Code to analyse the efficacy of the different SERL recruitment methods in enhancing response rates. These were tested in a 'pilot study' that took the form of a randomised control trial of a range of methods. The results were used to refine the final SERL recruitment approach.
A paper analysing the effects of different recruitment methods on recruitment outcomes using the SERL pilot study data - https://beta.ukdataservice.ac.uk/datacatalogue/studies/study?id=8666
- Pre-print: https://osf.io/f82b7/
- Published in Energy Research & Social Science: https://doi.org/10.1016/j.erss.2021.102312 (open access)
"Obtaining high-resolution energy consumption data from a large, representative sample of homes is critical for research, but low response rates, sample bias and high recruitment costs form substantial barriers. The wide-spread installation of smart meters offers a novel route to access such data, but in countries like Great Britain (GB) consent is required from each household; a real barrier to large-scale sampling. In this paper we show how certain study design choices can impact the response rate and sample bias in energy studies requesting access to half-hourly smart meter data and (optional) survey completion. We used a randomised control trial (RCT) with a 3 x 2 x 2 factorial design to test incentives, message content/structure and a 'push-to-web' approach in a large-scale pilot for the future 10,000+ GB sample. Up to 4 mailings were sent to 18,000 addresses, recruiting 1711 participants (9.5% response rate) from England and Wales. Our results and recommendations can be used to help future energy studies to achieve greater response rates and improved representation. UK-based researchers can apply to use our longitudinal smart meter and contextual datasets."
pilot_analysis_for_paper_v06.Rmd in the R folder contains the final version of the code used to create plots and statistical analyses used in the paper.
This code loads a data table 'results' consisting of 18,000 rows (one per address contacted in the pilot study) and 11 variables:
- consent_source ("online", "postal", "online+postal", "NA"); how a participant signed up
- cell (1:12); the test cell assigned to the address
- quintile (1:5) the Index of Multiple Deprivation (IMD) quintile
- region (7 of the regions in England and Wales)
- survey_status ("Completed", "Partially completed", "NA")
- survey_response_type ("online", "postal", "online + postal", "NA"); how survey response was provided
- consent_start_date(e.g. "2019-09-01"); when a participant signed up
- consent (TRUE or FALSE); whether the household gave consent to take part
- p2w (TRUE or FALSE); whether the address received the push-to-web treatment
- version (1 or 2); content version received by the household
- incentive ("None", "Thermometer", "Voucher")
Unfortunately the data cannot be made publicly available because SERL data is held in a secure data lab environment with quality-assured output checking processes. Any researcher interested in reproducing our results can apply for access to the data via the UK Data Service.