Most literature on knowledge utilization is premised on rationalist theories of organizations: bureaucracies draw on knowledge instrumentally, either to expand their power or to adjust policy output. This paper develops an alternative account of the functions of knowledge, arguing that organizations are just as likely to value knowledge as a source of legitimation, or as a way of substantiating their policy preferences. The prevalence of these different forms of knowledge utilization will depend on (a) features of the organization (its perceived stability and source of legitimation); and (b) features of the policy area (degree of contention and mode of settlement). The article applies this framework to explore the European Commission's use of knowledge in immigration policy, especially through the European Migration Network.

Since the late 1990s, the EU has sought to develop the so‐called ‘external dimension’ of cooperation on immigration and asylum: attempts to manage migration through cooperation with migration sending or transit countries. However, one can discern two rather distinct concepts of the ‘external dimension’. The first involves attempts to externalize traditional tools of domestic or EU migration control; the second to prevent the causes of migration and refugee flows, through development assistance and foreign policy tools. Both are based on different assumptions about how best to influence migration flows, and will have divergent impacts on migration flows, refugee protection and relations with third countries. It therefore makes a big difference which of the two is likely to emerge as the predominant pattern of cooperation in the future. This article looks at the factors influencing the emergence of both concepts, focusing on three central determinants: the potential of such approaches to meet migration policy goals; the institutional context; and domestic political—electoral pressures. It argues that the two last factors have militated in favour of the prevalence of externalization approaches to the detriment of longer‐term strategies of migration management, refugee protection and relations with third countries.

This article critically reviews theories of migration policy according to two criteria: methodological rigor and explanatory plausibility. It finds that political economy accounts are theoretically robust, but at the price of oversimplification. Neo-institutional theories offer more sophisticated accounts, but fall short on a number of methodological and explanatory counts. As an alternative, this article suggests a theory focusing on the functional imperatives of the state in the area of migration, which shape its responses to societal interests and institutional structures.Recent literature on the theory of migration policy has tended to be dominated by two overlapping themes: the question of why migration policies fail; and attempts to explain the inclusionary tendency of migration and integration policies. While the two issues are often treated together -and may indeed be similarly theorized -they are in principle separable. The first deals with the gap between (proclaimed) policy objectives and outcomes. It seeks to explain why states fail to achieve the goals set out in their stated migration policies (Hollifield, 1986(Hollifield, , 2000Castles, 2004; Cornelius et al. , 1992:3). The second theme is concerned with explaining the gap between the generally protectionist bent of public opinion in democratic states, and the more inclusionary policies that often emerge. While this may be understood as one dimension of the policy failure issue, it does in fact raise a rather different set of questionsquestions which may be obscured if we treat it as a subset of the first theme. Posing the question this second way implies a more narrow focus on the configuration of interests, ideas, and institutions that shape policy. It also allows us to address more specifically the question of why migration policy in liberal states does not succumb to the pressures for closure apparently endemic in democratic and welfare state systems (Bommes and Geddes, 2000;Hollifield, 2004).While there are any number of theories for explaining the first gap, recent attempts to theorize the second have been dominated by two schools: neoclassical political economy, as exemplified in the work of Gary P. Freeman; and various 1 Many thanks to Michael Bommes, Gary Freeman, James Hollifield, Mark Miller, and the anonymous reviewers at IMR for their very helpful comments on earlier drafts of this paper.

Political scientists are increasingly exhorted to ensure their research has policy 'impact', most notably via Research Excellence Framework (REF) impact case studies, and 'pathways to impact' statements in UK Research Council funding applications. Yet the assumptions underpinning these frameworks often fail to reflect available evidence and theories. Notions of 'impact', 'engagement' and 'knowledge exchange' are typically premised on simplistic, linear models of the policy process, according to which policy-makers are keen to 'utilise' expertise to produce more 'effective' policies. Such accounts overlook the rich body of literature in political science, policy studies, and sociology of knowledge, which offer more complex and nuanced accounts. Drawing on this wider literature, this paper sets out four different approaches to theorising the relationship: (1) knowledge shapes policy; (2) politics shapes knowledge; (3) co-production; and (4) autonomous spheres. We consider what each of these four approaches suggests about approaches to incentivising and measuring research impact.