Human resource safety practices and employee injuries.

Hypothesis Testing. The proposed associations were tested by regressing employee injuries on the predictor and control variables (industry type, organizational size). Support for the hypothesis is found if the beta weight is negative and the confidence interval around the beta weight does not include zero (a 90% confidence interval was used due to the small sample size).

RESULTS

Findings

Correlations and Descriptives. Descriptive statistics, reliability estimates, and correlations for all measures are reported in Table 1. Contrary to expectations, the correlations between the independent and dependent variables all had confidence intervals including zero.

Hypotheses. The first proposed hypothesis was tested by regressing injuries on the selection predictor variable and control variables. The standardized beta-weight for injuries is -.02, and has a confidence interval that includes zero (CI = -.28, .23), thus not supporting a link between selection for safety and reduced injuries (see Table 2).

The connections between general selection practices (prior work experience, personality-testing, intelligence-testing, drug-testing, alcohol-testing, and pre-employment physicals) and injuries were examined using additional regression analysis (Table 3). Of these practices, both prior work experience and drug-testing resulted in non-zero regression weights ([beta] = -.31; CI = -.54, -.08; [beta] = -.46; CI = -.83, -.08, respectively). These findings suggest that hiring for prior work experience and conducting pre-employment drug-testing are associated with lower numbers of workplace injuries, indicating partial support for Hypothesis 1b.

The second proposed hypothesis was tested by regressing injuries on the training predictor variable and control variables (industry type and organizational size). The results showed the association between training and injuries was low and positive ([beta] = .12) and the confidence interval included zero (CI = -.14, .38). Therefore, support for the hypothesis that training practices are positively associated with organizational safety outcomes was not found (see Table 2).

The third proposed hypothesis was tested by regressing injuries on the performance evaluation variable and control variables. The standardized beta weight for injuries is -.17, and has a confidence interval that includes zero (CI = -.42, .09), thus not supporting a connection between performance evaluation and reduced injuries (see Table 2).

Hypothesis 4a was tested by regressing injuries on the individual compensation predictor variable and control variables (see Table 2). The connection between individual compensation with injuries ([beta] = -.27) was negative and the confidence intervals around this beta-weight did not include zero (CI = -.51, -.03), thus supporting the hypothesis that individual safety rewards are associated with reduced injuries. Hypothesis 4b was then tested by regressing injuries on the group compensation predictor variable and control variables (see Table 2). The link between group compensation and injuries ([beta] = -.39) was negative and did not include zero (CI = -.63, -.15), thus indicating support for the hypothesis that group compensation practices are associated with reduced injuries. Further, the beta-weight of group compensation ([beta] = -.39) was larger than for individual compensation ([beta] = -.27), indicating that group compensation may potentially have a stronger association with injury outcomes than individual compensation.

The connection between individual and group compensation and injuries was tested further by regressing injuries on both individual and group compensation and the control variables simultaneously (see Table 4). The beta-weight for individual compensation ([beta] = -.12) was still negative, but the confidence interval around the beta-weight included zero (CI = -.38, .14). Whereas the beta-weight for group compensation ([beta] = -.33) was negative and the confidence interval around the beta-weight did not include zero (CI = -.61, -.06). Thus, when including both individual and group compensation, individual compensation's link with injuries no longer existed, indicating group compensation has the strongest association with reduced injuries.

DISCUSSION

Some of the findings in this study confirmed past findings and others provided interesting additions to the literature. Selecting for safety per se had not been studied much in the past, though some individual differences leading to injuries had been identified (Tiffin and McCormick, 1962; Arthur et at, 1991; Powell et at, 1971). Although failing to produce support for general selection of safety, this study found both prior work experience and drug-testing were associated with safety outcomes. The results for drug-testing added to the mixed findings, contrasting Normand et al.'s (1990), but supporting Spicer et al.'s (2003) findings. The work experience association supports Hansen's (1989) findings.

In contrast to several recent findings (Reber et al., 1993; Lehto and Salvendy, 1995; Harshbarger and Rose, 1991), this study did not find a connection between training and organizational safety outcomes. This lack of findings may be due to the training measure as it includes both supervisory and employee training. Although Harshbarger and Rose (1991) found that training was important at both levels, Harvey et al. (2001) found training was more important at the supervisory level. Training is also an aspect of safety more regulated by OSHA, perhaps decreasing the difference in the training practices across organizations. Performance evaluation also was not associated with safety in this study. This finding may be due to how performance evaluation was operationalized. Both formal performance evaluations and more general feedback questions were included, whereas most often studies examine the more general feedback and derive this measure from the employees (McAfee and Winn, 1989).

The examination of compensation's association with safety outcomes produced one of the key findings in this study. Consistent with past research, this study found that compensating for safety is associated with a lower number of injuries (Harshberger and Rose, 1991; McAfee and Winn, 1989; Zohar and Fussfeld, 1981). Although past research has found that a variety of compensation methods have positive effects (McAfee and Winn, 1989), individual and group safety compensation had not been compared previously.

Implications

Theoretical. This study's finding that group compensation has a stronger association with organizational safety outcomes than individual compensation suggests that researchers should perhaps revisit what creates a safe environment. Is a safer environment produced when employees work more as a group versus individually? If so, then perhaps group compensation needs to be an organization's focus. This contradicts past research that has been more supportive of individual compensation (Bartol and Locke, 2000). However, the importance of team interaction as noted by Hoffmann and Stetzer (1996) in reinforcing safety may make group pressure or shared norms more salient in safety, requiring reinforcement at the group level.

The findings that selecting specifically for safety does not have an impact on safety, but that more general selection practices (drug-testing and prior work experience) do have an impact, also have theoretical implications. Do selection practices need a direct link to outcomes sought, or may an indirect link be as effective? If so, organizations may be able to select for multiple organizational performance outcomes using a single, general selection practice.

Managerial. Some HR safety practices were found to have an association with employee injuries, indicating improvements in safety go beyond ergonomics and engineering aspects. Oftentimes safety outcomes are potential indicators of managerial practices (Ruth, 2004) rather than indicators of an ergonomically unsafe environment. Management needs to be able to indicate to employees that safety is more important than performance (Mullen, 2004). This can be done through a multitude of practices, a few of which have been indicated in this study, although these are far from comprehensive. Hiring for prior work experience and drug-testing are indicated to be good safety practices. This provides the prospect that organizations should consider the potential of altering their selection practices to take safety into account. However, using drug-testing as a selection practice has been noted in past research to have its limitations (Macdonald, 1997, 1995; Normand et al., 1990). Thus, organizations need to take into account a cost and benefit analysis, and examine the population that they will be screening prior to implementing this practice.

Further, compensation for safety (particularly at the group level) is found to be important. Organizations should seek rewards that help encourage employee support of safety practices. The importance of these rewards is discussed by Mullen (2004) who found that often employees would receive pressure to simply "get the job done." Organizations also need to maintain caution in not deterring employees from reporting injuries (Collinson, 1999). Overall, these findings encourage organizations to expand use of HR practices to help improve overall safety.

Limitations