Validating the Perceptions of Child Welfare Workforce Scale

Background and Purpose:  Most of the information the public receives about child welfare work is through media sources (Gainsborough, 2010, Landsman, 2001).  Not surprisingly, media attention of high profile cases leads the public to develop negative perceptions and mistrust of the child welfare system and the individual workers who handle cases (Ellett, Ellis, Westbrook & Dews, 2007).  These negative views towards child protection impact the agencies, individuals who work within the system, and the working conditions within agencies.  Specifically, public perception of child welfare has been shown to impact the psychological climate within agencies, worker behavior, worker burnout, job insecurity, and the recruitment and retention of a qualified workforce (Westbrook, Ellis & Ellett, 2006).  The current study seeks to better understand the validity of the Perception of Child Welfare Work (PCWW), a 29-question instrument designed to assess child welfare workers’ perceptions on how the public views them and the work they do.  This instrument has been used in studies that have assessed workforce issues in child welfare.

Methods:  Data for this study was obtained from a sample of 482 child welfare workers employed in voluntary child welfare agencies in a large northeastern city.  Voluntary agencies in this locale are private agencies that are under contract with the city and provide preventive services.  Workers in the sample represent various roles within the agencies and included administrators, supervisors, social workers, caseworkers, and case planners.  Cronbach’s alpha was used to assess the internal reliability of the scale and was found to be 0.72. A total of seven models were tested in this study based upon theoretical constructs identified in the literature.   Confirmatory factor analysis (CFA) was used to identify the best fitting model and better understand the validity of the PCWW.  Stata 12.0 was used to conduct the analysis.

Results:  The best fitting model consisted of a three factor solution comprised of fourteen observed exogenous variables.  The first factor, respect, was made up of four items.  Factor loadings for respect ranged from 0.65 to 0.86.  Coefficient alpha for this subscale was 0.83.  The second factor, stigma, consisted of four items whose factor loadings ranged from 0.60 to 0.74.  Coefficient alpha for this subscale was 0.82.  The third factor, value, consisted of six items with factor loadings ranging from 0.54 to 0.81.  Coefficient alpha for this subscale was 0.76. Correlations between the factors were 0.02, 0.45, and -0.43.  Fit statistics for the entire model indicate a good fit with the data (X² = 230.55, p=0.00; RMSEA=0.07; CFI=0.93; TLI=0.92).

Conclusions and Implications:  High factor loadings and small standard errors ranging from 0.021 to 0.037 for observed exogenous variables for each of the factors indicate the presence of convergent validity.  Additionally, correlations between the factors indicate that each of these is independent from the others, indicating good discriminate validity for the subscales.  The reliability and validity of this parsimonious model will allow researchers to consider the impact of public perceptions of child welfare in future studies.