1. Introduction
Reducing the contribution to atmospheric carbon dioxide is a challenging task and renewable energy is a major determinant to deal with it. The atmospheric concentration of carbon dioxide (CO
2), which is one of nine planetary boundaries (PBs), is considered an indicator to identify the risk of climate change [
1]. Steffen et al. [
2] warned that more than 450 ppm is beyond the high-risk area, which could be potentially devastating to the resilience system of Earth. Thus, increasing attention has been focused on renewables, which do not emit CO
2 in generating electricity, to reduce CO
2 emission in an effort to mitigate climate change. Among various renewable energy sources, concerns with household-scale electricity generation are centered on solar photovoltaics (PVs).
Numerous studies were conducted to find what drives social diffusion in the residential adoption of universal rooftop PVs [
3,
4,
5]. Most studies investigated smaller town residents who usually live in single-family houses, and others were based on multi-family buildings of a larger city focused on rooftop PVs [
6]. Only a few studies exist on residential PVs in an urban area. Thus, this study differs in that it examines household solar PVs installed on apartment balconies in a megacity region. It aims to obtain an expanded understanding of the spatial characteristics of residential solar PVs in an urban area with a multitude of apartment complexes, such as Seoul, South Korea.
As the national capital, Seoul leads an effort to expand the generation of renewable energy in South Korea. Since 2012, “One Less Nuclear Power Plant” has become a representative energy policy of Seoul [
7]. According to deeper citizen participation, the policy attempts to reduce energy consumption and increase electricity self-sufficiency by using renewable energy. In particular, Seoul declared the policy of “Solar City Seoul, 2022” in 2017, which intends to distribute solar panels to one million households (65% for mini-solar PVs installed on apartment balconies and 35% for rooftop PVs) by 2022.
At this juncture, with the case of Seoul, this study aims to analyze the adoption and diffusion of mini-solar PVs using annual panel data of 2014–2017. This empirical analysis focuses on peer effects while controlling for demographic, economic, built environment, and socio-political variables. It specifically has three objectives: (i) to typify the distribution pattern of the mini-solar PVs, which tends to be dispersed from neighboring previous adoptions; (ii) to explain the characteristics of apartment complexes with mini-solar PVs; and (iii) to investigate the proximity between new and previous adoptions in relation with peer effects “within” and “between” apartment complexes. This study will eventually provide implications for explaining the applicability of the renewables and spatial characteristics to diffuse them in an urban context.
2. Literature Review
Rogers [
8] argues that common rules are found among those who accept innovations, that is, diffusion of innovation occurs through the communication channel for the generation and sharing of information among the members of a similar social system. Rai and Henry [
9] highlight the importance of communication strategies with the rate and scale of adoption. In this respect, the individual’s behavior was found to be under the control of the social norm, which has been called social interaction, peer effect, imitation behavior, and herd behavior in the literature [
5,
10].
Regarding the diffusion of residential solar PVs (RSPVs), previous studies focused on the identification of the characteristics of the space in which peer effects are present and on the discovery of which attributes affect residential adoption. Bollinger and Gillingham [
3] were the first to confirm the peer effect of adoption of RSPVs. They found that in California, 0.78% of the possibility of a new adoption in a ZIP code is raised by one additional installation in that area. They further argued that the addition has a greater effect on larger households and commuters of more than a 30-minute distance. Regarding the peer effect, household size and commuting distance both increase the chance to see solar PV installations on the street.
Several UK studies analyzed the peer effect using data on solar PV adoption on the postal code scale. Richter [
11] reported that the social interaction effect is significant, albeit modest, and Snape [
12] presented that areas with similar solar PV installation capacity are clustered in space and the adoption rate increases over time. Likewise, in Germany, Dharshing [
5] conducted spatial regression and confirmed clustering at the county level.
From a different perspective, some studies analyzed how the distance between neighborhoods influence the adoption of solar PVs. Graziano and Gillingham [
13] conducted a panel analysis of a time series dataset of solar PV adoptions in Connecticut to evaluate how the spatial distribution for RSPVs is affected according to the buffer distances of 0.5 miles or less, 0.5–1 mile, and 1–4 miles at the census block group level. They found that about 0.44 PV systems on average are added by a new installation within 0.5 miles, suggesting the peer effect. Müller and Rode [
14] analyzed the physical distance between the locations of solar PVs using a binary panel logit model in Wiesbaden, Germany. The distance exerted a strong negative effect on the social interaction. Rode and Weber [
15] examined the peer effect through an epidemic diffusion model based on a dataset of RSPVs for 576,000 households in Germany. They selected raster widths of 0.5 km, 1 km, 4 km, and 10 km to explore the spatial effect and subsequently found that the effect is significant in the distance of 1 km and a larger variation is explained when the model considers spatial homogeneity.
While the adoption of solar PVs has been studied, focusing on spatial distribution, some of the studies may further expand our understanding at the individual/household level. Rai and Robinson [
16] specified multiple regression models based on survey data of RSPV adopters to analyze perception, informational gaps, and technological uncertainties. They tested various hypotheses regarding PV adopters’ decision-making and found the significant effects of neighboring PV installations, experiences with leased products, and contacts with previous adopters of RSPVs. A later study by Rai and Robinson [
4] developed an agent-based model of RSPV adoption and analyzed a household-level dataset of PV adopters in Texas. As a result, social networks between homogeneous individuals or groups were found to strongly affect each other. Recently, Rai and Henry [
9] recommended an agent-based model as a tool for analysis of the complex energy demand in relation to social influences and spatial constraints.
4. Results
4.1. Descriptive Statistics
Table 1 presents the descriptive statistics of all research variables. They appear to have enough variations for statistical inference.
Table 2 shows the annually new and accumulated adopters of mini-solar PVs. The mean of the new adopters increased to 5.79 in 2017 from 0.74 in 2014, denoting about an eight-time increase in 3 years. The accumulated number of households also increased by 12.8 times. The mean electricity consumption of each apartment complex qA log 2.572 (converted to 373.25 kWh) and was higher than the mean monthly household electricity consumption: 305 kWh in 2014, 304 kWh in 2015, and 309 kWh in 2016 [
44]. As for peer effects, the numbers of adopters within multi-buffers were as follows: 20.86 adopters in the 500 m band, 49.55 in the 1000 m band, and 85.62 in the 1500 m band.
4.2. Results Based on the Entire Sample
Table 3 presents the overall results of Model 1, which was based on the entire sample. Due to multicollinearity, the model excluded the variable of the number of adopters in the 0–1500 m radius (
Figure 6).
As a demographic characteristic, age was classified into the groups of 20–39, 40–69, and equal to or more than 70 years. The last 70+ age group was significant. Among the economic variables, subsidy from district governments, mean electricity consumption in the apartment complex, and the unit price of the property turned out to be significant. The subsidy by district governments also positively affected PV adoption. Lastly, the lower the mean electricity consumption and the unit price of the property, the more mini-solar PVs tended to be installed.
Regarding socio-political variables, for apartment complexes participating in the energy community program, the rate of mini-solar PVs installations increased by 12.5 times (≈ e^2.526). Additionally, district governments whose leaders were affiliated with the democratic party, entitled the “Democratic Party of Korea”, tended to achieve more mini-solar PVs in their apartment complexes.
Among the built environment variables, younger apartment complexes were equipped with more mini-solar PVs. Additionally, the greater the number of houses in the complex, the more mini-solar PVs it was likely to be equipped with.
As for peer effects, the density of households with mini-solar PVs had a positive impact on the number of households with new installations. The model also showed a significantly positive coefficient for the 500 m radius band. Meanwhile, for a step width of 0–1000 m the coefficient turned out to be negative, suggesting that the number of adopters sharply reduced for a distance between 500 and 1000 m. That is, in Seoul, mini-solar PVs installed in apartment complexes were clustered at a radius of about 500 m.
4.3. Energy Community
The Seoul energy community program comprises various activities for increasing energy citizenship of its residents, which was hypothesized to encourage them to adopt mini-solar PVs. In order to test this hypothesis, this study compared differences in analytical results between a group of 10 district governments with apartment complexes participating in the program and the other 10 not participating. Models 2 and 3 are based on the first and second groups, respectively.
As such, this study found that peer effects were quite different between the two groups. Model 2 presented similar results with Model 1. The density of households with mini-solar PVs and the number of adopters within the 500 m radius band had a positive effect on the new adoption. A 1% increase of the density of adopters increased the possibility of the adoption by 22%, and one adopter increase within the band led to the adoption by 0.16%. In contrast, based on Model 3, while the density of adopters had a positive effect on the new adoption, the number of adopters within the 1000 m radius band had a negative effect. This implies that districts without apartment complexes participating in the energy community program may be capable of delivering mini-solar PVs within each of the apartment complexes but the between-complex effect within the 500 m radius is not significant, while, if the radius exceeds 500 m, the adopters may rather decrease in number.
4.4. Hot/Cold Spots
In
Table 3, Models 4 and 5 show analytical results based on Local Moran’s
I to identify a spatial clustering of mini-solar PVs based on the density of adopters by block group in 2017. For local indicators of a spatial associated (LISA) relationship, we employed a spatial weight matrix according to the inverse distance spatial relationship, which was statistically significant, since the value of Local Moran’s
I was 0.0186 (
p < 0.001). A positive value denotes that clustering exists.
A clustering of hot spots was present in the northern (Dobong and Nowon Districts), northwestern (Eunpyeong District), and northcentral (Dongdaemun and Seongbuk Districts) parts of Seoul. In comparison, cold spots were largely distributed in the middle section of the city, particularly in the south side (Gangnam and Seocho Districts) and north side (Jongro, Jung, and Yongsan Districts) of the Han River.
Figure 7 shows the clustering of hot and cold spots in terms of the density of mini-solar PVs in Seoul. In this regard, we specified two models, each of which consisted of apartment complexes in the hot and cold spots.
Based on the hot spot subsample, Model 4 delivered quite different results from the other models. The variables of the 40–69 age group, the residential area, and the ICLEI membership were newly found to increase the adoption of mini-solar PVs. Additionally, apartment complexes that engaged in the energy community program had an 11.76-times higher demand for mini-solar PVs than those that did not engage. Regarding peer effects, a 1% increase in the density of adopters raised the new adoption rate by 20.66% and one adopter increase of the number of adopters increased by 0.13%. Notably, the peer effect was not significant in Model 5, that is, on apartment complexes in cold spots. The density of adopters also turned out to be insignificant.
In both models, the incentive, unit price of the property, and number of households were found to be significant. A 10,000 Korean won (approximately 10 USD) increase in incentives resulted in a higher possibility of PV adoption by 23.4% and 17.93% for Models 4 and 5, respectively. On the other hand, the unit price of the property was negatively associated with mini-solar PVs: Every 1,000,000 Korean won reduction in the apartment complex land price led to a 26.21% and 0.054% increase in mini-solar PVs for Models 4 and 5, respectively.
5. Discussion
Seoul declared the policy of “Solar City Seoul, 2022”, which aims at achieving “solar power capacity of 1 GW” and “solar powered houses of 1 million” by 2022. As part of this policy, Seoul plans to supply mini-solar PVs to apartments, specifically to 605,185 households by 2022 (an apartment is a dominant type of housing in this hyper-dense city). Currently, incentives for the mini-solar PV installation cover over 60% of the total cost, which will eventually impose a considerable burden on the local and city governments. For this reason, alternative policies have been suggested and studied to reduce the subsidy, but at the same time, concerns were raised, since the reduction may hamper the spread of mini-solar PVs [
18,
19]. At this juncture, focusing on peer effects, this study investigated different types of variables that possibly affect the delivery of mini-solar PVs.
Regarding age, a low proportion of the 70+ age group caused a higher adoption rate of mini-solar PVs, echoing the findings of Kellstedt et al. [
45] and Mills and Schleich [
46]. Additionally, the 40–69 age group was affected by the adoption of mini-solar PVs in hot spot areas. Likewise, in the previous studies [
23,
24,
25], residents in their 40s were found to have a higher tendency to install PVs. While the potential for the electricity generation of rooftop solar PVs is the highest, from 12 to 2 pm [
47], and the electricity cannot be saved and should be consumed immediately, elderly people and middle-aged full-time homemakers tend to stay longer during the daytime. Thus, this age-related result may suggest a better way of yielding more household cost savings through customized PV policies.
The mean electricity consumption and the unit price of the property (a proxy for the household’s income) had negative effects on mini-solar PV installations. Indeed, recent studies [
3,
14] argued that low-income households are more likely to install solar systems. Possibly, affluent households have a lower interest in mini-solar PVs because they can afford energy costs.
Peer effects turned out to be meaningful in distributing mini-solar PVs, which extends the findings of several previous studies [
3,
13,
14,
15,
48]. Janssen and Jager [
49], which found that an observability of innovation has a positive impact on the adoption of green products and those who have already accepted a technology have a higher tendency to accept new technologies. In relation to social desires, consumers feel satisfied by consuming the same product as their neighbors and occasionally participating in social comparison and imitation. If the density of mini-solar PVs is high in their neighborhood, people may regard mini-solar PVs as a socially acceptable or commonly applied norm and subsequently decide to install the PVs. Frederiks et al. [
39] called this a normative social influence.
In terms of the mini-solar PV distribution, peer effects have been identified “within” and “between” apartment complexes. The “within” effect was stronger. This may be because interactions between neighbors usually occur in the unit of the apartment complex, including energy community, as suggested in this study. These interactions would encourage non-adopters to install mini-solar PVs through word-of-mouth. With regard to the participation in the energy community, in Models 2 and 3, each of which consisted of 10 districts with and without apartment complexes in the energy community program, mini-solar PV adopters were clustered within a 500 m radius in areas in which most apartment complexes engage in the program, while a cluster was not formed in areas without such a complex. This implies that the participation in the energy community program assumes a more important role in spreading mini-solar PVs, which is in line with a finding on the peer effect “within” apartment complexes. That is, the word-of-mouth effect is likely to be stronger than the visual effect, supporting a mixed methods study conducted in Sweden [
50].
Regarding the visual effect, peer effects on the installation of mini-solar PVs were significant within a radius of 500 m between apartment complexes. The distance of peer effects appears to differ by region: 804 m (0.5 mile) in Connecticut [
13], 1 km in Germany [
15], and 610 m in Texas [
4]. That is, the distance was found to be shorter with the case of Seoul, partially because, in this hyper-compact city, a neighborhood with a certain number of households is usually defined as a shorter/smaller distance/area.
Meanwhile, membership in ICLEI was significant only in Model 4, which used a hot spot subsample. Notably, as shown in
Table 3, the model did not control the effect of the political affiliation. Thus, this suggests that the ICLEI membership has an effect, albeit modest, on mini-solar PV adoption in those areas.
Younger apartments were more likely to have mini-solar PVs. While the average life span of mini-solar PVs is currently 15–20 years, the longer they are used, the higher their economic benefits. Thus, those living in older apartments may be less interested in installing mini-solar PVs, since the apartments could be shortly reconstructed and about 100 dollars are required to relocate mini-solar PVs. In Seoul, if the age of an apartment complex reaches 30 years, it can be officially reconstructed, which is arguably among the most profitable financial investments. In this sense, it may be desirable to make it financially and/or technologically easier to transfer mini-solar PVs.
6. Conclusions
This study aimed at analyzing different factors on the adoption of mini-solar PVs in apartment complexes in Seoul. Particularly, we hypothesized that peer effects occur around apartment complexes in which mini-solar PVs are actively installed. Peer effects are exerted in the form of word-of-mouth by innovative early adopters and the visual effects of mini-solar PVs that are installed nearby. In particular, a social desire to gain satisfaction by following the innovative behavior of neighbors is considered the main driver for installing the PVs when their PV adoption rate is high and the PVs are installed in adjacent complexes. In order to evaluate these peer effects, this study employed four variables: the density of households with mini-solar PVs up to the preceding year in each apartment complex and the number of adopters within the 500 m, 1000 m, and 1 500m radius bands.
The major findings are that more mini-solar PVs could be installed in an apartment complex if the proportion of aged residents (age 70+) is low and the mean property value is low and when the district head is politically affiliated with the Democratic Party of Korea. Additionally, peer effects turned out to be significant: Specifically, the within-complex effect (word-of-mouth effect) is likely to be stronger than the between-complex effect (visual effect) in an urban context. Regarding the visual effect, new installations were positively affected by the density of previous mini-solar PVs and within a 500 m radius of those complexes in which mini-solar PVs are installed.
Participation in the energy community program was another factor on the mini-solar PV installation. The program consists of practical activities for energy education and conservation. Thus, it possibly increased the awareness of energy citizenship and positively affected the adoption of mini-solar PVs.
The Korean government helps to improve the regional electricity self-sufficiency by expanding the distributed generation, such as small-scale solar PVs. In 2017, the proportion of the distributed generation was only 12% but is planned to be expanded to 30% by 2040 [
51]. Mini-solar PV installation in apartment complexes can be used as a means for electricity self-sufficiency in urban areas. It allows for the production of electricity in residential areas and raises the resident’s energy consciousness. While public subsidies for the apartment PV installation are soon to be exhausted, local governments are recommended to facilitate social interaction/peer effects for the purpose of introducing new adopters of mini-solar PVs in their jurisdictions.