Individual and work group influences on 001



 



Individual and Work Group Influences on Turnover Intention Among Public
Child Welfare Workers: The Effects of Work Group Psychological Safety
Jean M. Kruzich, Joseph A. Mienko, Mark E. Courtney
PII:
DOI:
Reference:

S0190-7409(14)00098-X
doi: 10.1016/j.childyouth.2014.03.005
CYSR 2374

To appear in:

Children and Youth Services Review


Received date:
Accepted date:

7 February 2014
7 March 2014

Please cite this article as: Kruzich, J.M., Mienko, J.A. & Courtney, M.E., Individual and
Work Group Influences on Turnover Intention Among Public Child Welfare Workers: The
Effects of Work Group Psychological Safety, Children and Youth Services Review (2014),
doi: 10.1016/j.childyouth.2014.03.005

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.

AC


CE

PT

ED

MA

NU

SC
R

IP
T

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT


IP
T

Individual and Work Group Influences on Turnover Intention
Among Public Child Welfare Workers:
The Effects of Work Group Psychological Safety

SC
R

1. Introduction

High turnover rates and high levels of turnover intention have characterized social work in
child welfare settings for several decades (Nittoli, 2003). Findings from the federal Child and

NU

Family Service Reviews (CFSR) suggest that retention problems negatively affect outcomes
for children and families resulting in a lack of continuity in relationships between workers and


MA

families, a limited focus on child safety, and a negative impact on the timeliness of decisions
about safe and stable placements (GAO, 2003).
As the single most consistent predictor of turnover (Cho & Lewis, 2012; Dalton, Johnson

ED

& Daily, 1999), intention to leave is important not only as a precursor to turnover but as
an indicator of a workforce that may not be working at its full potential (Mor Barak, Levin,

PT

Nissly & Lane, 2006). Employees with unrealized turnover intentions are likely to resort
to other types of withdrawal behavior including absenteeism (Chang et al., 2013). Tham

CE

(2007), in a survey of public child welfare workers in Sweden, found that 48 percent of the
302 workers indicated an intention to leave in the coming year. Strand & Dore (2009),


AC

studying public child welfare staff in a state with a relatively low turnover rate of 8 percent,
found that 18 percent of the work force intended to leave in the next year and an additional
44 percent indicated they would prefer to leave the agency but salary and benefits were a
strong incentive to stay.
Findings from retention studies have failed to find demographic characteristics including
age, gender, race/ethnicity and education consistently related to retention. Previous study
findings have been mixed, with some studies discovering employees with social work degrees
are more likely to intend to stay (Ellett, Ellett & Rugutt, 2003; Scannapieco, Connell-Carrick
& Painter, 2007; Shim, 2010) while others have found employees with MSW degrees less likely
to intend to stay (Nissly, Mor Barak & Levin, 2005; Smith, 2005), and yet others detecting
no relationship between educational level and retention. Faller et al. (2010) found minority
(operationalized as non- white) child welfare staff had lower levels of work commitment,
1

ACCEPTED MANUSCRIPT

while others studies found no relationship between race and work withdrawal (DePanfilis &

Zlotnik, 2008) and still others concluding that Caucasian/White child welfare staff were less

IP
T

likely to stay (Shim, 2010).

Individual characteristics consistently related to intention to leave, and its opposite inten-

SC
R

tion to stay, are work attitudes including job satisfaction, (Mor Barak et al., 2006; Mor Barak,
Nissly & Levin, 2001; Moynihan & Pandey, 2007; Weaver, Chang, Clark & Rhee, 2007;
Freund, 2005; Auerbach, McGowan, Ausberger, Strolin-Goltzman & Schudrich, 2010), or-

NU

ganizational commitment (Boyas, Wind & Kang, 2012; Faller, Grabarek & Ortega, 2010;
Mor Barak et al., 2006), and burnout (Drake & Yadama, 1996; Kim & Lee, 2009 Huang,

Chuang & Lin, 2003; Schaufeli & Bakker, 2004; Strolin et al., 2005; Kim & Stoner, 2008;

MA

Ducharme, Knudsen & Roman, 2007; Boyas et al., 2012; Kim et al., 2012). Self-efficacy, included in a handful of studies, has also regularly been found to related to turnover intention

ED

(Chen & Scannapieco, 2010; Ellett, 2009; Ellett, Ellis, Westbrook & Dews, 2007; Kim & Lee,
2009; Weaver, Chang, Clark & Rhee, 2007).

PT

Psychological climate refers to individuals’ perceptions of their work environment and
how these perceptions drive their behavior and attitudes (James & James, 1989; Schneider,

CE

2006). Two psychological climate measures found to consistently predict turnover intentions
are perceived supervisor support (PSS) (Dickinson & Perry, 2003; Ellett et al., 2007; Nissly

et al., 2005; Fakunmoju, Woodruff, Kim, LeFevre & Hong, 2010; Mor Barak, Travis, Pyun

AC

& Xie, 2009; Tham, 2007; Faller et al., 2010; Strand & Dore, 2009; Landsman, 2001) and
perceived organizational support (POS) (Ellett, 2000; Tham, 2007; Rhoades & Eisenberger,
2002; Gillet, Gagné, Sauvagère & Fouquereau, 2012; Riggle et al., 2009; DePanfilis & Zlotnik,
2008).
While some studies of turnover intention in human service settings use conceptual models that explicitly consider both individual and contextual factors, with few exceptions (e.g.
Glisson & Green, 2006; Jinnett & Alexander, 1999; Smith, 2005), all work environment characteristics and individual attitudes are examined only at the individual level (Bliese & Jex,
2002). Failing to control for clustering at the level of the work unit or other organizational
levels limit our ability to explain child welfare staff attitudes and behaviors (Bliese & Jex,
2002; Chang, Wang & Huang, 2013). The failure to account for clustering at these levels may
2

ACCEPTED MANUSCRIPT

lead to biased results due to dependence of observations, an especially problematic outcome
in the estimation of uncertainty (e.g. p-values).


IP
T

The present study builds on existing research and theory to explain staff retention in
child welfare. It contributes to the workforce retention literature by (1) incorporating a

SC
R

model that assesses the influence of both individual and work group factors in child welfare
professionals’ intentions to stay in their current position, (2) introduces psychological safety
as a potential predictor of turnover intention, and (3) provides evidence for the utility of the

NU

psychological safety construct beyond multidisciplinary teams in health care setting. The
focus of this analysis is on the effects of perceived organizational support (POS), perceived

worker’s intention to stay.


ED

2. Theoretical Background

MA

supervisor support (PSS), and work unit psychological safety on front-line child welfare

Empirical and theoretical support for the association of psychological climate measures,

PT

work group psychological safety and worker intention to stay are summarized below.

CE

2.1. Psychological Climate Variables and Retention
An important distinction has been made between psychological climate (individual perceptions of the work environment) and organizational climate, including organizational work

AC


groups, and departments (collective perceptions of the work environment) (James & Jones,
1974). While general agreement exists on how psychological climate is defined, there is no
consensus on the content and specific dimensions comprising the construct (Parker et al.,
2003; Carr et al., 2003; Jones & James, 1979) nor whether climate is best conceived of as
a global or a facet-specific construct. Regardless, a wide range of facet-specific and global
work climate measures including justice, trust, civility, and employee voice have been found
to be related to staff retention.
2.2. Perceived Organizational Support (POS)
Organizational support theory (Eisenberger, Hungtington, Hutchison & Sowa, 1986;
Shore, Barksdale & Shore, 1995, e.g.) posits that in order to meet socio-emotional needs and

3

ACCEPTED MANUSCRIPT

to assess the organization’s readiness to reward increased efforts, employees form general beliefs concerning how much the organizations’ leadership values their contributions and cares

IP
T

about their well-being. Such perceived organizational support is assumed to be based on how
favorable an employee has been treated by the organization in the past (Eisenberger et al.,

SC
R

1986). Based on the norm of reciprocity, greater perceived organizational support is expected
to result in greater affective attachment and feelings of obligation to the organization (Shore
& Wayne, 1993).

NU

Current studies in social service settings provide evidence that perceived organizational
support plays an important role in the commitment process, helping to explain how basic
work experiences influence affective commitment and, ultimately, intention to stay. Tham

MA

(2007) examined several attitudes among child welfare workers and found the largest differences between public child welfare workers who intended to stay and those who intended

ED

to leave were in how they described the human resource orientation within the organization. Specifically, those who intended to leave seldom felt rewarded for a job well done or

PT

well taken care of, and perceived management as less interested in their health and wellbeing than those who intended to stay. Two additional studies in public child welfare found

CE

perceived organizational support, operationalized as administrative provision of adequate
resources to meet client needs and sensitivity to the needs and feelings of staff, to be a significant predictor of intention to remain employed in child welfare (Ellett, 2009; Westbrook

AC

et al., 2012). A third study in a public child welfare setting found no relationship between
job retention and organizational support (Smith, 2005).
Studies in the business literature provide further support for the relationship of perceived
organizational support to employee commitment. Rhoades & Eisenberger (2002), in their
meta-analysis, found perceived organizational support was negatively related to withdrawal
behaviors such as absenteeism, tardiness and turnover. Perceptions of supportive human
resources practices (e.g. participation in decision making, fairness of rewards, and growth
opportunities) have also been found to contribute to the development of perceived organizational support, and perceived organizational support was negatively related to withdrawal
(Allen et al., 2003). Eisenberger, Stinglhamber, Vandenberghe, Sucharski & Rhoades (2002)
found that employees who believed that the supervisor valued their contributions and cared
4

ACCEPTED MANUSCRIPT

about their well-being showed increased perceived organizational support, which in turn was
related to decreased turnover. Finally, a subsequent meta analysis provided additional sup-

IP
T

port with perceived organizational support accounting for nearly 25 percent of the variance

SC
R

in intention to leave (Riggle et al., 2009).
2.3. Perceived Supervisor Support (PSS)

Just as employees form global perceptions concerning their valuation by the organization (POS), they also develop perceptions of the degree to which supervisors value their

NU

contributions and care about their well-being (PSS) (Walumbwa et al., 2011). Child welfare studies consistently conclude that supportive supervision is associated with turnover

MA

intention, with the level of supervisory support significantly and inversely related to public
child welfare workers’ intentions to leave. Consistent with these findings, Dickinson & Perry
(2003) found PSS predicted intention to remain employed. Scannapieco & Connell-Corrick

ED

(2003) compared workers who stayed at the agency and those who left, and found those who
stayed spent more time with their supervisors than those who left. Thus, results from mul-

PT

tiple studies indicate that supervisors’ behavior has considerable potential to affect climate

CE

perceptions; perceptions that can influence employee’s intention to stay.
2.4. Team Psychological Safety

AC

Team psychological safety is defined as a shared belief that the team is a safe environment
for interpersonal risk taking (Edmondson, 1999). Psychologically safe teams are characterized by interpersonal trust, respect for the competence of all team members, and care and
concern about members as people. Psychological safety does not mean positive affect or
mutual liking but rather, a sense of confidence that the interpersonal consequences of wellintentioned risk will not be negative (Edmondson & Woolley, 2006). Psychological safety
has been found to influence work engagement (Vogelgesang, 2008; Rathert, Ishqaidef & May,
2009; May, Gilson & Harter, 2004), team learning (Edmondson, 1999; Schaubroeck et al.,
2011), involvement in continuous quality improvement efforts (Nembhard & Edmondson,
2006), and patient safety (Rathert et al., 2009).
Studies, primarily conducted in health care settings, affirm that employees’ perceptions of
their managers’ intentions and attitudes towards employees play an important role in shaping
5

ACCEPTED MANUSCRIPT

the degree of psychological safety. Leader behaviors identified as antecedents’ to psychological safety are leader inclusiveness, words and deeds exhibited that invite and appreciate

IP
T

others’ contributions in work groups (Nembhard & Edmondson, 2006) with inconsistency in
organizational conditions such as rewards, values, and evaluation systems predicting lower

SC
R

psychological safety (Lee et al., 2004).

A review of business and human service management literature identified only one study
that considered the influence of work unit psychological safety on turnover intention. Cal-

NU

lister (2006), investigated whether perceptions of department climate affected 308 science
and engineering faculty members’ intentions to quit. Departmental affective climate was
measured by aggregating four facets, one of which was psychological safety, and found a

MA

strong direct effect on intention to leave. In studies of a highly related construct, trust was
found to be negatively related to intention to quit Dirks & Ferrin (2002) and mediates the

ED

relationship between work-group incivility and turnover intentions (Miner-Rubino & Reed,
2010).

PT

2.5. Conceptual Model and Research Hypotheses
Based on a review of the literature, the hypothesized conceptual model links psychological

CE

climate measures with work unit climate. In this model psychological climate is represented
by two dimensions-perceptions of supervision quality and human resource orientation. Con-

AC

sistent with other studies, it was hypothesized that psychological climate measures would
be positively related to work group psychological safety (H1). Also employees experience
of work group psychological safety was expected to fully mediate the influence of supervisor quality and human resource primacy with employees having higher levels of work group
psychological safety having higher intentions to stay (H2). We also hypothesize an indirect
effect of psychological climate on worker intention to stay (H3).
3. Methods
3.1. Sample
The sample for this study was drawn from a population of public child welfare social
workers in Washington State in 2009. Workers were surveyed as a part of the implementa6

ACCEPTED MANUSCRIPT

tion of a statewide practice model for the child welfare system. All direct service workers
employed by the state were included in the survey. Workers received an email note from the

IP
T

agency director informing them of the survey and asking them to participate. Direct service
workers were defined as any worker with case management responsibilities focused on the

SC
R

investigation of allegations of child abuse or neglect, family preservation activities stemming
from an investigation of child abuse or neglect, or family reunification activities resulting
from the placement of a child in out-of-home care.

NU

Workers were informed that the survey was a part of the evaluation of the practice model
and that they were expected to complete the survey as part of their job. Subsequently they
were contacted by the agency’s research office and directed towards the web survey and were

MA

also told that offices that met a target response rate (90 percent) would be given a small
incentive payment by the agency that could be used for office social activities. Research office

ED

staff followed up directly with workers who had not completed the survey and informed unit
supervisors of their unit response rate; supervisors were not informed of which workers had

PT

completed the survey. Social worker surveys were collected via Survey Monkey, a web-based
survey tool that allows for the creation and deployment of survey instruments as well as the

CE

extraction of data from these instruments. Overall, the response rate for all social workers
in the agency (direct and indirect service workers) was 96 percent (n=1479). This number
includes workers who were not assigned to direct service roles and were thus not a part of

AC

this analysis (n=103). Administrative data was used for post hoc matching of social workers
in order to categorize respondents into supervisory work units. Some workers were not able
to be matched to a supervisory unit and were also excluded from this analysis (n=226).
Listwise deletion was used to exclude workers who did not respond to all of the identified
measurements and control variables (n=110). Based on these criteria, 1040 of the responding
social workers nested in 239 supervisory units were included in the study.
3.2. Measures
Workers responding to the survey items provided basic demographic information such
as age, gender, race, program areas (e.g. child protection, permanency planning, etc.),
and educational level. Additional standardized measurements were included in the survey

7

ACCEPTED MANUSCRIPT

addressing organizational psychological climate, psychological safety, and intention to stay
in the organization.

IP
T

Psychological climate measures were assessed using the QPS General Nordic Questionnaire for Psychological and Social Factors at Work (Lindström et al., 2000; Dallner, 2000) a

SC
R

tool developed to facilitate comparative epidemiological and work intervention studies. The
QPS Nordic focuses simultaneously on task, organizational, and individual level factors and
combines characteristics that have been traditionally measured by job stress and job redesign

NU

focused methods, such as the Job Content Questionnaire (JCQ) (Karasek et al., 1985) and
the Job Diagnostic Survey (JDS) (Hackman & Oldham, 1975).

MA

3.2.1. Perceived Organizational Support (POS)

Perceived organizational support involves employees’ global views about the extent to
which their supervisor values their contributions and cares about their well-being. Perceived

ED

organizational support was measured using the three item human resource primacy sub-scale
of the QPS Nordic General Nordic Questionnaire cited above. A representative item from

PT

the sub-scale reads as follows: “Workers are rewarded (money, encouragement) for a job well

CE

done”. Cronbach’s α for the sub-scale was .82.
3.2.2. Perceived Supervisor Support (PSS)

AC

PSS was originally conceptualized through two sub-scales from the QPS Nordic Questionnaire cited above: Quality of Supervision (three items), and Empowering leadership
(two items). The Quality of Supervision sub-scale assesses the extent to which leadership
is supportive, fair, and empowering. A sample item is, “If needed, can you get support
and help with your work from your immediate superior?” The empowering leadership subscale assesses the extent to which supervisors encourage participation in decision-making. A
representative item from the sub-scale reads as follows: “Does your immediate supervisor
encourage you to participate in important decisions?” The results of preliminary confirmatory factor analysis (using the sample outlined above) suggested that these sub-scales should
be collapsed into a single 5 item factor. Chronbach’s α for the combined sub-scale was .91.
We refer to this combined measurement as Supportive and Empowering Leadership (SEL).

8

ACCEPTED MANUSCRIPT

3.2.3. Team Psychological Safety (TPS)
Psychological safety was measured using the scale developed by Edmondson (1999). All

IP
T

seven of the original measures for psychological safety were included in the current survey and
no modifications were made to any of the items on the psychological safety measurement

SC
R

scale. The current survey measured psychological safety similarly to the Nordic derived
scales using a 5-point Likert scale with items ranging from 1 = never to 5 = always. This
measurement scale is a slight variation from the original Edmondson scale which utilized

NU

a 7-point Likert scale ranging from 1 = never to 7 = always. This change was made to
maintain congruence with the rest of the survey measurements. A representative item from

MA

this scale is as follows: “If you make a mistake in this unit, it is often held against you.”
The Cronbach’s α for psychological safety was .81.
3.2.4. Intention to Stay (ITS)

ED

Turnover intentions served as the dependent variable of the study and used the same
measures as Moynihan & Pandey (2007) in their study of turnover intention in twelve hu-

PT

man service organizations. Two items were used: “I often look for job opportunities outside
this organization”, and “I would be very happy to spend the rest of my career with this

CE

organization”. These items were each scored with a 5-point Likert scale ranging from 1 =
strongly agree to 5 = strongly disagree. Item (1) was reverse coded such that the measure-

AC

ment reflected an overall intention to stay with the organization. Cronbach’s α for turnover
intention was .75.

3.2.5. Control Variables
Based on our review of the literature several key control variables were identified as
important in our analysis of workers’ intention to stay with the agency. We specifically
included the age of the worker, and whether a worker had completed a graduate degree.
In addition we included minority status (i.e. non-white), professional social work training
(obtained a BA, BSW, or MSW), and at whether a worker was employed in child protection
or some other child welfare program. The program variable was included because child
protection investigative work is qualitatively different than other program areas in child
welfare and may differentially affect a worker’s intention to stay due to the levels of stress
9

ACCEPTED MANUSCRIPT

and conflict inherent in investigative work. Various two-way interactions between control
variables were assessed as a part of this analysis. All interactions had negligible effects

IP
T

on the parameter estimates and fit indices in our model. For the sake of simplicity and

SC
R

interpretation, we have chosen a model with no interactions as our final model.
3.3. Analysis Process

The first step of our analysis involved a review of descriptive statistics for our main
measurements and control variables. These statistics are displayed in Table 1. Median

NU

values for all scales are displayed along the diagonal of the table. We calculated median
values instead of means due to the ordinal nature of our data. Mean values are calculated

MA

for control variables which yields an average value for age and a proportion for all other
binary control variables. Spearman correlation coefficients and the associated significance
levels are also displayed in the table. In addition to these descriptive statistics we assessed

ED

the intraclass correlation coefficients (ICC) for all of our measurements to determine the
extent to which clustering was present at the supervisory unit level.

PT

The next step in our analysis involved the completion of a confirmatory factor analysis
(CFA) to assess the proposed measurement model. As indicated above, the results of the

CE

CFA suggested that quality of supervision and empowering leadership sub-scales should be
collapsed into a single measurement, supportive and empowering leadership. The rest of

AC

the proposed measurement model (i.e. psychological safety, intention to stay, and human
resource primacy) was confirmed by the CFA. The results of our descriptive analysis and the
CFA led us to test the proposed theoretical relationships using a structural equation model
(SEM). Parameter estimates were calculated using the weighted least squares means and
variance algorithm as implemented in Mplus. This algorithm was chosen due to the ordinal
nature of our manifest variables.
4. Results
4.1. Sample Description
The majority of respondents in the study sample were white (71 percent), over the age
of 40 (48 percent), and female (77 percent). Supervisory units ranged in size from 1 to
10

ACCEPTED MANUSCRIPT

9 workers with a mean of 4.4 and a median of 4. About 53 percent of the sample had
completed a graduate degree and 68 indicated that they had received a social work degree.

IP
T

Approximately 27 percent of the sample indicated that they were providing child protection

SC
R

services as opposed to other child welfare service areas (e.g. permanency planning, etc.).
4.2. ICC Results

The results of our ICC analysis indicated significant group-level clustering in the measurements of human resource primacy (ICC = .06, F = 1.35, p < .01), supportive and

NU

empowering leadership (ICC = .16, F = 1.93, p < .01), and psychological safety (ICC =
.22, F = 2.31, p < .01). No significant clustering was observed in the turnover intention

MA

construct (ICC = .03, F = 1.13, ns). Given that our primary interest in this analysis was
to better understand intention to stay, we decided to still test a single-level SEM. However,
given that it is still important to account for the group-level clustering of our other measure-

ED

ment variables, we use a sandwich estimator to calculate the standard errors in our model.
In this way, we do not seek to answer specific questions about the unit-level relationships

PT

between our measurement variables. However, we do control for the non-independence of
observations suggested by our ICC analysis and ensure that we can place a high degree of

AC

CE

confidence in any indicators of uncertainty in our results.

11

ACCEPTED MANUSCRIPT

1.

2.

3.

4.

5.

6.

7.

8.

9.



2. HRP

.32***



3. TPS

.53***

.24***



4. ITS

.24***

.41***

.25***



5. AGE

-.05(ns)

-.04(ns)

.03

.14***



6. PRF

.00(ns)

-.06(ns)

.03(ns)

-.1**

-.11***

7. EDU

.01(ns)

-.04(ns)

-.03(ns)

-.14***

-.03(ns)

.01(ns)



8. GDR

-.02(ns)

-.06(ns)

.00(ns)

.00(ns)

-.12***

.02(ns)

.04(ns)



9. PGM

-.05(ns)

-.01.03

-.09**

.01(ns)

-.03(ns)

.02(ns)

.00(ns)

-.05(ns)



10. RAC -.01(ns)

.00(ns)

-.09**

-.07*

2.3

4

3.5

4.2

SC
R


NU

Median

IP
T

1. SEL

10.

-.09**

-.01(ns)

.03(ns)

.00(ns)

.03(ns)



41.8

0.68

0.53

0.77

0.27

0.29

MA

Notes: SEL = supportive and empowering leadership, HRP = human resource primacy, TPS = team
psychological safety, ITS = intention to stay, AGE = worker age, PRF = professional affiliation (i.e.
social work degree) flag, EDU = graduate degree flag, GDR = worker gender, PGM = investigative

ED

work flag, RAC = minority status.
* p