“ WEE. c"

. .u E“... l: .
:8qu

a. Jung” 3.12

e .h
I “._4.
Pat:
.: z

t .1 .
iw as: an.
2 A .353: .
HM“. . :3. .13.... arnfifiu ,
3%. .WM».::!). 10!...
v; 1 .fl .31... an. :
. Win?!- 3
4.1%. . MPH...
.. 541.15»... .
"£4

1%
2‘75. i
xix-r .
".3“. at?
. 5.1:.

I:

.I ‘

u .3.
7

 

.
. fig:§
31%.. . .
. 1. . .5 ,
2.“..5. w
WAD-wail
NJ, aflrxo‘fikahlhfi
.i h... 3..“ Va... Hols.
I3 . Java? .

 

 

£7 a 1 5;... x.»
, I: . . .. 31.
MmurCJPn-Wflnfifivmmlewhfix xyl“
5 tv Vac ' .
. . . f
i..:...£.$§é.fifzg§ ,9me 3...
: .sk raft. 5.7!; : 150d”
.yht’atuct’rremfi
fiannnvj
It.
I‘D-'71.,
Azbciii:

flu ’dlflfflH-unfiuh.
are». q. www.mfid,
.mhuvn {51,347 «gov-6U!”
kw.” PM... “mum”. rm .mWJH “sure-m
I; g V l! 2‘. - I
. At<I.I'-.l F'.‘
v v A ‘

Al" (I ‘
if .1333».
saver? {13%).
I Ila: ; 1‘ ‘.
c. I;

39...}:

322112.
1:1- 1, 13
.33.. 35:74.

1.}!

. I’ll}; a :1}

.5 apt?!

ls st .21..

.tnfi'lnn cg ._
a? 1.75.1342, . 2.5 5 .mr
, I .-

. V . _. .r. 3;. . ,.

ct}. "Va h‘bh‘ :3 . v.2 .4...., ‘ .
fifififi; haw... V ..

 

.' R233

coo?

This is to certify that the
thesis entitled

THE RELATIONSHIP BE'IWEEN CONSTRUCTION
WORKER SAFETY AWARENESS AND PERSONALITY

presented by

Suzann 8. von Bemuth

has been accepted towards fulfillment
of the requirements for the

MS. degree in Construction Manage_ment

 

 

 

 

. gZ, .
Major Professor’s Signature
3/8/3 4

Date

 

MSU is an Affirmative Action/Equal Opportunity Institution

7 ———— -—~—- - 7————— ———— ---- *7

LIBRARY
Michigan State
University

 

 

 

 

.-c-D-3-D-0-.-C’l-I-o-.-C-o--o—n--.-A—---e--o-o-g'ei-t-o-o-o-Q-o-o-c-c-a-c-I-o-o-n-O-I-l-o--o—~—c-o-o-o-o-.-l-I-O-n-I-D-n—-_--

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/ClRC/DaleDue.indd-p.1

 

THE RELATIONSHIP BETWEEN CONSTRUCTION WORKER
SAFETY AWARENESS AND PERSONALITY

By

Suzann S. von Bemuth

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE
Construction Management Program

2006

ABSTRACT

THE RELATIONSHIP BETWEEN CONSTRUCTION WORKER SAFETY
AWARENESS AND PERSONALITY

By:

Suzann S. von Bemuth

Despite the industry’s best efforts to minimize risks, hazards, injuries and fatalities in
construction, it remains a hazardous occupation. Research into accident causation has
been performed and is ongoing, and new ideas surface everyday. This thesis is a foray
toward the integration of psychology into the predominantly engineering field of
construction by investigating the relationship between construction worker personality
and hazard perception. Thirty ironworkers were surveyed and data pertaining to their
sensitivity to hypothetical hazards, their strategy (risk response criterion) in reacting to
those hazards and their personality were measured. Signal Detection Theory was utilized
to determine sensitivity and response bias (strategy), and the Five Factor Model was used
to evaluate personality. Twenty-six out of thirty workers surveyed had a conservative
strategy. Multiple regressions performed on all variables relating to sensitivity (d')
resulted in an R2 = .34 and p = .06 for experience, Extraversion, Agreeableness,
Conscientiousness and Emotional Stability, taken together, and concerning strategy ([3)
were R2 = .37 and p = .017, for age, accidents, Agreeableness, and Emotional Stability,
taken together. The workers in this study who scored in the low range on the Five Factors
were found to have lower than average sensitivity, and tended to have more conservative

strategies. The null hypothesis (there is no relationship) was rejected for both d' and B.

ACKNOWLEDGEMENTS

I am indebted to my fiiends and family for their support, patience and love during this
endeavor, and I am grateful to my instructors, the staff of School of Planning, Design and
Construction, and my fellow students here at Michigan State University.

I would like to specifically and specially thank:

My parents:
Bob and Judy von Bemuth,
who always knew I would.

My partner and loving wife:
Kristin Grew,
who always knew I could
My committee members, who are also my fi'iends:

Dr. Tariq Abdelhamid, Dr. Mohammad Najafi, Dr. Jon Barley, and Mr. Paul Streng,
who gave me the chance.

Additional thanks to Mr. Art Ellul and Mr. Larry Kruth for their willingness to provide
input and support, and thanks for just being two great guys!

A very heartfelt thanks to Harry Gregson-Williams, for help with the notes.

iii

TABLE OF CONTENTS

LIST OF TABLES .......................................................................... vi
LIST OF FIGURES .......................................................................... vii
CHAPTER 1: INTRODUCTION ......................................................... l
1.1 Motivation .............................................................................. 2
1.2 Prior Research ......................................................................... 3
1.2.1 Sanders and McCormick Theory Categories .............................. 4
1.3 OHSA Accident Definition and Categories ....................................... 7
1.4 Problem Statement ................................................................... 8
1.5 Goals and Objectives ................................................................ 9
1.6 Scope of the Research ................................................................. 10
1.7 Thesis Overview ........................................................................ 10
CHAPTER 2: BACKGROUND ........................................................... 11
2.1 Accident Causation Research ...................................................... 12
2.2 Accident Causation in Construction ............................................... 16
2.2.1 Systems Model of Construction Accident Causation
(Mitropoulos et a1. 2005) ..................................................... 16
2.2.2 Incident Causation Model for Improving Feedback of
Safety Knowledge (Chua and Goh, 2004) .................................. 17
2.2.3 Corporate Culture (Molenaar et al., 2002) 17
2.3 Signal Detection Theory (SDT) ................................................... 18
2.4 The Use of SDT in Construction .................................................. 26
2.5 The Five Factor Model (F F M) .................................................... 27
2.6 People-Based Safety (Geller and Weigand, 2006) .............................. 30
2.7 The Five Factor Model and Safety in the Workplace
(Cellar et al., 2001) .................................................................. 34
2.8 The MMPI and the MBTI ........................................................... 37
2.9 Chapter Two Recap .................................................................. 39
CHAPTER 3: METHODOLOGY ....................................................... 40
3.1 Objective One ........................................................................ 41
3.2 Objective Two ......................................................................... 42
3.2.1 Objective Two Part A ......................................................... 42
3.2.2 Objective Two Part B ......................................................... 47
3.3 Additional Demographic Data ..................................................... 50
3.4 Methodology Summary ............................................................. 50
CHAPTER 4: SURVEY RESULTS AND DATA ANALYSIS ..................... 52

iv

4.1 Objective One: The Model .......................................................... 53

4.2 Data Collection ....................................................................... 56
4.3 SDT Analysis: Sensitivity and Strategy .......................................... 60
4.4 FFM Analysis ......................................................................... 61
4.5 Regression Analysis ................................................................. 63
4.6 Grouped Analysis (SDT) ............................................................ 64
4.7 Grouped Analysis (F FM) ........................................................... 69
4.8 Principle Component Analysis .................................................... 75
4.9 Multiple Regressions ................................................................ 75
4.9.1 d' versus Age, Exp., Acc., EX, A, C, ES and O ........................... 76
4.9.2 [3 versus Age, Exp., Acc., EX, A, C, ES and O ............................ 77
4.10 Discussion ........................................................................... 78
4.11 Results and Analysis Summary ................................................... 82
CHAPTER 5: SUMMARY AND CONCLUSIONS ................................... 83
5.1 Thesis Summary ..................................................................... 84
5.2 Conclusions ............................................................................. 84
5.3 Recommendations for Future Research ........................................... 88
5.4 Contributions of this Research ...................................................... 89
APPENDICES
Appendix A: Consent Letter and Survey Questionnaire ............................. 91
Appendix B: Survey Responses and Analysis (SDT) ................................ 108
Appendix C: Survey Responses and Analysis (FFM) ............................... 117
Appendix D: Regression Analysis: MIN ITAB Results .............................. 120
Appendix E: Normal Deviates and Ordinates for d’ and B 137
Appendix F: Multiple Regression Results ............................................. 139
Appendix G: Sample Distributions for the Five FactorRanges .................... 145 '
REFERENCES ............................................................................... 148

LIST OF TABLES

Table 2.1 The Four Outcomes of Signal Detection Theory ........................ 19
Table 2.2 Example Probabilities ...................................................... 24
Table 3.1 SDT Matrix for Detection of Unsafe Conditions in Construction. ...43
Table 3.2 Sample Survey Responses ................................................. 45
Table 4.1 Summary of Data ............................................................ 62
Table 4.2 Regression R, R2 and p-values ............................................. 64
Table 4.3 Average Variable Value Differences by Strategy Groups ............. 65
Table 4.4 Average Variable Value Differences by Sensitivity Groups .......... 67
Table 4.5 Five Factor Ranges .......................................................... 69
Table 4.6 d' and the Five Factors ...................................................... 70
Table 4.7 B and the Five Factors ....................................................... 72
Table 4.8 PCA Eigen Vector Matn'x .................................................. 75
Table 8.] Survey Responses and Analysis (SDT) .................................. 109
Table C.1 Survey Responses and Analysis (FF M) .................................. 118
Table 13.1 Normal Deviates and Ordinates for Calculating d’ and B .............. 138

vi

Figure 1.1
Figure 1.2
Figure 1.3
Figure 2.1
Figure 2.2
Figure 2.3
Figure 2.4
Figure 3.1
Figure 4.1
Figure 4.2
Figure 4.3
Figure 4.4
Figure 4.5
Figure 4.6
Figure 4.7
Figure 4.8
Figure 4.9
Figure 4.10
Figure 4.1 1
Figure 4.12

Figure 4.13

LIST OF FIGURES

Construction Worker Fatalities per 100,000 Workers ................. 3

Model of Contributing Factors in Accident Causation ................ 6

Accident Cause Levels .................................................... 7

Three Zones of Risk ........................................................ l4
Migration of Work Toward Loss of Control ............................ 15
Distribution of Detection Theory ......................................... 20
Sensitivity Ranges for d' ................................................... 21
Ranges for the Five Factors ................................................ 50
The ABCs of Accidents ..................................................... 54
The Expanded ABCs of Accidents ....................................... 5 6
Unsafe Situation with “Safe” or “I am not sure” Responses .......... 59
Safe Situation with “Unsafe” or “I am not sure” Responses .......... 60
Risky Group Averages Compared to Overall Averages ([3) ........... 66

Conservative Group Averages Compared to Overall Averages (0).. 66
Risky Group Averages Compared to Overall Averages (d ') ............. 67

Conservative Group Averages Compared to Overall Averages (d').68

Low Five Factor Scores Compared to Sensitivity (d') ................. 70
Moderate Five Factor Scores Compared to Sensitivity (d ’) ........... 71
High Five Factor Scores Compared to Sensitivity (d ') ................. 71
Low Five Factor Scores Compared to Strategy ([3) ..................... 72
Moderate Five Factor Scores Compared to Strategy (B) ............... 73

vii

Figure 4.14 High Five Factor Scores Compared to Strategy (B) ..................... 73

viii

Chapter 1
INTRODUCTION

1. INTRODUCTION

“What lies behind us and what lies before us are small matters compared to what lies
within us. ” (Ralph Waldo Emerson, circa 1880) In the context of this research, what lies
before us is our work future, hopefully accident free. What lies behind us is our work
history (hopefully also accident free,) life experiences and genetically heritable attributes.
What lies within us, our personality, has been shaped by what lies behind us, and will
have an effect on what lies before us. The aim of this research was to discover the link
between what lies within us (personality) and the way we see the world; specifically, how
construction workers interpret the safety of situations that surround them each and every

day on the jobsite.

1.1 Motivation

According to the Bureau of Labor Statistics (BLS), the construction industry employed
approximately 5.2% of all US. workers in 2003, yet accounted for 22.4% of occupational
fatalities in the same year (BLS, 2005). In 2004, the construction industry employed
approximately 6% of all US. workers, yet accounted for 10.5% of nonfatal occupational
injuries, and 24.3% of occupational fatalities (BLS, 2005). These statistics show that,
with respect to its representation of the total workforce, construction accounts for a
disproportionate segment of occupational injuries and fatalities. The 2005 BLS data show
1131 construction fatalities and 408,300 nonfatal injuries and illnesses. In fact, the US.
Department of Labor classifies construction as a high-risk occupation due to the

incidence of injuries and fatalities that occur in the construction industry every year.

In the recent past, the number of fatalities per 100,000 workers has dropped from a high
of 5.3 in 1994 to 4.0 in both 2002 and 2003. In 2004, an increase (the first since 1994)
was seen, from 4.0 to 4.1. These statistics are represented in Figure 1.1. The fact is that
the construction industry is a high-risk occupation. However, the question is, what is

being done to address these disparate statistics?

Construction Worker Fatalities

 

 

0 . . l a
1990 1 992 1 994 1 996 1998 2000 2002 2004 2006 .

 

Figure 1.1 Construction Worker Fatalities per 100,000 Workers (U.S.)

(Bureau of Labor Statistics, 2005)

1.2 Prior Research

The existing research literature has typically investigated accident type categorization and
the perceived causes of these accidents (Culver et a1, 1992; Hinze, 1996; Hinze et al.,
1998; Suraji et al., 2001). These investigations have mainly addressed “what” happens
concerning accidents, and “how” it happens. Few studies have addressed “why” the
accidents occur, however, three categories of “why” theories have been proposed by
Sanders and McCormick (1993). These categories are: 1. Accident Proneness Theories,

2. Job Demand vs. Worker Capability Theories and 3. Psychosocial Theories.

1.2.1 Sanders and McCormick Theory Categories

In category one, Accident Proneness Theories, Sanders and McCormick discuss the
theory that “some people are more prone to have accidents than others because of a
peculiar set of constitutional characteristics”, meaning there is some type of permanent
characteristic that certain people possess that makes them have more accidents than
others, and more accidents than they could be expected to have according to pure chance.
This particular type of theory was later challenged by McKenna (1983) on the grounds
that it assumed that all the workers were subject to identical job and environmental
factors. McKenna felt that the higher than chance occurrence of accidents for some
workers might be indicative of a higher risk occupation.

On the heels of the Accident Proneness Theories, the Accident Liability Theory
arose. This theory suggests that accident occurrence fluctuates with age. Multiple studies
(Broberg, 1984, National Academy of Science, 1982 and Shahani, 1987) found that
younger workers have more accidents than older workers, and though there is an increase
in accident occurrence for workers over 50, the rate of accident occurrence was still less
than that of the youngest workers. Some explanations for these findings are that younger
workers tend to be more inattentive, impulsive and reckless, have less discipline, and tend
to overestimate their own abilities more than older workers (Lampert 1974).

Sanders and McCormick define the theories in category two (Job Demand vs.
Worker Capability Theories) as including the basic assumption that “accident liability
increases when job demands exceed worker capabilities” (Sanders and McCormick,
1993). Simply put, the more demanding a job is, the more accidents can be expected.

The Adjustment to Stress Theory, also in-this second category, hypothesizes that accident

rates will increase when the level of stress exceeds the worker’s ability to cope with
stress. Sanders and McCormick identify these stressors as noise, poor illumination,
anxiety, lack of sleep, anger, etc. It is indicated that the research concerning these factors
is mixed, and not definitive.

Another theory in category two is the Arousal-Alertness Theory, and it involves
inappropriate arousal levels: either too high, such as excessive motivation, or too low,
such as boredom or apathy. A study by Brown (1990) indicates that care should be taken
to differentiate between stress and arousal, saying that the stress is defined as harmful,
but arousal is not necessarily so.

Psychosocial Theories (category three) includes the Goals-Freedom-Alertness
Theory, which indicates that when workers are given more control over work goals and
management is decentralized, accident occurrence is lower. This category also includes a
subset of theories, called psychoanalytical theories, which view accidents as “self
punitive acts caused by guilt and aggression” (Sanders and McCormick, 1993). It is
argued that this Freudian view of accident causation is generally dismissed as a poor
indicator of generalized accident causes.

Sanders and McCormick opine that no single theory or model of accident
causation is sufficient to represent the full scale of accident situations, and point to an
amalgamated model, developed by Sanders and Shaw, which they believe is broad
enough to encompass virtually every possible contributing factor from the other models
they reviewed, but specific enough to still be useful. The Sanders/Shaw model (as

represented by Sanders and McCormick) is included as Figure 1.2.

 

 

Physical Environment
Noise
Turpentine/Humidity
lllurnination

Work space dinensiom
Architectural aspects
Acceieration

Pollutants

Distractions

H

 

 

 

 

 

Management
Management
policies
Safety
orientation
Productive

Incentive
systens

Management
Decision
decentralization
Staffing
Resource
availability
Enployce
development
Coordination

Organmtioml
stnrcture

 

Equipment Design
Controls. displays
Conpatibility
Visibility

Guarding
Warnings

Electrical hazards
Mecinnical hazards
Themal hazards

Workplace arrangement

 

 

 

 

 

Work Itself

Pacing

Physical workload
Mental workload
Perceptual requirements
Motor skills requirenents
Repetitiveness/Boredom
Procedures/Methods
Shifi work

Work-Rest schedule

 

 

 

 

Social/Psychologcal

Environmern

Group norms
Organizational climte
Morale

Union rehtiom

lntergroup commnication

 

 

Worker/Coworker
Ability level
Alertness
Experience
Training
Persoml'ny

Physical capabilities
Age

Fatigue

Motivation
Alcohol/Drug
influence

illness

Intelligence
Ofilthe-job pressures
Job sat istaction

 

 

 

i—l

 

 

wmm>mZCJ

L

 

 

[WOH<>EU‘J

 

6%me

Figure 1.2 Model of Contributing Factors in Accident Causation (CFAC)
(Sanders and McCormick, 1993)

 

1.3 OHSA Accident Definition and Categories

The Occupational Safety and Health Administration (OSHA) defines an accident as
occurring “when a person or object receives an amount of energy or hazardous material
that cannot be safely absorbed” (OSHA, 2005). OSHA also offers Figure 1.3 as an
illustration of what it calls the three Accident Cause Levels: basic, indirect and direct.
The direct cause is the actual energy or hazardous material. Indirect causes usually result
from unsafe acts and/or conditions, and effect the direct cause. The basic causes are

factors such as poor management policies and decisions, or personal or environmental

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
     
 

factors.
Management Safety Policy Decisions
(BIQSISCES Personal Factors
Environmental Factors
INDIRECT Unsafe Act : Unsafe Condition
CAUSES <
DIRECT Unplanned Release of Energy
CAUSES and/or Hazardous Material

 

ACCIDENT

Personal Injury

Property Damage

 

 

 

Figure 1.3 Accident Cause Levels (OSHA Accident Investigation, 2005)

OSHA’s accident investigation guidelines specify procedures for accident investigators to
follow, and step 9a of these procedures requires the investigator to “determine why the
accident occurred.” This is a huge idea, encompassed in one little step.

In an attempt to clarify the OSHA accident categories, Hinze et a1. (1998)
investigated OSHA’s five basic cause categories: falls, struck-by, caught in/between,
electric shock and other. The objective was to create a coding system that would better
allow injuries to be grouped by the cause of the injury, and therefore allow for the
development of preventative strategies. A system was created that included 20 codes, and
proved to be a marked improvement over OSHA’s current system. Hinze et a1. (1998) felt
that further differentiation could be made, though better preventative measures could be

developed using the 20-code system.

1.4 Problem Statement

Although much progress has been made toward reducing the presence of hazards in
construction occupations, construction remains a high-risk industry. Reviewing the
information compiled by the Bureau of Labor Statistics, the construction indusz
accounts for a disproportionate number of the total yearly occupational fatalities, and the
rate of construction worker fatalities per 100,000 workers actually saw an increase from
2003 to 2004 (see Figure 1.1).

It seems apparent that the industry as a whole has begun to place safety in a
position of high priority, as evidenced by the decreasing trend in the rate of fatalities per
100,000 workers from 1992 to 2004. It is also apparent that the traditional safety
methods that brought this decrease to bear are not sufficient to eliminate construction

fatalities altogether. In light of the recent increase in fatalities, it is logical that research

should attempt to uncover new ways to prevent as many occupational accidents as
possible. There are many domains that have not been investigated, and it is only
beneficial to investigate as many as possible to preserve the lives and livelihood of the
construction industry workforce. This research has focused on one of these numerous
possibilities: the relationship and correlation between the personality and safety

awareness of construction ironworkers.

1.5 Goals and Objectives
The main goal of this research was to enhance the understanding of the factors that
influence the hazard identification of workers in construction situations. The following
objectives were proposed to realize this goal.
1. Develop a conceptual model to express the relationship between personality traits,
situations and the hazard recognition of the construction worker.
2. Investigate the relationship between personality traits and the hazard recognition

level of the construction worker.

It was and still is the author’s opinion that further understanding of the relationship
between workers’ personality and their sensitivity to hazards and hazardous situations
will go far in the ongoing quest to maximize the efficacy of safety training and accident
prevention measures, and provide a basis for further research in this area. The exploration
of these measures could prove to be quite valuable to the development of effective
managers and management strategies, the development, delivery and effectiveness of

training programs and beneficial introspection for workers themselves.

1.6 Scope of the Research

Ideally, the effect of the worker’s trade and type of work should be captured. This would
recognize that workers from different trades face very different occupational hazards, and
would attempt to identify and measure them- Though normalization across the industry
would be ideal, such normalization would be difficult, if not impossible. For this reason,
this study consulted a population of ironworkers, similar to the samples investigated by

Patel (2003) and Narang (2006).

1.7 Thesis Overview

This research has utilized the constructs of Signal Detection Theory and the Five Factor
Model to assess not only workers’ personality and their ability to identify a safe situation,
but to also produce two measurable data sets that could then be compared to show a
possible relationship between personality and safety awareness.

This thesis is comprised of five chapters. The first chapter offers a brief
introduction to the problem area and relevant research, as well as the motivation, goals,
objectives, and limitations of the proposed research. The second chapter gives
background information on Signal Detection Theory, the Five Factor Model and the
research application of personality in the arena of occupational safety. Chapter three
presents the methods proposed to achieve the goals and objectives presented in Chapter
One. Chapter Three also includes the expected results and contributions of this research.
Chapter Four presents the results of the survey and a discussion of the survey data
analysis. The fifth and final chapter offers the summary and conclusions drawn from the

research as well as suggestions for future research.

10

Chapter 2
BACKGROUND

11

2. BACKGROUND

In Chapter One, a brief overview of prior research concerning accident prevention was
presented, as well as the need for further research in this area. In this chapter, an
introduction to Signal Detection Theory (SDT) and the Five Factor Model (F FM) will be
given and the research concerning SDT and F FM individually will be presented. This

chapter concludes with a discussion of personality and safety research.

2.1 Accident Causation Research

As stated in section 1.3, accident causation has been a topic of extensive research, though
only a small segment has been devoted to the concept of ‘why’ accidents occur. This is a
very complex area, and it has been approached from very different angles. Various
researchers have attributed this ‘why’ to multiple factors, and the following is a brief
overview of some pertinent accident causation research, in addition to what was
discussed in section 1.2.

o McClay (1989): three elements of accident causes: the presence of hazards,
human actions, and the act of exceeding functional limitations (of humans or
equipment). McClay referred to this as a universal framework.

- Hinze (1996): high levels of distraction that a worker experiences on the job
increases the chance that an accident will occur. This distraction can be
attributed to various factors, similar to the stressors indicated in section 1.2
category two.

0 Reese and Eidson (1999): six factors can contribute to the occurrence of

construction accidents: actual physical hazards, environmental hazards, human

12

factors, lack of or poorly designed safety standards, failure to communicate
within and between trades.

0 Abdelhamid and Everett (2000): the ability of the worker to identify unsafe
conditions caused by management inadequacies or the worker’s own actions.

0 Suraji et a1. (2001): proximal and distal factors affect the way a worker responds

to on-the-job stimuli.

It must be stated, however, that the search for the causes of accidents is far fiom over,
and there are many sectors in which further research can take place. One such domain is
the relationship between situation and personality, which is discussed in depth in section
2.6 and 2.7, concerning the scholarship of Geller and Weigand (2006) and Cellar et a1.
(2001)

Rasmussen identified three work zones: the Safe Zone, where work can continue
normally with no undue hazard present; the Hazard Zone, where workers are “working at
the edge” (walking on thin ice); and the Loss of Control Zone, where accidents occur.
Rasmussen believed that the boundary between the Safe Zone and the Hazard Zone
should be identified to, and by, workers. Furthermore, he felt strongly that workers
should be trained in ways to remain in the Safe Zone, ways to recover from the Hazard
Zone back into the Safe Zone, and also how to identify, avoid, and ultimately recover
from the Loss of Control Zone. Figure 2.1 shows Rasmussen’s three zones of risk.
According to Rasmussen, workload and economic pressures can push the worker from

the Safe Zone toward the Hazard Zone, and eventually toward the Loss of Control Zone.

13

 

Boundary of
Unconditionally Safe

. Behavior
lrreversrble Loss of

Control Boundary

 
 
 
       
   
  
  

Organizational
Boundary to
Workload Economic Failure

Gradient

Hazard
Zone
Loss of
Control Zone

  

Cost
Gradient

 
 

Individual Boundary
to Unacceptable
Workload

 

\

Figure 2.1 Three Zones of Risk (Howell et al., 2002)

 

 

 

Figure 2.2 further illustrates the factors that can push a worker into and beyond the
Hazard Zone, and identifies the location of accidents. This figure shows that the pressure

to move from one Zone to another can be both internal and external.

14

 

  
     
   

      
   
 

   
   
 
  

   
 
    

 

Boundary to Functionally
Acceptable Behavror Boundary to
'gration Toward Frnancral
Least Effort Breakdown
Loss of
Control
Zone
Brownian movements
0 within space of proper
task performance
Local
Accidents
Management
The Boundary of Pressure Toward Boundary to
Safe Behavror as Etfrcienc Umcceptable
Defined by Safety Workload
Campaigns

 

 

 

Figure 2.2 Migration of Work Toward Loss of Control (Howell et al., 2002)

Based on the work of Rasmussen (1997), Howell et a1. (2002) proposed a new approach
to construction safety. This model recognizes that many factors affect the behaviors of
individuals where occupational safety is concerned. Howell et a1. (2002) stressed the
need to train workers to recognize hazardous situations, and suggested the following four
questions as tools to aid this recognition.

1. Where are you? (In what zone?)

2. What is the risk or hazard you now face?

3. What can be done to prevent the release of the hazard?

4. What can be done to reduce harm should the hazard be released?

15

With Rasmussen’s and Howell et al.’s emphasis on the worker’s ability to identify the
zones in mind, Patel (2003) employed a framework to assess how workers perceive safe
and unsafe situations. He used Signal Detection Theory, a method of decision-making
which is often used in manufacturing to detect defective products, to discern if the
workers could correctly identify these conditions. Patel found that 40 out of 42 subjects
had low to moderate sensitivity, suggesting that they might benefit from more safety
training. Patel proposed a few possibilities for future research, including similar analysis
of other populations, use of a larger population sample and the inclusion of other

variables such as archival injury history.

2.2 Accident Causation in Construction
2.2.1 Systems Model of Construction Accident Causation (Mitropoulos et al. 2005)
Based on the assumption that the current approach to construction safety is grounded in a
defensive strategy (barriers between workers and hazards), Mitropoulos et a1. chose to
investigate the work system factors that can contribute to hazardous situations and affect
worker behavior. This research identified some major limitations of the traditional
approach to safety; these limitations are as follows (Mitropoulus et al. 2005):
1. Reactive approach: Hazards should be avoided instead of managed or reacted to.
2. Conflict with production: Loss is still sustained, but it is loss of production, not
loss of health or life.
3. Uncertainty limits the effectiveness of defenses: The possible hazards in the field
of construction are innumerable, and it is not possible to plan for each and every

one.

16

4. Limited view of accident causation: The current view of accident causation
focuses on who is to blame for an accident, rather than interpreting safety as a
question of safe or unsafe situation, and that decisions on the job are often made
unconsciously and without an assessment of the possible hazard.

5. Limited learning: The traditional system of accident investigation focuses more

on who is liable than on a dynamic understanding of accident occurrence.

2.2.2 Incident Causation Model for Improving Feedback of Safety Knowledge
(Chua and Goh, 2004)
This scholarship addresses Safety Management Systems (SMS) on construction projects,
and specifically, the need to learn from SMS failure. Chua and Goh propose the Modified
Loss Causation Model (MLCM), which, if applied as a computer-based system and used
universally, could provide a uniform platform for incident investigation, data storage and
retrieval and future SMS planning. This model provides feedback in two distinct levels,
the first of which is feedback to the SMS system that failed, and the second is feedback to
the process through which future SMSs are created. This allows for continual

improvement in the area of construction safety.

2.2.3 Corporate Culture (Molenaar et al., 2002)

Molenaar, Brown, Caile and Smith compared the corporate culture of three of the safest
construction firms in the Denver area. The operant definition of ‘corporate culture’ used
in this study was “a pattern of basic assumptions invented, discovered or developed by a
given group as it learns to cope with its problems of external adaptation and internal

integration that has worked well enough to be valid and to be taught to new members as

17

the correct way to perceive, think and feel in relation to these problems” (Molenaar et al.,
2002)

The goal of the research was to assess upper management, middle management
and field personnel discrepancies with respect to company beliefs, values and behavior,
and to see how those discrepancies related to safety behavior (Molenaar et al., 2002). The
researchers developed a model, and created a survey based on the model to study several
characteristics of the company culture. They found that the company with the best safety
record of the three studied also had the most consistent safety culture, which was

identified as having the fewest culture discrepancies between company levels.

2.3 Signal Detection Theory

Signal Detection Theory (SDT) is a method used to assess the decision-making strategy
of a subject who must recognize and select a distinctly correct option. SDT has seen
extensive use in the manufacturing sector to identify and remove defective products.
SDT can be utilized to distinguish between two distinct ‘states of the world’ (Abdelhamid
et al., 2003), noise and signal. In a manufacturing setting, if the ‘state of the world’ is
signal (the product is defective), then the subject would chose either yes, the product is
defective (hit), or no, the product is not defective (miss). Similarly, if the ‘state of the
world' is noise, and the product is not defective, the response of the subject would either
be yes, the product is defective (false alarm) or no, the product is not defective (correct

rejection). Table 2.] further illustrates this decision matrix.

18

Table 2.1 The Four Outcomes of Signal Detection Theory

 

State of the World
SIGNAL NOISE
YES PHT FALSE ALARM

Response NO MISS CORRECT
REJECTION

 

 

 

 

 

 

 

 

 

Response bias, i.e., the tendency to give a certain response more often than another
response across multiple random trials, plays a significant role in the use of SDT. Some
subjects may tend to say yes more often than they say no, and consequently, have more
false alarms. Conversely, if the subject says no more often than yes, s/he would tend to
have fewer false alarms, but many misses. Each strategy has positives and negatives,
depending on the task in question, but neither is ideal.

In its basic form, Signal Detection Theory makes a few fundamental assumptions.
The first of which is that the value of d' (sensitivity) is not zero and is not negative. This
is because it assumes that a worker can discriminate, and that their discrimination is not
actively detrimental when compared to the ideal sensitivity. When observing a graphic
representation, such as presented in Figure 2.3, SDT assumes that the Signal and Noise
distributions never lie directly on top of each other, and that the Noise distribution is
always to the left of the Signal distribution. Another basic assumption is that the
variances of the distributions are uniform or at least similar. SDT allows for non-uniform

variances and for the value of d’ to be less than or equal to zero, but in these cases, either

the ratio of areas or the c criterion must be used to calculate B.

19

 

 

 

 

   
 

 

NO 1 Av YES
2 d' ' -

g Z1 Z2 ‘
6 Noise Signal

in

o \
a Correct H”
‘—' Re'ection
a. J .

4 '35.
3:22. P(X|N)
P<XIS> 555225;

 

 

 

 

 

 

 

vv‘vvv'vv

STIMULI
g Miss False Alarm

Figure 2.3 Distribution of Detection Theory (W ickens and Hollands, 2000)

As shown in Figure 2.3, Xc represents the point at which the subject makes a decision,
and indicates the subject’s response bias. The d' value is the subject’s sensitivity, or the
separation between the mean of the signal and noise distributions. With respect to
sensitivity, if the subject were completely unable to distinguish noise from signal, d'
would be 0. If the subject could infallibly distinguish noise from signal (perfect
discrimination), then, in theory, d' would be infinite. The value of d' is calculated by
adding 21 (the standard normal variable reflecting the possibility of a false alarm) and 22

(the standard normal variable reflecting the possibility of a hit). Both the values of Z1 and

20

Z2 can be found in standard statistical tables. Repeated trials indicate that average

sensitivity is reflected by a d' value equal to one (Wickens and Hollands, 2000).

d'=z.+ z2 (2.1)

For this study, the ideal value of d' was 4.6, if the worker were to have 11 hits and 7
correct rejections, and the absolute worst situation was 11 misses and 7 false alarms,

resulting in a d' value of — 4.6. The following table shows the range for low, moderate

 

 

 

and high sensitivity.
MODERATE
LOW
- 4.6 o .92 2.76 4.6
0% 50% 60% 30% 100%

Figure 2.4 Sensitivity Ranges for d' (Adapted from Patel, 2003)

The response strategy of an observer, whether they tend to say yes more often than no, or
vice versa, is represented by a value termed B. The value of [3 depends on the specific
location of X6 for a specific subject. This value of B is denoted as 13m! and represents
the decision criterion of the subject. At a given location of Xc, Bum,“ is the ratio of P(X|S)

to P(XIN), where P(X|S) represents the probability that Xc will be in a particular location

given signal and P(XIN) represents the probability that Xc will be in a particular location

21

given noise. Equation 2.1 shows the calculation of Bmt under the assumption that the
variances of the noise distribution and the signal distribution are the same and/or similar,

and that d’ is greater than zero. (Wickens and Hollands, 2000).
current: P(X|S)/P(XIN) (2.2)

Values of Bcumm relate to the number of misses and false alarms, such that a high
Bum.“ value indicates a high number of misses, and a low Beam“, indicates more false
alarms. Typically, mem is considered neutral when its value is equal to 1. Hence, a
13cm, greater than one(l) would indicate more misses than false alarms, and a Bowen, less
than one would indicate more false alarms than misses. In construction, misses are more
dangerous than false alarms, and a strategy that includes more misses than false alarms
will be referred to as ‘risky’ for this research. Likewise, a strategy that includes more
false alarms than misses will be termed ‘conservative’, though the traditional SDT
literature (Wickens and Hollands, 2000) uses the reverse of this terminology. ch can
also be used to characterize the strategy or bias of an individual subject by comparison to
[30,”, which is where [3 should be, given the situation the observer is under. [30,, is the ratio
of the probability of noise P(N) to the probability of signal P(S), the calculation for which

is shown in equation 2.2 (Wickens and Hollands, 2000).
130m: P(N)/P(S) (2-3)

If the value of cht is greater than the value of Bow, then the strategy of the

subject is more risky (Xc is farther to the right) because the subject tends to say no more

22

often than yes and the result is a high number of misses. If the value of Bean“: is less than
the value of Bopt, then the strategy of the subject is more conservative and X(; is positioned
farther to the left. This conservative strategy results in fewer misses and more false
alarms. Macmillan and Creelman (1991) suggest the use of the c criterion, included as
equation 2.3, to mitigate the effect of a very small or negative (1’ on the 13m. value, as
well as the effect of non-uniform variances. It is also suggested by Macmillan and

Creelman (1991) that the ratio of areas could be used as an effective measure of Bcumt.

c = -0.5[z(H) + z(FA)] (2.4)

The following example illustrates the use of SDT. This example is taken from
Patel’s 2003 thesis, and the SDT portion of this thesis research will closely parallel

Patel’s work.

Example:

A manufacturer produces DC motors using a process that has a 5% defective rate. After
receiving an increasing number of customer complaints, the manufacturer begins to
utilize an inspection system that detects 80% of the defective motors, but falsely rejects

good motors at a rate of 1%. What is the sensitivity of this inspection system?

Solution:

Table 2.1 can now be reconfigured, as shown below in Table 2.2, to reflect this example.

23

Table 2.2 Example Probabilities (Patel, 2003)

 

 

 

 

 

 

 

State of the World
SIGNAL NOISE
(Defective product) (Good product)
Response YES HIT = 80% FALSE ALARM = 1%
(Is the motor
defective? ) NO MISS = 20% CORRECT REJECTION = 99%

 

 

 

P(Noise) = P(Product is not defective) = .95 (given)
P(Signal) = P(Product is defective) = .05 (given)
P(Hit) = .80 (given)

P(Miss) = 1-P(Hit) = .20
P(False Alarm) = .01 (given)

P(Correct Rejection) = .99

SDT may be used as follows to determine the sensitivity and strategy of the inspection
process. To begin calculations concerning d' (sensitivity) the standard normal values of
Z1 and Z2 must be obtained. These can be obtained from standard statistical tables.

Using the probabilities of a false alarm and a miss, the Z values are as follows.

Z] = 2.326
22 = .842
From Figure 2.3, d' = Z, +22, therefore, d' = 2.326 + .842, d' = 3.168
As previously mentioned, average d' = 1, hence, the value of 3.168 = d' indicates

that the subject has a high sensitivity.

24

Looking back to equations 2.1 and 2.2, the calculation for [36mm requires the

determination of P(X|S) and P(XIN). However, according to Figure 2.2:

P(X|S) can be defined as the ordinate corresponding to Z2, and

P(XIN) can likewise be defined as the ordinate corresponding to 21
Using the table included in Appendix E:

P(X|S) [the ordinate corresponding to Z2] = .28

PO(|N) [the ordinate corresponding to Z.] = .027
Therefore, using Equation 2.1:

Bcumm = .28/.027 = 10.37

Bop. must be calculated to make a comparison to Bcumm, and this can be done using

Equation 2.2 as follows:
Bop, = .95/.05 = 19

Bum,“ is less than Bopt, indicating that the inspection process is risky (i.e. less misses and
more false alarms) according to standard use of the term in SDT literature. The Xc of the
process is positioned farther to the left. Consequently, the application of SDT to the
example of DC motor manufacturing has come to these conclusions: the process is more
sensitive than average to the difference between signal and noise, and thus, has a risky
strategy (in manufacturing), tending to indicate no more often than yes (each trial is

probabilistically independent of other trials) and resulting in more misses.

25

The above example demonstrates the use of SDT in assessing the ability of a process to
discriminate between two distinct ‘states of the world'. The specific ‘state of the world'
could be anything, such as “is the water wet or not”, “is the wine red or not”, and “is a
construction situation safe or not”.

It seems logical that SDT could be used to assess workers’ awareness of the safety
situations surrounding them in occupational settings, as well as the general strategies they
use when deciding if a situation is safe or not. The long-term use of such a tool could
prove useful as well: the same survey could be used initially to determine how sensitive
the workers currently are to jobsite situations and whether they could benefit from further
training. The strategies the workers employ can also indicate the type of training that may
be appropriate, and also suggest, to an extent, the possible learning and environment type
which should be used for each worker. The second use of SDT safety survey results can
be augmented by partnering it with the use of personality assessment tools, a discussion
of which is included in the next section. Furthermore, SDT could be used as a post-test
after trainings, both to indicate the effectiveness of the training, and, longitudinally, the
knowledge retention of the worker at certain intervals following the training. This last

assessment could be indicative of appropriate training frequency.

2.4 The Use of SDT in Construction

Patel (2003) utilized Signal Detection Theory to assess the occupational safety
competencies of 42 ironworkers. He conducted a survey based on standards for fall
protection containing hypothetical situations where the worker was asked to respond that
the situation was ‘safe’, ‘unsafe’ or ‘I don’t know’. It was found that approximately “95%

of the iron workers surveyed had a low to moderate sensitivity toward unsafe conditions”

26

(Patel, 2003). He also found that about half of the iron workers surveyed had more
misses than false alarms, which was considered a ‘risky’ decision making strategy.
Where age and years on the job were concerned, Patel’s regression analysis indicated that
sensitivity was not linearly correlated to age, and a moderate dependency between years
on the job and sensitivity was found, although no strong linear correlation between the

two was present.

2.5 The Five Factor Model

Borne on the shoulders of a German researcher named Baumgarten, Allport and Odbert
bought themselves a dictionary. In 1933, Baumgarten had released the fruits of many
years’ labor in the form an extensive list of descriptive German words that could be
construed to reflect the nebulous concept of ‘Personlichkeit’: personality.

Allport and Odbert poured over an unabridged English dictionary, seeking terms
that “distinguish the behavior of one human being from that of another” (Allport and
Odbert 1936). In the end, they collected more than 18,000 words. In an attempt to
organize their list, Allport and Odbert came up with four categories. They defined the
first category as “generalized and personalized determining tendencies - consistent and
stable modes of an individual’s adjustment to his environment” (Allport and Odbert,
1936), such as sociable, aggressive and fearful. The second category was temporary
states, activities and moods, such as afraid, rejoicing and elated. Category three consisted
of “highly evaluative syntax and judgments of personal conduct” (Allport and Odbert,
1936), such as excellent, worthy, average, irritating. The fourth category was created to

encompass all the heretofore-uncategorized terms, containing physical descriptions,

27

capabilities, talents; basically everything that could not easily be included in one of the
previous categories.

Norman (1967) took the terms collected by Allport and Odbert and re-categorized
them into seven groups: “biophysical traits, temporary states, activities, social roles,
social effects, evaluative terms, anatomical terms and physical terms”. In the process, he
eliminated ambiguous or obscure terms that he felt were not useful in describing
personality. Norman stated that a person could be described in a variety of ways:
enduring traits (irascibility), internal states (gleeful), physical traits (tall), activities
(jumping), effects on others (comforting), roles they play (parent) and the social
evaluation of their conduct (unacceptable).

Up to this point, the lexicon had been assigned to categories that the researchers
were assuming were mutually exclusive, but some researchers felt that the boundaries
between the categories were fuzzy, and that many of the descriptors could fit into more
than one category (Allen and Potkay, 1981). In a further effort to clarify the categories
for personality terms and develop a working taxonomy, Cattel (as cited in Pervis and
Johns, 1999), took a subset of the Allport/Odbert list, and began to develop his
multidimensional model of personality structure. Cattel eliminated nearly 99% of the
subset with which he began, and narrowed the field to a mere 35 variables. Pervis and
Johns (1999) indicate that Cattel’s decision to eliminate so many of the original terms
may have been because of the data analytic capabilities contemporary to Cattel’s
research: factor analysis of variable sets containing a large number of variables was
“prohibitively costly and complex” (Pervis and Johns, 1999). In 1970, Cattel and

colleagues released a 16 Personality Factor Questionnaire (16PF), and this quickly

28

became not only the object of much criticism, but also a readily available basis for further
research concerning viable, valid and reliable personality inventories and questionnaires.
One important direction for research stemming from the 16 factor model was the
inclination toward models with fewer, broader categories. One of the most accepted and
widely used personality factor models is the Five Factor Model. F iske (1949), Tupes and
Christal (1961), Norman (1963), Borgata (1964), Digrnan and Takemoto-Chock (1981),
and various others were intimately involved in the creation of the Five Factor structure,
as well as the independent reproduction of the structure based on Cattel’s categories.
Though there is some argument surrounding the actual names of the factors, such as
Extraversion sometimes referred to as Surgency, and Openness sometimes labeled
Intellect, the facets that the factors encompass are generally agreed upon by the experts in
the field. The Five Factors are as follows (each factor is succeeded by a few of its key
facets):
1. Extraversion (or Surgency): talkative, assertive, energetic
2. Agreeableness: good-natured, cooperative, trustful
3. Conscientiousness: orderly, responsible, dependable
4. Emotional Stability (vs. Neuroticism): calm, not prone to depression, not self-
conscious. [When Emotional Stability is reported as Neuroticism, any correlation
to this trait simply changes signs]

5. Openness (or Intellect): imaginative, independent-minded, receptive to new things

In 1981, Goldberg coined the phrase “The Big Five”, intending to reflect the broad,
encompassing nature of the factors themselves (Pervis and Johns, 1999). Goldberg felt

that the five traits represented personality on a global scale, the “broadest level of

29

abstraction” with each factor enveloping a large number of distinct and specific, though
related, personality characteristics (Pervis and Johns, 1999).

From this foundation, many inventories and questionnaires were developed. Of
these, there are three questionnaire types with appropriate validity, reliability and
acceptability to other researchers, which were considered for the personality assessment
segment of the proposed research. They are as follows:

1. The NEO F F I: Neuroticism, Extraversion and Openness Five Factor Inventory. In
general, this is the most commonly used of the three in a traditional questionnaire
setting. Originally published by Costa and Mche in 1992.

2. The TDA: Trait Descriptive Adjectives. This inventory is a list of single
adjectives that the survey responder indicates whether they believe each term
accurately describes themselves. Originally published by Goldberg in 1992.

3. '_f_he_BF_I_: Big Five Inventory. This inventory is laid out in short phrases, not the
full sentence structure of the NBC F F I, but longer and somewhat less ambiguous
than the single descriptors of the TDA. Originally published by John, Donahue

and Kettle in 1991.

The third form of the Five Factor Model was employed to assess the personality segment

of the research reported in this thesis. This is discussed further in Chapter 3.

2.6 People-Based Safety (Geller and Weigand, 2006)
E. Scott Geller, the Director of the Center for Applied Behavior Systems at Virginia Tech
in Blacksburg VA, was kind enough to grant the author of this thesis access to his most

recent scholarship concerning the relationship between occupational safety and

30

contemporary personality constructs, such as the Big Five. Dr. Geller has indicated that
though nothing definitive has been found in this area, much more research is needed,
warranted and encouraged. This paper is awaiting publication, and the information given
here is from the final draft version of the article, which was kindly supplied by Dr.
Geller.

In their article, Geller and Wiegand (2006) indicate the importance of three
distinct variables, and emphasize that these variables are interactive, dynamic and
reciprocal. The three domains identified by Geller and Wiegand are:

1. Environment: tools, equipment, climate of the workplace
2. Person: employee attitudes, beliefs, and personalities
3. Behavior: safe and at-risk work practices, as well as the willingness of another to

intervene for a coworker’s safety

Geller and Wiegand also recognize that situations and worker personalities are
interactive, an important basis for this research. The article states: “to achieve and
maintain an injury free workplace, employees need to address each of these domains
daily during the development, implementation, and evaluation of intervention strategies
to remove environmental hazards, decrease at-risk behaviors, increase safe behaviors, and
provide more user fiiendly or ergonomically sound work environments.” (Geller and
Weigand, 2006).

Geller and Weigand cite numerous studies which have evidenced that personality
factors influence safety-related behavior, many of which are also cited by Dr. Doug
Cellar and colleagues (2001), whose work is discussed in the next section. It seems that

the initial push to discover this relationship was in an attempt to define the injury-prone

31

personality. The methods used to this end were multiple and quite varied, and most of
the methods contained serious flaws, leading to inconsistent and ambiguous results
(Hadden, Suchman, & Klein, 1964; McKenna, 1983; Shaw & Sichel, 1971). Geller and
Weigand indicate that these flaws led to a significant amount of confusion about the
subject, and a low general opinion concerning the worthiness of the research. They
indicate, however, that the subject pops up about every ten or twelve years in the related
literature, and many contemporary researchers have continued to identify past failings
and miscommunications while calling for further study.

Geller and Weigand also ascertained that past safety researchers have focused on
the readily observable and reliably measured variables of environment and behavior
factors. It is important to note that these styles of study and subsequent management
strategies have not failed, and are the foundations of personality/safety research. In a
sense, these types of studies have pushed the American (and global) realm of
occupational safety systems to the current state. However, in light of the statistics
reported by the US. Bureau of Labor Statistics (see Figure 1.1) it seems apparent that the
number of occupational fatalities has reached a plateau, and has seen the first increase in
almost eleven years. This trend points to the fact that although the safety developments
of the past been successful, there is room and need for improvement. Geller and
Weigand term the concept “Low Hanging Fruit,” saying that past research in terms of
environment and behavior factors has been successfirl, and should, by all means, be
continued, but should not be taken as the bottom line of safety research. The two
researchers believe that one path for future safety research lies in the realm of

personality, and suggest the Five Factor Model as a most appropriate tool.

32

The article includes a discussion of personality traits themselves, and a detailed
discussion of the Big Five and the Five Factor Model. This information was covered in
the preceding section (2.5), and need not be repeated here.

Geller and Weigand (2006) also indicate that empirical research concerning
specific relationships between the Big Five and safety-related behavior is exceptionally
rare, and that they could only find one article that addressed these issues. The article is
“The Five Factor Model and Safety in the Workplace” (Cellar et al, 2001), which is
discussed in detail in the following section (2.7).

Personality trait researchers agree that traits are essentially immutable; people are
born with some degree of propensity toward certain personality characteristics (Ones and
Viswesvaran, 2002). Geller reiterates his theory that the environment, person and
behavior all play interactive and codependent roles in the expression of traits, and the
manifestation of traits can be manipulated through environmental conditions, behavior
intervention, and interpersonal dialogue.

The final statement of Geller and Weigand's article is as follows, and requires no
further explanation.

“The authors hope this article will serve two important functions: 1) to

increase awareness and understanding of the role personality can play in

both injury proneness and injury prevention, and 2) to stimulate the

systematic, empirical study of relationships between personality

predispositions and voluntary participation in eflorts to prevent

unintentional injury. ”

33

2.7 The Five Factor Model and Safety in the Workplace (Cellar et al., 2001)

In this study, Cellar and his colleagues administered the Revised NEO Personality
Inventory (N EO PI-R) to 202 undergraduate students. They also had the students
complete a self-report measure of personal workplace accident involvement histories. In
Cellar’s own words, “there have been few published studies that have examined the
relationships between personality and safety” and fewer still that have involved the Five
Factor Model as the personality perspective (Cellar et al., 2001).

Based on other research, Cellar et a1. (2001) felt that a more encompassing and
coherent representation of the subject could be extrapolated by focusing on safety
through the perspective of the Five Factor Model.

A few statistics on the population surveyed in this study are as follows:

0 Average age: 20.93 years, standard deviation: :t 4.96 years

0 Ethnic background: 60.4% Caucasian, 13.4% Hispanic, 12.4% African
American, 9.4% Asian American, 1.5% Native American, and 3% Other

0 69.3% held a job at the time of the study, though all participants had been
employed at some time

o The tenure of employment ranged from one to 25 years, averaging 4.9

years, standard deviation i 4.1 years

The version of the NBC PI-R that Cellar used consisted of 240 questions, judged
by the responder on a Likert scale of 1 — 5, with 1 indicating ‘strongly disagree’ and 5
being ‘strongly agree’. The respondents were asked to use this scale in rating the
accuracy with which the items in the NBC PI-R described the respondent personally.

Each participant also completed a personal safety behavior questionnaire, adapted from a

34

prior study by Arthur and Graziano in 1996 that assessed driving accidents. The NEO PI-
R creators, Costa and McCrae, reported their reliability (consistency with prior results)
levels as: Neuroticism = .92; Extraversion = .89; Openness = .87; Agreeableness = .86;
Conscientiousness = .90. These reliability levels from Cellar’s study were in line with
Costa and McCrae’s (Neuroticism = .90; Extraversion = .87; Openness = .89;
Agreeableness = .87; Conscientiousness = .90).

To complete the survey concerning workplace accidents, the participants were
asked to supply the number of workplace accidents where they were personally injured,
and categorize them as either ‘at fault,’ where the participant was at least partially
responsible for the occurrence of the accident, or ‘not at fault,’ where the participant was
not at all responsible for the occurrence of the accident. Though the participant must have
been personally injured for the occurrence to qualify as an accident, the injury did not
have to be significant, reported, or even reportable.

The participants were also asked to provide demographic information, including
age, gender, year in school, GPA, ethnic identification, and employment status.

The analyses of the data in this study included the examination of possible
relationships between each of the five factors, ‘at fault’ accidents, ‘not at fault’ accidents,
and the total number of accidents. In an effort to measure the ability of each factor to
predict accidents, namely, whether a particular participant might become involved in an
accident, additional regression analyses were performed. Cellar’s findings were as
follows:

0 Conscientiousness was found to have a statistically significant correlation to both

‘at fault’ (r = -.16, p < .05) and ‘not at fault’ (r = -.14, p < .05) accidents.

35

o Agreeableness was found to have a significant correlation to the total number of
accidents (the sum of ‘at fault’ and ‘not at fault’ accidents) (r = -.l3, p < .05).

0 When stepwise linear regression was performed, Conscientiousness alone proved
to be a good predictor of the number of ‘at fault’ accidents F(l,201) = 3.94, p <
.05 (R2 = .019) as well as the total number of ‘at fault’ and ‘not at fault’ accidents
F(l,201) = 5.02,p < .05 (R2 = .024).

0 When Agreeableness was included in the model to predict total work accidents, a
A R2 of .009 was found. The model also remained a statistically significant

predictor of total workplace accidents F(2,201) = 3.93, p < .05 (R2 = .033).

Based on these results, Cellar et al. concluded that Agreeableness alone did not account
for much incremental variance beyond that found when Agreeableness and
Conscientiousness were taken together. The article states that it should come as no
surprise that so much common variance was found between Agreeableness and
Conscientiousness, considering that the two factors themselves are highly related.

On the whole, Cellar et al. felt that their findings reflected a strong relationship
between personality factors and safe work practices and behaviors. They concede that
their sample possessed potentially limited work experience, as the mean age of the
participants was approximately twenty years and all were enrolled in at least one
university course. The article postulates that his findings concerning the
Conscientiousness and Agreeableness factors would likely be increased in strength when
using an older adult population. Cellar et a1. go on to say, “Through the extension of
safety research to include the Five Factor Model of personality, it is evident that

established personality inventories might be useful for predicting work-related accident

36

behavior by identifying broad behavioral patterns related to safety.” They indicate that
the identification of these patterns will assist in the development of training programs that
focus on altering certain behaviors, rather than trying to change the person (i.e., the
individual’s personality). It is also mentioned in the Cellar et al. paper that the use of
fixture safety/personality studies could lead to new selection practices, allowing for
employee selection based on the likelihood that the prospective employee will behave
safely in a work environment, and that using the results of the same types of studies could
lead to the development of more effective training programs to minimize the total number

of workplace accidents.

2.8 The MMPI and the MBTI
It should be noted that two very influential personality surveys have achieved certain
notoriety, but have had no mention in this thesis thus far. The first is the Minnesota
Multiphasic Personality Inventory, and the second is The Myers-Briggs Type Indicator.
In the 1930’s, a psychologist and a psychiatrist joined forces at the University of
Minnesota to develop the Minnesota Multiphasic Personality Inventory (MMPI). The
MMPI is the most frequently used clinical test, and is employed often in the legal system.
It is well researched and highly reliable, but it is very long, and very difficult to score. It
is also vulnerable to faking, because the intent of some items is very obvious (Karp &
Karp, 2000). The MMPI is a very in-depth analysis of an individual’s personality, and
contains 8 clinical scales and 15 content scales, all of which have individual scoring
strategies. Multiple versions of the MMPI are currently used, including those focusing
on adults, teens and children, and other differentiations. The MMPI has been in use for

more than 50 years, and has been subject to criticism that some of the items are “out of

37

date, sexist, awkward, or ambiguous,” additionally, “two items which contained religious
content specific to Christianity were found to be offensive to other religious sectors”
(Karp & Karp, 2000). These criticisms aside, the MMPI has seen tremendous use in the
academic and clinical environments, and is considered to be a staple of personality
research (Karp & Karp, 2000).

The Myers-Briggs Type Indicator (MBTI) is another well know personality
assessment tool. This is a model of personality development, and the four gradient
categories were termed ‘preferences’ by the MBTI creators, Isabel Briggs Myers and her
mother, Katharine Briggs (Personality Pathways, 2005). They began their own
adjustment of Jung’s model in the early 1940’s, and Myers-Briggs has been present in
personality lexicon ever since. The MBTI preferences are not traits, though they do
correlate to certain traits, and have been criticized by fellow psychologists as “too anti-
Freudian” (Pervin & Johns, 1992) to apply to the general population. The MBTI is based
on the work and theories of Dr. Carl Gustav Jung, a contemporary of Sigmund Freud and
a prominent champion and defender of Gestalt personality theory. The Myers-Briggs
Type Indicator has often been used as a personality assessment tool to aid in many work
situations, most notably, High Performance Work Systems. In this instance Myers-
Briggs was used as a measure to better allow members of a work team to mesh together
as a single entity for certain job purposes (Personality Pathways, 2005).

Though these two measures were not discussed as a part of this research, it would
be remiss to omit them completely fiom mention. These tests were both reviewed and
passed over for possible use as the personality assessment tool for this particular study for

various reasons: 1. The length of each respective questionnaire, 2. The depth to which the

38

results are indicated (multiple subcategories in the MMPI and gradient scales in the
MBTI), 3. The validity and reliability associated with each, and 4. The knowledge and

training required to accurately score the questionnaires.

2.9 Chapter Two Recap

This chapter presented a brief survey of the existing research concerning occupational
injuries and fatalities. It discussed the past safety research concerning accident causation
with specific emphasis on the work of Rasmussen and Howell et al., as well as the
hybridized research of Geller & Weigand (2006) and Cellar et al. (2001). This research
indicates that, though great strides have been made in the endeavor to minimize
occupational injuries and fatalities, the quest is far from over. Personality research and
accident causation research are both well-developed fields, however, they are not often
used in conjunction. This chapter also discusses the traditional use of Signal Detection
Theory, and how it is used to assess subject sensitivity and risk assessment (strategy).
Prior use of SDT in assessing sensitivity and strategy in construction workers was briefly
illustrated and further clarification of this subject will be presented in the following
chapter. This sample of the existing body of research on both subjects, independently
and jointly, suggests that further investigation of the relationship between personality and

safety is not only warranted, but necessary.

39

Chapter 3
METHODOLOGY

40

3. METHODOLOGY

Chapter one of this thesis offered a brief overview of prior research in the area of
accident prevention, and posited personality psychology as a prospective area for fixture
research into the causes of construction accidents. It also included the motivation, goals,
and objectives of this research. The second chapter gave background information on
Signal Detection Theory and discussed prior applicable use of SDT in the construction
industry. Chapter two also included historical information on the Big Five and the Five
Factor Model, as well as the research application of personality in the arena of
occupational safety. Chapter three presents the methods proposed to achieve the goals
and objectives stated in Chapter one.

As stated in Chapter one, the main goal of this research is to enhance the
understanding of the factors that influence construction workers’ ability to identify
hazards. The following sections break down the plan of action to attain this goal by

addressing each objective in turn.

3.1 Objective One

Develop a conceptual model to express the relationship between personality traits,
situations and the hazard recognition of the construction worker.

Proposed Method

Throughout the development of this research, multiple prior studies involving accident
causation and personality, both independently and jointly, were reviewed. These prior
research efforts have offered many models, and one model in particular lends itself quite

well to the proposed research. The model is an amalgamation of prior models, and was

41

developed by Sanders and Shaw and presented as Figure 1.2, in the form adapted by
Sanders and McCormick. The model developed to satisfy Objective One is in the same
spirit as the Sanders and Shaw model, though it was created with the specific intent to
illustrate the hypothetical connection between personality and situational awareness.

Discussion of the model and the model itself are included in Chapter 4.

3.2 Objective Two

Investigate the relationship between personality traits and the hazard recognition level of
the construction worker. This objective also involved an investigation of the relationship
of personality traits to the decision-making strategy of the worker. This objective was

twofold, and was addressed in appropriate steps, as follows.

3.2.1 Objective Two Part A
Develop and conduct a survey using Signal Detection Theory (SDT) to assess worker
safety awarenesses.
Proposed Method
At this point, it is appropriate to mention that if this objective had not been previously
executed in a similar fashion, it would be necessary to develop a technique to assess the
sensitivity and risk orientation of construction workers to unsafe conditions. However,
this research closely follows the methods used in Patel (2003), and Narang (2006). The
discussion that follows is a summary of this method, the purpose of which is to lay a
foundation for the above stated Objective Two Part A.

Patel (2003) conducted an assessment of construction worker sensitivity and risk

orientation using a modified version of Signal Detection Theory. In section 2.3, an

42

example concerning defective DC motor detection was given, and parallels were drawn
to the possible use of SDT in terms of construction safety. Patel used the terms ‘unsafe
condition’ and ‘safe condition’ to correspond with the SDT ‘states of the world,’ signal
and noise, respectively. A construction worker is faced with a hypothetical situation, and
must respond whether he/she believes the situation presented is either safe or unsafe. For
the purposes of the research conducted in the current study, the workers were asked to
respond whether they believe a situation is safe, unsafe or they are not sure. The matrix
presented in Table 3.1 reflects the following responses and is represented in Table 3.1:

0 gm: situation is unsafe; worker answers “Unsafe”

- Mis_s: situation is unsafe; worker answers “Safe”

0 M: situation is unsafe; worker answers “I am not sure”

0 False alarm: situation is safe; worker answers “Unsafe”

0 False alarm: situation is safe; worker answers “I am not sure”

0 Correct rejection: situation is safe; worker answers “Safe”

Table 3.1 SDT Matrix for Detection of Unsafe Conditions in Construction

(Adapted from Patel, 2003)

 

 

 

 

State of the World
SIGNAL NOISE
(UNSAFE condition) (SAFE condition)
RESIN“? YES HIT FALSE ALARM
(Is the condition
unsafe? ) NO MISS CORRECT REJECTION

 

 

 

 

 

 

43

As was discussed in chapter two, the SDT parameters of d' and Baum, can be respectively
used to assess subject sensitivity and risk assessment/response (strategy). Using the
adjusted SDT matrix in Table 3.1, d’ and Bow; can be used to assess the same
characteristics of construction workers. As in the example in chapter two, high values of
d' represent a significant ability to discriminate between safe and unsafe conditions, and
low values indicate a low ability to do so. The response bias (strategy) would normally
be indicated in the same way as in the chapter two example: the value of Burma compared
to the value of Bop, would indicate the worker’s strategy. However, a peculiarity in the
13cm, versus Born method was discovered, and will be discussed in greater detail at the
end of this section.

In Patel’s thesis, he indicated that the usual SDT application concerning strategy
would consider a worker with fewer false alarms than misses to employ a ‘risky strategy.’
He states, “In construction, the cost of a miss could be a fatality or serious injury.” The
author of this thesis is in complete agreement with Patel, and, as discussed in section 2.1,
addressed the situation by terrning the strategies as follows: the strategy in which the
worker has more false alarms will be ‘conservative’, and where more misses are present,
the strategy will be referred to as ‘risky.’ Additional discourse will be supplied when
appropriate to avoid semantic obfuscation.

In the current thesis, the sensitivity and strategy of construction workers has been
assessed using a self-report survey, because a live, real time study would not be feasible
for many reasons. The survey posed hypothetical situations to the worker and requested
one of three responses: “Safe”, “Unsafe”, or “I am not sure.” The corresponding SDT

assignments are shown in Table 3.1. The survey was modified from the one used by

44

Narang (2006), in that the portrayed situations were created using OSHA guidelines, and
each situation was specifically safe or unsafe according to OSHA. The survey is
comprised of 18 situations. Sample survey questions are as follows, and the survey in its
entirety is included as Appendix A:
o A crawler crane lifting steel columns, located at 30 feet from a 600 kV power
line.
0 When climbing a portable ladder to access an upper landing surface, the side rail
extends 3.5 feet above the upper landing surface.
0 While erecting steel beams using multiple lift rigging procedure, the steel

members are rigged at 6 feet apart.

The worker responses, hit, miss, false alarm, and correct rejection were converted to
probabilities to determine the sensitivity and risk orientation of the worker. The following
example is used to illustrate the SDT analysis, and is performed on the data obtained
from one of the workers who participated in the survey.

Table 3.2 Sample Survey Responses

 

 

 

 

 

 

 

State of the World
SIGNAL NOISE
(UNSAFE condition) (SAFE condition)
Response YES HIT = 8 FALSE ALARM = 6
(Is the condition
unsafe? ) NO MISS = 3 CORRECT REJECTION = 1

 

 

 

The following probabilities can be applied:

P(Noise) = P(Safe condition) = 7/ 18 =39%

45

P(Signal) = P(Unsafe condition) = 11/18 = 61%
P(Hit) = 8/11 = 73%

P(Miss) = l-P(Hit) = 3/11 = 27%

P(False Alarm) = 6/7 = 86%

P(Correct Rejection) = 1-P(False Alarm) = 1/7 = 14%

As outlined in section 2.3, d' (sensitivity) can be calculated as follows
Z] = -1.08 and Z2 = .61

Therefore, d' = -l .08 + .61 = -.47

This value of d' indicates that the worker has a low, in fact, negative, ability to
distinguish noise from signal; in other words, the worker’s sensitivity is actually worse
than not being able to distinguish noise from signal at all. This means that the basic SDT
equations for B cannot be used, because d' is negative. Referring again to section 2.3, the
course of action for this problem is to use the ratio of areas to produce the values of B.
For the sake of clarification, the B values in this research have been calculated using both
the corresponding ordinate method and the ratio of areas, and there were no discrepancies
in B values were found between the two methods. The ratio of the areas refers to the
areas under the curve representing miss and false alarm proportions. When using the ratio

of areas, the probability of a miss is divided by that of a false alarm.

Corresponding Ordinate Method for B
Ordinate corresponding to Z2 = .331
Ordinate corresponding to Z; = .223

current: .331/.223 = 1.484

46

B0131: P(Noise)/P(Signal) = 39/61 =.639

Bcumm is more than Bopt, which would make this a risky strategy. However, when
examined strictly by the definition of risky strategy, where more misses and fewer false
alarms are present, then it is safe to say that the Bum.“ versus Bop, comparison is not
correct, and that the procedure that should be used is indeed the ratio of areas. This is
because the worker had six false alarms and three misses, making him truly conservative.
When the ratio of areas is used, the point of comparison is one (1), because the B value
will be one when the pure count and probability of a false alarm is equal to that of miss,
with values greater than one indicating a risky strategy, and vice versa.

This particular idiosyncrasy of SDT was discovered during a pilot study for this
research. The survey that is included as Appendix A was given to three graduate students
in the Construction Management Program at Michigan State University to provide
preliminary data to the researcher. During the statistical analysis of the data from this
pilot study, it was found that the standard practices used to assign values for B were
creating a false assignment of strategy that did not match the pure counts of false alarms
and misses. This led to further research and discussion, and resulted in the discovery of
the method that has been used to analyze the actual data from the ironworkers in this

study.

3.2.2 Objective Two Part B
Conduct a version of the Five Factor Model (FF M) personality test with the same

population as surveyed in Objective Two Part A.

47

Proposed Method

As mentioned in section 2.3, the method selected to measure the personality segment of
this research was the Big Five Inventory (BF 1), an established questionnaire format of the
Five Factor Model. The reasons for the use of this model quite simple: the validity and
reliability is effectively identical for the TDA, the NBC F F I and the BFI. However, the
BF I uses a format that is more conducive to the study at hand, because it is not time
consuming and is relatively unambiguous, hence able to minimize confusion or apathy on
the part of the responder.

The personality questions that appear on the survey were selected fiom the 100
available on the International Personality Item Pool (IPIP, 2005) website. An equal
number of questions for each factor were included in the survey, which consisted of 30
items. It is appropriate to mention that some items in the BF 1 are scored in reverse
because the question is presented as a negative, and reversing the score is much less
confusing to the responder than asking the responder to reverse the location of ‘strongly
agree’ and ‘strongly disagree’ in the Likert scale. The following list contains examples of
BF I survey questions, and the personality section, which comprises the second half of the

actual survey that was given to the ironworkers in included in Appendix A.

I am the life of the party

I feel little concern for others

I am always prepared

I get stressed out easily

I have a rich vocabulary

48

As mentioned above, certain items are reverse scored because of negativity: essentially
the responder is asked how well each question applies to themselves, and occasionally a
question is asked that is the reverse of the factor being measured. For example, the
question is “I have little to say” or “I don’t talk a lot” and the response is “Very
Accurate.” In this case the responder shows very little of the factor being measured,
which is Extraversion, and would receive a score of one for that answer to either
question. Similarly, if the question is “I feel little concern for others” and the answer is
“Very Accurate” then the score is one, because the questions measures Agreeableness. A
person who says that feeling little concern for others is an accurate description of
themselves is probably not very agreeable.

The decision was made concerning this survey to drop the number of questions
from the usual 50 or 100 to thirty for several reasons. The first is that no validity is lost
even when the test has only 10 questions, or two per factor, (IPIP, 2005). Additionally,
though the safety portion of the survey is only 18 questions, it represents nine pages of
the entire survey. According to the IPIP information (2005) the need for 50 or more
questions is arbitrary in terms of validity and reliability of the survey, and participant
fatigue and apathy are increasing factors, respective to the length (real or perceived) of
the survey as a whole.

Similar to the scale used to assign groups via the range of sensitivities for SDT,

the following table can be constructed to categorize the Five Factor results.

49

HIGH

MODERATE

LOW

 

 

 

 

 

0% 50% 60% 80% 100%

Figure 3.1 Ranges for the Five Factors

3.3 Additional Demographic Data

In addition to the data collected via SDT and BF 1, certain demographic information,
including age, gender, years of experience and prior accident occurrence (accidents on
the job resulting in hospital visits or emergency medical care) were included in separate
analyses to both the SDT and the F F M results. These factors were regressed against the
SDT and FFM measures, and the results of those analyses are discussed in chapters four
and five. It should be noted that though there was no restriction on gender (or any other

variable), all of the participants in the survey were male.

3.4 Methodology Summary

This chapter outlined the procedures that were used to conduct this research, and
presented the modifications made to the survey tools. It discussed the changes in the
analysis of the SDT measure of B, due to the presence of negative (1’ values. This change

was further indicated by a pilot study done on three Master’s students at Michigan State

50

University before the survey was given to the ironworkers consulted in this study. A
complete statistical analysis was performed on the results of the pilot study, and the
limitations of using cht and Bop, became apparent. The decision was then made to use
the ratio of areas method suggested by Macmillan and Creelman (1991).

The reverse scoring for ‘negative’ questions in the personality section of the
survey was discussed, as was the decision to reduce the number of questions in this
section. This chapter also indicated that certain demographic information was collected

and regressed in the analysis.

51

Chapter 4
SURVEY RESULTS AND DATA ANALYSIS

52

4. Survey Results and Data Analysis

This chapter presents the survey data and its analysis according to the procedure outlined
in the previous chapter. As explained in chapter one, construction is an inherently
hazardous occupation, and this research was intended to broaden the scope of previuos
safety research by examining the relationship between personality factors and hazardous

situation recognition data as measured by Signal Detection Theory methods.

4.1 Objective One: The Model
As stated in chapter three, objective one was to develop a conceptual model to express
the relationship between personality traits, jobsite situations and the hazard recognition of
the construction worker. Figures 4.1 and 4.2 illustrate the model, with Figure 4.1
representing the basic model, and Figure 4.2 demonstrating a higher degree of detail.
While pursuing a degree in psychology some years ago, the author of this thesis
was enrolled in several educational psychology classes. These classes were specific to
the education of students (children and adults) with mental illness and disabilities. The
idea of the ‘ABCs of Behavior’ has been a long-standing tenet of behavioral psychology,
upon which special education psychology classes rely heavily. This idea helps a person
to understand his/her own behavior, as well as to provide insight into the behavior of
others. The common sense version of this Antecedent — Behavior —Consequence theory
is that everyone does what they do for a reason, and people do not do something if they
get nothing from it. It must be kept in mind, however, that the ‘something’ a person gets
from an action may not be something positive, but could be the removal of something

negative. In certain cases, it can be said that negative attention is more desirable to the

53

subject than no attention at all, if attention in general was the goal of the behavior. The

basic model developed for this thesis and the firlfillment of Objective One is as follows.

 

Behaviors can become or effect antecedents

 

 

 

T’l No Accident]

CONSEQUENCE

 

 

 

 

 

 

 

 

7
n...

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Consequences can become or effect antecedents

 

 

 

Figure 4.1 The ABCs of Accidents

Essentially, the model suggests that action (behavior) is preceded by a reason for that
action (antecedent), and once the action is carried out, it is followed by one or many
outcomes (consequences). For example: two children at day care would like to play with
the same truck. Child A has the truck, and Child B desires it. Child B now has a reason
or antecedent for his behavior: he wants the toy. His behavior, stemming from this
antecedent, will have consequences. The action Child B opts for could be to simply take
the truck from the other child. As any parent knows, most children of day care age (and
older) do not consider the consequences of their action before they do something. In this
case, the consequence might be that Child A cries when Child B takes the truck, and

maybe an adult would come take the truck back from Child B.

54

At this point the consequence becomes an antecedent, and Child B now can
choose to act in a variety of ways. Child B can cry, try to reclaim the truck, move on to
other toys, etc. Each level, each behavior, is an ever-blossoming, multi-petaled flower
that continues to grow and change. It is assumed and expected that the consequences of
actions will teach a lesson, and they do, but it may not always be the lesson that was
intended. For instance, Child B may choose to scream and cry when an adult takes from
him the truck he took from Child A, and the adult may give Child B a piece of candy.
This essentially becomes positive reinforcement for the child to scream and cry, because
the message that he received was “When I cry I get candy” as opposed to, “I should not
take someone else’s toys.” Admittedly, this example is elementary at best, but it should
help to clarify the complex interaction of the ABCs.

The ABCs of accidents work in essentially the same way as the example with the
children did, with one small change. The costs are much higher when an accident occurs.
Ideally, a worker in a potentially hazardous occupation should be aware of and able to
identify any possible hazards that may arise. However, this is not a perfect world, and
accidents are thus named because they are not expected, and generally were not foreseen.
It seems logical that a worker who looks ahead to possible consequences may be better
prepared for an accident, and/or better able to avoid one.

The expanded ABCs of accidents begin to explain the details of this research, in
that it expands upon the set of antecedents that are specifically at issue in this thesis. As
noted in the example with the children and their toy, consequences can become
antecedents, but it is important to also note that antecedents can and do work together,

and effect each other. In Figure 4.2, the two main antecedents that are being investigated

55

are personality and situation, and the bold bubble contains many examples of things that
can effect personality, or situation, or both. It is essential to understand that the defining
factor in this research is the worker: i.e. the measurements of personality and safety were
taken specifically through the filter of perception, namely that of the worker. Various
limitations inherent to this type of research make it incredibly difficult and unreliable,
and in some cases unethical, to measure personality and/or hazard recognition in real-
time, real-life studies. In light of this, the entire data set for this research was obtained

through self-report surveys.

 

  
  
 

 
    
 

Company climate Physical State

. . . Disability
Job Sat'sfaction -
Knowledg
_ General Cognition
actors outsrde the w

worker’s cornrol _
Weather/s rte
cond itions
Co- worker/s rpervisor
relations

      
 

 
     

      
 

 

 
 

 

 

 

Figure 4.2 The Expanded ABCs of Accidents

4.2 Data Collection

Thirty workers participated in the study by completing the developed survey, and no
questions were left unanswered. The sample size was reasonably large to allow the use
of the normal distribution to analyze the data. No restriction on age, gender, years of

experience, or any other criteria was imposed on the ironworkers who volunteered to

56

complete the survey. The aggregate survey is included as Appendix A. All workers who
participated in the survey did so voluntarily, and were not compensated or coerced in any
way. Each ironworker had received some form of safety training from their respective
companies, and it was reasonable to expect that the situations hypothesized in the survey
were within their scope of experience and training. The average age of the workers who
participated was 43 years and the average years of experience was 21.

The safety section of the survey addressed situations in which the worker must
identify whether the situation was safe or unsafe. The survey was adapted from the one
developed and implemented by Narang (2006), and presented 18 situations that an
ironworker would encounter during steel erection. The 18 safety questions were
comprised of absolutely safe conditions, absolutely unsafe conditions which presented a
significant personal risk, and situations that were considered to be unsafe by OHSA
standards, but carried a slightly lower personal risk. The workers were asked to choose
one of three responses during the SDT safety section, “Safe”, “Unsafe” or “I am not
sure.” The SDT coded responses (hit, miss, false alarm and correct rejection) are
included in Appendix B.

Figures 4.3 and 4.4 show the number of miss and false alarm responses to each
question. A miss occurs when the worker responds to an unsafe situation with a “Safe”
or “I am not sure” answer. A false alarm occurs when the worker responds to a safe
situation with an “Unsafe” or “I am not sure” answer. The two questions with the most

misses and the most false alarms, as well as the two questions with the least misses and

the least false alarms discussed below.

57

Question #7: Bolting steel members in place on the 3'“ floor, when the temporary
bracing is n_ot in place on the lower floors. 15 misses. Half of the ironworkers scored
misses on this question. This number of misses could mean that the situation is
ambiguous or unclear in some way, or that the workers do not consider the risk to be
significant. The workers may also simply not know if this is safe.

Question #3: While erecting steel beams using a multiple lift rigging procedure,
the steel members are rigged 6 feet apart. 14 misses. This question could represent an
intermittent risk, in that OSHA says it is not safe, but a worker may have to be six feet
tall or taller to even consider this a risk.

Question #9: An opening of 16 inches by 10 inches is left uncovered next to a
column. The concrete planks arrecast) are bearing (sitting) on the beam that connects to
this column. 27 false alarms. OSHA does not consider this to be a violation, though the
workers surveyed in this thesis overwhelmingly felt that this was not safe. This could
point to a possible need for revision in the OSHA standards.

Question #15: Ironworker climbs on the steel beam, when it is held by the crane
(tied with a choker), to bolt it in place. 21 false alarms. This is not an OSHA violation,
though 70% of the ironworkers surveyed felt that this was not safe. Again, possible
revisions may be indicated.

Question #13: Removing the fall protection while transferring fiom a beam to the
hoisting crane. 0 misses. Every single ironworker in this study recognized this as an
unsafe situation. In the author’s opinion, this is the most obvious unsafe situation in the
survey, and it is encouraging to know that the ironworkers surveyed were all aware of the

risk involved.

58

Question #14: A column is bolted in place with 3 anchor rods while beams are
being connected on the 2"“ floor. I miss. As was question #13, this was an obviously
unsafe condition, and the ironworkers overwhelmingly recognized this.

Question #2: When climbing a portable ladder to access an upper landing
surface, the side rail extends 3.5 feet above the upper landing surface. 6 false alarms. It
is possible that those who scored false alarms on this question simply were not aware of
the OSHA requirement for ladder overlap, but it is interesting to note that those who said
this was not safe were among the most conservative strategists in the sample.

Question #11: Operating a forinft on the 4'h floor when all perimeter cabling is
in place and precast concrete panels are being placed on the 6'hfloor. 7 false alarms. As
was the case with question #2, the workers who scored false alarms on this question were
much more conservative than those who did not. This may also be a somewhat

ambiguous question: there is not mention of what activity might be underway on the 5'11

floor of the building.
Questions with MISS Answers
16
l 14 : :- 73 ~ 1’ . ,, W7”
1 12‘ l l . . 1V. 7,
l 10 . A ‘ l ; 7 ' 7 fi
8 ’ i i i TW 1 Ti ’7 J L [Vi—l
‘ 6 l l I 1 l I l L, . a, r
‘6 1 . ! i? F I i I
2 . l l l J . ,Ll i i, i
. l 1 l i 1 1 1L 1, l 1...; 1 1 T1 ‘ L l
l 0
1 3 4 5 6 7 8 12 13 14 16

Question |

r _L _ 7 7 ,i i L __— ,7 fi.‘

Figure 4.3 Unsafe Situation with “Safe” or “I am not sure” Responses

59

201,,,_ l 1» ~ if ~77 fl»? #1,,

15 z. e

# of FALSE ALARMS

Figure 4.4 Safe Situation with “Unsafe” or “I am not sure” Responses

The personality section of the survey consisted of 30 questions that are further broken
down into five categories. Each category, Extraversion (EX), Agreeableness (A),
Conscientiousness (C), Emotional Stability (ES) and Openness (O) was represented by
six questions, and the total possible score for each category was 30, with values for each
question on a Likert scale of 1-5, as discussed in chapter three. The results of this section

of the survey are presented in Appendix C.

4.3 SDT Analysis: Sensitivity and Strategy

The SDT safety section of the survey contained 18 scenarios, 7 of which were safe and
11 were unsafe. If the worker correctly identified an unsafe condition as “Unsafe” then
the SDT representation was a hit (H). If the worker identified the unsafe situation as
“Safe” or “I am not sure” then the response was coded as a miss (M). If the situation was

safe and the worker correctly identified it as such, then the response was denoted as a

60

correct rejection (CR). A false alarm (FA) resulted when the situation was safe and the
worker identified it as “Unsafe” or “I am not sure.”

After the SDT status of the response was assigned (H, M, CR, FA), then the
respective probabilities were calculated. These calculations were performed on the data
for all 30 ironworkers to arrive at the SDT analysis, which is given in Table 4.1 along
with the calculations for the F FM data. Twenty-six of the 30 ironworkers surveyed were
conservative in strategy, 18 had low ability to discriminate between safe and unsafe
conditions, and 12 had moderate ability to discriminate. As mentioned in chapter two, the
ideal value of d' for this study was 4.6, if the worker were to have 11 hits and 7 correct
rejections, and the absolute worst situation was 11 misses and 7 false alarms, resulting in
a d' value of — 4.6. A d' value of 0, halfway between the ideal (best) value of 4.6 and the
worst value of — 4.6, is 50%. As illustrated in Figure 2.4, d' values between - 4.6 and
0.92 (60%) are classified as low, values between 0.92 and 2.76 (80%) are moderate, and
values greater than 2.76 are high sensitivity. In terms of B (strategy), values below one
are considered to be conservative, values above one are risky, with one representing an
absolutely neutral strategy, where the number of misses and false alarms are equal. The
average sensitivity (d') was 0.71 with a standard deviation of :1: 0.57 and the average

strategy (B) was 0.68 with a standard deviation of i 0.61.

4.4 FFM Analysis
As previously discussed, the F F M data was collected in the second section of the survey
by means of 30 questions to which the participant responded according to a choices

ranging from “Very Accurate” to “Very Inaccurate”. The complete data set for all 30

61

ironworkers is shown in Table 4.1, containing the SDT and F FM measures as well as the
collected demographic and historical data.

The scoring was performed according to a Likert scale of 1-5, 1 being “Very
Accurate” for questions that were not posed in the negative sense, and 5 being “Very
Accurate” for those questions that were posed in the negative sense. After the surveys
were scored the total score for each category was tallied. The total possible score for any
category was 30, because there were 6 questions for each category and the maximum

score for each question was 5.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.1 Summary of Data
Worker Age Exp. Acc. EX A C ES 0 d' B_ Strategyb Sensitivityd
1 42 20 0 23 26 25 16 24 1.16 0.16 ConservativelModerate
2 26 2 0 19 28 27 18 26 0.53 0.84 ConservativelLow
3 52 30 1 20 20 24 24 19 0.91 1.24 Risky Low
4 37 20 0 19 22 20 22 29 1.52 0.21 ConservativelModerate
5 49 30 0 20 20 24 24 19 1.17 0.93 ConservativeIModerate
6 50 31 0 9 24 2219 16 1.16 0.16 ConservativeIModerate
7 24 2 0 25 30 26 20 21 0.79 0.63 ConservativelLow
6 46 25 0 15 21 1617 16 1.16 0.16 ConservativeIModerate
9 48 20 0 15 22 23 28 17 0.53 0.84 ConservativelLow
10 49 32 0 15 21 22 16 21 .047 0.31 ConservativelLow
11 33 11 0 19 22 25 21 21 1.17 0.93 ConservativeIModerate
12 46 29 0 15 16 1617 12 0.44 0.47 ConservativeJLow
13 42 14 0 13 11 1516 16 095 0.52 ConservativelLow
14 26 6 0 19 22 1819 27 1.21 3.21 Risky [Moderate
15 49 29 0 21 24 21 19 22 0.79 0.13 ConservativelLow
16 50 26 1 21 22 27 21 20 1.17 0.93 ConservativelModerate
17 47 29 1 23 23 25 23 26 1.16 0.16 ConservativelModerate
16 39 16 0 20 19 21 16 16 0.68 1.55 Risky Low
19 42 19 0 23 26 25 16 23 1.52 0.21 ConservativelModerate
20 26 5 0 20 27 2716 26 0.00 1.33 Risky Low
21 43 20 0 21 21 2514 20 0.06 0.38 Conservative[Low
22 49 21 0 15 21 1817 16 0.74 0.32 ConservativeILow
23 52 33 1 15 21 2216 21 0.06 0.36 ConservativeILow
24 44 21 0 21 21 2514 20 0.44 0.47 ConservativelLow
25 43 20 0 23 26 27 18 23 1.09 0.42 ConservativeIModerate
26 52 30 1 20 20 25 24 16 0.53 0.84 ConservativelLow
27 46 20 0 16 23 24 27 16 0.53 0.64 ConservativelLow
26 31 11 0 16 23 24 22 22 0.79 0.63 ConservativelLow
29 52 30 1 21 19 23 25 18 1.17 0.93 ConservativelModerate
30 49 29 0 22 25 22 20 23 0.36 0.25 ConservativelLow
mg 43 21 0.2 19 22 23 20 21 0.71 0.66 ConservativelLow

 

 

 

 

 

 

 

 

 

 

 

 

 

 

62

The average score and standard deviations for the FFM factors were as follows:

Extraversion: sample mean: 18.87, sample standard deviation: :t 3.60
Agreeableness, sample mean: 22.27, sample standard deviation: i 3.54
Conscientiousness, sample mean: 22.87, sample standard deviation: :1: 3.29
Emotional Stability, sample mean: 19.97, sample standard deviation: i 3.44

Openness, sample mean: 20.67, sample standard deviation: :t 3.82

According to Figure 3.1, the high range for all five factors is 80% or more (a score of 24
or higher), the moderate range is 60% to 80% (a score of 18-23), and the low range is
anything below 60% (18 or less); the sample average for all five personality factors fell in

the moderate range.

4.5 Regression Analysis

Regression analysis was performed to determine the relationship between the workers’
safety knowledge and their personalities. Regressions were performed individually to
compare age, years of experience, and number of accidents to sensitivity and strategy.
Regressions were also individually performed to compare the FF M measures to the SDT
measures. The R2 and p-values for each regression model constructed are recorded in
Table 4.2, and the entire regression and scatter plots for each regression are included as

Appendix D.

63

Table 4.2 Regression R, R2 and p-values

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

variables R R2 (ti/3L P value
11’ and Age 0.316 0.1 0.884
d' and Exp 0.447 0.2 0.797
d' and Acc 1.095 1.2 0.057
d' and EX 3.507 12.3 0.572
d' and A 3.847 14.8 0.036
d' and C 2.280 5.2 0.226
d' and ES 2.665 7.1 0.155
d' and O 2.280 5.2 0.226
B and Age 3.962 15.7 0.030
B and Exp 3.834 14.7 0.036
B and Acc 0.548 0.3 0.770
B and EX 0.775 0.6 0.692
B and A 0.894 0.8 0.692
B and C 0.447 0.2 0.826
B and ES 1.817 3.3 0.340
B and O 1.095 1.2 0.558

 

As indicated in Table 4.2, the only variable combinations that had a p-value less than .05
were d' and Agreeableness, B and Age and B and Experience. The R2 values for these
items are low, but the fact that the p—value is less than (1 indicates that the finding itself is

significant, and there is a linear association between the two.

4.6 Grouped Analysis (SDT)

Considering the low levels of correlation and high p-values resulting from the regression
analysis, further analysis was conducted. The workers were split into groups according
to their strategy, risky v. conservative, and also according to their level of sensitivity: i.e.,
low, moderate or high. This further analysis was attempted to determine whether any

patterns or trends existed within particular sub-groups.

Table 4.3 shows the results of this grouped analysis. The table shows the average age,
experience, accidents, Extraversion, Agreeableness, Conscientiousness, Emotional
Stability, Openness and d' (sensitivity), in relation to B (strategy).

As shown below in Table 4.3, no difference between either strategy group and the
average of the 30 workers as a whole was greater than the respective standard deviation
for that variable. For example, the difference between the risky group and the average of
all workers for the variable Openness was 1.39; still within the standard deviation for the
entire sample, which was 3.82. This is further illustrated in Figures 4.5 and 4.6.

Likewise, the difference between the conservative group and the risky group across
all variables was less than the standard deviation for the respective variables. The
average for the risky group was lower than the average for the conservative group on all
variables with the exception of accidents, Extraversion and Openness. In other words,
the risky group was on average, younger, less experienced, less agreeable, less
conscientious, less stable emotionally, and had a lower sensitivity. However, the

differences in the averages for all variables between the two groups were too miniscule to

be considered definitive in any way.

Table 4.3 Average Variable Value Differences by Strategy Groups

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

B Age Exp. Acc. EX A C ES 0 d'
av all 43.07 21.23 0.20 18.87 22.27 22.87 19.97 20.67 0.71
tidy 36.25 15.25 0.25 19.75 22.00 19.75 19.75 22.00 0.70
a_vg cons 44.12 22.15 0.19 18.73 22.31 22.92 20.00 20.46 0.72
diff risky 6.82 5.98 0.05 0.88 0.27 3.12 0.22 1.33 0.01
diff cons 1.05 0.92 0.01 0.14 0.04 0.06 0.03 0.21 0.01
risky - consl -7.87 -6.90 0.06 1.02 -0.31 -3.17 -0.25 1.54 -0.02
[81. dev all i 8.67 9.14 0.41 3.60 3.54 3.29 3.44 3.82 0.57

 

65

 

 

30.00

25.00 1
g, 20.00 iuavg all 1
3 15.00 "Eavgiflylj

10.00

5.00

 

 

  

 

0.00 .
A C ES 0 d'

All Variables

Figure 4.5 Risky Group Averages Compared to Overall Averages (B)

35.00

30.00 -

 

25.00

 

20.00 -

 

15.00

Averages

10.00 >

5.00

 

 

 

0.00

 

All Variables ‘

Figure 4.6 Conservative Group Averages Compared to Overall Averages (B)

The same type of breakdown by groups was performed with respect to sensitivity, and as

in Table 4.3, Table 4.4 shows the average age, experience, accidents, Extraversion,

Agreeableness, Conscientiousness, Emotional Stability, Openness and B (strategy), but

66

related them to the sensitivity groups of low and moderate, as there were no results in the

highly sensitive range. Figures 4.7 and 4.8 show the variable averages by group.

Table 4.4 Average Variable Value Differences by Sensitivity Groups

 

 

45.00 ,- ..
40.00
35.00 ~
30.00 ,
25.00
20.00

15.00
1 10.00
I 5.00

 

 

M ‘Dafl alli
_‘ llavglow

Averages

 

 

0.00

 

Figure 4.7 Risky Group Averages Compared to Overall Averages (d')

67

1 50.00
1 45.00 LL #4.. #4. A_, is.
1 40.00

35.00 -.
30.00 ,
25.00 —
20.00
15.00
10.00
5.00
0.00 9

 

  
 

:Davg all
llavg modJ

 

Averages

 

 

Age Exp. Acc. EX A C ES 0 b
All Variables

Figure 4.8 Conservative Group Averages Compared to Overall Averages (d')

Similar to the results of grouping by strategy, no difference between either sensitivity
group and the average of the 30 workers as a whole was greater than the respective
standard deviation for that variable. For example, those with low sensitivity averaged
18.39 for the measure Extraversion, while the Extraversion average for the entire sample
was 18.87, the difference being well within the 3.6 standard deviation.

Additionally, the difference between the conservative group and the risky group
across all variables was less than the standard deviation for the respective variables. The
average for the low sensitivity group was less than the average for the moderate
sensitivity group on all variables. The average for the low sensitivity group was also less
than the average of the 30 workers as a whole for all variables. Again, as was the case
when grouped by strategy, the differences in the averages for all variables between the

two groups were too miniscule to be considered definitive in any way.

68

4.7 Grouped Analysis (F FM)

When considering the Five Factor variables, the maximum possible score on any variable

was 30, and the minimum possible score was 5. A scale was developed to allow

groupings to be made within the Five Factors, much as it was for the SDT measures. The

scale for the Five Factors was represented in Figure 3.2, and according to this scale, the

following table was constructed to assign ranges to the ironworkers’ F FM scores.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.5 Five Factor Ranges
c 3 Range

WorkerEX A C ES 0 d' B EX A C ES 0
1 23 26 25 18 24 1.16 0.16 Moderate High H'gh Low Moderate
2 19 28 27 18 26 0.53 0.84 Moderate High High Low igh
3 20 20 24 24 19 0.91 1.24 Moderate Moderate Moderate Moderate Moderate
4 19 22 20 22 29 1.52 0.21 Moderate Moderate Moderate Moderate High
5 20 20 24 24 19 1.17 0.93 Moderate Moderate Moderate Moderate Moderate
6 9 24 22 19 16 1.16 0.16 Low Moderate Moderate Moderate Low
7 25 30 26 20 21 0.79 0.63 High High High Moderate Moderate
8 15 21 18 1718 1.16 0.16 Low Moderate Low Low Low
9 15 22 23 28 17 0.53 0.84 Low Moderate Moderate High Low
10 15 21 22 18 21 -0.47 0.31 Low Moderate Moderate Low Moderate
11 19 22 25 21 21 1.17 0.93 Moderate Moderate High Moderate Moderate
12 15 18 16 17 12 0.44 0.47 Low Low Low Low Low
13 13 11 15 18 18 -0.95 0.52 Low Low Low Low Low
14 19 22 18 19 27 1.21 3.21 Moderate Moderate Low Moderate High
15 21 24 21 19 22 0.79 0.13 Moderate Moderate Moderate Moderate Moderate
16 21 22 27 21 20 1.17 0.93 Moderate Moderate High Moderate Moderate
17 23 23 25 23 26 1.16 0.16 Moderate Moderate High Moderate High ‘
18 20 19 21 18 16 0.68 1.55 Moderate Moderate Moderate Low Low
19 23 26 25 18 23 1.52 0.21 Moderate High High Low Moderate
20 20 27 27 18 26 0.00 1.33 Moderate High High Low High
21 21 21 25 14 20 0.06 0.38 Moderate Moderate High Low Moderate
22 15 21 18 17 18 0.74 0.32 Low Moderate Low Low Low
23 15 21 22 18 21 0.06 0.38 Low Moderate Moderate Low Moderate
24 21 21 25 14 20 0.44 0.47 Moderate Moderate fiqh Low Moderate
25 23 26 27 18 23 1.09 0.42 Moderate High High Low Moderate
26 20 20 25 24 18 0.53 0.84 Moderate Moderate High Moderate Low
27 16 23 24 27 16 0.53 0.84 Low Moderate Moderate High Low
28 18 23 24 22 22 0.79 0.63 Low Moderate Moderate Moderate Moderate
29 21 19 23 25 18 1.17 0.93 Moderate Moderate Moderate High Low
30 22 25 22 20 23 0.36 0.25 Moderate HigL Moderate Moderate Moderate

avg. all 19 22 23 20 21

69

(Figures in Appendix G show the distribution of low, moderate and high scores for each

of the Five Factors.)

As was performed with the SDT results, group analysis was executed with the FFM
results. The results are presented below in summary tables and figures, followed by a
brief explanation.

Table 4.6 d' and the Five Factors

EX

 

 

Low FFM vs. Sensitivity

 

I d' avg all
I d' avg low

Sensitivity

 

 

 

 

 

F ive Factors

 

 

 

Figure 4.9 Low Five Factor Scores Compared to Sensitivity (d')

70

 

1 Moderate FFM vs. Sensitivity

 

   

 

 

1
1
1
£17599“ 11
[ld'avg mod ;
’ '7' 1

1

ex A c as o 1

. 7 7 .7- 7 ..777 J

Sensitivity
o
a:
o

 

1.00
0.90
1 0.60
0.70
0.60 -
0.50
0.40
0.30 «
0.20 .
0.10
0.00

 

 

1 '5' avg all
, 1l d' avg high

 

Sensitivity

 

17 _ 7 _7 . 7 7

Figure 4.11 High Five Factor Scores Compared to Sensitivity (d')

71

Table 4.7 B and the Five Factors

EX A

 

1 Low FFM vs. Strategy
1
1 1.20 7 , 7- .77 7 , » , ~ » 1 1

1

1 ‘nbavgali H
1 leavglowj‘

 

 

 

 

  
 

 

Strategy
p
8
.1

 

 

 

 

 

1
1
F... 1...... [J1

Figure 4.12 Low Five Factor Scores Compared to Strategy (B)

72

Mod F FM vs. Strategy

 

 

1 0.60 7
0.70 7-
0.60 ~4

§ 0.50 7
E
m

0.40 ~—

 

 
  

1
1
1 1
1 _ .
1 1 11:1b7avg all—Wk
1 ; ab aw9M1
1 0.30 ~~ ‘

0.20 —

0.10 7
‘ 0.00 —

 

 

 

 

 

 

ES 0

 

1 7 ,7 7 777 7 77—7 ,7 ____1

Figure 4.13 Moderate Five Factor Scores Compared to Strategy (B)

 

1 High FFM vs. Strategy 1

 

 

  

 

[lubavgall ‘
j 1Ibavghigh1

 

 

C ES 0 1
Five Factors

Figure 4.14 High Five Factor Scores Compared to Strategy (B)

73

For both Tables 4.6 and 4.7, the first row represents the average d' or B for the entire

sample, regardless of personality factor. The second, third and fourth rows are the

average for those workers who scored low, moderate and high, respectively, on each

factor in the columns, according to the ranges shown in Figure 4.2. The fifth, sixth and

seventh rows indicate the difference between the corresponding mean in rows two, three

and four, respectively, and the first column. The eighth column is the low range mean

minus the high range mean, and the ninth column is the standard deviation of the entire

sample, regardless of personality.

Many of the values are within one standard deviation of their respective means,

but there are a few notable exceptions. These exceptions are as follows:

The average d' (sensitivity) of workers who scored less than 18 on Agreeableness
is nearly two standard deviations below the sample mean.

The difference between the mean d' (sensitivity) of those workers who scored in
the low range of Agreeableness and those who scored in the high range is almost
two standard deviations.

The d' (sensitivity) average for the low range was below the sample mean for all
five factors.

The d' (sensitivity) average for the high range was above the sample mean for all
five factors.

The low range B (strategy) categories for all factors excluding Conscientiousness
were below the sample mean.

The average B (strategy) of workers who scored in the low range of

Conscientiousness was above the sample mean for B.

74

o The average B (strategy) of workers who scored in the high range of Openness
was above the sample mean for B. (It is important to note that all differences in
mean for B are within one standard deviation of the mean. Mean B: .68 and

standard deviation B: i .61.

4.8 Principle Component Analysis

A Principle Component Analysis (PCA) was conducted on the B and d' values to
examine their interaction, primarily to determine if the two variables are independent, as
would be expected when using SDT. The PCA returned Eigen Values of 1.07 for d' and
.93 for B, with proportions of .535 and .465 respectively, indicating that d' and B are

indeed independent. The Eigen Vector Matrix is shown in Table 4.8.

Table 4.8 PCA Eigen Vector Matrix

 

Prin1 Prin2

 

dprime 0.707107 0.707107

 

 

 

 

 

Beta 0.707107 -O.70711

 

4.9 Multiple Regressions

Multiple regressions were also performed on the data obtained from the survey. Two
separate multiple regression were executed, one with d' versus age, experience, accident
count, Extraversion, Agreeableness, Conscientiousness, Emotional Stability and
Openness and the other with B versus the above mentioned variables. The complete

details of both regressions are included in Appendix F.

75

4.9.1 d' versus Age, Exp., Acc., EX, A, C, ES and O
The initial multiple regression of d' versus age, experience, accidents, Extraversion,
Agreeableness, Conscientiousness, Emotional Stability and Openness returned the

following regression equation:

d' = - 1.45 - 0.0212 Age + 0.0295 Exp. + 0.041 Acc. + 0.0514 EX+ 0.0820A - 0.0567 C

+ 0.0526 ES - 0.0051 0, with R2 = .35, R = .59, and p = .252.

At this point, the variables with p-values greater than .05 were removed fi'om the
analysis, leaving experience, Extraversion, Agreeableness, Conscientiousness and
Emotional Stability to compare to d'. These variables returned the following regression

equation:

d' = - 2.17 + 0.0129 Exp. + 0.0540 EX+ 0.0847A - 0.0575 C + 0.0510 ES with R2 =

.34, R = .59, andp = .06.

A second reduction was made based on p-values, and the variables with p-values greater
than .15 were removed from the analysis, leaving Extraversion, Agreeableness and
Emotional Stability in the regression. These three regressed against d' yielded the

following regression equation:
61’ = - 1.88 + 0.0336EX+ 0.0469A + 0.0456 ES with R2 = .26, R = .50, andp = .048.

It would appear that the second step regression equation is the most meaningful: though

the p—value is lower for the third regression equation, the R2 is not so significantly lower

76

as to justify the decrease in the R and R2 values. Also, the p-value was acceptable in the

second step regression.

4.9.2 B versus Age, Exp., Acc., EX, A, C, ES and O
The initial multiple regression of B versus age, experience, accidents, Extraversion,
Agreeableness, Conscientiousness, Emotional Stability and Openness returned the

following regression equation:

,8 = 3.85 - 0.0566Age + 0.0026 Exp. + 0.411 Acc. + 0.0156 EX- 0.0429A - 0.0299 C +

0.0449 ES- 0.0205 0, with R2 = .39, R = .62, andp = .171.

The variables with p-values greater than .05 were removed from the analysis, leaving
Age, Accidents, Agreeableness, and Emotional Stability to compare to B. These variables

returned the following regression equation:

,6 = 3.30 - 0.0502 Age + 0.324 Acc. - 0.0621 A + 0.0430 ES, withR2 = .37, R = .61, and

p=.017.

A second reduction was made, and the variables with p—values greater than .15 were
removed from the analysis, leaving Age and Agreeableness in the regression. These two

regressed against B yielded the following regression equation:
0 = 3.67 - 0.0390 Age — 0.0590 A, with R2 = .25, R = .50, andp = .021.

As was the case with the multiple regressions for d', it appears that the second step

regression equation is the most meaningful: though the p-value is lower for the third

77

regression equation, the R2 is enough lower to justify the decrease in the R and R2 values.

The p-value was also acceptable in the second step regression.

4.10 Discussion

With regard to the comparison by groups, sections 4.6 and 4.7, a few trends within this
sample can be asserted. Agreeableness, which was one of the factors with a significant p-
value when regressed against d'(sensitivity), indicated a deviation from the sample mean.
This was especially true for workers who scored in the low range of deviation, suggesting
that low Agreeableness is positively related to low sensitivity. This is also true for those
workers who scored low on all the five factors, though the trend is slightly weaker when
all five factors are considered. Additionally, the trend of workers who scored high on all
five factors scoring above average d' reinforces this tenuous relationship.

When considering strategy, Conscientiousness seemed to be the trend-defining
factor, though one worker had a very high B, and when this data point is removed, the
strategy of those who scored low on Conscientiousness tends to be lower, in fact below
one (conservative). Those workers who scored in the high range of Openness had a
slightly higher than average B, indicating a slightly riskier strategy.

These trends indicate a few notable concepts. The first is that the more Agreeable
a worker was, the more discrimination they showed. It is possible that the workers who
are less Agreeable care less about what others thinks or feel, and do not take safety
training to heart. It is also possible that the more Agreeable they are, the more willing
they are to please someone else by following safety rules on a jobsite, or by actively

trying to keep their work environment and work practices safe.

78

Concerning Openness, the same ideas can be postulated: perhaps the workers who
scored high in Openness are more willing to look at a new way to do things on the job,
instead of thinking that the old way is the only way, and no one should tell them how to
do their job. Though these trends are inconclusive and can be applied only to this
sample, these results merit further research into the relationship between safety awareness
or hazard recognition and personality.

The multiple regressions yielded the most useful information, indicating that the
variables with the strongest relationship to d' were experience, Extraversion,
Agreeableness, Conscientiousness and Emotional Stability when taken together. The

following regression equation describes the relationship between these variables and d':

d' = - 2.17 + 0.0129 Exp. + 0.0540EX+ 0.0847A - 0.0575 C + 0.0510ESwith

R2 = .34, R = .59, andp = .06.

This essentially means that the variables of Experience, Extraversion, Agreeableness,
Conscientiousness and Emotional Stability account for 59 % of the difference from the
mean for the measure of d'(sensitivity) when taken together. With this regression result,
the null hypothesis of Ho: R2 = 0 (there is no relationship) can be rejected in favor of the
alternative hypothesis Ha: R2 96 0 (there is a relationship).

The multiple regressions also indicated that the variables with the strongest
relationship to B were age, accidents, Agreeableness, and Emotional Stability when taken
together. The following regression equation describes the relationship between these

variables and B:

79

,0 = 3.30 - 0.0502 Age + 0.324 Acc. - 0.0621 A + 0.0430 ES, with R2 = .37, R =

.61, andp = .017.

In other words, this regression equation says that the variables of age, accidents,

Agreeableness, and Emotional Stability account for 61% of the difference from the mean

for the measure B (strategy) when taken together. With this regression result, the null

hypothesis of Ho: R2 = 0 (there is no relationship) can be rejected in favor of the

alternative hypothesis Ha: R2 i 0 (there is a relationship).

This research, like any investigation, has various inherent limitations, such as:

Both survey tools rely on self-reported data, which has been hotly debated
in the past (Winter et al., 1998 and Pervin and John, 1992). Personality
traits are observable, but the issue of observer bias casts a significant
shadow on this form of personality assessment. Safety behaviors could be
directly observed, and though observer bias would also be a factor, the
main issue here would be the obligation of the observer to intervene if s/he
were to anticipate an accident.

Only the traits that are represented by the Five Factors were measured.
This may not be the most appropriate tool for assessment, and this is a
topic for future investigation. Additionally, the various facets of the five
factors may prove to be useful in this type of research.

It is rarely the case where one isolated signal is the ‘state of the world.’ In
any situation, a person is required to make a series of judgments and
decisions that dictate the course of action that they ultimately choose.

These split second decisions are often unconscious, and difficult to

80

measure. SDT distills the process of a decision down to a choice that is
essentially yes or no, and allows for no other factors to be present in the
particular situations that were included in the survey for this research.
Fuzzy Signal Detection Theory attempts to address this difficulty.

The bulk of the population that responded to the survey was fiom the local
ironworkers union, and it is very possible that they represent a
homogenous segment of the total population. The world is not a closed
shop, and further research should look to diversify the sample population.
There is an inherent difference in what people say they would do, and the
actions they actually take. This could come to light as a defensive posture,
where a worker knows that the situation is unsafe, and answers that it is
so, but may choose not to act on that knowledge in real life situations. This
is actually a form of response bias, where the participant may answer in
the way they think they should, for whatever reason, and not in the way
they truly feel. Some individuals may feel that the risk of injury or death
is so minimal that they do not need to take safety precautions; for
instance, few people follow the safety guidelines for using a lawn mower
which include wearing long pants, boots, gloves and safety goggles. Of
course, there are always the select few who choose to trim their hedges
with a lawn mower. A partial solution may be to use Fuzzy SDT where
each scenario has three questions, answered on a continuum ranging from
“Very” to “Not at All” (Narang 2006):

1. How safe is this situation?

81

2. How certain are you about your answer?
3. How comfortable would you be working in this situation?
0 The sample size was very small: only 30 ironworkers were surveyed.
Ideally, a large sample should be obtained. The survey was also

administered only once, and should be subject to repetition.

4.11 Results and Analysis Summary
In this chapter, the survey data was analyzed using Signal Detection Theory, and the Five
Factor Model, to determine the hazard sensitivity and strategy of the ironworkers, as well
as their personalities. The results were further analyzed, using simple linear regression
analysis, comparison by groups and multiple regressions.

The objectives stated in chapter one were achieved using the methodology
outlined in chapter three and described in detail in this chapter. Chapter five discusses
the results and contributions, and concludes with a discussion of the research limitations

and suggestions for future research.

82

Chapter 5
SUMMARY AND CONCLUSION

83

5. Summary and Conclusions

5.1 Thesis Summary

The primary goal of this research was to enhance the understanding of the factors that
influence the hazard identification of workers in construction situations. The chapters
preceding this one presented the details of the background, concepts, existing literature,
prior research, and approach that this scholarship followed as well as the methods by
which the results were obtained and analyzed.

The first chapter outlined the problem area, and addressed the motivation for this
research. Chapter one also presented the motivation and objectives. Chapter two
discussed Accident Causation research, and introduced Signal Detection Theory and the
Five Factor Model. The third chapter discussed the methods used to achieve the
objectives of this research. Chapter four reported the data collected during the course of
this study, and the respective analysis. This chapter addresses the research findings in

further detail.

5.2 Conclusions

This thesis expanded on an approach to measure a worker’s ability to distinguish between
a safe condition and an unsafe condition on a jobsite. It also measured the worker’s
decision-making strategy. Furthermore, it compared the results of a five factor
personality assessment to the results obtained concerning workers’ discrimination and

strategy.

84

Analysis of the survey data indicated that experience, Extraversion,
Agreeableness, Conscientiousness and Emotional Stability were related to d' when
considered in a multiple regression relation (R2 = .34, R = .59, and p = .06). Likewise,
multiple regressions also indicated that the variables with the strongest relationship to B
were, respectively, age, accidents, Agreeableness, and Emotional Stability (R2 = .37, R =
.61, and p = .017). For both measures d' and B the null hypothesis of Ho: R2 = 0 (there is
no relationship) was rejected in favor of the alternative hypothesis Ha: R2 76 0 (there is a
relationship).

With this in mind, the average sensitivity of the workers was low, at 0.71 with a
standard deviation of i 0.57: 18 workers were below 0.92 and 12 were above. The
average strategy of the workers was conservative: only four out of the 30 surveyed were
risky (B > 1). The average strategy (B) of the 30 workers was 0.68, again with a large
standard deviation of i 0.61.

A distinct idiosyncrasy concerning the strategy calculation in Signal Detection
Theory was experienced during the pilot for this study. The result of this irregularity was
an inaccurate assignment of strategies: a worker with a conservative strategy could be
classified as risky if his/her sensitivity was very poor. In other words, if the d' value was
less than zero, the Bum,“ calculations were incorrect. In fact, the d' value could be zero,
below zero, or a small positive value approaching zero. It seems that the closer the value
gets to zero, the more distortion is present in the Bcumm calculations. Investigation into
existing SDT theory and research revealed solutions to this problem, one of which (the
ratio of areas) was used in this study. According to the basics of Signal Detection, the

theory assumes that subjects actually have the ability to discriminate, meaning that the

85

value of d’ (sensitivity) is not zero and is not negative. SDT also assumes that the
variances of the distributions are uniform or at least similar, and in this study, the
variances were not only unknown, there was no reason to assume that they were similar
at all.

As mentioned in the Discussion section of the previous chapter, the results from
the grouping analysis shown in sections 4.6 and 4.7 indicate that the risky group was, on
average, younger, less experienced, less agreeable, less conscientious, less stable
emotionally, and had a lower sensitivity. The average for the low sensitivity group was
less than the average for the moderate sensitivity group on all variables. Though the
grouping analysis was useful to determine trends within certain ranges of sensitivity and
strategy, the differences in the averages for all variables between the groups were too
miniscule to be considered definitive.

In the author’s opinion, the most useful information that resulted from this study
was that the majority of the workers were conservative in terms of strategy, meaning that
when they were unsure whether a situation was safe or not, they were more inclined to
view a safe situation as unsafe, rather viewing an unsafe situation as safe. The second
viewpoint, which epitomizes the risky strategy, is not desirable in a construction worker,
as it could easily lead to more accidents on the job site because workers were assuming
an unsafe situation to be safe. The author believes that it is more difficult to change
whether a person cares about a situation i.e., will the worker treat an unsafe situation
differently from a safe one, than it is to heighten the workers awareness of the dangers
any given situation may present. To be unaware of a hazard is entirely different than how

a person reacts to a hazard. To an extent, a conservative strategy can compensate

86

somewhat for a lack of knowledge; if a worker tends to view the work environment as
less safe, they may actually react to an unsafe situation in a very cautious manner, even if
they do not specifically know that the situation is not safe. Some experts would say that
erring on the side of caution is the route to take, and a worker with a conservative
strategy is essentially doing just that. To the author, it is comforting to know that such a
high number of workers in this study had a conservative strategy. However, it is
somewhat alarming that the workers’ average sensitivity was so low. This is a problem
that can be more easily addressed through training; a worker who cannot discriminate
between safe and unsafe situations can always benefit from simply being told that certain
things are not safe. For example, question #3 in the attached survey shows a tree rigging
of multiple steel members. The caption states that the members are rigged six feet apart;
however, OSHA requires that the members be rigged seven feet apart. Technically, this
situation is unsafe, though many workers may view it as safe, because a worker would
have to be six feet tall to be struck by a member while standing on level with the member
below. An answer of “Safe” to this question could be remedied through training.

The following section (5.3) makes various recommendations for future research
which would relate to studies done in the spirit of this thesis; looking for the relationship
between personality and safety awareness. However, it must be said that knowing what to
do with the relationship once it is established equally important. Parallel research into
adult learning styles, and training techniques based on learning styles and personalities
has been and should continue to be investigated. Knowing the personality ‘type’ of the
person who is to receive training can and often does effect the way trainings are

organized and executed. Within this psychological research, possible variables include

87

occupation, age, personality (of trainer and trainee), region of work, cultural background,
marital status and years of experience, and all of the intricate relationships between the
above variables. It is important to note that not all of the relationships between these
variables are known, and the complex interactions between variables work in ways that
have not even begun to be understood. Educational, Industrial/Organizational and various
other psychological disciplines have studied training in general, as well as specific
training aspects. The author’s recommendation in this continued research is to work very
closely with colleagues in other areas, as cooperation can often lead to quicker advances

in research and understanding.

5.3 Recommendations for Future Research

As mentioned in chapter four, future research should consider using Fuzzy SDT with a
third question, as well as a larger, more diverse population. This diversity should take
into account geographic regions, union/non-union, gender, country of origin, and size of
sample, among other aspects. Investigation into a way to observe workers in real-time, or
to minimize the disparity between self-report and action, should be considered. Further
research into locating the threshold of distortion concerning the standard calculation for
strategy should also be considered.

It is possible that the Five Factor Model is not the ideal research tool for assessing
personality. Other inventories and assessments should be considered as the psychological
measure in relation to safety. This is not to say that F F M should be discarded; only that
research into the appropriate tool should be as all-inclusive as possible.

This particular methodology may lend itself well to other trades within the

construction industry, and other industries themselves. The survey that was used in this

88

research was specific to ironworkers with a significant emphasis on fall hazards, and the
development and use of similar surveys in the course of further research would certainly

be a benefit to future understanding of accidents in general.

5.4 Contributions of this Research

As in any research, there always exists the possibility that there is no statistically
significant correlation between two variables. Though a correlation was found, it is by no
means a definitive result. With this in mind, this particular study attempted to fill a void
in the research on the subject, and any lack of a significant correlation in this study would
not and does not preclude the need for similar future research. This work has resulted in
the following contributions:

0 A statistically significant result indicating that a relationship does exist between
the variables studied (age, experience, accidents, Extraversion, Agreeableness,
Conscientiousness, Emotional Stability and Openness) and the two Signal
Detection Theory measures (d' and B).

o A revisitation to a technique designed to determine the sensitivity and risk
orientation of construction workers.

0 A survey combining safety and personality assessment specific to the construction
industry and ironworkers in particular.

I A platform from which to launch future research. The blind ‘stab in the dar ” has
been made, and future research can easily learn from, build on and replicate this
work.

0 The introduction of a significant social science issue into a largely engineering

field.

89

As the author of this thesis, I believe its greatest contribution is the fact that it was done,
that this journey was taken. To my knowledge, it is the first research of its kind with
respect to industry and topic, and will hopefully become but one of many studies and

scholarship in this area.

“Success is a journey, not a destination. ‘Ihe doing
is qften more important than the outcome”

firthurflshe (1943 — 1993 )

90

Appendix A
CONSENT LETTER AND SURVEY QUESTIONNAIRE

91

Participant Consent Form

OCCUPATIONAL SAFETY and PERSONALITY
Principal Investigator: Tariq S. Abdelhamid, PhD
Research Assistant: Suzie von Bemuth

The Construction Management Program at Michigan State University is conducting a
research project to assess the occupational safety knowledge and personality of
ironworkers. The research will help in improving the effectiveness of safety training
programs. You are being asked to participate in this project in your capacity as a
construction Ironworker.

As a participant in this research, you will be asked to complete an 18 question survey on
occupational safety rules related to fall protection, as well as the 30 question personality
survey.

Your assistance is voluntary and you may choose to terminate your involvement in this
study at any time during this project. If you are uncomfortable answering any part of the
survey, you may leave those sections unanswered. Your privacy will be protected to the
maximum extent allowable by law. Neither you nor your company will be identified by
name. The estimated time for the survey is 30-45 minutes. As a participant, you may
request a copy of this consent letter for your records.

If you have any questions about this project, you may contact Dr. Tariq Abdelhamid,
Construction Management Program, Michigan State University at (517) 432-6188, or Dr.
Mohammad Najafi, Construction Management Program, Michigan State University at
(517) 432-4937. If you have questions or concerns about your rights as a research
participant, please feel free to contact Peter Vasilenko, Ph.D., Director of the Human
Subject Protection Programs at Michigan State University: (517) 355-2180, fax: (517)
432-4503, email: irb@msu.edu, or regular mail: 202 Olds Hall, East Lansing, MI 48824.

I voluntarily agree to participate in this study.

 

 

 

Subject Name Occupation Signature Date
Witness Name Occupation Signature Date
[RE # 06-202

Category: Expedited
Approval Date: 3/27/2006
Expiration Date: 3/26/2007

 

 

 

92

SURVEY QUESTIONNAIRE

MSU Member: Suzie von Bemuth

MICHIGAN STATE UNIVERSITY

 

Occupational Safety and Personality Assessment

Date:

 

Construction Industry Experience (In Years):

 

Gender (Circle one): M F

Age:

Have you ever had a work related accident that required emergency medical care or a trip
to the hospital? (Circle one) Y N

If yes, how many?

 

SAFETY QUESTIONS‘

Please read the following scenarios and select your answer fiom the choices: “Safe, ”
“Unsafe, or “I am not sure. ” Please circle only one answer per question. The pictures
are intended as visual aids only, please answer the question using the written information
given. Please disregard other possible hazards present in the picture if they are not
specifically addressed in writing.

Please proceed to the questions on the next page.

There will be similar instructions at the beginning of the second segment of the survey.

93

1. A crawler crane lifting steel columns, located 30 feet from a 600 kV power line.

 

Is this situation (the distance between the crane and the power line)
A. Safe

B. Unsafe

C. 1 am not sure

2. When climbing a portable ladder to access an upper landing surface, the side rail

extends 3.5 feet above the upper landing surface.

"S‘.

   

.(“

   

Is this situation (the height that the ladder’s side rail extends above the
landing surface)

A. Safe

B. Unsafe

C. I am not sure

94

3. While erecting steel beams using a multiple lift rigging procedure, the steel
members are rigged 6 feet apart.

urn-l,

    

x
\

Is this situation (the distance between the rigged members)
A. Safe

B. Unsafe

C. I am not sure

4. Erecting 7"I floor columns, when all the 5‘” floor planks (precast concrete panels)
are ggt completely in place.

 

Is this situation (5th floor planking is incomplete during 7th floor
column erection)

A. Safe

B. Unsafe

C. I am not sure

95

5. Climbing a portable ladder that is set 1
foot out for every 5 feet climb (as shown in
figure).

Is this situation (the horizontal distance
compared to the vertical distance) 5

A. Safe

B. Unsafe

C. I am not sure

 

 

. An ironworker connecting 4"I floor beams, with an unprotected edge with no
decking in place on the lower floors while wearing a fall arrest harness that is

921 tied off.

  

Is this situation (the type of fall protection and the manner in which it
is used

A. Safe

B. Unsafe

C. I am not sure

96

7. Bolting steel members in place on the 3rd floor, when the temporary bracing is
n_ot in place on the lower floors.

  
   

Is this situation (temporary bracing is not in place)
A. Safe

B. Unsafe

C. I am not sure

8. A 40 ft. long beam being attached as a
part of a multi-rig assembly is bolted
with a single bolt at each end and the
choker (cable) is released. The steel

beam is detailed into place much later.

Is this situation (# of bolts)
A. Safe

B. Unsafe

C. I am not sure

 

97

9. An opening of 16 inches by 10 inches is left uncovered next to a column. The
concrete planks (precast) are bearing (sitting) on the beam that connects to this
column.

 

Is this situation (the hole is not covered)
A. Safe

B. Unsafe

C. I am not sure

10. Working on the 3"d floor of a building where the top of the perimeter cabling is

at 35” from the floor and the intermediate cable is 16” from the floor.

W.

  
 
 
 
 
  

Is this situation (the distance
fiom the top and intermediate
cables to the deck surface)

A. Safe

B. Unsafe

C. I am not sure

 

 

 

 

 

 

98

11. Operating a forklift on the 4"l floor when all perimeter cabling is in place and

precast concrete panels are being placed on the 6"I floor.

  
  

 

 

Is this situation (fork on the 4th floor, cabling is up, and panels going in
on the 6“I floor)

A. Safe

B. Unsafe

C. I am not sure

12. Beams and decking on the 7"I floor are being erected, when the bolting/detailing

on the 3" floor is incomplete.
'7 /.

 

Is this situation (The
number of floors
between the
bolting/detailing and
the erection)

A. Safe

B. Unsafe

C. I am not sure

99

13. Removing the fall protection
while transferring from a beam

to the hoisting crane.

Is this situation (removal of
fall protection during
transfer)

A. Safe

B. Unsafe

C. I am not sure

 

14. A column is bolted in place with 3 anchor rods while beams are being connected

( -

on the 2"d floor.

Is this situation - ,

A. Safe
B. Unsafe
C. I am not sure

 

100

15. Ironworker climbs on the steel beam, when it is held by the crane (tied with a
choker), to bolt it in place.

 

Is this situation (worker on the choker tied beam)
A. Safe

B. Unsafe

C. I am not sure

16. Working on 2,500 square feet of decking that has an unsecured connection.

Is this situation (decking
is unsecured)

A. Safe

B. Unsafe

C. I am not sure

 

 

 

 

101

17. A shaft opening 3 ft. X 1.5 ft., on the 3rd floor is covered by 3/4” ply and painted
with high visibility paint and marked with the word “HOLE”.

 

Is this situation (hole is covered and marked)
A. Safe

B. Unsafe

C. I am not sure

18. Working on a scaffold 5 feet above the lower level without a guardrail system.

     

I
.4 h i ‘
’ ' l i It...“

Is this situation (no guardrail, 5ft above surface)
A. Safe

B. Unsafe

C. I am not sure

102

PERSONALITY QUESTIONS2

0n the following pages, there are phrases describing people '5 behaviors. Please use the
rating scale below each question to describe how accurately each statement describes
you Describe yourself as you generally are now, not as you wish to be in the future.
Describe yourself as you honestly see yourself So that you can describe yourself in an
honest manner, your responses will be kept in absolute confidence. Please read each
statement carefully, and then circle the answer that is most appropriate for you. Please
circle only one answer per question.

1. I am the life of the party
a Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
e. Very Accurate

2. I feel little concern for others
a Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
e Very Accurate

3. I am always prepared

Very Inaccurate

Moderately Inaccurate

Neither Inaccurate nor Accurate
Moderately Accurate

Very Accurate

rap-99‘s»

4. I get stressed out easily

a. Very Inaccurate

b. Moderately Inaccurate

c. Neither Inaccurate nor Accurate
d. Moderately Accurate

e. Very Accurate

5. I have a rich vocabulary

Very Inaccurate

Moderately Inaccurate

Neither Inaccurate nor Accurate
Moderately Accurate

Very Accurate

sup-99‘s»

103

0‘!” {DP-99'!”

99.0

99.ng

I don’t talk a lot

Very Inaccurate

Moderately Inaccurate

Neither Inaccurate nor Accurate
Moderately Accurate

Very Accurate

I am interested in people

Very Inaccurate

. Moderately Inaccurate

Neither Inaccurate nor Accurate
Moderately Accurate
Very Accurate

I leave my belongings around

Very Inaccurate

Moderately Inaccurate

Neither Inaccurate nor Accurate
Moderately Accurate

Very Accurate

I am relaxed most of the time
a.
b.
c. Neither Inaccurate nor Accurate
d.
e.

Very Inaccurate
Moderately Inaccurate

Moderately Accurate
Very Accurate

10. I have difficulty understanding abstract ideas

a.
b.
c.

d.
e.

Very Inaccurate

Moderately Inaccurate

Neither Inaccurate nor Accurate
Moderately Accurate

Very Accurate

11. I feel comfortable around people

sup-99‘s»

Very Inaccurate

Moderately Inaccurate

Neither Inaccurate nor Accurate
Moderately Accurate

Very Accurate

12. I insult people often

0‘?”

9.0-9

Very Inaccurate

. Moderately Inaccurate

Neither Inaccurate nor Accurate
Moderately Accurate
Very Accurate

104

13. I pay attention to details
a. Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
e. Very Accurate

14. I worry about things
a. Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
e. Very Accurate

15. I have a vivid imagination
a. Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
e Very Accurate

16. I keep in the background

Very Inaccurate

. Moderately Inaccurate

Neither Inaccurate nor Accurate
Moderately Accurate

Very Accurate

0"!”

9.0-9

17. I sympathize with others’ feelings
a Very Inaccurate
b. Moderately Inaccurate
Neither Inaccurate nor Accurate
Moderately Accurate
Very Accurate

9.0-9

18. I make a mess of things
a. Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
e. Very Accurate

19. I seldom feel blue

Very Inaccurate

Moderately Inaccurate

Neither Inaccurate nor Accurate
Moderately Accurate

Very Accurate

sup-99‘s»

105

20. I am not interested in abstract ideas
a. Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
e Very Accurate

21. I start conversations
a. Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
e. Very Accurate

22. I am not interested in other people’s problems
a. Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
e. Very Accurate

23. I get chores done right away

a. Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
e. Very Accurate
24. I am easily disturbed

a Very Inaccurate

b. Moderately Inaccurate

c. Neither Inaccurate nor Accurate
d. Moderately Accurate

e. Very Accurate

25. I have excellent ideas
a. Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
e. Very Accurate

26. I have little to say
a. Very Inaccurate
b. Moderately Inaccurate
Neither Inaccurate nor Accurate
Moderately Accurate
Very Accurate

9.0-9

106

27. I have a soft heart

Very Inaccurate

Moderately Inaccurate

Neither Inaccurate nor Accurate
Moderately Accurate

Very Accurate

999.9ng

28. I often forget to put things back in their proper place

a. Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
e. Very Accurate
29. I get upset easily
a. Very Inaccurate
b. Moderately Inaccurate
c. Neither Inaccurate nor Accurate
d. Moderately Accurate
6. Very Accurate

30. I do not have a good imagination
Very Inaccurate

Moderately Inaccurate

Neither Inaccurate nor Accurate
Moderately Accurate

Very Accurate

sup-99‘s»

 

' This survey adapted from Narang (2006)
2 This survey adapted from IPIP (2005) http://ipip.ori.org

107

 

Appendix B
SURVEY RESPONSES AND ANALYSIS (SDT)

108

M FA CR H FA CR H M FA CR
X X
X ' X X

10
11
12
13
14
15
16
17 X
18 X
otal 10 1 4 3

X X
X X

X

X
7 4 3 4 7 4 2 5 10 1 3 4
. 0.91 0.09 0.57 0.43 0.64 0.36 0.43 0. 0.64 0.36 0.29 0.71 0.91 0.09 0.43 0.57

 

Table 8.] Survey Responses and Analysis (SDT)

109

10

11

12

13

14

15

16

17 X X

18 X X

otal 8 3 2 5 10 1 4 3 1 1

. 0.73 0.27 0.29 0.71 0.91 0.09 0.57 0.43 .73 .27 .43 .57 0.91 0. 0. 0.

 

Table 8.] (cont)

110

 

Table B.l (cont.)

1]]

 

1 1 1
0. 0. 0.1 0. 0.910. 0.710 0. 0 0

 

Table B.1 (cont.)

112

0 1
0. 0 0.71 .91 .09 .43 .57 0. 0 0.

 

Table B.1 (cont.)

113

 

Table B.1 (cont.)

114

10
11
12
13
14
15
16
17 X
18 X
otal 9 2 3 4 7 4 3 4
. 0.82 0.18 0.43 0.57 0.64 0.36 0.43 0.57 0. 0. 0. 0.

X
X

 

Table B.1 (cont.)

115

 

Table B.1 (cont.)

116

Appendix C
SURVEY RESPONSES AND ANALYSIS (FFM)

117

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Question

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5 5
M 3 4445423444443433443434 Mflflflfifl
4 4
M 3 3434435445244215444335 MWflMWW
3 3
.WI 3 1413152214241313122443 Mfififimwmm
2 2
M 4 4243412424212224422222 Mfimmflfl
1 4|
M 4 3344434335433343333544 Mwmufimufl
0 0
M 3 3433424234433333423333 MWflflmfl
M 5 5145541325412545415555 Mfiflnmfl
8 851878
W 3 3323343232232433343223 W12111
M 2 3255555252435553355535 Wfiwmwmm
M 4 4325433245322433314324 MQAAflmwmm
5 500449
W 4 3212424345534345443353 W22221
4 492029
3 300449
W 4 3212424345534345443353 W22221
M 4 4555525255444432424524 MWflNWE
M 3 4554524454334353435335 mafia/Nahum.
S S S S X S S S
E EOQACEOHACEOEACEOHACEO HACEO
4 9012345678901234567890
111111111122222222223

Table C.l Survey Responses and Analysis (FFM)

118

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

W16 W17 W18 W19 W20 W21 W22 W23 W24 W25 W26 W27 W28 W29 W30

 

 

Question

ES

ES

EX

ES

ES

EX

ES

ES

 

 

 

 

 

 

 

 

 

 

 

10
11
12
13
14
15
16
17
18
19
20
21

 

 

 

 

 

 

 

 

 

 

 

 

22
23

 

 

24
25

 

 

26
27

 

 

28
29

 

 

30

 

 

m
m
m
m
m
m
w.
m
m
m
m
m
m
m
m

 

Table C.1 (cont.)

119

Appendix D
REGRESSION ANALYSIS: MINITAB RESULTS

120

Regression Analysis: d' versus Age

The regression equation is

d’ = 0.792 - 0.0018 Age

Predictor Coef SE Coef T P
Constant 0.7920 0.5420 1.46 0.155
Age -0.00181 0.01234 -O.15 0.884
S = 0.576609 R-Sq = 0.1% R-Sq(adj) = 0.0%

Analysis of Variance

Source DF

Regression 1 0.0072

Residual Error 28 9.3094

Total 29 9.3166

Unusual Observations

Obs Age d’ Fit
10 49.0 -0.468 0.703
13 42.0 —0.955 0.716

MS F P
0.0072 0.02 0.884
0.3325

SE Fit Residual St Resid
0.128 —1.171 -2.08R
0.106 —1.671 -2.95R

R denotes an observation with a large standardized residual.

 

 

 

 

 

Scatterplot of d' vs Age
Scarlet-plot of d' vs Age
IHSd o a
O O CiOIIGD O
1.0‘ .
am e
T J
(15- 0 : 0
Q .
'U
0.04 o O
-0.5* '
-1.0~ ’
20 25 30 35 40 45 50 55
than

 

 

121

 

 

Regression Analysis: (1' versus Exp.

The regression equation is
d' = 0.649 + 0.0030 Exp.

Predictor Coef SE Coef T P
Constant 0.6494 0.2699 2.41 0.023
Exp. 0.00303 0.01170 0.26 0.797
S = 0.576141 R-Sq = 0.2% R-Sq(adj) = 0.0%

Analysis of Variance

Source DF SS MS F P
Regression 1 0.0223 0.0223 0.07 0.797
Residual Error 28 9.2943 0.3319

Total 29 9.3166

Unusual Observations

Obs Exp. d’ Fit SE Fit Residual St Resid
10 32.0 -0.468 0.747 0.164 -1.215 -2.20R
13 14.0 -O.955 0.692 0.135 -l.647 -2.94R

R denotes an observation with a large standardized residual.

Scatterplot of d' vs Exp.

 

 

 

 

 

 

 

Scatterplot of (1’ vs Exp.
1.5 o o
9 o o a see.
1 0- '
' o
o o . a
_ o o
0.5 0 . 3
in
0.0- o 9 o
-0.5‘ .
-1.0- °
I I I I I I r I
0 5 10 15 20 25 30 35
Exp.

 

 

 

122

Regression Analysis: d' versus Ace.

The regression equation
d‘ = 0.684 + 0.150 Acc.

Predictor Coef SE Co
Constant 0.6839 0.11
Acc. 0.1497 0.26
S = 0.573490 R-Sq = 1
Analysis of Variance
Source DF
Regression l 0.10
Residual Error 28 9.20
Total 29 9.31
Unusual Observations
Obs Acc. d' Fit
10 0.00 —0.468 0.684
13 0.00 -0.955 0.684

is

ef T P

71 5.84 0.000

18 0.57 0.572

.2% R—Sq(adj) = 0.0%

SS MS F P

76 0.1076 0.33 0.572

89 0.3289

66

SE Fit Residual St Resid

0.117 —1.152 —2.05R
0.117 -l.639 —2.92R

R denotes an observation with a large standardized residual.

 

 

 

 

 

 

Scattermt of d' vs Ace
Scatterplot of d ' vs Aoc.
1.5- 0
g o
1.0~ ‘3
0.51 E O
is
0.04 8 o
-0.5* 9
-1.04 '
0.0 0.2 0.4 0.6 0.8 1.0
Ace.

 

 

123

 

 

Regression Analysis: (1' versus EX

The regression equation is

d’ = - 0.330 + 0.0553 EX

Predictor Coef SE Coef
Constant -0.3303 0.5350
EX 0.05534 0.02787

3 = 0.540059 R—Sq = 12.3%

Analysis of Variance

Source DF SS
Regression 1 1.1500
Residual Error 28 8.1666
Total 29 9.3166

Unusual Observations

Obs EX d' Fit
6 9.0 1.1640 0.1678
13 13.0 -O.9550 0.3892

T P
—0.62 0.542
1.99 0.057
R—Sq(adj) = 9.2%
MS F P
1.1500 3.94 0.057
0.2917

SE Fit Residual St Resid
0.2921 0.9962 2.19RX
0.1909 —1.3442 —2.66R

R denotes an observation with a large standardized residual.
X denotes an observation whose X value gives it large influence.

Scatterplot of (1' vs EX

 

Scatterplotof d' vs EX

 

L5“

L0“

05~

OIF

-0.5d

 

-L0~

 

 

10 12

 

14 16 18 20 22 24 26

 

124

 

 

Regression Analysis: (1' versus A

The regression equation is

d’ = - 0.656 + 0.0615 A

Predictor Coef SE Coef
Constant -0.6563 0.6291
A 0.06153 0.02791

5 = 0.532475 R-Sq = 14.8%

Analysis of Variance

Source DF SS
Regression 1 1.3777
Residual Error 28 7.9388
Total 29 9.3166

Unusual Observations

Obs A d’ Fit
10 21.0 —0.4680 0.6359
13 11.0 —0.9550 0.0206

T P
—1.04 0.306
2.20 0.036
R-Sq(adj) = 11.7%
MS 8 p
1.3777 4.86 0.036
0.2835

SE Fit Residual St Resid
0.1034 —1.1039 —2.11R
0.3292 -O.9756 -2.33RX

R denotes an observation with a large standardized residual.
X denotes an observation whose X value gives it large influence.

Scatteflilot of d' vs A

 

Scatterplot of (1' vs A

 

1”5-

IHO-

(15‘

(L04

415‘

{Lon .

 

 

 

 

10 15

20 25 30

 

125

 

 

Regression Analysis: d' versus C

The regression equation is

d’ = — 0.185 + 0.0393 C

Predictor Coef SE Coef T P

Constant -0.1850 0.7326 -0.25 0.803

C 0.03931 0.03172 1.24 0.226

S = 0.561639 R-Sq = 5.2% R-Sq(adj) = 1.8%

Analysis of Variance

Source DF SS MS F P

Regression 1 0.4843 0.4843 1.54 0.226

Residual Error 28 8.8323 0.3154

Total 29 9.3166

Unusual Observations

Obs C d' Fit SE Fit Residual St Resid
10 22.0 -0.468 0.680 0.106 —l.148 —2.08R
13 15.0 —0.955 0.405 0.270 —1.360 —2.76RX

R denotes an observation with a large standardized residual.

X denotes an observation

Scattemlot of d' vs C

whose X value gives it large influence.

 

Scatterplot of d' vs C

 

1.54

1.07

0.5-

0.04

 

 

 

 

15.0

 

17.5

20.0

22.5

25.0

27.5

 

126

 

Regression Analysis: (1' versus ES

The regression equation is

d’ = — 0.163 + 0.0439 ES
Predictor Coef SE Coef T
Constant *0.1633 0.6079 —0.27
ES 0.04393 0.03002 1.46
S = 0.555964 R—Sq = 7.1% R-Sq(
Analysis of Variance
Source DF SS MS
Regression 1 0.6619 0.6619
Residual Error 28 8.6547 0.3091
Total 29 9.3166
Unusual Observations
Obs ES (1' Fit SE Fit
9 28.0 0.535 1.067 0.262
10 18.0 -0.468 0.627 0.117
13 18.0 -0.955 0.627 0.117

P
0.790
0.155

adj) 3

F

2.14

Residual
—0.532
—1.095
—1.582

.8%

0.15

St

P
5

Resid
-1.08
-2.02R
-2.91R

X

R denotes an observation with a large standardized residual.
X denotes an observation whose X value gives it large influence.

 

 

 

 

  

 

Scatterplot of d' vs ES
Scatherplot of (1' vs ES
151 o o
a z 0 o o o o
1.0-
0.5- '
. O
'0
0.04 ‘ 8
-0.54 .
-1.08 '
15.0 17.5 20.0 22.5 25.0 27.5

 

 

127

 

 

Regression Analysis: (1' versus 0

The regression equation is
d’ = 0.015 + 0.0338 0

Predictor Coef SE Coef T P

Constant 0.0148 0.5738 0.03 0.980

0 0.03383 0.02732 1.24 0.226

S = 0.561660 R-Sq = 5.2% R-Sq(adj) = 1.8%

Analysis of Variance

Source DF SS MS F P

Regression 1 0.4837 0.4837 1.53 0.226

Residual Error 28 8.8329 0.3155

Total 29 9.3166

Unusual Observations

Obs 0 d’ Fit SE Fit Residual St Resid
10 21.0 —0.468 0.725 0.103 -1.193 -2.16R
12 12.0 0.436 0.421 0.258 0.015 0.03 X
13 18.0 -0.955 0.624 0.126 —1.579 -2.88R

R denotes an observation wi

th a large standardized residual.

X denotes an observation whose X value gives it large influence.

 

 

 

 

 

 

Scatterplot of d' vs 0
Scatterplot of d ' vs 0
1.57 9 9
1.0-
0.5-
in

0.04 9 9 .

-0.5- 9

-1.0- 9

f I I I I
10 15 20 25 30
O

 

 

128

 

Regression Analysis: b versus Age

The regression equation is
b = 1.88 - 0.0280 Age

Predictor Coef SE Coef T P
Constant 1.8842 0.5389 3.50 0.002
Age —0.02798 0.01228 -2.28 0.030
S = 0.573397 R-Sq = 15.7% R-Sq(adj) = 12.6%

Analysis of Variance

Source DF SS MS F P
Regression 1 1.7086 1.7086 5.20 0.030
Residual Error 28 9.2059 0.3288

Total 29 10.9146

Unusual Observations

Obs Age b Fit SE Fit Residual St Resid
14 28.0 3.214 1.101 0.213 2.113 3.97R

R denotes an observation with a large standardized residual.

Scatterplot of b vs Age

 

Scatterplot of h vs Age

 

3.5-

3.0-

2.55

1.0‘

0.0 -

 

 

55

 

 

 

 

 

129

Regression Analysis: b versus Exp.

The regression equation is
b = 1.23 - 0.0258 Exp.

Predictor Coef SE Coef T P
Constant 1.2262 0.2700 4.54 0.000
Exp. —0.02577 0.01171 -2.20 0.036
S = 0.576489 R-Sq = 14.7% R-Sq(adj) = 11.7%

Analysis of Variance

Source DF SS MS F P
Regression 1 1.6090 1.6090 4.84 0.036
Residual Error 28 9.3055 0.3323

Total 29 10.9146

Unusual Observations

Obs Exp. b Fit SE Fit Residual St Resid
14 8.0 3.214 1.020 0.187 2.194 4.02R

R denotes an observation with a large standardized residual.

Seatterplot of h vs Exp.

 

Scatterplot of h vs Exp.

 

3.5~

2.54
2.0-
1.5~ 9
1.04

0.5..

 

 

0.0 -

 

 

.-
..
.1
.1
._.
fi
4

 

 

 

130

 

Regression Analysis: b versus Acc.

The regression equation is
b = 0.662 + 0.084 Acc.

Predictor Coef SE Coef T P
Constant 0.6623 0.1272 5.20 0.000
Acc. 0.0841 0.2845 0.30 0.770
S = 0.623373 R-Sq = 0.3% R-Sq(adj) = 0.0%

Analysis of Variance

Source DF SS MS F P
Regression 1 0.0339 0.0339 0.09 0.770
Residual Error 28 10.8806 0.3886

Total 29 10.9146

Unusual Observations

Obs Acc. b Fit SE Fit Residual St Resid
14 0.00 3.214 0.662 0.127 2.552 4.18R

R denotes an observation with a large standardized residual.

 

 

 

 

 

 

Scatterplot of b vs Acc.
Scatterplot of b vs Acc.
3.5 —
O
3.0~
2.55
2.0-
.n
1.5 ~ 9
. O
1.0-4
3 8
0.5 -
O
o
0'0- I I l I T I
0 0 0 2 0.4 0 6 0 8 1 0
Ace.

 

 

 

131

Regression Analysis: b versus EX

The regression equation is
b = 0.437 + 0.0128 EX

Predictor Coef SE Coef T P
Constant 0.4368 0.6167 0.71 0.485
EX 0.01284 0.03213 0.40 0.692

s = 0.622570 R-Sq = 0.6% R-Sq(adj) = 0.0%

Analysis of Variance

Source DF SS MS F P
Regression 1 0.0619 0.0619 0.16 0.692
Residual Error 28 10.8526 0.3876

Total 29 10.9146

Unusual Observations

Obs EX b Fit SE Fit Residual St Resid
6 9.0 0.158 0.552 0.337 —0.394 -0.75 X
14 19.0 3.214 0.681 0.114 2.533 4.14R

R denotes an observation with a large standardized residual.
X denotes an observation whose X value gives it large influence.

Scatterplot of b vs EX

 

 

Scatterplot of b vs EX
3.5-
O
3.0~
2.55
2.0-
.n
1.5- 9

IHO‘
so :8

 

 

 

 

__ '
05‘ 9 ‘
O . O
0'0‘ I r I I I I
10 12 14 16 18 20
EX

 

 

 

132

Regression Analysis: b versus A

The regression equation is
b = 1.03 - 0.0159 A

Predictor Coef SE Coef - T P
Constant 1.0337 0.7345 1.41 0.170
A —0.01593 0.03259 -0.49 0.629
S = 0.621698 R—Sq = 0.8% R-Sq(adj) = 0.0%

Analysis of Variance

Source DF SS MS F P
Regression 1 0.0923 0.0923 0.24 0.629
Residual Error 28 10.8222 0.3865

Total 29 10.9146

Unusual Observations

Obs A b Fit SE Fit Residual St Resid
13 11.0 0.523 0.859 0.384 -0.336 —0.69 X
14 22.0 3.214 0.683 0.114 2.531 4.14R

R denotes an observation with a large standardized residual.
X denotes an observation whose X value gives it large influence.

 

 

 

 

 

Scatterplot of b vs A
ScatterplotofbvsA
3.55
O
3.0-
251
2.0~
.n
1.5~ 9
. O
1.04
' 8 3 . e
—__—————_———__—————————_-—__7_"""1r*————————___________9
0.5" . . z .
o 9 o a . 8
0.0-‘1 I I I I
10 15 20 25 30
A

 

 

 

133

Regression Analysis: b versus C

The regression equation is
b = 0.858 - 0.0078 C

Predictor Coef SE Coef T P
Constant 0.8578 0.8137 1.05 0.301
C -0.00781 0.03523 —0.22 0.826
S = 0.623796 R—Sq = 0.2% R-Sq(adj) = 0.0%

Analysis of Variance

Source DF SS MS F P
Regression 1 0.0191 0.0191 0.05 0.826
Residual Error 28 10.8954 0.3891

Total 29 10.9146

Unusual Observations

Obs C b Fit SE Fit Residual St Resid
13 15.0 0.523 0.741 0.300 -0.218 -0.40 X
14 18.0 3.214 0.717 0.206 2.497 4.24R

R denotes an observation with a large standardized residual.

X denotes an observation whose X value gives it large influence.

Scatterplot of b vs C

 

Scatterplot of b vs C

 

3.5-
3.0-

2.5..

1.5~ 9

1.0~
3

 

_ e
0.5 o z
0

 

0.0-

 

 

 

15.0 17.5 20f0 22.5 25.0

 

134

 

Regression Analysis: b versus ES

The regression equation is
b = 0.037 + 0.0322 ES

Predictor Coef SE Coef T P
Constant 0.0368 0.6715 0.05 0.957
ES 0.03217 0.03316 0.97 0.340
S = 0.614111 R—Sq = 3.3% R-Sq(adj) = 0.0%

Analysis of Variance

Source DF SS MS F P
Regression 1 0.3549 0.3549 0.94 0.340
Residual Error 28 10.5597 0.3771

Total 29 10.9146

Unusual Observations

Obs ES b Fit SE Fit Residual St Resid
9 28.0 0.837 0.937 0.289 -0.100 —0.19 X
14 19.0 3.214 0.648 0.117 2.566 4.26R

R denotes an observation with a large standardized residual.
X denotes an observation whose X value gives it large influence.

Scatterplot of b vs ES

 

Scatterplot of b vs ES

 

3.5-

3.0-

2.5-J

2.0 r

1.0‘

0.5-

 

 

0.0 -

 

 

15.0 17.5 20.0 22.5 25.0 27.5

 

 

 

135

Regression Analysis: b versus 0

The regression equation is
b = 0.309 + 0.0179 0

Predictor Coef SE Coef T P
Constant 0.3088 0.6339 0.49 0.630
0 0.01791 0.03018 0.59 0.558
S = 0.620453 R—Sq = 1.2% R—Sq(adj) = 0.0%

Analysis of Variance

Source DF SS MS F P
Regression 1 0.1356 0.1356 0.35 0.558
Residual Error 28 10.7789 0.3850

Total 29 10.9146

Unusual Observations

Obs O b Fit SE Fit Residual St Resid
12 12.0 0.474 0.524 0.285 -0.050 -0.09 X
14 27.0 3.214 0.793 0.222 2.421 4.18R

R denotes an observation with a large standardized residual.
X denotes an observation whose X value gives it large influence.

 

 

 

 

 

Scatterplot of b vs 0
Scatterplot of b vs 0
3.5-
3.0-
2.53
2.0-
.n
1.5- 9
. O
1.04
O O 3 . . .
‘//.’i/”./
0.51 . 9 o
o 9 8 .
o o g o o
0.0-
T I I I
10 15 20 25
O

 

 

 

136

Appendix E
NORMAL DEVIATES AND ORDINATES FOR d' AND 8

137

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Normal Normal
p Deviate Ordinates p Deviate Ordinates
Z Z
0.01 2.326 0.027 0.26 0.643 0.325
0.02 2.054 0.048 0.27 0.613 0.331
0.03 1.881 0.068 . 0.28 0.583 0.337
0.04 1.751 0.086 0.29 0.553 0.342
0.05 1.645 0.103 0.3 0.524 0.348
0.06 1.555 0.1 19 0.31 0.496 0.353
0.07 1.476 0.134 0.32 0.468 0.358
0.08 1.405 0.149 0.33 0.44 0.362
0.09 1.341 0.162 0.34 0.412 0.367
0.1 1.282 0.176 0.35 0.385 0.371
0.1 1 1.227 0.188 0.36 0.358 0.374
0.12 1.175 0.2 0.37 0.332 0.378
0.13 1.126 0.212 0.38 0.305 0.381
0.14 1.08 0.223 0.39 0.279 0.384
0.1 5 1.036 0.233 0.4 0.253 0.386
0.16 0.994 0.243 0.41 0.228 0.389
0.17 0.954 0.253 0.42 0.202 0.391
0.18 0.915 0.263 0.43 0.176 0.393
0.19 0.878 0.272 0.44 0.151 0.394
0.2 0.842 0.28 0.45 0.126 0.396
0.21 0.806 0.288 0.46 0.1 0.397
0.22 0.772 0.296 0.47 0.075 0.398
0.23 0.739 0.304 0.48 0.05 0.398
0.24 0.706 0.311 0.49 0.025 0.399
0.25 0.674 0.318 0.5 0 0.399

 

 

Table E.l Normal Deviates and Ordinates for Calculating d' and [i

(Adapted from Patel, 2003)

138

Appendix F
MULTPILE REGRESSION RESULTS

139

Regression Analysis: 6' versus Age, Exp., Acc., EX, A, C, ES, 0

The regression equation is
d’ = — 1.45 — 0.0212 Age + 0.0295 Exp. + 0.041 Acc. + 0.0514 EX + 0.0820 A
— 0.0567 C + 0.0526 ES - 0.0051 0

Predictor Coef SE Coef T P
Constant —1.453 1.821 —0.80 0.434
Age -0.02117 0.04172 —0.51 0.617
Exp. 0.02948 0.03663 0.80 0.430
Acc. 0.0409 0.3503 0.12 0.908
EX 0.05141 0.03766 1.37 0.187
A 0.08200 0.04636 1.77 0.091
C -0.05666 0.05018 -1.13 0.272
ES 0.05260 0.03240 1.62 0.119
0 —0.00512 0.03795 -0.13 0.894

S = 0.537666 R-Sq = 34.9% R-Sq(adj) = 10.0%

Analysis of Variance

Source DF SS MS F P
Regression 8 3.2477 0.4060 1.40 0.252
Residual Error 21 6.0708 0.2891

Total 29 9.3185

Source DF Seq SS
Age 1 0.0077
Exp. 1 0.4719
Acc. 1 0.0797
EX 1 0.9176
A 1 0.7411
C 1 0.2596
ES 1 0.7649
0 1 0.0053

Unusual Observations

Obs Age d' Fit SE Fit Residual St Resid
10 49.0 -0.4700 0.5389 0.2796 —1.0089 —2.20R

R denotes an observation with a large standardized residual.

Regression Analysis: d' versus Exp., EX, A, C, ES

The regression equation is

d’ = — 2.17 + 0.0129 Exp. + 0.0540 EX + 0.0847 A - 0.0575 C + 0.0510 ES
Predictor Coef SE Coef T P
Constant -2.1689 0.9527 —2.28 0.032
Exp. 0.01294 0.01133 1.14 0.265
EX 0.05400 0.03296 1.64 0.114
A 0.08466 0.03770 2.25 0.034

140

C -0.05751 0.04324 -l.33 0.196
ES 0.05105 0.02883 1.77 0.089

s = 0.506378 R-Sq = 34.0% R-Sq(adj) = 20.2%

Analysis of Variance

Source DF SS MS F P
Regression 5 3.1645 0.6329 2.47 ‘0.061
Residual Error 24 6.1541 0.2564

Total 29 9.3185

Source DF Seq SS

Exp. 1 0.0218
EX 1 1.2790
A 1 0.8587
C 1 0.2012
ES 1 0.8038

Unusual Observations

Obs Exp. d' Fit SE Fit Residual St Resid
10 32.0 —0.4700 0.4867 0.1926 -0.9567 -2.04R
30 29.0 0.3600 1.2666 0.2277 —0.9066 -2.00R

R denotes an observation with a large standardized residual.

Regression Analysis: d' versus EX, A, ES

The regression equation is

d' = - 1.88 + 0.0336 EX + 0.0469 A + 0.0456 ES
Predictor Coef SE Coef T P
Constant -1.8755 0.8726 -2.15 0.041
EX 0.03360 0.03013 1.12 0.275
A 0.04694 0.03063 1.53 0.137
ES 0.04559 0.02783 1.64 0.113
S = 0.514827 R-Sq = 26.0% R—Sq(adj) = 17.5%

Analysis of Variance

Source DF SS MS F P
Regression 3 2.4273 0.8091 3.05 0.046
Residual Error 26 6.8912 0.2650

Total 29 9.3185

Source DF Seq SS

EX 1 1.1552
A 1 0.5609
ES 1 0.7112

141

Unusual Observations

Obs EX d' Fit SE Fit Residual St Resid
6 9.0 1.1600 0.4198 0.3392 0.7402 1.91 X
13 13.0 -0.9500 -0.1017 0.3256 —0.8483 —2.13R

R denotes an observation with a large standardized residual.
X denotes an observation whose X value gives it large influence.

Regression Analysis: b versus Age, Exp., Acc., EX, A, C, ES, 0

The regression equation is
b = 3.85 - 0.0566 Age + 0.0026 Exp. + 0.411 Acc. + 0.0156 EX - 0.0429 A
— 0.0299 C + 0.0449 ES - 0.0205 0

Predictor Coef SE Coef T P
Constant 3.853 1.912 2.01 0.057
Age —0.05662 0.04380 —1.29 0.210
Exp. 0.00256 0.03845 0.07 0.948
Acc. 0.4108 0.3678 1.12 0.277
EX 0.01559 0.03954 0.39 0.697
A —0.04293 0.04867 —0.88 0.388
C -0.02987 0.05269 -0.57 0.577
ES 0.04490 0.03401 1.32 0.201
0 -0.02048 0.03984 —0.51 0.613

s = 0.564465 R—Sq = 38.6% R-Sq(adj) = 15.2%

Analysis of Variance

Source DF SS MS F P
Regression 8 4.1990 0.5249 1.65 0.171
Residual Error 21 6.6910 0.3186

Total 29 10.8900

Source DF Seq SS
Age 1 1.7132
Exp. 1 0.0131
Acc. 1 0.8408
EX 1 0.1662
A 1 0.7757
C 1 0.0411
ES 1 0.5647
0 1 0.0842

Unusual Observations

Obs Age b Fit SE Fit Residual St Resid
13 42.0 0.520 1.233 0.443 -0.713 —2.04R
14 28.0 3.210 1.402 0.341 1.808 4.02R

R denotes an observation with a large standardized residual.

142

Regression Analysis: b versus Age, Acc., A, ES

The regression equation is
b = 3.30 - 0.0502 Age + 0.324 Acc. - 0.0621 A + 0.0430 ES

Predictor Coef SE Coef T P
Constant 3.302 1.183 2.79 0.010
Age -0.05022 0.01389 -3.62 0.001
Acc. 0.3242 0.2832 1.14 0.263
A -0.06214 0.03091 -2.01 0.055
ES 0.04301 0.03063 1.40 0.173
S = 0.523800 R—Sq = 37.0% R—Sq(adj) = 26.9%

Analysis of Variance

Source DF SS MS F P
Regression 4 4.0308 1.0077 3.67 0.017
Residual Error 25 6.8592 0.2744

Total 29 10.8900

Source DF Seq SS

Age 1 1.7132
Acc. 1 0.7684
A 1 1.0083
ES 1 0.5409

Unusual Observations

Obs Age b Fit SE Fit Residual St Resid
13 42.0 0.5200 1.2835 0.3698 —0.7635 —2.06R
14 28.0 3.2100 1.3462 0.2185 1.8638 3.92R

R denotes an observation with a large standardized residual.

Regression Analysis: b versus Age, A

The regression equation is
b = 3.67 - 0.0390 Age - 0.0590 A

Predictor Coef SE Coef T P
Constant 3.673 1.108 3.31 0.003
Age -0.03899 0.01323 -2.95 0.007
A -0.05902 0.03239 -1.82 0.079

s = 0.550142 R-Sq = 25.0% R-Sq(adj) = 19.4%

Analysis of Variance

Source DF SS MS F P
Regression 2 2.7183 1.3591 4.49 0.021
Residual Error 27 8.1717 0.3027

Total 29 10.8900

143

Source DF Seq SS
Age 1 1.7132
A 1 1.0051

Unusual Observations

Obs Age b Fit SE Fit Residual St Resid
13 42.0 0.520 1.386 0.385 —0.866 —2.20RX
14 28.0 3.210 1.283 0.227 1.927 3.85R

R denotes an observation with a large standardized residual.
X denotes an observation whose X value gives it large influence.

144

Appendix G
SAMPLE DISTRIBUTIONS FOR THE FIVE FACTOR RANGES

145

 

l__ .‘ 71 ,_ ‘ ‘1

  

Extraversion
\
3 3% ‘
‘ V 30% .
i f‘ 'fi
:D'OW \
1 W
a high J i
’7 I

L

i 67% ‘

fill ll_l-_l,lfl.lllil

Figure 4.6 Low, Moderate and High Extraversion Percentages

Agreeableness ‘

1 3% 1

* 3‘7
1
\

 

  

Figure 4.7 Low, Moderate and High Agreeableness Percentages

‘ Conscientiousness I

i
40% h‘ hJ

 

146

Figure 4.8 Low, Moderate and High Conscientiousness

 

Emotional Stabiity

 

  

1““1

Imod \

i @991

Figure 4.9 Low, Moderate and High Emotional Stability

i if 7707 7 ridlel'll'IieSSfir iifir i 7 7 i

3% 1

~ “ ‘
:1 low

1 L1

E‘E"

67%

Figure 4.10 Low, Moderate and High Openness

147

REFERENCES

148

REFERENCES

Abdelhamid, TS. and Everett, J. (2000). “Identifying Root Causes of Construction
Accidents.” Journal of Construction Engineering and Management. ASCE, 52-59.

Abdelhamid T.S., Patel B., Howell G.A. and Mitropoulos P. (2003). “Signal Detection
Theory: Enabling work near the Edge”. Proceedings of the I 1"' Conference of the
International Group of Lean Construction, July 2003, Virginia Tech, Blacksburg,
VA.

Broberg, E. (1986). Risk Estimation for Common Consumer Products. Proceedings of the
Human Factors Society 3 0‘” Annual Meeting. Santa Monica, CA: Human Factors
Society pp. 556-560

Brown, I. (1990). Accident Reporting and Analysis. In J. Wilson and E Corlett (eds.),
Evaluation of Human Work. London: Taylor & Francis, pp 755-778.

Bureau of Labor Statistics (2005). Career Guide to Industries, US. Department of Labor,
2004-2005 Edition. http://www.bls.gov/oco/cg/cgsoo3.htm (Sept 19, 2005).

Chua, D.K.H., and Goh, Y.M. (2004). Incident Causation Model for Improving Feedback
and Safety Knowledge. Journal of Construction Engineering and Management.
ASCE, July/August 2004. pp. 542-551.

Craik, C.H., Hogan, R., and Wolfe, RN. (1993). Fifty Years of Personality Psychology,
Plenum Press, New York, NY.

Cellar, D.F., Nelson, Z.C., Yorke, C.M., and Bauer, C. (2001). The Five Factor Model
and Safety in the Workplace: Investigating the Relationships Between Personality and
Accident Involvement. Journal of Prevention and Intervention in the Community. The
Hayworth Press, Inc. Vol. 22, No. 1, p. 43-52.

Culver, C., Hardesty, C., and Anderson, F. (1993). “Prevent Fatal Falls in Construction”.
Safety and Health, NSC, Itasca, IL.

Feist, J. (1994.) Theories of Personality, Harcourt Brace College Publishers, New York,
NY.

Geller, E.S., Wiegand, D.M. (2006). People Based Safety. Virginia Polytechnic
University. Unpublished Report.

149

 

Hadden, W., Suchman, E.A., & Klein, D. (1964). Accident Research: Methods and
Approaches. New York: Harper & Row.

Hinze, J., Pedersen, C., and Frdly, J. (1998). “Identifying the Root Causes of
Construction Injuries.” Journal of Construction Engineering and Management.
ASCE, 124(1), 67-69.

Hinze, J. (1996). “The Distraction Theory of Accident Causation”. Proc., Int Conference.
An Implementation of Safety and Health on Construction Sites. CIB Working
Commission W99.

Howell G.A., Ballard 6., Abdelhamid T. and Mitropoulos P. (2002). “Working Near the
Edge: A New Approach to Construction Safety.” Proceedings IGLC-I 0, August 2002,
Gramado, Brazil.

International Personality Item Pool (2001). A Scientific Collaboratory for the
Development of Advanced Measures of Personality Traits and Other Individual
Differences. http://ipip.ori.org/ (Sept 28, 2005).

Karp, C., & Karp, L. (2000) Minnesota Multiphasic Inventory: Questions to Ask.
http://www.falseallegations.com/mmpi-bw.htm. (Feb 22, 2006).

Lampert, U. (1974). Age and the Predisposition to Accidents. Archives des Maladies
Professionelles, 62, 173.

Macmillan, N., and Creelman, C. (1991). Detection Theory: A User ’s Guide. Cambrodge
University Press, New York, NY.

Maddi, S. (1996). Personality Theories, A Comparative Analysis, Brooks/Cole Publishing
Company, Pacific Grove, CA.

McCrae, R., and Costa, P. (1990). Personality in Adulthood, Guilford, New York.

McKenna, RP. (1983). Accident Proneness: A Conceptual Analysis. Accident Analysis &
Prevention, 15, 65-71.

McMartin, J. (1995). Personality Psychology, A Student Centered Approach, Sage
Publications, Thousand Oaks, CA.

McAdams, DP. (2001 ). The Person: An Integrated Introduction to Personality
Psychology, Harcourt College Publishers, Orlando, FL.

Mitropoulos, P., Abdelhamid, T., and Howell, G. (2005). Systems Model of Construction

Accident Causation. Journal of Construction Engineering and Management. ASCE,
July 2005. pp.816-825.

150

 

 

Molenaar, K., Brown, H., Caile, S., and Smith, R. (2002). Corporate Culture: A Study of
Firms with Outstanding Construction Safety. Journal of Professional Safety. ASSE.
July 2002. pp. 18-27

National Academy of Sciences (1982). Toward Safer Underground Coal Mines.
Washington, DC: National Academy of Sciences.

Narang, P. (2006). Analisis of Construction Safety Competencies using Signal Detection
Theory. MS. Thesis, Michigan State University, East Lansing, MI.

Norman, W.T. (1967). Toward an Adequate Taxonomy of Personality Attributes:
Replicated Factor Structure in Peer Nomination Personality Ratings. Journal of
Abnormal and Social Psychology, 66, 574-583.

Ones, D. S., & Viswesvaran, C. (2002). Industrial-organizational (I/O) psychology to
organizational behavior management (OBM): Join the family — Individual differences
matter. Journal of Organizational Behavior Management, 22(2), 41-57.

OSHA. (2005). US. Department of Labor. Bureau of Labor Statistics, Accident
Investigation Guidelines. http://www.osha.gov/SL TC/smallbusiness/sec6.html (Sept
26, 2005).

Patel, B.J. (2003). Assessment of Construction Workers Occupational Safety
Competencies Using Signal Detection Theory. MS. Thesis, Michigan State
University, East Lansing, MI.

Personality Pathways. (2005). Introduction to Myers-Briggs Personality Type & the
MBTI. http://www.personalitypathways.com/MBTI_intro.htmI. (Sept 10, 2005).

Pervin, L. A., and John, O. P. (1992). Handbook of Personality, Theory and Research.
The Guilford Press, New York, NY.

Rasmussen, J. (1997). Risk Management in a Dynamic Society: A Modeling Problem.
Safety Science Vol. 27 No. 2/3 183-213.

Reese, CD, and Eidson, J .V. (1999). Handbook of OHSA Construction Safety and
Health. Lewis Publishers, Boca Raton, FL.

Sanders, MS, and McCormick, B.J. (1993). Human Factors in Engineering and Design,
Seventh Edition, McGraw Hill, New York, NY.

Shaw, L., & Sichel, H. S. (1971). Accident Proneness. Pergamon Press. New York, NY.
Shahani, C. (1987). Industrial Accidents: Does age matter? Proceedings of the Human

Factors Society 31" Annual Meeting. Santa Monica, CA: Human Factors Society, pp.
553-557

151

 

Suraji, A., Duff, R., and Peckin, S. (2001). “Development of a Causal Model of
Construction Accident Causation”. Journal of Construction Engineering and
Management. July/August, 337-344.

Wickens, C., and Hollands, J. (2000). Engineering Psychology and Human Performance.
Prentice Hall Inc., Upper Saddle River, NJ.

Winter, D.G., John, O.P., Stewart, A.J., Klohnen, EC, and Duncan LE. (1998). Traits

and motives: Toward and integration of two traditions in personality research.
Psychological Review 105, 230-250.

152

   

u111111111111