Case study - Rent Arrears Prediction

Introduction

S2DS Project Partner Hackney Council provides social housing to over 20,000 tenants. As with other social housing providers, some tenants fall into arrears, and this is likely to become more acute following the rollout of Universal Credit. Of all new arrears cases however, only 16% go on to become the long-term cases that account for the majority of money owed. Currently, Hackney Council has no way of identifying which arrears cases are likely to be long-term, so all cases are treated equally.

Problems to Solve

Using data provided by Hackney Council’s Housing and Benefits services, the goal of the project was to build a predictive model that would allow Hackney Council to better understand the combinations of factors that are most closely associated with a high risk of falling into arrears, and then identify those cases that are most likely to become long term so as to better target interventions.

Results

A team of four data scientists delivered an automated data science pipeline, covering the cleansing of raw data, the creation of new features within it, and the training and assessment of models in an easily amendable and scalable way. Using machine learning, the team was able to accurately predict whether a tenancy is currently in long-term arrears or not, allowing the assignment of an arrears risk score to each tenancy.

Furthermore, the combination of factors that led to either high or low arrears risk scores were identified, allowing for the definition of risk groups, providing insights into how to improve the tenants’ situation. Finally, the model for predicting whether new cases of arrears will be long or short-term can help the council prioritise interventions, improving allocation of council resources.

Conclusion

The council now has a model to drive a data driven approach to understanding rent arrears and associated risks, helping them to better prioritise those that need more support in effectively managing their rental payments. It is hoped that this will facilitate a considerable reduction in the risk of long-term arrears cases, supporting both tenants and the council.

“I’ve been thinking as to how we could apply data science techniques to a number of areas of Housing services for some time and working with Pivigo on their S2DS programme made for a great pilot, giving us outputs to proactively help our tenants and maximise our income. The project has demonstrated that moving towards data driven management approaches doesn’t have to involve huge levels of up-front investment. We have released our work under an open source license and are actively seeking partners in other councils and housing associations to develop this model further. I would encourage other councils to think as to what problems they could solve with this approach and think that through the open development and sharing of our work have a great opportunity to push the sector forward”.

Tom Harrison, Digital Transformation Manager, Hackney Council