In the first data pilot for Arloesiadur, our project to develop an innovation data analytics platform for Welsh government, we have started to map research networks in Wales using data obtained from the Gateway to Research data, a repository of information about Research Council and Innovate UK funded projects.
In the first data pilot for Arloesiadur, our project to develop an innovation data analytics platform for Welsh government, we have started to map research networks in Wales using data from the Gateway to Research, a repository of information about Research Council and Innovate UK funded projects.
The potential policy relevance of this dataset is clear. Universities are a critical component of innovation systems that also include businesses and government – the famous “triple helix” idea. In addition to creating new knowledge and highly educated graduates, they also connect government and industry. With a better understanding of university networks, policymakers can put in place effective initiatives to support them.
Some burning questions include:
We begin exploring some of these topics in this blog. Before we get started, it is worth noting that this analysis is preliminary. The data pilots we are running until Autumn have the primary goal of exploring different datasets, the methods to analyse them, and the formats to communicate findings. We will use what we learn when we start building Arloesiadur in the Autumn. If you have any feedback or ideas, leave a comment below or send us an email at [email protected]. The code we have used to collect, analyse and visualise the data is available on GitHub.
Gateway to Research is a web portal and database with comprehensive information about research projects from the UK's seven research councils as well as Innovate UK, including:
The approach we have used to classify research topics into broader research domains (e.g. following the example above, to put "Artificial Intelligence” into Mathematics and Computing, or “Criminal Law and Criminology” into Social Sciences) is available at our GitHub repository.
In the rest of this blog, we present some preliminary findings from an initial exploration of the disciplines that Welsh universities specialise on, levels of collaboration across disciplines compared to the UK, and the geography and structure of research collaboration networks in different disciplines.
The heatmap below shows the research specialisation of Welsh organisations - grouped by local authority district - compared to the rest of Great Britain. We measure this by counting the number of organisations involved in research projects in each research domain and local authority, and comparing it to the UK’s average. Warmer colours indicate higher levels of specialisation. The heatmap excludes local authority districts with less than 20 organisations involved in projects, where the measures of specialisation are less likely to be robust.
The heatmap suggests that Wales has hotspots of research activity in all the “grand challenge areas” identified in its innovation strategy, including:
Several Welsh regions show high levels of activity in other academic disciplines like Social Sciences in Cardiff, or Arts and Humanities in Rhondda Cynon Taf (where the University of South Wales has two campuses).
Collaborating inside research disciplines makes it possible to specialise and accumulate knowledge. At the same time, combining ideas from different disciplines generates innovation. How to strike the right balance between both? This is another area of interest for policymakers that we can analyse with our data.
The heatmap below shows the levels of “discipline mixing’ in Wales compared to Britain. The basic idea is to compare the probability of finding organisations working in projects including research topics in the same research domain (the diagonal of the heatmap) or from different domains (off diagonal) in Wales with what happens elsewhere in Britain. As in the previous heatmap, warmer colours indicate that the levels of collaboration for a combination of disciplines are higher in Wales than in the UK.
The heatmap suggests that, in domain areas such as “Environmental Sciences”, “Physics” and “Mathematics and Computing”, Welsh organisations display a stronger propensity to participate in projects involving collaboration inside the same discipline.
We also find some instances of mixing across domains, in particular between Environmental Sciences and Life Sciences, and Environmental Sciences and Arts and Humanities. The seemingly striking finding of high levels of collaboration between Social Sciences and Physics in Wales seems to be caused by low levels of mixing between these two disciplines in Wales as well as the UK. A single project (Space Science for All) drives the result.
The interactive map above shows our preliminary analysis of the geographic scope of collaboration involving Welsh universities (1). The colour of each local authority district reflects the number of collaborative projects in the GtR data that the selected university has been involved with in that district (2).
The data do not take into account recent mergers in Wales, such as the University of Glamorgan and University of Wales Newport merging to become the University of South Wales. Projects linked to the new organisations’ predecessors are in Gateway to Research, but we have not yet linked this to the new organisations. This will alter the number of collaborations for new universities when it is taken into account of.
Collaborations between other universities and the old organisations are not affected by this. For instance, Cardiff University’s collaborations with the old University of Glamorgan would still be present in Cardiff’s count. In the case of University of Wales, Trinity Saint David we haven’t been able to locate any collaborations outside of the University’s own district (only one collaboration is recorded in the Gateway to Research data for the university).
Our initial exploration shows that Cardiff University has the highest levels of collaboration across the UK of any of the Welsh universities, with 1,050 collaborative projects having taken place across 151 districts. Swansea has the second highest count with 556 projects across 130 districts.
The highest number of collaborations in any one district outside of each university’s own district is 64, the number of collaborations between Cardiff University and organisations based in the City of Westminster (an area with a number of academic institutions including the University of Westminster, King’s College London and Imperial College). Aberystwyth and Swansea universities show similarly high numbers of collaborations with Westminster organisations (21 and 47 collaborative projects respectively).
It’s difficult to draw a distinct difference in geographic pattern between the universities. For obvious reasons, Cardiff and Swansea both collaborate strongly along the M4 corridor. Similarly, Bangor has collaborations stretching across the north of England and Aberystwyth has projects covering much of mid and west Wales.
The structure of Welsh research networks
Finally, we have started looking at the structure of Welsh research collaboration networks. How well connected are these networks? Who are the key players? Answering those questions can help policymakers assess the need for interventions to strengthen networks, and determine where to target them.
In these collaboration networks, the nodes represent organisations in the GTR data that participate in a project where there is at least one Welsh organisation. The links between them represent collaborative research projects. Each project has been allocated to a research domain, which makes it possible for us to look at different research domains separately.
In the three figures below, we focus on three research domains - Environmental Sciences, Mathematics and Computing and Social Sciences. For each of these, we plot one graph including all organisations (whether they are based in Wales or not), and another which removes non-Wales based organisation in order to see what is the level of integration - or fragmentation - in Welsh research networks. We colour Wales-based organisations in red, English organisations blue, and Scottish organisations green (we have not included organisations outside of Great Britain for now).
In the case of Environmental Sciences, there is a larger connected component involving collaborations between universities in Cardiff and public sector organisations including Welsh Government, Welsh Water or Countryside Council for Wales. The Mathematics and Computing network has a cluster of collaboration between Swansea University and several digital businesses. Interestingly, the BBC seems to play a connecting role, participating in two research projects involving Welsh-based organisations in different parts of the network. The Social Sciences network is somewhat less well connected: when we consider the network of Wales-based organisations, only a third of them participate in its largest component (compared with 48% of organisations in the Environmental Sciences network, and 43% in the Mathematics and Computing network).
We also find that when we remove non-Wales based organisations from the network, this leaves a significant number of Wales-based organisations ‘disconnected’. This happens with a third of the organisations in the Environmental Sciences, and Mathematics and Computing networks, and 42% of organisations in the Social Science network. This highlights the integration of Wales-based organisations in broader UK research networks, and raises an interesting question: Could there be scope to integrate them into Welsh research collaboration networks in future projects?
As we said at the beginning, this is an initial exploration of the data, but we feel we are only starting to scratch the surface. Other options we have not yet pursued include looking at individual researcher data, measuring research outputs, mapping international collaboration networks, tracking the evolution of collaboration over time, and benchmarking networks in Wales with other parts of the UK, for which we also have data. We can also perform more finely grained clustering analyses of research topics or abstracts to identify detailed research specialisations that may be of interest for technology foresight.
We are also planning to create maps of research networks allowing users to explore them interactively. We will also use the list of businesses that we have identified in this pilot as an input in other data pilots, including one where we will draw on different measures of innovative behaviour to find innovative businesses and entrepreneurs in Wales.
We will keep you posted on what we find.
The local authority district boundaries are based on TopoJSON data provided by Dr Martin Chorley of Cardiff University.
1. To produce the resulting map we geocoded the organisations within the GtR database, using their postcodes and address strings to get approximate latitude and longitude coordinates. We used these to place each organisation into a specific local authority area, allowing us to count the number of projects each Welsh university participated in, in each district. All of our methods are available on GitHub,) and a fuller methodology is available in the companion blog post
2. These won't all be academic projects, as the database also contains data about companies that have been involved in research and development, collaborative work, Knowledge Transfer Partnerships or Innovate UK funded projects.