Introduction

Since the twenty-first century, household energy consumption as a new energy consumption growth pole has increasingly been mentioned by scholars (Druckman and Jackson 2008; Joyeux and Ripple 2007; Tao et al. 2018; Zhang et al. 2017; Zhu et al. 2019). At the same time, household energy consumption has also caused more environmental and health problems (Muller and Yan 2018). According to statistics, in 2017, 2.8 billion people worldwide continued to rely on biomass, coal, or kerosene for cooking, which has led to household air pollution related to the premature deaths of 2.8 million people each year (IEA 2017). Therefore, promoting a change in energy use for rural household cooking and heating from traditional biomass energy, coal, and other high-polluting energy sources to clean, efficient, and safe modern energy sources is conducive to improving the environment and residents’ health.

Compared with developed countries, most developing countries, especially rural residents in developing countries, still rely heavily on fuelwood and other solid fuels. In the research on energy choices for household cooking in rural India, Pandey found that the use of firewood and chips increased from 75% in 2004–2005 to approximately 78% in 2007–2008 (Pandey and Chaubal 2011). Bahadur examined the trends and patterns of household energy consumption in Ghana, and the results showed that 47% of households in Bhutan used fuelwood in 2012 (Bahadur Rahut et al. 2016). As shown in Akpalu’s research in Ghana, each person used approximately 640 kg of wood fuel per annum in 2011 (Akpalu et al. 2011). Under the pressure of environmental pollution and the associated health problems, the Chinese government has taken various actions to promote the use of clean energy in rural areas, including improving energy efficiency and improving the rural energy structure. However, to better realize the improvement of the rural energy structure, the factors affecting the energy choices of rural residents must be understood. Household income is considered to have an important impact on residents’ energy choices. With increased income, a family will upgrade its energy choice from traditional fuels such as fuelwood to intermediate fuels such as coal and eventually to modern energy sources such as natural gas and electricity (Arnold et al. 2006, Iwona and Parikh 2018, Muyeye and Henk 2007, Muller and Yan 2018). In addition to income, the selective use of energy by households in the same income class or different income classes is affected by the price and availability of different energy sources (Baiyegunhi and Hassan 2014; Farsi et al. 2007; Lay et al. 2013). When there is free fuelwood or cheaper coal that is easily available to residents, they will tend to choose these options. Moreover, some studies have shown that characteristics such as education level, gender, and family scale will also have varying degrees of impact on family energy choices (Arabatzis and Malesios 2011, Baiyegunhi and Hassan 2014, Chen et al. 2016, Guta 2012, Heltberg 2004, Jiang and O’Neill 2004, OEzcan et al. 2013, Peng et al. 2010, Rao and Reddy 2007). For example, Pandey and Chaubal found that large-scale households tend to prefer low-quality fuels over clean fuels (Pandey and Chaubal 2011). In addition to socioeconomic factors, the process of energy selection is a manifestation of behavioral awareness. This choice behavior will be affected by an individual’s environment, including policy guidance, technological development, personal environmental awareness, role division, and other objective factors (Edwards et al. 2004; Gebreegziabher et al. 2012; Grieshop et al. 2011; Kar and Zerriffi 2018; Liao et al. 2019). Therefore, when the independent variables for this study were selected, energy-saving behavior, energy-saving awareness, and policy perception were also considered to be factors affecting household energy choices.

Obviously, the discussion of factors affecting residents’ energy choices is ongoing, and it has been difficult to reach a unified conclusion. Factors such as social environment, economic development, and personal characteristics play different roles under different assumptions (Hou et al. 2019; Sana et al. 2020). Masera proposed the energy-stacking model (ESM) to describe the dynamic evolution of household energy use (Bianca et al. 2014; Masera et al. 2000). He believed that energy stacks are common in urban and rural areas in developing countries. The energy stack corresponds to a variety of energy usage patterns; that is, modern energy sources may be used only as a partial substitute for traditional energy rather than a perfect substitute. The reasons why households choose multiple energy sources may be the occasional shortage of modern energy sources (Kowsari and Zerriffi 2011), the high cost of appliances related to purely modern energy sources, and other factors that lead to households’ incomplete use of modern energy sources. More importantly, the object of energy selection is not a single specific energy type but diversified energy types (Brauwer et al. 2019; Chen et al. 2019; Martey 2019). For example, some households will use electricity as the main energy source for daily life but will also choose other energy sources, such as coal and natural gas, as supplements, or some households may consume relatively little electricity and use more energy from sources such as coal or fuelwood. Compared with previous research on residents’ choice of a single energy source, this study provides a more realistic research basis; that is, residents’ energy choice is not a choice of a single energy source but a choice of different energy modes, as will be explained in detail in the “Materials and methods” section.

To address the question of the key driving factors for different energy selection modes and how to effectively promote the energy transition of rural households, this study selected 474 rural households in central China as the research objects and carried out the related research based on in-depth interview data. We found that even though Chinese rural areas have basically achieved a full power supply, the use of highly polluting energy sources such as biomass and coal still accounts for a large proportion of energy use. Through the multinomial logit model (MNL), we further explored the choice preferences of different household energy modes. The main contributions of this paper to the literature are as follows.

First, this research makes a preliminary and beneficial attempt to divide energy modes into 4 types based on the existing research and our research object. The modernization and cleanliness of these four energy mode types are gradually increasing. This classification method avoids the drawbacks of the single energy choice mode in previous studies, that is, the “one or the other” relationship, and is more in line with the actual situation.

Second, due to data limitations, rural households are relatively fragmented and difficult to monitor, and their energy consumption data, especially noncommercial energy data, are not easy to obtain, resulting in weak basic research on rural residents’ energy consumption. This research is based on an in-depth interview data of 474 rural households collected by experts. The results provide a certain reference value for energy selection and energy conversion in rural areas.

Third, this research proves that the family type has an important influence on residents’ cooking and heating energy choices. We find that different family types have different preferences for energy choices, and two-generation families are more inclined to choose the modern cleaning mode, which provides new ideas for the government to deal with rural energy issues.

Data

Questionnaire design

This research is based on the data of a survey conducted by our research group in 2017. The survey was conducted mainly on the topic of the current status of the energy consumption of rural residents in China, including three parts, as shown in Table 1. The first part focused on the basic information of the respondents, including the basic characteristics of rural family members (such as gender, age, and education), family characteristics (such as household size, income, and household type), housing conditions (living area and type of house), household appliances (electrical power, number of appliances), and transportation status (number and type of vehicles). The second part investigated the energy consumption of rural households, including types of household energy consumption, consumption and expenditure, and the current status of energy use for cooking, sanitation, heating, and refrigeration. The investigators helped family members recall their consumption expenditure on various energy sources in the preceding year and estimate their consumption of different energy varieties (such as coal, electricity, and gasoline) based on different local energy prices. The third part mainly examined the energy conservation initiative of residents, including energy conservation behavior, energy conservation awareness, and policy perception. To ensure the validity of the research, this paper uses SPSS24.0 to test the reliability and validity of part III. The results show that the Kaiser-Meyer-Olkin (KMO) value is 0.840, and the P value of Bartlett’s sphericity test is 0.000, which is significant. The Cronbach’s α coefficient of each latent variable is greater than 0.60, indicating that the scale has good reliability and validity. The inspection results are shown in Appendix 1 Table 6 and Appendix 2 Table 7.

Table 1 Scale composition
Full size table

The Likert 5-point scale was used to measure energy-saving behavior, energy-saving awareness, and policy perception in the third part. The higher the score is, the higher the frequency of energy-saving behaviors, the stronger the awareness of energy-saving, and the greater the sensitivity to energy-saving policies. Specifically, energy-saving behavior mainly refers to the energy-saving behaviors in terms of energy use for electrical appliances or transportation as well as persuading others to save energy. Energy-saving awareness mainly refers to energy-saving responsibility, willingness to purchase energy-saving products, and environmental protection knowledge (Lillemo 2014, Ma et al. 2011, Na et al. 2012). Policy perception mainly measures rural households’ understanding of and participation in energy-saving subsidy policies (Ma et al. 2011).

Study area and data collection

Taking into account the different geographical climates and different levels of economic development in rural areas in different regions, this article selects only the central and eastern regions of China as the research area. This field survey was conducted by random sampling and on-site interviews in the survey area from January to February 2017. A total of 483 questionnaires were issued, and 474 valid questionnaires were recovered, for an effective questionnaire rate of 98.1%. Table 2 shows the sample recovery situation in different villages and towns. For convenience, we labeled the survey locations sequentially as P1–P6.

Table 2 Data distribution in the survey area
Full size table

Hebei, Henan, and Shandong are neighboring provinces with similar geographical conditions. The climate is a temperate monsoon climate that has high temperatures and rainy conditions in summer and cold and dry conditions in winter. From the perspective of economic development, the total GDPs of Shandong, Henan, and Hebei ranked 3rd, 5th and 8th, respectively, in 2016. In the same year, the per capita disposable income of rural residents was approximately 13,000 yuan in Shandong Province and approximately 11,000 yuan in Hebei and Henan Provinces. In general, the geographical and economic conditions of the three provinces are similar. The impact of the geographical and hydrological environment was considered in selecting specific survey sites, and the survey villages are all located near the middle and lower reaches of the Yellow River. These villages and towns still have differences in natural endowment and environmental policies. Among them, Xinxianzhuang Village in Hebei Province has vigorously developed tourism in recent years and thus has paid more attention to environmental and ecological protection. Shicheng Village in Hebei Province belongs to the Beijing-Tianjin city group and is more vulnerable to policy influences. Mihe Town, Henan Province, was one of the top 1000 towns in China in terms of comprehensive strength in 2019, and its residents had a higher standard of living. Shangyu Township in Henan Province and Dawu Town in Shandong Province are both close to coal mining sites. The oil fields of Puyang County in Henan Province are rich in resources. Figure 1 shows the specific location of the survey area.

Fig. 1.
figure 1

Survey area map

Full size image

When we chose the villages for this research, considering the differences in resource endowments, we chose P3 and P6, two villages with rich coal resources, and P2 and P4, two villages with insufficient fossil fuel resources. From the perspective of policy impact, we selected two villages, P1 and P5, that are closer to the capital or provincial capital cities and thus may be more affected by policy. Based on the above considerations, we selected the residents of 6 villages as the research objects so that the research objects would be random and representative under limited conditions. Table 3 demonstrates the characteristics of these villages.

Table 3 Characteristics of research sites in this study
Full size table

Materials and methods

Estimation method

When rural households choose energy sources for life purposes, they face multiple choices rather than single choices. Therefore, the MNL can be used in a regression with energy choice as the dependent variable (Bahadur Rahut et al. 2016; Liao et al. 2019). In this model, rural households are the rational actors of energy consumption, and their decision-making is intended to maximize household utility. Any energy choice mode is the result of farmers’ self-selection under current conditions. If the mutually exclusive energy choice modes for household i are y = 1, 2, …, J (with J being a positive integer), suppose the random utility of energy choice mode j of household i is:

$$ {U}_{ij}={x}_i^{\prime }{\beta}_j+{\varepsilon}_{ij}\ \left(i=1,\dots, n;j=1,\dots, J\right) $$
(1)

In Eq. (1), xi is an explanatory variable that changes only with household i and does not change with energy choice mode j. Under the condition that household i chooses energy choice mode j, if and only if the utility of energy choice j is higher than that of other options, the probability that household i will choose mode j is:

$$ P\left({y}_i=j|{x}_i\right)=P\left({\varepsilon}_{ik}-{\varepsilon}_{ij}\le {x}_i^{\prime }{\beta}_j-{x}_i^{\prime }{\beta}_k,\forall k\ne j\right) $$
(2)

Assuming that {εij} in Eq. (2) is iid and obeys the I-type extreme value distribution, it can be proven that the probability of family i choosing mode j is:

$$ P\left({y}_i=j|{x}_i\right)=\left\{\begin{array}{c}\frac{1}{1+{\sum}_{k=2}^J\exp \left({x}_i^{\prime }{\beta}_k\right)}\ \left(j=1\right)\\ {}\frac{\exp \left({x}_i^{\prime }{\beta}_j\right)}{1+{\sum}_{k=2}^J\exp \left({x}_i^{\prime }{\beta}_k\right)}\ \left(j=2,\dots J\right)\end{array}\right. $$
(3)

In Eq. (3), the scheme corresponding to “j = 1” is the reference scheme, and the corresponding coefficient β1 = 0.

Let p = P(y = 1| x), since \( p=\frac{\exp \left({x}^{\prime}\beta \right)}{1+\exp \left({x}^{\prime}\beta \right)} \); then, \( \frac{p}{1-p}=\exp \left({x}^{\prime}\beta \right) \), and “p/(1 − p)” is called “probability ratio” or “relative risk,” which means the probability ratio of choosing this plan and not choosing this plan. The greater the probability ratio is, the stronger the preference for this plan.

Assuming that xj changes to xj + 1 and p accordingly becomes p∗, then the ratio of the new probability ratio to the original probability ratio, that is, the probability ratio, can be written as

$$ \frac{\frac{p^{\ast }}{1-p}}{\frac{p}{1-p}}=\frac{\exp \left[{\beta}_1+{\beta}_2{x}_2+\dots +{\beta}_j\left({x}_j+1\right)+\dots +{\beta}_k{x}_k\right]}{\exp \left({\beta}_1+{\beta}_2{x}_2+\dots {\beta}_j{x}_j+\dots +{\beta}_k{x}_k\right)}\approx \exp \left({\beta}_j\right) $$
(4)

Equation (4) shows that the probability ratio can be obtained by calculating exp(βj), which is used to explain the multiple of the probability ratio change caused by each unit increase of the variable xi.

The cooking and heating energy of rural households mentioned in this study specifically includes the energy used for daily cooking, heating in winter, and heating water for bathing. The survey found that there are four modes of energy choice in the cooking and heating of rural households, as shown in Fig. 2: (1) the biomass compound mode, using electricity and fuelwood at the same time; (2) the multiple compound mode, using electricity, coal, and firewood (biomass energy, including straw and firewood) simultaneously and mixing all available energy sources; (3) the modern commerce mode, using electricity and coal simultaneously (although the combustion of coal is also polluting, it is a better modern commercial energy source than firewood); and (4) the modern clean mode, using only clean modern commercial energy power and liquefied petroleum gas (LPG). Obviously, the modernization and cleanliness of these four energy modes are gradually increasing.

Fig. 2
figure 2

Energy choice mode for cooking and heating in rural households in central and eastern China

Full size image

MNL model and variables

On the basis of domestic and foreign research and considering field research, this study selected 11 variables that represent household wealth, household characteristics, energy conservation initiatives, heating demand, and natural endowment as potential influencing factors. On the basis of the MNL, the regression model for the selection of cooking and heating energy in rural households is constructed as:

$$ y={\alpha}_0+{\alpha}_1^1{X}_1^1+{\alpha}_1^2{X}_1^2+{\alpha}_1^3{X}_1^3+{\alpha}_2{X}_2+{\alpha}_3^1{X}_3^1+{\alpha}_3^2{X}_3^2+{\alpha}_4^1{X}_4^1+{\alpha}_4^2{X}_4^2+{\alpha}_4^3{X}_4^3+{\alpha}_5{X}_5+{\alpha}_6{X}_6+{\alpha}_7{X}_7+{\alpha}_8{X}_8+{\alpha}_9{\mathrm{X}}_9+{\alpha}_{10}{X}_{10}+{\alpha}_{11}{X}_{11}+\varepsilon $$
(5)

where y is the selected mode of cooking and heating energy for rural households, α0 is a constant term, and X1 is the family income class. The direct impact of income on energy use has been considered by many scholars. The energy ladder model (ELM) posits that as income increases, households will turn to more complex energy carriers. Although this hypothesis still needs to be fully verified, the general observations it elaborates regarding the strong dependence of household fuel use on income are very consistent (Damette et al. 2018; Liao et al. 2019). X2 is the number of household appliances. With the improvement of the economic situation, more families have the ability to purchase various appliances, and more clean energy is needed to power them. Traditional biomass energy will be insufficient to meet the energy needs of households with such appliances. This economic development will be reflected in the number of household appliances, which may also promote changes in the household energy structure (Barnes and Floor 1996; Han et al. 2018). X3 is the education level of the head of the household. The head of the household is the decision-maker for household affairs, and his or her education level may have an impact on household energy choices. However, existing studies have not yet reached a unified conclusion on the impact of education level on energy decision-making. This article takes the education level of the head of the household as a potential explanatory variable (Alem et al. 2016; Njong and Johannes 2011). X4 is the family type, and X5 is the leisure of the family members. Some studies have shown that there is a nonlinear relationship between household size and fuel choice, which means households with more people are less likely to use LPG than fuelwood (Rao and Reddy 2007). Different generations have different attitudes toward the environment, which determines the influence of family type on household energy choices (Bardazzi and Pazienza 2017). If more family members are idle at home, there may be more opportunities to collect biomass energy, which increases the uncertainty of the choice of multiple energy sources. Therefore, this article adds the family type and leisure of family members to the MNL model as a potential explanatory variable. X6 indicates energy-saving behavior, X7 represents energy-saving awareness, X8 refers to policy perception, X9 is heating demand, X10 and X11 are differences in natural endowment, and ε is a disturbance term. Considering that there may be a collinearity problem between the independent variables, before the multivariate logit regression, the Pearson correlation test on the selected independent variables are carried out. The results (Appendix 3 Table 8) show that there is no correlation between the variables or the correlation is weak.

The y values 1, 2, 3, or 4 represent choices of different energy choice modes, as shown in Fig. 2. This study more concisely named these modes mode 1, mode 2, mode 3, and mode 4, representing the modern clean mode, modern commerce mode, multiple compound mode, and biomass compound mode, respectively. Family income level is divided into four groups, and three dummy variables are utilized: \( {X}_1^1=1,{X}_1^2=1,\mathrm{and}\ {X}_1^3=1 \) indicate that the family is a poor family (3/4 income class), a rich family (2/4 income class), or a richer family (1/4 income class), respectively, and \( {X}_1^1={X}_1^2={X}_1^3=0 \) mean that the family is a poorer family (4/4 income class). Education level is a categorical variable: \( {X}_3^1=1\ \mathrm{and}\ {X}_3^2=1 \) indicate that the education level of the head of the household is junior high school or high school and above (including technical secondary school), respectively, and \( {X}_3^1={X}_3^2=0 \) mean that the education level of the head of the household is elementary school and below. Household type is also a categorical variable: \( {X}_4^1=1,{X}_4^2=1,\mathrm{and}\ {X}_4^3=1 \) indicate that the family is a single person, a couple with a family, or a two-generation family, respectively, and X41 = X42 = X43 = 0 mean a family with three or more generations. X5 indicates whether family members have free time. When X5 = 1, it means that someone is at home all day. Energy-saving behavior, energy-saving awareness, and policy perception are all represented by questionnaire statistical scores, with a full score of 10. The higher the score, the better the performance. The heating demand is characterized by the annual heating time of the household in hours. X10 and X11 both represent differences in natural endowment. X10 and X11 are the dummy variables of Hebi City and Tengzhou City, respectively; these cities are close to coal mining sites.

Results

Descriptive statistics on the current situation of energy choice modes for rural household cooking and heating

Figure 3 shows the current status of energy choices for rural household cooking and heating in central and eastern China. In the survey area, rural households have been fully energized, and all households used electricity for cooking and heating. Most households (75.3%) choose to use LPG, which is a clean energy source similar to natural gas or electricity, for cooking. Coal is an important commercial energy source for rural households, and 81% of households use coal for cooking and heating. Biomass energy still occupies an important position in the life of rural households. Approximately 41.8% of rural households collect biomass materials such as firewood and straw, which are used mainly for fire, heating, and water for sanitary uses.

Fig. 3
figure 3

Current status of single-energy use for cooking and heating in rural households in central and eastern China

Full size image

The choice of cooking and heating energy for rural households is shifting from traditional biomass energy to commercial energy, i.e., from mode 4 to mode 2, but the low quality and serious pollution are still problems. Figure 4 shows that most households’ cooking and heating energy choices are in the second or third stage, the multiple compound mode (mode 3 with 36.5%), which uses electricity, coal, and firewood simultaneously or the modern commerce mode, which uses electricity and coal simultaneously (mode 2 with 44.5%). At present, the cleanest and most efficient energy mode is the modern clean mode (mode 1), which uses only electricity and LPG. This mode is rarely selected, with only a 13.7 % utilization rate. There are also a small number of households (5.3%) that choose the biomass composite mode (mode 4). In this energy choice mode, farmers use biomass energy such as electricity and firewood simultaneously, and most of their consumption is of fuelwood. This mode can cause not only serious health and environmental problems but also energy waste due to low energy efficiency.

Fig. 4
figure 4

Current status of energy choice mode for cooking and heating in rural households in central and eastern China

Full size image

Results of MNL estimation

The regression results of the MNL for cooking and heating energy choices in rural households are shown in Table 4. The probability ratio can be calculated through the relative coefficients in the table. The larger the probability ratio is, the stronger the preference of farmers for this energy use mode. The results show that the number of electrical appliances, family type, family members’ leisure, heating demand, and natural endowment can significantly affect the choices of cooking and heating energy for rural households.

Table 4 Regression results of MNL for energy choice for cooking and heating in rural households
Full size table

The effects of education level, energy-saving behavior, and policy perception are relatively weak, and energy-saving awareness failed the significance test. The education level of the household head has limited the influence on the preference of cooking and heating energy sources in rural households. Only a junior high school education for the household head has statistical significance in the choice of the biomass compound mode. In general, energy conservation initiatives have little effect on the choice of cooking and heating energy for rural households. Families with high policy awareness are more inclined to choose modern business models, possibly because they are more sensitive to policies such as tiered tariffs.

Main factors of energy choice mode for rural household cooking and heating

According to the regression results, the energy choice mode for cooking and heating in rural households is affected mainly by family type, family members’ leisure, number of electrical appliances, heating demand, and natural endowment. Table 5 shows the distribution of the number of households with different energy choices among the five main influencing factors. This section will analyze and discuss the different factors.

Table 5 Distribution of energy choice modes for cooking and heating in rural households
Full size table
  1. 1.

    Family type

Different family types have associated preferences for mode 1, mode 2, and mode 3. In the distribution of various energy choice modes, single-person households and married couples do not choose mode 4, which uses a large amount of fuelwood. This choice may be related to factors such as insufficient manpower at home to obtain fuelwood. The current distribution of the energy choice mode is slightly different from the logit regression results. The former shows that double-income, no-kids (DINK) families have a preference for energy choice modes from small to large: mode 1, mode 3, and mode 2. The preferences of two-generation families are mode 1, mode 2, and mode 3. The results show that the most preferred mode for these two household types is the efficient and environmentally friendly mode 1. Therefore, when other conditions are the same, compared with single-person households and DINK families, two-generation families tend to choose mode 1. This also shows that a reasonable family population structure will facilitate the widespread use of clean energy (Jiang and O’Neill 2004; Li et al. 2008; Pandey and Chaubal 2011).

  1. 2.

    Family members’ leisure

Considering the health factors, families with members at home who have free time prefer mode 1. From the distribution of each energy mode, it is clear that for households with people at home all day, the choice of mode 1 is much higher than for households without any members at home all day (from 2.56 to 15.91%). The energy choice of households where no one is at home all day is concentrated in mode 2 (58.97%) and mode 3 (38.46%). Only a small number of households choose mode 2 (2.56%), which is far below the average. The regression results of the MNL model showed that under the same conditions, families with someone at home all day prefer mode 1 over mode 2 and mode 3. This is in line with the fact that the proportion of households with people at home choosing mode 1 is much higher than that of households with no people at home.

  1. 3.

    Number of appliances

The number of household appliances has a significant impact on the energy choice mode for cooking and heating. Increasing the number of household appliances will help families choose cleaner and more efficient modern commercial energy, such as LPG and electricity. Households with more than 20 electrical appliances do not choose mode 4. Obviously, when rural households have cleaner and more efficient modern energy devices, they will reduce the use of low-quality and highly polluting fuelwood. The more appliances a family has, the more they tend to choose mode 1, followed by mode 2. The data show that for every electrical appliance in the household, the probability of choosing mode 1 is 12% higher than that of choosing mode 2, 16% higher than that of choosing mode 3, and 18% higher than that of choosing mode 4.

  1. 4.

    Heating demand

Farmers with a strong heating demand are more inclined to choose mode 2 and mode 3 with coal. This study uses the annual heating hours of rural households to characterize heating needs. Judging from the distribution of energy choice modes in different heating demand segments, with the increase in heating demand, the proportion of mode 1 declines. Mode 4, which uses a large amount of fuelwood, has a large proportion (12.03%) only in the lowest heating demand segment. Heating all day for 60 days (1440 h) is a turning point in the energy choice for household heating. Families choose mode 3 when their heating days are less than 60. As the heating demand increases, the proportion of mode 2 increases. When the heating time is higher than 1440 h, most households choose mode 2. From this point of view, when the household heating demand in the survey area is low, households will first prefer electric heating devices or firewood for heating. When the household heating demand increases, rural households will reduce the use of fuelwood and turn to using coal instead. The results of the MNL model shown in Table 4 also indicate that for every 24 h of heating time increase, compared with choosing mode 1, the family’s preference for mode 3 increases by 1.68%, and the preference for mode 2 increases by 2.64%. From a policy perspective, in 2017, 10 national departments jointly formulated the “Clean Heating Plan for the Northern Region (2017–2021)”, which clearly promotes clean heating in rural areas, including the “Coal to Electricity, Coal to Gas” project in rural areas. The choice of heat sources for clean energy heating is more diversified, including solar energy, natural gas, industrial waste heat, and other clean heating sources. However, policy-makers and authorities also need to be aware of the prevalence of hollowing in rural areas, low willingness to perform clean heating renovations, and the high cost of a clean energy supply. It is necessary to further strengthen overall planning and management, accelerate the transformation of rural areas, provide “low-cost and clean energy supply”, implement the “rural centralized residence” policy, and develop other methods to build a long-term clean heating mechanism.

  1. 5.

    Natural endowment

Natural endowment affects the rural households’ preferences for cooking and heating energy choices. Specifically, the availability of coal and fuelwood in different regions affects cooking and heating energy choices. Among the research areas, Hebi belongs to the National Energy and Heavy Chemical Industry Base and the Huanghuaihai Plain Agricultural Comprehensive Development Zone, while Tengzhou is the national key coal development base, as it is close to coal mines and has a good natural endowment. However, the energy choices for cooking and heating between these two are quite different in terms of current situation and household preferences.

Regarding the distribution of energy choice modes, the proportion of households in Hebi that choose mode 1 was 38.24% higher, while the proportion of households that choose mode 3 was 14.71% lower than in other areas. However, in Tengzhou, the proportion of mode 3 is much higher than the average 41.8%, while the proportion of mode 2 is much lower than the average. From the perspective of energy choice preference, the energy choice modes of Hebi and Tengzhou are significantly different from those of other regions. Rural households in Hebi strongly prefer mode 1, while residents of Tengzhou strongly prefer modes 2 and 3. This may be due to the different production characteristics and availability of fuelwood in these two places. The energy choice modes for residents will be greatly affected by the convenience of local natural endowment (Arabatzis and Malesios 2011; Peng et al. 2010).

Discussion

Energy choices for rural household cooking and heating in China may be more in line with the ESM than with other models. In previous studies, researchers have used the ELM or ESM to explain the living energy choices of rural households; the ELM posits that rural households have an ordered preference for different energy sources. As income increases, households turn to more advanced energy carriers; at the same time, they give up less advanced alternatives and finally achieve complete conversion (Broadstock et al. 2016). The ESM holds that multiple energy sources are used for multiple purposes at the same time. Modern energy sources can be good but limited substitutes for traditional energy sources (Dickinson et al. 2019). As shown in Fig. 5, as the income class transitions, commercial energy consumption replaces noncommodity energy consumption gradually but not completely. In fact, when other variables are controlled, it is income level that does not have a continuous or significant effect on the choice of energy use for cooking and heating in rural households. Only rich families have a slight preference for mode 1 or mode 2 and are more inclined to choose a clean and efficient modern mode. This is consistent with the findings of Yonas Alem, Mohammad Soltani, and Liao Hua (Alem et al. 2016; Liao et al. 2019; Soltani et al. 2019).

Fig. 5
figure 5

Current situation of energy choices for cooking and heating in rural households of different income classes. a Current situation of single energy source. b Current situation of energy choice mode

Full size image

The results of this study show that the education level of the head of the household has a limited impact on the energy choices of rural households. The driving factors for the cooking and heating energy choices of rural residents in the central and eastern regions are mainly the number of appliances, family type, leisure of family members, heating demand, and natural endowment. These results provide new policy ideas for the use of clean energy in rural areas and the transformation of the energy structure. When a household has more than one electrical appliance, the possibility of choosing the modern clean mode is 18% higher than the possibility of choosing the biomass composite mode. From 2009 to 2013, China implemented the first policy of providing home appliances to households in the countryside, which greatly promoted the use of home appliances in rural areas and to a certain extent promoted the transformation of energy use in rural areas. Based on the safe service life of home appliances, these home appliances have begun to enter the replacement and renewal period. The new round of promotion policies for green home appliances and smart home appliances will promote the use of clean energy in rural areas and the upgrading of the energy structure. In addition, from the perspective of the heating demand of rural residents, when the heating demand increases, the use of fuelwood will be reduced, but at the same time, compared with the modern clean mode, households’ preference for the modern commerce mode will increase by 2.64%. Therefore, the rural centralized residence policy will facilitate the upgrading and transformation of energy in rural areas. However, how to effectively adopt central heating in rural areas is worthy of in-depth thinking. China, South Korea, Japan, and other Asian countries have tried to establish concentrated settlements in rural areas to improve the living standards of rural residents. Promoting the use of clean energy is an additional benefit of this policy, as evidenced by the study of Liu (Liu et al. 2020).

Conclusions and policy implications

Conclusions

This paper builds an MNL model on the basis of microsurvey interview data in rural areas in central and eastern China to study the factors that influence the choice of energy use in the cooking and heating of rural households. The findings are as follows:

  1. 1.

    At present, the cooking and heating energy choice for rural households is shifting from traditional bioenergy to commercial energy, but there are still problems of low-quality and serious pollution. The multiple compound mode and modern commerce mode are the main energy choice modes. Although the survey area has achieved 100% electricity, 41.8% of households still use biomass fuel. This may be related to the availability of coal and biomass fuels in different regions, that is, the impact of resource endowments on the choice of cooking and heating energy for rural households.

  2. 2.

    The number of household appliances and heating requirements has an impact on the choice of cooking and heating energy for rural households. Every piece of electrical appliance in the household, the probability of choosing modern clean mode is 12% higher than the modern commerce mode. And every 24 h of heating time increase, compared with choosing modern clean mode, the family’s preference for modern commerce mode increases by 2.64%.

  3. 3.

    Interestingly, the two factors of family type and people’s leisure at home have an important influence on the energy choice mode. Compared with single-person families and DINK families, two-generation families are more inclined to choose modern cleaning mode. Families with someone at home all day prefer the modern clean mode than other modes. Generally speaking, in multi-generational families, there are some people staying home for leisure. Out of concern for the health of family members, these families prefer the modern clean mode. Therefore, the government should comprehensively consider a reasonable family population structure when dealing with rural energy-use issues.

Policy implications

The energy use of rural residents is an important part of energy consumption. Promoting clean and efficient energy use in rural areas will bring many benefits to rural residents’ health, rural economic development, and reducing environmental pollution. The policy implications of this article are as follows:

  1. 1.

    A new round of household appliances upgrading policies needs to be put on the agenda. The use of new household appliances will force rural residents to abandon the use of high-polluting energy such as biomass, while also promoting the use of clean energy, which is beneficial to the environment and residents’ health.

  2. 2.

    A sound social pension mechanism will bring additional benefits, that is, to promote the use of clean energy. A sound social pension mechanism will improve the population structure of rural families and reduce the generation of DINK families. Benefiting from the consideration of the health of offspring members by multi-generation families, this will promote the use of clean energy.

  3. 3.

    Improve the heating demand in rural areas and implement a central heating policy. Centralized housing policy in rural areas is mainly based on the goal of accelerating rural economic construction, and it also provides convenience for centralized energy supply and optimization of energy structure.

This article uses in-depth interview data on the cooking and heating of rural households in the central and eastern regions of China and focuses on issues related to the choice modes of cooking and heating energy. The research area and the key variables are limited, so there may be the problem of general omission of variable bias, which is also an obstacle to in-depth and extensive research on this issue. For example, various cooking and heating equipment, energy supply-related infrastructure conditions, energy subsidies, and other policies may have an impact on the energy choice modes for cooking and heating in rural households. We conduct diverse investigations and studies on the current energy consumption of rural households. In the future, research on this topic will be more comprehensive and generate more specific and significant policy guidance.