Fear of street crime among Japanese mothers with elementary school children: A questionnaire survey using street montage photographs

ABSTRACT Since physical crimes committed on the street against children typically occur during school commutes, there is an urgent need to address Japanese mothers’ fear of such crimes. Although previous studies have proven that several street elements affect parents’ fear of crime, these elements have been evaluated individually, and there is little analysis of fear of crime using computer graphics (CG) to visualize the scenes. In this study, 96 street montage photos were processed using a systematic mechanism based on the theory of Crime Prevention Through Environmental Design (CPTED) to visualize the variation in factors assuming interaction effects. The results indicate that the presence of signs on windows, obtuse angles of corner lots, commercial building usage, and tables and chairs on the ground floor may moderate mothers’ fear of physical crimes against their children in certain scenes. Moreover, several interaction effects that promoted or moderated the main effects of other factors were observed. These findings suggest that, to reduce mothers’ fear of crime, the identified factors should be considered together, rather than separately\.


Introduction
According to the Japanese Police White Paper (National Police Agency 2019), there has been an extremely slow decline in the number of physical crimes committed on the street against children under the age of 13 over the past few years.These crimes typically occur in school commutes, from 7:00 am to 8:00 am and from 3:00 pm to 6:00 pm.Thus, school roads play an essential role in maintaining the safety of children.Furthermore, compared to the United Kingdom and other countries, Japanese elementary school children generally travel to school without parents accompanying them, which suggests that besides crime, there is also an urgent need to address psychological issues such as parents' fear of crime.
Interestingly, the theory of Crime Prevention Through Environmental Design (CPTED) has been developed through three generations by focusing on a situational approach to crime reduction (1st), social ecology (2nd), and human motivation and aspirations (3rd) (Mihinjac and Saville 2019).In this study, we adopted the first generation of CPTED (hereafter simply referred as CPTED), which was proposed and developed around 1970, and focused on its two principles of "surveillance" and "territoriality" in particular (Jacobs 1961;Jeffery 1971;Newman 1972).
Over the past two decades, several studies based on the theory of CPTED have shown that the physical space elements on school roads are related to elementary school children's safety.According to Matsunaga, Miyazaki, and Sumi (2009), while vacant land and the absence of windows on school roads, building entrances, and exits are associated with a higher crime rate against children, the presence of stores contributes to a lower crime rate.Ago, Matsunaga, and Sumi (2010) concluded that limiting the number of windows below the 2nd floor as well as walls over 1.5 meters will prevent the presence of suspicious people.Nevertheless, it is not only the aforementioned physical space elements but also other causes that affect the occurrence of crime.The opportunity to encounter victims also influences the actions of suspicious people, and crimes typically occur when time, space, and victims appropriately coincide (Matsunaga et al. 2012).
In addition to crimes committed on school roads, many researchers are concerned with parents' fear of crimes committed against their children.Parental restrictions when they feel unsafe about the neighborhood have been found to have an impact on children's outdoor activities (Amemiya et al. 2010).Extensive research has also shown that the physical space elements on school roads affect parents' fear of crime.Some store types (e.g., stores for dining and entertainment) make parents feel unsafe (Takayanagi and Akashi 2011), and the shape of streets and visibility of the building interior also influence parents' fear of crime (Deguchi and Yoshimura 2015).
However, as most previous studies adopted a questionnaire with a textual description, respondents were forced to answer it by relying only on their experience and imagination of the scene during the evaluation.This may cause inaccuracies, especially when evaluating parental fear of crime.Computer graphics (CG) have also been used to evaluate the effects of pedestrian attributes and types of shops on parents' fear of crime (Takayanagi 2014).The CG method referred to as "street montage" (Kang and Arima 2015) was used in the street evaluation and very effectively measured road width, cars, and visibility.However, there have been few quantitative analyses of fear of crime using CG to assess the elements of physical space on the street.In addition, previous studies are mostly limited to independently examining the effect of street elements.The mechanism by which some scenes cause parents' fear of crime is not fully understood, and the interaction effects of street elements remain unknown.

Research purpose and significance
The main aim of this research was to quantify mothers' fear of crime by evaluating the influence of selected street elements on their fear of crime.The principles of surveillance and territoriality of the CPTED were used as the theoretical basis for this study (Hino et al. 2011;Home Office & ODPM 2004).To this end, this study adopts three goals.The first is to visualize the variation of the factors of street elements through the method of street montage and to investigate how mothers' fear of physical crime against children varies as factors and levels change.The second is to systematically evaluate the influence of street elements rather than examining one or two factors individually, and consider the interaction effects as a whole mechanism in this study.The third is to consider both the dynamic and static aspects of surveillance and to investigate whether the increasing territoriality of the housing scale reduces or increases mothers' fear of crime.

Street montage
The target area was the H Elementary School Zone located in Adachi Ward, Tokyo, Japan.The area was selected because of the large number of precursor events against children as well as the size (75.9 hectares) from areas that are included in Adachi Ward, which has a relatively large number of crimes compared to other residential wards in Tokyo.The target area is a residential area consisting of four blocks of exclusively low-rise and one block of mid/high-rise residential zone of land-use zoning, in which the theory of CPTED is applicable (Lee, Park, and Jung 2016).
The elements in the street montage were determined based on the two principles of surveillance and territoriality in CPTED (see Table 1).Under the principle of surveillance, street elements were divided into two (Hino et al. 2011): "static" aspect that ensures visibility to the public space and "dynamic" aspect that facilitate human activity.This study focused on building windows and building usage of static and dynamic aspects, respectively.Under these two aspects, "transparency of stores," "windows facing public spaces," "use of the ground floor," and "use of corner lots" were selected as the main elements in this study.In particular, windows, corner lots, and the ground floor were the three main factors.The "exhibition and flowers," "open chat space," "defining territories," and "grassland" are selected as sub elements of this study.Factors that were selected based on the three main elements were important and were thus labeled as individual categories in the street montage.Factors selected based on four sub-elements, including exhibitions and flowers, defining territories, open chat spaces, grasslands, and houses, were used as common street elements and applied to the entire street montage.As a result, 18 factors were generated, of which 13 were selected from the main street element category and five from the sub-street element category.
Of the 18 factors adopted, two (grassland and use as open chat spaces) based on sub-elements were set as the level of the two main factors, considering their attributes.The sub factor grassland became a level of "paving," suggesting that no paving exists, and open chat space became a level of "presence of tables and chairs."The factors "existence of fence" of both corner lots and ground-floor categories were merged into one factor.Thus, 15 factors were retained (Table 2).Of these 15 varying factors in the street montage, seven were chosen as critical.According to previous studies, it is assumed that their interaction effects are more important than others, and they were thus prioritized in the arrangement of the experiment.
Twelve street scenes (three categories × four scenes each) were selected after the arrangement of the street montage.Each scene included three factors, and the interaction effects of the key factors were prioritized for evaluation.In the corner lots and ground-floor categories, all factors had two levels.In the window category, four out of the five factors had two levels, whereas "opacity of frosted glass", which showed the transparency of the window, had four levels (0, 33%, 66%, and 99%).The higher the opacity of the frosted glass, the lower the transparency of the window, making it more difficult to see through the window.The specific value of opacity in the street montage photo was controlled by changing the value in the opacity option of the frosted glass layer in Adobe Photoshop.
Each street scene in each category included eight montage photographs, and each photograph indicated different levels of the three factors.Among the 96 (8 × 12) montage photos, 8 × 10 photos contained three factors of two levels, and 8 × 2 photos contained one factor of four levels and two factors of two levels.

Questionnaire survey
The questionnaire was conducted online from 12 th to NaN Invalid Date NaN.The respondents were mothers of elementary school students, aged 30-49 years, and living in 22 wards of Tokyo, excluding Adachi Ward.A total of 325 survey responses were received.
Overall, 21 questions were included in the survey, which were divided into three parts: three questions on screening, six on personal attributes, and 12 on street montage.Respondents were asked to answer questions assuming that their eldest elementary school children may encounter a physical crime on school roads.In the street montage section, each question showed one of eight randomly selected montage photos.Based on the ten key issues, nine evaluation items were chosen in the questionnaire based on previous studies (see Table 3), and the options were divided into four levels: strongly agree, somewhat agree, disagree, and strongly disagree.
The preprocessing of the data collected from the questionnaire consisted of three steps.The first step involved checking and deleting invalid data gathered from 325 respondents, and 311 were eventually accepted.The second was the conversion of certain data in the personal attributes section for analytical purposes.The third content was the reverse coding of the results of the evaluation items, except for items I and II in the street montage section.

Statistical analysis
There are three prerequisites for ordinal logistic regression that should be confirmed before analysis.First, subjective questions were removed to avoid mutual interference.Second, the proportional-odds assumption is tenable.All the variables were converted into dummy variables.Data collected from the questionnaire was used for statistically testing the significance of the impact of the selected physical space elements on Japanese mothers' fear of crime by using three packages in R studio (version 4.0.5)-VGAM package, Hmisc package, and Brant test taking into account that certain tests may be inclined to reject the null hypothesis and lead to erratic results.For instance, when using the Hmisc package, an infinite value of xlim (limit of a variable) can lead to a model fitting failure.However, this can be fixed by conducting the Brant test, and the results are displayed as if the proportional odds assumption holds.As a result, the proportional odds assumption is tenable for all three methods, indicating that the ordinal logistic regression is applicable.Third, multicollinearity was tested using the car package (R).
The variable filter was used for the multiple regression analysis.The analysis was achieved by establishing several objects which represent the fitting linear models, and the objects are used in a stepwise algorithm with the default direction as "backward."Although all related factors and interactions were assumed in the original formula, only the significant factors and interactions were included.
After the variable filter, the command of the MASS package was used to execute the ordinal logistic regression analysis, and p-values and coefficients were calculated.The regression model is expressed as follows: where y i denotes a dependent variable representing an ordinal outcome with four categories (1-4) for each item in each scene.P y i � 4 ð Þ denotes the cumulative probability of y i less than or equal to i = 1,2,3.β j0 denotes the estimated intercept, β i;j the estimated coefficient, and x i;j the independent variables, including selected variables of personal attributes (school age, gender of school child, other transportation to school, other transportation to home) and the significant variables of the three factors in the street montage section.γj;k represents the estimated coefficient regarding the interaction term.

Window scene
Table 4 presents the results of the ordinal logistic regression of the street montage section for window scenes.Figure 1 shows the corresponding street montage photos, where significant main effects were   Furthermore, the scene in Window 1 indicates the existence of interaction effects.There is a situation in which either one or both factors exist in the scene, although the analysis result shows the same negative effect in both cases, the interaction effect could potentially be positive, considering the main effect.This is because the effects of both factors cannot be combined, owing to the upper limit on their effectiveness.Window 1-Item IV and Window 1-Item VI apply to this situation   (Table 5 and Figure 2).Considering that most of the main effects of Factor B (openness of window) are positive, the opacity of the frosted window (Factor A) is likely to have an adverse effect on children who could be watched over from surrounding roads (Item IV) and central buildings (VI).

Corner lot scene
Table 6 presents the results of the ordinal logistic regression of the street montage section for the corner lot scenes.In the corner lot category, factor N (paving) had positive effects on six items (items III -VII and IX) in corner lot 2 and six items (items I and V -IX) in corner lot 3 (see Figure 3).These results suggest that when the paving is white, mothers' fear of crime is reduced in different ways.Factor O (angle) has positive effects on items I and VIII in corner lot 2, indicating that when the angle is obtuse rather than acute, mothers tend to believe that the possibility of hiding a criminal is lower (I), and when the planting is well managed (VIII) in corner lot 2.

Ground floor scene
Table 7 shows the results of the ordinal logistic regression of the street montage section for ground-floor scenes.Factor G (presence of tables and chairs) has positive effects on Items IV and VII in the scene of ground floor 1, as Item V of ground floor 2 and Item VI of ground floor 3.This suggests that when tables and chairs are present on the ground floor and it is used as an open cafeteria, mothers are likely to believe that children could be watched over from surrounding roads (IV), helped by someone (VII) of ground floor 1, feel a bright environment (V) of ground floor 2, and that children could be watched over from the central building or land (VI) of ground floor 3. Factor J (commercial building usage) had positive effects on six items (III to VIII) of ground floor 2 and six items (I, II, V, VI, VIII, and IX) of ground floor 4.These results indicate that when the building was commercially used in the scene of ground floor 2, mothers tended to feel safer in many ways.The corresponding street montage photographs are shown in Figure 4.
The result of Ground floor 1-Item V (sense of brightness) shows that although the effects of factors H (fence) and G (tables and chairs) are both positive, the interaction effects of the two factors are negative (Figure 5).

Discussion
Several factors originally assumed to have a possible impact on mothers' fear of crime showed notably significant results.The result of the window category suggests that the openness of windows has a strong positive effect on mothers' fear of crime.Regardless of whether the window is actually open (Figure 1a) or used as a French window and can be accessed by pedestrians (Figure 1b), mothers tend to believe that physical protection is enhanced and the possibility for children to run away is higher in this scenario.This result suggests that parents' perceptions were consistent with the findings of previous studies in which windows were found to reduce crime and stranger danger risks on school routes (Ago, Matsunaga, and Sumi 2010;Matsunaga, Miyazaki, and Sumi 2009).In addition, it was found that the existence of signs that were originally supposed to influence surveillance actually has positive effects on physical protection.For corner lots, the result shows that an obtuse angle has positive effects on access and movement, as mothers feel that a criminal is unlikely to hide.This supports the findings of Deguchi and Yoshimura (2015) that the number of spots where parents feared for crime was lower on streets with good visibility.Interestingly, the results of the ground floor suggest that commercial building usage has a strong positive  influence on mothers' fear of crime in nine aspects.This result seemingly contradicts that of Takayanagi and Akashi (2011) in which stores made parents feel unsafe.However, unlike stores for dining and entertainment it targeted, the types of stores that can be located in the residential areas targeted in this study mitigated fear of crime.Tables and chairs used as open cafeterias enhance surveillance in four ways and mothers feel that children are being watched and could ask for help in this situation.Furthermore, this study investigated interaction effects.Although most results show the additional effect of the main effects.Notably, the results for  the window category, show there could be a possible negative effect of the opacity of the frosted window (factor A) to factor B (openness of window) on children being watched over from the surrounding roads (Item IV) and central building (VI) in Window 1, which shows the effect that is not reflected in the main effects.The negative interaction effects of H (fence) and factor G (tables and chairs) on Item V (sense of brightness) on ground floor 1 is also an interesting result.
Moreover, the results for the corner lots suggest that housing territoriality reduces mothers' fear of crime.The main effects in the corner lots suggest that pavement has a strong positive effect on mothers' fear of crime.When white paving is used instead of grassland, even though the territoriality of the housing is greater, mothers feel that children are safer in eight out of the nine evaluated aspects.However, too much territoriality could make it difficult for children to enter, even in an emergency.This limitation could be an important issue for future research.

Conclusion
This study investigated the mechanism by which mothers' fear of crime is manifested.The findings not only provide further in-depth implications on some street elements based on previous studies but also show that other factors of street elements, such as signs on the window, obtuse angle of corner lots, commercial building usage and the presence of tables and chairs on the ground floor may reduce mothers' fear of crime in several aspects in certain scenes.Moreover, the interaction effects of the street element factors were initially assumed.The existence of interactions is supported by the results, and some factors that enhance the main effects of the other factors, while others show the opposite effects which suggests that fear of crime should be comprehensively considered.
observed.Factor B (openness of window) has positive effects on Item III of Window 1 and items I -IV of Window 3, indicating that when windows are open in Window 1, mothers tend to feel that the possibility for children to run away is higher (III).When windows are open in Window 3, mothers feel that the possibility of hiding a criminal (I) and for children to be physically harmed is lower (II), the possibility for children to run away is higher (III), and that children could be watched from surrounding roads (IV).Factor D (existence of sign) has positive effects on Item III (Window 1), and indicates that when a sign exists in Window 1, mothers are likely to believe that the possibility for children to run away is higher (III).

Figure 1 .
Figure 1.Montage photos of the window scenes where significant main effects were observed.(a) Window 1 (Factor B: Openness of window -open) (b) Window 3 (Factor B: Openness of window -open) (c) Window 1 (Factor D: Existence of signs -existence).

Figure 2 .
Figure 2. A montage photo of the window 1 where negative interaction effect was observed (Factor A: opacity; Factor B: openness).

Figure 3 .
Figure 3. Montage photos of the corner lot scenes where significant main effects were observed.(a) Corner lot 2 (factor O: obtuse angle) (b) Corner lot 3 (factor N: white paving).

Figure 4 .
Figure 4. Montage photos of the ground-floor scenes where significant main effects were observed.(a) Ground floor 1 (Factor G: Presence of tables and chairs -yes) (b) Ground floor 3 (Factor G: Presence of tables and chairs -yes) (c) Ground floor 2 (Factor G: Presence of tables and chairs -yes) (d) Ground floor 2 (Factor J: Commercial building usage -yes).

Figure 5 .
Figure 5.A montage photo of the ground floor 1 where negative interaction effect was observed (Factor G: tables and chairs; Factor H: fence).

Table 1 .
Selection of elements and factors in street montage.

Table 2 .
Factors and their levels in street montage.

Table 4 .
The result of ordinal logistic regression of the window scenes.

Table 3 .
Items on fear of crime in the street montage section of the questionnaire.

Table 5 .
interaction effects in the window category.

Table 6 .
Result of ordinal logistic regression of the corner lot scenes.

Table 7 .
Result of ordinal logistic regression of ground floor scenes.