Cash Flow Forecasting Model for General Contractors Using Moving Weights of Cost Categories

Cash Flow Forecasting Model for General Contractors Using
Moving Weights of Cost Categories
Hyung K. Park1; Seung H. Han2; and Jeffrey S. Russell3
Abstract: This research introduces the development of a project-level cash flow forecasting model from a general contractor’s viewpoint.
While most previous models have been proposed to assist contractors in forecasting cash flow in the early stage of pretendering or the
planning phase, this paper aims to provide a tool that can be applicable during the construction phase based on the planned earned value
and the actual incurred cost on a jobsite level. The critical key to cash flow forecasting at this level lies in how to build a realistic cash-out
model. Toward the end, this paper adopts moving weights of cost categories in a budget that are variable depending on the progress of
construction works. In addition, it addresses time lags in accordance with the contractual payment conditions and credit times given by
suppliers or vendors. As for the cash-in model, net planned monthly earned values are simply transferred to the cash-in forecast with a
consideration of billing time and retention money. Validation of the proposed model involves applying realistic data from four ongoing
projects. Based on the results of comparative analyses, the writers conclude that the proposed model is more accurate and reliable, yet
simpler to field engineers who are generally not familiar with certain intricate financial knowledge.
DOI: 10.1061/ ASCE 0742-597X 2005 21:4 164
CE Database subject headings: Financial management; Forecasting; Contractors; Cost control; Construction industry.
Introduction
integrated schedule and cost items using a simulation model ap-
plied to the stochastic duration of the activities. However, it does
Background
not consider the time lag’s impact on costs, which is essential in
cash flow forecasting. The technique proposed by Sears 1981 is
Cash is the most important of a construction company’s re-
viewed accurately by manually integrating the schedule and cost
sources. More construction companies fail due to a lack of liquid-
items, but it requires considerable work and further, it does not
ity for supporting their daily activities than because of inadequate
consider the time lag between the expenditure and payment of a
management of other resources Singh and Lakanathan 1992;
related cost item. Navon’s model 1995, 1997 automatically in-
Navon 1994 . Russell 1991 pointed out that more than 60% of
tegrates the bill of quantity BOQ , cost estimate, and the sched-
construction contractor failures are due mainly to economic fac-
ule associated with a lower level of resources. However, if either
the BOQ or the schedule is altered due to various changes, inte-
tors. In an attempt to analyze the real business environment in the
gration is likely to be more complicated and time consuming.
construction industry, various forecasting methods have been ap-
Moreover, the main obstacle to automating the integration process
plied to cash flow management.
is compatibility between cost items of the BOQ and activity ele-
Numerous techniques for cash flow forecasting and manage-
ments of schedule.
ment differ in their levels of accuracy and detail, the degree of
In an attempt to improve the accuracy of a model for forecast-
automation in compiling them, and the method to integrate the
ing cash flow, Ashley and Teicholz 1977 suggested a cash flow
time and money elements. Some of the techniques are probabilis-
forecast based on detailed methods of cost flow. They classified
tic, but most of them are deterministic Navon 1995 . Rein-
the direct cost by a number of cost categories such as labor, ma-
schmidt and Frank 1976 proposed a model for cash flow fore-
terials, and equipment which are specified as percentages of total
casting in the early planning stage of a project. This model
cost. This approach is very realistic because it considers the na-
ture of the budget and cost. However, each of these cost elements
1
is assumed to be a fixed percentage of total cost over the project’s
PhD, General Manager, G.K CM Team, Daewoo Engineering &
duration. Moreover, this model does not consider the effect of
Construction Co. Ltd., Seoul, Korea. E-mail: parkhk@dwconst.co.kr
2
time lags on the costs. Also, Gates and Scarpa 1979 and Peer
Associate Professor, Dept. of Civil and Environmental Engineering,
Yonsei
Univ.,
Seoul,
Korea
corresponding
author .
E-mail:
1982 developed cash flow models in the conceptual and plan-
shh6018@yonsei.ac.kr
ning stages using algebraic formulations and polynomial regres-
3Professor, Dept. of Civil and Environmental Engineering, Univ. of
sions. However, none of these models considered time lags to the
Wisconsin, Madison, WI. E-mail: russell@engr.wisc.edu
costs and earned values.
Note. Discussion open until March 1, 2006. Separate discussions must
In reality, many factors exist during construction that may af-
be submitted for individual papers. To extend the closing date by one
fect the cash flow including time delays, cost overruns, uncon-
month, a written request must be filed with the ASCE Managing Editor.
firmed earned values, change orders, and changes of cost plan
The manuscript for this paper was submitted for review and possible
elements Bennett and Ormerod 1984 . The key points of cash
publication on November 2, 2004; approved on March 18, 2005. This
paper is part of the Journal of Management in Engineering, Vol. 21, No.
flow forecasts lie in how accurate, flexible, and comprehensive
4, October 1, 2005. ©ASCE, ISSN 0742-597X/2005/4-164–172/$25.00.
they are to be calculated and how effectively they consider uncer-
164 / JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / OCTOBER 2005

---

tain factors such as time delay, cost overrun, variation of cost, and
can be more accurate, flexible, and simpler to typical field engi-
earned value between plan and actual. Of course, it is impossible
neers on a job site.
to ensure that a project will definitely be as successful as initially
planned. Even though construction is in progress, cash flow fore-
casts cannot be determined precisely. As a result, most models
Cash Flow to General Contractor
and techniques aforementioned are found to have the following
problems: 1 they are not based on the construction stage, but
Typical Project Cash Flow
rather only on the planning or preliminary stages in the project
delivery process; 2 they do not consider time lags for the costs
Most construction projects are individual profit centers, each with
and earned values in forecasting cash flow; and 3 with regard to
its own cash cycle based on the costs of activities related to the
integration of cost items and activities, they are not compatible
project and on payments from a client, both of which are pre-
with each item and are rather complicated depending on when
scribed by a contract. Typical cash flow on a construction project
consists of: 1 cash out such as bid costs, preconstruction costs
change factors occur in the subsequent construction stage.
engineering, design, mobilization, etc. , materials and supplies,
Because cash flow is a reality, a cash flow forecast on a job site
equipment and equipment rentals, payments of subcontracts, labor
should be more precise than those during the preconstruction
and overhead; and 2 cash in such as billings less retentions ,
phases by addressing the uncertainties of the construction busi-
retentions, claims and change orders. The factors that affect cash
ness and jobsite procedures. The main objectives of this paper
flows are the duration of the project, the retention conditions, the
are: 1 to quantitatively study construction project cash flows; 2
times for receiving payments from the client, credit arrangement
to propose a forecasting cash flow model for construction projects
with suppliers or vendors, equipment rentals, and times of pay-
with a consideration of both variable cost weights and a time lag;
ments to subcontractors, etc.
and 3 to validate the proposed model and suggest guidelines for
Cash flow at the project level consists of a complete history of
implementing this cash flow forecasting system. In addition, this
all cash disbursement and all earnings received as a result of
paper provides implications to management by focusing more on
project execution. Many construction projects have negative net
how project managers or field engineers can benefit by using the
cash flows until the very end of construction when the final pay-
presented model.
ment is received or advanced payment is received before starting
the project. This is a typical situation when the final payment
Research Scope and Methodology
consists of retention money and the retention percentage is greater
than the profit percentage of the project.
Among a variety of types of construction projects, this research is
focused on bid projects. The proposed model is intended to be
Structure of Construction Budget
applicable to the construction stage in the project delivery process
from a general contractor’s viewpoint. Accordingly, the research
A budget structure in construction projects is constituted of cost
scope does not include investment projects such as Build–
accounts such as bills, sections, items, and resources. A budget
Operate–Transfer or Build–Operate–Own. Moreover, this re-
is a plan for allocating resources Meredith and Mantel 1995 .
search is relevant to the viewpoint of cash flow management at
Hendrickson and Au 1989 identified the fact that allocation of a
the project level, and subsequently project evaluation is per-
cost to the budget may be used to develop the cost function of an
formed by allowing contractors to reflect the capital cost or so-
operation. The basic idea in this method is that each expenditure
called, interest cost whenever negative cash flows occur.
item can be assigned to particular categories of operation. Ideally,
To achieve its end, methodology should possess several nec-
the allocation item of joint costs should be causally related to the
category of basic costs in an allocation process. Generally, a bud-
essary steps. As an initial step to meet the objectives, previous
get structure in construction projects is set up into labor, material,
research papers that deal with cash flow management are re-
equipment, subcontract, and indirect expenses. If a general con-
viewed to investigate problems with existing cash flow forecast-
tractor performs all the areas of job management on site, expenses
ing models. It should then suggest a new model of cash flow
for management and overhead cost become a higher burden to the
forecasting for a jobsite using a new algorithm. A numerical ex-
general contractor. To mitigate these costs, general contractors
ample is prepared to demonstrate and verify the computational
prefer to distribute a role of management to other participants. As
aspects of the model. The next step is to perform a simulation
an example, if a portion of a subcontract is increased, the general
using experimental data and to compare the model results to ex-
contractor is able to decrease the indirect expenses used for hiring
isting models proposed by other researcher. The last step of this
project personnel: the workers, supervisory personnel, and engi-
research is to validate the model. Although the model is devel-
neers associated with the project. From the general contractor’s
oped to offer a practical guideline to improve forecasting quality
viewpoint, labor and equipment costs are uncertain because pro-
in evaluating the cash flow on a job site, objectively assessing the
ductivity is extremely volatile and hard to measure. For this rea-
validity of the model in a real business scheme is quite difficult.
son, general contractors attempt to hire subcontractors to reduce
Accordingly, a comparative case study methodology is chosen as
job-management costs and to maximize their profit opportunity
a proper validation approach to the research features. Four
by concentrating their control ability on variable costs, uncertain
projects in progress, including one building project and three civil
time, and strict quality. Typically, the portions of subcontract cost
projects, with data compiled over a duration of 12 months are
range from 50 to 70%. Material, labor, equipment, and indirect
identified as the case study materials. Based on the results of
cost are arrayed between 25 and 35%, 5 and 15%, 10 and 25%,
comparative analyses, we measure to see if the proposed model
and 5 and 15%, respectively Oberlender 2000 .
JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / OCTOBER 2005 / 165

---

Jobsite Cash Flow Forecast Model
Cash-Out Model
Time Lag
The critical key to cash flow forecasting at the project level is
how to build a cash-out model. All resources to be incurred to
costs in a budget have different time lags. They are subject to
contracting procedures and a corporation’s payment policy to
other organizations. Accordingly, cash-out forecasts set the tone
for time lags. Cost categories are classified in order to compile
construction resources with similar time lags. Time lag, as used
here, is based on contracting payment conditions and credit times
given by suppliers or vendors.
Ahuja and Walsh 1983 also insist that there are delays be-
Fig. 1. Comparison of weights of costs during construction period
tween the dates of costs incurred and the dates of payment due.
These delays will vary depending on resource types and credit
arrangements as negotiated with subcontractors and suppliers.
For that reason, whenever costs are incurred in a periodic
This approach is maintained by a number of previous researchers
month, weights of cost categories relative to the remaining budget
Peterman 1973; Ashley and Teicholz 1977; McCaffer 1979;
are changed, even though neither the overall budget the forecast
Trimble 1982; Kenley and Wilson 1989; Navon 1995; Kaka
total cost nor the planning for execution is changed. Moreover, if
1996 .
a change of project amount or project duration occurred due to a
Different cost categories are defined for materials, labor,
change order or a change of contract conditions, weights of cost
equipment, subcontractors, indirect expenses site overhead , and
categories should also be adjusted Park 2001 .
depreciation items since these cost categories generally have dif-
Consequently, this implies that the next weight of a cost cat-
ferent time lags. If additional cost categories are needed, they can
egory to be applied will be set in accordance with the cumulative
be classified. As mentioned before, since payment conditions of
actual cost and the remaining budget. Thus, “the moving weights
subcontracts are controlled by general contractor policy, it can be
method” continuously changes over the project duration to pertain
noted that the general contractors entail 50–70% certainty in cash
to the remaining budget. Applying moving weights of cost cat-
flow forecasting regarding time lags. The only remaining prob-
egories to the remaining budget in a month time series reduces
lems are how to determine time lags of other cost categories and
the uncertainty of forecasting cash out for the remaining duration
how to plan a budget for each period.
of the project. This characteristic of a budget during the construc-
Jepson 1969 suggested that net cash flow for individual
tion period is illustrated in Fig. 2.
projects must be derived from “component” curves of inflow and
outflow profiles. Fondahl and Bacarreza 1972 claimed that total
Cash-In Model
costs can be broken down as to category since different cost re-
sources may have different cost curves or different time lags re-
Billing Time
lated to their payment.
Generally, earned values will be received on a monthly basis or
based on billing terms, but planning of earned values on a jobsite
Moving Weights of Cost Categories
is established by a monthly amount. Earned value planning is the
Ashley and Teicholz 1977 developed five cost curves for cost
basis for estimated cash-in values in actual cash flow analysis.
categories in their highway construction project. Fondahl and
Net planned monthly earned values are simply transferred to the
Bacarreza 1972 also applied three cost curves to their school
cash-in forecast, to be applied there with appropriate time lags.
project Curves 1, 2, and 3 . Curve 1 is based on the assumption
The billing period, the time between the dates of bill submittal
that the rate of expenditure will be uniform over the project du-
and the progress payment receipt, is stipulated in the contract. If a
ration. Curve 2 assumes that only 25% of the total cost is incurred
payment delay occurs due to the owner’s circumstances, the bill-
during the first half of the project duration and the remaining 75%
ing time of cash in can be adjusted in this model. In practice,
in the second half. Curve 3 assumes that 75% of the total cost is
billing terms in the contract should provide for a billing schedule
incurred in the first half of project duration. In their research, only
for owner and contractor, but those terms can be applied variously
depending on the owner’s financing situation.
field overhead and home office overhead costs were analogous to
Curve 1, which implies that only these costs were assumed to be
incurred at a uniform rate over the project duration.
In other words, all cost categories except field overhead and
home office overhead were not incurred at a uniform rate over the
project lifetime. Unless the curves of all cost categories are uni-
form, the relative weights of the different cost categories should
be changed whenever costs are incurred over the project duration.
If weights of cost categories are uniform over the project dura-
tion, curves of all categories should represent straight lines. The
concepts of the moving weights method and fixed weights method
are illustrated in Fig. 1.
Fig. 2. Characteristic of budget during construction period
166 / JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / OCTOBER 2005

---

Retention Money
Cash-in planning should consider the effects of retention money
and the billing period on earned values. Retention money is based
on a percentage of retention stipulated in the contract. A cumula-
tive cash-in curve is obtained from the cumulative earned value
curve by applying a retention rate and billing period. Generally,
contractors can improve cash flow by providing percent retention
schedules in contracts with subcontractors. Then, the retention
money is released when construction is completed and accepted.
If cash in is properly planned and manipulated by a model, it will
supply the funds necessary to meet the cash requirements of the
project without borrowing from other organizations.
Mathematical Algorithm of Model
Cash Out
The model algorithm for cash out can be represented by equa-
tions. In cash out, the cost categories in an initial budget can be
classified depending on the time lags of all resources in the bud-
get. After that, the following equation is applied:
initial weight wi = Ci ÷ TB
1
where i = cost categories; Ci= budget of individual cost categories;
and TB= initial total budget total costs .
Whenever deviation between actual and planned data occurs,
an adjustment of weight is calculated and applied to the next cash
Fig. 3. Process of model
planning. Since actual cost in accordance with initial weight of
each cost category in each month is not incurred, actual cost and
actual earned value should be reflected in the next weights of
TCt= actual cumulative total cost; and FCt+1,i= actual cash-out
individual categories. The weight is called the “moving weight”
flow.
in this research. Therefore, the next moving weight to be applied
is
Cash In
Earned value is converted to cash in by deducting retention and
moving weight
wi = Ci ÷ TB
2
applying billing time. This model considers that most contractors
where
Ci= remaining budget of individual cost category and
withhold retention from subcontractors at the same rate they are
TB= remaining total budget.
withheld by the owner. Therefore cash in should consider two
From Eqs. 1 and 2 , the constraints on weights of the indi-
kinds of retention money: contractors’ retention and subcontrac-
vidual cost categories can be represented by the following
tors’ retention. Hence, the cash in is calculated as follows:
equations:
CIt = Vt
1 − rc + rs
St
8
wi = 1
3
where CIt= cash in at the time t; Vt= earned value at the time t;
or
rc= contractual retention rate; rs= subcontractual retention rate;
and St= subcontract cost at the time t.
wi = 1
4
Depending on the contractual agreement, release of retention
where i = individual cost categories.
is prescribed in two ways: first at the completion of the contract
As a result, equations for the moving weights cash-out model
and second, at the end of the maintenance period. The model is
are as follows. In terms of this model, the algorithm can be con-
simulated by entering the figures of release of retention by con-
tinuously updated to the weight to be applied in each month over
tractors whenever the subcontract ends.
the project duration
Model Process
Ft+1,i = wt+1,i
Ct+1
5
Fig. 3 illustrates an integrated process of model to cash flow
forecasting. It consists of three steps designed for general contrac-
Ct,i − ACt,i
w
tors on a job site level to evaluate cash flow. The first step re-
t+1,i =
6
TBt − TCt
quires input data for evaluating each individual project, such as
planned earned values and budget cost to each month, cost cat-
AC
egories, weights, and time lags. If more cost categories due to
i =
FCt+1,i
7
different time lags are required, the users can classify separate
where Ft+1,i= forecast of individual cost categories of time series
cost categories.
in period t + 1; Ct+1= planned costs of time series in period t + 1;
The second step updates new weights to cost categories re-
wt+1,i= weights of cost categories of time series in period t + 1;
flected on actual cost. Also, forecast cash flow such as cash in,
AC,i= actual cumulative cost of individual cost categories;
cash out, cumulative cash flow, and capital cost are automatically
JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / OCTOBER 2005 / 167

---

Table 1. Example of Cash Forecasting at Start of Projecta
Planed
Time period
earned
Planned
Actual
Actual
Cash
Cash
Cumulative
Interest
days
value
budget
value
cost
in
out
Balance
balance
10%
Depreciation
0
0.00
0.00
30
1,097
1,072
—
—
—
83.08
−132.39
−132.39
−1.09
49.31
61
1,159
1,130
—
—
—
87.58
−139.56
−271.95
−2.24
51.98
91
1,104
1,095
—
—
—
84.86
−135.23
−407.18
−3.35
50.37
122
1,106
1,084
—
—
—
84.01
−133.87
−541.05
−4.45
49.86
152
982
975
—
—
5,448.00
75.56
5,327.59
4,786.53
39.34
44.85
183
627
620
—
—
—
987.66
−1,016.18
3,770.36
30.99
28.52
213
716
709
—
—
—
1,045.39
−1,078.01
2,692.35
22.13
32.61
244
997
972
—
—
—
1,035.10
−1,079.81
1,612.54
13.25
44.71
274
1,183
1,176
—
—
3,523.00
1,041.27
2,427.64
4,040.18
33.21
54.10
305
1,302
1,288
—
—
—
954.41
−1,013.66
3,026.52
24.88
59.25
335
1,173
1,152
—
—
—
632.71
−685.70
2,340.82
19.24
52.99
365
1,083
1,059
—
—
—
703.51
−752.23
1,588.60
13.06
48.71
396
—
—
—
—
—
851.96
−851.96
736.64
6.05
—
426
—
—
—
—
—
1,030.76
−1,030.76
−294.13
−2.42
—
457
—
—
—
—
—
1,128.93
−1,128.93
−1,423.06
−11.70
—
487
—
—
—
—
—
1,009.73
−1,009.73
−2,432.79
−20.00
—
518
—
—
—
—
—
928.21
−928.21
−3,361.00
−27.62
—
Sum
12,529
12,332
—
—
8,971.00
11,764.73
−3,361.00
—
129.30
567.27
Material
Labor
Depreciation
Equipment
Main
Sub-Con
Expense
Sum
Billing
Retainage
materials
time
Time lag days
150
0
0
0
0
150
0
120
0
Planned budget
4,033.80
456.28
567.27
27.13
0
6,775.20
472.32
12,332
Initial weight %
32.71%
3.70%
4.60%
0.22%
0.00%
54.94%
3.83%
100.00%
0%
aUnit=1,000 US dollars.
calculated. This stage is based on moving weights of each classi-
model since that is not generally considered as a job or project
fied cost category in each month. Moving weight is that weight to
cost. That is incurred at the company level and accordingly may
be applied to the next month that is adjusted and calculated by
be billed directly on a jobsite.
deducting the actual cost from the initial budget to the individual
classified cost category in each month. Therefore, a weight of
Illustrative Example
each budget of each individual cost category to the remaining
budget is to be changed every month.
To illustrate the new methodology proposed, we have conducted a
The final step provides feedback to estimate the new planned
simple case study. The illustrative case is composed with the fol-
earned values and budget for each month. Whenever deviations
lowing figures: project duration is 12 months, contract amount is
between planned and actual costs and earned values occur, they
US$12,529,000, and budget is US$12,332,000. Input variables for
are automatically distributed over the remaining duration if
the case are: 1 planned earned values PE , planned budget
needed. If deviations between them are considerably more or less
PB , actual earned value AE , and actual cost AC at each
than expected, the project manager must modify the initial plan-
month; 2 cost categories classified based on contract procedure;
ning to forecast cash flow.
3 weight percentage and credit time of each cost category; 4
As a basis for applying the proposed model, this paper sets up
billing time and percentage of retention to be stipulated in con-
basic assumptions: 1 in the initial time period, the planned
tract; and 5 percentage of interest to be applied by corporate
earned value to the contract amount and planned cost to the bud-
policy or decision. According to investigations by Singh and La-
get are not automatically generated each month. Instead they are
kanathan 1992 , the application of “S curves” for cash flow pro-
made independently by engineers on the jobsite by their own
jections can achieve an accuracy of approximately 88–97%. Sub-
method of planning. 2 Time lags of cost categories are based on
sequently, input data at each month are based on S curves.
corporate historical data and company policy. 3 Cost categories
The basic assumptions applied to the case study are: 1
classified at the start of a project have to continuously be used in
changes of AC and AE against PE and PB at each month are
order to maintain the degree of accuracy in moving weight over
addressed, but overrun of budget and delay of duration are not
the project duration. 4 This model is used to forecast cash flow
considered; 2 cost categories depending on time lags are simply
values at the close of each month last day of the month . 5
classified as labor 0 days , materials 150 days , rent equipment
Depreciation of company owned equipment is included in actual
0 days , depreciation of owned equipment 0 days , subcontract
cash transfer incurred cost in order to show cash flow at the
150 days , and field expense 0 days ; 3 billing time to earned
project level. 6 Home office overhead is not considered in this
value is 120 days and percent of retainage is 0% of earned value
168 / JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / OCTOBER 2005

---

Table 2. Example of Updated Cash Forecasting after “1” Montha
Time period
Planed
Planned
Actual
Actual
Cumulative Interest
days
value
budget
value
cost
Cash in
Cash out
Balance
balance
10%
Depreciation
0
0
0.00
30
1,097
1,072
1,090
729
—
87.48
−121.01
−121.01
−0.99
33.53
61
1,159
1,130
—
—
—
84.56
−136.54
−257.55
−2.12
51.98
91
1,104
1,095
—
—
—
81.94
−132.31
−389.86
−3.20
50.37
122
1,106
1,084
—
—
—
81.12
−130.98
−520.84
−4.28
49.86
152
982
975
—
—
5,441.00
72.96
5,323.19
4,802.35
39.47
44.85
183
627
620
—
—
—
654.38
−682.90
4,119.45
33.86
28.52
213
716
709
—
—
—
1,046.52 −1,079.13
3,040.32
24.99
32.61
244
997
972
—
—
—
1,035.43 −1,080.14
1,960.18
16.11
44.71
274
1,183
1,176
—
—
3,523.00
1,041.02
2,427.88
4,388.07
36.07
54.10
305
1,302
1,288
—
—
—
953.57 −1,012.82
3,375.25
27.74
59.25
335
1,173
1,152
—
—
—
631.29
−684.28
2,690.96
22.12
52.99
365
1,083
1,059
—
—
—
702.58
−751.29
1,939.67
15.94
48.71
396
—
—
—
—
—
854.55
−854.55
1,085.12
8.92
—
426
—
—
—
—
—
1,033.90 −1,033.90
51.22
0.42
—
457
—
—
—
—
—
1,132.37 −1,132.37
−1,081.15
−8.89
—
487
—
—
—
—
—
1,012.80 −1,012.80
−2,093.96
−17.21
—
518
—
—
—
—
—
931.04
−931.04
−3,025.00
−24.86
—
Sum
12,529
12,332
1,090
729
8,964.00
11,437.51 −3,025.00
—
164.08
551.49
Material
Labor
Depreciation Equipment Main materials Sub-Con
Expense
Sum
Retainage
Planned budget
4,033.80
456.28
567.27
27.13
0
6,775.20
472.32
12,332
0%
Actual cost after 1 month
72.97
8.97
33.53
1.60
0
535.01
76.91
729
Actual weight %
10.01
1.23
4.60
0.22
0.00
73.39
10.55
100.00
Remaining budget
3,960.82
447.32
533.74
25.53
0
6,240.19
395.41
11,603
New moving weight %
34.14
3.86
4.60
0.22
0.00
53.78
3.41
100.00
aUnit=1,000 US dollars.
each time; 4 in consideration of different time lags of cost cat-
actual cost US$729 is incurred, we can update new weights to
egories, cash flow is calculated each 30 days; and 5 whenever
cost categories. The proposed model employs weights that are
negative cumulative cash flow occurs, internal interest 10% is
updated based on weight percentage of cost categories to the re-
charged.
maining budget each month over the duration, while the tradi-
Table 1 represents the example of cash forecasting at the be-
tional approach is designed based only on weights to the initial
ginning of project “0” month made in accordance with the basic
total budget. For example, new moving weight of the material
conditions and aforementioned algorithm with considerations of
cost category 34.14% can be updated by dividing the remaining
time lags and moving weights. As an example, in time period of
budget of material US$3,960.82 to the remaining total budget
30 days, cash out US$83.03 is gaged only considering cost cat-
US$11,603 . Table 2 shows an example of the updated cash bal-
egories that have no time lags
planned budget
US$ 1,072
ance negative US$3,025,000 in accordance with these revised
sum of initial weights of labor, equipment, and expenses
moving weights.
7.75% . It expects that the final cash balance at completion will
be negative—around US$3,316,000. As the first month passes and
Validation of Model
Model validation includes measuring the accuracy of a model in
Table 3. Project Overview
describing the actual conditions of a problem to solve and in
Project name
Project overview
Project A: apartment
• 8–25 story seven building apartment
Table 4. Project Basic Dataa
• 490 unit
Items
Project A
Project B
Project C
Project D
• Area: 279 ha
Contract amount
46,648
94,465
79,632
51,257
Project B: industrial complex
• Width: 20 M four lane
Duration months
33
49.3
96
63
• Earth work: 19 million m3
Total budget dollars
36,486
82,946
72,989
42,221
• Joint venture project
Labor %
1.31
8.95
2.20
4.21
Project C: railway
• Total length: 11.432 km
Material %
33.73
21.28
15.30
3.08
• Bridge: 12 4,867 m
Equipment %
0.22
1.33
5.40
0
• Tunnel: 1 545 m
Subcontract %
49.14
63.47
61.70
83.58
• Stations: three stop
Depreciation %
4.6
0.36
0
0
• Joint venture project
Expense %
11.0
4.61
15.40
9.13
Project D: sewage treatment
• Treatment capacity: 80,830 t / day
aUnit=1,000 US dollars.
JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / OCTOBER 2005 / 169

---

Table 5. Time Lags for Each Cost Categorya
Table 6. Billing Time and Delay in Payment for Each Projecta
Project A
Project B
Project C
Project D
Project A
Project B
Project C
Project D
Labor
0b
0b
0b
0b
Billing time
120
30
90
60
Material
150
120
90
120
Delay in paymentb
0
0
150
30
Equipmentc
0b
0b
0b
0b
aUnit=days.
Subcontract
150
120
90
90
bDelay payment is total delay time from the client.
Depreciation
0b
0b
—
—
Expense
0b
0b
0b
0b
aUnit=days.
bThe last day of the month when cost is incurred.
lations are performed on each project in order to compare the
cEquipment cost is charged to expense of cash out in accounting
accuracy of forecasting models. Subsequently, 48 simulations per
perspective.
each project and a total of 192 simulations were performed for
four projects. In the comparative analysis the results of forecast-
evaluating the usefulness of the model in terms of its objectives to
ing are applied to cash flow each month instead of the cumulative
a larger case with similar problem contexts. Stated earlier, a com-
cash flow forecasting applied previously in experiments since pre-
parative case study methodology was used to validate whether the
vious cash flow affects subsequent cash flow.
model meets its development objective.
To compare the accuracy of two models, MWM and FWM, the
simulation is performed in accordance with the following two
types: 1 Type 1—planned data and actual data are identical to
Validation Procedures
each other. This type is used to determine the reliability of the
To verify the model, we performed simulation using empirical
model and compare the two methods under ideal conditions since
data from actual projects in progress. Simulation results based on
planned data are one of the most critical variables in this forecast-
the proposed model and existing model are compared. A simula-
ing cash flow model, and 2 Type 2—planned data and actual
tion template is implemented in a common spreadsheet
data are different as reported by the jobsite for 12 months. In this
package—Microsoft Excel™—for the simulation experiments.
case, the uncertainty of the construction job site is involved and
Considering different time lags of cost categories, cash flow is
the effect of planned data on the forecasting cash flow is consid-
calculated in monthly increments. The fixed weights method
ered. Finally, the following data are required for comparative
FWM —the current approach—applies fixed weights to cost cat-
analysis of the four projects in progress:
egories over project duration, whereas the moving weights
1.
Time lags of billing time and individual cost categories see
method MWM —the new proposed model—applies different
Tables 5 and 6 ;
weights each month using a new algorithm. The results of these
2.
Total contract amount and total budget reflect contract
two models were then compared to show the accuracy and con-
amount and budget changed during the construction stage;
sistency of the model. The stepwise procedures for the validation
3.
Monthly cost planning data and earned value planning data;
of the proposed model are as follows:
4.
Weights of cost categories in the budget;
1.
Four actual projects in progress including one building and
5.
Retention rate and capital cost rate; and
three civil infrastructure projects, with data compiled over a
6.
Actual cash flow such as cash in and cash out for each
duration of 12 months see Tables 3 and 4 ;
month.
2.
To obtain forecast cash flow data, MWM and FWM are ap-
plied to actual data through the simulation; and
3.
To compare the accuracy of MWM to FWM, the results of
Validation Results
the simulation are analyzed.
To simulate the dynamic cash flow forecasting, simulation ex-
Measurement of Accuracy
periments were conducted 12 times from the first to the twelfth
Mean absolute deviation MAD was used to measure the error
month by each method for individual projects and compared by
for each month’s forecasted cash flow by means of the two mod-
the two methods: MWM and FWM. In addition, 2 types of simu-
els: MWM and FWM. The MAD is a commonly used measure
Table 7. Example of Comparative Analysis for Project Aa
Actual cash flow
Forecasting cash flow
Mean absolute deviation
Time period
days
Cash in
Cash out
Cash flow
MWM
FWM
MWM
FWM
183
—
736.003
−736.003
−678.018
−707.169
57.985
28.834
213
—
900.032
−900.032
−875.830
−879.505
24.202
20.527
244
—
1,160.845
−1,160.845
−1,107.974
−1,112.959
52.871
47.886
274
4,357.000
733.015
3,623.985
3,567.408
3,523.455
56.577
100.530
305
—
810.016
−810.016
−919.695
−943.710
109.678
133.694
335
—
632.069
−632.069
−662.458
−697.197
30.388
65.128
365
—
909.009
−909.009
−1,041.238
−1,083.378
132.229
174.369
Average
66.276
81.567
MWM moving weights method; and FWM fixed weights method.
aSimulation 1—after 1 month, Type 1.
170 / JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / OCTOBER 2005

---

Table 8. Comparison of Mean Absolute Deviation MAD in Moving
Table 10. Reliability of Moving Weights Method Type 1 a
Weight Method MWM and Fixed Weights Method FWM
Type 1 a
Mean absolute
Contract
A / B
100
MAD
Project
deviation A
amount B b
%
MWM A
FWM B
B − A
MAPE
A
41.207
13,489
0.31
B
62.395
27,464
0.23
Project A
41.207b
68.010
26.803
65.04%c
C
27.094
4,539
0.60
Project B
62.395
93.356
30.962
49.62%
D
27.781
7,310
0.38
Project C
24.106
24.173
0.067
0.28%
Average
0.38
Project D
27.781
31.739
3.957
14.24%
aUnit 1,000 US dollars.
Average
32.30%
bContract amount means total earned values for 12 months.
aUnit=1,000 US dollars.
bIt is calculated by Eq. 9 through the 12 times of simulations from
In the same way, the average MAD can be calculated through
the 1st month to the 12th month refer to Table 7 for the case of
the 12 times of simulations from the first month to the twelfth
simulation 1 .
c
month. Based on the results of MAD for simulations, MWM is
MAPE= 26.803÷ 41.207= 65.04%.
more accurate than FWM. In the Type 1 simulation, the accuracy
is 0.28–65.04%, with an average of 32.30% higher than FWM in
that forecasts accuracy as the degree of variation by the following
the ideal condition, where planning is well established as reflected
equation. This measure is simply the measure of the absolute
on the construction jobsite, and continuously updated
see
deviations for all forecasts
Table 8 . In the case of Type 2, the accuracy is 1.14–4.54%, an
average of 2.13% higher see Table 9 . As a result, the degree of
1
MAD =
Acf
accuracy of MWM is an average of 17.21% higher than FWM.
n
ti − Fcfti
9
where Acf
Reliability
ti = actual value of cash flow, Fcfti = forecasting value of
cash flow at t month by the ith simulations, respectively;
Kenley and Wilson 1986 and Kaka and Price 1991 suggested
t = 1 – 12; i = 1 – 12; and n = number of observations.
that the error range of forecasting in the construction industry is
In comparative analysis, the mean absolute percent error
within ±3% of the contract amount. This is considered an accept-
MAPE method was estimated for comparing errors. It can be
able limit and demonstrates the reliability of the proposed model.
achieved by dividing the difference between MAD of MWM and
The error range of the forecasting is 0.23–0.6%, with an average
FWM by MAD of MWM. This fraction represents how large the
of 0.38% for four projects in Type 1, and 0.82–2.78%, with an
error of FWM is as compared to MWM. All absolute deviation
average of 1.79% for four projects in Type 2 see Tables 10 and
data between MWM and FWM are summed and divided by the
11 . Despite unavoidable errors in the planning data, the result is
number of observations, by the following equations:
thought of as being reasonably attained by applying the model for
forecasting. Consequently, the reliability of the MWM model is
t=12 i=12
1
mMAD
acceptable and well demonstrated based on simulation.
ti − fMADii
average MAPE =
100 %
10
n
mMAD
t=1 i=1
ti
Practical Implications to Industry
where mMADti= MAD value of MWM at t month by ith simula-
tion; fMAD
Financial management has long been recognized as an important
ti = MAD value of FWM at t month by ith simulation;
and n = number of observations.
management tool. A company can survive a transitional period
The cash flow data consisted of four detailed real projects in
without a profit, or even with a loss; however, it may fail due to
progress, together with their associated estimated monthly values.
lack of cash during the operation even if it has a good financial
The data were obtained from “D” Construction and Engineering
statement. In the viewpoint of corporate, cash flow forecasts
Company in Korea. During the data collecting, adjustment of
should be made at all stages of the project from the planning stage
budget and contract amount including inflation was applied to
to the operation and maintenance stage of a project. However,
specific projects. Table 7 shows the actual data to achieve the
cash flow forecasting and management are a dynamic process.
MAD for the specific case of simulation 1, which are estimated
Deviations of all projects at the corporate level may significantly
by Eq. 9 after the first month passes from the start of the project.
affect the firm’s financial status. Moreover, inadequate cash flow
forecasts to a certain project may drive a corporate into a crisis of
financial situations.
Table 9. Comparison of Mean Absolute Deviation MAD in Moving
Weight Method MWM and Fixed Weights Method FWM
Type 2
Table 11. Reliability of Moving Weights Method Type 2 a
MAD
Mean absolute
Contract
A / B
100
Mean absolute
Project
deviation A
amount B b
%
percent error
A
153.910
12,529
1.23
MWM A
FWM B
B − A
%
B
867.361
31,158
2.78
Project A
153.910
157.198
3.287
2.14
C
37.045
4,516
0.82
Project B
867.361
873.238
5.877
0.68
D
222.364
9,500
2.34
Project C
37.960
38.395
0.430
1.14
Average
1.79
Project D
222.364
232.462
10.099
4.54
aUnit 1,000 US dollars.
Average
—
—
—
2.13
bContract amount means total earned values for 12 months.
JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / OCTOBER 2005 / 171

---

Considering the real business world in construction industry,
actual data; 2 developing cash flow forecasts at the planning
various forecasting methods may be applied to cash flow. Some
stage using the relationship between cumulative earned value and
judgment on a jobsite is needed in the case of forecasting cash
cost categories; and finally 3 implementing a system at the
flow with respect to complexity and unexpected situations in con-
corporate level to monitor cash flow in a proactive and timely
struction industry. However, if jobsite engineers or project man-
manner.
agers rely only on judgment based on their experience without the
use of any mathematical forecasting techniques, they may not
References
make a good decision to forecast cash flow. Essentially, a good
forecasting technique needs to include both a historical trend-
based data supported method and competent judgments based on
Ahuja, H. N., and Walsh, M. A.. 1983 . Successful methods in cost
engineering, Wiley, New York.
construction experience and knowledge.
Ashley, D. B., and Teicholz, P. M. 1977 . “Pre-estimate cash flow analy-
In this respect, the proposed model suggest a practical and
sis.” J. Constr. Div., Am. Soc. Civ. Eng., 103 3 , 369–379.
easy approach for jobsite engineers and project managers who are
Bennett, J., and Ormerod, R. N. 1984 . “Simulation applied to construc-
generally not familiar with finance knowledge of forecasting cash
tion projects.” Constr. Manage. Econom., 2, 225–263.
flow using the regular reports. This model can be applied as part
Fondahl, J. W., and Bacarreza, R. R. 1972 . “Construction contract
of a project evaluation process and continuously updated to show
markup related to forecasted cash flow.” Technical Rep. Prepared for
deviations between plan and actual data through changes or infor-
Construction Industry Institute, Stanford Univ., Stanford, Calif.
mation from the jobsite. In addition, the fast and simple forecast-
Gates, M., and Scarpa, A. 1979 . “Preliminary cumulative cash flow
ing cash flow allows field engineers or project managers to sup-
analysis.” Cost Eng., 21 6 , 243–249.
port and to save time for decision making of strategy of cash
Hendrickson, C., and Au, T. 1989 . Project management for construc-
management to the corporation and projects.
tion: Fundamental concepts for owner, engineer, architects, and build-
ers, Prentice Hall, Englewood Cliffs, N.J.
Jepson, W. B. 1969 . “Financial control of construction and reducing the
element of risk.” Contact. J., April, 862–864.
Conclusions
Kaka, A. P. 1996 . “Towards more flexible and accurate cash flow.”
Constr. Manage. Econom., 14, 35–44.
A simple cash flow forecasting model MWM was developed to
Kaka, A. P., and Price, A. D. F. 1991 . “Net cash flow models: Are they
assist general contractors on jobsites during the construction
reliable?” Constr. Manage. Econom., 9, 291–308.
phase. The model was based on the general procedure of con-
Kenley, R., and Wilson, O. D. 1986 . “A construction project cash flow
model—An idiographic approach.” Constr. Manage. Econom., 4,
struction jobsites and the nature of a general contractor’s budget.
213–232.
The model included new methodology that was not addressed by
Kenley, R., and Wilson, O. D. 1989 . “A construction project net cash
previous researchers. A comparative case study methodology was
flow model.” Constr. Manage. Econom., 7, 3–18.
chosen as a proper validation approach to evaluate the benefits of
McCaffer, R. 1979 . “Cash flow forecasting.” Quantity Surveying, Au-
the proposed model. Four real projects were identified as the case
gust, 22–26.
study materials and the validity of the model was tested by actual
Meredith, J. R., and Mantel, S. J., Jr. 1995 . Project management—A
data from these projects in progress. The overall validation pro-
management approach, 3rd Ed., J Wiley, New York.
cedures were derived from a series of simulations by comparing
Navon, R. 1994 . “Company-level cash-flow management.” J. Constr.
the results of the proposed model with other models suggested by
Eng. Manage., 122 1 , 22–29.
previous researchers. Ultimately, the cash flow forecasting model
Navon, R. 1995 . “Resource-based model for automatic cash-flow fore-
was demonstrated to be a simple, accurate, and reliable forecast-
casting.” Constr. Manage. Econom., 13, 501–510.
ing tool for general contractors at the construction stage in com-
Navon, R. 1997 . “Cash-flow forecasting and management.” Proc., Con-
parison with FWM.
struction Congress, ASCE, New York, 1056–1063.
During the case studies, two issues were recognized as requir-
Oberlender, G. D. 2000 . Project management for engineering and con-
struction, 2nd Ed., McGraw–Hill, New York.
ing more research. The first is that the model is dependent on the
Park, H. K. 2001 . “Cash flow forecasting model using moving weights
planning of cost and earned value. If planning of cost and earned
of cost categories for general contractors on jobsite.” PhD disserta-
value are not accurate, the forecasted cash flow would not be
tion, Univ. of Wisconsin, Madison, Wis.
accurate. The second issue is, similar to the first, how to obtain
Peer, S. 1982 . “Application of cost-flow forecasting models.” J. Constr.
reliable variables at the jobsite level such as the release of reten-
Div., Am. Soc. Civ. Eng., 108 2 , 226–232.
tion money—because it can be applied depending on the duration
Peterman, G. G. 1973 . “A way to forecast cash flow.” World Constr.,
of a subcontract.
October, 17–22.
Despite several limitations, the proposed model presents a
Reinschmidt, K. F., and Frank, W. E. 1976 . “Construction cash flow
practical and easy approach for jobsite engineers and project man-
management system.” J. Constr. Div., Am. Soc. Civ. Eng., 102 4 ,
agers who are not familiar with extensive financial knowledge,
615–627.
Russell, J. S. 1991 . “Contractor failure: Analysis.” J. Perform. Constr.
just using regular reports without separate information at the job-
Facil., 5 2 , 163–180.
site. In addition, this model can be applied as part of a project
Sears, G. A. 1981 . “CPM/COST: An integrated approach.” J. Constr.
evaluation process considering internal interest capital cost at
Div., Am. Soc. Civ. Eng., 107 2 , 227–238.
the corporate level. The model can be continuously updated to
Singh, S., and Lakanathan, G.
1992 . “Computer-based cash flow
show deviations between planned and actual data through infor-
model.” Proc., 36th Annual Trans., AM. Assoc. of Cost Engineers,
mation changes from the jobsite. Encouraged by the results of
AACE, R.5.1–R.5.14.
current research, future procedural research will concentrate on:
Trimble, E. G. 1982 . “Micro computers in construction management.”
1 analyzing the impact on difference between planned data and
Building Technology Management, 2 2 , 11–13.
172 / JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / OCTOBER 2005

---