How can Chile move away from a high carbon economy?

How can Chile move away from a high carbon economy? José M. Cansino a,b, ⁎, Antonio Sánchez-Braza a , María L. Rodríguez-Arévalo c a Universidad de Sevilla, Departamento de Análisis Económico y Economía Política, Avda. Ramón y Cajal, 1, 41018 Sevilla, Spain b Universidad Autónoma de Chile, Chile c Universidad de Zaragoza, Departmento de Informática e Ingeniería de Sistemas, c/ María de Luna, 1, 50018 Zaragoza, Spain abstractarticle info Article history: Received 6 July 2016 Received in revised form 29 November 2017 Accepted 1 December 2017 Available online 5 December 2017 JEL classification: Q42 Q48 Q53 Q54 Q55 Q57 This paper quantitatively evaluates the performance of Chile's CO 2 emissions between 1991 and 2013 using a ‘complete decomposition’technique to examine emissions and their components. A decomposition analysis based on log-mean divisia index method (LMDI I) was conducted. Six decomposition factors were considered: Carbon Intensity effect (CI), RES penetration effect (RES), Energy Intensity effect (EI), Economy Structure effect (ES), Income effect (Yp) and Population effect (P). To know how these factors could influence each other in the future, the Innovative Accounting Approach (IAA) was used, including forecast error variance decomposition and Impulse Response Functions (IRFs). These two methodologies allow us to identify the drivers of CO 2 emission changes in the past (1991–2013), test policy measures and learn how these drivers could influence each other in the future, to evaluate whether the current measures meet the Paris commitments. The LMDI analysis results show that the Energy Intensity Factor is the main compensating factor of Chile's CO 2 emissions and the only effect with a clear trend to aid the economy's decoupling. IAA and IFRs results react similarly and confirm that carbon intensity reacts to shocks more significantly in the short term. The reaction to RES has the same and opposite behavior to shocks in ES and Yp, to disappear in the long term. © 2017 Elsevier B.V. All rights reserved. Keywords: CO 2 emissions Chile LMDI Innovative Accounting Approach 1. Introduction Energy related CO 2 emissions in Chile increased 178% for the 1990– 2013 period, which is well above the world average (56%) for OECD- American countries (104%) and all of OECD economies (9.4%) (IEA, 2017). These emissions derived only from fuel combustion; therefore, they are associated to anthropogenic actions. Unless Chile's energyrelated CO 2 emissions shares only 0.2% of total world, its percentage increased in 78.5% for the aforementioned period. Within the global context fighting against climate change, Chile decided to become a non-Annex-I Country after once the Kyoto Protocol was approved. Chile is highly vulnerable to climate change, as it meets seven of the nine characteristics listed by the United Nations Framework Convention on Climate Change (UNFCCC), as established at the Conference of the Parties: The country has low lying coastal areas, arid and semiarid areas, forests, territories that are susceptible to natural disasters, others that are prone to drought and desertification, urban areas with atmospheric pollution and mountain ecosystems (Ministry of Environment, 2011). It has been calculated that in Chile, environmental, social and economic losses derived from climate change could amount to 1.1% of the country's GDP in 2100 (CEPAL, 2012). Growing concerns about climate change in Chile explained the formalization of the pledge made in Copenhagen by registering it in Appendix II of the Copenhagen Accord (UNFCCC, 2009). Within the framework of COP21 in Paris, Chile decided to present its Intended Nationally Determined Contribution (INDC) defining its commitment in the battle against climate change, in terms of emission intensity (tons of CO 2 equivalent per unit of Gross Domestic Product –GDP–in millions of CLP$ 2011). Chile's commitment is based on sectorial analyses and mitigation scenarios developed with the Mitigation Actions Plans and Scenarios (MAPS) Chile project (Phase 2) (MAPs Chile, 2014), the National Greenhouse Gas Inventory results (1990–2010) and additional information provided by the Ministries of Environment, Energy, Agriculture and Finance, as well as observations received during the Public Consultation of the Intended National Contribution. The main goal of this paper is to evaluate whether or not the measures in force to meet the Paris commitments are adequate. To do that we consider results from the two methodologies mentioned below. These results allow us to identify past (1991–2013) drivers of changes in CO 2 emissions and test policy measures in that period. Together with this and regarding the main objective of the paper, the results also allow us to know how the identified drivers could influence each other in the future and if the Energy Economics 69 (2018) 350–366 ⁎Corresponding author. E-mail addresses: [email protected] (J.M. Cansino), [email protected] (A. Sánchez-Braza), [email protected] (M.L. Rodríguez-Arévalo). https://doi.org/10.1016/j.eneco.2017.12.001 0140-9883/© 2017 Elsevier B.V. All rights reserved. Contents lists available at ScienceDirect Energy Economics journal homepage: www.elsevier.com/locate/eneeco measures in force are well oriented towards these drivers. The analysis conductedfollowsfourstepsasisdetailedbelow. Firstly, the 1991–2013 decomposition of changes in CO 2 emissions is carried using a ‘complete decomposition’technique to examine emissions and their components. Six decomposition factors or effects have been considered: Carbon Intensity effect (CI), Renewable Energy Sources penetration effect (RES), Energy Intensity effect (EI), Economic Structure effect (ES), Income effect (Yp) and Population effect (P). This technique consists of using one type of index decomposition analysis (IDA) known as the log-mean divisia index method –LMDI I–(Ang, 2005). LMDI I-based studies are useful to understand the evolution of energy-related CO 2 emissions, and to identify the driving forces that have impacted these changes. Secondly, this paper offers the decomposition of changes in CO 2 emissions between 1991 and 2013 at the sectoral level; it traces changes in emissions derived from each consumption category per sector considered (energy, transport, industry, use of solvents and other products –USOP–, agriculture, and residuals and waste). Step three researches the long-run and short-run relationships among CO 2 emissions and their determinants and how the drivers identified could influence each other in the future; this paper implemented the Innovative Accounting Approach (IAA) taking Moutinho et al. (2016) and Robaina-Alves and Moutinho (2013) as the start point. It includes the forecast error variance decomposition and Impulse Response Functions (IRFs) applied to the factors where energy-related CO 2 emissions were decomposed. Despite both tools (LMDI and IAA) being well-known, the novelty of the analysis derives from i) using the second to know how the drivers identified could influence each other in the future, ii) adding a sectoral analysis (not only to industrial sectors) and iii) evaluating the measures in force by taking into account the results of causality obtained in this relationship and information regarding how long it takes to once again attain equilibrium after the long-run relationship has been shocked. Consequently, this paper allows one to look at both current and future measures and not just the past as is customary for IDA analysis. This paper contributes to the growing body of knowledge about the transition to low carbon economies. Additionally, we do not focus on CO 2 emission intensity as the key variable due to international agreements based on carbon emission intensity commitments, which have been severely criticized. Some experts point out that CO 2 emission intensity targets are used to hide the fact that a reduction in the carbon intensity could be achieved without eliminating a continuous increase of absolute carbon dioxide emissions (Stern and Jotzo, 2010; Cansino et al., 2015). Contributions to the literature derive from applying IAA to all economic sectors considered and not only for industry. To the best of our knowledge, no previous papers have been presented that focus on Chile's economy. By using these two methodologies, we are able to examine emissions and their components as well as analyze how these have evolved in the past. It also shows how their components are related to each other and how they can influence each other in the future. Finally, in light of the results, the fourth and final steps allow policy measures (past and current) to be examined, as well as the impacts of relevant facts about CO 2 emissions in the energy sector. To analyze past relevant policy measures, the entire period was divided into three sub-periods based not only on climate change actions (Sanhueza and Ladrón de Guevara, 2014), but also regarding events in the energy sector and the energy policy during the period considered. Related fossil fuels combustion energy represents 74.6% of total CO 2 -eq emissions up to 2010 (Ministry of Environment, 2016). The sub-periods considered are 1991–1996 (nearest to the signing of the Framework Convention at the UNFCCC summit in Rio until the beginning of massive natural gas imports from Argentina), 1997–2006 (from the relevant use of natural gas in the generation of electricity to the entry into force of the Kyoto Protocol), 2007– 2013 (when Argentina decided not to export natural gas and the end of the period with available data). Conducted research will put us in a position to evaluate if measures in force to meet the Paris commitments are adequate. The measures assessed are, for the most part, included on the long-term road map for the energy policy named Energia2050 (Ministry of Energy, 2016) and the National Climate Change Plan of Action 2017–2022 (PANCC 2017–2022 from the Spanish acronym for Plan de Acción Nacional de Cambio Climático) (Ministry of Environment, 2016). These results are interesting not only for researchers but also for policy-makers. In fact, this paper speaks directly to the authorities of Chile and the policy agenda regarding several issues, including energy security. The paper has been structured as follows: After the introduction, Section 2 provides details of methodology used. Dataset are exposed in Section 3. The results are presented in Section 4. In the light of our results, current political measures Paris Agreement oriented are examined in Section 5.Section 6 provides the main conclusions. 2. Methodology 2.1. LMDI analysis There are two main decomposition techniques to identify the drivers of changes in CO 2 emissions: Structural Decomposition Analysis (SDA) (Achão and Schaeffer, 2009; Zhang et al., 2011; Ang and Su, 2016; Cansino et al., 2016) and Index Decomposition Analysis (IDA) (Hoekstra and van den Bergh, 2003; Hatzigeorgiou et al., 2008; Ma and Stern, 2008; Andreoni and Galmarini, 2016). The latest comparisons between IDA and SDA have been shown in Su and Ang (2012).Comparatively, IDA has certain advantages over SDA. IDA enables decompositions for any aggregate (value, ratio or elasticity). Also, IDA requires less data than decomposition methods based on input-output analysis and it is useful when decomposing CO 2 emission changes between its various components (Cansino et al., 2015). This paper follows the Ang (2004) criteria that assessed the various decomposition methods. The latter concluded that LMDI I is a more recommendable method. However, it shows some limits with the proportionality test (Ang, 2004). LMDI I allows perfect decomposition (that is, without residuals) and provides a simple and direct association between the additive and the multiplicative decomposition form Ang and Liu (2007a). The IPAT equation is the starting point for the LMDI I conducted. The Impact = Population × Affluence × Technology (IPAT) equation is used to assess the contribution of CO 2 emission drivers. Specifically, the IPAT model (Commoner et al., 1971; York et al., 2002; Brizga et al., 2013)and the ‘Kaya Identity’(Kaya, 1990; Yamaji et al., 1991) are extended using IDA (Leontief and Ford, 1972; Rose and Casler, 1996) to assess the key drivers behind Chile's CO 2 emissions. Two recent papers focused on Chilean economy are due to Mundaca (2013) and Duran et al. (2015). The first conducted an interesting research using IPAT equation. The second carried out a decomposition of the energy consumption by Chilean industry but not of the CO 2 emissions as we do. Six-factor have been proposed to identify, quantify and explain the main determinant of the variation for total energy related CO 2 eq emissions in Chile between 1991 and 2013. Applying the decomposition proposed to six productive sectors, the total CO 2 emissions may be decomposed as follows: CO2¼X 6 i¼1 CIi∙RESi∙EIi∙ESi∙Yp ∙P¼X 6 i¼1 CO2i FFi ∙FFi Ei ∙Ei Yi ∙Yi Y∙Y P∙Pð1Þ CO 2i represents the energy related CO 2 emissions of sector i;FF i denotes the share of fossil fuels of sector i;E i denotes the energy consumption of sector i;Y i represents the output of sector i;Ydenotes 351J.M. Cansino et al. / Energy Economics 69 (2018) 350–366 the total output for the entire economy the same as in CO 2 ,andP represents the population. Changes in CO 2 emissions may be assessed by implementing additive or multiplicative decomposition. In this paper, an additive LMDI I analysis is carried out. The overall ratio of change in CO 2 emissions during the period 0 and t is decomposed as follows: ΔCO2¼CO2t−CO20¼ΔCI þΔRES þΔEI þΔES þΔYpþΔPð2Þ ΔCO 2 represents changes in aggregate CO 2 emissions in the economy from one period to another, being the right-hand side variables the representatives of the various contributing determinants as previously defined but now being referred to as changes. By considering the additive decomposition identity, the LMDI formula is calculated as Table 1 shows for each effect: The term w i (t) is the estimated weight for the additive LMDI I method and is defined as Ang (2005): witðÞ¼ CO2i;t−CO2i;0 lnCO2i;t−lnCO2i;0 ð9Þ Eq. (3) captures the Carbon Intensity effect (CI). Variable ΔCI shows the changes in CO 2 emissions from fossil fuels consumption in sector i (=CO 2i /FF i ), between the periods tand 0, respectively. Available statistical information does not offer fossil fuels consumption broken down by type of fuels so FF i is total fossil fuels by sector without distinction by different fuels. The CI factor captures the quality of fossil fuels and may also be used to evaluate the substitution between fossil fuels. We assume that the better a fossil fuel is, the less CO 2 it emits. Eq. (4) shows the RES penetration effect (RES). Variable ΔRES shows the share of fossil fuel consumption with respect to the total primary energy required in sector i(=FF i /E i ), between periods tand 0, respectively. The RES factor captures the Renewable Energy Source share in total primary energy matrix in each sector. Aspecific comment regarding RES needs to be made to better understand their link with CO 2 emission data. By carrying out a decomposition analysis, we could research the role of RES in Chile's energy matrix. However, one problem must be solved, which is linked to the fact that RES technologies are free or almost free of CO 2 emissions and we observe this as a crucial variable. To bridge this lack of information, we observed the evolution of the ratio between total fossil fuel consumption for total primary energy consumption. O'Mahony (2013) explained that theoretically the carbon emissions from renewables are zero as fossil fuels are not combusted during the generation of electricity (when used in energy supply) or in the delivering of energy services (when used in final consumption). As E i denotes the final energy consumption of sector imeasured in Tera calories it needs to be considered that not all of the primary energy sources use to generate final energy are fossil fuels. Rest of the total final energy consumption is generated by RES or nuclear plants. Being that nuclear energy in Chile is not part of the energy matrix, the role played by RES on CO 2 emissions trend can be derived from RES effect (IEA, 2017). This is why if the ratio FF i /E i is considered in the decomposition analysis to capture the penetration of RES into the Chilean energy matrix. A decline in values for the ratio of total fossil use on total energy use might show a higher share of RES in Chile's energy matrix (mainly hydropower). In other words, higher (lesser) values would mean lesser (higher) levels of RES penetration in the Chilean energy matrix. Eq. (5) presents the Energy Intensity effect (EI). Variable ΔEI shows the total primary energy required in comparison to the output in sector i (=E i /Y i ), between periods tand 0, respectively. EI factor is often used as a measure or aggregate proxy of the energy efficiency or technology level of a country's economy. It may be seen as a signal indicating the efficiency of the energy system and investments for energy saving (Goldemberg and Johansson, 2004; Voigt et al., 2014). Eq. (6) is the Economic mix or Economic Structure effect (ES). Variable ΔES shows the sectoral structure of Chile's economy between period tand 0, respectively. The ES factor indicates the relative weight of each sectoral output within the total output of the overall economy; it incorporates the relative impact of structural changes on Chile's economy in terms of CO 2 emissions for a given year included into the analysis. Eq. (7) is the Income effect (Yp). Variable ΔYp is the output per capita between period tand 0, respectively. Yp factor captures the income effect on CO 2 emission changes from energy consumption. Eq. (8) shows the Population effect (P). Variable ΔPshows the total population between period tand 0, respectively. The P factor allows the effects of population growth as a determinant for energy demand to be analyzed. To accommodate cases of zero value, Ang and Choi (1997),Ang and Liu (2007a),andAng et al. (1998) analyzed and proposed that the best way to handle this situation is by substituting zeros for a δvalue between 10 −10 and 10 −20 . This is known as the small value (SV) strategy. Ang and Liu (2007b) also showed that the SV strategy is robust when an appropriate value is used, and that it would provide satisfactory results even for highly extreme cases. 2.2. The Innovative Accounting Approach (IAA) Following Moutinho et al. (2016), different approaches have been used previously to account for decomposition effects. They highlight that Zhang and Cheng (2009) used in the VAR Granger Causality and Generalized Impulse Response to examine the causality among urban population, economic growth, energy consumption and CO 2 emissions. In a similar way, Lee and Chien (2010) applied the IAA to examine the relationship between energy consumption, capital stock and real income in G-7 countries. A more extended review of the literature is offered by Robaina-Alves and Moutinho (2013). The generalized forecast variance decompositions measure the contribution of each type of shock to the forecast error variance over different time horizons and estimates the simultaneous shocks stemming from other variables. This technique can also provide a rough analysis of how long it takes for a variable to revert back to equilibrium after the long run relationship has been shocked. IRFs show the dynamic responses of time series to a one-period standard deviation shock, and indicate the direction of the response to each of the shocks (Maghyereh, 2004; Park and Ratti, 2008; Ozkan and Ozkan, 2012; Shahbaz et al., 2013). Thus, a random shock in one standard deviation establishes a chain reaction over time in all VAR variables. IFR's calculate these chain reactions (Moutinho et al., 2016). Both computations are useful in assessing how shocks to economic variables reverberate through a system.This technique can also provide a roughanalysis of how long it takes for a variable to revert back to equilibrium after the long run relationship has been shocked. IRFs show the dynamic responses of time series to a one-period standard deviation shock, and indicate the direction of the response to each of the shocks (Maghyereh, 2004; Park and Ratti, 2008, Ozkan and Ozkan, 2012; Table 1 Formulae for each effect in the additive decomposition. ΔCI ¼X 6 i¼1 witðÞ∙ln CIi;t CIi;0  (3) ΔRES ¼X 6 i¼1 witðÞ∙ln RESi;t RESi;0  (4) ΔEI ¼X 6 i¼1 wit ðÞ ∙ln EIi;t EIi;0  (5) ΔES ¼X 6 i¼1 wit ðÞ ∙ln ESi;t ESi;0  (6) ΔYp ¼X 6 i¼1 witðÞ∙ln Ypt Yp0  (7) ΔP¼X 6 i¼1 witðÞ∙ln Pt P0  (8) 352 J.M. Cansino et al. / Energy Economics 69 (2018) 350–366 Shahbaz et al., 2013). Thus, a random shock in one standard deviation establishes a chain reaction over time in all VAR variables. IFR's calculate these chain reactions (Moutinho et al., 2016). We have implemented the IAA to know the long-run and short-run relationships between CO 2 emissions drivers and identify how they could influence each other in the future. IAA includes forecast error variance decomposition and Impulse Response Functions (IRFs). The IAA approach analyzes the causal dynamic relationship between various time series. One of the advantages of this method over other approaches used in the literature is that it allows us to incorporate an analysis for a forwards period, i.e. perform predictions of the behavioral relationships between time series from the selected sample period. In this paper, the generalized forecast variance decomposition approach estimates simultaneous shocks effects, which is determined from the estimation of the VAR model to test the strength of casual relationships between the determining factors of total CO 2 emissions, CI effect, RES effect, EI effect, ES effect, Yp effect and P effect. For instance, if the CI effect explains more of the forecast error variance of the RES effect, then we deduce that there is a unidirectional causality from the CI to the RES effect. The bidirectional causality exists if shocks in the RES effect also affect the CI effect in a significant way. If shocks happen in both series they do not have any impact on the changes in the CI and in the RES effects, then there is no causality between the variables. On the other hand, the relation among these six factors is analyzed with the IRFs. We determined how each decomposition factor of CO 2 emissions responds due to its shock and to shocks in other factors. Results from IFRs offer a rough analysis of how long it takes for the variable to go back to equilibrium after the long-run relationship has been shocked. 3. Database The emission data for CO 2 -eq stem from the official emission inventories that the government of Chile has sent to the UNFCCC (2016). The most recent year for which information is available is 2013. This data has been supplied by the Ministry of the Environment for this research. Energy consumption data –either for fossil fuels as for energy consumptionhas been taken from the energy balances published by the Ministry of Energy (CNE, 2015). All energy consumption data are measured in Teracalories. Energy balances available at Energia2050 were also considered (Ministry of Energy, 2016). Unfortunately, there are not time series of Gross Value Added available so Gross Domestic Product (GDP) data was used. GDP data come from National Accounting drafted by the Central Bank. All data used corresponds to real GDP data at constant prices for 2008 (BCC, 2016a).These GDP series, in real terms, have been built usingthe annual GDP deflator per activity class and the exchange rates for deflator values as of the linked series included in the databases within the aforementioned Central Bank National Accounting (BCC, 2016b). Total Chilean economy was grouped into the following six sectors: Energy, Transport, Industry, use of solvents and other products (USOP), Agriculture and Waste. Because of its relevance it might be noted that mining sector is included into the industrial sector. The criteria for grouping productive activities into these six sectors were twofold. First, to match official emission inventories information, energy balances and GDP data. Secondly, being near to sectors included in Chilean's INDC submitted to Paris in 2015. Finally, population data has been taken from the Central Bank of Chile (BCC, 2016c). Main figures of sectoral GDP are offered in Annex 1. Regarding Chile's commitments derived from Paris Agreement, when later changed to Use of the Land and Forestry sector (LULUCF) is excluded, the official document by the Chilean government defined two carbon intensity measures in its INDC, depending on whether it received international financing or not. The exact commitments were: a) By 2030, Chile is committed to reducing its CO 2 emissions per GDP unit by 30% below its 2007 levels, considering a future economic growth which allows the implementation of adequate measures to reach this commitment. This commitment considers an economic growth rate similar to the country's in the last decade, except for the most critical years of the international financial crisis (2008– 2009). The commitment affects the following sector included in the National Greenhouse Gas Inventory (1990–2010): Energy, which includes the generation and distribution of electricity, transportation, industry, mining, housing, among other fossil fuel consuming sectors, industrial processes, use of solvents and other products (USOP), agriculture, including the livestock sector and waste. b) Additionally, and subject to a grant from international cooperation, the country is committed to reducing its CO 2 emission per GDP unit by 2030 until it reaches a 35% to 45% reduction with respect to 2007 levels, considering, in turn, a future economic growth which allows the implementation of adequate measures to achieve such a commitment. The official document specifies that for the purposes of this commitment, an international monetary grant shall be deemed any grants which allow the implementation of actions having direct effects on greenhouse gas emissions within adequate time frames. Table 2 provides a quantified sample of Chile's commitments. 4. Results 4.1. Decomposition analysis The LMDI analysis results allow Chile's policy measures in force during the period under consideration to be examined. There are strong energy policy decisions which directly impact on some decomposition factors as CI. These are the cases of massive imports of natural gas from Argentina, but also decisions oriented to replacing such an energy source by coal when imports are interrupted. In an indirect way, these types of strong energy policy decisions also affect other factors, as is the case of population. When electricity is generated by thermal plants powered by coal, an increase of demand because of a higher population explains its role as a driver of CO 2 emissions. Another strong energy policy measure is the progressive elimination of leaded gasoline, which directly influences CI but in an indirect way on income and population factors when people go to gas stations and this fuel option is not available. Environmental stress (CO 2 -eq emissions) through income and population factors decreases when the quality of fuels increases. Related with this, a relevant question about energy and environmental policies is if they can act on income and population factors. It can be thought that those factors are beyond the control of such policies depending on the business cycle, income distribution patterns or demographic issues. However, there are many relevant measures acting on those two factors. For example, when, after 2007, it became mandatory for refrigerators and light bulbs to indicate their level of energy efficiency, consumers accessed a clear information that could change residential consumption levels, also modifying the EI trend. This type of ‘ecolabelling’policy is also relevant when it is regarded from the income factor perspective, as the better the level of Table 2 Chile's Intended Nationally Determined Contribution sent to COP21. Source: Government of Chile (2015). Carbon intensity per GDP (CLP$ 2011) Base Year: 2007 1.02 tCO 2 -eq/million CLP$ Target Year: 2030 (conditioned to economic growth) 0.71 tCO 2 -eq/million CLP$ Target Year: 2030 (conditioned to an international grant and economic growth) 0.56–0.66 tCO 2 -eq/million CLP$ 353J.M. Cansino et al. / Energy Economics 69 (2018) 350–366 energy efficiency, the higher the price of the device. A final example of energy policy measures altering the population factor is mentioned when referring to the Netbilling Act (Law no. 20.571), allowing selfconsumption of electricity from a small RES system installed at home and selling back part of this electricity to the grid. Focus is placed on those measures that could impact on decomposition factor values. A richer discussion could be held if the whole period were broken down into sub-period separated by relevant events based on what happened in the energy sector and the energy policy during the period of time considered and regarding the main climate change actions (Sanhueza and Ladrón de Guevara, 2014). The sub-periods considered were 1991–1996, 1997–2006 and 2007–2013. Fig. 1 and Table 3 show the main results of the LMDI analysis for the 1991–2013 period; a complete database has been drafted for all of the factors considered. Positive values imply that the factor or effect acts as a driver of CO 2 emissions (measured in gigatons –GT–in Fig. 1). When its sign is negative, this reveals that it works as a compensating factor. The specific analysis for each sector is carried out below. Results are presented by sub-periods. 4.2. From 1991 to 1996 This period is nearest to the signing the Framework Convention at the UNFCCC summit in Rio and runs up to the beginning of massive imports of natural gas from Argentina for power. On average, Tapio's index for this sub-period is equal to 0.58 (Tapio, 2005). In keeping with Vehmas et al. (2007) and Wang et al. (2016), this implies a weak decoupling between economic growth and CO 2 emissions. For this sub-period, the only factors that work as compensators for the increase in CO 2 emissions are CI and ES, but without surpassing absolute values for the rest of the factors that act as emission drivers. However, with the results obtained, we must be cautious because LMDI does not always capture the changes in energy consumption attributed to the energy consumption structure between sectors. We assume that these remain unchanged (Han et al., 2012). The carbon intensity factor –CI–was the most important inhibitory factor for the increase of CO 2 emissions. This showed a clear pattern for the sub-period evaluated; it seems to show that the energy matrix in Chile moved towards de-carbonization. When performing a sector analysis of the CI effect (Table 3a), it shows that for the energy sector, this factor had a negative sign, thus working against CO 2 emissions. The energy, USOP, Agriculture and waste sectors all acted as compensating sectors for CO 2 emissions. Transport and Industry sectors drove CO 2 emissions. The role of Transport sector could be partly explained because authorities created the Oil and Fuel Price Stabilisation Funds in 1991 (Law 19.030) and 2005 (Law 20.493) to avoid a negative impact of high crude oil prices (Mundaca, 2013). The industrial process sector registered a behavior that was similar to that of the transportation sector, but lesser in absolute terms. The explanation behind the positive sign of RES (Table 3b) is found in the behavior of the Energy sector. Between 1991 and 1996, the RES factor in this sector showed a quantitatively positive sign that was larger than the other sectors. For the same time span, the Agriculture and Industry sectors also worked to favor the increase in CO 2 emissions. For this sub-period, the EI factor failed to show a compensatingeffect on CO 2 emissions (Table 3c). Sanhueza and Ladrón de Guevara (2014) pointed out that in 1992, Chile received funding from the Global Environmental Facilities (GEF) for a project to deal with promoting the installation of services supporting efficient energy use, by using Energy Service Companies, through pilot projects in the country's copper mining sector. The ‘pilot’nature of this project might explain that despite the fact that it was correctly oriented in sectoral terms, it was not enough to have a strong impact on Chile's energy efficiency. Together with this pilot project and from the perspective of the domestic governance in the fight against climate change, it must be pointed out that a National Advisory Committee on Global Change was set up in 1996. Nevertheless, it only began to operate on a regular basis at the beginning of 1998. The ES factor acted as a compensating factor of CO 2 emissions. From 1991 to 1996, the weight of the Agriculture sector on Chile's GDP at constant prices reduced by 14.5%. This is coherent with its role as a compensating sector, shown on Table 3d. Other emitter sectors also reduced their weight in total GDP, such as mining, non-metallic and Pulp sectors (−8.51%), all of them included in the industry sector of Table 3d, and the Chemical sector, included in USOP sector. Detailed analysis is based on National Accounts provided by BCC (2016b). As was expected, the results for Income effect show a clear coupling between economic growth and emissions (Table 3 e and f). The population increase also works as a driver for CO 2 emissions. This result is in line with, for example, Hatzigeorgiou et al. (2010) and Cansino et al. (2015). From the sectoral perspective and similar to what happens with the Yp factor, the sectors where greater absolute values take on this factor are energy, transportation and agriculture. 4.3. From 1996 to 2006 The sub-period begins with an important change in Chile's primary energy and in the electricity generation matrix. This is a consequence Fig. 1. Main figures for factors considered in LMDI analysis 1991–2013. Source: Own elaboration. 354 J.M. Cansino et al. / Energy Economics 69 (2018) 350–366 of massive natural gas imports from Argentina. Another relevant fact of this sub-period was the rationing of the electrical supply due to extensive droughts in 1998 and 1999. Together with this, as mentioned, the National Advisory Committee on Global Change began to operate at the beginning of 1998. The period finalizes with a sudden interruption in the supply of natural gas. In addition to the CI factor, RES and EI factors act as emissions' compensators. The Tapio's index for this subperiod is equal to 0.33; this translates into a weak decoupling between economic growth and CO 2 emissions. For this second sub-period, CI continued to act as a compensating factor of CO 2 emissions, which increased the greater values in absolute terms. From Table 3a, it must be pointed out that the CI factor shows that the energy sector continued working as a compensating factor for CO 2 emissions together with Transport and USOP sector overweighting the role as a driver sector played by Agriculture, Industry and Waste. To better explain the role of the energy sector for this sub-period, it must be highlighted that the CI factor captures the substitution of fuels for cleaner types. Upon contemplating this aspect, one must consider two key factors that may be observed in Annex 2 and in the energy balances. The first is the decreased use of coal halfway between the onset and the end of the period; this contributes to explaining the behavior of this factor in the industrial processing sector for the latter years analyzed. Thus, there is a substitution effect between fossil fuels that correctly capture the CI effect as was expected. Coal is the most pollutant of all primary sources. The second factor is the major penetration of natural gas, as shown by the energy balances. This fuel, although fossil in origin, has a much lower CO 2 emission level than coal or crude oil derivatives. If in 1996 the import of natural gas in Chile was inexistent, by 2006 it reached 224,428 TJ. The generation of electricity powered by natural gas when from 307 GWh in 1996 to 11,438 GWh in 2006; electricity powered by coal decreased from 9499 GWh in 1996 to 7212 GWh in 2005 (IEA, 2017). In 1999, the installed wind power was seven times that of 1997. As seen on Table 3 a, the sectoral analysis for the CI factor shows that one of the sectors that behaves, in this sub-period, as a compensating factor for CO 2 emissions is the transport sector. To explain this behavior, one must consider that despite the increase in electrical consumption in the transport sector (Annex 3), the energy balances show the progressive elimination of leaded gasoline, which is one of the most pollutant crude oil derivatives. Its use was finally abandoned in 2003. Ceasing to use leaded gasoline helped move Chile's economy forward, away from a high carbon system. Technological advances in engines to reduce fuel consumption contributes to explaining the behavior of this factor in thetransport sector. Annex 3 shows that due to its importance, efforts to reduce CO 2 emissions are imposed on the highway and maritime transport sector. The next factor analyzed is the RES factor that acted as a compensating factor of Chile's CO 2 emissions for this second sub-period (Table 3b). The RES factor captures the impact of renewable energy sources on energy related CO 2 emissions for countries without nuclear plants. It must be highlighted that the new Ralco hydropower plant began operations in 2004; it accounted for 10% of all power generation in the Central-SIC interconnected System (which shares 77.7% of total power capacity in Chile). The electricity generated by hydropower went form 21,979 GWh in 2004 to 29,129 in 2006 (IEA, 2017). The sub-period under analysis included the best hydrologic years between droughts (2002–2005). Another aspect of the explanation for this result could be found in the Project titled ‘Removal of obstacles for the electrification of rural areas using renewable energies’(Ferón et al., 2016) implemented in the first part of 2000. This is coherent with the behavior show by the Agriculture and Waste sectors, as indicated in Table 3b. Besides hydropower, other relevant renewable energy sources in power generation include biomass (Pontt et al., 2008). The project mentioned above proposed a successful policy for rural electrification. The importance of such projects focusing on the rural sector must be underscored because it allows energy from biofuel and waste to be reduced. The main usage of such energy is for cooking and heating. In 1990, consumption was 3133 TPES of which 1809 was for the residential sector. Consumption in 2013 was 10,340 of which 3734 was for the residential sector. Despite the role of biomass as a renewable energy source, the negative impact of its emissions on the health of populations could be a key question on the political agenda (Pablo- Romero et al., 2015). The energy intensity factor was analyzed next; this acted as a compensating factor of Chile's CO 2 emissions (Table 3 c). However, it is an effect with an unclear trend to help this economy's decoupling (positive in the first and last sub-periods but negative between 1997 and 2006). Table 3 Decomposition effects by sectors and sub-periods (GT ΔCO 2 ). Source: Own elaboration. 3a 3b ΔCI 1991–1996 1997–2006 2007–2013 ΔRES 1991–1996 1997–2006 2007–2013 1. Energy −2583.055 −14,299.551 −11,786.221 1. Energy 72.170 47.317 86.054 2. Transport 333.893 −2288.281 2468.022 2. Transport 22.939 −3.458 6.974 3. Industry 4.359 1573.424 −2004.833 3. Industry 35.989 −405.411 244.558 4. USOP −56.641 −63.793 −405.532 4. USOP 26.966 32.595 83.280 5. Agriculture −2675.637 12,628.761 −9226.940 5. Agriculture 38.123 −534.954 711.132 6. Waste −406.098 823.346 −181.216 6. Waste 13.935 −93.836 66.086 Total −5383.180 −1626.094 −21,136.720 Total 210.122 −957.747 1198.084 3c 3d ΔEI 1991–1996 1997–2006 2007–2013 ΔES 1991–1996 1997–2006 2007–2013 1. Energy 2789.701 8475.080 −3518.954 1. Energy −575.103 −3767.583 15,184.900 2. Transport −973.755 −3018.860 −2313.688 2. Transport 1514.784 1885.640 −1121.288 3. Industry −130.048 −569.663 −483.257 3. Industry −137.604 −30.779 −136.375 4. USOP 2.221 3.958 347.761 4. USOP −5.034 0.180 −26.338 5. Agriculture −331.319 −19,684.966 7035.517 5. Agriculture −1035.478 3037.406 −1842.592 6. Waste 188.744 −557.640 −547.906 6. Waste −418.237 −416.322 220.662 Total 1545.544 −15,352.090 519.474 Total −656.674 708.542 12,278.970 3e 3f ΔYp 1991–1996 1997–2006 2007–2013 ΔP1991–1996 1997–2006 2007–2013 1. Energy 7142.261 9267.417 9377.220 1. Energy 2022.518 3596.371 3436.325 2. Transport 3551.695 4395.972 3800.987 2. Transport 1010.401 1701.450 1432.207 3. Industry 1131.657 1650.035 1134.603 3. Industry 324.525 611.849 433.664 4. USOP 26.036 27.885 29.455 4. USOP 7.303 10.381 11.458 5. Agriculture 3875.938 3418.246 2283.499 5. Agriculture 1100.506 1340.511 877.957 6. Waste 819.169 972.786 688.840 6. Waste 232.542 361.652 259.367 Total 16,546.757 19,732.340 17,314.603 Total 4697.796 7622.213 6450.978 355J.M. Cansino et al. / Energy Economics 69 (2018) 350–366 The sign of the EI effect is determined by the decrease of the electrical supply due to the serious droughts suffered in Chile in 1998 and 1999. For the most part, generating electricity in the country depends upon annual water flow, given that hydroelectric power plants in Chile have no inter-annual reservoir capacity. The only exception was the Laja Lake hydroelectric power plant. Hydroelectric power generation decreased by 15.8% in 1998 and 14.8% in 1999 (IEA, 2017). The crisis led to an electricity supply rationing program in November of 1998; however, the system continued to be vulnerable, in which case there were a series of disruptions in the electrical supply in January and February of 1999. In June of 1999, the government of Chile approved the electricity rationing decree that was in place until August 31 of that year. Between 2006 and 2008 and also linked to energy efficiency measures, two relevant policies were put into force by Chilean Authorities. The first was the Country Energy Efficiency Program instituted at the onset of 2005 and that began to operate as of December 1, 2008 (CNE, 2008). In line with these results, it could be said that this initiative helped to enhance Chile's Energy Efficiency together with others factors such as high prices of oil products. Derived from the data shown on Table 3c, there were four key sectors that defined the EI path: the Transport sector, the waste sector, the Agriculture (greatly affected by droughts) and the industry sector. Even so this latter had a quantitative contribution that was less important. In fact, although Industry worked as a compensating sector in the data provided by the Energia2050; it advised on the energy intensity trend in the mining sector. This latter accounts for approximately 14% of the total energy consumption (CNE, 2008). Generally speaking, between 2000 and 2010, the energy intensity of the mining sector went from 0.63 Koe/K CLP$ in 2003 to 0.89 in 2010. Different from first sub-period, the ES factor drove CO 2 emissions (Table 3d). The main change from a sectoral perspective is registered in the Agriculture sector which became a higher driver sector of CO 2 emissions. When considering the National Accounts, between 1997 and 2006, this sector increased 29.9%. The Transport sector responded in the same way, as it also increased either its weight on Chile's total GDP (10.68% between the beginning and the end of the sub-period) or its role driving CO 2 emissions. On the contrary, the Energy sector continued to work as a compensating sector as was the case in the previous sub-period. Indeed, its role increased but failed to balance the driver sectors. During these years, Chile was oneof the countries that registered the most Clean Development Mechanism (CDM) projects, because at the beginning of this period, a study on the strategic use of this market based instrument, the establishment of the Designated National Authority for the CDM in Chile and an intensive campaign promoting opportunities for implementing CDM projects were performed. Despite attempts by Chile's government to integrate these developments into national policies (mainly in the energy sector), integrating CDM projects into national policy was weak until the COP-15 in Copenhagen. For the previous sub-period, the results for Income and Population effects show a clear coupling between economic growth and emissions. Once again, these effects reveal as clear drivers of CO 2 emissions. Result of Income effect is in line with Mundaca (2013). 4.4. From 2007 to 2013 This sub-period began with the interruption of the natural gas supply from Argentina and concludes when the series ends. In 2007, natural gas imports were reduced by 51.5% due to Argentina's decision to interrupt its exports of this power source (CNE, 2015). Argentina, was experiencing internal political and economic turmoil, restricted the export of natural gas. Chile had already suffered a cut in natural gas supplies from Argentina in 2002 due to labour strikes. Utility companies alerted the Chilean Government to the sudden energy security crisis (Mundaca, 2013). Power production –mainly electricity–then counted on the use of coal and diesel fuel. That same year when Argentina interrupted the natural gas supply, Chile's coal imports increased by 38.9% while diesel reached 112.3%. Those facts are coherent with a Tapio's index value of 1.53; thus implying an expansive negative decoupling. Only the CI factor works as a compensator for CO 2 emissions. In 2009, the COP-15 in Copenhagen was conducted to negotiate a new climate treaty. National awareness of the climate change problem, establishing climate change policy and procedures for implementing this policy were, in these years, the most important actions. Another relevant fact for this sub-period was Chile's intention to become a member of the Organization for Economic Co-operation and Development (OECD) and the guidelines set out for its members. In 2010, the Ministry of the Environment and the Office for Climate Change (later renamed Departamento de Cambio Climático, meaning Climate Change Department) were created. During this sub-period, the National Climate Change Action Plan (PNACC from the name in Spanish: Plan Nacional de Acción Contra el Cambio Climático 2008–2012 (CONAMA, 2008). It was the first document to articulate climate change policies in Chile. CI continues to prove to be a compensating factor. The sectorial analysis shows that the energy sector continues to be the greatest influence on the total effect of CI, but it reduces its absolute value with respect to the previous sub-period. Reductions in natural gas imports forced the power system to use more diesel and coal. The country needed to replace the (no longer available) natural gas supply from Argentina as fast as possible to reduce possible black outs while at the same time avoid increasing power prices as much as possible. Between 2006 and 2007 natural gas imports were reduces by 51.5% and between 2007 and 2008, 72%. For these last two years, diesel imports increased 112% while coal imports reached 38.9% (IEA, 2017). Between 2006 and 2007 electricity powered by natural gas decreased 59.5%; this was compensated by a 444% increase in electrical power produced from crude oil derivatives 21.5% from coal. As of 2009 and 2010, Chile once again imported natural gas, mainly from Algeria, Equatorial Guinea, Egypt and Indonesia. To avoid another ‘Argentinean crisis’,Chile embarked on the construction of large-scale liquefied natural gas (LNG) terminals (Mundaca, 2013). Although Chile managed to progressively substitute natural gasfrom Argentina in favor of other supplier countries, natural gas imports in 2014 were still well below 1999 statistics. Coal imports saw an upturn in 2011 and since then, these have remained stable. Diesel imports, on the other hand, have maintained a similar level as registered in 2007 when this resource was used to substitute, in part, natural gas from Argentina. Together with the Energy, USOP, Agriculture and Waste sectors, the industry sector also favored the mitigation of CO 2 emissions regarding its value for CI effect on Table 3a. For the two previously considered sub-periods, the industrial sector operated as a driver of CO 2 emissions once disaggregated from the CI factor. Offshoring processes could help explain this for the case of industry. When analyzing ES results, one must go back to this issue. This is an interesting result because one needs to bear in mind that the industrial sector is the country's largest energy consumer; it accounts for approximately 36% of the total energy consumption. Together with the mining sector, Chile's industry demands 64% of all electricity generated in the country (CNE, 2012; Duran et al., 2015). As was pointed out between 2007 and 2013, the CI factor took on a negative sign for the agricultural sector proving to favor the mitigation of CO 2 emissions (see Table 3a). This behavior is due to the policies implemented such as codes for the rational use of fertilizers or the incorporation of new technologies for cleaner production; these have improved agricultural practices. The healthy use of fertilizers and the adoption of diversified crop systems have also contributed to this result. However, the possible substitution effect that the negative value of the CI factor points out flows in the opposite direction than the behavior of the RES penetration factor. The RES factor works asa driver from 2007 356 J.M. Cansino et al. / Energy Economics 69 (2018) 350–366 to 2013 as a whole. This is coherent with the fact that Chile has intensified the use of fossil fuels as an electricity generation source, thus reducing the relative weight of hydropower. Also, this is partially associated with drought. Between 2006 and 2007, the electricity generated by hydropower decreased 20.6% and 14.1% between 2009 and 2010 (Ministry of Energy, 2016; IEA, 2017). After the electrical supply was rationed in 1999 and natural gas imports from Argentina were interrupted in 2007, the main objective behind Chile's power policy was to attain a reliable power supply. Proof of this was that between 2010 and 2013, eight new coal-powered power plants began producing, with a 2147.9 MW capacity (these new plants were Nuevas Ventanas, Santa María, Campiche, Bocamina II, Andina, Angamos 1 and 2 and Hornitos). For that same period, the new capacity of the hydroelectric power plants was less (430.2 MW). Recently Agurto et al. (2013) has stressed the relationship between power generation and the business cycle in Chile. In fact, the power industry has often been considered to be a leading indicator because of the positive correlation between economic development and the trend in electricity consumption (Karanfil and Yuanjing, 2015). In 2010, a policy measure promoting non-conventional renewable energies (NCRES), Law No. 20.257 (MEFR, 2008) to encourage the use of RES was put into practice. This law states that energy generating companies in Chile, with a capacity of 200 MW or more, must ensure that 10% of their energy generation each year originates from NCRE sources. This required quota started at 5% between 2010 and 2014, with an increase of 0.5% in 2015 and the final quota of 10% set for 2024. Electrical energy distributors supplying regular consumers are required to reach the 10% quota by 2010. The positive sign showed by the RES factor does not support the relevant impact for these measures. The presence of other renewable technologies is still minimal for these years in Chile, and therefore, the greater or lesser contribution of hydroelectric power is what determines the sign for this factor in the sector analyzed. Wind power production began in 2001 (7 MWh). In 2013, the wind energy generation capacity was still a minority (554 MWh compared to 19,737 MWh from hydropower). Solar power (PV) began in 2013 with 8 MWh. Despite its scarcity, RES deployment in Chile is also interesting because it would help deal with a difficult situation due to the unreliable energy supply, exacerbated by periods of pronounced droughts and problems accessing natural gas from neighboring countries, mainly Argentina. This became clear with the 2006 gas crisis in Argentina (Ponzo et al., 2011; Duran et al., 2015). The EI factor acted as a driver for this sub-period (Table 3 c). This is specifically true for USOP and Agriculture sectors. Nevertheless, during these years, interesting initiatives were launched, which focused on energy efficiency. Between 2012 and 2014 36 energy-savings projects had been launched and financed by the Asociación Chilena de Eficiencia Energética (Chilean Association for Energy Efficiency). However, these programs allow for a 34.18 GW/year savings in comparison with the 66,639.88 GWh generated in Chile in 2013, based on the information supplied by the aforementioned association (AChEE) and the Ministry of Energy (2016). Although insufficient to attain a negative EI sign, it must be mentioned that other initiatives were also launched in those same years. For example, the Programa País de Eficiencia Energética (Country's Energy Efficiency Program) was launched at the end of 2006; it was mining initiative for clean energy (IMEL in Spanish acronyms) that counted on the collaboration of the Corporación de Fomento de la Producción (COFRO in Spanish acronyms, meaning Development Cooperation for Production) (CNE, 2008). The development of IMEL was based on volunteer agreements by which mining companies were committed to focusing a greater percentage of their demand though NCRES projects and energy efficiency. Measures focusing on the residential sector must also be mentioned. In 2006, the Ministry of Housing launched a regulatory program for the thermal conditioning of dwelling. Since 2007, there are mandatory norms to be complied with when insulating buildings. The residential sector (including business premises and offices, in addition to housing) represent 73% of all wood burned. As of 2007, it was mandatory that refrigerators and light bulbs indicate their level of energy efficiency. Despite the initiatives mentioned and coherently with the EI factor sign, the Energy 2050 document has recognized that the measures to promote energy efficiency in Chile have failed to attain the expected results. The ES factor, when considering Chile's economy in general, does not allow a clear reading to be extracted regarding its role as a driver or how to compensate CO 2 emissions (see Table 3 d). One must look at the breakdownofitsvaluessectorbysectortobeabletoidentifyhowthechangein the relative weight of each sector on the general economy impacts on CO 2 emissions. With this in mind, it is useful to show the output for the sectors of Chile's economy considered for the period analyzed (Annex 1). When the ES factor is broken down by sectors, although the Energy and Waste sectors failed favoring the mitigation of CO 2 emissions, the rest of the sectors succeed. The energy sector increased its share on total GDP up to 36% from 2007 to 2013. Particular interest might be paid to industry because its role as a compensating factor increased over the previous period. Total energy consumption in Chile has more than doubled since 1991 (Ministry of Energy, 2013). Chilean national GHG inventories offer more detailed information of certain disaggregated industries. This information allows the results to be discussed in a more stylized way. Unfortunately, this level of disaggregation is not offered by others key data sources such as energy balances. That is why productive sectors were grouped into six in order to conduct the LMDI analysis. However, information from it is useful when looking at the figures of Table 3d. The iron industry (which is the most important) registered continued decreases in the absolute volume of emission as of 2007. Others-Methanol registered improved behavior. Between 2006 and 2010, the volume of total emissions decreased by 70%. The production of lime registered a poor behavior, which is nevertheless, not correlated with the volume of emissions by the cement industry. It must be also noted that in the case of Industry this result could be due to offshoring processes of Chile's more polluting industry towards other countries where environmental legislation is less demanding. The analysis in this paper does not allow us to perform such an analysis. It would be necessary to develop a Multi-Regional Input Output analysis to be able to respond to this question, which is similar to those by Andrew and Peters (2013) and Mundaca et al. (2015). Nevertheless, this analysis exceeds the objective of this article. On the other hand, the agriculture sector, although it shows an erratic behavior, needs special attention as it works as a driver during most of the latter years considered. For the case of the Yp and P factors, coupling is also seen for the 2007–2013 sub-period. A positive value is obtained for all of the economic sectors analyzed (Table 3e and f). The sectors where greater absolute values take on this factor are energy, transport and agriculture. In spite of their importance, the main result of PNACC 2008–2012 was improving the GEI inventory information at the national and regional level (University of Chile et al., 2015). This information may be useful to correctly design measures in the future; mitigating CO 2 measures were limited. The Assessment Plan report deepened very little into the elaboration of impact indicators and the development of mitigating policies. 4.5. Innovative Accounting Approach To avoid errors due to the variability of the data, before estimating the VAR model we studied the statistical properties (mean, variance, autocorrelation, etc.) of the times series of the VAR model. Statistical properties all remain constant over time (Brooks, 2014). 4.5.1. Generalized variance decomposition Table 4 shows the results of the generalized variance decomposition over a ten-year period for decomposition factor of CO 2 emissions. The variance decomposition shows how much of the predicted error 357J.M. Cansino et al. / Energy Economics 69 (2018) 350–366 variance of a variable is described by innovations generated from each independent variable in a system, over several time horizons. The empirical evidence indicates that 57.37% of the CI effect is due to its own innovative shocks. The standard deviation shock in RES factor is the variable that better explains carbon intensity with a percentage of 17.80%. Other factors that explain the CI are the P and EI effects with 9.61% and 7.23%, respectively. The rest of the factors, i.e., ES and Yp, explain carbon intensity with a lower percentage 3.83% and 4.16%, respectively. Astrongandsignificant portion of 48.83% of RES effect is explained by its own innovative shock. Carbon intensity explains the RES factor with 31.96%. RES explains the carbon intensity factor with 31.96%. The EI factor is explained, almost in the same portion, by its own standard innovative shock and by the shock on carbon intensity effect, 33.65% and 31.46%, respectively. The structural composition of Chile's economy is mainly affected by the RES penetration effect (35.02%) and Energy Intensity effect (50.19%). This factor is explained solely by its own innovative shock by a low portion of 3.17%. A significant portion of the Yp factor is explained by its own shocks (63.53%), while the contribution of the RES factor and EI factor to the economic activity effect is 9.21% and 15.33% respectively. Finally, the population effect is explained by its own standard innovative shock with a portion of 55.99%. The standard deviation shock in the RES penetration effect and Energy Intensity effect are 18.15% and 11.77%, respectively. Taking 5% as a threshold, we may infer that there is bidirectional causality between the carbon intensity and RES penetration effect. This means that one way to reduce the CO 2 emission intensity would be by decreasing fossil fuel consumption with respect to the total primary energy, i.e., increasing the use of renewable energies. By also considering the reference of 5%, we may infer that there is unidirectional causality from CI to EI, from RES to EI and to ES and from EI to ES and to Yp. Common to this statistical evidence of the variance decomposed into CO 2 emission driver results allow us to highlight the following: 1. If the RES penetration effect decreases, then the carbon intensity decreases also. This is because the renewable energy source share in total primary energy is greater; therefore, the quality of fossil fuels is better. 2. The CI effect and RES penetration effect causes the Energy Intensity effect. Besides there is a unidirectional causality from RES penetration effect to Economy Structure effect. This means that if the share of the RES increases in the total primary energy matrix i.e., the carbon intensity and RES penetration effect decrease, the energy efficiency or the technology level of Chile's economy will be changed, which could highlight the importance of certain sectors in this economy. 3. The impact of structural change in Chile's economy in terms of CO 2 emissions and the income effects on CO 2 emission changes from energy consumption are changed by the energy efficiency of a Chile's economy due to the unidirectional causality of Energy Intensity factor to ES effect. 4.5.2. Impulse-response analysis With the aim of simulating the behavior of the variables involved in the study over time, the correlations among decomposition factors of CO 2 emissions were examined by the assistance of Impulse-Response Functions (IRF). Fig. 2 represents the IRFs for the six-factor of decomposition of CO 2 emissions. It can be seen that in the short term, carbon intensity reacts more significantly to shocks in itself and the RES penetration effect, compared to shocks in other variables. The reaction to RES has the same behavior and opposite to shocks in ES and Yp. Nevertheless, the latter ends up disappearing in the long term. The RES penetration effect has a significant reaction to a shock in the carbon intensity factor and to a shock in itself until it reaches the 4th time horizon, but dissipates in the long run. The response to shocks in other factors decreasing in the long run. The energy intensity reacts more sharply to shocks in RES and CI. From the 4th horizon, there is a bigger reaction to shock in CI than to shock in RES. The economy structure reacts to all factors; it is the strongest reaction to shocks in CI, RES and EI. After the 5th period, the response to these shocks decreases to dissipate in the long run. The income and population effects have a significant reaction to a shock in all variables in the sense explained in Section 4.1;consequently, acting on these factors through policy measures is recommended; consequently, acting on these factors through policy measures is recommended. The response to these shocks decreasing in the long run. The analysis of IRFs suggests the occurrence of the same causality relationships that were observed in variance decomposition analysis. Most of the factors have a significant reaction to a shock in the RES penetration and the Energy Intensity effect. These results show the importance of the mix level resource efficiency that can mitigate CO 2 emissions and energy consumption without compromising economic growth and change of population. These are useful findings to evaluate Table 4 Variance decomposition. Source: Own elaboration. Period CI RES EI ES Yp P Variance decomposition of CI 1 100.0000 0.000000 0.000000 0.000000 0.000000 0.000000 2 78.68459 14.12759 6.036208 0.919729 0.127448 0.104432 3 62.40732 22.55178 8.845984 3.332951 0.098039 2.763927 4 57.76604 21.65709 8.579019 4.489065 0.498203 7.010589 5 57.01630 19.66260 7.789011 4.334628 1.825365 9.372095 10 57.36806 17.79772 7.230208 3.835006 4.163441 9.605572 Variance decomposition of RES 1 24.89783 75.10217 0.000000 0.000000 0.000000 0.000000 2 35.78299 58.67517 0.601012 0.813575 0.925649 3.201607 3 35.17092 55.78895 3.919351 0.935910 0.882389 3.302483 4 31.20122 55.24450 6.638193 2.445270 1.001830 3.468993 5 30.21001 52.96596 7.062505 3.376165 0.956703 5.428663 10 31.95710 48.83296 6.714224 3.252251 2.540219 6.703241 Variance decomposition of EI 1 37.98300 20.27005 41.74695 0.000000 0.000000 0.000000 2 38.06957 15.78465 43.70422 0.778734 0.292270 1.370554 3 31.15882 22.20378 40.87566 3.613985 0.738783 1.408967 4 29.22731 22.33477 38.02486 5.497191 0.630469 4.285403 5 30.20942 20.68818 35.39482 5.663372 1.231191 6.813022 10 33.65725 19.02483 31.46369 5.049301 3.380212 7.424720 Variance decomposition of ES 1 0.334286 39.44275 57.48359 2.739378 0.000000 0.000000 2 3.034888 39.51262 54.23979 2.500600 0.000174 0.711930 3 5.054140 37.53578 53.93384 2.573431 0.014537 0.888271 4 5.986930 36.96307 53.21447 2.893571 0.034331 0.907630 5 6.613335 36.48545 52.32681 3.183683 0.040556 1.350168 10 8.783025 35.02054 50.18719 3.169882 0.674321 2.165047 Variance decomposition of Yp 1 0.673929 5.285425 6.412169 4.448004 83.18047 0.000000 2 1.399817 5.055993 9.761632 4.149186 78.70506 0.928312 3 1.557434 7.560474 13.84851 4.423392 71.60208 1.008107 4 1.596117 9.230652 15.66796 5.237689 67.03681 1.230772 5 1.911601 9.151710 15.87435 5.625918 65.40882 2.027602 10 3.894205 9.211015 15.33564 5.475892 63.53469 2.548558 Variance decomposition of P 1 3.371292 11.03700 4.339709 2.097658 0.082839 79.07150 2 7.175800 10.42025 6.891046 2.815543 0.262147 72.43521 3 6.940962 15.27302 10.09441 2.744607 0.761851 64.18515 4 6.337022 18.40638 11.91689 3.733337 1.182109 58.42426 5 6.340097 18.21802 12.14377 4.295290 1.146931 57.85589 10 7.975771 18.15517 11.77387 4.207951 1.889267 55.99798 358 J.M. Cansino et al. / Energy Economics 69 (2018) 350–366 Ang, B.W., 2005. The LMDI approach to decomposition analysis: a practical guide. Energy Policy 33 (7), 867–871. Ang, B.W., Choi, K.H., 1997. Decomposition of aggregate energy and gas emission intensities for industry: a refined Divisia index method. Energy J. 18 (3), 59–73. Ang, B.W., Liu, N., 2007a. Negative-value problems of the logarithmic mean Divisia index decomposition approach. Energy Policy 35 (1), 739–742. Ang, B.W., Liu, N., 2007b. Energy decomposition analysis: IEA model versus other methods. Energy Policy 35 (3), 1426–1432. Ang, B.W., Su, B., 2016. Carbon emission intensity in electricity production: a global analysis. Energy Policy 94, 56–63. Ang, B.W., Zhang, F.Q., Choi, K.H., 1998. Factorizing changes in energy and environmental indicators through decomposition. Energy 23 (6), 489–495. Bager, S., Mundaca, L., 2015. How smart are electricity users with ‘smart metering’?Abehavioural economics experiment. Paper Presented at the 38th International Association for Energy Economics (IAEE) International Conference, Antalya, Turkey, May 25–27. BCC, 2016a. Statistics Database, National Accounts, GDP Expenditure and Income, Linked Series Reference 2008, GDP by Class of Economic Activity. Banco Central de Chile, Santiago, Chile (http://si3.bcentral.cl/Siete/secure/cuadros/home.aspx?Idioma=en-US). BCC, 2016b. Statistics Database, National Accounts, GDP Expenditure and Income, Linked Series Reference 2008, GDP Deflactor. Banco Central de Chile, Santiago, Chile (http:// si3.bcentral.cl/Siete/secure/cuadros/home.aspx?Idioma=en-US). BCC, 2016c. Statistics Database, Employment, Wages and Demographics. Banco Central de Chile, Santiago, Chile (http://si3.bcentral.cl/Siete/secure/cuadros/home.aspx?Idioma= en-US). Bonilla, J.A., Cruz, A., Espinosa, M., Fabrikov, D., Fuertes, A., Garrido, J.M., Gómez, F., Sánchez, I., Segura, A., Valenzuela, F.J., Currás, J.M., 2012. Tecnologías de captura y almacenamiento de CO 2 . In: González, C.V., Cerón, M.C. (Eds.), I Certamen de Proyectos Educativos en Ingeniería Química en la UAL - Libro de actas, Editorial Universidad de Almería, Almería, pp. 18–29. Brizga, J., Feng, K., Hubacek, K., 2013. Drivers of CO 2 emissions in the former soviet union: a country level IPAT analysis from 1990 to 2010. Energy 59, 743–753. Brooks, C., 2014. Introductory econometrics for finance. Cambridge University Press, Cambridge. Cansino, J.M., Sánchez-Braza, A., Rodríguez-Arévalo, M.L., 2015. Driving forces of Spain's CO 2 emissions: a LMDI decomposition approach. Renew. Sust. Energ. Rev. 48, 749–759. Cansino, J.M., Román, R., Ordóñez, M., 2016. Main drivers of changes in CO 2 emissions in the Spanish economy: a structural decomposition analysis. Energy Policy 89, 150–159. CEPAL, 2012. La Economía del Cambio Climático en Chile. Comisión Económica para América Latina y el Caribe (CEPAL), Naciones Unidas, Santiago, Chile (http://repositorio.cepal.org/ bitstream/handle/11362/35372/S2012058_es.pdf?sequence=17). CNE, 2008. Informe Final de Evaluación Programa País de Eficiencia Energética. Comisión Nacional de Energía (National Commission of Energy), Chilean Ministry of Energy, Santiago, Chile (http://www.dipres.gob.cl/595/articles-141121_informe_ final.pdf). CNE, 2012. Balance Nacional de Energía, 2012. Comisión Nacional de Energía (National Commission of Energy), Chilean Ministry of Energy, Santiago, Chile (http:// energiaabierta.cne.cl/balance-energetico/). CNE, 2015. Balance Nacional de Energía, 2015. Comisión Nacional de Energía (National Commission of Energy), Chilean Ministry of Energy, Santiago, Chile (http:// energiaabierta.cne.cl/balance-energetico/). Commoner, B., Corr, M., Stamler, P., 1971. The Closing Circle: Nature, Man, and Technology. Knopf, New York. CONAMA, 2008. Plan de Acción Nacional de Cambio Climático PANCC (2008–2012). Comisión Nacional del Medio Ambiente, Santiago, Chile (http://www.mma.gob.cl/ 1304/w3-article-54787.html). Coronado, E., 2016. Impuesto Chileno Sobre Emisiones a la Atmósfera Provenientes de Fuentes Fijas (Un Estudio a la Luz de la Legislación Española). Doc. No. 2/2016. Instituto de Estudios Fiscales, Madrid. Duran, E., Aravena, C., Aguilar, R., 2015. Analysis and decomposition of energy consumption in the Chilean industry. Energy Policy 86, 552–561. Ferón, S., Heinrichs, H., Cordero, R.R., 2016. Sustainability of rural electrification programs based on off-grid photovoltaic (PV) systems in Chile. Energy Sustain. Soc. 6 (32) (19 pp). Goldemberg, J., Johansson, T.B. (Eds.), 2004. World Energy Assessment: Overview 2004 Update. UNDP, United Nations Development Programme, New York (http://www.leonardo-energy.org/world-energy-assessment-overview-2004- update). Government of Chile, 2015. Intended Nationally Determined Contribution of Chile Towards the Climate Agreement of Paris 2015. The Committee of Ministers for Sustainability and Climate Change, Government of Chile (http://www4.unfccc.int/ submissions/INDC/Published%20Documents/Chile/1/INDC%20Chile%20english% 20version.pdf). Han, Y., Fang, X., Liu, Y., Guo, D., Kan, S., 2012. A new decomposition method of influential factors on energy consumption intensity. Paper presented at the 2012. International Symposium on Computer, Consumer and Control (IS3C), National Chin-Yi University, Taichung, Taiwan, June 04–06 (http://ieeexplore.ieee.org/document/ 6228418/). Hatzigeorgiou, E., Polatidis, H., Haralambopoulos, D., 2008. CO 2 emissions in Greece for 1990–2002: a decomposition analysis and comparison of results using the Arithmetic Mean Divisia Index and Logarithmic Mean Divisia Index techniques. Energy 33 (3), 492–499. Hatzigeorgiou, E., Polatidis, H., Haralambopoulos, D., 2010. Energy CO 2 emissions for 1990–2020: a decomposition analysis for EU-25 and Greece. Energy Sources Part A 32 (20), 1908–1917. Hoekstra, R., van den Bergh, J., 2003. Comparing structural and index decomposition analysis. Energy Econ. 25 (1), 39–64. IEA, 2017. Statistics Report, Energy Balances. International Energy Agency, OECD/IEA, Paris (http://www.iea.org/statistics/). IEA and NEA, 2015. Projected Costs of Generating Electricity - 2015 Edition. International Energy Agency, OECD/IEA, and Nuclear Energy Agency, OECD/NEA, Paris and Issy-les- Moulineaux (https://www.iea.org/publications/freepublications/publication/projectedcosts-of-generating-electricity-2015-edition.html). Karanfil, F., Yuanjing, L., 2015. Electricity consumption and economic growth: exploring panel-specific differences. Energy Policy 82, 264–277. Kaya, Y., 1990. Impact of carbon dioxide emission control on GNP growth: interpretation of proposed scenarios. Paper Presented to the Energy and Industry Subgroup, Response Strategies Working Group. Intergovernmental Panel on Climate Change, Paris. Lee, C.C., Chien, M.S., 2010. Dynamic modelling of energy consumption, capital stock, and real income in G-7 countries. Energy Econ. 32 (3), 564–581. Leontief, W., Ford, D., 1972. Air pollution and the economic structure: empirical result of input–output computations. In: Brody, A., Carter, A.P. (Eds.), Input-Output Techniques. North-Holland Publishing Company, Amsterdam-London, pp. 9–30. Ma, C., Stern, D.I., 2008. China's changing energy intensity trend: a decomposition analysis. Energy Econ. 30 (3), 1037–1053. Maghyereh, A., 2004. Oil price shocks and emerging stock markets: a generalized VAR approach. Int. J. Appl. Econ. Quant. Stud. 1 (2), 27–40. Malvik, H.V., Hannisdahl, O.H., Wensaas, G.B., 2013. The future is electric! The EV revolution in Norway - Explanations and lessons learned. ECEEE Summer Study Proceedings, Panel 4 Transport and Mobility: How to Deliver Energy Efficiency, 4-102-13, 995-1005. European Council for an Energy Eficient Economy, Stockholm (http://proceedings.eceee.org/visabstrakt.php?event=3&doc=4- 102-13). MAPS Chile, 2014. Opciones de Mitigación para Enfrentar el Cambio Climático: Resultados de Fase 2. Ministerio del Medio Ambiente, Santiago, Chile (http:// www.mapschile.cl/). MEFR, 2008. Ley N°. 20.257, que Introduce Modificaciones a La Ley General de Servicios Eléctricos Respecto de la Generación de Energía Eléctrica con Fuentes De Energías Renovables No Convencionales. Ministerio de Economía, Fomento y Reconstrucción, Santiago, Chile (http://www.cne.cl/archivos_bajar/20_257_1.pdf). Ministry of Energy, 2013. Action Plan of Energy Efficiency 2020. Chilean Ministry of Energy, Santiago, Chile (http://www.amchamchile.cl/UserFiles/Image/Events/octubre/ energia/plan-de-accion-de-eficiencia-energetica2020.pdf). Ministry of Energy, 2016. Energía 2050. Política Energética de Chile. Chilean Ministry of Energy, Santiago, Chile (http://www.energia2050.cl/es/energia-2050/energia-2050- politica-energetica-de-chile/). Ministry of Environment, 2011. 2ª Comunicación Nacional de Chile ante la Convención Marco de las Naciones Unidas sobre Cambio Climático. Chilean Ministry of Energy, Santiago, Chile (http://www.mma.gob.cl/1304/articles- 50880_docomunicadoCambioClimatico.pdf). Ministry of Environment, 2016. Plan de Acción Nacional de Cambio Climático 2017– 2022. Chilean Ministry of Energy, Santiago, Chile (http://portal.mma.gob.cl/wpcontent/uploads/2016/04/Anteproyecto-PANCC-2017-2022-FINAL-2016-04- 18.pdf). Moutinho, V., Madaleno, M., Silva, P.M., 2016. Which factors drive CO 2 emissions in EU- 15? Decomposition and innovative accounting. Energy Effic. 9 (5), 1087–1113. Mundaca, L., 2013. Climate change and energy policy in Chile: up in smoke? Energy Policy 52, 235–248. Mundaca, L., Román, R., Cansino, J.M., 2015. Towards a green energy economy? A macroeconomic-climate evaluation of Sweden's CO 2 emissions. Appl. Energy 148, 196–209. Mwasilu, F., Justo, J.J., Kim, E.K., Do, T.D., Jung, J.W., 2014. Electric vehicles and smart grid interaction: a review on vehicle to grid and renewable energy sources integration. Renew. Sust. Energ. Rev. 34, 501–516. O'Mahony, T., 2013. Decomposition of Ireland's carbon emissions from 1990 to 2010: an extended Kaya identity. Energy Policy 59, 573–581. O'Ryan, R., Díaz, M., Clerc, J., 2010. Energy Consumption, Greenhouse Gas Emissions and Mitigation Options for Chile, 2007–2030. Program for Environmental Economics and Management (PROGEA). Department of Industrial Engineering, University of Chile, Santiago, Chile (http://www.energycommunity.org/documents/Aplicacion% 20de%20LEAP%20en%20Chile,%202010.pdf). Ozkan, F., Ozkan, O., 2012. An analysis of CO 2 emissions of Turkish industries and energy sector. Reg. Sect. Econ. Stud. 12 (2), 65–85. Pablo-Romero, M.P., Román, R., González-Limón, J.M., Praena-Crespo, M., 2015. Effects of fine particles on children's hospital admissions for respiratory health in Seville, Spain. J. Air Waste Manage. Assoc. 65 (4), 436–444. Park, J., Ratti, R.A., 2008. Oil price shocks and stock markets in the US and 13 European countries. Energy Econ. 30 (5), 2587–2608. Pontt, J., Pontt, C., Guiñez, C., 2008. Estudio de Contribución de las ERNC al SIC 2025. Informe Sectorial Final sobre Potencial de la Biomasa para la Generación Eléctrica en Chile. Universidad Técnica Federico Santa María, Valparaíso, Chile (http:// www.neim.utfsm.cl/arch/20080808-02-Biomasa.pdf). Ponzo, R., Dyner, I., Arango, S., Larsen, E.R., 2011. Regulation and development of the Argentinean gas market. Energy Policy 39 (3), 1070–1079. Robaina-Alves, M., Moutinho, V., 2013. Decomposition analysis and Innovative Accounting Approach for energy-related CO 2 (carbon dioxide) emissions intensity over 1996–2009 in Portugal. Energy 57 (1), 775–787. Rose, A., Casler, S., 1996. Input-output structural decomposition analysis: a critical appraisal. Econ. Syst. Res. 8 (1), 33–62. Sanhueza, J.E., Ladrón de Guevara, F.A., 2014. A case study of Chilean mitigation actions. Climate Dev. 6 (Sup.1), 34–42. 365J.M. Cansino et al. / Energy Economics 69 (2018) 350–366 Shahbaz, M., Hye, Q.M.A., Tiwari, A.K., Leitão, N.C., 2013. Economic growth, energy consumption, financial development, international trade and CO 2 emissions in Indonesia. Renew. Sust. Energ. Rev. 25, 109–121. Stern, D.I., Jotzo, F., 2010. How ambitious are China and India's emissions intensity targets? Energy Policy 38 (11), 6776–6783. Su,B.,Ang,B.W.,2012.Structural decomposition analysis applied to energy and emissions: some methodological developments. Energy Econ. 34 (1), 177–188. Tapio, P., 2005. Towards a theory of decoupling: degrees of decoupling in the EU and the case of road traffic in Finland between 1970 and 2001. Transp. Policy 12 (2), 137–151. UNFCCC, 2009. Report of the Conference of the Parties on its Fifteenth Session, Held in Copenhagen From 7 to 19 December 2009. United Nations Framework Convention on Climate Change, Bonn (http://unfccc.int/meetings/copenhagen_dec_2009/items/ 5262.php). UNFCCC, 2016. Submitted biennial update reports (BURs) from non-Annex I Parties. United Nations Framework Convention on Climate Change, Bonn (http://unfccc.int/national_ reports/non-annex_i_natcom/reporting_on_climate_change/items/8722.php). University of Chile, Poch Ambiental, Adapt-Chile, 2015. Evaluación de Término del Plan de Acción Nacional de Cambio Climático 2008–2012. (Santiago, Chile, http:// portal.mma.gob.cl/wp-content/uploads/2014/11/Evaluacion-PANCC-2008-2012- Fase-I-Licitacion_conAnexos.pdf). Vehmas, J., Luukkanen, J., Kaivo-oja, J., 2007. Linking analyses and environmental Kuznets curves for material flows in the European Union 1980–2000. J. Clean. Prod. 15 (17), 1662–1673. Vivas, F.J., De Las Heras, A., Segura, F., Andújar, J.M., 2016. Aproposalofenergymanagement strategy on hybrid renewable system with hydrogen backup. Paper Presented at the 7th International Renewable Energy Congress (IREC) International Conference, Hammamet, Tunissia, March 22-24. Voigt, S., De Cian, E., Schymura, M., Verdolini, E., 2014. Energy intensity developments in 40 major economies: structural change or technology improvement? Energy Econ. 41, 47–62. Wang, Q., Li, R., Jiang, R., 2016. Decoupling and decomposition analysis of carbon emissions from industry: a case study from China. Sustainability 8 (10), 1–12. Yamaji, K., Matsuhashi, R., Nagata, Y., Kaya, Y., 1991. An integrated system for CO 2 /energy/ GNP analysis: case studies on economic measures for CO 2 reduction in Japan. Presented at the Workshop on CO 2 Reduction and Removal: Measures for the Next Century. International Institute for Applied Systems Analysis, Laxemburg, Austria. York, R., Rosa, E., Dietz, T., 2002. Bridging environmental science with environmental policy: plasticity of population, affluence, and technology. Soc. Sci. Q. 83 (1), 18–34. Zhang, X.P., Cheng, X.M., 2009. Energy consumption, carbon emissions, and economic growth in China. Ecol. Econ. 68 (10), 2706–2712. Zhang, Y., Zhang, J., Yang, Z., Li, S., 2011. Regional differences in the factors that influence China's energy-related carbon emissions, and potential mitigation strategies. Energy Policy 39 (12), 7712–7718. 366 J.M. Cansino et al. / Energy Economics 69 (2018) 350–366