Is administrative data collection identical to a survey?

No. Both can produce statistics, but this collection relies on administrative records rather than primarily collecting answers through a statistical questionnaire.

Can the collection use more than one administrative source?

Yes. Combining sources may improve coverage or detail, but introduces harmonization and linkage questions that need evaluation.

Are published figures necessarily final when the first extract arrives?

No. Late records and source corrections may require revisions under a documented policy.

AI Glossary / Statistics and probability / Intermediate

Administrative data collection

Definition

Administrative data collection is the organized acquisition and processing of data from administrative sources for statistical purposes. In the source glossary's sense, it encompasses collection, processing, storage and dissemination. It uses records generated by operational or regulatory activities rather than relying solely on direct statistical questioning of respondents.

What it means

Administrative data are the records themselves; an administrative data collection is the statistical activity that turns relevant records into a defined information resource. Its boundaries include which sources are obtained, which units and periods are covered, and what statistical outputs are produced.

The original administrative system may exist to register businesses, administer benefits or deliver a service. The statistical collection reuses information from that system for a specified analytical purpose. Reuse can reduce respondent burden, but it requires agreement about definitions and careful evaluation of whether operational records support the intended statistics.

How it works

Define the statistical purpose, target population, required variables and reporting period. Identify administrative sources and assess their coverage, operational definitions, update practices and suitability. Establish an appropriate basis for access and arrangements for transfer, security, retention and communication of source changes.

Acquire a documented extract or feed, validate its structure and reconcile received totals with the provider. Transform operational records into statistical units, address duplicates and revisions, and evaluate missing or inconsistent fields. If sources are linked, assess linkage quality and preserve information needed to understand unmatched records.

Store the resulting data with metadata and processing history, then produce outputs using defined methods and disclosure controls. Evaluate quality throughout the process rather than only when publishing. A recurring collection also needs procedures for late records, revised extracts and changes in eligibility or recording systems. These changes can alter statistical results without a corresponding change in the real-world activity being measured.

Three practical examples

1. A statistical office receives business-registration extracts and processes them into a maintained business register with documented activity and coverage rules.

2. An education reporting program obtains enrollment records from institutions, harmonizes definitions and produces statistics for a clearly specified reference period.

3. A public-service analyst combines authorized administrative extracts to describe program participation, checking linkage errors and differences between applications, participants and service events.

Think of it this way

Think of a collection as a carefully managed pipeline from operational records to statistical information. The source supplies raw material, but the pipeline must preserve meaning, remove processing mistakes and explain what the finished output represents. A fast pipeline is not sufficient if it silently changes units or drops part of the intended population.

Worked example

Consider an original monthly collection designed to count active permits at month-end. A provider sends 10,000 rows. Validation identifies 400 exact duplicate export rows, leaving 9,600 distinct permit records. Of these, 600 have expired before the reference date, leaving 9,000 active permits under the stated definition.

Suppose those active permits belong to 8,200 distinct holders. The collection can report 9,000 active permits and 8,200 represented holders, but those are different measures. The ratio is 9,000/8,200, approximately 1.098 permits per represented holder.

If 200 additional active permit records arrive after the publication cutoff, the handling depends on the collection's documented revision policy. A revised permit count would be 9,200 if all 200 are valid, distinct and in scope. The holder count cannot simply be increased by 200, because the late permits may belong to already represented holders. This example shows why validation, timing and statistical-unit definitions are part of collection design.

Illustrative case study

Fictional case study: Registerview produces quarterly statistics from a licensing system. Its first extract contains multiple versions of each application and mixes pending applications with active licenses. The statistical team works with the source owner to document status histories and the meaning of each timestamp.

The team defines a reference-date rule, keeps the appropriate record versions and reconciles output counts before publication. It also establishes a process for receiving late corrections and explaining revisions. A metadata note describes who is covered by the licensing requirement and who falls outside it.

This fictional case illustrates a statistical collection built around administrative processes. It does not establish that the resulting statistics cover every business or activity in the wider population.

Common misconceptions and mistakes

Obtaining an administrative file is not the same as completing an administrative data collection. Statistical processing, quality assessment and documentation remain necessary. Likewise, repeated receipt of records does not guarantee consistent time-series definitions when the source system changes.

Another mistake is assuming that administrative collection has no respondent burden or error. It may reduce direct statistical requests, but providers still need to prepare and maintain useful feeds. Administrative eligibility rules, incomplete records and processing decisions can create limitations even where sampling is not involved.

Business use

A well-designed administrative collection can supply recurring information more efficiently than repeatedly asking respondents for facts already recorded. It can support planning, monitoring and the preparation of datasets for responsible analytical use.

Assign ownership for source relationships, data definitions, quality checks and publication decisions. Maintain a clear record of transformations so an unexpected result can be traced to the extract, the processing rules or actual activity. For AI applications, preserve the distinction between operational status and the target concept used for training or evaluation. Review whether the collection still supports its purpose when administrative policies or technology change, rather than treating an established feed as permanently fit for use.

Check your understanding

A collection receives 1,000 records but discovers that 100 are duplicate copies. Can it report 1,000 distinct events?

Answer: No. If the remaining records each represent one distinct in-scope event, the count is 900. Further checks are still needed to establish scope and meaning.

Related terms

Administrative data
Record linkage
administrative source
statistical collection
register
data validation
reference period and revision policy

Learn next

map one collection from source agreement to publication, identifying its statistical units, scope rules, validation checks and revision points. Test how a changed source field or delayed extract would affect the published measures.

Sources

Primary sources: OECD, OECD Glossary of Statistical Terms, Administrative data collection: https://www.oecd.org/content/dam/oecd/en/publications/reports/2008/09/oecd-glossary-of-statistical-terms_g1gh9ad7/9789264055087-en.pdf

Australian statistical data integration guidance, Data management: https://statisticaldataintegration.abs.gov.au/topics/data-management

Australian statistical data integration guidance, Quality assurance: https://statisticaldataintegration.abs.gov.au/topics/data-management/data-quality/quality-assurance

Previous: Administrative data Next: Admissible decision rule