Engineering Computations

Risk analysis model for regional railroad investment

Sunduck Suh, Wonho Suh, Jung In Kim,
Article information:
To cite this document:
Sunduck Suh, Wonho Suh, Jung In Kim, (2017) "Risk analysis model for regional railroad
investment", Engineering Computations, Vol. 34 Issue: 1, pp.164-173, doi: 10.1108/EC-10-2015-0323
Permanent link to this document:
http://dx.doi.org/10.1108/EC-10-2015-0323
Downloaded by Universitas Negeri Semarang At 23:40 18 April 2017 (PT)

Downloaded on: 18 April 2017, At: 23:40 (PT)

References: this document contains references to 5 other documents.
To copy this document: permissions@emeraldinsight.com
The fulltext of this document has been downloaded 54 times since 2017*
Users who downloaded this article also downloaded:
(2017),"Generalized finite element analysis of high-rise wall-frame structural systems", Engineering
Computations, Vol. 34 Iss 1 pp. 189-210 http://dx.doi.org/10.1108/EC-07-2016-0266
(2017),"Study of laser displacement measurement data abnormal correction algorithm", Engineering
Computations, Vol. 34 Iss 1 pp. 123-133 http://dx.doi.org/10.1108/EC-10-2015-0325

Access to this document was granted through an Emerald subscription provided by emerald-
srm:603747 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as
well as providing an extensive range of online products and additional customer resources and
services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the
Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for
digital archive preservation.

*Related content and download information correct at time of download.


EC
34,1
164
Received 25 October 2015
Revised 26 November 2015
Accepted 10 January 2016
Downloaded by Universitas Negeri Semarang At 23:40 18 April 2017 (PT)
Engineering Computations
Vol. 34 No. 1, 2017
pp. 164-173
© Emerald Publishing Limited
0264-4401
DOI 10.1108/EC-10-2015-0323
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/0264-4401.htm
Risk analysis model for regional
railroad investment
Sunduck Suh and Wonho Suh
Department of Transportation and Logistics Engineering,
Hanyang University, Ansan, Republic of Korea, and
Jung In Kim
Department of Civil and Environmental Engineering, Stanford University,
Stanford, California, USA
Abstract
Purpose – The purpose of this study is to model risks in financial analysis. These risks associated with
uncertainties in the projects should be properly addressed to ensure proper decision regarding the projects.
Performance indicators should be developed and assessing risks has high priority. All these activities
comprise appraisal, and based on these, a proper course of action should be recommended.
Design/methodology/approach – To analyze the attractiveness of a project for foreign regional railroad
investment or participation, the project should be analyzed in a systematic way. First, the project’s goals and
objectives should be evaluated for compatibility. Also, criteria of acceptability for stakeholders should be
checked against output from the project. Usually, a project can have many alternatives, and impacts of each
alternative should be analyzed in terms of quantitative and qualitative forecasts of impacts. Benefits and costs
need to be counted in proper units of measurement per goals and objectives.
Findings – This paper shows that risk modeling can reflect uncertainty in decision-making and provide
robustness of modeling process and improved communication. Also, challenges are presented in using risk
analysis.
Originality/value – To overcome the shortcomings of traditional mathematical optimization model in
identifying best sets of projects for private application, the proposed model finds ways to incorporate risk
management components for the optimization procedure. Based on simulation results, a brute force solution
procedure using enumeration can be used. Another approach is recommended to use the genetic algorithm
process to reduce the number of alternatives to search to reach an optimal solution.
Keywords Risk analysis, Financial analysis, Regional railroad investment, Risk model
Paper type Research paper
Introduction
To analyze the attractiveness of a project for foreign investment or participation, the project
should be analyzed in systematic way. First, the project’s goals and objectives should be
evaluated for compatibility. Also, criteria of acceptability for stakeholders should be checked
against output from the project. Usually, a project can have many alternatives, and impacts
of each alternative should be analyzed in terms of quantitative and qualitative forecasts of
impacts. Benefits and costs need to be counted in proper units of measurement per goals and
objectives. Performance indicators should be developed for evaluation. Additionally, it is
important to assess risks (uncertainties) in project evaluation and analysis. All these
activities comprise appraisal, and based on these, a proper course of action should be
recommended.
Preparation of this article was sponsored by a grant NRF-2014R1A1A2054793 and Transportation &
Logistics Research Program ID-97344 of Korean Government.
 Project development process follows certain well-known paths; pre-investment phase, Risk analysis
implementation phase, operation phase and decommissioning. Pre-investment studies model
usually cost from 0.5 to 3 per cent of the project cost; in a large project, it can be somewhere
between 0.5 to 1 per cent. It has opportunity study, pre-feasibility study, feasibility study
which is more of optimization study and various support studies, such as market/marketing,
raw material, Environmental Impact Assessment (EIA), location, scale of economy,
technology selection and equipment selection, etc. Project team comprises project economics,
markets, process engineer, construction engineer, organizational, finance and accounting 165
experts, EIA experts and social and political experts. Accuracy of opportunity study is
considered within 25-30 per cent and approximately within 10 per cent for in-depth study.

Financial analysis
Financial analysis is a basis for “profitable” project definition, and a very important step to
Downloaded by Universitas Negeri Semarang At 23:40 18 April 2017 (PT)

evaluate project viability for foreign investment and participation. It is based on a pro-forma
financial model which uses expected figures. It can produce financial indicators, like return
on equity and value of company and share, and can be used to test different financing
arrangement.
The starting point of a financial modeling is to have a reasonable split on equity vs debt.
Equity portion can be reduced down 10 per cent, but this case is very rare. Lenders usually
like to see debt: equity less than 2. In this sense, 30:70 might be the most common in practice.
It needs to start with equity sources, as debt is usually easier than equity. One also likes to
seek for a multilateral agency or government debt, including export credit agency. It is
necessary to see if project financing is possible. For private financial evaluation, use of
current price is recommended, but it involves issue of forecasting inflation systematically.
A financial model has many components which are arranged in usually different sheets.
In the general input part, economic or basic information is listed. Investment cost
components should have infrastructure (including track), routing/scheduling (R/S), system
and station/depot. Operations and maintenance (O&M) cost should be linked with operation
plan. Usually financing sources and financing cost need a separate sheet. Farebox revenue is
dependent on demand forecast, which, in turn, has a relationship with fare and operation
requirement including R/S requirement and operation pattern. Station development revenue
and extra-ordinary revenue can be significant depending on situation. Output should include
financial indices and summary.
Financial model should best be commenced in conceptual stage, but to do that, it has some
major issues. Availability of capital for fixed investment, preoperational expenditures and
working capital and maintaining liquidity for planning horizon are major concerns in the
planning process. Also, whether the proposed capital structure and debt service plan is
adequate should be properly evaluated. Questions like “Are investors’ criteria satisfied”? “Is
financial obligation satisfied”? and “Are financial indicators acceptable with reasonable
risk”? should be answered from the financial modeling effort.
A good financial model should be easy to understand, have a clear logical flow, should use
easy-to-understand Excel functions, be objectives-driven, be decision supporting, should
have a proper level of detail, be error-free in terms that it should not have unintended
consequences/unacceptable implicit assumption and calibration should be easy.

Risk modeling
Financial modeling uses pro-forma financial sheets which uses expected values. Therefore,
the decision based on the financial model is also subject to uncertainties included in expected
values. These uncertainties should be addressed properly to ensure “correct” decision
regarding the projects. Traditionally, scenario planning has been used, but one cannot put
 EC the probability of a specific scenario happening. Risk modeling can reflect uncertainty in
34,1 decision-making with objective means. It can provide robustness of modeling process and
improved communication. It can handle combinations and probabilities of factors and can
evaluate contingent claims and non-linear behaviors. Limitations of traditional Excel
structure, however, present challenges in using risk analysis. The risk modeling process
involves, defining risk and risk analysis, building a risk model, risk identification and
166 prioritization, risk analysis and modeling and risk monitoring and controlling.
Risk analysis specifically considers different level of risks involved in the analysis.
Uncertainty (risk) needs to be addressed systematically. Efficient risk management involves
steps to quantify size of risk and proper allocation based on the premise “give risks to the one
who can control the risks BEST”.

General input
Downloaded by Universitas Negeri Semarang At 23:40 18 April 2017 (PT)

General input section depicts current situation regarding tax rate, energy cost structure and
gross national income among others. It also has a scenario on inflation; a scenario on growth
rate, discount rate, system configuration, project timing including staging; and a what–if
analysis frame.

Cost
Investment cost should address different items of cost components. Infrastructure cost
estimation can be facilitated if a preliminary design is possible or at least in a kilometer of
earthwork, bridge, tunnel or viaduct is available. Critical information is unit price.
Cost for R/S can be estimated once the number of train-sets is determined based on
operation requirement and reserve ratio and unit price of one train-set. Of course, the
assumption of procurement (local/foreign) and any technology transfer requirement affects
overall cost.
Cost of systems can be estimated based on per kilometer unit cost. If the figures are
available, it is recommended to estimate by component. Cost of station and depot depends on
the location, size and unit cost.
O&M cost can be estimated in many different methodologies, from most simplified
estimate on unit cost per train-kilometer to a more sophisticated spreadsheet model. It should
be responsive to operation pattern and other major variables, like labor vs machine in
ticketing, and energy cost. One needs to develop a broad picture of operation activities first,
like how to sell ticket, etc.
Financing cost depends on funding source used, amount and terms and conditions. One of
the challenges is that this financing cost is based on monthly activity. One needs to check
whether threshold value for Debt Service Coverage Ratio (DSCR) is higher than 1.2, for
example. Draw-down order is important to keep track of amount and sources.

General input
Farebox revenue is the most important factor in evaluating the viability of railroad projects.
The following financing cost should be calculated monthly; farebox revenue should be
somehow calculated monthly. This requirement will put a heavy burden on revenue
forecasts. One needs to pay attention to the evaluation criteria of the bidding, but fare level
might be the most important criteria. Elasticity of demand, study for range of fare level, time
of day (TOD) and seasonal variation with different level of fare should be clearly studied.
Revenue should be responsive to level of service (LOS) provided and linked to operation
requirement which has cost implication. If needed, profit maximizing fare (or operation level)
should be studied.
 Station development revenue can be significant depending on location and size. The Risk analysis
revenue from station development has the issue of timing of realization, and the risk size is model
usually bigger than farebox revenue. Extra-ordinary revenue can be from parking, office and
small shops, but it should not be double-counted with station development. One needs to
consider local potentials. When no detailed breakdown is possible, a certain percentage of the
total revenue can be also used.
Output should provide summary by scenario and major financial indicators and value of 167
concession company.

Practical aspects of risk modeling – simulation with @Risk

There are some practical factors to consider when implementing risk modeling in practice.
Uncertainties involved in estimated figures should be explicitly considered with a procedure
Downloaded by Universitas Negeri Semarang At 23:40 18 April 2017 (PT)

such as Monte Carlo simulation. Other consideration is dependency among variables. How
one can handle dependency in variables in simulation and how they affect output is a fairly
complicated issue.
Probability distributions related to value of variables can come in many different formats;
uniform, triangular, Program Evaluation and Review Technique (PERT), binomial, normal,
lognormal, Poisson and discrete. Or one can use a probability distribution fitted from
observed data.
Simulation with different probability distribution and explicit consideration of
dependency can be easily modeled using commercial programs, for example @Risk from the
Palisade Corporation (2015). Because the financial sheets are usually developed using Excel
program, the @Risk program, being an add-in for Excel, can be easily integrated with Excel
spreadsheets. Figures 1 and 2 show sample input and output from the @Risk program.

Figure 1.
Sample input from
@Risk
 EC
34,1

168
Downloaded by Universitas Negeri Semarang At 23:40 18 April 2017 (PT)

Figure 2.
Sample output from
@Risk

Mathematical models for analyzing investment priorities

When there is no budget limitation, one can choose all the projects that are profitable in the
portfolio. Adding non-profitable projects to the portfolio will lower the total profit, so one
needs to add only profitable projects to the portfolio. In reality, however, this kind of
unlimited budget situation does not happen. Usually, there is a budget constraint. In a
corporate environment, selecting the most profitable project out of many alternatives within
a budget constraint is called a “capital rationing” problem.
This model is designed to identify subset of projects out of the many profitable projects
according to priority. Two different approaches can be, at least in theoretical sense, used
depending on the assumptions used. Combinations of alternative are mutually exclusive in
both cases:
(1) all projects are independent; and
(2) projects are NOT independent, but interrelated.

The first case can be readily addressed with a class of models, such as a budget allocation
problem, capacity planning problem, job shop scheduling (an optimization problem in which
ideal jobs are assigned to resources at particular times) and portfolio allocation of risk
minimization as it is or with some form of modification. Traditionally, the most familiar form
of model to address this situation is a “knapsack problem”.
The second case is somewhat more complicated. The situation can arise when we need to
evaluate a railroad link in terms of national railroad network. All rail link candidates are not
independent in this case. Theoretically, one must evaluate all possible combinations of each
alternative being chosen or not. For example, when there are ten link alternatives, the
number of analysis should be done will be 1,023 (210 ⫺ 1). One needs to evaluate all these
combinations and should select combinations of links which give best achievement of goals
 and objectives. This requires usually heavy computation resources, especially when there Risk analysis
are large numbers of alternatives. Instead of full enumeration of all possible combination of model
alternatives, more efficient procedure is available under the scheme of a network design
problem as indicated by Magnanti and Wong (1984).
When there are uncertainties involved as we have seen in the previous chapter or
goals and objectives are multi-attribute, these mathematical optimization models are not
really efficient under the situation. In this study, pure mathematical models to address
the projects’ selection in terms of priority are discussed first; limitations of pure single 169
objective mathematical model are discussed; and a different model, simulation
optimization model, is proposed to address uncertainty involved in all numbers used in
the analysis.

Selecting independent projects with budget constraint

Downloaded by Universitas Negeri Semarang At 23:40 18 April 2017 (PT)

This will be a typical capital budgeting problem. Each individual project is not mutually
exclusive, but a combination of projects will be. The combination of projects should be
within budget. To explain this problem, four sample projects are considered as listed in
Table I.
This problem can be solved by testing all 15 combinations of projects, i.e. 1, 2, 3, 4, 1-2, 1-3,
1-4, 2-3, 2-4, 3-4, 1-2-3, 1-2-4, 1-3-4, 2-3-4 and 1-2-3-4, and identifying the combination which
gives the highest net present value (NPV) value. When there is a constraint in resources,
projects with higher requirements are not possible. This problem can be re-casted as an
optimization problem as shown in Figure 3.
When the number of projects is large, one can use the optimization software, such as
LINGO (LINDO Systems, 2015) or other standard 0-1 integer linear programming program.
Model and solution output from the LINGO program is shown in Figure 4. The model can
also be solved with Solver in an Excel spreadsheet program (Figure 5).
As was seen in the previous section, the NPV of project combination of independent
projects can be summed over all the projects in the combination. If the projects are dependent
on each other, this assumption does not apply in this process. A major difference with the
independent project is that the NPV of combinations of dependent projects should be
separately analyzed for each combination. For example, the NPV of project combination of 1
and 2 can be different from the sum of NPV of Projects 1 and 2 done individually compared
against a do-nothing situation. Project impacts should be evaluated for each combination of
projects.
This requires higher computation requirements in practice. For example, when there
are ten project alternatives, project impacts should be calculated for 1,023 combinations.
In the independent case, ten impacts were calculated, and the impact combinations were
assembled based on ten impacts. When the number of project alternatives is high, this
introduces a heavy computation burden. For example, if the potential project

End of year Project 1 Project 2 Project 3 Project 4

0 ⫺10,000 ⫺15,000 ⫺18,000 ⫺25,000

1 2,800 3,600 4,300 5,400 Table I.
2 2,800 3,600 4,300 5,400 Sample project
– – – – – example (average of
– – – – – five simulations with
10 2,800 3,600 4,300 5,400 different random
NPV (12%) 5,821 5,341 6,296 5,511 seeds)
 EC
34,1

170
Downloaded by Universitas Negeri Semarang At 23:40 18 April 2017 (PT)

Figure 3.
Optimization problem
example

Figure 4.
LINGO model and
output

alternatives are 20, the number of cases to be evaluated will be more than one million.
Full enumeration of more than a million cases is difficult to carry out in practice.
These problems have been addressed as a network design problem. There are many
algorithms to tackle this kind of problem, but basically, this is an NP-complete problem. The
larger the number of alternatives, the computation requirement increases non-polynomially.
Two different classes of problems are being used:
 Risk analysis
model

171
Downloaded by Universitas Negeri Semarang At 23:40 18 April 2017 (PT)

Figure 5.
Excel solver output

(1) Continuous Network Design Problem (NDP): When investment variables can have
continuous values.
(2) Discrete NDP: When investment variables can have either 0 or 1.

The first continuous NDP problem can be simplified as a piecewise linear programming
problem or can be formulated as a non-linear problem. The second discrete NDP problem is
usually formulated as a 0-1 integer problem, linear or non-linear. In the national railroad
network context, the problem can be cast as shown in Figure 6.
Different approaches are being used to solve for the NDP problems depending on
specific formulation. Besides optimizing solution algorithms, search algorithms, such as
genetic algorithm, are also being used widely recently. For each combination of project
alternative, the impacts should be analyzed. This is usually done in the framework of
network analysis. The network analysis can be simplified mathematically, for example,
 EC as a multi-commodity flow problem, but basically, it involves solution of network
34,1 assignment to understand impacts on vehicle-kilometer or travel time. In practice, the
impacts are being usually analyzed with the help of commercial network analysis
programs, such as Cube (Citilabs, 2015) or TransCAD (2015). Figure 7 shows a
framework which is tested in this study to analyze the effects of combinations of
dependent projects. In this study, TransCAD was used as a procedure to facilitate the
172 case study by using data prepared in the TransCAD format.

Conclusion and discussion

The mathematical optimization models introduced in this paper can handle a large number
of projects and evaluate a subset of projects under budget constraints. Although these
models are popular in practice, they have limitation when applied to private procurement
project selection.
Downloaded by Universitas Negeri Semarang At 23:40 18 April 2017 (PT)

The first issue with these models is that they usually use a single objective, but as was
shown in this paper, project evaluation involves many variables with a different level of
uncertainty. Under this condition, traditional models have limitation to address the potential
risk. Therefore, one needs to use methods that can explicitly consider variables with a
different level of uncertainty. As seen in the risk management modeling section, variables
with a different level of uncertainty can be modeled with simulation procedure, such as
Monte Carlo simulation. To account for the risk component in an optimization model, various
approaches can be devised. For example, solutions from an optimization model can be used

Figure 6.
NDP formulation

Figure 7.
NDP solution
framework proposed
 as an initial solution and uncertainty can be applied in the second stage. Although this Risk analysis
approach cannot guarantee “optimal” solutions, the proposed method can serve the purpose model
in heuristic way.
To overcome the shortcomings of traditional mathematical optimization model in
identifying best sets of projects for private application, we need to look for ways to
incorporate risk management components to optimization procedure. Based on simulation
results, a brute force solution procedure using enumeration can be used. Another approach is
to use a genetic algorithm process to reduce the number of alternatives to search to reach an 173
optimal solution.
Especially for foreign regional railroad investment uncertainties always exist and they
influence the process of the investment. The proposed model provides a systematic method
to analyze the attractiveness of a project focusing on foreign regional railroad investment.
Although the proposed approach cannot eliminate the uncertainty, the model offers insights
Downloaded by Universitas Negeri Semarang At 23:40 18 April 2017 (PT)

in to how to address the risk in regional railroad investment.

References
Citilabs (2015), “Cube software”, available at: www.citilabs.com (accessed 25 April 2015).
LINDO Systems (2015), “Lingo software”, available at: www.lindo.com (accessed 25 April 2015).
Magnanti, T. and Wong, R. (1984), “Network design and transportation planning: models and
algorithms”, Transportation Science, Vol. 18 No. 1.
Palisade Corporation (2015), “@RISK”, available at: www.palisade.com/ (accessed 25 April 2015).
TransCAD (2015), “Transportation planning software”, available at: www.caliper.com (accessed 25
April 2015).

Corresponding author
Wonho Suh can be contacted at: wonhosuh@hanyang.ac.kr

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: permissions@emeraldinsight.com