Spaces Eliciting Negative and Positive Emotions in Shrinking Neighbourhoods: a Study in Seoul, South Korea, Using EEG (Electroencephalography)

Abstract

Although shrinking neighbourhoods are places where urban citizens experience negative emotions, some evidence suggests that people in some shrinking neighbourhoods feel less negative emotions than in other areas. Nevertheless, empirical studies that analyse environmental and personal elements that affect people’s emotions in a shrinking neighbourhood remain insufficient. This is rather surprising, considering an increasing interest in the effects of negative emotions on individuals’ health. Thus, this study used electroencephalography (EEG) to examine the impacts of environmental and personal characteristics on people’s emotional levels in a shrinking area of Seoul, South Korea. A multilinear regression model was used to analyse emotional valence levels between sites with different urban designs and management levels. The results revealed that people felt positive emotions at sites where both urban design factors and their management were both satisfactory at appropriate levels. The results also found that people who had lived or worked in the neighbourhood for a long time and were women experienced more positive emotions than visitors and men. This finding implies that a shrinking neighbourhood can maintain a sense of satisfaction as long as the area is carefully managed. Revealing the emotional effects of environmental and personal characteristics in a shrinking neighbourhood can be used for planning practices and policy-making to create healthy and liveable urban neighbourhoods.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11524-022-00608-8.

Introduction

A shrinking area in a city is widely regarded as a place where people feel intense negative emotions. Emotion is an immediate physiological response to perceived events or stimuli and is distinct from feelings; the conscious experience of emotions and physical sensations (e.g. hunger, pain, and pressure); or moods, a mix of feelings, and emotions that last for hours, days, or even months (e.g. gloomy and romantic). As noted in previous studies, continuous depopulation and job loss in shrinking communities trigger the following issues: a rise in poverty for individuals and a community [1], a reduction in opportunities for building social capital and social networks [2, 3], an increase in vulnerability to crime, and even the creation of a strong neighbourhood stigma involving the area being viewed as abandoned and unwanted [4]. These circumstances result in people feeling negative emotions (e.g. anxiety and fear) about living in or visiting such neighbourhoods.

However, some evidence suggests that people do not always feel unpleasant emotions in shrinking neighbourhoods. This argument is supported by studies examining the relationship between urban shrinkage and happiness in 38 major US cities and 439 districts in Germany [5, 6]. The studies found no statistically significant differences in citizen happiness between shrinking and growing cities. This finding implies that there are some shrinking neighbourhoods where people feel less intense negative emotions than in other shrinking areas.

Despite this contrary discussion, the causes influencing citizens’ emotional differences between shrinking areas are little known. This is rather surprising, given that negative emotions critically affect a citizen’s physical and mental health. As proven in previous studies, daily exposure to negative emotions has a direct, negative effect on cardiovascular activity and the immune system [7]. Moreover, citizens’ emotions of anxiety and fear associated with public spaces restrict physical activities and social interactions, further evoking loneliness and social isolation, significantly associated with an increased risk of early mortality. [8]

Previous psychological studies have investigated the emotional impacts of various factors that can occur in shrinking urban areas, such as economic crisis [9], a decrease in social capital [10], a lack of public infrastructure [11], political unrest and restrictions [12], and even religious activities [13]. Nevertheless, empirical studies examining the emotional effects of specific environmental and personal characteristics reported in shrinking communities still remain insufficient.

Therefore, this study aimed to analyse the impacts of environmental and personal characteristics in a shrinking area on individuals’ emotional levels. We adopted electroencephalography (EEG) to detect emotions, regarding an increasing interest in measuring people’s psychological data in response to different environments [14, 15, ¹⁶]. We measured 45 participants’ emotional valence levels in 12 different spots in Nangok-dong, a shrinking area in Seoul, South Korea. We categorised the experiment spots into 7 sites based on their environmental characteristics and used a multilinear regression method to identify differences in the emotional levels. Through these processes, this study addressed the following question: What environmental and personal characteristics affect individuals’ negative or positive emotions in a shrinking neighbourhood?

This paper is structured as follows: Sect. 2 discusses the theoretical and empirical studies on how environmental and personal features affect individuals’ emotions in shrinking areas and the EEG methods used in previous studies to analyse emotional levels; it also introduces research hypotheses; Sect. 3 introduces the personal characteristics of the test subjects, the experiment sites and their environmental characteristics, data collection, and analysis methods; and Sect. 4 presents the results of our analysis model to answer the research questions. This paper culminates in discussing this study’s major findings, implications, limitations, and suggestions for future research.

Background and Context

Emotions and Environmental Characteristics

The present study categorises the environmental features that influence individuals’ emotions in a shrinking neighbourhood from two angles: urban design (e.g. green spaces and inactive frontages) and their management (e.g. management of abandoned houses and vandalism). On the one hand, studies on urban planning and environmental psychology have documented urban design features that affect people’s emotions. Overall, the literature suggests that natural environments, pedestrian-friendly designs, mixed land-uses, and promoted social interactions are the core characteristics for generating positive emotions, experiences, and pedestrian satisfaction. For example, according to the book City: Rediscovering the center, physical elements—trees, open spaces, attractive buildings and street designs, mixed-use developments, and social interactions—should be regarded as important for higher user satisfaction in public spaces [17]. Similarly, the book Great Streets underlined the importance of built environment attributes for better pedestrian experiences, such as trees, diverse designs and usages of buildings, many doors and windows, and public spaces [18].

The emotional impacts of urban design have been examined in other studies. For instance, a study has reported a statistically significant association between active building with good street design (e.g. a front entrance) and stress, a psychological and physiological response that causes negative emotions [19]. A survey-based study also discovered a strong link between urban open spaces and stress [20]. Another study using the EEG method reported that test subjects felt more anxiety when viewing traffic scenes than when viewing green urban settings [21]. Similarly, there is evidence that natural scenes positively influence the person’s psychophysiological states [22].

On the other hand, previous studies on urban shrinkage have focused on environmental management as a critical factor affecting citizens’ emotions. As stated in previous studies, shrinking communities face diverse issues, such as depopulation, job loss, increased crime vulnerability, weakened social capital, and inhabitants’ further-reduced willingness to manage their personal and community environments [1, 2, 23]. The breakdown of social control in a community causes an increase in the neighbourhood’s physical disorder (e.g. vandalism, abandoned houses, and illegal dumping) and social disorder (e.g. drunkenness and fighting and arguing in public spaces). Previous studies have highlighted that the proliferation of physical and social disorders is a significant environmental factor that causes people to feel a negative mood, such as depression, as well as stress that causes negative emotions, such as fear and anxiety [24, 25].

Well-managed environments in shrinking areas can improve people’s emotions and experiences. Previous studies found a statistically significant association between Crime Prevention Through Environmental Design (CPTED), which is a design strategy to create safer neighbourhoods, and an increased sense of physical safety and psychological comfort [26, 27]. Accordingly, well-managed interventions, such as the attractive design of fences and street lights, have been widely applied in shrinking areas as a strategy to reduce people’s negative emotions of fear and anxiety.

Emotions and Personal Characteristics

Previous studies suggest that age, residence experience, and gender affect the degree of negative emotions in shrinking areas. Increased job losses push young people to move away from a shrinking neighbourhood. Compared to the young, older people felt higher residential satisfaction in the same place [28]. They also revealed that a negative relationship existed between the length of living in a place and their satisfaction [28]. This finding raised a discussion that longer residences in a shrinking area raise the inhabitants’ awareness of shrinking, which lowers residential happiness. Furthermore, there is evidence that females have a higher sense of place attachment than males [29]. Given a significant association between positive emotions and place attachment, this result implies that females may feel more positive emotions in a shrinking neighbourhood than males [30].

Emotion Classification, Measurement, and EEG

Emotion classification studies have been approached from one of two perspectives: discrete or dimensional models. Discrete emotion models argue that each emotion is a fundamentally distinct construct with a specific representation [31], whereas dimensional emotion models propose that an emotional state can be described as a point in a multi-dimensional space [32]. In particular, among various dimensional models, the currently dominant one is the two-dimensional model, such as the circumplex and vector models. Most of these dimensional models contain a significant dimension of valence that determines a specific emotion as negative or positive.

The discrete model has been criticised on some points. One is that people rarely feel the “full-blown” basic emotions predicted by the discrete emotion model (e.g. joy, sadness, and fear), and there are many “borderline” emotions that the discrete model cannot explain in reality [3]. This argument is supported by a study that found the discrete model’s much lower reliability in categorising emotionally ambiguous examples than the dimensional model [33]. Regarding the issue, some authors favour the idea that emotional response measurements reflect dimensions rather than discrete states [34].

Psychophysiological studies based on spatial distribution analysis of EEG signals have mostly focused on the frontal asymmetry of the two cerebral hemispheres to detect the valence level. The theory of emotional valence posits that the right hemisphere is dominant for negative emotions in terms of EEG signal amplitude, whereas the left hemisphere is dominant for positive emotions [35, 36, ³⁷]. This idea has been widely embraced and expanded in previous studies that examined subjects’ emotions by comparing the inactivation of the two frontal hemispheres in response to stimuli [38, 39, ⁴⁰]. Studies have stated that inactivation of the left frontal lobe is linked to negative emotions, while activation of the right frontal lobe is associated with positive emotions [41, 42, ⁴³]. The inactivation of the two frontal hemispheres can be used by analysing EEG signals. For example, high alpha activity (8–12 Hz on the EEG frequency band), which is associated with relaxed states, and low beta activity (13–30 Hz on the EEG frequency band), associated with stress and anxiety, are both indicators of frontal inactivation. ^{44, 45, 46} F3 and F4 are prefrontal cortical positions that are important in emotion regulation and cognitive function control; hence, they have been widely used to assess EEG signals. Given these principles, previous studies have analysed people’s negative or positive emotions by calculating their valence levels as in Eq. (1) [47, 48]:

$$Valence=\frac{\alpha \left(F,4\right)}{\beta \left(F,4\right)}-\frac{\alpha \left(F,3\right)}{\beta \left(F,3\right)}$$ 1

where $\alpha \left(F,4\right)$ represents the alpha-wave power extracted in the F4 position of the right frontal lobe and $\beta \left(F,3\right)$ is the beta-wave power extracted in the F3 position of the left frontal lobe (see Fig. 1).

Fig. 1.

Electroencephalography (EEG) equipment and electrode locations: a a portable EEG device, EMOTIV EPOC + ; b F3 and F4 electrode locations for the valence measurement

Research Hypotheses

Based on the findings of literature reviews, we propose research hypotheses as follows:

• Hypothesis 1: Emotional valence levels in open spaces with trees or active social interactions would be positive, whereas those in heavy traffic volumes or inactive frontages would be negative in a shrinking neighbourhood.

• Hypothesis 2: Emotional valence levels amid abandoned houses, vandalism, or a central point of people loitering or smoking would be negative, while emotions in sites with installed CPTED strategies would be positive in a shrinking neighbourhood.

• Hypothesis 3: Emotional valence levels of older people, residents, and females would be more positive in a shrinking neighbourhood than younger people, visitors, and males.

Setting and Methods

Research Methodology Process

This research method comprises 5 steps. First, we selected 12 spots in which the environmental characteristics covered in the literature reviews were intensively located in a shrinking neighbourhood. Second, we collected the raw EEG data of each participant in the spots. Third, we calculated the valence levels with the refined EEG data, showing the degree of negative or positive emotions. Fourth, we classified the 12 spots into 7 sites based on a similarity of their environmental qualities. Fifth, we conducted a multilinear regression to compare the differences in emotional levels between the 7 environmental conditions. After these steps, we analysed the results to suggest implications for the planning of shrinking neighbourhoods.

Participants

A total of 45 participants randomly recruited via online and field promotions participated in the experiments (42% male and 58% female) (see Fig. 2). Their average age was 31.6 years (S.D. = 12.917). The participants consisted of 20 insiders living or working in a community and 25 outsiders who had never visited the site before. The participants’ principal occupations were students (55%), white-collar workers (20%), and other occupations (30%), including those with part-time employment and homemakers.

Fig. 2.

Pictures of experiments and a recruiting poster

Before each experiment, all the participants were informed about the study (e.g. the experiment’s purpose and procedures, their right to not participate or to drop out, the privacy of their data, and the data-disposal method and plan). Only participants who signed an informed consent form could be involved in the EEG experiment; we have kept the signed consent forms.

Study Area and Context

As the study area, we selected Nangok-dong, a major shrinking neighbourhood located the Gwanak district of Seoul, South Korea. Demographic and economic statistics indicate urban shrinkage in Nangok-dong. According to Seoul Statistics, from 2000 to 2020, there was a continuous decrease (21.34%) in the area’s population, from 35,665 to 28,054 [49]. The number of businesses also showed a similar pattern, reducing by 14.97%, from 1,122 in 2006 to 954 in 2019. In Nangok-dong, the percentages of residents over 65 years old (18.85% in 2020), with a disability (5.4% in 2017) and recipients of National Basic Living Security (5.0% in 2018) were considerably higher compared to those of Gawank-gu (15.47%, 3.9%, and 2.9%, respectively) and Seoul (15.82%, 4.0%, and 3.0%, respectively). Those of vulnerable social groups lived amid poor environmental management conditions, as evidenced by the percentage of deteriorated houses (16.4% in 2017). This concentration of marginalised social groups and the collapse of commercial functions within the shrinking community have inhibited the inflow of new residents [50].

We selected 12 spots in Nangok-dong in order to perform the EEG experiments. Experiment spot selection was conducted based on direct observations, discussions with local planners and landscape architects, and reference to a government planning report entailing residents’ complaints about the built environments of the community. Because the EEG experiment required a significant amount of time and intense work, from participant recruitment to EEG measurement with the headset, conducting the experiments in many different spots was limited in reality. Instead, we identified the 12 major spots in Nangok-dong that exemplify the environmental characteristics that influence people’s emotions, as reported in literature reviews (i.e. social interactions, open spaces, inactive frontages, social disorders, rubbish, vandalism, abandoned houses, vehicle-oriented environments, and CPTED). These characteristics represent the environmental qualities seen in typical low-rise shrinking residential areas in South Korea, as well as those widely documented in other neighbourhoods around the world, such as Baltimore’s Westside, Tokyo’s Okutama, and Manchester’s Moss Side.

Data Collection

The EEG field experiments were conducted in the 12 spots from March 16, 2018, to May 20, 2018. The EMOTIV EPOC + , a non-invasive EEG headset device, was used for the data collection, as it has 14 sensors that satisfy the 10–20 system, an internationally recognised method to apply the location of scalp electrodes in an EEG study. The raw EEG data were extracted with a frequency range of 128 Hz using the application EmotivPRO v1.8.1.

The participants, while wearing the EEG headset, spent time (e.g. standing and slowly walking) in the spots with the guidance of a researcher who stayed 2 to 3 m behind them. To control the generation of artificial noise from their muscle movements, all the test subjects were instructed not to speak, touch their faces, or move excessively. They were asked to rest for 3 min before beginning an experiment in a different spot. The field experiments were conducted on sunny days, at temperatures between 10 and 22 °C and from 10 a.m. to 6 p.m. on Fridays and Saturdays.

Data Preprocessing

Artificial noise was removed from the raw data using EEGLAB v.14.1.2 and MATLAB R2016a. Raw EEG data can contain extrinsic and intrinsic noise generated by external sources (i.e. environmental noises) and internal sources (i.e. eye, muscle, and heart movements) [51, 52]. To eliminate the noise, we first removed the epoch baseline, which might cause artefacts at the beginning and end of the EEG signal. Then, we applied high and low band filtering (0.1 Hz–30 Hz) in order to remove the most common extrinsic noise [53]. Other channels not investigated in the research were eliminated, except for F3 and F4 channels, because interpolated channels might interfere with the independent component analysis (ICA). Following that, we eliminated significant and noticeable artefacts, possibly generated by muscular movements, by visual inspection. Finally, we used the ICA to remove intrinsic noise, as its effectiveness has been widely proven in previous studies [54 - 56].

Using the refined EEG data, we analysed the emotional valence levels of the 45 participants in the 12 spots (n = 540). Using the fast Fourier transform (FFT) algorithm, we extracted the alpha-wave (8–12 Hz) and beta-wave (13–30 Hz) power at the specific scalp positions (F3 and F4) from the refined data. The extracted alpha- and beta-wave power were normalised by subtracting the participants’ baseline mean power, obtained within the first 20 s before the experiments to examine changes in their valence levels before and during the experiment [57, 58]. The alpha- and beta-wave powers were transformed into valence values according to Eq. (1). As a result, positive valence levels indicate the participants’ relatively positive emotional levels compared to before the experiment, while negative ones indicate their comparatively negative emotional levels.

Clustering

We categorised the 12 spots into 7 sites (see Fig. 3). In the study area, there were spots where multiple environmental characteristics were present in a location, which created an ambiguity issue when analysing individuals’ emotional responses to each environmental condition. Thus, we conducted a cluster analysis using Stata SE 16.1 in order to solve the issue and present one location’s representative characteristics. Two hierarchical clustering methods were used for the cluster analysis: the ward’s linkage method and the average-linkage method using Euclidean distance. Seven groups of environmental characteristics were selected as they showed the highest clustering effectiveness, with a pseudo-F statistic value of 20.73 for both methods.

Fig. 3.

Location and pictures of the study area, experiment spots and sites

Tables 1 and 2 show each site’s predominant environmental characteristics: site 1 for a less-crowded commercial street with active social interaction (SI), site 2 for a less-crowded open space (OS), site 3 for an open space with social disorder (OSSD), site 4 for inactive frontages with CPTED (IFC), site 5 for inactive frontages with vandalism (IFVSD), site 6 for abandoned houses with vandalism (AHV), and site 7 for vehicle-oriented environments (VOEs).

Table 1.

Environmental characteristics of experiment sites

Site	Social interaction	Open space	Inactive frontages	Social disorder	Rubbish	Vandalism	Abandoned houses	Vehicle-oriented environment	CPTED
1	1	0	0	0	0	0	0	0	0
2	0	1	0	0	0	0	0	0	0
3	1	1	0	1	1	0	0	0	0
4	0	0	1	0	1	0	0	0	1
5	0	0	1	0	1	1	0	0	0
6	0	0	0	0	1	1	1	0	0
7	0	0	0	0	0.7	0	0	1	0

1 indicates the presence of each variable in a site, while 0 shows the lack of each variable

Table 2.

Definitions and descriptive statistics of key variables

Variable	Definition	N	Mean (S.D)	Min	Max
Dependent variable
Emotion	Valence level	540	-0.034 (0.307)	-2.303	2.337
Personal characteristics
Residential status	Residents or workers in the community, 1; visitor, 0	540	0.444	0	1
Gender	Male: 1; Female: 0	540	0.494	0	1
Age	Age of participant (years old)	540	31.6 (12.917)	12	65
Environmental characteristics
Social interaction (SI)	Presence of people sitting, socialising and staying amid less-crowded commercial streets (0, no; 1, yes)	12	0.167	0	1
Open space (OS)	Presence of a less-crowded open space (0, no; 1, yes)	12	0.083	0	1
Open space with social disorder (OSSD)	Presence of open space in which children play but teenagers smoke and loiter in a group, and people drink alcohol (0, no; 1, yes)	12	0.083	0	1
Inactive frontages with CPTED (IFC)	Presence of inactive street frontages with high retaining walls and a fire extinguisher, CCTVs, emergency alarms, wall paintings and stickers for crime prevention (0, no; 1, yes)	12	0.167	0	1
Inactive frontages with vandalism (IFV)	Presence of inactive street frontages with high retaining walls and broken street lights (0, no; 1, yes)	12	0.167	0	1
Abandoned houses with vandalism (AHV)	Presence of abandoned houses with broken windows and street lights (0, no; 1, yes)	12	0.083	0	1
Vehicle-oriented environment (VOE)	Absence of pedestrian footpaths on streets with high traffic volume (0, no; 1, yes)	12	0.25	0	1

The mean shows the average level of each variable in the experiment area (12 spots)

Data Analysis

Multilinear regression was conducted to predict the impact of environmental and personal characteristics on the participants’ emotional levels while in the shrinking neighbourhood of Nangok-dong.¹ The dependent variable was the valence level, indicating the participants’ relative positive or negative emotional levels compared to before the experiment; the independent variables consisted of his or her personal characteristics (residential status, gender, and age) and the 7 sites of environmental characteristics. We selected the SI site, which has been widely stated in other studies as an essential factor that promotes positive emotions, as a reference category for estimating the differences in emotional levels between the reference site and the other sites. The model equation is as follows:

$$\begin{array}{c}Y={\beta }_0+{\beta }_1\left(R,e,s,i,d,e,n,t\right)+{\beta }_2\left(M,a,l,e\right)+{\beta }_3\left(A,g,e\right)+{\beta }_4\left(O,S\right)+{\beta }_5\left(O,S,S,D\right)\\ +{\beta }_6\left(I,F,V\right)+{\beta }_7\left(I,F,C\right)+{\beta }_8\left(A,H,V\right)+{\beta }_9\left(V,O,E\right)+ϵ\end{array}$$ 2

where $Y$ is coded as the participants’ average emotional level (i.e. valence level) in the sites; ${\beta }_0$ is constant, the adjusted mean of the SI site; ${\beta }_1,{\beta }_2,{\beta }_3$ represent personal characteristics of resident, male, and age; ${\beta }_4,{\beta }_5,{\beta }_6,{\beta }_7,{\beta }_8,{\beta }_9$ indicate the emotional-level differences between each site and the reference category; and $ϵ$ is the residuals.

Results

Table 3 presents the multilinear regression results. A significant regression equation was found (F(9,530) = 3.48, p = 0.000) with an R-squared value of 0.085, so 8.5% of the variation in individuals’ emotions could be explained by the model containing the independent variables. We also found a nearly medium effect size in the regression model (Cohen’s $f^2$= 0.093).² This model’s data met the assumptions of linearity, collinearity, and homogeneity of variance. Although the standardised residuals were approximately normally distributed in the Q-Q plot, the assumption of normality was not satisfied in the Shapiro–Wilk test. Therefore, robust standard errors were used to fix the normality of this model.

Table 3.

Multilinear regression results for the impacts of personal characteristics and environmental characteristics on individuals’ emotional valence levels

Variable	Emotion (valence level)
	Coef	(Robust S.E.)	p >|z|
Personal characteristics
Residential status	0.058*	(0.022)	0.010
Gender	-0.072*	(0.030)	0.015
Age	0.002	(0.001)	0.111
Environmental characteristics
Open space (OS)	0.004	(0.071)	0.950
Open space with social disorder (OSSD)	-0.117**	(0.038)	0.002
Inactive frontages with CPTED (IFC)	-0.122**	(0.042)	0.004
Inactive frontages with vandalism (IFV)	-0.136**	(0.046)	0.003
Abandoned houses with vandalism (AHV)	-0.140**	(0.042)	0.001
Vehicle-oriented environment (VOE)	-0.179***	(0.043)	0.000
Constant	0.016	(0.048)	0.742

*p < 0.05; **p < 0.01; ***p < 0.001

Regarding environmental characteristics, although there was no statistically significant difference in the emotional valence levels between the SI and OS sites, the results revealed statistically significant differences in the valence levels between the SI site and the other 5 sites, such as OSSD, IFC, IFV, AHV, and VOE. The average valence levels in each site were more negative by 0.117 (OSSD), 0.122 (IFC), 0.136 (IFV), 0.140 (AHV), and 0.179 (VOE) in valence levels than those in the SI site.

Figure 4, the estimated marginal means, shows each site’s emotional valence levels in the shrinking neighbourhood of Nangok-dong. The results indicate the answer to hypothesis (1) and (2), the environmental characteristics affecting negative or positive emotions. On average, the participants felt positive emotions only in 2 sites: OS (margin = 0.080) and SI (margin = 0.075), which is consistent with previous studies [17, 18, 20]. In contrast, the participants, on average, felt intense negative emotions while amid poor levels of urban design or management, such as IFV (margin = -0.061), AHV (margin = -0.065), and VOEs (margin = -0.104); these results support evidence from other studies (see Supplementary Table A1) [19, 21, 24, 25]. Interestingly, negative emotional levels, on average, were also measured in spots where the factors of urban design and management complexly existed, such as OSSD (margin = -0.042) and IFC (margin = -0.046).

Fig. 4.

Predictive marginal means of environmental characteristics with 95% CIs

In terms of personal characteristics, Table 3 also shows the answers to hypothesis (3) related to the effects of personal characteristics on someone’s emotions. Residential status and gender are statistically significant predictors of his or her emotional reactions, while age is insignificant. The research participants who had either living or working experience in an area felt more positive emotions, by 0.083 (p = 0.010) in their valence level, than those who had never visited the site before the experiment. This result is inconsistent with the argument in a study that longer living experiences in a shrinking area rather lower residential happiness by increasing the inhabitants’ awareness of the negative community changes caused by urban shrinkage [28]. Table 3 also shows that males felt more negative emotions, by 0.072 (p = 0.015) in the valence level, than females; this result supports previous studies that have suggested that women feel more positive emotions than men as they feel a higher neighbourhood attachment than men [29].

Discussion and Conclusion

Although shrinking neighbourhoods are places where people feel negative emotions in cities, some evidence indicates that people experience less negative emotions in some areas than in other areas. Nevertheless, empirical studies analysing the impacts of environmental and personal characteristics on individuals’ negative or positive emotions in a shrinking area remain insufficient. Analysis of the emotional effects of such characteristics may reveal the environmental and personal reasons for the differences in citizens’ emotions between the shrinking communities; this will contribute to creating healthy and liveable urban neighbourhoods.

This study used EEG to measure research participants’ emotional valence levels amid various environmental conditions in a shrinking area and used a multilinear regression model with the collected EEG data to assess emotional impacts. This study revealed two significant findings on environmental and personal attributes that affect people’s emotions in a shrinking area.

The finding on environmental characteristics is that people felt positive emotions at sites where both urban design factors and their management were both satisfactory at appropriate levels. People’s average emotions on other sites, where either one of the two factors was absent, were negative (i.e. open space with social disorder, inactive frontages with CPTED or vandalism, abandoned houses, and vehicle-oriented environments). This finding is supported by the two results. First, although the participants felt positive emotions in a less-crowded open space, they contrastively experienced negative emotions in another similar open space often used by teenagers who were smoking and loitering or older people drinking alcohol. This contrary result means that citizens are likely to feel negative emotions even in a well-designed space if it involves signs of social disorder, including anti-social behaviour.

Second, although the participants felt slightly more positive emotions in the space of inactive frontages with CPTED than in the space of similar inactive frontages with vandalism, the average emotional level in the space with CPTED was still negative objectively. These results revealed that the presence of CPTED could not utterly neutralise people’s negative emotions while in the shrinking area, which is inconsistent with previous studies [26, 27]. From this perspective, it is difficult to deny the possibility that CPTED was not implemented in the area as effectively as it should have been. The presence of CPTED among deteriorated or unmanaged environments could create a strong contrast in the context of environmental management; this contrast could make pedestrians feel anxiety and fear when passing by specific spots that reveal a shrinking area’s crime hotspots. Given these results, while each urban design and environmental management affect citizens’ positive emotions separately, balancing between the factors at appropriate levels is key to effectively delivering their positive functions to shrinking neighbourhoods.

The finding on personal characteristics is that people who had lived or worked in the neighbourhood for a long time and women felt more positive emotions than visitors and men. This finding is consistent with arguments from previous studies: Long-term residents and women have a greater sense of place attachment than short-term residents and men [28, 29], and a sense of place attachment is significantly associated with positive emotions in places [30]. People with relatively higher place attachment or positive emotions towards their shrinking cities (e.g. long-term residents and women) reported a higher willingness to stay in their communities, meaning a lower probability of leaving them in the near future [28]. In other words, even if a neighbourhood shrinks, it can be primarily composed of people who have positive emotions and a sense of attachment to the area. The finding implies the possibility that a shrinking neighbourhood can maintain a sense of satisfaction as long as the area is carefully managed. Of course, this study has a limitation on relatively insufficient sample sizes, as we used EEG for emotion measurements. Therefore, this discussion needs to be further investigated in future studies with larger sample sizes.

Limitations and Future Research

This present study has some limitations. The range of environmental characteristics dealt with is limited; there can be other variables that affect individuals’ emotions while in shrinking spaces. Therefore, it is expected that parallel studies of other environmental characteristics, such as street connectivity, a view-corridor, building or population density, or visible sky ratios, would enhance the study’s generalisability.

This study analysed the degrees of negative or positive emotions, but it did not distinguish between various emotional states. Specific emotional states can be classified by analysing emotional arousal levels, which indicate a person’s activity or passivity, and valence levels. Therefore, it is necessary to analyse valence and arousal levels for future studies to reveal the correlations between emotional states and environmental and personal characteristics in shrinking neighbourhoods.

This present study analysed emotional reactions only in a shrinking area in Seoul, South Korea; therefore, there can be criticism that it is difficult to generalise the results to other countries with regard to environmental and cultural differences. However, the environmental elements classified in this study appear universally in deteriorated environments of a typical shrinking neighbourhood, such as vandalism, abandoned houses, inactive street facades, and social disorders. These elements were chosen based on the results of previous research on these factors’ impacts on individuals’ emotions, perceptions, and experiences in other countries. In this respect, the findings can be sufficiently generalisable and can be applied to shrinking neighbourhoods in other countries after modifying them for their contexts.

Implications

Our findings have several policy implications in the context of the development of shrinking urban areas. Creating more pedestrian-friendly environments that promote social networks in shrinking communities and managing them well are core strategies to improve people’s emotions and mental and physical health.

Regarding the negative impacts of physical disorders on emotions shown in the results, abandoned properties, underused spaces, and leftover spaces should be well managed or gradually transformed into community facilities. The local government could merge these spaces and transform them into various facilities, such as green spaces, children’s parks, and even community-based gardening areas for growing food. This strategy could help eliminate potential spots for social disorders such as crime, drugs, group smoking, and drinking, which create an intense atmosphere of stress and social disharmony. Moreover, inactive street frontages, which this study found provoke negative emotions, need to be transformed into active frontages on narrow and shaded streets, including sufficient windows, entrances, terraces, and street width for future development. These strategies could enhance people’s perceived safety and increase social interactions in the community.

However, these strategies should be based on social engagement and participation by gradually empowering a broader civic society. The process of demolishing undesirable properties and places linked to the past can inadvertently erase the emotional attachment that bonds residents to their shrinking neighbourhood [23].

Here, residents’ neighbourhood attachment and positive emotions are important assets for urban renewal projects in the future. Their emotional connections to the community promote their engagement in environmental revitalisation projects as well as ultimately reinforce the attachment and participation of other people in the community [59]. Therefore, their positive emotions and community attachment should be actively utilised in conjunction with various public government financial and policy support and local planning experts.

Supplementary Information

Below is the link to the electronic supplementary material.