Comparing residents' fear of crime with recorded crime data-case study of Ostrava, Czech Republic

International Journal of Geo-Information Article Comparing Residents’ Fear of Crime with Recorded Crime Data—Case Study of Ostrava, Czech Republic JiˇríPánek 1,* , Igor Ivan 2and Lucie Macková1 1Department of Development and Environmental Studies, Palacky University Olomouc, 771 46 Olomouc, Czech Republic 2Department of Geoinformatics, VSB—Technical University of Ostrava, 708 00 Ostrava, Czech Republic *Correspondence: [email protected] Received: 2 August 2019; Accepted: 4 September 2019; Published: 8 September 2019   Abstract: The fear of crime is an established research topic, not only in sociology, environmental psychology and criminology, but also in GIScience. Using spatial analysis to analyse patterns, explore hotspots and determine the significance of respective surveys is one reason for the increase in popularity of such research topics for geographers, cartographers and spatial data scientists. This paper presents the results of an intensive online map-based questionnaire with 1551 respondents from the city of Ostrava, Czech Republic. The respondents marked 3792 points associated with the fear of crime over a ten week period. The perception data were compared with recorded crime data acquired from police department records for the years 2015–2018. This paper explores the spatial autocorrelation from perceived hotspots and from recorded crime hotspots. Our findings fit into the literature confirming results about the locations that most frequently attract fear, but there is still room for more investigations regarding the links between recorded crime and the fear of crime. Keywords: VGI; participatory mapping; crime mapping; PPGIS 1. Introduction Understanding people’s responses to the fear of crime [ 1 ] is important for understanding behaviours which reflect this concern. The fear of crime is one of the essential concepts of Criminology, and is often described as an emotional response to potential victimisation [ 2 ]. Historically, the fear of crime was defined as the likelihood of the risk [ 3 , 4 ]. In other words, how likely it is to become a victim of a crime. However, recent studies [ 5 , 6 ] link the fear of crime to a person’s emotional response. Moreover, it is important for policymakers and municipalities, who are able to act upon these concerns. Importantly, the fear of crime affects individuals and their well-being, and it may even influence their behaviour as regards to frequenting the respective sites. On the other hand, there has been a vast amount of research focusing purely on analysing and exploring patterns in recorded crime data [ 7 – 19 ] relating to the time of day and land use. This research combines the perceived crime data with recorded crime data in one analysis in order to explore the attributes of various parts of the selected city—Ostrava, Czech Republic. Geographic information systems (GIS) are well-positioned to integrate the data on people’s perceptions of the urban environment with police crime reports. Using GIS can enable the police and local communities to address the fear of crime and recorded crime in a targeted manner. The residents’ perceptions of crime might not be reflected in reported crimes and so the police could have a new tool with which to pinpoint underreported criminal activity [ 20 ] and to improve prevention communication in areas where crime is underestimated. This research focuses on the Czech city of Ostrava. The city was selected for this case study because it was involved in the project titled “Effective Methods of Identification, Assessment and ISPRS Int. J. Geo-Inf. 2019,8, 401; doi:10.3390/ijgi8090401 www.mdpi.com/journal/ijgi ISPRS Int. J. Geo-Inf. 2019,8, 401 2 of 15 Monitoring of Safety Risk Areas Using Spatial Micro-data” run by the VSB-Technical University of Ostrava. Involvement in the project is linked with prior collaboration with this participating university and their willingness to share crime occurrence data and other indicators that would help to identify safety risk areas in general. The city of Ostrava was selected for this case study following almost one-year’s preparation and cooperation in the area of perception of fear mapping with the local administration. The city was selected based on the authors previous experience with participatory mapping, the knowledge of local realities and the second worst crime intensity in the country (after Prague). Another reason for selecting the city for this case study was the availability of recorded crime data for Ostrava, as this was the first time such data were released for any large city in the Czech Republic. Ostrava is the third largest city in the Czech Republic and it is located close to the north east border with Poland. It was founded in 1267 and it is a coal-mining city. The wider conurbation of the region has a population of approximately 500,000. Since the 1990s, after the rapid decline of the city’s industrial sectors such as iron, steel and coal-mining, Ostrava has been transformed into a modern cultural city with numerous theatres, galleries and other cultural facilities. The long-term aim of the project is to compare the subjective/emotional fear of crime perception with reported crime incidents. A comparison of such combined data has never been done in the Czech Republic prior to this research. Hence, this gives a unique insight into the Czech fear of crime/recorded crime landscape. The aim of the paper is to investigate whether the fear of crime occurs in the areas where the recorded crime is and vice versa. The data are not compared with the population density, as such, the data do not exist for the Czech Republic. There are data of where people live, but this is a density of sleeping people, and it is quite different from real-time population density in the city. From September to November 2018, an online participatory mapping survey similar to [ 21 ] was launched focusing on the perceptions of the residents of selected cities as regards to safety and crime. The survey design was created by the authors and the municipal representatives, and included mapping the sites where people felt unsafe, the reasons why and the possible actions taken by the residents. Drawing from the tradition of environmental psychology and building on the experience of participatory mapping, the authors analyse the results with the aim of: (1) Identifying statistically significant hotspots from the responses in each city; (2) analysing whether there are any similarities in these hotspots within the cities; (3) comparing perceived crime hotspots with recorded crime hotspots. The paper is structured as follows: (1) A literature review on fear of crime introduces the research topic; (2) the methods used for data collection are presented; (3) the results of the case study of Ostrava are outlined and followed by (4) a discussion concerning the comparison of our results with other research. The paper ends with a short conclusion that may be read as an invitation for future research collaboration in the area of the fear of crime analysis. The raw data collected are available for browsing at the link provided towards the end of the paper. The data from the police are not public and therefore, cannot be browsed or shared. 2. Background Research has identified crime and anti-social behaviour as important influences on the residents’ well-being [ 22 , 23 ]. The fear of crime is a major dimension in the formation of the quality of life, yet it is often disregarded from a public policy standpoint [ 24 ]. Many studies have suggested a link between the fear of crime, social disorder and serious crime (e.g., [ 25 ]), although “broken windows” theory was criticised to focus only on the physical and forgetting about social factors influencing the fear of crime [ 26 ]. The examples of this critique can be found in [ 27 ]. The fear of crime emerged as a stand-alone concept in the 1960s, but it has been defined and measured in different ways [ 28 – 30 ]. The fear of crime is often understood as a negative emotional reaction demonstrated as the fear of criminality [ 3 ] or victimization [ 31 ]. In this paper, the authors understand the fear of crime as an emotional response towards places in a city where respondents do not feel safe, in accordance with the main question in the questionnaire (see more in Methods and Results): “Mark the places where you feel unsafe”. The fear of crime can be perceived differently by different social groups (including ISPRS Int. J. Geo-Inf. 2019,8, 401 3 of 15 age, gender, ethnicity, class and disability), and this has been discussed in various studies [ 32 , 33 ]. Curtis [ 34 ] asserted that representing feelings or beliefs about the urban environment can enable the understanding of their impact on behaviour. The studies by critical and feminist scholars have contributed new insights to the understanding of the emotions related to the perception of fear and crime in the urban environment [ 35 – 37 ]. It has been found that victimisation (being previously a victim of crime) and physical vulnerability, such as age or gender, greatly influence the fear of crime [38]. Research also shows that specific types of crime correlate with particular characteristics of urban spaces [ 39 ]. Therefore, it is important to consider these aspects when planning urban design. For example, Stankevice et al. [ 40 ] demonstrated that specialised areas and greenery in dense residential areas contribute to crime prevention. However, if these areas are combined with local centres, commercial or industrial areas, they become even more attractive to criminals. Similarly, Hillier and Sahbaz [ 41 ] argued that mixed-use street segments are relatively safe in the majority of cases and that an increased residential population neutralises the risks associated with sparse residence in mixed-use areas. Moreover, the lack of residents in public spaces and the discontinuity of public spaces increase crime because of the lack of oversight by other residents. This is in line with the pioneering findings of Jacobs [ 42 ], who claimed that urban spaces with mixed use lead to less crime because they provide more natural surveillance. The studies also show [ 7 ] that crime hotspots can change over time, but the papers also discuss the spatial periodicity of the crime itself [43]. Crime and the fear of crime can be concentrated in hotspots, which can have direct effects on the experiences of pedestrians [ 44 ]. The different types of land use might attract different types of crime [ 7 ]. Monteiro [ 45 ] found that areas with a concentration of trade tend to attract robbery and burglary. Sypion-Dutkowska and Leitner [ 10 ] found that certain areas, such as alcohol outlets, clubs and discos, cultural facilities and municipal housing, attracted crime. Moreover, commercial buildings also indirectly attracted crime. The strong influence of land use was limited to within 50 m of the surrounding area. Railways, bus stations and other public transportation nodes may also be perceived as less safe because they can encourage crimes where target density is crucial, such as pickpocketing or mugging [ 46 ]. As has been argued by Clarke [ 47 ], the fear of crime that prevents many from using public transport has a serious impact on revenues. Research from Mexico City found that around one-third of public transportation users felt unsafe or not very safe, and that public transportation is an important, albeit neglected, dimension of policies targeting the quality of life [ 24 ]. Matijosaitiene et al. stated that in Manhattan, New York, the open spaces and outdoor recreational areas generate larceny. This usually happens in periods of darkness due to the lower numbers of visitors, and during the day because of the larger number of strangers. As with the transportation hubs, the presence of strangers in urban spaces makes them less safe [ 7 ]. Thus far, most of the literature compares the fear of crime locations within one city with their respective land use, or different times of the day, or proximity to various points of interest (transportation hubs, parks, pubs, open spaces, etc.). Some research compares the fear of crime in different neighbourhoods of one city [ 48 ], but the comparison of the fear of crime with recorded crime data has been quite limited [49]. 3. Methods and Results The data collection was carried out by the researchers from September to November 2018 using social networks as well as official channels of the city administration. A web-based digital map (see Figure 1) designed by the authors as a simple crowdsourcing webpage was used to collect the data. The language of the survey was Czech (the webpage was translated into English only for the needs of this paper). For the data collection, the points were identified as the optimal feature class, and beside sketching polygons [ 50 ], the use of points was the predominant method for spatially-explicit preference mapping. The authors preferred the points over polygons as they tend to be easier for respondents to grasp and usually have a higher completion rate in the mapping activity [ 51 ], although it is necessary to have a higher amount of answers. On the other hand, polygon areas outside hotspots may represent “potential spatial errors from using polygons” ([51] p. 239). ISPRS Int. J. Geo-Inf. 2019,8, 401 4 of 15 ISPRS Int. J. Geo-Inf. 2019, 8, x FOR PEER REVIEW 4 of 15 Figure 1. Print screen of the mapping application interface used for data collection. In this research, there was only one spatial question—Where do you feel unsafe? This was followed by a fairly intensive questionnaire with non-spatial questions regarding the times of day respondents do not feel safe, the reasons for their feelings, how often they visit places where they feel unsafe and how they react to their unpleasant feelings. Furthermore, the demographics were collected concerning age, gender, the neighbourhood where they live and how safe they rate their neighbourhood compared to other areas of the city. Nevertheless, this paper analyses only the spatial part of the questionnaire. In total, the data set has 1551 respondents, who marked 3792 points during the ten week data collection period. Gender distribution is more or less balanced with a slight over-representation of women (5l.6%). On the other hand, the age distribution does not fully reflect the normal distribution as Ostrava has several universities with a large number of students. Nevertheless, the previous research about emotional mapping [52] proved that there are no significant differences between the answers of university students and the general population. Whether this is the case in Ostrava has not yet been proved. Most respondents felt unsafe (Figure 2) at the main train station, which is near the city centre and also surrounded by low income housing and derelict buildings. Similarly, people felt unsafe at another train station (Ostrava-Svinov), which is in the north-west of the city. The main street in Ostrava´s neighbourhood, Poruba was also perceived as unsafe. In the city centre, people felt unsafe in the areas leading from the train station to the Forum Nová Karolina shopping centre and along the famous Stodolní street, which has many bars and pubs. The southern part of the city had several key hotspots where people felt unsafe—among them, the Ostrava-Vítkovice and Ostrava-Hrabůvka areas. This may have mainly been caused by cheap local hostels and localities with a higher occurrence of drunks and drug addicted people. Figure 1. Print screen of the mapping application interface used for data collection. In this research, there was only one spatial question—Where do you feel unsafe? This was followed by a fairly intensive questionnaire with non-spatial questions regarding the times of day respondents do not feel safe, the reasons for their feelings, how often they visit places where they feel unsafe and how they react to their unpleasant feelings. Furthermore, the demographics were collected concerning age, gender, the neighbourhood where they live and how safe they rate their neighbourhood compared to other areas of the city. Nevertheless, this paper analyses only the spatial part of the questionnaire. In total, the data set has 1551 respondents, who marked 3792 points during the ten week data collection period. Gender distribution is more or less balanced with a slight over-representation of women (5l.6%). On the other hand, the age distribution does not fully reflect the normal distribution as Ostrava has several universities with a large number of students. Nevertheless, the previous research about emotional mapping [ 52 ] proved that there are no significant differences between the answers of university students and the general population. Whether this is the case in Ostrava has not yet been proved. Most respondents felt unsafe (Figure 2) at the main train station, which is near the city centre and also surrounded by low income housing and derelict buildings. Similarly, people felt unsafe at another train station (Ostrava-Svinov), which is in the north-west of the city. The main street in Ostrava ´ s neighbourhood, Poruba was also perceived as unsafe. In the city centre, people felt unsafe in the areas leading from the train station to the Forum Nov á Karolina shopping centre and along the famous Stodoln í street, which has many bars and pubs. The southern part of the city had several key hotspots where people felt unsafe—among them, the Ostrava-V í tkovice and Ostrava-Hrab˚uvka areas. This may have mainly been caused by cheap local hostels and localities with a higher occurrence of drunks and drug addicted people. ISPRS Int. J. Geo-Inf. 2019,8, 401 5 of 15 ISPRS Int. J. Geo-Inf. 2019, 8, x FOR PEER REVIEW 5 of 15 Figure 2. Perception of unsafe places in Ostrava. Recorded Crime Data The map based crime statistics are in Europe and often are available only on a country or regional level [53]. In the Czech Republic, the most detailed resolution of recorded crime data available so far were police districts. In Ostrava, it would be nine districts. For this research, the recorded crime data was provided by the Police Presidium of the Czech Republic. The data covered the period from January 2015 to July 2018 and contained all recorded crime offences which occurred in Ostrava and within a 5 km buffer area beyond its borders. There were 257,381 incidents in total. Each incident was classified according to a crime offence category, class and subclass, and localised in space using geographical coordinates, and in time. The pre-processing of data consisted of filtering out crime offences that were localised at police stations because of unknown actual localisation (e.g., during travel on a tram), and this applied to 4% of all the data. There were several reasons for the selection of the five categories/classes of criminal offences (violent crime, burglaries, extremism, personal theft and car robbery). Firstly, these crime offences were recommended for use in this study by the police and the Ostrava authorities. Secondly, the total number of incidents in each of the selected categories/classes were high enough for further spatial analysis, compared to many other categories/classes which had a limited number of incidents (e.g., sexual offences, arson, fraud). Thirdly, the spatial distribution of events is not influenced to the same extent by external factors as some other categories/classes (e.g., traffic accidents, driving offences). Fourthly, the selected categories/classes can be compared with the categories defined in the fear of crime data. Spatial Analytics To analyse the spatial distribution of crime and the fear of crime data, the method of spatial autocorrelation was used to detect how the incidents tend to be clustered in space. All events were aggregated to a hexagonal grid with the hexagon sides 40 m long. This side was selected after a sensitivity analysis comparing Moran’s I for hexagon sides from 20 to 120 m (with 20 m steps). Table Figure 2. Perception of unsafe places in Ostrava. 3.1. Recorded Crime Data The map based crime statistics are in Europe and often are available only on a country or regional level [ 53 ]. In the Czech Republic, the most detailed resolution of recorded crime data available so far were police districts. In Ostrava, it would be nine districts. For this research, the recorded crime data was provided by the Police Presidium of the Czech Republic. The data covered the period from January 2015 to July 2018 and contained all recorded crime offences which occurred in Ostrava and within a 5 km buffer area beyond its borders. There were 257,381 incidents in total. Each incident was classified according to a crime offence category, class and subclass, and localised in space using geographical coordinates, and in time. The pre-processing of data consisted of filtering out crime offences that were localised at police stations because of unknown actual localisation (e.g., during travel on a tram), and this applied to 4% of all the data. There were several reasons for the selection of the five categories/classes of criminal offences (violent crime, burglaries, extremism, personal theft and car robbery). Firstly, these crime offences were recommended for use in this study by the police and the Ostrava authorities. Secondly, the total number of incidents in each of the selected categories/classes were high enough for further spatial analysis, compared to many other categories/classes which had a limited number of incidents (e.g., sexual offences, arson, fraud). Thirdly, the spatial distribution of events is not influenced to the same extent by external factors as some other categories/classes (e.g., traffic accidents, driving offences). Fourthly, the selected categories/classes can be compared with the categories defined in the fear of crime data. 3.2. Spatial Analytics To analyse the spatial distribution of crime and the fear of crime data, the method of spatial autocorrelation was used to detect how the incidents tend to be clustered in space. All events were aggregated to a hexagonal grid with the hexagon sides 40 m long. This side was selected after a ISPRS Int. J. Geo-Inf. 2019,8, 401 6 of 15 sensitivity analysis comparing Moran’s I for hexagon sides from 20 to 120 m (with 20 m steps). Table 1 is comparing Moran’s I, z-value (using 999 random permutations) and the number of hexagons. It is evident that the Moran’s I value remains very similar (from 0.28 to 0.35) with the increasing hexagon side. The highest Moran’ s I is for 20 and 40 m long hexagon side, while the total number of hexagons for 40 m side is four times smaller than 20 m side, which is what supports the selection of 40 m long hexagon sides. Table 1. Change of Moran’s Iof the fear of crime data with increasing length of the hexagon side. Hexagon Side Moran’s Iz-Value N of Hexagons 20 0.35 737.20 1,217,705 40 0.35 301.04 314,944 60 0.33 177.86 135,660 80 0.28 146.95 76,500 100 0.31 112.78 49,200 120 0.33 114.23 34,200 Table 2summarises the global spatial autocorrelation using Moran’s I[ 54 ] and using queen adjacency that is defined as: I=       n Pn i(yi−y)2        ×        Pn iPn jwij(yi−y)yj−y Pn iPn jwij         (1) where iand jrefer to different hexagons, yis the data value in each, y represents the overall mean, w ij represents an element from the spatial weights matrix (equal to 1 if hexagons share the border or 0 if hexagons do not share the border). For more details, see, e.g., [55]. Table 2. Moran’s Iof recorded crime data and the fear of crime data. Type Category Moran’s Iz-Value Registered Crime Violent Crime 0.11 105.86 Burglary 0.11 101.19 Car Robbery 0.25 245.34 Personal Theft 0.14 224.80 Extremism 0.06 55.31 Fear of Crime Personal Theft 0.30 284.82 Car Robbery 0.20 206.29 Assault 0.26 241.79 Verbal Attack 0.30 269.07 Sexual Attack 0.18 187.14 Offense by drunk/drug person 0.32 257.58 Offense by homeless person 0.33 260.32 All events 0.35 301.04 All Moran’s Ivalues are statistically significant (see z-values in Table 2using 999 random permutations), although the values are rather small because many hexagons are without any event. The strongest autocorrelation from the recorded crime data is in the case of car robbery, while the other four categories have a very similar level of spatial autocorrelation. The fear of crime data show typically higher spatial autocorrelation with Moran’s Iof all events equal to 0.35. The category with ISPRS Int. J. Geo-Inf. 2019,8, 401 7 of 15 the strongest autocorrelation is the fear of offences committed by drunk or homeless people, with 0.32 and 0.33 respectively. To find areas in the city with a higher concentration of recorded crime events or the fear of crime events, a local spatial autocorrelation was used. A cluster map (sample in Figure 3) was created for all the above-selected categories of the crime data and the fear of crime data, in order to augment the significant locations with an indication of the type of spatial association [ 56 ]. Each hexagon in these cluster maps was classified into one of five categories: - High-high (dark red)—high value (number of events) in the hexagon is surrounded by high values (number of events) in surrounding hexagons relative to the overall mean; - Low-low (dark blue)—low value in the hexagon is surrounded by low values in surrounding hexagons relative to the overall mean; - High-low (light red)—high value in the hexagon is surrounded by low values in surrounding hexagons relative to the overall mean; - Low-high (light blue)—low value in the hexagon is surrounded by high values in surrounding hexagons relative to the overall mean; - Not-significant 0.75 the result is not statistically significant. ISPRS Int. J. Geo-Inf. 2019, 8, x FOR PEER REVIEW 7 of 15 significant locations with an indication of the type of spatial association [56]. Each hexagon in these cluster maps was classified into one of five categories: - High-high (dark red)—high value (number of events) in the hexagon is surrounded by high values (number of events) in surrounding hexagons relative to the overall mean; - Low-low (dark blue)—low value in the hexagon is surrounded by low values in surrounding hexagons relative to the overall mean; - High-low (light red)—high value in the hexagon is surrounded by low values in surrounding hexagons relative to the overall mean; - Low-high (light blue)—low value in the hexagon is surrounded by high values in surrounding hexagons relative to the overall mean; - Not-significant 0.75 the result is not statistically significant. Hexagons classified as members of high-high clusters represent the core areas with a significantly higher number of events and high-low cluster highlights the locations with spatial outliers (higher number of events) compared to surrounding hexagons (see Figure 3). Figure 3. Cluster map of fear of crime events (all events). A combination of two different cluster maps was used to compare the locations with higher intensities of crime and the fear of crime and to find areas where the fear of crime is justified by a higher intensity of recorded crime (Figure 4). Each cluster map classified all hexagons into five different categories, so the total number of combinations for the two cluster maps was 25 categories. In order to make the final visualisation comprehensive, categories were organised into nine resulting groups. The most important category is the combination of high-high × high-high clusters because these represent the only hexagons with a significantly high level of recorded crime and the fear of crime in the hexagon itself and in the adjacent hexagons (description in Table 3). All hexagons that were not classified into a high-high cluster in any of the two combined cluster maps were grouped into one class that was not displayed on the final maps. These hexagons represent the locations with a low or no significant intensity of events (recorded crime and fear of crime). The only exception is the combination of high-low × high-low classes where a hexagon contains a higher number of events Figure 3. Cluster map of fear of crime events (all events). Hexagons classified as members of high-high clusters represent the core areas with a significantly higher number of events and high-low cluster highlights the locations with spatial outliers (higher number of events) compared to surrounding hexagons (see Figure 3). A combination of two different cluster maps was used to compare the locations with higher intensities of crime and the fear of crime and to find areas where the fear of crime is justified by a higher intensity of recorded crime (Figure 4). Each cluster map classified all hexagons into five different categories, so the total number of combinations for the two cluster maps was 25 categories. In order to ISPRS Int. J. Geo-Inf. 2019,8, 401 8 of 15 make the final visualisation comprehensive, categories were organised into nine resulting groups. The most important category is the combination of high-high × high-high clusters because these represent the only hexagons with a significantly high level of recorded crime and the fear of crime in the hexagon itself and in the adjacent hexagons (description in Table 3). All hexagons that were not classified into a high-high cluster in any of the two combined cluster maps were grouped into one class that was not displayed on the final maps. These hexagons represent the locations with a low or no significant intensity of events (recorded crime and fear of crime). The only exception is the combination of high-low × high-low classes where a hexagon contains a higher number of events (recorded crime and fear of crime) compared to the overall mean value. While there were only a few members of this class, they were grouped with the high-high ×high-high class. The remaining combinations of a high-high cluster in one cluster map and an insignificant, low-low, low-high or high-low in the second cluster map make up the other eight categories. These eight categories can be divided into two groups, depending on whether the recorded crime or the fear of crime (hues of green were used in the maps below) belongs to the high-high cluster. In cases where the fear of crime belonged to the high-high cluster, the hues of blue were used in the maps below. These hexagons represent the locations where the fear of crime is not justified by the recorded crime data. In contrast, the green hues highlight hexagons with a high-high cluster membership in recorded crime data only. At these locations, people did not declare a higher level of the fear of crime despite the significantly high level of criminal activity. The hue of blue or green corresponds to the intensity of the non-high-high variable (fear of crime or recorded crime). ISPRS Int. J. Geo-Inf. 2019, 8, x FOR PEER REVIEW 9 of 15 Figure 4. Cluster map comparing the fear of personal thefts and thefts as recorded crime incidents. The combinations of the different types of crime and the fear of crime were also tested, and their correlations were proved. The highest correlation from all combinations (r = 0.22) is for the combination of fear of being attacked by a drunk or drug addicted person(s) and personal theft (recorded crime). In the eastern part of the map (Figure 5), two locations are visible, that were previously described, but have a higher number of red hexagons. Another hot spot area is in the western part of the map (Figure 5), around Mariánské square, an area with the presence of socially excluded persons. A larger area with an unjustified fear of crime can be found in the central part of the map. The main reason for the increased level of fear is a cheap local hostel. Figure 4. Cluster map comparing the fear of personal thefts and thefts as recorded crime incidents. ISPRS Int. J. Geo-Inf. 2019,8, 401 9 of 15 Table 3. The description of classes used in a combined cluster map. Combinations (Fear of Crime ×Recorded Crime) Description high-high ×high-high High level of fear in the hexagon and adjacent hexagons (compared to the overall mean value value) and high level of crime intensity in the hexagon and in adjacent hexagons (compared to the overall mean value) high-high ×not significant High level of fear in the hexagon and adjacent hexagons (compared to the overall mean value value) and not a significant level of crime intensity in the hexagon and in adjacent hexagons (compared to the overall mean value) high-high ×low-low High level of fear in the hexagon and adjacent hexagons (compared to the overall mean value) and low level of crime intensity in the hexagon and in adjacent hexagons (compared to the overall mean value) high-high ×low-high High level of fear in the hexagon and adjacent hexagons (compared to the overall mean value) and low crime intensity in the hexagon together with high intensity of crime in adjacent hexagons (compared to the overall mean value) high-high ×high-low High level of fear in the hexagon and adjacent hexagons (compared to the overall mean value) and high crime intensity in the hexagon together with low intensity of crime in adjacent hexagons (compared to the overall mean value) The combined cluster map comparing the fear of personal thefts, with thefts as recorded crime incidents shows that the city centre of Ostrava has two high crime density areas. The first is along Stodoln í street, famous for its large number of pubs and concentration of young (and often drunk) people during Friday and Saturday nights. The second place highlighted in red, is the area of the public transport hub (combination of tram, trolleybus and train) with a supermarket, a cheap hostel, and a hazardous area below the bridge with a high occurrence of aggressive individuals or groups. In general, the green hues predominate on the map, and this is caused by a significantly higher number of recorded crime events than the selected locations based on the questionnaires. Hexagons with a darker green or blue colour are evidently closer to red hexagons, and the light green colour represents hexagons with a very small number of events. The combinations of the different types of crime and the fear of crime were also tested, and their correlations were proved. The highest correlation from all combinations (r=0.22) is for the combination of fear of being attacked by a drunk or drug addicted person(s) and personal theft (recorded crime). In the eastern part of the map (Figure 5), two locations are visible, that were previously described, but have a higher number of red hexagons. Another hot spot area is in the western part of the map (Figure 5), around Mari á nsk é square, an area with the presence of socially excluded persons. A larger area with an unjustified fear of crime can be found in the central part of the map. The main reason for the increased level of fear is a cheap local hostel.