Content uploaded by Cheuk Hang Au
Author content
All content in this area was uploaded by Cheuk Hang Au on Oct 22, 2021
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=utca20
Journal of Information Technology Case and Application
Research
ISSN: 1522-8053 (Print) 2333-6897 (Online) Journal homepage: http://www.tandfonline.com/loi/utca20
Establishing a three-step model of designing the
polling stations for shorter queue and smaller
waiting time: a case study using computer
simulation
Cheuk Hang Au, Zewen Xu, Lin Wang & Walter S.L. Fung
To cite this article: Cheuk Hang Au, Zewen Xu, Lin Wang & Walter S.L. Fung (2018): Establishing
a three-step model of designing the polling stations for shorter queue and smaller waiting time: a
case study using computer simulation, Journal of Information Technology Case and Application
Research, DOI: 10.1080/15228053.2017.1433494
To link to this article: https://doi.org/10.1080/15228053.2017.1433494
Published online: 05 Feb 2018.
Submit your article to this journal
View related articles
View Crossmark data
Establishing a three-step model of designing the polling stations
for shorter queue and smaller waiting time: a case study using
computer simulation
Cheuk Hang Au
a
, Zewen Xu
b
, Lin Wang
b
, and Walter S.L. Fung
b
a
Department of Decision Sciences and Managerial Economics, The Chinese University of Hong Kong, Shatin, Hong
Kong;
b
Department of Computing, The Hong Kong Polytechnic University, Hunghom, Hong Kong
ABSTRACT
This article tries to address the issue of lengthy queues at the polling
stations during elections using simulation. It is known that the long queues
may harm the voting experience and turn voters away. Moreover, the
queuing problems have been repeating in different countries. We tried to
use computer simulation to address this problem, with a polling station in
the 2016 Hong Kong Legislative Election as the study target. We found and
addressed the bottlenecks of the station successfully by reallocating the
resources in the polling station in the simulations. Based on our results, we
establish a three-step model for shortening the polling stations queues and
waiting time more generally.
Introduction
Polling station queues are getting more common during the elections in many countries (Spencer
& Markovits, 2010), including the United States, the United Kingdom, France, Australia and Hong
Kong. The queues are getting longer and thus have produced congestion and increased the waiting time.
These can discourage voters and harm the confidence of the public in elections. Thus, it has become a
frequent topic of both the government and researchers from different fields to look for a solution
(Murata & Konishi, 2013; Smith, Monfort, & Blumberg, 2015;Stewart,2015). Despite the earlier efforts
of addressing the problem, the applications of simulation software in shortening polling station queues
have not been commonly seen.
We tried to use computer simulation to re-design the polling stations for solving the congestions
and shortening the queues, at a polling station in the 2016 Hong Kong legislative election as our
study target. Our approach has utilized the various advantages of simulations software such as
providing greater insights with lower cost and risk (Hunt, Robson, Lemelin, & Mcintyre, 2010), and
hopefully to provide a solution to polling station congestion with low cost and effort in other
contexts as well. Our case study also helps to explain the complexity of the polling station situations
which may not be captured through other research methods (Zainal, 2007). Based on our findings,
we have established a Three-Steps process model for designing the polling stations.
The paper is structured as follows. Firstly, we present our literature review on polling station
management, computer simulations software and queuing theories and modelling. It is then followed
by the explanation of research method before we describe our proposed simulation design.
Thereafter, we go to the discussion and present our Three Steps model. Finally, we narrate limita-
tions of our study and suggest future research directions followed by conclusion. The structure is
CONTACT Cheuk Hang Au allen.au@link.cuhk.edu.hk Department of Decision Sciences and Managerial Economics, The
Chinese University of Hong Kong, Shatin, Hong Kong
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/UTCA.
© 2018 Cheuk Hang Au, Zewen Xu, Lin Wang and Walter S.L. Fung
JOURNAL OF INFORMATION TECHNOLOGY CASE AND APPLICATION RESEARCH
https://doi.org/10.1080/15228053.2017.1433494
similar to previous studies of building simulations for other contexts, which first built the models
that imitate the existing systems for looking at the current problems, followed by building the
simulation models for possible improvement and testing the models. (Liu, Duan, & Liu, 2016; Zhu,
Zhang, Chu, He, & Li, 2014)
Research question
How can we solve the problem of long queues and congestions at the polling station with the assistance
of technologies such as computer simulations?
It is critical to fix the issues of the long polling station queues, as it would discourage voting and
harm the confidence of the public in the elections (Stewart, 2015). However, such issues have not
been adequately addressed. While voting simulations were employed previously (Murata & Konishi,
2013), a process model is more helpful for practitioners to follow and scholars to study further. The
proposed approach could contribute to shortening the queue and reducing the waiting time with
limited extra cost using different combinations of technologies for future elections.
Literature review
Polling stations queue and management
The management of polling stations, particularly the queues, is frequently studied in electoral studies
(Coles, 2004;Durán et al., in press; Grant, 1980). Long queues in the polling stations may discourage
voting, decrease the voters’confidence in elections and increase the time spent by voters (Murata &
Konishi, 2013; Stewart, 2015). Table 1 shows a selected list of findings about polling stations queues
and management.
In 2012 United States Presidential election, the media widely reported the existence of long queue of
voters. Many voters had to wait for hours after the close of the polling stations to cast their ballots.
Barack Obama, who was re-elected as the US President, remarked on the importance of fixing the
problems in his victory speech (Stewart, 2015). Another example was in 2010 UK General election
when the chaotic queues turned some voters away. The electoral commission blamed the problems on
Table 1. Previous research related to polling stations queues and management.
References Main Findings Empirical/Conceptual
Factors and Consideration of Polling Stations Queues and Management
Stewart (2015) The long queues are due to the discrepancy between the number of
voters who show up and the resources assigned to accommodate them.
Empirical
Coles (2004) The polling stations should have adequate staff to control the access to
the polling station and ensure order and clam.
Empirical
Clark (2015) Over-allocation of voters to some polling stations, unrealistic planning
expectations and inadequate training of staff are common causes of long
queues in polling stations.
Conceptual
Durán et al. (in press) The optimized match between voters and polling stations may be
achieved by using linear programming and queuing theory.
Empirical
Impact of Polling Stations Queues and Management
Smith et al. (2015) Wait times and lines were the greatest barriers to efficient and enjoyable
voting.
Empirical
Elklit and Reynolds (2002) The election administration, including the polling station management,
have a direct bearing on how the sense of political efficacy develops in
individuals, the development of legitimacy and progression towards
democratic consolidation, particularly in developing countries.
Empirical
Kornblith and Jawahar (2005) Long lines in polling stations would create an ominous experience for
the voters and harm their rights.
Empirical
Yang, Fry, Kelton, and Allen
(2014)
The long lines at the polling stations may bring excessive wait times for
voters and charges of voter inequity.
Empirical
Clark (2015) The too long queues in the polling stations may eventually turn the
voters away.
Conceptual
2C. H. AU ET AL.
poor planning and inadequate staffing arrangements and urged for an improvement (BBC, 2015).
Other cases of congestion in polling stations included 2016 Australian Federal Election (Bell, 2016;
Laskie, 2016) and overseas polling stations of 2017 French Presidential election in London (Mitchell,
2017) and Montreal (Wang, 2017).
In other domains, computer simulations have been seen for solving the queue challenges
(Alomair et al. 2015). Thus, it may be possible to utilize simulations software to address the queuing
issues during the election.
About simulation software
Simulations are the imitations of the operation of real-world processes, systems or scenarios over
time, which have been used in different contexts such as testing new business process and technol-
ogies for performance optimization, testing, training, education, and video games (Hunt et al., 2010;
Zeigler, Kim, & PräHofer, 2009).
Advantages of using simulation software include: providing greater insights, identifying the
“trouble spots”or “bottlenecks”and being a less risky and less costly approach for testing the
potential solutions (Hunt et al., 2010; Lawson, Itami, Gimblett, & Manning, 2006). Given the benefits
of using the simulation software, we may see extensive applications in different domains such as
defence, airports, manufacturing, engineering and healthcare (Alomair et al. 2015). Previous scholars
have made different arguments for selecting desired simulation software, which have been listed in
Table 2.
Among different simulation software, Flexsim, MATLAB, ExtendSim, EcoSimPro and
Automation Studio are considered as the more popular options (Krahl, 2009; Lofberg, 2004;
Nordgren, 2003).
Queuing theory and queueing system
Queuing theory is the mathematical study of waiting lines or queues. Based on the theory, models
are built to imitate the characteristics of a part of another system embodying the original, primarily
for predicting queue lengths and waiting time (Leks & Gwiazda, 2015; Sundarapandian, 2009). Such
applications have been seen in different aspects, including medical studies (Wang et al., 2017),
Table 2. Previous argument related to the selection criteria of simulations software.
References Arguments Empirical/Conceptual
Sahay and Gupta (2003) The primary drivers of selecting simulation software include technology,
cost and pricing, features, customisation, and support and services. The
secondary drivers include vendor vision, industry covered and the
strength of vendor.
Empirical
Nikoukaran, Hlupic, and Paul
(1998)
To evaluate simulation software, users should consider the vendor,
model and input, execution, animation, testing and efficiency and output
of the software.
Conceptual
Cochran and Chen (2005) Desired features for Simulation software may be categorized into
programming features, simulation features and environment features.
The decision- making processes may be designed using fuzzy logic.
Empirical
Davis and Williams (1994) The adoption of simulation should consider the cost,
comprehensiveness, integration with other systems, documentation,
availability of training packages, ease of use, hardware and installation
requirement, and confidence-related issues.
Empirical
Tewoldberhan et al. (2002) The consideration of selecting simulation software may be grouped into
different categories, including model development, input, vendor,
execution, animations, testing efficiency, output and user experience.
Conceptual
Banks (1991) It is recommended for the developers to consider the time requirement,
model control, system complexity, output, accuracy, training required
and environment suitability when selecting the simulation software.
Conceptual
JOURNAL OF INFORMATION TECHNOLOGY CASE AND APPLICATION RESEARCH 3
transportation (Wahed, Faghri, & Li, 2017), production (Leks and Gwiazga, 2015) and banking
industry (Zaragoza & Mota, 2016).
An isolated queue consists of one or more servers, together with buffers, which supply room to
accommodate customers and is driven by its traffic. Queue management concerns about the number
of customers in the queue (including customers waiting in line and the customer in service), waiting
time (the time a customer spent for waiting after arrival) and the system time (the sum of the waiting
and service times of a customer) (Li & Li, 2009).
Descriptions of queueing systems often include the system configuration, behaviour and service
discipline of each server, size and arrival patterns of the input (Li & Li, 2009). Among different
queue models, M/M/1 queue model is one of the most popular and elementary ones. It has been
adopted and studied by scholars in queuing management and customer management (Balbo &
Vigliotti, 2013; Haviv & Oz, 2016).
In M/M/1 queue, arrivals occur at rate λaccording to a Poisson process. Service times have an
exponential distribution with rate parameter μ, where 1/μis the mean service time. In more recent
articles, log-normal distribution has become more common in modelling service times (Adedigba,
2005). There is a single server serving one customer at the same time, on a first come, first-served
basis. Upon the service completion, the customer leaves the queue, and the number of customers in
the system reduces by one. The buffer is of infinite size, so there is no limit on the number of
customers it can contain (Haviv & Oz, 2016).
When all other conditions are kept the same, but there are more than one servers, the queuing
model is considered as an M/M/c queue model. Sometimes it is considered as a generalization of the
M/M/1 queue (Harrison & Patel, 1994).
Research method
Considering the advantages of using simulations software in solving queue problem and the research
question (Walsham, 1995), a case study was conducted using simulation. While previous articles
suggested the applications of linear programming and other mathematics-based methods (Durán
et al., in press), simulations involve less statistical knowledge for their applications and provide a
visualized image for users to understand and to illustrate the solutions to other stakeholders (Banks,
1991; Cochran & Chen, 2005). Earlier articles from different disciplines have indicated that compu-
ter simulation may be used as a research method for predicting the outcome or testing the
hypotheses (Dooley, 2001; Gipps, 1981; Stasser, 1988).
Based on our research objectives, two criteria formed the basis of case selection. First, we should
select a polling station that had long queues and serious congestion. It will allow us to develop the
simulations and a process model based on a real case to enhance the potential theoretical and practical
contributions of our study. If we can address the most serious congestion, deductively the less serious
congestion in other polling stations can be solved. Second, we should choose a case where the real
information can be obtained so that we can build the simulations which are closer to the reality. We
chose the Tai Koo Shing West polling station (also known as “Eastern District JPC Club House”polling
station given its location) in the 2016 Hong Kong legislative election for simulation, which was one of
the most congested polling stations in the said election. It was reported that many voters had to wait for
up to four hours after the official closing time of the polling stations. The polling station in Tai Koo
Shing West was said to be one of the most congested polling stations and had drawn attention from the
public. This satisfies our first case selection criterion. Additionally, the Electoral Affairs Commission
(EAC) of the Hong Kong Government has been providing much detailed information online for various
elections in Hong Kong. This satisfies our second criterion.
We first obtained the information from EAC via their website and email requests. Based on the
information obtained and earlier articles, we made some assumptions and simplifications in the
unknown aspects of the polling stations. We then built the simulation to restore the original
situation in the selected polling station. Then, we built the possible improvement models by
4C. H. AU ET AL.
relocating different components (such as polling boxes, ballot issuing tables and voting compart-
ments) and adjusting different parameters (such as time spent in polling boxes and voting compart-
ment, the voters’arrival pattern and turnout rate) in the model. Based on our comparison on
simulation software and the two selection criteria, we have chosen Flexsim software for our research
for simulating and modelling the scenarios, given it has frequently been used for dynamic discrete
event system modeling and visual simulation and monitoring of some system (Liu et al., 2016) and
recommended according to previous research. (Chen, Hu, & Xu, 2013; Qian & Sun, 2013).
Moreover, it is a cheaper option, easier to use and supportive of animations. All these characteristics
are desired for our research. Having designed and run the simulations, we conducted data analysis
based on the simulation results and established our polling station design process model.
Case description
The 2016 Hong Kong legislative election was conducted on 4th September 2016. Each voter was
entitled to one vote in his/her geographical constituency based on his/her registered residential
address, and another vote in the functional constituency.
1
There were 571 ordinary polling stations
and 3,779,085 registered voters. All stations opened at 7.30 a.m. and closed at 10.30 p.m. After the
stations had been closed and all voters in the stations finished casting their votes, most of the polling
stations would be converted into counting stations (Chan, 2017). Most voters have a vote in both
constituencies, but a limited number of them opted out in the functional constituency, or had their
functional constituencies having only one valid candidate and thus uncontested (Chan, 2017;
HKGov., 2016).
The large election turnout and the long queue
The election turnout was the highest since the first legislative election in 1991 (39.15%) with
geographical constituencies at 58.28% (or 2.2 million) (HKGov., 2016). It might be explained by
various incidents and factors, such as the rise of the localist camp
2
after Umbrella Revolution
3
in
2014 (Chung, 2014), the Causeway Bay booksellers’disappearance
4
(Chan, 2016), the Project
ThunderGo,
5
Leung Chun-ying–UGL agreement controversy,
6
ban on pro-independence candidates
controversy
7
and pro-democracy candidates’suspension of electioneering
8
(Carey, 2017).
Thus, the high turnout rate might be an explanation of the long queues. There were long voters’
queues at some of the polling stations, including the one in Tai Koo Shing West and Belcher in Hong
Kong Island, as well as in Cheung Kwan O and Tai Wai in New Territories East (SCMP, 2016). In
Taikoo Shing West, it was reported that about 1400 voters needed to wait after the official closing
1
In the political systems of Hong Kong, a functional constituency (FC) is a professional or special interest group involved in the
electoral process. Eligible voters in a FC may include natural persons or other designated legal entities.
2
The Localist camp is a loose term referring to various political groups that advocates Hong Kong independence in the light of the
conflicts between Hong Kong and China.
3
Triggered by Beijing’s restrictive decision the Hong Kong electoral reform, Umbrella Revolution refers to a series of street protests
in Hong Kong in late 2014.
4
The Causeway Bay booksellers’disappearances are a series of international disappearances concerning five staff members of
Causeway Bay Books. It was widely believed that they were taken away and detained by the China Government.
5
Introduced by Benny Tai, a pro-democracy scholar, The Project ThunderGo was to target for achieving a majority for the anti-
establishment force in the legislative council in 2016.
6
In October 2014, it was reported that Leung Chun-ying, Chief Executive of Hong Kong, had signed an agreement with UGL, an
Australian engineering firm, in relation to its takeover of DTZ Holdings, a UK company in which Leung was the company director.
UGL undertook to pay Leung £4 million. The public are concerned about the nature of payment, potential conflict of interests,
relevant systems of declaration of interests and taxation implications.
7
A controversy sparked during the 2016 Hong Kong Legislative Council election as the government banned six potential localist
candidates from running the election, which was widely seen as screening based on the political point of view.
8
Before the polling day, six pro-democrat candidates and one non-aligned independent openly suspended their campaigns in the
hope of deflecting support to their pan-democrats allies who were seen as standing a better chance.
JOURNAL OF INFORMATION TECHNOLOGY CASE AND APPLICATION RESEARCH 5
time of the polling stations, and the last voters cast their ballots at 2:30 a.m. of the next day, which
was 4 hours after the official closing time.
Previously the voters in Taikoo Shing West were divided into two separate polling stations, but in
2016 they were assigned to the same polling station (Tong, 2016). A local councillor criticized the
arrangement and claimed that over a hundred of voters gave up because of being unable to put up
with the long queue (Chiu, 2016). Such a case may be considered as a problem of over-allocation of
voters previously highlighted by Clark (2015). The impact of long queues in the mentioned station in
2016 was similar to the previous studies about creating the ominous experience for the voters and
harming their rights (Kornblith & Jawahar, 2005; Smith et al., 2015).
Operating manuals and post-election report for legislative election
For each legislative election, the Hong Kong Government would publish a set of operating manuals
for all polling stations staff before the polling day, and the post-election reports to the public
explaining their investigation on electoral-related complaint and evaluation of the elections arrange-
ment after the polling day. According to the manuals, briefing and training sessions were arranged
for polling staff. Instructions of polling stations set-up in the manual covered the detailed require-
ment of different aspects, such as voting compartments, publicity display stands, ballot boxes, ballot-
issuing desks, designated area for candidates and their election agents/polling agents, directional
signs and notices, such as “No Staying Zone”(“NSZ”) signs. Besides, rehearsal should be conducted
after setting up the polling stations.
It was suggested in the manuals that the ballot boxes should be placed near the polling station exit
and, as far as possible. However, there were no suggestions about the exact number and position of the
ballot boxes, voting compartments and ballot-issuing desks within the polling stations. Thus, there
were gaps between the ideal and actual station arrangement, contributing to longer queue and
congestion. The post-election report for the 2004 election highlighted the over-crowdedness or
prolonged queuing outside the polling stations. Fourteen related complaints were launched, with the
related polling stations located in different places. (Electoral Affairs Commission, 2004). Our review
suggested while EAC assigned adequate staff in the polling stations in previous elections, the 2016
election was not the first occurrence of serious congestion. Apparently, the EAC has not adequately
addressed the queues at the polling stations.
Government information and figures for the polling in taikoo shing west
According to the Electoral Affairs Commission (EAC), 8949 registered voters were assigned to the
polling station in Tai Koo Shing West in 2016, with 7246 of them (81.0%) being a voter in District
Council (Second) Functional Constituency. 5493 voters (61.38%) cast their vote in the station.
In the polling station, there were four ballot-issuing desks, eight ordinary voting compartments,
one voting compartment for wheelchair users, and a set of ballot boxes. In 2012, there were three
ballot-issuing desks, six ordinary voting compartments, one voting compartment for wheelchair
users, and a set of ballot boxes. However, the reasons for such combination were not provided in the
manuals or the replies of our email from EAC.
The voting process
The ballots were in paper form, and the steps of voting were as follows,
(1) After entering the designated polling stations, voters showed their Hong Kong identity cards
for verifying identifications.
(2) Voters obtained the ballot and a piece of cardboard with a chop bearing a”✓”from polling
station staff.
6C. H. AU ET AL.
(3) Voters got into a voting compartment and chopped on the circle of their chosen candidate
(or candidate list) with the given chop on the ballot.
(4) If voters made any error in marking a ballot or advertently spoiled a ballot, they might
return that ballot to the Presiding Officer and ask for a replacement.
(5) Voters folded the ballots once according to the pre-folding by polling staff (if any), and put
the ballot into ballot box of the corresponding constituencies. The ballot for geographical
constituencies should be put into the blue ballot box; the ballot for District Council (Second)
functional constituency should be put into the white ballot box, while the ballot for other
functional constituencies should be put into the red ballot box.
Unlike the elections in other countries where early voting or proxy arrangements may be
available, all voters in Hong Kong must physically visit the polling station in person to cast their
votes (HKGov, 2016).
Assumptions of simulation based on case scenario
It is necessary for us to make some assumptions because some data are not available or cannot be
collected by the government due to the election confidentiality issues.
We assumed the voters’arrival pattern followed an exponential distribution, while the services time of
ballot issuing desks followed a lognormal distribution (Adedigba, 2005). The voters’inter-arrival time, as
the parameter of the exponential distribution of the arrival pattern, depended on the turnout.
We assumed that when the voters arrived at the polling station, they would go to the ballot
issuing desks with the shortest queue. . The mean and the standard deviation for services time of
ballot issuing desks were 60 seconds and 6 seconds respectively.
For the time of chopping on the ballot and the time of putting the ballot into the boxes, we
assumed a normal distribution applied, with the mean and standard deviation values shown in
Table 3:
Voters could choose any voting compartments available after getting the ballot as long as the
compartment is free. We could only assume above time figures because the government could not
collect such data due to confidentiality issues. We also assumed that the proposed changes would
make no effect on the candidates and their representatives as observers in the polling stations.
Results
Model for initial design of the polling station
We considered the polling station queues as an isolated, multistage queueing system. Firstly, we built
the simulation to represent the situation of Tai Koo Shing West polling station (As shown in
Figure 1)
The layout was based on the information from the government and site inspection. The voters
entered the station from the left door, line up in the first queue, and went to the queues in front of
ballots issuing desks. After registration, they would get the ballots and went to the voting compart-
ments. The process is shown in Figure 2.
Table 3. The assumed time requirements of chopping the ballot and putting the ballot into boxes (in
Seconds).
Mean Standard Deviation
Chopping on ballot (ordinary voters) 25 2
Chopping on ballot (wheelchair voters) 35 3
Putting the ballot into the boxes 8 1
JOURNAL OF INFORMATION TECHNOLOGY CASE AND APPLICATION RESEARCH 7
The process of going to voting compartments is shown in Figure 2. There was a queue area in
voting compartments.
After chopping on the ballot (as shown in Figure 3), the voters would go to the voting box to put
the ballots and leave (as shown in Figure 4).
Only the eligible voters would use the voting compartment meant for wheelchair users, as shown
in Figure 5:
We assumed that 5496 voters would arrive within the official open hours of the polling station
(15 hours, or 54000s). We tried to restore the original situation in the station, and we assumed 300 of
them were wheelchair users. The average inter-arrival time of ordinary voters is 15 hours × 3600
seconds ÷ 5196 voters = 10.392 seconds. The average inter-arrival time of wheelchair voters is
15 hours × 3600 seconds ÷ 300 voters = 180 seconds. Figure 6 shows the simulation of the initial
model.
Figure 1. Layout of the initial model.
Figure 2. The image of leaving the ballots issuing tables.
8C. H. AU ET AL.
Stage 1 was the stage when the voters entered the polling stations, while stage 2 is the stage of
waiting at the ballots issuing tables. Stage 3 is the stage of waiting to get into the voting compart-
ment, while Stage 4 is the stage of queueing for putting the ballots into the boxes.
Figure 3. Chopping in the voting compartment.
Figure 4. The image of putting the ballots into voting box.
Figure 5. Voting area only for wheelchair users.
JOURNAL OF INFORMATION TECHNOLOGY CASE AND APPLICATION RESEARCH 9
“TotalNum of Exit Voter”represents how many people have voted, while “TotalNum of Entry
Voter”represents the number of voters who came to the station.
Tables 4 and 5show the statistical results of the initial models.The sum of average waiting time in
4 stages equals to the waiting time that a voter needs to wait before completing the whole voting
process. According to the average content and the waiting time information, we found that in stage 1
the waiting time was unreasonable and indicated a high level of congestion.
The first improved model
Based on the initial model, we found the bottleneck when issuing ballots. Voters had to wait for a
long period before getting the ballots and registration. However, after getting the ballots, voters did
not need to wait for getting into a voting compartment. There were too few ballot-issuing desks, but
too many voting compartments. As a result, we established a new model of six ballot issuing desks
and six ordinary voting compartments. The voting compartment for wheelchair users was kept.
Figure 7 shows the the layout of the first improved model, while Figures 8 and 9show the
simulations of the first improved model.
We created three scenarios differed by turnouts for simulations to test the capacity of the
improved model. The simulation time for these scenarios are the same (i.e., 15 hours, 7.30 a.m. to
Figure 6. Simulation of the initial model.
Table 4. Statistical results of the initial model –No. of voters.
No of Voters
VotingBox1 3587
TotalNum of Exit Voter 3586
TotalNum of Entry ordinary Voter 5216
TotalNum of Entry wheelchair Voter 298
No. of Voters that cast their vote after the closing time of polling station 1927
10 C. H. AU ET AL.
10.30 p.m.), and there were 300 wheelchair voters for these scenarios. The other simulation settings
are shown in Table 6.
The definitions of the stages are the same as that of the initial design, except stage 2 refers to
queue 2–7, and stage 3 refers to queue 8–14 in the 1st improved model.
Tables 7 and 8show the statistical results of the first improved model. The average waiting time,
the average number of voters in the queues, and the number of voters that had to vote after the close
of polling station significantly dropped. The statistical results demonstrated that the “1st improved
model”model could shorten the queue and the waiting time, and thus could handle cases smoothly
for turnout not higher than 70%. However, as the turnout increased to 80%, the congestion occurred
again.
The second improved model
To address the congestion in the 1st version when the turnout reaches 80%, we modified our 1st version
by introducing an additional set of ballot boxes. We did not add new voting compartments or ballot-
issuing desks because the area size of the polling station may not allow. Besides, the official information
indicated that there would be only one set of ballot boxes in most scenarios. We proposed an additional
set of ballot boxes would occupy less extra space and thus can be accommodated more likely.
Table 5. Statistical results of the initial model –average waiting time and average content.
Average Content (No. of Voters
waiting)
Average Waiting Time (By Queue,
in seconds)
Average Waiting Time (By Stage,
in seconds)
Stage 1 Queue1 891.9 8576.2 8576.2
Stage 2 Queue2 19.1 1130.4 1131.6
Queue3 19.1 1137.6
Queue4 19 1128.8
Queue5 19 1129.8
Stage 3 Queue6 0 1.9 2.3
Queue7 0 2
Queue8 0 1.9
Queue9 0 2.5
Queue10 0 2.4
Queue11 0 1.8
Queue12 0 2.3
Queue13 0 2.3
Queue14 0 3.2
Stage 4 Queue15 0 1.9 1.9
Total 9712
Figure 7. Layout of the first improved model.
JOURNAL OF INFORMATION TECHNOLOGY CASE AND APPLICATION RESEARCH 11
Figure 8. Simulation of the first improved model (1).
Figure 9. Simulation of the first improved model (2).
12 C. H. AU ET AL.
Given that congestion occurs in the 1st model only when the turnout reaches 80%, we would only
test one scenario, which is identical to scenario three tested in the previous model.
Figures 10 and 11 show the layout and simulations of the second improvement model respec-
tively, while Table 9 shows the simulation settings of the second improved model.
The definition of the stages is the same with that of the 1st improved model, except in stage 4 we
have an additional queue (queue 16), which was also for waiting to put the ballots into the ballot
boxes like queue 15.
The average waiting time slightly dropped, but the number of voters that had to vote after the
close of polling station slightly increased.
Tables 10 and 11 show the statistical results of the second improved model. While the above
results may not exclude the option of an extra set of ballot boxes for shortening the queue, such
Table 6. Simulation settings of the first improved model.
Scenarios Turnout
No. of ordinary
voters
Average inter-arrival time (Ordinary
Voters, in seconds)
Average inter-arrival time (Wheelchair
voters, in seconds)
Scenario 1 60% 5196 15 hours × 3600 seconds ÷ 5196
voters = 10.392 seconds
15 hours × 3600 seconds ÷ 300
voters = 180 seconds
Scenario 2 70% 5964 15 hours × 3600 seconds ÷ 5964
voters = 9.054 seconds
15 hours × 3600 seconds ÷ 300
voters = 180 seconds
Scenario 3 80% 6859 15 hours × 3600 seconds ÷ 6859
voters = 7.782 seconds
15 hours × 3600 seconds ÷ 300
voters = 180 seconds
Table 7. Statistical results of the first improved model –No. of voters.
Scenario 1 Scenario 2 Scenario 3
Turnout 61% 70% 80%
VotingBox1 5283 5346 5376
TotalNum of Exit Voter 5282 5345 5375
TotalNum of Entry Ordinary Voter 5227 6106 7052
TotalNum of Entry Wheelchair Voter 298 302 305
No. of Voters that cast their vote after the closing time of polling station 242 1062 1981
Table 8. Statistical results of the first improved model –average waiting time.
Scenario 1 Scenario 2 Scenario 3
Average waiting time
(in seconds)
Average
Content
Average waiting time
(in seconds)
Average
Content
Average waiting time
(in seconds)
Average
Content
Queue1 694.5 72.5 3531.5 425.3 6388.6 881.9
Queue2 997.8 16.7 1113.1 18.7 1109.2 18.6
Queue3 961.4 16 977.5 16.4 1142.3 19.2
Queue4 628.7 10.3 1030.4 17.3 1115.7 18.8
Queue5 1046.1 17.6 1029.8 17.2 1106.3 18.6
Queue6 649.2 10.4 1133.5 19.1 1134.2 19.1
Queue7 910.4 15.1 1115.9 18.6 1122.2 19
Queue8 6.1 0.1 6.2 0.1 5.4 0.1
Queue9 5 0.1 5.1 0.1 5.5 0.1
Queue10 5.7 0.1 7.3 0.1 5.2 0.1
Queue11 7 0.1 6.2 0.1 6 0.1
Queue12 5.5 0.1 5.8 0.1 5.3 0.1
Queue13 6 0.1 7.3 0.1 6.6 0.1
Queue14 2.7 0 3.3 0 3.3 0
Queue15 4.4 0.4 4.6 0.5 4.5 0.4
Stage 1 694.5 3531.5 6388.6
Stage 2 865.6 1066.7 1121.7
Stage 3 5.4 5.9 5.3
Stage 4 4.4 4.6 4.5
Total 1569.9 4608.7 7520.1
JOURNAL OF INFORMATION TECHNOLOGY CASE AND APPLICATION RESEARCH 13
Figure 10. Layout of the second improved model.
Figure 11. Simulation of the second improved model.
14 C. H. AU ET AL.
options may not always be feasible. In Hong Kong, it would require the witness of candidates or
their polling agents when using a new empty ballot box. Concurrently using more than one ballot
box of the same type may not be allowed in some polling stations, or may lead to controversies
related to the fairness of the election.
Alternative statistical assumption
To ensure the effectiveness and the generalizability of our solution, we have tested the model with
different statistical distribution, which also served as a compensation of the lack of involvement of
other stakeholders and real data.
For example, when we were considering the servicing time of the ballot issuing tables, we tried our
model using both exponential distribution and the lognormal distribution, since both might be seen in
previous studies. For the time spent in putting the ballot into the boxes, we have tried normal
distribution and exponential distribution. Besides, we also tried different mean value or parameters.
The reported waiting time and the queue length, regardless of the combination of the adopted statistical
assumptions, were shortened when using our improved models but the extent varied. Based on this, we
can confirm that our approach may help shortening the overall queue length and waiting time.
Table 9. Simulation settings of the second improved model.
No. of voters
Average inter-arrival time (Ordinary Voters,
in seconds)
Average inter-arrival time (Wheelchair
voters, in seconds)
6859 ordinary voters, 300
wheelchair voters
15 hours × 3600 seconds ÷ 6859
voters = 7.782 seconds
15 hours × 3600 seconds ÷ 300 voters = 180
seconds
Table 10. Statistical results of the second improved model (1).
No of Voters
VotingBox1 2656
VotingBox2 2731
TotalNum of Exit Voter 5386
TotalNum of Entry ordinary Voter 7117
TotalNum of Entry wheelchair Voter 263
No. of Voters that cast their vote after the closing time of polling station 1994
Table 11. Statistical results of the second improved model (2).
Average Content Waiting Time(By Queue, in seconds) Waiting Time(By Stage, in seconds)
Stage 1 Queue1 870.1 6245.2 6245.2
Stage 2 Queue2 19.2 1134.3 1124.6
Queue3 18.9 1124.4
Queue4 18.8 1103.5
Queue5 19 1129.2
Queue6 18.9 1121.3
Queue7 19 1134.6
Stage 3 Queue8 0.1 6.5 5.9
Queue9 0.1 6.2
Queue10 0.1 7.2
Queue11 0.1 5.9
Queue12 0.1 7
Queue13 0.1 6.5
Queue14 0 1.9
Stage 4 Queue15 0.1 1.5 1.6
Queue16 0.1 1.6
Total 7377.3
JOURNAL OF INFORMATION TECHNOLOGY CASE AND APPLICATION RESEARCH 15
Discussion
In general, we first modelled the original situation of the congested polling station with long queues,
followed by building new models for finding better solutions. Simulating the original polling station
can help finding the bottlenecks. In our case, by adjusting the ratio of the number ballot-issuing
desks to the number of voting compartments (from 1:2 to 1:1), we can address the bottleneck and
shorten the waiting queue. In our second improved model, we introduced an additional set of voting
boxes, but it had relatively limited impact. Though the overall waiting time further reduced, but the
queue size was slightly longer.
However, in designing election stations there is a need to consider other issues as well, such as
confidentiality and fairness. Sometimes, the arrangement may not be acceptable to the candidates or
the political parties or may need in advance consultation and agreement despite its advantages of
shortening the queues.
Our trials of alternative statistical assumption of different time requirement and arrival pattern
can contribute to the generalization of our results. We may confirm that, despite the potential
discrepancy between our model and the reality, we have demonstrated how to solve the potential
congestion in polling stations.
Polling station design process model
Our results affirmed that designing the polling stations with simulation software may avoid long
queue and congestions, and thus decrease the discontent of the public in electoral matters. Based on
our results, we have proposed establishing a 3-step process model for setting up the polling stations,
which is illustrated in Table 12 and Figure 12.
Stage 1: the predicting stage
To begin, it is first needed to predict needed resources and time in different steps of polling with the
measurement of polling station venue size, which is to address the previous resources misallocation
problem (Stewart, 2015). We termed this stage the “Predicting Stage”.
The government needs to visit the proposed polling stations and to measure the size of the
stations. The site visit personnel should have a basic measurement of size and dimension of the place
Table 12. The antecedents, activities and outcome of different stages.
Stage 1 - Predicting Stage 2 - Planning Stage 3 - Polling
Antecedents
●The need of estimating
the time and resources
requirement
●The need of seeking optimum resources allocation in the
polling station
●The need of getting real data
for future polling station
betterment
●Conduct the election smoothly
Activities
✓Site visit, measurement
and photo-taking
✓Mock polling station
establishment, with actors
to vote
✓Use simulations to find the need time and expected
length of queue of different combination of voting com-
partments and various supplies
✓Using sensors and other
technologies for
✓Detection of voters’arrival and
departure.
✓Time needed for each step of
the voting processes.
Outcome
✓Estimated time and
resources requirement
✓Desired polling station layout with minimum expected
length of queue and waiting time
✓Real data collected for future
use and evaluation
16 C. H. AU ET AL.
and should be aware of any fixed furniture that could not be moved easily. Ideally, photos may be
taken for later references. Besides, the government should establish a mock polling station for
finding the time needed in different steps of voting by mocking the entire process, including but
not limited to,
●Services time of ballot issuing desks
●Time of chopping on ballots (ordinary voters/wheelchair voters)
●Time of putting the ballot into the boxes.
When collecting these time data, it is desirable to mock the process for a few more time and to
take the average time and standard deviations needed for each step (Adedigba, 2005). If it is
infeasible to visit all the stations, the government should at least visit the polling stations which
are more likely to have congestion. For example, when the area of the polling station is significantly
reduced, or when the number of voters allocated to the polling station significantly increase (like our
study target), it is easy to predict an increase of the chances of congestion or long queue. The
department responsible for holding elections should be alerted and take precautions, and these steps
are proposed in response to the recommendation of Hunt et al. (2010) and Lawson et al. (2006).
At the end of this stage, the time and resources needed for the polling stations may be estimated
with more concrete data. These estimations will be brought to the next stage for building the
simulations and planning the actual polling stations.
Stage 2: the planning stage
After the predicting stage, it is needed to plan the layout of the polling station with optimized
resources and supplies allocation for shortening the queue and waiting time. We termed this stage
the “Planning Stage”. The need of optimizing the allocations was highlighted by Stewart (2015) and
Elklit and Reynolds (2002).
Based on the time measurement results and the site-visit information, simulations should be built to
mock different combinations of setting the tables, voting compartments and other components, thus to
find a set-up that would produce the shortest queue and require the minimum time for the voters to cast
their vote. Similarly, if it is not feasible to take these steps for all polling stations, the government should
at least take the said steps on the polling stations which have a higher likelihood of congestion.
The outcome of this stage is the desired polling stations layouts for all polling stations of the
election. The layout should help minimize the length of the queues and waiting time. These layouts
should be brought to the next stage for implementation in the polling day.
Stage 3: the polling stage
The desired layouts based on the simulations should be brought and implemented in the polling day
for conducting election smoothly. Given the earlier simulations were built based on estimated data,
deviations may exist and should be replaced by real data. During the polling day, some real data may
Figure 12. Activities and data input/output in the process model.
JOURNAL OF INFORMATION TECHNOLOGY CASE AND APPLICATION RESEARCH 17
be collected for future simulation and election evaluation as long as the voters’privacy and election
confidentiality are not harmed. We term this stage the “Polling Stage”.
For example, sensors may be installed at the entrance and exit of the polling station for detecting
the time a voter enter and leave the polling stations. These data can help finding the number of
voters concurrently in the polling station at different times, and thus help to build the voters’arrival
and departure pattern.
On top of sensors at the entrance and exit of the polling station, the government may also
measure the time needed for individual voters in each step by using RFID tags (Ni, Liu, Lau, & Patil,
2004). However, such measurement should be made only based on the agreement of the voters. The
government may recruit voters to participate on a self-voluntary basis. In general, anonymity and
confidentiality of election must be maintained when collecting these data.
The outcome of this stage is the real data of polling station, which may be used for building
simulations which is closer to actual scenarios in the future. The data collected may be processed
using different data analytical techniques.
Limitation and further research direction
Our study has a few limitations that should be addressed. An important limitation, in particular,
concerns the issue of generalizing from a single case (Walsham, 2006). In the future, we may extend
our research with simulating other polling stations and collecting more data about the time needed
in different steps. Secondly, we should also take the feedback from different stakeholders, such as
voters and candidates, into consideration when we design the polling stations. Finally, yet impor-
tantly, we are looking forward to seeing a full implementation of our model for designing the polling
stations so that researchers can look for betterment.
Conclusion
This paper presented a new approach of using IT application for solving the issue of congestions in
the polling stations, an issue that has been highlighted by scholars from different disciplines. We
have narrowed the research gap of using computer simulation in polling stations management. Based
on some widely-adopted assumptions, we re-designed the polling stations and successfully proposed
a solution for the congestion and shortened the waiting time in our selected case.
Our process model may be summarized as three stages, namely predicting, planning and polling.
The predicting stage serves for collecting measurement figures and time data of each step of the
voting processes. These figures and data are used for simulations in the planning stage. After trying
different combinations in the simulations in the planning stage, the desired layout of all polling
stations may be found and should be implemented in the polling day. During the polling day, some
real-time data such as voters’arrival, departure and duration of different steps may be collected.
The simulation can help finding the optimized polling station layout for shortening the waiting time
and queue. Thus, it would enhance the overall voters’experience in the election and would help to
regain voters’confidence in the election. These are critical for a reputed election and democratic society.
Given we have chosen a highly congested station for our test, we hope it may solve the queues of the
polling stations, and thus bring a better voting experience in both Hong Kong and other countries.
Acknowledgments
We would like to express our gratitude to the Electoral Affairs Commission of the Hong Kong government for
providing the information and the reviewers as well as the editors for providing valuable comments to help us revise
the manuscript.
18 C. H. AU ET AL.
Notes on contributors
heuk Hang Au is a master student at the Chinese University of Hong Kong. His research focuses on Social Media, e-
Commerce, IS Education and IT Applications in Social Science. Related publications can be found in the Journal of
Information Technology Case and Application Research (JITCAR), the International Journal of Game-based Learning
(IJGBL) and many IEEE conferences proceedings.
Zewen Xu is a master student at the Hong Kong Polytechnic University. Her research focuses on the digital world, e-
trading and IT applications in real-world problems. Related publications can be found in the Journal of Information
Technology Case and Application Research (JITCAR).
Lin Wang is a master student at the Hong Kong Polytechnic University. Her research focuses on e-Commerce and
other IT applications in business. Related publications can be found in the Journal of Information Technology Case and
Application Research (JITCAR).
Walter S. L. Fung is a teaching fellow at the Hong Kong Polytechnic University. His research focuses on Innovation
and entrepreneurship, Information Security Management, Knowledge Management, Computer Simulation, and
Technology and Social Science. Related publications can be found in the Journal of Information Technology Case
and Application Research (JITCAR) and some reputed IEEE Journals.
ORCID
Cheuk Hang Au http://orcid.org/0000-0002-2121-8573
References
Adedigba, A. I. (2005). Statistical distributions for service times. Saskatoon: University of Saskatchewan.
Alomair, Y., Ahmad, I., & Alghamdi, A. (2015). A review of evaluation methods and techniques for simulation
packages. Procedia Computer Science,62, 249–256. doi:10.1016/j.procs.2015.08.447
Balbo, G., & Vigliotti, M. G. (2013). On the analysis of a M/M/1 queue with bulk services. The Computer Journal,58,
57–74. doi:10.1093/comjnl/bxt118
Banks, J. (1991, December). Selecting simulation software. In B. L. Nelson, W. D. Kelton, & G. M. Clark (Eds.),
Proceedings of the 23rd conference on Winter simulation (pp. 15–20). Phoenix, AZ: IEEE Computer Society.
BBC. 2015. General election 2015: ‘No repeat’of polling day queues [Online]. Retrieved from http://www.bbc.com/
news/uk-politics-31702994
Bell, A. 2016. Voters frustrated by lengthy queues [Online]. Retrievd from http://www.dailytelegraph.com.au/newslocal/
news/federal-election-2016-voters-frustrated-by-lengthy-queues-at-polling-booths-around-sydney/news-story/
f918d83dbc22e11915c60bd4238c4566
Carey, J. M. (2017). Electoral formula and fragmentation in Hong Kong. Journal of East Asian Studies,17(2), 215–231.
Chan, C. (2017). Legalizing politics: An evaluation of Hong Kong’s recent attempt at democratization. Election Law
Journal: Rules, Politics, and Policy,16(2), 296–305. doi:10.1089/elj.2016.0407
Chan, C. P. (2016). Post-umbrella movement: Localism and radicalness of the Hong Kong student movement.
Contemporary Chinese Political Economy and Strategic Relations,2(2), 885.
Chen, L.-H., Hu, D.-W., & Xu, T. (2013). Highway freight terminal facilities allocation based on flexsim. Procedia-
Social and Behavioral Sciences,96, 368–381. doi:10.1016/j.sbspro.2013.08.044
Chiu, A. (2016). 嚴正投訴及譴責: 2016 年立法會選舉太古城西選區票站安排(Serious complaints and condemna-
tion: The arrangement of 2016 legislative council elections taikoo shing west constituency polling station) [Online].
Retrieved from https://andrewkychiu.files.wordpress.com/2016/09/dcletters-16-013_complain_letter_to_eac_cmab.
pdf
Chung, H. F. (2014). A tale of two societies: Fragments of an ethnography on umbrella revolution. Hong Kong
Anthropologist,7,1–19.
Clark, A. (2015). Public administration and the integrity of the electoral process in British elections. Public
Administration,93(1), 86–102. doi:10.1111/padm.2015.93.issue-1
Cochran, J. K., & Chen, H. N. (2005). Fuzzy multi-criteria selection of object-oriented simulation software for
production system analysis. Computers & Operations Research,32(1), 153–168. doi:10.1016/S0305-0548(03)00209-0
Coles, K. A. (2004). Election day: The construction of democracy through technique. Cultural Anthropology,19(4),
551–580. doi:10.1525/can.2004.19.issue-4
Davis, L., & Williams, G. (1994). Evaluating and selecting simulation software using the analytic hierarchy process.
Integrated Manufacturing Systems,5(1), 23–32. doi:10.1108/09576069410050314
Dooley, K. (2001). Social research methods (4th ed.). Upper Saddle River, NJ: Prentice Hall.
JOURNAL OF INFORMATION TECHNOLOGY CASE AND APPLICATION RESEARCH 19
Durán, G., Giormenti, M., Guajardo, M., Pinto, P. M., Rey, P. A., & Stier-Moses, N. E. (in press). Improving access to
voting: Matching voters to polling stations optimally Electoral Studies.
Electoral Affairs Commission. (Ed.). (2004). Report on the 2004 legislative council election held on 12 september 2004.
Hong Kong: Hong Kong Government.
Elklit, J., & Reynolds, A. (2002). The impact of election administration on the legitimacy of emerging democracies: A
new comparative politics research agenda. Commonwealth and Comparative Politics,40(2), 86–119. doi:10.1080/
713999584
Gipps, P. G. (1981). A behavioural car-following model for computer simulation. Transportation Research Part B:
Methodological,15(2), 105–111. doi:10.1016/0191-2615(81)90037-0
Grant, F. H., III. (1980). Reducing voter waiting time. Interfaces,10(5), 19–25. doi:10.1287/inte.10.5.19
Harrison, P., & Patel, N. (1994). Performance modelling of communication networks and computerarchitectures.
Wokingham: Addison-Wesley.
Haviv, M., & Oz, B. (2016). Regulating an observable M/M/1 queue. Operations Research Letters,44, 196–198.
doi:10.1016/j.orl.2016.01.002
HKGov. (2016) 2016 legislative council election - home. Elections.gov.hk. Retrieved August 20, 2017, from http://www.
elections.gov.hk/legco2016/eng/index.html
Hunt, L. M., Robson, M., Lemelin, R. H., & Mcintyre, N. (2010). Exploring the acceptability of spatial simulation
models of outdoor recreation for use by participants in public participation processes. Leisure Sciences,32(3), 222–
239. doi:10.1080/01490401003712515
Kornblith, M., & Jawahar, V. (2005). Elections versus democracy. Journal of Democracy,16(1), 124–137. doi:10.1353/
jod.2005.0011
Krahl, D. (2009, December). ExtendSim advanced techology: Discrete rate simulation. In Winter simulation conference
(pp. 333–338). Austin, TX: Manuel Rossetti and Ray Hill.
Laskie, A. (2016). Australian electoral commission sorry for long queues outside polling stations. The Sydney Morning
Herald.
Lawson, S. R., Itami, R. M., Gimblett, H. R., & Manning, R. E. (2006). Benefits and challenges of computer simulation
modeling of backcountry recreation use in the Desolation Lake Area of the John Muir Wilderness. Journal of
Leisure Research,38(2), 187. doi:10.1080/00222216.2006.11950075
Leks, D., & Gwiazda, A. (2015). Application of FlexSim for modelling and simulation of the production process.
Selected Engineering Problems,2,51–56.
Li, G., & Li, V. (2009). New queueing theory. Hong Kong: Sunny Interactive.
Liu, T., Duan, Y., & Liu, Y. (2016). Simulation and optimization of the AS/RS based on Flexsim. In J. Hung, N. Yen, &
K. Li (Eds.), Frontier Computing. Lecture Notes in Electrical Engineering, (pp. 855–863). Singapore: Springer.
Lofberg, J. (2004, September). YALMIP: A toolbox for modeling and optimization in MATLAB. In Jia-Yush Yen and
Gary G. Yen (Eds.), Computer aided control systems design, 2004 IEEE International Symposium on (pp. 284–289).
New Orleans, LA: IEEE.exte.
Mitchell, J. (2017). French citizens in London blast ‘three hour wait’to vote in election. Evening Standard.
Murata, T., & Konishi, K. (2013). Making a practical policy proposal for polling place assignment using voting
simulation tool. SICE Journal of Control, Measurement, and System Integration,6, 124–130. doi:10.9746/jcmsi.6.124
Ni, L. M., Liu, Y., Lau, Y. C., & Patil, A. P. (2004). LANDMARC: Indoor location sensing using active RFID. Wireless
Networks,10(6), 701–710. doi:10.1023/B:WINE.0000044029.06344.dd
Nikoukaran, J., Hlupic, V., & Paul, R. J. (1998, December). Criteria for simulation software evaluation. In D. J.
Medeiros, E. F. Watson, J. S. Carson, & M. S. Manivannan (Eds.), Proceedings of the 30th conference on Winter
simulation (pp. 399–406). Washington, DC: IEEE Computer Society Press.
Nordgren, W. B. (2003, December). Flexible simulation (Flexsim) software: Flexsim simulation environment. In D.
Ferrin & D. J. Morrice (Eds.), Proceedings of the 35th conference on Winter simulation: Driving innovation (pp.
197–200). New Orleans, LA: Winter Simulation Conference.
Qian, Z., & Sun, H. (2013). A simulation study on production logistics balance based on petri net+ Flexsim. In M.
Liang & H. Nagana (Eds.), International Conference on Education Technology and Information System (ICETIS
2013). Sanya, China: Atlantis Press.
Sahay, B. S., & Gupta, A. K. (2003). Development of software selection criteria for supply chain solutions. Industrial
Management & Data Systems,103(2), 97–110. doi:10.1108/02635570310463429
SCMP. (2016). Hong Kong Legco elections live: Polls finally close in hotly contested races. South China Morning Post.
Smith, M. A., Monfort, S. S., & Blumberg, E. J. (2015). Improving voter experience through user testing and iterative
design. Journal of Usability Studies,10, 116–128.
Spencer, D. M., & Markovits, Z. S. (2010). Long lines at polling stations? Observations from an election day field study.
Election Law Journal, 9(1), 3–17.
Stasser, G. (1988). Computer simulation as a research tool: The DISCUSS model of group decision making. Journal of
Experimental Social Psychology,24(5), 393–422. doi:10.1016/0022-1031(88)90028-5
Stewart, C., III. (2015). Managing polling place resources. Caltech/MIT Voting Technology Project Report.
Sundarapandian, V. (2009). Probability, statistics and queuing theory. New Delhi, India: Phi Learning.
20 C. H. AU ET AL.
Tewoldeberhan, T. W., Verbraeck, A., Valentin, E., & Bardonnet, G. (2002, December). Software evaluation and
selection: An evaluation and selection methodology for discrete-event simulation software. In J. L. Snowdon & J. M.
Charnes (Eds.), Proceedings of the 34th conference on Winter simulation: Exploring new frontiers (pp. 67–75). San
Diego, CA: Winter Simulation Conference.
Tong, E. (2016). Hong Kong LegCo Election 2016: Thousands queued to vote early hours of morning amid record
turnout. Hong Kong: Free Press.
Wahed, M. A., Faghri, A., & Li, M. (2017). An innovative simulation model for the operations of a multipurpose
seaport: A case study from Port of Wilmington, USA. International Journal of Simulation and Process Modelling,12,
151–164. doi:10.1504/IJSPM.2017.083530
Walsham, G. (1995). Interpretive case studies in IS research: Nature and method. European Journal of Information
Systems,4(2), 74–81. doi:10.1057/ejis.1995.9
Walsham, G. (2006). Doing interpretive research. European Journal of Information Systems,15(3), 320–330.
doi:10.1057/palgrave.ejis.3000589
Wang, A. B. (2017). French voters around the world turn out in droves to vote in presidential election. The
Washington Post.
Wang, X., Pan, Y., Song, L., Huang, X., Liu, M., & Liu, A. (2017). Modelling and application for eclampsia with
SimMom. International Journal of Simulation and Process Modelling,12, 103–110. doi:10.1504/IJSPM.2017.083526
Yang, M., Fry, M. J., Kelton, W. D., & Allen, T. T. (2014). Improving voting systems through service-operations
management. Production and Operations Management,23(7), 1083–1097. doi:10.1111/poms.2014.23.issue-7
Zainal, Z. (2007). Case study as a research method (pp. 9). Johor Bahru, Malaysia: Jurnal Kemanusiaan.
Zaragoza, M. A., & Mota, I. F. D. L. (2016). Tactics for approaching cash optimisation in bank branches. International
Journal of Simulation and Process Modelling,11, 492–503. doi:10.1504/IJSPM.2016.082913
Zeigler, B., Kim, T., & PräHofer, H. (2009). Theory of modeling and simulation. Amsterdam [u.a.]: Academic Press.
Zhu, X., Zhang, R., Chu, F., He, Z., & Li, J. (2014). A Flexsim-based optimization for the operation process of cold-
chain logistics distribution centre. Journal of Applied Research and Technology,12, 270–278. doi:10.1016/S1665-
6423(14)72343-0
JOURNAL OF INFORMATION TECHNOLOGY CASE AND APPLICATION RESEARCH 21
