This is the Pre-Published Version.

© Emerald Publishing Limited. This AAM is provided for your own personal use only. It may not be used for resale, reprinting, systematic distribution,
emailing, or for any other commercial purpose without the permission of the publisher.
The following publication Zhao, Y., Qi, K., Chan, A.P.C., Chiang, Y.H. and Siu, M.F.F. (2022), "Manpower forecasting models in the construction
industry: a systematic review", Engineering, Construction and Architectural Management, Vol. 29 No. 8, pp. 3137-3156 is published by Emerald and is
available at https://dx.doi.org/10.1108/ECAM-05-2020-0351
1 MANPOWER FORECASTING MODELS IN THE CONSTRUCTION INDUSTRY:
2 A SYSTEMATIC REVIEW
3
4 Yijie Zhao, Kai Qi, Albert P.C. Chan, Yat Hung Chiang, Ming Fung Francis Siu
5
6 ABSTRACT
7
8 Purpose – This paper aims to make a systematic review of the manpower prediction model of
9 the construction industry. It aims to determine the forecasting model's development trend,
10 analyse the use limitations and applicable conditions of each forecasting model, and then
11 identify the impact indicators of the human resource forecasting model from an economic point
12 of view. It is hoped that this study will provide insights into the selection of forecasting models
13 for governments and groups that are dealing with human resource forecasts.
14 Design/methodology/approach – The common search engine, Scopus, was used to retrieve
15 construction manpower forecast-related articles for this review. Keywords such as
16 "construction", "building", "labour", "manpower" were searched. Papers that not related to the
17 manpower prediction model of the construction industry were excluded. A total of 27 articles
18 were obtained and rated according to the publication time, author and organization of the article.
19 The prediction model used in the selected paper was analysed.
20 Findings – The number of papers focusing on the prediction of manpower in the construction
21 industry is on the rise. Hong Kong is the region with the largest number of published papers.
22 Different methods have different requirements for the quality of historical data. Most
23 forecasting methods are not suitable for sudden changes in the labour market. This paper also
24 finds that the construction output is the economic indicator with the most significant influence
25 on the forecasting model.
26 Research limitations/implications – The research results discuss the problem that the
27 prediction results are not accurate due to the sudden change of data in the current prediction
28 model. Besides, the study results take stock of the published literature and can provide an
29 overall understanding of the forecasting methods of human resources in the construction
30 industry.
31 Practical implications – Through this study, decision-makers can choose a reasonable
32 prediction model according to their situation. Decision-makers can make clear plans for future
33 construction projects specifically when there are changes in the labour market caused by
34 emergencies. Also, this study can help decision-makers understand the current research trend
35 of human resources forecasting models.
36 Originality/value – Although the human resource prediction model's effectiveness in the
37 construction industry is affected by the dynamic change of data, the research results show that
38 it is expected to solve the problem using artificial intelligence. No one has researched this area,
39 and it is expected to become the focus of research in the future.
40

1
41 1. INTRODUCTION
42
43 The labour supply is determined by demographic factors (demographic structure and labour
44 force participation rate) (Aaronson et al., 2006; Toder and Solanki, 1999). However, a specific
45 industry's labour demand is more determined by economic activities or social division of labour
46 (Ngai and Pissarides, 2005; Schaible and Mahadevan-Vijaya, 2002). This may lead to an
47 imbalance between labour demand and supply. For alleviating this imbalance, the market will
48 regulate itself automatically and the government will also issue incentive policies to improve
49 this situation (International Labour Organization, 2017).
50
51 The construction industry constitutes a significant portion of the economy in many countries
52 (Betts et al., 2015). Despite advances in mechanisation and technology in recent years, the
53 construction industry is commonly viewed as craft-based and labour-intensive, making it
54 heavily reliant on the availability of manpower (Chan et al., 2006). Unbalanced manpower
55 supply and demand results in a labour shortage or a surplus. Labour shortage leads to the
56 employment of unqualified workers, a heavy workload for existing workers to finish
57 construction on time, and increased payments to maintain the current workforce and to attract
58 new workers; whereas, labour surplus leads to the firing of qualified workers, which can
59 negatively affect their families and social stability (Ho, 2010). Government bureaux, training
60 institutions, and the industries concerned have paid attention to labour resources planning and
61 related matters. Accurate manpower forecast is becoming increasingly critical for policy
62 formulation and human resources planning in the construction industry. A series of qualitative
63 methods and quantitative methods have been developed to balance worker demand and worker
64 supply for broad occupations in healthcare, the military, construction, tourism, and service
65 industries (Chan et al., 2006; Reeves and Reid, 1999; Reifels et al., 2014; Safarishahrbijari,
66 2018; Shen and Huang, 2008; Wu et al., 2017). Few scholars, however, have put their emphasis
67 on conducting a systematic review of the human resource prediction model within the
68 construction industry. As the motivation of manpower forecasting varies among different
69 regions, it is difficult to select the most applicable prediction models according to their regional
70 characteristics. The objectives of this review are thus to (i) review the papers from 1990 to
71 2020 and explore the development trend of manpower forecasting model in the construction
72 industry, (ii) analyse the use limitations and applicable conditions of each forecasting model,
73 and (iii) identify the impact index of the human resource forecasting model from an economic
74 perspective.

2
 75 2. LITERATURE REVIEW
76
77 Manpower forecasting is essential to human resource management. To deal with the future
78 challenge, stakeholder of construction projects and manager of companies need to forecast
79 manpower demand depends on the working schedule (Noe et al., 2017). Accurately predicting
80 manpower demand and supply has a positive impact on many industries (Garza et al., 2013).
81 The earliest record of manpower planning focused on the unbalance between war industry and
82 civilian needs (Pate, 1943). Since it was first proposed by Pate in 1943, manpower planning
83 has been introduced into other industry. For example, in the medical industry, Simonds (1976)
84 predicted the manpower demand in 1976 to ensure the operation of the medical system. For the
85 aviation industry, manpower planning is essential for safety and scheduling punctuality (Tang
86 and Hsu, 2019). Davis (1987) firstly developed a model to simulate the manpower requirement
87 of a construction project in 1987. The model can help construction projects to minimize unused
88 staff and avoid over-use of staff. It is necessary for the construction industry to predict
89 manpower demand and supply because of the high complexity of construction activities and
90 manpower composition (Li, 2019). For the construction industry, accurately forecasting
91 manpower demand and supply can help construction projects save cost and minimize the
92 construction period. A growing number of studies have focused on manpower forecasting
93 within the construction industry in recent years.
94
95 Application of Delphi method in manpower forecasting
96
97 As a popular and typical qualitative method, Delphi method engages a panel of experts
98 answering questionnaires in several (usually two or more) rounds until some form of consensus
99 is reached (Ameyaw et al., 2016). Delphi method is a simple and flexible method and can help
100 researchers to get repetition and statistic result (Landeta, 2006). Besides, some manpower
101 factors are difficult to quantify and lack data such as identifying the critical factors (Chang-
102 Richards et al., 2017). Delphi method is particularly important in this case. Thus, the Delphi
103 method is selected by some scholars as the main method to predict manpower demand and
104 supply. Kwak et al. (1997) used the Delphi technique to identify factors affecting the demand
105 and supply of clinical laboratory manpower in urban academic health centres. The identified
106 factors were then structured and subjected to the Analytic Hierarchy Process (AHP), in order
107 to identify the relative importance and criticality of the identified factors. The classical Delphi
108 method can be further modified to different forms such as modified Delphi, policy Delphi,
109 technological Delphi, real-time Delphi, argument Delphi, and online Delphi for a range of
110 purposes (Hasson and Keeney, 2011). For example, Solnet et al. (2014) adopted the modified
111 Delphi process. Drawn from a comprehensive literature review, they provided the expert panel
112 with pre-selected factors that impact on the tourism workforce. Some criticisms of qualitative
113 methods arise because of its time-consuming characteristics (spending a large amount on costly
114 execution time) and its dependency on expert opinions (leading to subjective and biased results)
115 (Parker and Caine, 1996; Safarishahrbijari, 2018).
116
117 Existing methods in manpower forecasting
118
119 Quantitative methods include a variety of mathematical models, such as time series forecasting,
120 regression analysis, multiplier approaches, and stock-and-flow models, grey models, and linear
121 differential equations. Time series models estimate future trends based on historical data or
122 previously observed values. Among time series modelling techniques, vector error correction
123 (VEC), Box-Jenkins, exponential and Markov processes have been commonly used in
124 manpower forecasting applications, including construction workforce demand (Wong et al.,

3
125 2005, 2007), personnel in the energy technology industry (Hsu et al., 2012), physician and
126 nurse supply (Kinstler et al., 2008; Santric-Milicevic et al., 2013), and call centre workforce
127 requirements (Shen and Huang, 2008). Time series models are heavily dependent on historical
128 data and are used to examine the relationship solely between time and past behaviour and then
129 to extrapolate that trend into the future (Wong et al., 2012). The univariate time series
130 projection is a relatively reliable, simple, and inexpensive method for analysing underlying
131 trends, cyclic, and periodic elements (Bartholomew et al., 1991). These methods are limited by
132 the assumption that analysing past trends themselves is good enough for short-term forecasting
133 purposes. No other variables besides time need be included in the equations because it is not
134 the objective to find the relationship between the dependent variable and these other variables.
135 Consequently, they can be beneficial for short-term prediction (i.e., one year ahead).
136 Regression analysis is another popular method for workforce forecasts, which assess the impact
137 of independent variables on a dependent variable. Regression models have been widely used
138 for a range of workforce forecasting situations, such as mental health workforce capacity
139 (Reifels et al., 2014) and personnel demand in construction projects (Agarwal, et al., 2013).
140 The regression analysis has the advantages in describing the influence of factors (i.e.,
141 independent variables or predictors) on the dependent variable (such as manpower demand).
142 However, this method is usually criticised when the regression function does not contain the
143 parameter of interest. Moreover, regression analysis is unable to model highly dynamic systems
144 involving personnel entering and exiting (e.g., recruitment and attrition) over time
145 (Safarishahrbijari, 2018). Along with the attention to employment planning, an increasing
146 number of mathematical models and simulations have been developed to deal with workforce
147 forecasts and their reliability and their applicability has been proved in actual situations,
148 including multiplier models (Chan et al., 2006; Sing et al., 2012), stock-and-flow techniques
149 (Crettenden et al., 2014; Fraher et al., 2013; Sing et al., 2011), grey model (Ho, 2010), and
150 linear differential equations (Jiang and Begun, 2002).
151
152 Effective manpower forecasting can help the construction industry in the planning and
153 allocation of labour resources. The time series projection (Wong et al., 2005; Wong, et al.,
154 2011), labour multiplier approach (Chan et al., 2006; Proverbs et al., 1999; Sing et al., 2012),
155 grey model (Ho, 2010) and stock-and-flow techniques (Sing et al., 2011) have been developed
156 and applied to forecasting a construction workforce. Wong et al. (2012) conducted a literature
157 review of construction workforce forecasting models. Their review focused on labour demand
158 only. They included general models developed by government, labour departments, and
159 construction related-organisations, but methods specifically designed for the construction
160 industry were not examined in detail.
161
162 Motivations of conducting manpower forecasting reviews
163
164 The motivations underlying research endeavours into construction manpower forecasting
165 modelling in different regions were triggered by particular events. The reasons are summarised
166 in Table 1. It is worth to notice that the increase in the number of new projects and sudden
167 changes in the labour market caused by emergencies are the two main reasons, as well as labour
168 shortage and economic downturn. As it is believed that an increasing number of manpower
169 studies have been conducted in recent years due to labour shortage problems suffered
170 worldwide, it is necessary to review the relevant literature in a systematic and comprehensive
171 manner in order to identify potential methods useful for forecasting manpower in construction.
172

4
173
174 Table 1. Motivations of modelling construction manpower
Publication year Authors Locations Events
Rosenfeld and Tremendous surge of immigration resulted in the sudden change of labour
1993 Israel
Warszawski market.
Assessing the need of manpower in preconstruction since no benchmark can
1995 Persad et al. US
be found.
Bell and South Carolina Department of Transportation (SCDOT) undertook two major
2003 US
Brandenburg construction programs resulting in increased volume of work.
Australia suffered labour shortage in construction industry since the early
2014 Liu et al. Australia
twenty-first century.
Young persons were not attracted to the construction sector and there was lack
2014 Vilutienė et al. Lithuanian
of skilled manpower.
Hong Kong was severely hit by the downturn of the property market in
2016 Chan et al. Hong Kong
construction sector.
Chang-Richards et 2008 global financial crisis and the earthquakes which happened in 2010 and
2017 New Zealand
al. 2011 in Christchurch, New Zealand, resulting in a skills shortage.
Hong Kong expands its building stock requiring management for both new
2017 Lai Hong Kong
and ageing buildings.
175
176 3. METHODOLOGY
177
178 A systematic literature review was carried out to identify studies relevant to manpower
179 forecasting in the construction industry. The literature review is a typical methodology for
180 acquiring knowledge on specific issues in construction management (CM) research (Chan et
181 al., 2004; Darko et al., 2017; Yi and Chan, 2013). The common search engine, Scopus, was
182 used to retrieve construction manpower forecast-related articles for this review. First, a
183 comprehensive desktop search was performed under the “title/abstract/keyword” field in
184 Scopus. The search keywords included “construction”, “building”, “housing”, “real estate”,
185 “civil”, “infrastructure”, “manpower”, “labour”, “workforce”, “employment”, “forecast”,
186 “prediction”, “estimation”, and “plan”. The combinations of search keywords were varied
187 using the Boolean logic (AND) or (OR). Only articles written in English were searched. The
188 subject areas consisted of engineering, environment, business, management, computer science,
189 mathematics, decision sciences, economics, econometrics and finance, and social sciences.
190 Only papers published in peer-reviewed journals were selected for this review. Dissertations,
191 theses, conference proceedings, textbooks, non-full texts, non-peer reviewed publications were
192 excluded. Review articles and reference lists were further retrieved to ensure relevant studies
193 had not been missed.
194
195 The search was performed in March 2020. Online article records were imported to EndNote
196 X7 (Thomson Reuters, Philadelphia, PA, USA). Each title and abstract was assessed. When
197 information in the title and abstract was insufficient to determine the articles that should be
198 selected, the full text was reviewed. In accordance with the above selection criteria, 27 peer-
199 reviewed journal articles published between January 1980 and March 2020 related to
200 construction manpower forecasting were selected for this review. The paper selection process
201 is illustrated in Figure 1. Annual publication trends in journals were extracted and analysed by
202 descriptive statistics and contribution assessments (Abidoye and Chan, 2017).
203
Figure 1: PRISMA diagram for extracting the most relevant article

5
 Records identified through Scopus database
Identification searching (manpower/labor/workforce/employment)
and (forecast/prediction/estimation/plan)
Time: Jan 1980 to Mar 2020
(n=40,641)

Limited to construction area, journal article and

Screen

English
ing

(n=1,865)

Records excluded (n=1,657):

Records screened for eligibility based on title
• Not related to manpower forecast (n=1533)
(n=208)
• Duplicates (n=124)
Eligibility

Records excluded (n=178):

Articles assessed for eligibility based on abstract
• Not within the construction industry (n=152)
(n=30)
• Not related to manpower forecast (n=26)

Records excluded (n=5):

Full text articles assessed for eligibility
• Conceptual manpower forecast, no qualitative/
(n=25)
quantitative methods used (n=26)

Studies identified from the reference list of selected

papers
Included

(n=2)

Final extraction
(n=27)
204
205 To evaluate the contributions of the authors and their affiliations (university, institution or
206 research centre) to the research area, the approach proposed by Howard et al. (1987) was
207 adopted, used in a number of similar construction management-related review studies related
208 to green building (Darko and Chan, 2016), construction labour productivity (Yi and Chan,
209 2013), hedonic pricing modelling in property price appraisal (Abidoye and Chan, 2017), and
210 construction and demolition waste management (Yuan and Shen, 2011). Howard et al. (1987)
211 considered that the author contributions were unequal in a multi-authored article. They
212 assumed that the first author had the highest contribution, followed by the second, the third,
213 and so on. Their proposed formula is given as Equation (1).
1.5n − i
Score= (1)
n
∑ i =11.5n −i
214 where n = the number of authors of the article, i = the order of the specific author.
215
216 Consequently, author credits can be appropriately divided in a multi-authored paper. Assuming
217 a one-point score for each article, Table 2 illustrates a detailed score distribution for authors.
218 Using the score matrix, the accumulated scores for authors and institutions were calculated,
219 compared, and discussed.
220
221 Table 2. Score matrix for multi-authored papers [Data extracted Ke et al. (2009)]
Number of Order of specific author
authors 1 2 3 4 5
1 1.00
2 0.60 0.40
3 0.47 0.32 0.21
4 0.42 0.28 0.18 0.12
5 0.38 0.26 0.17 0.11 0.08
222

6
223 4. RESULTS
224
225 4.1 Overview of manpower forecasting publications
226
227 Publications grouped by year
228
229 Figure 2 presents the distribution over time of the manpower forecasting-related papers
230 between 1990 and March 2020. The earliest paper on human resource planning in the
231 construction industry was published in 1993. Three papers were published from 1995 to 1999,
232 two papers were from 2005 to 2004 and four papers were from 2000 to 2014. Twelve papers
233 were published from 2010 to 2014, and the number of papers published reached the peak during
234 this period. Before 2014, the number of papers published was on the rise. It can be inferred that
235 the research on the manpower prediction model of the construction industry has received more
236 and more attention in this period. There was a decline between 2015 and 2020, with five related
237 articles published. The reasons can be summarized as the problem of labour shortage in the
238 construction industry has been alleviated.

7
239
Figure 2: Number of papers published in 5-year period
Number of papers published in 5-year period
14
12
12
Number of publications

10
8
6 5
4
4 3
2
2 1
0
1990-1994 1995-1999 2000-2004 2005-2009 2010-2014 2015-2020
Time

240
241 Publications grouped by geographical area
242
243 The geographical area studied (Table 3). It seems that Hong Kong has been given the most
244 attention, with 17 articles published. This could be attributed to the Hong Kong Government’s
245 greater need for planning manpower and the data availability because of the known shortage
246 of skilled labour in Hong Kong (Chan et al., 2006; Sing et al., 2012). In addition to Hong Kong,
247 more and more regions paid attention to the development of manpower forecasting models for
248 the construction industry, and related studies were conducted in eight regions. Two articles
249 have been published in Britain and two in the United States. In US, the two studies used
250 regression analysis to predict manpower demand (Bell and Brandenburg, 2003; Persad et al.,
251 1995). The motivations of these studies are the increased construction work (Bell and
252 Brandenburg, 2003) and the lack of benchmark in assessing manpower demand in
253 preconstruction (Persad et al., 1995). In UK, Ball and Wood (1995) adopted regression analysis
254 to estimate the demand of construction workers. Motivated by the shortage of construction
255 workforce since the early 21st century in Australis, Liu et al. (2014) used time series and
256 multiple regression analyses to forecast manpower needs. Agarwal et al. (2013) in India also
257 adopted regression model to predict employees in building constructions. Both Vilutienė et al.
258 (2014) in Lithuania and Chang-Richards et al. (2017) in New Zealand adopted Delphi method
259 in their article. Most studies outside Hong Kong used regression analysis and Delphi method
260 to forecast construction workforce. The steps, techniques and related software in regression
261 analysis are easy to acquire, making it user-friendly to many quantitative studies. Whereas,
262 Delphi method is an effective alternative when historical data is lacking, making it widely used
263 in regions where data acquisition is difficult.
264
265 Table 3. Study area focus
Study area Number of articles 1990-1994 1995-1999 2000-2004 2005-2009 2010-2014 2015-2020
Hong Kong 17 — 1 1 4 8 3
United States of America 2 — 1 1 — — —
United Kingdom 2 — 1 — — 1 —
Australia 1 — — — — 1 —
New Zealand 1 — — — — — 1
Lithuania 1 — — — — 1 —
Israel 1 1 — — — — —
India 1 — — — — 1 —
Iran 1 — — — — — 1

8
266
267 Publications grouped by authorships
268
269 Table 4 lists the authors who have published more than six articles and the literature scores.
270 Table 5 lists the name, class of institution, number and score of related articles published by
271 their affiliations. A total of 46 authors were identified. Among them, 5 contributed to at least
272 six papers. Wong JMW, Chan APC and Sing MCP were the most active scholars with 6–7
273 papers and scores over 2.40. It can be seen that most of the authors who have published more
274 than six articles belong to the three affiliations with the highest scores listed in Table 5.
275
276 Table 4. Authors’ involved in at least three papers
Authors Studies Affiliations Publications Score
Wong JMW Chan et al., 2003; The Hong Kong Polytechnic University. 7 2.75
Chan et al., 2006;
Wong et al., 2005, 2007, 2008, 2010, 2011.
Chan APC Chan et al., 2003; The Hong Kong Polytechnic University. 7 2.45
Chan et al., 2006;
Wong et al., 2005, 2007, 2008, 2010, 2011.
Sing MCP Liu et al., 2014; Curtin University; 6 2.42
Sing et al., 2011; City University of Hong Kong;
Sing et al., 2014; The Hong Kong Polytechnic University.
Sing et al., 2012;
Sing, et al., 2015;
Sing et al., 2016.
Love PED Liu et al., 2014; Curtin University. 6 1.68
Sing et al., 2011;
Sing et al., 2014;
Sing et al., 2012;
Sing et al., 2015;
Sing et al., 2016.
Chiang YH Chan et al., 2006; The Hong Kong Polytechnic University. 7 1.52
Chan et al., 2003;
Wong et al., 2005, 2007, 2008, 2010, 2011.
277
278 Author affiliations included universities, government bodies, and companies. Most (89%) were
279 from universities, which accords with the studies of Yi and Chan (2013) and Abidoye and Chan
280 (2017), who reported that university researchers constituted the most to construction
281 management research. Among these institutions, The Hong Kong Polytechnic University has
282 the highest score, with 9 points, and 10 articles published, followed by Curtin University, and
283 the City University of Hong Kong. Other institutions have scores of one or less. The research
284 on predicting manpower in the construction industry in these three universities is relatively
285 complete.
286
287 Table 5. Authors’ affiliations
Affiliations Class of Institution No. of publications No. of authors Score
The Hong Kong Polytechnic University, Hong Kong University 9 10 9.00
Curtin University, Australia University 6 6 3.95
City University of Hong Kong, Hong Kong University 7 5 3.87
South Bank University, UK University 1 2 1.00
University of Wolverhampton, UK University 1 2 1.00
Vilnius Gediminas Technical University, Lithuania University 1 4 1.00
Art University of Isfahan, Iran University 1 2 1.00
National Institute of Construction Management and Research, India University 1 2 0.79
University of Texas at Austin, US University 2 2 0.72
Massachusetts Institute of Technology, US University 1 1 0.60
Clemson University, US University 1 1 0.60
The University of Auckland, New Zealand University 1 1 0.47
Texas Department of Transportation, US Government body 1 1 0.47
Technion-Israel Institute of Technology, Israel University 1 1 0.40
University of Canterbury, New Zealand University 1 1 0.32
New Zealand Lifelines Committee, New Zealand Government body 1 1 0.21
Indian Institute of Technology Madras, India University 1 1 0.21
Gherzi Eastern Ltd., India Company 1 1 0.21

9
 Birmingham City University, UK University 1 1 0.18
288
289 4.2 Forecasting models used
290
291 Constraints and attributes of manpower forecast models
292
293 In order to clarify the advantages, disadvantages and constraints of the different prediction
294 models or methods used in the selected literature, statistics analysis of the prediction models is
295 carried out. The strengths and the constraints applying to the methods are presented in Table 6.
296 Most forecasting models can be used for both supply and demand sides. The labour multiplier
297 approach can only be used for manpower demand forecasting, whereas, a stock-flow model is
298 applicable to manpower supply prediction. As a qualitative method, Delphi method has been
299 well applied in many fields. The limitation of this research method can be concluded that the
300 results effectiveness depends on the expert’s experience or the subjective guesses, despite the
301 drawbacks, this method is beneficial when historical data are unreliable (Gnatzy et al., 2011).
302 Therefore, Delphi method is suitable for studying the changes in the labour market under
303 emergencies. Regression model, time series analysis and econometrics model are all suitable
304 for short-term prediction. However, these models are not easy to model highly dynamic data
305 (Liu et al., 2014). The labour force multiplier method (LFM) is a method that commonly used
306 to forecast labour demand, and it is also a better choice for the Hong Kong Government (Chan
307 et al., 2006). Specifically, the LFM approach is proposed based on the assumption that
308 construction output is related to labour demand, with the condition that the government has a
309 clear plan. Compared with the three above mentioned methods, LFM can make a more accurate
310 prediction. However, in contrast, this method depends on the accuracy of building output data.
311 Another forecasting method is stock-flow model, this simulation method is suitable to simulate
312 the construction industry's real labour market. The accuracy of stock-flow model prediction
313 results depends on the grasp of reality (Sing et al., 2011). Finally, the Gray model can establish
314 a prediction model through incomplete information, which does not require the high integrated
315 data source (Ho, 2010). Overall, the regression model is the most widely used prediction
316 method, because the regression models is better at describing the influence of individual factors.
317
318 It is worth clarifying that some studies used two or more forecasting techniques can avoid the
319 limitation of using a single prediction. Ho (2013) and Liu et al. (2014) adopted time series and
320 multiple regression analyses. Wong et al. (2011) compared time series, multiple regression and
321 an econometric model (cointegration analysis and VEC). Sing et al. (2014) and Wong et al.
322 (2010) adopted both qualitative and quantitative techniques in their studies [structured
323 interview and multipliers and stock-flow in Sing et al. (2014); focus group and linear regression
324 in Wong et al. (2010)]. Considering the demand and supply aspects of human resources forecast,
325 a more accurate result can be generated (Sing et al., 2014).
326

10
327
328 Table 6. Summary of manpower forecast models
Approach Methodology Strength Constraints Demand Supply Study area Studies
Delphi method Collecting Particularly useful Dependent on the Yes Yes New Zealand, Chang-Richards et
opinion/view of when historical knowledge and Hong Kong, al., 2017;
workplace data are experience of Lithuania Lai, 2017;
supervisors and unavailable involved experts Sing et al., 2014;
employers Vilutienė et al.,
through 2014;
questionnaires, Wong et al., 2010.
interviews,
expert
assessment or
focus groups
Regression model Establishing the Advantageous in Reliable only when Yes Yes US, Bell and
relationship describing the the determinant Hong Kong, Brandenburg, 2003;
between influence of factors variables are properly Australia Chan et al., 2003;
determinant selected and Ho, 2013;
variables and estimated; unable to Agarwal et al., 2013;
forecasting model highly Liu et al., 2014;
variable dynamic data series Persad et al., 1995;
Wong et al., 2010;
Wong et al., 2008,
2010, 2011.
Time series analysis Projection of the Simple, reliable, Cannot capture the Yes Yes Hong Kong, Ho, 2013;
[by vector error trend of variables and inexpensive dynamic behaviour; Australia Liu et al., 2014;
correction (VEC), over short period appropriate for short- Wong et al., 2005,
exponential of time based on term prediction only 2010, 2011.
smoothing, ARIMA, historical data
or Box-Jenkins
technique]
Econometric model Establishing the Able to investigate Only appropriate for Yes Yes UK, Ball and Wood,
(co-integration time series the dynamic co- short-term prediction; Hong Kong 1995;
analysis and VEC) model followed movement among complex procedures Wong et al., 2007,
by the testing of variables and 2011.
cointegration capture the
among variables relationship
between workforce
demand/supply and
the associated
factors
Labour multiplier Establishing the Straightforward, Dependent on the Yes No Hong Kong, Chan et al., 2006;
approach linear easily understood accuracy of industry Israel Proverbs et al.,
relationship by practitioners output data and on 1999;
between the the assumption that Rosenfeld and
industry output there exists a direct Warszawski, 1993;
and manpower relationship between Sing et al., 2014;
demand per unit construction output Sing et al., 2012;
and manpower Sing et al., 2015.
demand
Stock-flow model Creating a stock Provide a Require No Yes Iran, Dabirian et al.,
(system dynamics) flow by first systematic view of computerised Hong Kong 2018;
determining the manpower analytical solutions Sing et al., 2011;
current structures; consider Sing et al., 2014;
workforce model variables Sing et al., 2016.
dynamically over
time; appropriate
for longer-term
forecast (5-10
years).
Gray model Generation of Address issues The homogeneous- Yes Yes Hong Kong Ho, 2010.
[GM (1,1)] gray sequences with small sample exponent simulative
size and/or poor deviation exists in
information GM (1, 1) model;
model improvements
are usually needed in
practice for the
accuracy of results
329

11
330 Economic indicators
331
332 A series of economic indicators have been applied in workforce forecasting models in the
333 construction industry. The key indicators mentioned three times or more are listed in Table 7.
334 Construction output was the most used indicator in the forecasting model (n = 9), followed by
335 project expenditure (n = 7) and labour deployment (n = 6). Labour wages and labour
336 productivity were identified five times each. Several studies also considered materials price (n
337 = 4), bank rate (n = 3), and GDP (n = 3) in their models.
338
339 Table 7. Key economic indicators in construction manpower forecast model
Economic indicators Studies Method involved No. of studies (n)
Construction output Ball and Wood, 1995; Econometric model 9
(gross value of construction work) Ho, 2013; Regression model
Liu et al., 2014; Time series analysis
Sing et al., 2014; Labour multiplier approach
Sing et al., 2012; Stock-flow model
Sing et al., 2016;
Wong et al., 2010;
Wong et al., 2007, 2011.
Project expenditure Bell and Brandenburg, 2003; Regression model 7
Chan et al., 2006; Labour multiplier approach
Chan et al., 2003;
Agarwal et al., 2013;
Persad et al., 1995;
Sing et al., 2012;
Wong et al., 2008.
Labour deployment Chan et al., 2006; Labour multiplier approach 6
(man-days or man-hours) Proverbs et al., 1999; Regression model
Rosenfeld and Warszawski, 1993; Stock-flow model
Sing et al., 2012;
Sing et al., 2015;
Sing et al., 2016.
Labour wage in construction Ball and Wood, 1995; Econometric model 5
Liu et al., 2014; Regression model
Wong et al., 2005, 2007, 2011. Time series analysis
Labour productivity Liu et al., 2014; Regression model 5
(gross construction output per man- Wong et al., 2005, 2007, 2008, 2011. Time series analysis
hour, construction output divided by Econometric model
total employed person and median
hours of work)
Material price Ball and Wood, 1995; Econometric model 4
Liu et al., 2014; Time series analysis
Wong et al., 2007, 2011.
Bank rate Liu et al., 2014; Econometric model 3
(interest rate) Wong et al., 2007, 2011. Time series analysis
GDP Sing et al., 2014; Stock-flow model 3
Sing et al., 2012; Labour multiplier approach
Vilutienė et al., 2014.
340
341 Of the various factors, construction output was identified as the most significant for
342 determining the employment level in the construction market (Table 7, mentioned by 9 studies).
343 At the same time, many prediction models take construction output as the main variable, which
344 reveals that the accuracy of construction output data has a great impact of the prediction result.
345 This is consistent with Wong et al. (2007). Construction output has a positive impact on
346 employment level, for example, a larger construction investment results in a higher workforce
347 demand (Briscoe and Wilson, 1993; Wong et al., 2011). Shrinkage in construction output can
348 severely affect job opportunities in the construction industry (Wong et al., 2007). Construction
349 output always accounts for some part of GDP in both developed and developing countries.
350 Several studies adopted GDP as a surrogate for construction output in explaining changes of
351 manpower demand (Sing et al., 2014; Sing et al., 2012; Vilutienė et al., 2014).
352

12
353 Project expenditure and labour deployment (man-hours or man-days) are the two important
354 factors in the multiplier approach, which is based on the assumption that construction projects
355 require the same level of labour per unit of project expenditure (Chan et al., 2006). The labour
356 wage can be negatively related to employment (Wong et al., 2011). A high labour wage in
357 construction may encourage construction companies to promote pre-fabrication for on-site
358 work (Ball and Wood, 1995). High labour costs will reduce manpower demand in an open
359 economy (Ross and Zimmermann, 1993). Wong et al. (2007) advocated that a positive
360 relationship between employment and wages exists (i.e. a larger number of employment
361 reflects a higher wage level). In practice, this positive relationship has been observed in both
362 construction (Wong et al., 2007) and manufacturing sectors (Crane and Nourzad, 1998).
363
364 Because of the labour shortage, companies are expected to offer higher wages to attract suitable
365 personnel (Williams, 2004). Therefore, the supply side of labour markets should be assessed
366 to achieve a comprehensive explanation of labour wage determination and wage elasticity. A
367 negative relationship was found between employment level and material prices (Liu et al.,
368 2014). Variables such as labour wages, capital costs, and material and fuels inputs are input
369 cost factors. A high input cost reduces the workforce demand level in the construction industry
370 (Ball and Wood, 1995; Briscoe and Wilson, 1993; Liu et al., 2014; Ross and Zimmermann,
371 1993). As the price of capital increases, companies substitute labour for capital, which is related
372 positively with bank rate (Wong et al., 2007). The decrease in bank rate could lead to an
373 increase in investment, thereby contributing more job opportunities (Briscoe and Wilson, 1993).
374
375 These above variables can be intimately correlated (Wong et al., 2005). Under a situation of
376 stable workforce supply, an increase in employment level might lower the
377 unemployment/underemployment rates, thus boosting labour wages level due to shortages in
378 specific occupations. Labour wage increases would consequently stimulate the employers to
379 cut costs by introducing technologies that are dependent less on labour and more on capital
380 (Ehrenberg and Smith, 2016; Wong et al., 2005). On the other hand, the rising labour
381 productivity as a consequence of technological progress is recognised as a significant indicator
382 of the level of labour employment (Rosenfeld and Warszawski, 1993; Wong et al., 2005). The
383 rising efficiency in the production process caused by technological improvements could cut
384 down the requirement for labour (Gruneberg, 1997). Construction labour productivity,
385 therefore, is usually taken as a proxy for the level of technological advancement (Nakanishi,
386 2001).
387
388 5. DISCUSSION
389
390 Studies of manpower modelling have been growing over time. Previous papers generally
391 focused on one specific model and selected the factors based on the motivations of their study
392 area. Studies on manpower planning and modelling is extensive and incongruent, and thus lack
393 of guidance on further development of human resource forecasting model in the construction
394 industry. This literature review compares and assesses these prediction models and the factors,
395 and consequently identifies the features that needed to improve the effectiveness in developing
396 a forecasting model, which will enhance the techniques of manpower planning for scholars and
397 human resource practitioners.
398
399 Approaches and techniques are becoming more comprehensive and complex. Quantitative
400 approaches dominate demand based and supply based models. Manpower forecasting takes
401 place at the industry and project levels. The former (i.e., industry level) is concerned with
402 manpower forecasting for the industry as a whole. Econometric analysis has been applied to

13
403 estimate how the various social-economic and political variables affect employment levels on
404 an industry-wide basis. Multiple regression modelling and time series analysis have been
405 employed as basic forecasting tools. Historical statistical data published by government units
406 and institutions have been used. The latter (i.e. project level) is more concerned with short term
407 time scales for project forecasting than the industrial models, which are often focused on the
408 long-term development of the economy and the provision of training and education
409 programmes (Briscoe and Wilson, 1993). The project-level models consider expenditures/costs
410 and labour deployment of a specific project in manpower forecasting. Project expenditures and
411 labour deployment (man-hours or man-days) are used in multiplier forecasting approaches in
412 deriving fixed coefficients for labour demand, the coefficient relating to a type of construction,
413 whether building, civil engineering or housing works (Chan et al., 2006).
414
415 5.1 Knowledge gaps
416
417 Change of system parameters
418
419 Most models are developed on the basis that behaviour is fixed or follows a certain pattern.
420 Types of behaviour may change in response to alterations in exogenous factors or new events.
421 An example was the global financial crisis in 2008, which had a significant impact on the world
422 economic climate (Jiang and Liu, 2011; Liu et al., 2014). Jiang and Liu (2011) considered the
423 global economic turbulence as an exogenous variable within their forecasting model. Another
424 example of exogenous factors is government policy, which has been examined in several
425 studies (Chang-Richards et al., 2017; Dabirian et al., 2018; Liu et al., 2014; Sing et al., 2011;
426 Sing et al., 2014; Sing et al., 2016; Wong et al., 2010). Inflation is viewed as a hidden
427 behavioural relationship that has been ignored by the majority of authors (Chan et al., 2006).
428 Inflation can have a potential impact although its influence is usually kept hidden when not at
429 a high level (Briscoe and Wilson, 1993). Sing et al. (2012) used an adjustment factor for
430 contract value to take inflation into account in their model. Therefore, consideration of a range
431 of scenarios with exogenous factors and hidden factors is essential, especially for a long-term
432 forecast. The sensitivity of a forecast is evaluated by establishing a number of possible
433 scenarios instead of relying on a single “fixed point” scenario (Briscoe and Wilson, 1993).
434
435 Constrained by limited data
436
437 A major problem associated with manpower forecasting is the availability and accessibility of
438 reliable data. The data on key indicators or variables can be lacking or inadequate in terms of
439 suitability, accuracy and timeliness. In the construction industry, information related to
440 employment for different occupations at the regional level is difficult to obtain and always
441 unreliable (Briscoe and Wilson, 1993). Several private firms collect industry data such as
442 industry projections and material costs; however, these are usually for-profit ventures, limiting
443 their applications to research and development (Rasdorf et al., 2016). Some government
444 agencies and institutions, such as the US Census Bureau and Bureau of Labour Statistics,
445 provide a wide range of data at no cost, such as employment level, labour wage, and
446 construction output. Researchers and organisations even investigate individual projects and
447 quantify labour data at the project level (Bell and Brandenburg, 2003; Dabirian et al., 2018).
448 However, access to accurate and consistent data at the regional level for the whole construction
449 industry is always elusive. As indicated in Table 2, Hong Kong is the most investigated region.
450 This has been attributed to the urge of the Hong Kong government to make manpower forecasts
451 to assist policy formulation and construction investment planning, which has consequently
452 resulted in many workforce consultancy and research studies (Chan et al., 2006). For

14
453 consultancy projects, the government provides high quality data in terms of accuracy,
454 consistency, completeness, and usefulness in construction, which accurately and effectively
455 conveys information. In other regions, for example, open data on employment level is always
456 annually available only in consolidated form for the complete construction industry
457 (occupational data are unavailable); additionally, there are changes in data categorisations and
458 collection methods which result in inconsistency; cash flows for a project are usually
459 unavailable for privacy reasons (Rasdorf et al., 2016). Consequently, it is recommended that
460 the construction community, including owners, contractors, government agencies and
461 researchers should provide improved high quality data provision resources and data (Rasdorf
462 et al., 2016).
463
464 5.2 Future studies
465
466 Current manpower forecasting mainly adopts conventional statistical techniques such as time
467 series modelling, regression analysis and multiplier approaches. Conventional statistical
468 methods require high quality data and large amounts of it, but may not capture the dynamic
469 changes. Since the 1990s, artificial intelligence (AI) techniques including artificial neural
470 network (ANN), genetic algorithm (GA), fuzzy logic and case-based reasoning, began to be
471 applied to solve complex construction management issues, such as forecasting construction
472 demand, optimising site operations, estimating tender bids, and estimating labour productivity
473 (Bee Hua, 2008). AI techniques have yet to be examined in relation to employment prediction
474 and planning in the construction industry. Future studies are needed into the adoption of AI
475 techniques for construction labour market prediction, of critical importance to construction
476 manpower planning.
477

15
478 6. CONCLUSION
479
480 This paper conducts a critical review of the existing manpower prediction model in the
481 construction industry. The results show that the total number of papers concerned with the
482 prediction model of human resources in the construction industry is increasing. The number of
483 regions that focus on manpower prediction is also arising gradually, among which, Hong Kong
484 generates the largest number of related papers. Most of the published papers are from colleges
485 and universities, followed by the government and enterprises.
486
487 In addition, this paper consists of a review of the concepts and features of workforce demand-
488 forecasting approaches in the construction industry. Reliable forecasting of manpower demand
489 and supply will help the construction market to provide a stable and well-planned supply of a
490 well-trained workforce. Appropriate manpower planning has been a significant contributor to
491 the recovery of the construction economy. This review has enhanced understanding of
492 workforce forecast methodologies and labour economic issues within the construction industry.
493 Specifically, this paper discusses the applicability and limitations of each prediction method.
494 It is found that different methods have different requirements of the quality of historical data,
495 and most of the forecasting methods are not suitable for emergencies in the labour market.
496 Except for the grey model, most quantitative forecasting models require higher accuracy of
497 historical data.
498
499 Then, this paper discusses the influential indicators in various forecasting models from an
500 economic perspective. Key factors including construction output, project expenditure, labour
501 deployment, labour wage, labour productivity, material price, bank rate, and GDP affecting the
502 construction labour market were identified and discussed. This study contributes as a reference,
503 for future studies of manpower planning in different regions and identifies the properties of
504 required data in terms of availability, coverage, continuity, sample size, and speed of
505 publication and stresses the importance of providing sufficient resources to ensure that data.
506 Although many scholars have focused on the manpower prediction model, few literatures
507 verify the model prediction accuracy. Secondly, most of the forecasting models are also limited
508 by the low quality of the original data, and the existing forecasting models are difficult to adapt
509 to the dynamic changes in the labour market. In the future, the use of artificial intelligence
510 prediction technology may solve this problem.

16
511 ACKNOWLEDGEMENT
512
513 The authors gratefully acknowledge the Development Bureau of the Government (Contract No.
514 WQ/088/17) of Hong Kong (SAR) and Research Grants Council GRF (F-PP6M) of Hong Kong
515 (SAR) for providing the funding enabling the commissioning of this study.
516

17
517 DATA AVAILABILITY STATEMENT
518
519 All data, models, and code generated or used during the study appear in the submitted article.
520

18
521 REFERENCES
522
523 Aaronson, S., Fallick, B., Figura, A., Pingle, J. F., and Wascher, W. L. (2006). The recent decline in the
524 labor force participation rate and its implications for potential labor supply. Brookings Papers
525 on Economic Activity, 2006(1), 69-154.
526 Abidoye, R. B., and Chan, A. P. (2017). Critical review of hedonic pricing model application in property
527 price appraisal: A case of Nigeria. International Journal of Sustainable Built Environment, 6(1),
528 250-259.
529 Ameyaw, E. E., Hu, Y., Shan, M., Chan, A. P., and Le, Y. (2016). Application of Delphi method in
530 construction engineering and management research: a quantitative perspective. Journal of Civil
531 Engineering and Management, 22(8), 991-1000.
532 Ball, M., and Wood, A. (1995). How many jobs does construction expenditure generate? Construction
533 Management and Economics, 13(4), 307-318.
534 Bartholomew, D. J., Forbes, A. F., and McClean, S. I. (1991). Statistical techniques for manpower
535 planning (second ed.). England: John Wiley and Sons.
536 Bee Hua, G. (2008). The state of applications of quantitative analysis techniques to construction
537 economics and management (1983 to 2006). Construction Management and Economics, 26(5),
538 485-497.
539 Bell, L. C., and Brandenburg, S. G. (2003). Forecasting construction staffing for transportation agencies.
540 Journal of Management in Engineering, 19(3), 116-120.
541 Betts, M., Robinson, G., Burton, C., Leonard, J., Sharda, A., and Whittington, T. (2015). Global
542 construction 2030: A global forecast for the construction industry 2030. London: Global
543 Construction Perspectives and Oxford Economics.
544 Briscoe, G., and Wilson, R. A. (1993). Employment forecasting in the construction industry. UK:
545 Ashgate Publishing.
546 Chan, A. P., Chiang, Y., Mak, S. W., Choy, L. H., and James, M. (2006). Forecasting the demand for
547 construction skills in Hong Kong. Construction Innovation, 6(1), 3-19.
548 Chan, A. P., Scott, D., and Chan, A. P. (2004). Factors affecting the success of a construction project.
549 Journal of Construction Engineering and Management, 130(1), 153-155.
550 Chan, A. P., Wong, J. M., and Chiang, Y. (2003). Modelling Labour Demand at Project Level—An
551 Empirical Study in Hong Kong. Journal of Engineering, Design and Technology, 1(2), 135-
552 150.
553 Chang-Richards, Y., Wilkinson, S., Seville, E., and Brunsdon, D. (2017). Effects of a major disaster on
554 skills shortages in the construction industry: Lessons learned from New Zealand. Engineering,
555 Construction and Architectural Management, 24(1), 2-20.
556 Crane, S. E., and Nourzad, F. (1998). Improving local manufacturing employment forecasts using
557 cointegration analysis. Growth and Change, 29(2), 175-195.
558 Crettenden, I. F., McCarty, M. V., Fenech, B. J., Heywood, T., Taitz, M. C., and Tudman, S. (2014).
559 How evidence-based workforce planning in Australia is informing policy development in the
560 retention and distribution of the health workforce. Human Resources for Health, 12(1), 7.
561 Dabirian, S., Abbaspour, S., and Ahmadi, M. (2018). An Improved Evaluation of Construction Project’s
562 Labor Need Based on Project Workflow. International Journal of Innovation, Management and
563 Technology, 9(1), 49-55.
564 Darko, A., and Chan, A. P. (2016). Critical analysis of green building research trend in construction
565 journals. Habitat International, 57, 53-63.
566 Darko, A., Zhang, C., and Chan, A. P. (2017). Drivers for green building: A review of empirical studies.
567 Habitat International, 60, 34-49.
568 Davis, C. (1987). Optimization of future manpower requirements in a multi-discipline consultancy.
569 Construction Management and Economics, 5(1), 45–56.
570 Ehrenberg, R. G., and Smith, R. S. (2016). Modern labor economics: Theory and public policy. New
571 York: Routledge.
572 Fraher, E. P., Knapton, A., Sheldon, G. F., Meyer, A., and Ricketts, T. C. (2013). Projecting surgeon
573 supply using a dynamic model. Annals of Surgery, 257(5), 867-872.
574 Garza, A., Kumara, S., Martin, J. D., and Finke, D. A. (2013). Multi-agent system for manpower
575 modeling. IIE Annual Conference. Proceedings, 187-196.

19
576 Gnatzy, T., Warth, J., von der Gracht, H., and Darkow, I.-L. (2011). Validating an innovative real-time
577 Delphi approach - A methodological comparison between real-time and conventional Delphi
578 studies. Technological Forecasting and Social Change, 78(9), 1681–1694.
579 Gruneberg, S. L. (1997). Construction economics: An introduction. London: Macmillan International
580 Higher Education.
581 Hasson, F., and Keeney, S. (2011). Enhancing rigour in the Delphi technique research. Technological
582 Forecasting and Social Change, 78(9), 1695-1704.
583 Ho, P. H. (2010). Forecasting construction manpower demand by Gray model. Journal of Construction
584 Engineering and Management, 136(12), 1299-1305.
585 Ho, P. H. (2013). Forecasting the manpower demand for quantity surveyors in Hong Kong.
586 Construction Economics and Building, 13(3), 1-12.
587 Howard, G. S., Cole, D. A., and Maxwell, S. E. (1987). Research productivity in psychology based on
588 publication in the journals of the American psychological association. American Psychologist,
589 42(11), 975.
590 Hsu, C.-C., Chen, S.-H., and Hsien, B.-C. (2012). The manpower forecast model of the energy
591 technology industry. Journal of Statistics and Management Systems, 15(4-5), 499-517.
592 International Labour Organization (ILO). (2017). ILO Labour Force Estimates and Projections: 1990–
593 2030. Methodological Description. Geneva.
594 Jiang, H., and Liu, C. (2011). Forecasting construction demand: A vector error correction model with
595 dummy variables. Construction Management and Economics, 29(9), 969-979.
596 Jiang, H. J., and Begun, J. W. (2002). Dynamics of change in local physician supply: An ecological
597 perspective. Social Science and Medicine, 54(10), 1525-1541.
598 Ke, Y., Wang, S., Chan, A. P., and Cheung, E. (2009). Research trend of public-private partnership in
599 construction journals. Journal of Construction Engineering and Management, 135(10), 1076-
600 1086.
601 Kinstler, D. P., Johnson, R. W., Richter, A., and Kocher, K. (2008). Navy Nurse Corps manpower
602 management model. Journal of Health Organisation and Management, 22(6), 614-626.
603 Kwak, N., McCarthy, K. J., and Parker, G. E. (1997). A human resource planning model for
604 hospital/medical technologists: An analytic hierarchy process approach. Journal of Medical
605 Systems, 21(3), 173-187.
606 L Agarwal, A., Rajput, B., Mangesh, A., and Satpute, E. (2013). Model formulation to estimate
607 manpower demand for the real-estate construction projects in India. Organisation, Technology
608 and Management in Construction: An International Journal, 5(2), 828-833.
609 Lai, J. H. (2017). Building operation and maintenance: manpower in Hong Kong. Facilities, 35(3/4),
610 220-241.
611 Landeta, J. (2006). Current validity of the Delphi method in social sciences. Technological Forecasting
612 and Social Change, 73(5), 467–482.
613 Lawrence, M., Goodwin, P., O'Connor, M., and Önkal, D. (2006). Judgmental forecasting: A review of
614 progress over the last 25 years. International Journal of Forecasting, 22(3), 493-518.
615 Li, H. (2019). Estimation of stadium construction schedule based on big data analysis. International
616 Journal of Computers and Applications, 41(4), 268-275.
617 Liu, J., Love, P. E., Sing, M. C., Carey, B., and Matthews, J. (2014). Modeling Australia’s construction
618 workforce demand: Empirical study with a global economic perspective. Journal of
619 Construction Engineering and Management, 141(4), 05014019.
620 Love, P. E., Holt, G. D., and Li, H. (2002). Triangulation in construction management research.
621 Engineering, Construction and Architectural Management, 9(4), 294-303.
622 Nakanishi, Y. (2001). Dynamic labour demand using error correction model. Applied Economics, 33(6),
623 783-790.
624 Ngai, L. R., and Pissarides, C. A. (2005). Trends in labour supply and economic growth. Center for
625 Economic Performance, London School of Economics.
626 Noe, R. A., Hollenbeck, J. R., Gerhart, B. A., and Wright, P. M. (2017). Fundamentals of human
627 resource management (7th ed.). New York: McGraw-Hill Education.
628 Parker, B., and Caine, D. (1996). Holonic modelling: human resource planning and the two faces of
629 Janus. International Journal of Manpower, 17(8), 30-45.
630 Pate, J. E. (1943). Mobilizing Manpower. Social Forces, 22(2), 154–162.

20
631 Persad, K. R., O'Connor, J., and Varghese, K. (1995). Forecastng engineering manpower requirements
632 for highway preconstruction activities. Journal of Management in Engineering, 11(3), 41-47.
633 Proverbs, D., Holt, G. D., and Olomolaiye, P. (1999). A method for estimating labour requirements and
634 costs for international construction projects at inception. Building and Environment, 34(1), 43-
635 48.
636 Rasdorf, W., Hummer, J. E., and Vereen, S. C. (2016). Data collection opportunities and challenges for
637 skilled construction labor demand forecast modeling. Public Works Management and Policy,
638 21(1), 28-52.
639 Reeves, G. R., and Reid, R. C. (1999). A military reserve manpower planning model. Computers and
640 operations research, 26(12), 1231-1242.
641 Reifels, L., Naccarella, L., Blashki, G., and Pirkis, J. (2014). Examining disaster mental health
642 workforce capacity. Psychiatry: Interpersonal and Biological Processes, 77(2), 199-205.
643 Rosenfeld, Y., and Warszawski, A. (1993). Forecasting methodology of national demand for
644 construction labour. Construction Management and Economics, 11(1), 18-29.
645 Ross, D. R., and Zimmermann, K. F. (1993). Evaluating reported determinants of labor demand. Labour
646 Economics, 1(1), 71-84.
647 Safarishahrbijari, A. (2018). Workforce forecasting models: A systematic review. Journal of
648 Forecasting, 37(7), 739-753.
649 Santric-Milicevic, M., Vasic, V., and Marinkovic, J. (2013). Physician and nurse supply in Serbia using
650 time-series data. Human Resources for Health, 11(1), 27.
651 Schaible, W., and Mahadevan-Vijaya, R. (2002). World and Regional estimates for selected key
652 indicators of the labour market. Geneva: International Labour Organization.
653 Shen, H., and Huang, J. Z. (2008). Forecasting time series of inhomogeneous Poisson processes with
654 application to call center workforce management. The Annals of Applied Statistics, 2(2), 601-
655 623.
656 Simonds, S. K. (1976). Health Education Manpower in the United States. Health Education
657 Monographs, 4(3), 208-225.
658 Sing, C.-p., Love, P., and Tam, C. M. (2011). Stock-flow model for forecasting labor supply. Journal
659 of Construction Engineering and Management, 138(6), 707-715.
660 Sing, C.-P., Love, P. E., and Tam, C.-M. (2014). Forecasting the demand and supply of technicians in
661 the construction industry. Journal of Management in Engineering, 30(3), 04014006.
662 Sing, C.-p., Love, P. E., and Tam, C. M. (2012). Multiplier model for forecasting manpower demand.
663 Journal of Construction Engineering and Management, 138(10), 1161-1168.
664 Sing, C., Chan, H. C., Love, P., and Leung, A. (2015). Building Maintenance and Repair: Determining
665 the Workforce Demand and Supply for a Mandatory Building-Inspection Scheme. Journal of
666 Performance of Constructed Facilities, 30(2), 04015014.
667 Sing, M. C., Love, P., Edwards, D., and Liu, J. (2016). Dynamic modeling of workforce planning for
668 infrastructure projects. Journal of Management in Engineering, 32(6), 04016019.
669 Solnet, D. J., Baum, T., Kralj, A., Robinson, R. N., Ritchie, B. W., and Olsen, M. (2014). The Asia-
670 Pacific tourism workforce of the future: Using Delphi techniques to identify possible scenarios.
671 Journal of Travel Research, 53(6), 693-704.
672 Tang, C. H., and Hsu, I. S. (2019). Manpower supply planning for aircraft line maintenance under
673 stochastic incidents. Journal of the Chinese Institute of Engineers, 42(4), 333-345.
674 Toder, E. J., and Solanki, S. (1999). Effects of Demographic Trends on Labor Supply and Living
675 Standards. Washington: Urban Institute Press.
676 Vilutienė, T., Podvezko, V., Ambrasas, G., and Šarka, V. (2014). Forecasting the demand for blue-
677 collar workers in the construction sector in 2020: the case of Lithuania. Economic research-
678 Ekonomska istraživanja, 27(1), 442-462.
679 Williams, R. D. (2004). The demand for labour in the UK. Labour Market Trends, 112(8), 321-330.
680 Wong, F. K., Hui, E. C., Wong, J. T., and Wan, J. K. (2010). The impact of urban renewal to the labour
681 force in Hong Kong. Facilities, 28(13/14), 611-640.
682 Wong, J., Chan, A., and Chiang, Y. (2012). A critical review of forecasting models to predict manpower
683 demand. Construction Economics and Building, 4(2), 43-56.

21
684 Wong, J. M., Chan, A. P., and Chiang, Y. (2005). Time series forecasts of the construction labour
685 market in Hong Kong: the Box‐Jenkins approach. Construction Management and Economics,
686 23(9), 979-991.
687 Wong, J. M., Chan, A. P., and Chiang, Y. (2007). Forecasting construction manpower demand: A vector
688 error correction model. Building and Environment, 42(8), 3030-3041.
689 Wong, J. M., Chan, A. P., and Chiang, Y. (2008). Modeling and forecasting construction labor demand:
690 Multivariate analysis. Journal of Construction Engineering and Management, 134(9), 664-672.
691 Wong, J. M., Chan, A. P., and Chiang, Y. (2010). Modeling construction occupational demand: Case
692 of Hong Kong. Journal of Construction Engineering and Management, 136(9), 991-1002.
693 Wong, J. M., Chan, A. P., and Chiang, Y. (2011). Construction manpower demand forecasting: A
694 comparative study of univariate time series, multiple regression and econometric modelling
695 techniques. Engineering, Construction and Architectural Management, 18(1), 7-29.
696 Wu, D. C., Song, H., and Shen, S. (2017). New developments in tourism and hotel demand modeling
697 and forecasting. International Journal of Contemporary Hospitality Management, 29(1), 507-
698 529.
699 Yi, W., and Chan, A. P. (2013). Critical review of labor productivity research in construction journals.
700 Journal of Management in Engineering, 30(2), 214-225.
701 Yuan, H., and Shen, L. (2011). Trend of the research on construction and demolition waste management.
702 Waste management, 31(4), 670-679.

22