Interpretation of the analysis results
The analyses conducted in this study found that, in addition to economic factors, a lower population density, high proportion of young people, larger number of household members, and fewer years living at the same address were demographic and living environmental factors that enhanced the diffusion ratio of PV systems. These findings are generally consistent with those of previous studies targeting other regions. Nevertheless, while the positive influence of a larger proportion of young residents on diffusion is in line with the results by Drury et al. (2012), it contradicts those of Dharshing (2017), where young people were defined as less than 20 years old. A deep observation of the age structure groups in census blocks with broad diffusion of PV systems indicates a higher population proportion in their 30s and early 40s that are parents rearing children. Thus, we can interpret the results to be consistent with the explanation by Kwan (2012), referring to Pickett-Baker and Ozaki (2008), that in areas where the population is composed largely of those aged 35–45, there is a relatively high installation rate, as this age group has greater motivation to purchase green products.
While the panel analysis in Sect. 2 considered an independent variable expressing residents’ income level, the spatial econometric analysis in Sect. 3 did not address this factor due to the unavailability of data at the census-block level. However, some demographic attributes considered as independent variables in the latter analysis could be associated with income level to a certain extent. Actually, a correlation analysis indicated that a ward with relatively high-income level tends to have a higher proportion of young people, a lower proportion of elderly people, and a lower proportion of residents of 20 years or more (see the Appendix 2). Therefore, we may interpret that the positive and negative impacts of a higher proportion of young people and a longer duration of residence, respectively, arise partly from high- and low-income levels, respectively.
The spatial econometric analysis also indicated a potentially positive spatial spillover influence on diffusion in a census block through lower ratios of detached houses and lower population densities in nearby census blocks. Census blocks with lower ratios of detached houses, and accordingly higher ratios of multiple dwellings, yet with lower population densities, suggest that such census blocks contain a station or arterial street together with a surrounding commercial district. That is, conveniences of daily life such as transportation and shopping provided by the surrounding environment seem to affect diffusion. Relatively wealthier people are assumed to gather and live in residential areas located near a station and commercial facilities, considering the potentially high land price of such areas (Ministry of Land, Infrastructure, Transport and Tourism 2018). We cannot state this with certainty, however, since this spatial econometric analysis did not deal with economic factors directly.
The analysis detected a statistically significant neighbor effect only for census block contiguity distances of ≤ 500 m, and ≤ 1000 m, and distances in between. These distances fall within the range implied by the three prior studies that targeted all of Germany (Rode and Weber 2016; Rode and Müller 2016) and Connecticut in the US (Graziano and Gillingham 2015), which analyzed the distance significantly influenced by the neighbor effect. It is still worth noting that the present study yielded similar distances by applying the SDM while simultaneously considering the spatial spillover of social attributes in Kyoto City, where half of census blocks are densely populated, with more than 10,000 people per km2. Rai and Robinson (2013), who investigated the neighbor effect in detail through surveys of residents, broke down the effect into a passive effect through witnessing PV panels being installed by neighbors, and an active effect through word-of-mouth communications with neighbor PV system adopters. The distances detected in this study seem to be consistent with their finding.
In other studies attempting to examine the neighbor effect in Japan, the municipality-level data analysis by Seki et al. (2013) verified that, in explaining the diffusion ratio in a municipality, adopting the mean diffusion ratio within the range of 30 km from the municipality as an independent variable improved the explanatory power; meanwhile, the prefecture-level data analysis by Seki et al. (2014) observed no significant neighbor effect of PV diffusion. Considering the more than tenfold difference in the range of the distance to define neighbors between that study and ours, the neighbor effect they verified might have been the result of other mechanisms than those confirmed in this study. Targeting only Kyoto City and covering an area of 828 km2, we were unable to detect such a long-range neighbor effect.
Implications from questionnaire surveys in Kyoto City
Here, we attempt to provide more background and greater context for our analysis by analyzing the results of two questionnaire surveys regarding citizens’ installation of PV systems in Kyoto City.
(1) Survey targeted at PV installers
Kyoto City government conducted questionnaire surveys targeted at the installers of PV systems subsidized by the city. The Global Environment Policy Office, where one of this study’s authors (TT) is employed, was charged with implementing the surveys. The survey conducted in FY 2013 obtained 891 responses out of 1580 questionnaires distributedFootnote 6 (Global Environment Policy Office of Kyoto City 2013).
We extracted and sorted the respondents’ answers to a multiple-choice questionnaire regarding their incentive to install PV systems.Footnote 7 As shown in Table 3, a high proportion of responses indicated that a house manufacturer was their incentive, as the responses included individuals who live in newly built houses where PV systems were preinstalled. In principle, the choices presented should not be considered independent of each other; in all likelihood, a person would not pursue PV installation only based on a single incentive, but rather on a hierarchical series of decision-making steps driven by multiple incentives. Although this study did not investigate further the decision-making mechanisms, Table 3 does indicate that communication with a peer (i.e., information from a peer or PV installation by a peer) constituted an influential incentive to a certain degree. This fact could be an explanation for the neighbor effect detected in Sect. 3.
(2) Attitude survey from citizens
A basic questionnaire survey on environmental behavior conducted by a study group on global warming countermeasures in Kyoto City in 2015 (Global Environment Policy Office of Kyoto City 2015a, b) is another useful source of information. The survey sought to assess overall trends in environmental awareness and the activities of citizens and was not limited to PV installers. One of the current study’s authors (TK) was involved in designing the questionnaire and another (TT) was engaged in processing the data. The survey was administered to a random sample of citizens; 3000 questionnaires were distributed, and there were 1058 responsesFootnote 8 (Global Environment Policy Office of Kyoto City 2015b).
Using the survey response data, we performed a correlation analysis between the set of answers regarding respondents’ intentions to install PV systems in their homesFootnote 9 and answers to five sets of questions regarding: (1) installations by respondents’ relatives, neighbors, and business peers; (2) the frequency of conversations about global warming and energy-saving behavior among respondents’ family members who lived together; (3) the frequency of similar conversations with respondents’ child(ren); (4) the frequency of such conversations with respondents’ neighbors and/or business peers; and (5) the frequency of such conversations with respondents’ friends.Footnote 10 Answers were selected on a five-point Likert scale, but we computed a Pearson product-moment correlation coefficient treating the Likert scale as an interval scale.
Table 4 shows the correlation coefficient obtained. The results confirmed a positive, statistically significant correlation between respondents’ intentions to install and every one of the factors listed above. This suggests that the installation of PV systems near a respondent’s home and conversations with family members influenced the respondents’ intentions to install. We should note that, however, the intention does not necessarily lead to PV installation, owing to a so-called intention–behavior gap.
It is worth mentioning a finding from a prior analysis by the Global Environment Policy Office of Kyoto City (2015b) that used the same survey data and indicated that the frequency of conversations on global warming and energy-saving behavior was especially high in households with a child(ren) in the fourth and fifth grades. This potential influence of children is likely at least partly attributable to environmental education activities in elementary schools. In fact, as of FY 2005, Kyoto municipal elementary schools established a program to prompt children to contribute to global warming countermeasures through lessons followed by assignments to put countermeasures into practice at home, including but not limited to keeping a household environmental account book (Global Environment Policy Office of Kyoto City 2014). We concluded that parents’ conversations with offspring could positively influence their intention to install a PV system. This can serve as another explanation regarding the cause of the positive relationship between the proportion of young people and the diffusion ratio yielded by the analysis in Sect. 3.
This analysis revealed a tendency for steady PV system deployment to occur in areas convenient for daily life, characterized by ample living environments with low population density, and a large proportion of multiple households raising one or more children. Nevertheless, even in such areas, the diffusion ratio is still below 10%, except in a few census blocks. To further accelerate diffusion, it would be effective to provide information to promote system diffusion selectively to the residents in areas expected to have a greater diffusion potential, especially districts located in the centers of these areas. Given the questionnaire results that promotion by PV dealers increases the opportunities for residents to consider system installation, we recommend providing residents with information from the government on the subsidy and other advantages of PV installation through dealers. Considering a potential indirect influence of promoting communications among residents with children in local communities within an area of a few hundred meters squared, other effective measures could include the activities of organizations such as the Solar Community Organization described by Noll et al. (2014) to facilitate PV installation (Palm 2016), and strengthening communications to children in school through energy education.
We also suggest that policy makers identify the many areas with no PV diffusion and without effective neighborhood influences and investigate their barriers to diffusion. Taking a longer view, we would recommend considering policy measures to extend the areas with demographic and living environmental conditions suitable for the diffusion identified by our analysis. Such policy measures might include supporting schemes for families with young children to acquire detached houses or to live in detached houses owned by their parents (grandparents for the young children) with necessary home renovations, thereby extending residential areas with detached houses occupied by families with young children and more household members. Such policy measures would need to be carefully examined beyond the specific issues of PV deployment.
The spatial statistical analysis also implied that certain social conditions tended to limit PV diffusion, especially in densely built areas with small detached houses. Such areas include zones highly vulnerable to disasters such as earthquakes and fire. We expect future opportunities for living environment improvements, as disaster prevention measures would also facilitate diffusion.