Workplace Safety Climate Surveys for City Center and Cosmopolitan Construction Projects, Las Vegas, Nevada (Final Report)

Contents

Acknowledgments
Detailed contents of the Final Report
A. Introduction
B. Methods
C. Results

1. Construction Workers
2. Foreman
3. Superintendents
4. Executives
5. Comparison of 4 groups

D. Conclusions and Recommendations
Appendices
References

B. METHODS

Four separate safety climate surveys were developed for administration on the City Center and Cosmopolitan construction projects. The climate surveys were created, in English and Spanish, containing 37 safety-related questions about perceptions about the General Contractor (Perini), about employers (sub-contractors), and about individuals' perceptions of safety on the job. Workers, foremen, superintendents and top management executives who volunteered for the survey were given the option of completing either the English or Spanish version of the survey. Of the 1,035 workers who described themselves as Hispanic 730 (70%) completed the English version of the survey, compared with 305 (30%) who completed the Spanish version.

Each of the 37 questions has a 6-point Likert scale response ranging from strongly disagree to strongly agree. Demographic questions such age, years worked in construction, work status (apprentice or journeymen), trade, and city / home local, and ethnicity were included in the survey. There is also an open-ended question where persons responding to the survey can provide additional comments about the job and actions that could be taken to improve safety. The survey was pilot tested with workers on the site for two weeks to evaluate the specificity and sensitivity of the questions and the viability of the protocol for integration into the Smart Mark/OSHA 10 hour training classes.

The safety climate surveys were distributed during the OSHA 10 hour classes managed by CPWR as specified in the MOU between Perini and the Southern Nevada Building Trades Council. The worker and superintendent surveys were explained and administered to workers by CPWR trainer/coordinators and at weekly superintendent meetings, respectively. A separate schedule for administering the survey was established for foremen and top management executives. Foremen received the survey during a safety culture training session, and top management executives were directly provided the survey and a self-addressed mailing envelope to return to CPWR. Appendices 1-1 to 1-4 contain the final version of the climate surveys used for workers (both in English and in Spanish), foremen, superintendents, and top management executives. All four surveys contained 24 comparable questions on demographics, questions about the general contractor (Perini), questions about the subcontractor, and questions about individual perceptions about jobsite safety. Appendix 1-5 contains the mapping of questions across the surveys to standardize the data analysis. These surveys were shared with Perini and the head of the Southern Nevada Building Trades Council as part of the participatory process to evaluate the safety climate of the City Center and Cosmopolitan construction site.

The protocol for administering the surveys was as follows. First, a CPWR trainer/training coordinator provided background information describing the purpose of the survey to collect information about safety on the site and the voluntary nature of the survey. There was no individual information (personal identifiers) collected. Second, participants were provided the survey and asked to complete it in ten minutes. Once completed, the survey was collected by the trainer/coordinator and put in a sealed self-addressed envelope and sent directly back to CPWR for data entry and analysis.

Between the dates of July 8 and October 21, 2008, a total of 3,781 individuals completed the survey, 3,482 of whom were craft workers attending the OSHA 10 hour training classes at either the City Center or Cosmopolitan sites. Of those attending the training, 3,035 (87%) of the workers completed the safety climate survey. One hundred and thirty-four foremen (96%) (hired by Perini) completed the survey, sixty-one (43%) (hired by Perini) superintendents completed the survey, and 17 (100%) top management executives completed the survey. The overall response rate for the four groups was 86%. Once the surveys were completed and received by CPWR, individual surveys were given a unique identifier and sent to BizCore, Langhorne, PA. for data entry. Survey data were entered into an excel database and were reviewed for quality control by CPWR. Data analysis was performed by CPWR and collaborators at Colorado State University and Illinois Institute of Technology.

Statistical Analysis

There are three components to the analysis of the safety climate survey data. First, we provide a descriptive analysis which examines the demographic characteristics of the people who responded to the survey, and the aggregate/collective responses to each of the questions. Second, a set of psychometric analyses were conducted to identify factors which stand out as important indicators of the safety climate and other related safety issues. Finally, a series of inferential statistics were used to examine mean differences of these factors and to identify important contributing factors of safety performance.

Descriptive Analysis

Numerical values were assigned to each of the survey responses, so that strongly disagree = 1, disagree = 2, somewhat disagree = 3, somewhat agree = 4, agree = 5, and strongly agree = 6. Lower means scores indicate stronger disagreement whereas higher mean scores indicate stronger agreement for each of the 37 questions. Data are presented as percentages and mean ± standard deviation (STD). The student t-test was used to compare means between two groups. The analysis of variance (ANOVA) method was used when comparing means of more than two groups (Howell, 2007)¹ All statistical analysis was performed using SAS for Windows Version 9.1 (SAS Institute Inc., 2003).

Psychometric Analysis

As shown in Appendix 1, each survey consists of a set of specific questions targeted to each surveyed group. To facilitate the analysis and discussion of the survey results, and to also provide a clearer view of the big picture, it is common to empirically identify a small set of underlying themes based on the larger number of survey items by means of a factor analysis technique (Gorsuch, 1983)². These underlying themes or factors (as a statistical term) can be considered as indices of meaningful concepts. Specifically, we used factor analysis to identify the structure of the relationships among the survey items based on data, which was collected from four surveys designed for workers, foremen, superintendents/assistant superintendents/project managers, and Perini management executives. The structures of these survey items were further verified by two additional statistical methods (parallel analysis and item analysis). Parallel analysis is a statistical technique which provides an empirical criterion to evaluate the factor analysis results (Horn, 1965)³ and to assist us determine the number of factors that should be retained based on the survey items. Item analysis is an empirical method to analyze to what extent a survey item is related to its correspondent factor (Allyn & Yen, 2001)⁴ . After that, survey items, which correlated strongly with each other and at the same time did not correlate strongly with other items, were grouped together into factors. Some of these factors included multiple items, while others had a single item. The sum of the grouped item scores represents the score of the factor of interest (e.g., Perini safety climate score).

Based on the above analyses, 16 factors, 15 factors, 13 factors, and 12 factors were identified for the worker, foremen, superintendent, and the executive surveys, respectively. Definitions of the factors in the four surveys are provided in Appendix 2-5. In addition, the correspondent survey items of each factor for the surveys, and descriptive statistics (alpha coefficient, possible range, observed range, mean, standard deviation, and number of respondents) are reported in Tables 2, 4, 6, and 8, respectively. Alpha coefficients range from 0 to 1, with values closer to 1 indicating that the set of items measures well a common underlying theme (i.e., factor). Calculation of an alpha coefficient involves correlations among items. As a result, alpha coefficients cannot be computed for factors consisting of a single survey question. The mean of each factor is calculated by summing the individual responses to the set of items for each factor and computing their arithmetic average. For example, the average Perini Safety Climate score based on the 2,817 workers who responded to all seven items is 29.7, as shown in Table 2 below. The standard deviation of each factor score is an index that shows how widely the respondents' scores differ from the mean. When the distribution of respondents' scores exhibits a bell-shape curve, as a rule, about 2/3 of these scores are expected to fall within the range of the mean score minus one standard deviation and the mean score plus one standard deviation. For instance we can estimate that approximately 2/3 of all scores on the Perini Safety Climate falls between 22.2 (i.e., 29.7 – 7.5) and 37.2 (i.e., 29.7 + 7.5), assuming the distribution of Perini Safety Climate scores exhibits a bell-shape curve. Finally, sample size is the number of respondents who respond to all items in a given factor. Thus, it varies from one factor to another. For example, 2,817 workers responded to all seven items comprising Perini Safety Climate, while 2,843 workers responded to all eight items comprising Foreman Safety Management, as shown in Table 2.

Limitations

This report documents our final examination of the climate survey data. Of the 9,300 workers on the two sites (as of December 19, 2008) it was initially estimated that seventy percent had not received OSHA 10 hour training. This survey was only provided to those workers who had not previously received OSHA 10 hour training.

---

¹ Howell, D. C. (2007). Fundamental Statistics for the Behavioral Sciences, (6th ed.)
² Gorsuch, R. L. (1983). Factor analysis (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum.
³ Horn, J. L. (1965). A rationale and test for the number of factors in factor analysis. Psychometrika, 32, 179-185.
⁴ Allyn, M. J., & Yen, W. (2001). Introduction to Measurement Theory. Waveland Press.

---

Table of Contents