Existing researches relevant to residential energy consumption behaviors and low carbon perceptions under LCCP are as follows.

First, policy contents and effects of LCCP. China has implemented three batches of LCCP to decrease energy consumption and carbon emissions in cities. Scholars have summarized policy measures in five areas: planning design, supporting policies, emissions inventories, low-carbon industries, and residential energy consumption (Wang et al., 2018; Westman and Broto, 2018; Peng and Bai, 2021). Many pilot cities have launched initiatives to decrease residential energy-related emissions, such as popularizing scientific knowledge, promoting low-carbon lifestyles, conducting campus education, and organizing energy-saving activities (Zhao et al., 2016; Li et al., 2018). Evaluation of policy effects in LCCP has been carried out by establishing indicator systems (Peng and Deng, 2021), factor decomposition (Qu and Liu, 2017; Cai et al., 2018), and causal inference (Hong et al., 2021; Shen et al., 2021; Liu et al., 2022). Among them, many scholars have considered LCCP as a quasi-natural experiment and adopted the DID model to examine causal relationships between LCCP and urban emissions (Huo et al., 2022), green growth (Cheng et al., 2019), eco-efficiency (Song et al., 2020), and corporate emissions (Chen et al., 2021). Despite being one of the major policy contents in LCCP, there are no researches to analyze the policy effects on residential energy consumption and carbon emissions.

Second, studies on accounting and influencing factors of residential energy-related emissions. Some works have calculated residential energy-related emissions by means of surveys (Li et al., 2019), input-output method (Fan et al., 2012; Xia et al., 2019), or consumer lifestyle approach (Chen et al., 2019). Nevertheless, these results are weak in accuracy and authenticity. Other researches measure residential energy consumption and carbon emissions from household living, such as appliances (Miao, 2017), housing (Ma et al., 2022a), cooking (Zhang et al., 2017), and transportation (Lin and Du, 2015). In addition, several works focusing on carbon emissions from heating, lighting, appliances, and cooling in commercial buildings (Xiang et al., 2022) are also very enlightening. The research scope ranges from urban clusters (Ma et al., 2022b) to national (Zhang et al., 2022a,b) to global (Xiang et al., 2022). Meanwhile, methods to identify the influencing factors of residential energy-related emissions include the log-mean divisia index (LMDI) method (Ma et al., 2022a), stochastic impacts by regression on population, affluence, and technology (STIRPAT) (Miao, 2017), latent dirichlet allocation (Wu et al., 2021), and driving force analysis (Shen et al., 2018). However, these methods are mainly based on conditions, such as economic development, population scale, household income, and technological progress, and have not provided deeper insight into the interrelationship between residential energy-related emissions and local policies.

Third, surveying and profiling for residential low-carbon perceptions and behaviors. To better understand the relationship between low-carbon perceptions and behaviors of residents, many scholars and institutions have conducted surveys and interviews since the 1990s (Wang and Zhou, 2020). Although complete information could not be obtained, survey research is increasingly deepening the understanding of residential low-carbon perceptions and behaviors (Yang et al., 2021). Researchers measured subjective factors such as psychological distance (Wang et al., 2019), emotional characteristics (Lehman et al., 2019; Galway and Beery, 2022), political orientation (Luo and Zhao, 2019; Gregersen et al., 2020), education level (Wang and Zhou, 2020), and social norms (Sörqvist and Langeborg, 2019). External environmental factors include climate experience (Bø and Wolff, 2020; Sambrook et al., 2021; Yang et al., 2021), local conditions (Cianconi et al., 2020; Zhang et al., 2020), and environmental management (Marshall et al., 2019), and so on. Upon these, classical theories on perception and behavior have grown, such as the value-belief-norm theory (Kiatkawsin and Han, 2017; Sarkis, 2017), attitudinal-behavior-circumstance (ABC) theory (Ding et al., 2018), and TPB (Huang and Ge, 2019). TPB underlines that intentions determine individual behaviors through behavioral attitudes, subjective norms, and perceived behavioral control (Tian et al., 2022). TPB builds a bridge between the perceptions and behaviors of residents with more explanatory power. Therefore, this study utilizes TPB as a theoretical framework to explain how LCCP can shape residential low-carbon perceptions and subsequently drive mitigation behaviors.

Promoting low-carbon lifestyles and mitigating energy consumption are indispensable policy contents of LCCP. As a comprehensive policy, it contains design planning, supporting policies, monitoring systems, industrial greening, and low-carbon living (Li et al., 2018; Westman and Broto, 2018; Peng and Bai, 2021). Overall, LCCP significantly decreased energy consumption and carbon emissions in pilot cities, and by a bigger margin than in other cities (Wang et al., 2015). For residential energy-related emissions, pilot cities also explore various initiatives to carry out, such as supplying clean energy, green transportation systems, improving building energy efficiency, and popularizing low-carbon ideas (Zhao et al., 2016; Ma et al., 2021). At the same time, residents have reacted positively to low-carbon policies by supporting emissions mitigation policies and taking low-carbon behaviors (Tian et al., 2022). Therefore, as policy priorities of LCCP, residential energy consumption and carbon emissions would be adjusted by pilot policies.

To better understand the influence mechanisms of LCCP on residential energy-related emissions, this study introduces TPB to diagnose the relationship between residential low-carbon perceptions and behaviors. TPB emphasizes that residential behaviors are derived from the combination of three perceptions: behavioral attitudes, subjective norms, and perceived behavioral control (Ding et al., 2018). This study utilizes TPB as a theoretical framework to investigate how LCCP influences residential behavioral attitudes, subjective norms, and perceived behavioral control and subsequently drives residential energy-related mitigation emissions behaviors.

First, behavioral attitude is a personal positive or negative affective tendency toward particular behaviors. It is influenced by the combination of rational evaluations and likelihood assessments of behavioral outcomes. Attitudes have remarkable effects on low-carbon behavioral intentions and realistic behaviors, and other related attitudes (attitudes toward science and climate) also influence residential low-carbon behaviors (Ding et al., 2018). Residents in cities are willing to buy environmental-friendly products, even if they pay more, as well as strongly support climate policies (Yang et al., 2014; Huang and Ge, 2019; Tian et al., 2022). Especially in some pilot cities, residential behavioral attitudes are influenced are reinforced by disclosing environmental information, establishing low-carbon museums, and encouraging waste recycling (Zhao et al., 2016; Ma et al., 2021). Therefore, LCCP seeks to influence residential behavioral attitudes and in turn motivate emissions control behaviors.

Next, subjective norms are external pressures that individuals sense. These pressures generated by other individuals or society make individuals form their own judgments about whether they should do or not. Individual low-carbon perceptions and behaviors would be influenced by psychological distance (Wang et al., 2019), emotional characteristics (Lehman et al., 2019; Galway and Beery, 2022), political orientation (Luo and Zhao, 2019; Gregersen et al., 2020), education level (Wang and Zhou, 2020), and social norms (Sörqvist and Langeborg, 2019). In China, where the government dominates the narrative and policy agenda of every societal challenge, concerns about climate change are focused on the environment and health. As a result, climate skepticism is not prominent, and education has a strong influence on the subjective factors of Chinese residents (Ding et al., 2018; Wang and Zhou, 2020; Yang et al., 2021). In the development of LCCP, social media propaganda, low-carbon school education, and National Low-carbon Day activities are conducted as initiatives to foster the low-carbon atmosphere in society (Zhao et al., 2016; Ma et al., 2021). Therefore, LCCP cultivates residential behavioral habits by configuring social subjective norms.

Finally, perceived behavioral control refers to the degree of difficulty individuals feel about performing a particular behavior. Perceived behavioral control includes perceptions of facilitators or hindrances (such as convenience, economy, and time conditions), as well as perceptions of the degree of influence of these factors. Concretely, individual behaviors will be adapted when conditions such as spatial planning, transportation design, housing conditions, and public services are within local conditions (Zhang et al., 2020). When residents have experienced climate change impacts, or their local environment is under serious climate threat, they would formulate higher perceptions of climate change and will be more likely to implement low-carbon behaviors (Wang and Zhou, 2020). Pilot cities have also influenced residential perceived behavioral control through carbon inclusive policies, such as subsidizing green travel, developing public transportation, rewarding energy and water conservation, and establishing credit systems (Li et al., 2018). Therefore, LCCP tries to promote residents to implement low-carbon behaviors by improving positive factors of behavioral perceptions.

This study focuses on carbon emissions resulting from residential energy consumption behaviors in cities. In order to obtain accurate data, residential household living energy consumption can provide a source for carbon emissions accounting. Zheng et al. (2011) and Zhao et al. (2012) classified residential energy consumption activities into four categories: electrical appliances, central heating, private transportation, and activities utilizing fuels. Miao (2017) calculated residential energy-related emissions for 216 cities in China from the perspective of private transportation and house-based energy consumption. Particularly, house-based energy consumption includes three major energy consumption sectors: (1) electricity consumption; (2) central heating; and (3) natural gas, liquefied petroleum gas, and coal gas uses. In this study, residential energy-related emissions due to household living energy consumption are defined as those resulting from residential electricity, central heating, residential gas, and private transportation, according to existing studies and available data. Through accounting above four categories of energy consumption and carbon emissions, residential energy-related emissions in cities are aggregated. Accounting methods are shown in Table 1.

Residential energy-related emissions (REE_it) are the sum of residential electricity, central heating, residential gas, and private transport. i and t are dummy variables for city and year, respectively.

To compare changes in residential energy-related emissions after LCCP, the DID model is adopted to examine policy effects. The National Development and Reform Commission of China issued three batches of LCCP notifications in 2010, 2012, and 2017, respectively. Since the second batch of notifications was issued at the end of 2012, considering policy lag, the pilot time is set to 2013. Three batches of pilots included 6 provinces, 77 cities, and 4 counties in total. Considering that, some of the pilot areas are provinces, prefecture-level cities within the pilot provinces would be taken as pilot cities (Yao and Shen, 2021). Total 120 prefecture-level cities included in the pilots are considered as research subjects. Therefore, in this study, pilot cities are regarded as the treatment group and non-pilot cities as the control group. The policy effects of LCCP on residential energy-related emissions will be examined by setting up the DID model:

where REC_it is CO₂ emissions due to residential energy consumption in cities; did_it is a dummy variable used to recognize pilot cities; Control_jit represents a series of control variables with j types; α, β, and δ are series of estimated regression coefficients; ν_t, γ_i, and ε_it represent year fixed effects, city fixed effects, and random error terms; i and t are dummy variables for city and year, respectively.

Control variables include (1) GDP, which is usually served as a reference to measure the economic development of cities; (2) population, which represents the population level of cities; (3) technology, which reflects the technological innovation capacity of cities; (4) finance, which measures the saving capacity of residents in the financial sector of cities; and (5) industrial structure, which measures the structural composition of local industrial development. With reference to the existing study, the above control variables are represented by GDP per capita, population density, ratio of science and technology expenditure in GDP, ratio of deposits in financial institutions in GDP, and ratio of secondary industry, respectively (Song et al., 2020; Yu and Zhang, 2021; Huo et al., 2022).

All 281 prefecture-level cities in China from 2004 to 2020 were taken as research samples. Data on central heating for household living were obtained from the China Statistical Yearbook of Urban Construction. Private urban car ownership is obtained from the traffic data of statistical yearbooks of prefecture-level cities. The rest of the data were acquired from the China City Statistical Yearbook. As mentioned earlier, some of the data are ratio, and non-ratio type data have been logarized (Huo et al., 2022; Liu et al., 2022) here. Missing values were filled in using interpolation (Song et al., 2020; Jia et al., 2021). The descriptive statistics of variables are shown in Table 2.

Using equation (2), benchmark regression results for the policy effects of LCCP can be obtained, which is shown in Table 3. As observed from column (1), the coefficient β representing LCCP is significantly negative. It indicates that LCCP would significantly decrease carbon emissions from residential energy consumption. Upon this, control variables were added to the regression analysis. Column (2) means that the regression result is still negative and significant at a higher level after adding the control variables. For further investigation of the policy effects of LCCP, this study replaces the dependent variable total energy-related consumption related to electricity, central heating, gas, and private transportation generated carbon emissions in cities (TEE). TEE is not just limited to household energy consumption but includes total electricity consumption, central heating, gas consumption, and private transportation in cities. In columns (3) and (4), it is found that the coefficients β are still significantly negative. Thus, these results above preliminarily confirm the theoretical Hypothesis 1, that LCCP would decrease carbon emissions from residential energy consumption.

Satisfying the parallel trend test is a prerequisite for the DID model (Hong et al., 2021; Jia et al., 2021; Liu et al., 2022). Parallel trends can test that there is no obvious difference in the trend of residential energy-related emissions between pilot cities and non-pilot cities before the implementation of LCCP. This study will adopt the event study method for parallel trend testing (Jia et al., 2021).

In Equation (3), D_it denotes a dummy variable for whether city i is a pilot city in various years t. Other variables are the same as in equation (2). Equation (3) focuses on the coefficient λ_t, which represents estimated coefficients of the policy effects of LCCP for pilot cities. Since the first batch of LCCP was implemented in 2010 and the time range of sample data is 2004–2020, this study picked 6 years before and after LCCP as the test time span. The pilot years are 2010, 2013, and 2017. In this study, the parallel trend test will be conducted for these three pilot periods together. The relative pilot years are used as the horizontal coordinates of Figure 1, with reference to the existing literature (Jia et al., 2021).

Figure 1 displays the results of the parallel trend test. Before the implementation of LCCP (left side of the dashed line), the policy effect is not significant and there is no common trend in coefficient values. However, after implementing LCCP (right side of the dashed line), the estimated coefficient values of λ_t, are significantly negative and the absolute values show a gradual increasing trend. The solid dots indicate the estimated coefficients λ_t of equation (3), and the error bar are upper and lower 95% confidence intervals. Accordingly, this study considers that the sample passes the parallel trend test. The implementation of LCCP can achieve a significant and increasingly strong policy effect of decreasing residential energy-related emissions.

Both benchmark regression and robustness tests have been able to show that LCCP can decrease carbon emissions from residential energy consumption. How does the policy effect work, in other words, by what mechanisms would LCCP influence residential energy consumption and carbon emissions? Based on TPB, residential low-carbon behaviors are influenced by perceptions. Therefore, this study proposes Hypotheses 2–4 in 2.2. These three hypotheses explore the influencing mechanisms of LCCP on residential energy-related emissions in terms of behavioral attitudes, subjective norms, and perceived behavioral control, respectively. In order to test the above hypotheses of influence mechanisms, this study adopts the mediating model to validate (Chen et al., 2021; Liu et al., 2022).

First, LCCP may strengthen the behavioral attitudes of residents to promote emissions mitigation behaviors. When residents have positive or negative affective tendencies toward low-carbon behaviors, they are more likely to display emissions mitigation behaviors or not. For example, those who show supportive attitudes toward low-carbon consumption intentions are more likely to purchase environmentally friendly products and support consumption-side emissions mitigation policies (Yang et al., 2014; Ding et al., 2018; Tian et al., 2022). To validate the mechanism, this study used the amount of per capita garbage harmlessly treated in cities to measure residential behavioral attitudes. The reason for this is that domestic garbage harmlessly treated represents the willingness of residents to engage in environmentally friendly behavior (Ek and Miliute-Plepiene, 2018). If residents are able to actively participate in domestic garbage harmlessly treated, then it indicates that they have positive attitudes toward environmental improvement behaviors. At the same time, domestic garbage harmlessly treated can also decrease a part of GHG from residential non-energy consumption (Chen et al., 2020). In Table 5, column (1) demonstrates that LCCP increases the amount of per capita domestic garbage harmlessly treated with a strong significance.

Second, LCCP would establish subjective norms for residents to promote mitigation behaviors. Subjective norms can form external pressures on residents. These pressures may stem from many subjective factors such as psychological distance, emotional characteristics, political orientation, and education level. Subjective norms play a crucial role in residential energy consumption mitigation behaviors due to the influence of traditional collectivist values culture in China (Webb et al., 2013; Mancha and Yoder, 2015; Chen, 2016). In China, education is the largest subjective factor influencing the behaviors of residents (Wang and Zhou, 2020). Therefore, the ratio of higher education to the total urban population in cities was chosen to validate the mechanism of subjective norms, as shown in Table 5. As demonstrated in column (1), LCCP significantly increases the ratio of higher education in the total urban population.

Third, LCCP could adjust the perceived behavioral control of residents to promote emissions mitigation behaviors. When residents perceive the degree of difficulty in implementing low-carbon behaviors, such perceptions can facilitate or hinder the occurrence of behaviors. In particular, spatial planning, transportation design, housing conditions, public services, and other conditions in cities could influence residential energy consumption and emissions mitigation behaviors (Zhang et al., 2020). Therefore, many pilot cities have adopted initiatives such as strict building energy efficiency standards, subsidizing the purchase of new energy vehicles, improving public transportation services, and rewarding energy and water saving behaviors to influence residential behavior perceptions (Liu and Qin, 2016; Zhao et al., 2016; Li et al., 2018). To validate the influence mechanism of perceived behavioral control, this study uses the number of buses owned per 1,000 people in cities. This is because buses are one of the low-carbon options for travel for residents in cities. Moreover, the number of buses could also reflect cities' efforts to mitigate emissions in the transportation sector. Column (3) demonstrated that LCCP increased bus ownership per 1,000 people and was significant.

Through the above mechanism analysis, it is clear that LCCP could strengthen behavioral attitudes, establish subjective norms, and adjust perceived behavioral control to shape residential low-carbon perceptions and subsequently drive emissions mitigation behaviors. Such results confirm that Hypotheses 2–4.

The vastness of China gives birth to distinct differences between cities in terms of geographic conditions, economic development, and institutional arrangements. The benchmark regression robustness test does not take into account these differences in location and size. For a deeper understanding of the differences in policy effects of LCCP across cities, this study examines the heterogeneity of geographic location and city size.

Cities located in various regions have differences in energy structure, economic development, emissions levels, and policy implementation standards. In this study, Chinese cities are divided into eastern, central, and western regions, as shown in Table 6. Columns (1–3) demonstrate the differences in policy effects of LCCP among these three regions. The estimated coefficients for eastern and central regions are significantly negative. It indicates that LCCP within eastern and central regions can help to decrease residential energy-related emissions. It is possible that due to the high level of economic development, concentration of educational resources, and availability of infrastructure in eastern and central regions. These contribute to the low-carbon perception of residents, who are willing to implement energy saving and emissions mitigation behaviors (Shen et al., 2021; Yang et al., 2021). Column (3) demonstrates that the estimated coefficient for the western region, although also negative, is not significant. Economic growth is still a priority in the western region, and insufficient management of energy consumption and lack of low-carbon perceptions in society, which in turn leads to the ineffectiveness of carbon emissions mitigation among residents. Thus, the implementation of LCCP in the western region is more difficult and policy effects are not effective.

Bigger cities, due to their economic development, massive population, and dense transportation, simultaneously generate the “big city disease,” such as environmental pollution, energy consumption, and resource scarcity (Cheng et al., 2019). Insufficient development, inadequate infrastructure, and environmental problems in smaller cities may also have an impact on policy effects. According to the Notice on Adjusting the Size of Cities issued by the State Council of China in 2014, this study divides pilot cities into two groups: megacities and large (bigger) cities, and medium and small (smaller) cities, as shown in Table 6. Columns (4) and (5) show the regression results in these two groups of cities. The coefficient estimates are all significant, which indicates that LCCP could achieve residential energy-related emissions mitigation in both bigger cities and smaller cities. By comparing the results in these two columns, it can be observed that the estimated coefficients are larger and more significant for bigger cities. These suggest that LCCP could be more helpful in mitigating residential energy-related emissions in bigger cities. The possible reason for this is that residents in bigger cities face more climate threats, such as extreme heat, health loss, and energy shortage. Therefore, they are more concerned about environmental issues and have stronger low-carbon perceptions, which in turn help to drive energy-saving behaviors (Huo et al., 2022). In contrast, smaller cities are still in the development stage and have relatively modest emissions levels, thus residential low-carbon perceptions are not as strong as bigger cities. Meanwhile, the infrastructure in smaller cities is insufficient, and the cost of emissions mitigation for residents is higher than inputs, resulting in fewer residential low-carbon perceptions and energy-saving behaviors.