Purpose: This study used Structural Equations Modeling (SEM) to empirically validate a conceptual framework that links child welfare (CW) employees' intentions to remain employed (IRE) to personal characteristics, and organizational factors framing the CW work setting. Our approach was on employee retention as an alternative to more traditional studies of employee burnout,

Methodology: A 198 item survey of all professional CW staff in a state agency was completed by 1423 employees (62.3%) in the fall of 2002. The survey included measures of intentions to remain employed in CW, self-efficacy beliefs (SE), human caring (HC) and professional organizational culture (POC). A two-stage, multivariate model using LISREL (Jorskog & Sorbom, 1996) was completed to confirm the factor structure of the survey variables and to empirically validate the conceptual framework guiding the study (Ellett, Ellett, & Rugutt, 2003). Primary research questions framing these analyses were:

1. is there confirmatory evidence for the measurement models originally developed for the survey variables using exploratory factor analysis? 2. which variables in a SEM provide the strongest and most meaningful linkages to CW employees' IRE?

LISREL was first used to statistically confirm the measurement model for each of the survey measures identified in previous exploratory factor analyses. Subsequently, LISREL was used to develop and validate a SEM linking personal and organizational variables to IRE.

Results: Confirmatory factor analyses showed that: a) 9 factors demonstrated perfect fit to their measurement models; (b) the fit statistics for the remaining survey factors were all within acceptable limits (Bentler, 1993; Schumaker & Lomax, 1996) with Alpha reliabilities ranging from .63 to .94. SEM results showed that professional commitment (an element of the HC measure) had the most direct effect on the IRE dependent measure [standardized regression coefficient (SRC) =.71], followed by the POC.measure (SRC=.20). Interestingly, the SRC linking the SE and IRE measures was -.11. All three SRC's were statistically significant (p<.001). The HC, POC and SE measures demonstrated moderately strong, positive relationships with each other (r=.31 to .38, p<.001).

Implications: The results have implications for theory development, future research, policy making, and practice in CW. This study presents a rich alternative to recently published articles that continue to focus on burnout as a the cause of CW employee turnover (Drake & Yadama, 1996). This study, the largest known statewide study of its kind, changes the empirical and theoretical focus to understanding personal and organizational factors that contribute to CW employee retention. We believe this conceptual and empirical focus can yield new information important for developing and strengthening the holding power of CW organizations for employees (and this reduce turnover). The measures developed can be used in a variety of future studies in CW, and the SEM developed is an empirically-derived framework for future studies of the important, national workforce problem of retaining CW employees. The finding that professional commitment has the most direct effect on retention has practical implications for the preparation, selection, mentoring, and retention of CW professionals. These implications and associated issues are discussed in the full paper.