Become a columnist. Or write an article. Or head back to the homepage.

Columns

Data Quality and the Cops

The Irish Government is currently spending thousands of euro of taxpayer's money to establish why the Gardai (Police) in Waterford were able to record higher crime detection rates than any other garda division in the Irish Republic.

The implication that has been put forward in some parts of the media is that the local gardai and their superiors are to blame for this because they encouraged "inflation" of these figures to help boost their career and promotion prospects. The Government has initiated an enquiry to root out how and by whom these figures were recorded.

Remember - the Government is spending thousands of euro on this enquiry, on top of the cost of actually gathering the data in the first place. This is a total waste of public funds at a time when the government can little afford excessive expenditure (or so they tell us). Furthermore, coverage of this issue has undermined faith in the overall crime statistics as reported by the Gardai and Department of Justice.

I submit that I can identify the root causes of this problem in the next 560 words or so and that this article should be downloaded by a civil servant in the Department of Justice and presented to the minister forthwith as the detailed report.

My understanding of the problems in Waterford is that the "crimes detected" figure was significantly higher than in other garda districts. This was due to the gardai in the Waterford district defining the concept "crimes detected" differently to other districts (based on a synopsis of the issue in the Irish Times of Wednesday 07/08/2002).

Fundamentally the root cause (as I see it based on reading newspaper articles on this issue) is that while the Department of Justice and the Garda Siochana decided to implement this reporting, there does not appear to have been any clear definition of the key concepts that were to be reported on - for example when should a crime be classified as "detected".

When is a crime detected? Is it when the authorities have satisfied themselves that they have identifed who committed the crime irrespective of arrest, prosecution and/or criminal sanctions (imprisonment or fines etc)? Or is it when there is actually an arrest and prosecution? Once that is defined, it is easy to report on the data and also to identify errors. In short, it is easier to ensure that the processes by which the data is gathered are statistically stable and will perform within the expected ranges. In the absence of a clear definition of the valid values of data to be used (e.g. "A crime is detected when the judicary impose a criminal sanction") it is impossible to say that the process by which the data is gathered is statistically stable and as such all data should be regarded as what W.Edwards Deming calls a "special cause". In Demings view of Total Quality Management, the entire incident in Waterford is an example of a special cause. One of the key drives of TQM is to eliminate special causes of error and bring the process by which the data is created into a state of statistical stability.

Without any clear definition of the "acceptable" domain values for a piece of data, you are left with data producers who will, as is human nature, get creative and improvise. If you are lucky the people may apply the "reasonable man" test (as defined by Alfred Lord Denning). Usually they will just improvise in a manner that will make them look good.

The key steps to be taken to improve the situation within the crime statistics reporting are:
1) The department of Justice and the senior officers in the Garda Siochana need to define clearly what a "Detected crime" is. All data creators from civilian support staff through to the commissioner need to be trained in these concepts.
2) No-one in any garda division should be held responsible for errors, omissions or improvisations they may have made in the absence of such a definition. At most they should be given a minor reprimand for not consulting other divisions to see what the established practice was.
3) Quality control reporting should be undertaken at random intervals in randomly selected divisions on a sample of randomly selected records. This will indicate early on the statistical stability of the information gathering processes.
4) The plug should be pulled immediately on the Waterford enquiry as, on the basis of the facts available to me, it will not determine anything over and above what I have put forward in this article and the funds being spent on it could be better spent on buying equipment for a primary school somewhere.
5) The government should make total data quality management part of its drive to cut costs. Industry estimates agree that data quality accounts for at least 10% of the bottom line costs in any organisation. How much could be done if 10% could be shaved off the cost-base of every government body?

by

Daragh O'Brien
18th September 2002

Discuss this Article

Topics