This thesis examines how computer modelling matters for policy-making by looking at two case studies of European fisheries management. Based on documentary analysis and ethnographic interviews and observations, the main case is located within the European Union (EU) and centred around the flatfish fishery in the North Sea with a supplementary one from outside the EU and focused on the North East Arctic cod fishery in the Barents Sea. As in other much-contested areas of public policy, fisheries officials in the EU and neighbouring countries seek to develop a universalistic and objective ground by which to depoliticise management decisions. In this sense, modelling has long become their preferred approach to produce policy relevant representations of the otherwise hidden dynamics of a fishery. Social constructivists in the field of Science and Technology Studies (STS) studying the modelling used in areas of policy-making such as, for instance, climate change have questioned whether models are the right tools for this job given that the modelling may conceal large uncertainties about their accuracy and relevance to policy-making. Some of these scholars argue for producing ‘good’ models for policy-making, and thus more robust policies, by constructively engaging the non-modellers or non-specialists in the quality assurance of the modelling. ‘Fisheries Studies’ literature suggests, however, that modelling can contribute to policy resilience despite its well-known limitations to produce accurate fish counting. It follows that models are doing something else than providing policy-salient real-looking representations. How may modelling count differently for policy-making in fisheries and beyond? Drawing on the ‘co-production’ of science and social order framework from STS, the thesis puts forward three related arguments. First, that the technologies designed to depoliticise decision-making, including modelling, become spaces for political work by policy-makers, stakeholders and scientists. Second, that the role of computer modelling for policy stems from how representational validity and political usefulness are produced together. Third, that the role of modelling for policy is mediated by virtue of being assessed together with other technologies for depoliticising as part of a whole sociotechnical infrastructure to allow evidence-based decisions. As a distinctive contribution, this thesis thus questions the presumption in many social constructivist accounts that modelling alone becomes central to the policy process and its outcomes. The significance of modelling for policy-making should be understood in terms of its contribution to processes of sociotechnical framing. Narratives that foreground the former and background the latter show an analytical bias that needs turning around.

It is conventional to argue that the autonomy and reputation of regulatory agencies depend on their expertise. Yet the studies on how independent regulatory agencies (IRAs) create and deploy their knowledge capacity are few and far apart. Normatively, the justification for delegating decision-making powers to IRAs is that they operate by using technical analysis and expertise rather than political considerations. But yet again, although delegation has been discussed as a design principle, systematic evidence on the conditions under which IRAs make use of knowledge and how is still scarce. The literature on knowledge utilization portrays a rather complex link between expertise and policy, where relevant knowledge is not always reflected in policy outcomes and plays several functions besides facilitating the solution of policy problems. Unfortunately, scholars of IRAs have not exploited the insights of this literature yet. This dissertation addresses the under-explored question of the usage of economic knowledge by IRAs. We identify four possible uses of expertise: instrumental (i.e., to solve problems); strategic (e.g. to advocate a position); symbolic (e.g., to gain legitimacy), and non-use. Our aim is to explain under which conditions a certain usage is more likely to occur. To do so, we draw on the methodological device of explanatory typologies (Elman 2005). Specifically, we select two explanatory dimensions that reflect both the context and the content of policy: the level of conflict in the policy arena, and the degree problem tractability. We use different combinations of these two dimensions to derive four hypotheses on the possible uses of expertise mentioned above. The elusive nature of knowledge utilization makes the identification and measurement of these different usages highly dependent on an in-depth understanding of the institutional, organisational, and political context in which a regulatory decision is taken. We have thus opted for a qualitative approach based on case studies and process tracing (Bennett 2010; Brady 2010; Freedman 2010) to appraise the four hypotheses. Empirically, we performed three case studies on regulatory policy decisions taken by the UK Office of Communications (Ofcom) between 2005 and 2010. We find that, given certain scope conditions, the prevalent use of economic analysis is instrumental - a finding that contradicts previous research that labelled instrumental learning as extremely rare, if not a sort of technocratic utopia. Other uses still exist however, and given other scope conditions regulators can be strategic and symbolic in their approach to knowledge and expertise. This is not surprising if we accept the notion that regulators operate in a policy environment that is eminently but not exclusively technical: to survive in a (at least partially) political environment, regulators have to deploy usages of knowledge that deviate from the instrumental type.

The current welfare-to-work reform in Britain is activating lone parents with older children and marks a step-change in the treatment of lone parents. While there has been some support for using age of child as selection criterion for the activation of lone parents, it is not clear whether this equates to selecting by ‘ability to work’ if interpreted as ability to obtain a job. The commitment of the current government to evidence-based-policy-making and the large amount of research available in this area form the justification for carrying out a policy appraisal of this aspect of the current welfare-to-work reform. The potential and likelihood to make substantial progress towards the lone parent employment and the child poverty target of the selection criteria will be assessed and compared to alternative approaches. Five selection models are identified in the international policy review: selection by age of child, transition status, employability or by caseworkers and finally, a voluntary model. The analysis is based on a critical discussion of the available evidence, an international policy review and secondary analysis of the Families and Children Study. I argue that the current approach of selecting lone parents by the age of child is unlikely to result in substantial progress towards the lone parent employment target and instead likely to create a substantial group of long-term unemployed lone parents. Alternative approaches, such as using different selection criteria that take into account the employability of lone parents are more likely to make progress towards the employment and child poverty target.

The term “evidence-based” is ubiquitous in practice and policy-making settings around the world; it is de rigueur to claim this approach. This dissertation is an inquiry into the work of evidence-based policy-making with a particular focus on the social practices of policy work/ers involved with developing policies relating to chronic disease at the Ministry of Health in British Columbia (B.C.), Canada. I begin with an examination of tensions in the policy-making literature germane to the relationship between knowledge, its production, and policy-making: the environment into which evidence-based policy-making emerged in the 1990s. Drawing on the theorising of knowledge, discourse, and power – particularly from Foucault’s work – for the analytic approach, I present the commitment to claims of “evidence-based” practices found in key government policy framework documents and policy workers’ accounts of their practices, gathered through interviews. I then show the unravelling of this commitment in those accounts. This research reveals how the policy frameworks construct chronic disease as a financial burden on the health care system and direct policy workers to develop policies with this construction in mind. The discourses associated with evidence-based policy-making narrow how policy workers can think about evidence and its production to positivist, scientific methods and numerical measures that will provide proof of cost cutting. Proponents of evidence-based policy-making laud it as keeping politics and ideology out of the policy-making process. However, the policy workers I interviewed reveal the power relations organising their deeply political work environment. Furthermore, the minutiae constituting policy-making practices produce a “managerialist approach to governance” (Edwards, Gillies, and Horsley, 2015, p. 1) in which people with chronic disease are noticeable by their near-absence. When they do appear, they are responsibilised to decrease the burden on the health/care system and the economy. I argue that as a governing project with an appearance of failure, given the many cracks in the commitment to the claim and the practices of being evidence-based, the discourse of evidence-based policy-making is actually quite successful. It has continuous effects: people are separated (so-called apolitical policy workers into imagined neutral space and decision-makers into political space), knowledge is divided, costs and responsibilities are downloaded to individuals, and evidence-based discourses appear in countless settings. The governing works. / Graduate

京都大学 / 新制・課程博士 / 博士(医学) / 甲第24285号 / 医博第4901号 / 新制||医||1061(附属図書館) / 京都大学大学院医学研究科医学専攻 / (主査)教授 髙折 晃史, 教授 小杉 眞司, 教授 佐藤 俊哉 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM