Safety climate — shared employee perceptions of how much an organization genuinely prioritizes safety — is one of the most heavily used leading indicators of workplace injuries, yet its causal predictive validity remains incompletely established. Johnson (2007) addressed, with a prospective design in heavy manufacturing, the paper's own stated problem that the predictive validity of safety climate "has not been firmly established," showing that the 16-item Zohar Safety Climate Questionnaire predicted subsequent safety behavior and accident experience. Later meta-analyses and cross-validation studies substantially established the correlational predictive validity and mediation of the pathway in aggregate, refined the temporal window over which prediction is valid, and extended the instrument to many industries and countries with contrasting factor-structure findings; the same literature left open, as field-level questions, whether the mediated pathway from climate through behavior to injury frequency holds prospectively with adequate statistical power and in which causal direction, whether the instrument's measurement model generalizes across industries and cultures, whether safety climate should be scored as one global factor or separable dimensions, and whether improving climate actually causes injury reduction. What remains open is the causal translation and the measurement model: whether improving safety climate produces sustained reductions in injuries through behavior change, together with a generalizable measurement model and settled empirical criteria for uni- versus multi-dimensional scoring.

Background

Safety climate denotes employees' shared perceptions of the priority that safety genuinely receives in an organization, relative to competing goals such as productivity. The construct was introduced by Dov Zohar in 1980 and is typically measured by questionnaires; the most widely used instrument family descends from Zohar's work, including the 16-item Zohar Safety Climate Questionnaire (ZSCQ) from Zohar and Luria's (2005) multilevel model, which distinguishes organization-level climate (policies and management commitment) from group-level climate (direct supervisory practices). A leading indicator, in this literature, is a measurable condition that predicts future safety outcomes — as opposed to lagging indicators such as recorded injury rates, which only report the past.

The scientific problem is that an instrument can correlate with outcomes without being a credible basis for prediction or intervention. Johnson (2007) addressed this in a heavy-manufacturing organization: 292 employees at three sites completed the 16-item ZSCQ, and safety behavior and accident experience were then tracked for five months and analyzed with structural equation modeling (a statistical technique that estimates networks of hypothesized causal paths among measured and latent variables, allowing tests of mediation — the claim that climate affects injuries indirectly, through safety behavior). The study found the questionnaire psychometrically sound, able to predict behavior and accident experience, and trimmable to 11 items with little loss of explanatory power; the paper itself framed this contribution as "additional support" for safety climate as a predictive indicator of safety-related outcomes, against its stated problem that predictive validity "has not been firmly established." However, the evidence came from a single organization and a purely observational design that could not exclude reverse causation — injuries feeding back into perceptions — or third-variable explanations, and later validation attempts reported contrasting factor structures for the instrument across domains: three supervisor factors in Johnson's heavy-manufacturing data and a superior single-factor fit in another safety-critical domain (as reviewed by Curcuruto et al. 2018), leaving the multi-factor versus single-factor representation unresolved. Predictive validity had therefore been suggested, not firmly established.

Subsequent literature substantially consolidated the correlational picture. Meta-analyses (studies that statistically aggregate effect sizes across many primary studies) by Clarke (2006), Christian et al. (2009), Beus et al. (2010), and Nahrgang et al. (2011) established that safety climate, especially group-level climate, is among the strongest situational predictors of safety behavior and accidents, and supported a meta-analytic climate → behavior → accident path model; Beus et al. (2010) also found the effects on injuries to be small, sometimes matched by reverse injury → climate paths, with perceived management commitment the most robust facet. Zohar (2010), reviewing thirty years of research, judged safety climate validated as a leading indicator across industries and countries and redirected the field toward antecedents — a program since quantified by He et al. (2019). Bergman et al. (2014) showed that a climate assessment's predictive relationship with severe safety-critical incidents expires within roughly three months, so aggregation windows and survey cadence matter. Cross-validation and meta-analytic work in rail (Curcuruto et al. 2018), construction (Alruqi et al. 2018), and across industries (Jiang et al. 2019; Casey et al. 2022) found the instrument's dimensionality to vary by context, with industry-specific scales capturing genuine nuance but no settled criteria for when to score climate as one global factor or as separable supervisory and organizational dimensions. On the causal side, Zohar and Polachek (2014) demonstrated in a randomized field study that feedback to supervisors on their safety-versus-productivity discourse improved safety climate, safety behavior, and independently audited performance; Lee et al. (2019), systematically reviewing the safety-climate intervention literature, located only 19 studies, mostly quasi-experimental, whose typical endpoint was an improved climate score rather than reduced injuries.

Problem Statement

Does safety climate function as a causally valid leading indicator of workplace injuries? Building on Johnson's (2007) prospective study in heavy manufacturing — which addressed the paper's stated problem that the predictive validity of safety climate "has not been firmly established" — and on the open field-level questions documented in the subsequent literature cited below, the problem is to specify and justify the research program — its study designs, analyses, and decision criteria — that would (1) test, with adequate statistical power at the group level, whether the mediated pathway safety climate → safety behavior → injury frequency holds prospectively, including whether climate predicts injuries or injuries reshape climate perceptions; (2) establish which candidate measurement model for the Zohar Safety Climate Questionnaire — three-factor, trimmed 11-item, single-factor, or industry-specific — generalizes across industries, occupational groups, and cultural contexts, together with empirical criteria for choosing between uni-dimensional and multi-dimensional scoring for prediction and intervention; and (3) establish whether improving safety climate produces measurable, sustained improvements in safety behavior and reductions in injury outcomes. These three components are jointly required — none alone resolves the task, and (1) and (3) furnish complementary observational and interventional evidence for the causal-translation question — and together they serve one objective: determining whether safety climate can be converted from an established correlate of safety outcomes into a validated, actionable leading indicator whose improvement can be shown to prevent injuries.

References

  1. Johnson, S. E. (2007). The predictive validity of safety climate. Journal of Safety Research, 38(5), 511–521. doi:10.1016/j.jsr.2007.07.001, https://doi.org/10.1016/j.jsr.2007.07.001
  2. Zohar, D., & Luria, G. (2005). A multilevel model of safety climate: Cross-level relationships between organization and group-level climates. Journal of Applied Psychology, 90(4), 616–628. doi:10.1037/0021-9010.90.4.616, https://doi.org/10.1037/0021-9010.90.4.616
  3. Clarke, S. (2006). The relationship between safety climate and safety performance: A meta-analytic review. Journal of Occupational Health Psychology, 11(4), 315–327. doi:10.1037/1076-8998.11.4.315, https://doi.org/10.1037/1076-8998.11.4.315
  4. Christian, M. S., Bradley, J. C., Wallace, J. C., & Burke, M. J. (2009). Workplace safety: A meta-analysis of the roles of person and situation factors. Journal of Applied Psychology, 94(5), 1103–1127. doi:10.1037/a0016172, https://doi.org/10.1037/a0016172
  5. Beus, J. M., Payne, S. C., Bergman, M. E., & Arthur, W. (2010). Safety climate and injuries: An examination of theoretical and empirical relationships. Journal of Applied Psychology, 95(4), 713–727. doi:10.1037/a0019164, https://doi.org/10.1037/a0019164
  6. Zohar, D. (2010). Thirty years of safety climate research: Reflections and future directions. Accident Analysis & Prevention, 42(5), 1517–1522. doi:10.1016/j.aap.2009.12.019, https://doi.org/10.1016/j.aap.2009.12.019
  7. Nahrgang, J. D., Morgeson, F. P., & Hofmann, D. A. (2011). Safety at work: A meta-analytic investigation of the link between job demands, job resources, burnout, engagement, and safety outcomes. Journal of Applied Psychology, 96(1), 71–94. doi:10.1037/a0021484, https://doi.org/10.1037/a0021484
  8. Zohar, D., & Polachek, T. (2014). Discourse-based intervention for modifying supervisory communication as leverage for safety climate and performance improvement: A randomized field study. Journal of Applied Psychology, 99(1), 113–124. doi:10.1037/a0034096, https://doi.org/10.1037/a0034096
  9. Bergman, M. E., Payne, S. C., Taylor, A. B., & Beus, J. M. (2014). The shelf life of a safety climate assessment: How long until the relationship with safety-critical incidents expires? Journal of Business and Psychology, 29(4), 519–540. doi:10.1007/s10869-013-9337-2, https://doi.org/10.1007/s10869-013-9337-2
  10. Alruqi, W., Hallowell, M., & Techera, U. (2018). Safety climate dimensions and their relationship to construction safety performance: A meta-analytic review. Safety Science, 109, 165–173. doi:10.1016/j.ssci.2018.05.019, https://doi.org/10.1016/j.ssci.2018.05.019
  11. Curcuruto, M., Griffin, M. A., Kandola, R., & Morgan, J. I. (2018). Multilevel safety climate in the UK rail industry: A cross validation of the Zohar and Luria MSC scale. Safety Science, 110, 183–194. doi:10.1016/j.ssci.2018.02.008, https://doi.org/10.1016/j.ssci.2018.02.008
  12. Jiang, L., Lavaysse, L., & Probst, T. M. (2019). Safety climate and safety outcomes: A meta-analytic comparison of universal vs. industry-specific safety climate predictive validity. Work & Stress, 33(1), 41–57. doi:10.1080/02678373.2018.1457737, https://doi.org/10.1080/02678373.2018.1457737
  13. He, Y., Wang, Y., & Payne, S. C. (2019). How is safety climate formed? A meta-analysis of the antecedents of safety climate. Organizational Psychology Review, 9(2–3), 124–156. doi:10.1177/2041386619874870, https://doi.org/10.1177/2041386619874870
  14. Lee, J., Huang, Y.-H., Cheung, J. H., Chen, Z., & Shaw, W. S. (2019). A systematic review of the safety climate intervention literature: Past trends and future directions. Journal of Occupational Health Psychology, 24(1), 66–91. doi:10.1037/ocp0000113, https://doi.org/10.1037/ocp0000113
  15. Casey, T. W., Hu, X., Kanse, L., & Varhammar, A. (2022). A tale of six climates: Reflections and learnings after the development of six industry-specific safety climate scales. Journal of Safety Research, 82, 151–158. doi:10.1016/j.jsr.2022.05.006, https://doi.org/10.1016/j.jsr.2022.05.006