Introduction

Demand management, including energy conservation and improving efficiency, is crucial to reducing greenhouse gas (GHG) emissions (Dincer & Rosen, 1999; Omer, 2008). Energy demand in the residential sector is increasing worldwide, with a relatively high share of GHG emissions (Nejat et al., 2015). The residential sector accounted for 23.8% (2.2Gtoe) of global energy consumption (excluding non-energy use) in 2021 (IEA, 2023). In this respect, the potential savings in this sector are significant and will become even more critical in the future (Pablo-Romero et al., 2017). In addition, as secondary energy (Chand, 2020; Kavgic et al., 2010), electricity can cause social conflict and environmental load in the production and distribution process, so managing demand factors is an essential task from the perspective of environmental justice (Lim et al., 2014).

Korean residential energy consumption and intensity indicators are similar to or slightly below those of developed countries (IEA, 2023; World Bank, 2024; Kosis, 2021). Korean residential energy consumption per capita is 0.4 toe/cap in 2020, which is higher than the developing country group (non-OECD, 0.22 toe/cap) and slightly lower than the developed country group (OECD average, 0.5 toe/cap) or the EU (0.54 toe/cap). In terms of energy intensity, energy consumption per GDP or floor area in 2020 is 12.7 toe/M$, which is lower than China (16 toe/M$) or India (17 toe/M$) and similar to that of the EU (12 toe/M$) and US (13 toe/M$). Residential energy consumption per floor area is 13.6 koe/m2, higher than the United States (12 koe/m2) and lower than Europe (14 koe/m2) (González-Torres, et al., 2022).

However, since Korea is one of the major countries that uses a lot of fossil energy, more efforts to reduce consumption are needed considering greenhouse gas emissions. GHG emissions from fuel combustion in the Korean residential sector were 31.9 MtCO2 in 2019, ranking 14th in the world, and 76.2 MtCO2 (ranking 11th) considering the conversion sector altogether (IEA, 2021). Electricity accounted for 26.8% of Korean residential energy consumption in 2019, the second after city gas for heating and cooking (KEEI, 2020). Korean household electricity consumption was 76,303 GWh in 2020, 4.5 times higher than 16,960 GWh in 1990, and its compound annual growth rate has been 5.1% over the past 30 years (KESIS, 2021a).

In Korea, approximately half of the population lives in apartments (KOSIS, 2021), and residential apartment complexes are becoming a new target of demand management policies (Choi et al., 2018; Ko et al., 2020). Looking at the characteristics of energy consumption in the residential sector in Korea, city gas and oil are mainly used for heating and cooking, and electricity is generally used for lighting and home appliances. Apartments have the advantage of relatively high insulation performance compared to other houses. Therefore, when managing energy demand in apartments, it is necessary to focus on using home appliances rather than heating and cooking energy.

Currently, policy instruments for energy consumption in apartments, such as building energy efficiency certification (Ji et al., 2021) and mini solar PV distribution (Gunderson & Yun, 2021), are used by central and local government units. Ji et al. (2021) point out that building energy efficiency rating certification is mainly carried out in apartments, but this system does not consider the number of home appliances used. Although the building efficiency certification system is partially related to the use of air conditioners in the summer and electric heaters in the winter, it does not fully reveal the electricity consumption patterns of residents. In this regard, residents’ perceptions and behavior may be essential for demand management in addition to technological alternatives (Gram-Hanssen, 2011).

In Korea, apartments are a socially preferred housing type (Kim & Han, 2012; Kwon & Baek, 2013), young, wealthy households tend to reside there, and they are more likely to own a lot of electrical appliances. As more and more apartment complexes are being built and the number of households living in apartments is expected to increase, an electricity demand management strategy is needed to respond to this trend.

Recently, programs have been designed that exchange voluntary participation and economic incentives. For example, demand response (DR) programs allow utility companies to control home energy consumption remotely, or consumers can reduce it at the utility company’s request (Seo & Hong, 2014). Under the DR system, consumers receive incentives to reduce consumption for real-time demand management according to precontracted conditions (Ko et al., 2020). Therefore, it is essential to understand the factors in detail and guide apartment residents on reducing consumption.

The expected contributions of this study in the academic and policy fields are three things as follows:

  1. 1)

    This research provides a case study focused on residential electricity consumption of households living in apartments (similar to condominiums in western countries), which can be applied to countries or regions that favor high-rise and high-density housing types. Economic growth and urbanization in developing countries in other regions will also facilitate the spread of high-rise apartment buildings, and electricity consumption by those living in these housing types will continue to grow. Policy measures based on many empirical case studies can respond to changing housing trends and residential energy consumption

  2. 2)

    This study contributes theoretical discussions of energy practices through empirical analysis. From a social practice perspective, residential electricity consumption consists of the technical and behavioral aspects of the appliance uses. There have been cases in the literature dealing with the impact of home appliances. Still, there is a lack of cases comprehensively considering the complex factors of home appliance use, with some exceptions

  3. 3)

    This study can provide primary information needed to identify targets for public promotion. Through group comparison and quantile regression analysis, this study presented Korean households’ characteristics of home appliance use and the correlation between each home appliance and electricity consumption. Demand management in the household sector requires improving home appliances’ efficiency and habits. Efforts to identify core home appliances that require demand management and identify policy targets for each home appliance can increase effectiveness

Study background and literature review

Socially preference for apartments in Korea

According to the OECD Affordable Housing Database, the proportion of apartment/flat type in total residential dwelling unit stock is 75% in Korea, the highest among the 43 countries (OECD, 2022); however, the “apartments” in the Korean Building Act do not correspond to the housing stock classification of OECD statistics. Apartments only mean multi-family housing with five or more floors in Korea. Multi-family housing with four floors or less is classified as a “tenement house” or “Multi-household house,” not an apartment (Table 1).

Table 1 Korean legal housing type classification
Full size table

Even if the definition of apartments is limited to apartments with five or more floors, the housing stock and number of households living in apartments gradually increase (Fig. 1). According to the census survey, more than half of Korean households lived in apartments in 2020, and the proportion of apartments among dwelling stock was 63% including vacant houses. Korean household size decreased from 3.1 to 2.4 between 2000 and 2020, but during the same period, apartment household size showed a gradual decrease from 2.9 to 2.7. The floor area per person is gradually increasing, from 20.5 to 28.3 m2 (KOSIS, 2021).

Fig. 1
figure 1

Housing stock and households ratio by housing type. Source: KOSIS (2021). Note: “Multi-unit houses” are legally classified as “detached housing” in Korea and counted on a building basis rather than on a unit basis in the housing statistics of the census survey. Thus, there is a difference in the statistics of detached / multi-unit houses in 1.a and 1.b

Full size image

The increasing trend of apartments is due to the preference for social status and economic interests. Apartments in Korea are similar to condominiums in Western countries (Chang, 2022). Korean people socially recognize them as middle-class housing rather than low-cost rental dwellings. In general, apartments are considered more expensive and luxurious than other housing types, and most people tend to perceive owning an apartment as one of their life goals (Kim & Han, 2012). Apartments mean more than just housing as good investment assets in Korea (Shin & Yi, 2019). In economic growth, it emerged to replace old detached houses and supply new modern housing on a large scale (Shin et al., 2016). Since the 2000s, apartment branding like “well-being” and “luxurious” has reinforced the social preference for apartments by the construction industry (Kwon & Baek, 2013).

Apartment as a target of residential demand management

Social status and lifestyle are closely related to residential energy consumption patterns (Lutzenhiser & Gossard, 2000). Households living in apartments would have the affluent lifestyle and economic power to afford the relatively high housing prices compared to other houses. Figure 2 shows the difference in electricity consumption by housing type, with households living in apartments consuming relatively more electricity. Considering apartments’ social status and future trends in Korean society, policy attention is needed on the residential electricity demand for apartments.

Fig. 2
figure 2

Comparison of electricity consumption estimates by housing type in 2017. Note: The electricity consumption is based on general purpose, excluding night heating. Source: KESIS (2021b)

Full size image

The public also needs to reduce electricity consumption and wants to know specific ways to save. According to data of 2017 Korea Energy Consumption Survey (MOTIE, 2018), sharing information is the most desired energy-saving means for apartment residents, as shown in Fig. 3. Korean apartment households (39.2%) often said they need practical information to obtain energy savings. They also said improving awareness nationwide (24.5%) and disseminating energy conservation devices (22.6%) are essential. Therefore, it is necessary to develop a strategy for providing information tailored to each target: to whom and what information to provide. Steg (2008) stated that providing customized information is vital for promoting home energy conservation.

Fig. 3
figure 3

Energy conservation measures preferred by Korean apartment residents. Source: MOTIE (2018)

Full size image

Daily routinized electricity consumption

Electricity consumption at home results from routine and habitual practices using a variety of appliances that provide different services (Shove & Walker, 2014; Wilson & Dowlatabadi, 2007). How many household appliances are owned and used is related to various needs and wants (Shove, 2003). People will increasingly rely on home appliances to meet their needs in the future. Therefore, policymaking requires considering the effect of home appliance ownership and usage behaviors (Carlsson-Kanyama et al., 2005).

Gram-Hanssen (2011) suggested that home energy practices have four components: “1) embodied habits, 2) rules and knowledge, 3) engagement or meaning, 4) technologies.” The habitual practice, shared information and values, and technical characteristics such as specifications and energy efficiency of owned home appliances determine home energy consumption. Analyzing the impact of household appliances requires considering behavioral dimensions such as frequency, duration, and technical attributes. There should be a focus on the high consumption of home appliances and specific technology, perhaps by consumer groups associated with the appliances. Furthermore, it is also necessary to change behavior through information and incentives for the group that mainly uses each home appliance (Kavousian et al., 2013).

Therefore, it is essential to understand factors to provide appropriate saving information to Korean apartment households. For example, it is necessary to focus on who prefers which appliances and how much each device affects electricity consumption.

Review of prior research

Studies on the electricity consumption of apartments in Korea

Some of the previous studies dealing with Korean apartment energy consumption tend to focus on the physical properties of the building (Jang et al., 2015; Ji et al., 2021; Kim & Suh, 2021; Suh & Kim, 2019). Jang et al. (2015) analyzed the effect of building age on energy consumption by comparing a group of apartment complexes built at different times. They show that electricity consumption is less affected by the physical properties of the building, unlike heating energy. Ji et al. (2021) analyzed over 2.6 million apartments registered in the National Building Energy Database. They examined the effect of reducing energy consumption depending on whether or not they had a building energy efficiency certificate.

In the residential sector, energy efficiency certificates are closely linked to the insulation of the building, which in turn is closely linked to heating and cooling energy, accounting for a large part of home energy consumption. While electricity is used for air conditioners for cooling and other appliances used in Korea, it is not common for heating. As Ji et al. (2021) indicate, the energy efficiency certificate does not fully deal with the household’s electricity consumption by appliances. Therefore, we need to focus on home appliance use to identify household electricity consumption patterns and suggest electricity demand management measures.

Other studies analyzed the impact of the number of home appliances or usage patterns on electricity consumption (Choi et al., 2012; Park & Kim, 2012; Seo & Hong, 2014). Park & Kim (2012) conducted a survey targeting 140 apartment households and analyzed the effect of ventilation on electricity consumption. Choi et al. (2012) selected four apartment complexes based on whether they were residential-commercial complexes and the shape of the buildings. They compared the number of home appliances owned and energy-saving behaviors. Lee et al. (2019) analyzed the impact of housing characteristics and the number of home appliances on electricity consumption for 71 apartment households. Seo & Hong (2014) monitored electricity consumption for two weeks in the summer for 30 households living in an apartment in Daegu, Korea, collected technical and behavioral data on home appliances, and created a daily load profile. Although these studies provide valuable information on the relationship between home appliances and electricity consumption, they are limited to relatively small samples.

Estimating the effect of home appliances

Regression analysis captures the impact of electricity consumption on home appliances. Leahy & Lyons (2010) used Ireland’s Household Budget Survey data to suggest a relationship between ownership of nine home appliances and electricity consumption. Owning a freezer, dryer, dishwasher, or vacuum cleaner has been shown to increase electricity consumption. Bedir et al. (2013) surveyed 323 households in the Netherlands and divided home appliances into general, cleaning, cooking, hobby, and ventilation according to their intended use. They classified variables into duration of use and number. General home appliances and cooking appliances were found to have a significant impact considering the number of home appliances, and general home appliances and hobby home appliances were found to have a significant impact on the model of daily usage time.

However, household electricity consumption is not homogeneous. Some studies have attempted the estimation of appliances using quantile regression considering the heteroscedasticity of residential electricity consumption (Belaid & Rault, 2021; Frondel et al., 2019; Huang, 2015; Kim, 2020).

Belaid & Rault (2021) used survey data on household energy expenditure in Egypt. Using quantile regression analysis, they estimated the impact of income, household size and composition, occupation type, housing type, and size, and the presence or absence of an air conditioner on energy expenditure. Households with air conditioning consume more energy than households without, and the difference becomes more considerable in groups with higher spending.

Huang (2015) estimated household electricity consumption in a quantile regression model based on four datasets (1981, 1991, 2001, and 2011) from Taiwan’s FIES (Family Income and Expenditure Survey). The variables used in this study are divided into household head characteristics, common household characteristics, housing characteristics, geographical location, and home appliances (number of 5 types of appliances). Among home appliances, the number of televisions, air conditioners, and washing machines had a significant effect on electricity consumption over all periods, and the effect was greater for households with high electricity consumption than for households with low electricity consumption.

Frondel et al. (2019) used data from the German Residential Energy Survey (GRECS) to estimate the consumption contribution of individual home appliances using quantile regression. There are three main types of home appliance variables they used: number of home appliances (TV set, refrigerator, freezer, etc.), use cycle (washing, dishwashing, etc.), and ownership (for less common home appliances). This study showed that the number of refrigerators, freezers, and TVs strongly impacts electricity consumption, and the effect varies depending on the quantiles.

Using the same data as the Household Energy Standing Survey (HESS) database, which is the data source for this study, Kim (2020) estimated the effect of the annual usage time of TV, washing machines, air conditioners, refrigerators, and rice cooker using a quantile regression model. In this study, the effect of TV was insignificant in any quantile, but the effect of the refrigerator was significant. The marginal effects of washing machines and rice cookers were relatively greater in households with low electricity consumption, while the marginal effects of air conditioners and refrigerators were relatively greater in households with high electricity consumption.

Niu et al. (2016) analyzed how the types of appliances each household operates affect electricity consumption. According to this study, the greater the income, the more diverse the types of products held, and the impact of diversity of appliances tends to increase in higher consumption groups.

The differences between existing research and this study are as follows. This study integrates the comprehensive properties of home appliances into one factor and considers them in the analysis. Factoring and integrating household living behaviors is consistent with the perspective of social practice theory. Ben & Steemers (2018) and Sanquist et al. (2012) identified factors related to home appliances through factor analysis. In particular, Sanquist et al. (2012) incorporated household appliance technical and behavioral indicators and compared them according to the urbanization level. Considering home appliances’ technical and behavioral attributes are important for consumption, previous literature does not sufficiently meet the theoretical discussion.

We examine the usage characteristics of home appliances in a wide range of apartment households through a group comparison of factors derived from home appliance variables and estimate differentiated impacts according to energy consumption distribution based on quantile regression. Overall, this study can provide the primary information needed to identify the target of promoting savings policies for apartment households. Demand management in the home sector requires improving the efficiency and habits of home appliances, and here, efforts to identify policy targets for each home appliance can increase the effectiveness of demand management policies.

Method

Data source

The Household Energy Standing Survey (HESS),Footnote 1 conducted by the Korea Energy Economic Research Institute since 2011, provides information on the energy consumption, household details, housing, and home appliances of 2520 Korean households. The survey results are shown online in the National Energy Statistical Information System. This study used 2016–2018 data (KESIS, 2021a, b).

Because this study focuses on Korean apartment households, 3486 samples whose housing type was an apartment (out of 7560 samples in the 3-year survey sample) were the main analysis targets. After excluding 27 items with missing values due to nonresponse or rejection of responses in occupation and education variables, the total sample size was 3459. HESS is a survey that allows the replacement of samples. If households deviate from the sample, they are replaced with households with similar housing and household characteristics within the same region (KEEI, 2018).

Factor analysis and quantile approach

This study explores the relationship between home appliances, socio-economic attributes, and residential electricity in Korean apartment households. Accordingly, 34 variable data of 8 major home appliances were collected, and the data are related to the physical and behavioral properties of the home appliances (Table 2).

Table 2 34 variables of 8 main appliances in the study
Full size table

The main research flow is presented in Fig. 4. First, factor analysis integrates home appliances’ variables into each factor (factors related to home appliances). The variables indicating the ownership and use of a home appliance consist of count, size or capacity, usage frequency, and operating time. In particular, the greater the number of home appliances, the greater the usage and longer the usage time. Suppose entering variables directly into a regression model while ignoring the similarities between these closely related variables. In that case, multicollinearity may occur due to strong correlations between sub-variables within each instrument. Therefore, considering this correlation, sub-variables are integrated for each home appliance to ensure clarity and prevent multicollinearity. Factor analysis enables the consideration of various characteristics while efficiently reducing the number of variables.

Fig. 4
figure 4

Flow chart of analysis

Full size image

The factor-extracting method is the principal component factoring (PCF), and the rotation method is varimax. The number of factors was determined based on eigenvalues > 1, and a variable with a communality less than 0.4 was deleted (Costello & Osborne, 2005). The suitability of data for factor analysis was performed with Kaiser–Meyer–Olkin (KMO) and Bartlett’s test of sphericity.

Factor scores can be estimated as a result of factor analysis. They have a mean of 0 and a variance of 1, and the covariance between factors is 0; each is independent. The scores represent the relative spacing for a particular factor. For example, if a specific group’s score on the TV factor is relatively high, that group tends to watch large TVs more often than other groups. This study uses the factor score secondarily for group comparison and regression analysis.

Comparing factor scores for each home appliance with socio-economic variables makes it possible to determine which group prefers a specific home appliance more. For example, preferences for specific home appliances may vary depending on household size, aging level, and income. This study also analyzes the impact of home appliances on electricity consumption per capita, considering control variables such as household and housing characteristics. The presence or absence of heteroscedasticity influences the decision about which model to use. The ordinary least square (OLS) regression result would be adopted if heteroscedasticity did not occur in residuals. If heteroscedasticity were confirmed, the quantile regression result would be adopted. Concerning homoscedasticity, if the variance of errors in the OLS model is not constant, the parameter estimates can be biased, causing a validity problem (Hayes & Cai, 2007).

The conditional quantile approach can flexibly analyze the effect of each variable and provide relatively robust and efficient estimates for the asymmetric distribution of outliers and dependent variables (Koenker & Hallock, 2001; Yu et al., 2003). In this study, the electricity consumption of the household sample may be heterogeneous because of household composition, income level, dwelling size, home appliance use, and regional differences. A quantile regression enables more detailed analysis for specific data locations, such as households with high or low consumption (Buhai, 2005; Hao & Naiman, 2007). A quantile regression estimated according to the conditional τ quantile can be expressed according to this equation:

$${E}_{i}={\beta }_{\tau }{\prime}{X}_{i}+{\mu }_{\tau i}, {Q}_{\tau }\left({y}_{i}\left|{X}_{i}\right.\right)={\beta }_{\tau }{\prime}{X}_{i}, \tau \in \left[\mathrm{0,1}\right], {Q}_{\tau }\left({\mu }_{\tau i}\left|{X}_{i}\right.\right)=0\left(i=\mathrm{1,2},\dots ,n\right)$$

where Ei is the electricity consumption per capita as the dependent variable, and Xi is the (k × 1) vector of the independent variables composed of home appliances, household, and housing-related variables. τ is the conditional quantile with a value between 0 and 1, τ of 0.5 means median quantile. βτ is the (k × 1) regression coefficient vector by quantile, and μτi is the error term. Qτ(yi|Xi) represents the τth conditional quantile of the dependent variable (electricity consumption per capita), given the independent variable Xi. The quantile regression was conducted from the 10th to the 90th quantile to examine the influence of home appliance factors from low to high consumption levels. Regional and year dummy variables were added to the OLS and quantile regression models to consider the regional and temporal differences.

Descriptive statistics

Home appliance variables

Table 3 shows descriptive statistics for the home appliance data used in the study. Home appliance variables are divided into technical characteristics and behavioral characteristics. The type of the former variables is the number of products owned, size, and capacity, and the type of the latter variables is frequency and time of use. The variables related to behaviors are the average usage pattern of households according to specific time units such as yearly, weekly, daily, and hourly by each appliance. Therefore, usage behavior variables do not represent the exact results of use but rather the regular use pattern of each household. All home appliance variables were constructed by summing the home appliance values of each household, taking into account the variation in the number of home appliances used by each household. For example, the number and size of TV in television-related variables refer to the sum of the number and size of all televisions in the house, respectively.

Table 3 The ownership, size, and use behaviors of home appliances in Korean apartments
Full size table

Figure 5 details the level of home appliance ownership. Ventilating fans, refrigerators, televisions, and washing machines are necessities with close to 100% penetration. Over 70% of the sample owns two or more electric fans and refrigerators. Almost every household has a washing machine, but very few have two or more. The others are optional appliances that 14 to 30% of the sample did not own.

Fig. 5
figure 5

Home appliance ownership in Korean apartments. Note: The error bar means the standard deviation of the average value of each home appliance

Full size image

Figure 6 shows the annual usage time distribution, calculated based on the average usage time and frequency data of all appliances in the same category in each household. The rice cooker data were calculated based on the daily use of the warming function and the number of cooking times per week. If a household cooked more than eight times a week, it was assumed that the warming function was used daily for a week.

Fig. 6
figure 6

Annual hours of use by home appliance in Korean apartments. Notes: 1. The result is the sum of the hours of use by all home appliances in 2018 data set for limited usage frequency data. 2. Data with zero usage time or outliers are excluded

Full size image

As the refrigerator is operated every hour of the year, their usage times are much higher than those of other home appliances. Many households have two or more refrigerators, so it was expected that the difference in accumulated usage time for each device would be greater and their impacts on electricity consumption would be more significant.

The rice cooker was the most used except for the refrigerator, but the variance was quite different across households. As a multifunctional home appliance, the television also has a relatively large variance. Washing machines and vacuum cleaners are used for specific purposes, and their usage time is short.

Dependent and control variables

The dependent variable, annual electricity consumption per capita, was calculated by dividing a household’s yearly electricity consumption by the number of household members. The average annual electricity consumption of the 3459 apartment household samples used in the analysis was 3246.8 kWh (std. dev., 938.7 kWh), and the annual electricity consumption per person was 1158.6 kWh (std. dev., 595.3 kWh). The average value of Korean household electricity consumption was 65,495 GWh in this period, and the consumption per household and per capita were 3298.4 kWh and 1277.8 kWh, respectively.

The control variables are related to demographic and housing characteristics. Specific categories were created from these variables to fit the purpose of the study during the data preprocessing (see Table 4). The household size was classified according to the number of household members. The bigger the household size in the apartment, the larger the proportion of the sample.Footnote 2 The proportion of single-person households was relatively low in this sample as single-person households do not mainly reside in the apartment complex in Korea.

Table 4 Electricity consumption, household, and housing characteristics in Korean apartments
Full size table

For the age of household heads, 46 samples in their 20 s were only 1.3% of the total sample, so they are combined with 30 s. Therefore, age groups are divided into four subgroups. About 65% of the household heads were in their 40 s and 50 s. For elderly members over 65 years old, as another age variable, there are no 65 + older people in most household samples (84%).

In the HESS data, household income consists of nine intervals and measures the total annual income of all household members in Korean won. The difference between intervals is 1 million won monthly, roughly 1000 U.S. dollars. In the household distribution by income interval, more than half of the households belonged to the middle-income intervals, 3 to 5.

The educational level of household heads ranges from less than high school to a graduate degree. Bachelor’s degrees account for the largest proportion of the total. More than half of apartment households’ heads have studied at the university level. Household heads’ job types are employed, self-employed, and unemployed. Only 13% of household heads do not work, and most households maintain job stability.

Housing occupancy variables are classified according to ownership. Rent covers occupying without owning, including company housing, public rent, and free of charge. In general, a significant number of households in the sample live in apartments they own (76%). Floor area, a continuous variable along with electricity consumption, averages 93 square meters. Although there is some variation, you can see many medium-sized ones. The 84-square-meter type is the most popular among new apartments in Korea.

When looking at descriptive statistics, Korean apartment households have relatively many members, live in large houses they own, and are middle class with incomes above a certain level. In addition, the head of the household is middle-aged, highly educated, and engaged in certain economic activities.

Results

Integration of home appliance-related technical and behavioral dimensions

Factor analysis increases the clarity of analysis by integrating similar variables and avoids multicollinearity in regression analysis. Eight factors were extracted based on the eigenvalue(> 1), and the minimum eigenvalue was 1.657, resulting from factor analysis of home appliance-related variables. The result explains 72.8% of the total variance (Table 5). The use frequency of the washing machine and uptime of rice cookers for keeping warm were excluded as their communality was less than 0.4. The number of variables for each factor was three or four, and the minimum value of all factor loadings was 0.662. The Bartlett test of sphericity result was significant at 65,115.68 (df 496, p = 0.000), and the KMO value was 0.8, ensuring the validity of the analysis. Cronbach’s alpha value of all factors exceeded 0.6, which can be regarded as a reliable level.

Table 5 Factor analysis results
Full size table

The group segmentation of home appliance use

This section explores which groups prefer each home appliance. Figure 7 shows the results comparing home appliance-related factors and sociodemographic variables. Higher factor scores for home appliances are associated with owning and using more home appliances more frequently. A high score on a particular home appliance factor in a specific subgroup is highly related to the high use of that home appliance.

Fig. 7
figure 7

Comparison of appliance factors by socio-demographic variables in Korean apartments. Notes: 1. Fan, electric fan; AC, air conditioner; Washer, washing machine; Vacuum, vacuum cleaner; Fridge, refrigerator; R.Cooker, rice cooker; TV, television; PC, personal computer. 2. The factor scores indicate a relative position; the higher the value of a particular home appliance, the more it is owned and used in specific subgroups

Full size image

As the number of household members increases, the factor scores for home appliances generally tend to grow. Larger families require more devices and use them more frequently. What is noteworthy is that the factor scores for TVs in two-person households and washing machines in three-person households tend to be higher, contrary to the general trend.

Two major trends appear regarding the household head’s age and the number of older adults. Younger households prefer PCs, air conditioners, washing machines, and vacuum cleaners, while more senior families prefer TVs and refrigerators. The contrasting distribution of PCs, TVs, and refrigerators shows that generational differences are notable variables when reviewing home appliances.

These four devices preferred by young people (PC, air conditioner, washing machine, and vacuum cleaner) are relatively preferred by the highly educated but are less relevant to the unemployed. In particular, the air conditioner and computer factor scores responded significantly to educational level. TVs, refrigerators, and electric rice cookers are not less affected by differences in education or job type. However, households with highly educated people who graduated from graduate school do not often use electric cookers.

Households that rent seem to own and use fewer home appliances than those that own homes. Perhaps this trend has something to do with housing costs.

Factor scores for computers, air conditioners, refrigerators, and fans respond relatively strongly to household income. The higher the income, the lower the ratio of electricity bills to pay, so running the air conditioner in the summer can lower the indoor temperature and provide thermal comfort. Households with larger refrigerators are more likely to have the financial ability to purchase one.

Impact of home appliance use on electricity consumption

OLS and quantile regression

Table 6 shows the results of OLS and quantile regression analysis. Although the direction of the coefficients is consistent between OLS and the quantile model, the size and significance of the coefficients are often somewhat different between the two models. The Breusch–Pagan test for heteroskedasticity suggests that the OLS regression model has strong heteroscedasticity in the residuals (χ2 = 2240.53, p = 0.0000) and that adopting the quantile regression model is appropriate.

Table 6 OLS and quantile regression estimates
Full size table

Single-person households have higher electricity consumption per capita compared to other household sizes in the OLS model. A similar trend regarding household size is also found in the quantile regression model. In addition, the quantile regression model shows that the higher the quantile, the greater the impact of single-person households.

The influence of relatively older households compared to younger households with household heads under 40 is insignificant in OLS model, but in the quantile regression model, the differences in age of household heads were often significant. However, it is difficult to conclude from these results that older people consume more electricity. It may be related to the fact that as the head of the household gets older, the number of household members increases and grows and their electricity consumption increases (McLoughlin et al., 2012). In this respect, households with two or more elderly people consume less electricity than households without elderly people. Perhaps we can refer to the discussion that elderly households spend a long time at home but use fewer home appliances (Brounen et al., 2012). Similarly, the difference in the number of elderly people is insignificant in the 80th quantile, which has higher electricity consumption.

In the OLS model, the educational level of the household head is not significant, but the variable is significant in the lower quantiles in the quantile model, and the influence of the level of graduate school education is especially noticeable. However, except for graduate school graduates, the difference between low and high educational attainment was minimal in the median and the upper quantiles.

Regarding job type, both OLS and the quantile model show that there is little difference in electricity consumption according to the job type of the householder. Occupation does not seem to have a significant effect on electricity consumption, at least in Korean apartment households.

Income is an indicator closely related to electricity consumption, and it is generally known that high-income households consume more electricity (Huang, 2015; Sanquist et al., 2012). The difference between the upper income and the lowest income interval is generally significant. The quantile regression model shows that the difference between each consumption section and the reference section increases from the bottom to the middle quantile. However, the difference decreases or is insignificant in the upper quantile with high consumption. More empirical research on differences in quantiles and income in apartment samples is needed.

Home ownership was not a significant variable in OLS, but its influence was confirmed in quantile regression. Except for the upper quantile, the rental group consumes less than households that own and reside in their homes. As previously discussed in “The group segmentation of home appliance use,” this is similar to the fact that differences in home appliance use were revealed depending on occupancy classification.

Lastly, there is a significant relationship between house size and electricity consumption, and the impact of house size becomes greater in groups with higher consumption. Ten square meters promote 14.1 kWh in the lower quantile group, but 10 square meters promote 25.7 kWh in the upper quantile group.

OLS and quantile models show that home appliances significantly impact electricity consumption. In particular, air conditioners have the most significant coefficients in the middle and upper quantiles of the OLS and the quantile models. Excluding air conditioners, electric rice cookers, TVs, and refrigerators for the upper quantile have a relatively strong influence.

Influence distribution of home appliances across the quantiles

Figure 8 visualizes the impact of the appliances in Table 6 on electricity consumption, expanded to overall quantiles (0.1–0.9th). The x-axis of the graph represents the range from 0.1 to 0.9 of the quantiles used in the quantile regression model. The y-axis of the graph is a regression coefficient that represents the impact of each home appliance on electricity consumption. The difference in impact on electricity consumption according to the quantile for each home appliance is confirmed through this graph. The coefficients are decreasing, constant, increasing, or U-shaped according to quantile distribution.

Fig. 8
figure 8

Coefficient patterns of home appliances by quantiles. Note: Fan, electric fan; AC, air conditioner; Washer, washing machine; Vacuum, vacuum cleaner; Fridge:, refrigerator; R.Cooker, rice cooker; TV, television; PC, personal computer

Full size image

The vacuum cleaner is a decreasing type with less impact on consumption as the quantile increases. Its coefficients are insignificant in the higher quantile and may be associated with the lowest usage time. In households with low consumption, the effect of vacuum cleaners on electricity consumption is significant because the frequency of use of other home appliances is low.

The refrigerator’s coefficient decreases to the 40th quantile and then maintains a steady level. It operates every hour and constantly consumes a certain amount of electricity, so its effect is greater than other appliances in the lower quantile. Interestingly, even above the median quantile households, whose consumption is relatively high, more household appliances may be used frequently, but the coefficients of refrigerators remained high.

Fans, washing machines, and computers had a constant impact regardless of consumption level. Electric fans are generally less expensive to purchase and use and are used longer than air conditioner, from early summer to autumn. Also, the cooling effect increases when an electric fan is used with an air conditioner. For this reason, almost all households regularly use electric fans.

The washing machine is another essential appliance used to wash clothes and bedding. Most households have one washing machine, and the deviations in the number of units and the time of use among samples were relatively small. Regardless of electricity consumption level, the number of washing machines and their usage time can be constant. Computers had a consistent effect except in the 90th quantile. Computers were the fourth most used per year, excluding refrigerators, but only 70% of the sample households owned computers. Although computers are multifunctional appliances used for working, studying, searching for information, and entertainment, the relatively low level of ownership seems to limit the difference in coefficients by quantiles.

Television and air conditioner have more significant impacts as household consumption increases. The differences between minimum and maximum coefficients were 6.1 times for television and 1.8 for air conditioners. Almost all households own a television, one of the appliances with a long usage time per year, as it is a multifunctional appliance. The difference in television usage time between household samples was also relatively large. The purpose of television use may be simple or complex depending on the household, which affects the usage time deviation and makes it more influential in the higher quantiles.

Air conditioners also have a high impact on consumption in the 70–90th quantile. The annual usage time of air conditioners is lower than that of other home appliances because they are mainly used in summer. However, it is well-known for relatively higher electricity consumption; the impact can vary greatly depending on the number of air conditioners and how much they are used.

Lastly, the rice cooker showed a U-shape, with a strong consumption influence in the low and high consumption groups and a low influence in the middle. Its usage time was the longest, with the most significant deviation among all home appliances in this study. These characteristics may also affect the coefficient difference between quantiles, but more empirical studies are needed.

Discussion

Policy implications of focusing on home appliances

The use of home appliances is closely related to electricity consumption. In reality, we use home appliances daily at home, and the result of those actions is energy consumption. This repeated pattern of behavior is a routinized habit. Therefore, rather than the electricity consumption itself, what is important is to look at which home appliances increase consumption and why we own and use more of such home appliances in the social, economic, cultural, and regional context.

Discriminating major home appliances and interpreting the social and cultural relevance associated with those appliances is meaningful from a policy perspective. In designing and operating a policy on electricity consumption, it is very important to determine who the target group is. Setting clear policy targets can contribute to maximizing policy effectiveness. In particular, given that energy consumption is heterogeneous across all households, demand management must be viewed separately according to consumption level. Quantile regression analysis shows the impact of home appliances broken down according to electricity consumption distribution. Policymakers need to provide customized information and information on incentives to prevent further increases in consumption among heavy consumption groups above the median quantile level.

Sociocultural context related to the use of major home appliances

Which home appliances consume more electricity in apartment households in Korea? Why do Koreans place more value on and use home appliances? What creates such preferences? This study could not answer all of these questions, but it did at least identify some notable connections to home appliances. This study pays attention to the greater influences of air conditioners, rice cookers, TVs, and refrigerators. These home appliances have a larger influence than other appliances or tend to have a greater influence in groups with higher quantiles.

Different climatic conditions, electricity price levels, and social and cultural perceptions of air conditioning create differences in impact on consumption depending on the region. For example, the coefficient of air conditioners is not significant in Germany, possibly because only 0.4% of households have it (Frondel et al., 2019). In Korea, summer is hot and humid, and air conditioner controls humidity and cooling. In high-density metropolitan cities centered on apartments, which are standard housing in Korea, the temperature is higher because of the heat island effect, so air conditioning is more frequent. In addition, compared to other housing types, it may not be easy to ventilate an apartment. Even with the same apartment type, depending on the apartment complex’s structure, direction, and height, the ventilation effect or the degree of heat absorption may lead to differences in the use of air conditioners. In addition, as temperatures are expected to continue to rise due to the climate crisis, there is a possibility that demand for cooling will continue to increase, so there is a need to pay more attention to how air conditioners are used to manage electricity demand.

The impact of electricity consumption through television tends to be greater in households with high consumption. Similar results were found in other quantile regression results in Egypt (Belaid & Rault, 2021), Taiwan (Huang, 2015), and Korea (Kim, 2020). Meanwhile, in Germany, the effect of the 75th quantile was lower than that of the median quantile (Frondel et al., 2019). OTT (Over The Top) services may be one of the reasons for the strong influence of television. It has spread rapidly among Korean families since the 2010s. OTT providers such as YouTube and Netflix provide a platform where users can easily access various video content (Dwyer et al., 2018; Shon et al., 2021). TV also has a relatively strong association with older people and retirees. This trend seems to be familiar not only in Korea but also in other regions. In the United States, it has been reported that older adults spend three times more time watching TV than younger adults (Depp et al., 2010). As mentioned earlier, although older people spend much time living there, the variety of home appliance use may be low due to little familiarity with the latest technology and adherence to existing lifestyle habits. Perhaps watching TV may be a relatively easy activity for older people with limited mobility to spend their leisure time.

Previous studies suggested that refrigerators and freezers consume significant amounts of electricity (Frondel et al., 2019; Leahy & Lyons, 2010; Zhou & Teng, 2013). This study also found that refrigerators have considerable influence across quantiles. Refrigerators are products that operate 24 h a day, 365 days a year, and the preference for large refrigerators is increasing, which significantly impacts electricity consumption and needs to be paid attention to in the future. In Korea, many households do not need large refrigerators due to the decrease in household members and the development of delivery culture. However, large refrigerators are highly preferred due to the allocation of space for refrigerator placement during apartment construction and the display effect of refrigerators as symbols of social status (Ryu & Yun, 2023).

The importance of refrigerators in Korea is also connected to food culture. Korean food culture, which values the traditional fermented food called kimchi, can affect having additional refrigerators for specific purposes. In Korea, kimchi refrigerators have been developed to store kimchi separately. They began to be distributed in the early 1990s and gradually became mainstream (Cho, 2006). TVs and refrigerators are also highly associated with older people. Older people are more familiar with traditional foods such as kimchi than younger people, and they tend to cook and store their food so they may need more extensive and additional refrigerators.

In Asia, rice-centered food culture and attitudes toward health and nutrition are strongly related to using rice cookers (Son et al., 2013). The pursuit of convenience in the meal preparation process may have contributed to the social adoption of machines that cook and keep food warm (Yu et al., 2017; Zhang et al., 2020).

This study’s quantile regression model shows rice cookers significantly impact consumption through a U-shaped pattern. However, no particular associations with other socioeconomic variables were found.

Future strategies for managing electricity demand in the residential sector

Based on the above discussion, this section contains several suggestions for policymakers regarding electricity demand management.

First, there is a need to clarify the target of information provision for each home appliance. For example, as previously presented in “The group segmentation of home appliance use,” PCs and air conditioners are owned and used more in households where the head of the family has a higher level of education, has a higher income, and has a larger number of household members. Personal computers and electric fans are used more in middle-aged households. Households with four or more members, consisting of parents and children in their teens or 20 s, may need individual fans in each room and spend much time on personal computers for social media, games, and learning. It is consistent with these appliances’ relatively low usage in households whose heads are 65 years or older. Younger households mainly use washing machines and vacuum cleaners. They provide cleanliness, and it is estimated that households with young couples and their children need these services more. A Swedish case suggested that younger people prefer laundry services more than older people (Carlsson-Kanyama et al., 2005). Clarifying the information-providing group may increase the likelihood that recipients will also be interested.

Second, after clarifying the information provision target, developing a customized manual for each target is necessary. For example, young households with children may be receptive to a saving method for home appliances related to cleanness and comfort. More educated and higher-income households with children may need a way to save electricity using a personal computer and air conditioner. For elderly households, it may be practical to recommend energy-efficient televisions and refrigerators.

In addition to behavioral corrections to increase the energy efficiency of home appliances, the manual will need to compare the technical properties of each home appliance easily. For example, behavioral correction can include storing cooked food in the refrigerator and then reheating it when necessary rather than keeping food warm in an electric rice cooker for a long time. In addition, providing information on the power consumption of electric rice cookers will be useful in increasing understanding. According to the survey on home appliance ownership of Korea Power Exchange in 2019 (KPX, 2020), the power consumption of an electric rice cooker is 1050 W, less than air conditioners (1598 W) but much higher than refrigerators (42.2 W), televisions (151.5 W), electric fans (43.7 W), and computers and monitors (376.9 W).

Conclusion

The apartment is the most popular housing type in Korea and has recently been the target of energy-saving policies. This study focuses on the influence of home appliances on electricity consumption and the relationship between these factors and other socio-demographic variables.

The results of this study utilizing a combination of factor analysis and quantile regression have three main features. First, each home appliance’s number, specifications, and behavioral aspects (such as how often they are used) were integrated through factor analysis to improve the validity of the analysis. Second, the segmentation result shows that the ownership and usage behaviors of home appliances depend on socio-demographic groups. Through this approach, preferences for home appliances can be classified according to specific groups, and these groups can become primary data for setting targets for energy-saving consulting for households in apartment complexes. Third, the influence of each home appliance is not constant. According to the quantile regression analysis, it may differ depending on the consumption scale. In particular, this study identified the major home appliances that significantly impact higher-consuming households.

Not all households living in certain housing types are homogeneous, and lifestyles may differ depending on the scale of electricity consumption, household composition, income, and education level. This indicates that prioritizing each target group using home appliances and effectively providing customized guidelines is necessary. The number of households living in apartments will continue to increase in rapidly developing countries, so it is necessary to conduct similar studies that examine the factors related to energy consumption in apartment environments. These detailed studies on home appliances’ technical and behavioral aspects will contribute to making the segmented information provision strategies according to each target group for residential energy demand management.