Transforming Social Survey data Collection in the UK opportunities and challenges. Ian O Sullivan and Ed Dunn Social Survey Statistical Redesign ONS

Transcription

1 Transforming Social Survey data Collection in the UK opportunities and challenges Ian O Sullivan and Ed Dunn Social Survey Statistical Redesign ONS

2 The world is changing Digital revolution Across public and private sector Online shopping, online banking, online everything... Driving licences, car tax, public services in the round Data revolution Loyalty cards, road traffic Data is the new oil ONS leadership Better Statistics, Better Decisions

3 It will drive a revolution for ONS Data is our core business Already on the journey Electronic Data Collection, admin data It will change the nature of our surveys Why collect what already exists? We will still need surveys but they will look very different in 5 years

4 Data Collection Transformation To rebalance ONS data collection activity significantly toward wider, more integrated use of administrative data sources, thereby reducing our reliance on large population and business surveys.further, the remaining survey operations will become more efficient through a move from paper and personal interview-based collection to online data collection.

5 Social Survey Transformation Data integration Administrative data use alongside surveys Online Surveys Mixed mode Survey rationalisation & redesign Field force modernisation

6 As Is - Current Social Survey Model SURVEYS Frame (PAF) SURVEY 1 CAPI CATI SURVEY 2 CAPI PRIMARY OUTPUTS (survey based) PRIMARY OUTPUTS Secondary outputs (model based?) Secondary Analysis in SSD SURVEY 3 CAPI CATI PRIMARY OUTPUTS Secondary Analysis in ONS Data delivery to customers Secondary Analysis

7 Portfolio of continuous ONS social surveys and several GSS surveys; smaller ad hoc surveys Multiple topics, multiple designs, multiple approaches; Blaise based Harmonised but not integrated; individual designs Silos reflective of individual analytical requirements and reflective of history and fear of discontinuity 7

8 What is the desired future state? Multi source, multi mode data collection Surveys, admin, big data Surveys digital by default where sensible Balance moving increasingly from surveys to nonsurvey Supported by common services / IT systems For ONS and external use Saving money

9 Non-survey data Opportunities Improve survey coverage Replace survey questions Targeted sampling Improved sample frames Improved quality New outputs More timely data Reduced burden on respondents

10 Non-survey data Challenges Concept alignment Definitions of admin data and survey question Quality e.g. timeliness, coverage, coherence Quality Assurance of Admin Data (QAAD) Process maps to understand data journey Linking rates Security of ongoing supply

11 Move On-Line - Opportunities Transform NOT Translate Take LFS back to a Labour Market Survey Respondent Centric Design Transform engagement Quality Savings

12 Move On-Line - Challenges Self-completion with voluntary surveys Designing to keep pace Engaging with respondents Stakeholders breaks in series Design constraints Are there really costs savings?

13 The real opportunity/challenge!

14 Option 1 LFS design with PCS boost Master Wave (200k HH sample) Equal probability sampling for LFS waves and other ONS surveys PCS boost (150k) Boost provides weighting to stratify the sample based on Hard to Count (HtC) LFS Wave 2 LFS Wave 3 APS OPN Expenditure LFS Wave 4 LFS Wave 5 Core master wave randomly sampled - provides a frame for other ONS surveys Boost needs to be large enough to meet all PCS requirements (HtC weighting, dependant sampling etc)

15 Option 2 Master Wave designed for PCS Stratified PCS Sample (350,000 HH) Address Register Population Spine Admin data is used to stratify the PCS sample Post stratification used to re-weight for other ONS surveys LFS Wave 1 LFS Wave 2 LFS Wave 3 LFS Wave 4 LFS Wave 5 APS EXP OPN Use administrative records to stratify the PCS sample Admin data needs to be of high quality to re-weight PCS sample for LFS and other surveys

16 Option 3 Master Wave based on ADC Characteristics Survey Characteristics Survey Sample (700,000 HH) Stratified PCS sample within Master Wave Equal probability sampling for other ONS Surveys Address Register Population Spine LFS Wave 1 LFS Wave 2 LFS Wave 3 LFS Wave 4 APS EXP OPN LFS Wave 5 Master wave based on larger random sample of households Use admin data to check that PCS requirements met by master wave