Statistical Analysis on Factors in Reducing Construction Site Accident
Frequency Rate in Hong Kong
W F Poon
Labour Department, The Government of the Hong Kong Special Administrative Region
and
Gary C.H. Ma
Industrial Centre, Hong Kong Polytechnic University
and
Dr. John K.L. Ho
Department of Manufacturing Engineering and Engineering Management,
City University of Hong Kong
Abstract
A set of Factories & Industrial Undertaking (Safety Management) Regulations is going through
the legislation process and will soon be introduced to Hong Kong. The new Regulations will
require employers to implement 14 safety elements in a safety management system. This paper
describes the investigation on the effectiveness of the 14 elements, if the elements are applied to
reduce construction site accident rate in Hong Kong.
The investigation was carried out by two parts. The first part of the statistical analysis was
utilised the independent sample t-test to evaluate the relationship between the means of
accident frequency rate and the corresponding safety element. The second part of analysis
applied the multiple linear regression models to examine the combined effect of the safety
elements in reduce site accident frequency rate. The findings and the recommendations are
drawn at the end of the paper.
Key words: safety legislation, statistical analysis, site accident frequency rate
Introduction
The number of construction related accidents in Hong Kong remained at a very high level.
Although, the construction industry employed less than 8% of the total industrial workforce, it
had consistently accounted for more than one third of all industrial accidents over the last ten
years. The situation was unacceptable.
The Hong Kong government, at its 1995 consultation paper, suggested changing from the
traditional enforcement approach to a safety management system approach in tackling
workplace safety and health issues.
A 14-elements safety management system framework is proposed under the drafted Factories &
Industrial Undertaking (Safety Management) Regulations. The 14 elements are:
1.
2.
3.
4.
A safety policy which states the commitment of the enterprise to safety and health at work
A structure to assure implementation of the commitment to safety and health at work
Training to equipment personnel with knowledge to work safely and without risk to health
In-house safety rules to provide instruction for achieving safety management objectives
5. A programme of inspection to identify hazardous conditions and for the rectification of any
such conditions at regular intervals or as appropriate
6. A programme to identify hazardous exposure or the risk of such exposure to the workers
and to provide suitable personal protective equipment as a last resort where engineering
control methods are not feasible
7. Investigation of accidents or incidents to find out the cause of any accident or incident and
to develop prompt arrangement to prevent recurrence
8. Emergency preparedness to develop, communicate and execute plans prescribing the
effective management of emergency situations
9. Evaluation, selection and control of sub-contractors to ensure that sub-contractors are fully
aware of their safety obligations and are in fact meeting them
10. Safety committees to identify, recommend and keep under review measures to improve the
safety and health at work
11. Evaluation of job related hazards or potential hazards and development of safety
procedures
12. Promotion, development and maintenance of safety and health awareness in a workplace
13. A programme for accident control and elimination of hazards before exposing workers to
any adverse work environment
14. A programme to protect workers from occupational health and hazards
The 14 elements safety management system is new to Hong Kong construction industry. This
research investigated the effectiveness of these 14 elements in reducing the site accident
frequency rate through questionnaire survey.
Questionnaire Survey
Literature reviewed have indicated that factors affecting site accident frequency rate were:

Company characteristics;

Safety management process elements; and

Human behavioural factor towards safety.
For the development of the questionnaire, emphasis was made only on the company
characteristics and safety management process elements. Human behavioural factor that
mainly reflecting personal altitude towards safety was excluded because of the irrelevant to the
research objective. The survey questionnaire, embedded with 14 elements of the safety
management system framework, was divided into two parts:

The first part of the questionnaire was structured to collect background information of the
construction companies. Information, such as types of construction companies, nature of
work conducted, size of employees, workers’ turnover rate and percentage of
subcontracting, would be collected.

The second part of this questionnaire was devised to collect information about the methods
used by the construction companies and their respective site accident rate.
200 construction companies listed in the Hong Kong Construction Association Ltd. were
randomly selected for the questionnaire survey. 62 companies completed and returned the
questionnaire. The result of the questionnaire survey was tabulated in Table 1.
Table 1: Results of questionnaire survey
Survey questions
Small company size (20 or few employees)
Medium company size (21-99 employees)
Large company size (100 or more employees)
0 – 20 % of subcontracting work
21 – 40 of subcontracting work
41 – 60 of subcontracting work
61 – 80 of subcontracting work
81 –100 of subcontracting work
Front line management responsible and accountable for safety
Middle management responsible and accountable for safety
Top management responsible and accountable for safety
With company safety policy in place
Without company safety policy
No safety organisation
Having a safety organisation but without clear responsibility
Having a safety organisation with clear responsibility
No safety committee
Ad hoc committee after accident
With a standing safety committee
No safety audit will be conducted
Audit in every 6 months
Audit in every year
Without any emergency preparedness programme at all
With an emergency preparedness programme
Without any health assurance programme
With health assurance programme in place
Without accident investigation programme at all
With accident investigation programme in place
No job hazard analysis programme in place
Job hazard analysis involved with general checklist only
Job hazard analysis with a comprehensive programme
No safety orientation at all
Safety orientation with handouts only
Safety orientation with a comprehensive course
Without personal protection programme at all
With personal protection programme in place
Without any process control programme at all
With process control programme in place
No safety promotional programme at all
Safety promotional programme attended by the line management
Safety promotional programme attended by the top management
Without any accident reporting programme
With accident reporting programme in place
Without any selection of subcontractors programme
With selection of subcontractors programme in place
Without any in-house safety rules
With in-house safety rules
Table 1: Summary of survey results (continue)
Response
9
25
28
13
11
9
19
10
20
24
18
54
8
20
23
19
16
25
21
33
22
7
35
27
55
7
29
33
54
7
1
28
20
14
47
15
57
5
26
27
9
3
59
61
1
39
23
Average
accident
frequency rate
per 1000
workers
246.59
235.95
90.07
48.48
103.64
187.22
207.69
323.87
303.01
155.21
47.49
157.10
269.58
299.10
148.63
65.25
298.81
184.47
59.38
241.21
106.75
47.40
228.85
97.42
178.00
121.44
275.32
80.48
190.69
47.32
11.56
272.07
122.32
41.13
207.84
58.13
181.50
58.88
270.57
120.20
39.98
383.94
160.82
172.96
89.37
201.29
121.30
Survey questions
General inspection programme without any report and follow-up action
Goal setting inspection programme without any report and follow-up action
General inspection programme with report and follow-up action
Goal setting inspection programme with report and follow-up action
No safety training at all
Safety training for some supervisors
Safety training for all supervisors
Regular safety training to all members of the project team
Special training provided in addition to regular training
Response
24
3
19
16
26
9
5
11
11
Average
accident
frequency rate
per 1000
workers
273.93
131.74
155.59
44.65
277.75
164.53
126.44
81.91
36.78
These data were taken for statistical analysis to identify the essential elements that would
contribute to the reduction of the accident frequency rate.
Statistical Analysis Methods
The selections of statistical analysis methods were based on the nature of the collected data and
the expected outcomes of the analysed result.
Independent Sample t-test
Due to the nature of the collected data, independent sample t-test was adopted because of its
capabilities to analyse the ordinal data (i.e. the data have a rank or order all categories according
to some criterions). Independent sample t-test examines the significance in difference in the
means of dependent variable for independent variables.
Multiple Linear Regression Model
Multiple linear regression model evaluates the combined effect of a set of independent
variables in affecting the dependent variable. The concept of multiple linear regression analysis
is identical to that of a simple regression analysis except that two or more independent variables
are used simultaneous to explain the dependent variable. The regression model is in the
following form:
AFR = Constant + B1X1 + B2X2 + …+ BnXn
(Equation 1)
Where Bn is the partial regression coefficient for Xn; and Xn is the independent variables
determine the outcome of the dependent variables.
Stepwise Regression
Due to the practical difficulty to describe numerous variables to include into the regression
model, stepwise regression techniques are used to remove the insignificant variables. The
technique of the stepwise regression is capable to look into different combinations of
independent variables in affecting the dependent variable. The selection criteria are based on
the partial regression coefficients, the coefficients of R2 and the value of the Adjusted R2. The
value of the adjusted R2 is the percentage of variation of the dependent variable that can be
explained by these significant independent variables.
The computer software “SPSS” was adopted to solve the complicated equations.
Results of the Statistical Analysis
The first part of the statistical analysis involved the independent sample t-test, which was used
to compute the relationship between the means of accident frequency rate and the
corresponding safety elements implemented. A 95% confidence level was set as the criteria for
the analysis. It meant that if the calculated P-value was less than 0.05, then the difference
between groups under tested was statistically significant. For the safety element with its entire
test groups were found to have P-value less than 0.05, then the element was found significantly
correlation with the accident frequency rate rate.
The test results and remarks on the elements' significance with the accident frequency rate were
summarised in Table 2.
Table 2: Summary of Independent T-test result
Tested Item
Company size:
 small Vs medium
 medium Vs large
 large Vs small
Subcontracting work:
 0% - 20% Vs 21% - 40%
 21% - 40% Vs 41% - 60%
 41% - 60% Vs 61% - 80%
 61% - 80% Vs 81% - 100%
 81% - 100% Vs 0% to 20%
Management involvement:
 front line management Vs middle
management
 middle management Vs top management
 top management Vs front line management
Safety policy:
 safety policy Vs no safety policy
Safety organisation:
 without clear allocation of responsibility Vs
without any safety organisation at all.
 responsibility clearly defined Vs
responsibility was not defined.
 responsibility clearly defined Vs without
any safety organisation at all.
Safety Committee
 ad hoc safety committee after accident Vs
no safety committee
 ad hoc safety committee after accident Vs
permanent safety committee
 permanent safety committee Vs no safety
committee
Significance in
the Lavene’s test
for the equality
of variance
P-valu
e
Means of
accident
frequency
rate
0.330
0.185
0.073
0.784
0.000
0.000
Not different
Different
Different
0.111
0.132
0.585
0.753
0.007
0.002
0.005
0.552
0.003
0.000
Different
Different
Not different
Different
Different
0.06
0.000
Different
0.004
0.000
0.000
0.000
Different
Different
0.039
0.001
Different
0.074
0.000
Different
0.511
0.000
Different
0.028
0.000
Different
0.729
0.000
Different
0.002
0.000
Different
0.001
0.000
Different
Correlation
with accident
frequency
rate
Insignificant
Insignificant
Significant
Significant
Significant
Significant
Table 2: Summary of Independent T-test result (continue)
Tested Item
Safety audit:
 safety audit every year Vs no safety audit
 safety audit every year Vs safety audit every
six months
Emergency preparedness programme:
 emergency preparedness Vs no emergency
preparedness
Health assurance programme:
 health assurance Vs no health assurance
Accident investigation programme:
 accident investigation Vs no accident
investigation
Job hazard analysis:
 job hazard analysis with general checklist
Vs no job hazard analysis
 job hazard analysis with general checklist
Vs comprehensive job hazard analysis
programme
 comprehensive job hazard analysis
programme Vs no job hazard analysis
Safety orientation programme:
 safety orientation (with handout only) Vs no
safety orientation
 safety orientation (with handout only) Vs
safety orientation (comprehensive course)
 safety orientation programme
(comprehensive course) Vs no safety
orientation
Personal protection programme:
 personal protection Vs no personal
protection
Process control programme:
 process control Vs no process control
Safety promotional programme:
 safety promotion attended by line
management Vs no safety promotion
 safety promotion attended by the line
management Vs safety promotion attended
by top management
 safety promotional attended by top
management Vs no safety promotion
Accident reporting programme:
 accident reporting Vs no accident reporting
Selection of subcontractors programme:
 selection of subcontractors Vs no selection
of subcontractors
In-house safety rules:
 in-house safety rules Vs no in-house safety
rules
Significance in
the Lavene’s test
for the equality
of variance
P-valu
e
Means of
accident
frequency
rate
0.012
0.000
Different
0.247
0.025
Different
0.035
0.000
Different
0.398
0.224
Not different
0.003
0.000
Different
0.001
0.000
Different
0.035
Different
0.116
Not different
0.005
0.000
Different
0.022
0.000
Different
0.001
0.000
Different
0.000
0.000
Not different
0.006
0.000
Different
0.173
0.000
Different
0.017
0.000
Different
0.001
0.000
Different
0.097
0.001
Different
0.476
Not different
0.007
Different
Correlation
with accident
frequency
rate
Significant
Significant
Insignificant
Significant
Insignificant
Significant
Insignificant
Significant
Significant
Significant
Insignificant
Significant
0.182
Table 2: Summary of Independent T-test result (continue)
Significance in
Means of
Correlation
Tested Item
Safety inspection:
 general inspection with no report and no
follow up Vs goal setting inspection with no
report and no follow up
 general inspection with report and follow up
action taken Vs goal setting inspection with
no report and no follow up
 general inspection with report and follow up
action taken Vs goal setting inspection with
report and follow up taken
 general inspection with no report and no
follow up action taken Vs goal setting
inspection with report and follow up taken
Safety training programme:
 safety training for some supervisors Vs no
safety training
 safety training for some supervisors Vs
safety training for all supervisors
 safety training for all supervisors Vs safety
training for the whole project team
 regular safety training for the whole project
team Vs special training in addition to
regular training
 special training in addition to regular
training Vs no safety training.
the Lavene’s test
for the equality
of variance
P-valu
e
accident
frequency
rate
0.074
0.009
Different
0.151
0.619
Not different
0.002
0.000
Different
0.000
0.000
Different
0.688
0.001
Different
0.152
0.316
Not different
0.419
0.023
Different
0.039
0.002
Different
0.001
0.000
Different
with accident
frequency
rate
Insignificant
Insignificant
The second part of the statistical analysis used the same questionnaire survey results to undergo
another statistical analysis with the multiple linear regression models to evaluate the combined
effect of the safety elements in reducing site accident frequency rate. Stepwise regression was
firstly adopted to remove the insignificant variable.
After processing of the stepwise regression analysis, five elements, namely:
(1) accident/incident investigation programme,
(2) safety inspection programme (general inspection programme with report and follow up
action taken),
(3) accident/incident reporting programme,
(4) safety orientation programme (with handout only) and
(5) safety orientation programme (with comprehensive course),
were found to have the combined effects in reducing the site accident frequency rate (Table 3
referred). The R-value calculated for the combination of all these 5 elements was 0.923. Thus
the adjusted R2-value was 0.84 with a standard error of the estimate equalled to 46.1328.
Table 3: Stepwise Regression to Entered/Removed Variables
Model
Variables Entered
1
Accident/Incident Investigation Programme
2
3
Safety Inspection Programme (General inspection
programme with report and follow up action taken)
Accident/Incident Reporting Programme
Stepwise Criteria
Probability-of-F-to-enter <= 0.050
Probability-of-F-to-remove >= 0.100
Probability-of-F-to-enter <= 0.050
Probability-of-F-to-remove >= 0.100
Probability-of-F-to-enter <= 0.050
Probability-of-F-to-remove >= 0.100
Table 3: Stepwise Regression to Entered/Removed Variables (Continue)
Model
Variables Entered
Stepwise Criteria
4
Safety Orientation Programme (with handout only)
5
Safety Orientation Programme (with
comprehensive course)
Probability-of-F-to-enter <= 0.050
Probability-of-F-to-remove >= 0.100
Probability-of-F-to-enter <= 0.050
Probability-of-F-to-remove >= 0.100
These five elements were further undertaken the multiple linear regression analysis in respect
of the accident frequency rate as the dependent variable. The coefficients calculated (see Table
4) were applied to Equation 1 and gave rise to a new formula:
Accident frequency rate = 383.943 – 81.342 A/IIP – 80.163 SINSPD3 – 112.379 A/IRP –
76.098 SORIEND1 – 74.652 SORIEND2
Table 4: Results of the Coefficients
Unstandardized
Coefficients
Bn
(Partial regression coefficient)
(Constant)
383.943
A/IIP
-81.342
SINSPD3
-80.163
A/IRP
-112.379
SORIEND1
-76.098
SORIEND2
-74.652
Model
Standardized
Coefficients
Beta coefficient
Order of the
contribution to
reduce AFR
Collinearity Statistics
Tolerance
VIF
-0.355
-0.307
-0.211
-0.311
-0.273
1
2
3
4
5
0.215
0.237
0.933
0.267
0.249
4.642
4.125
1.072
3.740
4.014
Where A/IIP denotes Accident/Incident Investigation Programme
SINSPD3 denotes Safety Inspection Programme (General inspection programme with
report and follow up action taken)
A/IRP denotes Accident/Incident Reporting Programme
SORIEND1 denotes Safety Orientation Programme (with handout only)
SORIEND2 denotes Safety Orientation Programme (with comprehensive course)
Interpretation of the results
Independent sample t-test
The T-test results indicated that the elements had significant correlation with the site accident
rate were:
(1) safety policy,
(2) safety organisation,
(3) safety audit,
(4) safety committee,
(5) emergency preparedness,
(6) in-house safety rule,
(7) safety orientation,
(8) safety promotional programme,
(9) accident reporting programme,
(10) accident investigation programme,
(11) personal protection programme,
(12) process control programme and
(13) management involvement
Apart from the factor of "management involved on safety", the identified safety elements were
also embraced in the drafted safety management framework.
Stepwise Regression
The adjusted R2 is 0.840 which means 84% of the variation of the accident frequency rate could
be explained by the five process elements in the regression model.
Discussion
The independent sample t-test identified elements that have individual influence on the site
accident rate. However, in the real environment safety elements are not isolated but are
affecting each other. Thus the multiple linear regress model was used to identify the
combination of safety element that would have most effects on the site accident rate. This
argument was later proven correct as the safety inspection program that was excluded by the
t-test came second on the most combined contribution factors list.
In the following context, explanations were given on how and why these 5 factors have
significant relationship with the site accident frequency rate in the Hong Kong construction
industry.
Relationship between accident/incident reporting & investigation programme and the site
accident frequency rate
Accident/incident reporting & investigation programme was found to be the most significant
contributor to reduce site accident frequency rate. It provides useful information to prevent
similar accident in recurring. The information gathered from accident/incident investigations
are also useful and effective mechanism to formulate the corrective actions. However, as the
information could only be gathered after an accident had actually happened or had learned
experience from others, it is only re-active indicator to reflect site safety performance.
Relationship between safety inspection programme and the site accident frequency rate
Safety inspection programme has been identified as the second most effective contributor on
reduction of site accident frequency rate and this result is consistency with many previous
research works. The advantage of safety inspection is to expose potential hazard in the work
site before an accident happens, thus, it is regarded as a pro-active site safety performance
indicator.
However, the effectiveness of safety inspection to combat site accident frequency rate is very
much relying on the comprehensiveness of the inspection programme and the competency of
the personnel carrying out the inspection. In this connection, consideration and emphasis have
been made on the setting of questionnaire. According to the research finding, general safety
inspection programme with report and follow up action is the most effective tool to reduce site
accident rate.
Relationship between safety orientation programme with the site accident frequency rate
Safety orientation programme is particular important to those new workers working in a
construction site.
Commissioner of Labour (1993) stated that new workers are the most vulnerable to site
accidents. It is because new workers are not familiar with site environment and the potential
hazard as well. Hence, there is a need to organise orientation for this high causality group.
From the research finding, it is observed that the effect of safety orientation programme (with
handout only) is better than the programme with comprehensive course. This indicates that no
matter how comprehensive the programme is, it can only improve site accident frequency rate
to a limited extent. As the majority of the new workers to a construction site are coming from
low level of education group, it is difficult to break in with various information to their mind in
a relatively short period of time. On the contrary, simple diagrams or pictures in showing the
site working environment and potential hazard may be more impressive than any other
comprehensive course in such a short duration of time.
In additional to the five significant factors, there are also several interesting observations from
the returned survey questionnaires.
Relationship between management involvement in safety and the site accident frequency rate
The survey result indicates that management involvement in safety is significantly correlated
with the site accident frequency rate. Furthermore, the average site accident frequency rate is
exceptionally good when the top management is directly responsible and accountable for safety
in the work site. This result is expected because top management commitment is the necessary
condition for the success of any safety programme. According to much literature, it reveals that
many accidents are the result of management failure. Hinze and Raboud (1988) found that all
successful safety management programmes must be supported by top management. Hence,
management commitment and support are essential to bring accident rates down.
Relationship between employees’ size of company and the site accident frequency rate
The survey shows that larger companies have generally better average site accident frequency
rate than that of the smaller companies. Although, it cannot prove the significant correlation
between them, it is generally believed that larger companies may have better capability of
combating site accident frequency rate. Possibly, larger companies may have sufficient
resources to implement more structured and formalised safety programmes in enhancing safety
in the work sites.
Relationship between the level of subcontracting work and the site accident frequency rate.
According to the survey the average site accident frequency rate is directly related with the
level of subcontracting work. Lai (1987) attributes the high site casualty rates to the use of
labour-only subcontractors.
In Hong Kong the subcontracted labour is highly mobile. This high mobility of subcontracted
labour, makes the worker less familiar with the site environment and the potential hazard, is a
constituting factor to the high accident rate. Furthermore, due to the financial incapability,
many small subcontractors are unable to implement comprehensive safety programmes. As
such, it is hard to uphold their safety awareness. Besides, the subcontracted labour is rewarded
according to the number of pieces of work that have been completed, therefore, the workers
often make light of safe working practices in the pursuit of productivity.
In addition, due to the deficiency of the existing safety regulations, the principle contractor is
held liable under the majority of the existing construction sites safety regulations. This is
reflected by the number of prosecution made to subcontracted labour is relatively much smaller
than the one to the principle contractor. This information has given a wrong message to the
subcontracted group that site safety is not their responsibility. As such, this subcontracted
labour may make light their legal liability and not co-operated with the principle contractors.
This is the reason why this subcontracted labour is difficult to be controlled and hence
constituted for the high casualty rate in the Hong Kong construction industry.
Conclusion
The new safety management regulation is an effective tool to reduce the site accident rate in the
construction industry. This paper has statistically proved that 12 out the 14 elements of the
proposed safety management system framework were effective tool to reduce the site accident
frequency rate.
The research further identifies accident reporting programme is the third contributor in
reducing accident rate, but this element has not been included in the proposed safety
management system framework. It is recommended that the government to consider this
element in the final version of the safety management system.
Since not all the safety process elements stipulated in the new regulations have significant
correlation with reduction of accident frequency rate, the government is urged to enforce the
new safety management regulation in stages so that it will not become a heavy burden to the
industry.
References
Hinze, J.,and Raboud, P.,(1988) Safety on large building construction projects, Journal of
Construction Engineering and Management, ASCE, 114(2), pp.286-293.
Hong Kong Government (1993) Commissioner for Labour Annual Report, Government Printer,
Hong Kong.
Labour Department (1999) A Guide to Safety Management, Hong Kong
LAI Man-yi (1987) A Review of the Subcontracting Systems in the Hong Kong Construction
Industry, MSc thesis, The University of Hong Kong, Hong Kong.