The complete model

Voluntary Turnover

Every factor researchers have brought together to explain voluntary turnover, combined from 3 study models into one ranked driver set — the model the priority engine reasons over.

25 drivers · 3 contributing models · Handbook entry →

§1The drivers, by consensus

Ranked by how many independent models include each driver (consensus), then pooled effect. “In N of 3” is a robustness signal — a driver in many models is one many researchers agreed matters; a driver in one is a single lead worth following.

DriverConsensusPooled rGradeHeterogeneity
Organizational Commitment3 / 3−0.27AI²≈0.98
Job Satisfaction3 / 3−0.23AI²≈0.99
Turnover Intention2 / 3+0.48AI²≈1.00
Job Search2 / 3+0.40A
Met Expectations2 / 3−0.16A
Person Organization Fit1 / 3−0.29A
Job Embeddedness1 / 3−0.26A
Perceived Job Alternatives1 / 3+0.23A
Emotional Exhaustion1 / 3+0.21A
Overall Job Performance1 / 3−0.21A
Role Clarity1 / 3−0.21A
Tenure1 / 3−0.20A
Work Engagement1 / 3−0.20A
Job Involvement1 / 3−0.19A
Perceived Organizational Support1 / 3−0.19A
Work Family Conflict1 / 3+0.19A
Psychological Contract Breach1 / 3+0.18A
Compensation1 / 3−0.17A
Organizational Justice1 / 3−0.17A
Role Ambiguity1 / 3+0.15A
Role Conflict1 / 3+0.15A
Internal Mobility1 / 3−0.11A
Age1 / 3−0.09A
Pay Level1 / 3−0.09A
Pay Satisfaction1 / 3−0.06

“Pooled r” is the meta-analytic prior where one exists; drivers marked study-reported have a model but not yet a pooled prior — a concrete acquisition target. This is the value-of-information view in miniature: the priority engine can rank which driver to measure first by consensus, strength, and how much is still unknown.

§2The models it’s built from

Each contributing study model — the researcher’s theory of this outcome — with its size and grade, so the meta-model is auditable back to its sources.

Source studyPredictorsGrade
doi:10.1111/peps.1222619A
doi:10.1177/01492063000260030511A
doi:10.1111/j.1744-6570.1993.tb00874.x2A
Composed live from the registry. The complete model is the union of all study-level models of this outcome; study-level models are derived from the effect corpus and re-derived every loop cycle.