Crime Prevention Based on the Strategic Mapping of Living Conditions
Round 1
Reviewer 1 Report
This paper presents a GIS-based model for the measurement and mapping of an index of living conditions in urban residential areas across Sweden. The idea is good. But the result is obtained by simply adding several factors with same weight. The paper lacks a detailed analysis of these variables and their importance. And here are still many issues to be considered carefully.
- English should be very carefully edited. This version is very hard to learn.
- The model should be introduced in Abstract section.
- In line 33: “However” is not used properly.
- In line 70: is SCB short for Statistics Sweden?
- In lines 165-168: it is hard to understand.
- In Table 1: you should check the table carefully, especially “Classification method and Classes”. Such as “Class 1; > 49”, “Class 3 ≤ 23”….
- In Table 1: it is hard to understand, class 3: ≤23, class 4: ≤10, class 5:>3. Is there any relationship between these classes? What about value which is smaller than 3.
- In Table 1: how do you get these values, such as 49, 23, 10, 3, 43……
- In Table 1: why the Income is from 2016? Others are from 2018.
- In line 162: “with the values in the lowest class being converted to the value 1” means what? From Table 1, I cannot find out the relationship between the lowest class and 1.
- why the model in Figure 2 is different from the process of experiments.
- In Table 2: “Classification method and classes”: where are classes? And what is No of grids?
- The caption of Table 2 need to be modified.
- In lines 264-266: it is hard to understand.
- In Figure 4-6: when was the police assessment made?
- In line 390: “Another reason for these not having been included in the police’s assessment may be that some areas with poorer living conditions have a limited crime problem according to the police’s criteria.” Has the criteria been investigated to support your ideas? Or the model failed in these areas?
Author Response
Point 1: (Major) This paper presents a GIS-based model for the measurement and mapping of an index of living conditions in urban residential areas across Sweden. The idea is good. But the result is obtained by simply adding several factors with same weight. The paper lacks a detailed analysis of these variables and their importance. And here are still many issues to be considered carefully.
Response 1: Thank you for the comments and review. Very good comments overall. As a review should be. We have conducted a more detailed literature review according to your comments focusing on social disorganization theory and collective efficacy theory and more clearly related these theories / approaches to our perspective on living conditions. We also see our perspective as an extension of this criminologically and geographically oriented research, which is also mentioned in the text, including references to other studies that have developed similar index measures to support the geographic analysis of crime. This review also includes more key references - both from international perspective and Swedish context, and edits throughout the paper. We have also added references that better justify the variables we use when measuring living conditions. The review also includes corrections after minor comments (every minor commented on by the authors) from you, own corrections (e.g. in figures and tables),and a language review.
Point 2: English should be very carefully edited. This version is very hard to learn.
Response 2: This version has been language reviewed.
Point 3: The model should be introduced in Abstract section.
Response 3: Not sure about this comment, we have gone through the abstract a couple of times and we think it is clear that the model is introduced and mentioned in several places?
Point 4: In line 33: “However” is not used properly.
Response 4: "However" has been removed
Point 5: In line 70: is SCB short for Statistics Sweden?
Response 5: Correct, the Swedish name for the authority – Statistiska Centralbyrån - has been added to explain the abbreviation.
Point 6: In lines 165-168: it is hard to understand.
Response 6: We agree, these lines were unnecessarily complicated, one sentence has been simplified to: “This division is based on Statistics Sweden's official classification of income intervals (gross income before tax):”
Point 7: In Table 1: you should check the table carefully, especially “Classification method and Classes”. Such as “Class 1; > 49”, “Class 3 ≤ 23”….:
Response 7: Good comment. The division of the classes' intervals has been clarified, e.g. Class 3: 23 - 49, instead of ≤ 23, also some errors have been corrected.
Point 8 In Table 1: it is hard to understand, class 3: ≤23, class 4: ≤10, class 5:>3. Is there any relationship between these classes? What about value which is smaller than 3.
Response 8: See comment above, the division of the classes' intervals has been clarified.
Point 9: In Table 1: how do you get these values, such as 49, 23, 10, 3, 43…… In Table 1: why the Income is from 2016? Others are from 2018:
Response 9: See comment above, the division of the classes' intervals has been clarified. The classification in different class intervals is based on what the distribution of values looks like. The most common classification method in this study is based on standard deviations from the mean value, when this method does not work, other classification methods are used, e.g. Statistics Sweden's class standard division of income according to people's purchasing power (quantiles + 1 added class) is used. The classification methods for each variable are mentioned in Table 1. Regarding the difference between the years for the collection of the variables for the years 2016 and 2018, the latest available year was used for each variable. Average income differences in the geographical grids across Sweden differ very little between years that are close to each other, e.g. 2016-2018, therefore 2016 values for income across Sweden are judged to be reliable in the model.
Point 10: In line 162: “with the values in the lowest class being converted to the value 1” means what?
Response 10: This line (previous line 162) has been reformulated till: “the values in the lowest class being assigned the numerical value 1” We have also tried to clarify which elements the steps contain by reformulating a sentence before: “The four steps, which include value assignment and equal weighting, are as follows:”
Point 11: From Table 1, I cannot find out the relationship between the lowest class and 1.
Response 11: Good point, the reference to Table 1 is moved to a more relevant sentence (three lines forward).
Point 12: why the model in Figure 2 is different from the process of experiments.
Response 12: Figure 2 shows how the different concepts and conditions (theoretically and conceptually) are related to each other. Figure 2 also indicates the steps of the process, but without describing the calculation details (as described in the method section). We agree that the reference to Figure 2 (which is theoretical) is a bit unclear. The following sentence has been added instead of "(see Figure 2)": “These conditions are described more extensively in the theoretical model in section 3 and Figure 2 below”
Point 13: In Table 2: “Classification method and classes”: where are classes? And what is No of grids?
Response 13: See clarifications above. The classification methods mentioned in Table 1 (standard deviations and quantiles) are conventional techniques used in GIS. The number of grids refers to the number of mapped 250 x 250 meter grids for each variable used. This is reformulated in the caption of Table 2 (see below).
Point 14: The caption of Table 2 need to be modified.
Response 14: Thanks for the comment; we have reformulated the caption of Table 2 to: “Intermediate descriptions, classification of variables, classification methods, classes, number of mapped 250 x 250 meter grids for each variable, motivation of classes, source and comments.”
Point 15: In lines 264-266: it is hard to understand.
Response 15: Clarifications with examples have been made in the text. Edits to Figure 2 have also been made.
Point 16: In Figure 4-6: when was the police assessment made?
Response 16: 2018-2019, this is mentioned in the material section, but we have also added it to the caption.
Point 17: In line 390: “Another reason for these not having been included in the police’s assessment may be that some areas with poorer living conditions have a limited crime problem according to the police’s criteria.” Has the criteria been investigated to support your ideas? Or the model failed in these areas?
Response 17: The comment is good, we have reformulated the sentence and added a forward-looking line to the discussion: “Another reason for these not having been included in the police’s assessment may be that some areas with vulnerable living conditions have a limited crime problem according to the police’s criteria. If this is the case, it represents a positive outcome, and these areas need to be studied in more detail in order to establish which local conditions are restricting the development of crime. A possible way forward is to develop the model by also including variables that can measure e.g. the institutional capacity of society or the collective efficacy of the local community.”
Author Response File: 
Author Response.docx
Reviewer 2 Report
This paper is well written and examines how 'living conditions' may relate to areas with high crime exposure. I very much enjoyed reading this.
I recommend a few minor revision
1) The first is linked to the theoretical underpinning of this work. Whilst you identify several socio-economic factors that might relate to increased levels of crime (references 1-16) I think you should also reflect how this relates to some of the crime-place theories that underpin this. For example - how does this study relate/add value to earlier works on social disorganisation, and more recently on social cohesion and collective efficacy. How similar are your variables - and given Swedish context are you measuring any of these. I think this is important as these have been widely studied - and I would expect some discussion of why you are examining 'living conditions' as opposed to the more established theories. Are you proposing this as an alternative way of identifying high crime areas, or complimentary alternative or other.
2) A second issue to potentially discuss - perhaps as further research is crime is often linked to opportunities afforded by routine activities- linked to location of different land use types. It may be that you 61% overlap, and 38% high crime areas would be improved if you took into account local land use - some of this crime may be more to do with mobility and population movements within places - and some may be better explained by living conditions/socio-economic factors
3) It would be useful to say a bit more about the data used and methods the police use in creating their crime exposure measure. Is this based on combination of recorded crime data and police intelligence. Does in include all crime types or is it primarily driven by violent crime, or volume crime such as burglary. Generally spatial patterns of crime differ by crime type - as do underlying predictive variables. Again this might be explanation for 61% overlap, as living conditions might be better predictor for more residential/local crime as opposed to those linked to mobility (further away from where people live).
Author Response
Point 1: This paper is well written and examines how 'living conditions' may relate to areas with high crime exposure. I very much enjoyed reading this. I recommend a few minor revision The first is linked to the theoretical underpinning of this work. Whilst you identify several socio-economic factors that might relate to increased levels of crime (references 1-16) I think you should also reflect how this relates to some of the crime-place theories that underpin this. For example - how does this study relate/add value to earlier works on social disorganisation, and more recently on social cohesion and collective efficacy. How similar are your variables - and given Swedish context are you measuring any of these.
Response 1: Thank you for the comments and review. Very good comments overall. We have conducted a more detailed literature review according to your comments focusing on social disorganization theory and collective efficacy theory and more clearly related these theories / approaches to our perspective on living conditions. We also see our perspective as an extension of this criminologically and geographically oriented research, which is also mentioned in the text, including references to other studies that have developed similar index measures to support the geographic analysis of crime. This review also includes more key references - both from international perspective and Swedish context, and edits throughout the paper. We have also added references that better justify the variables we use when measuring living conditions.
Point 2: I think this is important as these have been widely studied - and I would expect some discussion of why you are examining 'living conditions' as opposed to the more established theories. Are you proposing this as an alternative way of identifying high crime areas, or complimentary alternative or other.
Response 2: Good comment: we have added a section in the introduction that relates this paper's approach to previous criminological approaches; “This paper’s approach to living conditions is based on previous studies and theories of social stress [refs], which have been operationalized into geographically measurable variables and indices in order to analyze relationships between living conditions and expressions of social disorder. For example, different indices of living conditions have previously been applied to and found to explain geographical differences in arson and crime in large Swedish cities [refs]. Thus, the theoretical and methodological approach used in this paper may be seen as an extension of existing criminological approaches, such as those based on social disorganization and collective efficacy theory, which ad-dress similar geographical differences in crime and which use similar indices to measure vulnerability [refs]”
Point 3: A second issue to potentially discuss - perhaps as further research is crime is often linked to opportunities afforded by routine activities- linked to location of different land use types. It may be that you 61% overlap, and 38% high crime areas would be improved if you took into account local land use - some of this crime may be more to do with mobility and population movements within places - and some may be better explained by living conditions/socio-economic factors
Response 3: Thanks for this comment: This perspective has been added to the discussion section as a proposal for further research!
Point 4: It would be useful to say a bit more about the data used and methods the police use in creating their crime exposure measure. Is this based on combination of recorded crime data and police intelligence. Does in include all crime types or is it primarily driven by violent crime, or volume crime such as burglary. Generally spatial patterns of crime differ by crime type - as do underlying predictive variables. Again this might be explanation for 61% overlap, as living conditions might be better predictor for more residential/local crime as opposed to those linked to mobility (further away from where people live).
Response 4: Good comment again: It is, as you mention, a combination, but with a focus on more subjective assessments, which is stated in the paper. The challenge is, (work in progress within the police), to make the assessment more data-driven and linked to people's living conditions, police observations and other important data that can be collected from societal actors, e.g. incident reports from municipal officials and the rescue service. This is a perspective that may be developed in the next step/paper.
Author Response File: Author Response.docx
Reviewer 3 Report
This is a highly timely manuscript not only addressing a somewhat burning issue in Sweden but also in several other countries. It also demonstrates clearly how careful conduct (data, basic methods, etc.) of the use of GIS can provide significant input to the understanding of un-wanted social behaviour in the society. It is also easy to see how such an approach can benefit in attempts to understand and control for extreme behaviour such as terrorims or recruitments for such activities. Shedding light in the shadows early on is perhaps the best way forward in providing citizens with safe living environments - one of the basic functions os any society.
Author Response
Thank you for the comment!
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
You should check the full text carefully. And here are still some issues to be considered carefully.
1.In Table 1: you should check the table carefully, especially “Classification method and Classes”. Such as “Class 1; > 49”, “Class 3 ≤ 23”….
Why does this error still exist?
2.In Table 1: it is hard to understand, class 3: ≤23, class 4: ≤10, class 5:>3. Is there any relationship between these classes? What about value which is smaller than 3.
I noticed that you have clarified the division of the classes' intervals, but does this have any effect on the experimental results? Additionally, why are the boundary values of variables different. Specifically, in income, Class 1: 0 - 163,617; Class 2: 163,618 - 280,169; Class 2: 280,170 - 400,577; but in other variable, such as employment rate, Class 1: ≤ 43; Class 2: 43 - 59; Class 2: 59 - 75.
3.In Table 2: “Classification method and classes”: where are classes? And what is No of grids?
You'd better check it carefully. In Table 2, I can only find the classification method, but can not find the classes. And is “No of grids” the same as “No. of grids”?
4.The caption of Table 2 need to be modified.
Are you sure you have reformulated? You'd better check it carefully. And where is “living conditions index”.
Author Response
Dear reviewer,
Thank you for being so careful in your review, great job! We also apologize for missing or misunderstanding some of your comments in the first review round. We hope we understood correctly this time.
1.In Table 1: you should check the table carefully, especially “Classification method and Classes”. Such as “Class 1; > 49”, “Class 3 ≤ 23”….
Why does this error still exist?
Response 1: To clarify the classification intervals for all used variables, all characters (≤ etc) in Table 1 are changed to numbers in intervals, e.g. 49.1 – 100 % (with the exception of median income class 5,” > 500,000 SEK” has been changed to "over 500,000 SEK". Decimals have also been added to clarify the class divisions, e.g. 49.1
2.In Table 1: it is hard to understand, class 3: ≤23, class 4: ≤10, class 5:>3. Is there any relationship between these classes? What about value which is smaller than 3.
Response 2: See response 1
I noticed that you have clarified the division of the classes' intervals, but does this have any effect on the experimental results? Additionally, why are the boundary values of variables different. Specifically, in income, Class 1: 0 - 163,617; Class 2: 163,618 - 280,169; Class 2: 280,170 - 400,577; but in other variable, such as employment rate, Class 1: ≤ 43; Class 2: 43 - 59; Class 2: 59 - 75.
Response 2: We hope that response 1 can help clarify some of this. The boundaries within the variable divisions are based on each variable and dataset and the value distribution across Sweden. In most cases, we have used classification according to standard deviations. With regard to the income interval, it is based on Statistics Sweden's official classification of the median income intervals (gross income before tax) of the population aged 20 years and over. This is also clarified in the text (on page 5 and in table 1). Decimals have also been added to clarify the class divisions. New tests have been performed and they show the same results as before (ie there are no effect on the experimental results).
3.In Table 2: “Classification method and classes”: where are classes? And what is No of grids?
You'd better check it carefully. In Table 2, I can only find the classification method, but can not find the classes. And is “No of grids” the same as “No. of grids”?
Response 3: Thank you so much for being observant. We must admit that we missed this comment in the first review round. We apologize for this. The table 2 should not contain classes. The class division for the index values for living conditions is shown in Figure 3 and in the maps in Figures 4 -7 . No. of grids means number of mapped 250 x 250 grids in for each index (see response 4 below).
4.The caption of Table 2 need to be modified.
Are you sure you have reformulated? You'd better check it carefully. And where is “living conditions index”.
Response 4: The caption is now modified to: “Classification methods for different indices (economic, family and living conditions) and the number of mapped 250 x 250 meter grids per dataset” to better fit the contents of Table 2.
Best regards,
the authors
Author Response File: Author Response.docx
