You are on page 1of 13

ASERS

J
ournal of Advanced Studies
in Finance

Biannually
Volume VII
Issue 1(13)
Summer 2016

ISSN 2068 – 8393


Journal DOI
http://dx.doi.org/10.14505/jasf
Journal of Advanced Studies in Finance

is an advanced e-publisher struggling to bring further worldwide


learning, knowledge and research. This transformative mission is
realized through our commitment to innovation and enterprise, placing
us at the cutting-edge of electronic delivery in a world that increasingly considers the dominance of
digital content and networked access not only to books and journals but to a whole range of other
pedagogic services.
In both books and journals, ASERS Publishing is a hallmark of the finest scholarly publishing
and cutting-edge research, maintained by our commitment to rigorous peer-review process.
Using pioneer developing technologies, ASERS Publishing keeps pace with the rapid changes
in the e-publishing market.
ASERS Publishing is committed to providing customers with the information they want, when
they want and how they want it. To serve this purpose, ASERS publishing offers digital Higher
Education materials from its journals, courses and scientific books, in a proven way in order to engage
the academic society from the entire world.

1
Summer 2016
Volume VII
Issue 1(13)

Editor in Chief
Laura GAVRILĂ (formerly) ŞTEFĂNESCU
Spiru Haret University, Romania Contents:
Co-Editor
Rajmund MIRDALA
Technical University of Kosice, Slovak
1 The Role of Asymmetries and Banking Sector
Indicators in the Interest Rate Pass-Through in Malta
Republic
Brian MICALLEF, Noel RAPA,Tiziana Marie GAUCI …5
Editorial Advisory Board
Mădălina Constantinescu
Spiru Haret University, Romania External Financing Patterns for Small and Medium
Rosaria Rita Canale
2 Firms in Eastern Europe and Central Asia: Does
University of Naples Parthenope, Italy Financial and Institutional Development Matter?
Francesco P. Esposito Octavian PORUMBOIU …14
Allied Irish Bank, Group Market Risk
Management
Lean Hooi Hooi How to Stabilize the Currency Exchange Rate
Universiti Sains Malaysia, Malaysia 3
Sergey BLINOV …38
Terence Hung
United International College, Hong Kong
Renata Karkowska Assessment of Economic Indicators for Evaluation of
Faculty of Management, University of 4 Financial Performance
Warsaw, Poland
Mihaela COCOŞILĂ …70
Kosta Josifidis
University of Novi Sad, Serbia
A Note on Credit Spread Forwards
Ivan Kitov
Russian Academy of Sciences, Russia 5 Markus HERTRICH ...77
Piotr Misztal
The Jan Kochanowski University in Kielce, Applications of Simulation - based Methods in
Faculty of Management and Finance: The Use of ModelRisk Software
6
Administration, Poland
Hamed HABIBI, Reza HABIBI ...82
Andreea Pascucci
University of Bologna, Italy
Rachel Price-Kreitz
Ecole de Management de Strasbourg,
France
ASERS Publishing
Daniel Stavarek http://www.asers.eu/asers-publishing
Silesian University, Czech Republic ISSN 2068-8393
Laura Ungureanu Journal's Issue DOI:
Spiru Haret University, Romania http://dx.doi.org/10.14505/jasf.v7.1(13).0
Wing-Keung Wong
Department of Economics, Institute for
Computational Mathematics, Hong Kong
Baptist University

2
Call for Papers
Volume VII, Issue 2(14), Winter 2016

Journal of Advanced Studies in Finance

The Journal aims to publish empirical or theoretical articles which make significant contributions
in all areas of finance, such as: asset pricing, corporate finance, banking and market microstructure, but
also newly developing fields such as law and finance, behavioral finance and experimental finance.

The Journal will serve as a focal point of communication and debates for its contributors for the
better dissemination of information and knowledge on a global scale.

The primary aim of the Journal has been and remains the provision of a forum for the
dissemination of a variety of international issues, empirical research and other matters of interest to
researchers and practitioners in a diversity of subjects linked to the broad theme of finance.

The Editor in Chief would like to invite submissions for the 7th Volume, Issue 2(14), Winter 2016
of the Journal of Advanced Studies in Finance (JASF).

Journal of Advanced Studies in Finance is indexed in EconLit, RePEC, CABELL's Directories,


EBSCO, ProQuest, CEEOL databases.
All papers will first be considered by the Editors for general relevance, originality and
significance. If accepted for review, papers will then be subject to double blind peer review.

Deadline for Submission: 15th October , 2016


Expected Publication Date: Dcember, 2016
Web: www.asers.eu/journals/jasf/
E-mails: jasf@asers.eu asers2010@yahoo.co.uk

3
Journal of Advanced Studies in Finance
Volume VII Issue1(13) Summer 2016

DOI: http://dx.doi.org/10.14505/jasf.v7.1(13).06

Applications of Simulation-Based Methods in Finance:


The Use of ModelRisk Software
Hamed HABIBI
Faculty of Science and Engineering, Curtin University, Perth, Australia
hamed.habibi@postgrad.curtin.edu.au
Reza HABIBI
Iran Banking Institute, Central Bank of Iran, Tehran, Iran
r_habibi@ibi.ac.ir

Suggested Citation:
Habibi, H., Habibi, R. (2016). Applications of Simulation-Based Methods in Finance: The Use of ModelRisk Software,
Journal of Advanced Studies in Finance, (Volume VII, Summer), 1(13): 82-89. DOI:10.14505/jasf.v7.1(13).06. Available
from: http://www.asers.eu/journals/jasf/curent-issue.
Article’s History:
Received May, 2016; Revised June, 2016; Accepted July, 2016.
2016. ASERS Publishing. All rights reserved.
Abstract:
This paper has two parts. The first part considers the statistical arbitrage detection using the simulation-based
approaches. The statistical arbitrage is the opportunity of attaining gain at future with a high probability with zero investment
at the current time. Simulated methods relies on the direct use of three famous theorems in the field of stochastic process,
namely (i) the arc-sine laws, (ii) the first passage of time, and (iii) optional sampling theorem. It is very important for investors
to use the simulated approaches by user-friendly software like the ModelRisk of Excel which is done in this note. In the
second part, the conditional NPVaR (CNPVaR) of a cash flow stream in the presence of exchange rate risk. The distribution
of risk factors are assumed to be a specified location-scale distribution. The application of dynamic programming and
quadratic programming are studied.
Keywords: arc-sine laws; CNPVaR; dynamic and quadratic programming; exchange rate risk; first passage of time; optional
sampling theorem; statistical arbitrage.
JEL Classification: C15, C53, C61.
1. Introduction
This paper contains two parts. The first part studies the statistical arbitrage based on simulated
algorithms. The arbitrage opportunity is, by constructing a portfolio, attaining the risk free profit at future with zero
investing at the current time. The statistical arbitrage is achieving the profit at future with a high probability
(Bondarenko 2003). The current note considers the statistical arbitrage detection using the simulation-based
approaches. Simulated methods relies on the direct use of three famous theorems in the field of stochastic
process, namely (i) the arc-sine laws, (ii) the first passage of time, and (iii) optional sampling theorem. These
theorems may have many financial applications. For example, if there are strong criteria such that the logarithm
of a price of stock follows a Brownian motion, then 𝜏𝑎 (the first passage) is the first time that the logarithm of
price hits the level 𝑎. Therefore, it is very important (for investors, for example) to simulate them by user-friendly
software like Excel. The ModelRisk (Vose company) is an adds-in software of Excel designed to perform
simulation, sensitivity analysis, optimization and risk analysis with applications in finance models, see Vose
(2010). This note considers the simulation-based approach to detect statistical arbitrage opportunities. The
second part studies the conditional NPVaR (CNPVaR) of a cash flow stream in the presence of exchange rate
risk.

82
Journal of Advanced Studies in Finance
Volume VII Issue1(13) Summer 2016
Part 1 The statistical arbitrage
The rest of this part is organized as follows. In the next section, the simulation-based approaches and
some simulation results are presented. A real data analysis is considered in section 3. To these ends, consider a
portfolio containing a long position in a stock 𝑆𝑡 and 𝛾 portion of a riskless bond 𝐵𝑡 , with risk free rate growth 𝑟
in short position. The value of portfolio is 𝑣𝑡 = 𝑆𝑡 − 𝛾𝐵𝑡 .
2. Simulated methods
This section is organized for presenting simulation-based approaches.
2.1. Arc-Sin law
The first approach uses the arc-sin law which is given by (Calin 2012). The probability that a Brownian
2 𝑡
motion 𝐵𝑡 does not have any zeros in the interval (t1 , t 2 ) is equal to 𝑃(𝐵𝑡 ≠ 0, 𝑡 ∈ (𝑡1 , 𝑡2 )) = 𝜋 arcsin √𝑡1 .
2
To use the arc-sin law, it is assumed that 𝛾 is a time varying parameter, that is, 𝛾 = 𝛾𝑡 . Let 𝑣0 = 0, then, 𝛾0 =
𝑆0
, let the probability of having possible arbitrage is pt = P(vt > 0). It is assumed that 𝑆𝑡 follows a Black-
𝐵0
σ2
(μ− )t+σwt
Scholes formulae, thus St = S0 e 2 , where 𝑤𝑡 is the Wiener process. Thus, pt = P(St > γt Bt ) =
σ2
(μ− )t+σwt γ
P (e 2 > γ t ert ).
0
σ2 wt 1 γ −1 γ
Let μ = r + 2
. Then, pt = P ( > σ t log (γ t )). Define rt = σ t log (γ t ) or equivalently, γt =
√t √ 0 √ 0
−σrt √t
γ0 e . Then, 𝑝𝑡 = 𝑁(𝑟𝑡 ), where 𝑁 is the CDF of standard normal distribution. Suppose that as t → ∞,
then rt → ∞ and therefore, 𝑝𝑡 → 1. Next, suppose that t → ∞, then rt → 0 and √trt → 0 therefore, pt →
1
0.5. However, the Arc-Sin law presents a better result. Notice that for t ∈ (t1 , t 2 ), then pt = (2) P(|wt | >
1 1 t
|rt |) ≥ ( ) P(|wt | > |rt |, t ∈ (t1 , t 2 )) ≈ arcsin √ 1 . These results are correct if 𝑤𝑡 follows a Wiener
2 π t 2

process. As follows, the behavior of pt is studied for rt = t, et , e−t in Table 1.


Table 1 - The behaviour of pt

r rt = t rt = et rt = e−t p1 p2 p3

0.001 0.001 1.001001 0.999 0.500399 0.841587 0.841103

0.002 0.002 1.002002 0.998002 0.500798 0.841829 0.840861

0.003 0.003 1.003005 0.997004 0.501197 0.842071 0.840619

0.028 0.028 1.028396 0.972388 0.511169 0.848118 0.834571

0.029 0.029 1.029425 0.971416 0.511568 0.84836 0.83433

0.03 0.03 1.030455 0.970446 0.511966 0.848602 0.834088

0.031 0.031 1.031486 0.969476 0.512365 0.848843 0.833846

0.032 0.032 1.032518 0.968507 0.512764 0.849085 0.833604

0.06 0.06 1.061837 0.941765 0.523922 0.855845 0.826843

0.061 0.061 1.062899 0.940823 0.52432 0.856086 0.826602

0.062 0.062 1.063962 0.939883 0.524719 0.856327 0.826361

0.063 0.063 1.065027 0.938943 0.525117 0.856568 0.82612

0.09 0.09 1.094174 0.913931 0.535856 0.863061 0.819623

83
Journal of Advanced Studies in Finance
Volume VII Issue1(13) Summer 2016

It is seen that as t → ∞, then pt → 1, for rt = t, et and pt → 0.5 for rt = e−t .


2.2 The first passage of time
Here, another approach is proposed using the first passage time theorem. Let 𝜏𝑎 be the first time the
2
|a| −a −3
Brownian motion Bt hit a. Then, τa has a Pearson (5) distribution given by density fτa (x) = e 2x x2 ,x>
√2π
0. Next, we suppose that γt = γ0 is fixed. The aim is to find the stopping time τ such that P(vτ > 0) = 1.
w σ2
Hence, the stopping time τ is the first time that t t hits the level a = r + 2 − μ. Again, this first passage is
simulated using ModelRisk software. While μ = r = 0.05, σ = 0.25, the level is 0.03125. Here, using a Monte
w
Carlo simulation (with 1000 repetition), 1000 paths of t t t = 0.001(0.001)1, are generated and the empirical
estimate of τa is obtained 0.05 with standard deviation 0.196. The histogram estimation of its density is given by
Figure 1. A Pearon (5) distribution is fitted with a = 0.03125. The SIC, AIC, HQIC and LR goodness of fit are
8182.5, 8192.38, 8188.66 and 1, respectively. The following Table 2, gives the mean and standard deviation of
τa for various values of parameters.

Figure 1. Histogram plot of τa


The surface of mean and standard deviation of τa are given in Figures 1 and 2, respectively. It is seen as
μ gets large and σ is small, the results are more accurate.

Table 2 - The mean and standard deviation of τa

E(τa )

μ/σ 0.100 0.200 0.250 0.300 0.350 0.400 0.500

0.01 0.033 0.036 0.03 0.043 0.034 0.023 0.060

0.03 0.049 0.055 0.03 0.01 0.06 0.061 0.057

0.05 0.014 0.041 0.05 0.05 0.023 0.017 0.024

0.07 0.0247 0.03 0.031 0.06 0.035 0.0158 0.0253

0.09 0.021 0.0286 0.0246 0.027 0.043 0.015 0.039

0.1 0.0433 0.048 0.0286 0.041 0.049 0.022 0.043

0.12 0.039 0.021 0.027 0.087 0.053 0.04 0.026

84
Journal of Advanced Studies in Finance
Volume VII Issue1(13) Summer 2016

stdev(τa )

μ/σ 0.1 0.2 0.25 0.3 0.35 0.4 0.5

0.01 0.116 0.152 0.142 0.161 0.151 0.161 0.203

0.03 0.176 0.192 0.213 0.225 0.206 0.22 0.193

0.05 0.099 0.178 0.196 0.183 0.104 0.123 0.091

0.07 0.093 0.146 0.124 0.203 0.154 0.047 0.118

0.09 0.077 0.145 0.109 0.143 0.16 0.039 0.172

0.1 0.16 0.178 0.124 0.174 0.187 0.111 0.176

0.12 0.16 0.1 0.11 0.25 0.19 0.168 0.128

Figure 2 - Mean Surface of τa

Figure 3 - Stdev surface of Tau


2.3. The optional sampling theorem
The theorem is as follows. Let Mt , t ≥ 0, be a right continuous Ғt -martingale (Ғt is a filtration) and τ be a
stopping time with respect to Ғt . If either one of conditions of the following conditions holds: 1. τ is bounded, i.e.
∃ N < ∞ such that τ ≤ N; 2. ∃ c > 0, such that E[|Mt |] ≤ c, ∀t > 0, then E(Mτ ) = E(M0 ). In this case, it
w
is assumed that Mt∗ = t t is not a martingale with respect to Ғt . Therefore, an important question is if E(Mτ∗ ) =

85
Journal of Advanced Studies in Finance
Volume VII Issue1(13) Summer 2016
E(M1∗ ) = 0. Again, this question is answered through simulations. Assuming μ = r = 0.05, σ = 0.25, the
histogram of Mτ∗ is plotted as follows in Figure 4. The following Table 3 gives the Monte Carlo estimate of E(Mτ∗ )
for various selections of μ and σ. The surface of mean and standard deviation of Mτ∗ are given in Figures 4 and
5, respectively. For almost all cells, the standard deviation is high and the mean of Mτ∗ has sharp changes in sign
and values, indicating there is no robust inference about the Mτ∗ , because it is not a martingale. In the next
section, a real data set is analyzed and statistical arbitrage is detected.
3. Real data set
In this section, existence the statistical arbitrage in stock of Intel corporation, a multinational technology
company, is surveyed. The daily stock price are collected for period of study 20th February 2015 to 18th February
2016, including 250 log-returns. They are taken from

Figure 4 - Histogram plot of Mτa


Table 3 - The mean and standard deviation of Mτ∗

E(Mτ∗ )
𝜇/𝜎 0.1 0.2 0.25 0.3 0.35 0.4 0.5
0.01 0.254 -0.945 -0.0255 -0.696 -0.252 -6.18 -2.18
0.03 -0.42 6.65 0.388 4.27 -0.164 -1.17 -1.43
0.05 -2.86 1.52 0.89 3.125 -0.84 7.54 -2.76
0.07 1.17 -1.59 -1.47 0.75 4.27 -0.82 4
0.09 0.046 -1.75 1.75 -1.66 -0.5 0.185 -3.49
0.10 0.695 1.024 0.438 -1.47 1.52 0.95 1.89
0.12 -0.37 2.56 1.4 4 4.88 2.09 3.89
stdev(Mτ∗ )
𝜇/𝜎 0.1 0.2 0.25 0.3 0.35 0.4 0.5
0.01 29.42 27.26 24.94 26.27 22.9 24.06 23.85
0.03 26.41 24.85 25.9 23.61 25.64 24.52 27.17
0.05 26.66 23.21 24.72 22.73 26.24 25.3 25.35
0.07 23.3 24.51 25.11 24.64 25.04 27.13 23.91
0.09 22.06 23.49 24.02 24.65 23.46 23 25.16
0.1 22.68 24.14 25.56 26.76 25.9 22.3 22.42
0.12 21.22 25.46 24.47 23.85 22.66 23.66 25.71

86
Journal of Advanced Studies in Finance
Volume VII Issue1(13) Summer 2016

Figure 5 - Mean surface of Mτ∗

Figure 6 - Stdev surface of 𝑀𝜏∗


Google-finance website. Time series plot of return series is given by Figure 7. The p-value chi-squared
normality test is 0 which shows the log-returns (ri ) are not normally distributed. It is seen that a t 3 (student-t with
3 degrees of freedom) is suitable for returns. The S0 = 34.41 and B0 = 100, then γ0 = 0.3441. The risk free
rate is 0.19 taken from https://www.treasury.gov. Here, it is assumed that γt = γ0 . The probability of having
statistical arbitrage is pt = P(St > γt Bt ) = P(S0 ∏ti=1(1 + ri ) > γ0 Bt ) = P(∏ti=1(1 + ri ) > ert ) =
∑ti=1 log(1+ri )
P( t
> 𝑟) ≈ P(r̅t > 𝑟).
Time series plot of log-return
0.05
0.0
-0.05
-0.10

0 50 100 150 200 250

Figure 7 - Time series plot of log-returns


The last equality uses this fact that log(1 + ri ) ≈ ri . A central limit theorem implies that r̅t has
approximately normal distribution with mean 3 and variance 6⁄t. The plot of pt is given as follows in Figure 8.

87
Journal of Advanced Studies in Finance
Volume VII Issue1(13) Summer 2016

The probability of statistical arbitrage

1.00
0.95
0.90
0.85
0.80
0.75
0.70

0 50 100 150 200 250

Figure 8 - The probability of statistical arbitrage


Part 2 CNPVaR for foreign exchange risk
Rockafellar and Uryasev (2000) proposed an approach to optimizing a financial portfolio by minimizing
conditional value-at-risk (CVaR) rather than minimizing value-at-risk (VaR). Ye and Tiong (2000) studied the
NPVaR method in infrastructure project investment evaluation. Thus, Wang and Gao (2012) applied the NPVaR
criteria in foreign exchange risks of international construction projects. In this note, the minimizing CNPVaR for
projects, in the presence of foreign exchange risk. Following Wang and Gao (2012), the NPV of project involving
ai f i
the foreign exchange risk is: NPV = ∑∞ i=1 (1+r)i , where a i is the exchange rate of i-th currency to reference
currency and fi is the cash flow at time i base on i-th currency. The ai is random variable while r is non-random
interest rate. Variables fi can be interpreted as the difference of volume of cash-in-flow and cash-out-flow of i-th
currency. The aim of this note, is to maximize the CNPVaR with respect to {fi }.
However, in a exchange shop, for example, in each time, a portfolio of currencies are sold or bought. In
this note, a simple case is considered where at time i in project, the volume of cash flow of fi of i-th currency is
achieved. The rest of the current part is organized as follows. In section 4, the CNPVaR is formulated for a
location-scale distribution is obtained. In section 5, the application of quadratic and dynamic programming is
studied.
4. CNPVaR in location-scale distribution
Suppose that all variables ai are independent and normally distributed with mean μi and variance σ2i .
Then, NPV is normally distributed with the mean (μ) and variance σ2 (assuming σ2 < ∞), where:
∞ ∞
μi fi 2
σ2i fi2
μ=∑ and σ = ∑ .
(1 + r)i (1 + r)2i
i=1 i=1

Following Sarykalin et al. (2008) for normally distributed variables the CVaR (here, the CNPVaR) is given
1
by: CNPVaR = μ + k1 (α)σ, where k1 (α) = 2π exp(−(1 − α)(erf −1 (2α − 1))2 ) where erf(z) =

1 z 2
∫ e−0.5z . It is easy to see that erf −1 (x)
= N −1 (x + 0.5), where N−1 is the quantile function of normal
√2π 0
distribution.
For example, assuming α = 0.01, then k1 (α) = 3.047 or k1 (0.05) = 1.264.
∂CNPVaR μ k (α)σ2 f μi (1+r)i σ f μ σ2j
Notice that: = (1+r)
i 1 i i
i + σ(1+r)2i = 0, then, fi = (α)σ2
. Therefore, fi = μi × × (1 + r)i−j .
∂fi −k1 i j j σ2i

88
Journal of Advanced Studies in Finance
Volume VII Issue1(13) Summer 2016
fi
Remark 1. When there is one currency in NPV, then = (1 + r).
fi−1

Remark 2. It is not reasonable to assume the normal distribution for ai ′s. Usually they have distributions
fi
with fat tails, like student-t or generalized error distribution (GED). However, when (1+r) i are

bounded, using the central limit theorem, the above results are satisfied.
1 α
Remark 3. Following Acerbi (2002), the CNPVaR is given by CNPVaR α = α ∫0 NPVaR β dβ. Using the
Monte Carlo estimate the above integral is estimated.
𝑎𝑖 −𝜇
Remark 4. Assuming a scale-location family for 𝑎𝑖 , like the 𝑔0 ( 𝜎
), the 𝑁𝑃𝑉𝑎𝑅𝛽 is 𝜇 − 𝑘𝛽 𝜎, where
𝑘𝛼 is the 𝛼-th quantile of distribution of 𝑔0 .
1 𝛼
Remark 5. Considering the above remarks, then 𝐶𝑁𝑃𝑉𝑎𝑅𝛼 = 𝜇 + 𝑘̅ 𝜎, where 𝑘̅ = ∫0 𝑘𝛽 𝑑𝛽. Again,
𝛼
𝑓𝑖 𝜇 𝜎2
𝑓𝑗
= 𝜇 𝑖 × 𝜎𝑗2 × (1 + 𝑟)𝑖−𝑗 .
𝑗 𝑖

5. Quadratic and dynamic programming


Hereafter, the application of quadratic programming is studied. According to the Remark 5, CNPVaR α =
μi σ2i
∑∞ ∗
i=1 μi fi+ k̅ (∑∞ σ ∗2 2 0.5
i=1 i if ) , where μ ∗
= i and σ
(1+r)i
∗2
i = . Considering some constrains like
(1+r)2i
∗2
∑∞ ∗
i=1 μi bi ≤ b and ∑∞
i=1 σi di ≤ d, this problem reduces to a quadratic programming.

Example 1:
−1 α
Consider two cash flows where g 0 (x) = e−x , x > 0. Notice that k̅ = ∫0 ln(1 − β)dβ which is 1 +
α
(1−α)ln(1−α)
α
. For α = 0.05, it is 0.02543. Suppose that μ1 = μ2 = σ1 = σ2 = 1, and r=0.05. Then, the linear
f f f2 f2
1
programming is: maxZ = 1.05 2
+ 1.1025 + 0.02543√1.1025
1
+ 1.2155
2
such that f1 + f2 ≤ 1 and f′s are
positive. Then, f1 = 1 and f2 = 0 and Z=0.9766. Next, to brows the dynamic programming setting here, let:
CNPVaR n,α = ∑ni=1 μ∗i fi + k̅ (∑ni=1 σ∗2 2 0.5
i fi ) and 𝜌𝑛 = CNPVaR α = (∑∞ ∗2 2 0.5
i=1 σi fi ) − (∑ni=1 σ∗2 2 0.5
i fi ) .
Then, max(CNPVaR α ) = max(CNPVaR n,α + 𝜌𝑛 ), which defines the dynamic programming approach for
maximizing the CNPVaR α .
References
[1] Acerbi, C. 2002. Spectral measures of risk: a coherent representation of subjective risk aversion. Journal of
Banking and Finance, 26:1505–1518.
[2] Bondarenko, O. 2003. Statistical Arbitrage and Securities Prices. Review of Financial Studies, 16: 875–919.
[3] Calin, O. 2012. An introduction to stochastic calculus with applications. On-line lecture notes.
[4] Rockafellar, R. T., and Uryasev, S. 2000. Optimization of conditional value-at-risk. Journal of Risk, 3: 21-41.
[5] Sarykalin, S., Serraino, G., and Uryasev, S. 2008. Value-at-risk vs. conditional value-at-risk in risk
management and optimization. Tutorials in Operations Research. C @ Informs: 270-298.
[6] Vose, D. 2010. Risk analysis: a quantitative guide. Wiley.
[7] Wang, X. Q., and Gao, B. 2012. Dynamic measurement and evaluation on foreign exchange risks of
international construction projects. Proceedings of the 2012 IEEE IEEM. USA.
[8] Ye, S., and Tiong, R. L. K. 2000. Npv-at-risk method in infrastructure project investment evaluation. Journal
of Construction Engineering and Management, 3: 227-233.

89
Journal of Advanced Studies in Finance
Volume VII Issue1(13) Summer 2016
ASERS

Web: www.asers.eu
URL: http://www.asers.eu/asers-publishing
E-mail: asers@asers.eu asers2010@yahoo.co.uk
ISSN 2068 – 8393
Journal DOI: http://dx.doi.org/10.14505/jasf
Journal’s Issue DOI: http://dx.doi.org/10.14505/jasf.v7.1(13).00

90

You might also like