Risk perceptions and experience in child protection decision-making: a comparative study of student social workers in Wales and Aotearoa New Zealand

Decision making in child protection is the product of interacting factors between workers, organizations, families, and macro social structures. Individual perceptions of risk, safety, and harm, as one piece of this complex puzzle, are important to understand. This article reports on a comparative study of social work students in two countries: Wales and Aotearoa New Zealand. Using a mixed methods survey and a staged vignette (in which the situation becomes progressively more serious), we found there were similarities between respondents from the two countries in their perceptions of risk, safety and harm, and their reasoning processes. Beneath this broad consensus, respondents from Wales rated the level of harm to the children lower at earlier stages but were more likely to say the case should meet the threshold for statutory intervention. Risk-averse respondents were more likely to conclude the children experienced serious harm and also that the case should meet the threshold for statutory intervention. These differences largely disappeared by the concluding stage. Qualitative analysis shows that the reasoning processes used to explain risk, safety, and plan goals were similar between the two countries. Some nuanced differences emerged in relation to a risk-averse group from Aotearoa New Zealand emphasizing the importance of continuing engagement with professional services as a sign of change. Implications are discussed, particularly for workforce development and the needs of newly qualified social workers.


Introduction
Making judgements and decisions is a core part of social work practice (Taylor & Whittaker, 2018;Taylor, 2017).These judgements and decisions can have significant and long-lasting consequences for the children and families involved.Making effective judgements and decisions is essential for ensuring that family needs are met, and children protected from abuse and neglect.Social work decisions should be based on sound judgements, accurate assessments, and a critical analysis of the situation (Author 1 2013), while conforming with relevant legal and practice requirements (Ebsen et al., 2023).Given the values of the global social work profession, they should also be made fairly and consistently, in the interests of social justice, whilst accounting for the individual needs of children, their parents and carers (Sewpaul, 2013).Yet the quality of social work judgement and decision-making has been criticized in many jurisdictions around the world (Author, 2022;Author, 2016;Munro, 1999Munro, , 2011)).For example, studies have found decision-making is variable in relation to similar family situations, reflecting elements such as differences in values or experience of the decision-maker (Fluke et al., 2018;Author 1, 2018;Author 2, 2017).Decision-making is also a contributing factor to socioeconomic and ethnic disparities in rates of contact with the child protection systems in both the UK and Aotearoa New Zealand (Bywaters et al., 2015;Dickens et al., 2007;Author, 2019a, b).
When assessing the quality of social work judgements and decisions, it is important to recognize the various evaluative criteria that may be used (Hood et al., 2022).These include i) accuracy (the extent to which judgements and decisions are corroborated by external knowledge and empirical events), ii) consistency (the extent to which different professionals make similar judgements and decisions in relation to similar cases), iii) outcomes (the extent to which judgements and decisions result in positive improvements for children and families), iv) the principles of good practice (the extent to which judgements and decisions are made in accordance with normative and professional standards), and v) equity (the extent to which people from different socio-demographic groups are treated similarly and fairly).It is equally important to recognize that judgements and decisions result from a complicated interaction of personal, case-related, external, and organizational factors (Baumann et al., 2014;Fluke et al., 2020;Lauritzen et al., 2018).For example, individual practitioners are influenced and constrained by their organizational context, including thresholds, demand and supply issues, institutional cultures and social 'sense-making' processes, as well as the preferences of managers within hierarchical bureaucracies (Baumann et al., 2014;Hood et al., 2019;Author, 2019;Platt & Turney, 2014).The organizational context contains heuristics based on pattern recognition that serve to embed certain types of responses to certain types of cases, as well as group decision-making processes and specific assessment tools that further affect decision outcomes (Gore et al., 2018;Platt & Turney, 2014).Factors external to the organization, such as macro level policies, also shape judgment and decision outcomes.For example, a national orientation towards family support or child protection changes the assumed aims of the system, and with it, reasoning rationales, thresholds and the pattern of resource distribution within a country.This affects the framing of decisions and the scope of what is considered possible, desirable and acceptable (Gilbert et al., 2011;Križ & Skivenes, 2013;Skivenes & Thoburn, 2017).The individual social worker operates within a complex ecology, yet the views and experiences of the individual are themselves important factors to consider.Perceptions of risk, safety and harm are intrinsic to social work judgements and decisions, especially in the context of child protection.
It should be no surprise that decision outcomes are variable, even when case characteristics are held constant or made similar via vignette-based research (Fluke et al., 2016;Author, 2014;Regehr et al., 2010).This inevitable feature of complex processes represents a serious social justice issue.While it is challenging to balance competing principles in response to diverse circumstances, there should be a priori a basic level of consistency, irrespective of decision-maker, ethnicity, geography (at least within the same country) and socioeconomic circumstances (Enosh & Bayer-Topilsky, 2014;Author, 2014).
Within this context, the judgement of the individual social worker is affected by how they perceive the behaviour, situation and presentation of families reported to child protection services, particularly how these elements are codified as indicating risk, harm, or safety.The interplay between these concepts, information about the family and the mechanisms of information-gathering have all received attention within the literature.Specifically, the influence of personal and professional values (Taylor, 2017;Taylor & White, 2006), levels of experience, and attitudes towards family preservation relative to child safety affect perceptions of risk (Horwath, 2007;Fluke et al., 2016).For example, those with more positive attitudes towards foster care are more likely to recommend it, independently of case characteristics (Benbenishty et al., 2016).
Attitudes towards risk also interact with theoretical concepts used to interpret behaviour.Author (2017) found that practitioners who were more risk-averse in their initial judgement were more likely to draw on trauma-related concepts to construct a plausible vision of future significant harm to the child.This linkage was important to the justificatory logic that led to recommendations for further intervention.On the other hand, practitioners with less risk-averse initial judgements (those who were more 'risk-friendly') were likely to emphasize the current social needs of the family.Similarly, Kriz and Skivenes (2013) found that when presented with the same vignette, social workers from different parts of the world had different perceptions of risk.Those with more protectionist systems of child welfare (e.g., California) rated risk lower than those with more supportive orientations (e.g., Norway).The authors argued that this was due to a higher risk threshold for intervention in protectionist systems, where preventive work is less emphasized or available.Ethnographic studies of decision-making in social work have also shown how casework judgements are often framed in certain ways, for example that the case is 'worrying' or 'a child protection case'.This initial framing informs subsequent interpretations of new information, which could be an example of anchoring or first impression bias (Helm, 2016).Pithouse et al (2012) found that this type of labelling typically happens very early on in the decision-making process, when information is relatively limited.
Studies have also found that more experienced workers tend to have lower perceptions of risk, compared to less experienced workers.This generalization includes experienced professionals compared with students (Fleming et al., 2015).Fluke (2016) found that staff with more experience tend to emphasize family preservation over immediate child safety but were also more likely to be in managerial positions, and no longer directly responsible for their own case load.Less experienced and caseload carrying workers tend to emphasize child safety (Fluke et al., 2020).Alternatively, de Haan et al. (2019) reported little difference between case recommendations made by student social workers and qualified workers (with de facto more experience).They found instead that in relation to decisions about child removal, the worker's beliefs about the parent's ability to change and attitudes towards out-of-home placements were associated with different types of decisions.
Previous studies, mostly involving qualified practitioners, sometimes in comparison with student social workers, have found that initial perceptions of risk (e.g., risk-aversion vs risk-friendly) may be formed quickly, based on limited information, following a notification (or referral) to child welfare servicesand that these initial judgements influence subsequent assessments and decision recommendations (Križ & Roundtree-Swain, 2017).However, we do not know whether the same thing might hold true for groups of social work students, and whether different levels of pre-qualifying social care experience make any difference.In this study, we compared the responses of student social workers from Wales and Aotearoa New Zealand (Aotearoa New Zealand) to the same unfolding case vignette, to explore differences and similarities based on i) country of origin, ii) levels of existing social care experience and iii) initial perceptions of risk.Our research questions were as follows: 1. Are initial perceptions of risk in relation to a child protection case vignette associated with respondents' country of origin?
2. Are initial perceptions of risk associate with respondents' levels of existing social care experience? 3. Are initial perceptions of risk associated with subsequent identification of risk and safety factors?4. Are initial perceptions of risk, country of origin or levels of existing social care related to respondents' identification of the main aim of intervention with the family, or what they recommend for the children's care plan?

Methods
We completed a study of social work judgement and decision-making, by recruiting students from social work qualifying programmes in Wales and Aotearoa New Zealand.We chose these countries for pragmatic reasons because they are the home locations of the two authors and gave us ready access to populations of social work students.However, they also represent broadly similar historic 'child protection' orientations as per those described above, though with significant differences due to historical and cultural differences, and the layering of policy reforms over time (see Hyslop, 2022 for a discussion of this in Aotearoa, and x for a Welsh discussion).Once recruited, we asked respondents to complete a mixed-methods survey, based on an unfolding case vignette ( The school report that the mother has been seen hitting the children in the playground outside.The children tell their teacher they are hit regularly at home, including with objects.They say their mother is sad and cries all the time.They also talk about visiting their dad and having fun.

Development of the vignette
The case vignette was a replication of one used previously in another study in Aotearoa New Zealand (Author).The vignette was written by author 2 and a colleague, based on their experience with multiple cases from research and practice.It was then evaluated by a focus group of currently practising child welfare social workers.They were asked to ascertain its authenticity, recognisability as 'familiar' or 'realistic'; and that it was likely to hit most practitioners' thresholds for various initial decisions on the child protection decision-making continuum.Minor changes were made following this process.The vignette contains information that can be interpreted as child maltreatment, although no details that are likely to be distressing for those with social care or social work experience.

Sampling
Invitations to take part in the study were advertised to social work students on qualifying programmes in Wales and Aotearoa New Zealand.In Wales, this included a mixture of BA and MA-level courses (n=4), involving approximately 100 students per cohort (total approximately 400).Social work qualifying programmes in Wales are generalist in nature (they do not focus on specific areas of practice, such as child protection).Twenty-five students participated in the study.In Aotearoa New Zealand the survey was circulated to all social work programmes (n=17) via the national social work education committee and through personal emails.There are a mixture of university and polytechnic programmes currently operating, and the survey was aimed at years three and four BSW students and applied MSW students.At least five of those programmes circulated it to their qualifying students, including programmes from both sectors.Numbers range between programmes from approximately 40 to 120 per cohort, so the survey was circulated to a minimum of approximately 400 students.Before taking the survey, respondents had to self-identify as a social work student, read an information sheet, and complete a consent form.In total, 81 respondents started the survey, and 65 completed at least half of the questions, representing 16% of the sampled group (meeting our threshold for inclusion in the study).

Ethics
Participation in the study was voluntary and at any point respondents were able to exit the study by closing their internet browser; however, once they had answered 50% of the questions, their data were retained for use in the study.Ethical approval was independently granted by the relevant ethics committees at both authors' universities.

Quantitative data analysis
Data from the survey were downloaded into Excel (version 16) and SPSS (version 25) for descriptive analysis.Following a similar process to that used in a previous study in Aotearoa New Zealand (Author, 2016), we categorized respondents as being either 'riskaverse' or 'risk-friendly', based on their initial response to the first stage of the vignette.We then used an independent samples t-test for stage 1 of the vignette to see whether respondents' ratings of risk were significantly different based on country of origin or existing levels of social care experience (research questions 1 and 2).We then used a series of paired sample t-test to see whether initial ratings of risk at stage one were related to subsequent riskratings in stages two, three and four (research question 3).

Qualitative data analysis
Qualitative data analysis consisted of separating the open-text responses into four groups: risk-averse from Aotearoa New Zealand, risk-averse from Wales, risk-friendly from Aotearoa New Zealand and risk-averse from Wales.Responses to the questions were analyzed via risk-averse and risk-friendly groups in each country and organized into deductive codes created by the question format, that is: risk factors, safety factors, problem causes, plan goals, essential changes and intervention aims (research questions 3 and 4).Each risk-averse and risk-friendly group was then subject to content analysis (Kleinheksel et al., 2020) and thematic analysis to identify prominent themes firstly through simple counts of responses, then via a deeper reading for overall patterns of more abstract themes (Vaismoradi et al., 2016).Further coding was undertaken to identify differences by nationality within each group (risk-averse vs. risk-friendly).

Sample description
In total, 65 respondents completed ≥ 50% of the survey, and were included in the analysis (  Of those who did not complete the survey in full, 11 were female (84.6%), 1 male (and 1 did not provide a response to the question about gender); 5 were from Wales (38.5%) and 8 from Aotearoa New Zealand (61.5%); 5 had no previous experience in social care (38.5%), 3 had a low level of previous experience (23.1%), and 4 had a moderate level of previous experience (30.8%), and 1 did not answer the question about previous social care experience.These proportions are all broadly comparable to the overall nature of the sample, and thus we judge that this level of attrition would not have significantly altered the final results.

Categorizing respondents
We categorized respondents based on their initial judgement of risk, their levels of previous social care experience, and country of origin (table 4).Using their response to the first question in the surveybased on this information, how would you rate the level of risk of harm to the children?-respondents were categorized as 'risk-friendly' (those who responded a little risk, or somewhat risky) or 'risk-averse' (those who responded substantial risk or high risk).None of the respondents selected no risk (such that there was an equal chance for each respondent of being included in either group).Most respondents were categorized as risk-averse.Respondents from Aotearoa New Zealand were more likely to be categorized as risk-averse, compared to those from Wales.
In relation to experience, we grouped together respondents with no or low social care experience (between zero and two years) and those with moderate or high social care experience (three or more years).Overall, most respondents were less experienced.

Overall responses to the case vignette
We analyzed responses to survey questions in relation to country of origin and riskaversion vs risk-friendliness.(The questions and possible responses can be seen in table 2, above.)For all these, a higher number on the Likert scale indicates a higher level of concern (e.g., about the level of risk) or increased certainty (e.g., about the presence of abuse or neglect, or that the case should meet the threshold for intervention).In relation to safety, the higher the rating, the less safe the respondent considered the child to be.Overall, as you would expect given the vignette design, levels of concern increased as the survey progressed (figure 1).(Respondents were not asked about how unsafe the children were as part of stage one.) Figure 1.The overall proportion of respondents at each stage of the vignette who said the risk was high / moderate, the children were unsafe, the children were being abused or neglected, and that the case should meet the threshold for statutory intervention.

Are initial perceptions of risk associated with respondents' country of origin?
Comparing the two countries, a greater proportion of respondents from Aotearoa New Zealand than Wales rated the level of harm to be higher at the outset of the vignette, but this was reversed by stage 4 (figures 2 and 3).At stage one, there was a significant difference between respondents based on country (Wales vs Aotearoa New Zealand) in relation to the question about harm (Wales; M=3.58, SD=.717.Aotearoa New Zealand; M=3.73, SD=.679), t (62)=2.174,p=.033 (table x).However, there were no significant differences in relation to the other questions about risk, abuse or neglect and thresholds.Thus, the answer to research question 1 -are initial perceptions of risk in relation to a child protection case vignette associated with respondents' country of origin?-is no.Are initial perceptions of risk associate with respondents' levels of existing social care experience?
Comparing between respondents with less social care experience (between zero and two years) and those with more social care experience (three plus years), a higher proportion of the more experienced group had lower initial perceptions of risk, harm, safety and whether the children were being abused or neglected.Perceptions of whether the notification (or referral) should meet the threshold for statutory intervention were similar throughout (figures 4 and 5).% agreement that the children are not safe.

NZ Wales
At stage one, for the question about risk, there was a significant difference between respondents with less social care experience (M=3.87,SD=.615) and those with more social experience respondents (M=3.36,SD=.700); t(62)=3.078,p=.003 (table x).There were no significant differences at stage one in relation to the other questions (about harm, abuse or neglect and thresholds).Thus, the answer to research question 2 -are initial perceptions of risk associate with respondents' levels of existing social care experience?-is also yes.

Are initial perceptions of risk in relation to a child protection case vignette associated with subsequent identification of risk and safety factors?
When comparing between respondents categorised as risk-averse or risk-friendly, as you would expect, a greater proportion of those in the former category gave consistently higher ratings in relation to safety, harm, abuse or neglect, and thresholduntil stage 4, when they largely converged (figures 6 and 7).
Figures 6 and 7.The proportion of respondents categorised as 'risk-averse' and 'riskfriendly' who rated the risk as moderate or high, and whether they met the threshold for statutory intervention, at different stages of the vignette.% agreement that the children are not safe.

Less experienced More experienced <Paired samples t-test here>
A series of paired t-tests were conducted to determine if initial ratings of risk at stage one of the vignette resulted in different ratings of risk at stages two, three and four.Results for the comparison between stages one and two showed that the mean risk rating was not statistically significantly different between the stages (t = -1.240,df=62, p = .220)at a significance level of 0.05 (  Results for the comparison between stages one and four showed that the mean risk rating was statistically significantly different between the stages (t = -9.613,df=54, p = <.001) at a significance level of 0.05.A 95% confidence interval for the true difference in population means resulted in the interval of (-1.187, -.777) Thus, the answer to the first part of research question 3 (Are initial perceptions of risk associated with subsequent identification of risk?) is primarily yes (for stages three and four, but not stage two).
Qualitative content analysis was also undertaken to examine differences in risk and safety perceptions between the risk-averse and risk-friendly groups in each country, especially in the early stages of the vignette, where perceptions differed the most.At stage one, the most prominent themes were similar between the risk-averse and risk-friendly groups, through emphases differed.In terms of what respondents felt indicated 'safety factors', the most prominent themes were: that Max (one of the children in the vignette) was more settled in school; all the children in the family were attending school regularly; that school staff were aware of police involvement; and that the father had left the family home.Beneath this broad consensus, some findings differed between the risk-averse and riskfriendly groups.The risk-averse group from both countries emphasized police involvement as a sign of safety, compared to the risk-friendly groups.Respondents from Aotearoa New Zealand tended to want more information before forming a judgement about safety, compared to those from Wales.
At stage two, where risk perceptions were first explored in the survey using openended questions, the findings were also mixed.At this stage, all groups identified the parent's own traumatic backgrounds as a moderate to strong theme, the third most noted risk factor in each group.The risk-averse group (both in Wales and Aotearoa New Zealand) had high consensus around the most important risk factors, namely financial stress and poverty, and exposure of the children to intimate partner violence (IPV).Other similarities in this group included intergenerational trauma (this held true across all four groups) general stress and exhaustion and the presence of sibling violence.However, the Aotearoa New Zealand riskaverse group identified a lack of community and family support as a key theme and noted the unsafe community neighborhood as moderately important, whereas this was not so for riskaverse Welsh respondents, who did not mention the neighborhood at all.
In the risk-friendly group (both Wales and Aotearoa New Zealand), responses were more diverse, with responses spread across multiple risk categories compared to the riskaverse group.This may be an artefact of the group having fewer members overall, with less chance of saturation than the risk-averse group (which was larger).The most emphasized risk factors overall were poverty and financial stress, lack of family and community supports, and the parents' traumatic childhoods.However, only the Aotearoa New Zealand respondents noted the exposure to IPV as an important risk factor, something scarcely mentioned at all by Welsh respondents; the latter did however note the unsafe neighborhood, which was not mentioned at all by the Aotearoa New Zealand risk-friendly respondents.
These findings show a high degree of consensus between the risk-averse and friendly groups relating to the importance of financial stress and poverty as a risk factor, but with diverse nuances for all other themes.The most notable difference was that the risk-averse group emphasized exposure to IPV, while the risk-friendly group emphasized lack of family and community support.

Are initial perceptions of risk, country of origin or levels of existing social care related to respondents' identification of the main aim of intervention with the family, or what they recommend for the children's care plan?
As well as considering indicators of risk, safety, harm, and abuse or neglect, respondents were asked what actions they would recommend at each stage.Respondents were asked to select from a list (table 5), and to use a free-text box to note any additional actions ('other').In addition, actions were evaluated through several qualitative questions that asked respondents about their goals for a family plan, intervention aims, and the essential changes they perceived as important for the family or their situation.In terms of plan goals, there was a high degree of consensus between all four groups (Wales / Aotearoa New Zealand, risk-averse / risk-friendly) that the most important goal was to support Shannon's (the mother's) mental health.All groups also focused on ending or stopping the IPV, though there were differences in emphasis, with some framing this as Dan (the father) needing to seek help, while others framed it more as both parent's responsibility.
Beneath these overarching similarities, the risk-averse group was more focused on safety for the children than the risk-friendly group, while the risk-friendly group was more intent on improving parenting skills.The Aotearoa New Zealand respondents were more likely to have some alternative care as part of the plan goals, while those from Wales had a focus on daily routines for the children, including feeding, school attendance, and hygiene, which the Aotearoa New Zealand respondents did not mention at all.
Those in the risk-averse group had clearer and more detailed aims for the intervention, compared with the risk-friendly group, whose stated aims tended to be vaguer and more diverse.For the risk-averse group, the main aim was to keep the children safely with their mother, with supports of various kinds in place, for the children to be safe and happy, for them to have supervised contact with their father, and for their basic needs to be met.The risk-friendly group had a similar focus on support, reducing stress and safety to the risk-averse group, but it was not as pronounced, and there were a greater number of diverse miscellaneous aims as well.Although all groups noted the risk factor of poverty, addressing this did not feature highly in the aims of any respondents, being mentioned by just a few respondents, and not at all in the risk-friendly group from Aotearoa New Zealand.
In terms of the essential or 'bottom-line' changes, the main finding is again one of consensus rather than difference, though one striking finding did emerge.Firstly, the main themes showed that all respondents felt (in order of descending importance) that the essential changes were: completion of parenting and IPV courses; that either Shannon alone or both parents should be engaged in counselling to address their own trauma, mental health and communication; no police call-outs for IPV; a clear parenting plan in place around residence (custody) and contact (access); and for parents to show 'proof' or demonstration of parenting behavioral change, otherwise unspecified.The only major difference was the emphasis placed by the risk-averse group from Aotearoa New Zealand on the importance of continuing engagement with professional services or 'cooperation' as a sign of change.Overall, there was no apparent difference between the risk-averse and risk-friendly groups in the findings on essential changes.

Limitations
Before discussing these findings, it is worth noting four important limitations to the study.First, while the use of a case vignette allows us to compare between different respondents, nonetheless there are limitations to the use of vignettes (Matza et al., 2021).There are no objective standards against which to compare the vignette (for example, whether it is authentic or not), and vignettes are by their nature limited in terms of their content compared to real-life scenarios.We also do not know whether respondents would behave differently, and to what extent and in what ways, if they were faced with a similar situation in real-life (Taylor, 2006).Second, respondents in our study experienced 'survey fatigue', meaning that while our initial sample consisted of 81 student social workers, the final sample included 65, and 13 of these did not complete 100% of the survey.This limits our ability to generalize from the findings.However, relative to other social work vignette studies, our sample is at least comparable, with examples we could find ranging from 22 to 201 (Harris et al., 2022;Williams & Soydan, 2005).Third, the study took place during the Covid-19 pandemic, with various legal restrictions and statutory guidance being applicable in the two countries.It is hard to say how this might have affected the respondents' responses to the vignettes.Possibly they might have felt more worried than usual, if they imagined the children were out-of-school and going unseen by professionals working in universal services.Finally, the student social workers who took part were a self-selecting, small sample, and thus we cannot claim (and are not claiming) that they are representative of any larger population.This again limits the ability to generalize from our findings.

Discussion
The main finding of this exploratory study is the degree of consensus between the responses of social work students from Wales and Aotearoa New Zealand to the case vignette, despite the two locations being approximately 18,000km apart.In addition, the two countries are self-evidently varied in terms of culture, history, population demographics, legal systems, and approaches to social work education.Despite these important differences, social work students from Wales and Aotearoa New Zealand gave relatively similar ratings in relation to safety, harm, and to question of whether the children were being abused or neglected, and this may reflect broadly similar welfare orientations despite other substantial social differences (Gilbert et al., 2011).Beneath this broad similarity, some fine-grained differences are evident.The primary between-country differences related to perceptions of risksimilar at the outset of the vignette, but higher in Wales at stages two and three.Subtle differences in thresholds for actions were also perceptible beneath the broad consensus findings.Respondents from Aotearoa New Zealand generally had a higher perception of risk but did not en masse reach the threshold for intervention until stage four, compared to Welsh respondents.This may suggest more emphasis on a child welfare orientation operating in Aotearoa New Zealand compared to Wales (Križ & Skivenes, 2013), and would be consistent with Bunting et al's (2018) view of the UK as having an increasing orientation towards child protection, whilst Wales has one of the highest rates of children in care in the world (Wood & Forrester, 2023).Thus, when considering our first research question, we did find some association between perceptions of risk and country of origin.
Equally, as found in other studies comparing more and less experienced workers or comparing between qualified social workers and student social workers, levels of existing experience was a distinguishing factor between our respondents (Bartelink et al., 2018).For example, Benbenishty et al (2002) found that less experienced workers and students tend to perceive higher levels of risk compared with more experienced and trained workers.Thus, when considering our second research question, we did find some association between perceptions of risk and existing levels of social care experience.This suggests that experience can be a factor acting in concert with several others to produce judgement outcomes, within the student population as well as for qualified practitioners.As child protection in the UK and in Aotearoa New Zealand has increasingly become the first port of call for newly qualified social workers, who will have less experience than most other groups of workers, it is perhaps not entirely remiss to suggest this might be a systemic problem.For example, 63% of Oranga Tamariki staff have less than three years' experience, up from 53% five years ago (Oranga Tamariki, 2021).The more inexperienced workers there are in these teams, the more likely families are to encounter a risk-averse response upon being referred (all else being equal).As a matter of policy, one should be asking how services in both countries might attract and retain more experienced workers to return and / or stay in the child protection field, and how to ensure less experienced workers are supported effectively as they develop their experience.
However, perhaps the most useful way of grouping these respondents, rather than country of origin or existing experience, can be made based on their initial judgements of the level of risk to the childrenthat is, whether they were risk-averse (initially rating the risk as moderate or high) or risk-friendly (initially rating the risk as being non-existent or low).This reflects the findings of previous studies (Author), which used the same risk-averse and riskfriendly groupings, and similarly found it to be a helpful way of understanding subsequent judgements and decision-making.In other words, to predict what each respondent would say about the case vignette, it would be more helpful to know their initial risk rating than it would to know whether they were from Wales or Aotearoa New Zealand, while knowing about levels of social care experience would help you predict whether someone was likely to be risk-averse or risk-friendly.Of course, as the vignette unfolded, we saw increasing convergence between these two groups, as more information became known.Yet in a reallife case, an initial perception of the risk as being high vs low would inform what, if anything, subsequently happened, as initial case categorizations shape decision pathways and processes (Lonne et al., 2020).Respondents in the risk-friendly group would have been more likely to close the case without further investigation, compared to the risk-averse group, and so no further information would have been 'discovered'.In relation to our third research question, while risk-averse and risk-friendly respondents identified some similar risk and safety factors, including the parent's own traumatic backgrounds, and financial stress, those in the risk-averse group were more likely to interpret ambiguous information as indicating risk.This reflects findings made by Portwood (1998), who discovered that practitioners with more limited experience of child maltreatment casework were more likely to interpret uncertain information as indicative of abuse / neglect.In this case vignette, information was provided to respondents that there have been police callouts to the home and the family may have frequent arguments.It is noteworthy that this information was interpreted by the riskaverse group as indicating the potential risk of IPV, whereas the risk-friendly group were less likely to identify IPV as a concern and more likely to emphasize the lack of family and community support.
When considering our final research question, and the actions recommended by different respondents, the primary stand-out finding is that of broad consensus between the two countries.This could suggest that even when initial judgements about risk are different, as well as judgements about harm and safety, once the threshold for statutory intervention is crossed, the availability of resources and typical patterns of proceduralised responses become more important, and in a sense limited in terms of what can and will be done with and for the family.Another point worth considering is that the risk-averse respondents were more likely to recommend interventions of various kinds, including interventions that might be thought of (by some) as more progressive, such as Family Group Conferences, compared to those that might be considered more oppressive (by some), such as care arrangements.If so, this would be in contrast to Roberts' (1970) finding that more experienced workers (who tend to be less risk-averse) can be more pessimistic about the likelihood of positive future outcomes, and thus are more likely to recommend quicker and more legalistic interventions (such as child removal).However, this point is complicated by the statutory requirement (in Aotearoa New Zealand) and growing expectation (in Wales) that Family Group Conferences are legally mandated before any removal process can occur (apart from immediate safety warrants).Risk-aversion did not necessarily predict the form of intervention, although it could be said to predict a more interventive stance generally.Finally, while a high proportion of our respondents noted the family's financial problems, students from Aotearoa New Zealand were more likely to talk about the lack of community-based support, while students from Wales barely mentioned this factor.Even when respondents did explicitly comment on the family's deprived socioeconomic circumstances, this did not translate into their recommendations for intervention.Typically, interventions were focused on parenting behavior, and individual psychology, rather than social circumstances.Is this a limitation of a casework approach to child protection, or does it tell us something specific about the approaches taken in neo-liberal democracies that prioritize individualistic responses in statutory services, leaving social determinants relatively untouched?(Bywaters et al., 2019;Featherstone et al., 2016;Hyslop & Author, 2018;Author et al., 2019).
Having said this, such an argument does presuppose that risk-friendliness is 'better' than risk-aversion, which may not always be true.Perhaps more experienced workers are becoming desensitized to the abusive and neglectful experiences of children, in a way that their less experienced colleagues are not.Alternative explanations are more convincing, however, as experience has been associated with greater familiarity with risks and hazards, so that they are less 'alarming', especially for "risks that fall outside of everyday experience" (Fleming et al., 2015(Fleming et al., , p. 2298)).Some risk factors that appear serious to less experienced workers may not be associated with any increased likelihood of serious harm over the longer term, at least not when early help and prevention services are available.Time undoubtedly provides a feedback loop that shapes heuristics regarding case outcome expectations (Klein, 2015;Taylor, 2016;a Taylor, 2016b).Fluke et al (2016) similarly suggest this explanation of experience as influencing: "an individual staff member's understanding of what happens over time to the children and families served by their agency, which may influence their perceptions and beliefs regarding the efficacy of child safety and family preservation efforts" (Fluke et al., 2016, p.214).However, while this may increase the perception of risk that the system itself may cause (reducing perceptions of risk without intervention), it may also be distorted by a lack of information about false negatives.As there is no objective measure of child abuse and neglect incidence or definition, (despite consensus around the most egregious examples), we have no external standard with which to compare risk-aversion and riskfriendliness (Cradock, 2014).Nonetheless, when there are policy aims in both countries to safely reduce the number of children in care, within that context it is worth considering the role that risk-aversion can play in the involvement of more and more families with the child protection system.

Conclusion
This study examines risk perceptions of social work students in Wales and Aotearoa New Zealand.By comparing the two countries, as well as risk-averse and risk-friendly groups of respondents, areas of divergence and consensus are identified.It shows that the ways risk is constructed in both countries is remarkably similar, while pointing out subtle areas of difference and divergence.The important role of experience as shaping risk attitudes is highlighted.Implications are that there is a need for greater consensus around how risk should be constructed within a given national context, greater critical analysis of how terms such as risk and safety are encoded in practice, and careful attention to worker professional development and education, given how influential experience is in shaping risk aversion or friendliness.Understanding which really benefits children and their families is important to understand in the context of the intervention of the child protection system.

Figures 2 and 3 .
Figures2 and 3.The proportion of respondents from Wales and Aotearoa New Zealand who rated the risk as moderate or high, and said the children were 'not safe', at different stages of the vignette.

Figures 4
Figures4 and 5. Risk and safety ratings for respondents with differing levels of experience, across vignette stages.
table 1), administered via Qualtrics.At each stage of the vignette, respondents answered questions about levels of risk, harm, and safety, identified risk and protective factors, and made recommendations about what they would do next (table 2).Respondents addressed the questions using a variety of Likert-scales, with labels appropriate to each item, as well as open free-text questions.The survey was available between September 2020 and March 2021.

Table 2 :
An overview of the questions in the survey, and responses for each one.
table3) and 52 completed 100% of the survey (answering questions in relation to all four stages of the vignette).All respondents declared themselves to be current social work students, registered on qualifying University programmes.Of these, 40 (61.5%) were from Aotearoa New Zealand and 25 (38.5%) from Wales.Most were female (81.5%) and aged between 18 and 34 (56.9%).Most were White British (30.8%) from Wales or Pākehā (New Zealander of European descent; 38.5%) from Aotearoa New Zealand.More than half had either zero years or 1 to 2 years' experience in social care (60%).In the tables below, note that sub-category totals do not always add up to 65, because not every respondent answered all questions.

Table 3 :
Description of the sample.

Table 4 :
Break-down of respondents into categories of risk-averse and less experienced.

Table x
. The results of an independent samples t-test for respondents from Wales vs Aotearoa New Zealand at stage one in relation to the question: If there was no intervention and things continued as they are, how would you rate the level of harm to the children over time?

Table x
. The results of an independent samples t-test for more vs less experienced respondents at stage one in relation to the question: How would you rate the risk of harm in relation to the children in this case?

Table x
table x).
. The results of a paired samples t-test for stages one and two in relation to the question: How would you rate the risk of harm in relation to the children in this case?

Table x .
The results of a paired samples t-test for stages one and three in relation to the question: How would you rate the risk of harm in relation to the children in this case?

Table 5 :
Frequencies of different recommendations at each stage of the vignette.