Views!
HUNDREDS of daily LOCAL views - THOUSANDS of monthly LOYAL readers! THE MOST POPULAR ATTRACTION IN HARTIES!
Published:  2014-12-29 Views:  1097
Author:  karelnel
Published in:  Security/Sekuriteit
Harties Crime Statistics Up To 2013- Part 8 : Analysis

Harties Crime Stats up to 2013 – Part 8 : Analysis

The previous articles in this series portrayed the official crime stats of the SAPS at the end of 2013 in graphical format. Although many things about the huge data-set is not known, it nevertheless contains  very valuable information. Apart from verifying the accuracy of the source data, what can one do with such a data-set to make the underlying trends visible?

Firstly one can plot the data, as it represents a time series. One can usually see easily whether there are any trends over time and if so, there may be assignable reasons that can be sought for. If the data and the trends are fairly consistent over the years it implies that the specific crime mechanisms are somehow embedded in the fabric of the society, e.g. the size and composition of the stratum of society involved in crime. If the data is erratic and no trend can be distinguished, it may be that the data sample size is too small or that the crime incidents are opportunistic, random events executed in a hit-and-miss fashion. (No established pattern yet)

Secondly one can calculate the annual average values. Although this cannot be used as a predictor of future expectations, it does give an indication of the magnitude of the specific type of crime. In comparing averages over a period of time one can with a fair amount of confidence determine which crime types happens more frequent than others on a consistent basis. This parameter can become an important building brick in focusing the crime-fighting effort on the crimes where it can make the greatest possible improvement.

Thirdly one can sort the data (averages). Invariably one will get what is called a ‘Pareto-distribution’ or some approximation thereof. This is a fairly universal distribution (also called the 20/80 distribution) which indicates that 20% of the individuals makes 80% of the total impact. For example – 20% of society normally possess 80% or more of all resources of that society. Figure 1 below will indicate that this distribution can also be found in the Hartbeespoort crime data: the first 8 categories combined represent more than 80% of all crimes. It is abundantly clear from the graph that the statistics of crime  can be drastically improved by reducing the first two categories, although that will not necessarily reduce the impact of crime on the society because for example, the first category is a relatively petty type of crime. This distribution however already gives an indication where crime combatting measures should focus.

 
   

Fig 1: 10 Year Average Ranking of Crime types for Hartbeespoortdam

A more sensible way to determine the most significant crime types for the community can be derived from the field of risk analysis and – management. In risk management it is customary to determine the probability of a risk occurring and multiply that figure with the severity of the consequences if that risk occurs and that is called the impact of the risk.  This product is then ranked and the most important ones are addressed. Analogous the frequency of the occurrence of a crime type can be multiplied by the severity of the crime type and that product can be called the “Impact” of that crime on society. The crimes with the greatest “Impacts” should then be addressed first. How can the severity of the consequences of a specific type of crime be determined? According to literature it is a difficult task but one way is to use the legal penalties of the type of crime as an indication. A better way would be to call on a panel of knowledgeable people to do a pairwise comparison analysis to determine the relative impacts of the different types of crimes, perhaps in terms of a ‘severity index’ (rated figure out of 10 or 100).

Example: Assume that such a severity analysis was done on the crime types, and the results were as found in Table 1 below. When the annual average and the severity index are multiplied and the results sorted, a slightly different Pareto distribution is found – indicating the ranking of crime IMPACTS. See Figure 2. This gives a more sensible indication on which crimes to concentrate to optimally reduce the impact of crime on society when resources are limited. As the high impact categories are reduced by effective policing, the other categories will become relatively more important and will gradually receive more attention until all crime types have been addressed over a period of time.

Table 1: Annual averages and Severity Indices for Crime Types (Example)

Crime Type

Annual Average

Hartbeespoortdam

Severity Index

(Assumption)

Murder

9

10

Total Sexual Crimes

39

8

Attempted murder

21

8

Assault with the intent to inflict grievous bodily harm

149

7

Common assault

184

6

Common robbery

47

5

Robbery with aggravating circumstances

106

6

Arson

10

4

Malicious damage to property

162

4

Burglary at non-residential premises

117

4

Burglary at residential premises

581

5

Theft of motor vehicle and motorcycle

96

3

Theft out of or from motor vehicle

190

2

Stock-theft

12

2

Illegal possession of firearms and ammunition

11

4

Drug-related crime

56

4

Driving under the influence of alcohol or drugs

26

3

All theft not mentioned elsewhere

607

4

Commercial crime

95

4

Shoplifting

27

3

Carjacking

7

4

Truck hijacking

2

4

Robbery at residential premises

28

4

Robbery at non-residential premises

11

3

Culpable homicide

18

8

Public violence

1

3

Crimen injuria

26

2

Neglect and ill-treatment of children

2

5

Kidnapping

2

5

 

 

 
   

Figure 2: Crime Impact Ranking for Hartbeespoort based on assumed severity indices

Up to this point the stochastic nature of the crime statistical data was not considered.  The employment of statistical variability analyses is a necessary and powerful way of understanding and using the data for  forecast and early warning purposes. Data (like crime incidences) can vary within certain limits without implying a significant variation – this is called statistical variability. If variability falls outside these limits, then the changes become significant. A further article will deal with these methods. Sufficient to say that the Normal distribution is used on many occasions as a handy approximation of the nature of such type of data, and therefore the parameter of the ‘standard deviation’ is used. In a nutshell – if a standard deviation is calculated for a data-set, and an upper limit is calculated by adding 2 times the standard deviation to the average of the data set, then any point that falls beyond this upper limit  does  vary significantly from the rest with a confidence level of about 95%. This is a powerful tool to determine whether there are significant changes in crime trends when the incidence figures vary.

The standard deviation can also be used to calculate variations for smaller or larger time periods. When a standard deviation is calculated for a year (12 months) as x , then the standard deviation for a month can be expected to be x/√(12). This principle can be used to construct charts for the most important crime types which can be updated on a monthly basis and will give early warning when trends change significantly.

CONCLUSION:

This article explored a few fairly simple but powerful analysis techniques that could be employed by a crime analysis work-group for the benefit of the society. A following article will deal further with the measurement of crime and the fighting of crime.

---000---

KPJ Nel

2014-12-29


Directors: S.A.M Smit & J.F Koster - Copyright ©2018 HartiesNet (Pty) Ltd - our Disclaimer