Does talking the talk matter? Effects of supervisor safety communication and safety climate on long-haul truckers’ safety performance

Accident Analysis and Prevention 117 (2018) 357–367

Contents lists available at ScienceDirect

Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap

Does talking the talk matter? Effects of supervisor safety communication and safety climate on long-haul truckers’ safety performance

T

⁎

Yueng-hsiang Huanga, , Robert R. Sinclairb, Jin Leec, Anna C. McFaddenb, Janelle H. Cheungd, Lauren A. Murphye a

Risk Control Services, Liberty Mutual Group, Hopkinton, MA, USA Clemson University, Clemson, SC, USA c Kansas State University, Manhattan, KS, USA d Oregon Health and Science University, Portland, OR, USA e Northeastern University, Boston, MA, USA b

A R T I C L E I N F O

A B S T R A C T

Keywords: Safety climate Supervisor safety communication Safety performance Long-haul truck drivers Workplace injury

This study examines the distinct contribution of supervisory safety communication and its interaction with safety climate in the prediction of safety performance and objective safety outcomes. Supervisory safety communication is defined as subordinates’ perceptions of the extent to which their supervisor provides them with relevant safety information about their job (i.e., top-down communication) and the extent to which they feel comfortable discussing safety issues with their supervisor (i.e., bottom-up communication). Survey data were collected from 5162 truck drivers from a U.S. trucking company with a 62.1% response rate. Individual employees’ survey responses were matched to their safety outcomes (i.e., lost-time injuries) six months after the survey data collection. Results showed that the quality of supervisor communication about safety uniquely contributes to safety outcomes, above and beyond measures of both group-level and organization-level safety climate. The construct validity of a newly-adapted safety communication scale was demonstrated, particularly focusing on its distinctiveness from safety climate and testing a model showing that communication had both main and moderating effects on safety behavior that ultimately predicted truck drivers’ injury rates. Our findings support the need for continued attention to supervisory safety communication as an important factor by itself, as well as a contingency factor influencing how safety climate relates to safety outcomes.

1. Introduction Safety climate is generally defined as employees’ shared perceptions of their organization’s policies, procedures, and practices in regards to the value and importance placed on safety (Zohar, 1980, 2000). According to Zohar (2008, 2010), safety climate should be measured using a framework that distinguishes between organization-level (employees’ perceptions of top management commitment to and prioritization of safety) and group-level (employees’ perceptions of direct supervisor or workgroup commitment to safety) safety climate perceptions (Huang et al., 2013). The two safety climate components reflect distinct referent points, top management and direct supervisors, that serve as important cues for employees’ safety-related perceptions. Supervisory communication practices are another workplace factor that may uniquely contribute to safety, above and beyond safety climate (Sinclair et al., 2014). Supervisors are the main channel through

⁎

which safety policies and procedures are communicated to subordinates in a “top-down” fashion. Communication also has “bottom-up” effects related to whether subordinates are willing and able to share safetyrelated concerns with their supervisors. Our study examined the distinct and interactive effects of safety communication and safety climate on safety performance and objective safety outcomes. Since supervisors are often workers’ main source of information about safety concerns (especially in the lone worker context as described below), we focused on group-level safety climate perceptions based on employees’ perceptions of their direct supervisors. We tested our hypotheses after controlling for organization-level safety climate, which strengthens the inference that any observed effects are attributable to supervisory communication. We see our study as making two main contributions to the literature. Our first contribution to the literature is to extend prior literature concerning the distinct and interactive roles of safety climate and safety

Corresponding author at: Risk Control Services, Liberty Mutual Group 71 Frankland Road, Hopkinton, MA 01748 USA. E-mail address: [email protected] (Y.-h. Huang).

https://doi.org/10.1016/j.aap.2017.09.006 Received 1 November 2016; Received in revised form 4 April 2017; Accepted 5 September 2017 Available online 28 February 2018 0001-4575/ © 2017 Published by Elsevier Ltd.

Accident Analysis and Prevention 117 (2018) 357–367

Y.-h. Huang et al.

for workers to perform in a safer manner (Zohar et al., 2015). In line with this reasoning, as well as other studies of safety-specific supervisor leadership (e.g., Conchie et al., 2012; Kelloway et al., 2006; Mullen and Kelloway, 2009) we expect group-level safety climate (i.e., employees’ perceptions about supervisors’ commitment to safety) to be related to both safety performance and lost time injury.

communication in safety performance outcomes. Extensive research, summarized in several meta-analytic reviews, demonstrates the importance of safety climate in predicting safety outcomes (Beus et al., 2010; Christian et al., 2009; Clarke, 2006a, 2010; Griffin and Neal, 2000; Nahrgang et al., 2011). A smaller body of literature shows similar benefits of safety communication in empirical tests of communication measures (Griffin and Neal, 2000; Hofmann and Stetzer, 1998; Kath et al., 2010; Parker et al., 2001; Zohar and Luria, 2003) and evaluations of communication-focused interventions (e.g., Kines et al., 2010). Relatively few studies, however, have examined the distinct effects of safety communication and safety climate on outcomes. Additionally, despite their close conceptual relationship, no research that we are aware of has tested interactions between safety communication and safety climate. Such research is important in understanding both the nomological network of safety climate and the contingencies that affect the relationship between safety climate and outcomes. Given these concerns, we examine: (1) the empirical distinctiveness of safety communication and safety climate with respect to construct validity (i.e., distinct factor structure), (2) incremental predictive validity in a model in which safety behavior mediates the relationship of communication and climate with outcomes, and (3) whether safety communication and safety climate interact in their prediction of safety outcomes. Our second contribution concerns the nature of our sample. We focus on long-haul truck drivers who have received relatively little attention in safety literature despite the importance of safety for these workers. Transportation-related incidents are the number one cause of workplace fatalities in the United States, and truckers have a disproportionate share of those incidents (Bureau of Labor Statistics, 2014). From a theoretical perspective, long-haul truckers are an example of lone workers for whom safety climate-related processes may operate differently than for workers aggregated into larger units (e.g., Huang et al., 2013; Olson et al., 2009) and for whom communication with their supervisor may be especially important, as the supervisor often is their only link to the broader organization. Our study extends prior research on truckers’ driving safety by investigating the effects of both safety communication and safety climate. We link these antecedents to two outcomes relevant to truckers: (1) self-reports of safe driving performance and (2) an objective measure of days lost to injuries.

Hypothesis 1. Group-level safety climate is positively related to safety performance. Hypothesis 2. Group-level safety climate is negatively related to lost time injury.

1.2. Supervisory safety communication Our approach to studying communication focuses on the quality of safety communication between supervisors and subordinates. Supervisors who communicate effectively about safety may have employees who have a better understanding of safe behavior and the possible outcomes of unsafe behavior (Michael et al., 2006). Moreover, subordinates who perceive themselves as able to talk with their supervisor about safety issues may be more likely to report unsafe conditions prior to accidents. Thus, we define supervisory safety communication as subordinates’ perceptions of the extent to which their supervisor provides them with relevant safety information about their job (i.e., top-down communication) and the extent to which they feel comfortable discussing safety issues with their supervisor (i.e., bottomup communication). Several studies demonstrate the importance of safety communication. For example, Zohar and Luria (2003) found that informing supervisors of the number of safety-related exchanges they had with subordinates increased their number of safety-related communications and decreased unsafe behavior. Additionally, Hofmann and Stetzer (1998) found that safety communication moderated the relationship between informational cues of work-related accidents and causal attributions. The authors noted that safety communication from supervisors encouraged upward communication regarding safety (i.e., subordinates voicing safety concerns to management), which may affect how employees view the causes of safety events at work. Taken together, this literature supports both the importance of safety communication for safety-related outcomes and the role of the organization in fostering effective safety communication. Few studies, however, have examined the joint effects of safety communication and safety climate in predicting safety outcomes. Researchers often examine safety communication as a facet of safety climate (e.g., Griffin and Neal, 2000) or as an outcome of safety communication interventions (e.g., Kines et al., 2010; Zohar and Polachek, 2014). What is missing from this literature are studies that examine the distinct effects of safety climate and safety communication in relation to safety outcomes. The distinction between safety communication and safety climate is an important theoretical and empirical issue in this research stream. Some older models of safety climate treated safety communication as a component of safety climate. For example, Griffin and Neal (2000) treated safety climate as a higher order composite of safety training, management values, safety inspections, and safety communication. Other research, however, treats safety communication either as an antecedent or as a consequence of safety climate. For example, Hofmann and Stetzer (1998) argued that safety climate influences organizational communication practices about safety issues, such that negative safety climates would lead to less open communication about safety issues. In contrast, other researchers have conceptualized communication as an important influence on safety climate (e.g., Kines et al., 2010; Zohar, 2010; Zohar and Polachek, 2014). Zohar (2010) argued for a symbolic social interactionist perspective on the development of climate in that climate arises through employees’ sense-making processes about safety issues in the organization, which are shaped by

1.1. Safety climate with lone workers The increased use of technology and the changing nature of work have led to more workers working alone in isolated locations. Longhaul truck drivers are an excellent representation of lone workers, as they are often on the road and are only required to report to their dispatchers and/or supervisors a few times a day (mostly over the phone or using an electronic device). In fact, they may not have face-toface conversations with their supervisors for weeks at a time. While safety climate is typically referred to as shared perceptions among employees (Neal and Griffin, 2004; Zohar and Luria, 2005), lone workers, including truck drivers, usually do not interact with their supervisors and coworkers. In fact, Huang et al. (2013) found that truck drivers’ safety climate perceptions were not shared within their work groups. Insufficient statistical evidence for aggregation and the nature of lone working strongly suggest that truck drivers’ safety climate perceptions are best understood at the individual level, sometimes referred to as psychological safety climate (Christian et al., 2009; Huang et al., 2013). Safety climate perceptions relate to employees’ expected consequences of safe or unsafe performance (Zohar, 2010). According to Zohar et al. (2015), safety climate perceptions that are more positive can lead to better safety performance, due to the instrumentality of and valence for positive performance outcomes. If safe performance is perceived to result in supervisory recognition or support, a more positive safety climate will emerge and, thus, promote stronger motivation 358

Accident Analysis and Prevention 117 (2018) 357–367

Y.-h. Huang et al.

accidents and injuries. This implies that safety performance should mediate the relationship of safety climate and safety communication with injury outcomes. The mediating effects of safety performance have been supported in many studies (Christian et al., 2009; Zohar et al., 2014). To our knowledge, the role of communication in this general model has not been tested. We assume that the same relationship would hold, such that safety performance mediates the relationship between safety climate and lost time injury, as well as between safety communications and lost time injury.

supervisory safety communication. Because there are plausible arguments for safety communication both as an antecedent and as a consequence of safety climate, we conceptualize communication as a correlate of climate. We assume that safety communication is conceptually distinct from safety climate and that communication encompasses organizational policies and practices that may either influence or be influenced by safety climate. This leads us to expect positive safety climate perceptions to be associated with better supervisory safety communication, but we do not make a specific causal claim concerning the relationship.

Hypothesis 8. Safety performance mediates the relationship between group-level safety climate and lost time injury.

Hypothesis 3. Group-level safety climate is positively related to safety communication.

Hypothesis 9. Safety performance mediates the relationship between supervisory safety communication and lost time injury.

Good safety communication practices should facilitate safer behavior and fewer resulting injuries at work. We contend that the effects of communication on the safety performance relationship are realized through increased knowledge and information. Cigularov et al. (2010) referred to this effect as the “education function” of communication. In top-down communication, supervisors pass along information about company safety policies and procedures, leading subordinates to experience increased task clarity and enabling them to adjust their behavior accordingly (Clarke, 2006b). Additionally, when supervisors regularly communicate with their subordinates, it may increase subordinates’ comfort in expressing safety concerns (bottom-up communication) to their supervisors (Hofmann and Morgeson, 1999), enabling the supervisors to proactively address safety concerns.

2. Method 2.1. Participants The data used for this study came from a single trucking company located in the United States. This company was non-unionized and used contracts with a single customer for an entire trailer load, rather than multiple customers for mixed freight in each trailer. The data were gathered as part of a larger research program focused on developing and validating safety climate measures for truck drivers. Three prior studies by the research team have supported the extrinsic motivation argument for the effects of safety climate (Zohar et al., 2015), offered preliminary validity evidence for the climate scale (Huang et al., 2013), and examined agreement between supervisors and subordinates about safety climate perceptions and behavior (Huang et al., 2014). All of these prior studies used participants from a first wave of data collection. The present study used data from a second wave of data collection (collected approximately two years after the first wave) and was constructed so that none of the data nor any of the participants from this study overlapped with the previously published data; all data in this study are unique to this study. The second-wave survey was sent to 8308 truck drivers, 5162 of whom returned surveys, resulting in a response rate of 62.1%. We removed 77 participants who completed less than 50% of the survey. Most participants (90.5%) had not participated in the first data collection (annual turnover was approximately 75% per year during the years of our data collection). In order to keep this study fully distinct from the prior research, we removed an additional 485 participants (9.5%) who had participated in the first survey, leaving us with a sample of 4600 participants for the analyses described below. All participants were non-unionized long-haul truck drivers. Their average age was 47.7 years (SD = 10.7), their average tenure as a professional driver was 12.4 years (SD = 33.4), and their average employment at the current company was 6.1 years (SD = 6.0). Driver gender was not gathered as part of this study. In our past experience, the vast majority (typically more than 90%) of long-haul truck drivers are men. In the present company, 7% of the drivers are women; the company did not want to include gender as a variable partly to minimize perceived concerns that female participants might have about being identified.

Hypothesis 4. Supervisory safety communication is positively related to safety performance. Hypothesis 5. Supervisor safety communication is negatively related to lost time injury. 1.3. Moderating effects of safety communication Supervisory safety communication also may be a contingency factor affecting safety climate outcomes; the protective effects of safety climate may be increased or diminished depending on the nature of safety communication. The education function of communication described above suggests that when supervisor safety communication quality improves, the effects of safety climate should be enhanced. When safety communication is poor, the effects of safety climate may be diminished as subordinate safety knowledge may be lower and safety concerns may not be addressed. In other words, organizations that prioritize safety still may not realize the benefits of a strong positive safety climate unless they also have good communication practices in place. This reasoning leads us to hypothesize that supervisory safety communication moderates the relationship of group-level safety climate with safety performance and injuries. Hypothesis 6. Supervisory safety communication will moderate the relationship of group-level safety climate with safety performance such that better safety communication will strengthen the positive relationship between safety climate and safety performance. Hypothesis 7. Supervisory safety communication will moderate the relationship of group-level safety climate with lost time injury such that the negative relationship between safety climate and lost time injury is stronger when safety communication is better.

2.2. Data collection procedure The company requires all truck drivers to take a 20-min web-based safety training program every spring. The company allowed the research team to use this opportunity to recruit drivers to complete a webbased survey. At the end of the training, all participants received an invitation to complete the survey and an informed consent form. When drivers agreed to participate and clicked the link on the web to accept the survey, they were linked to a web site hosting the survey. The

1.4. Mediating effects of safety performance Safety scholars have drawn from the job performance literature to distinguish proximal and distal outcomes of safety climate (Griffin and Neal, 2000; Neal and Griffin, 2004; Neal and Griffin, 2006). This literature assumes that safety behavior (i.e., performance) is a proximal outcome of climate which influences more distal outcomes such as 359

Accident Analysis and Prevention 117 (2018) 357–367

Y.-h. Huang et al.

supervisor) through in-vehicle radio devices and cell phones. Thus, even though multiple drivers may share the same dispatcher (supervisor), they have little to no opportunity to interact with their coworkers and often do not know any other drivers with the same supervisor. This suggests that shared perceptions are unlikely among drivers. This is consistent with the findings in Huang et al.’s (2013) study where the Intra class correlation (ICC)1 was smaller than 0.10 and the ICC2 was smaller than 0.70 for organizational- and group-level safety climate, neither of which supported aggregating the data (Bliese, 2000; Lee et al., 2014). Note that ICC1 indicates the amount of variance in individual truck driver-level responses that can be explained by dispatcher-level properties while ICC2 indicates the reliability of the mean safety climate scores of the truck drivers within a group (by dispatcher). Additionally, over 90% of the supervisors in the current study only had one subordinate participating in the project. Taken together, both the design of the job, and past available empirical evidence, provided support for examining truck driver climate perceptions as individual psychological safety climate.

survey took about 15 min to complete. Participants were asked to enter their individual company ID as an identifier to match their survey responses to company injury records. The company did not, however, have access to the survey data, and the research team did not have access to participants’ names or other identifying information, ensuring that the participants’ data would remain confidential. At the completion of the study, the company received survey feedback at the company level only. Five $100 gift cards were provided via lottery as incentives to encourage participation. Participants were asked to provide contact information (e.g., email or phone number) at the end of the survey if they wanted to be entered in the lottery. Six months after survey implementation, the company provided lost time injury data from all drivers to the research team. Employees’ company IDs were used to match employees’ survey responses with lost time injury data. As compared with cross-sectional designs, a prospective design such as this allows more stringent testing of the predictive validity of the study’s variables. 2.3. Measures

2.3.2. Safety communication The safety communication scale was adapted from Cigularov et al. (2010), Cox and Cheyne (2000), and Hofmann and Morgeson (1999). The scale consists of eight items reflecting top-down communication (e.g., “Safety information is always brought to my attention by my immediate supervisor”) and bottom-up communication (e.g., “I feel comfortable discussing safety issues with my immediate supervisor”). Construct validity of the scale was examined using confirmatory factor analyses (CFAs). Following Hu and Bentler (1999), we used comparative fit index (CFI) greater than 0.95 and root mean squared error of approximation (RMSEA) smaller than 0.08 as criteria for goodness of fit of the measurement model. We examined the second-order hierarchical model for safety communication. This model had two dimensions of safety communication (i.e., top-down and bottom-up) and an overarching higher-order factor indicating overall safety communication. Also, covariances were allowed among the three negatively worded items (DiStefano and Motl, 2006; Merritt, 2012). The models demonstrated satisfactory model fit (CFI = 0.99; RMESA = 0.05). Means, standard deviations, and factor loadings of the eight scale items are presented in Table 1. We also examined the discriminant validity of the quality of the safety communication factor from the three sub-factors of the grouplevel safety climate scale (i.e., safety promotion, delivery limit, and cell phone disapproval). As shown in the bottom half of Table 2, we conducted a series of confirmatory factor analyses testing a 4-factor model with a separate factor for communication and each of the three safety climate factors, in comparison to three 3-factor models, each of which specified the communication items loading onto one of the safety climate dimensions. The goal of these analyses was to demonstrate improved model fit by the inclusion of the distinct safety communication

All items were rated on a 5-point Likert scale (1: strongly disagree, 5: strongly agree). 2.3.1. Safety climate We used a trucking industry-specific safety climate scale to assess safety climate perceptions. This context-specific measure is tailored toward the trucking industry with items specifically intended to capture the relative priority of safety. That is, rather than simply asking about perceptions of safety, the items follow Zohar’s (2010) recommendation to assess drivers’ perceptions about how the organization/supervisor manages the trade-off between performance and safety concerns. The scale consists of two 20-item subscales measuring organization- and group-level safety climate (total of 40 items). Example items include: “My company uses any available information to improve existing safety rules” (organization-level safety climate subscale) and “My supervisor compliments employees who pay special attention to safety” (grouplevel safety climate subscale). Prior research has demonstrated the strong psychometric properties of the scale and offered both construct validity and criterion-related validity evidence to support its use in trucking samples (Huang et al., 2013). In the current study, internal consistency estimates (Cronbach’s α) for the organization- and group-level safety climate subscales were satisfactory (0.92 & 0.94, respectively). We note that our hypotheses focused only on the group-level safety climate measure, which refers to perceptions about the supervisor’s (i.e., dispatcher’s) priority for safety. We included the organization-level measure as a control to ensure that any obtained effects related to the group measure could not be attributed to a failure to account for organization-level safety climate. Each driver primarily communicated with his/her dispatcher (i.e., Table 1 Items of the supervisor safety communication scale. #

Items

Mean (S.D.)

Standardized Factor Loadings Factor 1

1 2(R) 3 4(R) 5 6 7(R) 8

I feel comfortable discussing safety issues with my supervisor. I try to avoid talking about safety issues with my supervisor. I feel that my supervisor openly accepts ideas for improving safety. I am reluctant to discuss safety-related problems with my supervisor. I feel that my supervisor encourages open communication about safety. Safety information is always brought to my attention by my supervisor. My supervisor does not always inform me of current concerns and issues. There is good communication here about safety issues which affect me.

4.32 4.29 4.01 4.08 4.27 4.13 3.97 4.18

(1.00) (1.02) (1.13) (1.17) (.98) (1.02) (1.17) (1.03)

**

0.78 0.64** 0.74** 0.51** 0.91** – – –

(0.02) (0.02) (0.02) (0.02) (0.02)

Factor 2 – – – – – 0.82** (0.02) 0.64** (0.02) 0.83** (0.02)

Notes. (R) indicates reverse-worded item; S.D. = standard deviation. The factor loadings are based on the two-factor alternative model. Factor 1: bottom-up communication (Cronbach’s α = 0.85). Factor 2: top-down communication (Cronbach’s α = 0.80). ** p < 0.01

360

Accident Analysis and Prevention 117 (2018) 357–367

Y.-h. Huang et al.

variable, relative to simple accident/injury counts, is that it can convey information about injury severity, as more serious injuries generally are associated with more lost days of work.

Table 2 Construct and discriminant validity of the quality of safety communication measure. χ2 (df, significance)

CFI

Safety communication measurement model 2nd order hierarchical 217.40 (16, 0.99 model p < 0.01) Discriminant validity of safety communication and safety 4-factor model 6,475.70 (341, 0.93 p < 0.01) 3-factor model A 13,422.13 (344, 0.85 p < 0.01) 3-factor model B 10,011.85 (344, 0.89 p < 0.01) 3-factor model C 8,494.17 (344, 0.90 p < 0.01)

RMSEA (90% confidence interval)

2.3.5. Controls We included four control variables in the current study. First, we controlled for organization-level safety climate (Mean = 4.08, SD = 0.70). Second, we controlled for the frequency of safety communication using a single item asking participants how often they communicated with their supervisor about safety concerns. We judged it important to control for the frequency of communication so as to assure that the communication effects are in fact attributable to the quality of communication rather than the frequency, given that past research has shown that the frequency of supervisory communication also is an important consideration (cf. Kacmar et al., 2003). Participants rated this item on a 4-point scale with response anchors of 1 = rarely, 2 = once in a while, 3 = often, and 4 = all the time (Mean = 2.55, SD = 0.90). Third, based on Kockelman and Kweon (2002), we controlled for miles driven per year with an objective measure obtained from company human resources records (Mean = 22,660.34 miles, SD = 14,273.44). Finally, following Frone (1998), we controlled for driving experience with a self-reported measure of professional tenure (Mean = 12.38 years, SD = 33.44). Becker (2005) noted that including unnecessary control variables can lower statistical power and recommended comparing results with and without inclusion of control variables. Given our substantial sample size, we had no concerns about statistical power. Additionally, analyses with and without the control variables were largely the same, so for the sake of brevity we report only the analyses with the control variables included.

0.052 (0.046−0.059) climate 0.063 (0.061–0.064) 0.091 (0.090–0.092) 0.078 (0.077–0.079) 0.072 (0.070–0.073)

Notes: CFI = Comparative Fit Index; RMSEA = Root Mean Square Error of Approximation. 2nd order hierarchical model was based on two dimensions of quality of safety communication and Co-variances among the three negatively worded items were specified. 4-factor model: This model included four correlated factors: one dimension quality of safety communication and three subfactors of group-level safety climate. 3-factor model A: This model included three correlated factors: Items for quality of safety communication were merged with the safety promotion subfactor of the group climate measure + two additional group climate subfactors. 3-factor model B: This model included three correlated factors: Items for quality of safety communication were merged with the delivery limits subfactor of the group climate measure + two additional group climate subfactors. 3-factor model C: This model included three correlated factors: Items for quality of safety communication were merged with the cell phone disapproval subfactor of the group climate measure + two additional group climate subfactors.

factor. The model with a distinct communication factor outperformed any of the 3-factor models. Specifically, merging safety communication items with one of the safety climate factors resulted in non-overlapping RMSEA confidence intervals, compared to the models with distinct safety communication factor(s), indicating significant model fit change. The 4-factor model with a separate communication factor was the best overall fitting model (CFI = 0.93, RSMEA = 0.06). Taken as a whole, these findings supported the construct validity of the safety communication scale. We also note that the correlation between the two communication dimensions was 0.78. Since the current study does not focus on the differential effects of the two components of safety communication but rather on the overall quality of safety communication, we treated this as a single scale for the purpose of testing our hypothesized model. Internal consistency (Cronbach’s α) of the final communication scale was 0.90. Thus, the safety communication scale has appropriate reliability and validity.

2.4. Analysis Regression analyses were conducted to test Hypothesis 1 through 6 utilizing R 3.0.1 (R Core Team, 2013). Variance inflation factor (VIF) values were smaller than 5 (range = 1.00–4.74) for all independent variables (i.e., organization- and group-level safety climate perceptions, safety communication, group-level safety climate and safety communication interaction term, frequency of safety communication, miles driven per year) in the regression models predicting either safety performance or lost time injury, indicating no serious concern of multicollinearity (Rogerson, 2001). To test the hypothesized mediation effects (Hypothesis 7 and Hypothesis 8), we conducted path analyses using Mplus version 6.0 software (Muthén and Muthén, 2006). We used zero-inflated Poisson regression (Miaou, 1994) in the regression model for testing Hypothesis 2, Hypothesis 5 and Hypothesis 7 because the dependent variable in this model (i.e., lost time injury) was a count variable that was not normally distributed (overdispersion with excessive zeros). Because we did not have a priori hypotheses about the sub-dimensions of the climate and communication constructs, we tested the path model using a single latent variable representing safety climate and a second latent variable representing safety communication. We used Monte Carlo simulations to estimate the statistical significance of indirect effects. Specifically, confidence intervals of the estimated indirect effects were calculated by the parametric residual bootstrapping method based on 20,000 replications (Pituch et al., 2006) in R 3.0.1. If the range of the replications of the indirect effect estimate does not include zero, it can be concluded that the indirect effect is significantly different from zero at the p < 0.05 level. The final path model was selected considering model parsimony, statistical significance of path coefficients, and goodness of fit indicated by Akaike information criteria (AIC) and Bayes information criteria (BIC). Smaller AIC and BIC suggest better fit (e.g., Schreiber et al., 2006). Traditional model fit indexes such as comparative fit index (CFI) and root mean square error of approximation (RMSEA) were not available because lost time injury was a count variable and χ2 statistics could not be calculated.

2.3.3. Safety performance/safety behavior Six items adapted from Huang et al. (2005) were used to assess truck drivers’ self-reported driving safety performance. A sample item is “I always comply with the posted speed limits.” The Cronbach’s α was 0.66, which is slightly below the 0.70 criteria of acceptable internal consistency (Nunnally and Bernstein, 1994). However, in a previous study (Zohar et al., 2014), the same scale demonstrated a Cronbach’s α of 0.75 (N = 3578) and was significantly correlated with a prospective measure of near-miss frequencies. Also, the scale items regarding both safety compliance and participation behaviors are of good content validity (Huang et al., 2005). Moreover, the Cronbach’s α of 0.66 was greater than the unacceptably low internal consistency criterion (< 0.50; George and Mallery, 2003; Kline, 2000). 2.3.4. Lost time injury (LTI) Lost time injury, the objective safety criterion used in this study, was operationalized as the number of lost work days (over the past six months) due to injury and was measured six months after the survey implementation. One notable advantage of using the lost time injury 361

Accident Analysis and Prevention 117 (2018) 357–367

Y.-h. Huang et al.

Table 3 Means, standard deviations, correlations, and Cronbach’s αs of the study variables.

1. 2. 3. 4. 5. 6. 7. 8.

Organization-level safety climate Group-level safety climate Quality of safety communication Safety performance Lost time injury Miles driven per year Driving experience Frequency of safety communication

Mean (S.D.)

1

2

3

4

5

6

7

8

4.08 4.22 4.16 4.44 0.53 22660.34 11.71 2.55

(0.92) 0.77** 0.70** 0.51** −0.01 −0.01 −0.02 0.34**

(0.94) 0.83** 0.54** −0.003 0.01 −0.01 0.45**

(0.90) 0.54** −0.002 0.01 −0.03 0.45**

(0.75) 0.01 0.00 0.02 0.18**

– −0.02 0.02 0.01

– 0.04** −0.08**

– −0.05**

–

(0.70) (0.73) (0.82) (0.64) (7.50) (14273.44) (9.30) (.90)

Notes: The lost time injury variable is a positively skewed count variable (unit = day). Values in parentheses are Cronbach’s α; S.D. = standard deviation ** p < 0.01

3. Results

3.2. Path analysis

Table 3 presents means, standard deviations, and internal consistency estimates (Cronbach’s α) of the study variables and their intercorrelations. Correlations between the focal study variables supported Hypothesis 1 regarding the relationship between group-level safety climate and safety performance (r = 0.54, p < 0.01); Hypothesis 3, concerning the relationship between safety communication and grouplevel safety climate (r = 0.83, p < 0.01); and Hypothesis 4 regarding the relationship between safety communication and safety performance (r = 0.54, p < 0.01). Hypothesis 2 and Hypothesis 5, which regarded the relationship of lost time injuries with safety climate (Hypothesis 2) and safety communication (Hypothesis 5), were not supported by the correlations, as neither correlation was significant. This could be due to the small variance of the lost time injury variable, as most respondents (98%) reported zero lost time injuries. Therefore, when lost time injury was used as an outcome variable in regression or path analysis, we used zero-inflated Poisson regression (Miaou, 1994).

We examined a series of path models to confirm the hypothesized moderation and mediation effects. We first tested the baseline partial mediation model (Model 1 in Table 4) in which all direct and indirect effects of the independent variables (e.g., group-level safety climate, quality of safety communication, and their product term), as well as control variables, were included. In order to test the mediation effect of safety performance in Model 1, we compared this model to an alternative model (Model 1′ in Table 4) in which the link between safety performance and lost time injury was not specified (i.e., by fixing the path coefficient to zero). In other words, no mediation was assumed in Model 1′ because all independent variables were specified to have only direct effects on safety performance and lost time injury. As shown in Table 4, Model 1 was preferred over Model 1′ based on smaller AIC and BIC statistics. Also, consistent with our hypothesized mediation model, the link between safety performance and lost time injury in Model 1 was significant with a coefficient estimate of −0.34 (S.E. = 0.16, p < 05). In Model 1, the moderation effect of the quality of safety communication on the relationship between group-level safety climate and safety performance was supported (coefficient = 0.05, S.E. = 0.02, p < 0.05), but the moderation effect of safety communication on the relationship between group-level safety climate and lost time injury was not supported (coefficient = 0.59, S.E. = 0.39, p = 0.14). Also, the direct effect of the quality of safety communication on lost time injury was non-significant (coefficient = 0.05, S.E. = 0.20, p = 0.80). Therefore, Model 2 was created by omitting the two non-significant paths, [group-level safety climate × quality of safety communication → lost time injury] and [quality of safety communication → lost time injury], for model parsimony. In Model 2 (Table 4), the direct effect of group-level safety climate on lost time injury was not significant (coefficient = −0.46, S.E. = 0.29, p = 0.11), and this path was excluded in the more parsimonious Model 3 (Table 4). In Model 3, paths between all of the focal study variables (except for some control variables) were statistically significant at p < 0.01. Although the AIC and BIC values of Model 3 were higher than those of Models 1 and 2 in general, suggesting poorer fit, Model 3 was chosen as the final model due to its greater parsimony. This full mediation path model is in line with the path model of Zohar et al. (2014) in which safety performance fully mediated the relationship between trucking safety climate and frequency of near misses. Table 5 presents the direct and indirect effects of Model 3 and Fig. 3 is its graphical illustration. Group-level safety climate (coefficient = 0.22, S.E. = 0.03, p < 0.01) and quality of safety communication (coefficient = 0.21, S.E. = 0.03, p < 0.01) had significant direct effects on safety performance. The relationship between grouplevel safety climate and safety performance was moderated by the quality of safety communication (coefficient = 0.05, S.E. = 0.02, p < 0.05) such that the positive impact of group-level safety climate on safety performance tended to be greater when the quality of safety

3.1. Regression analysis Fig. 1-A illustrates the regression model for testing Hypothesis 1, Hypothesis 4 and Hypothesis 6, and Fig. 2-A illustrates the regression model for testing Hypothesis 2, Hypothesis 5 and Hypothesis 7. After controlling for the effects of organization-level safety climate, driving experience, miles driven per year, and frequency of safety communication, main effects of group-level safety climate on safety performance (B = 0.22, S.E. = 0.03, p < 0.01) and lost time injury (B = −1.32, S.E. = 0.07, p < 0.01) were statistically significant, respectively supporting Hypothesis 1 and Hypothesis 2. Similarly, the main effects of the quality of safety communication on safety performance (B = 0.21, S.E. = 0.02, p < 0.01) and lost time injury (B = −0.32, S.E. = 0.05, p < 0.01) were statistically significant, respectively supporting Hypothesis 4 and Hypothesis 5. We also obtained significant interactions between group-level safety climate and the quality of safety communication in prediction of safety performance (B = 0.05, S.E. = 0.02, p < 0.01) and in prediction of lost time injury (B = −0.72, S.E. = 0.05, p < 0.01), supporting Hypothesis 6 and Hypothesis 7. Fig. 1-B illustrates that the positive impact of the group-level safety climate on truck drivers’ safety performance tended to be greater as the quality of safety communication got higher. The predictors in this regression model explained 32% of the variance in safety performance. Fig. 2-B represents the moderation effect of the quality of safety communication on the relationship between group-level safety climate and lost time injury. The negative impact of group-level safety climate on lost time injury tended to be greater as the quality of safety communication got higher. The predictors in this Poisson regression model explained 7% of the variance in lost time injury.

362

Accident Analysis and Prevention 117 (2018) 357–367

Y.-h. Huang et al.

Fig. 1. -A A graphical illustration of the regression model with safety performance as a dependent variable. Fig. 1-B Group-level safety climate and safety performance relationship by the quality of safety communication (QSC).

communication was greater. These findings are consistent with the results of the regression model for testing Hypothesis 1, Hypothesis 4 and Hypothesis 6 (Fig. 1-A). At the same time, safety performance was negatively linked to lost time injury, as expected (coefficient = −0.37, S.E. = 0.13, p < 0.01). All of the indirect effects were significant, and the results supported Hypotheses 8 and 9 regarding the mediation effect of safety performance.

as well as a contingency factor influencing how safety climate relates to outcomes. 4.1. Contributions Our first contribution to the literature was to show that the quality of supervisor communication about safety uniquely contributes to safety outcomes above and beyond measures of both group-level and organization-level safety climate. We developed a measure of safety communication and showed that it had strong psychometric properties, including its structural distinctiveness from all three dimensions of a measure of group-level safety climate and its relationship with both measures of safety performance and lost time injuries. We also showed that supervisory safety communication predicted safety outcomes above and beyond several theoretically-relevant control variables, most notably including measures of organization-level safety climate and the frequency of supervisor safety communication. Our findings are consistent with past literature that treats communication as either an antecedent or a consequence of safety climate, rather than as an aspect of climate itself. In this view, supervisory safety communication may be viewed as a set of supervisory practices that can contribute to the development of a strong positive safety climate and

4. Discussion Improving long-haul truckers’ driving safety is an important challenge for safety scholars. Long-haul truckers are a special case of lone workers whose occupations have been under studied in past safety literature. Our study extended the literature on lone workers’ safety by investigating how the quality of supervisor-subordinate safety communication influenced truck drivers’ safety performance and lost time injury. We demonstrated the construct validity of a newly-adapted safety communication scale, particularly focusing on its distinctiveness from safety climate, and tested a model showing that communication had both main and moderated effects on safety behavior that ultimately predicted truck drivers’ injury rates. Thus, our findings support supervisory safety communication as a distinct influence on safety outcomes 363

Accident Analysis and Prevention 117 (2018) 357–367

Y.-h. Huang et al.

Fig. 2. -A A graphical illustration of the regression model with the lost time injury as a dependent variable. Fig. 2-B Group-level safety climate and lost time injury relationship by the quality of safety communication (QSC).

discuss safety matters with their supervisor appear to enhance the benefits of safety climate. Conversely, poor communication practices may undermine supervisory efforts to promote safety, even when subordinates perceive that safety is a priority. Future research should continue to examine communication as a constraining or enhancing factor for safety climate. Secondly, our findings contribute to the literature by extending prior research on safety climate with lone workers. Lone workers are theoretically interesting in that they may not develop aggregated perceptions of safety climate and their relationships at work are almost entirely mediated by a supervisor; in such settings, supervisors may play a particularly important role in safety outcomes. On the other hand, truckers’ supervisors are presented with additional challenges in managing safety because most of their management efforts are done remotely, through telephonic or computer-based communication. In such settings, supervisors may need different kinds of efforts to manage safety. Truckers also may face challenges with managing potential tradeoffs between safety and performance. Supervisors are likely to exert a tremendous influence on how drivers manage these tradeoffs, and communication is likely central to those efforts. A notable quality of the climate scale we used is that it specifically asks participants how their supervisor manages tradeoffs between safety and other priorities in the trucking context.

that are likely enhanced by a strong positive safety climate. Thus, in Zohar’s (2010) formulation, safety climate reflects perceptions about the relative perceived priority of safety, whereas communication is one of several organizational practices likely to influence climate perceptions. These practices include both efforts to provide subordinates with safety-relevant information (top-down communication) and efforts to encourage subordinates to discuss safety matters with their supervisor (bottom-up communication), particularly those involving potentially hazardous situations. Our findings support a view of climate and communication as constructs that are structurally distinct but related. We also showed that climate and communication interact. The positive relationship between climate and safety performance was stronger when supervisor safety communication was of higher quality. Similarly, the negative relationship between climate and lost time injuries was stronger when the quality of safety communication was higher. Interpretations of these results must be made with caution because of the potentially ambiguous relationship between climate and communication. Because both measures were gathered at the same time point for this study and because there are reasonable competing theories for the causal direction of their relationship, we did not hypothesize a specific causal pathway. However, one potential interpretation is that safety communication is an important boundary condition for safety climate. That is, efforts to share information with subordinates and encourage subordinates to 364

Accident Analysis and Prevention 117 (2018) 357–367

Y.-h. Huang et al.

Table 4 Testing the mediating effect of safety performance on the relationships between the group-level safety climate/quality of safety communication and lost time injury. Model 1

Model 1′

Model 2

Model 3

Focal study variables: Standardized Coefficient (S.E.) GSC → SP 0.22 (0.03)** COMM → SP 0.21 (0.03)** GSC × COMM → SP 0.05 (0.02)* GSC → LTI −0.69 (.32)* COMM → LTI 0.05 (0.20) GSC × COMM → LTI 0.59 (0.39) SP → LTI −0.34 (0.16)* Control variables: Standardized Coefficient (S.E.) MILES → SP −0.00 (0.00) EXP → SP 0.002 (0.001) FSC → SP −0.08 (0.01)** OSC → SP 0.18 (.02)** MILES → LTI −0.00 (0.00) EXP → LTI 0.02 (0.01)* FSC → LTI 0.66 (0.21)** OSC → LTI 0.79 (0.33)*

0.22 (0.03)** 0.21 (0.03)** 0.05 (0.02)* −0.60 (0.31) −0.10 (0.16) 0.49 (0.38) –

0.22 (0.03)** 0.21 (0.03)** 0.05 (0.02)* −0.46 (0.29) – – −0.34 (0.12)**

0.22 (0.03)* 0.21 (0.03)* 0.05 (0.02)* – – – −0.37 (0.13)*

−0.00 (0.00) 0.002 (0.001) −0.08 (0.01)** 0.18 (0.02)** −0.00 (0.00) 0.02 (0.01)* 0.63 (0.21)** 0.56 (0.27)*

−0.00 (0.00) 0.002 (0.001) −0.08 (0.01)** 0.18 (0.02)** −0.00 (0.00) 0.02 (0.01) 0.49 (0.23)* 0.56 (0.28)*

−0.00 (0.00) 0.002 (0.001) −0.08 (0.01)** 0.18 (0.02)** −0.00 (0.00) 0.01 (0.01) 0.30 (0.20) 0.35 (0.25)

Model fit summary Log-likelihood # of free parameters AIC BIC Sample-size adjusted BIC

−3239.95 18 6515.89 6625.06 6567.86

−3251.89 17 6537.79 6640.88 6686.87

−3286.02 16 6604.31 6701.07 6650.23

−3209.15 19 6456.31 6571.53 6511.16

Notes: OSC = Organization-level safety climate; GSC = Group-level safety climate; COMM = Communication; SP = Safety performance; LTI = Lost time injury; AIC = Akaike information criteria; BIC = Bayes information criteria; S.E. = Standard error. Model 1: The baseline partial mediation model. Model 1′: An alternative model of the Model 1; the SP and LTI were two separate endogenous variables and no mediation assumed. Model 2: The baseline partial mediation model; the two paths [COMM → LTI] and [GSC × COMM → LTI] were omitted as their coefficients were statistically non-significant in Model 1. Model 3: The full mediation model (final model); the path [GSC → LTI] was omitted as its coefficient was statistically non-significant in Model 2. * p < 0.05. ** p < 0.01.

4.2. Practical implications

Table 5 Direct and indirect effects of the final path model (Model 3 in Table 4). Standardized Coefficient (S.E.)

Indirect Effect (S.E.)

20,000 Bootstrapping 95% C.I.

Direct Effects GSC → SP COMM → SP GSC × COMM → SP SB → LTI

0.22 (0.03)** 0.21 (0.03)** 0.05 (0.02)* −0.37 (0.13)**

– – – –

– – – –

Indirect Effects H8: GSC → SP → LTI

–

[−0.15–−0.02]

H9: COMM → SP → LTI

–

H8 & 9: GSC × COMM → SP → LTI

–

−0.08 (0.03)** −0.08 (0.03)** −0.02 (0.01)*

Path

Our findings reinforce past research on the central role that supervisory communication plays in fostering organizational safety (e.g., Hofmann and Morgeson, 1999; Tan-Wilhelm et al., 2000). Indeed, a small but growing literature demonstrates empirical evidence that communication-focused interventions can benefit safety. Examples of specific past intervention studies include coaching construction foremen to include more safety issues in their daily exchanges with workers (Kaskutas et al., 2013; Kines et al., 2010) and providing manufacturing supervisors with feedback on their communication style (Zohar and Polachek, 2014). Future efforts to develop such interventions may benefit from including the top-down and bottom-up distinction incorporated in our safety communication measure by developing specific training components focused on each aspect of safety communication. Our results also provide additional empirical support for the importance of safety climate in the workplace, especially in the lone worker context and high-risk industries like trucking. Moreover, the industry-specific safety climate scale used in the current study can be

[−0.14–−0.02] [−0.04–−0.001]

Fig. 3. A graphical illustration of the final path model (Model 3 in Table 4; detailed results of the indirect effects testing are presented in Table 5).

365

Accident Analysis and Prevention 117 (2018) 357–367

Y.-h. Huang et al.

may underestimate the actual relationships. These distributional issues may be one reason we only explained six percent of the variance in this outcome. Future studies can take different approaches, such as using lost time injury for longer periods of time or using alternative objective outcomes with more variance (e.g., the amount of cost due to minor to major road safety incidents, frequency of near misses). It also is important to recognize that even small effects can reflect impactful findings when the effects accumulate over time (Bliese et al., 2011).

used by trucking companies to assess specific dimensions that warrant improvements. Specific leverage points can be identified using an industry-specific scale so that intervention suggestions can be more actionable and less vague. 4.3. Limitations and future directions As with any research, our study has some limitations that highlight important avenues for future research. First, the study used a crosssectional design for all data except the lost time injury outcome. This design limits the strength of potential causal inferences that may be drawn concerning the relationship between safety communication, climate, and performance and is one reason we did not make directional hypotheses concerning the climate-communication relationship. From a theoretical standpoint, it is more plausible that safety performance mediates the effects of communication and climate on injuries than would be the opposite causal paths (i.e., climate and communication mediating the effects of performance on outcomes). Moreover, climaterelated safety intervention research (e.g., Zohar and Polachek, 2014) provides additional empirical support for the inference that changes in climate exert a causal influence on safety performance. It seems likely that the relationship is reciprocal, such that climate and communication have mutually beneficial effects. However, this proposition needs to be empirically tested (Sinclair et al., 2014). Ideally, researchers might obtain measures of climate, communication, and performance at multiple time points, testing multiple models and comparing the various possibilities of relations between these measures over time. Another potential limitation is that the drivers we studied worked for a company where safety was already a high enough priority that the company was willing to invest time and effort into supporting safety climate research. The company also already had several safety-related policies in place, such as engine restrictions that limited the drivers to 60 miles per hour, safety-focused employee recognition programs and annual training, as well as more frequent meetings and teleconferences. Such circumstances may limit the amount of variability in safety climate or safety communication measures as well as in the safety outcome variables and, thus, constrain the possible observed effects, making our study a conservative test of our hypotheses. Moreover, there may be generalizability concerns as drivers from the same company may be situated in similar work environments, culture, and/or organizational structures. It may be challenging to address this issue in future research, as companies that do not prioritize safety may be less willing to support safety-focused research. One possible approach would be sampling employees through alternate sources such as through labor unions (e.g., Sinclair et al., 2010). We also relied on self-reports for most of the study variables, raising potential concerns about common method variance and social desirability effects. Self-reports are, in our view, not controversial for measures intended to capture employees’ own perceptions of their work environment, as is the case for the communication and climate measures. There may be more concern with the use of self-reported measures of safety performance, although self-reports are still the most commonly used measure of performance in the safety literature (e.g., Huang et al., 2007). Despite their common use in the literature, selfreports of performance are subject to impression management concerns, and future research should continue to supplement these measures with alternate sources of data. A strength of our study is that we were able to link the self-reported data to lost time injuries experienced over a six-month post-survey time period. The lost time injury variable does have some weaknesses that should be recognized. We measured lost time injuries reported over a relatively narrow time frame of six months. Although no clear criteria exist for setting an appropriate time interval in prospective research, it certainly is possible that six months was an insufficient time period to have enough injuries to study because of the low base rates of serious work injuries. This means that the parameter estimates in our study

Acknowledgements The authors wish to thank the following team members for their invaluable assistance: Dov Zohar (Technion – Israel Institute of Technology), Mo Wang (University of Florida), Marvin Dainoff, Michelle Robertson, Susan Jeffries, Angela Garabet, and Peg Rothwell (Liberty Mutual Research Institute for Safety) for data collection, analysis and general assistance. References Becker, T.E., 2005. Potential problems in the statistical control of variables in organizational research: a qualitative analysis with recommendations. Organ. Res. Methods 8, 274–289. Beus, J.M., Payne, S.C., Bergman, M.E., Arthur, W., 2010. Safety climate and injuries: an examination of theoretical and empirical relationships. J. Appl. Psychol. 95 (4), 713–727. Bliese, P., Adler, A.B., Castro, C.A., 2011. Research-based preventive mental health care strategies in the military. In: Adler, A.B., Bliese, P.D., Castro, C.A. (Eds.), Deployment Psychology: Evidence-based Strategies to Promote Mental Health in the Military. American Psychological Association, Washington D.C, pp. 103–124. Bliese, P., 2000. Within-group agreement, non-independence, and reliability. In: Klein, K., Kozlowski, S. (Eds.), Multi-level Theory, Research, and Methods in Organizations. Jossey-Bass, San Francisco: CA, pp. 349–381. Bureau of Labor Statistics, 2014. Census of Fatal Occupational Injuries Summary. U.S. Department of Labor, Washington, D.C(Retrieved from http://www.bls.gov/news. release/cfoi.nr0.htm). Christian, M.S., Bradley, J.C., Wallace, J.C., Burke, M.J., 2009. Workplace safety: a metaanalysis of the roles of person and situation factors. J. Appl. Psychol. 94, 1103–1127. Cigularov, K.P., Chen, P.Y., Rosecrance, J., 2010. The effects of error management climate and safety communication on safety: a multi-level study. Accid. Anal. Prev. 42, 1498–1506. Clarke, S., 2006a. The relationship between safety climate and safety performance: a meta-analytic review. J. Occup. Health Psychol. 11 (4), 315–327. Clarke, S., 2006b. Safety climate in an automobile manufacturing plant: the effects of work environment, job communication and safety attitudes on accidents and unsafe behaviour. Personnel Rev. 35, 413–430. Clarke, S., 2010. An integrative model of safety climate: linking psychological climate and work attitudes to individual safety outcomes using meta-analysis. J. Occup. Organ. Psychol. 83 (3), 553–578. Conchie, S.M., Taylor, P.J., Donald, I., 2012. Promoting safety voice with safety-specific transformational leadership: the mediating role of two dimensions of trust. J. Occup. Health Psychol. 17, 105–115. Cox, S.J., Cheyne, A.J.T., 2000. Assessing safety culture in offshore environments. Saf. Sci. 34, 111–129. DiStefano, C., Motl, R.W., 2006. Further investigating method effects associated with negatively worded items on self-report surveys. Struct. Equ. Model. 13, 440–464. Frone, M.R., 1998. Predictors of work injuries among employed adolescents. J. Appl. Psychol. 83, 565–576. George, D., Mallery, M., 2003. Using SPSS for Windows Step by Step: A Simple Guide and Reference. 11.0 Update, 4th edition. Allyn and Bacon, Boston, MA. Griffin, M.A., Neal, A., 2000. Perceptions of safety at work: a framework for linking safety climate to safety performance, knowledge, and motivation. J. Occup. Health Psychol. 5 (3), 347–358. Hofmann, D.A., Morgeson, F.P., 1999. Safety-related behavior as a social exchange: the role of perceived organizational support and leader-member exchange. J. Appl. Psychol. 84, 286–296. Hofmann, D.A., Stetzer, A., 1998. The role of safety climate and communication in accident interpretation: implications for learning from negative events. Acad. Manage. J. 41 (6), 644–646. Hu, L., Bentler, P.M., 1999. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct. Equ. Model. 6, 1–55. Huang, Y.H., Roetting, M., McDevitt, J.R., Melton, D., Smith, G.S., 2005. Feedback by technology: attitudes and opinions of truck drivers. Transp. Res. Part F: Traffic Psychol. Behav. 8, 277–297. Huang, Y.H., Chen, J.C., DeArmond, S., Cigularov, K., Chen, P.Y., 2007. Roles of safety climate and shift work on perceived injury risk: a multi-level analysis. Accid. Anal. Prev. 39, 1088–1096. Huang, Y.H., Zohar, D., Robertson, M.M., Garabet, A., Lee, J., Murphy, L.A., 2013. Development and validation of safety climate scales for lone workers using truck drivers as exemplar. Transp. Res. Part F: Traffic Psychol. Behav. 17, 5–19.

366

Accident Analysis and Prevention 117 (2018) 357–367

Y.-h. Huang et al.

model for lone workers: results of the Safety & Health Involvement for Truckers (SHIFT) pilot study. J. Occup. Environ. Med. 51, 1233–1246. Parker, S.K., Axtell, C.M., Turner, N., 2001. Designing a safer workplace: importance of job autonomy, communication quality, and supportive supervisors. J. Occup. Health Psychol. 6, 211–228. Pituch, K.A., Stapleton, L.M., Kang, J.Y., 2006. A comparison of single sample and bootstrap methods to assess mediation in cluster randomized trials. Multivariate Behav. Res. 41, 367–400. R Core Team, 2013. R: A Language and Environment for Statistical Computing, 3.0.1. R Foundation for Statistical Computing, Vienna, Austria(Retrieved from http://www.Rproject.org/). Rogerson, P.A., 2001. Statistical Methods for Geography. Sage, London. Schreiber, J.B., Nora, A., Stage, F.K., Barlow, E.A., King, J., 2006. Reporting structural equation modeling and confirmatory factor analysis results: a review. J. Educ. Res. 99, 323–338. Sinclair, R.R., Martin, J.E., Sears, L.E., 2010. Labor unions and safety climate: perceived union safety values and retail employee safety outcomes. Accid. Anal. Prev. 42, 1477–1487. Sinclair, R.R., Stanyar, K.R., McFadden, A.C., Brawley, A.M., Huang, Y.H., 2014. The role of communication in occupational safety and health management. In: Miller, V.D., Gordon, M.E. (Eds.), Meeting the Challenges of Human Resource Management: A Communication Perspective. Routledge, New York, NY, pp. 179–191. Tan-Wilhelm, D., Witte, K., Liu, W.Y., Newman, L.S., Janssen, A., Ellison, C., Yancey, A., Sanderson, W., Henneberger, P.K., 2000. Impact of a worker notification program: assessment of attitudinal and behavioural outcomes. Am. J. Ind. Med. 37, 205–213. Zohar, D., Luria, G., 2003. The use of supervisory practices as leverage to improve safety behavior: a cross-level intervention model. J. Saf. Res. 34, 567–577. Zohar, D., Luria, G., 2005. A multilevel model of safety climate: cross-level relationships between organization and group-level climates. J. Appl. Psychol. 90, 616–628. Zohar, D., Huang, Y.H., Lee, J., Robertson, M.M., 2014. A mediation model linking supervisory leadership and work ownership with safety climate as predictors of truck driver safety performance. Accid. Anal. Prev. 62, 17–25. Zohar, D., Huang, Y.H., Lee, J., Robertson, M.M., 2015. Testing extrinsic and intrinsic motivation as explanatory variables for the safety climate-safety performance relationship among long-haul truck drivers. Transp. Res. Part F Traffic Psychol. Behav. 30, 84–96. Zohar, D., 1980. Safety climate in industrial organizations: theoretical and applied implications. J. Appl. Psychol. 65, 96–102. Zohar, D., 2000. A group-level model of safety climate: testing the effects of group climate on microaccidents in manufacturing jobs. J. Appl. Psychol. 85, 587–596. Zohar, D., 2008. Safety climate and beyond: a multi-level multi-climate framework. Saf. Sci. 46, 376–387. Zohar, D., 2010. Thirty years of safety climate research: reflections and future directions. Accid. Anal. Prev. 42 (5), 1517–1522. Zohar, D., Polachek, T., 2014. Discourse-based intervention for modifying supervisory communication as a leverage for safety climate and performance improvement: a randomized field study. J. Appl. Psychol. 99, 113–124.

Huang, Y.H., Robertson, M.M., Lee, J., Rineer, J., Murphy, L.A., Garabet, A., Dainoff, M.J., 2014. Supervisory interpretation of safety climate versus employee safety climate perception: association with safety behavior and outcomes for lone workers. Transp. Res. Part F: Traffic Psychol. Behav. 26, 348–360. Kacmar, K.M., Witt, L.A., Zivnuska, S., Gully, S.M., 2003. The interactive effect of leadermember exchange and communication frequency on performance ratings. J. Appl. Psychol. 88, 764–772. Kaskutas, V., Dale, A.M., Lipscomb, H., Evanoff, B., 2013. Fall prevention and safety communication training for foremen: report of a pilot project designed to improve residential construction safety. J. Saf. Res. 44, 111–118. Kath, L.M., Marks, K.M., Ranney, J., 2010. Safety climate dimensions: leader-member exchange: and organizational support as predictors of upward safety communication in a sample of rail industry workers. Saf. Sci. 48, 643–650. Kelloway, E.K., Mullen, J., Francis, L., 2006. Divergent effects of passive and transformational leadership on safety outcomes. J. Occup. Health Psychol. 11, 76–86. Kines, P., Andersen, L.S., Spangenberg, S., Mikkelsen, K.L., Dyreborg, J., Zohar, D., 2010. Improving construction site safety through leader-based verbal safety communication. J. Saf. Res. 41, 399–406. Kline, P., 2000. A Psychometrics Primer. Free Association Books, London. Kockelman, K.M., Kweon, Y.J., 2002. Driver injury severity: an application of ordered probit models. Accid. Anal. Prev. 34, 313–321. Lee, J., Huang, Y.H., Robertson, M.M., Murphy, L.A., Garabet, A., Chang, W.R., 2014. External validity of a generic safety climate scale for lone workers across different industries and companies. Accid. Anal. Prev. 63, 138–145. Merritt, S.M., 2012. The two-factor solution to Allen and Meyer’s (1990) affective commitment scale: effects of negatively worded items. J. Bus. Psychol. 27, 421–436. Miaou, S.P., 1994. The relationship between truck accidents and geometric design of road sections: poisson versus negative binomial regressions. Accid. Anal. Prev. 26, 471–482. Michael, J.H., Guo, Z., Wiedenbeck, J.K., Ray, C.D., 2006. Production supervisor impacts on subordinates' safety outcomes: an investigation of leader-member exchange and safety communication. J. Saf. Res. 37, 469–477. Mullen, J., Kelloway, E.K., 2009. Safety leadership: a longitudinal study of the effects of transformational leadership on safety outcomes. J. Occup. Organ. Psychol. 20, 253–272. Muthén, L.K., Muthén, B.O., 2006. Mplus Software 6.0. Muthén & Muthén, Los Angeles. 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. J. Appl. Psychol. 96 (1), 71–94. Neal, A., Griffin, M.A., 2004. Safety climate and safety at work. In: Barling, J., Frone, M. (Eds.), The Psychology of Workplace Safety. American Psychological Association, Washington D.C, pp. 15–34. Neal, A., Griffin, M.A., 2006. A study of the lagged relationships among safety climate, safety motivation, safety behavior, and accidents at the individual and group levels. J. Appl. Psychol. 91, 946–953. Nunnally, J.C., Bernstein, I.H., 1994. Psychometric Theory, 3rd edition. McGraw-Hill, New York. Olson, R., Anger, W.K., Elliot, D.L., Wipfli, B., Gray, M., 2009. A new health promotion

367