Putting context centre stage: evidence from a systems evaluation of an area based empowerment initiative in England

Abstract It is now widely accepted that context matters in evaluations of the health inequalities impact of community-based social initiatives. Systems thinking offers a lens for uncovering the dynamic relationship between such initiatives and their social contexts. However, there are very few examples that show how a systems approach can be applied in practice and what kinds of evidence are produced when this happens. In this paper, we use data from ethnographic fieldwork embedded within a multi-site mixed method evaluation to demonstrate how a systems approach can be applied in practice to evaluate the early stages of an area-based empowerment initiative – Big Local (funded by the Big Lottery Fund and delivered by Local Trust). Taking place in 150 different local areas in England and underpinned by an ethos of resident-led collective action, Big Local offers an illustration of the applicability of a systems approach to better understand the change processes that emerge as social initiatives embed and co-evolve within a series of local contexts. Findings reveal which parts of the social system are likely to be changed, by what mechanisms, and with what implications. They also raise some salient considerations for knowledge generation and methods development in public health evaluation, particularly for the evaluation of social initiatives where change does not necessarily happen in linear or predictable ways. We suggest future evaluations of such initiatives require the use of more flexible designs, encompassing qualitative approaches capable of capturing the complexity of relational systems processes, alongside more traditional quantitative methods.


Background
In public health evaluation, context is broadly defined as 'the circumstances or events that form the environment within which something exists or takes place' (Poland et al., 2006, p. 59) -i.e. a policy or initiative -and may include political, organisational, economic and social features (Blamey & Mackenzie, 2007). It is now widely accepted that context matters when evaluating initiatives in order to generate policy relevant evidence for their effects on inequalities in health (Poland, Frohlich, & Cargo, 2008;Shoveller et al., 2015). Engaging with context is important for understanding geographical variation in health outcomes and life expectancy that may in part be a legacy of past experiences, socio-economic conditions and demography. It is also important for understanding the effects of action taken to address social determinants of health inequalities (Whitehead & Popay, 2010). By definition, such actions can be broadly conceived of as social interventions, as they are, in effect, purposeful attempts to change social contexts (as well as economic and environmental factors) that may in part influence patterns of lifestyle, behaviours and health outcomes (Poland et al., 2008). Despite its importance, a more nuanced interpretation of context is still not routinely integrated into the evaluation of initiatives in the public health field (Datta & Petticrew, 2013). This is particularly the case in relation to social aspects of context (Shoveller et al., 2015) despite the recognition that social relations lie 'at the heart of any social intervention' (Poland et al., 2008, p. 301).
A number of reasons can be offered to explain why context has remained relatively under-theorised in public health research. Advances by theory-driven evaluators have, undeniably, given context more prominence (Blamey & Mackenzie, 2007;Marchal, van Belle, van Olmen, Hoerée, & Kegels, 2012) with a view to teasing out specific contextual conditions conducive to triggering mechanisms for change (Pawson & Tilley, 1994). Yet context is still often 'conceptualised as [something] external' to a public health policy or programme (Barnes, Matka, & Sullivan, 2003, p. 269). This may risk an artificial separation, when the relationship between context and initiatives is, in actuality, more blurred (Shoveller et al., 2015). A further issue stems from the fact that researchers and practitioners tend to conceptualise initiatives in terms of their activities and component parts (Hawe, Shiell, & Riley, 2009). Evaluations have assessed the impact of such initiatives through linear causal chains giving prominence to experimental designs using quantitative methods. Such evaluation approaches may be inappropriate where initiatives are introduced into multifaceted social, economic and political environments as they leave little scope for capturing insights into the complexities of how such dynamic systems function (Diez Roux, 2011). A systems lens -on the other hand -conceives of initiatives as an event introduced in a series of diverse systems (Hawe, Shiell, et al., 2009). This directs attention to understanding 'what changes flow from perturbations in a system' (Petticrew, 2015, p. 3) when initiatives and contexts couple, embed and co-evolve (Hawe, 2015a). Durie and Wyatt (2013), applying such thinking to a case study of a community initiative 'C2' , suggest that systems are characterised by 'dynamic behaviour(s)' (p. 176), requiring attention to how they adapt over time and to the feedback loops (both dampening or amplifying) between system components in order to understand their effects. A relational focus is necessary in order to understand the dynamics of complex systems. We must study the relationships between component parts within the system, and between the whole system and the wider environment within which it sits. This requires more flexible evaluation designs including the use of qualitative methods in order to track the complexities of shifting social processes (Hawe, Bond, & Butler, 2009).
Thinking about interventions and contexts as systems is particularly relevant in the evaluation of area-based initiatives. Such initiatives have been promoted for decades (in the UK at least) as a means of ameliorating deprivation and thereby reducing social and health inequalities (Lawless, Foden, Wilson, & Beatty, 2010;Thomson, 2008). They are a type of policy initiative tackling multiple determinants of health through combined actions on social, economic and environmental conditions and are typically rolled out into a large number of disadvantaged geographical localities that are often socially diverse (Thomson, 2008). As South and Phillips (2014) argue, where such initiatives are underpinned by empowerment processes (e.g. through communities gaining greater control and taking action themselves), the task of separating out the initiative from its context is likely to be even more difficult. This is because the processes of change that unfold in such initiatives are, in effect, changes in dimensions of the social context itself (e.g. social relationships, cohesion and identities). The nature of change that flows from community action may also be unpredictable or disruptive as initiatives and contexts co-evolve (South & Phillips, 2014;Trickett, 2009). Indeed, some argue such disruption is required if initiatives are going to break the pattern of entrenched poor health and health inequalities (Hawe, 2015b).
In this paper, we use data drawn from ethnographic fieldwork embedded within a mixed method evaluation to illustrate how a systems approach can be practically applied in evaluating the early stages of an area-based initiative -Big Local. Findings are used to reveal how a systems approach can help us better understand change processes that emerge as social initiatives embed and co-evolve within a series of local contexts.

The Big Local initiative
Funded by the Big Lottery and managed by a charitable trust, Local Trust, Big Local is arguably one of England's largest area-based initiatives to date. Through the initiative, 150 relatively disadvantaged geographical areas in England have been allocated over £1 million each for a period of 10 years or more to support residents in making their area a better place to live. The funding is accompanied by a range of advice and support from Local Trust and its delivery partners. A core principle underpinning the initiative is that decision-making is put directly into the hands of local residents. By enhancing community control, this resident-led programme therefore has the potential to reduce health inequalities.
Whilst there is flexibility in how Big Local areas use or spend their funding to meet local priorities, they all operate within a shared framework that flows from involving local people in Big Local, forming a partnership led by residents, developing a plan and reviewing it over time (see Box 1). There is no formal requirement for residents to engage with local government structures or public sector organisations. However, Big Local partnerships are encouraged to collaborate and network with professionals and other organisations in order to achieve local priorities or deliver activities in their plans (Local Trust, 2015, July).

The evaluation study
The Communities in Control study is an ongoing independent mixed method evaluation of the impact of the Big Local initiative on health inequalities and their social determinants. The research team comprise staff from five academic members of the national Institute for Health Research (nIHR) School for Public Health Research (SPHR). Data presented in this paper derive from the first phase of the Communities in Control study, which focuses on early implementation of Big Local and how the initiative is unfolding on the ground. Ethics approval for the research methods was obtained from Lancaster University Research Ethics Committee (on 3 February 2014).
Ethnographic fieldwork was conducted by five research teams in 10 disparate geographical areas that were funded under the Big Local initiative and with national delivery partners. Areas were selected to reflect diversity in local context (based on socio-demographic characteristics including population size, socio-economic status, age structure and ethnic diversity) and to include areas that were experiencing Box 1. big local proposed outcomes and framework. big local outcomes: • Communities will be better able to identify local needs and take action in response to them. • People will have increased skills and confidence, so that they continue to identify and respond to needs in the future.
• the community will make a difference to the needs it prioritises. • People will feel that their area is an even better place to live. big local framework: • Getting people involved • exploring your big local vision • Forming your big local partnership • Creating a big local plan • Delivering your big local plan • Collecting the evidence • reviewing your big local plan and partnership Source: http://localtrust.org.uk different stages of the Big Local framework. Fieldwork took place in two overlapping waves between november 2013 and november 2015. All research teams followed a common protocol, allowing for flexibility to suit local settings and for adaptations to be made as the research progressed. The objective was to provide a 'thick' description of Big Local, the factors shaping the initiative in local areas, and the emerging or potential mechanisms influencing changes in community control and other social determinants of health. We took an open-ended approach to defining local context. Working with residents we aimed to identify which aspects of the local historical and contemporary environmental, political, social and economic fabric were most important in shaping local identities and relationships. This included but was not limited to: different perceptions of community held by local stakeholders; trust and conflict between local people and local agencies; and the nature of other policy or local-level initiatives in the area. In total, the research team completed 138 interviews with Big Local residents (n = 62), workers (n = 49) and other stakeholders (n = 27); 18 group exercises (including focus group discussions and participatory mapping exercises) and more than 440 hours of observation at partnership meetings, away days, workshops, events, projects and other community spaces. Informed consent was sought for all fieldwork activities. Fieldwork methods were complemented by a review of 30 Big Local plans. This produced a large volume of data. All research materials (field notes, transcripts, documents) were entered into the data management software nVivo 10 and were coded using a common framework so they could be cross-referenced during more detailed analyses.
In-depth analysis of materials from across all three parts of the study (neighbourhood fieldwork, the review of plans and national fieldwork) was based on comparative narrative synthesis across fieldwork sites and methods. Cross-case analysis was initiated through the sharing of memos and a series of face-to-face data analysis workshops where early findings were presented by each individual research team as a way of identifying patterns and relationships in the data. Analysis then continued through a combination of data tabulation and narrative techniques until an overall story had emerged to describe the findings (Popay et al., 2006). This involved an iterative process of review and refinement until agreement was reached about a set of general propositions in relation to the cross site data (Yin, 2009). After there was a general agreement on the main themes, systems-thinking was drawn upon to help structure ongoing analyses. All quotes and excerpts used to illustrate the themes have been anonymised in order to protect research participants' identities. Codes provide the following information: fieldwork area (FA1-10 or 'national' for fieldwork with national-level delivery partners) -research method -participant role (details in Box 2). For example, 'FA1 -interview -PM04' indicates that the participant was recruited from fieldwork area '1' , the quote is from an interview and the person's role was as a partnership member (PM).
Big Local is still a relatively young initiative. When fieldwork began areas were either developing their first plans or were, on average, within the first 12 months of delivering them. Even so, we were able to observe how the initiative was embedding within local systems (manifesting itself within thinking and practice (Hawe, 2015a)) and how context and initiative were co-evolving during the early stages. Below, we start by describing the process of embedding and then explore how aspects of the pre-existing context shape the changes -in both the initiative and context -that ensue (i.e. their co-evolution). Findings offer insights into how a systems approach can reveal how aspects of social context influence, Box 2. Key to participant codes. lClr = local councillor SG = member of big local steering group (interim partnership) PM = big local partnership member WO = local employee of public or voluntary sector organisation DP = national delivery partner and are influenced by, the development of collective action taken by communities; as well as how these might impact on social and health inequalities.

The embedding of Big Local within each local system
Each Big Local system may be defined by the relationships between the settings where Big Local activity took place, the positions and roles that shaped activity within these settings and the various resources (including skills, ideas and knowledge as well as material resources) that were exchanged. During the early days, most Big Local-initiated activity concerned the formation of the local partnership -the group, consisting largely of local residents, who are responsible for developing and delivering the local Big Local programme. As these partnerships were formed we observed shifts in social connections and relationships as residents came together and engaged with each other, and with other organisations, in order to take ownership of their Big Local initiative. new communities of interest began to emerge as residents (and partner organisations) started to identify and work towards a shared vision for their area. At first, most Big Local-related activity was initiated by the partnership and focused on organising outreach and stakeholder events in order to connect with a wide range of local residents and to explore their aspirations for the local area. …we had like these ideas clouds, they were very simple it was just like a cloud, and a bit like a think bubble and ask people what do you want in the FA6 Ward?… I think we collected probably about 400 ideas clouds… It was really good, we had a really high level of engagement, it felt very active…the idea was to look at those ideas, suggest some priorities and some recommendations and then pitch it back to the wider group. (FA6-interview-LCLR1) Over time, the nature of partnership activities evolved to address identified priorities. A wide variety of different projects were initiated, including: fashion shows (FA5), dog shows (FA5), cooking events, growing/planting projects (FA2, FA3, FA10), family fun days and galas (FA6, FA8), music and dance performances and community arts projects (FA3, FA7). Often events were specifically designed to engage with groups that the partnership considered to have remained relatively marginalised from community life or previous initiatives. For example, partnership members in FA7 were preparing for a community safety event in one of their most 'notorious' estates in order to engage with young people who were believed to be at risk of committing crime. Similarly, in FA5, an area where dog attacks had made the national news, a dog show was held to promote responsible ownership. In some areas, as time went on, responsibility for delivering the identified projects started to diffuse beyond the partnership, with wider sections of the community helping drive the initiative forward. These additional volunteers would help co-manage an ongoing programme of events, becoming increasingly responsible for initiating ongoing action. In other areas, activity was mainly driven by the partnership, with extra volunteers being sought on a one-off basis to assist with specific projects. There were seen to be many benefits to bringing a wider section of the local community into driving Big Local-related activity, not just for the Big Local project but for the local residents themselves.
It's provided some people with the skill sets to be able to implement this kind of work locally … It's kind of professionalised them in a way. They've had to put in rules of governance, and I think that that, those kind of skills will be like long lasting and will leave a legacy in the area, so I think that's quite important in terms of them being active citizens and making decisions about their area. (FA7-interview-LCLRO1) There was often a desire among partnership members to start making noticeable changes in the local area so residents would gain confidence, seeing that they can make a difference by doing things for themselves, and to create a sense of pride among the wider community. In FA10, for example, a local planting initiative was awarded a Britain in Bloom prize, inspiring the local residents to initiate their own gardening club.
The planters have gone in and whatever and you do feel there's a bit of momentum … you know I feel quite positive. (FA10-interview-W03) As well as this focus on what some residents referred to as 'quick wins' (FA3-interview-SG03, for example), most partnerships also began discussing how Big Local could bring about lasting change to their neighbourhood after the initiative was officially finished. Sustainability was often framed not just in terms of reinvesting the money but by changing the way local people viewed, and behaved towards their local area. In some areas the balance between investment in material structures and the more difficult task of bringing about long-term social change (i.e. feelings of togetherness and self-confidence) triggered lively debate.
And people didn't want that kind of project (redevelopment of the high street), they wanted a more, a project that actually helped people, it wasn't only about shop fronts. (FA6-interview-LCLR03) … it was really important to have something visible that will last, rather than one-off events which would be a waste of money. (FA4-interview-SG19) Some partnerships maintained a firmly resident-led model. In others, as time went on, residents began to work closely with various local organisations. We observed how professional and resident behaviours started to shift (both passively and actively) as both adapted to new ways of working. In FA10, for example, a council worker described how special considerations had been put in place in an attempt to modify the behaviours of elected councillors so they behaved as more equal partners. …we started having a very small steering group that just really was coming out with ideas and we had elected members involved because we wanted them to understand that it was about listening to local people and I think that's been a success too but it took a while and so what I would do is have separate meetings with them to talk about what was happening as well as their involvement so it didn't become political or so they could learn that it was about listening rather than leading which is often the case you know when a council is involved with something. But this wasn't about leading this was about encouraging, it was about empowering and it was about supporting. So it was a slightly different way of doing things. So we had to put in you know different things in place to ensure that that happened really. (FA10-interview -W02) It was not just relationships and behaviours that were being shifted by Big Local; resources were also transformed. In FA3, for example, there was evidence of the initiative causing displacement and redistribution of council and Big Local resources (particularly human resources) as both parties negotiated their shared goals and started to pool resources in order to achieve them. …so there is three councillors on the exec we have got support from the council's neighbourhood services team… so they have been really helpful. So that is FA3-C1-1, that you will have seen at the meeting and she is absolutely superb, she and her team. (FA3-interview-PM03) The process of embedding of the initiative within each local system led to various important changes. During fieldwork we observed how Big Local was reaching ever wider sections of the community, how residents and professionals were co-adapting to meet shared goals, how early successes were beginning to reinforce action (or negative experiences dampen action) and how residents were starting to think about sustainability beyond the initial 10 years. We identified various aspects of social context that influenced the development of collective action within each local system. These are explored below.

Physical boundaries
As human beings, we are all embedded in numerous communities based around various common interests. Big Local is primarily about improving the places where people live, so it makes sense that the money is allocated to a specific geographical area (identified by the funders in negotiation with the local council). By doing this, however, the initiative effectively defines the physical -and consequently social -boundaries within which activities take place. Sometimes the boundaries of these new 'Big Local communities' were contiguous with pre-existing social groupings as experienced by local residents (a ward or housing estate, for example). At other times -particularly in the larger Big Local areas 1 -it might bring together a number of previously disparate sub-communities (perhaps separated by a major road). We observed how the cohesiveness of these new Big Local communities greatly influenced the speed at which residents were able to engage with one another, develop a shared vision and bring about change in their local area. Key to this was whether or not people within the area had a strong shared sense of place, a shared history, identity or experience of working together. For example, in FA4, the boundary created by Big Local maps onto a coherent pre-existing community -easily recognisable by local residents and with a strong sense of shared history and a common purpose (to address the area's social and economic decline.) Here, Big Local provided a new impetus for residents and community organisations to come together and to take action around a common endeavour.
There's loads of like-minded people out there but they don't know each other and I think that's going to be a big part of it; getting everyone together focusing on this, you know, this opportunity. It's the only way we can make the most of it. (FA4-interview-SG06) On the other hand, four of our fieldwork sites (FA2, FA3, FA6 and FA9) effectively comprised several distinct neighbourhoods. Tensions emerged as sub-areas identified competing priorities. In some areas, members of the partnership were keen to ensure that each sub-area was represented. …[you have to] make sure that you're getting an equal representation in terms of when it comes to the generation of ideas from all these different places. (FA6-inerviewt-PM02) Others (in FA2 and FA9, for example) felt the Big Local resource should be prioritised for certain sub-areas, often those residents who were considered to be most in need. In FA2, local residents struggled with the scale of the area produced by the Big Local boundary. While there was initial success in bringing in residents from a local association from one neighbourhood, many of those involved ultimately wanted to concentrate their energies on the estate where they lived rather than the Big Local area as a whole.
I wanted to bring people that were on the resident committee to the Big Local. And they did, they attended, and a few of them joined for a long time. But a lot of them have wanted to concentrate back more on the estate, even though it's a small project they felt that they wanted to use their energies back on the estate. (FA2-interview-PM04) Such tensions slowed the process of Big Local embedding within local networks and bringing about change. Sometimes this was due to difficulties for the partnership in identifying, approaching and engaging with the wider community. In FA6, for example, there seems to have been tension among partnership members over reaching the right local residents. A binary distinction was drawn between the usual suspects (i.e. resident activists) and a wider population of what might be considered ordinary residents. This latter group sometimes seemed to be idealised as more genuinely representative and aware of their neighbourhood -if only they could be engaged and their knowledge and energy harnessed.
A member of the partnership talks about some of the other members being a bit 'difficult' . She provides something of the history of Big Local, describing how a previous worker had been heavily involved with Big Local and had been instrumental in writing the plan and moving it forward. Another member of the partnership suggests that there had, however, been tension around what was written and how decisions had been made, with residents of a particular estate not agreeing with what was in the plan. She suggests that it was felt that only certain interests were being perused, and mostly those of the more 'middle class' side of the ward. (FA6-partnership meeting observation) In other areas, partnership members saw no problem in re-envisaging the Big Local boundary to meet their needs. In FA10, a common view was that in order to be inclusive there should be fluidity in how the boundary was enacted to reflect the constant movement of people in and out of the space to use resources and facilities. …we don't try and make it hard and fast the boundary at all and in fact what we find is that people go out of the area to things to activities or to find resources or services and people come in. (FA10-interview-W01) There was also evidence of non-geographically bound communities of shared interest emerging across different Big Local areas as groups of residents linked-up to share learning and knowledge to address common challenges. In part, this was facilitated by networking and learning structures (including faceto-face events) put in place by the Big Local programme nationally as well as through the network of Big Local reps who could facilitate information sharing about what was happening in different Big Local areas.
So it would be an event with 25 people, sometimes more, sometimes less, in different areas. Usually based around a specific topical question and it's about helping them exchange learning from each other really. We sometimes have external speakers as well but the idea is … you can accelerate the rate of change if you get people in community programmes together. (national-interview-DP7) As Big Local partnerships established themselves, evidence was also emerging of these cross-area networks beginning to form more organically at the local level.
Often you see someone saying 'oh yeah the [Big Local area] partnership have just gone to visit [other Big Local area] and they discovered this, that and the other' and, you know, and they'd be like 'does anyone know anything about X,Y or Z' but it reveals that someone, a group of people have gone, got in touch, just sorted it out, jumped in a car, done all the logistics and spent 3 hours learning about something for someone else and that's happened and that's independent.  national partners defined the emergence of such networks as a positive marker, as the above respondent suggested, '[it] tells me they [Big Local residents] are in, they're thinking control' (national-interview-DP7).
Across sites there was a two-way process of negotiation between Big Local and the physical boundaries of the areas in which it is delivered. The development of collective action within Big Local was shaped by pre-existing communities of shared interest and the introduction of Big Local also forced local residents to reconsider how they perceived and acted within local boundaries. Both processes impacted on the kinds of changes and actions that ensued.

Organisational and decision-making structures
Across our field sites we observed some considerable diversity in the way that Big Local came to be positioned within pre-existing organisations and decision-making structures. Quite often this depended on the level of self-organisation among residents before the introduction of Big Local and the quality of relationships between pre-existing organisations and residents. In areas with pre-existing community activism such as tenant and residents associations, local charities or church groups, partnerships would very quickly embed and work through established networks between a wide range of local stakeholders. In FA3, for example, the structures for Big Local were embedded in a pre-existing community partnership. The close and effective working relationship between this partnership and the city council (as well as other local organisations) meant that Big Local activity became increasingly embedded within council structures. The relationship between the two was reinforced as council and Big Local (human) resources were displaced, redistributed (and pooled) as both parties negotiated and worked towards shared goals. The relationship has consolidated over a series of encounters, with both the Big Local partnership and the council becoming increasingly dependent on each other to meet their co-identified goals. This was evident, for example, when discussing a proposed housing development project.
I have also got a commitment from the council that if I can make a success of it they will match fund and better so we could get a lot of money and eventually we can spread wider.  In contrast, in those areas where there were few pre-existing links between organisations, or where local residents did not view these organisations positively (perhaps due to negative experiences in the past), the partnership would sometimes operate on a strictly resident-led model. The process of forging external relationships with powerful others would take more time. In FA9, for example, poor relationships and pre-existing factions meant it seemed important that residents had fully established their partnership and gained a sense of identity and confidence before council workers and elected councillors were invited to join. The implication seemed to be that local residents wanted to protect their project from becoming subsumed under the agenda of other, more powerful, organisations. Similarly, in FA2, residents recognised the need to engage with existing power-holders (in order to achieve their aspirations for the area) but maintained an arms' length relationship -inviting elected members to attend the partnership meetings, but not offering them a vote. This approach was reciprocated, with the council-led community forum not having made any particular attempts to engage with Big Local. Over time, this stance gradually evolved, with the decision being taken to allow a local councillor voting rights and the Big Local partnership providing regular updates to the council-led area forum.
My determination has always been that it should be locally led, not politically led, and so that's one reason we didn't have any councillors with a voting right. Although now we've agreed that they should have…. (FA2-interview-PM03) A common observation across all sites was the central role that pre-existing relationships (and their quality) played in shaping the networks through which Big Local was acting and the direction and speed with which these networks were developing. Different kinds of action emerged as residents either focused on identifying their own goals or co-identified them with their (more powerful) allies.

Past experience and collective memory
In a complex initiative such as Big Local -in which change is dependent on interactions between groups of people -past experiences shape what is, and is not, possible, particularly in the early stages of the initiative. For example (as alluded to in the previous section), residents' previous experiences with and attitudes towards external agencies appeared to greatly influence the extent that residents on the partnerships were prepared to engage with them. In FA4, the perception that money coming into the area had not directly benefited local residents, led the partnership to be suspicious of the motivations of external organisations even when they may be offering to help.
[Partnership members] feel frightened that this might be taken away or somebody else will get the money.  In other areas (such as FA1, FA5 and FA10), where partnership members approached Big Local with no such fears, proactive links were being made with local schools, youth work agencies, job club agencies, informal community groups and arts and craft groups. This resulted in multiple interactions between actors, which worked to maximise actions in priority areas. non-resident stakeholders were seen not only to bring skills/extra powers/leverage to meet local priorities but there was also a focus on the potential gains from joining with others to achieve common goals. In these cases, what appeared to be important was that the expected behaviours of other organisations were seen to match the ethos and priorities of the Big Local partnership (and other local residents.) Shared memories also shaped expectations of what could be achieved by a new community initiative such as Big Local. The impact of an apparently slow pace of change was mediated by people's experience of similar previous initiatives. For many, it led to demotivation, resulting in people dropping out of the process or worries about the wider public perceptions of the partnerships ability to bring about change for the whole community (FA1, FA2, FA4, FA8 and FA9). When residents perceived they were failing to address identified priorities, belief in their ability for collective action would be diminished.
Yeah, I think they were very keen at the beginning and then, you know, there was nothing to keep their keenness going… and then because there was a lack of support, for want of a better word, and nothing, there was always meetings and nothing out of it. We started losing interest, which was a shame. (FA6-focus group discussion) In FA4 and FA8 memory of previous community initiatives, considered to have been ineffective, contributed to this sense of anxiety and collective cynicism. On the other hand, in FA2, representation on the partnership of individuals with experience of community and youth work allowed them to be pragmatic about the speed of change that was realistic in a community initiative.
By the end of 2015, hopefully, it would have really made an impact. I mean it's not through lack of contact, not through lack of publicity or marketing, but it takes time really. (FA2-interview-PM03) As Big Local progressed and new experiences accrued, the influence of previous experiences became increasingly diluted. Sometimes, positive experiences would reinforce residents' belief in their ability to bring about positive change.
He's an amazing volunteer and local resident who's just not only learnt to use IT but also took the Health and Safety Certificate on behalf of the partnership to enable us to do our Big Week [an Open Week which was seen as a turning point in terms of group confidence]. So that's amazing I don't think he ever had an ambition to do it but he took it and he said right I'll have a go and actually he's really proud of it and he's so proud of it he gave that as a quote and that's going to be in our magazine [for the whole area]. (FA10-interview-W02) Shared memories were important in shaping how context and initiative began to co-evolve through mutually reinforcing (positive and negative) feedback loops. They impacted on trust, speeding up or slowing down the process of collective action that Big Local put in train. They also influenced expectations, shaping the impact of ongoing experiences. In each area, we witnessed the creating of new memories and narratives through the recounting of stories, both celebratory and cautionary, about the Big Local journey so far. We found that people were keen to look back and see how far they had come. This process of creating new shared memories seemed to be important in moving them forward.

What does the systems lens add?
While the relevance of a systems orientated approach to public health has been recognised by researchers for at least two decades (Shiell & Hawe, 1996) there is little guidance on, and few worked examples of, how the concepts and theory can be applied practically (Sautkina et al., 2014). The findings presented here offer insights into how a systems approach, that foregrounds social context, can help reveal the changes and processes that might lead to enhanced collective action by communities, and how these might impact on social and health inequalities and their social determinants. They reveal which parts of the system are likely to be changed, by what mechanisms, and with what implications. For example, they reveal how fragmentations in the community can be linked to structural, spatial and social characteristics that influence perceptions of different sub-group identities and experiences. This complicates attempts to ensure that local initiatives represent the community and carries the risk that certain sections of the community, perhaps those that are already most marginalised, may not benefit from the initiative. They show how tensions emerge as residents (with or without the involvement of local organisations) struggle to agree on a shared vision for their local areas, which might contribute to burn out. On a more positive note, they reveal how residents in those areas that do have positive early experiences might feel increased self-esteem and this might reinforce action leading to a feedback loop that will maximise positive benefits over time. They also reveal that through the processes of the initiative, the residents involved are also starting to bring about small shifts, transforming the preexisting context into which the initiative is being delivered, with the effect of amplifying or dampening the potential for ongoing action for change.
The experience of applying systems thinking within this public health evaluation raises some salient points for knowledge generation and methods development in this field. By focusing more explicitly on context using a systems lens, this foregrounds attention to what usually is in the background of evaluations (Hawe, Bond, et al., 2009). During Big Local's early stages, pre-existing social conditions (collective memories and pre-existing decision-making structures, for example) strongly influenced what relationships and connections were made and the ways that the initiative is developing into different forms and at different speeds. Particularly relevant to area-based initiatives is how the initiative's geographical boundaries set at the local level interacted with social relationships within these and the extent that residents could form shared identities. During the fieldwork's timeframe, we also began to observe changes arising from perturbations in local neighbourhood systems (Petticrew, 2015). For example, local actors were beginning to define new ways of working, causing resultant shifts in behaviours, resources and connections. As Ogilvie et al. (2011) suggest, viewing the initiative as an event made the research team more alert to tracking the 'less immediately tangible actions' (p. 213) than a more conventional approach to conceptualising and evaluating an initiative would be able to do.

Opportunities and challenges
A key aspect of this study's originality lies not only in the use of a systems approach; but also in the way that a large corpus of ethnographic data has been brought together and analysed across a multitude of diverse settings, not as individual cases but as one large data-set that allowed greater purchase for understanding the system changes that Big Local was bringing about. For evaluation, this allows greater transferable insights into what might happen if a similar initiative were introduced into a new system. In-depth qualitative methods have been promoted as an approach to help capture complexity in community initiatives (Durie & Wyatt, 2013). However, there are few examples of ethnographic data being analysed through a systems lens. Within the fieldwork design, drawing on ethnography and deploying a team based across England enabled the capture of emergent and non-linear processes. Data generation methods such as long-term observation allowed the tracking of relationships unfolding over time in diverse spaces within neighbourhoods rather than relying only on conventional stakeholder interviews. At a practical level having researchers located in proximity to local fieldwork sites who could respond flexibly (attending meetings at short notice, for example) also meant that local developments could, to an extent, be tracked in real time.
A tension in our analysis emerged, however, in our attempts to balance attention to each individual local context (at each field site) whilst delivering an overall story about the initiative and the change mechanisms embedded within. In making interpretative statements it was important that we remained sensitive to the specific point each area was in their own Big Local journey. Further challenges arose in conducting ethnographic fieldwork where the phenomena of interest are not bounded in time and space and when certain aspects of the design (the duration of fieldwork and where it would take place, for example) were pre-determined. There were difficulties in setting the boundaries for research when each Big Local system was defined by the fuzziness of system boundaries and the importance of processes that were operating within the wider social, political, economic and cultural environment. We experienced a tension in how we represented local processes during a time of significant macro-level political and economic change. It was during our team deliberative sessions that we were able to explore temporal flows of relationships beyond the boundaries of our situated accounts. Even so, despite the long-term nature of our engagement at each field site (more than two years in some cases), our research still presents a snapshot of ongoing dynamic processes that are unfolding in a longer term initiative (of 10 or more years). Future studies might want to consider in more detail how we plug the gap between systems theory and research practice to better take account of interactions between the primary system being studied and the wider systems within which it is embedded.
Inevitably, as ethnographers, we were part of the field we were studying. Three particular issues illustrate this. First, residents were invited to help shape the research process in their local area and the way in which their local context was represented through this process. Second, hearing about findings during feedback sessions may have contributed to the empowerment of residents by offering them reassurance that they shared common challenges. Third, it is possible that the approach we took to negotiating access and working with the residents might have contributed to the shifting of power, allowing them to challenge our case for being there.

Conclusion
The use of a systems lens to analyse ethnographic data generated across a number of different sites and over a relatively lengthy period of time has enabled us to shed light on the ways that area-based community empowerment initiatives might embed within and co-evolve with social contexts. It reveals the unpredictable dynamic processes that emerge leading to, and away from, the development of collective action and, potentially, health inequalities. It also reveals the diversity of relations that need to be explored in order to understand the potential for longer term positive (or negative) impacts on social determinants and health inequalities. Future studies seeking to evaluate the impact of community-based social initiatives should pay close attention to pre-existing decision-making structures and processes, collective identities within diverse communities, the relationships between organisations and the residents they serve, and how these relationships are shaped by previous experiences and memories. The pre-existence of fragmented social groupings (based on divergent collective identities), in particular, has emerged in our analysis as an important factor that may potentially exacerbate inequalities when a social initiative is introduced. In public health, systems approaches may be particularly relevant for evaluating social initiatives over the longer term where change does not necessarily 'happen' in linear or predictable ways. They might help us understand what changes to expect in what situations and also to understand why things might not be moving in the expected direction. Note 1. Population sizes across the 150 BL areas range from approximately 1577-17,300, with an average of 6800.