Figuring it out:

Getting to know police crime
statistics

Crime statistics are, and always will be, a source of debate. The moratorium on the release of police crime statistics announced in August 2000 has fuelled this debate and increased the number of questions about crime trends and the accuracy of the statistics. Available evidence suggests that at national, provincial and area level the statistics are accurate, bearing in mind the limitations facing police statistics the world over.

After nearly a year, the moratorium on the release of crime statistics was lifted on May 31st 2001. The new statistics will come under intense scrutiny. To appreciate what the new figures are saying, it will be important to understand how police statistics are collected, what they tell us, how they should be used and what their limitations are.

While the moratorium has undoubtedly harmed the legitimacy of the police and their crime statistics, it is also true that public knowledge about official crime statistics is limited.

What do police statistics tell us?

The most important feature of police crime statistics is that they tell us about crimes that the public report and the police record. As a result, crime statistics often say more about reporting patterns and police procedure than about actual crime levels. Nevertheless, police statistics provide useful information on the following:

How crime levels change over time

By analysing the number of crimes recorded by the police in a given time period and in a particular geographic area, changes in crime levels over time can be tracked. The significance of the trends will depend on how much data is available: the higher the numbers being analysed and the longer the time period, the more accurate the trends are likely to be.

How crime rates compare in different areas

Police data can be used to compare crime rates in different suburbs, cities, provinces and countries. However, in each area police statistics must be supplemented with data on population size, the number of registered vehicles, or the number of stock owned, depending on the crime types that are being compared.

This is the most accurate way of comparing crime between two areas because it takes into account the potential for victimisation. It is inaccurate to compare the raw number of murders in Soweto with those in Benoni, because there may be many more people living in Soweto. This means the potential for murder is much higher there than in Benoni.

About the nature of certain crimes

The police record a range of details about when and where crimes occur, who the victims are and, when they are known, who the perpetrators are. In some cases information is also recorded about the types of weapons used, and other factors such as the use of alcohol or drugs by the victim or perpetrator. This information helps to determine whether, for example, most assaults are related to alcohol consumption or the number of murders committed with a gun is increasing.

What police statistics do not tell us

Official crime data, like that gathered through any other means, have their limitations. It is ultimately the responsibility of those who use police statistics to be aware of these gaps and endeavour to fill them in their analysis of crime. Police crime statistics do not tell us:

About crimes not reported to the police

Because the police depend on the public to report offences, it is unrealistic to expect the police to have information about all crime. Crime types that are typically underreported include petty theft, fraud, corruption, sexual assault and crimes against children. Victim surveys — which ask a representative sample of people in a particular area about crimes they have experienced over a specific time period — can supplement police data to fill some of these gaps.

About the causes of crime

On their own, police statistics do not show what causes crime. These statistics, like those gathered through victim surveys, indicate trends and areas for further investigation. For example, an analysis of the number of murders recorded on a monthly basis in South Africa shows a consistent and dramatic peak over the Christmas and Easter holiday periods. This shows that murder may be related to the gathering of people, friends and family over holiday periods, together with the consumption of alcohol. The police data however neither proves that this is the case, nor that alcohol actually causes crime.

About police performance

On their own, crime statistics are not always helpful in assessing police performance, for three reasons:

Crime statistics do not necessarily reflect what the police are doing about crime and how effective these strategies are. For example, a six-month police operation to gather intelligence, investigate and arrest members of a car theft syndicate may not have a major impact on vehicle theft statistics unless the syndicate’s activities were confined to a small, clearly identified locality and the statistics were analysed in this locality.



There are a range of factors that influence crime trends that have little to do with police action, such as income inequality, the proportion of young people in the population, good parental supervision, and levels of school attendance.



An increase in some types of recorded crime does not always mean the police are performing poorly — the opposite may be true. For example, increases of recorded domestic assault or sexual assault may be a positive indicator — police may be providing victim support, applying the new Domestic Violence Act, or may have earned the trust and confidence of victims in their area. Crimes such as drunk driving and drug related offences rely almost entirely on the police for their detection and recording. High levels of recorded crime would consequently be a positive performance indicator for the police.

Crime statistics should not be used in isolation to measure police effectiveness — they should form part of a package of measurement tools that are based on the objectives and activities of the police in a particular area. It is also important to analyse not only the crime statistics, but also the number of arrests and convictions.

Are the SAPS statistics accurate?

The accuracy of statistics should be assessed in terms of validity (i.e. do they tell us about the true crime picture?) and reliability (i.e. do they provide information in a consistent manner?). Although a full assessment of SAPS statistics has not been undertaken, some points can be made based on available information.

Validity

One way of assessing whether police data is valid is to compare it with victim survey results. This enables a check on whether crime levels are increasing or decreasing as per the official data, and the extent to which police figures reflect the range of crimes being committed.

Only one national survey — the national Victims of Crime survey of 1998 — has been conducted in South Africa. More surveys of a similar nature are needed to determine whether crime levels are changing as indicated by the police statistics.

The national survey did provide information on the extent to which victims report crime to the police. The findings showed that broadly, reporting patterns were similar to those in other countries. Most victims reported crimes such as car hijacking, vehicle theft, burglary and murder. Less serious property crimes such as attempted burglary, mugging, and theft of a bicycle, and personal violent crimes such as assault and sexual offences were reported by fewer than half the victims.

Reliability

Once the public have reported a crime, how the police handle the information can affect the reliability thereof. This was identified as a problem by the SAPS in 1997 when a ministerial committee of inquiry into the collection, processing and publication of crime statistics was launched (see Crime Index Volume 2 Number 3 1998).

The essence of the committee’s findings was that the absence of an ‘information culture’ in the police underlies the problems affecting crime statistics, particularly how information is processed at station level. Sources of human error also included insufficient training, inadequate resources and computer support at station level and a high turnover of skilled personnel.

The reliability of the data was also affected by several technical problems: the complexity of the system, the enormity of the crime database, network difficulties, computer downtime, faulty programmes and inadequate data validation. Some of the recommendations of the committee have been implemented by the SAPS.

Importantly, the committee found that the routine and widespread public scepticism about police data in South Africa was unfounded. Far from recommending a moratorium on the release of statistics, the committee suggested that more regular releases of crime information to the public were required.

The police statistics are probably reliable at national, provincial and area level (because the overall numbers are large enough to be unaffected by a few mistakes). The committee of inquiry found that the major problems relate to:

The recording of exactly where crimes happened.



The classification of certain crimes, e.g. aggravated vs. common robbery, serious assault vs. attempted murder.



Updating the SAPS data sheet that provides information about the outcome of a case once it has been to court (i.e. whether a conviction was achieved, whether the case was withdrawn, etc.)

The SAPS Crime Information Analysis Centre (CIAC) has in fact done particularly well to produce comprehensive, consistent and geographically detailed information about all 28 serious crime types. These data are available on the SAPS web site: www.saps.org.za. Compared to other developing countries, members of the public in South Africa have much for which to be grateful regarding our official crime database.

Antoinette Louw
Institute for Security Studies

This article is sponsored by the European Union