Devolution risks simply changing where UK policy learning happens rather than how. In his third blog in the series, Geoff White warns that regional centralisation could replace national control, creating new spatial silos. To prevent this, he argues we need to rewire policy learning into a networked system: building regional capacity for collaborative and comparative learning that brings together local partners, evidence and experience across time, places and policy functions.

In previous blogs I’ve argued that decades of centralisation have left the UK with a weak policy learning culture in which valuable lessons have been lost in central archives and from institutional memory. In this blog I suggest that devolution could either risk reproducing that learning weakness at a different spatial level or provide the opportunity to rewire and strengthen the learning model.
Devolution may shift the geography of decision-making, but not necessarily the degree to which policy learning remains or, at least, is seen to be over centralised. Survey evidence from Scotland and Wales, for example, found that devolved governments were perceived by some citizens, particularly those in peripheral areas, as a form of regional centralism, displacing rather than resolving issues of centralised control.
It does not follow, therefore, that devolution in England will automatically strengthen the decentralised capacity to experiment, adapt and learn. Moreover, when resources are limited, devolved authorities will naturally focus on their own patch and struggle to integrate across very different policy domains. This could result in devolution creating spatial as well as functional silos, further restricting the circulation of knowledge and the capacity to learn.
However, devolution also creates the potential for different and more effective modes of decentralised policy learning if accompanied by processes that connect learning across time, place and function. This opportunity is reinforced by recent revisions to the Green and the Magenta Books which, respectively, emphasise the importance of place-based strategies and ‘test and learn’ approaches to evaluation.
In their framework on policy learning, political scientists Dunlop and Radaelli show that how governments learn is shaped by the degree to which the problems they face are tractable (such as how complex they are) and the strength of organised and recognised groups represented at the policymaking table. The modes include learning through established hierarchies, exchanges of expertise and evidence, negotiation between interested parties with different objectives and priorities, and engagement of wider stakeholders and citizens. I argued in my first blog that policy learning in the UK has for many decades been dominated by the central hierarchical mode.
Devolution creates the opportunity to broaden the modes of learning and to bring about more effective policy learning as follows:
- Hierarchical learning: Devolved authorities are expected to evaluate the effectiveness and social value of what they do. Mayoral Strategic Authorities (MSAs) have to work within an agreed national outcomes framework and collaborate with national evaluators where required. Rather than seeing this as a hierarchical imposition, MSAs should make it into an interactive process to demonstrate how to use and benefit from Patton’s developmental and real-time evaluation approach. This uses systems thinking and complexity concepts in a continuous and adaptive process of learning about ‘whether we’re doing the right things’ as well as ‘whether we’re doing things right’.
- Evidence-based learning: Devolved authorities are closer to the context in which policy operates. They can generate and test evidence about what is appropriate and what works in their specific places rather than relying on evidence generated elsewhere – allowing them to policy learn rather than policy borrow. This can be done systematically by identifying and promoting regional and local areas of research interest, signalling knowledge gaps to higher education institutions, and mobilising them to co-design and evaluate policies built for their local and regional conditions.
- Learning through bargaining: Devolution creates more scope for direct negotiation between local partners rather than being mediated through hierarchical relationships with central government departments and agencies. Their interaction can change not only the policy bargain but also their understanding of the problem itself and the range of policy solutions.
- Reflexive learning: Bringing different stakeholders together around a particular place and its problems (through, for example, citizen assemblies and action learning sets) may challenge the assumptions embedded in centrally designed policy, including the boundaries between employment, skills, health, transport, investment and regeneration. It can expose different perspectives on the problems themselves and the relationships between them.
The different modes of learning need to be applied across and between devolved authorities – through comparative learning. The evidence and experience in one place can inform another if differences in context are properly understood. Mayoral and other devolved authorities need to be able to compare their experiences systematically, identify patterns and understand why similar interventions in one place do not always lead to the same outcomes in another. It’s not just a matter of copying best practice but of learning what happened, why and what might be transferable and what is context specific.
Action will be needed to avoid the potential loss of scale economies from devolution. This could be facilitated through the Evaluation Accelerator Fund which explicitly targets building evaluation capabilities across national civil service professions and local government. Devolved authorities and central government should use the Fund and other means to pursue the following:
- Central hosting of data for counterfactual analysis (e.g. setting up synthetic controls for different intervention categories) and synthesising evidence from the past and across geographies through the Office for National Statistics and/or the What Works Centre for Local Economic Growth;
- Permanent or temporary secondments of skilled staff from the Government Economics Service and Government Statistical Service;
- Twinning settlements or other horizontal arrangements between devolved authorities at different points in their devolution journey to share and develop good practice methods and evaluation evidence.
Rewiring the state of policy learning is not simply a question of transferring existing learning modes to a regional level. It requires a fundamental shift in the way governments generate, exchange and apply knowledge about the appropriateness and effectiveness of policy design and delivery.
The table below suggests what that shift might look like, moving from the current dominant hierarchical model to the networked approach that devolution makes possible.
| Aspect of learning | Escaping the shadow of hierarchy | To a devolved network of learning |
| What it is for | Establishing best practice about what works in getting things done effectively and efficiently | Continuously assessing whether the ‘right’ things are being done and adapting the policy and strategic case |
| Where it happens | Learning within departments and agencies | Learning across time, space and functions (comparative learning) |
| Its focus | Programmes and projects | Policy portfolios, packages and experiments |
| What it relies on | Primarily formal evidence | Experience and expertise as well as formal evidence |
| How it is transferred | Disseminating knowledge and intelligence mostly vertically | Circulating knowledge and intelligence through meshes and networks |
| What happens to it | Storing the knowledge | Building and synthesising a living body of comparative knowledge |
Devolution offers the opportunity to break free from the shadow of hierarchy. But shifting powers from the centre could simply change where learning happens rather than how it happens. Without deliberate reform to our state of learning, we risk fracturing a centralised hierarchy into isolated, sub-national silos: repeating past mistakes and failing to ask whether we have specified problems in the right way and are doing the right things to address them. Wider modes of learning need to be woven into the fabric of devolved government from the outset.
The challenge is to make devolution more than a redistribution of power, using it instead to build a truly networked, adaptive and collaborative learning state.
The views and opinions expressed in this post are those of the author(s) and not necessarily those of the Bennett School of Public Policy.