On this page · 10 sections
A safety climate score can help reveal how workers experience safety priorities in everyday work. It can also conceal important differences if the survey, participants and scoring rules disappear behind a single number. The useful question is what the result tells the organization to investigate and improve.
This article explains how to read safety climate results alongside evidence about actual work. NOSACQ-50 provides a specific example of a research-based questionnaire; the interpretation questions and fictional calculation below are my practical discussion aids, not a new validated instrument.
What a safety climate score measures
The NOSACQ-50 developers describe safety climate in terms of a workgroup’s shared perceptions of management and worker safety policies, procedures and practices. A survey asks how those priorities are experienced. It does not directly test a machine guard, measure airborne exposure or establish whether an isolation procedure functions.
A favourable result can inform a discussion about leadership and working arrangements. It cannot replace the evidence needed for a particular protective decision. Likewise, an unfavourable result deserves examination without treating it as a diagnosis of an individual worker or proof of misconduct.
| Evidence | Question it can address | What it does not establish alone |
|---|---|---|
| Safety climate responses | How do respondents experience safety priorities, communication and participation? | That every hazard is controlled or that all workers share the same experience. |
| Control-performance evidence | Does the specified protection function under the relevant operating conditions? | That workers trust the reporting or decision process. |
| Injury and incident records | What events were recorded during a defined period and exposure? | That an absence of recorded harm proves effective controls or open reporting. |
Identify the instrument before interpreting the number
Kines and colleagues (2011) developed NOSACQ-50 as a safety climate questionnaire. The current developer description identifies 50 items across seven dimensions. Three concern management priorities, empowerment and fairness; four concern worker commitment, risk priorities, communication and confidence in safety arrangements.
Read the instrument name, language version, intended respondents and dimensions. A short local poll, a modified questionnaire and the original NOSACQ-50 are not automatically interchangeable. Changing wording, response options or included items may change the measurement. Document the changes and assess their implications rather than inheriting a validation claim from the source instrument.
Prior validation supports a specified instrument and its studied uses. It does not independently validate this article, a new dashboard formula or every application in another workforce. Be equally precise about whether a research paper measured perceptions, self-reported behaviour, observed performance or injuries.
Check coding, missing responses and dimension means
The developer’s scoring instructions use a four-point response scale, including positively and negatively worded items. Negatively worded items are reverse-coded: a response of 1 becomes 4, 2 becomes 3, and so on, using 5 minus the response. This is a scoring transformation, not a correction of the participant’s opinion.
The instructions calculate a person’s mean within each dimension from answered, correctly coded items, then average eligible person-level means for that dimension. Responses are excluded for a dimension when fewer than half its items are answered. A blank response is not zero and should not silently become agreement, disagreement or a favourable answer.
These details matter before any comparison. Check the actual version and scoring instructions, preserve the seven-dimension profile, and explain any departure from the prescribed method. A reliability statistic concerns measurement consistency; it does not demonstrate that workplace controls function.
Ask whose experience the result represents
Record who was eligible, who was invited, who responded and which responses were usable. Consider differences among shifts, contractors, tasks, languages and supervisory roles. A high response count from one group cannot supply the views missing from another.
A mean also does not show how closely people agree. The same average can come from similar responses or sharply divided experiences. Examine the distribution and relevant group structure before describing an organizational score as a shared climate. Obtain suitable statistical support when making group-level or uncertainty claims.
Keep manager and worker perspectives distinguishable where the instrument and assessment plan require it. A difference can help identify a question about how decisions are experienced; it is not a basis for deciding which group supplied the “correct” opinion.
Worked example: one average, two different experiences
This is an original fictional example. The numbers are invented, are not NOSACQ-50 field results and do not describe an actual employer. Suppose a hypothetical four-point questionnaire produces the following results for one dimension. All responding workers have complete responses, the same scoring rules apply, and each respondent has equal weight.
| Group | Invited | Responded | Dimension mean |
|---|---|---|---|
| Shift A | 100 | 80 | 3.60 |
| Shift B | 40 | 20 | 2.40 |
| Combined respondents | 140 | 100 | 3.36 |
The combined mean is (80 × 3.60 + 20 × 2.40) ÷ 100 = 3.36. That calculation weights respondents equally; it does not give the two shifts equal weight. Looking only at 3.36 would hide the lower reported experience on Shift B.
The overall response rate is 100 ÷ 140, or approximately 71.4%. It is 80% on Shift A and 50% on Shift B. Those proportions identify missing participation; they do not establish whether nonrespondents would answer similarly. Without response distributions and an appropriate analysis, the table also does not establish statistical significance or agreement within either shift.
Dividing 3.36 by 4 produces 84%, but that is not 84% compliance, 84% of hazards controlled or an 84% probability of avoiding an injury. The scale measures a different construct. These invented groups also do not establish a privacy threshold for publishing real subgroup results.
Treat benchmarks as context, not certification
NFA’s interpretation guidance explicitly distinguishes participants’ perceptions from actual conditions. It presents preliminary interpretation guidelines for individual dimensions and warns that its international database is not a representative sample. A position relative to that database is not a population ranking or an independent safety certification.
Check whether a comparison uses the same instrument, dimension, respondent role, language, industry and operating context. Describe important differences. The fictional example above does not apply the developer’s interpretation categories to its invented questionnaire or propose a new pass mark. Keep the result attached to the question it can answer. A numerical conversion or a favourable comparison cannot turn perceptions into a verified percentage of legal conformity or control effectiveness.
A stable total can conceal changing work
In their published abstract, Niu and Leicht (2024) report stable overall construction safety climate scores across project phases alongside item-level sensitivity and interview findings about trade-specific and phase-specific challenges. Their combined approach shows why an overall average needs contextual interpretation. This discussion is limited to the abstract; it is not a full appraisal of that study.
For repeated surveys, record changes in the work, respondent mix, instrument and collection method. Ask whether a change in score could reflect different participation or conditions. If several interventions occurred together, a before-and-after mean alone does not isolate the contribution of each one.
Even where research reports an association with behaviour or injuries, preserve the study’s actual outcome, design and uncertainty. A fitted statistical relationship does not automatically supply an individual prediction or demonstrate that raising a survey score will cause a specified reduction in injuries.
Protect participation and discuss the result
The developer’s practical guidelines frame the survey as a basis for dialogue and development, with voluntary participation, an intention to act and results made understandable to workers. Treating low scores as a contest or a reason to pressure respondents works against that purpose.
Explain the purpose, access arrangements and intended use before collecting answers. Avoid promising anonymity that the collection method cannot provide. NFA’s ethics guidance warns about identification through small occupational groups and cautions against reporting groups smaller than 20. A larger count does not by itself guarantee privacy when other details identify people.
Discuss the findings without trying to identify who gave an unfavourable answer. Use protected ways to explore the work conditions behind a pattern, and report back on decisions. The fictional table is a teaching illustration, not permission to release similarly structured employer data.
Connect the finding to action and separate verification
Suppose a group reports difficulty raising maintenance concerns. The next step is to examine how a report reaches someone with authority, resources and a response obligation. A higher survey score later may describe a change in perceptions; checking actual reports and equipment condition addresses a different question.
OSHA’s program-evaluation guidance includes safety opinion surveys among possible indicators while also calling for examination of reporting processes, hazard controls and worker participation. The questions below are my practical interpretation of that distinction, not a validated checklist or mandatory form.
| Record | Question | Evidence to seek |
|---|---|---|
| Finding and scope | What dimension and respondent group produced the concern? | Instrument version, dates, usable responses, distribution and assessment limits. |
| Work explanation | What conditions could explain the pattern? | Protected worker discussion, task context and relevant reporting or maintenance records. |
| Action | What change will address the identified condition? | Accountable owner, decision authority, resources and a defined intended result. |
| Verification | Does the revised process or control function? | Checks of the actual process or protective measure, alongside worker experience. |
| Follow-up | What changed, and what remains unresolved? | Comparable measures, participant context, decisions communicated and review triggers. |
The companion articles on audit-score interpretation and research applicability apply the same discipline to different evidence. Keep climate, audits, exposure measurements and incident records identifiable when considering them together. Combining them into a dashboard does not, by itself, validate a new overall rating.
The current Jenkins Industry Safety Journal collection is not peer reviewed. AI assistance supported development and editing of this article. The fictional example and discussion aids have not undergone independent validation, and no original workforce study or independent technical review is claimed.
References and source record
This is a focused interpretation guide. Instrument descriptions and scoring are based on the developer’s guidance; the discussion of Niu and Leicht is limited to the published abstract available through the author’s institutional record. The references do not represent an exhaustive literature review, and no participant dataset was collected or analysed for this article.
- Kines, P. A., Lappalainen, J., Mikkelsen, K. L., Olsen, E., Pousette, A., Tharaldsen, J., Tómasson, K., & Törner, M. (2011). Nordic Safety Climate Questionnaire (NOSACQ-50): A new tool for diagnosing occupational safety climate. International Journal of Industrial Ergonomics, 41(6), 634–646. Instrument identity and development checked against the developer’s institutional abstract.
- Niu, M., & Leicht, R. M. (2024). Evaluating the safety climate in construction projects: A longitudinal mixed-methods study. Buildings, 14(12), 4070. Abstract consulted through Penn State’s institutional publication record.
- National Research Centre for the Working Environment. (n.d.). Safety Climate Questionnaire—NOSACQ-50.
- National Research Centre for the Working Environment. (n.d.). Analysing NOSACQ-50 data.
- National Research Centre for the Working Environment. (n.d.). Interpreting the Nordic Occupational Safety Climate Questionnaire NOSACQ-50 results. Includes benchmark update dated August 3, 2026.
- National Research Centre for the Working Environment. (n.d.). NOSACQ-50 soft guidelines.
- National Research Centre for the Working Environment. (n.d.). Ethics in questionnaire studies—NOSACQ-50.
- Occupational Safety and Health Administration. (n.d.). Program evaluation and improvement. Recommended Practices for Safety and Health Programs. Web guidance checked October 26, 2026.