Originally posted on Caterpillar Safety Service’s Safety Culture WORLD blog and reposted here with Caterpillar’s permission.

Caterpillar’s Safety Perception Survey had its beginnings in the late 1970s when Dr. Dan Petersen and Dr. Chuck Bailey teamed to study what really made a difference in safety performance. Their research took place in two parts. The first focus was on what “conventional safety wisdom” said made a difference, i.e., “procedural – engineered,” OSHA regulations and the like. This five-year effort showed that “The 12 most commonly used criteria in standard safety program audits are poor measures of program effectiveness.”

Another piece of the first study was input from safety professionals on processes and questions that they thought would make a difference in observable safety performance. As a part of this discovery effort approximately 1,500 potential survey questions were distilled down to 73 questions that the study showed to have “a statistically significant response.”  The validation approach compared question responses from a poor safety record (high injury rate) organization to those of a good safety record organization.

All the questions that had no significant statistical difference in response between the very stratified (in terms of safety record) organizations were eliminated from the survey. These validated questions were then mapped to 20 safety processes that, when existent in a safety culture, should consistently deliver observably better safety practices. These observably relevant questions and their 20 “management systems” were then re-tested in a second phase of the nine-year study.

After reviewing the data from the first study a second, multi-year “verification study” focused on positive reinforcement of correct safety activities and “the quality of the 20 management systems which have an effect on human behavior relating to safety” was executed.  They found that “The most successful safety programs are those which recognize and deal effectively with employee and supervisor behavior and attitudes which affect safety.” 

Additionally, the research team began to engage in problem-solving efforts around the weak scoring processes that the survey showed to exist. In this phase of the study, question responses from another set of good and poor safety record organizations were once again compared to confirm the validity of high question scores to good safety performance (low injury rate). The end result was a strong statistically valid correlation between these 20 high-scoring processes, improvement efforts to low-scoring processes, training of supervisors in these improved safety management systems and consequently the percent of observable safe worker behaviors and subsequent lower injury rates.

And thus the conclusion that when present at higher percentages, these questions and the associated processes they map to, were statistically valid indicators of the “human factor” and associated worker activities and therefore a culture of better safety performance (lower injury rates). Or as their study conclusions stated: “Application of a 20-category survey technique developed by this study provides a reliable measure of safety program effectiveness.”