Natural England - the government’s advisor on the natural environment, sought to understand whether Small Area Estimation (SAE) techniques could be utilised on data collated through their annual Monitor of Engagement with the Natural Environment (MENE) quantitative survey, to enable improved accuracy and precision of survey-based estimates of the extent that local populations engage with the natural environment.
City Science Response
Harnessing the extensive data science capabilities of our in-house research team, City Science assessed the suitability of the MENE dataset for Small Area Estimation by analysing over 420,000 accumulated responses from 9 years of survey data.
Research was undertaken into the potential of auxiliary information, such as Census data, to power the model based on our team’s experience of handling a variety of datasets and rigorous analytical approach. Generalized linear mixed-effects regression modelling was performed, with a particular focus on the potential bias of using survey and auxiliary data. Modelling was delivered through the R STAN modelling software.
In order to fit within annual budgetary constraints, City Science rapidly mobilised our delivery team and delivered the project within a tight timescale. A technical report collating output from our commission was provided to the Natural England along with a client walkthrough of our findings. The generated R code was provided to Natural England, enabling the client to replicate and expand on our work where desired in the future.
City Science’s Research team concluded that implementing Small Area Estimation was eminently suitable for the MENE dataset. Our findings identified the possibility to reduce the MENE survey size by up to 20% without impacting the quality of the survey’s findings at Local Authority level. We also illustrated the potential to use census data as auxiliary information to enhance Small Area Estimation down to Middle Layer Census Super Output Area (MSOA) levels.