Comparing Local Authority Rates of Children in Care: A Survey of the Children’s Social Care Workforce in Wales

The rate of children in care in Wales is one of the highest in the world and has increased considerably in the past two decades. Whilst many factors may be driving these increases, there is considerable variation between local authorities. This article presents findings from a survey completed by children’s social care workers in Wales ( n ¼ 792). It compares the views, values and responses to case study vignettes of workers in authorities with increasing to those with decreasing care rates over five years (2016–2020). Statistically significant differences were found relating to the values and the practices of workers, with workers in lo-cal authorities with reducing rates having stronger pro-family values, less risk averse responses to case vignettes, more confidence in the decisions made in their local authority and being more positive about support for practice. The findings indicate that variations in local authority values and practices may influence the rate of children in care and that some with significant social problems seem able to avoid the large numbers of children in care found in other authorities. The challenge faced in Wales and the UK is how local authorities can learn from one another to ensure consistency and quality in services.


Introduction
The care rate in Wales increased by 89 per cent between 2003 and 2021, with more than 1 per cent of all children currently in care (StatsWales, 2021).The rate of increase has been substantially higher in Wales than in England (Department for Education, 2021;StatsWales, 2021).There are also considerable variations in care rates between local authorities in Wales (Hodges, 2020).The focus of this article is to explore some of the factors that may be contributing to the variation in care rates across local authorities in Wales.
Why are care rates so high in Wales?
Care rates in Wales may be high for multiple and complicated reasons.High care rates are strongly connected to deprivation (Bywaters et al., 2020).A recent study in England (Bennett et al.,2 0 2 1 ) found that between 2015 and 2020, a 1 per cent increase in child poverty was associated with five additional children entering care per 100,000 and in total, it was found that only 8.1 per cent of care entries were linked to rising child poverty.
Another factor to consider is the recognition over recent decades of new forms of harm to children (Care Crisis Review, 2018).These include heightened recognition of issues already identified, such as parental substance misuse or domestic abuse, as well as new forms of harm that services were either not aware of or did not respond to as strongly, such as child sexual or criminal exploitation (Pearce, 2014;Robinson et al., 2019).Some additional factors to consider include the increase in care rates in the early 2000s in Wales and other UK nations, which has been linked to increased media, political and professional attention following high-profile deaths such as that of Peter Connolly and Victoria Climbie (Elliott, 2020).

Variations between local authorities
Whilst the overall increase in care rates is striking, it masks great variation in care rates between local authorities in Wales.For instance, a child in Torfaen is five times more likely to be in care than one in Carmarthenshire (StatsWales, 2021).Whilst some of the differences between local authorities are due to issues such as levels of deprivation, substantial variation still exists even when this is controlled for (Hodges, 2020).
The changes in care rates have also been different between local authorities.To illustrate the extremes, in the past five years, four local authorities have seen increases of 40 per cent or more, whilst two have seen decreases of more than 20 per cent (StatsWales, 2021).A study in England (Wijedasa et al.,2 0 1 8 ) found that economic and service quality factors were associated with decreases in local authority rates of children in care over a five-year period.These factors included: a decrease in the proportion of low-income families in the area, participation in the Department for Education's Innovation Programme and better OFSTED judgements.Furthermore, Bennett et al. (2020) found that unemployment was associated with rising rates of children in care.They found that for each percentage point increase in unemployment rate, an estimated additional 9 children per 100 000 per year became looked after the following year.
There is also a substantial area of research about how social workers make decisions (e.g.Christiansen and Anderssen, 2010;Bartelink et al., 2018;Helm, 2022), and theories about how variations in decision-making could influence variations in the rate of children in care (McGhee et al., 2018;Thomas, 2018;Mason et al., 2021).For instance, Spratt et al. (2015) found some evidence of confirmation bias in social worker decisionmaking in Northern Ireland.The research demonstrated the importance of social worker's priori assumptions which can cause them to selectively use case histories in ways which may confirm their own biases about families.Benbenishty et al.'s (2015) international comparison study in Israel, the Netherlands, Northern Ireland and Spain, found that the mother's wish regarding removal had no impact on judgements and decisions, yet social worker attitudes towards birth families, significantly influenced decision-making regarding the level of intervention recommended.With such high stakes, it is crucial that social worker decisionmaking is consistent and transparent (Holt and Kelly, 2014).
Whatever the underlying pressures are increasing rates of care, it is important to try to understand how some local authorities are resisting this.If we understand what some local authorities are doing to keep rates of children in care down, then it may be more possible for other local authorities to learn from them and make achievable changes, than for instance, responding to increased general concerns regarding risk or deprivation.
Using responses from a survey of the children's social care sector in Wales, this article attempts to identify what differences there may be between local authorities with reducing care rates compared with those with increasing rates over a five-year period (2016)(2017)(2018)(2019)(2020).

Are high care rates a cause for concern?
Whilst there is an assumption in recent policy commentary that care rates are too high and need to be reduced, some children need to be in care and some outcomes for children improve once they are in care (Rutter et al., 2007;Sinclair et al., 2019).Yet we want, as a society, to give families every opportunity and support to stay together.At most we can suggest that the rate should not be too high or too low, but that establishing the 'right' level is difficult (Cordis Bright, 2013).
Nonetheless, there is a growing concern about the substantial increase in the rate of children in care in Wales, as there is for England, which has seen a 22 per cent increase between 2003and 2021(Department for Children, Schools and Families, 2003;Care Crisis Review, 2018;Department for Education, 2021).This is partly about the harms that care can cause to children, families and communities, as well as its potentially limited capacity to reverse the effect of previous harms experienced with birth families (Greeson et al.,2 0 1 1 ).There is also evidence of long-term harm to parents once they have had children removed (Broadhurst and Mason, 2020;Griffiths et al.,2 0 2 1 ).Care is also a very expensive option (Elliott, 2017), and there are legitimate questions about whether it is always the most cost-effective way of helping children.
Whilst it is difficult to answer the question whether the care rate is 'right', one helpful source of evidence on this matter might be those working within the system.Do they feel the rate of children in care is right?And are there differences between local authorities in how confident practitioners are about this?This study seeks to answer such questions as part of its examination of differences between local authorities.It thus seeks to explore whether there are differences in values, practices and views between workers in local authorities with increasing or reducing care rates, and to investigate in which local authorities workers feel more confident about the decisions made and the support provided to enable good practice.

Method
The research uses self-report data from an online survey of the children's social care sector in Wales (see Supplementary Material S1 for a copy of the survey).The survey included closed and open-ended questions, although this article largely reports on the findings of the closed questions.Its main focus is the comparative analysis of differences in responses between local authorities with increasing and decreasing rates of care in the last five years.See the full report for further descriptive statistics and qualitative analysis of the survey (Forrester et al., 2021).

Ethics
Ethical permission was approved by Cardiff University's Research Ethics Committee.All respondents gave informed consent to take part in the study.

Survey development
The survey was developed with a steering group consisting of heads of children's services, Welsh Government officials and academics in Wales.Topics for inclusion were identified through a review of the literature, consultation with the steering group and meetings with a group of mothers with experience of children's social care.The questionnaire took approximately 15 min to complete and was available in both Welsh and English.

Participants
The online survey was circulated to heads of service children's services and Social Care Wales in November 2020, who shared it with their staff and network.In general, the survey was representative of the workforce, and the sample was large enough to allow considerable confidence in the findings.Data show that 82.3 per cent and 17.7 per cent of social workers in Wales are female and male, respectively (Social Care Wales, 2020).In this sample of qualified social workers, 84.5 per cent are female and 15.3 per cent male.Likewise, the age profile is also representative.Most qualified social workers in this study are aged between 35-44 (29.3 per cent) and 45-54 (26.2 per cent); the average age of a qualified social worker in Wales is 46.Where this sample differs slightly from the Welsh average is for ethnicity: 95.9 per cent of qualified social workers in this sample are white, compared with the national average of 88 per cent (Social Care Wales, 2020).The majority of respondents worked directly with children and families, for instance in child in need or child protection teams, with those less directly involved still playing key roles, such as in fostering or adoption services.
The local authority response rate was calculated by dividing the number of respondents in each local authority, by the number of children's social care workers employed in each local authority, available publicly in social workforce statistics (StatsWales, 2019).Respondents who described their role as 'administrative, other' and leaders (who are categorised separately by StatsWales) were excluded.The response rate ranged from 5.5 per cent to 58.0 per cent, meaning that some local authorities are more represented than others in the findings (see Figure 1 for the distribution of response rates).One local authority had a response rate of 2.2 per cent and was excluded from the comparative analysis due to concerns about representativeness.

Measures
Fifty-one closed questions, covering three key areas, are included in the analysis presented in this article.These were: (1) respondents' views on the factors influencing and driving the increase in care rates; (2) values relating to birth families and out-of-home care and (3) attitudes and risk culture in practice.One open-ended question asked which practice framework their local authority uses.
Respondents' views on the factors influencing and driving the increase in care rates These questions were scored from 1 ¼ strongly disagree to 5 ¼ strongly agree and covered views on: (1) key factors believed to be influencing the increase in care rates; (2) the influence of agencies outside children's social care and (3) children's social care culture and organisation.
Factor analysis was used to group questions that related to each other.A smaller number of factors allows for a simpler presentation of findings and a clearer comparative analysis.

Values relating to birth families and out-of-home care
This component was developed from the work of Davidson-Arad and Benbenishty (2010).Three value subscales were chosen and modified from this work based on: (1) keeping children at home, where possible and safe to do so; (2) positive views of foster care, which can be interpreted as out-of-home care in general and (3) favouring reunification and minimal time spent in out-of-home care (see Supplementary Material S2 for a breakdown of questions included under each subscale).In each of these areas, both positive and negative attitudes were included (reverse coding was used to create an index for each attitude).
Respondents were asked to indicate their agreement with each item on a five-point scale, from 1 ¼ strongly disagree to 5 ¼ strongly agree and the mean scores for each subscale were calculated.Cronbach's alpha was used to measure the internal consistency of the scale (Cronbach, 1951).The Cronbach's alpha score for subscale 1 was 0.6, for subscale 2 it was 0.6 and for subscale 3 it was 0.5.A scale reliability coefficient of 0.6 means the scale has an average internal consistency.A coefficient of 0.5 has adequate though relatively low internal consistency but is still a usable measure.

Attitudes and risk culture in practice
To explore whether there are differences in practice between local authorities in terms of risk culture and attitudes, two case study vignettes were developed.This is a difficult issue to research, as there are no existing measures.The vignettes were based on genuine cases and piloted with social workers in England to identify two vignettes that produced a range of responses.It was expected that differences in responses would be found between local authorities with increasing or reducing care rates.
Each case study had an immediate and a longer-term response.Responses to the case studies were ranked from 1-3 based on the level of risk appetite, with 3 being the most 'risk focused' (i.e.most likely to remove the child from home) and 1 being 'family focused'.Mean responses to the case studies were calculated per respondent and named their 'risk score'.Internal consistency of the risk scores across case studies was calculated using Cronbach's alpha (Cronbach, 1951).The combined risk score had relatively low, though acceptable, internal consistency (Cronbach's alpha ¼ 0.5).This means that an individual's answer to one question did tend to relate to their answer to other case study questions, but that the relationship was relatively weak.Responses to each case study, as well as the combined case study score, were therefore tested.Stronger differences were found in response to Case study 2 and therefore this was used in most analyses.See full report (Forrester et al.,2 0 2 1 ) for more information about how the case studies were developed.

Comparative analysis
Local authorities with increasing versus decreasing rates of care Comparing local authorities with increasing or decreasing rates of care was identified as a key interest topic by the steering group.Publicly accessible aggregate data on the rate of children in care per 10,000 by local authority were sourced from the Welsh Government's website (StatsWales, 2021).Local authorities with increasing rates of care, greater than the Welsh average (21 per cent) in the last five years, and those with a decrease greater than 10 per cent, were identified.This was to ensure increases and decreases were substantially enough to warrant comparison.
A risk with such a comparison is that local authorities with unusual rates of care five years ago (e.g.very high or very low) might be identified as 'increasing' or 'decreasing', whilst they are in fact just returning to the mean and thus it becomes a comparison of high or low deprivation.To account for this, we used a linear regression equation to calculate the care rate that might be expected based on the level of deprivation (The Welsh Index of deprivation, 2019, sourced from Berry, 2020) in each authority, and authorities that varied substantially (15 per cent in the opposite direction you would expect) from this were identified and excluded.The local authorities belonging to the two groups can be found in Table 1.

Factor analysis
Exploratory factor analysis (EFA) was used to understand the underlying structure of the data and reduce data dimensionality.Fourteen factors were uncovered in the data.The number of factors to retain initially was decided using a combination of the eigenvalue rule, scree test and parallel analysis.These suggested that the first eight factors should be retained.They explained 93.5 per cent of the variance in the data.
Results from the EFA were used to select parameters for the confirmatory factor analysis (CFA).CFA was then used to test the validity and reliability of factors and to determine factor loadings.Variables with factor loadings of over 0.4 were selected for the initial CFA model (Mehmetoglu and Jakobsen, 2017).The model was estimated using maximum likelihood.Factor/scale reliability was assessed using Raykov's (1997) reliability coefficient.Factors with reliability coefficients less than 0.7 were dropped from the model.Factor convergent and discriminant validity were assessed using STATA's command condisc.T h i sl e f tf o u rl a t e n t variables (see Table 3 for grouped questions).Model fit was tested using the Chi-squared (v 2 ) test, the standardised root mean-squared residual, root mean-squared error of approximation, the comparative fit index and the Tucker-Lewis index.The tests concluded that the model fit was good.Labels for the four factors were decided by the researchers based on the variables included, whether they made sense conceptually, and were informed by the qualitative analysis of the free text survey data.Estimated values of the latent variables were calculated using STATA's post estimation command Predict.To make the factor scores easier to interpret, they were rescaled to between 1 and 5 to reflect the scoring of the survey questions.
For the comparative analysis, initial bivariate analysis (t-tests and Chisquared tests) was conducted followed by logistic regression using STATA V.15 (StataCorp, 2017).

Results
Of the 792 respondents who completed the survey, 718 were workers and 74 were senior managers-approximately 18 per cent of the children's social care workforce (StatsWales, 2019).Of these, 584 were qualified social workers-approximately 34 per cent of those working in children's social care (StatsWales, 2019).
This section presents descriptive statistics about sector worker views on the rate of children in care; the findings from the factor analysis; the value subscales relating to birth families and out-of-home care and the case study vignettes exploring attitudes and risk culture in practice.It will then compare these measures between local authorities with increasing and decreasing rates of children in care.

Views on the rate of children in care
Respondents were asked a closed question about whether they felt the overall rate of children in care in Wales, and in their local authority, was too high or too low (Table 2).Of those that responded, most felt that the rate of children in care in Wales and in their local authority was too high or much too high (67.9 per cent, n ¼ 473 and 56.7 per cent, n ¼ 394, respectively).Respondents were more likely to think that the care rate in Wales was too high than in their own local authority, with four in ten believing that the rate was 'about right' in their authority (39.6 per cent) and more than one-quarter thinking the same for Wales as a whole (28.6 per cent).Very few respondents thought the care rate was too low in either Wales or their local authority.
A factor analysis: views on the factors influencing and driving the increase in care rates The survey asked multiple Likert scale questions about what could be influencing the rate of children in care.To simplify the results, a factor analysis was conducted (see Forrester et al. (2021) for descriptive statistics for all survey questions).The factor analysis led to the identification of four factors (see Table 3 for factor loadings and included survey questions), and to the creation of a 'score' ranging from 1 to 5 for each: 1. Confidence in local authority: A higher score means that respondents are more confident that their local authority keeps children safe and, where possible, at home (M ¼ 4.2, SD ¼ 0.6).2. Cuts to services: A higher score means that respondents place more emphasis on cuts to services affecting whether a child goes into care (M ¼ 3.9, SD ¼ 0.9).3. Support for practice: A higher score means that respondents are more likely to believe that their local authority has working practices in place to support the local authority's vision of practice (M ¼ 4.0, SD ¼ 0.6).4. Influence of community agencies: A higher score means that workers place more emphasis on community agencies such as police, schools and health services as factors influencing the rate of children in care in their area (M ¼ 3.4, SD ¼ 0.8).In addition, there were three pairs of questions that were strongly correlated with one another and seemed to be conceptually related but could not be tested statistically as factors as they only related to two questions.A factor must have at least three variables which are strongly correlated to be included in the model (Rahn, 2021).Therefore, three new variables were created by calculating the mean score for each of the three pairs.For ease of presentation, we refer to these as factors: 5.The influence of courts: A higher score means that respondents place more emphasis on the influence of courts and local authority solicitors on care rates in their local authority (M ¼ 4.0, SD ¼ 0.8).6.The influence of workers and managers: A higher score means that respondents place more emphasis on workers and managers as influencing care rates in their local authority (M ¼ 4.2, SD ¼ 0.7).7. Parental factors: A higher score means that respondents are more likely to identify parental alcohol, drugs or domestic abuse as increasing care rates (M ¼ 4.0, SD ¼ 0.7).

Value subscales: values relating to birth families and out-of-home care
The mean score across the whole sample for subscale 1-keeping children at home, where possible and safe to do so-was 3.1 (SD ¼ 0.6).Scores range from 1 to 5, the higher the score, the more strongly a respondent associates with that value.For subscale 2-positive views of foster care, which can be interpreted as out-of-home care in general-the mean score was 2.3 (SD ¼ 0.5) and for subscale 3-favouring reunification and minimal time spent in out-of-home care-the mean score was 3.0 (SD ¼ 0.7).The results suggest, in general, that respondents were slightly more in favour of keeping children with their families and away from out-of-home care.However, the purpose of this element of the study is mainly comparative-to see whether there are differences between local authorities in their values.

Case study vignettes: Attitudes and risk culture in practice
In In response to Case study 2, most participants chose to immediately 'Start protective action whilst trying to engage Joanne in protecting Dawn' (51.0 per cent, n ¼ 382), and over the next 6 months to 'Work with Dawn and Joanne in the community' (75.3 per cent, n ¼ 563).The average risk score for Case study 2 was 1.8 (SD ¼ 0.5).The combined mean risk score for both case studies was 1.8 (SD ¼ 0.5).
For a more detailed analysis of the case studies, see the full report (Forrester et al., 2021).

Comparing the responses of local authorities with increasing or reducing care rates
Using the measures listed above, bivariate statistics tested for differences between responses from local authorities with increasing versus decreasing rates of care over the past five years.There were no significant differences found between the two groups in the following values: positive views of foster care (subscale 2) or favouring reunification (subscale 3).There were also no significant differences in the following views as influencing and driving the increase in care rates: cuts to services (factor 2); influence of community agencies (factor 4); influence of the courts (factor 5); influence of workers and managers (factor 6) or parental factors (factor 7).
Where significant differences were found, a multiple logistic regression model tested these associations further.Respondents from local authorities with increasing care rates were less likely to be against removing a child at risk from home (values subscale 1); more risk averse in their decision-making (Case study 2 risk score) and less confident that their local authority keeps children safe and where possible, at home (factor 1).The demographic characteristics-age and whether they were a qualified social worker-were initially controlled in the model, but these made very little difference and were therefore excluded.See Table 4 for odds ratios.
Respondents from local authorities with increasing rates were also less likely to feel that their local authority had the procedures in place to support the local authority's vision of practice, such as suitable training (factor 3), however due to collinearity with factor 1, this variable could not be included in the regression model.When factor 1 was replaced with factor 3, the odds ratio for factor 3 was 0.51 (95 per cent CI: 0.31-0.84).

Differences in practice framework use
Practice frameworks have become increasingly popular in recent years in the UK.They include both whole-system approaches such as signs of safety or restorative practice and models of practice applied within specific teams or services.It is also possible for an individual to have a specific framework they use.However, in general, the way that practice frameworks are understood is as a whole-system commitment to particular ways of working.
The degree to which practice frameworks are used seemed inconsistent, with workers from the same local authority often giving different responses.However, sufficient respondents identified signs of safety and the outcomes framework as the practice framework they use, to allow comparison of local authorities with increasing and decreasing care rates.Signs of safety is a well-known model for child and family social work based on solution focused principles (Turnell and Edwards, 1999).The outcomes framework is a collaborative approach to agreeing outcomes with families based on the principles of the 2014 Social Services and Well-Being Act (Wales).Chi-squared tests showed that 31.4 per cent (n ¼ 22) of respondents from local authorities with decreasing care rates stated they used the outcomes framework compared with 2.9 per cent (n ¼ 12) in local authorities with increasing rates, and this difference was significant (v 2 (1, N ¼ 473) ¼ 72.4,p < 0.001).There were no significant differences between the groups for signs of safety, with 15.3 per cent (n ¼ 63) stating they use it in the increasing care rate group and 17.1 per cent (n ¼ 12) stating they use it in the decreasing group (v 2 (1, N ¼ 473) ¼ 0.1, p ¼ 0.750).It is worth noting that these results are based on data collected from an open question and 55.0 per cent of the data were missing.

Discussion
This article reports findings from a survey of the children's social care sector in Wales.There was a high level of agreement amongst children's social care workers that care rates are too high.Respondents felt this was true for both the country as a whole and for their local authority.For those working in children's social care, the drivers of increasing care rates are multifaceted, encompassing both what is happening for the families that they work with and the public service response to this.
The article compares the views, values and attitudes of respondents from local authorities with increasing and decreasing rates of children in care.Compared with respondents from local authorities with increasing care rates, respondents from local authorities with decreasing rates were more confident that their local authority keeps children safe and, where possible, at home; had values that were more positive about birth families; were less risk averse in response to case studies; were more likely to feel that their local authority had the procedures in place to support the local authority's vision of practice, such as training and supervision and were more likely to use the outcomes framework.
This suggests that the difference in care rates between local authorities over time is likely to be due, at least in part, to differences in the practices of children's social care.It also suggests that the other factors identified by respondents as influencing care rates (e.g.increased need) may be issues for Wales as a whole, but are unlikely to explain the differences between local authorities.
This study adds to the debate about the increasing rates of children in care in the UK and the disparities in rates of children in care between the nations (Featherstone et al., 2019;Bywaters et al., 2020).It attempted to fill the gaps found in other literature about what is contributing to the differences in rates of care between local authorities that cannot be explained by deprivation (Hodges, 2020).It supports the hypothesis of others that differences in care rates are due in part to differences in operational procedures and practices (McGhee et al., 2018;Thomas, 2018;Mason et al., 2021).

Implications for social work
A key finding was that the local authorities with reducing care rates did not feel children were being left at risk.Workers in local authorities with reducing care rates were more confident that their local authority was getting decisions right and protecting children who needed it, whilst preventing others from entering care.This suggests that from the perspective of workers it is possible for care rates to be reduced without children being left at inappropriate levels of risk or suffering harm.On the other hand, in those local authorities with increasing care rates, workers and managers are more likely to feel that there are children in care, who should not be.This supports the argument that local authorities might be safely able to reduce the rate of children in care (Westlake et al., 2018;Guggenheim, 2021).
The results provide some indicators of how this might be achievable.First, respondents in local authorities with reducing care rates identified better support for practice.It seems likely that better support for good practice within an authority keeps more children at home.Second, there was some evidence that this was related to the use of practice frameworks.In particular, the development of an outcomes framework in one of the local authorities with reducing care rates was a feature of the practice there.Interestingly, the other local authority with reducing care rates also used a practice framework, based on systemic practice (Forrester et al., 2013).The fact that signs of safety, a licenced model with a solution-focused approach to child protection, did not seem to be associated with reduced care rates perhaps suggests that practice frameworks are not a simple panacea for improving and supporting practice.Indeed, developing the outcomes framework required the local authority to put in the hard work to develop their own tailored framework and it is very possible that this is what made the framework 'work', by creating the energy and sense of ownership needed.
It was interesting to observe differences in responses to the case study practice vignettes.This provides suggestive evidence that there may be actual differences in how difficult decisions are made in local authorities, though further research would be needed to validate such measures.
Finally, there were significant differences in the values found in local authorities with reducing care rates.Respondents from these local authorities were more likely to value the importance of keeping children with birth families.This points to an important but possibly challenging finding.Key issues for children's social care such as influencing care rates may not be primarily about evidence-it is possible that values-based practice is more important (Bisman, 2004).In other words, perhaps the values found across a local authority are crucial in explaining differences between local authority care rates that are not explained by socioeconomic differences.Yet this begs another question-which is how to create and shape values?The findings from this study cannot answer this question, though the importance of support for practice and the potential contribution of practice frameworks may be important.Training and supervision that enhances practitioners' awareness of the impact that their attitudes, and the context in which they are embedded, have on their judgements and decisions could be beneficial (Benbenishty et al., 2015;Bartelink et al., 2018).
In recent years, a number of practice frameworks have been developed, with some indication that some of these frameworks may reduce care rates.For instance, systemic practice (Forrester et al., 2013), family safeguarding using motivational interviewing (Forrester et al., 2017;Rodger et al., 2020) and restorative practice (Mason et al., 2017) were all implemented in local authorities that reduced the rates of children in care.It may be that a common feature of these frameworks is a shared set of values that focus on working with families and supporting social workers to keep children at home, if possible.

Strengths and limitations of the research
Collecting data through a relatively brief online survey allowed for views to be obtained from a large number of children's social care workers.The number of responses and the response rate, at nearly one-fifth of the children's social care workforce and more than one-third of qualified social workers in the sector in Wales, makes this one of the largest surveys of people working in social care undertaken in the UK.
Yet the ability of the survey to obtain a large number of responses is related to one of its main limitations.Many of the responses were precoded which limited more in-depth analysis.Furthermore, most of the measures used were exploratory; there are not standardised measures for many of the issues of interest.For instance, this is the first time that responses to case studies have been used in this way (as far as the authors are aware).These were partially successful in identifying differences in decision-making, but would benefit from further research to establish validity.
There are also challenges in knowing which comparison groups or outcome measures to choose for the analysis.There are likely to be many variables influencing care rates.Therefore, unpicking the specific contribution of, for instance, staff values is difficult because many other factors will be influencing care rates.The challenge is identifying true signals amidst the cacophony of other noises.Furthermore, there were only two local authorities with decreases in care rates significant enough to warrant inclusion in the decreasing rates of care group.It is possible that there could be key factors unique to these settings that have not been measured and could be influencing the findings.

Conclusions
There are many factors influencing care rates and they interact in complicated ways that we do not understand well (Care Crisis Review, 2018).The study found important similarities and differences between local authorities with decreasing care rates compared with those with increases.The ratings of pressures on families and services were very similar, whilst there were significant differences in factors that were related to the internal workings of children's social care.This included satisfaction with elements of the service and the values and practice decisions of respondents.This suggests that although all local authorities experience multiple pressures that might lead children to be in care, they respond to them in different ways.Some local authorities seem able to mitigate some of the factors that drive increasing care rates.
A key issue arising from this is about equity.In Wales, a child in Torfaen is five times more likely to be in care than one in Carmarthenshire and there are other variations that do not seem to be wholly explained by differences in deprivation.The study findings suggest that some of this difference is about different values and practices within local authorities.The challenge faced in Wales and the UK is how local authorities can learn from one another to ensure consistency and quality in services.

Figure 1 :
Figure 1: Box plot of local authority response rates.

Table 1 .
Local authorities with increasing and decreasing rates of care selected for comparison a One local authority excluded owing to 15 per cent lower than predicted care rate.

Table 2 .
Responses to the question 'Overall the rate of children in care is ...'

Table 3 .
Standardised factor loadings and indicator reliability aThe amount of the variance in an indicator explained by the latent variable.***All factor loadings p 0.001.
response to Case study 1 (see Supplementary Material S3 for case study descriptions), most respondents chose to immediately 'Work with family on a child protection basis with the intention of getting them to understand your concerns' (62.1 per cent, n ¼ 488), and over the next 6 months to either 'work with the family in the community' (43.2 per cent, n ¼ 339) or 'Work with the family whilst Anna is in alternative care with a view to returning her home' (40.6 per cent, n ¼ 319).The average risk score for Case study 1 was 1.7 (SD ¼ 0.8).

Table 4 .
The odds of being a respondent from a local authority with increasing versus decreasing care rates over the last 5 years Comparing Local Authority Rates of Children in Care Page 13 of 21 Downloaded from https://academic.oup.com/bjsw/advance-article/doi/10.1093/bjsw/bcad097/7080092 by Cardiff University user on 15 August 2023