Journal Archive

Suicide may be defined differently, but the general definition of suicide is ‘an act of killing oneself’. A wider definition is ‘an attempt to kill oneself or a psychological condition to do so’ (You, 2014). Considering the wider definition, suicide may be understood as a result of consecutive actions including suicide ideation, attempted suicide and suicidal behavior (Harwood et al., 2002).

Scholars also point out the need for understanding suicide as not only a behavioral result but also a process leading to suicide (Shin et al., 2014). In particular, scholars have indicated that attention should be paid to the presence of a past experience of suicide ideation when considering suicide among the elderly. This is because, although the presence of a past experience of suicide ideation does not always lead to actual suicide, a person who has experienced suicide ideation has a higher risk of death by suicide, and continued suicide ideation is an important factor leading to suicide (Park, 2015; Ronald et al., 2015).

A previous report showed that the presence of more experiences of suicide ideation is more likely to lead to actual suicidal behavior, especially among the elderly (Miller et al., 2001). Therefore, the presence of a past experience of suicide ideation may be an important predictive factor in the suicide rate among the elderly. Therefore, in the present study we consider the experience of suicide ideation as a major predictive factor of suicide among the elderly.

Suicide ideation and actual suicide affect the people around the victim (Bearman and Moody, 2004; Lee et al., 2010; Seo and Lee, 2015; Robbins and Alessi, 1985). Because of the spatial dependence, the trend is not localized to a specific region; it is highly probable that spatial clustering may occur through interactions with neighboring regions.

Spatial dependence refers to the interactions between geographically proximate individuals, groups, and areas. Upon the existence of spatial dependencethere is a similarity regarding a certain phenomenon between geographically proximate localities (Anselin and Bera, 1998). In other words, the characteristics of individuals, groups or regions that are geographically close are similar due to the spillover effect (Tobler, 1970), and the spatial dependence is represented by the clustering of geographically adjacent spaces.

According to Morenoff (2003), the spatial clusters may be formed as a result of an artifact by the spatial clustering of predictive variables. Therefore, spatial clustering of elderly suicide may be caused by the spatial clustering of the predictive variables related to the elderly suicide in the spatial units. For instance, the suicide rate among the elderly may be high in two geographically adjacent regions, which may be because the policy circumstances regarding elderly welfare are relatively poor in regions due to their low financial independence. This suggests that after the predictive variables of the suicide rate among the elderly are controlled, the spatial dependence of the suicide rate among the elderly should no longer be significant.

Next, the neighborhood effect of spatial clustering is caused by spatial externality. The neighborhood effect refers to the effect that independent causal relationships have on social performance through social networks and interactions (Jencks and Mayer, 1990). According to the neighborhood effect, regions within a certain distance have network of similar probabilities. As a result, the culture, ideology and social network of the regions within a certain distance have more mutual influences in comparison with regions out of the range.

In this regard, the First Law of Geography according to Tobler (1970) states, “Everything is related to everything else, but near things are more related than distant things.” In the context of elderly suicide, the law suggests that elderly suicide may have an affect between individuals and between regions, and that spatial clustering of the suicide rate may occur even if the predictive variables are controlled.

Suicide contagion is categorized into indirect suicide contagion through media, which is called the Werther effect, and direct suicide contagion in the middle of direct social relationships. Previous studies on the direct suicide contagion showed that suicide inside a social network is a highly influential factor to suicide-related behavior of others (Robbins and Conroy, 1983; Thoits, 1985; Brent et al., 1989). It was particularly highlighted that suicide by an acquaintance who has been in a high level personal relationship has a more direct effect on the increase of the likelihood of suicide-related behavior by the people around the person (Bearman and Moody, 2004). This suggests that suicide may be caused by not only personal reasons, but also by external factors.

In addition, the elderly have a relatively small range of activity due to physical restrictions (Robert and Li, 2001; Glass et al., 2003). This means that the elderly may be more affected by the interactions between adjacent regions than people of other generations. Therefore, the spatial clustering of elderly suicide may be the result of the homogeneous regional responses to the issues that have a negative impact on the mental health conditions of the elderly.

Until now, studies on elderly suicide have focused on the economic or psychological factors of individuals, including income level and depression (Kim and Kim, 2011; Kim and Kwon, 2013; Kim and Kim, 2014; Park and Kim, 2017). Several studies have recently been conducted considering not only the individual factors but also environmental factors (Noh, 2013). Since individuals’ behavior is affected by environmental factors, suicide ideation and behavior are also expected to be affected by environmental factors (Lee and Oh, 2008).

Previous studies on the spatial distribution of suicidal risks have shown that the ratio of economically disadvantaged residents is higher in underdeveloped areas (Latkin and Curry, 2003; Kuramoto et al., 2013). In addition, it was reported that the suicide rate is higher in rural areas than in urban areas (Chang et al., 2011). This may be because rural areas have experienced spatial, social and cultural severance more than urban areas in the rapid industrialization process. Hence, this severance might have acted as a factor that increases the suicide rate.

Previous studies show the possibility of spatial interactions between adjacent regions with regard to the suicide rate among the elderly. The degree of underdevelopment or characteristics of regions may be mutually affected between regions through the spillover effect mentioned above. Therefore, spatial dependence and clustering may also be found in elderly suicide ideation and actual suicide.

In addition to the discussion on the spatial dependence in elderly suicide, attention should be paid to the spatial heterogeneity of elderly suicide ideation and suicide rate according to gender. Generally, the rates of suicide ideation and suicidal intent are higher among women, but the actual suicide rate among men is double or triple the suicide rate among women (Kim and Kim, 2016). Hence, despite the presence of spatial dependence of elderly suicide ideation and suicide rate, different spatial clusters may be formed because of the differences between men and women.

First, with regard to suicide ideation among the elderly, elderly women may be more spatially dependent than elderly men because women are affected by others more than men (Jun and Namgung, 2018). According to Graziano et al. (1993), this study suggest, in the context of suicide ideation, elderly women are more affected spatially by others’ opinions and behavior.

In addition, women generally provide a higher level of social support to others through social networks than men, and maintain closer emotional relationships with others (Antonucci, 2001; Chang, 2010). On the contrary, men tend to form more diverse and wider network of human relationships to acquire jobs and promote their career (Fischer and Oliker, 1983; Wellman et al., 1983). Recalling that elderly men generally experience loss of their social roles, the social network of elderly men is considered to be much smaller than that of elderly women.

Previous studies on the functions of social networks according to gender showed that the relationships of elderly women with friends and acquaintances are in the type of ‘facing each other,’ of which the main functions are interests in others and social and emotional exchanges. On the contrary, the relationships of elderly men are in the type of ‘standing beside,’ of which the main functions are intellectual exchange rather than emotional support (Wright, 1982; Parker and de Vries, 1993; Walker, 1995). In summary, in the context of elderly suicide ideation, elderly women, who are more active in social and emotional exchanges are more exposed to the impact of others’ opinions and behavior. This suggests that spatial dependence in suicide ideation may be higher among elderly women than among elderly men.

Next, the suicide rate among the elderly may be more spatially dependent among elderly men than elderly women. Bass and Lambert (2004) reported that men have more negative recognition about environmental factors and are more affected by them in comparison with women. This negative recognition and the impact of the environmental factors may be particularly higher among elderly men who have experienced social retirement due to the loss of the gender roles expected of men. For many years, men have been expected to be more adventurous than women (Girard, 1993), and play the role of a family head to support the family financially (Chuang and Huang, 2007). Accordingly, as the external compression caused by the environmental change continues, men who are restrained in the stereotypical gender role experience a higher level of stress (Kawachi and Berkman, 2001).

In addition, there is a difference between men and women in actual suicidal behavior. While men execute more aggressive actions in suicidal behavior, such as the use of firearms and falling (Kim, 2013), women have a substantially lower level of actual suicidal behavior compared to their suicidal intention because of the high labeling effect on suicide (Ji, 2012).

In summary of the discussions described above, elderly suicide ideation may be more spatially dependent among elderly women, who are highly influenced by others and have stronger social network with others. Spatial dependence in their mental and emotional stress may be higher among elderly women. On the contrary, the actual suicide behavior may be more spatially dependent among elderly men, who have strongly negative recognition about environmental factors due to their stererotyped gender role and aggressiveness. Therefore, because of the ‘Werther effect’ through mass media or the impact of the suicide of an acquaintance in a social relationship, the spatial dependence in actual suicide may be stronger among elderly men than elderly women.

Previous studies show that suicide ideation and actual suicide among the elderly are affected by economic and social factors(Kim and Kim, 2011; Kim and Kwon, 2013). Other studies were conducted based on time-series data to analyze the patterns in which the levels of suicide spread and suicidal risks are fixed (Roh, 2017; Joo and Lee, 2014).

Some of the studies about elderly suicide were conducted by considering the spatial factors (Roh, 2017; Joo and Lee, 2014). However, elderly suicide has not been sufficiently studied from the viewpoint of spatial dependence. These studies have a consensus about the spread of suicide due to its social contagion as well as the possibility of the clustering in spatial units. The studies have provided spatial patterns of suicide that occurs within spatial units, but the analysis is still descriptive.

The analysis based on spatial dependence must be based on contextual interpretation. As discussed above, elderly suicide in a specific region may have mutual effects on the adjacent regions in various aspects, and spatial dependence and spatial clustering may thereby take place. In addition, according to Tobler’s First Law of Geography, geographically close regions have similar characteristics. This means that spatial dependence and spatial clustering of suicide ideation and actual elderly suicide may be different among different regions. This also means that different suicide-related contexts should be taken into consideration for different regions.

On the other hand, a limitation of the previous studies conducted by considering the spatial factors related to elderly suicide is that they failed to consider the presence of elderly suicide ideation, which may function as an important predictive factor. There is a report that the elderly have a higher probability of carrying out actual suicide, as they have more experiences of suicide ideation (Miller et al., 2001). This suggests that there should be a consideration on the spatial pattern of elderly groups who have had suicide ideation. When clusters related to the elderly suicide ideation are formed in spatial units, it is highly probable that the spatial clusters related to elderly suicide may be formed at the clusters as well.

Based on the discussions described above, we verified not only the spatial dependence of elderly suicide in different regions, but also the difference in the pattern of the spatial dependence according to gender. Through this, the present study may provide a new perspective in relation to elderly suicide with respect to the space. Furthermore, the results obtained from the present study may be used to establish more systematic and meaningful policies for preventing suicide. The research hypotheses set up in the present study based on the discussions above are described below.

• Hypothesis 1: Local suicide ideation and actual suicide among the elderly are spatially dependent.
• Hypothesis 2: The level of spatial dependence in local suicide ideation is higher for elderly women than elderly men.
• Hypothesis 3: The level of spatial dependence in local actual suicide is higher for elderly men than elderly women.

The scope of the present study is described below. First, the spatial scope of the present study included 225 administrative units in Korea (‘si,’ ‘gun’ and ‘gu’) except the island regions, which were expected to lack spatial dependence due to the absence of adjacent regions (Jeju-do, Ulreung-gun and Ongjin-gun in Incheon).

The subjects of the present study were the elderly aged 60 years or older. From a systematic point of view, those aged 65 years or older are considered as elderly citizens. However, the studies that analyzed the subjective recognition of age by individuals pointed out that the subjective recognition of age may be different from the objective criteria of the elderly (Jeong and Song, 2012; Song and Kim, 2013). According to data from the National Pension Research Institute, the age groups recognized as the elderly vary from those between 60 to 64 years of age or those between 65 to 70 years of age. David (2001) reported that one of the universal symptoms that are generally found among the elderly is geriatric depression. Considering that negative mental health conditions, such as depression, can lead to suicide, as described above, in our study we appropriately defined the elderly as the people aged 60 years or older who may have geriatric depression.

Next, the variables of local elderly suicide ideation used in the present study were selected by using data from the Korea Community Health Survey, which is conducted each year by 253 public health centers under the supervision of the Korea Center for Disease Control and Prevention. The data includes information related to the health of residents in local communities, such as smoking, drinking, physical activities, mental health and quality of life. Therefore, we considered the data as suitable for determining the presence of an experience of elderly suicide ideation. However, the survey item related to suicide ideation is included in the data in a period of 4 years. Hence, we analyzed the data of 2017, the latest survey data including items related to suicide ideation.

Actual elderly suicide in different localities was analyzed by using data from the Causes of Death Statistics for 2017. The Causes of Death Statistics according to the Statistics Act and the Act on Registration of Family Relations are based on the death notices submitted to the administrative offices of the registered places throughout Korea (eup, myeonn, dong, si and gu), and includes information about personal characteristics such as gender, age, marriage, education and occupation, as well as the causes of death such as disease, accidents and intentional self-injury (suicide). Therefore, we considered that the data is suitable for analyzing local actual suicide among the elderly throughout Korea, and reconstructed the data for our analysis by selecting the individuals whose cause of death was intentional self-injury (suicide).

The analytical methods of the present study are described below. First, local suicide ideation and actual suicide among the elderly, which were used as the key variables in the present study, were reprocessed into ‘the number of individuals per 100,000 in the elderly population’ from the analytical data. The Community Health Survey data used for the analysis of local suicide ideation among the elderly, is a dataset from the survey that includes a question whether a respondent had an experience of suicide ideation. Thus, the local suicide ideation among the elderly was constructed as the ratio of the elderly who experienced suicide ideationto the entire elderly population within each locality. Given that the very small portion to the entire elderly population, then, the local elderly suicide ideation was was converted into the ratio per 100,000.¹⁾ The data of the Causes of Death Statistics used for the analysis of local actual elderly suicide was prepared with reference to the death notices submitted to the registration offices of the individuals. For the local actual suicide among the the elderly, the ratio of the elderly whose cause of death was suicide to the entire elderly population in the locality was converted into the ratio per 100,000.²⁾

Second, to investigate spatial dependence in the present study, we established spatial data for the administrative units except for the island regions mentioned above. Subsequently, to verify the hypotheses of the present study, we used Moran’s I index to investigate the spatial dependence of elderly suicide ideation and actual suicide in 2017. Moran’s I index that is employed to analyze spatial dependence has a value between –1 and 1, and a value closer to 1 indicates the presence of adjacent regions showing more similar characteristics (Lee and Shim, 2011).

Third, Moran’s I index has limitations when used to investigate the detailed spatial dependence of small areas. Hence, spatial dependence in local units was investigated to analyze the spatial distribution of the elderly according to gender for the hypotheses 2 and 3.

The Local Indicator of Spatial Association (LISA) analysis is generally used in parallel to verify the spatial autocorrelation of local units, allowing for a more systematic spatial analysis (Anselin, 1995). Spatial correlations are classified by the LISA into 4 types: 1) the HH (High-High) type where there are high values around a high value of a specific region; 2) the HL (High-Low) type where there are low values around a high value; 3) the LH (Low-High) type where there are high values around a low value; and 4) LL (Low-Low) type where there are low values around a low value. The LISA analysis was performed to investigate the spatial clustering of suicide ideation and actual suicide among the elderly by comparing it between men and women.

Prior to the main analysis of the present study, the analysis of the spatial dependence of local suicide ideation and actual suicide among the elderly, we reviewed the overall characteristics of the elderly in relation to suicide. We analyzed the mean values of elderly suicide ideation and suicide in all localities, and performed a t-test with regard to local suicide ideation and suicide among the elderly according to gender. Table 1 shows the analytic results.

The analysis shows that the pattern of local elderly suicide ideation and actual suicide was different between men and women. Local suicide ideation was experienced more by women (12933.63) than men (7953.66). On the contrary, local actual suicide was higher among men (71.31) than women (22.96). An additional t-test was performed to verify the significance of the differences. The analysis showed that the t-values for local suicide ideation and local actual suicide were –10.776 and 19.820, respectively, indicating that the differences are statistically significant. The results shown in Table 1 are consistent with the previous report that suicide ideation is higher among women but actual suicide is higher among men (Kim and Kim, 2016). The results also show that even if spatial clusters of elderly suicide rate are formed at adjacent regions, spatial dependence may be different between men and women.

Table 2 shows the overall characteristics of local suicide ideation and actual suicide among the elderly in the 16 provinces except Jeju-do. The analytic results showed that local suicide ideation and actual suicide among the elderly according to gender in the 16 provinces were similar to the results shown in Table 1. The t-test shows that there is a statistically significant difference between men and women in both local suicide ideation and actual suicide in most of the provinces except Gwangju and Ulsan.

Table 2 also indicates that local elderly suicide ideation and actual suicide are relatively high in provinces having a higher ratio of non-urban and rural areas. Among the 16 provinces, the ratio of elderly having an experience of suicide ideation was higher in the order of Jeonbuk (15393.22), Chungnam (15044.65) and Sejong (12888.89). Local actual suicide was higher in the order of Gangwon (62.53), Chungnam (55.44) and Gyeongnam (51.89). Local elderly suicide ideation and actual suicide of the 225 localities are presented in the Appendix. The localities that showed higher ratios of the elderly having an experience of suicide ideation were Cheonan-si (28750.00), Boryeong-si (23275.86) and Dangjin-si (22740.52). Local actual suicide was higher in the order of Sacheon-si (148.64), Pyeongchang-gun (113.94) and Yangsan-si (111.76).³⁾

The results obtained from the analysis of the 16 provinces and 225 localities are consistent with the previous reports that suicide ideation and actual suicide among the elderly are relatively high in non-urban or rural areas (Kim, 2009; Bae and Eom, 2009; Song et al., 2010; Chang et al., 2011; Son, 2012).

We examined Hypothesis 1, which is directed to test the spatial dependence in local suicide ideation and actual suicide among the entire elderly including both men and women. For this, we calculated Moran’s I index of local suicide ideation and actual suicide among the elderly. Moran’s I index is calculated by defining the spatial weighted matrix, which represents the contiguity between localities, and by measuring the data similarity between localities.

The spatial dependence may be calculated by the queen method, where the contiguity is measured with reference to the regional boundary, or by the rook method, where the contiguity is measured with reference to the regional vertexes. A clear theory has not been established about the selection of the references determining adjacent regions (Chi & Zhu, 2019). Chi & Zhu (2019), in relation to the selection of a spatial weighted matrix, suggested a data-centered approach based on the level of spatial dependence and statistical significance. Accordingly, we employed the spatial weighted matrices prepared by the two methods to calculate Moran’s I index, and then performed the analysis based on Moran’s I index which was calculated by the rook method and which showed higher values.⁴⁾ Table 3 shows the Moran’s I index of global suicide ideation and suicide rate of the entire elderly population and the gender groups according to the contiguity setting.

As shown in Table 3, local suicide ideation and actual suicide had a positive spatial dependence with the Moran’s I values of 0.220 and 0.212, respectively. The results were consistent with the previous reports about suicide among the elderly (Joo, 2014; Choi, 2016; Jeon, 2016; Roh, 2017). Joo (2014) and Roh (2017) identified the spatial patterns of elderly suicide in different age groups, and verified the level of spatial clustering of elderly suicide each year in terms of the Moran’s I index. Joo (2014) reported that the Moran’s I index for yearly elderly suicide was between 0.18 and 0.22. Roh (2017) reported that the Moran’s I index for yearly suicide among the elderly was between 0.17 and 0.35. Other studies have shown that the Moran’s I value for subjective mental health (Jeon, 2016) and depression (Choi, 2016) in relation to suicide was 0.15 and 0.34, respectively.

Considering that the range of the Moran’s I index, being between –1 and 1 and the results from previous studies, spatial dependence of local elderly suicide found in the present study was not very high. Since spatial dependence is caused by physical accessibility, the local spatial dependence may be lower than spatial dependence at a personal level. The results of the present study may reflect this difference. Nevertheless, the Moran’s I values calculated in the present study are statistically significant, and the similar Moran’s I values from previous studies could support the validity of the analytic results obtained in this study.

The analytic results of the present study mean that there is spatial dependence in local suicide.; thus, geographically proximate localities have similar attributes. The results also mean that elderly suicide ideation and actual suicide are affected by the situation of adjacent localities than that of remote localities. In other words, elderly suicide ideation and actual suicide are spatially clustered at similar levels among nearby localities having similar internal characteristics within a certain range. This may be understood as a result of the cultural and ideological influences given on adjacent localities through the network formed among the adjacent localities. In addition, the results support Hypothesis 1, which is that “Local suicide ideation and actual suicide among the elderly are spatially dependent.”

Although Moran’s I index is useful in investigating the level of spatial dependence in global spatial units, it has limitations in analyzing the clustering in a specific region. Therefore, we additionally performed a LISA analysis to measure the spatial dependence of local suicide ideation and actual suicide of the total elderly population at the local level. Figure 1 shows that the HH types of local elderly suicide ideation and actual suicide are generally found in Gangwon and Chungcheong provinces.⁵⁾ This supports the previous reports that the elderly suicide ideation and actual suicide at the local level are higher in non-urban and rural ares than in urban areas (Chang et al., 2011; Son, 2012; Song et al., 2010; Kim, 2009; Bae and Eom, 2009).

Figure 1. 
LISA cluster maps of local suicide ideation and actual suicide among the elderly

In comparison with the elderly living in urban areas, the elderly in rural areas are more vulnerable due to the decrease of social activities and the lack of social support networks. In addition, social and emotional isolation increases depression as well as the possibility of suicide ideation or suicide behavior (Chang et al., 2011; Son, 2012; Song et al., 2010; Kim, 2009; Bae and Eom, 2009). The localities of the HH type found in the analytic results, are also the less urbanized areas. Hence, the spatial clustering of the localities may be understood as the result of a similar level of spatial, social and cultural severance in contrast to urban areas.

The results of the present study also show that the spatial clustering with regard to actual suicide are overlapped or extended at localities showing spatial clustering with regard to suicide ideation. For example, in Gangwon and Chungcheong provinces where the HH type of suicide ideation was found, the HH type of suicide rate was also found to be overlapped or clustered. This suggests that the intensification of the influence of the local network between the localities caused the homogenization of the local residents. The experience of suicide ideation is the most important predictive factor of suicide behavior (Park, 2015; Ronald et al., 2015), and localities within a certain range have a similar network of probability. As the mutual influences increase among the localities through the network, the residents within a specific range are highly likely to be homogenized in their mind set. Therefore, the stress environment may be recognized similarly in the regions surrounding a certain region where the local residents are vulnerable to a stressful environment.

Next, to examine Hypotheses 2 and 3 of the present study, we analyzed the difference in spatial dependence between local suicide ideation and actual suicide among the elderly between men and women.

First, we analyzed the global spatial dependence of local elderly suicide ideation according to gender. Table 3 shows that the Moran’s I values of local suicide ideation of elderly men and elderly women were 0.179 and 0.210, respectively. The Moran’s I values showed that the two groups have positive spatial dependence in common. Nevertheless, it should be noted that the Moran’s I value is higher in elderly women than in elderly men.

It was mentioned earlier that the spatial dependence of elderly suicide ideation may be higher in elderly women than in elderly men. Women are affected by others’ opinions and behavior more than men (Jun and Namgung, 2018). With regard to social networks, women provide more social support to others and maintain emotionally close relationships compared to men (Antonucci, 2001). Therefore, the analytic results show that elderly women may be more vulnerable to mental and emotional stresses that can affect suicide ideation, such as depression; thus, their spatial dependence may be higher. In addition, the analytic results support Hypothesis 2 that the level of spatial dependence in local suicide ideation is higher for elderly women than elderly men.

Next, the LISA analysis was used to analyze the difference of the spatial clustering of local elderly suicide ideation between men and women. As shown in Figure 2, the spatial clustering of suicide ideation was found in the non-captial region in both elderly men and elderly women. More specifically, the HH type of local suicide ideation was found in the Chungcheon, Gangwon and Jeonra regions in both elderly men and elderly women. However, the male elderly group showed more spatial clusters in the Chungcheong and Jeonra regions, while female elderly groups showed more spatial clusters in the Gangwon region.⁶⁾

Figure 2. 
Gender difference in spatial dependence on local suicide ideation

Next, we analyzed the global spatial dependence of local actual suicide among the elderly according to gender. As shown in Table 3, the Moran’s I values of local actual suicide of elderly men and elderly women was 0.177 and 0.092, respectively. That is, spatial dependence in local actual suicide is lower in elderly women while being relatively higher in elderly men.

As mentioned earlier, spatial dependence in local elderly suicide may be higher in men than in women. Generally, men have more negative recognition about environmental factors in comparison with women (Bass and Lambert, 2004). This trend may be stronger among elderly men who have experienced social retirement. Moreover, men, who are generally more aggressive than women, are more likely to actually carry out suicide (Joiner, 2005). On the contrary, the labeling effect generated by the actual execution of suicide is stronger on elderly women than on elderly men (Ji, 2012). Therefore, the substantial execution of suicide by elderly women, as well as its spatial dependence, may be lower in the group of elderly women who are more strongly affected by others.

In summary, the analytic results show that the spatial dependence of actual suicide by the ‘Werther effect’ through mass media or the impact of the suicide of an acquaintance may be stronger among elderly men than elderly women. Therefore, the results support Hypothesis 3 that the level of spatial dependence in local actual suicide is higher for elderly men than elderly women.

Next, using the LISA analysis, we analyzed the difference of the spatial clustering in the local actual suicide among the elderly between men and women. Figure 3 shows that, as in the case of suicide ideation, the spatial clustering of the actual suicide in both groups was found in the non-capital region. Specifically, the spatial clusters of the actual suicide in the two groups are concentrated in the Gangwon and Chungcheong regions. The HH type was found mostly in Gangwon in the male elderly group, while it was found mostly in Chungcheon in the female elderly group.⁷⁾

Figure 3. 
Gender differences in spatial dependence on actual suicide

Despite the discussions described above, the weakened social network of women by geographical integration may be understood as one of the causes of spatial clustering of the HH type of suicide in elderly women. The localities where the HH type of suicide ideation and suicide rate of elderly women are found the Chungcheong region, including Sejong-si, Cheongju-si and Gongju-si. In particular, Sejong-si was formed by integrating former Yeonggi-gun, parts of Gongju-si, and former Cheongwon-gun, which was a remaining part of Cheongju-si in the past. The new inflow of population and the rapid development of areas incurred by the reorganization of the administrative areas must have caused a drastic change in the living environment of the indigenous residents. Hence, the level of social networks of elderly women that lasted a long time, might have decreased by the new population inflow and the change of residential areas.

As discussed above, in the actual execution of suicide, the labeling effect of suicide is stronger on elderly women than elderly men. Therefore, the spatial clusters of the suicide rate of elderly women in the localities may have formed because the influence of their social networks decreased due to the reorganization of administrative areas, which failed to offset the influence of the labeling effect related to suicide by elderly women.