Anyone who has worked in Australian research and innovation policy for more than a decade will recognise the pattern. The Chief Scientist’s The Chance to Change report (2000), the Cutler review (2008), the National Innovation and Science Agenda in 2015, Australia 2030 (Innovation and Science Australia, 2017) and the Strategic Examination of R&D last year each diagnosed the system with care, set national ambitions and recommended a suite of measures.
Each report is thoughtful. We have had forty years of vigorous agreement.
The diagnoses overlap: Business investment in R&D is too low, collaboration between universities and firms is thin, the funding architecture is fragmented, and the system is reviewed more often than it is governed.
The remedies overlap too: Reform the tax incentive, consolidate the programs, fund translation, build precincts, set a national strategy with a lead minister.
A reader who placed the recommendations of 2000, 2008, 2017 and 2025 side by side would find a core of proposals that recur with minor variation, each presented as new.
What the documents share is a near silence on execution. Which organisation will administer each measure, with what staff and systems, how a firm or a researcher will experience it, how it interacts with the dozens of measures already in place, and how anyone will know in five years whether it succeeded. These questions go to an implementation stage the inquiries treat as somebody else’s task.
The program record shows the result. Commercialisation assistance and early-stage finance have cycled through the Innovation Investment Fund (IEF) and COMET, Commercialisation Australia, the Entrepreneurs’ Infrastructure Programme, the Entrepreneurs’ Programme and the Industry Growth Program over a quarter of a century.
Each transition discarded the administrative learning, provider relationships and client familiarity the previous scheme had built up.
Lost knowledge
The explanation is a knowledge gap before it is a funding or political one. The implementation problem is among the best-documented in public policy. Pressman and Wildavsky showed fifty years ago that the number of decision points between an announcement and an outcome predicts failure better than the quality of the design (Pressman & Wildavsky, 1984).
Lipsky showed that the front-line officers who administer a program are, in effect, its final policymakers (Lipsky, 1980).
Australia once taught this knowledge as a discipline, with chairs in public administration, a standard national text and an apprenticeship inside departments that both designed and ran programs.
The knowledge went missing in two ways.
Displacement. The reform program of the 1980s and 1990s, Australia’s version of new public management, moved delivery out of departments into agencies, contracts and consultancies, and redefined what remained as generic management.
Once delivery is a matter of contracts and performance indicators, there is nothing distinctive left to teach.
The university discipline was absorbed into policy studies, policy became the prestige track in the public service, and the apprenticeship that passed administrative judgement from one generation to the next lost its home.
The Thodey review described the result as a hollowing out of capability (Thodey, 2019). The Robodebt Royal Commission showed what follows when policy, legal advice and delivery operate as strangers (Holmes, 2023).
Non-encounter. Research and innovation policy is written by scientists, economists and business advocates, and the panels that review the system are drawn from the same three groups. None of these professions trains its members in implementation.
A scientist assumes that a well-argued case produces action. An economist assumes that incentives do.
Unfortunately, neither of these approaches produces action without an organisation, authority, people and time. The knowledge was never displaced from these advocates. They never had it..
Economists without a theory of change
Economists deserve particular attention in this critique. Market economics is a theory of how prices and incentives settle, and its policy branch is a theory of how governments correct the failures of that settlement; it does not have a theory of change.
How an institution moves from one condition to another, who has to act, through which organisation and against what resistance, is outside the model, and an adviser trained in the model does not ask.
The method says whether a measure is worth doing but does not say how the measure is to be done. It also treats the organisation that has to do it as a cost.
An earlier generation of Australian economists was different. Coombs and the Department of Post-War Reconstruction, and the generation of Giblin, Copland and Wilson they drew on, had a theory of change because they were institutional economists first. They thought in organisations, employment and regions, and they built institutions to give their analysis effect (Coombs, 1981).
Research and innovation policy shows the consequence of the shift from that tradition. The tax incentive suits the market economists because it needs almost no administration.
The things the field actually needs, collaborative programs, precincts, intermediaries and translation infrastructure, are institutional, and they need exactly the knowledge the market economists do not have.
Retrieval, not restoration
None of this is a case for restoring the administrative state of the 1950s. The reform era answered real problems. The case is for retrieving the knowledge, and the retrieval is easier than it looks, because the same knowledge has been rediscovered under new names in the places innovation policy takes its cues from.
Mazzucato and Kattel argue that mission-oriented policy depends on dynamic capabilities inside the public sector (Kattel & Mazzucato, 2018). Jennifer Pahlka’s Recoding America (2023) and Dan Davies’ The Unaccountability Machine (2024) have taken the implementation argument to a general readership. The charge of nostalgia does not apply to an argument built on sources from the past five years.
What the next review can do
There are several things a review can do once it has received its terms of reference and before it writes a word of recommendation.
Carry out the implementation assessment before the recommendations are settled and publish it as part of the report. For each proposed measure: which organisation administers it, through what systems, how a firm experiences it, and what evidence will show in five years whether it succeeded. A recommendation the panel cannot answer for is not ready, however compelling the diagnosis.
Treat machinery design as part of the brief. Before recommending a new organisation, show that the function cannot be performed by an existing one with the authority and resources to do it, and count the decision points the new one adds.
Ask for delivery experience on the panel or co-opt it. Adding someone who has run a large program, a research office or a State delivery agency changes the questions a review asks.
Recommend how capability survives the transition. Where the review proposes that a program be renamed or consolidated, specify what learning, which provider relationships and which client data transfer to the successor, and require the evaluation of the outgoing program to be published before the incoming one begins.
Australia’s innovation field has agreed on what is wrong for forty years and has written the same list of remedies five times over.
The list remains undelivered because the people who write it, and the panels that endorse it, have never been asked to show how the organisations that have to deliver it will operate.
The next review can be the first to ask that question of itself, and the four steps above are where it might start.
This is an extract from a longer essay which is available on the Acton Institute website.
John H Howard is Executive Director of the Acton Institute for Policy Research and Innovation. His book Hidden in Plain Sight: How places create value from innovation is published this month.
Do you know more? Contact James Riley via Email.