The complete model

Turnover Intention

Every factor researchers have brought together to explain turnover intention, combined from 16 study models into one ranked driver set — the model the priority engine reasons over.

30 drivers · 16 contributing models · Handbook entry →

§1The drivers, by consensus

Ranked by how many independent models include each driver (consensus), then pooled effect. “In N of 16” 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
Job Satisfaction3 / 16−0.52A
Organizational Commitment3 / 16−0.48AI²≈1.00
Work Family Conflict3 / 16+0.32AI²≈1.00
Procedural Justice2 / 16−0.45A
Work Engagement2 / 16−0.45AI²≈0.96
Work Life Conflict Wlc2 / 16+0.38AI²≈0.99
Work Interference With Family Wif2 / 16+0.27AI²≈0.99
Affective Commitment1 / 16−0.56A
Psychological Contract Breach1 / 16+0.54A
Job Embeddedness1 / 16−0.47B
Person Job Fit1 / 16−0.46A
Organizational Justice1 / 16−0.44B
Psychological Contract1 / 16−0.43A
Burnout1 / 16+0.41B
Perceived Organizational Support1 / 16−0.40AI²≈1.00
Distributive Justice1 / 16−0.40B
Career Development1 / 16−0.39C
Transformational Leadership1 / 16−0.38A
Interpersonal Justice1 / 16−0.36A
Perceived Supervisor Support1 / 16−0.36A
Supervisor Support Ss1 / 16−0.36A
Informational Justice1 / 16−0.36A
Servant Leadership1 / 16−0.34A
Person Organization Fit1 / 16−0.32A
Family Work Conflict1 / 16+0.26A
Interactional Justice1 / 16−0.24B
Person Group Fit1 / 16−0.22A
Colleague Support1 / 16−0.19A
Power Distance1 / 16+0.09B
Collectivism1 / 16−0.05B

“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:10.1108/k-01-2022-01195A
doi:10.3389/fpubh.2025.16428434A
doi:10.1111/j.1744-6570.2005.00672.x3A
doi:10.1177/10596011073133073A
doi:10.1002/nop2.8723B
doi:10.1111/j.1744-6570.1993.tb00874.x2A
doi:10.15240/tul/001/2020-1-0072A
doi:10.1155/2023/33566202B
doi:10.1037/a00189382B
doi:10.3389/fpsyg.2022.8725682A
doi:10.1016/j.tate.2021.1034252B
doi:10.1006/jvbe.2001.18422A
doi:10.1037/0021-9010.86.3.4252A
doi:10.1111/jan.148462B
doi:10.1037/a00221702A
doi:10.1186/s12912-023-01496-22A
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.