Using Administrative Data for Statistical Purposes Stephen Penneck

Using Administrative Data for Statistical Purposes Stephen Penneck Office for National Statistics United Kingdom

The Three Primary Sources • Administrative • Censuses • Surveys

Aspects of Quality Dimensions of quality Administrative data Sample surveys Relevance Definitions and coverage will be relevant to the admin system. Good source for detailed analysis. Surveys can be designed to be relevant to the analytical need. Quality is constrained by sample sizes Accuracy Subject to non-sampling error. Not under the control of statisticians Subject to sampling as well as nonsampling error. Under statistical control Timeliness Some sources (eg tax data) less timely than surveys Many administrative sources very quick. Surveys subject to response times. Accessibility Depends on legal structure. May also be technical and institutional barriers Under direct control of the statistical agency Comparability Dependent on changing administrative definitions over time Under direct control of the statistical agency Coherence Often enables data linking if common identifiers exist Depends on common registers

Costs • Administrative data can replace survey data and will reduce respondent load • But need to include quality costs

Confidentiality • NSIs have strong record of confidentiality • New analysis possible through data linkage • Public confusion between administration / policy use and statistical use

The Three Challenges • Quality – how good are administrative data? • Costs – do they offer cost savings? • Confidentiality – how to overcome public perceptions?