Strategy of Energy Conservation and Emission Reduction in Residential Building Sector: A Case Study of Jiangsu Province, China

,


Introduction
Te continually increasing world energy consumption is creating many challenges to the ecology [1,2] through its contribution to climate change, with the IPCC's (2018) report on the impact of global warming of 1.5 °C above preindustrial levels, for example, being of major concern.Residential energy consumption (REC) is a signifcant contributor, accounting for 21% of the world's total energy consumption according to the latest global energy data from the International Energy Agency (IEA).
Along with the rapid development of its economy, China's energy consumption has continued to grow rapidly in recent decades [3], surpassing the United States to become the world's largest energy consumer, with 20.3% of the world's energy consumption since 2010 [4,5] and with its REC accounting for 16.8% of total national energy consumption and 16.4% of global REC [6].While REC will continue to increase in the future, related studies have pointed out that per capita REC will increase at an average annual rate of 1.5% before 2040 and subsequently slow to 0.94% [5].
To help limit global warming to 1.5-2 °C by 2100, China has committed to having its CO 2 emissions peaking around 2030 [7,8] and achieving carbon neutrality by 2060 [9].As the residential sector is one of the most potential energysaving sectors [10,11], the Chinese government has proposed to change household energy to be cleaner and more efcient in its clean energy strategy [5].Te efective control of REC is a crucial part of China's CO 2 emissions peak and carbon neutralization plan.
In response, this study analyzes the REC in Jiangsu Province, China, as a case study.Jiangsu is a rapidly developing economy over the past two decades that has been at the expense of high energy consumption [12], which is a typical development pattern for most developing regions.Based on the STIRPAT model, panel data of 13 of the province's cities from 2001 to 2019 are explored to understand the infuence mechanism underlying the high growth of REC and establish the impact factors involved.Te results show that the average number of persons per household, per capita housing construction area, urbanization rate (urbanization in this paper refers to the process of transforming rural population into urban population), and cooling degree days have a signifcant positive efect on REC, while per capita housing construction area, residential water consumption, and residential liquefed petroleum gas (LPG) consumption have a signifcant negative efect.Te results of this study reveal the change of REC and its impact factors, helping to control energy consumption in residential sectors from the city level, which could be used for reference to other cities both in China and around the world.

Literature Review
2.1.Bibliometric Analysis on REC.VOS viewer, a computer program for constructing and viewing bibliometric maps, is used to make a bibliometric analysis and network analysis of the articles collected by the Web of Science Core Collection bibliographic database to analyze the main research areas in the feld of REC.Te query used was TS � ("residential energy consumption" OR "residential energy use" OR "household energy consumption" OR "household energy use"), which identifed 1,179 relevant articles published before 2021.Te terms extracted from the title and abstract of the publications are fltered for a minimum of 30 occurrences through the text mining function of the VOS viewer software [13].Te terms in the same cluster are marked in the same color (see Figure 1).
Te VOS viewer groups are termed into three categories by cluster analysis, where each cluster is marked with a diferent color.Cluster 1 (marked in red) focused on the occupant behavior as well as its interfering factors.Du et al. [14], for instance, explored the impact of occupant behaviors on energy consumption in high-rise residential buildings, while Wolske et al. [15] examined recent fndings on social infuence in energy behavior and discussed how this can result in peer efects.Cluster 2 (marked in blue) mainly contains studies of the raw material of residential energy, including fuel and LPG; Chen et al. [16] proposed set of regression models to quantify fuel consumption for the residential sector based on temperature-related variables and socioeconomic parameters, and Bhandari and Pandit [17] used a Long-range Energy Alternative Planning System (LEAP) tool to analyze the LPG demand for residential cooking from 2015 to 2035.Cluster 3 (marked in green) concerns not only on energy saving but also on emission reduction.Tis includes the study of Fan and Lei [18] accessing the key factors that afect the residential CO 2 emissions in Beijing from 1995 to 2015, based on a newly built decomposition model with generalized Fisher index and MP model to fnd that energy consumption intensity is a decisive factor in inhibiting residential CO 2 emission, and Zhang et al.'s [19] quantifcation of the indirect efects on energy usage and PM 2.5 emissions of urban and rural residents' lifestyles in China during 2005-2015.

Methods Used to Analyze REC.
Te methods used to analyze REC can be divided into two diferent but complementary categories, i.e., top-down models and bottom-up models.Te bottom-up method analyzes regional and national REC by the data for a representative set of individual houses [20].Te accuracy of the bottom-up model results depends heavily on the quality of the data source, which is difcult to test.Compared to the bottom-up models, the topdown models use national or regional time-series data [21], refecting the macroeconomic situation.
A STIRPAT (Stochastic Impacts by Regression on Population, Afuence, and Technology) theory, frst proposed by Richard York, has been widely used in the study of building energy consumption to analyze the factors impacting on energy consumption [22].Te STIRPAT model is a typical top-down model used to analyze REC at the macrolevel, analyzing the impact of population, afuence, and technological level.Hasanov and Mikayilov [23] use the STIRPAT model to demonstrate the relationship between diferent population age groups and residential electricity consumption in Azerbaijan over the period of 2000-2012, and Liddle [24] estimates the residential electricity consumption, using the STIRPAT model to analyze the U.S. state-based panel data.Te STIRPAT model is also used to explore the impact of urbanization on the energy  consumption of the residential sector, transportation sectors, industrial sectors, and commercial sectors, based on data from the Association of South East Asian countries over the period of 1995-2013 [25].With the promotion of energy conservation and emission reduction in China, REC has aroused much research interest in the country's academic feld, and an increasing number of studies use the STIRPAT model to analyze REC.Wang and Yang [26], for instance, use the balanced panel data of 29 Chinese provinces from 1998 to 2014 to investigate the nonlinear relationships between urbanization and REC based on the STIRPAT framework, and Dong et al. [27] combine an extended STIRPAT with a seemingly unrelated regression to explore the determinants of urban REC per capita and rural REC per capita based on 2007-2016 China provincial data.
All the articles reviewed above are at the national, provincial, or state level, however, with no analysis of REC by the STIRPAT model at the city level, resulting in the lack of a comparatively microscopic perspective to cope with REC issues.More specifc studies focusing on the city-scale REC are therefore needed.

Te Study Area. Te focus of this study is on Jiangsu
Province located in the eastern coastal center of the Chinese mainland (see Figure 2) and one of the most developed areas in China [28].As with many other Chinese provinces, since the turn of the 21st Century, its economy's continued rapid development [29] has created serious ecological problems [30], which have led to the need to consider the issue of saving energy and protecting the environment when pursuing economic development.
Referring to other studies using the STIRPAT model [31][32][33], the number of impact factors is set to six.Due to this limitation, some of the factors extracted from the literature need to be deleted.Te specifc process is as follows.

Research Workfow.
Te main work includes the extraction of infuential factors, weight calculation, and modeling analysis (see Figure 3).Te REC impact factors are extracted from the related literature, and then their weight is calculated based on the panel data of 13 cities in Jiangsu Province by the IWCM.Finally, the STIRPAT model is built to analyze the impact of these factors on REC.

Te STIRPAT Model.
Te IPAT model refects the infuence of human activities on the environment and is a widely accepted formula for analyzing the impact of population, afuence, and technological level on the environment [34,35].Te equation is as follows: where I represents the environmental impact, P represents the population, A represents the afuence, and T represents the technology.
Te IPAT theory tacitly accepts that the contribution of diferent factors to the environmental is the same, that is, diferent factors afect environmental quality in proportions, and the various factors are mutually independent.However, according to references [35,36], these assumptions are too idealized and the relationship between the various factors and environmental quality is not simply linear but often involves interactions or superposition nonlinear efects.
Te STIRPAT model introduces the exponential form into the model so that it can study the nonproportional efects of diferent factors on the environment [35,37].Te model is as follows: where a is the proportional constant term; b, c, and d are the elastic coefcients of population, afuence, and technology; and e is the residual value.Taking the natural logarithm on both sides of equation ( 6) gives the following equation: where the regression coefcient refects the elastic relationship between the explanatory variable and the explained variable.Te value of the regression coefcient is the percentage change in the dependent variable caused by a 1% change in the independent variable when the other independent variables remain unchanged [37][38][39].
Based on the literature review and understanding of the knowledge of the REC impact factors (see Table 1), cooling degree days (CDD) and heating degree days (HDD) are added to the original three dimensions of the STIRPAT model [40][41][42][43] to give the following equation: where EC it denotes the energy consumption per household, AP it is the average persons per household, pCA it is the per capita housing construction area, UR it is the urbanization rate, R&D it is the rate of the expenditures on research and development to GDP, and CDD it and HDD it are the CDD and HDD of the city i in year t, respectively (see Table 2).α is the intercept term, ε is the model error term, and β is the regression coefcient.When the independent variable changes 1%, the dependent variable will change β%.

Independent Weight Coefcient Method.
As some of the impact factors selected from the literature are related to each other to a certain extent, the factors containing information coincidence are deleted by the IWCM.Tis is an objective weighting method, which refects the amount of information contained in the index according to the collinearity between each index and other indexes, so as to calculate the weight of each index [64].IWCM mainly judges the correlation between one variable and other variables by calculating the complex correlation coefcient R value [64].Te stronger the correlation between one variable and other variables, the greater the complex correlation coefcient R value from the regression analysis, indicating that the index is more collinear with other indicators.Te strong collinearity represents a high degree of information repeatability, and it will be given a lower weight.
Te equation for calculating the complex correlation coefcient R is as follows:

􏽱
. (5) Te reciprocal of the complex correlation coefcient R i is recorded as follows: Te impact factor weight W i is converted into a decimal value in the range [0, 1] by normalization, with Before using IWCM to calculate the weight of each factor, it is necessary to standardize the data by where F ij is the jth initial value of the ith impact factor, F ′ is the standardized value, and F max is the maximum value of F. IWCM is then used to select the impact factors to be analyzed by the STIRPAT model (see Figure 4).
To avoid the situation where the estimation result is distorted or the model is unable to be estimated accurately because of a high correlation between explanatory variables, it is necessary to test the multicollinearity of the constructed regression model [65].Te variance infation factor (VIF) is often used for this purpose.It is generally believed in statistics that there is a positive relationship between the VIF value and multicollinearity, recognizing that the regression model may have the problem of multicollinearity if the VIF value is more than 10 [66].
Te unit root test judges the stability of the panel data by checking whether there is a unit root in the panel data series.If the data have a unit root, it is a nonstationary time series [67].If the data cannot pass the stability test, there is a great possibility that there will be t-test failure and pseudoregression, in the analysis of the multiple linear regression models.Here, the LLC test is used to test the stability of the panel data.

Data Source.
Given that the ofcial statistical yearbook of each region is only updated to 2020, the time range of this study is from 2001 to 2019.Te original data are derived from the Jiangsu Statistical Yearbook (2001-2019) and the 2001 to 2019 statistical yearbooks of the province's prefecture-level data.To unify the physical unit measurement of diferent energy, the collected energy consumption is converted from physical quantities to standard coal equivalents [68,69].Te climatic data of the cities refer to the daily average temperature data of urban meteorological stations.Te CDD and HDD are calculated according to the ASHRAE Handbook [70].

IWCM Results.
According to the weight, the average persons per household, per capita housing construction area, urbanization rate, the rate of the expenditures on research and development to GDP, CDD, and HDD are selected as the variables of the STIRPAT model (see Table 3).

Panel Data Overview.
Te time variable of panel data is from 2001 to 2019, and the panel variable is 13 prefecturelevel cities in Jiangsu Province with 247 samples in total.Te mean, standard deviation, minimum, and maximum values of variables are shown in Table 4.

Results of the Multicollinearity and Stability Tests.
All the VIF values of the explanatory variables are less than 10, and therefore the regression model passed the multicollinearity test (see Table 5).
Data for the variables in the multiple linear regression models are stable (see Table 6).

Model Results
. Te value of R 2 is 0.911, meaning that the model's degree of ft is high (see Table 7).
Tese results show that the average number of persons per household, per capita housing construction area, urbanization rate, and CDD have a signifcant impact on REC, while the rate of expenditure on research and development to GDP and HDD are not signifcant.It is shown that when the average number of persons per household, per capita housing construction area, urbanization rate, and CDD increased by 1%, the average energy consumption per household increases by 0.72%, 0.64%, 2.15%, and 0.25%, respectively.

Implications for the Evolution of REC in Developing
Regions.It is apparent that Jiangsu's REC has an obvious increasing trend from 2001 to 2019 (see Figure 5).Te residential electricity and natural gas consumption of each city all increased signifcantly during this period, while the residential liquefed petroleum gas gradually declined.Te changing pattern of REC in Jiangsu Province is the epitome of the energy conservation and emission reduction process  in the most developing regions in the globe, with a continuous and signifcant increase in REC and a tendency to be cleaner and more sustainable in the early-middle stage [11], [71].Te experience of Jiangsu has a certain referential value for the developing regions, and this section analyzes the REC in Jiangsu as a case to provide implications for the evolution of REC in the developing regions.Jiangsu is a typical hot summer and cold winter province in the middle and lower reaches of the Yangtze River [72,73], and air-conditioning is used for cooling and heating in most areas of the province [74].Tis means that domestic electricity consumption accounts for a large proportion in the REC structure.Tis dependence on electric heating in winter has attracted extensive attention from the country's social circles recently, suggesting the need to start a pilot demonstration project of clean and low-carbon heating in some of the province's cities, which may change its REC structure.
Although natural gas is not a satisfactory energy type for most countries that have achieved carbon peaking and carbon neutralization, its degree of exploitation and utilization in China is relatively low and coal is still the main source of energy supply.As a transitional energy type, natural gas can reduce SO 2 and dust emissions by nearly 100%, CO 2 emissions by 60%, and N 2 O emissions by 50% [75].Natural gas helps to reduce acid rain, slow the greenhouse efect, and fundamentally improve the quality of the environment [76,77].Terefore, to realize the goal of carbon peaking and carbon neutralization in China, it is suggested that full play is given to the role of natural gas as a bridge.Te use of natural gas by the residential sector can reduce the consumption of coal and oil and greatly alleviate the problem of environmental pollution.Natural gas was not very popular until the national promotion of natural gas in the Eleventh Five-Year Plan (2006-2010), when some newbuilt residential buildings began to use pipeline natural gas instead of canned LPG for cooking.Te proportion of natural gas consumption in the province's REC structure has increased, gradually replacing that of LPG since 2006 (see Figure 6).Te province has basically realized the popularization of natural gas use, and the priority in energy saving and emission reduction will change in the future.To release the stress on the environment, therefore, the province could promulgate policy to promote the use of renewable energy [11] and limit residential natural gas consumption [78].

Analysis of the Impact Factors.
Te positive impact of the average number of persons per household on REC is also found in other studies [53,[79][80][81][82]. Population growth will increase the demand for energy in the residential sector, and more residents consume more energy in daily life.
Consistent with Zhang and Li [28], Fan et al. [44], and Chen et al. [83], a positive relationship is found between the urbanization rate and REC in the present study.Te infuence of the urbanization rate on REC is related to economic development and the degree of urbanization [26,84].
Based on the study of 136 countries, Wang and Lin [84] argued that if the process of urbanization is not accompanied by corresponding economic growth, the increase in urbanization rate will even lead to the reduction of REC.
When the urbanization rate reaches a certain level, its impact on REC will no longer be signifcant [44,84].Jiangsu Province has maintained a rapid speed of economic development over the past two decades [12], and so there is a positive relationship between urbanization rate and REC.Terefore, as the province's urbanization rate has reached 72%, its impact on REC is expected to gradually decrease in the future.Per capita housing construction area that has a signifcant positive impact on REC is also found in other studies [43,85].Tis has a signifcant impact on lighting and air conditioning [53,86], and together with the number of residents, has a signifcant impact on the number and use intensity of household appliances (e.g., lighting and air conditioning) [87], leading to more energy being consumed to meet work needs [53].Moreover, when the per capita housing construction area increases to a certain extent, its impact on REC will be negative because of houses becoming vacant: Tonipara and Runst [86] demonstrated that the foor area has an inverted "U" efect on REC, fnding that REC decreases when the foor area exceeds about 100 m 2 .Terefore, as Jiangsu's per capita housing construction area has increased continually since 2000, the increase of per capita housing construction area is expected to have a negative impact on REC when it reaches a certain value in the future.
According to Li et al.'s study of the impact of climate change on REC in China [88], the impact of hot summers on REC is more signifcant than cold winters, owing to the diferences of heating systems and temperature tolerance of residents in diferent areas.A higher humidity has a negative efect on the thermal comfort of the human body [89], and compared with residents in hot and humid areas, those in hot and arid areas have a higher tolerance of the thermal environment [90].Jiangsu Province is in the ecotone of temperate and subtropical zones [91], and its air humidity is high [92].Terefore, the temperature tolerance of Jiangsu's residents is low, and they tend to consume more energy to improve the thermal comfort of the living environment when they feel hot.

Impact on Specifc Residential Consumption.
To investigate the infuence of the selected factors on specifc residential consumption, this study builds the STIRPAT models for Jiangsu's 2001-2019 domestic electricity consumption, domestic water consumption, household LPG consumption, and natural gas household consumption (see Figure 7).

Journal of Environmental and Public Health
Te results show that CDD has a signifcant impact on only domestic electricity consumption, which is because residents often use air conditioners to improve thermal comfort [93] and the running time of air conditioners has a direct impact on the daily electricity consumption.Water is mainly used for cooking, cleaning, and drinking, while LPG and natural gas are mainly used for cooking in Jiangsu's residential buildings.A high temperature may change people's living habits and have a certain impact on water consumption, LPG consumption, and natural gas consumption.With the improvement in the living standards, people often use air-conditioning cooling to obtain a suitable temperature environment, and so CDD has no signifcant efect on water household consumption, LPG household consumption, and natural gas household consumption.It has been recognized above that electricity consumption accounts for a large proportion in Jiangsu's REC structure, and so CDD has a signifcant efect on REC because of its signifcant efect on electricity consumption.
Te average number of persons per household and urbanization rate positively afects domestic electricity consumption, water household consumption, LPG household consumption, and natural gas household consumption.Te impact of per capita housing construction area on electricity consumption and LPG consumption is positive, while negative on water consumption and natural gas consumption.Tere is no signifcant impact of the rate of expenditures on research and development to GDP and HDD on electricity consumption, water consumption, LPG consumption, and natural gas consumption.

Conclusions
Tis innovative study develops an extended STIRPAT model to identify the factors with a signifcant impact on RED at the city level based on the data of 13 cities in Jiangsu Province, China, from 2001 to 2019, and several new fndings are obtained.Te model results show that when the average number of persons per household, per capita housing construction area, urbanization rate, and CDD increased by 1%, the energy consumption per household will increase by 0.72%, 0.64%, 2.15%, and 0.25%, respectively.CDD has a signifcant impact only on domestic electricity consumption and no signifcant efect on water consumption, LPG consumption, and natural gas consumption.Per capita housing construction area has a negative efect on the household consumption of LPG and natural gas.
In addition to the results of the STIRPAT model, it is also found that (1) domestic electricity consumption accounts for a large proportion of REC structure, but its energy consumption structure is likely to change in the future; (2) the proportion of natural gas in the REC structure is expected to gradually replace that of LPG; (3) Jiangsu's urbanization rate has reached 72% at present, and the impact of urbanization on REC is expected to gradually decrease in the future; (4) the impact of per capita housing construction area is also forecasted to change when the area reaches a certain value because of its inverted "U" efect on REC; and (5) the impact of hot summers on REC is found to be more signifcant than cold winters in temperate and subtropical zones with high air humidity.
For the regions that has almost realized the cleaning transforming of its energy consumption structure in residential sectors, the next step to improving REC quality is to promote the use of renewable energy and develop appropriate technologies.Since the increase in urbanization rate and per capita housing construction area will not lead to a signifcant increase of REC after reaching a certain value, a policy to improve residents' living standards would be benefcial for energy conservation and the felicity index of residents.
Te study is limited by not considering amount of straw, coal, frewood, and other carbon emitters consumed in rural residential buildings as it is not contained in ofcial statistics (although unlikely to be signifcant).Its scope is also limited to Jiangsu Province, China, which means that there will also be some limitation of the generalizability of its fndings.Furthermore, similar studies are therefore needed to examine the extent of this as well as further expanding the scope of the research and analyzing REC from such diferent perspectives as macro and micro and long-and short-term.

Data Availability
Te original data are derived from the Jiangsu Statistical Yearbook (2001-2019) and the 2001-2019 statistical yearbooks of the province's prefecture-level data.

Conflicts of Interest
Te authors declare that they have no conficts of interest.

Figure 1 :
Figure 1: Co-occurrence of terms clusters by a VOS viewer.Note: the size of node represents the number of occurrences of a term, and the length of the connection between the two nodes represents their relationship.

Figure 4 :
Figure 4: REC factors in Jiangsu Province based on the STIRPAT model.

3 Figure 7 :
Figure 7: Summary of the regression coefcients.Note: the factors with nonsignifcant regression coefcient are set to 0.

Table 1 :
Preliminary table of REC cluster factors.

Table 4 :
Descriptive statistics of the STIRPAT model variables.

Table 5 :
Multicollinearity test results of the STIRPAT model variables.

Table 6 :
LLC unit root test results of the STIRPAT model variables.

Table 7 :
Regression results of the STIRPAT model.