S7 Ep28: Why civil service reform fails (and what actually works)

Every civil service reform plan opens with the same list of complaints: poor performance, low motivation, weak accountability. Across six African countries and three decades, governments launched 131 separate reform efforts; not one fully achieved what it set out to do.

Martin Williams spent more than a decade working alongside Ghana's civil service before writing a book called Reform as Process that analyses the lessons from his experience, and the rest of the 131 reforms. For example, 34 programmes across six countries tried to link civil service pay to performance; none delivered. One lesson is that formal rules and accountability systems cannot govern what matters in a civil service: innovation, adaptation, co-ordination, the willingness to act on the spirit of a rule rather than its letter. Meaningful reforms often require no money at all. They require changing expectations from inside, starting small and building credibility, decentralising the leadership of change, and treating new formal rules as a last resort rather than a first step.

The book behind this episode:

Williams, Martin J. 2026. Reform as Process: Implementing Change in Public Bureaucracies. New York: Columbia University Press. Open-access PDF available at uplopen.com.

To cite this episode:

Phillips, Tim, and Martin J. Williams. 2026. "Why civil service reform fails (and what actually works)." VoxDev Talks (podcast).

Assign this as extra listening. The citation above is formatted and ready for a reading list or VLE.

About the guest

Martin J. Williams is Associate Professor of Organizational Studies and Associate Professor (by courtesy) of Political Science and Public Policy at the University of Michigan, and Associate Faculty at the Blavatnik School of Government, University of Oxford. His research spans the politics and management of policy implementation, public service delivery, and bureaucratic reform, with a sustained focus on sub-Saharan Africa. He previously worked as an economist in Ghana's Ministry of Trade and Industry as an Overseas Development Institute Fellow, and as a Senior Researcher at the Economic Policy Research Institute in Cape Town. Reform as Process has been shortlisted for the Douglass North Award for best book in institutional and organizational economics.

Research cited in this episode

Non-verifiable tasks. In organizational economics, a verifiable action is one where a third party (an auditor, a judge, an administrative tribunal) can determine objectively whether it was performed correctly. Non-verifiable tasks are those where no such determination can be made; they include innovation, adaptation, co-ordination across teams, and acting on the spirit of a rule rather than its letter. Williams draws on this framework, which originates in contract theory, to explain why formal accountability systems consistently fall short: they can only govern verifiable outputs, leaving the full range of non-verifiable tasks unaddressed and, in many cases, actively crowded out.

Performance-linked incentive systems. Williams's dataset covers 34 separate reform efforts across Ghana, Kenya, Nigeria, Senegal, South Africa, and Zambia that attempted to tie civil service pay or progression to measured performance. Not one delivered sustained differentiated incentives on an ongoing basis; only two achieved even partial delivery of rewards, and none delivered sanctions based on measured performance. Williams argues this is not isolated implementation failure but reflects a structural incompatibility between formalised performance metrics and the non-verifiable nature of much civil service work. Managers respond rationally: they set soft targets, award uniform scores, and the process becomes a tick-box exercise.

Projectization of reform. Williams uses this term to describe the dominant approach: treating change as a time-bound, discrete intervention with its own budget, acronym, and implementing team, conceived separately from the organisation's core work. This approach systematically distorts reform goals towards formally measurable outputs (new policies, new laws) rather than sustained behavioural change, undermines credibility by signalling a predetermined end date, and reinforces the perception among civil servants that reform is a temporary performance before things return to normal.

Continuous improvement. Williams draws an analogy with physical fitness: achieving a target and then stopping does not sustain the gain. High-performing organisations, in the public and private sectors alike, treat improvement as an ongoing process embedded in daily work, not a periodic project handed to a specialist unit. Starting small is not an absence of ambition; it is how credibility is built and larger changes become possible. Williams argues civil service reform should be reconceived on these terms, with performance improvement treated as the job of everyone in the organisation.

Decentralised reform leadership. The dominant model of reform leadership, Williams argues, is a visionary leader driving a top-down plan. This model is counterproductive. It personalises reform in ways that guarantee reversal when the leader moves on, and it cannot reach the day-to-day interactions among the thousands of individuals and hundreds of teams that determine how a civil service actually works. A more effective model is catalysing rather than forcing: creating conditions in which teams can identify and solve their own problems, escalate issues, co-ordinate with each other, and act on ideas for improvement without fear of being ignored or penalised.

More VoxDev Talks episodes

How government analytics can improve public sector implementation, in which Daniel Rogger and Christian Schuster discuss their efforts to use the data that already exists in governments to better understand how they function.