Problematizations in Health Policy

This article directs attention to the significance, for health promotion advocates, of reflecting on how “problems” are constituted, or brought into existence, as particular sorts of problems, within policies and policy proposals. To this end, it introduces a poststructural analytic strategy called “What’s the Problem Represented to be?” (WPR approach), and contrasts this perspective to the ways in which “problems” are commonly conceptualized in health policy analyses (e.g., “a problem stream,” “wicked problems”). Such a perspective offers a significant rethinking of the conventional emphasis on agenda setting and policy-making processes in considering the meaning of success or failure in health policy initiatives. The starting point is a close analysis of items that are “successful,” in the sense that they make the political agenda, to see how representations of “problems” within selected policies limit what is talked about as possible or desirable, or as impossible and undesirable. This form of analysis thus enables critical reflections on the substantive content of policy initiatives in health policy. The article takes a step back from policy process theories, frameworks, and models to offer reflections at the level of paradigms. Highlighting potential dangers and limitations in positivism, interpretivism, and critical realism, it uses international, Australian, and South Australian examples in health policy to explore what poststructural policy analysis contributes to understanding the broad political influences shaping contemporary modes of rule.

presumed problematic conditions to problematizations, how "problems" are constituted-given shape and meaningwithin policies. 2 The article proceeds in two parts. In the first, four paradigms in health policy research-positivist, interpretive, critical realist, and poststructuralist-are examined for their grounding assumptions and political implications. Specific policy process theories and frameworks-respectively, comprehensive rationalism, Kingdon's (1984Kingdon's ( /2003) "problem stream," "wicked problems," and the WPR approach-are used to illustrate these premises, with a focus on how "problems" are conceptualized within these frameworks. A second section, headed "Applying WPR," directs attention to health policy analyses that ask questions about how "problems" are conceptualized within policy texts. It proceeds to contrast these approaches with a WPR form of analysis, using examples from Australian public policy. The article concludes with brief reflections on the possible uses and feasibility of including a WPR form of policy analysis in health policy research.
Throughout the article, attention is drawn to the political implications of theoretical perspectives. With Annemarie Mol (2002), the case is made that our methods "are not a way of opening a window on the world, but a way of interfering with it. They act, they mediate between an object and its representations" (p. 155, emphasis in original). In this view, paradigms matter and researchers are involved necessarily in an ontological politics (Mol, 1999), creating specific social and political realities (Bacchi & Rönnblom, 2014). It follows that researchers have a responsibility to reflect critically on the realities their methods create, enjoining a self-problematizing approach to methodological commitments.

Paradigms in Health Policy
Currently, a good deal of attention is directed to the political science models available to health policy researchers to assist them in navigating the difficult path of policy development (Bernier & Clavier, 2011;Clavier & de Leeuw, 2013;de Leeuw, 2007). In this article, I take a step back from identified frameworks and theories to offer comments at the level of paradigms (see Alvesson & Sandberg, 2011). I undertake this form of analysis because I wish to highlight common paradigmatic assumptions that cut across theoretical perspectives among health policy researchers. For example, critical realist premises underpin theory-based evaluation, which is associated with a "theory of change/action," systems theory, complexity theory, and logic programs, methods endorsed in recent World Health Organization (WHO, 2015) publications. Identified paradigmatic assumptions will be interrogated through the lens of "problems"-how "problems" are conceptualized-within specific frameworks, for example, comprehensive rationalism, Kingdon's (1984Kingdon's ( /2003 "problem stream," "wicked problems," and Bacchi's (2009) WPR approach. I take as my starting point Lucy Gilson's edited methodology reader for Health Policy and Systems Research (HRSR), published by the WHO in 2012. Gilson's (2012) heuristic has been selected for two reasons: First, it highlights the importance, for health policy research, of reflecting on the ways in which researchers understand the world and their views on how to "know" about "it" (p. 35); and, second, it directs attention to two increasingly popular perspectives in health policy research, interpretivism and critical realism. Unfortunately, Gilson neglects to consider a poststructuralist paradigm, which will be introduced following comments on positivism, interpretivism, and critical realism, with WPR as exemplar. To forecast the argument, I suggest that paradigmatic assumptions within interpretivism and critical realism pose challenges to those who wish to enlist these perspectives in a program of substantive structural change, giving pause to those who deploy them.

Positivism
Generally positivism remains the dominant paradigm within medical research, relying upon an assumption that independent (objective) scholars can access information, or evidence, about "the real." Facts and values are held to be separate and distinct. The categories of "research material" and "researcher" are treated as closed and fixed. The main research involves discovering associations among identified factors to discover "what works." This model of evidencebased medicine lies behind randomized control trials (RCTs), which are located at the summit of the "hierarchy of evidence." RCTs and systematic reviews of RCTs provide the main sources of research produced by the Cochrane Collaboration (Bacchi, 2009).
To gain credibility, social scientists try, at times, to mimic the medical model. Duflo (Duflo, Dupras, Kremer, & Sinei, 2007;Duflo & Kremer, 2003) is only the most obvious case where this mimicry takes place. More generally, there are attempts to set up "experiments" where "variables" can be tested and causality determined. The social science equivalent of the Cochrane Collaboration is the Campbell Collaboration.
The mantra "what works?" has, as an unstated premise, some notion of the problem being addressed. In other words, "what works?" implies that something requires "fixing" or "solving." Problems are exogenous to the analysis. Their status as independent phenomena requiring intervention is unquestioned.
Comprehensive rationalism. We can see this form of reasoning operating within policy studies in what has been described as comprehensive rationality or the rational comprehensive model (Bacchi, 1999). This approach endorses the view that there is a real world that is accessible to objective description and analysis. The process of making policy is set out in clearcut "stages": agenda setting, formulation, implementation, and evaluation. There is an assumption that there is some readily identifiable social/economic problem that needs "addressing" and that policy makers get together and do their best to come up with a policy that will "deal with" this problem. A further assumption is that they will approach this task rationally and come up with the best solution given cultural, political, and economic constraints. There is a presumed "decision space" at the outset of the policy process where problems are identified. The real work, as it were, consists in finding solutions-often, technical solutions-to those problems. The emphasis on measurement and technical expertise leads to the labeling of this approach, in some accounts, as "technical rationalism." The idea in comprehensive rationalism, noted above, that it makes sense to describe the policy-making process as a series of "stages," is popular in health policy research, though typically those who employ this schema insist that the model is not intended to imply strict separation among the identified steps (Walt et al., 2008). For example, the WHO's (2015) Training Manual for Health in All Policies (HiAP), offered here as an up-to-date and representative example of mainstream health policy theorizing, lists the following stages in policy making: • Agenda setting (identify the problem, conduct research, formulate policy); • Policy formation (develop policy options, negotiate, formulate policy); • Policy implementation (implement and enforce policy); and • Policy review (monitoring, evaluation, and reporting). (p. 57; emphasis added) The authors of the Manual remind readers that the complex nature of HiAP-that it is "an inherently political process that involves the reallocation of resources including power and responsibilities"-means that the process is "not necessarily linear" (WHO, 2015, p. 58).
Notably, for the purposes of this article, the initial "stage" of agenda setting is described as involving "problem identification," implying that problems exist separate from the policy process and need only to be named. The remainder of the document, however, tells a different story about "complex social problems," which are called "wicked," and about the processes of problem redefinition and reframing (WHO, 2015, pp. 54-56).
The adoption of a "stages" heuristic in this WHO Training Manual, therefore, does not necessarily signal a positivist orientation. Other paradigms, specifically interpretive and critical realist paradigms, are at work in the same document, and require further analysis, especially given the growing popularity of these paradigms in the health policy field. Nonetheless, it is worth remembering that, regardless of stated orientation, the heavy reliance on epidemiological measurement and frequent appeals to (hard) "science" (Bacigalupe, Esnaola, Martín, & Zuazagoitia, 2010, p. 505;Puska, 2007, p. 328;Sihto, Ollila, & Koivusalo, 2006, p. 4) indicate the continuing strength of positivism in health policy research.

Interpretivism
Interpretivism turns its primary attention to the social actors in policy research, both those conducting the research and those who are research "subjects." Interpretivists are located in a hermeneutic tradition that sees people's self-interpretations as central to understanding social organization. In contrast to positivism, the category of "the subject" is opened up to reflect on how subjects create meaning in their lives. The positivist objective of establishing distance from research subjects is countered by an emphasis on how subjects interpret their experiences (Gilson, 2012, p. 35). Attention shifts from the positivist objective of predicting outcomes to an interpretive focus on understanding social interactions. Rather than attempting to isolate "facts" from "values," as in positivism, people's beliefs and intentions become central to the research exercise.
An interpretive approach to policy making can be traced to the development of a political rationalist tradition in 1970s policy studies (Bacchi, 1999). Theorists in this tradition object to the impression conveyed by technical rationalists that policy is a straightforward matter of finding technical answers (solutions) to readily identified problems. They are much more sensitive to the give and take of politics, to the shifting of perspectives and positions, and to the role played by politics, here meaning party politics and bureaucratic politics, in decision making. Importantly, they address the need to talk about the role of values in policy making.
Political rationalists counter the emphasis on expertise in comprehensive rationalism with a commitment to ensuring that the process of decision making is as open as possible. Hence, they are pluralists. They are also incrementalists, believing that the "limited cognitive capacity of the human being" requires that analysis should be simplified in all possible ways, "for example, by proceeding step by step through trial and error rather than trying to comprehend a problem in its entirety" (Lindblom, 1980, p. 35).
It is possible to identify among early political rationalists, such as Lindblom (1980) and Wildavsky (1979), similar concerns and themes to those dominating health policy discussions and reports today. Specifically, they expressed growing disquiet about the complexity, uncertainty, and political maneuvering surrounding policy processes. The political climate of the 1970s highlighted an erosion of consensus in American politics-where these debates played out-leading to the conclusion that "the nature of the problem is itself in doubt" (Rein & Schön, 1977, p. 237). Importantly for this article, therefore, political rationalists objected to the lack of attention in technical rationalism to that "most crucial aspect" of policy development described as "problem formulation" or "problem setting" (Rein & Schön, 1977, pp. 235-236).
While this naming of problems as critical components of the policy process is salutary, for political rationalists, "problem setting" is limited to the analytic task of forming problems in ways that make them manageable. Dery (1984), for example, dedicates an entire book to "problem definition," which in his view requires political scientists to be concerned "with the production of administratively workable and politically realistic ideas for solving social problems" (p. 38). Wildavsky (1979) concurs that "in public policy . . . creativity consists of finding a problem about which something can and ought to be done. In a word, the solution is part of defining the problem" (p. 3). This pragmatic orientation reduces the space for contesting how "problems" are constituted in policies.
Along similar lines, Rein and Schön (1977) introduce the language of framing to clarify what "problem setting" involves. "Framing," they explain, refers to the process by which "worries, arising in problematic situations, can be converted into the orderly formation of problems" (Rein and Schön, 1977, p. 238), a goal echoed in the sub-title of their important book on framing theory (Schön & Rein, 1994): "Towards the resolution of intractable policy controversies." Rein and Schön (1977) set as criteria for this analytic task a principle of consistency and a principle of testability-"the theory or model contained in problem setting should be subject to empirical test; it should be capable of disconfirmation" (p. 249). The appeal to empirical testing indicates a lingering positivism despite the sensitivity to the impact of competing values in political processes. Moreover, despite expressed doubts about the very "nature of the problem" (Rein & Schön, 1977, p. 237), there is still a sense that the goal is "solving social problems" (Dery, 1984, p. 38), as if these are readily identifiable and outside the policy-making process.
The language of problem definition and problem framing proliferates in contemporary health policy analysis, as does the concept of "wicked problems," coined by Rittel and Webber (1973). An important task for this article is considering how "problems" are conceptualized in these usages. "Wicked problems" is tackled in the section on critical realism, where, I suggest, it belongs. The remainder of this section pursues contemporary interpretations of problem definition and framing in the work of John Kingdon (1984Kingdon ( /2003 and in the WHO (2015) HiAP Training Manual, introduced above.
Kingdon's "three streams." Although Kingdon's book introducing his "three streams" approach to policy development was first published over 30 years ago, in 1984, it has recently been endorsed by Colebatch, a leading interpretivist (Lancaster, Ritter, & Colebatch, 2014), and continues to be one of the main political science frameworks adapted by health policy researchers (Baum, Lawless, & Williams, 2013;Leppo, Ollila, Peña, Wismar, & Cook, 2013). Kingdon's focus is policy making not policy analysis. That is, he sets out to describe the forces at work getting "conditions" onto the political agenda as "problems." In the Foreword to the second edition, Thurber (2003) describes the goal as understanding "the complexity and dynamics of how the national agenda is set" (p. vii).
Kingdon distinguishes among three separate streams within policy making, which he labels "problem, policy, and politics." A novelty in his approach is his insistence on the separation of these three streams-each has a life of its own. So, for example, Kingdon insists that policies are generated in the policy stream with no necessary connection to the problem stream. In fact, he argues that proposals more or less wait for a problem to become available. This argument leads him to question the emphasis in much policy theory on "problem solving" (Kingdon, 1984(Kingdon, /2003). His concern, by contrast, is to explain how items make it onto the policy agenda through "windows of opportunity" and due to the skills of "policy entrepreneurs." In Kingdon's (1984Kingdon's ( /2003 view, a key political process is how "conditions" become problems, which he states signals that someone has decided to do something about them (p. 109). This "becoming" problem, he argues, always has "a perceptual, interpretive element" (p. 110). There is a short section in the book on problem definition and the language of framing is also used. As Lancaster et al. (2014) describe, for Kingdon, policy making is conceived as "an ongoing process of managing the problematic" (p. 148).
This very description signals the ambiguity around the notion of "problem" in Kingdon. What is "the problematic"? How is this determined? At times, Kingdon (2003) refers to problems as if they have an independent status. For example, he says that "problems are not simply the conditions or external events themselves" (p. 109; emphasis added), implying of course that, to an extent, they are exactly those conditions or events. He tells us, "Problems are brought to the attention of people in and around government" (p. 19), suggesting that they exist exogenously, though he also states, "The data do not speak for themselves" (p. 94). The exact status of problems therefore is left hanging due to the primary concern with agenda-setting-having an issue reach the policy agenda. For Kingdon, a problem is an issue that achieves problem status! This argument, for which Kingdon is most often cited (Exworthy, 2008;Baum et al., 2013), leaves behind the recognition of contestation over the understanding of the problem in Kingdon's discussion of problem definition and framing. It comes to appear that the primary political exercise is having an issue taken up on a policy agenda, downplaying or rendering irrelevant the key dimension of how the problem is to be conceived. This ambivalence about the nature or status of a "problem" is reflected in other interpretivist accounts.
While "issues" are treated as "equivocal discussion topics that are named, blamed and claimed as disputants argue about them" (Dewulf et al., 2009, p. 170), they appear also as products of "multiple perspectives on how to address a particular problem" (Koon et al., 2016, p. 3

, emphasis added).
Here it is important to note that, for Kingdon, as for other interpretivists (Fischer, 2003;Hoppe, 2011), reflecting the hermeneutic influence, problem definition and framing are understood primarily as activities instigated by social actors. Referring specifically to interpretive research on framing, Koon et al. (2016) note the focus on "how actors create meaning in the policy process and how they package these meanings for instrumental and expressive purposes" (p. 7). As an example of this tradition, Rochefort and Cobb (1994) consider how governments respond to the competing problem definitions produced by "issue advocates" (p. 15; for discussion, see Bacchi, 1999, pp. 36-37, 45). Similarly, for Kingdon, policy making is "an arena where a variety of participants with different perspectives, power and roles contest both what the policy problem is and what (or indeed whether anything) should be done about it" (Lancaster et al., 2014, p. 149).
By contrast, as discussed below, in the place of a sovereign subject who can access "true" meaning (Foucault, 1972, p. 54), poststructural policy analysis considers how governmental problematizations produce particular kinds of provisional "subject". Social actors are understood to be in continual formation; hence, they form part of what must be "interpreted" rather than the starting point of interpretation. In line with this perspective, whereas interpretivists tend to focus on the problematizations that people produce, poststructural policy analysis interrogates the governmental problematizations that constitute what "subjects" can become (Bacchi & Goodwin, in press).
As a result, in poststructural policy analysis, the analytic focus shifts from the competing perspectives of policy actors-how they understand a "problem"-to the problematizations (the ways in which "problems" are produced and represented) in governmental policies and practices (Bacchi, 2015). Whereas for Kingdon, the goal is achieving problem status for a "condition," through its inclusion on a government's agenda, for analytic strategies such as WPR (developed below), the fixing of "troubling conditions" (Bacchi, 2009, p. xi) as "problems" or "social problems" instigates critical analysis of how those "problems" are constituted as "problems" of a particular sort within policies. For Kingdon, achieving problem status for a "condition" is the goal; for WPR, such an "achievement" initiates a critical analytical process (see examples of application of WPR later in the article).
The HiAP Training Manual (WHO, 2015), as signaled above, serves as a useful example of an interpretive perspective. It draws explicitly on Kingdon's three streams model to reflect on the possibility of creating "windows of opportunity" (p. 58). Attention is also directed to "framing," described as "how an issue is defined, which can in turn influence how the issue is viewed (non-issue, problem, crisis, etc.), who is considered responsible and the cause and possible solutions" (p. 59). The particular concern is encouraging "policy stakeholders" (i.e., social actors) to engage in "redefining or reframing the problem" to allow "for new ways of understanding, which can encourage new stakeholders to engage in the policy process" (p. 59). Again, drawing on Kingdon, "policy champions/policy entrepreneurs," those who "proactively promote policy reforms," also "frame discussion of the issue, build consensus, attract resources, and seize and create opportunities to move the reform forward" (p. 61). Alongside this brief acknowledgment of how "problems" involve interpretation, the bulk of the numerous references to policy problems in the HiAP Manual imply that their status is uncontentious (e.g., "strong evidence of the problem," p. 32; "the magnitude of the problem," p. 59).
This variation in the meanings of the term problem signals, I suggest, the need for more attention to the implications of specific usages. While interpretivists helpfully alert analysts to the ways in which social actors give "problems" specific meanings, the political implications of how "problems" are constituted within policies are not generally considered. However, several authors use the language of framing to refer to the meanings produced within a policy rather than meanings imputed by social actors. These contributions and how WPR differs from them are considered in the second part of the article. Before undertaking this analysis, it is necessary to examine both the increasingly popular critical realist paradigm and a poststructuralist alternative. Gilson (2012) usefully locates critical realism "somewhere between" positivism and interpretivism (pp. 35-36). Theorists within this paradigm postulate a reality existing "independently of social actors" while accepting that the interpretations of those actors can influence that reality, and that a "range of individual, group, organizational and societal processes and structures" influence human action.

Critical Realism
Critical realism has had a significant influence in health policy research, primarily through the approach to program evaluation developed by Pawson andTilley (1997, 2004;Haigh, Harris, & Haigh, 2012;Hunter & Killoran, 2004;Pawson, Greenhalgh, Harvey, & Walshe, 2005;Sanderson, 2000). "Realist evaluation," or sometimes "realistic evaluation" (Pawson & Tilley, 2004, p. 3), the focus of analysis in this article, has links to theory-based evaluation-an approach with a long history in organization studies-to Weiss's (2000) "theory of change" and to "logic programs" (Brickmayer & Weiss, 2000;Leeuw, 2003;Taylor-Powell, Jones, & Henert, 2003). This cluster of theories alters the positivist evidence-based question "what works?" asking instead "what works for whom in what circumstances and in what respects, and how?" (Pawson & Tilley, 2004, p. 2). This question has great appeal among health policy researchers, as it recognizes the need to examine the "contexts" within which policy interventions operate (Baum et al., 2014, p. i135;Clavier & de Leeuw, 2013, p. 16). However, it becomes necessary to reflect on how that formulation-"what works for whom in what circumstance?"-conceptualizes contexts, subjects, and problems.
Importantly, the primary focus in realist evaluation is on the behaviors of social actors. Programs (policies) are described as theories that raise hypotheses about how people will behave in specific circumstances. As Pawson et al. (2005) explain, "Such conjectures are grounded on assumptions about what gives rise to poor performance, inappropriate behavior and so on, and how changes may be made to these patterns" (p. S1:22). Programs, we are told, "only work through the stakeholders' reasoning and knowledge of that reasoning is integral to understanding its outcomes" (p. S1:22). This insistence that "at least part of the explanation" for the successes and failures of interventions can be traced to "reasoning and personal choices of different actors and participants" indicates a link to rational choice theory, acknowledged by Pawson (2002, p. 356). To test hypotheses about "behavioural mechanisms," Leeuw (2003) turns to "the state of the art within the social/behavioural/economic sciences" (p. 8).
As the poststructural scholars Glynos and Howarth (2007, p. 84) point out, this focus on individuals and their behaviors betrays a methodological individualism, which, I suggest, ought to be of concern to health policy researchers who promote the social determinants of health (SDH). Indeed, given the focus on individual behaviors, the popularity of critical realist forms of evaluation in health policy research may inadvertently contribute to the phenomenon of "lifestyle drift"-the emphasis in much contemporary health policy on links between lifestyle "factors" and poor health-which so disturbs these researchers (Baum, 2011;Hunter, Popay, & Tannahill, 2010;Schrecker, 2013). Pawson and Tilley (2004) emphasize that "realists regard programmes as rather sophisticated social interactions set amidst a complex social reality," creating a link to complexity theory, systems theory, and "wicked problems," detailed below (p. 6). To "deal with" such "intricacy" (complexity), they call upon "science" to develop an "analytic framework to break down systems into their key components and processes"-"mechanisms," "contexts," and "outcomes"called CMOCs (context-mechanism-outcome pattern configurations) (pp. 6, 9). Programs provide resources to social actors; how those actors interpret and act upon the "intervention stratagem," or "mechanism," then determines "success" or "failure" (p. 6), returning us to the motivations and behaviors of individuals (i.e., methodological individualism).
The researcher (with the stakeholder) posits the "potential processes through which a program may work." These processes consist of hypotheses or conjectures about how people will behave, as mentioned above. They "pinpoint the ways in which the resources on offer may permeate into the reasoning of the subjects" (Pawson & Tilley, 2004, p. 7). The evaluator's task is to test these hypotheses empirically, drawing on a wide range of "evidence," both qualitative and quantitative (p. 11). On this point, Pawson and Tilley insist, "Realist research is absolutely conventional, and pleased to be so, in utilizing the time-honored 'research cycle' of hypothesis testing and refinement" (p. 10).
This evaluation method is underpinned by an understanding of "causation" as "complex." There is an explicit challenge to the positivist stance on "causal laws"-described as a "successionist" model of causality dependent on "constant conjunctures"-and an endorsement of a "generative" model (Pawson et al., 2005, p. S1:21-22). To make a "causal inference" between "two events (X and Y)," "one needs to understand the underlying mechanism (M) that connects them and the context (C) in which the relationship occurs" (p. S1:21-22). Recalling that mechanisms refer to hypotheses about individuals' behaviors, this understanding of causality relies upon the same methodological individualism highlighted and queried above. Moreover, the commitment to scientific testing of hypothesized "causal mechanisms" indicates a residual positivism, despite protestations to the contrary (see Glynos & Howarth, 2007, p. 32).
As described by Glynos and Howarth (2007), the realist view of causality as necessarily operating through social actors appears in descriptions of "causal chains" (p. 90), which "link" to individuals in the last instance (see also "causal pathways, " Exworthy, 2008, p. 319). Indeed, many of the explanatory models of social change operating in public health and SDH research position the individual as the ultimate target-for example, "causes of the causes" (Marmot & Allen, 2014, p. S517), "social determinants of behaviours and lifestyles" (Marmot, Allen, Bell, Bloomer, & Goldblatt, 2012, p. 1025, and Dahlgren and Whitehead's (1991) model of the main determinants of health with individual lifestyle factors positioned near the center of the concentric circles of "influence" (Exworthy, 2008, p. 320). Freeman (2006) appropriately characterizes the last of these as belonging "securely in a rationalist, hierarchical, or modernist conception of cause, one that can confidently separate dependent from independent and proximal from distal variables" (p. 65).
In these explanations, "contexts" tend to be treated as "mediating factors," located somewhere along the "chain." Hart and Moore (2014) note the tendency in epidemiology to constitute social factors ("contexts") as intermediaries, "held at arms-length from causation" (p. 406). Krieger (2008) cautions against distinguishing between the "murky realm of 'distal'" causal factors and "proximal" causal factors, which positions structural or social interventions "at a distance" (p. 223; see also Freeman, 2006, on upstream/downstream). This caution appears relevant to "causal chains" as well. Logic programs, which have clear links to theory-based evaluation and critical realism, incorporate "external factors" (or "environment" or "context") as part of the analysis. However, such factors attract little attention and the primary focus remains on the individual subject (Taylor-Powell et al., 2003). Moreover, there are as many disputes about the nature of a "context" as there are about the character of a "problem" (think, for example, of debates about "globalization"), cautioning against reliance on an uncontested notion of context (Dilkes-Frayne, 2014).
In terms of approaches to policy making, realist evaluation again appears to be quite conventional. "Typically," according to Pawson and Tilley (2004), there are a series of "phases" (p. 13). The first phase, they explain, involves "preliminary analysis of the problem" and "problem identification," reminding us somewhat disturbingly of comprehensive rationalism (Pawson et al., 2005, p. S1:22; emphasis in original). The critical realist, Norman Fairclough (2013), declares himself a "moderate constructivist," in which the point of critique is to ask "what the problems really are" (p. 185). In his account, problematizations represent the views of various groups of social actors about the nature of those problems. This understanding of problematization sits at some distance from the focus in Foucault-influenced poststructuralism and a WPR analysis on the implicit problematizations within policies, described in the following section.
In addition, in realist evaluation, the room for critical analysis is constrained because problems tend to be prescribed by those commissioning the evaluation. Consider, for example, Pawson et al.'s (2005) realist evaluation of performance measurement. Where a WPR analysis would identify such a program as a governing technology and hence deserving of critical analysis (see below), for Pawson et al. (2005), the starting point is simply to measure, rate, and ("sometimes") rank "the performance in question" (p. S1:22). Similarly, in Tilley's (2004Tilley's ( , 2010 work on situational crime prevention and "problem-oriented policing" (emphasis added), it is not possible to raise questions about the meaning of "crime" (compare Bacchi, 2009, Chapter 5). Rather, the assumed existence and character of so-called "problem behaviors" are unquestioned, suggesting a worrying normative agenda in realist evaluation.
Illustrating the possibly didactic and integrative character of programs that assume problems, in comments on a theorybased evaluation of some U.S. community initiatives, Weiss (2000) concludes, Now that services are more convenient and more appropriate for the residents' needs, they attend regularly and do what they are supposed to do (take medication, attend homework help programs, and so on). The end result is improved functioning of the families and healthy, productive adolescents. (p. 105, emphasis added) In this stance, health "becomes a metaphor for the success of the state with respect to those governed" (McQueen, Wismar, Lin, & Jones, 2012, p. 8). These limitations on the kinds of questions that can be raised about policies and programs appear again in the turn to "complexity," "systems," and "wicked problems," pursued below.
"Wicked problems." The language of "wicked problems" has proliferated in health policy research alongside the uptake of critical realist forms of evaluation. The link between the two is complexity, a concept with diverse associations (Urry, 2005). For critical realists, as Baum et al. (2014, p. i133) note, quoting Pawson, Greenhalgh, Harvey, and Walshe (2004, p. iv), "social (or in this case policy) interventions are 'complex systems thrust amidst complex systems,'" linking this perspective to systems theory (de Savigny & Adam, 2009, p. 30). Hunter (2013) refers to "the complex reality and nature of wicked problems" (p. 149, emphasis added).
"Wicked problems" is conceptual shorthand used to characterize "messy," "fuzzy" problems, which are described as multi-causal and requiring intersectoral interventions, such as HiAP. The concept can be traced, in its initial formulation, to the urban planners, Rittel and Webber (1973)

. A more recent incarnation is Ritchey's (2011) Wicked problems-Social Messes, referenced in the WHO's (2015) HiAP
Training Manual. From the Australian Public Service Commission (2007, p. 1) to the WHO (2015, pp. 51-4), there is unanimity that "wicked problems"-the prime example is obesity-occur in "health areas." "Wicked problems" are characterized differently in specific sites, with important political implications. The WHO's (2015) HiAP Training Manual, for example, describes wicked problems as "subjective" (p. 55), a characterization that appears in Ritchey (2011, p. 20) but not in Rittel and Webber's (1973) original formulation. Such a characterization seems dangerously depoliticizing.
The WHO makes the point that, for "complex" or "wicked" problems, "interventions must be flexible as outcomes may be unforeseen" (WHO, 2015, p. 52). At the same time, however, it is held to be possible to "model" "wicked problems" through methods such as General Morphological Analysis (GMA; Ritchey, 2011; see also Tremblay and Richard, 2011, p. 381). Complexity and hence "wicked problems" are thus deemed to be manageable through the kind of logic models introduced above, models traceable to logframe matrices and "problem trees" (AusAid, 2005). As in realist evaluation, these models work through the behaviors of social actors, reinforcing a focus on lifestyles that ought to be a concern for proponents of SDH. Ferlie, Fitzgerald, McGivern, Dopson, and Bennett (2011) describe "challenging behaviour change objectives" as a core feature of a "wicked problems" (p. 322) approach to policy (see also Australian Public Service Commission, 2007, p. 32).
Systems thinking claims to offer "an approach to problem solving that views 'problems' as part of a wider, dynamic system" (de Savigny & Adam, 2009, p. 33, emphasis added). However, the space to consider just where these "problems" come from and how they are understood is severely constrained. In Ritchey's (2011) GMA, the "problem area" is set by the "principal client," and stakeholders are invited to consider "different aspects of the problem complex" (pp. 64-67). He offers the case study of "youth and criminality" in Sweden as a "consequence of social exclusion" (p. 81). As with "performance measurement" and "situational crime prevention" in Pawson and Tilley's work (see above), the key referents "youth," "criminality," and "social exclusion" are taken as given, restricting the critical interrogation that is possible. In effect, the term wicked problems, therefore, reinforces the conventional, pervasive view of policy as reacting to problems that must be solved. De Savigny and Adam (2009) describe systems thinking as a problem-solving approach. Moreover, the objective of GMA is to create a "common problem concept" (Ritchey, 2011, p. 28), reducing the possibility of contestation. By contrast, poststructuralism, as described below, creates a space for questioning taken-forgranted concepts and categories.

Poststructuralism
There is no single poststructural theory. The WPR approach, offered as exemplar in this section (below), draws upon a Foucault-influenced poststructural perspective. In this view, there is a focus on the plurality of practices that produce hierarchical and inegalitarian technologies of rule. Commitments to some notion of truth are replaced by an emphasis on discourses, or knowledges, and the constitutive role of knowledge practices. This emphasis on a plurality of practices means that "things," including objects and subjects, are not essences. Rather, they are seen to be contingent and in process, always developing and subject to change. We saw above how this approach to the subject as provisional and in formation leads to important differences from interpretivism, which takes the subject as the starting point for meaning making. The focus on the constitutive character of practices also produces a sharp distinction from critical realist premises. In Foucault-influenced poststructuralism, realities emerge in practices. Hence, a singular reality, assumed in critical realism, is deemed to be a political creation rather than an ontological given.
It has been suggested (Cilliers, 1998) that poststructuralism and complexity are natural allies, given the emphasis in the latter on plurality and contingency. However, such an argument ignores the political ethic of poststructualism. As Dillon (2000) suggests, "different techniques themselves entail different ethics or ways of being" (p. 7). Specifically, he finds that "much complexity thinking remains indebted to the modern project of science," as displayed in the models above. The political focus in these accounts becomes how we can influence people to behave in desired/"desirable" ways instead of how we can produce a just society. The suggestion developed in this article is that some of these models, specifically those associated with realist evaluation, may actively undermine those justice goals by closing off the space to consider how "problems" are constituted in policies.

Bacchi's WPR approach.
A WPR approach to policy analysis undertakes a task markedly different from the numerous policy process theories offered to assist advocates to get a policy onto the agenda or to explain "why and how policies fail or succeed" (de Leeuw, Clavier, & Benton, 2014, p. 8). By contrast, the starting point is a close analysis of items that make the political agenda to reflect upon the overall shape of policy initiatives, what they encompass and what they leave out. In addition, a WPR approach challenges the conventional and pervasive view that policies address problems. I describe this view as reactive in the sense that policies are conceived to be reactions to presumed problems or problematic situationswith the term "problem" operating as a catch-all conceptual shorthand for diffuse and often unspecified issues.
In contrast, a WPR approach describes policies as productive or creative-constituting (making come into existence) "problems" as particular sorts of problems. The analysis starts from policy proposals and recommendations to see how they represent or constitute the "problems" they purport to address. Governing, it is argued, takes place through these problematizations. We are not talking here about competing understandings or perceptions of a problem presumed to exist, but the texts and discourses that produce "problems" of particular sorts within policies (Reekie, 1994, p. 463). Discourses, as understood in this perspective, consist of socially produced forms of knowledge that constitute "the real." Importantly, in this understanding, references to governing extend well beyond political institutions, party politics, networks, and even social movements to encompass the full range of knowledges (or discourses) and sites involved in societal administration. Because the practices and theories of "experts," researchers, and professionals from diverse fields, including psychology, epidemiology, health promotion, and political science, are involved in governing, they become targets for critical analysis. Governing in this broad sense is described as a "problematizing activity" (Rose & Miller, 1992, p. 181), in which "policy cannot get to work without first problematizing its territory" (T. Osborne, 1997, p. 174).
The key to understanding how governing takes place, therefore, is to study how governing practices, understood broadly, problematize issues.
To say that policies create "problems" as particular sorts of problems does not mean to suggest that governments set out to produce homelessness or poverty, or even to deliberately represent homelessness or poverty in particular ways. Rather, the proposition is that the specific policy or policy proposal contains within it an implicit representation of the "problem," referred to as a problem representation (Bacchi, 2009). This proposition relies upon a simple idea: That what we propose to do about something indicates what we think needs to change and hence what we think is problematicthat is, what the "problem" is represented or constituted to be. Following this logic, it becomes possible to "read off" how the "problem" is constituted from examining a specific policy proposal. To conduct research in this way, one starts from the proposed "solution" (the policy) and asks-"if the suggestion is that this form of change or intervention is required, what is the 'problem' represented (constituted) to be?" We saw an example above in the case of training programs for women.
Crucially, representations of "problems" are not images or imagined states; they are interventions. As Shapiro (1988) explains, "Representations do not imitate reality but are the practices through which things take on meaning and value" (p. xi). Anderson and Harrison (2010) concur, As things and events they [representations] enact worlds, rather than being simple go-betweens tasked with re-presenting some pre-existing order or force. In their taking-place they have an expressive power as active interventions in the co-fabrication of worlds. (p. 14) A problem representation therefore is the way in which a particular policy "problem" is constituted as the real (Bacchi, 2012, p. 151).
Problematizations thus become part of how we are governed. That is, governing takes place through the ways in which "problems" are constituted in policies. Put in other words, we are governed through problematizations, rather than through policies, signaling the importance of critically interrogating problem representations. To undertake this task, a WPR approach offers several forms of interrelated questioning and analysis, which can be followed sequentially or applied as part of an integrated analysis (adapted from Bacchi, 2009, p. 48): • • Question 1: What's the "problem" of (e.g., "discrimination," "problem gamblers," "drug use/abuse," "domestic violence," "absenteeism," "anti-social behavior") represented to be (constituted to be) in a specific policy or policies? • • Question 2: What presuppositions-necessary meanings antecedent to an argument-and assumptions (ontological, epistemological) underlie this representation of the "problem" (problem representation)? This question involves a form of Foucauldian archaeology (Foucault, 1972). • • Question 3: How has this representation of the "problem" come about? This question involves a form of Foucauldian genealogy (Foucault, 1971(Foucault, /1977. • • Question 4: What is left unproblematic in this problem representation? Where are the silences? • • Question 5: What effects (discursive, subjectification, and lived) are produced by this representation of the "problem"? • • Question 6: How and where has this representation of the "problem" been produced, disseminated, and defended? How has it been and/or can it be questioned, disrupted, and replaced? • • Step 7: Apply this list of questions to one's own problem representations.
Adopting a genealogical perspective (Question 3) ensures that policies are considered in both temporal and spatial context, always with an eye to the contestable nature of contexts (discussion above). Examining the effects or implications of specific problem representations (Question 5) involves consideration of how they may make it difficult to raise certain issues (see also Question 4), how "subjects" are produced (or constituted) within them (subjectification), and the lived effects that accompany them. In this way, WPR provides a form of critical evaluation of policies and policy proposals.
This approach is poststructural in the sense that it takes nothing for granted in the "objects," "subjects," or "problems" that form the basis of policy analysis. Rather, these "things" are understood as shaped, or constituted, through practices. Practices, in this vocabulary, relate to a "zone or space of governmental intervention": "To focus on practices is not to focus on the hard edge of (to adopt a superannuated vocabulary) the 'real-concrete,' but upon the leading edges of governmental problematisation" (T. Osborne, 1997, p. 176).
This perspective has significant implications for the ways in which we approach the policy field. Taken-for-granted categories and concepts-"drugs," "addiction," "crime," "youth," "skill," "human capital," "wellbeing"-become things to interrogate through scrutinizing the governmental practices and associated mentalities or rationalities that produce them (Dean, 1999). In addition, "subjects" are not presumed to exist as sovereign social actors; rather, who we become forms part of a continuing process, in which policies play an active role through making certain subject positions available. The concept "subject position" refers to the kinds of "subject" that it is possible to become in specific discourses/knowledges (e.g., "the consumer," "the caring mother," "the delinquent," "the problem gambler," "the literate citizen"). For example, in other work, I describe policies as gendering practices to capture the active, ongoing, and always incomplete processes that constitute (make come into existence; Jones, 1997, p. 265) "women" and "men" as specific kinds of unequal political "subject" (Bacchi, in press).
Broadening the scope of the analysis to encompass the governing knowledges of "experts" and professionals brings new questions to policy analysis. Instead of assuming that government is limited to legislative enactments, we consider how such enactments rely on professional knowledges, for example, on psychological theories of child development or psychiatric diagnostic criteria (see Fraser, Moore, & Keane, 2014). Consider, for example, the policy implications of the categorization of homosexuality as "sexual deviance" in the 1952 edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM; Bacchi, 2009, p. 235) and the inclusion of gambling as a diagnosable "behavioral addiction" in the most recent edition of the DSM (American Psychiatric Association, 2013, section 312.31). A poststructural skepticism about the taken-for-granted status of such knowledges leads to questions about their genealogy and their political implications/effects. As Foucault (1981Foucault ( /1994 elaborates, this form of critique does not consist in saying that things aren't good the way they are. It consists in seeing on what type of assumptions, of familiar notions, of established, unexamined ways of thinking the accepted practices are based. (p. 456) In line with a poststructural sensitivity to knowledge creation, the WPR approach encompasses an undertaking to apply the questions in the approach to one's own policy proposals in a practice of self-problematization (Step 7, above). Such a practice reflects recognition that each of us is immersed in the governing knowledges, the "unexamined ways of thinking," of our age and that those among us positioned as "experts" or professionals have a responsibility to reflect on the political implications of the knowledges we deploy. The goal is to create the space to reflect critically on all proposals for change, including one's own recommendations, to govern "with a minimum of domination" (Foucault, 1987, p. 129

Applying WPR
This section of the article offers two exemplars to illustrate how to apply a WPR analytic strategy to specific policy examples: "Parental Engagement With Literacy" and "Income Quarantining." Prior to undertaking this analysis, it is necessary to consider some novel applications of framing theory.

Alternative Framing Approaches
In the discussion of interpretivism above, the point was made that many scholars operating within this paradigm tend to focus on how social actors are involved in problem framing and definition. However, some important contributions in the health policy field (and elsewhere) explicitly analyze how "problems" are conceptualized within policy documents, suggesting useful connections to the kind of analysis offered by a WPR approach. These contributions move beyond the majority of policy process theories that focus on agenda setting and policy development to reflect on the substantive content of health policies. It is important therefore to examine some of these contributions, to clarify their project and how WPR differs. The examples I use (Bambra, Smith, Garthwaite, Joyce, & Hunter, 2011;Freeman, 2006;Smith et al., 2009) take as their target how "health inequalities" or "health inequities" are conceptualized in selected policies, a central question for health policy research, as noted by Embrett and Randall (2014, pp. 153-154). Smith et al. (2009) examine national policy statements on health inequalities in England, Scotland, and Wales since 1997, specifically to see how the "policy problem" of "health inequalities" is "framed" differently within these varied texts. Here the focus is not on social actors "framing" an issue for "instrumental and expressive purposes" (Koon et al., 2016, p. 7), as observed in earlier interpretive accounts-or at least not directly-but on how the policy text itself gives meaning to "equity," which is recognized to be a contested concept (see Baum et al., 2013, p. 210). To make their comparisons among conceptions of "health inequalities," the authors draw on the well-known distinction developed by Graham and Kelly (2004) among approaches that conceptualize "health inequalities" as either a "problem" of "health disadvantage," or "as a health gap," or as "social gradient." Theoretically, Smith et al. (2009) locate themselves within the "linguistic turn" (p. 220) and utilize critical discourse analysis, associated with Fairclough (2013). They set their task as identifying "how the [policy] texts present the causes of, and solutions to, health inequalities," and express disappointment that rhetorical commitments to tackling "wider determinants" translate into quite limited interventions (Smith et al., 2009, p. 220 ff). Freeman (2006) asks similar questions about "how the problem of health equity is constructed in different countries" and also takes policy texts as his primary research material (p. 51). He uses the language of "problem definition" and "framing," tracing the latter to cognitive psychology (p. 59 fn 13; see also Dewulf et al., 2009). With Smith et al. (2009), Freeman locates himself theoretically within the "linguistic turn," highlighting connections between framing and rhetoric, and linking his analysis to the interpretive tradition of scholars such as Rein and Schön (see above), Fischer (2003), and Stone (1988;Freeman, 2006, p. 52). Usefully Freeman explores assumptions about disease causation and the policy process that operate within selected policy texts, acknowledging the importance of "deep-rooted beliefs about rationality and linearity" (p. 64). Bambra et al. (2011) explore the "different ways in which the 'problem' of health inequalities is conceived of" (p. 401) in three landmark British reports on the topic, those by Black (1980), Acheson (1998), andMarmot (2010). With Smith et al. (2009), they indicate disappointment with the recommendations in the reports. They express strong reservations about Marmot's reliance on psychosocial, rather than material, factors as explanations of "health inequality" (p. 403). They also have concerns about the Marmot Review's "capabilities" discourse, which, they argue, could be translated into policy in ways which merely shift the responsibility for poor health onto individuals and communities who fail to develop the social networks required to ensure "resilience" against health problems. (p. 403) WPR, as outlined earlier, does not analyze "problems" and "solutions" separately in the way that these contributions do. Rather, it begins with a postulated solution and identifies the problem representation implicit within it. Hence, the level of critique differs. The target is not the rhetorical distance between descriptions of "problems" and "recommendations," which are judged to be limited or disappointing. The project starts from recommendations to see how the "problem" is constituted within them.
In this form of analysis, what is of most interest and concern are continuities within policies, across statements of "problems" and "solutions," continuities that rest on deepseated ontological and epistemological premises (Question 2 above). The issue is not the rhetorical ploys of governments judged to be reluctant to deliver substantive change but "unexamined ways of thinking" (Foucault, 1981(Foucault, /1994) that underpin specific policy proposals and shape "problems" as particular kinds of problems. In contrast to the three examples in this section, a WPR analysis does not focus primarily on language (see Bacchi & Bonham, 2014). Rather, governing is deemed to take place through the discourses, or knowledges, on which policy proposals rely. The usefulness of this form of analysis is illustrated through two examples: The South Australian (SA) Government's Health Lens Analysis (HLA) on education, called "Parental Engagement with Literacy," and the Australian Federal Government's compulsory income management policy, described as "income quarantining."

Example 1: "Parental Engagement With Literacy"
The HLA, "Parental Engagement With Literacy," is an outcome of the SA Government's uptake of "Health in All Policies" as a way to improve population health. As Newman (2011) describes, "The HiAP approach is based on [government departments in South Australia] working together to achieve win-win outcomes that enable both improved population health outcomes and the realisation of other sectors' goals" (p. 14). Targets are derived from the State Government's Strategic Plan (SASP), with the understanding that "the Senior Officers Groups will identify an appropriate policy focus for each Health Lens Analysis, approve project proposals and endorse the final project recommendations" (Government of South Australia, 2013, p. 5).
The Family Engagement With Literacy project involved a partnership between the SA Department of Education and Child Development (DECD) and the Department of Health and Ageing (DHA; Government of South Australia, 2013). It addressed the SASP Target 6.12: "By 2010, 93% of students in Year 3 to achieve the national benchmarks in reading, writing and numeracy" (Newman, 2011, p. 11). The aim of the project was to raise parental engagement with literacy to improve literacy outcomes for children in the early years of schooling, and ultimately improve their health, with a particular focus on low socio-economic families. (Government of South Australia, 2013, p. 8) The "program logic" behind the initiative rested on the premise that "increased literacy leads to improved health" (Baum et al., 2014, pp. i135-i136). "Theoretical causal pathways" link the "short-term HiAP objective of increasing parental engagement in literacy at home" to "improved literacy and, eventually, to improved health" (Baum et al., 2014, p. i136).
A WPR analysis begins with the recommendation or proposal, in this instance that of increasing "parental engagement with literacy." Based on this proposal, the "problem" is constituted to be both lack of parental engagement and lack of literacy. Subsequent questions in the WPR approach (see above) identify grounding assumptions within these problem representations, genealogies of the identified problem representations, and reflections on silences and effects, always with an eye to contestation and debate around the interventions and how they represent the "problem." Such questions produce the following critical reflections.
The expressed goal to increase parental engagement with their children's literacy, which is underpinned by psychological development theory (K. Osborne, Baum, & Brown, 2013, p. 11), rests upon assumptions about the role and responsibility of parents in producing "literate" citizens. According to Rose (2000), this form of "responsibilisation" operates at the "pole of morality," a point strengthened through the explicit targeting of "low socio-economic families": "It seeks to govern a polity through the micro-management of the selfsteering practices of its citizens" (p. 193).
"Literacy" meanwhile is a key indicator associated with contemporary governance objectives of securing the identity of citizens and the nation (Kelly, 2015), and developing a competitive and productive workforce (Salter, 2013, p. 12). Kelly (2015) shows how literacy forms part of a security apparatus, "a way of contingently positioning the capabilities of human subjects and populations in relation to the needs of forms of rule" (p. 178). Notably, "as a measurable and culturally normative activity," literacy possesses the capacity to exclude. Kelly's examples are remote Indigenous communities and prospective migrants.
Instead of accepting statistical correlations between "health" and "literacy" as drivers of policy formulation, therefore, a WPR analysis raises critical questions about "the frameworks of sense and obviousness with which policy is thought, talked and written about" (Ball, 2006, p. 44). "Parental engagement" and "literacy," along with the proliferation of new literacies, such as "financial literacy" (Bastian & Coveney, 2013, p. 165), "Asia literacy" (Salter, 2013), and "health literacy" (Green, Tones, Cross, & Woodall, 2015), and the theories upon which they rest, are treated as technologies of rule that require critical scrutiny.
Example 2: "Income quarantining" Compulsory income management, or "income quarantining," is an Australian Federal Government initiative designed to direct welfare income away from certain goods, such as cigarettes, alcohol, and gambling, deemed to be deleterious to the health and well-being of targeted groups. First applied in Indigenous communities, it has been broadened to encompass other categories of welfare recipient (Bletsas, 2012).
In a 2013 review of "evidence relating to 'what works' to influence the social and economic determinants of Indigenous health," K. Osborne et al. (2013) highlight the "difficulty in evaluating complex social policy such as income management" (p. 32). The review raises concerns about the policy, specifically from a social determinants perspective, arguing that compulsion is "in tension with the need to fully involve Indigenous Australians as equal partners in taking action to improve their wellbeing" (K. Osborne et al., 2013, p. 32) Other critical reports, including a HIA (Health Impact Assessment) of the Northern Territory Emergency Response (see Bacchi, 2009, pp. 116-120), of which compulsory income management was a part, are mentioned alongside findings of "positive consequences" and a "mixed reaction" among some Indigenous communities. The review concludes that "the evidence of the benefits and disadvantages of this approach may be clearer after the policy has been in place for a longer period of time" (K. Osborne et al., 2013, p. 32).
A WPR analysis brings different questions to "evaluation." The suggestion is that, rather than evaluating policies in terms of their abilities to "solve" problems, we need to study the ways policies constitute "problems." The proposition is that governing takes place through the formation of "problems," that is, through problematizations. In these terms, the proposal to quarantine the income of welfare recipients, Indigenous and otherwise, produces the "problem" as inappropriate use of income. As Bletsas (2012) notes, This policy reveals an individualistic understanding of poverty: It directs itself at altering the behaviour of individuals and the way they spend their income, not at the wider context in which such "individual" decisions are made. (pp. 39, 44) Moreover, targeting the behaviors of specific groups of "significant disadvantage" (Macklin, 2009in Bletsas, 2012) produces "poverty" as a residual condition restricted to "dysfunctional communities" who need to be "advanced in a developmental sense: So they can be caught up to the rest of 'us,' the affluent, western, mainstream" (Bletsas, 2012, p. 47).
As illustrated in this example, a WPR form of analysis extends beyond forms of critique that offer structural explanations of poverty to question how the individualistic way of constituting the "problem" in compulsory income management shapes "subjects" and lives. It also prompts investigation into the genesis of that particular representation of the "problem," challenging its taken-for-granted status as a mode of governing, and questioning its normalizing and integrationist effects. In this form of analysis, problem representations are treated as political interventions that need to be contested at the level of what they produce.

Conclusions: The "Project Trap" and Deep Evaluation
The article advances a poststructuralist mode of theorizing, which, it suggests, opens up a new set of questions for health policy researchers. In the place of tracking policies through "stages" of development, it recommends critical analysis of the categories and knowledges that shape current governing practices. To instigate this form of investigation, it promotes the study of problematizations-how "problems" are constituted within policies and policy proposals.
Doubtless, this form of analysis faces all sorts of challenges, particularly in research settings where, often, "problems" are set by those in positions of authority. In the article examples of this practice appear in Pawson et al.'s (2005) "realist review" where problems are set by those commissioning the research, in Ritchey's (2011) GMA where the "problem area" is set by the "principal client" (pp. 64-67), and in South Australia's HiAP program (Government of South Australia, 2013) where Senior Officers Groups identify an "appropriate policy focus" (p. 5).
We have here versions of what March, Smyth, and Mukhopadhyay (1999) call the "project trap," where analytic frameworks (e.g., gender analysis, SDH) remain "narrowly applicable to programmes and projects" (p. 49), subservient to wider policy objectives. Pat Armstrong (2002), for example, describes the limitations imposed on gender analysis by an inability to question the embrace of privatization by the Canadian government and how this mode of governing shapes health "problems." Elsewhere, with Joan Eveline (Bacchi & Eveline, 2010a, p. 30), I introduce the idea of "deep evaluation" as an intervention to challenge the limitations imposed by the ex post character of forms of policy analysis such as Emancipation Impact Assessments (Bacchi & Eveline, 2010b, p. 52) and HIAs, and by "inside government" positioning (Delany et al., 2014, p. 8;Staudt, 2003). It includes as critical foci: (a) the meanings attached to key concepts (e.g., "equity"); (b) how the "problem" is represented (i.e., applying the WPR approach); (c) how "contexts" are represented.
For researchers, an even greater challenge involves the need to subject their own proposals to a WPR form of analysis (see above). As forecast in the introductory comments, self-problematization is critical because, as with government policy, our classificatory selections actively constitute social reality (Bletsas, 2012). In this view, contra Bambra et al. (2011, p. 405), who suggest that researchers focus "less on describing the problem and more on ways to solve it," researchers are called upon to contest the current emphasis on "problem-solving," and to subject their own problematizations to the kind of critical scrutiny a WPR analysis recommends. The task becomes considering the extent to which recommended policy proposals, including one's own proposals, either reproduce or disrupt modes of governing that install forms of marginalization and domination.