Electricity restructuring, greenhouse gas emissions efficiency and employment reallocation

Energy Policy 92 (2016) 468–476

Contents lists available at ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

Electricity restructuring, greenhouse gas emissions efficiency and employment reallocation Dongha Kim, Jinook Jeong n School of Economics, Yonsei University, Seoul, Republic of Korea

H I G H L I G H T S







GHG emissions efficiency and employment reallocation have a feedback effect. GHG emissions efficiency improvement increases the employment shares of the related industries except mining. Employment share increases in those industries have negative effects on emissions efficiency. Electricity restructuring increases overall employment except farming.

art ic l e i nf o

a b s t r a c t

Article history: Received 21 September 2015 Received in revised form 4 February 2016 Accepted 4 February 2016 Available online 15 February 2016

This paper focuses on demonstrating the feedback relationships between greenhouse gas emissions efficiency in the electricity sector and employment reallocation with a consideration of the effects of electricity restructuring and socio-demographic factors. We postulate the construction, information, manufacturing, utilities, and mining sectors as a group of emissions efficiency-related industries and identify the mutual relationships. The emissions efficiency positively influences the job shares of these industries except mining, whereas increased employment in these industries has a negative effect on the emissions efficiency. Electricity restructuring has a positive effect on overall employment, however, it does not have a statistically significant effect on the emissions efficiency. Additionally, population aging and educational attainment have positive effects on the emissions efficiency, and a higher proportion of rental households has a negative influence on it. Increases in renewable energy and nuclear energy generation have elastic effects on enhancing emissions efficiency. & 2016 Elsevier Ltd. All rights reserved.

JEL classification: Q43 Q48 Q52 Keywords: GHG emissions efficiency Employment reallocation Electricity restructuring Feedback effects Socio-demographic changes

1. Introduction This study assesses a feedback effect between greenhouse gas emissions efficiency (GHG-EE) in the electricity sector and employment reallocation (ER) in the U.S. with a consideration of effects of electricity restructuring and socio-demographic factors. The electricity market comprises the four main areas of generation, transmission, distribution, and retail market. Since the 1990s, a competitive environment has initiated regarding generation and the retail market. However, the transmission and distribution markets have remained under regulation for the stable operation of the electrical power systems. The effects of electricity market deregulation have been thoroughly examined because the n

Corresponding author. E-mail address: [email protected] (J. Jeong).

http://dx.doi.org/10.1016/j.enpol.2016.02.009 0301-4215/& 2016 Elsevier Ltd. All rights reserved.

electrical power system is a major infrastructure sustaining economic activities. Although the 2000 California electricity crisis led to a suspension of restructuring schemes in seven other states, 15 states continued to deregulate their wholesale markets and consumers' retail options.1 The main purpose of electricity restructuring is an improvement of operational efficiencies because the naturally integrated system has yielded side effects and inefficient management (Joskow, 1997). In other words, price reductions are expected to result from the competitive environment, since electricity itself is a homogeneous product. Also, more efficient firms will enter the 1 Arizona, Arkansas, California, Montana, Nevada, New Mexico, and Virginia suspended deregulation; Delaware, District of Columbia, Illinois, Maine, Maryland, Massachusetts, Michigan, New Hampshire, New Jersey, New York, Ohio, Oregon, Pennsylvania, Rhode Island, and Texas continued their restructuring policies.

D. Kim, J. Jeong / Energy Policy 92 (2016) 468–476

electricity market and incumbents put more efforts to retain their competitiveness after the deregulation. As electricity has played a more important part in recent lifestyle, structural changes in the electricity sector could affect the overall economy. In addition, the United Nations Framework Convention on Climate Change was forged in 1992 to stabilize GHG emissions and the Kyoto Protocol was established in 1997 to oblige the abatement of the emissions to a specific level in each regulated country. Despite several controversies about the historical responsibilities of developed and developing countries, environmental regulations and policies pertaining to GHG emissions or the applications of renewable energy sources have increased globally. Although the U. S. has not formally participated in any global mandatory controls related to environmental issues, gradual implementations of environmental legislations have proceeded, such as Renewable Portfolio Standards (RPS). From this perspective, restructuring the electricity sector, which accounts for about 40 percent of national energy consumption (US Energy Information Administration, 2015), could also affect GHG-EE through several channels of the demand and supply sides. Furthermore, electricity restructuring and GHG-EE could matter to employment. Recent efficiency gains from technology progress may destroy jobs, because automation and mechanization have replaced workers. As the information age has progressed, job destruction can be severe because small numbers of workers can be hugely productive in the growing service sector. Nevertheless, the overall patterns of job destruction and job creation depend on the cost of renovation, and vary across sectors (Mortensen and Pissarides, 1998) and categories of work (Autor, 2015). Improvements in GHG-EE can positively influence job creation not only in the electricity sector but also in the related industries. This is because the process of the improvements in the electricity sector is highly labor-intensive (Rifkin, 2014). In the generation area, there is a need to install additional equipment or facilities in existing thermoelectric power plants, for instance, integrated gasification combined cycle (IGCC) power system or carbon capture and storage (CCS). Also, there should be an expansion of power plants for renewable energy sources, such as solar, wind and nuclear, instead of the traditional ones (Sims et al., 2003). The transmission and distribution system would be replaced to secure a capacity for distributed generation and information-based management system, so called smart grid. Finally, in the retail area, an installation of new information and communication technology devices would be achieved for real-time measurement and collection of data on electricity use and the emissions, such as smart metering (Gungor et al., 2011). Also, newly growing electric products, such as plug-in hybrid electric vehicle, would be important factors to GHG-EE in the electricity retail market (Hu et al., 2015). Since these replacements and developments are expected to gradually progress over the short and mid-term (Rifkin, 2014), GHG-EE development could generate numerous jobs in electricity and related-sectors in this period. This study postulates construction, information, manufacturing, utilities, and mining as the related industries, and figures out the effects of GHG-EE in the electricity sector on employment in these industries as well as overall employment. Numerous studies consider the influences of electricity restructuring on GHG emissions (such as CO2) and job creation. However, these studies limit their focus only to description of possible scenarios and outcomes without empirical analysis. In addition, employment, which relates to the economic activities of these industries, could simultaneously have significant influences on GHGEE. However, there are few studies dealing with this possible feedback effect between GHG-EE and employment reallocation. In addition, this study includes the effects of socio-demographic changes on the GHG-EE. Not only in developed countries,

469

but also in many of developing countries, the proportion of the elderly shows a steep increase. In 2015, most of developed countries, such as U.S., Russia, and U.K, show 20–24 percent as the proportion of population aged 60 or over, and some European countries, such as France, Germany, and Italy, show 25–29 percent. Most of developing countries, such as China and Republic of Korea, show 10–19 percent of the proportion. In 2050, however, most of countries, except some countries of Africa, and South and Central Asia, will become super-aging society (UNDESA Population division, 2015). Since the dramatic changes of demographic structure affect a sustainability of national economic system and policies, many economic literature have diagnosed the possible negative effects of the aging society on the economy (Cutler et al., 1990; Gruber and Wise, 2001). In this respect, some papers deal with immigration as a practical solution to retain working-age population and economic growth (Blank, 2016). Among socio-demographic changes, this study focuses on population aging, and increasing educational attainment, rental households and one-person households, related to electricity consumption. In this study, we aim to identify a feedback relationship between GHG-EE and ER by constructing a simultaneous equation system. Also, we examine the determinants of the implementation of electricity restructuring, which affects both GHG-EE and ER, to apply to the identification of the mutual relationship. To this end, socio-demographic, economic, and electricity-related variables are considered. This study is presented as follows. Section 2 provides a brief overview of relevant literature. Section 3 describes the data used in the analysis and the estimation model. Section 4 presents the empirical results and Section 5 discusses the conclusions and policy implications from the results.

2. Literature review Since electricity services and power systems are significant parts of social overhead capital, stable operation is the first priority for a sustainable economy. Hence, the power systems were initially owned by the central government, and thus electricity markets had natural monopoly properties under an integrated system. Under the Federal Public Utility Regulatory Policies Act of 1978 and Energy Policy Act of 1992, Qualifying Facilities and Independent Power Producers (as defined under those laws) could act as new entrants in the electricity market. Furthermore, between 1990 and 2000, 16 states ultimately opened their wholesale markets and retail options to consumers. Due to the importance of electrical services (described above), many studies have investigated the influences of electricity restructuring on the US economy in the years around 2000. Nevertheless, there are few studies which empirically deal with the effects of socio-demographic factors on deregulation of electricity market. As for energy efficiency, however, Zarnikau (2003) and Bollino (2009) show that the higher income and education attainment, the higher willingness to pay (WTP) for energy efficiency. Also, age has a negative impact on the WTP for the efficiency (Zarnikau, 2003) and homeowners present the higher mean of WTP than rental households (Bollino, 2009).2 In addition, as the main purpose of restructuring is a reduction of electricity price through operational efficiency improvement under competition, high electricity price and inputs prices, such as natural gas price, can be an important impetus for deregulation (Fagan, 2006). 2 Zarnikau (2003) also acknowledged that homeowners will be positive to energy efficiency improvement, which contradicts to estimation results in the paper.

470

D. Kim, J. Jeong / Energy Policy 92 (2016) 468–476

and Sohngen (2014) present that RPS could achieve lower carbonintensity, using panel data. However, Yin and Powers (2010) find the ineffectiveness of RPS and point out the weakness when efficiency of RPS is uniformly measured by a dummy variable. In this respect, this study investigates the feedback relationships between GHG-EE in the electricity sector and ER in electricity-related industries under consideration with electricity restructuring, which could not be satisfied by the existing literature. Also, this study can enlarge our understanding of the influences of recent socio-demographic changes on GHG-EE. The set of variables used in the estimation model (described in Section 3) were chosen based on the above relevant literature.

Regarding GHG emissions, Palmer (1999) investigates the possible effects of electricity restructuring on GHG emissions according to the demand and supply sides of electricity. On the demand side, the introduction of competition in the wholesale market could draw more efficient entrants so that they induce lower average retail prices. In turn, lower prices lead to higher demands for electricity, and emissions thereby increase.3 Furthermore, the deregulation could lead to a higher possibility of implementing real-time pricing or retaining demand-side management programs, which reduce GHG emissions. On the supply side, deregulation could reduce emissions through expansions of nuclear and renewable energy generation and a change to the composition of inter-regional electricity trading. Akorede et al. (2010) point out the importance of a distributed generation system on the improvement to GHG-EE. Also, Mansur (2007) suggests that utilities may change their strategic behavior in the competitive environment after electricity restructuring and under several environmental regulations, and this change can abate their emissions. However, Burtraw et al. (2000) emphasize that the competitive environment after the deregulation could induce the higher utilization of cheaper inputs, such as coal, which emit more CO2 , even though the improvement to GHG-EE after restructuring. In the perspective of employment, Markiewicz et al. (2004) point out that expenditures on employment after restructuring depend on the extent to which utilities can transfer costs to retail prices. In their empirical analysis, the results depend on whether they are investor-owned utilities or municipal utilities (MUs), and they find positive effects of restructuring on job creation in the case of MUs. Also, Roland-Holst (2008) finds that an increase in overall energy efficiency can generate positive outcomes for job creation in California. Related to socio-demographic factors, using an input–output model, Kronenberg (2009) suggests that an aging society could increase energy consumption mostly in the health and education sectors, which does not reduce GHG emissions. In addition, Tonn and Eisenberg (2007), York (2007), and Brounen et al. (2012) provide evidence that several socio-demographic changes, such as population aging and an increase of one-person households, could increase energy consumption and GHG emissions. However, Gough et al. (2011), O'Neill et al. (2010), and Dalton et al. (2008) present aging society would decrease GHG emissions, even though one-person households (or lower household size) consistently increase GHG emissions. With regard to rental households, Druckman and Jackson (2008) show that the effects of rental households on GHG emissions depend on the relative magnitudes between lower energy consumption and less incentive to use energy efficient measures. Also, Vassileva et al. (2012) provide that the GHG-EE of rental households depends on whether tenants separately pay their energy bills, and their monthly incomes. Finally, regarding public awareness of environmental issues, Diamantopoulos et al. (2003) presents that most of reviewed literature suggested the positive linkage between education level and the awareness (or actual practice) for environmental issues.4 Regarding RPS, there are complex arguments about how measure the efficiency of RPS, and the effects of RPS on an expansion of renewable energy sources. Menz and Vachon (2006) and Adelaja et al. (2010) suggest that RPS has positive effects on an expansion of wind capacity, using cross-sectional data, and Sekar

This study focuses on the mutual relationships between GHGEE in the electricity sector (such as CO2) and ER in related sectors, and influences of electricity restructuring, recent socio-demographic changes, and other related factors on these relationships. To this end, a balanced panel dataset of 48 states was constructed from 2000 to 2012.5 Table 1 presents a description of the data. RSTRC is a dummy variable that indicates whether the state implements electricity restructuring.6 Because no state has enacted an additional deregulation since California's electricity crisis in 2000, RSTRC is time invariant for the period covered by our dataset. This variable is included in both equations (GHG-EE and employment). MPCO2 measures GHG-EE in this study. It is computed as total generation divided by CO2 emissions from the electricity sector. Computed this way, we interpret the improvement of GHG-EE intuitively. In other words, increases in total generation given the level of CO2 emissions or decreases in CO2 emissions given the total generation are interpreted as improvement of the efficiency or increases of MPCO2. Table 2 shows the summary statistics of explanatory variables in this study. According to the literature reviewed above, in the GHG-EE equation, RPS, OADR, BCHER, RENT, LIVALON, ELPR, NGEID, ELISP, GWSGEN, NCGEN, GSPPC, HDD, and CDD are explanatory variables. RPS is a dummy variable indicating whether the state implements RPS. RPS is a policy that imposes a mandatory combination of renewable energy sources at a certain rate when regulation targets use energy sources. Hence, RPS could be beneficial to enhancements of GHG-EE. However, it is controversial whether RPS, based on market principles, has practical effectiveness for enhancing GHG-EE or reducing the level of emissions. OADR, RENT, LIVALON, and BCHER are socio-demographic variables. OADR is the old age dependency ratio, computed as the share of the population over age 65 relative to the share of the population from 15 to 64 years old. OADR has been steadily increasing in the dataset. In addition, rental households (RENT), oneperson households (LIVALON) and educational attainment (BCHER) are increasing (Fig. 1). The variable BCHER is used as the effect of education or a proxy for knowledge of environmental issues. There are several possible explanations for the link between an aging society and electricity demand. Elderly people may spend more time in their home and live in energy-inefficient one

3 Beginning with Khazzoom–Brookes Postulate (Khazzoom, 1980; Brookes, 1990), many studies have dealt with ‘rebound effect’ in theoretical and empirical approaches. Most of them concludes that the rebound effect is empirically small (e.g. Greening et al., 2000), and the positive net rebound effect depends on the emission intensity of new energy sources and fuel elasticity of substitution (Saunders, 2000). 4 However, they proves the linkage partially in some parts of tests.

5 The time span is chosen by the data availability of socio-demographic variables. Also, Alaska, District of Columbia, and Hawaii are eliminated for a construction of a balanced panel dataset. 6 Markiewicz et al. (2004) use three alternative dummy variables of electricity restructuring for comparison: formal hearings of restructuring but failed legislation, passed legislation, and start of retail access. This study uses the second definition of electricity restructuring.

3. Data and methods

D. Kim, J. Jeong / Energy Policy 92 (2016) 468–476

471

Table 1 Data description. Variable

Definition

Unit

Source

RSTRC RPS MPCO2

Dummy(1 ¼ restructuring, 0 ¼otherwise) Dummy(1 ¼ RPS, 0 ¼otherwise) MWh/ metric tons

EIA DSIRE EIA

OADR BCHER OWNER RENT LIVALON GSPPC ELPR DELPR NGEID ELISP PARTY GWSGEN NCGEN HDD CDD DYPC WAGE NETAX INCEXP

Electricity restructuring Renewable Portfolio Standard Emission efficiency ¼total generation/ CO2 emission from electricity sector Old age dependancy ratio¼ over 65 population/15–64 population Over bachelor degree Type of households ¼owner Type of households ¼rent Householder living alone GDP by State Per Capita Average electricity price by state Difference of electricity price ¼State's average electricity price – US average electricity price Natural gas price in the electric power sector Net Interstate flow of electricity State governor's political party Renewable energy generation (geothermal, wind and solar) Nuclear generation Heating degree days Cooling degree days Disposable personal income per capita Wage and salaries Tax on production and import less subsidies Insurance trust expenditure by state government

% % of Total population % of Total household % of Total household % of Total household Dollars Dollars Dollars Dollars per million Btu Million KWh Dummy(1 ¼ democrats, 0¼ otherwise) % of Total generation % of Total generation # of days # of days Dollars Thousand dollars Million dollars Thousands dollars

Census Census Census Census Census BEA EIA EIA EIA EIA NGA EIA EIA NOAA NOAA BEA BEA BEA Census

NFEMPL CNSTR INFO MANU MINING UTIL

Nonfarm employment Construction Information Manufacturing Mining Utilities

% % % % % %

BEA BEA BEA BEA BEA BEA

of of of of of of

total total total total total total

employment employment employment employment employment employment

Table 2 Summary statistics.

OADR BCHER RENT LIVALON ELPR DELPR NGEID ELISP GWSGEN NCGEN CDD HDD GSPPC NETAX INCEXP WAGE DYPC

Mean

Stdev.

Min

Max

0.192 0.264 0.318 0.273 8.326  0.263 5.795  253.3 0.016 0.280 1124.6 5126 42 844 18 165 6.326 120.15 31 963

0.025 0.048 0.042 0.020 2.687 2.434 1.938 24 095 0.033 0.154 782.4 1984 9070 21 993 2.994 140.23 5978

0.127 0.143 0.234 0.169 4.170  4.130 0.000  67 901 0.000 0.042 94 434 24 095 1194 1.420 6.835 19 499

0.281 0.393 0.475 0.323 18.07 8.250 11.81 90 210 0.248 0.808 3545 9674 82 010 129 870 18.375 900.51 51 087

since they are more reluctant to change and utilize new devices, so that an aging trend could lead to increased electricity consumption. However, the level of activities by elders is lower than that of younger people (Hamza and Gilroy, 2011; Willis et al., 2011), so that it could suggest relatively lower electricity consumption itself. In other words, the former case suggests a negative influence on aggregate GHG-EE, whereas the latter improves it. Higher retail price (ELPR) could induce fewer incentives for utilities to improve their GHG-EE. However, renewables or nuclear generation can be competitive on the supply side (Palmer, 1999) and it could lead to decreased electricity consumption on the demand side, which can offset the degradation as well. Higher generation costs from an increase in the input prices (such as NGEID) provide incentives to reform their generation systems toward better efficiency, whereas an increase of natural gas price

could lead to a switch of the input from gas to coal (Burtraw et al., 2000), which emits more GHG. In addition, high net interstate flow of electricity (ELISP) is beneficial to the aggregate GHG-EE within a state because states can avoid emitting CO2 in their own regions by trading electricity with other regions.7 GWSGEN and NCGEN generate less CO2 , which benefits GHG-EE, however, HDD and CDD, which lead to use of electrical devices for cooling and heating, are expected to negatively influence it. Finally, the effects of GSPPC could be ambiguous because it advances the efficiency while leading to higher electricity consumption. In the employment equation, NETAX, INCEXP, WAGE, and 7 However, this advantage will be offset by increases of GHG emissions from other states in the national perspective.

472

D. Kim, J. Jeong / Energy Policy 92 (2016) 468–476

Fig. 1. The graph of socio-demographic variables.

GSPPC are explanatory variables for controlling aggregate economic behaviors and fluctuations, and thus we can extract the effect of GHG-EE on the employment. Higher NETAX imposes burdens on agents to enlarge their employment expenditures, whereas, INCEXP lessens that burden. WAGE and GSPPC could have ambiguous impacts on employment by sector. In other words, economic growth and efficiency gains do not directly lead to an increase in the total number of jobs and differently influence ER in each specific industry. This study also identifies the determinants of electricity restructuring. It allows us to correct a sample selection bias by applying the inverse Mill's ratio (IMR) to each equation. Based on the reviewed literature, OADR, BCHER, OWNER, LIVALON, DYPC, DELPR, NGEID, and PARTY are considered as the determinants. We expect that higher OADR, which means the higher conservative tendency to the drastic change, would result in a negative effect on the possibility of restructuring. In addition, BCHER, OWNER, and LIVALON could be favorable to the derived results of restructuring, such as a reduction of retail prices or the efficiency improvements. Since the discussion on the necessity of deregulation is raised as an economy grows, we expect a positive effect of disposable income (DYPC) on restructuring. A positive value of DELPR implies that a state's electricity price is higher than the national average price. Thus, this state is more likely to deregulate the market (Ardoin and Grady, 2006). From the perspective of suppliers, higher input costs require suppliers to increase their retail prices to retain profitability. However, in a competitive environment, the transfer of costs to customers is not sustainable. Hence, NGEID negatively influences the possibility of restructuring.8 Additionally, we account for the effects of the political affiliation of the state 8 In the beginning of deregulation, the state governments allowed the implementation for stranded cost recovery. However, it was temporary alternatives in the short-term, and thus it will not affect the decision-making from utilities even though they are aware of these schemes in advance.

governor (PARTY) as a dummy variable indicator, coded 1¼ Democratic Party affiliation and 0 ¼otherwise.9 All socio-demographic variables are obtained from the US Census Bureau. The US Energy Information Administration provides the electricity-related data. The US Bureau of Economic Analysis provides information on the economy, and climatic data are obtained from the National Oceanic and Atmospheric Administration. Last, information on the affiliations of state governors and RPS are obtained from the National Governors Association and the Database of State Incentives for Renewable and Efficiency, respectively. To achieve the goals of this study, we construct a simultaneous equation system regarding restructuring as follows:

rit⁎ = zit β1 + ηi + uit ,

i = 1, …, N , t = 1, …, T;

ri = 1[rit⁎ > 0]

(1)

yit = riϕ2 + hit δ + xit β2 + ai + ft + Tr + ϵit

(2)

hit = riϕ3 + yit θ + witβ3 + bi + ft + Tr + eit

(3)

1[·] is the indicator function where ri, or RSTRC, equals one if its latent variable rit⁎ , which is determined by the vector of explanatory variables, satisfies the argument, and it is zero otherwise. yit represents emissions efficiency, MPCO2 and hit is a proportion of each related industry's employment.10 As we postulate, hit represents, respectively, non-farm (NFEMPL), construction (CNSTR), information (INFO), manufacturing (MANU), mining 9 As the main interest of this study is GHG-EE, we set 1 for Democrats, which have more favorable to environmental issues. 10 Since Davis and Haltiwanger (1990, 1991), ER has been defined as the sum of the absolute value of job creation and destruction, which are weighted by the size of each specific sector, using firm-level data. In this study, however, because we use aggregate state-level data, we simply use the share of each electricity-related industry's employment.

D. Kim, J. Jeong / Energy Policy 92 (2016) 468–476

(MINING) and utilities (UTIL) employment in each estimation model. zit , x it and wit are vectors of explanatory variables that can contain common elements. In this study, as stated above, zit consists of OADR, BHCER, OWNER, LIVALON, DYPC, DELPR, NGEID and PARTY. xit comprises RPS, OADR, BCHER, RENT, LIVALON, ELPR, NGEID, ELISP, GWSGEN, NCGEN, CDD, HDD and GSPPC. In addition, wit comprises GSPPC, NETAX, INCEXP and WAGE. β1, β2, β3, δ, θ , ϕ2 and ϕ3 are unknown parameters or vectors of parameters. ηi , ai and bi are unobserved individual specific effects. In addition, ft and Tr are time-specific dummy and time-trend variables, respectively. We assume that all explanatory variables in each equation are strictly exogenous with their idiosyncratic error, uit, ϵit and eit, respectively. Moreover, all explanatory variables, individual specific effects and error terms in Eqs. (1)–(3) are exogenous with other idiosyncratic errors in Eqs. (2) and (3) and we assume that individual specific effect, ηi, in Eq. (1) is exogenous with its idiosyncratic error, uit:

E(uit |zit − s , ηi ) = 0

t = 1, …, T ,

Table 3 Result of RE probit estimation. RSTRC OADR BCHER OWNER LIVALON DYPC DELPR NGEID PARTY CONS

s = 0, …, t

E(ϵit |zit , ηi , uit , xit − s , wit − s, bi , ft − s , Tr , eit ) = 0 E(eit |zit , ηi , uit , xit − s , ai , wit − s, ft − s , Tr , ϵit ) = 0 Three main problems exist for estimation to achieve a consistent estimator. First, by construction, there are endogeneity problems caused by hit in Eq. (2) and by yit in Eq. (3). Second, Maddala and Lee (1976), Heckman (1976, 1978, 1979), and Barnow et al. (1980) argue that sample selection bias should be corrected when we consider a switching rule in dummy variables and Berg (1991) serve as an example. Hence, the IMR of Eq. (1) after random effects (RE) probit estimations are employed to correct the bias in the above setting. To solve these problems and obtain consistent estimates of the effects of all of the variables at the same time, we first apply the approach taken by Lee et al. (1980). This method solves the endogeneity problem of sample selection bias by inserting IMR into each equation.11 The structural form of Eqs. (2) and (3) are transformed into their reduced forms, which contain only exogenous variables with idiosyncratic error, at which point a fitted value of each dependent variable in the equations, MPCO2HAT and JOBHAT, can be obtained. By substituting MPCO2HAT for yit in Eq. (3) and JOBHAT for hit in Eq. (2), we obtain consistent estimators. However, we cannot calculate a coefficient for the time-invariant variables, such as RSTRC, if we apply within or first difference transformations to solve the incidental parameter problem (IPP) in the analysis of panel data, such as fixed effects (FE) estimations, because time-invariant variables are eliminated in the process of transformation. We deal with this problem using two alternatives to calculate the coefficients of the time-invariant variables. First, we assume that all explanatory variables are exogenous with unobserved heterogeneity and then apply RE estimation. Under the additional assumption, unobserved individual specific effects do not bring about the IPP, so we can estimate the coefficients without transformation. Second, we apply Hausman and Taylor (1981) estimation with loosened additional assumptions. With this approach, we calculate the coefficients of the time-invariant variables by using some of the explanatory variables, which are exogenous with unobserved individual specific effects, as instrumental variables and then apply FE estimation to solve IPP. The main difference between the two methods is whether we assume exogeneity of all of the explanatory variables or some of them with the unobserved effects to calculate the 11 Wooldridge (1995) also suggests using IMR to correct selectivity bias in panel data analysis.

473

 7.100nnn (3.623) 0.813 (2.268) 4.106nn (1.730) 15.549nnn (5.577) 0.963n (0.579) 4.075nnn (0.380)  0.493nn (0.207) 0.511nnn (0.135)  15.691nnn (5.766)

Notes: The values in parentheses are standard errors. n nn

p < 0.10 . p < 0.05. p < 0.01.

nnn

coefficients and apply transformation to avoid IPP.12 We compare the two results using the Hausman test to identify whether the additional assumption about the extent of exogeneity in the model can be justified.

4. Results and discussions Table 3 shows the results of the RE probit estimation of Eq. (1). All of the coefficients are consistent with our expectations, although BCHER is not statistically significant. OADR has a negative impact on the possibility of electricity restructuring, whereas BCHER, OWNER, and LIVALON have positive effects on it. These results show that the elderly are more reluctant to drastic changes, in this case, electricity restructuring. Also, people who are aware of and related to the benefits of the deregulation have positive attitudes to this policy. DYPC, which represents states' overall economic status, also influences positively. Similarly, DELPR positively influences electricity restructuring, whereas NGEID has a negative impact on the implementation of the restructuring. Last, we interpret that states' governors who are affiliated Democrats have more positive attitudes toward an implementation or a continuation of the deregulation than non-Democrats.13 Table 4 shows the feedback effect between GHG-EE and ER under the results of the Hausman Taylor (HT) estimation.14 Not only overall employment (except the farm sector), but also the information, manufacturing, and mining sectors have mutual relationships with GHG-EE. Also, we find positive effects of GHG-EE improvement on job share increases in the construction and 12 As a set of exogenous explanatory variables with individual effects, we assume that time, dummy variable, trend, CDD, and HDD are exogenous in the HT estimation. 13 The political economy issue this estimate may imply is not a major interest of this paper. Interested authors are referred to Lyon and Yin (2010), Ka and Teske (2002), Ardoin and Grady (2006), Campbell (1996), and Hlasny (2013), among others. 14 Because some of the data of CNSTR, INFO, MANU, UTIL, and MINING are missing due to confidentiality (although reflected in aggregate data), we use a cardinal spline method to interpolate these data. In each job classification, there are no more than one or two missing points in each of four states among 624 observations, which will not influence the main results. As there is too much missing data in MINING, Delaware and Maine are excluded from the estimation of MINING.

474

D. Kim, J. Jeong / Energy Policy 92 (2016) 468–476

Table 4 Feedback effect between GHG-EE and ER. MPCO2

NFEMPL

 38.472n (22.245)

JOBHAT

MPCO2

MPCO2

INFO

 16.795nn (6.587)

 0.509 (1.357) 0.003n (  0.002)

MPCO2HAT

CNSTR

0.031nnn (0.010)

MPCO2

MANU

 1.406n (0.754) 0.009nnn (0.003)

MPCO2

UTIL

MINING

 9.306n (5.331)

 31.746 (29.454) 0.063nnn (0.013)

MPCO2

0.001n (0.000)

 0.008n (0.005)

Notes: The values in parentheses are standard errors. n nn

p < 0.10. p < 0.05. p < 0.01.

nnn

Table 5 Results of HT estimation on GHG-EE. NFEMPL RSTRC

0.589 (0.454) RPS  0.081nn (0.032) OADR 4.595nnn (1.541) BCHER 3.808nnn (1.076) RENT  0.084 (1.172) LIVALON 1.284 (1.148) ELPR  0.164 (0.157) NGEID 0.002 (0.048) ELISP  0.000 (0.002) GWSGEN 1.408nnn (0.483) NCGEN 1.236nnn (0.330) GSPPC 0.409 (0.297) CDD  0.071n (0.038) HDD  0.032 (0.107) CONS 32.26n (19.26) IMR 0.054 (0.040) N 624 Chi2 281.88nnn

CNSTR

INFO

MANU

UTIL

MINING

 0.025  0.296  0.045n (0.025) 4.840nnn (1.533) 3.778nnn (1.109)  1.710nn (0.864) 1.208 (1.157)  0.016 (0.141)  0.035 (0.044)  0.002 (0.002) 0.834nn (0.341) 1.006nnn (0.310)  0.014 (0.163)  0.042 (0.035) 0.038 (0.104)  1.051 (2.055) 0.070n (0.039) 624 279.41nnn

0.116 (0.278)  0.065nn (0.026) 8.697nnn (2.209) 3.920nnn (1.086)  1.018 (0.826) 1.859 (1.185 0.011 (0.138) 0.005 (0.046)  0.001 (0.002) 0.958nnn (0.333) 0.889nnn (0.316) 0.233 (0.179)  0.055 (0.035) 0.030 (0.096)  4.169 (2.409) 0.648n (0.039) 624 284.69nnn

0.008 (0.272)  0.063nn (0.026) 5.852nnn (1.635) 3.922nnn (1.088)  1.494n (0.795) 1.588 (1.174) 0.021 (0.139)  0.021 (0.044)  0.001 (0.002) 1.006nnn (0.346) 0.924nnn (0.315) 0.114 (0.165)  0.055 (0.035) 0.030 (0.096)  2.526 (2.199) 0.075n (0.039) 624 281.19nnn

0.052 (0.289)  0.054nn (0.026) 5.488nnn (1.657) 3.536nnn (1.119)  1.373n (0.817) 1.066 (1.147)  0.024 (0.137)  0.037 (0.044)  0.002 (0.002) 0.885nnn (0.339) 1.015nnn (0.310)  0.011 (0.144)  0.043 (0.035) 0.034 (0.097)  1.059 (2.009) 0.071n (0.039) 624 279.02nnn

0.105 (0.309)  0.057nn (0.026) 2.537 (  1.895) 2.903nn (1.343)  1.242 (0.837) 1.833 (1.363)  0.089 (0.170)  0.060 (0.047)  0.003 (0.002) 1.044nnn (0.356) 1.070nnn (0.317) 0.095 (0.152)  0.028 (0.037)  0.006 (0.104)  1.485 (2.110) 0.094n (0.049) 598 250.16nnn

Notes: The values in parentheses are standard errors. The results of time dummy and trend are excluded due to the page constraint. n

p < 0.10. p < 0.05. nnn p < 0.01. nn

utilities sectors, although we cannot identify the feedback effect. GHG-EE in the electricity sector has a positive effect on increases to the job share of all of the electricity-related industries, except MINING.15 Specifically, a one percent improvement in

15 We acknowledge that the employment results we show are so intuitive that a simpler analysis than our regression model may lead us to the same conclusion. The time span of our data (2000–2012) is rather short to capture accurate macroeconomic effects, either. However, the regression analysis has merits, too. The simultaneous equation system enables us to identify the complex multi-directional effects in electricity industry. Furthermore, our model resolves the potential sample selection bias. We anticipate that the employment stimulus effects can be better analyzed in the future when the socio-demographic data are accumulated sufficiently.

GHG-EE increases the job share in manufacturing by 6.3 percent and increases the job share by 3.1 percent in the construction sector. The information and utilities industries increase as much as 0.9 percent and 0.1 percent, respectively, by the efficiency increase. Although the share of the mining sector is decreased by about 0.8 percent, overall employment (NFEMPL) is increased by about 0.3 percent with GHG-EE improvement. We speculate that GHGEE is enhanced by the installation of additional equipment, such as CCS or an IGCC power system, to existing power plants or the expansion of renewables or nuclear power plants. Hence, manufacturing and construction are sensitively influenced by GHG-EE improvement. Moreover, we propose that, because systematization of emissions data and efficient management of GHG emissions are needed for the efficiency enhancements, the shares of related industries' employment are enlarged. Conversely, with the gradual replacement of inputs for electricity generation from traditional energy sources, such as coal and natural gas, to renewable or nuclear energy sources, the share of the mining sector is reduced as GHG-EE increases. On the other hand, employment increases in related sectors have negative effects on GHG-EE because these industries require more electricity as their economic activities increase. In the GHG-EE equation of the each industry (Table 5), we find that RPS has a negative effect on GHG-EE. Since RPS is based on market principles to reduce the costs of emissions, the practical effectiveness of RPS to reduce the absolute volume of emissions or to improve GHG-EE is controversial.16 Among the socio-demographic variables, OADR has a positive effect on GHG-EE of about 4.595–8.697. A higher proportion of people educated beyond bachelor degrees has a positive effect on GHG-EE of about 2.903– 3.922. However, rental households have a negative effect on EE of about  1.710 to  1.373. Among the explanatory variables, we find no statistically significant effects of ELPR, NGEID, ELISP, GSPPC, and HDD on GHG-EE. Only CDD has a negative influence on GHG-EE because most of the cooling devices are highly dependent on electricity compared to heating devices. Additionally, GWSGEN and NCGEN have almost unit elasticity and positive effects on GHG-EE of about 0.834–1.408 and about 0.889–1.236, respectively. Electricity restructuring has a statistically insignificant effect on GHG-EE improvement. It implies that electricity restructuring may not lead to direct improvement of GHG-EE, even though it can influence other related factors, such as electricity price or expansion of nuclear and renewable energy power plants. Also, most of the literature point out that rebound effects from better operational efficiency after restructuring are empirically small. Thus, the results about the effect of the deregulation on GHG-EE in electricity sector could be intuitive. 16 Thus, this finding needs to be interpreted with caution. When we use alternative definitions of RPS variable, such as the requirement schedules on legislation, the statistical significance is unstable. As an anonymous referee points out, California is so dominant in the sample, and the estimation result is sensitive to how we define RPS. We are grateful to the referee for the suggestion.

D. Kim, J. Jeong / Energy Policy 92 (2016) 468–476

Table 6 Results of HT estimation on ER. NFEMPL 0.011nnn ( 0.004) GSPPC 0.008nnn (0.002) NETAX  0.005nnn (0.002) INCEXP  0.001 (0.001) WAGE 0.013nnn (0.003) CONS 0.690nnn (0.030) IMR 0.000 (0.000) N 624 Chi2 814.28nnn RSTRC

CNSTR

INFO

MANU

UTIL

MINING

 0.040nn (0.019) 0.003 (0.013)  0.024nn (0.011) 0.007 (0.005) 0.117nnn (0.016)  1.824nnn (0.187)  0.006nnn (0.002) 624 431.59nnn

0.004 (0.011) 0.000 (0.004)  0.014nnn (0.003)  0.003nn (0.001) 0.027nnn (0.005)  0.326nnn (0.055)  0.001nn (0.001) 624 414.03nnn

 0.057n (0.035)  0.010 (0.017)  0.110nnn (0.014)  0.007 (0.006) 0.272nnn (0.022)  3.605nnn (0.252)  0.005nn (0.003) 624 890.35nnn

0.001 (0.001)  0.002nnn (0.001) 0.000 (0.000)  0.000n (0.000) 0.003nnn (0.001)  0.042nnn (0.008) 0.000 (0.000) 624 133.82nnn

 0.011 (0.013) 0.001 (0.006) 0.004 (0.006)  0.003 (0.002) 0.022nnn (0.008)  0.427nnn (0.090) 0.002nn (0.001) 598 144.98nnn

Notes: The values in parentheses are standard errors. The results of time dummy and trend are excluded due to the page constraint. 0.000 represents a number below 0.000, not an exact zero. n

p < 0.10. p < 0.05. nnn p < 0.01. nn

Table 7 Results of Hausman test. MPCO2 Chi2 P-value

12.20 (0.994)

NFEMPL 7.37 (0.987)

CNSTR 22.43 (0.214)

INFO n

26.05 (0.099)

UTIL

MINING

14.25 (0.713)

25.68 (0.107)

Notes: The result of MANU is not available due to failure of the asymptotic assumptions. n

p < 0.10.

In the employment equation (Table 6), most of the employment proportions are positively influenced by WAGE, from about 0.3 percent to 27.2 percent. A one percent increase of NETAX has a negative effect on most of the related industries' employment, from about 0.5 percent to  11 percent. GSPPC has a positive effect on overall ER, although not in the utilities sector. This result suggests that reliance on automated processes in electrical power systems may reduce employment in the utilities sector. Unlike the case of GHG-EE, electricity restructuring has a positive effect on overall employment of about 1.1 percent, despite decreases in construction and manufacturing employment of about 4 percent and 6 percent, respectively. Last, IMR plays a part in correcting sample selection bias in most of the results. For comparison, we conduct RE estimations and find that most of the relationships and significance levels are similar to the results of the HT estimation. Specifically, the results of Hausman test in Table 7 show whether the differences between the two results are statistically significant. Most of coefficients from the two alternative estimations are statistically similar, except for INFO and MINING marginally, at p < 0.10 level of significance. In other words, fundamental assumptions of estimation are needed to carefully analyze and the results should be conservatively interpreted using HT estimations.

475

but also in the related industries, construction, information, manufacturing, utilities, and mining. To assess the relationships, we construct a simultaneous equation system with a consideration of a switching rule for electricity restructuring, and socio-demographic factors, economic and electricity-related variables, and climatic conditions are dealt with. To achieve consistent coefficients, we apply HT estimation and RE probit estimation following Lee et al. (1980) approach. It allows us to deal with the endogeneity problem and sample selection bias at the same time, and also obtain the coefficient of the time-invariant dummy variable. The results of estimation show feedback relationships between GHG-EE in the electricity sector and employment in the information, manufacturing, and mining sectors, as well as overall ER. Specifically, GHG-EE improvement in the electricity sector increases the shares of employment in the information and manufacturing sectors, whereas it decreases the share of the mining sector. The improvement also leads to the increase in the construction and utilities sectors, although there was no feedback relationship. However, employment in all these industries has a negative effect on the GHG-EE. With respect to the socio-demographic factors, aging and education positively influence the GHGEE, whereas rental households have a negative effect on the efficiency. Furthermore, we identify a positive effect of the renewables and nuclear energy generation, and a negative effect of CDD (representing the climatic condition) on the efficiency. The efficiency of an additional environmental policy, RPS, cannot derive a definite answer for GHG-EE in electricity sector. Regarding employment, there are positive effects of WAGE and negative effects of NETAX, which interpretations are intuitive. In addition, even though an increase in economic status (GSPPC) positively affects overall employment in the period of dataset, the specific effect is ambiguous by industry. Of course, the estimation results are on the partial effect of each variable, not the overall effect. The net emissions effect would depend on the complex interaction of the involved variables. Based on these results, policy makers should consider this feedback effects between GHG-EE in electricity sector and employment allocation. Although policies that target GHG emissions abatement could be justified since they could create related jobs, economic activities from these derived jobs could offset the effectiveness of GHG-EE enhancement. This study only focuses on the GHG emissions in electricity sector. However, technical progress could abate GHG emissions not only from electricity but also from other energy sources in each industry, for instance, emissions-intensive manufacturing sector. Thus, future papers would deal with the dynamics of GHG emissions from overall energy use and specific job creation and destruction issues, based on relatedtechnology development, in the long-term perspective. In addition, providing incentives to improve efficiency of electricity use for rental households could be essential to abate GHG emissions from the electricity sector. Last, we empirically identify the positive effect of electricity restructuring on overall employment. Since electricity restructuring is expected to gradually increase, although it has recently halted, this understanding of the mutual relationships between electricity restructuring, GHG emissions, and employment is beneficial under the trend of enhancing environmental regulation and feeble engines of economic growth.

5. Conclusions and policy implications Acknowledgments By considering the influence of the electricity sector on the economy, we postulate feedback relationships between GHG-EE in the electricity sector and ER not only in the overall employment

We are grateful to two anonymous referees for their valuable comments. This work was supported by the National Research

476

D. Kim, J. Jeong / Energy Policy 92 (2016) 468–476

Foundation of Korea Grant funded by the Korean Government (NRF-2013S1A3A2053586).

References Adelaja, A., Hailu, Y.G., McKeown, C.H., Tekle, A.T., 2010. Effects of renewable energy policies on wind industry development in the US. J. Nat. Resour. Policy Res. 2 (3), 245–262. Akorede, M.F., Hizam, H., Pouresmaeil, E., 2010. Distributed energy resources and benefits to the environment. Renew. Sustain. Energy Rev. 14 (2), 724–734. Ardoin, P.J., Grady, D., 2006. The politics of electricity restructuring across the American states: power failure and policy failure. State Local Gov. Rev. 38 (3), 165–175. Autor, David H., 2015. Why are there still so many jobs? The history and future of workplace automation. J. Econ. Perspect. 29 (3), 3–30. Barnow, B.S., Cain, G.G., Goldberger, A.S., 1980. Issues in the analysis of selectivity bias. In: Stromsdorfer, E., Farkas, G. (Eds.), Evaluation Studies Review Annual, vol. 5, p. 4359. Berg, S.V., Jeong, Jinook, 1991. An evaluation of incentive regulation for electric utilities. J. Regul. Econ. 3, 45–55. Blank, R.M., 2016. What drives American competitiveness? Ann. Am. Acad. Polit. Social Sci. 663 (1), 8–30. Bollino, C.A., 2009. The willingness to pay for renewable energy sources: the case of Italy with socio-demographic determinants. Energy J. 30 (2), 81–96. Brookes, L., 1990. The greenhouse effect: the fallacies in the energy efficiency solution. Energy Policy 18 (2), 199–201. Brounen, Dirk, Kok, Nils, Quigley, John M., 2012. Residential energy use and conservation: economics and demographics. Eur. Econ. Rev. 56 (5), 931–945. Burtraw, Dallas, Palmer, Karen, Heintzelman, Martin, 2000. Electricity Restructuring: Consequences and Opportunities for the Environment. Discussion Paper Resource for the Future: 00–39. Campbell, H.E., 1996. The politics of requesting: strategic behavior and public utility regulation. J. Policy Anal. Manag. 15 (3), 395–423. Cutler, D.M., Poterba, J.M., Sheiner, L.M., Summers, L.H., Akerlof, G.A., 1990. An aging society: opportunity or challenge?. Brook. Pap. Econ. Act., 1–73. Dalton, M., O'Neill, B., Prskawetz, A., Jiang, L., Pitkin, J., 2008. Population aging and future carbon emissions in the United States. Energy Econ. 30 (2), 642–675. Davis, S.J., Haltiwanger, J., 1990. Gross job creation and destruction: microeconomic evidence and macroeconomic implications. NBER Macroecon. Ann. 5, 123–168. Davis, S.J., Haltiwanger, J., 1991. Gross Job Creation, Gross Job Destruction and Employment Reallocation. No. w3728. NBER. Diamantopoulos, A., Schlegelmilch, B.B., Sinkovics, R.R., Bohlen, G.M., 2003. Can socio-demographics still play a role in profiling green consumers? A review of the evidence and an empirical investigation. J. Bus. Res. 56 (6), 465–480. Druckman, A., Jackson, T., 2008. Household energy consumption in the UK: a highly geographically and socio-economically disaggregated model. Energy Policy 36 (8), 3177–3192. Fagan, M.L., 2006. Measuring and explaining electricity price changes in restructured states. Electr. J. 19 (5), 35–42. Gough, I., Abdallah, S., Johnson, V., Ryan-Collins, J., Smith, C., 2011. The Distribution of Total Greenhouse Gas Emissions by Households in the UK, and Some Implications for Social Policy. LSE STICERD Research Paper No. CASE152. Greening, L.A., Greene, D.L., Difiglio, C., 2000. Energy efficiency and consumption the rebound effect a survey. Energy Policy 28 (6), 389–401. Gruber, J., Wise, D., 2001. An International Perspective on Policies for an Aging Society. No. w8103. NBER. Gungor, V.C., Sahin, D., Kocak, T., Ergüt, S., Buccella, C., Cecati, C., Hancke, G.P., 2011. Smart grid technologies: communication technologies and standards. IEEE Trans. Ind. Inform. 7 (4), 529–539. Hamza, N., Gilroy, R., 2011. The challenge to UK energy policy: an ageing population perspective on energy saving measures and consumption. Energy Policy 39 (2), 782–789. Hausman, J.A., Taylor, W.E., 1981. Panel data and unobservable individual effects. Econometrica 49 (6), 1377–1398. Heckman, J.J., 1976. The common structure of statistical models of truncation, sample selection and limited dependent variables and a simple estimator for such models. Ann. Econ. Soc. Meas. 5 (4), 475–492. Heckman, J.J., 1978. Dummy endogenous variables in a simultaneous equation

system. Econometrica 46 (4), 931–959. Heckman, J.J., 1979. Sample selection bias as a specification error. Econometrica 47 (1), 153–161. Hlasny, V., 2013. Commissioner Selection and Deregulation of Gas Utilities. SSRN 2029542. Hu, X., Moura, S.J., Murgovski, N., Egardt, B., Cao, D., 2015. Integrated Optimization of Battery Sizing, Charging, and Power Management in Plug-in Hybrid Electric Vehicles. http://dx.doi.org/10.1109/TCST.2015.2476799. Joskow, P.L., 1997. Restructuring, competition and regulatory reform in the US electricity sector. J. Econ. Perspect. 11 (3), 119–138. Ka, S., Teske, P., 2002. Ideology and professionalism electricity regulation and deregulation over time in the American states. Am. Polit. Res. 30 (3), 323–343. Khazzoom, J.D., 1980. Economic implications of mandated efficiency in standards for household appliances. Energy J. 1 (4), 21–40. Kronenberg, Tobias, 2009. The impact of demographic change on energy use and greenhouse gas emission in Germany. Ecol. Econ. 68 (10), 2637–2645. Lee, Lung-Fei, Maddala, G.S., Trost, R.P., 1980. Asymptotic covariance matrices of two-stage probit and two-stage tobit methods for simultaneous equations models with selectivity. Econometrica 28 (2), 491–503. Lyon, T.P., Yin, H., 2010. Why do states adopt Renewable Portfolio Standards?: an empirical investigation Energy J. 31 (3), 133–157. Maddala, G.S., Lee, L., 1976. Recursive models with qualitative endogenous variables. Annu. Econ. Soc. Meas. 5 (4), 525–545. Mansur, Erint, 2007. Do oligopolists pollute less? Evidence from a restructured electricity market. J. Ind. Econ. 55 (4), 661–689. Markiewicz, Kira, Rose, Nancy, Wolfram, Catherine, 2004. Has restructuring improved operating efficiency at US electricity generating plants? CSEM WP 135. Menz, F.C., Vachon, S., 2006. The effectiveness of different policy regimes for promoting wind power: experiences from the states. Energy Policy 34 (14), 1786–1796. Mortensen, Dale T., Pissarides, Christopher A., 1998. Technological progress, job creation, and job destruction. Rev. Econ. Dyn. 1 (4), 733–753. O'Neill, B.C., Dalton, M., Fuchs, R., Jiang, L., Pachauri, S., Zigova, K., 2010. Global demographic trends and future carbon emissions. Proc. Natl. Acad. Sci. 107 (41), 17521–17526. Palmer, Karen, 1999. Electricity restructuring: shortcut or Detour on the road to achieving Greenhouse gas reduction?. Clim. Issues Brief 18. Rifkin, Jeremy, 2014. The Zero Marginal Cost Society: The Internet of Things, The Collaborative Commons, and the Eclipse of Capitalism. Palgrave MacMillan, New York. Roland-Holst, D., 2008. Energy efficiency, innovation, and job creation in California. Department of Agricultural and Resource Economics, UCB. Saunders, H.D., 2000. A view from the macro side: rebound, backfire, and Khazzoom–Brookes. Energy Policy 28 (6), 439–449. Sekar, S., Sohngen, B., 2014. The Effects of Renewable Portfolio Standards on Carbon Intensity in the United States. Resources for the Future Discussion Paper, (14–10). Sims, Ralph E.H., Rogner, Hans-Holger, Gregory, Ken, 2003. Carbon emission and mitigation cost comparisons between fossil fuel, nuclear and renewable energy resources for electricity generation. Energy Policy 31 (13), 1315–1326. Tonn, Bruce, Eisenberg, Joel, 2007. The aging US population and residential energy demand. Energy Policy 35 (1), 743–745. United Nations, Department of Economic and Social Affairs (UNDESA), Population division, 2015. World Population Prospects: The 2015 revision. U.S. Energy Information Administration (EIA), 2015. Monthly Energy Review May 2015. U.S. Department of Energy. Washington, DC 20585. May. Vassileva, I., Wallin, F., Dahlquist, E., 2012. Analytical comparison between electricity consumption and behavioral characteristics of Swedish households in rented apartments. Appl. Energy 90 (1), 182–188. Willis, K., Scarpa, R., Gilroy, R., Hamza, N., 2011. Renewable energy adoption in an ageing population: heterogeneity in preferences for micro-generation technology adoption. Energy Policy 39 (10), 6021–6029. Wooldridge, J.M., 1995. Selection corrections for panel data models under conditional mean independence assumptions. J. Econom. 68 (1), 115–132. Yin, H., Powers, N., 2010. Do state renewable portfolio standards promote in-state renewable generation. Energy Policy 38 (2), 1140–1149. York, Richard, 2007. Demographic trends and energy consumption in European Union Nations, 1960–2025. Social Sci. Res. 36 (3), 855–872. Zarnikau, J., 2003. Consumer demand for ‘green power’ and energy efficiency. Energy Policy 31 (15), 1661–1672.