Independence and impact: a typology of researcher-user relationships for policing and crime prevention

ABSTRACT The transition from ‘Mode 1’ to ‘Mode 2’ knowledge-production has created an explicit focus on research impact which is reflected in the funding and organisation of research, the relationships between research and users, and the focus of research studies. It has also led to efforts to understand pathways to impact, although these studies have so far had little to say specifically about crime-related research. This paper comprises an effort to address this gap and explore variations in research-user relationships within research into policing and crime prevention. An ESRC-funded project researching the crime consequences of the Covid-19 pandemic, of which we were team members, faced several challenges in terms of data access, data collection and the need to deliver strong academic research outputs, as well as maintaining a clear focus on real world research impacts. This led to a co-productive relationship between the researchers and a range of external partners. There is a dearth of literature theorising the nature of such researcher-user relationships, despite some rich accounts of individual experiences. This paper uses examples drawn from the writings of those who have been involved in research impacts (or who have been thwarted in their attempts), to set alongside the project on which we have been working. From this we present a typology of researcher-user relationships, that we hope will further theoretical discussion in the field and might usefully be applied more broadly to other areas of criminological interest.

life or adaptations to routine activities made to Covid-19 by individuals or by organisations. Finally, it tried to anticipate possible longer-term changes in crime patterns brought about because of breaks in previously routinely reproduced criminogenic or counter-criminogenic conditions. Involvement in this project led us to consider our project's approach and experience in relation to differing researcher-user relationships more generally.
Despite a wealth of literature about the 'impact agenda', and pathways to impact within other disciplines, there has been surprising little theoretical attention to understanding impact within crime-related fields of research. Based on a review of case studies within policing and crime prevention, we present a typology of researcher-user relationships as a starting point to understand the nature and context of criminological research impacts.

Background: the impact agenda
There has been increased emphasis in recent years on the broader social, cultural, environmental, and economic benefits accruing from research. This 'impact agenda' is reflected in the UK in the Research Excellence Framework (REF) introduced by HEFCE in 2009 and implemented in both the REF 2014 and 2021. Similar initiatives have been applied on a global scale, acknowledging the widespread transition within research from Mode 1 to Mode 2 knowledge. In writing about the relationship between science and society, Gibbons (1994) first introduced the distinction between Mode 1 and Mode 2 knowledge-production and use. Traditionally, Mode 1 research was undertaken with no impact in mind, although eventually it might be picked up and used by interested third parties. In Mode 1 research, therefore, there is little interaction between researchers and end-users of findings. More recently Mode 2 has increased in importance. Its orientation is essentially applied. Many 'experts' (both academic and non-academic) work together to 'co-produce' knowledge often to provide solutions to mutually identified problems (Greenhalgh 2016a). Mode 2 involves complex, multifaceted relationships between researchers and research-users wherein real-world impact is a central aim of the research endeavour. This is not to imply that Mode 2 research necessarily involves abandoning scientific principles or methodological rigour, but rather that Mode 2 research is orientated towards greater application, social distribution, and greater accountability (Nowotny et al. 2003). Nonetheless, concerns were soon raised about the potential for research to become predominantly funded and controlled by the research-user. Ziman (2000) feared that research objectivity and scientific rigour within the physical sciences would be compromised by a more applied focus permitting greater involvement of external sources of influence. There has also been considerable debate about Mode 2 research within the social sciences and in particular limitations of 'impact culture', as supported by the REF (Research Excellence Framework). For example, the restrictive assumptions of the REF may lead researchers to feel obliged to prioritise easily measurable short-term impacts, at the expense of long-term impacts made in more complex and diffuse ways (Greenhalgh 2016b). Also noted are the restrictions Mode 2 research can place on researchers' critical agendas. Research findings that run contrary to policymakers' priorities and interests can easily be ignored, and their impact minimised (Colley 2014). Some researchers feel they must temper their criticism of current policies and practice for policymakers and practitioners to be receptive to their research (Laing et al. 2018).
Crime-related researchers have been concerned about potential conflicts with their role as independent and critical social commentators (Crawford 2020a). Some have argued thus that they should avoid being too beholden to the differing interests and objectives of criminal justice agencies and strive to remain free from government and political pressures (McAra 2017). At the same time criminologists are often drawn to the discipline by a desire to make a difference in terms of tangible social change (Rock 2014). This involves finding a way practically to help reduce crime harms, whilst retaining an independent critical voice and avoiding collusion with state institutions and authorities. 1 The degree to which criminologists co-work with the police and policymakers is therefore contentious within the discipline.
Those who have worked closely with the Home Office, with a direct policy-focus to their research, have been met with considerable animosity (Laycock 2011, Hough 2014, Mayhew 2016. Often dismissed as 'administrative criminology' (or 'official criminology') their work has been deemed unquestioning of policy and practice (Young 1992, Hope 2011. Whilst the move to Mode 2 knowledge production has increased researchers' needs to be impactful, there are debates about not only how this can be more effectively achieved, but also how to do so without compromising the researcher's own ideals. McAra (2017) raised concerns about how criminologists are increasingly becoming 'problem-solvers' commissioned by government and criminal justice agencies to research to their agendas. Others have argued in favour of the benefits of co-production (Crawford 2020a, Davies 2021, whilst also recognising that in practice researcher-user relationships can be inherently contested due to their often-disparate ideologies, agendas, priorities, and pressures (Crawford 2017) and may be marked by tensions to do with power differences and competing professional interests (Crawford 2020b).

Understanding impact
There are diverse pathways to crime-related impact, differing conditions that can facilitate or hinder impact, and a variety of roles played by researchers in relation to potential research-users. The issue, however, remains undertheorized (Williams 2020) which makes it difficult for crime-related researchers to know how to optimise the potential impact of their research.
Within other disciplines (most notably health research) there is a growing body of literature on potential pathways and conditions for impact (e.g. Greenhalgh 2016a and b). Muhonen et al. (2020) analysed 60 research case studies from across the Humanities and Social Sciences to develop a typology of 12 pathways to impact. Although not discussed in any explicit detail, within these pathways a range of different relationships between researchers and research-users are evident. Spaagen and van Drooge (2011) highlighted the importance of the quality (as opposed to just the frequency or regularity) of researcher-user interactions. They argued that 'productive interactions' were those that embraced a Mode 2 approach to user-engagement and involvement, and that these interactions facilitated implementation of research into policy and practice. Other models identify different core types of research stakeholders (Mostert 2010), potential impact mechanisms (Schneider 2019), and barriers to impact creation (Molnar 2016).
This cross-disciplinary work touches upon areas of interest to crime-related research, in particular the types of engagement with various types of research-users. The route to impact in both health and crime-related research can be long and complex, and change can often only be implemented at ground level if leverage can be found higher up. This is well discussed within crime prevention in the work of Sampson et al. (2010) but has not yet been applied specifically to research on impact. These researcher-user relationships seem to be a significant contextual feature of impact creation, but the nature of these relationships, and the conditions and mechanisms for impact remain unexplored.
The more extensive research within health studies provides a useful methodological and analytical framework. Understanding impact requires detailed accounts focused on processes and outcomes of the research. As outcomes can be diffuse and long-term, researchers recommend using a purposive sample of retrospective case studies (Palinkas 2015). The rich qualitative descriptions they provide are useful in trying to unpack specific impact processes, and in highlighting the importance of context in shaping patterns of possibility (Bell et al. 2011). Such narrative analysis is favoured over more systematic and quantitative meta-analyses (Greenhalgh, Thorne, and Malterud 2018).
Researchers have also tended to adopt a realist view of impact (Pawson 2005) as it sensitises us to underlying causal mechanisms for impact that are usually unobservable (Coombs and Meijer 2021). The realist approach has been widely adopted within health impact research (Greenhalgh 2016b) and its benefits are widely acknowledged (Molnar 2016). We follow the better developed literature on health impact research in using a realist framework, understood in broadly Pawsonian terms, to examine researcher-user relationships as a contextual condition for impact, drawing mostly on participant accounts.
A major previous attempt to theorise criminologists' often difficult relationship with policy and policymakers was Loader and Sparks (2010) discussion of 'public criminology', identifying various 'types' of criminologist by their 'styles of criminological engagement.' They categorised criminologists into one of five different roles -(i) the scientific expert, (ii) the policy advisor, (iii) the observer-turned-player, (iv) the social movement theorist, and (v) the lonely prophet. Critics have, however, argued that their stance is elitist and insular (Christie 2011, Groombridge 2011, and critical of those researchers who work with the government and stakeholders (Morgan 2011). Furthermore, the typology served to pigeon-hole researchers into hard separate camps, failing to recognise the nuances of those relationships, and the tendency of researchers to adopt different roles in different research contexts throughout the course of their career (Currie 2011).
Rather than attempting to cover the whole gamut of crime-related research, we focus on policing and crime prevention, an area in which we have direct experience. The literature on academic-police and crime prevention partnerships has grown in the past few decades. This is partly due to the influence of situational crime prevention, problem-oriented policing (POP), and evidence-based policing (Knutsson and Tompson 2017). In addition, the establishment of the College of Policing in the UK in 2012, has foregrounded the role of ongoing research-based learning in the development and professionalisation of the police.
Such developments may represent a shift from researchers being critical of, to working with, the police (Bradley and Nixon 2009). This does not mean that these relationships are without their difficulties (Crawford 2020). They can be highly contested due to differing organisational cultures (Stanko 2007), and a fluid socio-political landscape (Colley 2014, McAra 2017. Moreover, these relationships vary. Rojek et al. (2012) distinguish between researcher-police co-operation, co-ordination, and collaboration, whilst Guerette et al. (2019) note variations in degrees of formality.
Although there is an emerging literature giving accounts of relationships between crime research and potential users, there has been no systematic empirical investigation beyond individual research projects or programmes. This paper draws together what can be learned from the examples we have to date and attempts to formulate a theoretical framework for future research. As a well-known and frequently utilised distinction within the literature, we draw on the Mode 1 and Mode 2 knowledge production as a framework for analysing the full spectrum of 'pure' to applied research.
In the next section we explore variations in research-user relationships within crime-related research as possible key contextual differences for the types of mechanism activated and forms of impact produced. Table 1 presents a typology of researcher-user interfaces along a spectrum broadly going from Mode 1 to Mode 2 science. We have used a spectrum of Mode 1 to Mode 2 knowledge production as a starting point as this is well recognised within the broader literature on 'impact culture' within the social sciences and has been used by other criminologists to reflect upon the shifting climate of research (see, for example, Crawford, 2020). Using this broad distinction, we adopted a realist approach to attempt to 'unpack' contextual conditions that may both facilitate and impede the

Research-user relationships as contexts for impact
activation of mechanisms for generating research impacts, focusing specifically on the relationships between the researcher and research users. We draw on a sample of case studies, following other work of impact in different research and policy domains. It needs to be borne in mind that such case study accounts may be produced by interested parties, aiming to celebrate or castigate particular kinds of work (see, for example, Clarke 2020 andWalters 2008 respectively). This necessarily introduces some biases as we can only sample cases from those published and therefore publicly available, and these may not reflect more typical experiences. It may be that unusually negative or positive experience drove some authors' discussions. Highly reflective accounts of co-working are also still quite neglected in academic literature (see for example, Lumsden 2017). Table 1 distinguishes 'ideal types' 2 of researcher-user relationships, under four main categories. At the left end lies pure 'Mode 1' research, where curiosity-driven research is not orientated to impact, results are reported in learned journals targeting other researchers, and producers have little or no direct interest in impact. At the other end of the spectrum lies research fully in Mode 2. It describes commissioned studies that are controlled by the research-user, orientated exclusively to satisfying their interests. In contrast to Loader and Sparks (2010) our analysis is at the level of a specific research project or programme, rather than at the level of a specific researcher, as most researchers will participate in multiple types of researcher-user relationships depending on context and opportunity. In the sections to follow we will go through the typology drawing on brief case examples from the literature from within policing and crime prevention, whilst also illustrating some variability within types using further case literature.
We begin with 'independent' research, under which we include two subtypes: 'scientist' and 'critic'. We then go on to 'collaborator' followed by 'in house' and 'contract' alongside their subtypes. Our recent work within the crime harms of Covid-19 project fell largely within the 'collaborator' category.

Independent (a) Scientist
The archetypal independent scientist conducts the purest form of Mode 1 knowledge-production. 'Blue skies' research with little direct thought for societal impact is undertaken, orientated exclusively to enhancing knowledge within a field. The researcher maintains a high level of independence, setting the direction of the research and the specific research questions, with funding typically provided by an external research council or the researcher's affiliated Higher Education Institution (HEI). Findings are reported in academic books and journal articles. This research may have potential applications, but they only emerge later and through the work of other researchers and end-users. The pathway to impact can therefore be uncertain, long and diffused. We should stress that independence is not the same as objectivity. Many scientists, independent or otherwise, may have their own biases! There are numerous examples of independent scientist crime-related research, and more specifically within policing and crime prevention. For example, patterns of repeat victimisation (RV) were initially identified by Sparks and colleagues. Findings from a victimisation survey were reported in a book and peer-reviewed journal article (Sparks et al. 1977, Sparks 1981, but received little attention outside of academia. The potential practical importance of RV was later recognised and applied in relation to domestic burglary (e.g. Forrester 1990, Farrell andPease 1993). Even then it failed to make much impact on policing strategy and practice until attention to it was built into a policing performance indicator in 1999 (Farrell 2000). Despite the volume of evidence finding that strategies based on repeat victimisation do significantly reduce crime (Pease 1998) across a diverse range of offence types including domestic abuse (Kennedy 2020), it has had surprisingly little long-term impact on policing practice (Kennedy 2015). Furthermore, the impact was only possible when Forrester et al.'s research, conducted under the auspices of an in-house, government-funded research unit, translated Mode 1 knowledge of RV into practical preventive action, and the government then incentivised the police to put it to use. Impact deriving from Mode 1 research findings is contingent on recognition of their practical significance and their demonstrated application to solve problems. The potential research-user may not be well placed to assess the quality and rigour of research, fail to recognise potential applications or benefits, or even do more harm than good if they misunderstand or misapply the research.
Of course, much Mode 1 work will have no impact. It will be ephemeral, flawed, quickly forgotten, or overtaken by better studies or only of interest to other members of the research community.

(b) Critic
The role of the critic is also relatively independent from external influences in that funding is typically from research councils or the researcher's HEI, and the research is generally defined and directed by the researcher with some critical agenda in mind. The aim is typically to analyse flaws in a specific institution, area of policy or law, to raise awareness of injustices, inequalities, and poor practices. It might not be as independent as the scientist in that some funding and input might come from interested external parties such as NGOs, charities, or activist groups. Some key examples can be found in the earlier critical work on policing in the 1970s and 1980s (e.g. Hall 1978, Lea andYoung 1984), as well as research highlighting racism and sexism within crime control and policing (Gordon 1984, Stanko 1985. Of primary importance to the researcher is the ability to maintain a critical voice, and be free from external influences, often with an emancipatory agenda. Dissemination tends to follow Mode 1 research in that outputs are usually peer-reviewed journal articles and books, aimed at a scholarly readership. Critical research can of course have significant impact, but that impact is uncertain, and the pathway can be long and complex. Critical research must often be both timely and newsworthy (Murray 2017) to hit a 'window of opportunity' when there is institutional or public dissatisfaction with the status quo (Tonry 2010). Whilst there are strategies that can be used to try and maximise impact such as 'horizon scanning' of emerging political trends, or co-working with lobbyists and media, impact can be quite happenstance, or be slow and incremental (with the research 'paving the way' for some future revival of interest).
As an example, Matthews (2008) book Prostitution, Politics and Policy provided a comprehensive summary and critique of the regulation and policing of sex workers, particularly in the UK. It was concerned with the increased criminalisation of prostitution and critical of the current policy. Published at a time when the issue of prostitution was high on the social and political agenda this scholarly text proved influential with the UK government. In 2014 Matthews went on to advise the government in their All-Party Parliamentary Group on Prostitution and the Global Sex Trade and contributed to the subsequent Home Office publication Shifting the Burden: Inquiry to Assess the Operation of the Current Legal Settlement on Prostitution in England and Wales which debated various ways forward in terms of policy.
Working with those at the heart of the problem may, however, sometimes require too much compromise for the critic, especially when research users are not (yet) receptive to critical advice they do not want to hear (Colley 2014, McAra 2017.

Collaborator
The second type of researcher-user relationship is collaborative. Bradley and Nixon (2009) distinguish between researchers who are critical of the police, and those who work with them. Recent discussion of impact creation notes a further shift in police-academic relationships towards the co-production of knowledge and shared crime reduction initiatives (Crawford 2017, Bacon et al. 2021). We propose a more nuanced set of distinctions in police-researcher relationships and argue that whilst there may be shifts over time, each furnishes a distinctive context, facilitating greater (or lesser) impact through differing mechanisms.
Collaborative research embraces Mode 2 knowledge production, with its explicit concern with impact. There have been various reflexive discussions of it including its advantages and disadvantages, which we draw upon below.
(a) Problem-solver This is police-led, in the sense that the research begins with a police-identified problem. The researcher is brought in as an expert consultant by the police to help solve specific crime and / or operational problems. This has become a more common form of researcher-user relationship with the increasing influence of POP; developed by Goldstein in 1979 to tackle entrenched patterns of crime (and other issues within the police remit) by adopting an analytic approach.
The SARA (Scanning, Analysis, Response, Assessment) process, now widely used to implement POP Spelman 1987, Sidebottom andTilley 2011) puts research at the heart of policing. Researchers are sometimes drafted in to help with specific problems. Goldstein and Clarke have written about examples in which they have participated whereby they had a great deal of autonomy, given their reputation and seniority. For example, work with police in North Carolina to problemsolve thefts from vehicles parked in city centre car parks, and thefts of appliances from construction sites Goldstein 2003a and2003b). Clarke and Goldstein helped find pinch-points for the problems and the police were persuaded to try them, but getting the promising measures enacted involved further steps. The cases highlight difficulties getting 'buy-in' to an intervention from nonpolice partners, the need for long term commitments to such initiatives, and the huge personal investment needed by police officers to continue the work. Braga et al. (2001) were similarly involved in an innovative partnership with police, and other stakeholders in Boston to problem-solve a growth in youth homicides. While the crime problem itself was identified by the police, the researchers acted as experts to rethink the strategy and suggested a deterrence approach (Operation Ceasefire) whereby highly publicised strict crackdowns on gang violence by the most chronic offenders would be used to relay the serious consequences to other local gangs and gang members, and lead to behaviour change. After implementation there was a large reduction in youth homicides, gun crime and gang violence. This example, however, also highlights the difficulty in effecting long-term change, as the Ceasefire intervention was discontinued in 2000, and by 2006 youth violence in Boston was on the rise again leading to fresh consultation and problem-solving (Braga 2020). Previously successful crime reduction effects therefore declined as impetus was lost, personnel changes were made, and resources were directed elsewhere once the problem was deemed 'solved'.
Although the researcher might have considerable say on the analysis of the problem, and the direction of the implementation, this is not equal co-production. The researcher has no power over implementation. Whether and how that happens may depend on many factors outside their control, such as the resources, politics, and priorities of the police and other stakeholders. In addition, proposed solutions from researchers may not always be well received, and without acceptance from the police and other parties, impact is hindered. This is illustrated in work related to the example above, where the researchers advocated an Operation Ceasefire approach to reduce cases of domestic violence (Kennedy 2020). However, the suggestion was rejected by local agencies and found no traction in the US for over a decade. 3 (b) Co-producer This is neither researcher-nor police-led but is a joint enterprise. Co-production assumes an equal and reciprocal relationship between researchers and research-users, so that the shaping of the research agenda is based upon shared understandings and all parties assume ownership of the research (Crawford 2017). Co-production varies considerably in terms of the number and nature of involved partners, and the engagement in co-production at all stages of research and application. There is a diverse range of possible funding arrangements, but most co-productive partnerships are, at least initially, supported by some research council funding. For example, the Scottish Institute for Policing Research (SIPR), a co-led academic and police research network is funded by ACPOS and the Scottish Research Funding Council (Fyfe and Richardson 2018). The Australian Research Council has also provided linkage grants to enable co-productive partnerships between universities and industry partners (Bradley and Nixon 2009).
Co-productive research partnerships offer a Mode 2 compromise position for researchers striving to make real-world impacts while maintaining some academic independence (see Davies 2021). Despite this, the literature on researchers' experiences of co-production is still quite limited (Lumsden 2017). Co-production ideally comprises a forum in which 'problems' can be both raised and solved (McAra 2017), and in which the researcher can challenge existing practices (Crawford 2020). In practice such a relationship can sometimes be marked by some conflict, given the different perspectives, interests and cultures of researchers and users. Innes et al. (2019) describe the co-productive relationship as 'messy' and 'tarnished', as compromises must be made throughout the research process, but note that this is a necessary requirement and should not be viewed as an insurmountable problem.
The crime harms Covid-19 project is a good illustration of co-production, although it also incorporated aspects of partnership as noted below. It required co-operation with user-organisations both in the conduct of the research and for the achievement of impact. It was framed and conducted firmly with a Mode 2 orientation. Hence it was proposed, funded, conducted and outputs reported with an orientation to impact, although several papers have been published in academic journals. 4 Aspirations to achieve impact were expressed in the use of an advisory board with members from the police, College of Policing, and the Home Office, who helped focus the work on priorities; and in the more informal contacts with service delivery organisations such as Neighbourhood Watch, and several individual police services. Contributions were made to two series of briefing papers, the first one of which was set up specifically to ensure timely reporting of findings to user organisations. It comprised brief statistical bulletins tracking emerging trends in specific crimes. The other series was set up independently to the project but complemented its objectives and several contributions were made to it. This series was a theoretically informed (albeit speculative) effort to identify expected crime problems that would arise from Covid-19 and what could be done to try to pre-empt them. At the time of writing there are 15 papers in the first series and 28 in the second. Many members of both series have been further disseminated through Police Insight.
It is too early realistically to gauge practical impacts of the Covid 19 project on policy or practice. Our funding and the relationships forged with service delivery organisations enabled us to conduct research that has Mode 1 attributes (i.e. the scope to define research parameters, production of research of significant academic interest, and adherence to academic standards of quality control). In addition, efforts at framing the research in ways orientated to meeting user-organisation's needs incorporated a Mode 2 orientation, and some of the work has been disseminated in Mode 2 focused ways to try to maximise the chances that it can inform decision-making. Control was not ceded to third parties, and members of the research team have not been inhibited in publication due to dependencies on potential user-organisations.
The most widely documented case of co-production is that of the N8 Policing Research Partnership (N8 PRP) funded from 2013 by HEFCE, involving 8 universities and 11 police forces in the North of England. N8 PRP aimed to harness the 'collaborative advantage' whereby sharing views, methods, evidence, resources, and experiences would facilitate knowledge production that could not be achieved by any single organisation acting alone (Crawford 2017). Reaching a consensus from such a varied group of researchers and research users, often from diverse organisational bases, is inherently difficult, and conflicts are therefore commonplace.
Conceptually, co-productive partnerships aspire to equal power between partners in terms of the way research is framed, evidence-collection, forms of analysis, implemented solutions, and outcome evaluation. But in practice, one partner may still dominate at any or all stages, and certain ideas may find greater purchase among research-users if they are consistent with their pre-existing views and procedures. This is a common experience in any form of researcher-police partnership as the police are a powerful (as well as accountable) organisation with a tendency to downplay dissenting viewpoints (Crawford 2017(Crawford , 2020a(Crawford , 2020b. Researchers have expressed concerns that the findings and solutions may necessarily be tempered if they are to be accepted and applied by research users, limiting their critical role (McAra 2017, Fyfe and Richardson 2018). Crawford (2017Crawford ( , 2020aCrawford ( , 2020b, however, maintains that conflicts can be managed, and criticality retained to achieve 'independent interdependence' whereby a researcher can effectively maintain their critical distance and autonomy whilst working with police partners, to the benefit of both. Greater impact is possible when all partners have been involved throughout the process and have shared ownership of the research due to the time and personal investment put into the enterprise. Research council funding, and the independence researchers are often able thereby to maintain, tends to lead to published research outputs and scholarly dissemination of the work. Bacon et al. (2021), however, question the sustainability of any research partnership funded through external grants with a limited time frame. But such partnerships may be sustainable after initial start-up funding if members work together to bid for more money.

(c) Partner
We come now to researcher-led partnerships. In this the researcher generally sets the direction for the research, looking for police with whom to pursue it. The relationship is not co-productive, but there are likely to be aspects of co-production and compromise as the researcher will be reliant upon the police and other research-users for co-operation and access to data. This was certainly the case with the Covid-19 project. All parties are impact-driven and therefore have an interest in co-operating in research that is not only academically robust and original but also useful for improving police policy and practice. This type is more likely when the research is externally and independently funded through a research council, but good pre-existing relationships between partners are generally essential in bringing it off. Cockbain (2015) worked with six UK police forces from 2009 to 2014 as part of her PhD research into the sex trafficking of British children. Her university department had strong existing contacts within the police which she acknowledges were beneficial in negotiating access to highly sensitive police data. Whilst wanting to produce intellectually significant and academically robust research, she also tailored her research to the police's unmet research needs, so she was able to maintain her researcher independence whilst negotiating access. The work resulted in her PhD and scholarly publications, but also proved particularly impactful. For example, it was cited in government reports, led to consultations with SOCA and the NSPCC, and fed into professional training. Cockbain attributes the success of this partnership to the forward-thinking of her police partners and her ability to take advantage of an issue that was particularly relevant to the police at the time, but underresearched.
Not all such partnerships are successful, nor are the research findings always so well received. Murray (2017) writes about her experiences working for the Scottish Centre for Crime and Justice Research, whilst conducting her PhD on Stop and Search rates in Scotland. The research produced troubling findings for the Scottish Police and Scottish Government, which they subsequently sought to suppress, but were unable to as the PhD was funded independently by the ESRC. Nevertheless, the report led to major legal reforms after the published findings were picked up by the media. The price paid for such critical independence may be that the researcher does not find a receptive audience in requests for future research co-operation.
Academic freedom in police-academic partnerships is valuable in enabling the labels, definitions, and practices of the powerful to be challenged (Lumsden 2017). Even as an independently funded researcher, however, the reliance on maintaining a good working relationship with partners, and the emphasis placed upon this by universities, in some cases can be felt to significantly undermine a researcher's critical role (Lumsden and Goode 2018).

In-house
The next types of relationship represent more formal arrangements of the researcher as someone employed within the research-user setting, both at the level of a criminal justice agency such as the police, or within a policy-making environment. Here we see a greater shift towards applied research, and scope for greater control by government organisations or police bodies. Although this does not necessarily mean a loss of all independence and criticality on the part of the researcher, as many critics of in-house working have suggested (Young 1992, Hope 2011. In fact, within these types of relationships we again see great scope for variability and a more complex and nuanced set of relationships than previous literature has implied.

(a) Research unit
The Home Office Research Unit began in 1960. Situational crime prevention (SCP), which emerged in the Unit, has had a substantial impact on policy. SCP was rooted in critical research reviews of established crime control practice, which were undertaken and published within the Unit (for example, Clarke and Hough 1984). SCP proposed a radically different understanding of crime and approaches to its prevention as it treated offending as a function of opportunity rather than as an aberrant product of social injustices or psychopathology (Mayhew 1976). It thus marked a major shift from conventional ways of understanding and controlling crime despite having its genesis within a governmental research unit.
SCP has been impactive on crime prevention policy and practice in the UK. It has informed a range of government programmes, including Safer Cities, The Crime Reduction Programme, and at the time of writing Safer Streets. What the Research Unit at the Home Office shows is that there are circumstances in which in-house research units can be sites for original research and impact. It had considerable control over the direction of its research, its recommendations, and dissemination of its outputs (Hough 2014, Mayhew 2016. Its impact was, of course facilitated by the fact that it operated within the Home Office and was orientated to helping achieve its objectives. Such critical pioneering research within an in-house unit is also relatively rare. The rise of evidence-based policing and the professionalisation of the police have also led to individual police forces employing their own in-house research teams. Dawson and Williams (2009) write about their experiences working within a civilian research unit within the Metropolitan Police. The Metropolitan police set the research priorities for the research unit, and the unit also provided a conduit for external researchers to gain access and collaborate. In addition, the researchers were responsible for the dissemination of research to police officers to better inform their operational strategies and practice. Their reflection on the experience is largely positive, although greater critical reflexivity is perhaps unlikely whilst holding an employed position within the police. They do note the conflicts which arise with other frontline and policy staff, and how these relationships inevitably affect whether the research is ever acted upon.

(b) 'Pracademic'
As well as civilian researchers, recent years have seen the growth of the 'pracademic' or practitioner-turned-academic who conducts research as part of their practitioner role, often balancing frontline work with running their own small research projects (Braga 2016). These are often highly skilled practitioners from within an organisation, who have some level of academic training, and who have been inspired by their own experiences on the ground to change things from within. They may go on to complete a doctorate or eventually join academia full time. Sherman (2022) gives an account of this and refers to the many academically reputable publications that have been produced from pracademics associated with Cambridge University. Holgersson (2015) provides a detailed account of his own experiences within the Swedish Police. Whilst he felt well-placed and well-informed to identify and conduct research into areas of unmet needs, he frequently experienced obstruction and harassment from other police officers. His position, however, also meant that his views at times were more accepted because (unlike external researchers) he did not have to deal with setbacks related to organisational culture and hostility towards academics (Bradley and Nixon 2009). Holgersson notes his position left him vulnerable when his research agenda did not match that of his employers, and when his findings were seen to be critical of their practices. Wanting to improve policing, but constrained by organisational politics and culture, he sought media publicity to raise awareness and foster public demand for police change.
As researchers who work regularly with police and other agencies we know of many examples of 'pracademics' but whilst still employed within those organisations, candid reflective accounts such as Holgersson's are scarce. Pracademics tend to have a low to moderate level of independence, research problems are likely to be a mixture of organisationally defined priorities and those identified by the researcher based on their own operational experience. The direction research takes is likely to be constrained by the service's interests and resources to support it. As insiders, pracademics will have invaluable insights into problems and potential solutions and will have far greater research access than any external researcher (Braga 2016). But they will usually be heavily reliant on the organisation endorsing their findings and supporting recommended changes.
(c) 'Embedded researcher' The final form of in-house researcher is the 'embedded researcher' or 'embedded criminologist', the latter term coined by Petersilia (2008) in describing her own role as a special advisor to the California Department of Corrections and Rehabilitation. Although not focused on policing and crime prevention, Petersilia (2008) provides a rich reflective account of her experiences including the changes she helped to bring about, and the constraints and challenges she faced. She highlighted how working within the system at senior level provided opportunities to influence policy and practice, in her case helping to shift the emphasis of the department towards the community reintegration of offenders. But she also details the limits of that influence, and the importance of timing, finding the right 'windows of opportunity' when policymakers and legislatures were more open to change in that direction (Tonry 2010). Petersilia's (2008) account strikes a chord with the experiences of Betsy Stanko who after a distinguished academic career raising awareness of female victims of violence, transitioned to a role as special advisor to the Metropolitan Police. Stanko (2007) describes the varied roles she played in her career and her desire to draw on her research experience to make a more direct impact. She notes the mismatch she often saw between evidence and practice, and the lack of in-depth understanding of violence among policymakers. Like Petersilia she was met with some hostility from her academic peers who cautioned that she had 'gone to the dark side,' sacrificing her objectivity and independence (Stanko 2007). Both of their accounts highlight the need to make compromises but feel that these are necessary to make a meaningful impact, which could not be made on the 'outside'.
The embedded researcher tends to exert considerably more direct influence than the other two types of in-house role as their appointment is based on their seniority and reputation, and they are already trusted within the organisation. They have a key advisory role with greater scope to define the nature of the problem and bring research findings to the table. This gives them a distinctive standing, enabling them to make considerable impact (Braga and Davis 2014), but they are often no longer the primary producers of research and subsequently they may act more as gatekeepers and advisors. Guerette et al. (2019) argue that unlike pracademics embedded researchers can retain their standing within academia, enabling much greater objectivity and independence. But as they are employed by the organisation, there are still limits on this, and they face many political and organisational constraints (Petersilia 2008).
Despite the rigour and soundness of their evidence-based advice, embedded researcher's work will not always be well received or acted upon, and they have no power to implement change themselves. Whilst they will have much better access to senior decision-makers, any research outputs will need to be negotiated with the organisation, and it is likely they will sign non-disclosure agreements severely limiting any critical commentary on their experiences. Knutsson (2015) provides an account of his experiences working within the Swedish Police. He was met with opposition, obstruction, and a reluctance to publish his findings. He discusses the ethical dilemmas embedded researchers may face when the loyalties expected by the host organisation conflict with the researcher's own moral standards. The increased opportunity to reach senior decisionmakers from an internal position necessarily brings limitations in terms of criticality and freedom to disseminate findings and experiences. Whilst impact can be made, given the right timing, and framing of research, findings and subsequent discussions about their application may remain private.

Contract research
Contract research represents the most 'Mode 2' form of research. Here, organisations commission external researchers to conduct specified pieces of work. Such contract research is becoming commonplace within policing and crime prevention. Contracted researchers may be academics (such as those within a university-based research unit), or private consultants.

(a) Commissioned
Both policymakers and police have a history of commissioning research to undertake discrete pieces of research, systematic reviews, and evaluations. The decline of the Home Office RDS directorate and closure of the Police Research Group saw a shift to commissioned research that may lack the critical agenda of the earlier in-house research units (Rock 2014, Mayhew 2016. The research produced may better match the research-user's interests and be framed in a way most useful to policymakers and practitioners (Hough 2014). But commissioned research can give greater power to research-users who can control not only the definition and direction of the research, but also whether the findings are publicly released and acted upon. Less room is left for the researcher's critical voice, and impact may be more to do with fine-tuning policy than rethinking and challenging the status quo (Morgan and Hough 2007). As researchers are often reliant on securing future commissions from the same user-organisations, they may be reluctant to openly criticise them. The reflective accounts and case studies that do exist tend therefore to be from academics who have faced particularly negative experiences. Brown (2015) reflects on a series of four evaluations of organised crime policing commissioned by the UK Home Office. Whilst he recognises the advantages of the role, in particular the access to data that an outsider would not be able to obtain, he also faced obstacles, for example, attempts of senior officers to manipulate and obstruct the research, and ultimately failure to accept and publish the evaluation findings. When the research is user-funded the organisation holds the intellectual copyright and therefore controls publication. Hope (2008) similarly reports findings of his evaluation into a burglary reduction initiative being suppressed by the Home Office.
Commissioned research usually allows the researcher little independence. Critics of government research contend that such research can potentially be manipulated for political gain (Hope 2008(Hope , 2011, and by some is seen to legitimise the criminalisation and marginalisation of vulnerable sectors of society (Walters 2008). The user-organisations, whether government or a service delivery agency such as the police, can have considerable power over the researcher, making it difficult for them to speak out about worrying aspects of the research or be transparent about any unintended negative consequences of the policies or practices being examined (Lumsden 2017).

(b) Commercial
Private or commercial consultant contracts with research-users represent the purest form of Mode 2 research production. The private consultant will tend to have no critical interest in the research-user's practices, and the relationship is often one of 'commissioner' and 'client'. This is increasingly the type of relationship with Home Office research as policy staff 'contract out' research and evaluation projects to private sector consultants (Mayhew 2016), such as special advisors, business consultants, and market research firms (Rock 2014). In some cases, the government may out-source the labour for the administration and data collection, such as the British Crime Survey since 2001 and the Commercial Victimisation Survey since 2015, leaving in-house policy staff to analyse, interpret and publish the findings, although the data are made available to third-party researchers for their own analytic purposes.
Accounts from those working within non-university consultancy firms are absent from the literature. This is not unsurprising as unlike the academic researcher, the private consultant would have little to gain from the publication of critical reflections on the research process. They will often work to tight deadlines and on multiple projects simultaneously, so may have insufficient time to devote to such accounts. Although in this category the researcher has less control over the focus of the research and the end project outcomes, this does not necessarily mean that the work lacks value and credibility. There are numerous examples of well-conducted, rigorous research by private consultants, but the researcher tends to be little concerned with end impacts. The priority lies in delivery of the research as specified in the contract. Table 2 draws together the typology reported above. Column 2 lists the main ways in which research contributes to impact within the differing forms of research-user relationship outlined in Table 1. Column 3 indicates conditions fostering those impact-generating contributions and Column 4 conditions inhibiting them. There is no single route to impact, routes vary, and contextual contingencies are at work in the achievement of impact.
In practice idiosyncratic contextual conditions and the intentional behaviours of those involved in the relationships mean that though we hope that the broad picture painted here captures important impact condition configurations, those configurations do not operate in a mechanical or deterministic way. Moreover, individual researchers migrate from involvement in one type of researcher-user relationship to another and, indeed, may be involved in several simultaneously in different research projects, or indeed even within one project.

Conclusion
This paper adopted a broadly realist approach to identifying a variety of researcher-user relationships as contextual conditions favouring or inhibiting impact realisation, using policing and crime prevention as a relatively well-documented example. The paper notes a broad move from Mode 1 to Mode 2 research involving increased attention to impact on the part of researchers and increasing appetite for research amongst users. Drawing on available published participant accounts of research and impact, this paper has shown that differing forms of relationship between researchers and users, along a Mode 1 to Mode 2 spectrum, create varying contextual conditions for impact pathways, some of which may involve uncomfortable compromises for researchers whose academic freedom may be jeopardised. However, no form of relationship either guarantees impact or rules it out. Situational contingencies, the attributes of individuals involved, and the intentional behaviours of those involved make a difference, and the type of relationship may well affect the form that impact takes. It may also be that Mode 2 research has a higher likelihood of facilitating impact, but that more independent Mode 1 research can have more far-reaching and fundamental impacts.
It is also important to recognise that routes to impact can be complex and non-linear and be heavily influenced by factors outside of the researcher's control. The sources of funding for the research, for example, may have substantial effects on a researcher's ability to achieve meaningful and sustainable impacts. Often research funding may come directly from a user organisation, which may limit the researchers' agency to ensure findings are acted upon and implemented. This paper alludes to a project in which the authors have participated where efforts have been made to maximise real-time impact with research-users via a combination of Mode 1 and Mode 2 research outputs. It makes use of diverse case study examples from within policing and crime prevention although we acknowledge they may not have been representative. From within this literature, however, common features regarding the researcher-user relationship were evident, and these were felt by the researchers to have a varying effect on the impact (or lack of impact) of that given project.
It is important to note, however, that these relationships are just one potential contextual factor that we chose to focus on here, and it is likely that there is great variability within these relationships. In addition, there are likely to be many other contributory contextual factors such as the interpersonal relationships between individual researchers and users, their respective past experiences of coworking, and therefore the level of researcher-user trust. We see the typology outlined in this paper as a starting point enabling future research to continue to explore how contextual factors affect the mechanisms behind research impact. Very little is known about the specific impact pathways within crime-related research, and we hope that the framework presented in this paper will stimulate debate and further investigation.
One limitation of the current research is the fact that we can only rely on accounts that are published and therefore available to us. It is likely that researchers are much more likely to write a reflective account of their experience if it is contrary to their usual experiences, particularly if the experience is negative resulting in conflict or a feeling that potential impact has been minimised in some way. For example, there are likely to be experiences where researchers have been critical of the research user organisation, and that criticism has been positively responded to, but in such cases they may not feel it is as worthy of discussion. We are probably much less likely to find these accounts. In the case of contract research, where the research is a commercial venture, there are very few accounts of the research experience, so some accounts are notably (but unsurprisingly) absent from the literature.
The typology proposed in this paper comprises 'ideal types'. Ideal types attempt to provide informative simplifications that illuminate social processes. Unsurprisingly, therefore, we found that actual projects and experiences can cross categories. Researchers change roles dependent on the research project, and the nature of the researcher-user relationship can evolve within the timeframe of a single project and/or collaboration (see, for example, Davies 2018Davies , 2021. The current focus has been on the nature of those relationships. In future work we hope to build upon this further to explore models of specific pathways to impact. Whilst there is a considerable body of literature on this in health sciences, for example, models of research-user relationships and associated pathways to impact are under-researched in crime-related research. Notes 1. See, for example, Left Realist criminologists who have voiced concerns about the risks of further criminalising already marginalised sectors of society by working too closely with the government and criminal justice agencies (Young 1992, Lea 2015). 2. The use of ideal types was originally proposed by Max Weber (see Weber 1949). Ideal types are neither average types nor normatively desirable types. Rather they are simplifications that attempt to abstract key elements and their relationships in ways that are informative. They are not intended to capture the full complexity and heterogeneity of actual cases. 3. The researchers felt that it proved unpopular for political reasons. It was finally trialled in 2009 in North Carolina, was evaluated to be successful, and then replicated elsewhere. 4. For a list of publications and outputs please refer to the project website https://covid19-crime.com/

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by the Economic and Social Research Council under the UKRI Open Call on Covid-19 [ES/ V00445X/1].