Measuring Respondent Burden to Statistical Surveys JohnMark Frost

Measuring Respondent Burden to Statistical Surveys John-Mark Frost, Sarah Green, Jacqui Jones & Denise Williams UK Office for National Statistics

Aim of Presentation • To provide an overview of the: - Drivers for the UK pilot of using the Standard Cost Model to measure respondent burden from statistical surveys - Issues highlighted by this work

Overview of Presentation • Measuring respondent burden from statistical surveys - Drivers - Measurement • Standard Cost Model for measuring respondent burden from statistical surveys - Drivers Measurement Pilot Issues • The way forward

Drivers for Measuring Respondent Burden from Statistical Surveys PMIs Prime Ministers’ Instructions on the Control of Statistical Surveys Code of Practice UK Code of Practice for Official Statistics Report annually the estimated costs (for example, on businesses, service providers, or the public) of responding to statistical surveys …

Measurement of Respondent Burden from Statistical Surveys Measurement PMIs Code of Practice Simple method: - used during survey control reviews - 20% of survey sample - postal self-completion questionnaire Calculated by: (No. of questionnaires (100% response)) * (estimated completion time) * (appropriate hourly rate) + Up-rating for respondent re-contact Recognised weakness in method: “questionnaire did not adequately capture all activities involved in complying with the regulation”

SCM to Measure Total Respondent Burden in Europe Measurement PMIs Code of Practice Statistics Netherlands Standard Cost Model - European approach to measure administrative burdens for all government information requests - Including survey participation Standard Cost Model approach: SCM - Highlights impact of international legislation - Transparent measurement - Measures administrative burden - Data collected in face-to-face interviews

SCM to Measure Total Respondent Burden in the UK Measurement PM’s Code Administrative Burden Reduction Project of - Total UK administrative burden - Includes survey participation - Burden associated with statistical surveys forms only a small part of the total - Use SCM Practice SCM ABRP

Drivers for Measuring Respondent Burden from Statistical Surveys Measurement PM’s Need to change method Code of ! Practice Need consistency = Use SCM ABRP

Introduction of the SCM to Measure Respondent Burden from Statistical Surveys • ABRP used face-to-face interviews with very small sample sizes - Not feasible or robust enough for measuring compliance on a regular basis • A paper-based version of the SCM developed to measure costs of complying with statistical surveys - Implement quantitative rather than qualitative approach • The paper-based version was piloted in ONS

SCM to Measure Respondent Burden from Statistical Surveys • Questionnaire developed - 10 questions Breakdown of responding activity Information available from ‘business-as-usual’ External costs Survey irritants - For example: unclear definitions or available information did not match information requested • Standard Cost Model formula: mean weighted cost per questionnaire + re-contact uplift * number of questionnaires in survey sample * survey frequency

Pilot for Measuring Respondent Burden from Statistical Surveys Using the SCM • Pilot highlighted large differences in estimates when compared to: - Face-to-face interview methodology - Previous ONS methodology • Differences to previous ONS methodology are driven by: - Different information in models Change in pay rates Different uplifts for overheads Adjustment to internal costs for business-as-usual

Issues with the SCM for Measuring Respondent Burden from Statistical Surveys • Figures surrounded by confidence intervals of the same magnitude as previous methodology • Method is hugely resource intensive - Respondents - Data producers • Difficulties in estimating accurately - Breakdown of activities - Associating costs with respondents

Questions Raised from Pilot Results • Statistical surveys form only small fraction of total administrative burden - Does measurement need to be consistent? • Are such resource costs beneficial? - The figures are purely estimates • Would a simpler method be: - Fit-for-purpose? - As robust?

Way Forward for Measuring Respondent Burden from Statistical Surveys • Recommendation to use a simpler model: - More proportionate - Is concerned with consistently measuring change over time rather than accurate levels • ONS has worked with producers of UK official statistics to develop such a model - Currently proposals are being finalised to take to the National Statistician for approval

Thank you John-Mark. Frost@ons. gov. uk