16. Census Data Quality

During the creation of this document the following methods have been used to ensure and assess the quality of the data presented.

1. Peer review of the collection, extraction and analysis processes.
2. Repeat entry for samples of manually entered data.
3. Analysis to check internal consistency of data.
4. “Sanity checks” against other work and previous census documents
5. Sanity checks by department specialists against existing knowledge and reports.

The quality of the data presented here falls into several main categories.

Excellent prison and inmate management data

The key census data is passed to the Department from Courts and is the basis on which the prison muster is created. The same definitive source of data is used in the census that is used for managing inmates. This high quality data includes the following: inmate identities, the facilities in which they reside, the offences that have resulted in their current sentences, and the sentence data itself.

Good imported historical data

Of almost the same high quality is the historic criminal record data. While this data is sourced from the IOMS database, its transmission entailed various import processes, and the distilling of information from older records. This data potentially contains some errors; it is not used for formal inmate management purposes, but is retained for other purposes. This generally high-quality data includes historic sentencing detail from which information such as sentence counts and earliest custodial sentence age are derived.

Good routine demographic data

Key demographic data such as gender, age, ethnicity, etc are collected routinely, some of which is information volunteered by offenders themselves. The IOMS system does not enforce the collection of this data but where it is recorded (almost 100%) it is regarded as sound. (Although it is always possible for inmates to be deliberately misleading as evidenced by the number of aliases recorded in the database). Other data, such as risk scores, while generally accepted as accurate, are susceptible to system instability, and individual errors have been observed.

Representative data

Some data collected by the Department is peripheral to the inmate’s management, and as such staff may omit to cover it in reception interviews, and inmates may not always report it accurately; follow-up checking is not done nor indeed is it always realistic. An inmate who states that his income prior to imprisonment was derived from wages is taken at his word. Given that inconsistencies have been uncovered in data of this type, it should be regarded as indicative only, representing the approximate proportions of inmates in various sub-groups, rather than exact numbers.

The following data should be regarded as representative rather than precise:

Marriage status; Child dependents and child care; Gang membership; Income sources; Educational qualifications.