Predicting Retention: Using Engagement and Exit Survey Data to Understand Workforce Turnover in Child Welfare

Background and Purpose:
Child welfare agencies face mounting challenges due to high staff turnover and low employee engagement, which threaten service effectiveness and continuity. This study aims to identify predictive factors from employee engagement surveys that are associated with workforce retention. Using Generalized Linear Modeling (GLM), we analyze responses from current and former employees to uncover key items and themes that differentiate those who stay from those who leave. The research seeks to answer: What engagement survey items predict worker retention? How do these predictors differ between those who remain and those who exit? Findings will inform targeted strategies to improve retention.

Methods:
The study utilizes data from an ongoing employee engagement survey administered by a large statewide child welfare agency, alongside an exit survey completed by employees leaving the voluntarily or involuntarily. The sample consists of roughly 2,000 employees who separated from employment over a two-year period. Survey items covered key areas such as organizational support, leadership effectiveness, job satisfaction, workload, and opportunities for growth. Statistical analysis included paired and independent samples t-tests to explore mean differences between engagement and exit responses and Generalized Linear Modeling (GLM) to identify items most strongly associated with retention. By comparing engagement patterns of those who left versus those who stayed, the study highlights specific predictive factors to inform retention-focused strategies.

Results:
Initial analysis reveals significant differences in perceptions across multiple employee subgroups. Notably, distinct attitudinal patterns emerged between employees who indicated on the engagement survey that they would not leave but ultimately did, and those who said they would stay and remained employed. Similarly, contrasts were found between those who expressed an intention to leave but did not, and those who both indicated they would leave and followed through. Additional variation in engagement scores was observed between individuals who exited voluntarily versus those who were separated involuntarily. These subgroup comparisons underscore the complexity of employee intent and behavior, and the importance of examining both expressed intent and actual turnover. Generalized Linear Modeling further identified several areas—particularly around organizational trust, communication, supervisor support, and workload—as highly predictive of retention behavior.

Conclusions and Implications:
As child welfare organizations continue to grapple with retaining a stable and engaged workforce, leadership must be equipped with a deeper understanding of the factors influencing employee departure. This study provides evidence that employee perceptions that play a significant role in predicting turnover. One practical implication is the development of organizational policies that promote structured career development plans and transparent advancement opportunities, which can boost engagement and decrease voluntary exits. Currently, the literature lacks comprehensive studies that integrate engagement and exit survey data over time using predictive analytics, particularly within the child welfare sector. This study fills that gap and offers actionable insights for data-informed policy and practice decisions aimed at improving workforce stability.