Purpose and Scope of the Study

Two knowledge prerequisites for effective safety management on construction projects are: (1) an understanding of the number and nature of the safety variables that are relevant to any given project situation, and (2) an understanding of how to organize and combine these variables for effective and adaptive use. There are many human, technical, and environmental variables that influence project safety performance. The variables may be classified according to the level at which they come into play. A myriad of variables influence safety performance at the task level. They change as the nature of the construction work changes; therefore, they may be regarded as temporal variables. Examples of these variables are: guarding interior floor openings on elevated decks, cement finishers wearing gloves to protect against cement dermatitis, or using a trench box for protection against trench cave-ins. Temporal safety variables are activated by a few relatively enduring safety variables at the project organizational level. Previous research indicates that there are five principal enduring project safety variables at this level: management commitment; motivational controls; technical controls; worker group processes; and workers’ characteristics. The variables at the project organizational level are, in turn, influenced by factors outside the project organization. Important external factors are: the safety cultures of the general contractor and subcontractor firms, their safety budgets for the project, owner-emphasis on safety, liability concerns, union influence, the influence of insurance carriers, and the impact of OSHA. This study will focus on the interrelationships among the five enduring safety variables at the project organizational level (in italics) and their impact on safety performance. In order to carry out the causal analysis, the variables will need to be measured; therefore, an important aspect of the research will be to determine reliable and valid measures of the variables.

Approach

Two types of models of the five variables and their effects on safety performance will be hypothesized in the study. The first model will be a measurement model that will specify the five enduring safety variables and the empirical measures that operationally define them. The second model will be a causal model that will establish the causal relationships among the variables. The procedure will be to statistically test the hypothesized models against a correlation matrix constructed from empirical measures of the variables. The tests will be conducted using structural equation modeling - a multivariate technique for analyzing a series of dependence relationships among latent and observed variables simultaneously. The proposed models will be compared to competing models to demonstrate that no better-fitting models exist. When good-fitting (and nomologically valid) models are found, the structural parameters of the models will be estimated, and the results will be compared with past research findings. The models and analysis procedure are explained in more detail in the Analysis Section starting on page 11.

Data Source

The data will be collected through a cross-sectional survey conducted at the worksites of several construction companies. The unit of analysis will be workers. It is anticipated that approximately 200 workers will be surveyed. Workers will be asked to respond to a number of questions about the five enduring variables and safety conditions on their projects on a self-administered questionnaire. With most questions, they will be asked to indicate their perceptions on Likert-type rating scales. Demographic information will also be collected from project management personnel.

STATEMENT OF THE PROBLEM

Construction safety performance is influenced by many variables: individual, technical, environmental, and organizational at the micro, meso, and macro levels. The problem of running a safe job is complicated by the fact that the nature of the work, the environment that it is conducted in, and the people involved constantly change. The safety requirements can be totally different from one construction task to another, and the requirements constantly change as the work moves from one stage to another. As the physical environment is transformed, new hazards and obstacles are created for workers as they move about the site. New workers are continuously arriving on the site to take the place of workers who have completed their specialized tasks. They are vulnerable to accidents, until they become aware of hazards on the site and learn how to cope with them. Adding to the problem is the fact that most sites involve multiple employers.

Philip Leather (1987) researched building construction safety from a work design perspective. His synoptic view of the problem was that a multi-factor analysis approach was necessary. He also offered an explanation of why that idea is not prevalent in the construction industry. He writes:

The study of construction safety reveals a complex array of diverse, yet often closely interrelated, factors and relationships. ... a recurring obstacle to developing effective strategies for improving the construction industry’s safety record has been its overriding acceptance of single variable analysis, in particular, its preoccupation with the concept of carelessness as a unifactorial and unqualified explanation of accidents. ... What is needed, instead, is a scheme of understanding and explanation which emphasizes the multiplicity and complexity of accident causation, especially the interrelation of individual, organizational, and job variables.

An extensive literature search was performed prior to writing this proposal. Much has been written about the five enduring variables and their impact on safety performance. Also, many accident causation models involving the interaction of the variables have been formulated by researchers. Six of these models were used as support in hypothesizing the multicausal model in this proposal. Taken together, the six models seemed to account for the important organizational, individual, environmental, and technical factors that are involved. Another desirable feature of the models was that they were geared towards management action. As the research progresses, and more is learned about the enduring variables and how they interact to cause good or poor safety performance, the hypothesized causal model may be changed. For now, the six models described below seem to provide a satisfactory conceptual basis for the study.

Heinrich’s Domino Theory Model

In the 1920’s, Herbert W. Heinrich proposed a theory of accident causation based on the examination of thousands of insurance records of industrial accidents. His was the first comprehensive effort by anyone to explain the industrial accident phenomena scientifically. Before Heinrich, people believed that industrial accidents were a matter of fate. In his book, Industrial Accident Prevention (Heinrich 1941), which is now regarded as a classic, he conceptualized a domino theory of accident causation that states:

	Injuries are caused by accidents.
	Accidents are caused by unsafe acts and conditions.
	Unsafe acts and conditions are caused by the faults of persons.
	Faults of persons are caused by the social environment and ancestry.

Heinrich attempted to show that the accident sequence could be interrupted by removing any one of the dominos in the sequence. Furthermore, he stated that the bull’s eye of the accident prevention target was the unsafe act of a person or a mechanical or physical hazard.

Bird’s Modified Domino Theory Model

Many researchers felt that Heinrich’s theory attributed too much cause to factors internal to workers and neglected the importance of external factors. Around 1970, Frank E. Bird, a researcher with the International Loss Control Institute, revised Heinrich’s domino theory (Bird and O’Shell 1973). Bird’s model was a simple revision, but it was an important insight, because it introduced the thought of managerial error into the accident causation sequence. Bird’s modified domino theory is not as widely accepted by construction managers as Heinrich’s model, probably because Heinrich’s model lets them "off the hook". Blaming workers is easier and less costly than training workers, changing how an operation is performed, or making environmental modifications. Ironically, although Heinrich seemed to emphasize the fault of the worker, a careful study of his writings leaves the reader with the impression that the notion has been over-emphasized. Bird’s revised domino theory is:

	Injuries are caused by accidents.
	For every accident there are immediate causes that are related to operational errors.
	Operational errors are only symptoms of deeper underlying or basic causes related to management errors.
	The absence of a system of effective control permits the existence of the factors referred to as basic causes.

Kjellen and Larsson’s Accident Research Model

Kjellen and Larsson (1981) developed an accident causation model as a result of research work within the Occupational Accident Research Unit (OARU) of the Royal Institute of Technology in Stockholm, Sweden. It was developed to serve as a common conceptual framework for the members of the OARU and as a basis for research into the development of a systematic safety management system. The model operates on two levels: the accident sequence and the underlying determining factors. The accident sequence is a chain of deviations in the planned production process or work environment that results in an injury or property loss. The determining factors are structural properties of the production system that influence the accident sequence indirectly and change slowly in comparison with it. The researchers divided the determining factors into three categories: (1) physical/technical, (2) organizational/economic, and (3) social/individual.

Dawson, Poynter, and Stevens’ Hazards Control Model

Dawson, Poynter, and Stevens (1983) studied the safety programs of eight petrochemical facilities in Great Britain. On the basis of what they found, they proposed a safety management model designed around technical controls and motivational controls. They defined technical controls as controls that are employed against specific hazards. They might involve modifying physical or technical characteristics of the working environment, modifying specific behavior patterns of individuals, or restructuring the way workers and the environment interact. They defined motivational controls as controls that are concerned with the development and maintenance of general safety awareness and management support of technical controls. Motivational controls realistically address the need to purposely manage the motivation to implement technical controls in an environment where the chief concern is the generation of profits. The three principal elements of motivational controls they identified are: (1) setting a safety tone for the organization, (2) definitions of safety responsibility, and (3) developing mechanisms of accountability for safety performance.

Dedobbeleer’s Safety Behavior Model

Dedobbeleer (1985) studied the factors that influence construction workers’ safety behavior on nine building construction projects in the Baltimore, Maryland metropolitan area. She hypothesized a safety behavior model where workers’ safety behavior depended on three primary factors: (1) predisposing factors that related to workers’ safety knowledge, attitudes, and other personal characteristics, (2) enabling factors that related to the availability of safety training, safety equipment, and safety instructions, and (3) reinforcing factors that related to management’s attitude toward safety, foremen’s enforcement of safe conditions and practices, and co-workers’ attitudes toward safety. A basic premise of her study was that workers’ safety behavior is determined by the combined influence and interaction of these three primary factors. The study found that a combination of predisposing and enabling factors explained 51% of the variance in workers’ safety behavior. The study also found that reinforcing factors affected safety behavior, but indirectly, through predisposing factors.

LaFlamme’s Four-Level Systems Model

LaFlamme (1990), a Canadian researcher, devised a four-level model based on a systems approach. The four levels are: (1) work organization, (2) working situation, (3) accidental sequence, and (4) the accident. According to LaFlamme, work organization is a spatial variable and working situation, accident sequence, and the accident are temporal variables. The work organization level involves structural background factors (human and technical) that influence safety performance. The factors at this level concern the design, organization, implementation, and control of work processes. The factors at the second level, working situation, concern the nature of the tasks to be performed, the work environment, the machines and tools required, and the characteristics of the persons who will do the work. The third level, accident sequence, starts when a disturbance occurs in the working situation (system). The sequence can be interrupted by any of the components in the system involved. An example would be an alert foreman who averts an electrical accident by pointing out an overhead power line to a mobile crane operator. If the accident sequence is interrupted, the system will recover to a safe state again. If the sequence is not stopped, it will end as an accident (the fourth level). The accident can result in an injury, property loss, or near-miss.

RESEARCH OBJECTIVES

1. Determine reliable and valid measures of the following latent variables: management commitment, motivational controls, technical controls, workers’ group processes, workers’ personal characteristics, and safety performance.

2. Analyze the causal relationships among the above variables.

HYPOTHESES

The following hypotheses will be statistically tested in the study:

H1: The variation in motivational controls, technical controls, and safety performance is largely accounted for by explanatory causes explicitly included in the hypothesized multicausal model.

H2: The existence of a strong link between workers’ personal characteristics and safety performance is more myth than reality.

H3: The quality of motivational and technical controls is strongly linked to the degree of management commitment.

H4: The greatest determinant of a high degree of safety performance is the control of unsafe conditions.

DATA COLLECTION

Sample: Setting and Subjects

A cross-sectional survey will be conducted among workers, supervisors, safety coordinators, and site managers. The primary unit of analysis for the study will be construction workers: their attitudes and beliefs about safety actions and conditions on their projects. They represent a homogeneous group of respondents who should be knowledgeable about the questions, since the issues directly affect their well-being. The plan is to survey somewhere between 150-300 workers, since the recommended ratio of sample size to free parameters in a structural equation model is between 5:1 and 10:1, depending on the number of free parameters in the model (Harris 1990). The reader will note that there are 30 parameters in the proposed causal model on page 15 (23 l parameters and 7 g parameters). The exact number of subjects to be interviewed will be determined once it is known how many of the parameters in the causal model will be free parameters. The statistical treatments performed in the study will only involve data that are collected from the workers. Management and supervisory personnel from each of the projects will also be surveyed, but only to acquire qualitative information about safety practices, the type of employees, and other project characteristics. The latter information will be an aid in discussing the results of the statistical analysis.

The study will be a multiple-site, multiple-company study. Eight to ten construction sites will be surveyed. An effort will be made to survey an equal number of subjects from each site. It is anticipated that most of the sites chosen for the survey will be located in the Southeast. To increase the explanatory power of the data and generalizability of the results, the research will cover five or six companies and will purposely attempt to focus on examples of excellent safety programs and programs that are not as good. The study will focus on non-residential building construction projects with a value greater than $3,000,000. The surveys will be conducted during the foundation, framing, and closing-in stages of construction, because this is usually when there is much high-risk activity taking place, and concomitantly, a time when concern for workers’ safety is highest.

Survey Instruments

The survey instruments will be a structured questionnaire and an interview form. The questionnaire will be designed to collect information from workers. The interview form will be used to obtain information from site managers, supervisors, and safety officers. The workers’ survey instrument will be pre-tested on a small sample of workers to verify the appropriateness and clarity of the questions. Workers will be asked to give their perceptions regarding the 23 x measures and the 3 y measures described below. The measures were determined from an extensive study of safety literature. Several questions will be asked about each of the measures. To obtain scores for the measures, responses for each question will be given values ranging from 1 to 5. The value 5 will be given for high agreement with a statement, and a value of 1 will be given for disagreement with a statement. Therefore, if a measure consisted of four questions, the summed score for the measure could range from a low of 4 to a high of 20. A higher score will generally mean that there is a higher level of safety. It is expected that 75-125 questions will be asked, depending on the final design of the questionnaire. The questionnaire should take workers between 30-45 minutes to complete. It will be administered at the worksites, with the researcher present to provide assistance if necessary. The questionnaire will be designed to be self-administered, so it will also be possible for workers to complete it at home or during lunch time, without the researcher being present. The management/supervisor questionnaire will be less involved and should only take 15-20 minutes to complete. It will not be scored. The purpose will be to obtain background information about the different construction sites.

Empirical Measures

This section contains descriptions of the empirical measures. They are organized according to the theoretical constructs that they relate to. The references that were consulted in developing the measures are listed at the end of each section.

Management commitment

x₁ - Site manager’s safety attitude will be measured by workers’ perceptions of the site manager’s attitude about safety, in particular the degree to which safety is valued in comparison with high production. An active and consistent involvement with site safety on the site manager’s part will also be important for a high score on this variable. Essentially, this variable will measure the values, principles, and priorities of the person who is governing the construction work.

x₂ - Safety inspections and reporting will be measured by workers’ perceptions of the frequency and thoroughness of safety inspections, and whether the inspections lead to prompt and effective correction of hazards. The variable will also be measured by the frequency of safety tours by the site manager and senior management officials from the main office.

x₃ - Safety planning will be measured by workers’ perceptions of the degree to which hazards in production operations are anticipated by management and steps are taken to eliminate or mitigate them. Preparatory measures could include special training in how to cope with the hazards.

x₄ - Line accountability will be measured by workers’ perceptions of the extent to which the safety attitude and performance of supervisors is scrutinized by management.

x₅ - Control of subcontractors will be measured by workers’ perceptions of how successful the site manager is in getting subcontractors to work harmoniously together and to perform the work in a safe manner.

x₆ - Owner interest in safety will be measured by workers’ perceptions of the owner’s interest in or indifference toward safety conditions on the project.

x₇ - Status of project safety officer will be measured by workers’ perceptions of the status and authority given to the safety officer or safety representative on the site, and the extent to which his or her recommendations are acted upon.

References consulted: Adams 1976; Leather 1987; Fredin, Gerdman, and Thorson 1974; Niskanen 1994; Dodd 1986; Dawson, Poynter, and Stevens 1983; Cohen 1977; DeJoy 1985; Zohar 1980; Helander 1991; Matilla 1989; Dunbar 1975; Robinson 1982; Dedobbeleer and Beland 1991; Rundmo 1992; Levitt and Samelson 1987; Liska, Goodlow, and Sen 1992.

Motivational controls and group processes

x₈ - Supervisor’s leadership will be measured by workers’ perceptions of how capable they think their supervisor is, how well they interact and communicate with their supervisor, how concerned they think their supervisor is with their well-being, whether they think their supervisor respects their contribution, and how much say they think their supervisor has with his or her superiors.

x₉ - Supervisor’s safety attitude will be measured by workers’ perceptions of how important safety is to their supervisor, as demonstrated by the amount of advice and information that the supervisor gives on safety and the extent to which the supervisor encourages safe behavior. Also measured will be workers’ perceptions of the willingness of their supervisor to put high production ahead of following safety rules.

x₁₀ - Group and interpersonal processes will be measured by workers’ perceptions of how well-coordinated their group is in performing the work and dealing with safety hazards. For example, does the group jointly decide on work procedures and safety rules? Also measured will be the degree of negative or positive peer pressure that exists as measured by the reactions workers expect from their co-workers when they use safety equipment, use extra care, or point out unsafe conditions. Do their co-workers deride the value of working safely, or do they support and encourage safe behavior?

x₁₁ - Safety training and orientation will be measured by workers’ perceptions of the provisions that have been provided for safety instructions at the time of initial employment and the quality of ongoing safety training. The questions will also measure workers’ perceptions of the extent to which the lack of safety training opens the way for the substitution of informally determined practices in place of proven safety standards and behaviors.

References consulted: Weber 1992; Leather 1987; Konczal 1979; Guastello 1993; Vandenput 1970; Cohen 1977; DeJoy 1988; Komaki, Heinzmann, and Lawson 1980; Matilla, Dedobbeleer and Beland 1991; Hyttinen, and Rantanen 1994; Ray, Purswell, and Bowen 1993; Helander 1991; Robinson 1982; Andriessen 1978; Hinze and Gordon 1981; Levitt and Samelson 1987; Liska, Goodloe, and Sen 1992.

Technical controls

x₁₂ - Orderliness of the worksite will be measured by workers’ perceptions of the tidiness of the site, the orderliness of work processes, and the adequacy of lighting.

x₁₃ - Physical hazards controlled will be measured by workers’ perceptions of the effectiveness of the material controls that exist to protect them against falls, falling objects, electrical shock, fire and explosion, overexertion, and other hazards of a physical nature.

x₁₄ - Mechanical hazards controlled will be measured by workers’ perceptions of the effectiveness of the material controls that exist to protect them against being struck or run over by moving vehicles, caught in moving machinery, injured by power tools, and other hazards of a mechanical nature.

x₁₅ - Health hazards controlled will be measured by workers’ perceptions of the effectiveness of the material controls that exist to protect them against contact with harmful chemical substances, hearing loss, lead poisoning, inhalation of asbestos fibers, and other threats to their health.

x₁₆ - Safety equipment provided will be measured by the workers’ perceptions of the availability of proper ladders, scaffolds, personal protective gear, grounded tools, proper slings, and other appropriate safety equipment.

References consulted: Bush 1975; Fredin, Gedman, and Thorson 1974; Helander 1991; Matilla 1989; Niskanen and Saarsalmi 1983.

Workers’ characteristics

x₁₇ - Knowledge of safety and health will be measured by how well workers answer a test consisting of a number of true-false and multiple-choice type questions on safety practices.

x₁₈ - Experience and qualifications will be measured by asking workers to indicate the number of years they have worked in the building construction industry and the number of years they have worked at their present tasks, or tasks that are similar in nature.

x₁₉ - Attitude toward safety will be measured by asking workers to indicate the degree of care and attention they use to ensure their own and their workmates’ safety, what they think about the benefits of safe practices, whether they think accidents are a matter of fate, their perceptions of how careful they think their workmates are, and what they think of their foreman as a role model for safety.

x₂₀ - Risk-taking tendency will be measured by asking workers to indicate their tendency to take risks in order to do their work or increase their output and, thereby, please their supervisors. The workers will also be asked about their willingness to take risks in order to appear confident in the eyes of their fellow workers. They will also be asked to indicate if they believe they think they can take greater risks than their fellow workers, because they are better coordinated, more physically fit, or more savvy. Finally, they will asked about their attitude toward using personal protective equipment.

x₂₁ - Stability and reliability will be measured by asking workers to give their age, the length of time they have worked for their present employer, their marital status, level of education, and indicate whether or not they own a home.

x₂₂ - Alcohol and substance abuse will be measured by workers’ perceptions of how serious a problem alcohol and substance abuse is on their project, and second, how serious a problem it is in their crew. The workers will not be asked to indicate whether they personally use alcohol or drugs on the job, as it is unlikely that anyone would admit it.

x₂₃ - Stress will be measured by workers’ perceptions of pressure they feel due to the work pace, conflict and frustration with their job, and/or disharmony at home.

References consulted: DeBobes 1986; Keyserling 1983; Cohen 1977; Wuebker 1986; Helander 1991; Landeweerd, Urlings, DeJong, Nijhuis, and Bouter 1990; Mckenna 1983; Cameron 1975; Dedobbeleer and Beland 1991; Andriessen 1978; Hinze and Parker 1978; Liska, Goodloe, and Sen 1992.

Safety performance

y - Safety performance will be measured in three different ways: (1) observation by the researcher, (2) workers’ descriptions, and (3) near-miss rate. The reason for measuring safety performance in three different ways is that it is apt to be a difficult variable to measure. The three methods of measurement will help ensure that this important variable is evaluated successfully.

For measure number 1, the researcher will make a one-time observation of safety conditions on each site. A checklist will be developed that will include 30-40 items related to physical hazards, mechanical hazards, health hazards, and safety behavior. The researcher will tour the sites and note the status of the items. Each item will be marked as being fully-controlled, partially-controlled, uncontrolled, or not observed if the item either was not seen or was not relevant at the time. A safety index will be used to express the percentage of items under control on each site.

For measure number 2, workers will be asked 15-20 questions concerning their safety behavior and the state of conditions in very specific aspects of their work. The workers will not be asked to evaluate their behavior and safety conditions on the site, but to describe it in very specific terms. A safety performance index for each worker will be computed by adding up their responses to the questions. An advantage of this measure over measure 1 is that it will reflect the state of project safety over a period of time rather than at one point in time.

For measure number 3, workers will be asked about the number of near-misses they have experienced since being on the project. The number of near-misses that are reported by each worker will be added up and divided by total man-hours of exposure to get a near-miss rate for the project. The reader may wonder why number of injuries is not used instead of near-misses. There are two reasons. First, on most projects there are usually not enough injuries to draw valid conclusions about safety performance. To illustrate, on a construction project with 50,000 man-hours of exposure, the expected number of injuries is 4. It can be shown, statistically, that the project could experience 0 - 8 injuries (a 100 percent swing each way) and still be within the 95% confidence limits (Jacobs 1970). Second, there is evidence that many injuries go unreported (Accident 1992, BE&K 1993, Reid 1987). Number of near-misses as a measure does not have these problems, because there are many more near-miss incidents than injury-producing incidents (Hubbard 1988, ), and there is not an incentive to underreport.

ANALYSIS

Structural Equation Models (General)

Structural equation models (SEM’s) have their own conventions, which can be confusing. For this reason, the components and symbols will be briefly explained before proceeding with an explanation of the models shown in Figures 1 and 2.

Latent variables are represented with a circle or oval. Latent variables cannot be directly observed, but their effects on manifest variables can be measured. Manifest variables are represented with a square. A one-way arrow between two variables indicates a postulated direct influence of one variable on another. A two-way arrow indicates that the variables may be correlated. Eta (h ) designates a latent endogenous variable. Endogenous variables are variables that are measured within the system (the system being the model, in this case), and their values are affected both by other variables in the system and also by variables outside the system. Ksi (x ) designates a latent exogenous variable. Exogenous variables are those measured outside the system; they can effect the behavior of the system, but are not themselves affected by fluctuations in the system. Phi (f ) designates the degree of correlation between two variables; lambda (l ) designates a factor loading; gamma (g ) designates a path coefficient; zeta (z ) designates a residual effect; and delta (d ) designates a measurement error (Everitt 1984).

The Measurement Model

Model description

The measurement model in Figure 1 hypothesizes the relationship between the latent variables x ₁, x ₂, x ₃, and x ₄ and the empirical measures that define them (x₁ - x₂₃). The model posits that x ₁ (management commitment) is a unidimensional construct defined by the measures x₁ - x₇; x ₃ (motivational controls) is a unidimensional construct defined by the measures x₈ - x₁₁; x ₄ (technical controls) is a unidimensional construct defined by the measures x₁₂ - x₁₆; and x ₂ (workers’ characteristics) is a unidimensional construct defined by the measures x₁₇ - x₂₃. The measured variables in the model are standardized; that is, they have a mean of zero and a standard deviation of 1. Parameters f ₁ - f ₅ indicate the strengths of the correlations among the latent variables. The factor loadings l ₁ - l ₂₃ are analogous to standardized partial regression coefficients or beta weights. They indicate the strength of the measures in representing the values of the latent variables. The variables d ₁ - d ₂₃ represent the postulated errors of measurement.

SEE APPENDIX A - Figure 1. Hypothesized Measurement Model

When a measurement model fits the data set well, the multiple indicators of each latent variable should converge to measure a single underlying construct; that is, the set of measures of each construct should be unidimensional. The c ² goodness-of-fit statistic is used to assess dimensionality. Under the hypothesis that the specified model is a true reflection of reality, small r values (r < .10) indicate that the observed data does not adequately fit the hypothesized model. A reason for the bad fit could be that the sets of measures for the latent variables are not unidimensional. An examination of the reliabilities of the measures can provide additional insights.

Measurement error (reliability)

The measures of the constructs should be reliable; that is, as free from random error as possible. The reliability of a measure is the proportion of the measure that is free from error. As a rule, all indicators will contain some measurement error; i.e., unexplained variance which cannot be attributed to the effects of the latent variable. Using Figure 1 for an example, the variance in the measure x₁ contains true-score variance resulting from the influence of x ₁ and error variance resulting from d ₁. The proportion of the variance in x₁ due to x ₁ is the reliability of the measure. If there had been no errors in measurement shown, it would have meant that the variation in the indicators was assumed to be completely explained by the construct x ₁ - an unlikely eventuality. Some researchers feel that it is difficult to justify a proposed indicator of a latent variable in exploratory research when more than 50 percent of its variance is a result of measurement error (Hughes, Price, and Marris 1986).

Reformulation of the measurement model

A good-fitting model can often be achieved by successively modifying the hypothesized model (Hughes, Price, and Marris 1986). With this procedure, the hypothesized measurement model is compared with a model that has been restricted in some way. For example, if the latent variable x ₁ in the model in Figure 1 turns out to be highly correlated with the latent variable x ₃, it may be difficult to support the hypothesis that they represent distinct concepts. As a test, the model in Figure 1 can be compared with a model in which the correlation between the two constructs is restricted to 1. A difference-of-c ² test provides statistical evidence of whether the model in Figure 1 is a significant improvement over the restricted model. If the test shows that the model in Figure 1 fits the data set significantly better than the restricted model, support is provided for having two constructs rather than one construct. Other types of restrictions can be tried, until a point is reached where no further significant improvement can made in the model. The desired end result is a measurement model that consists of an optimum number of latent and indicator variables, and one that is in accord with the realities of the phenomenon under study.

The Causal Model

Model description

The multicausal model in Figure 2 specifies the hypothesized pattern of relationships among the five latent variables ‘management commitment’, ‘motivational controls’, ‘technical controls’, ‘workers’ characteristics’, and ‘safety performance.’ The hypothesized relationship between x ₁, h ₁, h ₂, and y is based on the assumption that ‘management commitment’ indirectly impacts on ‘safety performance’ through its impact on ‘motivational controls’ and ‘technical controls.’ The hypothesized relationship between x ₂, h ₂, and y is based on the assumption that ‘workers’ characteristics’ impacts directly on ‘technical controls’ and ‘safety performance.’ The parameters g ₁ - g ₅ represent the strengths of the hypothesized causal relationships among the latent variables x ₁, x ₂, h ₁, h ₂, and the measured variable y. They are called path coefficients, and like the l parameters in the measurement model, they are analogous to standardized partial regression coefficients. The parameters z ₁, z ₂, and z ₃ represent residual influences which cannot be attributed to variables shown in the model. Large residuals indicate that important causal variables have been left out. The multiple coefficient of determination, R², measures the proportion of variance that is accounted for by explanatory causes included in the model.

Fit of model to data

A good model should fit the sample data well. In terms of the equation

model + residual = data,

the goal is to keep the residual as small as possible. The adequacy of the fit between model and data is determined by comparing the correlations implied by the causal model against the correlations expressing actual relations among the empirical measures. The model is "confirmed" when the correlations that are predicted by the model correspond to the correlations among the empirical measures. If the model is not confirmed, it can be reformulated and the model-fitting procedure tried again.

Model-Fitting Computer Programs

Simple structural equation models involving a small number of empirical measures can be solved by hand quite easily. The procedure involves the algebraic solution of a small set of simultaneous path equations. However, as the number of empirical measures goes up, the number of path equations increases rapidly. (The number of equations is given by n (n - 1) ¸ 2, where n is the number of empirical measures.) As a result, model-fitting computer programs are used to solve "real-world" SEM models. Several computer programs have been developed for this purpose. The two most popular programs are EQS and LISREL. EQS is distributed by BMDP Statistical Software, Inc. (Bentler 1988), while LISREL is distributed by Scientific Software (Jöreskog and Sörbom 1993). Both programs are available for use on mainframes, workstations, and personal computers (DOS and Windows versions)

SEE APPENDIX B - Figure 2. Hypothesized Multicausal Model

It is informative to have a general idea of how the computer programs work. The large sets of path equations are solved by an iterative trial and error process. An arbitrary set of initial path coefficients serves as the starting point. The correlations implied by the path coefficients are calculated and compared to the correlation matrix of the data set. If the fit is poor (and it usually is at the start of the procedure), the computer program changes one or more of the initial path coefficients in a direction that improves the fit, and repeats the process. The cycle is repeated again and again, each time modifying the set of path coefficients to improve the agreement between the implied and observed correlations. Eventually, a point is reached where the set of path coefficients in the model cannot be improved on. This is the optimum solution. The computer program then generates a c ² statistic that detects the degree of fit between the model and the data set to which it was applied. If the c ² statistic is sufficiently small, the hypothesized model is said to be "confirmed" by the data (Loehlin 1987).

References consulted: Loehlin 1987; Li 1975; Asher 1983; Hughes, Price, and Marris 1986; Everitt 1984; Biddle and Marlin 1987.

EXPECTED RESULTS

At the beginning of this proposal, it was stated that there were two knowledge prerequisites for the implementation of effective safety management systems: (1) an understanding of the number and nature of the variables that are relative to any given project situation, and (2) an understanding of how the variables can be combined for effective and adaptive use. The two knowledge requirements will be pursued in a pragmatic, hypotheses-testing framework through the development and testing of structure equation measurement and causal models. If two models can be developed that are consistent with the observed data, they should yield considerable information about how safety variables at the project organizational level affect safety performance. These results can then be compared with existing studies.

There is much controversy about how the variables under study influence safety performance. The research will attempt to explain the impact of the variables in a clear and straightforward fashion. Understanding how these variables interact will provide important information for developing effective safety interventions.

The researcher is surmising that, at the project organizational level, there are five dimensions to the project safety function. The question is: Do management commitment, motivational controls, technical controls, workers’ group processes, and workers’ characteristics constitute five distinct, yet correlated constructs? Are the empirical measures fully accounted for by the five constructs, or are there other variables that should be considered? Perhaps there are fewer or more than five constructs. The study should help answer these questions.

The study should also provide information about the empirical measures that operationalize the constructs. It is desirable to know which measures are least affected by extraneous factors. In other words, which measures are the most reliable indicators of the true value of the constructs. The study should also indicate whether or not some of the measures should be combined.

An important reason for carrying out the study is to learn how safety variables at the project organizational level induce change in safety performance. This information will help safety practitioners control and anticipate project safety performance. Also of interest is how causal effects are transmitted, whether directly or indirectly. The parameters that are determined for the multicausal model should shed light on this question. The question of which has the greater effect - management commitment or workers’ characteristics - should be determined. Do motivational controls, technical controls, and group processes indeed act as intervening variables? Perhaps there are other variables that have not been considered that mediate between management commitment and safety performance. It is of interest to know how much of the variation in safety performance is explained by motivational controls, technical controls, group processes, and workers’ characteristics. Are there other important factors that account for the variation in safety performance? The study should help answer all these questions.

SIGNIFICANCE

LIST OF REFERENCES

	Aaker, David A. and Bagozzi, Richard P. (1979) "Unobservable Variables in Structural Equation Models with an Application in Industrial Selling." Journal of Marketing Research, 16, 147-158.
	"Accident reports off mark." (1992). Engineering News Record, August 3, 9-10.
	Adams, Edward E. (1976) "Accident Causation and the Management System." Professional Safety, October, 26-29.
	Anderson, James C. and Gerbing, David W. (1982) "Some Methods for Respecifying Measurement Models to Obtain Unidimensional Construct Measurement." Journal of Marketing Research, 19, 453-460.
	Anderson, James C. and Gerbing, David W. (1988) "Structural Equation Modeling in Practice: A Review and Recommended Two-Step Approach." Psychological Bulletin, 103(3), 411-423.
	Anderson, J. M. (1992) "Managing Safety in Construction." Proceedings of the Institute of Civil Engineering, 92(3), 127-132.
	Andriessen, J. H. T. H. (1978) "Safe Behavior and Safety Motivation." Journal of Occupational Accidents, 1, 363-376.
	Asher, Herbert B. (1983) Causal Modeling, Beverly Hills, CA: Sage Publications.
	Bagozzi, Richard P., Youjae, Yi, and Phillips, Lynn W. (1991) Assessing Construct Validity in Organizational Research, 36, 421-458.
	Barnett, Ralph L. and Brickman, Dennis B. (1986) "Safety Hierarchy." Journal of Safety Research, 17, 49-55.
	Bartinuson, J. and Weinstein, S. (1978) Occupational Hazards of Construction - A Manual for Building Trades Apprentices, A Labor Occupational Health Program Publication, Center for Labor Research and Education, Institute of Industrial Relations, Berkeley: University of California.
	"BE&K hits safety snag." (1993) Engineering News Record, December 13, 12.
	Bentler, P. M. (1988) Theory and Implementation of EQS, A Structural Equations Program, Los Angeles: BMDP Statistical Software.
	Bentler, Peter M. (1993) EQS Structural Equations Program Manual, Los Angeles, CA: BMDP Statistical Software, Inc.
	Bentler, P. M. and Bonett, Douglas G. (1980) "Significance Tests and Goodness of Fit in the Analysis of Covariance Structures." Psychological Bulletin, 88(3), 588-606.
	Bentler, Peter M. and Wu, Eric J. C. (1993) EQS/Windows User’s Guide, Los Angeles, CA: BMDP Statistical Software, Inc.
	Biddle Bruce J. and Marlin, Marjorie M. (1987) "Causality, Confirmation, Credulity, and Structural Equation Modeling." Child Development, 58(1), 5-17.
	Bird, Frank E. and O’Shell, Harold E. (1973) "The Principles of Loss Control Management." Selected Readings in Safety, Widner, Joanne T., ed., Macom, GA: Academy Press.
	Bollen, Kenneth A. (1989). Structural Equations with Latent Variables, New York, NY: John Wiley & Sons, Inc.
	Brown, R. L. and Holmes, Harold (1986) "The Use of a Factor-Analytic Procedure for Assessing the Validity of an Employee Safety Climate Model." Accident Analysis and Prevention, 18(6), 455-470.
	Bruening, John C. (1989) "Pizzagalli Construction: Performance-Oriented Safety Pays Off." Occupational Hazards, (June), 45-48.
	Bush, Vincent G. (1975) Safety in the Construction Industry, Reston, VA: Reston Publishing Company.
	Butani, Shail J. (1988) "Relative Risk Analysis of Injuries in Coal Mining by Age and Experience at Present Company." Journal of Occupational Accidents, 10, 209-216.
	Cameron, Colin (1975) "Accident Proneness." Accident Analysis and Prevention, 7, 49-53.
	Center for Excellence in Construction Safety. (1989) Report of the National Forum on Construction Safety and Health Priorities (April 23-25), West Virginia University.
	Chelius, James R. (1991) "Role of Workers’ Compensation in Developing Safer Workplaces." Monthly Labor Review, September, 22-25.
	Chhokar, Jagdeep S. and Wallin, Jerry A. (1984) "Improving Safety Through Applied Behavior Analysis." Journal of Safety Research, 15, 141-151.
	Churchill, Gilbert A. Jr. (1979) "A Paradigm for Developing Better Measures of Marketing Constructs." Journal of Marketing Research, 16, 64-73.
	Cloutier, Esther. (1994) "The Effect of Age on Safety and Work Practices Among Domestic Trash Collectors in Quebec." Safety Science, 17, 291-308.
	Cohen, Alexander. (1977) "Factors in Successful Occupational Safety Programs." Journal of Safety Research, 9(4), 169-178.
	Cooper, M. D., Phillips, R.A. , and Robertson, I.T. (1993) "Improving Safety on Construction Sites by Psychologically Based Techniques." European Review of Applied Psychology, 43(1), 33-37.
	Cornish, J. W. (1969) "Construction Safety: A Management Responsibility." New Zealand Engineering, 24(11), 359-359.
	Craske, S. (1968) "A Study of the Relation Between Personality and Accident History." British Journal of Medical Psychology, 41, 399-404.
	Cronbach, L. J. (1951) "Coefficient Alpha and the Internal Structure of Tests." Psychometrica, 6, 297-334.
	Davids, Anthony and Mahoney, James T. (1957) "Personality Dynamics and Accident Proneness in an Industrial Setting." 41(5), 303-306.
	Dawson, Sandra, Poynter, Philip, and Stevens, David (1982) "Activities and Outcomes: Monitoring Safety Performance." Chemistry and Industry, 20, 790-796.
	Dawson, Sandra, Poynter, Philip, Stevens, David (1982) "Strategies for Controlling Hazards at Work." Journal of Safety Research, 13, 95-112.
	Dawson, S., Poynter, P., and Stevens, D. (1983) "How to Secure an Effective Health and Safety Program at Work." Omega The International Journal of Management Science, 11(5), 433-446.
	Dawson, S., Clinton, A., Bamford, M., and Willman, P. (1985) "Safety in Construction: Self Regulation, Industrial Structure and Workforce Involvement." Journal of General Management, 10(4), 21-38.
	DeBobes, Leo (1986) "The Work Environment: the Psychological Factors in Accident Prevention." Personnel Journal, January, 34-38.
	Dedobbeleer, Nicole. (1985) Factors Related to Construction Workers’ Safety Practices. Doctoral Dissertation, School of Hygiene and Public Health, Johns Hopkins University, Baltimore, MD.
	Dedobbleleer, Nicole and German, Pearl. (1989) "Safety Practices in Construction Industry." Professional Safety, (January), 33-38.
	Dedobbeleer, Nicole and Beland, Francois (1991) "A Safety Climate Measure for Construction Sites." Journal of Safety Research, 22(2), 97-103.
	DeJoy, David M. (1985) "Attributional Processes and Hazard Control Management in Industry." Journal of Safety Research, 16(2), 61-71.
	DeJoy, David M. (1987) "Supervisor Attributions and Responses for Multicausal Workplace Accidents." Journal of Occupational Accidents, 9, 213-223.
	DeJoy, David M. (1990) "Toward a Comprehensive Human Factors Model of Workplace Accident Causation." Professional Safety, (May), 11-16.
	DeVellis, Robert F. (1991) Scale Development - Theory and Applications, Newbury Park, CA: Sage Publications, Inc.
	Dial, Cortney M. (1992) "Incident-Focused Managers." Professional Safety, (April), 37-45.
	Diekemper, Roman F and Spartz, Donald A. (1970) "A Quantitative and Qualitative Measurement of Industrial Safety Activities." ASSE Journal, (December), 12-19.
	Dodd, L. A. (1986) "Safety Implications in Construction." Proceedings of the Institute of Civil Engineers, Part I, 80(February), 33-40.
	Dunbar, Roger L. M. (1975) "Managerıs Influence on Subordinates Thinking About Safety." Academy of Management Journal, 18(2), 364-369.
	Dunn, G., Everitt, B., and Pickles, A. (1993) Modeling Covariances and Latent Variables Using EQS, New York, NY: Chapman & Hall.
	Dwyer, T. and Raftery, A. E. (1991) "Industrial Accidents are Produced by Social Relations of Work: A Sociological Theory of Industrial Accidents." Applied Ergonomics, 22(3), 167-178.
	Everitt, B.S. (1984) An Introduction to Latent Variable Models, New York, NY: Chapman and Hall.
	Eyssen, Gail M. and Hoffmann, Joan E. (1980) "Managersı Attitudes and the Occurrence of Accidents in a Telephone Company." Journal of Occupational Accidents, 2, 291-304.
	Ferguson, John C., McNally, Michael S, and Booth, Richard F. (1984) "Individual Characteristics as Predictors of Accidental Injuries in Naval Personnel." Accident Analysis and Prevention, 16(1), 55-62.
	Fishbein, Martin. (1967) Attitude Theory and Measurement, New York, NY: John Wiley and Sons, Inc
	Fishbein, Martin and Ajzen, Icek. (1975) Belief, Attitude, Intention, and Behavior: An Introduction to Theory and Research, Menlo Park, CA: Addison-Wesley Publishing Company.
	Fraser, C. (1980) COSAN Userıs Guide, Toronto: Ontario Institute of Studies in Education.
	Fredin, Hans, Gerdman, Peter, and Thorson, Jan. (1974) "Industrial Accidents in the Construction Industry." Scandinavian Journal of Social Medicine, 2(2), 67-77.
	Gans, George M. Jr. (1981) "Construction Manager and Safety." American Society of Civil Engineers, Journal of the Construction Division, 107(2), 219-226.
	Griew, Stephen (1960) "A Study of Accidents in Relation to Occupation and Age." Ergonomics, 2, 17-23.
	Griffiths, D. K. (1985) "Safety Attitudes of Management." Ergonomics, 28(1), 61-67.
	Grimaldi, John V. (1970) "The Measurement of Safety Engineering Performance." Journal of Safety Research, 2(3), 137-159.
	Grondstrom, Roland (1980) "Serious Occupational Accidents - An Investigation of Causes." Journal of Occupational Accidents, 2(4), 283-289.
	Grunberg, Leon. (1983) "The Effects of the Social Relations of Production on Productivity and Workersı Safety: An Ignored Set of Relationships." International Journal of Health Services, 13(4), 621-634.
	Guastello, Stephen J. (1993) "Do We Really Know How Well Our Occupational Accident Prevention Programs Work?" Safety Science, 16, 445-463.
	Gun, R. T. and Ryan, C. F. (1994) "A Case-Control Study of Possible Risk Factors in the Causation of Occupational Injury." Safety Science, 18, 1-13.
	Haas, J. (1977) "Learning Real Feelings: A Study of High Steel Ironworkers Reactions to Fear and Damage." Sociology of Work and Occupations, 4(2), 147-170.
	Hair, Joseph F. Jr., Anderson, Rolph E., Tatham, Ronald L., and Black, William C. (1992) Multivariate Data Analysis with Readings, 3rd Ed., New York, NY: Macmillan Publishing Company.
	Hale, Andrew R. and Glendon, A. Ian (1987) Individual Behavior in the Control of Danger, New York, NY: Elsevier Science Publishers.
	Hammer, Willie. (1985) Occupational Safety Management and Engineering, Englewood Cliffs, NJ: Prentice-Hall, Inc.
	Hans, Mick. (1993), "OSHA Recordkeeping: Is It Time For a Change?" Safety + Health, (November), 50-54.
	Hansen, Curtiss P. (1988) "Personality Characteristics of the Accident Involved Employee." Journal of Business and Psychology, 2(4), 346-365.
	Hansen, Curtiss P. (1989) "A Causal Model of the Relationship Among Accidents, Biodata, Personality, and Cognitive Factors." Journal of Applied Psychology, 74(1), 81-90.
	Harris, Michael M. and Schaubroeck, John. (1990) "Confirmatory Modeling in Organizational Behavior/Human Resource Management: Issues and Applications." Journal of Management, 16(2), 337-360.
	Hayduk, L. E. (1987) Structural Equation Modeling with LISREL: Essentials and Advances, Baltimore: The Johns Hopkins University Press.
	Heinrich, H. W. (1941) Industrial Accident Prevention, 2nd Ed., New York, NY: McGraw-Hill Book Company.
	Heinrich, H. W., Petersen, P. E., and Nestor Roos, D. B. A. (1980) Industrial Accident Prevention, 5th Ed., New York, NY: McGraw-Hill Book Company.
	Helander, Martin G. (1991) "Safety Hazards and Motivation for Safe Work in the Construction Industry." International Journal of Industrial Ergonomics, 8, 205-223.
	Hellander, Martin G. (1991) "Safety Hazards and Motivation for Safe Work in the Construction Industry." International Journal of Industrial Ergonomics, 8, 205-223.
	Helm, Jolene DeAnn. (1977) Personality Predictors of Accident Tendencies in Construction Workers. Doctoral Dissertation, College of Health, University of Utah, Salt Lake City, Utah.
	Hidley, John H. and Krause, Thomas R. (1994) "Behavior-Based Safety: Paradigm Shift Beyond the Failures of Attitude-Based Programs." Professional Safety, October, 28-32.
	Hinze, Jimmie. (1978) "Turnover, New Workers, and Safety." American Society of Civil Engineers, Journal of the Construction Division, 104(4), 409-417.
	Hinze, Jimmie. (1981) "Human Aspects of Construction Safety." American Society of Civil Engineers Journal of the Construction Division, 107(1), 61-72.
	Hinze, Jimmie (1987) "Qualities of Safe Superintendents." Journal of Construction Engineering and Management, 113(1), 169-171.
	Hinze, Jimmie and Pannullo, John. (1978) "Safety: Function of Job Control." American Society of Civil Engineers Journal of the Construction Division, 104(2), 241-249.
	Hinze, Jimmie, and Parker, Henry, W. (1978) "Safety: Productivity and Job Pressures." American Society of Civil Engineers Journal of the Construction Division, 104(1), 27-34.
	Hinze, Jimmie, and Gordon, Francine. (1979) "Supervisor-Worker Relationship Affects Injury Rate." American Society of Civil Engineers Journal of the Construction Division, 105(3), 253-262.
	Hinze, Jimmie and Harrison, Charles. (1981) "Safety Programs in Large Construction Firms." American Society of Civil Engineers Journal of the Construction Division, 107(3), 455-467.
	Hinze, Jimmie and Figone, Lori A. (1988) "Subcontractor Safety as Influenced by General Contractors on Small and Medium Sized Projects." Source Document 38, Construction Industry Institute, University of Texas at Austin.
	Hinze, Jimmie and Talley, Donna Marie. (1988) "Subcontractor Safety as Influenced by General Contractors on Large Projects." Source Document 39, Construction Industry Institute, University of Texas at Austin.
	Hinze, Jimmie and Raboud, Paul. (1988) "Safety on Large Building Construction Projects." Journal of Construction Engineering and Management, 114(2), 286-293.
	Hogan, Joyce and Hogan Robert. (1989) "How to Measure Employee Reliability." Journal of Applied Psychology, 74(2), 273-279.
	Hoskisson, Robert E., Hitt, Michael, A., Johnson, Richard A., and Moesel, Douglas D. (1993) "Construct Validity of an Objective(Entropy) Categorical Measure of Diversification Strategy." Strategic Management Journal, 14, 215-235.
	Houghton, C. S. Porter nee and Corlett, E. N. (1989) "Performance Differences of Individuals Classified by Questionnaire as Accident Prone or Non-Accident Prone." Ergonomics, 32(3), 317-333.
	House, Robert J. and Rizzo, John R. (1972) "Toward the Measurement of Organizational Practices: Scale Development and Validation." Journal of Applied Psychology, 56(5), 388-396.
	Hovden, Jan and Tinmannsvik, Ranveig Kviseth. (1990) "Internal Control: A Strategy for Occupational Safety and Health. Experiences from Norway." Journal of Occupational Accidents, 12, 21-30.
	Hubbard, R. K B. and Neil, J. T. (1986) "Major-Minor Accident Ratios in the Construction Industry." Journal of Occupational Accidents, 7, 225-237.
	Hughes, Marie Adele, Price, Leon R., and Marris, Daniel W. (1986) "Linking Theory Construction and Theory Testing: Models with Multiple Indicators of Latent Variables." Academy of Management Review, 11(1), 128-144.
	Hunter, John E. and Gerbing, David W. (1982) "Unidimensional Measurement, Second Order Factor Analysis, and Causal Models." Research in Organizational Behavior, 4, 267-320.
	Hyttinen, Marita. (1993) "The Building Site from the Viewpoint of Work Psychology and Safety." Psykologia, 28(3), 187-195.
	Jacobs, Herbert H. (1970) "Towards More Effective Safety Measurement Systems." Journal of Safety Research, 3(3), 160-175.
	James, Lawrence R. and Jones, Allan P. (1974) "Organizational Climate: A Review of Theory and Research." Psychological Bulletin, 81(12), 1096-1112.
	James, Lois A. and James, Lawrence R. (1989) "Integrating Work Environment Perceptions: Explorations into the Measurement of Meaning." Journal of Applied Psychology, 74(5), 739-751.
	Jaworski, Bernard J. and Kohli, Ajay K. (1993) "Market Orientation: Antecedents and Consequences." Journal of Marketing, 57, 53-70.
	Jorëskog and Sörbom. (1993) LISREL 8: Structural Equation Modeling with the SIMPIS Command Language, SSI Scientific Software International, Hillsdale, NJ: Lawrence Erlbaum Associates.
	Kamp, John. (1991) "Preemployment Personality Testing for Loss Control." Professional Safety, (June), 38-40.
	Keats, Barbara W. and Hitt, Michael A. (1988) "A Causal Model of Linkages Among Environmental Dimensions, Macro Organizational Characteristics, and Performance." Academy of Management Journal, 31(3), 570-598.
	Kerr, William. (1957) "Complementary Theories of Safety Psychology." The Journal of Safety Research, 45, 3-9.
	Kharbanda, Om P. and Stallworthy, Ernest A. "Company Culture: Its Role in an Industrial Society." Industrial Management and Data Systems, 91(2), 42-46.
	Kim, Jae-On, and Mueller, Charles W. (1978) Introduction to Factor Analysis - What It Is and How to Do It, Beverly Hills, CA: Sage Publications, Inc.
	Kjellen, Urban. (1982) "An Evaluation of Safety Information Systems at Six Medium-Sized and Large Firms." Journal of Occupational Accidents, 3, 273-288.
	Kjellen, Urban. (1983) "Application of an Accident Process Model to the Development and Testing of Changes in the Safety Information Systems of Two Construction Firms." Journal of Occupational Accidents, 5(2), 99-119.
	Kjellen, Urban. (1984) "The Role of Deviations in Accident Causation and Control." Journal of Occupational Accidents, 6, 117-126.
	Kjellen, Urban. (1984) "The Deviation Concept in Occupational Accident Control-I." Accident Analysis and Prevention, 16(4), 289-306.
	Kjellen, Urban and Hovden, Jan. (1993) "Reducing Risks by Deviation Control - A Retrospective into a Research Strategy." Safety Science, 16, 417-438.
	Kjellen, U. and Larsson, T. J., (1981) "Investigating Accidents and Reducing Risks - A Dynamic Approach." Journal of Occupational Accidents, 3, 129-140.
	Kletz, Trevor A. (1993) "Accident Data - the Need for a New Look at the Sort of Data that are Collected and Analyzed." Safety Science, 16, 407-415.
	Kohli, Ajay K. and Jaworski, Bernard J. (1990) "Market Orientation: The Construct, Research Propositions, and Managerial Implications." Journal of Marketing, 54, 1-18.
	Komaki, Judi, Barwick, Kenneth D., and Scott, Lawrence R. (1978) "A Behavioral Approach to Occupational Safety: Pinpointing and Reinforcing Safe Performance in a Food Processing Plant." Journal of Applied Psychology, 63(4), 434-445.
	Komaki, Judi, Heinzmann, Arlene T., and Lawson, Loralie. (1980) "Effect of Training and Feedback: Component Analysis of a Behavioral Safety Program." Journal of Applied Psychology, 65(13), 261-270.
	Komaki, Judith L., Collins, Robert L., and Penn, Pat. (1982) "The Role of Performance Antecedents and Consequences in Work Motivation." Journal of Applied Psychology, 67(3), 334-340.
	Konczal, Edward F. (1979) "The Supervisors Role in Accident Prevention." Supervisory Management, July, 31-34
	LaBar, Gregg. (1992) ""Breaking New Ground in Construction Safety." Occupational Hazards, (May), 58-63.
	LaCroix, David V. and DeJoy, David M. (1989) "Causal Attributions to Effort and Supervisory Response to Workplace Accidents." Journal of Occupational Accidents, 11, 97-109.
	LaFlamme, Lucie. (1990) "A Better Understanding of Occupational Accident Genesis to Improve Safety in the Workplace." Journal of Occupational Accidents, 12, 155-165.
	Landeweerd, Jan A., Urlings, Ilse J. M., Dejong, Ad H. J., Nijhuis, Frans J. N., Bouter, Lex M. (1990) "Risk Taking Tendency Among Construction Workers." Journal of