Social work research in the UK: a view through the lens of REF2021

The Research Assessment Exercise was introduced in 1986 to measure research quality and to determine the allocation of higher education funding. The renamed Research Excellence Framework (REF) has become an important barometer of research capacity and calibre across academic disciplines in UK universities. Based on the expert insights of REF sub-panel members for Unit of Assessment 20 (UOA20), Social Work and Social Policy, this article contributes to understanding of the current state of UK social work research. It documents the process of research quality assessment and reports on the current social work research landscape, including impact. Given its growing vigour, increased engagement with theory and conceptual frameworks, policy and practice and its methodological diversity, it is evident that social work research has achieved considerable consolidation and growth in its activity and knowledge base. Whilst Russell


Introduction
The Research Excellence Framework (REF) has significant implications for the quality and vitality of UK research, for academic disciplines, universities, university departments, future research funding and academic careers. The REF usually takes place every six years; the previous exercise was completed in 2014, but REF2021 did not report until June 2022 following delays caused by the pandemic. The assessment process is delivered through panels and sub-panels which take responsibility for Units of Assessment (UOAs). UOA20 covered Social Work and Social Policy and included some, but not all, research in Criminology. Units submitting to the REF were assessed on quality of research outputs, impact (research-generated effects, or benefits beyond academia, demonstrated by case studies) and environment (demonstrated by environment statements and additional data).
Universities and their staff commit significant energy and resources to the REF process and, with the shift to assessing impact introduced for REF2014, social work organisations and policy makers also contribute by providing information and supporting statements to evidence impact. It is therefore essential that all opportunities for learning offered by the REF exercise are captured.
This article has been produced by both academic and impact assessors who served on Sub-panel 20 for REF2021, with the aim of disseminating and reflecting on the picture of social work research and impact provided by REF assessment. We acknowledge that UOA20 does not capture all social work research produced in the UK: numerous research outputs are not submitted for REF assessment and some universities take the strategic decision to return their social work research to another UOA such as Allied Health Professions (UOA3) or Sociology (UOA21). Members of Sub-panel 20 encountered challenges in distinguishing social work from social policy and criminology research: much of the work assessed was interdisciplinary and, for many outputs, disciplinary boundaries appeared porous. Nevertheless, opportunities for painting a picture of social work research across the UK are rare and this account benefits from the fact that some social work assessors contributed to the work of the subpanel in both 2014 and 2021, enabling us to comment on change and growth.
Research England, part of UK Research and Innovation (UKRI) which provides government funding for research, managed the REF process and acted on behalf of the Scottish Funding Council, the Higher Education Council for Wales and the Department for the Economy in Northern Ireland. Formal REF outputs include reports to universities and their departments, ratings tables which are broken down by UOA and by institution, and panel reports which include overviews of the work of individual sub-panels [see REF2021 (2022b for Sub-panel 20's overview report]. The REF has strong governance processes so, whilst other organisations may use REF data to produce their own material (e.g. university marketing campaigns or university 'league tables' published in the national press) these are not endorsed by the REF. This article represents the views of social work assessors who served on Subpanel 20; our comments should not be attributed to UKRI or the chairs of Sub-panel 20 or Main Panel C. We offer 'high-level' conclusions about the assessment process and the broad profile of social work research and impact that the REF provides. We do not comment on specific outputs, researchers or institutions.
We begin by considering the implications of the REF for universities, their staff and for society more broadly. We then move to explaining the REF process, in particular the workings of Sub-panel 20. We provide accounts of social work research and impact as viewed through the lens of the REF before reaching some conclusions. Discussion of research environment, the third element of REF assessment, is woven through. Each academic assessor on Sub-panel 20 reviewed only the outputs allocated to themself; none saw the full array. Whilst we have been able to draw on Sub-panel 20's published overview report (REF2021, 2022), REF rules required us to destroy all notes and records. Instead, we have engaged in iterative consultation with all seven social work academic assessors on Sub-panel 20, eliciting their responses to a questionnaire and their comments on successive drafts of this article. This dialogical process has enabled us to paint a rich picture of social work research submitted to Sub-panel 20. Since REF impact case studies, and impact scores for each UOA submission, are in the public domain (REF2021, 2022a), we have been able to return directly to these to inform this account.  (Coe and Kernohan, 2022).

Why the REF matters
The implications of additional research funding are significant. Research councils and other funders can use REF outcomes as a benchmark of research quality. A key change since REF2014 is that UK access to European Union research funding is restricted following Brexit. Hence, QR funding's contribution to the total research pot is now greater and may be relied on more heavily. Arguably, the REF process provides accountability for public investment, so that the benefits of university research can be demonstrated to funders and society as a whole. Along with other forms of audit, it creates a performance incentive for universities.
Within universities, REF results are used to inform institutional decisions on resource allocation. Departments that have improved their position in the league tables may expect to be rewarded, whilst departments whose ranking has fallen and who lack other sources of income (particularly teaching income) may be vulnerable to cuts. There are also implications for staff contracts. Following the Stern Review (2016), in REF2021, all staff with 'significant responsibility for research' were entered for REF assessment, a rule change that resulted in 76,132 academics submitting at least one research output, compared with 52,000 in 2014.
Universities developed their own internal systems for assessing the quality of outputs well ahead of the REF submission deadline. In some institutions, staff with 'significant responsibility' for research but with outputs either scarce or judged internally to be of low quality may have been 'invited to consider' changing to 'teaching only' contracts in order to ensure a high-quality submission overall. The REF results do not specify the ratings allocated to individual researchers' outputs, so offering protection for individual careers. For staff included in the REF, particularly those chosen to submit an impact case study, advantages may include opportunities for promotion or career advancement elsewhere.
The value of the REF is not uncontested. Some critics regard it as emblematic of an expanding audit culture (Torrance, 2020) that 'constructs an illusion of intellectual excellence and innovation whose true purpose is the neutralization of the university as a centre of independent knowledge creation and learning' (O'Regan and Gray, 2018, p. 533). The University and Colleges Union (UCU, 2022) considered the REF 'a flawed, bureaucratic nightmare . . . a drain on the time and resources of university staff, and funding often entrenches structural inequalities'. Others are more qualified in their appraisals. Whilst also criticising the neoliberal underpinnings of the REF, MacDonald (2017) recognises that it can provide institutional space and incentive for impactful research for the public good. Manville et al. (2021) highlight the burden imposed by the REF, but report that over half the academics surveyed claimed it had not significantly influenced their own research. At the other end of the spectrum, Whitfield (2023) expresses confidence in the REF as a mechanism for ensuring accountability for government research funding which has strengthened the quality of UK research.
As suggested by the UCU, QR funding allocations may reinforce some existing inequalities between universities: wealthier institutions tend to perform well and so receive more QR funding. Russell Group (a self-selected group of twenty-four older and better-endowed universities) and older universities cluster at the top of the rankings. But this cannot be taken for granted and in social work and social policy, some newer institutions performed well in 2021.
For the world outside universities, the REF's emphasis on impact and requirement to produce impact case studies has had the effect of strengthening communication and collaboration between university-based researchers and a wide range of partners from the practice and policy sectors. As an applied discipline, social work research has always maintained close links with research users, but the assessment of impact has sharpened researchers' interest in how their research is disseminated, taken up and translated into social change. With research translation now a major preoccupation for academia, university research is increasingly likely to be engaging in ongoing dialogue with a variety of stakeholders and contributing to local, regional and national priorities.
The process of assessment UOA20 was one of twelve sub-panels that sat under REF2021's Main Panel C which covered the social sciences. This structure allowed for planning, calibration and management of the task of assessing very large amounts of material to be carried out across disciplines, with the aim of achieving consistency and adherence to common practices and standards. Chairs were appointed to the REF panels and sub-panels following open advertisement and interviews and sub-panel members were nominated by learned bodies and other associations such as the Association of Professors of Social Work, Joint Social Work Education Council, the British Association of Social Work, the British Society of Gerontology and some of the major charities. Impact assessors and research users were similarly nominated and came from a range of practice, policy, research and knowledge transfer organisations. Sub-panel members were appointed in stages, allowing sub-panels to ensure they included a wide range of expertise and were sufficiently diverse and representative. The REF's own analysis of appointments to all REF2021 sub-panels found significantly increased representativeness since 2014 (REF2021/01, 2021). Sub-panel 20 included twenty-six academic assessors (eight of whom were predominantly social work academics), three research users, six impact assessors, one panel adviser and one panel secretary. Twenty-four sub-panel members (63 per cent) were women and fourteen (37 per cent) were men. Twenty-one per cent of the sub-panel were from Black and Minoritised ethnic communities. Institutions from each of the UK's four nations were represented as were Russell Group and post-1992 universities (approximately seventy-eight institutions, mostly former polytechnics granted university status by the 1992 Further and Higher Education Act).
Sub-panel 20 met between November 2019 and March 2022. Restrictions imposed by the pandemic meant that most meetings were held online, but this did not seem to impede communication or decisionmaking. Planning and preparation started early and sub-panel members were able to contribute to developing guidance for submitting units-this was deliberately broad and inclusive reflecting the disciplines' porous boundaries-planning workload management and working methods. All sub-panel members undertook unconscious bias training and participated in calibration exercises designed to ensure consistency in grading all components of submissions. In keeping with REF guidance (REF2021a, 2019) regarding the reliability of citation data, especially in applied research fields, Sub-panel 20 made no use of bibliographic metrics to inform the assessment of outputs, relying solely on expert review. A grading process underpinned by assessor commentary was adopted with continuous monitoring of assessors' rating patterns and particular attention given to borderline grades.
In total, seventy-six universities submitted research to sub-panel 20 -15 more than in REF2014. Of these, 39.5 per cent were pre-1992 and 60.5 per cent were post-1992 universities. Fourteen institutions submitted to UOA20 for the first time, indicating expansion of research in these fields.
Likewise, the numbers of full-time equivalent (FTE) staff with significant responsibility for research who were returned to UOA20 increased by 61.8 per cent from REF2014. This growth is in part attributable to changes in REF procedures: REF2014 did not require all such staff to be submitted, whereas REF2021 did. However, it also indicates the vigour of the UOA20 disciplines. The sub-panel received a total of 5,158 outputs for review, an increase of 8 per cent from REF2014.
All outputs were reviewed by two assessors. One took responsibility for reviewing all outputs for an institution, so providing a full overview of that university's unit of submission. The second was allocated on the basis of their expertise. A high number of interdisciplinary outputs was assessed and where expertise was felt to be lacking, in consultation with the sub-panel's interdisciplinary adviser, outputs were cross-referred to appropriate sub-panels. Where gradings differed substantially, paired assessors discussed them and reached a joint decision. Very occasionally, a third assessor was brought in. The same approach was employed to assess impact case studies which were assessed jointly by an impact assessor and the academic assessor with responsibility for the submitting institution. Environment statements were assessed in larger groups of three including the allocated institutional assessor and two other panel members. The overall quality profile for each unit of submission to the REF2021 was calculated using a formula weighting of 60 per cent for outputs, 25 per cent for impact and 15 per cent for environment.
Social work and social policy: UOA20 results As noted above, it was not always easy to allocate outputs to one specific discipline. However, if the number of outputs allocated to a second assessor with social work expertise is used as a crude indicator of discipline, Sub-panel 20 assessed approximately 1,400 outputs for social work, 2,220 for social policy and 1,130 for criminology. Social work outputs therefore constituted approximately 30 per cent of the work assessed. In total, 225 impact case studies were assessed across all three disciplines, up from 190 in 2014. About one-third of these were predominantly social work, although this categorisation is very approximate since many were multidisciplinary. Table 1 shows the final profiles for outputs, impact and environment across the whole UOA. As in REF2014, the UOA as a whole performed particularly well in respect of impact with over three-quarters of case studies rated 3* (internationally excellent) or 4* (world-leading). Similarly, three-quarters of the outputs assessed were rated 3* or 4*. Direct comparison between REF2021 and REF2014 results is difficult, due to changed rules regarding the ratio of outputs to returned FTE staff, and because REF2021 covered seven years (1 January 2014-31 December 2020) whilst REF2014 covered six years (January 2008-December 2013). Nonetheless, these results represent a slight improvement on those of REF2014. Of the 4,844 outputs submitted to the sub-panel, 577 (12 per cent) were attributed to early career researchers, signalling the continued growth and sustainability of the research community. Similarly, the environment statements submitted to Sub-panel 20 reported a 20.9 per cent increase in PhD completions over the REF period, as seen in Table 2.
Other indicators of growth found across the sub-panel included an average increase in universities' social work and social policy research funding of £4.2 million per annum across the REF period.
Following changes in the REF guidance, the environment statements reported on equality, diversity and inclusion (EDI) issues more fully than in REF2014. The stronger statements assessed provided detail on progress made since REF2014 regarding under-representation of staff with protected characteristics (i.e. age, disability, gender reassignment, marriage and civil partnership, pregnancy and maternity, race, religion or belief, sex and sexual orientation as defined by the 2010 Equality Act).
Significant progress towards, or the achievement of, key EDI benchmarks was found in some cases, although this tended to be better evidenced for gender than ethnicity or other protected characteristics. However, there were some examples of positive practice in relation to research students: for instance, one institution had targeted PhD studentships on Black, Asian and Minoritised students. The final ranking table for UOA20 was not dissimilar to that for REF2014, with the top end dominated by the better-resourced Russell Group and older universities. However, it was notable that more post-1992 institutions featured in the top thirty universities for social work and social policy than in 2014, indicating the increasing strength of some of these departments. As a group, institutions submitting to UOA20 for the first time produced a generally lower profile than the overall subpanel profile: success in the REF is, to some extent, shaped by experience in and the resource devoted to crafting the submission. However, these new submissions displayed some key strengths and demonstrated expansion of the disciplines.

The research landscape
Turning to UK social work outputs submitted to Sub-panel 20, the overarching impression is that social work has continued to grow in maturity, sophistication and confidence as a research discipline and interdisciplinary field. Increased breadth and diversity were demonstrated in the range of contemporary issues explored and approaches taken. Assessors saw examples of outstanding excellence amongst outputs based on applied and empirical, conceptual and literature-based research. Where substantive themes or methodologies were familiar from REF2014, these were often treated with greater refinement and criticality. There were also strong signs that substantial funder investment can bear fruit in research of world-leading quality. Whilst centres of excellence in particular sub-fields produced some outstanding outputs, excellent research and 4* outputs were identified in all submitting units, including those newly submitting to the REF.

Output quality assessment
The three criteria used for appraising the quality of outputs were 'originality, significance and rigour' (see Box 1).
The overall quality of each output was judged according to the following starred level ratings: Four star: quality that is world leading in terms of originality, significance and rigour. Three star: quality that is internationally excellent in terms of originality, significance and rigour but which falls short of the highest standards of excellence. Two star: quality that is recognised internationally in terms of originality, significance and rigour.
One star: quality that is recognised nationally in terms of originality, significance and rigour. Unclassified: outputs that fall below the quality levels described above or do not meet the definition of research used for the REF.

Demonstrating research quality
Whilst claims to originality, significance and rigour were commonly flagged in outputs assessed, their validity needed to be demonstrated and not merely asserted. Originality was strongest where, for example, a new model was introduced for conceptualising or responding to a familiar social work issue or new questions explored in uncharted territory. It was less evident where a tried and tested approach was being replicated or there was significant overlap between outputs within the same submission. Significance could be both under-sold, commonly as an afterthought, or over-sold, say by drawing far-reaching implications from an exploratory study. Conversely, compelling significance could be demonstrated where clear contributions to knowledge, implications for law, policy, practice and/or research were well integrated. We discuss demonstration of rigour in more detail below. However, we saw instances where limited funding or tight deadlines appeared to have led to hurriedly written reports or lack of fit between funder agendas and realworld practicalities (such as challenges with implementing initiatives to be evaluated) compromised the suitability and effectiveness of the research approach taken.
Box 1. Research quality criteria 1. Originality: . . . the extent to which the output makes an important and innovative contribution to understanding and knowledge in the field. Research outputs that demonstrate originality may do one or more of the following: produce and interpret new empirical finding or new material; engage with new and/or complex problems; develop innovative research methods, methodologies and analytical techniques; show imaginative and creative scope; provide new arguments and/or new forms of expression, formal innovations, interpretations and/or insights; collect and engage with novel types of data and/or advance theory or the analysis of doctrine, policy or practice, and new forms of expression.
2. Significance: . . . the extent to which the work has influenced, or has the capacity to influence, knowledge and scholarly thought, or the development and understanding of policy and/or practice.

Topics examined
As in REF2014, there was a preponderance of social work research on children and families, particularly child safeguarding, looked after children and young people and care leavers. In contrast, there was less attention to early help and support (services provided for children and families as soon as a problem appears). Some outputs highlighted the socio-economic, political and cultural determinants of challenges these children and families face, others focused on their individual and immediate characteristics and contexts. Many outputs discussed evaluation of new, sometimes innovative, practice or policy initiatives and their impacts; others explored pathways through and outcomes from servicesas-usual. A prominent theme was child protection threshold decisionmaking, with relationship-based and trauma-informed practice, online harms and child trafficking more recently emerging themes. Amongst outputs addressing social work and social care with adults and older people, established foci on ageing and dementia, mental health, disabilities, care and carers, and palliative or end-of-life care, were complemented by growth in attention to sexuality and LGBTQþ groups, and gerotechnology. Other themes included isolation and loneliness, work and retirement and cultural and environmental gerontology. Several research themes cross-cut service user groups. Prominent amongst them was interpersonal violence against children, women and older people. Cross-cutting themes were also commonly interdisciplinary and spoke to 'big' contemporary social work and policy challenges: inequalities, poverty and social exclusion, migration and digital transformation. For example, there was clear growth in outputs addressing social work with refugees, asylum seekers and migrants; with Black and Minoritised communities and on digital technologies-their opportunities and risks, impacts on practice and technological value as research tools. Intergenerational and spatial perspectives were also foregrounded.
As in REF2014, an array of outputs considered the social work profession: its governance and regulation; professional identity, ethics and values; organisational cultures; supervision and support; social workers' well-being and professional education.

Use of theory and conceptual frameworks
There was wide variation in the extent to which outputs engaged with theory or concepts. A minority were written expressly to use a particular theoretical or conceptual lens to scrutinise social work issues. Elsewhere, variable recourse to theory could reflect significant differences in funding level and requirements. Outputs from government-or agencycommissioned research were more often targeted towards practice or policy application than theorisation. In contrast, larger-scale research council or charity funding could provide the platform for outputs whose sophisticated application of theory significantly enhanced their contributions to knowledge as well as practice and policy.
Relatively, few social work outputs introduced entirely original theories or concepts, but a considerable number drew on theory from diverse disciplines. Some familiar theories-such as attachment or ecological theories, the social model of disability or specific practice theories-were applied with increased criticality and nuance. Others were more newly applied to social work issues, including theories of social justice and rights applied to safeguarding; post-colonial theory and concepts of social exclusion, cultural competence or empowerment applied to work with minority communities and migrants; organisational learning and systems theories applied to innovation and development and psychosocial and communication theories applied to relationship-based practice.

Scale and ambition
Many social work outputs were based on small scale, local and qualitative research, as is typical within the discipline. These bear witness both to high levels of local stakeholder partnership that underpin their funding, focus and design, and to alignment with person-centred social work values and engagement with lived experience. However, their preponderance may also reflect low funding, traditional lack of UK social work research capacity or appetite for quantitative research, and paucity of robust, suitable large-scale datasets.
Nonetheless, there was encouraging evidence that the scope and scale of the UK social work research landscape is shifting. Some excellent collaborative international and comparative research addressed social work concerns globally (including in the Global South). There were also some outstanding outputs from large, well-funded, multi-method and/or longitudinal projects. Where there was substantial external funding from major national research funders, there was scope and scale for groundbreaking research and world-leading outputs. Such projects could also offer capacity development, with early career researchers on occasion leading authorship of excellent publications.

Methodologies and methods
We noted the increasing range, diversity and sophistication of methodologies and methods used. Empirical research outputs ranged from: small scale, qualitative studies to longitudinal analyses of large datasets and multi-method evaluations to in-depth ethnographies and cross-country comparisons to local action research projects. Naturally, methodological quality varied: some methods were applied inappropriately; others were poorly applied or were too thinly explained. Assessors saw plentiful use of tried and tested methods such as interviews and focus groups, with relatively fewer novel methods developed and well demonstrated. Despite this, the growth in qualitative and quantitative methodological range, competence and confidence was impressive.
First, it was clear that qualitative research strengths already demonstrated in REF2014 had developed further, including increased use of innovative and creative-such as visual, digital and arts-based-methods. There was also some excellent ethnographic research, at times making sophisticated use of multiple methods to explore lived experience and practice in diverse settings. UK social work research has long prioritised user voice and co-production-arguably as a path-leader in social and health sciences. Assessors observed participatory research continuing to flourish, through co-produced work with people with dementia, mental health needs and physical impairments, with LGBTQþ and excluded communities and with vulnerable children and young people. The highest quality studies reflected critically upon the complexity of service users' engagement with services and research, deploying methods to enable meaningful participation across all research stages.
Second, enhanced methodological range and sophistication were demonstrated in areas considered weaker in social work research submitted to REF2014. These included: sound use of economic analysis to evaluate cost-effectiveness/benefits of initiatives; rigorous randomised-controlled trials or quasi-experiments examining the effectiveness of interventions and digital technology used for large-scale surveys. The growth and improved rigour of quantitative and longitudinal research was especially striking, featuring amongst others robust multivariate analysis of large administrative or cohort/panel datasets that could afford, at scale, cross-sectional evidence of patterns of needs and services, or longitudinal evidence of outcomes over time. Large-scale and experimental quantitative methods could also be deployed to excellent effect within mixed-method designs. For example, mixed methods randomised-controlled trials could shed light on the contexts and mechanisms of change that contributed to outcomes; qualitative interviews with service users or providers could make sense of patterns of need, service use or outcomes identified through large data analysis.

Research impact
What is impact?
Research impact has become an increasingly important element in REF assessment: its contribution to the overall rating of submissions increased from 20 per cent in REF2014 to 25 per cent in REF2021. Impact was reported mainly in the format of impact case studies which were expected to provide a clear, coherent narrative that included an account of which audiences, constituencies, groups, organisations, places, services or sectors had benefited, been influenced or been acted upon as a consequence of research. REF2021 defined impact as an 'effect on, change or benefit to the economy, society, culture, public policy or services, health, the environment or quality of life, beyond academia' (REF2021, 2019a, p. 68). So any positive impact beyond an institution's own teaching and advancement of academic knowledge was within scope of this broad definition.
Impact was understood to include, but was not limited to, an effect on, change or benefit to: the activity, attitude, awareness, behaviour, capacity, opportunity, performance, policy, practice, process or understanding; of an audience, beneficiary, community, constituency, organisation or individuals; in any geographic location whether locally, regionally, nationally or internationally (REF2021, 2019a, p. 90).
The guidance also acknowledged additional impacts arising from research, such as holding public or private bodies to account, which may have resulted in a proposed change not taking place, or made a contribution to critical public debate. Assessors considered whether the claims made for impact were supported by evidence and indicators. They also examined the coherence and clarity of narratives describing the sequence of activities and events leading to underpinning research being acted upon. This 'pathway to impact' proved to be a key element in the assessment. REF documentation acknowledged 'there are multiple and diverse pathways through which research achieves impact'; moreover, the relationship between research and impact can be 'indirect or non-linear' and can be 'foreseen or unforeseen'. Impact may also be achieved by 'individuals alone, through inter-institutional groups, to groups including both academic and nonacademic participants' (REF2021, 2019b, p. 63).
The two key components of impact were articulated in REF guidance as 1. 'Reach' is understood as the extent and/or diversity of the beneficiaries of the impact. Reach is defined and assessed in terms of the extent to which the potential constituencies, number or groups of beneficiaries have been reached (REF2021, 2019a, p. 52). So, this is designed to distinguish between, for example, a study that had an impact on a small number of people in one geographical area, with another that could have affected policy impacting on a whole country. Having said that, it is not designed to be assessed in purely geographic terms, nor in terms of absolute numbers of beneficiaries. 2. 'Significance' is understood as the degree to which the impact has enabled, enriched, influenced, informed or changed the performance, policies, practices, products, services, understanding, awareness or well-being of the beneficiaries (REF2021, 2019a, p. 52). Therefore, this is designed to capture something about the scale and importance of the claimed impact.
Assessors produced an overall judgement about reach and significance of impacts, rather than assessing each criterion separately, and neither was privileged over the other.

Quality and range of impacts
Over three-quarters of the impact case studies considered as predominantly social work were judged to have achieved impact of outstanding or very considerable reach and significance. Many were interdisciplinary and high levels of impact were achieved at local, regional, national and international levels.
Amongst the wide variety of social work case studies, some were based on the work of one or two researchers whilst others were collaborative. Some were founded on one-off studies, others on a long-term commitment to one research area, such as adoption practice. They demonstrated a wide range of practice and policy outcomes, with impact reaching practitioners, policy makers, local authorities, membership organisations, voluntary sector groups, service users, excluded groups and communities.
The following comprise the main types of impact demonstrated: Government policy including legislation, statutory guidance-such as changing formal legislation about rules on contact. Regulation-such as the inclusion of new criteria into the inspection frameworks used by Ofsted. Practices and the way in people approach their work-such as developing new 'models' of social work practice by delivering training programmes across a number of local authorities or professional groups. Products, resources and technologies-such as new toolkits for social workers or managers of residential homes, and physical objects for interacting with children.
Increased understanding, including public or professional awareness-such as identifying positive benefits of aspects of the care system traditionally seen as problematic.
Increased well-being, engagement, safety or health of service users-such as helping people who have experienced abuse overcome feelings of stigma and disempowerment through a process of ethical engagement in the research and policy-making process.
Excellence was found across the different impact types and beneficiaries, including examples of challenging the status quo and influencing the terms of debate. However, some case studies seemed to confuse dissemination with impact by, for example, simply citing numbers of downloads or participants attending training sessions, rather than specifying what had actually changed. Impact could also be over-claimed, such as where contributions affected teaching within but not beyond the submitting institution, or changed practice appeared attributable to the roll-out of an intervention, rather than to the research evaluating it. Impact was frequently assisted by researchers engaging with key stakeholders and planning for impact from the outset. Research council and universities' own Impact Acceleration funding had supported ongoing relationships with key stakeholders in a number of cases, and including stakeholder representatives in research teams was also effective. Participation in formal governance processes, such as responses to consultations and attendance at select committees were important ways to influence policy makers, and using a range of digital technologies, including communication campaigns, could amplify impact wider. The strongest case studies were able to show how research findings had been taken on by users in ways that shifted practices and/or policy agendas beyond original expectations. However, sub-panel members also noted some challenges in providing evidence. The links between the research and the impact claimed were not always clear, nor the beneficiaries always easily identifiable.
Notwithstanding some shortcomings, there were notable improvements overall in the presentation of case studies, indicating that planning and support to achieve external impact is becoming more established. There had clearly been a continuing commitment of resources and investment to enable staff to develop and engage in impact activities, reflecting a growing maturity of impact strategies in institutions (McKenna, 2021).

Discussion
The enhanced contribution of impact to the assessment process in REF2021 played to social work's strengths. As an applied discipline promoting social change and development (International Federation of Social Workers, 2014), social work research is committed to achieving impact. Increasingly, universities are devoting energy and resources to building collaborations with research users and those who use services; these help to ensure research findings are used to refine policy, improve services, practices and their outcomes and widen public understanding. This approach needs to be adopted consistently across institutions: the relationships that support impact creation require nurturing over time.
There were indications that UK social work research is becoming less parochial and that increased success in capturing funding has produced some rigorous, large-scale studies that address complex, sometimes global, challenges. Social work research appears more confident in wielding a range of methodologies to good effect, whilst maintaining attention to inequalities and power disparities. Funders need to take on board the message that high-quality social work research is delivered when sufficient time and resources are invested.
As university research centres and institutes build their size and expertise, an inevitable tendency towards specialisation may increase the divide between research on children and on adults. However, some research fields, such as mental health, interpersonal violence, inequalities or poverty, may succeed in bridging this space. The REF's emphasis on interdisciplinarity may also be valuable here.
In terms of social work's interdisciplinary partnerships, new research partners in fields such as geography, art and design and informatics are being added to the list of traditional research partners. Social work has always been a porous discipline and this has enabled it to forge new partnerships and embrace new thinking and fields of study.
For those assessing social work research, its open boundaries pose challenges in defining and reporting on what is contained within its territory. For REF purposes, social work's inclusion in a sub-panel shared with social policy and criminology seems appropriate and allowed for interdisciplinary assessment of outputs and impact that were themselves frequently interdisciplinary. However, for a complete picture of UK social work research to emerge from the REF, all social work units would need to submit to UOA20; we would encourage those who returned their research to other UOAs to consider doing so in the next REF.
All outputs and impact case studies submitted to REF2021 are now available on its website (https://results2021.ref.ac.uk/outputs# and https:// results2021.ref.ac.uk). This offers the fullest record available at present of social work research in the UK. This material could usefully be categorised and reviewed in a more systematic manner than has been possible here to identify current themes, trends and gaps in social work research. Knowledge of which questions or issues are under-researched, for instance, would be helpful.
The REF process is founded on the principle of peer review and this method of assessment is familiar and acceptable to most academics. In REF2021, in line with the recommendations of the Stern Review (2016), outputs were treated as the product of the submitting unit rather than of individual researchers. This is consistent with the collaborative and teambased approaches common in social work research, and contrasts with a focus on individual researchers that characterises other approaches to identifying excellence in social work research (see e.g. Hodge and Turner, 2023). Whilst the cost, burgeoning scale and influence of the REF may attract critics, its increased attention to impact has shifted university strategies in ways that have benefited social work research. Moreover, the REF's focus on EDI issues has potential to effect change in higher education, and this drive aligns closely with social work research agendas and values.

Conclusion
During the global pandemic, there was considerable media coverage and public interest concerning the experience of marginalised groups, as well as other issues that have traditionally been the focus of social work research but have attracted limited attention from other constituencies. This shift in public discourse offers an opportunity for social work research to reach new and wider audiences and to accelerate the pace of social change. We have described impact as an attainable goal for social work research, whilst noting the importance of support and infrastructure to achieve it. This infrastructure does not need to be confined to universities driven by the competitive ethos of the REF and its rewards. There are strong arguments for establishing wider cross-cutting mechanisms that can bring together researchers, practitioners and policy makers regionally or nationally, to identify how available research can be harnessed to tackle urgent social problems and where new or more knowledge is required to inform the development of solutions. The picture of social work research and impact offered by the REF could provide the basis for the work of such forums.
Social work research in the UK is growing in strength and influence. It is beginning to attract substantial investment from new funders, notably some that previously have focused solely on health research. The REF provides valuable evidence that social work research can achieve the highest levels of excellence and real change when it is adequately resourced. Opening up sources of health funding to social work researchers has the potential to power large and impactful research studies; this may further diversify the nature of social work research itself. Social work research also has the capacity to offer ideas and methods that are attractive to researchers in other disciplines. REF2021 represents an important staging post in the cycle of social work research's development. The resources and energy invested in the REF exercise should be utilised to explore future directions and themes for the discipline.