Research highlights

• Created and applied new statistical methodologies for complex social science contexts. 
• Opened up access to non-specialists in policy making organisations with user-friendly statistical software packages and training programmes.
• Worked directly with organisations to guide policy and practice decisions, including for a new police funding strategy in the UK.

Breakthrough for better decision making 

The successful development of economic, education and social policy often rests on analysis of complex multilevel datasets that are difficult to work with, especially for non-specialists.   

The University of Bristol’s Centre for Multilevel Modelling (CMM) has addressed this challenge by incorporating its cutting-edge statistical research into its widely used multilevel modelling software package, MLwiN, as well as face-to-face training programmes and training materials.  

This work has significantly increased the number of non-academics and non-specialists equipped to apply these techniques – transforming the application of new statistical practices within policy making organisations and empowering them to improve the accuracy of evidence-based decision making. 

Simplifying access  

The data underlying decision making in social science contexts often has a complex hierarchical, nested or clustered structure. For example, complex interacting dependencies between national, local authority, school and teacher effects might impact educational outcomes in ways that policy development teams need to account for. 

For many years, CMM has developed and applied new statistical methodologies to fit such complex models – for example, the centre has worked on methodological research on the problems of missing data and measurement errors in birth cohort studies and longitudinal surveys.  

Crucially, it has also translated its statistically novel and computationally intensive methodology developments into several user-friendly statistical software packages to give applied researchers direct access to these new methods.  

CMM supports the software packages with extensive user documentation, plus training that it provides via face-to-face user workshops and online training materials.

For example, CMM provides the Learning Environment for Multilevel Modelling (LEMMA) online multilevel modelling course. LEMMA has over 37,000 users and brings the methodology and software to a very wide audience. 

As a result of these efforts, the flagship MLwiN package now has over 15,000 users within over 190 organisations, including the UK Home Office and the US FDA Office of Acquisitions and Grants Service.  

Impact on policy development 

With adoption now so widespread, CMM’s statistical research and software have been cited in over 25 reports by organisations including the World Health Organisation, Public Health England and several UK and international government departments – covering areas such as children’s services, community services, welfare services, public health, energy, climate change, agriculture and defence.  

The Centre also works directly with organisations to guide policy and practice decisions.  The team collaborated with the Higher Education Funding Council for England (now Office for Students), for example, to answer several questions relating to fair admission to university for different ethnic groups.  

Notably, CMM also worked with the UK government Home Office to analyse the geographical predictors of crime and incidents. The commission included training government statisticians to ensure they could replicate analysis with each new year of data - and also apply multilevel modelling techniques to future policing policy.  

Ultimately, this led to a substantial report that informed UK police funding strategy based on new, previously unavailable insights.