Development and evaluation of a safety culture survey for occupational safety

TR Number



Journal Title

Journal ISSN

Volume Title


Virginia Tech


The present study includes the development, large-scale administration to workers at four industrial plants, and evaluation of the Safety Culture Survey (SCS). The SCS consists of three scales: the Actively Caring Scale (ACS), the Safety Perception Scale (SPS), and the Risk Propensity Scale (RPS). The ACS measures person factors related to one's propensity to actively care for the safety of others. Actively caring (AC) refers to employees caring enough about the safety of their coworkers to act on their behalf. In other words, AC refers to continually looking for environmental hazards and unsafe work practices and implementing appropriate corrective actions when unsafe conditions or behaviors are observed. Included in the ACS is the RAC (reported AC) subscale. The RAC focuses on person, behavior, and environment issues. The RAC also categorizes various levels of AC (i.e., whether employees feel they should, are willing to, or often actively care). The SPS measures employees’ opinions and attitudes about their current safety climate. The scale addresses a variety of safety perceptions, including management concern for safety, peer support for safety, and personal responsibility for safety. The RPS measures person factors hypothesized to relate to an individual's propensity to engage in risky behaviors which increase the likelihood of a "near miss" or an injury. The RPS also includes the injury index subscale (i.e., reports of work-related injuries and illnesses).

A stepwise multiple regression found the ACS subscales to predict over 50 percent of the variance in RAC scores. Furthermore, the construct validity of the AC model was supported in a general way. A factor analysis revealed one AC factor and two correlated risk propensity factors. Also, the ACS subscales were more highly correlated with each other than with subscales from the RPS (i.e., variables hypothesized not to predict AC).

There were two interesting interactions found among SCS variables. The interaction between focus of AC (behavior, person, environment) and level of AC (should, willing, often) indicated employees were most willing to AC from a behavior-focus, yet least likely to report they often did AC from a behavior-focus. In addition, employees who perceived an unsupportive safety climate (i.e., those with low SPS scores) and who perceived a high level of risk on the job were less likely to AC compared with employees who perceived an unsupportive safety climate and who perceived a low level of risk on the job. This indicated the importance of efforts to increase the visible support of safety efforts as well as assessing safety perceptions before introducing interventions to increase the salience of work-related hazards.

Furthermore, a stepwise regression to predict injury rate (i.e., injury index scores) with RPS scores was disappointing, predicting only 5.4 percent of variance in injury index scores. However, when injury index scores were divided into high, medium, and low, significant differences were found among RPS subscale scores.

In conclusion, the SCS is presented as a reliable and valid research tool. It can also be used as an applied tool for industry to assess the levels of person factors related to AC behaviors, to assess the perception of management, peer, and personal responsibility for safety efforts, and to help evaluate the effects of interventions designed to bring about a safer workplace.