'I'm So Stressed!': A Longitudinal Model of Stress, Burnout and Engagement among Social Workers in Child Welfare Settings.

The well-documented day-to-day and long-term experiences of job stress and burnout among employees in child welfare organisations increasingly raise concerns among leaders, policy makers and scholars. Testing a theory-driven longitudinal model, this study seeks to advance understanding of the differential impact of job stressors (work-family conflict, role conflict and role ambiguity) and burnout (emotional exhaustion and depersonalisation) on employee disengagement (work withdrawal and exit-seeking behaviours). Data were collected at three six-month intervals from an availability sample of 362 front line social workers or social work supervisors who work in a large urban public child welfare organisation in the USA. The study's results yielded a good model fit (RMSEA = 0.06, CFI = 0.96, NFI = 0.94). Work-family conflict, role ambiguity and role conflict were found to impact work withdrawal and exit-seeking behaviours indirectly through burnout. The outcome variable, exit-seeking behaviours, was positively impacted by depersonalisation and work withdrawal at a statistically significant level. Overall, findings, at least in the US context, highlight the importance of further examining the development of job burnout among social workers and social work supervisors working in child welfare settings, as well as the utility of long-term administrative strategies to mitigate risks of burnout development and support engagement.

Introduction

To meet its service-focused mission, the social work profession is reliant on competent and engaged workers. Yet, social workers are considered high-risk for job stress and burnout. Although no definitive statistics exist on the prevalence rates of stress and burnout in social work, findings from previous studies suggest that social workers are prone to experiencing both (Arrington, 2008; Lloyd ). Stress and burnout have attracted the attention of scholars and practitioners because of potential hazards posed to social workers' well-being and effectiveness (e.g. job satisfaction: Evans ; organisational commitment: Lambert ; turnover: Acker, 2012; Kim and Stoner, 2008).

Challenges resultant from job stress and burnout are also, in particular, experienced by social workers in child welfare settings (Lizano and Mor Barak, 2012; Travis and Mor Barak, 2010; Sprang ). Child welfare—which is characterised by high job demands, low wages, high caseloads and need for quality supervision—has been described as one of the most challenging professions (AECF, 2003; Kim, 2011). Although a preponderance of published studies are in North America, these challenges are shared in child welfare settings across the world (Burns, 2011). Yet, limited empirical evidence exists to provide guidance on which types of stressors, and subsequent burnout, influence employee disengagement among employees (particularly social workers) in child welfare settings.

What is the prolonged impact of different types of stressors on an individual's proclivity to quit? Which dimensions of burnout—emotional exhaustion or depersonalisation—have the most impact on one's propensity to psychologically retreat from work activities? This study addresses these questions by examining which stressors and burnout factors ‘tip the scale’ or increase the likelihood of disengagement over time. We specifically test a theory-driven longitudinal model of antecedents to engagement among social workers and social work supervisors based in a public child welfare organisation in the USA. This is an important line of inquiry because job stress and engagement link to organisational effectiveness (Blacksmith and Harter, 2011; Louis ). Building a quality-driven and engaged workforce requires in-depth analyses of stress-related factors that may hinder worker effectiveness (Simona ).

Conceptual model and research hypotheses

Definitional clarity

Work stressors

Broadly defined, work stress is the ‘harmful physical and emotional responses that occur when the requirements of the job do not match the capabilities, resources, or needs of the worker’ (NIOSH, 1999, p. 6). Occupational role stress emerges as individuals experience conflict, ambiguity or overload in work-related roles (Dobreva-Martinova ; Tetrick, 1992). Work–family conflict is conceptually defined as inter-role conflict resulting from an incompatibility in work and family responsibilities (Greenhaus and Beutell, 1985; Michel ). Conceptualised in Robert Kahn's (1964) seminal work, role conflict refers to the conflicting demands that employees face, whereas role ambiguity relates to the lack of clarity in the expectations of others.

Job burnout

Chronic work-related stressors have prolonged effects on employees resulting in burnout (Lee and Ashforth, 1996; Maslach ; Schaufeli ). Burnout is a multidimensional construct, yet, from a theoretical and practical perspective, ‘the overall concept of burnout may sometimes help us to advance our knowledge in a more thorough way than research on the separate, underlying dimensions’ as well as to streamline study findings (Brenninkmeijer and VanYperen, 2003, p. i17). Despite this, we focus on two dimensions of burnout. Emotional exhaustion and depersonalisation are stress responses resulting from prolonged exposure to overload and social conflict in the workplace (Maslach, 1998; Maslach ).

Emotional exhaustion, central to job burnout, refers to feelings of exhaustion resulting from overwhelming demands (Maslach ). Feelings of emotional exhaustion subsequently lead to depersonalisation. Depersonalisation is a protective state of cynicism that spurs dissonance, either cognitive or emotional, with others as a coping strategy for work demands and exhaustion (Halbesleben and Buckley, 2004).

Disengagement

Employee disengagement can be a psychological form of ‘leaving’ one's job while remaining employed (Burris ). Hence, work withdrawal and exit-seeking behaviours are considered forms of disengagement. Work withdrawal is a way of ‘zoning out’ or a form of job neglect by psychologically detaching from work-related tasks or organisational activities (Hopkins ; Kidwell and Robie, 2003; Rusbult ; Withey and Cooper, 1989). Examples of withdrawal behaviours include not putting forth concerted effort in work tasks, avoiding co-workers or misusing organisational time (Kidwell and Robie, 2003; Naus ). When engaging in exit-seeking behaviours (herein referred to as exit), employees contemplate quitting by actively looking for a new job (Travis and Mor Barak, 2010). As such, exit reflects a form of psychological disengagement (Burris ; Naus ). Thus, work withdrawal and exit as forms of disengagement are well-documented predictors of turnover (Griffeth ), which may result in lowered levels of innovation and worker creativity (Krueger and Killham, 2006).

Primary model and research hypotheses

Figure 1 depicts relationships among work stressors, job burnout and disengagement over time. The proposed conceptual model reflects the long-term negative effects of work stressors on employee disengagement through job burnout. Relationships among study variables are supported by theoretical and empirical evidence. Role theory (Biddle, 1986; Kahn ) offers insight into how job stress affects worker outcomes including burnout and disengagement. Burnout is postulated to develop due to chronic stressors in the workplace resulting from incongruence between the worker and the job (Maslach, 2003) making role theory a strong point of departure in helping explain antecedents to burnout (Cordes and Dougherty, 1993). According to role theory, organisational roles are ascribed within the context of organisational structures, processes and norms (Dobreva-Martinova ) and are determined by employees' position or status (Biddle, 1986). Organisational context and status levels can influence how well employees perform and whether they experience conflicting demands and ambiguous job situations (Dobreva-Martinova ). Using role theory as a frame, burnout and associated disengagement can emerge as employees experience challenges in balancing competing needs of work and family or home life (work–family conflict), incompatibility with work roles (role conflict) and lack of clear expectations about one's work roles (role ambiguity) (Kim and Stoner, 2008; Rupert ).

Figure 1

Study conceptual model

Burnout

Chronic exposure to work-related stressors is a well-documented contributor to burnout (Boyas and Wind, 2010; Kim and Stoner, 2008; Lloyd ; Maslach, 1998, 2006). Researchers found that work–family conflict is among occupational stressors that incite job burnout (Allen ; Maslach, 2005; Schaufeli ). Positive relationships between role conflict and ambiguity and job burnout have been documented (Cordes and Dougherty, 1993; Lee and Ashforth, 1996; Örtqvist and Wincent, 2006). Accordingly, occupational demands associated with work–family conflict, role conflict, role ambiguity and generalised job stress are linked to a greater propensity for burnout (Nissly ; AECF, 2003; Boyas and Wind, 2010; Kim, 2011; Regehr ). Thus:

Hypothesis 1: Role conflict, role ambiguity, and work–family conflict at Time 1 will have a positive relationship with burnout dimensions at Time 2.

Relationship between burnout dimensions

Distinguishing the relationship between emotional exhaustion and depersonalisation serves as a starting point to understanding the differential impact of burnout on disengagement. Researchers found that emotional exhaustion elicits a worker response resulting in depersonalisation of clients being served and co-workers to create an emotional distance from the stressful work condition (Maslach ):

Hypothesis 2: Emotional exhaustion leads to depersonalisation.

The mediating impact of job burnout

Researchers found that job burnout adversely influences work-related engagement (International Labour Office, 1993; Maslach ) and high turnover rates (Mor Barak ; Zlotnik ). In a meta-analysis of 115 empirical studies, Swider and Zimmerman (2010) found that burnout factors correlated with absenteeism, turnover and job performance. As stated by Montero-Marín , workers experiencing burnout are characterised by ‘a lack of personal involvement in work-related tasks, leading one to give up as a response to any difficulty’ (p. 11). From this vantage point, work withdrawal is a manifestation of burnout factors and can result in work behaviours that seemingly reflect a decision to limit work-related activities. As such, job burnout creates a cognitive distance that buffers against stressful work experiences (Lee and Ashforth, 1996; Maslach ). This leads to reduction in motivation or an increase in frustration with the job, which may foster development of attitudes and behaviours that reflect disengagement and psychological withdrawal in the workplace (Halbesleben and Buckley, 2004):

Hypothesis 3a: Emotional exhaustion and depersonalisation at Time 2 will positively impact disengagement at Time 3.

Hypothesis 3b: Workplace stressors at Time 1 will have a positive and indirect relationship, through burnout dimensions at Time 2, on disengagement at Time 3.

Relationship between engagement behaviours

Withey and Cooper (1989) found that job neglect (similar to work withdrawal) and exit were related because both are forms of detaching oneself from the job. As people withdraw, they may have greater propensity to move towards leaving (exit). Hence, withdrawal can result in exit, particularly when alternatives are either perceived as or actually not an optimal choice (Withey and Cooper, 1989). In a cross-sectional examination of public child welfare, we found a direct and positive relationship between work withdrawal and exit-related behaviours (Travis and Mor Barak, 2010). Although it does not represent ‘exit’ as conceptualised in this study, actual turnover has been found to be precipitated by limited engagement in work-related tasks and organisational activities (Griffeth ; Withey and Cooper, 1989). Thus, we hypothesised:

Hypothesis 4: Work withdrawal will positively impact exit-seeking behaviours.

Competing model

Overall, job-related stressors are well-documented predictors of undesirable worker outcomes (e.g. counterproductive work behaviours and intention to leave) (Mor Barak ; Travis and Mor Barak, 2010; Williams ). Increased harmony between work and life demands and lessened work–family conflict have been leading predictors of engagement-related factors (Allen ). Thus, we tested a competing model that explored direct relationships between work stressors and disengagement. By examining results from the primary and competing model, we were able to determine whether stressors directly affected disengagement or work through burnout.

Methods

Design and context

This three-wave longitudinal panel study was conducted as a part of a large-scale study under the auspices of a university-based training centre for children and family services employees. It builds on previous research on employee responses to undesirable work situations and retention in child welfare organisations (Nissly ; Travis and Mor Barak, 2010; Travis ; Lizano and Mor Barak, 2012). Data were collected at three six-month intervals from an availability sample of 362 employees at large urban public child welfare agency using a self-report questionnaire. Ninety-five per cent of the sample described their job position as social workers or social work supervisors. In addition, 50 per cent reported having a formal social work education or training at either the bachelor or graduate levels.

Participants were recruited while attending one of several child welfare skill and knowledge building trainings. The voluntary nature and study details were explained during the informed-consent process. Free lunch was offered as compensation for participation. Respondents completed the baseline questionnaire in a vacant training room. Questionnaires with unique identifiers were mailed to participants at six- and twelve-month time points. The same measurement instrument was used for all data collection. Table 1 presents demographic information for the study sample. A University's Institutional Review Board approved all study protocols, and all participants provided written informed consent.

Table 1

Participant demographics

	%	M	SD
Age		36.83 (range: 21–70)	11.46
Gender
Male	16.3
Female	83.4
Race/ethnicity
Caucasian	30.7
Latino(a)/Hispanic	29.9
African American/black	21.9
Asian/Pacific Islander	12.2
Other	4.2
Tenure (years)		6.16 (range: < 1–36)	6.73
0–5	62.4
6–10	18.5
11–15	10.9
16–20	6.9
21–25	0.7
≥ 26	0.7
Position
Direct service provider	84.8
Supervisor	15.2
Educational background
Bachelor's degree	34.3
Master's degree	63.10
Ph.D.	0.8
Other	1.1

Measures

All measures were captured using a six-point Likert scale. Response options for all items ranged from 1 = strongly disagree to 6 = strongly agree. Reverse coding was used to ensure that higher-scale scores corresponded with higher levels of work–family conflict, role conflict, emotional exhaustion, depersonalisation, work withdrawal or exit. A composite score was created for each scale.

Work–family conflict

Beatty's (1996) three-item work–family conflict scale was used to measure conflict between work and family demands. Scale items included ‘My job keeps me away from my family too much’ and ‘I have a good balance between my job and my family life’. This scale yielded a Cronbach's alpha reliability value of 0.75.

Role conflict and role ambiguity

As facets of work stress, role conflict and role ambiguity were measured using Rizzo, House and Lirtzman's (1970) scale items. Eight role conflict and six role ambiguity items were used. Scale items for role conflict and ambiguity included items such as ‘I have clear planned goals and objectives for my job’ and ‘I receive incompatible tasks from two or more people’. Role ambiguity and role conflict, respectively, had Cronbach's alpha coefficients of 0.83 and 0.85.

Emotional exhaustion and depersonalisation were measured using the Maslach Burnout Inventory (MBI; Maslach and Jackson, 1981). Two of three subscales corresponded to dimensions of emotional exhaustion (nine items) and depersonalisation (five items). The third subscale, personal accomplishment, was excluded based on theoretical and empirical reasons. Theoretically, personal accomplishment more accurately describes a personality characteristic and not a symptom of job burnout (Cordes and Dougherty, 1993). Empirically, MBI's personal accomplishment subscale has demonstrated a weaker relationship with the other two dimensions of job burnout (Lee and Ashforth, 1996); when used longitudinally, it has not demonstrated invariance over time (Kim and Ji, 2009). The two MBI subscales yielded a Cronbach's alpha reliability coefficient of 0.94 (emotional exhaustion) and 0.81 (depersonalisation).

Employee disengagement

Levels of disengagement were measured using Rusbult job neglect (six items) and exit scales (four items), which measure an individual's self-reported generalised tendencies to engage in withdrawal or exit-seeking behaviours. These scales were tested in different variations across three studies by Rusbult and yielded reliability coefficients ranging from 0.69 to 0.82 for job neglect and from 0.76 to 0.97 for exit. In the current study, the job neglect and exit scales respectively yielded a Cronbach's alpha coefficient of 0.73 and 0.89.

Covariates

Respondents' self-reported age, gender and organisational tenure were derived from demographic questionnaire items. The covariate of gender was dummy coded in preparation for analysis (male = 0, female = 1).

Analytic procedures

Data preparation

Missing data due to attrition and other extraneous variables are considered commonplace and are particularly prevalent in longitudinal studies (Little and Rubin, 2002). Patterns of missing data among the central study variables were examined prior to analysis. Three hundred and sixty-two respondents participated in the study at baseline, 187 participated at Time 2 and 133 participated in all three waves of data collection. Preliminary examination suggested that data were missing at random as defined by Rubin (1996). To test whether job stress, work–family conflict, emotional exhaustion or depersonalisation predicted missing data, a series of logistic regression models were conducted. Logistic regression analyses yielded no significant relationships among study variables and patterns of missing data. Furthermore, relationships among study covariates (demographic characteristics) and missing data were also examined via logistic regression, with results suggesting a null relationship between age, sex and job tenure and whether or not a participant missed a data collection point.

Using IBM's AMOS 18 software, a full information maximum likelihood (FIML) estimate approach was chosen to address missing data in the sample due to the potential bias that older methods such as case deletion and mean imputation can introduce (Schafer and Graham, 2002).

Study variables were examined for normality prior to analysis. Deviation from normality among the central study variables was minimal and within the 3.0 maximum skewness value (absolute value) recommended when conducting structural equation modelling using FIML estimation methods during analysis (Kline, 2010). Skewness among study variables ranged from –0.09 to 0.51, whereas kurtosis ranged from –1.24 to 0.019.

Descriptive statistics

Descriptive statistics and bivariate correlations for the central study variables are presented in Table 2.

Table 2

Descriptive statistics and interitem correlation values

	M	SD	1	2	3	4	5	6
1 Work–family conflict(T1)	9.68	4.14	1
2 Role conflict(T1)	30.45	8.51	0.262**	1
3 Role ambiguity(T1)	14.41	5.04	0.186**	0.156**	1
4 Emotional exhaustion(T2)	30.11	10.76	0.253**	0.192*	0.146	1
5 Depersonalisation(T2)	11.23	5.02	0.121	–0.015	0.146	0.436**	1
6 Work withdrawal(T3)	14.27	5.34	0.117	0.207*	0.118	0.424**	0.358**	1
7 Exit-seeking behaviours(T3)	12.65	6.65	0.18	–0.014	0.222*	0.478**	0.401**	0.458**

* p ≤ 0.05; ** p ≤ 0.01.

Path analysis

A form of structural equation modelling, path analysis was employed to test the hypothesised conceptual model. Path analysis is an approach that models multivariate interrelationships of observed variables and allows for the simultaneous testing of multiple relationships (Bollen, 2005). A two-step model-testing process was employed, yielding two models. The first model tested the theorised direct and indirect relationships between workplace stressors, emotional exhaustion, depersonalisation and the outcomes of exit and work withdrawal over time. The impact of control variables on the outcomes of interest was also tested in Model 1. A second trimmed model was analysed after removing all insignificant paths from Model 1. The model presenting only significant paths is shown in Figure 2.

Figure 2

Path analytic model of workplace stressors, burnout and disengagement over time (standardised coefficients)

Results

Fit indices

Multiple indices were used to assess the fit of the primary and competing model as recommended in the structural equation modelling literature (Kline, 2010). Although we used the chi-square test, other fit indices were also used because the chi-square test is known to be sensitive to sample size (Tabachnick and Fidell, 2007). Root mean square error of approximation (RMSEA) was used to assess model approximation to population parameters (Kline, 2010), whereas the comparative fit index (CFI) and the normed fit index (NFI; Bentler, 1990) were used to examine the improvement of fit between the full and trimmed models.

Full model

The full hypothesised conceptual model yielded a statistically significant chi-square value of X2 (12) = 27.25, p = 0.007. However, the remaining fit statistics met criterion suggesting good model fit (RMSEA ≤ 0.05, NFI and CFI ≥ 0.90). The full model included an RMSEA value of 0.06, a CFI value of 0.96 and an NFI value of 0.94.

Competing model

The competing model yielded a statistically significant chi-square value of X2 (13) = 30.3, p = 0.007. However, the remaining fit statistics again suggested an acceptable model fit (RMSEA = 0.05, CFI = 0.90, NFI = 0.85).

Path coefficients

Primary model

Standardised coefficients for the model paths are presented in Figure 2. Table 3 also provides the full and competing model path analysis results. Findings suggest that work–family conflict (β = 0.24) and role conflict (β = 0.16) significantly and positively impacted emotional exhaustion, with work–family conflict having a greater influence on development of emotional exhaustion than role conflict. Work–family conflict and role conflict accounted for 13.6 per cent of the explained variance in emotional exhaustion. No statistically significant relationships were found between role ambiguity and burnout components (i.e. emotional exhaustion and depersonalisation). Emotional exhaustion significantly impacted depersonalisation (β = 0.45) and accounted for 22.3 per cent of its variance, whereas no other predictor variables were found to have a statistically significant relationship with depersonalisation.

Table 3

Full and competing model path analysis results

Outcome variable	Independent variable	b	SE b	B
Full model
Emotional exhaustion(T2)	Work–family conflict(T1)	0.65	0.20	0.24***
Emotional exhaustion(T2)	Role conflict(T1)	0.21	0.10	0.16*
Emotional exhaustion(T2)	Role ambiguity(T1)	0.26	0.16	0.12
Depersonalisation(T2)	Emotional exhaustion(T2)	0.21	0.04	0.46***
Depersonalisation(T2)	Work–family conflict(T1)	0.02	0.11	0.02
Depersonalisation(T2)	Role conflict(T1)	–0.06	0.05	–0.10
Depersonalisation(T2)	Role ambiguity(T1)	0.10	0.09	0.10
Work withdrawal(T3)	Emotional exhaustion(T2)	0.21	0.05	0.42***
Work withdrawal(T3)	Depersonalisation(T2)	0.10	0.10	0.09
Work withdrawal(T3)	Tenure(T1)	0.05	0.09	0.06
Work withdrawal(T3)	Sex(T1)	–1.70	1.15	–0.12
Work withdrawal(T3)	Age(T1)	–0.13	0.05	–0.27*
Exit(T3)	Work withdrawal(T3)	0.35	0.12	0.28***
Exit(T3)	Emotional exhaustion(T2)	0.16	0.06	0.27**
Exit(T3)	Depersonalisation(T2)	0.25	0.12	0.19*
Exit(T3)	Age(T1)	0.05	0.07	0.08
Exit(T3)	Sex(T1)	0.63	1.41	0.04
Exit(T3)	Tenure(T1)	–0.12	0.11	–0.12
Trimmed model
Emotional exhaustion(T2)	Work–family conflict(T1)	0.67	0.20	0.25***
Emotional exhaustion(T2)	Role conflict(T1)	0.23	0.10	0.17**
Depersonalisation(T2)	Emotional exhaustion(T2)	0.20	0.04	0.46***
Work withdrawal(T3)	Emotional exhaustion(T2)	0.22	0.04	0.46***
Work withdrawal(T3)	Age(T1)	–0.10	0.04	–0.23**
Exit(T3)	Work withdrawal(T3)	0.33	0.11	0.26**
Exit(T3)	Depersonalisation(T2)	0.23	0.12	0.18*
Exit(T3)	Emotional exhaustion(T2)	0.17	0.06	0.29**

*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001. Full model R2: emotional exhaustion (T2) = 0.14, depersonali-sation(T2) = 0.22, work withdrawal(T3) = 0.29, exit(T3) = 0.35. Trimmed model R2: emotional exhaustion(T2) = 0.12, depersonalisation(T2) = 0.21, work withdrawal(T3) = 0.26, exit(T3) = 0.34.

A statistically significant positive relationship was found between emotional exhaustion and outcomes of work withdrawal (β = 0.42) and exit (β = 0.27), with a stronger relationship existing between emotional exhaustion and withdrawal. Depersonalisation was found to have a positive and statistically significant relationship with exit (β = 0.19) but not work withdrawal. A positive relationship between withdrawal and exit was also found (β = 0.28). Age was the only control variable found to be positively related to any outcome of interest, with a negative relationship with withdrawal (β = –0.27). The path analytic model was found to account for approximately one-third of the variance in withdrawal (28.9 per cent) and exit (34.6 per cent). No significant direct paths were found between the predictor variables of work–family conflict, role conflict and role ambiguity (Time 1) and the outcome variables of withdrawal and exit (Time 3).

The competing model yielded the same statistically significant paths as the primary model, demonstrating consistency among the relationships of variables in both models. The beta coefficients varied only slightly in the competing model. Work–family conflict (β = 0.25) and role conflict (β = 0.17) positively impacted emotional exhaustion. Emotional exhaustion significantly impacted depersonalisation (β = 0.46), withdrawal (β = 0.46) and exit (β = 0.29). Depersonalisation positively impacted exit (β = 0.18) and withdrawal (β = 0.26). Age was the only control variable demonstrating a statistically significant relationship with withdrawal (β = –0.23). No significant relationships were found between work–family conflict, role ambiguity and role conflict and the outcome variables of withdrawal and exit.

Discussion

Summary

Findings provide a nuanced perspective of interrelationships among workplace stressors, burnout and engagement over time. Key findings point to the central role that burnout plays in disengagement among front line social workers or social work supervisors in a child welfare setting. Previous studies examining stress and burnout among human service and child welfare social workers generally have not tested the components of the job stress construct (e.g. role conflict, role ambiguity) (Boyas and Wind, 2010; Kim and Stoner, 2008). Instead, studies often use a composite and this potential masks the unique contribution of each stress component to job burnout. This study has sought to fill this gap.

The hypothesised relationship (Hypothesis 1) between work stressors and burnout was partially supported. Work–family and role conflict were positive and significant predictors of emotional exhaustion—a finding congruent with the job burnout literature (Allen ; Maslach, 2003; Peeters ). Nevertheless, our finding of a null relationship between role ambiguity and burnout was novel and unexpected. Job burnout is theorised to be a result of chronic exposure to work stressors (Maslach, 2003, 2006)—a relationship that has received evidentiary support in previous studies (Kim and Stoner, 2008; Lee and Ashforth, 1996).

None of the three stressors (role conflict, role ambiguity and work–family conflict) directly affected depersonalisation. Each stressor was indirectly linked to depersonalisation through emotional exhaustion at statistically significant levels. Based on our conceptual model, these findings suggest that work stressors lead to feelings of cynicism (depersonalisation) through emotional exhaustion.

As posited in Hypothesis 2, emotional exhaustion was a significant correlate of depersonalisation. This is consistent with previous research (Ashill and Rod, 2011; Diestel and Schmidt, 2010). For example, in a longitudinal study of German nursing home and civil servant employees, Diestel and Schmidt (2010) found empirical support for the causal relationship between emotional exhaustion and depersonalisation.

Several facets of Hypothesis 3a were supported with emotional exhaustion positively impact both work withdrawal and exit over time; however depersonalisation was only a significant correlate of exit. Job burnout theories offer insight into these findings. The accumulative stress may place an emotional tax on workers, which can have adverse implications for engagement. Study findings also supported Hypothesis 3b. That is, burnout served as a mediator of work stressors on employee disengagement over time. In the primary model tested, we found that only stressors (work–family conflict and role conflict) affected disengagement (withdrawal and exit-seeking behaviours) through the burnout dimension of emotional exhaustion. Acker (2012) had similar findings in which emotional exhaustion served as mediator of role stress and intent to quit among a study of mental health providers (including social workers).

When testing the competing model, we found that none of the stressors directly predicted exit and work withdrawal over time. This is contrary to Hopkins et al.'s (2010) findings that showed job stress was the strongest predictor of work and job withdrawal behaviours. Yet, Hopkins examined a composite of stress (i.e. emotional exhaustion, role conflict and role overload) which differs from this study's emphasis on distinguishing types of stressors.

As hypothesised (Hypothesis 4), work withdrawal served as an antecedent of exit-seeking behaviours. This replicated our findings in a cross-sectional study of child welfare workers (Travis and Mor Barak, 2010). Griffeth found in a meta-analyses of factors leading to turnover that lack of engagement precedes intention to leave and actual quitting.

Finally, we found that age was a significant correlate of withdrawal. In this, younger workers were more likely to report engaging in work withdrawal. In this accord, as proposed by Ng and Feldman (2009), generational differences can be explained by one's expectations and perceptions of the job itself as well as the labour market. Older workers may have greater flexibility in intra-organisational contract negotiation and problem resolution processes and, as a result, have less motivation to engage in withdrawal behaviours. Also, Ng and Feldman offer ‘self-preservation’ (p. 1068) as an explanation, in which older workers may be more weary of the perceived impact of age stereotypes on one's ability to ascertain a job elsewhere. This maturity may also spur a focus on putting others before oneself or cultivates greater suspension of self in these contexts.

Study strengths and limitations

Findings should be interpreted within the context of the study's limitations. The study may be limited in generalisability based on use of availability sampling. The study setting, a large public child welfare organisation in an USA-based urban locale, limits the external validity and precludes generalising findings in rural or suburban areas as well as other regions in the world. Data for the present study were drawn from self-reports, posing a threat of common source bias. Moreover, some scales used to reflect study constructs have known limitations and more study is needed within the context of child welfare or social work settings.

Despite these limitations, the study has strengths that allow our findings to contribute to a greater understanding of role of stress and burnout in the development of the disengaged worker. The study's longitudinal design afforded us the ability to test the complex interrelationships of stress, burnout and disengagement over time. Additionally, examining separate components of stress, burnout and disengagement painted a picture of the complex interrelationships among variables of interest.

Implications

Improving engagement goes beyond simply asking the right questions. Engaging employees requires a year-round focus on changing behaviours, processes and systems to anticipate and respond to your organisation's needs. From the leadership team to the front line employees, all levels within an organisation must commit to making these changes (Gallup, 2010).

This study has practice and research implications for human resource development in child welfare organisations and among social work professionals. At the practice level, the above quote emphasises the importance of developing structural mechanisms to support worker engagement at all levels. Policies and practices that may inadvertently foster burnout and disengagement need to be evaluated based on the impact on employee job performance, business results and client outcomes. At the same time, developing timely organisational-level interventions to deal with conflict associated with work–family fit and competing demands of work roles is essential. This may include helping workers express concerns and emotions as well as active coping skills as mechanisms for increasing engagement and dealing with role stress (Ng and Feldman, 2012; Stalker ).

Amidst finding ways to prevent burnout and disengagement, leaders may consider what is working to help employees thrive. This is particularly relevant in child welfare settings in which workers experience emotional exhaustion, but are also satisfied with their jobs (Stalker ). Through the strengths-based lens, leaders may gauge best practices and leverage organisational resources to create innovative solutions, build resilience, increase intention to stay and foster engagement.

Finally, researchers are encouraged to distinguish between forms of work stressors and burnout to allow for examination of the differential impact of distinct stressors and burnout dimensions on well-being. Concurrently, there is a need to advance understanding of supportive mechanisms that mitigate disengagement and bolster engagement among social work professionals. Exploring factors that encourage productive work attitudes and behaviours is a fruitful line of inquiry (Collins, 2008) and may inform interventions that promote an engaged workforce and meet the mission of social service organisations.

Conclusion

Given the positive link between engagement and the provision of quality services (Markos and Sridevi, 2010), service effectiveness can be indirectly impacted due to employees not fully engaging in job activities. This study offers a deeper examination of dimensions of work stressors and burnout as antecedents to disengagement among social workers and social work supervisors. Findings indicate that emotional exhaustion is an indirect and direct threat to engagement, highlighting the importance of further examining the development of job burnout among social workers in child welfare settings and administrative strategies that may mitigate potential obstacles to optimal organisational effectiveness.