Establishing a Collaborative Data Culture
Collaboration is essential to achieving ecda’s ambition to share data and combine it with the experience of people and communities across Essex, to create powerful insight and inform policy decisions.
The University of Essex strengthens ecda’s work by providing academic rigour and challenge, exchanging skills and creating unique knowledge that is shared with communities and partners across Essex.
Philipp Broniecki, Senior Research Officer at the ESRC Business and Local Government Data Research Centre, University of Essex, shares his contribution to ecda's homelessness project.
In March 2020 ecda initiated a project in partnership with Colchester Borough Council that will help us to prevent people in Essex from becoming homeless. Understanding what makes one person higher risk of becoming homeless than another will enable Councils to identify when is the most effective time intervene and with what level of support, which will hopefully prevent people from becoming homeless.
The University’s early contribution to the homelessness project helped to shape the research questions that the project seeks to answer by looking at what is feasible with the appropriate data held by Essex County Council and Colchester Borough Council. Senior Researchers from the Business and Local Government Data Research Centre at the University of Essex advise on research design, such as the trade-off between information contained in the data (the level of observation) and the amount of available data (the number of rows in a data table). The role of the University in this initial stage of the homelessness work is to help with the methodological approach of the project by advising on the differences between forecasting outcomes and identifying causes of outcomes. Furthermore, the university looks to contribute to the project by advising on the modelling approaches to forecast the risk of homelessness and to help with coding models in statistical software.
Discussions of data feasibility consider which data is better suited for predicting the risk of homelessness than to finding the causes of homelessness. Together the ecda project team, made up of analysts and researchers from Essex County Council, the University of Essex and Colchester Borough Council, opted for an approach that ensures the data is observational and cannot identify individuals across various data sets. Furthermore, the team agreed that the level of information in the data should be kept as high as possible. More informative data is more disaggregated data and could mean fewer observations and potentially fewer correlates of homelessness are included in the analysis.
External research suggests that there is a wealth of data available from across multiple organisations in Essex that could relate to homelessness, such as council tax arrears, benefit recipients and social care support. Machine learning – learning from data – is well suited for the purpose of forecasting across the data available from multiple organisations. Once anonymised and extracted the data is placed into the careful hands of analysts across the team who use algorithms to determine the risk of homelessness as accurately as possible.
As the homelessness project progresses throughout 2020 the University hopes to continue its knowledge exchange with colleagues from Essex County Council and Essex Police and provide further insights on predictive algorithms. There are numerous algorithms that could be useful for forecasting the risk of homelessness. Instead of using just one algorithm, ecda could for instance, keep learning from the data by using model averaging approaches – the premise of model averaging is that all models are useful, and the right mix is often superior to any one individual algorithm. With expertise in writing code in programming languages such as R, the University hopes to support the project through further application of algorithms.
page updated 22/07/20