Professional Documents
Culture Documents
Abstract: Disparities in worker risk tolerance may create barriers to implementing safety management systems and improving safety per-
Downloaded from ascelibrary.org by University of Liverpool on 02/16/20. Copyright ASCE. For personal use only; all rights reserved.
formance. At present, it is unclear if and to what extent construction safety risk tolerance vary across broad geographic regions. To better
understand patterns in these sociocultural constructs, a survey of building trade contractors and subcontractors was administered. Using
principal component analysis and K-means clustering, the determinants of risk tolerance were analyzed for 11,997 construction workers
from 17 countries via controlled sampling for equal representation. The analysis showed that risk tolerance is influenced and linked by
individual and sociocultural determinants, i.e., affective associations, control beliefs, safety culture, and risk-taking attitudes. Differences
and distinct groupings were observed when the derived global risk tolerance scores were compared to country-specific risk-tolerance scores.
This study contributes to the literature by empirically identifying determinants of risk tolerance and quantifying cross-cultural disparities in
risk tolerance. It was found that the natural grouping of countries, based on their risk-tolerance determinants, coincides with their ancestral
heritage and socioeconomic systems. The results can be used to inform policymakers, stakeholders, safety professionals, and industry leaders
to improve safety decisions in the workplace, promote strong situational awareness, design structural policies, and implement safety
programs. DOI: 10.1061/(ASCE)CO.1943-7862.0001789. © 2020 American Society of Civil Engineers.
Author keywords: Construction; Safety; Risk tolerance; Situational awareness; Cross-culture; Fundamental social cause.
Introduction During the last two decades, the prevention of serious incidents
and fatalities has been at the forefront of project planning. Safety
Globalization has facilitated the rapid growth of international con- management systems have been commonly adopted by con-
struction markets via advancement in communications, materials, struction companies, which are sets of interrelated or interacting
technology, and transportation. In 2016, the construction industry elements to establish systematic safety and health policy and ob-
accounted for approximately 6% of the global gross domestic prod- jectives to proactively implement interventions and improve out-
uct (GDP) with total annual revenues of $13 trillion (Gerbert et al. comes (Robson et al. 2007). Despite these efforts, serious injury
2016). In the US alone, construction spending increased by 3.1% to and fatality rates in the industry have plateaued for the past 5 years
$1.2 trillion in 2017, representing approximately 6.5% of the (ILO 2018). With increasing societal expectations and environmen-
national GDP (US Census 2017). Already a sizeable contributor tal pressures, the financial and reputational consequences of major
to the global GDP, the industry is expected to grow with estimated accidents have never been greater (Hopkins 2010; Rasmussen
revenues of $15 trillion by 2025 (Oxford Economics 2018). As a 1997). Therefore, much improvement is still needed.
result, an increasing number of construction companies are explor- High-quality construction safety management systems are re-
ing international opportunities. quired for competitiveness in global and domestic markets. Most
The construction industry has allocated significant effort and safety management systems are dependent on the situational aware-
resources to promote the safety and well-being of its workers. ness of the workforce, which requires strong hazard recognition
skills, reasonable risk perception and tolerance, and risk-averse
1
Ph.D. Candidate, Dept. of Civil, Environmental, and Architectural En- decision making (Hallowell 2010). Among these elements of situa-
gineering, Univ. of Colorado at Boulder, 1111 Engineering Dr., Boulder, tional awareness, risk tolerance remains comparatively elusive
CO 80309 (corresponding author). ORCID: https://orcid.org/0000-0002 with relatively less scientific literature. In the context of global
-4692-8416. Email: rico.salas@colorado.edu construction organizations, risk tolerance is a key challenge. For
2
Beavers Professor of Construction Engineering, Dept. of Civil, Envir-
example, it is not yet understood how diversity in construction
onmental, and Architectural Engineering, Univ. of Colorado at Boulder,
1111 Engineering Dr., Boulder, CO 80309. Email: matthew.hallowell@ workers’ backgrounds, demographics, and national cultures impact
colorado.edu risk tolerance. Disparities in risk tolerance at the individual worker
3
Professor, Dept. of Civil, Environmental, and Architectural Engineer- level may create barriers for training, hazard avoidance, and other
ing, Univ. of Colorado at Boulder, 1111 Engineering Dr., Boulder, safety efforts. Thus, in the design of safety management systems,
CO 80309. Email: balajir@colorado.edu individual risk tolerance should be considered to achieve the
4
Assistant Professor, Dept. of Civil and Construction Engineering, desired safety objectives (Robson et al. 2007).
Western Michigan Univ., 1903 W. Michigan Ave., Kalamazoo, MI 49008. In terms of individual situational awareness, risk tolerance is
Email: siddharth.bhandari@wmich.edu
defined as an individual’s subjective understanding of and behavior
Note. This manuscript was submitted on May 8, 2019; approved on
August 29, 2019; published online on February 6, 2020. Discussion period regarding danger that could possibly lead to injury. Because of its
open until July 6, 2020; separate discussions must be submitted for indi- subjective nature, risk tolerance is nuanced and challenging to
vidual papers. This paper is part of the Journal of Construction Engineer- clearly define, measure, and manage. Understanding cross-cultural
ing and Management, © ASCE, ISSN 0733-9364. differences in risk tolerance could help to catalyze a shift in cultural
Understanding Risk Perception and Tolerance as There are many definitions of culture, and the landscape is
Sociocultural Constructs changing, but for the purpose of this research it is the collective
Risk is a measure of potential loss, typically expressed in terms of mechanisms of managing risk and uncertainty. Understanding the
the product of accident likelihood and magnitude of loss (Eaton sociocultural constructs of individual risk tolerance, in relation to
and Little 2011; Pinheiro et al. 2011). Risk is often associated with situational awareness, help in the framing and strategizing of this
the possibility that an undesirable state of reality may occur due research and explaining the risk tolerance of construction workers
to natural events or human actions (Reason 1997). Because safety across national cultures. This study focused on risk tolerance but
often involves making decisions under uncertainty, safety risk risk perception was considered to be its antecedent.
management is not simply an objective measurement of danger.
Rather, each social group and each individual within the group Risk Perception and Tolerance in the Construction
may have highly varied risk tolerance based on culture, personal Industry
experiences, and rationality (Reason 1997; Rohrmann and Renn
2006; Rundmo 2011). Construction workers generally consider construction work to be
There is an element of risk inherent in all construction safety risky in nature (Man et al. 2017). Some studies have identified
decisions because of the uncertainty associated with dynamic con- that a lack of safety awareness, work pressure, coworker attitudes,
struction sites. Endsley and Jones (2004) showed that situational and socioeconomic factors are reasons for risk-taking behaviors
awareness is an essential element in decision-making processes (Choudhry and Fang 2008; Hinze and Gambatese 2003). A com-
and comprises three distinct levels: perception, comprehension, mon finding is that there are differences in risk perception
and projection. Perception is the processing of information, status, across organizational levels (e.g., management versus workers)
attributes and dynamics of relevant elements in the environment. and national cultures (Hallowell 2010; Korkmaz and Park 2018;
Then, comprehension involves the synthesis and integration of the Perlman et al. 2014). For instance, Hallowell (2010) found that
perceived information through the processes of pattern recognition, workers have a much higher risk tolerance than their managers
interpretation and evaluation to understand how it will impact the in the US Pacific Northwest. Additionally, Korkmaz and Park
individual’s goals and objectives. Finally, projection is the ability of (2018) suggested that differences in the risk perception of South
individuals to predict future events and action in their environment Korean and foreign construction workers are influenced by their
(Endsley and Jones 2004). management’s commitment to safety.
With the situational awareness model in mind, risk recognition Such dissimilarities in risk perception are influenced by safety
policy and standards, management norms, workgroup norms, and
is the identification of hazards that are present or can be anticipated
personal attitude toward risks (Choi et al. 2017; Hallowell 2010;
(Pinheiro et al. 2011). The identified risks are processed through
Korkmaz and Park 2018). Choi et al. (2017) shows that individual
perception. Risk perception is defined as an individual’s subjective
risk perception is manifested through a perceived workgroup norm,
judgment of the frequency and severity of particular risks through
which mediates the relationship between the management norm
processing of information restricted by a known space and period
and safety and risk-taking behavior. Individual risk perceptions re-
of time (Hallowell 2010). Thus, risk tolerance is the personal
garding decisions and/or actions are influenced by information and
appraisal of risk and corresponding feelings of comfort or discom-
feedback from others.
fort given the risk. These constructs include the extent and control
Risk perception in the construction industry is well-researched
of uncertainty and confidence in the evaluation of the situation
as evidenced by the reviewed literature. The literature pertinent to
(Williams and Noyes 2007).
risk perception relies on the understanding of concepts, with a fo-
Risk tolerance relates to the quantity of risk, qualitative features
cus on workplace safety contexts (i.e., organizational structure, pol-
of hazards, perceived benefits of risk taking, personal relation to the
icies, and work experience) and the benefit/payoff. However, the
hazard, and feelings of acceptability. Risk tolerance is driven by
understanding of risk tolerance, as a mediator of risk perception
individual circumstances such as the type of work or form of a haz-
and safety actions, is limited. In addition, the available literature
ard; personal characteristics such as affective associations, expo-
provides limited evidence of the shaping of and disparities in risk
sure history, and feelings of control; perceived consequences
tolerance in construction organizations and projects in broader indi-
such as the probability of injury or death; and social and cultural
vidual and sociocultural contexts.
factors such as having an economic perspective, skepticism and
safety culture (Rohrmann and Renn 2006). How an individual re-
acts to the situation depends on how they see controllable situations Point of Departure
as opportunities or how they tolerate uncontrollable situations
(Krueger and Dickson 1994). Thus, every decision is made with Existing research efforts have focused on the understanding of risk
a subjective assessment of risk and is often followed by a selection perception in the context of workplace safety and culture, such as
complete the survey while others completed the survey but with (Thompson 2010). The PCA methodology serves two distinct
missing responses. These observations were also dropped from purposes: data summation and data reduction. Data summation
the sample. After data cleaning and processing, the sample com- analysis groups a highly correlated set of variables into a
prised 11,997 workers from 17 countries. The demographic singular component (James et al. 2013). Data reduction
dispersion of the sample was large and geographically dispersed analysis identifies a set of representative variables for each
and the included countries have different economic and sociocul- component that can be combined to produce a set of components
tural norms and systems. Fig. 2 and Table 1 (Panel A) show the representing the original much larger set of variables (James
distribution of the sample by country. et al. 2013).
The majority of the respondents were nonsupervisory workers PCA was applied to the following variables, with the unit of
(80%) who have formal training. They skewed young in age analysis of an individual worker, as described in Table 2: (1) Indi-
(20–49 years) with 3–15 years of work experience. Most were mar- vidual Context—Affective Association and Controllability Belief
ried (72%) and had children (62%). Among these workers, 89% did (variable codes IA1–IA5 and ICo1–ICo7, respectively); (2) Social
not report experiencing work-related injuries. Only 5% reported and Cultural Context—Societal View and Safety Culture (variable
experiencing work-related injuries beyond first aid (i.e., medical codes SE2–SE3 and SS1–SS11, respectively); and (3) Attitude to-
treatment or a day away from work). Panels B and C of Table 1 ward Risk Taking (variable code AR1–AR2). Principal components,
show the descriptive statistics of the sample. comprising correlated determinants, were identified using the
Kaiser criterion. The Kaiser criterion was used to drop all compo-
nents with eigenvalues equal to or less than 1.0 which is the eigen-
Analytical Approach value equal to the information accounted for by an average single
R Core Team (version 2019), an open-source software environment item (Thompson 2010). The potential components (i.e., high load-
for statistical computing and graphics, was used to analyze the data ing) were examined using Cronbach alpha in establishing an effec-
(R Project version 2018). To explore the individual characteristics tive representation of variables. The value of the Cronbach alpha
and sociocultural context of the construction worker’s risk toler- varies and the higher the positive number (at least 1.0), the more the
ance, a global principal component analysis (PCA) was conducted. set of variables being tested is correlated with the latent factor
PCA is used as a data aggregation and variable reduction technique (Greene 2000). Then, the averages of the loading factors of the
and the resulting principal components represent the original larger highest loading components from the global PCA were calculated
set of fractional variables. PCA involves a linear/nonlinear mapping and assigned as global risk-tolerance scores. There are no good or
of the data to a lower-dimensional space such that the variance of bad risk-tolerance scores as these numbers were derived from latent
the data in the low-dimensional representation is maximized variables (i.e., components) represented by unobserved constructs
(James et al. 2013). In PCA, latent variables represent unobserved (Thompson 2010). This method is commonly used in ranking in-
constructs and are referred to as components or dimensions stitutions (e.g., top universities, colleges, and hospitals) or countries
Belief and Safety Culture determinants with the following high Figs. 4(a and b) and Table 4 show the local risk-tolerance scores
loading variables: coworker/friend work-related injuries (code: by country based on Components 1 and 2. Singapore (2.147) and
IA2), control of safety (code: ICo3), breaking critical rules (code: South Korea (2.111) ranked highest in the local risk-tolerance score
ICo6), reasons for breaking critical rules (code: ICo7), following in the first component of the local PCAs. Colombia (1.641) and
rules and procedures (code: SS1), high risk work without training Turkey (1.654) had the lowest scores in the first component.
(code: SS4), working at heights without fall protection (code: SS6), In the second component, Singapore (1.377) ranked first while
safety rules and training (code: SS7), coworker unsafely working Spain (0.918) was last. This result indicates that individuals in the
(code: SS9), PPE availability (code: SS11), and risk choices higher-ranked countries have a higher tolerance for risk while those
(code: AR2). from lower-ranked countries have a lower risk tolerance.
The high loading components from global PCA identified the The K-means clustering grouped Mexico, Colombia, Brazil,
important determinants of risk tolerance of construction workers Chile, and Spain into the first cluster based on their local risk
from 17 countries. The results suggest that risk tolerance is highly scores; Australia, Japan, Korea, Hong Kong, Singapore, and China
influenced by Affective Association, Control Belief, Safety in the second cluster; and Turkey, India, Thailand, Malaysia,
Culture, and Risk-taking Attitudes. Indonesia, and Vietnam were in the third cluster. Note that these
clusters were analytically created and not directed by the research
team. The K-means clustering of Component 2 grouped the risk-
Local Risk-Tolerance Scores tolerance scores of the sampled countries into three clusters which
PCA was performed on the represented 17 countries using their coincide with the geographical regions (i.e., Asia Pacific, Latin
country-specific determinants with the mean of the loading factors America, and Europe). Figs. 5(a and b) show the K-means
in each component calculated and assigned as their local risk- clustering results of Components 1 and 2.
tolerance score. As previously discussed, there are no good or The local PCA identified the important determinants of risk
bad risk-tolerance scores. The scores are a representation of the tolerance in each country; however, there were disparities in the
unobserved constructs of the determinants. However, the scores risk tolerance when the scores were compared. Interestingly,
provide the rank (for this case, the represented countries) of the the clustering of country-specific risk scores indicates that the char-
latent variables (i.e., components). The analysis generated from low acteristics and sociocultural background are influenced and shared
to high risk-tolerance scores. not only within the country but also among several countries.
Fig. 4. Local risk-tolerance scores using (a) Component 1; and (b) Component 2.
rules and procedures training (code: SS7), coworker unsafely unsafely working (code: SS9), risk-taking behaviors (code:
working (code: SS9), and risk choices (code: AR2). The risk- AR1), and risk choices (code: AR2). The results indicate that
tolerance scores of Cluster 3 countries were influenced by the there were determinants common to the countries within a
following: training on high-risk work (code: SS4), coworker cluster that influence the risk tolerance of construction workers.
Fig. 5. K-means clustering of (a) Component 1 risk-tolerance scores; and (b) Component 2 risk-tolerance scores.
Hence, multinational construction companies can better strategize Aside from strengthening structural policies and building safety
to customize plans and programs in implementing a safety manage- cultures, targeted interventions, such as education, training and
ment system based on geography, national culture, subcultures, and engagements, are needed in countries where risk tolerance is
social norms. high. Though some countries may have regimented policies and
adequate equipment, tools, and PPE, workers may still be vulner- organizations better understand, identify, and manage safety
able to serious injuries or fatalities. Continuous communication and risks. It may also accelerate the development of a robust safety
engagements are needed to address risky behaviors that may cause culture.
affective reaction to risk. Equally important, workers in countries
with low risk tolerance will require similar interventions, in
addition to adequate equipment, tools, and PPE, in performing Study Limitations
their jobs. In locations where risk tolerance is high, construction
safety policies should focus on compliance education/engagements There are some important limitations within the data and as-
and incentives. However, in locations where risk tolerance is sumptions employed in this research. First, the data are mostly
low, policies should focus on access to proper equipment/tools, representative of building and services construction companies
plans/procedures, and safety education/training. Again, these resulting in the possibility that small enterprises are underrepre-
construction safety policy interventions and resource allocation sented. The survey and data collection were administered by the
may change depending on the location, culture, and background owners and/or representatives of the construction companies,
of workers and their mechanisms in addressing risk and which may have resulted in some biases or failure to generalize
uncertainty. Hence, broadening the suite of structural policies to other populations. To address these issues, appropriate data
considering worker’s risk tolerance may help construction analysis techniques were applied to test the consistency and
Korkmaz, S., and D. J. Park. 2018. “Comparison of safety perception be- Rohrmann, B., and O. Renn. 2006. “Risk perception research: An introduction.”
tween foreign and local workers in the construction industry in Republic In Cross-cultural risk perception: A survey of empirical studies, 13–43.
of Korea.” Saf. Health Work 9 (1): 53–58. https://doi.org/10.1016/j Boston: Springer.
.shaw.2017.07.002. Rundmo, T. 2011. “Employee images of risk.” J. Risk Res. 4 (4): 393–404.
Krueger, N., and P. Dickson. 1994. “How believing in ourselves increases https://doi.org/10.1080/136698701100653259.
risk taking: Perceived.” Decis. Sci. 25 (3): 385. https://doi.org/10.1111/j Sitkin, S. B., and A. L. Pablo. 1992. “Reconceptualizing the determinants
.1540-5915.1994.tb01849.x. of risk behavior.” Acad. Manage. Rev. 17 (1): 9. https://doi.org/10.5465
Link, B. 2008. “Epidemiological sociology and the social shaping of pop- /amr.1992.4279564.
ulation health.” J. Health Social Behav. 49 (4): 367–384. https://doi.org Thompson, B. 2010. Exploratory and confirmatory factor analysis.
/10.1177/002214650804900401. Washington, DC: American Psychological Association.
Man, S. S., A. H. S. Chan, and H. M. Wong. 2017. “Risk-taking behaviors US Census. 2017. Construction spending: Historical value put in place.
of Hong Kong construction workers: A thematic study.” Saf. Sci. Washington, DC: US Census.
98 (Oct): 25–36. https://doi.org/10.1016/j.ssci.2017.05.004. Williams, D., and J. Noyes. 2007. “How does our perception of risk
Oxford Economics. 2018. The economic significance of the meetings and influence decision making? Implications for the design of risk informa-
events industry to the U.S. economy, 1–52. Oxford, UK: Oxford tion.” Theor. Issues Ergon. Sci. 8 (1): 1–35. https://doi.org/10.1080
Economics. /14639220500484419.
Perlman, A., R. Sacks, and R. Barak. 2014. “Hazard recognition and risk Zhao, D., A. P. McCoy, B. M. Kleiner, T. H. Mills, and H. Lingard. 2016.
perception in construction.” Saf. Sci. 64 (Apr): 22–31. https://doi.org/10 “Stakeholder perceptions of risk in construction.” Saf. Sci. 82 (Feb):
.1016/j.ssci.2013.11.019. 111–119. https://doi.org/10.1016/j.ssci.2015.09.002.