Background

This data article is a downscaled replication of my MSc Thesis using updated data, completed in Rmarkdown. The aim is to show how deprivation indices affect different crime rates in london. The deprivation indices are provided on an Lower Layer Super Output Areas level (from here on refered to as LSOA(s) - more on these below) and include indices such as income and employment deprivation, more on this data here: https://www.gov.uk/government/statistics/english-indices-of-deprivation-2019

The crime data we are using is found in the london DataStore and provides LSOA level reported crime data found here: https://data.london.gov.uk/dataset/recorded_crime_summary

All data used in this analysis is completely open source.

The question I ask is, does deprivation have any relation to reported crime, and if so which types of deprivation have the strongest relation.

Lower Layer Super Output Areas (LSOA)

Before we get into the different deprivation indices, I think its important that we understand what LSOAs are, considering the indices are created for these areas. LSOAs were created with the specific function of collection of “small-area” statistics, as such, the LSOAs are intrinsically designed to be small areas of population size between 650-1500 as of today there are 32,844 LSOAs in the UK. Below is a histogram showing the distrubtion of populations in LSOAs thorughout the UK

Deprevation Indices

Deprivation indices are provided by the the UK government on an open source basis and rank each LSOA form 1-32844, 1 being the most deprived area and 32844 being the least deprived. The indices are; Income, Employment, Crime, Barriers to Housing and Service, Education, Health and Living Environment. In addition to a Multiple Deprivation Index which weights the above indices.

A more complete overview of this data can be found here: https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/835115/IoD2019_Statistical_Release.pdf

Here we will focus on economic indicators such as employment depriavtion (Emp) and income deprivation (Inc) and barriers to housing and service deprivation ()

Crime Data

The Crime Data (shown below) shows LSOA Code the borough name and the major/minor category of crime and then the number of reported occurences for each month between 2020-03 and 2022-02.

datatable(head(CrimeDat),options = list(scrollX = TRUE))

This analysis will initially focus on major categories of crime rather than the sub minor categories of crime, hopefully we will be able to explore minor categories in the future. However, For the sake of time we wont be looking at all of the major categories of crime (As I did in my thesis), given there are many categories and many indicators I feel I wont be able to analyse all of these as in depth as they would deserve. So the Major categories of crime I will be looking at specifically will be; Violence Against the Person, Arson and Criminal Damage & Drug Offences (crimes often focused on in the media).

The Major Categories of crime are detailed below.

unique(CrimeDat$Major.Category)
##  [1] "Arson and Criminal Damage"           
##  [2] "Burglary"                            
##  [3] "Drug Offences"                       
##  [4] "Miscellaneous Crimes Against Society"
##  [5] "Possession of Weapons"               
##  [6] "Public Order Offences"               
##  [7] "Robbery"                             
##  [8] "Theft"                               
##  [9] "Vehicle Offences"                    
## [10] "Violence Against the Person"

Transforming the Data

Our aim here is to show how deprivation indicators may or may not affect reported crime stats, as such I believe it is important to see if there are any other variables in our data which may be affected the reported crime statistics and adjust for them accordingly - Having already worked with this data, and in the interest of not going on forever, what will be adjusting for is population.

We will look at how population affects crime rates to see if any adjustments need to be made for this (it may be the case that LSOAs with larger population may have higher crime rates)

Population density has a significant enough affect on recorded crime in each of our crime stats, so we will adjust for this by taking the number of crimes per 1000 people in a given LSOA, this normalising will reduce the affect population has on the recorded crime stats.

Analysis

So how do the different deprivation indicators & crime rates relate? We will be using OLS to observe any relations. First we will look at how population affects crime rates to see if any adjustments need to be made for this (it may be the case that LSOAs with larger population may have higher crime rates)

Given that we are working with social analysis we are not expecting the deprivation indicators in these models to be able to account for all the variance in the data, however we can see there is at least some negative trend between the crimes per 1k population and the ranking in almost all the indexes most significantly in the multiple deprivation, income deprivation and employment deprivation.

Given the large variability in the condition in LSOAs we can often get quite noisy data e.g.some LSOAs may be home to major transport routes and see large amount of people pass through which could inflate crime rates it may help to look at cohorts of crime rates which is so helpully provided to us in deciles, to help reduce some of this noise. A good example of this variability is the LSOA Westminster 013E which has almost 4x the reported crimes as the population of the lSOA ~9000 reported crimes in an area of only 2500 people.

When looking at cohorts (shown in bar charts) it becomes evident that for all the crimes we looked at we see a higher crime rate in the lower deciles (1 being the bottom 10% and 10 being the top 10%) in many cases the bottom 20% of deprivation deciles will have a crime rate per 1k ~ double of the crime in the top 20%. Whilst looking individual LSOAs we saw a much less distinct relationship, looking at these cohorts however we can see that the deprivation decile one belongs to can have an impact on the crime rates per 1K people quite significantly.

In short when looking at deciles economic indicators can play a role in the crime rates one could expect to face between 2020 and 2022

The fun part

Making a map. It can often be more insightful to conduct such analysis spatially, and give people a more personal sense of the data.For this LSOA Data we have been lucky enough to have been provided with the polygonal data for the boundaries for each LSOA and we will be using the R library Leaflet to plot this.

## Warning in OGRSpatialRef(dsn, layer, morphFromESRI = morphFromESRI, dumpSRS =
## dumpSRS, : Discarded datum OSGB_1936 in Proj4 definition: +proj=tmerc +lat_0=49
## +lon_0=-2 +k=0.999601272 +x_0=400000 +y_0=-100000 +ellps=airy +units=m +no_defs

Here we have a map of all LSOAs in color coded for multiple deprivation decile, you can hover over each tile and get the crime rate per 1k population for all crimes, the name of the LSOA and the deprivation decile to which it belongs.

So there we have a totally not comprehensive analysis of the relationship between crime rates and deprivation indicators in London. And a cool map.

If you think any of the code behind this document could help you out hit me up ==>