31,70 1.096,29 528,64 2.432,94 1.798,77 2.024,17 317,35 1.024,42 162,64 D S614,48 1.363,09 1.473,27 645,71 2.243,92 1.251,44 444,34 1.249,17 777,26 1.686,65 858,07 343,26 1.414,16 833,32 443,25 D A 1.190,55 656,46 321,95 420,99 665,34 1.025,64 990,12 261,12 1.798,97 172,48 355,97 1.118,45 400,18 542,41 1.058,87 D Total 1.403,482 493,933 1.808,828 1.700,869 1.306,941 1.142,687 1.241,999 1.797,363 2.620,402 718,687 1.155,107 2.688,593 1.563,533 203,807 1.622,050 D P 33,09 64,76 136,22 126,80 140,52 141,33 154,80 177,02 116,04 51,21 38,65 32,71 18,98 42,67 12,63 D M 0,06 0,01 1,22 0,43 1,81 1,36 0,02 0,10 0,01 0,05 1,14 0,02 0,07 0,21 5,87 Services D I 91,67 56,25 365,79 307,58 279,86 74,01 582,97 255,23 172,56 270,91 335,25 11,56 404,00 861,55 934,83 D S 2,50 53,38 53,44 86,46 110,86 76,21 109,42 102,55 249,62 113,75 127,79 148,09 113,24 21,97 60,07 D A 241,23 134,69 63,64 82,83 136,40 219,06 210,69 57,09 429,30 44,10 96,97 326,66 121,07 156,88 302,86 D Total 368,539 196,594 113,715 604,113 669,437 511,966 108,078 477,811 108,938 391,827 405,871 157,447 430,878 769,514 1.170,870 D P 59,24 115,70 256,66 243,59 259,94 250,80 272,21 304,20 195,04 87,56 62,99 50,33 29,35 68,45 19,93 D M 0,32 0,05 0,21 0,05 0,07 0,36 0,00 0,01 0,19 0,23 1,13 0,01 0,08 0,00 0,03 Residential D I5,42 210,29 161,80 26,02 415,01 295,26 501,91 270,25 328,57 505,66 731,05 1.369,44 144,77 161,28 28,02 D S 134,36 186,26 406,61 344,41 363,66 101,42 95,17 64,72 626,76 478,26 465,40 1.019,25 571,12 194,69 263,01 D A 431,85 240,64 119,91 159,12 252,32 388,74 370,51 98,10 721,58 75,41 158,05 502,59 187,26 251,69 477,98 D Total 620,354 332,354 945,185 773,192 460,973 446,069 45,643 128,881 428,609 996,298 1.292,634 102,083 643,045 172,733 167,049 J.M. Cansino et al. / Energy 181 (2019) 760e770 767 Table A.4 Decomposition effects when blocking activity effect Sectors Effects 2000e2001 2001e2002 2002e2003 2003e2004 2004e2005 2005e2006 2006e2007 2007e2008 2008e2009 2009e2010 2010e2011 2011e2012 2012e2013 2013e2014 2014e2015 Agriculture D A 87,140 44,826 23,136 34,947 54,524 74,910 68,310 17,755 114,887 10,546 22,237 82,355 34,271 47,783 82,088 g POB 0,137 0,481 2140 1531 1030 0,645 0,735 3101 0,270 1161 0,399 0,100 0157 0,272 0042 g MIX 0,003 0000 0,508 0,004 0006 0,003 0000 0,005 0025 0,019 0,104 0,003 0000 0,001 0115 g INT 1565 1274 25,421 13,675 0,724 6254 0,992 9585 2089 15,905 2678 6947 6311 1005 2432 g EST 1571 1019 4017 4829 6972 0,623 1176 6459 0,077 4448 5063 2254 11,478 1438 1716 g TOTAL 3001 1812 24,053 10,381 5224 4988 0,919 11,949 0,920 9315 9244 5796 6010 2715 4304 Meaning SDE SDE NDE NDE SDE SDE NDE SDE WDE SDE NDE NDE NDE SDE SDE Industry D A 921,727 514,154 254,555 333,282 520,523 735,810 658,715 173,419 1.116,736 103,596 220,267 707,820 265,656 361,801 649,739 g POB 0,137 0,481 2140 1531 1030 0,645 0,735 3101 0,270 1161 0,399 0,100 0157 0,272 0042 g MIX 0,000 0000 0,002 0000 0,000 0006 0,000 0001 0,001 0,001 0002 0,001 0000 0,001 0004 g INT 0,497 0532 5943 1653 0,324 6598 1577 9161 1762 21,130 3849 0,718 3445 3093 3640 g EST 0,280 0299 1095 1788 0,721 0606 1279 3272 1491 13,901 5664 1065 1840 0,361 0,320 g TOTAL 0,914 0350 6991 1396 0,634 6565 1033 8332 1984 9391 0,418 0,752 2449 3005 3365 Meaning NDE WDE NDE NDE NDE SDE NDE SDE SDE NDE WDE NDE NDE SDE SDE Energy D A 210,98 110,48 53,20 69,19 109,20 185,12 201,72 52,45 375,35 37,09 75,85 265,47 108,14 149,70 294,87 g POB 0,14 0,48 2,14 1,53 1,03 0,65 0,73 3,10 0,27 1,16 0,40 0,10 0,16 0,27 0,04 g MIX 0,00 0,00 0,00 0,00 0,01 0,36 0,00 0,01 0,00 0,00 0,02 0,02 0,02 0,00 0,03 g INT 1,50 1,08 3,05 0,93 0,09 9,64 2,63 6,05 0,34 36,71 10,19 6,90 15,04 0,03 2,84 g EST 0,50 1,19 4,68 1,98 0,56 0,92 0,08 2,26 0,17 35,39 4,72 0,88 16,09 0,42 4,30 g TOTAL 1,86 1,79 3,78 2,58 1,51 9,73 1,97 0,31 1,44 0,85 4,09 7,14 0,22 0,12 1,45 Meaning SDE SDE NDE NDE NDE NDE SDE NDE NDE NDE SDE NDE WDE NDE NDE Transport D A 1.190,55 656,46 321,95 420,99 665,34 1.025,64 990,12 261,12 1.798,97 172,48 355,97 1.118,45 400,18 542,41 1.058,87 g POB 0,14 0,48 2,14 1,53 1,03 0,65 0,73 3,10 0,27 1,16 0,40 0,10 0,16 0,27 0,04 g MIX 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,01 0,00 0,00 g INT 0,56 1,35 7,05 3,04 3,31 0,69 0,03 4,20 0,29 14,11 5,05 1,81 0,79 1,89 0,15 g EST 0,52 2,08 4,58 1,53 3,37 1,22 0,45 4,78 0,43 9,78 2,41 0,31 3,53 1,54 0,42 g TOTAL 0,18 0,25 4,62 3,04 0,96 0,11 0,25 4,88 0,54 2,17 1,24 0,40 1,91 0,62 0,53 Meaning NDE WDE NDE NDE NDE NDE NDE SDE NDE SDE SDE WDE SDE WDE NDE Services D A 241,23 134,69 63,64 82,83 136,40 219,06 210,69 57,09 429,30 44,10 96,97 326,66 121,07 156,88 302,86 g POB 0,14 0,48 2,14 1,53 1,03 0,65 0,73 3,10 0,27 1,16 0,40 0,10 0,16 0,27 0,04 g MIX 0,00 0,00 0,02 0,01 0,01 0,01 0,00 0,00 0,00 0,00 0,01 0,00 0,00 0,00 0,02 g INT 0,38 0,42 5,75 3,71 2,05 0,34 2,77 4,47 0,40 6,14 3,46 0,04 3,34 5,49 3,09 g EST 0,01 0,40 0,84 1,04 0,81 0,35 0,52 1,80 0,58 2,58 1,32 0,45 0,94 0,14 0,20 g TOTAL 0,53 0,46 2,79 6,29 3,91 1,34 1,51 10,37 2,25 10,88 6,19 1,52 1,56 5,91 2,87 Meaning NDE NDE SDE NDE NDE NDE SDE NDE NDE NDE NDE NDE SDE SDE NDE Residential D A 431,85 240,64 119,91 159,12 252,32 388,74 370,51 98,10 721,58 75,41 158,05 502,59 187,26 251,69 477,98 g POB 0,14 0,48 2,14 1,53 1,03 0,65 0,73 3,10 0,27 1,16 0,40 0,10 0,16 0,27 0,04 g MIX 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,01 0,00 0,00 0,00 0,00 g INT 0,01 0,87 1,35 0,16 1,64 0,76 1,35 2,75 0,46 6,71 4,63 2,72 0,77 0,64 0,06 g EST 0,31 0,77 3,39 2,16 1,44 0,26 0,26 0,66 0,87 6,34 2,94 2,03 3,05 0,77 0,55 g TOTAL 0,44 0,38 6,88 3,86 0,83 0,15 0,88 0,69 2,59 15,21 6,18 1,80 1,43 1,69 0,65 Meaning NDE NDE NDE NDE NDE NDE WDE NDE NDE NDE SDE NDE SDE SDE WDE J.M. Cansino et al. / Energy 181 (2019) 760e770768