Research That Matters (January 17 - 20, 2008)


Diplomat Ballroom (Omni Shoreham)

Refining the Prediction of Turnover Risk in Child Welfare Workers

Steven L. McMurtry, PhD, University of Wisconsin-Milwaukee, Susan J. Rose, PhD, University of Wisconsin-Milwaukee, M.J. Brondino, PhD, University of Wisconsin-Milwaukee, and Joshua P. Mersky, PhD, University of Wisconsin-Milwaukee.

Purpose: Turnover among child welfare staff remains alarming high in many agencies, reaching up to 90% annually. In addition to threatening service quality, turnover increases costs as agencies must constantly train replacement staff. Also, since child welfare is an entry point for many beginning social workers, quitting these jobs may increase their risk of leaving the profession. Research indicates that turnover arises from many factors that act in complex ways. This paper reports results from a study that measured a richer array of predictor variables than prior works and tested a path model that refines existing knowledge of how these variables operate.

Methods: Data are analyzed from a cross-sectional survey of 293 employees of a large, urban child welfare system. The response rate was 57%, and respondents included both public- and private-agency employees. Information was collected on variables in six categories. With example variables shown in parentheses, the categories were: Job Stressors (amount of work, role conflict), Interpersonal Variables (supervisor support, loyalty to others), Identification with Work (job rewardingness, job commitment), Personality Characteristics (psychological hardiness), Workplace Factors (promotion opportunities, organizational supportiveness), and exogenous controls (age, race/ethnicity). Each category was tested for its capacity to predict turnover risk and three intermediate variables--emotional exhaustion, organizational commitment, and job satisfaction. Measures included existing scales such as the Personal Values Survey, Job Descriptive Index, and Maslach Burnout Inventory, plus new scales developed to measure constructs such as job difficulty and perceived workplace safety. The latter scales were constructed through exploratory factor analysis and internal consistency checks.

Results: After testing and refining an initial path model, a more parsimonious version was estimated. All relevant indices indicated that this model fit the data well (÷2(10)=13.71, p>.18; CFI>.99; TLI>.98; RMSEA<.04; SRMR=.02). Job Stressor variables were strongly predictive of emotional exhaustion, while Personality Characteristics were weakly so. Workplace Factors and Identification with Work were modestly predictive of job satisfaction and more strongly predictive of organizational commitment. Among intermediate variables, organizational commitment had a moderate positive effect on job satisfaction, while the effect of emotional exhaustion on job satisfaction was moderate but negative. Turnover risk increased along with scores for emotional exhaustion, whereas increases in job satisfaction and organizational commitment decreased turnover risk. About 46% of the variation in turnover risk, 28% in organizational commitment, 50% in job satisfaction, and 58% in emotional exhaustion was explained by the model.

Implications: The results help resolve conflicts in prior findings, identify new predictors, and illuminate pathways taken by factors affecting turnover risk. The effect of emotional exhaustion on turnover risk was particularly strong, as was the effect of job stressors on emotional exhaustion. Efforts to mitigate these stressors may thus prove fruitful. Results also suggest that organizational commitment is slightly more predictive of turnover risk than is job satisfaction, and variables that predict organizational commitment (e.g., promotion opportunities) do so more strongly than those affecting job satisfaction. Finally, the model shows that first-level predictors operate on turnover risk through intermediate variables rather than directly.