Different effects of main influence factors on household energy consumption in three typical rural villages of China

Yang, Ruiliang et al. Article Different effects of main influence factors on household energy consumption in three typical rural villages of China Energy Reports Provided in Cooperation with: Elsevier Suggested Citation: Yang, Ruiliang et al. (2018) : Different effects of main influence factors on household energy consumption in three typical rural villages of China, Energy Reports, ISSN 2352-4847, Elsevier, Amsterdam, Vol. 4, pp. 603-618, https://doi.org/10.1016/j.egyr.2018.09.006 This Version is available at: https://hdl.handle.net/10419/243543 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ Energy Reports 4 (2018) 603–618 Contents lists available at ScienceDirect Energy Reports journal homepage: www.elsevier.com/locate/egyr Research paper Different effects of main influence factors on household energy consumption in three typical rural villages of China Ruiliang Yang ∗, Jiangmin He, Sha Li, Wen Su, Yue Ren, Xinyu Li Key Laboratory of Modern Electromechanical Equipment Technology, Tianjin Polytechnic University, Tianjin 300387, China article info Article history: Received 21 May 2018 Received in revised form 14 September 2018 Accepted 20 September 2018 Available online 11 October 2018 Keywords: Energy consumption Influence factor Ordinary least squares (OLS) Quantile regression (QR) abstract Three typical rural villages are chosen as study areas to study the different effects of main influence factors on rural household energy consumption using the ordinary least squares (OLS) model and the quantile regression (QR) model. Three typical rural villages include an affluent rural village in the richest province, a well-off rural village in the most energy-rich province and an out-of-poverty rural village in the most population province. The OLS results show family size and household income are significant at the 10 % level in all study areas, while air conditioner and refrigerator are only significant at the 5 % level of the affluent rural village. Family size has a positive effect on energy consumption, while household income has different effect in different areas. The QR results fluctuate around QLS results with large amplitude, indicating a different pattern of the effect of main independent variables on energy consumption using different models. The slopes are significantly different for all study areas at 10% significant level, which shows the distribution heterogeneity in studying the relationships between energy consumption and influence factors. This study is useful to develop more appropriate energy policy according to the main influence factors in different rural villages. ©2018 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction With the rapid urbanization in China, the rural population was declining in recent years, whereas rural household energy consumption was rising sharply. Although the rural population declined from 745.4 million in 2005 to 603.5 million in 2015 (NBS, 2017), energy consumption of rural residents increased from 115.5 million tons coal equivalent (tce) in 2005 to 211.8 million tce in 2015 (NBS,2016), with an annual growth of 6.6%. The average annual household energy consumption of rural residents increased from 155 kg coal equivalent (kgce) in 2005 to 351 kgce in 2015, approximately 226.5%. If China can maintain the average annual urbanization rate of 1% and reach an urbanization rate of approximately 70% by the end of 2030, there will still be more than 400 million rural population, which means enormous rural household energy consumption due to rising average annual household energy consumption of rural residents (Yu and Guo,2016). To reduce the enormous rural household energy consumption by the large rural population, it is necessary to further study the effect of main influence factors on rural household energy consumption. Some researchers analyzed the effect of main influence factors on household energy consumption at the macrolevel, with the data came from the government statistics such as ∗Corresponding author. E-mail address: [email protected] (R. Yang). China Statistical Yearbook. The stochastic impacts by regression on population, affluence and technology (STIRPAT) model is the most widely applied at the macro-level to research the relationship between household energy consumption and its influence factors (Ding et al.,2016). GDP per capita, average temperature, and population are the popular influence factors used in this model (Ding et al.,2016;Miao,2017; Li, 2015). Logarithmic mean Divisia index (LMDI) technique is also frequently applied to the field of energy consumption (Nie and Kemp,2014;Chung et al.,2011;Zhang and Guo,2013), which can analyze the influence factors on energy consumption at the macro-level. Population, floor space, energy mix, appliance, and climate effect may be chosen as influence factors in the LMDI decomposition analysis. Residential energy model global (REMG) model (Daioglou et al.,2012), bottom-up model for residential energy use (van Ruijven et al.,2011), piecewise linear model (Liu et al.,2016), panel smooth transition regression (PSTR) model (Lee and Chiu,2011), consumer lifestyle approach (CLA) (Wei et al.,2007), stochastic impacts by regression on population, affluence, and technology (STIRPAT) model (York,2007), quantile regression (QR) model (Kaza, 2007), local area resource analysis (LARA) model (Druckman and Jackson, 2007) are also applied to research the influence factors on household energy consumption. However, due to the lack of detailed statistics on rural household energy consumption, researching the main influence factors on household energy consumption at the macro-level focuses on a province-level or city-level, and the conclusions are applied to https://doi.org/10.1016/j.egyr.2018.09.006 2352-4847/©2018 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 604 R. Yang et al. / Energy Reports 4 (2018) 603–618 Table 1 The summary of factors influencing the rural residential energy consumption in recent studies. Authors Sample Data period Data source Rural/urban Influence factors Methods How do the factors act on the household energy consumption/ electricity consumption? Macrolevel Ding et al. (2016) 30 provinces in China 1997– 2013 China Energy Statistical Yearbook (1998–2014), China Statistical Yearbook (1998–2014), National Meteorological Information Center Rural and urban GDP per capita, urbanization level, annual average temperature STIRPAT model Significant positive impact (GDP per capita), significant negative impact (annual average temperature), uncertainty (urbanization level) Liu et al. (2016) 540 observations in 30 Chinese provinces 1995– 2012 China Energy Statistical Yearbook (1996–2013), China Statistical Yearbook (1996–2013) Rural and urban Income Piecewise linear model Significant impact on residential electricity consumption (income) Li and Lin (2015) A balanced panel data set of 73 countries 1971– 2010 World Bank, UN Rural and urban Urbanization, industrialization STIRPAT model Significant impact (urbanization, industrialization) Nie and Kemp (2014) China 2002– 2010 China statistical yearbook 2011, China energy statistical yearbook (2003–2011) Rural and urban Population, floor space, energy mix, energy demand from appliances effect LMDI technique The most important factor (energy demand from appliances effect), the second most important factor (floor space), the third most important factor (population), negligible factor (energy mix) Zhang and Guo (2013) China 1991 to 2010 China Statistical Yearbook, China Energy Statistical Yearbook Rural Population, energy intensity, per capita net income, energy mix LMDI technique The critical factor (per capita net income), dominant role (energy intensity), followed effect (population), very minor role (energy mix) Daioglou et al. (2012) India, China, South East Asia, South Africa and Brazil 2007 World Development Indicators of the World Bank, World Bank, World Health Organization Rural and urban Population, household expenditure, population density, household size, temperature REMG model Primary drivers of energy use (population, household expenditure, population density, household size, temperature) van Ruijven et al. (2011) India 1971– 2003 OECD Environmental Outlook scenario, National Sample Survey Organization (NSSO) of the Ministry of Statistics Rural and urban Population, household size, income and temperature Bottom-up model Primary drivers of energy use (population, household size, income and temperature) Lee and Chiu (2011) 24 OECD countries 1978– 2004 The World Development Indicators Rural and urban Real income, electricity price, and temperature PSTR model Significant impact (real income), negative or U-shaped relationship (temperature), no influence (electricity price) (continued on next page) the average situation rather than the actual rural village. Because rural resident is defined ‘‘the resident living outside of urban area’’ (NBS,2017), many non-farmer population are often classified as rural population in Chinese official statistics, which leads to great R. Yang et al. / Energy Reports 4 (2018) 603–618 605 Table 1 (continued). Authors Sample Data period Data source Rural/urban Influence factors Methods How do the factors act on the household energy consumption/ electricity consumption? Kaza (2010) 4382 observations in USA 1987– 2005. Residential Energy Consumption Survey Data from the Energy Information Administration Rural and urban Heating degree days, heating area, cooling degree days, cooling area, housing size, household size, age of house, neighborhood density, income, price, ownership status, housing type QR approach Significant impact (heating degree days, heating area, cooling degree days, cooling area, housing size, household size, age of house, income, price) Pachauri and Jiang (2008) India, China 1999/ 2000 Various Indian official and ministerial publications, National Bureau of Statistics (NBS). Rural and urban Income, urbanization, energy access, energy price, geographic variation, climatic factors, local resource, availability, customs and tastes demographic factors Comparison by figures or tables. The most important factors (income, urbanization, energy access, energy price) York (2007) European Union Nations 1960– 2000 World Bank Rural and urban Population size, age structure, economic development, urbanization Elasticity STIRPAT model Clear effect (population size, age structure), substantial contribution (economic development, urbanization) Wei et al. (2007) China 1999– 2002 China Energy Statistical Yearbook, China Statistical Yearbook, China Industrial Statistical Yearbook, China Population Statistical Yearbook, China Rural Statistical Yearbook Rural and urban Home energy use; food, education, cultural and recreation services; personal travel; clothing, medicine and medical services; transport and communication services; household facilities and services; miscellaneous commodities and services CLA The most energy-intensive for urban residents (home energy use, food, education, cultural and recreation services), the most energy-intensive for rural residents (home energy use; food, education, cultural recreation services; personal travel) Druckman and Jackson (2008) 7000 sampled households in the UK. 2004– 2005 Expenditure and Food Survey (EFS), Rural and urban Income levels, type of dwelling, tenure, household composition and rural/ urban location, population LARA Strongly influenced (income levels), also extremely important (type of dwelling, tenure, household composition and rural/ urban location, population) (continued on next page) limitations of studying China’s rural energy-related issues using the average situation. Some researchers analyzed the effect of main influence factors on household energy consumption in China at the micro-level, 606 R. Yang et al. / Energy Reports 4 (2018) 603–618 Table 1 (continued). Authors Sample Data period Data source Rural/urban Influence factors Methods How do the factors act on the household energy consumption/ electricity consumption? Microlevel Ding et al. (2017) 187 valid questionnaires in Jiangsu province of China Uncertainty Questionnaire and interviews Rural and urban Energy-saving knowledge; environmental values; environmental responsibility; comfort preference; group psychology; low-carbon energy-saving willingness; habits and lifestyle; publicity, education and information; policies and regulations; economic cost, social technology Linear regression model The factor with the greatest influence for urban residents is low-carbon energy-saving willingness, the greatest influence on rural residents is publicity, education and information Yu and Guo (2016) 3404 households in 12 provincial units of China 2014 Questionnaires Rural Floor area, family member, No. of refrigerator, paid electricity bill monthly, household yearly income, dwelling age, No. of televisions, calculated electricity price, household head’s education, TV labeling tier one, refrigerator labeling tier one, regional heterogeneity Stochastic frontier model Significant effect on electricity consumption (floor area, family member, No. of refrigerator, paid electricity bill monthly, household yearly income) Not Significant impact on electricity consumption (dwelling age, No. of televisions, calculated electricity price, household head’s education, TV labeling tier one, refrigerator labeling tier one, regional heterogeneity) Niu et al. (2016) 1128 survey questionnaires in western China 2012 Questionnaires Rural and urban Price of electricity, per capita income, the price of electrical appliances, the diversity of appliances, household size OLS and quantile regressions Significantly effects on electricity consumption (per capita income, the price of electrical appliances, the diversity of appliances), little effects on electricity consumption (household size), no effects on electricity consumption (price of electricity) (continued on next page) with the data mainly from questionnaires. For example, based on 165 responses from the rural region of Zibo city in Shandong province, Liu et al. regarded age, education, household size, income, and ownership of air conditioners as main influence factors R. Yang et al. / Energy Reports 4 (2018) 603–618 607 Table 1 (continued). Authors Sample Data period Data source Rural/urban Influence factors Methods How do the factors act on the household energy consumption/ electricity consumption? Liu et al. (2013) 165 responses from Shandong province 2011 Questionnaire Rural Age, education, household size, income, ownership of air conditioners Linear regression model Positively correlated with the five explanatory variables at 10% significant level Ping et al. (2013) 478 households in eastern part of QinghaiTibet Plateau 2004 Participatory Rural Appraisal (PRA), Physical Monitoring (PM) Rural Altitude, household size, annual income, education level and livelihood Regression analysis Important influence factors on energy consumption per capita (altitude, household size and annual income), Leading factors of electricity consumption (altitude, annual income, education level) Niu et al. (2012) 719 sample data in the Western Loess Plateau of China 2009– 2010 Questionnaire Rural and urban Income level, energy attributes Crossquadratic model Main impact (the income level), litter impact (energy attributes) Wang et al. (2002) 384 households in 12 villages of 4 towns in Jiangsu Province, China 1997 Questionnaire Rural Annual per capita income, per capita straw yield, number of persons and pigs per household, attitude towards energy consumption Stratification sampling method Main factors (annual per capita income, per capita straw yield, number of persons and pigs per household, attitude towards energy consumption) on household energy consumption using the linear regression method (Liu et al.,2013). Based on 187 valid questionnaires from Jiangsu province, Ding et al. studied the differences in the factors influencing energy-saving behavior of urban and rural residents 608 R. Yang et al. / Energy Reports 4 (2018) 603–618 Fig. 1. The location of investigative areas. using multiple regression analysis (Ding et al.,2017). Based on data from 1128 survey questionnaires in western China, Niu et al. analyzed six factors on residential electricity consumption using the QR approach (Niu et al.,2016). The greatest advantage of the micro-level is completely closed to the actual rural village instead of the average situation. However, the rural villages studied above were mostly randomly selected areas, and the conclusions were quite different, see Table 1. Table 1 summarizes the factors influencing household energy consumption in recent studies. It can be seen from Table 1 that the main influence factors of rural household energy consumption (appear in one or more documents, and are generally considered to have a significant effect) are: household income (or GDP per capita or household expenditure), family size, household size, location (or annual average temperature, climatic factor), air conditioner, refrigerator, education level, price of electrical appliances, energy price, urbanization level, industrialization. The factors influencing household energy consumption in different documents were inconsistent, which means it is necessary to further study the factors in the typical rural area. For this propose, this paper analyzes the effect of main influence factors on rural household energy consumption in typical rural villages not only using the ordinary least squares (OLS) model, but also using the QR model, which can tease out the effects of various influence factors on the distribution on the energy consumption in different areas instead of focusing on the conditional average (Chung et al.,2011). This paper differs from existing articles in three major respects. First, study areas are three typical rural villages in three typical provinces. Due to China’s extensive territory, there is sharply different life styles among rural regions, which means there may be different influence factors in different rural regions. To distinguish real main influence factors on the energy consumption, the study areas must be typical areas. Three typical rural villages include an affluent rural village with a strongly developed agricultural economy in the richest province, a well-off rural village with a mixed economy of agriculture and local industry in the most energy-rich province, and an out-of-poverty rural village with an overwhelmingly agricultural economy in the most population province. Second, this paper not only analyzes the influence factors on energy consumption using the OLS and QR models, but also discusses the different effects of all independent variables and the significant Table 2 Brief profiles of the household characteristics. Item Tianping Village Dapu Village Shihuiyao Village All Effective number of households 397 96 170 663 Effective population 1821 396 778 2995 variables using these models. Third, the Kruskal–Wallis test is used in this paper to compare the differences of energy consumption and six influence factors in study areas. 2. Survey 2.1. Investigative areas This study was undertaken in three typical rural villages: Tianping Village, Dapu Village and Shihuiyao Village (see Fig. 1). The three selected villages are far apart to reduce the impact of cities on rural areas. Because the villages in the same province always have similar energy policy and energy consumption habit, one village is chosen from one province. Tianping Village is an affluent rural village with a strongly developed agricultural economy in northwest Tianjin, the richest province with the highest per capita gross domestic product in China (NBS,2017). Tianping Village has the distinctive features of nursery stock production, and almost all residents lived by stock cultivation and transportation. Dapu Village is a well-off rural village in central Shanxi province, China’s largest coal producing and storage province. Like most of China’s rural villages, Dapu Village has a mixed economy with agriculture and local industry. Almost all Dapu Village residents lived by farmland and coal industry. Shihuiyao Village is an out-of-poverty rural village in southern Henan province, the most population province. Shihuiyao Village has an overwhelmingly agricultural economy, and almost all Shihuiyao Village residents lived by farmland. To pursue a better life, many adults of Shihuiyao Village township went to coastal boomtown for work, the old men and children stayed at home as ‘‘empty nest’’. 2.2. Sample selection and data acquisition Survey on residential energy consumption in the rural village was undertaken by a survey team in January 2016. Half of total household was chosen in each village as samples. The questionnaires were applied by home visits with the help of the local government and the local junior high school. Questionnaire topics include: energy type, energy consumption, household information, house information and other energy related information. Incomplete or obviously false samples were abandoned. Table 2 shows brief profiles of household characteristics. 2.3. Energy type and influence factors Energies used by the residents in this survey include electricity, liquefied petroleum gas (LPG), coal, gasoline, diesel, biogas, solar energy, stalk, and fireweed. To allow convenient energy consumption comparison, all energies are indicated by their standard coal equivalent (China Nation Standardization Management Committee,1998). Fig. 2 shows the average annual energy consumption for study regions. The average annual energy consumption for a household in Tianping Village, Dapu Village, Shihuiyao Village are 2301.0, 2363.7, 1030.1 kgce, respectively. The main energies used in Tianping Village, Dapu Village, and Shihuiyao Village are coal and gasoline, coal and stalk, and stalk and diesel, respectively. R. Yang et al. / Energy Reports 4 (2018) 603–618 609 3. Methodology General statistical characteristics of energy consumption in study areas are shown in Table 3. The average annual energy consumption in Shihuiyao Village is obviously below than that in Tianping Village and Dapu Village. The skewness and kurtosis of energy consumption in study areas are larger than zero, meaning energy consumption in study areas is right advertence and relatively centralized. The kurtosis coefficients of energy consumption in investigative regions are all lower than 3, meaning less outliers (Joanes and Gill,1998). Fig. 3 shows the kernel densities of household energy consumption. 3.1. The main influence factors The main influence factors in this paper are chosen from analyzed main influence factors in Table 1, i.e., household income (or GDP per capita or household expenditure), family size, household size, air conditioner, refrigerator, education level, price of electrical appliances, energy price, urbanization level, industrialization, location (or annual average temperature, climatic factor). However, for study areas, the above factors may be modified according to the actual situation. First, due to universal nine-year compulsory education in China from 1986, the investigative homeowners were junior high school education or slightly lower than junior high school education, though some homeowners appeared to be illiterate for a long time without reading and writing. Thus, the education levels of the homeowners are basically the same, and are not considered in this study. Second, the electrical appliances and commercial energy were all purchased from the market, with prices fluctuating in different periods, which meant it is meaningless to research the effect of electrical appliances prices and commercial energy price on energy consumption using the electrical appliances and commercial energy from different sources. Thus, the price of electrical appliances and energy price are not considered in this study. Third, urbanization level and industrialization are very effective to discuss urban-related issues, but failed to study the rural problems. Thus, urbanization level and industrialization are not considered in this study. Fourth, three study areas are considered separately rather than together, so the location is not considered as the main influence factor in this study. In summary, the main influence factors in this study are household size, household income, family size, household age, air conditioner and refrigerator. Table 4 shows the brief profiles of six influence factors of residential energy consumption. The average house areas, family size, household age and the percentage of refrigerator in three investigative areas are close, whereas household income and the percentage of air conditioner in three investigative areas are significantly different. The average household income of Tianping Village residents, Dapu Village residents and Shihuiyao Village residents are 41.9 ±34.4, 27 ±20.2, 18.4 ±12.3 thousand Yuan, respectively. The average household income of the surveyed Tianping Village residents was significantly higher than that of the other two regions, while the average household income of the surveyed Shihuiyao Village residents is significantly lower than that of the other two regions. The percentages of air conditioner in Tianping Village, Dapu Village and Shihuiyao Village are 58.9%, 8.3% and 67.1%. The percentage of air conditioner in Dapu Village is significantly lower than that of the other two regions. 3.2. Methodology The OLS and QR models are used to estimate the effect of six influence factors (household income, family size, household size, air conditioner, refrigerator, and household age) on energy consumption. The linear model takes the following form: yi=xi′β+εi(1) where yiis the dependent variable, representing household energy consumption; xiis six chosen independent variables, including household area (HA), household income (HI), family size (FS), household age (HHA), air conditioner (AC), and refrigerator (RE). HA, HI, FS, HHA are normal variables, while AC and RE are dummy variables. AC =1/RE =1 means air conditioner/refrigerator is used by investigative resident, while AC =0/RE =0 means air conditioner/refrigerator is not used. εiis error term. βis unknown parameter, and can be estimated by the OLS model: min n ∑ i=1(yi−xi′β)2(2) βcan also be estimated for any quantile τby the QR model (Niu et al.,2016): min ⎡ ⎣∑ i∈{i:yi≥xiβ} τ⏐ ⏐yi−xi′β⏐ ⏐+∑ i∈{i:yi<xiβ} (1−τ)⏐ ⏐yi−x′ iβ⏐ ⏐⎤ ⎦(3) where 0 ≤τ≤1. 4. Results 4.1. The results of OLS and QR models The results of OLS and QR models using six independent variables are presented in Table 5. The OLS results indicate that the significant variables at 10% significant level in Tianping Village, Dapu Village, Shihuiyao Village and all investigative area are HI, FS, AC, RE; HI, FS; HI, FS; and HI, FS, respectively. FS is a significant variable, and has a positive effect on energy consumption. HI is also a significant variable, and has a positive effect on energy consumption of Tianping Village residents, but has a negative effect on energy consumption of Dapu Village residents and Shihuiyao Village residents. There were many migrant workers with more money but less time at home in Dapu Village and Shihuiyao Village, which made income and energy consumption different trend. For Tianping Village residents, owning an air conditioner or refrigerator can result in a significant increase of household energy consumption. Different to the OLS model, the QR model seeks to estimate the conditional quantile function models in which quantiles of the conditional distribution of the response variable are expressed as functions of observed covariates (Niu et al.,2016). The 5th, 10th, 20th ... 80th, 90th and 95th quantiles are estimated in Table 6, which helps to better understand the effects of the distribution of independent variables. Fig. 4 represents a summary of QR results for the six independent variables in investigative rural villages. In Fig. 4, the solid line shows the QR estimates for the quantiles ranging from 0.05 to 0.95, the shaded area denotes 90% confidence interval for the QR estimates, the dashed lines and the dotted lines denotes the coefficient and 90% confidence band of the OLS estimates, respectively. It can be seen from Fig. 4 and Table 5 that the QR results can change across the quantiles, and fluctuate around the QLS results with large amplitude, indicating a different pattern of the effect of six independent variables on energy consumption using two models. The effects of HI on energy consumption are significant positive at all quantiles for Tianping Village residents and all investigative residents, while are negative and statistically significant at most quantiles for Dapu Village residents and Shihuiyao Village residents. For Tianping Village residents, the positive effects of HI are low at bottom quantiles, and high at top quantiles. For Dapu Village 610 R. Yang et al. / Energy Reports 4 (2018) 603–618 Fig. 2. The average annual energy consumption in study regions. Table 3 General statistical characteristics of energy consumption in study areas. Sample Maximum (kgce) Minimum (kgce) Mean (kgce) 25th percentile (kgce) Median (kgce) 75th percentile (kgce) Skewness Kurtosis Tianping Village 397 4589.6 824.6 2301 1693.5 2104.8 2852 0.53822 2.6232 Dapu Village 96 3962.0 606. 8 2363.7 1823.6 2331.7 2722.8 0.11074 2.9132 Shihuiyao Village 170 3335.6 79.3 1030.1 393.0 763.7 1622.6 0.74486 2.8243 All 663 4589.6 79.3 1984.2 1306.1 1889.9 2599.2 0.22873 2.6229 Fig. 3. Kernel density estimate of energy consumption in study areas after eliminating outliers. residents, HI has a positive effect on energy consumption below 10th, then has more and more negative effects from 10th to 95th. The effects of HI on energy consumption for Dapu Village residents are statistically significant at 5th, 70th to 95th. For Shihuiyao R. Yang et al. / Energy Reports 4 (2018) 603–618 617 Table 8 Comparison of computed coefficient for Tianping Village residents by QR model using the six independent variables and the significant variables. Variables HI FS AC RE QR model using the six independent variables QR model using the significant variables Difference percentage (%) QR model using the six independent variables QR model using the significant variables Difference percentage (%) QR model using the six independent variables QR model using the significant variables Difference percentage (%) QR model using the six independent variables QR model using the significant variables Difference percentage (%) OLS 5.5327**** 6.0244**** 8.89 132.9606****137.6406**** 3.52 218.4098** 229.2016*** 4.94 207.3721** 206.707** 0.32 (1.2346) (1.2173) (34.2246) (34.2496) (85.2301) (85.2300) (84.7642) (85.1041) Quantiles 0.05 4.9134*** 5.1149*** 4.10 97.7839** 104.8899** 7.27 188.7508*185.5213 −1.71 148.7397 169.4997 13.96 (1.5562) (1.6529) (43.1412) (46.5054) (107.4353) (115.7284) (106.8480) (115.5576) 0.10 4.6403*** 4.916**** 5.94 161.8098****152.74**** −5.61 261.0276*** 282.84*** 8.36 116.3973 113.16 −2.78 (1.4549) (1.4819) (40.3336) (41.6951) (100.4435) (103.7581) (99.8944) (103.605) 0.20 4.1702*** 4.3763*** 4.94 161.5526****170.0762**** 5.28 240.5192** 250.3893*** 4.10 155.5460 105.7 −32.05 (1.5331) (1.3700) (42.5000) (38.5468) (105.8385) (95.9235) (105.2599) (95.7819) 0.3 3.5232*** 3.9015*** 10.74 91.0295*** 113.9696**** 25.20 143.3511*129.8391 −9.43 124.3274 67.2696 −45.89 (1.2361) (1.2221) (34.2670) (34.3852) (85.3358) (85.5675) (84.8693) (85.4411) 0.4 4.8272**** 5.0364**** 4.33 129.6012****142.35**** 9.84 181.1876*133.4727 −26.33 110.1938 59.2273 −46.25 (1.3370) (1.3718) (37.0638) (38.5953) (92.3008) (96.0442) (91.7962) (95.9024) 0.5 6.9707**** 7.8317**** 12.35 158.9553****171.5833**** 7.94 239.6707** 238.8667** −0.34 149.7197 173.7 16.02 (1.6812) (1.6729) (46.6079) (47.0683) (116.0686) (117.1292) (115.4341) (116.9563) 0.6 7.3090**** 8.5382**** 16.82 159.4485*** 174.3088*** 9.32 300.5054** 296.8706** −1.21 190.6122 290.0647** 52.18 (1.8587) (1.9602) (51.5276) (55.1521) (128.3203) (137.2457) (127.6188) (137.0431) 0.7 8.4451**** 9.5586**** 13.19 129.9750** 138.8777** 6.85 230.1383 242.5075 5.37 146.5787 169.1504 15.40 (2.1720) (2.1143) (60.2118) (59.4868) (149.9467) (148.0327) (149.1270) (147.8142) 0.8 11.1429**** 12.4580**** 11.80 110.8519 122.3387*10.36 184.7438 184.8048 0.03 227.6133 237.8971 4.52 (2.4539) (2.3244) (68.0268) (65.3983) (169.4084) (162.7433) (168.4823) (162.5031) 0.9 7.6342** 9.9575**** 30.43 73.0562 53.475 −26.80 124.1993 57.275 −53.88 536.4136*** 561.85*** 4.74 (2.9596) (2.6384) (82.0468) (74.2312) (204.3228) (184.724) (203.2059) (184.4513) 0.95 7.0240** 9.0411**** 28.72 27.5248 56.8889 106.68 291.1356 191.4 −34.26 697.3974*** 503.6444*** −27.78 (3.3179) (2.6613) (91.9788) (74.8782) (229.0565) (186.3341) (227.8043) (186.0591) Note: *p <0.1. **p <0.05. ***p <0.01. ****p <0.001. Table 9 Energy comparisons in three rural regions by Kruskal–Wallis test. Region 1 Region 2 Electricity Gasoline Solar energy Coal Diesel Stalk LPG Fireweed Energy consumption Tianping Dapu ⊗ ⊗ ⊗ • ⊗ ⊗ ⊗ • • Shihuiyao ⊗ ⊗ • ⊗ ⊗ ⊗ ⊗ ⊗ ⊗ Dapu Shihuiyao • • • ⊗ ⊗ ⊗ • ⊗ ⊗ Note: •denotes no significant difference at 5% significant level; ⊗denotes significantly difference at 5% significant level. size, household income, family size, household age, air conditioner and refrigerator) in three typical townships (Tianping Village, an affluent rural village; Dapu Village, a well-off rural village; Shihuiyao Village, an out-of-poverty rural village) using the OLS and QR models. The OLS results indicate that the significant variables at the 10% level in Tianping Village, Dapu Village, Shihuiyao Village and all investigative villages are household income, family size, air conditioner, refrigerator; household income, family size; household income, family size; and household income, family size, respectively. The QR results fluctuate around the QLS results with large amplitude, indicating a different pattern of the effect at different quantiles. As the significant variable in three study villages, household income in three investigative rural regions is significantly different, while family size of the Tianping Village resident is close to that of Shihuiyao Village resident, but significantly different to that of Dapu Village resident. Air conditioner and refrigerator of the Tianping Village resident, significant variables for Tianping Village resident, are close to that of the Shihuiyao Village resident, but different to that of Dapu Village. Table 10 The comparisons of six influence factors by Kruskal–Wallis test. Region 1 Region 2 HA HI FS HHA AC RE Tianping Dapu • ⊗ ⊗ ⊗ ⊗ ⊗ Shihuiyao • ⊗ • ⊗ • • Dapu Shihuiyao • ⊗ • • ⊗ • Note: •denotes no significant difference at 5% significant level; ⊗denotes significantly difference at 5% significant level. The out-of-poverty rural village consumed more biomass energy and less total energy, and energy consumption is positively correlated with family size, but negatively correlated with household income. The reason is migrant workers mean higher incomes and less energy consumption. Thus, local governments should rationally arrange energy supply according to the local population, further encourage farmers to continue to use bio-energy on an existing basis, and improve the cleanliness and convenience of bioenergy. At the same time, local governments should strive to create local jobs and reduce migrant workers. 618 R. Yang et al. / Energy Reports 4 (2018) 603–618 The well-off rural village consumed more coal due to local abundant coal resources, and energy consumption is positively correlated with family size and household income. Thus, local governments should guide rational energy consumption, use cleaner energy sources such as bio-energy and solar energy, and reduce dependence on heavily polluting coal. Moreover, the local government should provide a more rational energy subsidy program based on the rural population, guide farmers to local employment, in particular, allow local farmers to enjoy the benefits of the local energy industry. The affluent rural village consumed more commercial energy and cleaner coal owing to high income, more air conditioner and refrigerator, and government benefits. The energy consumption is positively correlated with family size, household income, air conditioner and refrigerator. Thus, the affluent local government should rationally guide the local population’s energy consumption, strengthen energy education for local farmers, provide market access standards for high-energy equipment, and encourage more affluent farmers to buy more energy efficient products. Moreover, local government should supply clean energy with low pollution instead of energy with high pollution, and energy equipment with high combustion efficiency instead of low combustion efficiency. The energy consumption in all investigative area is positively correlated with family size and household income, which is different to that in Dapu Village and Shihuiyao Village. So, it is unreasonable to use the overall data instead of separate regions. To control the rapid growth of rural energy consumption, the central government should guide residents to rational energy consumption when improving rural household income. Furthermore, because climate has a great influence on energy consumption, three study areas in this paper are all within or close to the cold zone climate region. Further studies will extend these conclusions and corresponding policy implications to other climate zones. Acknowledgments This research is supported by Tianjin Natural Science Foundation, China (No. 16JCZDJC31400) and the China Scholarship Council, China (File No. 201808120029). References China Nation Standardization Management Committee, 1998. General principles for calculation of the comprehensive energy consumption. China Standard Press, Beijing, China, p. GB/T 2589. Chung, W., Kam, M.S., Ip, C.Y., 2011. A study of residential energy use in hong kong by decomposition analysis, 1990–2007. Appl. Energy 88, 5180–5187. Daioglou, V., van Ruijven, B.J., van Vuuren, D.P., 2012. Model projections for household energy use in developing countries. Energy 37, 601–615. Ding, Y., Qu, W., Niu, S., Liang, M., Qiang, W., Hong, Z., 2016. Factors influencing the spatial difference in household energy consumption in China. Sustainability 8, 1285. Ding, Z., Wang, G., Liu, Z., Long, R., 2017. Research on differences in the factors influencing the energy-saving behavior of urban and rural residents in China–A case study of Jiangsu Province. Energy Policy 100, 252–259. Druckman, A., Jackson, T., 2008. Household energy consumption in the UK: A highly geographically and socio-economically disaggregated model. Energy Policy 36, 3177–3192. Joanes, D.N., Gill, C.A., 1998. Comparing measures of sample skewness and kurtosis. J. R. Stat. Soc. D Stat. 47, 183–189. Kaza, N., 2010. Understanding the spectrum of residential energy consumption: A quantile regression approach. Energy Policy 38, 6574–6585. Lee, C., Chiu, Y., 2011. Electricity demand elasticities and temperature: Evidence from panel smooth transition regression with instrumental variable approach. Energy Econ. 33, 896–902. Li, K., Lin, B., 2015. Impacts of urbanization and industrialization on energy consumption/CO2 emissions: Does the level of development matter? Renew. Sust. Energ. Rev. 52, 1107–1122. Liu, Y., Gao, Y., Hao, Y., Liao, H., 2016. The relationship between residential electricity consumption and income: A piecewise linear model with panel data. Energies 9, 831. Liu, W., Spaargaren, G., Heerink, N., Mol, A.P.J., Wang, C., 2013. Energy consumption practices of rural households in north China: Basic characteristics and potential for low carbon development. Energy Policy 55, 128–138. Miao, L., 2017. Examining the impact factors of urban residential energy consumption and CO2emissions in China – Evidence from city-level data. Ecol. Indic. 73, 29–37. NBS, 2016. China Energy Statistical Yearbook 2016. China Statistical Press, Beijing. NBS, 2017. China Statistical Yearbook 2017. China Statistical Press, Beijing. Nie, H., Kemp, R., 2014. Index decomposition analysis of residential energy consumption in China: 2002–2010. Appl. Energy 121, 10–19. Niu, S., Jia, Y., Ye, L., Dai, R., Li, N., 2016. Does electricity consumption improve residential living status in less developed regions? An empirical analysis using the quantile regression approach. Energy 95, 550–560. Niu, S., Zhang, X., Zhao, C., Niu, Y., 2012. Variations in energy consumption and survival between rural and urban households: a case study of the western loess plateau, china. Energy Policy 49, 515–527. Pachauri, S., Jiang, L., 2008. The household energy transition in India and China. Energy Policy 36, 4022–4035. Ping, X., Li, C., Jiang, Z., 2013. Household energy consumption patterns in agricultural zone, pastoral zone and agro-pastoral transitional zone in eastern part of Qinghai-Tibet Plateau. Biomass Bioenerg. 5 (8), 1–9. van Ruijven, B.J., van Vuuren, D.P., de Vries, B.J.M., Isaac, M., van der Sluijs, J.P., Lucas, P.L., Balachandra, P., 2011. Model projections for household energy use in India. Energy Policy 39, 7747–7761. Tsinghua University Building Energy Conservation Research Center, 2016. 2016 Annual Report on China Building Energy Efficiency. China Construction Industry Press, Beijing, China. Wang, X., Dai, X., Zhou, Y., 2002. Domestic energy consumption in rural china: a study on sheyang county of jiangsu province. Biomass Bioenerg. 22, 251–256. Wei, Y., Liu, L., Fan, Y., Wu, G., 2007. The impact of lifestyle on energy use and CO2 emission: An empirical analysis of China’s residents. Energy Policy 35, 247–257. York, R., 2007. Demographic trends and energy consumption in European Union Nations, 1960–2025. Soc. Sci. Res. 36, 855–872. Yu, Y., Guo, J., 2016. Identifying electricity-saving potential in rural China: Empirical evidence from a household survey. Energy Policy 94, 1–9. Zhang, M., Guo, F., 2013. Analysis of rural residential commercial energy consumption in China. Energy 52, 222–229. Zhu, H., Guo, Y., You, W., Xu, Y., 2016. The heterogeneity dependence between crude oil price changes and industry stock market returns in China: Evidence from a quantile regression approach. Energy Econ. 55, 30–41.