Professional Documents
Culture Documents
Editorial Note 1
Articles REVAAM Model to determine a company's value by multiple valuation and linear regression
analysis
9
Niki Geiersbach
Reasoning Disparities Between HK and US Managers 131
Isola Oluwabusuyi
A Study on Using Business Intelligence for Improving Marketing Efforts 141
ISBN: 978-1-4251-8179-6
ISSN: 1918-2325
2010 Business Intelligence Journal 1
Article 5: Application of Ants Colony System for Bankruptcy Prediction of Companies listed in Tehran Stock
Exchange
Author: 1 - Vali Khodadadi Shahid Chamran University of Ahwaz.
2 - Abolfazl (Parviz) Zandinia Allameh Tabatabaei University.
3 - Marzieh Nouri MA in accounting.
e-mail: m_n592007@yahoo.com.
Article 6: TheValue Relevance of Earnings Components in Two Different GAAPs
Author: 1 - Zahra Nikbakht Payam Noor University (PNU), Kohpayeh, Esfahan, Iran.
email: Zahra.Nikbakht77@gmail.com
2 - Abbas Aflatooni Department of Accounting, College of Economics and Social
Sciences, Shahid Chamran University of Ahvaz, Ahvaz, Iran.
email: abbasaflatooni@gmail.com
Article 7: AWord on Game Theory Application to Corporate Finance
Author: Jos Carlos Arias (PhD, DBA) Vice-Chancellor (Rector) of the Isles Internationale
Universit, School of Doctoral Studies of the European Union (Brussels, Belgium) and Senior
Partner of Secured Assets Yield Corporation Limited (London, UK)
email: josecarlosarias@saycocorporativo.com
Article 8: The Impact of International Business on the Global Economy
Author: Niki Geiersbach Candidate to PhD in Management Science at the School of Doctoral
Studies of the European Union.
Article 9: Reasoning Disparities Between HK and Us Managers
Author: Dr. Isola Oluwabusuyi (DBA - Argosy University) Leading instructor at the School of
Business and Accounting, Brown Mackie College, Atlanta.
Article 10: A Study on Using Business Intelligence for Improving Marketing Efforts
Author: 1 - Arpan Kumar Kar Doctorate student Information systems area XLRI School of
business and Human Resources.
email: arpan.kar@astra.xlri.ac.in
2 - Ashis Kumar Pani Faculty Information systems area XLRI School of business and
Human Resources.
email: akpani@xlri.ac.in
3 - Supriya Kumar De Faculty Information systems area XLRI School of business and
Human Resources.
email: skde@xlri.ac.in
Article 11: The Effect of Corporate Governance Structure on the Extent ofVoluntary Disclosure in Iran
Author: 1 - Vali Khodadadi Department of Accounting, College of Economics and Social Sciences,
Shahid Chamran University of Ahwaz, Ahwaz, Iran.
2 - Soheila Khazami M.A in Accounting, Islamic Azad University, Shushtar branch,
Khuzestan, Iran.
3 - Abbas Aflatooni Department of Accounting, College of Economics and Social
Sciences, Shahid Chamran University of Ahwaz, Ahwaz, Iran.
In order to make contact with any of the Authors referred to above, please forward your request to: edit.bij@saycocorporativo.
com, including BIJs edition (BIJ Volume 3, Number 2, July 2010), articles and authors names with your requirement. BIJs
Editor will be glad to submit your requests or inquiries before authors.
should be used and abbreviations should sections can include subheadings, and when
be avoided. No literature should be cited. appropriate, both sections can be combined.
Following the abstract, about 3 to 10 key The Acknowledgments of people,
words that will provide indexing references grants, funds, etc should be brief.
to should be listed. Tables should be kept to a minimum
A list of non-standard Abbreviations and be designed to be as simple as
should be added. In general, non-standard possible. Tables are to be typed double-
abbreviations should be used only when spaced throughout, including headings and
the full term is very long and used often. footnotes. Each table should be on a separate
Each abbreviation should be spelled out and page, numbered consecutively in Arabic
introduced in parentheses the first time it is numerals and supplied with a heading and
used in the text. a legend. Tables should be self-explanatory
The Introduction should provide without reference to the text. The details
a clear statement of the problem, the of the research methods should preferably
relevant literature on the subject, and the be described in the legend instead of in the
proposed approach or solution. It should be text. The same data should not be presented
understandable to colleagues from a broad in both table and graph form or repeated in
range of disciplines. the text.
Materials and methods should Figure legends should be typed in
be complete enough to allow possible numerical order on a separate sheet. Graphics
replication of the research. However, only should be prepared using applications
truly new research methods should be capable of generating high resolution GIF,
described in detail; previously published TIFF, JPEG or PowerPoint before pasting in
methods should be cited, and important the Microsoft Word manuscript file. Tables
modifications of published methods should should be prepared in Microsoft Word. Use
be mentioned briefly. Capitalize trade Arabic numerals to designate figures and
names and include the manufacturers upper case letters for their parts (Figure 1).
name and address. Subheadings should be Begin each legend with a title and include
used. Methods in general use need not be sufficient description so that the figure is
described in detail. understandable without reading the text
Results should be presented with clarity of the manuscript. Information given in
and precision. The results should be written legends should not be repeated in the text.
in the past tense when describing authors References: In the text, a reference
findings. Previously published findings identified by means of an authors name
should be written in the present tense. Results should be followed by the date of the
should be explained, but largely without reference in parentheses. When there are
referring to the literature. Discussion, more than two authors, only the first authors
speculation and detailed interpretation of name should be mentioned, followed by et
data should not be included in the Results al. In the event that an author cited has had
but should be put into the Discussion section. two or more works published during the
The Discussion should interpret the same year, the reference, both in the text and
findings in view of the results obtained in in the reference list, should be identified by
this and in past studies on the topic. State a lower case letter like a and b after the
the conclusions in a few sentences at the end date to distinguish the works.
of the paper. The Results and Discussion
Smith (2000), Wang et al. (2003), (Kelebeni, Short Communications are limited to a
1983), (Usman and Smith, 1992), (Chege, maximum of two figures and one table. They
1998; Chukwura, 1987a,b; Tijani, 1993, should present a complete study that is more
1995), (Kumasi et al., 2001) limited in scope than is found in full-length
papers. The items of manuscript preparation
References should be listed at the end of listed above apply to Short Communications
the paper in alphabetical order. Articles with the following differences: (1) Abstracts
in preparation or articles submitted for are limited to 100 words; (2) instead of a
publication, unpublished observations, separate Materials and Methods section,
personal communications, etc. should not research methods may be incorporated into
be included in the reference list but should Figure Legends and Table footnotes; (3)
only be mentioned in the article text (e.g., Results and Discussion should be combined
A. Kingori, University of Nairobi, Kenya, into a single section.
personal communication). Journal names
are abbreviated according to Chemical Ab- Proofs and Reprints
stracts. Authors are fully responsible for the
accuracy of the references. Electronic proofs will be sent (e-mail
attachment) to the corresponding author as
Examples: a PDF file. Page proofs are considered to
be the final version of the manuscript. With
Papadogonas TA (2007). The financial the exception of typographical or minor
performance of large and small firms: clerical errors, no changes will be made in
evidence from Greece. Int. J. Financ. the manuscript at the proof stage. Because
Serv. Manage. 2(1/2): 14 20. BIJ will be published online without access
restrictions, authors will have electronic
Mihiotis AN, Konidaris NF (2007). Internal access to the full text (PDF) of the article.
auditing: an essential tool for adding Authors can download the PDF file from
value and improving the operations of which they can print unlimited copies of
financial institutions and organizations. their articles.
Int. J. Financ. Serv. Manage. 2(1/2): 75
81. Copyright
Abstract
This paper shows an alternative model to the widely used method of multiple valuation (or relative
valuation) in order to calculate the value of a company by using either the Price Earnings (PE) and/or
the Enterprise Value to Earnings Before Interest, Taxes, Depreciation and Amortization (EV/EBITDA).
When calculating multiples, analysts tend to consider average multiples within an industry and apply them
directly to the target company; however, we believe that this practice is not considering differences among
the companies being compared, although they belong to the same sector or industry. REVAAM Model
uses linear regression to calculate adjusted PE and EV/EBITDA multiples by taking into consideration
profitability factors for each multiple in order to differentiate companies in the samples. Calculations
are based on public data for US companies, but could be further expanded to other markets. Not only
REVAAM Model provides a better estimate to relative valuation analysis than simply using average
multiples, but it could be used to compare under/overvalued companies or sectors, and also analyze
multiple value changes over time as the intrinsic fundamentals change.
Acosta-Calzado Luis G., Acosta Calzado C., Nurrieta-Romo H. - REVAAM Model to determine a companys value by multiple valuation and linear regression analysis
10 Business Intelligence Journal July
we will consider the use of the PE and EV/ aPE = the value of the intercept calculated
EBITDA multiples. by the regression analysis
bROE = the value of the coefficient for
Design ROE
ROE = the return on equity, calculated
As we said before, the best way to as the net income divided by the companys
determine a companys value is by using beginning equity value
the DCF method, and in order to do so we bNM = the value of the coefficient for Net
need to forecast the free cash flow, either Margin
to the firm or to the equity shareholders. Net M arg in = calculated as the net
This process has to be detailed and requires income of year t divided by the sales of year
knowledge of the companys financials and f = the sum of errors derived from the
the expected growth. In consequence, we regression analysis
propose to adjust the PE multiple by using
Equation 2
the Return on Equity5 (ROE) and the Net
Margin6 as the indicators of profitability for EV/EBITDA = aEVEBITDA + bROC * ROC + bOM *
equity shareholders of the companies being Pre - tax Operating M arg in + f
analyzed. In contrast, for the EV/EBITDA
multiple we propose adjusting by using Where
the Return on Capital7 (ROC) and the Pre-
tax Operating Margin8. These adjustments EV/EBITDA =the EBITDA multiple
should be made using multiple regression calculated as the Enterprise Value or
analysis, in order to minimize errors, where Value of the Net Operating Assets divided
the independent variables are the multiples by the Earnings before Interests, Taxes,
and the other factors are the dependent Depreciation and Amortization
variables. These relationships are shown in aEVEBITDA = is the value of the intercept
the equations below: calculated by the regression analysis
bROC = the value of the coefficient for
Equation 1
ROC
P/E = aPE + bROE * ROE + bNM * Net M arg in + f ROC = the return on capital, calculated
as the after-tax operating income (EBIT*(1-
Where, tax)) divided by the companys beginning
equity value plus the debt balance
P/E = the price earnings ratio calculated bOM = the value of the coefficient for Net
as the price per share divided by net earnings Margin
per share Pre - tax Operating M arg in = calculated
5
Return on Equity measures the rate of return on shareholders
as the pre-tax operating income of year t
equity and is calculated as the Net Income divided by the book divided by the sales of year
value of equity f = the sum of errors derived from the
6
Net Margin measures the profitability and is calculated as the net
income divided by revenues
regression analysis
7
Return on Capital measures the rate of return on the companys
capital and is calculated as operating earnings after taxes divided
by the book value of equity plus the book value of debt
8
Pre-tax Operating Margin or EBIT measures profitability before
interests and taxes and is calculated as operating revenues less
operating expenses plus non-operating income
Acosta-Calzado Luis G., Acosta Calzado C., Nurrieta-Romo H. - REVAAM Model to determine a companys value by multiple valuation and linear regression analysis
12 Business Intelligence Journal July
Industries 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Advertising 30 33 37 32 34 32 36 39 28 36
Aerospace/Defense 39 64 76 72 66 70 73 68 65 67
Air Transport 36 44 42 43 45 45 56 48 43 44
Auto & Truck 19 20 22 21 24 29 31 27 19 22
Auto Parts 55 64 61 62 59 58 64 55 53 54
Bank 211 489 505 558 546 537 598 551 521 527
Beverage 32 41 43 41 37 41 48 43 40 41
Biotechnology 0 75 90 84 89 87 105 102 107 121
Building Materials 50 65 68 62 60 57 60 59 64 67
Chemical 125 141 149 144 136 143 154 143 137 145
Coal 0 5 7 8 10 12 16 17 17 21
Computers 399 467 630 535 530 533 572 518 445 462
Diversified Co. 92 102 105 103 117 118 134 106 117 126
Drug 286 285 288 276 304 306 334 367 341 337
Educational Services 27 35 36 34 37 38 37 38 33 38
Electric Utility 88 82 75 77 74 74 75 71 67 66
Electrical Equipment 81 91 94 86 91 91 94 85 82 87
Electronics 150 208 211 194 189 186 196 187 181 192
Entertainment 91 106 126 116 117 118 132 129 115 130
Environmental 50 79 85 77 84 91 96 88 78 91
Financial Svcs. 184 223 238 231 232 244 269 294 295 296
Food Processing 86 115 114 104 103 110 123 122 108 121
Furniture and Furnishings 33 32 35 36 37 36 38 38 33 35
Healthcare Information 32 32 35 35 31 35 34 37 28 33
Homebuilding 54 56 48 44 33 34 41 35 31 28
Hotel/Gaming 52 84 94 80 76 76 84 74 67 74
Household Products 29 32 32 32 29 26 31 27 25 23
Human Resources 0 21 28 27 27 30 35 34 30 30
Industrial Services 173 211 205 190 199 207 230 194 166 168
Information Services 0 18 30 29 32 36 41 37 33 29
Insurance 86 84 96 115 120 127 137 135 121 124
Internet 315 446 422 330 346 365 389 320 260 295
Investment Co. 45 41 39 39 36 36 35 31 41 44
Machinery 134 167 172 153 147 149 153 135 123 130
Manuf. Housing/RV 20 19 20 18 18 16 19 17 17 15
Maritime 14 19 19 23 27 39 46 51 55 53
Industries 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Medical Services 156 207 226 197 194 184 186 177 159 162
Medical Supplies 182 224 244 236 261 261 279 273 251 264
Metal Fabricating 38 45 45 37 37 41 37 36 34 36
Metals & Mining 36 44 55 54 75 77 82 77 77 79
Natural Gas 71 72 72 70 66 65 64 69 75 75
Newspaper 18 20 19 20 19 19 18 17 15 15
Office Equip/Supplies 28 33 33 31 27 27 26 24 25 25
Oilfield Svcs/Equip. 71 87 87 88 92 98 110 112 111 113
Packaging & Container 36 42 41 34 34 37 36 34 32 31
Paper/Forest Products 48 47 43 40 38 40 42 38 37 39
Petroleum 134 175 170 162 177 178 208 210 211 222
Pharmacy Services 10 14 15 15 13 15 20 18 18 21
Power 15 30 33 29 33 36 55 69 76 87
Precious Metals 30 37 42 48 60 62 67 83 74 78
Precision Instrument 85 108 114 102 103 104 104 102 89 98
Property Management 0 0 0 0 0 0 0 11 16 20
Publishing 43 44 42 40 42 47 50 39 26 30
R.E.I.T. 143 152 151 144 134 122 143 146 143 143
Railroad 16 15 15 16 17 18 20 15 14 15
Recreation 81 86 88 73 77 74 84 72 63 65
Restaurant 90 93 94 85 83 82 81 74 67 68
Retail (Special Lines) 232 240 255 255 238 237 227 219 206 213
Retail Automotive 10 18 17 12 26 27 25 15 15 15
Retail Building Supply 11 10 10 8 8 10 9 8 7 7
Retail Store 80 117 119 90 89 90 91 72 67 75
Securities Brokerage 27 27 24 27 25 31 32 30 31 30
Semiconductor 113 126 123 128 138 135 138 152 136 139
Shoe 26 26 27 24 23 22 24 19 18 19
Steel 48 48 45 41 36 40 46 38 32 35
Telecom. Equipment 112 140 146 124 119 122 136 123 109 115
Telecom. Services 200 252 234 180 176 188 196 173 163 164
Thrift 124 116 146 242 221 221 248 233 233 227
Tobacco 11 12 12 11 12 11 11 10 11 12
Toiletries/Cosmetics 17 20 24 21 22 20 21 20 22 19
Trucking 45 42 40 39 35 37 38 31 32 33
Water Utility 13 15 16 16 16 16 16 15 15 15
Wireless Networking 69 79 76 63 65 66 73 73 56 60
Total 5587 6959 7320 6913 6973 7092 7659 7249 6752 7036
Acosta-Calzado Luis G., Acosta Calzado C., Nurrieta-Romo H. - REVAAM Model to determine a companys value by multiple valuation and linear regression analysis
14 Business Intelligence Journal July
The following graphs show the relationships 2009 of all the companies analyzed. At
between P/E and ROE and Net Margin first glance, it appears to be no linear
and between EV/EVITDA and ROC and relationship at all.
Pre-tax Operating Margin for the year
EV/EBITDA EV/EBITDA
multiple multiple
1800 1800
1600 1600
1400 1400
1200 1200
1000 1000
800 800
600 600
400 400
200 200
0 0
0 1000 2000 3000 4000 5000 6000 7000 0 1000 2000 3000 4000 5000 6000 7000
ROC Pre-tax Operating Margin
The basic linear regression model assumes From equations 1 and 2, we apply natural
that the contributions of the different logarithms to obtain:
Equation 3
independent variables to the prediction of
the dependent variable are additive and Ln (P/E) = aPE + bROE * Ln (ROE) + bNM *
they tend to follow normal distributions. Ln (Net M arg in) + f
Equation 4
In our case, the relationships between our
variables may be multiplicative and also Ln (EV/EBITDA) = aEVEBITDA + bROC * Ln (ROC) +
they have highly skewed distributions bOM * Ln (Pre - tax Operating Margin) + f
(positive values). Hence it may be
possible to make their distributions Now, scatter plots for the relationships
more normal-looking by applying the between the variables transformed
logarithm transformation, as shown in to natural logarithms show a linear
the following equations: relationship.
Ln(P/E) Ln(P/E)
8 8
6 6
4 4
2 2
0 0
-2 -2
-4 -4
-6 -6
-8 -6 -4 -2 0 2 4 -8 -6 -4 -2 0 2 4
Ln(ROE) Ln(Net Margin)
Ln(EV/EBITDA) Ln(EV/EBITDA)
8 7
6
6
5
4 4
2 3
2
0
1
-2 0
-1
-4
-2
-6 -3
-8 -6 -4 -2 0 2 4 -8 -6 -4 -2 0 2 4
Ln(ROC) Ln(Pre-tax Operatin Margin)
It has to be noted, that the number of P/E ratio. Although the R2 coefficients do
companies that can be used is reduced by not present high values, the F-test shows that
transforming variables to natural logarithms we can use the linear model. For all years,
due to the fact that negative values for our ROE is a significant variable (not equal to
variables cannot be transformed to natural zero), but that is not the case for Net Margin
logarithms. This is a limitation for the model. in years 2004, 2007 and 2008. For purposes
of comparison we will use both factors in
Results all years.
Acosta-Calzado Luis G., Acosta Calzado C., Nurrieta-Romo H. - REVAAM Model to determine a companys value by multiple valuation and linear regression analysis
16 Business Intelligence Journal July
Concept 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Multiple correlation factor 0.2715 0.4975 0.5090 0.5882 0.6216 0.5984 0.5673 0.5344 0.4400 0.5324
R-square coefficient 0.0737 0.2475 0.2591 0.3459 0.3864 0.3581 0.3218 0.2856 0.1936 0.2834
F test value 799.74
111.064 545.417 503.213 800.975 964.306 890.183 622.593 367.977 543.500
Critical F value 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Intercept 3.2672 1.7281 1.6491 2.0136 1.9820 1.9316 1.8731 1.9007 1.2959 1.5055
Prob Intercepcin 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
ln(ROE) -0.2856 -0.5912 -0.5763 -0.5552 -0.5558 -0.6222 -0.5305 -0.5242 -0.4720 -0.4952
Prob ln(ROE) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
ln(Net Margin) 0.3490 0.0744 0.0406 0.0297 -0.0212 0.0436 -0.0652 -0.0242 0.0066 -0.0841
Prob (Net Margin) 0.0000 0.0000 0.0444 0.0449 0.1424 0.0051 0.0000 0.1398 0.7257 0.0000
In the case of the EV/EBITDA relationship other factors that need to be considered in
established in equation (4), we obtained the our analysis. As with the Net Margin for
results in the table below. The F-test also equation (3), there are some years (2003,
indicates that the linear regression model 2005 and 2007) in which Pre-tax Operating
is adequate. R2 values are smaller than for Margin is not significant.
equation (3); this may mean that there are
Concept 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Multiple correlation factor 0.1125 0.2354 0.2566 0.2196 0.2201 0.1790 0.1844 0.2308 0.2535 0.2192
R-square coefficient 0.0127 0.0554 0.0658 0.0482 0.0485 0.0321 0.0340 0.0533 0.0643 0.0481
F test value 22.458 113.001 127.863 93.823 95.170 61.088 68.194 98.852 116.734 86.831
Critical F value 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Intercept 1.9749 1.4948 1.3906 1.6710 1.7411 1.8215 1.8645 1.7209 1.0705 1.5180
Prob Intercepcin 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
ln(ROC) -0.1282 -0.2349 -0.2316 -0.1768 -0.1377 -0.1462 -0.1260 -0.1995 -0.2607 -0.1388
Prob ln(ROC) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
ln(Pre-tax Op Margin) 0.0795 0.0649 0.0566 -0.0042 -0.0682 -0.0023 -0.0334 -0.0016 0.0597 -0.0141
Prob (Pre-tax Op Margin) 0.0001 0.0001 0.0001 0.7744 0.0000 0.8795 0.0143 0.9147 0.0004 0.2998
Now that we have established our the P/E and EV/EBITDA and compare them
regression parameters for equations (3) and to arithmetic averages.
(4) by year, we now calculate for each year
As can be seen in the numbers and in the P/E and EV/EBITDA in each year are above
following graphs simple average values for those calculated from the regressions.
Graph 9. P/E regression vs P/E average Graph 10. EV/EBITDA regression vs EV/EBITDA
average
In the case of P/E ratios, average values to obtain an equity value for the company
are about 2.2 times the values obtained from being analyzed.
regression analysis, and for EV/EBITDA
average values are 1.57 times the regression Graph 11. Regression Multiples over time
values.
Another important aspect can be derived 30.00
P/E
Acosta-Calzado Luis G., Acosta Calzado C., Nurrieta-Romo H. - REVAAM Model to determine a companys value by multiple valuation and linear regression analysis
18 Business Intelligence Journal July
25.00
EV/EBITDA regression
EV/EBITDA average
20.00
15.00
10.00
5.00 4.7
-
Information Services
Medical Supplies
Machinery
Advertising
Oilfield Svcs/Equip.
Telecom. Services
Pharmacy Services
Computers
Medical Services
Metal Fabricating
Trucking
Petroleum
Securities Brokerage
Diversified Co.
Building Materials
Tobacco
Hotel/Gaming
Chemical
Telecom. Equipment
Aerospace/Defense
Toiletries/Cosmetics
Maritime
Property Management
Biotechnology
Investment Co.
Thrift
Office Equip/Supplies
Household Products
Precious Metals
Insurance
Environmental
Beverage
Coal
Power
The solid line in Graph 12 is the average multiple that analysts use when
calculated EV/EBITDA for each sector doing relative valuation.
using the corresponding average ROC and Sectors whose average EV/EBITDAs
average Pre-tax Operating Margin in 2009. are above the REVAAMs EV/EBITDAs
Dots represent the simple EV/EBITDA are considered to be overvalued, whereas
80.00
P/E regression
70.00 P/E average
60.00
50.00
40.00
30.00
20.00
10.00
-
Healthcare Information
Precision Instrument
Entertainment
Recreation
Machinery
Water Utility
Food Processing
Advertising
Electrical Equipment
Telecom. Services
Computers
Pharmacy Services
Petroleum
Diversified Co.
Internet
Hotel/Gaming
Publishing
Telecom. Equipment
Aerospace/Defense
Electric Utility
Retail Automotive
Retail Building Supply
Industrial Services
Thrift
Office Equip/Supplies
Household Products
Semiconductor
Insurance
Auto Parts
Beverage
Steel
Coal
Power
Similar to Graph 12, the solid line in Taking into consideration results from
Graph 13 is the calculated P/E for each both graphics 12 and 13, we find common
sector using the corresponding average sectors that have the same results. However,
ROE and average Net Margin in 2009, and there are sectors that show opposite results
dots represent the arithmetic average of P/E in both analysis, such as the Electric Utility,
multiples in each sector. Retail Automotive, REIT, Manufacturing
Comparing average P/Es to REVAAMs Housing/RV, Trucking and Telecom Services
P/Es, we can observe in Graph 13 that 27 sectors.
sectors are undervalued, 13 sectors are We have used the REVAAM Model
within a 5% range of fair price, and 32 in comparing market-wide multiples in
are overvalued. Thrift/Banking, Precious specific years and through time, and also by
Metals, Tobacco, Healthcare Information, comparing different sectors. Now we will
Insurance and Educational Services are the apply the model to a specific company that
most overvalued sectors, while Maritime, belongs to the Drug sector (Pharmaceutical
Steel, Manufacturing Housing/RV, Office Laboratories), and we will use a US based
Equipment and Supplies and Metal company called Perrigo Co. (Ticker PRGO).
Fabricating are the most undervalued. The table below shows the parameters
Acosta-Calzado Luis G., Acosta Calzado C., Nurrieta-Romo H. - REVAAM Model to determine a companys value by multiple valuation and linear regression analysis
20 Business Intelligence Journal July
for Perrigo Co. at the end of each year and REVAAM Model in the specific Drug
the actual P/E and EV/EBITDA multiples sector. The same is done to the multiples of
that the company is trading at, and then the whole Drug sector.
we calculate these multiples by using the
The following graphs (14 and 15) show Perrigo was trading at the end of year 2009,
the REVAAM calculations for the P/E and and simple average multiples for the Drug
EV/EBITDA multiples respectively and the sector at the same date.
comparison against the current multiples
Graph 14. Perrigos and Drug sector P/E Graph 15. Perrigos and Drug sector EV/EBITDA
calculation calculation
Current P/E REVAAM P/E Av. Sector P/E REVAAM Sector P/E Current EV/EBITDA REVAAM EV/EBITDA
80.00 25.00
Av. Sector EV/EBITDA REVAAM Sector EV/EBITDA
70.00
20.00
60.00
50.00 15.00
40.00
30.00 10.00
20.00
5.00
10.00
- -
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
As it can be seen, in both graphs the individual companies, such as the one used
REVAAM multiples (both P/E and EV/ as example, we may be understating its value
EBITDA) for the Drug sector are very close if we use only the average sector multiple.
to the simple averages of those multiples. In our experience most (over 60%) Mergers
However, when looking at the multiples and Acquisitions are valued using simple
at which Perrigo was trading compared average multiples.
to calculated REVAAM multiples for the As stated initially, the preferred
company, it seems that Perrigo was trading valuation method for the authors is the
at lower multiples than it should had been use of DCF analysis, however relative
or it was undervalued. In the case of the P/E valuation may give a quick insight of the
multiple this is true for all years except for intrinsic value. Just considering one, two
2000 and 2008, and for the EV/EBITDA or more financial factors that differentiate
multiple Perrigo was undervalued for all companies valuations, is a better proxy to
years except for 2007, 2008 and barely 2009. what a DCF analysis will show for different
Another important conclusion is that, companies. The model also lets us compare
in general, Perrigo is a company whose market valuations to those derived from
REVAAM multiples are over the sector the REVAAM Model. As seen in the recent
REVAAM multiples, suggesting that it crises the market sometimes is not rational,
outperforms other companies in the sector, and this model let us take into consideration
and may be a good investment in the long objective financial and operational factors
run. without the inertia of bull or bear markets.
Additionally, the REVAAM Model provides
Conclusions different depth levels of analysis, from
the whole market to sector and company
REVAAM Model does present several specific calculations, and allows for time
limitations: such as a) it considers few comparison based on the prevailing factors
factors to define the multiples, b) due at each period.
to logarithms transformation it cannot Further work in this model may include
consider companies with negative values determining other significant factors that
for the factors, c) it is based on historical may affect valuations to be included in
data and does not take into consideration the regression analyses, also developing a
future growth, d) it requires at least 30 version for mature markets and markets in
companies for the linear regressions to be development, that take into consideration
significant, and some sectors do not include risk differences among countries.
that number of companies, e) is applied to
public companies because of information References
availability and f) may not apply to markets
where there are few public companies in Schreiner, Andreas. 2007. Equity
order to be statically significant. Valuation Using Multiples: An
However, we believe that it provides a Empirical Investigation Dissertation der
better estimate to relative valuation analysis Universitt St. Gallen (HSG)
than simply using average multiples. We
have seen that even for sector analysis some Damodaran, Aswath. 2002. Investment
valuations could be 50% percent off (both Valuation: Tools and Techniques for
positive and negative), and when valuing
Acosta-Calzado Luis G., Acosta Calzado C., Nurrieta-Romo H. - REVAAM Model to determine a companys value by multiple valuation and linear regression analysis
22 Business Intelligence Journal July
McKinsey & Company Inc., et al. 2005. Pagano, Robert R. 1998. Estadstica para
Valuation: Measuring and Managing the las Ciencia del Comportamiento, 5
Value of Companies, Edition, University Edicin.
Edition.
Beaver, W. 1967. Financial ratios as
Damodaran, Aswath. 2010. The Dark Side predictors of failure, Empirical Research
of Valuation: Valuing Young, Distressed, in Accounting. (1966), Selected
and Complex Businesses, 2nd Edition, studies, supplement to the Journal of
Pearson Education, Inc. Accounting Research, January, 1966.
Medlen, Craig. 2007. The bubble machine: Greene W. 1993. Econometric Analysis,
relative capital valuation, distributive 2th Edition, Mc. Millan Publishing Co.
shares and capital gains An article from:
Journal of Economic Issues. Mandala G.S. 1983. Limited Dependent and
Qualitative Variables in Econometrics,
Iwanowski, Raymond J. 1995. An Econometric Society, Monographs
investors guide to floating-rate notes: number 3, Cambridge, University.
Conventions, mathematics and relative
valuation (United States fixed-income Bloomberg Professional monitoring service.
research. Portfolio strategues)
Yahoo Finance (http://finance.yahoo.com)
Hoover, Scott. 2005. Stock Valuation: An
Essential Guide to Wall Streets Most Capital IQ (http://www.capitaliq.com)
Popular Valuation Models McGraw-Hill
Library of Investment and Finance. Value Line (http://www.valueline.com)
Abstract
In this paper an attempt is made to develop hybrid models of three layer feed forward back propagation
artificial neural network (ANN) and autoregressive integrated moving average (ARIMA) for forecasting
the future index value and trend of Indian stock market viz. SENSEX, BSE IT, BSE Oil & Gas, BSE 100 and
S& P CNX Nifty. Simulations have been done using prices of daily open, high, low and close of SENSEX,
BSE IT, BSE Oil & Gas, BSE 100 and S& P CNX Nifty. These are chosen as input data values and output
is the forecasted closing price of SENSEX, BSE IT, BSE Oil & Gas, BSE 100 and S& P CNX Nifty for the
next day and future trend. Simulation results of hybrid models are compared with results of ANN based
models and ARIMA based models. Convergence and performance of models have been evaluated on the
basis of the simulation results done on MATLAB6.1 and SPSS13.0.
Merh N., Saxena Vinod P., Raj Pardasani K. - A Comparison between Hybrid Approaches of ANN and ARIMA for Indian Stock Trend Forecasting
24 Business Intelligence Journal July
(b) prices move in trend (c) history repeats Next objective is to compare the results
itself. The use of technical analysis (Malkiel, and trends of actual and predicted values
1996; Skabar and Cloete, 2001) goes of above mentioned indices.
against the conservative academic opinion,
which regards this behavior as irrational To evaluate the performance of the
given the efficient markets hypothesis. models by calculating average absolute
According to Efficient Market Hypothesis error (AAE), root mean square error
(Samanta and Bordoloi, 2005; Fama, (RMSE), mean absolute percentage
1965 and Fama, 1970) all available market error (MAPE), mean percent square
information are factored immediately into error (MPSE) and compare their values.
the formation of stock price and therefore,
the best predictor of future stock price is To compare the results and trends
the latest stock price is the latest available between the traditional hybrid and new
price. Technical analysis, (Murphy, 1999) hybrid approach for forecasting stock
refers to the various methods that aim to market in Indian context.
predict future price movements using past
stock prices and volume information. It is To compare the trends between models
(Mills, 1990; Priestley, 1988) possible that based on single approach based of ANN/
non-linear models are able to explain this ARIMA and hybrid approach using
residual variance and produce more reliable techniques of ANN and ARIMA.
predictions of the stock price movements.
In the present study a hybrid models Financial Time Series
have been developed using techniques of
both artificial neural network (ANN) and A time series is a sequence of vectors
autoregressive integrated moving average Xt = (xt - n, ... ..., xt - i, ... ..., xt - 2, xt - 1) where xt - i
(ARIMA) and their performances are represents past value that varies with time.
compared to the ANN and ARIMA models. In stock prices also the data are typical
In the study MATLAB 6.1.0 and SPSS example of time series. Models in the study
13.0 for windows have been used for predict the future trends on the basis of the
simulations. past values. In the current study open, high,
low and close of BSE 30 (SENSEX), BSE
Purpose of the Research IT, BSE Oil & Gas, BSE 100 and S&P CNX
Nifty are used for the prediction of next
Main objective of the study is to develop stock value and future trend. The price of
models for forecasting the next days volume traded is not a significant predictor
close value of SENSEX, BSE IT, BSE Oil for the time series as during learning in the
& Gas, BSE 100 and S& P CNX Nifty. In ANN model most of the prices against traded
the study hybrid model using techniques volume are zero as the data is not available
of feed forward back propagation with respect to volume traded. Thus traded
supervised learning artificial neural volume is dropped as one of the variables or
network and autoregressive integrated predictor.
moving average are developed and
results are compared.
Merh N., Saxena Vinod P., Raj Pardasani K. - A Comparison between Hybrid Approaches of ANN and ARIMA for Indian Stock Trend Forecasting
26 Business Intelligence Journal July
Table 1. Input and Output Time Series for the Study. Table 3. Control Parameters used in the Hybrid
Model (Error Forecast through ARIMA to ANN) in
Output the Study.
Index Name Input Variables
Variable
BSE 30 (SENSEX) ANN (input for ANN will be
from april Next day Model Index ARIMA
Open High Low Close residual from ARIMA)
16,2004 to april close price
16,2009 ANN(5-5-1)Linear-log
BSE30 ARIMA(p,d,q)= sigmoid-log sigmoid, a =0.5,
BSE IT (SENSEX) ARIMA(1,1,1) h=0.001, Epochs=1000, Error
from april Next day Tolerance=0.01
Open High Low Close
16,2004 to april close price
16,2009 ANN(5-5-1)Linear-log
ARIMA(p,d,q)= sigmoid-log sigmoid, a =0.9,
BSE Oil & Gas BSEIT
Tolerance=0.001
ANN(4-4-1)Linear-log to ANN model. Hence in the first case data
BSE Oli & sigmoid-log sigmoid, a =0.9,
Gas h=0.01, Epochs=1000, Error
ARIMA(1,0,1) set contains four different time series with
Tolerance=0.001 respect to open, high, low and close price
ANN(4-4-1)Linear-log whereas in the second case data set contains
sigmoid-log sigmoid, a =0.9,
BSE 100
h=0.01, Epochs=1000, Error
ARIMA(1,0,0) time series generated as the residual from
Tolerance=0.001 ARIMA model which are represented in the
ANN(4-4-1)Linear-log following diagrams (Figure 1 and Figure 2).
sigmoid-log sigmoid, a =0.9,
S&P CNX
NIFTY h=0.01, Epochs=1000, Error
ARIMA(1,1,1) Input data for all Indices consist of
Tolerance=0.001 data collected from http://www.cmie.com
(CMIE Prowess database release 3.1, Date
of Downloading: May 15, 2009).
F = 0.9 - 0.1
xmax - xmin
Figure 1. Three Layer Feed-forward Neural
such that ymax = 0.9 and ymin = 0.1
Network with One Hidden Layer for Hybrid Model
(3)
(Error Forecast through ANN to ARIMA).
Offset = (ymin - (F) * xmin)
(4)
When data is loaded in the ANN (Artificial
Neural Network), it must be preprocessed Offset = (0.1 - (F) * xmin)
from its numeric range into the numeric
range that the ANN can deal with efficiently. The input data is preprocessed by using
In this process, proper transformation of data the following scaling function and the output
simplifies the process of learning and may data is again processed using reverse scale.
improve the generalizability of the learned
(5)
results (Kim and Lee, 2004).
F * X + Offset
where X is the data which is to be processed
Figure 2. Three Layer Feed-forward Neural ymax = 0.9 and ymin = 0.1
(7)
Network with One Hidden Layer for Hybrid Model
(Error Forecast through ARIMA to ANN). ymax - ymin
F=
xmax - xmin
Where A is the scaled value which to be
In artificial neural network model at the
hidden and output layer log sigmoid transfer converted in the original numeric value.
(8)
function has been used and its range lies in
[0, 1]. So before providing the data to the A - ymin
val =
F + xmin
network, the data is to be preprocessed or
scaled. Once the input scaled data is processed
through the artificial neural network and the Back propagation algorithm (supervised
output obtained, it is descaled. Methods for learning) is chosen for the training of the
preprocessing and post processing data are neural network, in which finite number of
as described as follows (Merh, Saxena and pairs of input patterns Xk and target output
Pardasani, 2008): patterns Yk in the form (X1, T1), (X2, T2), ...
Merh N., Saxena Vinod P., Raj Pardasani K. - A Comparison between Hybrid Approaches of ANN and ARIMA for Indian Stock Trend Forecasting
28 Business Intelligence Journal July
..., (Xk, Tk) are initialized. Training dataset sigmoid as a neuron signal function is often
for artificial neural network module consists justified by the biological argument that it
of approximately 1218 samples. The dataset represents the behaviour of the firing rate of
used for prediction consist of 30 days closing action potential, averaged over a population
price value. In model I (Figure 1) for of neurons (Kumar, 2004; Koch, 1999).
training open, high, low and close price has More specifically, for the input layer the
of about 1218 sample been provided to the values are as follows:
(9)
model and next day close is provided as the
target. While for the prediction next thirty xik = /un = 0 wuik pik, i = 1, 2, ... ..., n
(10)
days forecasted values have been generated
from the model. In model-II (Figure 2) f (xik) = xik, i = 1, 2, ... ..., n
1218 residuals generated from ARIMA are where xik is i th component of the input vector Xk
provided for training and residual of next and k is the iteration index
day is the target for training. In the current
research work control parameters learning Input for the hidden layer is as follows:
rate, momentum, epochs, error tolerance,
(11)
target and number of hidden layers are
taken randomly. z =/
k
h
n
i=0
k k
w x , h = 1, 2, ... ..., m
ih i
Neural Network Architecture and Output generated by the hidden layer will
Design be as follows:
(12)
k
while for the output layer and the hidden f (y kj ) = 1/ (1 + e ), j = 1, 2, ... ..., p
-y j
for the development of the overall network differentiable processing element function
structure. These factors include learning for the hidden layer PEs would allow chain
rate, momentum factor, initial weight, error rule of partial differentiation to be used to
function, bias function and target output. calculate weight changes for any weight in
For the estimation of above said parameters, the network. This was realized while solving
various strategies are required rather it credit assignment problem (Barto, 1984;
is an art to decide these parameters. The Minsky, 1961; Sinencio and Lau, 1992).
current work without emphasizing on these Multilayer learning (in this case three
strategies assumes learning rate, momentum, layer learning) for the Figure 3 can be
epochs, error tolerance and target randomly. explained as follows (Sinencio and Lau,
1992):
Learning Process The output error across the entire FZ
processing element is found by the following
Back propagation algorithm (supervised cost function:
(19)
learning) is chosen for the training of the
neural network, in which finite number of /= / 1
2
q
j=1 (bkj - z j) 2
pairs of input patterns Xk and target output Following equation computes the output
patterns Tk in the form (X1, T1), (X2, T2), ... of FZ processing element z j :
(20)
..., (Xk, Tk) are initialized.
Instantaneous error Ek for kth learning zj = / p
y Wij
1=1 i
where (22)
TK = (e1k, e2k, ... ..., e kp) T = (t1k - f (y1k), ... ..., where ri = / n
h=1 akh vhi
t kp - f (y kp))
(17)
The weights adjustment is performed
Mk = 1 / pj = 1 (t kj - f (y kj )) 2 by moving along the cost function in the
2
opposite direction of the gradient to a
Mean Square Error on the entire training minimum (minimum is the input /output
set T on a particular neural network will be mapping producing the least amount of
as follows: the total error). The connection weights
(18)
between the FY and FZ processing elements
M = 1 / Qk = 1 (M) k are adjusted and for the two layers the error
Q
correction learning will be explained by the
where M is the mean square error and following equations:
(23)
k is the iteration index. We have a set of Q
training vector pairs. 2E/2wij = 2/2wij 6 21 / qj = 1 (bkj - z j) 2 @
(24)
Merh N., Saxena Vinod P., Raj Pardasani K. - A Comparison between Hybrid Approaches of ANN and ARIMA for Indian Stock Trend Forecasting
30 Business Intelligence Journal July
A positive constant valued learning rate range be 0.1 to .0001, depending upon the
(h) has been added to adjust the amount of application.
change with each move down the gradient.
Adjustments to the connection weights Use of ARIMA in Hybrid
between the FX and FY processing elements Modelling
are derived using the chain rule (Sinencio
and Lau, 1992) The ARIMA model is representative
of linear models and has achieved great
(26)
popularity since the publication of Box
2E/2vih = (2E/2yi) (2yi /2ri) (2ri /2xh) (2xh /2vhi) Jenkins classic book: Time-Series Analysis:
(27)
Forecasting and Control (Box and Jenkins,
= / (bkl - yl) yl whl f' (ri) akh
p
i=1
1976; Zhang, 2001). Different stages of
designing and implementing an ARIMA
A positive constant valued learning rate (p, d, q) based model are identification,
(b) has been added to adjust the amount of estimation, diagnosis and forecasting. If
change with each move down the gradient. seasonal component is also included then the
model becomes ARIMA (p, d, q) (P, D, Q)s
where s is number of periods per session and
P, D, Q are seasonal components. Financial
time series can be stationary or non-
stationary. For converting a time series to a
stationary series it is necessary to difference
the time series d times. ARIMA (p, d, q) is
applying ARMA (p, q) to a stationary series
where p is the order of auto regression,
q is the order of moving-average and d is
the order of differencing. It is also known
Figure 3. Three Layer Neural Network With set of as Box Jenkins (BJ) Method. Training
Inputs and Outputs. dataset ARIMA consists of more than 1218
samples.
From equations 25 and 27 weight
adjustment equation becomes
(28)
Hybrid Model of ANN and ARIMA
(ANN_ARIMA)
wijnew = wijold - h (2E/2wij)
and It may be reasonable to consider a
(29)
time series to be composed of a linear
w hinew = w hiold - b (2E/2vhi) autocorrelation structure and a non-linear
component. In the proposed model there are
Large learning rate can lead to oscillations mainly two stages. In the first stage ANN is
during learning. A refined method of used to forecast the future value of the close
increasing learning rate is to introduce price and then the residual generated is
a momentum term into weight update provided to the ARIMA which will forecast
procedure, where (learning rate) > 0. The the error forecast. In the second stage the
training process will continue until the error predicted close price by ANN is summed
tolerance reaches some specific value let the
Merh N., Saxena Vinod P., Raj Pardasani K. - A Comparison between Hybrid Approaches of ANN and ARIMA for Indian Stock Trend Forecasting
32 Business Intelligence Journal July
Figure 5. Sequence Graph of Residual Generated from ANN Model Indicates that for Indices BSE 30, BSE IT,
BSE Oil& Gas, BSE 100 and S&P CNX Nifty Input Time Series are Stationary.
Values of ACF, PACF, P-Value 0 and best fit was found having minimum
(significance) and LB Statistics (Q*) for Akaike Information criteria (AIC) value.
the maximum lag 10 also indicates that the The best fit was selected after applying
residual series is stationary do not needs any various p and q values the model. Table 5
differencing or transformation using natural shows the best fit model for five cases. To
log (Table 4). It is also found that in case check the adequateness of the model the
of S&P CNX Nifty after taking first order ACF and PACF plots of the error between
differencing results are better. the input series and its fit were generated.
Table 4. ACF, PACF, Significance Value and LB Table 5. Best Fit ARIMA Model for the Given Data
-Statistics Q* Value for all the Indices Taken in The Set.
Study at Ten Degrees of Freedom.
ARIMA
Index p d q
(p, d, q)
LB - Statistic
Significance
(P-Value)
table (Table 6). To determine whether the and ANN is used to model the residuals
time series is white noise or not, the Box- summation of the ARIMA forecasted stock
Ljung Q* statistic is compared with the price and the error forecast generated by
chi-square distribution with (h-m) degrees ANN produces the final forecasted value.
of freedom. Here h is the number of lags Let Xt be the time series for t = 1, 2, 3
and m is the number of parameters. Table which generates Nt as the forecasted series
6 clearly demonstrates that the Q* statistic by implementing ARIMA and the residual
has same distribution as chi-square with (h- (et) will be
(32)
m) degrees of freedom. The plots and the
autocorrelations generated indicate that the et = Xt - Nt where t = 1, 2, 3, 4...
model fits well.
Later ANN is used to model et (residual of
Table 6. LB Statistics and P-Value of the Error or the ANN forecasting), which will generate a
Residual Generated from ARIMA Model for the series of forecast, let it be Pt. Studies (Zhang,
Verification of White Noise. 2003; Sallehuddin, et al., 2008) indicate
that, this step can be said as a process of
error generation of time series prediction
LB - Statistics Q*
Is Series White
(h-m) Degrees
Significance
of Freedom
Noise
Index
multivariate time series forecasting for
ARIMA model (Sallehuddin, et al., 2008).
Final forecasting Xp will be as follows:
BSE 30(SENSEX) (10-1)=9 10.209 0.334 Yes
BSEIT (10-1)=9 10.438 0.316 Yes (33)
BSE Oil & Gas (10-2)=8 5.234 0.732 Yes X p = Nt + Pt
BSE 100 (10-1)=9 6.240 0.716 Yes
S&P CNX Nifty (10-2)=8 2.930 0.939 Yes Framework of the proposed model is
shown in figure 6.
Forecasting Using Hybrid ANN As seen in the Figure 6 the first step is
and ARIMA the input time series selection. In the present
model five data sets BSE 30, BSE IT, BSE
Figure 4 outlines the forecasting Oil & Gas, BSE 100 and S&P CNX Nifty
procedure. The above mentioned ARIMA each has been used. In the next step the
models are not used for forecasting purpose. time series is used for model identification
The fit of the input (residual of ANN) series and the 30 samples of hold out data for
generated by the ARIMA models is summed forecasting through ARIMA. This forecasts
with the predicted close values generated the next thirty days closing price of SENSEX
by ANN. This results in the final predicted from February 26, 2009 to April 16, 2009.
value. Residual generated by ARIMA is provided
to ANN for residual analysis. Finally the
Hybrid Model of ARIMA and ANN residual predicted is added to the ARIMA
(ARIMA_ANN) forecasts thus giving the final forecasted
values.
The model has same components as the
ANN_ARIMA model. The difference is that
here ARIMA is used to forecast the values
Merh N., Saxena Vinod P., Raj Pardasani K. - A Comparison between Hybrid Approaches of ANN and ARIMA for Indian Stock Trend Forecasting
34 Business Intelligence Journal July
differencing it was non-stationary (Figure 7 BSE 30(SENSEX) 0.018 -0.007 0.191 321.255
(a), Figure 7 (b), Figure 7 (c), Figure 7 (d)). BSEIT 0.005 -0.011 0.353 308.625
If the series is transformed only by taking BSE 100 0.022 -0.004 0.114 329.811
differencing or natural logarithm then also S&P CNX Nifty 0.021 -0.011 0.222 318.425
series is not becoming stationary. After BSE Oil & Gas is not fitting in ARIMA modelling.
Figure 7. (a) Sequence Graph of BSE 30 Close Price from April 16, 2004 to April 16, 2009. Figure 7 (b)
Sequence Graph of BSE IT Close Price from April 16, 2004 to April 16, 2009.
Figure 7. (c) Sequence Graph of BSE 100 Close Price from April 16, 2004 to April 16, 2009. Figure 7 (d)
Sequence Graph of S&P CNX Nifty Close Price from April 16, 2004 to April 16, 2009.
Figure 8. Sequence Graph of Transformed Input Time Series for Indices BSE 30, BSE IT, BSE Oil& Gas, BSE
100 and S&P CNX Nifty. Series is Transformed Using Natural Logarithm and Differencing of First Order.
After analyzing ACF, PACF and AIC value. The best fit was selected after
performing various iterations by taking applying various p and q values the model.
different p and q parameter values with d as Table 8 shows the best fit model for five
0 and best fit was found having minimum cases. To check the adequateness of the
Merh N., Saxena Vinod P., Raj Pardasani K. - A Comparison between Hybrid Approaches of ANN and ARIMA for Indian Stock Trend Forecasting
36 Business Intelligence Journal July
model the ACF and PACF plots of the error Residual derived from suitable ARIMA
between the input series and its fit were model is supplied to ANN 5-5-1 model for
generated. residual analysis and residual is forecasted
as shown in Figure 2. Control parameters
Table 8. Best Fit ARIMA Model for the Given Data for models are explained in Table 3. Various
Set. errors derived are explained in table 10.
ARIMA
Index p d q
(p, d, q)
Results and Simulation
BSE 30(SENSEX) 1 1 1 ARIMA(1,1,1)
Comparisons
BSEIT 0 1 2 ARIMA(0,1,2)
BSE 100 1 1 1 ARIMA(1,1,1)
Comparisons and performance of above
S&P CNX Nifty 1 1 1 ARIMA(1,1,1) models have been evaluated by calculating
BSE Oil & Gas is not fitting in ARIMA modelling.
errors between the actual close price and the
In the process of diagnosis it is verified predicted close price generated by all the
whether the residual or error generated if models (Hybrid ARIMA_ANN and Hybrid
white noise or not. The autocorrelations ANN_ARIMA). In the study average
specifying Q*-value, P-value (significance) absolute error (AAE), root mean square
at different degrees of freedom having a error (RMSE), mean absolute percentage
lag of maximum 300 are shown in the table error (MAPE)and mean square percent
(Table 9). To determine whether the time error MSPE) have been calculated. These
series is white noise or not, the Box-Ljung measures help in calculating and comparing
Q* statistic is compared with the chi-square accuracy of different techniques. It will
distribution with (h-m) degrees of freedom. not only help in measuring a particular
Here h is the number of lags and m is the techniques reliability, usefulness but also
number of parameters. Table 10 clearly help in searching the optimal technique
demonstrates that the Q* statistic has same (Hanke and Wichern, 2007).Figure 9
distribution as chi-square with (h-m) degrees displays the comparison between the actual
of freedom. Plots and the autocorrelations close of BSE 30 (SENSEX), BSE IT, BSE
generated indicated that the model fits well. Oil & Gas, BSE 100 and S&P CNX Nifty
forecasted close price generated by the
Table 9. LB Statistics and P-Value of the Error or models discussed above during Feb 26,
Residual Generated from ARIMA Model for the 2009 to April 16, 2009.
Verification of White Noise. Table 10. Comparison between the Errors of Hybrid
Model of ANN_ARIMA and Hybrid ARIMA_ANN
LB-Statistics Q*
Is Series White
(h-m) Degrees
Models.
Significance
of freedom
(P-Value)
Noise
Errors
Model
Index
Index
AAE RMSE MAPE MSPE
BSE30
(SENSEX)
77.8306 102.1076 0.8200 0.011496
BSE 30
(300-2)=298 289.956 0.620 Yes
(SENSEX) BSEIT 5.8395 7.1693 0.2591 0.001001
ANN_ARIMA
Errors
Model
Index
AAE RMSE MAPE MSPE
BSE30
(SENSEX)
164.0544 198..9514 1.7619 0.0451531
ARIMA_ANN
Merh N., Saxena Vinod P., Raj Pardasani K. - A Comparison between Hybrid Approaches of ANN and ARIMA for Indian Stock Trend Forecasting
38 Business Intelligence Journal July
Conclusion References
In the current research an attempt was made Abu- Mostafa, Y. S., et al. (2001). Neural
to study whether a hybrid model achieves Networks in Financial Engineering,
better results than an individual mode? IEEE Transactions on Neural Networks,
Two hybrid models were developed using 12(4): 653-656.
techniques of ANN and ARIMA comparison
is also made between the conventional hybrid Afolabi, Mark O. and Olude, Olatoyosi
approach using ARIMA, ANN and proposed (2007). Predicting Stock Prices Using a
hybrid approach using ANN, ARIMA. A Hybrid Kohonen Self Organizing Map
hybrid model has been developed using (SOM), Proceeding of 40th International
techniques of back propagation artificial Conference on Systems Sciences, 1560-
neural networks (ANN) and ARIMA for 1605.
analyzing BSE 30 (SENSEX), BSE IT, BSE
Oil & Gas, BSE 100 and S&P CNX Nifty Barto, A., (1984). Simulation Experiments
for forecasting the next days closing price with Goal-Seeking Adaptive Elements,
and predicting the stock trend. In the current Air Force Wright Aeronautical
study, BSE 30 (SENSEX), BSE IT, BSE Oil Laboratory, Technical Report, AFWAL-
& Gas, BSE 100 and S&P CNX Nifty open, TR-84-1022.
high, low, and close prices have been taken
as the input data set and the output generated Box, G.E.P. and Jenkins, G.M., (1976). Time
is the predicted close price of the same. Series Analysis: Forecasting and Control.
Results show that prediction of San Francisco, CA: Holden-Day.
hybrid ANN_ARIMA model for BSE 30
(SENSEX), BSE IT and S&P CNX Nifty are Brabazon, T. (2000). A Connectivist
better than hybrid ARIMA_ANN whereas Approach to Index Modelling in
in the case of BSE 100 results of ARIMA_ Financing Markets, Proceedings of Coil
ANN are marginally better then hybrid / EvoNet Summer School. University of
ANN_ARIMA. Hybrid Model ARIMA_ Limerick, Briys, E., Bellalah, M., Mai, H.
ANN(Error Forecast through ARIMA to M. and Varenne, F. de (1998). Options,
ANN) exhibited over-fitting problem. Futures and Exotic Derivatives. England:
Various trials were done by modifying the John Wiley & Sons Ltd.
learning rate and momentum. Finally the
learning rate was reduced and momentum Fama, E.F. (1965). The Behaviour of Stock
changed accordingly to resolve this problem. Market Prices, Journal of Business, 38:
Results also demonstrate that for BSE Oil 34-105.
& Gas ANN and hybrid ANN_ARIMA are
able to manage the input data set and predict Fama, E.F. (1970). Efficient Capital Markets:
the future closing price, whereas ARIMA A Review of Theory and Empirical Work,
and ARIMA_ANN fails to forecast future Journal of Finance, 25: 1465-1468.
values.
Hanke, John E. and Wichern, Dean W.
(2007). Business Forecasting. New
Delhi: Pearson Education, Inc, 79-80.
Kate A., Simth, Jatinder, N. D, Gupta Minsky, M. (1961). Steps Towards AI,
(2000). Neural Networks in Business: Proceeding of the IRE, 49: 5-30.
Techniques and Applications for the
Operations Researcher, Computer and Merh, Nitin, V.P. Saxena and Kamal Raj
Operations Research, 27: 1023-1044. Pardasani (2008), Artificial Neural
Network for Stock Market Forecasting,
Kim, Kyoung-jae and Lee, Won Boo (2004). Nirma University Journal of Business
Stock Market prediction using Artificial and Management Studies (NUJBMS),
Neural Networks with Optimal Feature 2(3 & 4): 3-19.
Transformation, Neural Computing and
Application, 13: 255260. Murphy, J. J. (1999). Technical Analysis of
the Financial Markets: A Comprehensive
Koch, C., (1999), Biophysics of Guide to Trading Methods and
Computation: Information Processing in Applications. New York: Institute of
Single Neuron, Oxford university Press, Finance, 4-5.
New York.
Priestley, M. B. (1988). Non-linear and Non-
Kumar, Satish (2004), Neural Networks: stationary Time Series Analysis. London:
A Classroom Approach, Tata McGraw- Academic Press.
Hill Publishing Company Limited, New
Delhi. Refenes, A. P., (Ed) (1995). Neural Network
in the Capital Markets. England:John
Lawrence, S., Tsoi, A. C., and Giles, C. L. Wiley & Sons Ltd, 149-161.
(1996). Noisy Time Series Prediction
Using Symbolic Representation and Sallehuddin, Roselina., Shamsuddin, Siti
Recurrent Neural Network Grammatical Mariyam. and Hashim, Siti Zaiton Mohd
Inference. Technical report UMIACS- (2008). Hybridization Model of Linear
TR-96-27 and CS-TR-3625. Institute for and Non-linear Time Series Data for
Advanced Computer Studies, University Forecasting. Second Asia International
of Maryland. Conference on Modelling & Simulation,
IEEE Conference Proceedings, DOI
Malkiel, B.G. (1996). A Random Walk 10.1109/AMS.2008.142, 597 602.
Down Wall Street. New York: W. W.
Norton, 6th Edition. Samanta G.P. and Bordoloi, Sanjib (2005).
Predicting Stock Market- An Application
Miao, K., Chen, F., and Zhao, Z. G. (2007). of Artificial Neural Network Technique
Stock Price Forecast Based on Bacterial Through Genetic Algorithm, Finance
Colony RBF Neural Network, Journal of India, 19(1): 173-188.
QingDao University, 20 (2): 5054 (in
Chinese). Sinencio, Edgar Snchez and Lau, Clifford
(1992). Artificial Neural Networks:
Merh N., Saxena Vinod P., Raj Pardasani K. - A Comparison between Hybrid Approaches of ANN and ARIMA for Indian Stock Trend Forecasting
40 Business Intelligence Journal July
Paradigms, Applications and Hardware Yao, J.T., and Poh, H. L. (1996). Equity
Implementations. New York: IEEE Press, Forecasting: A Case Study on the KLSE
16-17. Index. Neural Networks in Financial
Engineering, Proceedings of the Third
Skabar, Andrew and Cloete, Ian (2001). International Conference on Neural
Neural Networks, Financial Trading Networks in the Capital Markets. World
and The Efficient Markets Hypothesis, Scientific, Singapore, 341353.
Australian Computer Society, Inc. ,
Twenty-Fifth Australasian Computer Yudong, Zhang and Lenan, Wu (2009).
Science Conference (ACSC2002), Stock Market Prediction of S&P 500 via
Melbourne, Australia, Conferences in Combination of Improved BCO Approach
Research and Practice in Information and BP Neural Network, Expert Systems
Technology, 4: 241 249. with Applications, 36: 88498854.
Termizel, Taskaya T., and Casey, M.C. Zhang, Defu., Jiang, Qingshan and li, Xin
(2005). A Comparative Study of (2004). Application of Neural Networks
Autoregressive Neural Network Hybrids, in Financial Data Mining, International
Neural Networks, 18( 5-6): 781-789. Journal of Computational Intelligence,
1(2): 116-119.
Tseng, Fang-Mei., Yu, Hsiao-Cheng
and Tzeng, Gwo-Hsiung (2002). Zhang, G. Peter (2003). Time Series
Combining Neural Network Model with Forecasting Using a Hybrid ARIMA and
Seasonal Time Series ARIMA model, Neural Network Model, Neurocomputing,
Technological Forecasting & Social 50: 159 175.
Change, 69(1): 7187.
Zhang, Guoqiang Peter (2001). An
Windrow, B, D Rumelhart, E. and Lehr, M.A Investigation of Neural Networks
(1994). Neural Networks Applications for Linear Time-Series Forecasting,
in Industry, Business and Science, Computers and Operations Research, 28:
Communications of the ACM, 37(3): 93- 1183-1202.
105.
Annexure
Table 11. Comparison between the Actual and Table 12. Comparison between the Actual and
Predicted Values Generated by ANN, ARIMA, Predicted Values Generated by ANN, ARIMA,
Hybrid ANN_ARIMA and Hybrid ARIMA_ANN Hybrid ANN_ARIMA and Hybrid ARIMA_ANN
Models for BSE 30 (SENSEX). Models for BSE IT.
BSE 30 (SENSEX)
Actual Values of
Using Following Model Using Following Models
BSE IT
Date
Date
ARIMA_
ARIMA_
Hybrid
Hybrid
Hybrid
Hybrid
ARIMA
ARIMA
ARIMA
ARIMA
ANN_
ANN_
ANN
ANN
ANN
ANN
26-FEB-2009 8954.90 9001.30 8875.57 8954.90 8940.26 26-FEB-2009 2110.70 2092.10 2082.62 2110.70 2132.37
27-FEB-2009 8891.60 9070.80 8908.93 8872.95 9003.59 27-FEB-2009 2096.20 2107.70 2094.06 2098.99 2136.73
02-MAR-2009 8607.10 9033.20 8875.15 8535.06 8873.20 02-MAR-2009 2057.60 2090.00 2088.89 2055.41 2099.85
03-MAR-2009 8427.30 8653.80 8558.21 8256.00 8305.54 03-MAR-2009 2019.20 2063.80 2062.57 2014.56 2046.14
04-MAR-2009 8446.50 8403.70 8441.90 8355.44 8342.52 04-MAR-2009 2032.80 2042.70 2028.77 2026.67 2000.29
05-MAR-2009 8197.90 8388.60 8338.91 8215.11 8358.90 05-MAR-2009 2011.00 2051.10 2023.89 2010.42 2042.54
06-MAR-2009 8325.80 8199.10 8304.31 8249.14 8178.69 06-MAR-2009 2072.30 2035.40 2039.78 2066.23 2041.72
09-MAR-2009 8160.40 8128.70 8270.04 8211.33 8306.92 09-MAR-2009 2025.30 2073.00 2050.24 2031.89 2084.09
12-MAR-2009 8343.80 8095.30 8228.51 8356.54 8134.24 12-MAR-2009 2073.50 2046.40 2045.77 2065.17 2035.71
13-MAR-2009 8756.60 8217.80 8415.51 8856.50 8546.12 13-MAR-2009 2190.30 2079.40 2090.56 2195.74 2122.25
16-MAR-2009 8943.50 8663.90 8769.14 9160.10 9125.61 16-MAR-2009 2199.20 2186.20 2166.94 2215.38 2281.13
17-MAR-2009 8863.80 9019.10 8938.47 8976.20 9128.24 17-MAR-2009 2161.30 2197.30 2181.95 2160.81 2228.77
18-MAR-2009 8976.70 8960.20 8922.82 8914.27 8863.54 18-MAR-2009 2184.80 2154.40 2176.88 2179.35 2165.62
19-MAR-2009 9001.70 9060.60 8926.35 9008.33 8995.96 19-MAR-2009 2219.20 2181.80 2191.26 2224.70 2213.85
20-MAR-2009 8966.70 9145.20 9018.28 8943.02 9109.10 20-MAR-2009 2217.40 2226.60 2209.45 2222.30 2252.05
23-MAR-2009 9424.00 9103.50 9070.86 9352.24 9034.63 23-MAR-2009 2271.60 2227.70 2232.91 2269.44 2255.47
24-MAR-2009 9471.00 9534.10 9420.80 9599.84 9789.33 24-MAR-2009 2263.00 2287.00 2250.89 2270.07 2300.15
25-MAR-2009 9667.90 9619.70 9536.24 9642.53 9601.84 25-MAR-2009 2286.00 2257.30 2280.31 2281.23 2307.30
26-MAR-2009 10003.10 9832.50 9742.85 10022.48 9889.88 26-MAR-2009 2361.20 2303.70 2297.32 2366.34 2317.39
27-MAR-2009 10048.50 10116.90 9963.24 10117.08 10238.85 27-MAR-2009 2337.80 2373.80 2325.16 2345.84 2403.54
30-MAR-2009 9568.10 10202.60 9891.09 9540.60 9991.71 30-MAR-2009 2234.30 2344.70 2314.72 2227.54 2342.29
31-MAR-2009 9708.50 9748.20 9663.57 9453.42 9356.06 31-MAR-2009 2285.70 2221.10 2270.98 2269.79 2205.26
01-APR-2009 9902.00 9870.90 9686.73 9886.04 9791.43 01-APR-2009 2358.10 2300.70 2290.72 2370.45 2319.84
02-APR-2009 10348.80 10103.00 10102.59 10361.30 10333.22 02-APR-2009 2454.70 2364.60 2389.35 2461.61 2470.73
06-APR-2009 10534.90 10369.70 10329.97 10633.71 10591.58 06-APR-2009 2454.70 2450.60 2416.18 2467.47 2482.51
08-APR-2009 10742.30 10513.10 10614.78 10808.71 10835.05 08-APR-2009 2498.40 2436.00 2474.98 2497.07 2516.52
09-APR-2009 10803.90 10612.20 10668.07 10896.04 10811.00 09-APR-2009 2499.30 2470.40 2469.76 2509.08 2508.05
13-APR-2009 10967.20 10669.70 10907.09 11044.27 11058.25 13-APR-2009 2487.00 2469.20 2488.09 2489.93 2532.51
15-APR-2009 11284.70 10728.30 11063.38 11404.30 11138.88 15-APR-2009 2444.60 2453.50 2486.54 2446.88 2500.39
16-APR-2009 10947.40 10825.60 11058.48 11171.08 11295.20 16-APR-2009 2391.80 2435.00 2450.64 2390.08 2423.57
Merh N., Saxena Vinod P., Raj Pardasani K. - A Comparison between Hybrid Approaches of ANN and ARIMA for Indian Stock Trend Forecasting
42 Business Intelligence Journal July
Table 13. Comparison between the Actual and Table 14. Comparison between the Actual and
Predicted Values Generated by ANN and Hybrid Predicted Values Generated by ANN, ARIMA,
ANN_ARIMA Models for BSE Oil & Gas. Hybrid ANN_ARIMA and Hybrid ARIMA_ANN
Models for BSE 100.
Actual Values of
BSE Oil & Gas
BSE 100
Models
Date
Date
ARIMA_
Hybrid
Hybrid
ARIMA
ARIMA
ANN_
ANN
ANN
ANN Hybrid ANN_ARIMA
26-FEB-2009 6165.30 6053.20 6165.30 26-FEB-2009 4537.10 4390.20 4513.22 4537.10 4547.03
27-FEB-2009 6064.10 6137.10 6123.74 27-FEB-2009 4516.40 4385.90 4519.83 4612.21 4545.90
02-MAR-2009 5886.00 5979.70 5896.71 02-MAR-2009 4377.20 4374.40 4506.72 4462.31 4505.59
03-MAR-2009 5775.60 5867.90 5756.51 03-MAR-2009 4293.50 4297.20 4355.14 4295.33 4244.34
04-MAR-2009 5842.20 5781.00 5805.64 04-MAR-2009 4311.00 4252.50 4290.44 4308.59 4243.66
05-MAR-2009 5625.30 5842.50 5611.64 05-MAR-2009 4193.50 4244.80 4265.99 4231.65 4289.49
06-MAR-2009 5724.10 5694.90 5667.44 06-MAR-2009 4242.20 4222.50 4240.53 4208.74 4179.18
09-MAR-2009 5633.90 5806.60 5596.39 09-MAR-2009 4160.40 4198.80 4209.41 4173.25 4212.41
12-MAR-2009 5814.60 5719.40 5750.73 12-MAR-2009 4246.80 4191.40 4182.86 4221.76 4150.97
13-MAR-2009 6088.50 5846.80 6058.02 13-MAR-2009 4437.00 4215.80 4289.33 4473.13 4354.95
16-MAR-2009 6298.30 6096.20 6322.54 16-MAR-2009 4541.40 4292.60 4455.78 4685.67 4605.23
17-MAR-2009 6206.60 6323.90 6265.31 17-MAR-2009 4504.60 4368.60 4539.99 4666.87 4626.68
18-MAR-2009 6308.50 6133.80 6331.08 18-MAR-2009 4562.40 4385.00 4529.72 4651.10 4505.22
19-MAR-2009 6355.20 6318.10 6407.08 19-MAR-2009 4575.40 4426.40 4535.12 4691.10 4570.97
20-MAR-2009 6368.00 6313.40 6415.65 20-MAR-2009 4558.60 4434.60 4594.27 4655.78 4637.39
23-MAR-2009 6775.40 6353.10 6823.19 23-MAR-2009 4775.60 4410.30 4600.63 4856.47 4569.78
24-MAR-2009 6772.80 6848.90 6892.79 24-MAR-2009 4784.80 4580.40 4779.09 5023.05 4955.80
25-MAR-2009 7020.90 6751.70 7099.39 25-MAR-2009 4886.00 4703.90 4818.57 5019.31 4830.35
26-MAR-2009 7190.70 6982.90 7304.55 26-MAR-2009 5036.40 4749.40 4921.18 5155.17 4989.71
27-MAR-2009 7181.80 7049.10 7311.15 27-MAR-2009 5091.60 4969.10 5030.01 5278.78 5147.70
30-MAR-2009 6992.40 7087.20 7119.09 30-MAR-2009 4861.00 5065.70 5030.27 4940.90 5094.32
31-MAR-2009 7053.00 6919.90 7133.05 31-MAR-2009 4942.50 4841.20 4891.13 4808.99 4740.67
01-APR-2009 7256.00 6977.40 7344.40 01-APR-2009 5028.40 4859.90 4942.23 5094.47 4996.14
02-APR-2009 7671.80 7074.20 7795.20 02-APR-2009 5253.10 4953.00 5117.91 5363.00 5206.57
06-APR-2009 7747.60 7609.40 7960.75 06-APR-2009 5348.20 5243.00 5243.84 5543.93 5379.26
08-APR-2009 7998.80 7757.70 8191.81 08-APR-2009 5467.30 5369.50 5436.97 5535.91 5542.77
09-APR-2009 7978.10 7672.40 8175.60 09-APR-2009 5500.80 5381.90 5385.92 5564.59 5418.97
13-APR-2009 8083.20 7780.20 8296.84 13-APR-2009 5593.10 5484.30 5575.56 5670.65 5692.91
15-APR-2009 8273.40 7787.80 8499.08 15-APR-2009 5780.00 5525.50 5658.77 5850.96 5681.28
16-APR-2009 7901.40 7795.90 8172.23 16-APR-2009 5588.10 5566.30 5644.33 5754.09 5767.80
ARIMA_
Hybrid
Hybrid
ARIMA
ARIMA
ANN_
ANN
ANN
26-FEB-2009 2785.70 2750.90 2763.00 2785.70 2755.55
27-FEB-2009 2763.70 2763.60 2750.75 2798.50 2775.04
02-MAR-2009 2674.60 2753.90 2732.20 2684.40 2753.26
03-MAR-2009 2622.40 2715.00 2672.34 2577.57 2624.76
04-MAR-2009 2645.20 2681.50 2649.49 2585.52 2608.59
05-MAR-2009 2576.70 2677.90 2619.97 2545.20 2623.93
06-MAR-2009 2620.20 2665.70 2591.73 2550.42 2558.31
09-MAR-2009 2573.20 2662.30 2595.33 2530.72 2624.30
12-MAR-2009 2617.40 2660.30 2619.95 2550.34 2607.91
13-MAR-2009 2719.30 2666.10 2663.82 2674.98 2664.47
16-MAR-2009 2777.30 2698.20 2724.20 2785.15 2781.09
17-MAR-2009 2757.50 2745.60 2759.13 2783.97 2812.85
18-MAR-2009 2794.70 2759.20 2787.89 2790.69 2795.22
19-MAR-2009 2807.20 2775.90 2776.86 2817.22 2792.45
20-MAR-2009 2807.10 2789.70 2787.23 2817.32 2823.17
23-MAR-2009 2939.90 2791.10 2890.55 2944.78 2914.73
24-MAR-2009 2938.70 2881.10 2959.90 3013.36 3010.07
25-MAR-2009 2984.40 2958.10 2932.74 3018.16 2921.73
26-MAR-2009 3082.30 2984.10 3036.67 3101.88 3088.70
27-MAR-2009 3108.60 3087.60 3067.77 3166.06 3119.52
30-MAR-2009 2978.20 3147.50 3024.81 2999.19 3066.33
31-MAR-2009 3021.00 3061.20 3018.87 2943.59 2982.25
01-APR-2009 3060.40 3041.10 3011.76 3041.89 3020.31
02-APR-2009 3211.10 3074.30 3170.05 3221.46 3221.02
06-APR-2009 3256.60 3203.60 3249.85 3328.29 3291.46
08-APR-2009 3342.90 3276.90 3237.04 3376.84 3252.16
09-APR-2009 3342.10 3296.40 3365.83 3384.32 3472.01
13-APR-2009 3382.60 3324.20 3378.66 3416.34 3365.20
15-APR-2009 3484.20 3332.80 3406.00 3525.46 3412.40
16-APR-2009 3369.50 3348.10 3432.62 3460.00 3512.79
Merh N., Saxena Vinod P., Raj Pardasani K. - A Comparison between Hybrid Approaches of ANN and ARIMA for Indian Stock Trend Forecasting
44 Business Intelligence Journal July
Abstract
One firm valuation method is to use discounted cash flow. In this paper the valuation method of Capital
Cash Flow discounted at the Weighted Average Cost of Capital (WACC) before tax is represented and, as
a proof to its efficiency in Iran market, it is compared with two common methods, i.e. Free Cash Flow
discounted at the weighted average cost of capital after tax and the adjusted present value. For the same
purpose, 54 firms from among those companies listed on Tehran Stock Exchange were selected as sample
and their financial information for three-year financial period from 2004 to 2006 were collected and
analyzed by paired Students t-test. Research results showed that using an appropriate discount rate will
make the value calculated by Capital Cash Flow method become twice as much when using two methods
mentioned above.
Dastgir M., Khodadadi V., Ghayed M. - Cash Flows Valuation Using Capital Cash Flow Method Comparing it with Free Cash Flow Method and
Adjusted Present Value Method in Companies Listed on Tehran Stock Exchange
46 Business Intelligence Journal July
Dastgir M., Khodadadi V., Ghayed M. - Cash Flows Valuation Using Capital Cash Flow Method Comparing it with Free Cash Flow Method and
Adjusted Present Value Method in Companies Listed on Tehran Stock Exchange
48 Business Intelligence Journal July
1. Discounting operating Free Cash Flow tax shield which is very far from its correct
at the weighted average cost of capital value.
In a survey, Fernandez took measures
2. Discounting equity Free Cash Flow at for firms valuation through four methods
the cost of equity of: Discounting operating Free Cash Flow
at the weighted average cost of capital;
3. Adjusted present value discounting equity Free Cash Flow at the
cost of equity; discounting the Capital
4. Discounting the Capital Cash Flow at Cash Flow at the unlevered cost of capital;
the unlevered cost of capital Adjusted present value.
The research results indicated that firms
All above methods are based on a similar valuation with using above four methods
theory: levered firms value is equal to will give similar responses. This conclusion
unlevered firms value plus present value of is logic as all of these methods analyze
tax shield arising from financing due to debt. a similar fact with similar assumptions.
However, none of above methods can be These methods difference in institutions
defined without discount rate for valuation. valuations is resulted from difference in tax
These discount rates depend upon firms shield calculation. Hence, their difference
leverage policy. The firms leverage policy in term of various methods of tax shield
determines that what discount rate should calculation is presented below:
be applied for the proper estimation of tax
shield value and, consequently, firms value. Value of Tax Shield (VTS)
Therefore, all valuation methods should
start from a clear assumption about leverage Valuable researches have been conducted
policy of firm. about cash flows valuation. Discrepancy
In this survey, different hypotheses among these methods and different theories
about levered firms policy and that how on firms valuation using discounted cash
they are affected by following factors are flows is arising from difference in the
investigated: calculation of tax shield. Interest paid on
debt is a cost subtracted from profit but no
1. Applicable discount rate tax is deducted for it, while tax is received
on dividend or the accumulated profit
2. Present value of tax shield dependent upon tax share. Thus, in presence
of debt, total payments to the debt holders
3. Applicability of above methods and shareholders will be higher.
In this paper, different theories are
Then, different valuation methods and presented for calculating present value of tax
the way to select from among them are shield. Upon analysis of the results obtained
explained. Lastly, through model presented from these theories it will be proved that
by Booth (2002) and Fernandez (2004) it VTS is not the present value of the tax shield
was shown that how using wrong methods discounted at a certain rate, rather it is the
will lead to mistakes which despite of their difference between two present values:
being small, they will have major impacts present value of taxed paid in the unlevered
on calculations. Using wrong formulas can company minus the present value of the
result in an estimation of the present value of taxes paid in the levered company. The
Dastgir M., Khodadadi V., Ghayed M. - Cash Flows Valuation Using Capital Cash Flow Method Comparing it with Free Cash Flow Method and
Adjusted Present Value Method in Companies Listed on Tehran Stock Exchange
50 Business Intelligence Journal July
problem
D
in their
E
paper: for the calculation of reduce their taxes through other tools (e.g.
and E + D they used book value of depreciation).
t-1 t-1
E +D
t-1 t-1 t-1 t-1
Miller argues that the optimum debt not use a valuation method similar to that
structure of collective companies indicates used by a firm with default debt.
the existence of such a structure for each In this respect, in firms with target debt
of them individually. Thereafter, Miller ratio the Free Cash Flow will be discounted
introduces personal income tax and at the following rate:
corporate income tax. According to Miller,
WACC = Ku - 8
D+EB
; E
D Kd T (1 + Ku)
firms value when no debt is assumed is 1 + Kd
equal to:
Et - 1 + Dt - 1 = FCFt + Kd TDt - 1
Ku - g Ku - g
Vu = FCF (1 - TPA)
Ku
(10) Miles & Ezell [1985]
Then he adds that firms attempts
made for making increase in their debts is The following formula shows the
inconsistent with market balance. Increase relationship between the levered beta (L)
in debts causes changes in rate of return with the asset beta (u) (assuming a risk-
on debt and rate of return to equity and, free debt and a debt beta of zero):
thus, firms value under such conditions is
independent of rate of debts. bL = bU + DbU ;1 - TRF E /E
1 + RF
(7) Miller & Scholes [1978] (11) Chambers, Harris & Pringle
[1982]
When the rate of income tax is higher
than rate of income on capital profit, most This group of researchers compared
investors will have to pay taxes higher than four valuation methods for discounted cash
their receivable dividend. Miller and Scholes flows: equity cash flow discounted at the
conclude that if corporate repurchases its required return to levered equity (Ke), Free
shares, it will make no preference between Cash Flow discounted at WACC, Capital
dividend and realized gain on capital. Cash Flow discounted at WACC before tax
According to these researchers, firm value is and Adjusted present value methods. They
independent of the policy of firms dividend argue that in case of target debt the first three
payment. methods will give similar results. But, when
there is no target debt, these methods will
(8) De Angelo and Masulis [1980] result in different values. Only the Adjusted
present value will give the same results as
De Angelo and Masulis extended Millers other three methods under both conditions.
study. With considering that the tax final Of course, firms are simply analyzed for one
rate varies for different firms, they predict financial period. The reason of such result
that firms, instead of using debt, try to for their researches was a mistake: they had
Dastgir M., Khodadadi V., Ghayed M. - Cash Flows Valuation Using Capital Cash Flow Method Comparing it with Free Cash Flow Method and
Adjusted Present Value Method in Companies Listed on Tehran Stock Exchange
52 Business Intelligence Journal July
(18) Inselbag and Kaufold [1997] And, this value will be lower when levered
costs exist.
They believe that in a case where
monetary value of debt is fixed, the value Research Design
of tax shield (VTS) may be given by Myers
method. If the firm has a constant debt/value The objective of this research is to
ratio, the value of tax shield (VTS) will be introduce Capital Cash Flow and to compare
calculated through Miles and Ezells. it with two methods of Free Cash Flow and
According to these researchers, present Adjusted present value. Therefore, research
value of tax shield of firms planning based variables are presented as follows:
on amount of nominal debt is higher than
firms emphasizing on debt ratio. This theory Studied Variables
cannot be accepted for two reasons: firstly,
no firm has output operating cash flow for Three studied variables including Capital
having target debt ratio (instead of a target Cash Flow, Free Cash Flow and adjusted
debt outstanding). And, secondly, as we present value are introduced as follows:
know, the tax shield is the difference between
two present values of taxes in levered and Capital Cash Flow
unlevered firms.
Inselbag and Kaufold argue that the risk of Capital Cash Flow includes all cash flows
target debt ratio is higher than that of having paid or payable to investors. In this method,
target debt outstanding. If so, the present Capital Cash Flows are the cash flows
value of taxes paid by levered firms should available for all holders of the companys
be higher than that of firms with target debt securities equivalent to the equity cash flow
and, in consequence, the present value of tax after deduction of companys assets tax.
shield for the second-group firms should be Capital Cash Flow is calculated as
lower and this is opposed to the theory cited follows:
by above-mentioned researchers. Capital Cash Flow = Net Income +
Depreciation - Capital expenditures ! 9
(19) Copeland, Koller and Murrin
Working Capital + Interest
[2000]
Net profit contains tax savings due to
The studies of this group of researches interest expense of debt. Then, the impact of
confirmed the theories of Harris & Pringle cash flow adjustments including depreciation
(1985) and Myers (1974) relating to present expense, capital expenditures and capital
value of tax shield and they concluded that turnover are taken into account. At last,
we leave it to the readers judgment to upon addition of cash interest, the Capital
decide which approach best fits his or her Cash Flow is obtained which indicates the
situation. after-tax cash flow for investors. For the
same reason, it is discounted at the before-
(20) Fernandez [2001] tax weighted average cost of capital which
is calculated as follows:
This researcher believes that where the
levered costs are zero, the value of tax shield WACCAT = E Ke + D Kd
V V
is equal to tax rate multiplied by debt value.
Where D/V is the debt-to-value ratio; E/V present value. This method was introduced
is the equity-to-value ratio, and KD and KE by Mayers and separates firms value into
are the respective expected debt and equity two parts: unlevered operating cash flows
returns. and cash flows depending upon project
On the other hand, capital assets financing.
pricing model (CAPM) should be used for Where:
calculating the required returns.
APV = / tn= 0 FCF + / tn= 0 Int. t (Tc) t
1 + Kg (1 - Kd)
Ke = RF + be RP
Int. t = expense interest of debt in time t
Ke = RF + be RP
Tc = effective income tax rate
Where Rf is the risk-free rate, Rp is the Other components of this equation were
risk premium, Kd and Ke are the debt and represented before.
equity betas, respectively. In this model, firms value is initially
calculated with using firms capital expense
Free Cash Flow in the absence of debt and, then, the present
value of financing tax savings due to debt
The basic assumption in determination is added to it. The reason of this separation
of firm value with using Free Cash Flow is is to allow using different discount rates
that a trading units value is resulted from depending upon risk rate for two parts.
its power in making the operating cash flow
and other cash flows due to investment. This Methodology of Research
is calculated as follows:
Considering the structure of hypotheses
Free Cash Flow = earnings before interest
and the method used for data finding and
and taxes + estimated taxes + collection within the finite time of research,
cash adjustments statistical population was selected for three
consecutive years from 2004 to 2006 from
Since tax shield arising from interest among companies listed on Tehran Stock
expense of debt has not been taken into Exchange.
account in the calculation of Free Cash Cochrans (1977) sample size formula1
Flow, it is necessary to use the weighted was used for determining sample size
average cost of capital after tax deduction. of firms for stratified random sampling.
Where: Finally, 54 firms were selected as sample
within three years of testing course.
WACCAT = E Ke + D Kd (1 - T) In next step, all data and information
V V
needed for calculating firm value by three
T= effective tax rate represented methods were extracted from
Other components of this equation were financial information existing in stock
represented before. exchange, sites and other resources.
Dastgir M., Khodadadi V., Ghayed M. - Cash Flows Valuation Using Capital Cash Flow Method Comparing it with Free Cash Flow Method and
Adjusted Present Value Method in Companies Listed on Tehran Stock Exchange
54 Business Intelligence Journal July
With considering the structure of survey between two methods of Capital Cash
hypotheses and the collected data, paired Flow and Free Cash Flow. Therefore, in
Students t-test was applied for data analysis. general:
Hypothesis H0 is accepted. That is, -level
Research Findings of 5%, using Capital Cash Flow method in
the cash flows valuations would result in
Researchs First Hypothesis Testing similar results of Free Cash Flow.
Paired Differences
Std. Deviation
the Difference
Sig. (2-tailed)
Variables
95%
Dower
Upper
Mean
Confidence
Df
t
Interval of
Sig. (2-tailed)
Variables
Dower
Upper
Mean
Df
t
CCF & APV 0.76 0.89 0.12 0.032 0.16 -0.63 53 0.530
Correlation N Variables
0.99 54 CCF & FCF Table 4. Paired Samples Correlation
Firms valuation with using discounted Booth, Laurence, 2007, Capital cash flow,
cash flows is an effective and extensive topic APV & valuation, European financial
in financial decision making and investment management, Vol.13, No.1, pp 29-48
opportunities which, in author opinion, has
Dastgir M., Khodadadi V., Ghayed M. - Cash Flows Valuation Using Capital Cash Flow Method Comparing it with Free Cash Flow Method and
Adjusted Present Value Method in Companies Listed on Tehran Stock Exchange
56 Business Intelligence Journal July
Chambers, D.R., R.S. Harris and J.J. Pringle Fernandez, Pablo, 2007, Valuing
(1982), Treatment of Financing Mix in companies by cash flow discounting: ten
Analyzing Investment Opportunities, methods and nine theories, Emerald
Financial Management (Summer), pp. managerial finance, Vol.33, No.11,
24-41. pp853-876
Fernndez, Pablo (2001), The correct Harris, R.S. and J.J. Pringle (1985), Risk-
value of tax shields. An analysis of 23 Adjusted Discount Rate Extensions
theories, Working Paper No. 276051, form the Average-Risk Case, Journal of
Social Science Research Network. Financial Research (Fall), pp. 237-244.
Lewellen, W.G. and D.R. Emery (1986), Miller, M. and M. Scholes (1978), Dividend
Corporate Debt Management and the and Taxes, Journal of Financial
Value of the Firm, Journal of Financial Economics (Dec.), pp. 333-364.
Quantitative Analysis (December), pp.
415-426. Modigliani, F. and M. Miller (1958), The
Cost of Capital, Corporation Finance and
Luehrman, Timothy A. (1997), Whats It the Theory of Investment, American
Worth: A General Managers Guide to Economic Review, 48, pp. 261-297.
Valuation, and Using APV: A Better
Tool for Valuing Operations, Harvard Modigliani, F. and M. Miller (1963),
Business Review, (May-June), pp. 132- Corporate Income Taxes and the Cost
154 of Capital: A Correction, American
Economic Review (June), pp. 433-443.
Massari, Mario and Roncaglio, Francesco
and Zanetti, Laura, 2007, On the Myers, S.C. (1974), Interactions of
equivalence between APV and the WACC Corporate Financing and Investment
approach in a growing leveraged firms, Decisions Implications for Capital
European financial management, Vol.14, Budgeting, Journal of Finance (March),
No.1, 2007, pp 152-162 pp. 1-25.
Miles, J.A. and J.R. Ezzell (1980), The Ruback, Richard S. (1995), A Note on
Weighted Average Cost of Capital: Capital Cash Flow Valuation, Harvard
Perfect Capital Markets and Project Life: Business School, 9-295-069.
A Clarification, Journal of Financial and
Quantitative Analysis (September), pp. Ruback, Richard S. (2002), Capital
719-730. Cash Flows: A Simple Approach to
Valuing Risky Cash Flows, Financial
Miles, J.A. and J.R. Ezzell (1985), Management, Summer, pp. 85-103.
Reformulating Tax Shield Valuation:
A Note, Journal of Finance, Vol. XL, 5 Taggart, R.A., Jr. (1991), Consistent
(December), pp. 1485-1492. Valuation and Cost of Capital. Expressions
With Corporate and Personal Taxes,
Miller, M.H. (1977), Debt and Taxes, Financial Management (Autumn), pp.
Journal of Finance (May), pp. 261-276. 8-20.
Dastgir M., Khodadadi V., Ghayed M. - Cash Flows Valuation Using Capital Cash Flow Method Comparing it with Free Cash Flow Method and
Adjusted Present Value Method in Companies Listed on Tehran Stock Exchange
58 Business Intelligence Journal July
Tham, joseph and Wonder. X, 2002, Non- Tham, joseph, 2000, Practical equity
conventional WACC with risky debt valuation: A simple approach, Wp
and risky tax shields, working paper available on the social science research
available on the social science research network (SSRN)
network (SSRN)
Abstract
For this work, we developed a General Equilibrium Model for the Mexican economy where the role of
the indirect tax rate on value added is important. The model is used to analyze the effect of the tax reform
proposed by the President in 2001. Without any doubts, the results obtained are important for discussion
of actual themes in economic politics in Mexico. Particularly, evidence suggests that applying a tax rate
of 15% on food, medicines, educational services and press and editorial products will represent 93,000
million pesos of additional resources for the government, of which 11,079 million pesos could be used to
compensate the first three deciles of the population in order to maintain their original consumption level.
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
60 Business Intelligence Journal July
the 63 productive branches in the Mexican For the production sector, both
economy, whose goods are tradable. The simulations show a decrease in education
estimation of the elasticities constitutes one services. However, for the fiscal policy
of the most important contributions for this without subsidy this effect is less and also
work. promotes a greater production of goods in
The simulation results of both fiscal the construction and installation sectors.
policies can be summarized as follows: Besides, the fiscal policy without subsidy
For private consumption, an increase in the generates more value added and the
return on capital, an increase in the subsidy reallocation of production factors is less
to the first population deciles through wages strong as a consequence of a greater increase
and the increase in the price of compounded in the return on capital, of 1.1%, than in the
goods, affect the income and consumption fiscal policy with subsidy.
decisions of population deciles. In both Total investment increases more in the
simulations, income raises due to the increase fiscal policy with no subsidies for the first
in return on capital and to the reallocation of three population deciles because government
production factors. But changes in income saving is greater than in the fiscal policy with
are greater for the simulation of the second subsidies. Gross fixed investment increases
fiscal policy where the first population more for the fiscal policy with subsidy for
deciles are benefited by the subsidy in the first three population deciles because the
their income through wages. Changes in return on capital corresponding to the rent
income and in price of compounded goods price of capital pieces is greater due to the
led to a decrease in total consumption for increase on wages.
all population deciles. In both simulations
the most affected population deciles are the 2. Model
first three in terms of their consumption,
and the least affected is the tenth population 2.1. Considerations about the model
decile; however, it is important to point out
that the second fiscal policy, where deciles The following model represents a small
with lower income are protected, generates open economy that is price acceptant, with
a larger decrease in total consumption than five economic sectors: companies, families,
the policy without the subsidy. For the first government, foreign sector and investment
fiscal policy, the impact in total consumption sector.
for the first three deciles is 11,079 million Equations within the model include
pesos. endogenous and exogenous variables, as
In the government sector, revenue from well as parameters. Endogenous variables
the collection of income tax, tax on capital will be identified by capital letters;
and value added tax is 13.19% greater for exogenous variables will be in capital letters
the policy without subsidy, which represents and hats, and parameters in lower-case or
an additional 93,300 million pesos to the Greek letters
original equilibrium, whereas for the second
fiscal policy (with subsidy) 20,000 million 2.2. Price System
pesos more are collected respect to the
original equilibrium. Four prices are computed in the price
system: price for domestic goods (Pi ) ,
price for compounded goods (Pi ) , price of
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
62 Business Intelligence Journal July
m
imports (Pi ) and the international price of for final consumption goods which are
we
exports (Pi ) . The first one is obtained from made of intermediate inputs from other
the zero-gain equation for companies1. sectors and of value added. The decision
of how many domestic goods are produced
(1)
(Xis) is calculated by aggregating, through a
vai (1 + tiiva)` wK j ` wl j Leontief function, the intermediate inputs
1-a a i i
n
Pi = d 1 - ai ai
+ / a j,i Pj and demand for labor and capital.
ci j=1 (5)
(8)
by the subsidy on exports for good i (si) .
wL ^1 - aih a i
(4) VAi c m
Ki = wk ai
we P id ci
Pi = (9)
^1 + sih TC
2.3. Supply VAi c wK ai 1-a i
m
wL ^1 - aih
Li =
ci
In the supply side we consider n
different productive sectors in the economy Where ci corresponds to the efficiency
1
This consideration derives from the assumption that in a small
factor of sector i.
open economy companies are price acceptant, so that revenue Total demand for capital (Kd) and labor
for each additional unit produced is equal to its cost the sum of demands in each sector i.
(10) (15)
n S
n
K d = / Ki S g = Zg - / Yig Pi
i=1
i=1
(11)
n
Ld = / Li 2.5. Demand
i=1
(19)
Additionally, government saving (Sg) H
corresponds to the income not used for Y ic = / Y ic h
consumption. h=1
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
64 Business Intelligence Journal July
(20) (24)
H
P imq
S p = / S hp
ic
Pk = c Pic m c imq m
h=1 ic imq
where:
2.6. Investment
ic = Preference parameter for construction good.
imq = Preference parameter for machinery and equipament
For the investment side a new system
was used, which was not found in other Consequently, price for new capital
general equilibrium problems in the revised pieces (Pk) in the economy will be a weighted
literature2. Total investment is considered average of the prices of its two inputs:
not only to be the sum of public, private construction price and that of machinery
and foreign savings, but also equal to the and equipment. This price and the rent price
fixed gross investment (IBF) plus inventory of capital services allows to determine the
variation (YiVE) . return on capital (t) and the value of fixed
(21)
gross investment (IBF).
S The vector of demand for goods that fixed
I = S p + S g + F TC
gross investment generates is the following:
The value of fixed gross investment made
(25)
in the economy (IBF) responds to a Tobins BF
q process, by which the amount of fixed YiIBF = ii I
Pi
gross investment is proportional to the
difference between return on capital in the Where qi represents the preference
domestic economy (t) and the real interest parameter of sector i for fixed gross
rate of foreign financial assets %
(r*) . investment goods.
(22)
The vector of demand for total investment
S will correspond to a percentage qi of total
I BF = d ` t - r*j
investment amount (I/Pi) for each investment
Return on capital (t) refers to the good i.
(26)
rent price of capital pieces or the price of
productive services of capital earned by its YiI = zi I
Pi
owners (wK), relative to its price (Pk).
(23)
Therefore, demand for inventory
t = wK variations will correspond to the difference
PK
between the demand of total investment
Due to prices of construction goods (Pic) and the demand for fixed gross investment
and to prices of machinery and equipment goods, allowing demand for inventory
(Pimq), and considering that the marginal variations to adjust to changes in demand
cost of producing new capital pieces in of total investment and in demand for fixed
the economy must match capital price, and gross investment goods.
(27)
applying Shepards theory, we obtain:
YiVE = YiI - YiIBF
2
See: Serra (1979), Cordova (1991), Balzarotti and Cicowiez
(2004) and Cicowiez (2004)
quantity demanded of exports of good i (Ei) calculate the demand of domestic goods
is determined by the foreign price of exports (Di) and the demand for export goods (Ei).
%
of good i(Pi*) , by the international price of To compute the first one, we need to find
exports of good i (Piwe) and by the price the ratio of domestic use (UDi) through
elasticity of exports (hi). a minimization problem to determine
(28)
the quantity of domestic and imported
S Si*
hi
goods needed by companies to produce
Ei = Ei f Pwe p
compounded goods.
Pi (32)
and it adjusts to maintain equilibrium in the internal demands for domestic produced
balance of payments. goods (Di) and of the demand for exports of
(30)
good i.
n S n S (34)
BP = / Pi* Mi - / Piwe Ei - F = 0 Xid = Di + Ei
i=1 i=1
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
66 Business Intelligence Journal July
In the supply side four parameters are good j used to produce one unit of good i. Its
calculated for each of the 23 sectors. The value is obtained from equation (6):
first one corresponds to the amount of
(37)
aggregated value used in the production of
one unit of good i (vai). For this, we need to II j,i
a j,i =
work out its value from equation (5): Xis
The efficient factor parameter (ci) is
(36)
obtained from the equation (7), working out
vai = wK Ki +s wL Li its value we obtain:
Xi (38)
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
68 Business Intelligence Journal July
(43)
in sector i respect to the total production IBF VE
costs in that sector (ai). Its value is obtained zi = Pi Yi + Pi Yi
I
from the relative price expression relative to
the equilibrium of factors. From the equations for the foreign sector,
we obtain just one parameter mi from the
wK = 1 - ai Li expression of relative equilibrium prices of
wL ai Ki
domestic and imported goods.
where ai is:
(39) 1
P d = ni M iv i
ai = wL Li P m 1 - ni D v1 i
wL Li + wK Ki i
In the private consumption sector two Where the value of mi is the following:
parameters are calculated: the preference
(44)
of each group h for good i (bih) and the 1
percentage each group h allocates of their Pid D iv i
ni =
income to consumption (gh). 1
Pim M + Pid D iv
vi
1
(40)
i
i
c
bih = nPi Yi h
From the equations of market equilibrium
/ Pi Yic h of goods we get parameter bi for each
i=1
productive sector from the minimization
The second parameter is obtained from problem used to obtain demand for domestic
dividing total consumption of group h by goods, where:
(45)
total income of group h:
(41) vi - 1 vi - 1
vi
Pid Di + Pim Mi = bi ` ni Di j vi - 1
n
vi
+ ^1 - nih Mi vi
/PY i
c
ih
gh = i=1
And working out the value of bi we get:
Z hp
For the investment sector we also calculate bi = Pid Di + Pim Mi
vi
two parameters. One is the preference of
vi - 1 vi - 1
` ni Di vi
+ ^1 - nih Mi vi j vi - 1
solving is complex, so we need software demand for capital and labor, as well as the
that helps us solve this type of equations. quantity produced for year 2000 for each of
It could be argued that the calculation of the 23 productive sectors shown in Table 2.
the general equilibrium is not necessary The level of detail of the private demand
because all prices will be equal to one allowed obtaining data for consumption by
since the parameters of the utility function each population decile of total income per
and production were calibrated under this capita for the 23 final goods. At the same
assumption; however, obtaining these time, GAMS showed data about income
results will show that the model is indeed and savings for each population decile.
well calibrated. If the quantities obtained for In the investment sector we got quantities
the original equilibrium do not match those demanded for exports and imports for the
used in the social accounting matrix and/or 23 sectors due to the estimation of price
the prices were not equal to one, an error has elasticities for exports and substitution
occurred in the calibration process to obtain elasticities between domestic and imported
the equilibrium. products, for the 63 tradable goods in the
The software used is General Algebraic Mexican economy.
System (GAMS), which is specifically At the same time, GAMS produced
designed for linear, non-linear and multiple results for intermediate demand of the
integration optimization4. 23 productive sectors corresponding to
The calculation of the general equilibrium intermediate inputs used in the fabrication of
in the Mexican economy for year 2000, the final goods. Finally, data on prices allowed
following assumptions were satisfied: to check that prices from the calculation of
the general equilibrium by GAMS were the
a. Supply is equal to demand in all markets. same as those found in the social accounting
matrix.
b. No sector has positive benefits. All results for the general equilibrium are
shown in Exhibit II.
c. All actors fulfill their budgetary
constraint. 5. Simulations of Fiscal Policies
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
70 Business Intelligence Journal July
the first decile and 2% to 7% for the second the subsidy through wages in the first three
and third deciles. population deciles. Decrease in demand
The first simulation has as main goal for capital among sectors is explained by
to expose the effects of the 15% tax on the different utilization intensities of that
production, consumption, investment, trade, factor. Table 3 shows that the Non-alcoholic
tax colleting and on gross domestic product. beverage sector absorbed all the capital
This will be helpful since the policy was quantity freed by other sectors. Demand
proposed by the President and was sent for capital was 61.3% more in the second
for review to the Congress in 2001, but simulation than in the first one.
has neither been approved or rejected due In the case of the labor market, for
to the unknown economic effects of such both simulations we see an increase in the
policy. The second simulation, besides of demand for labor in all sectors except for
having the same taxing as the first one, is the education service sector, due to the
important because it can prove if protecting substitution among production factors.
lower income population deciles through Nevertheless, changes in quantity demanded
subsidies, as mentioned by different political of labor are smaller in the first simulation,
actors, will help to implement the fiscal but in the same direction as in the second
policy proposed by the President. simulation. Table 4 shows that for the
For both simulations, the description construction and installation sector, the
the effects is detailed for all 23 sectors in increase in quantity demanded of labor was
the economy. The following tables show larger for the first simulation in 3.9% than
the percentage variations obtained respect the second simulation.
to the initial equilibrium and also include
the variations between the first and second Table 3. Capital Demand
simulations.
Simulation Simulation
Sector % Change
1 2
5.1. Effects on Production Sector
sec1 -0.1% -0.3% 0.2%
sec2 -0.1% -0.2% 0.2%
In both simulations we get distortions in
sec3 -0.1% -0.2% 0.1%
the price of compounded goods and salaries.
For both cases, wages have no change since sec4 0% -0.4% 0.4%
they are fixed in order to calculate the new sec5 17.6% 78.9% -61.3%
equilibriums; nevertheless, the economy sec6 0% 0% 0%
experienced and increase in the capital sec7 -0.3% -0.6% 0.3%
retribution that affected the decisions on the sec8 0% -0.4% 0.4%
quantity demanded for production factors sec9 -0.5% -0.5% 0%
for the fabrication of compounded goods. sec10 -0.6% -1.1% 0.6%
In the case of capital markets, the increase sec11 -0.3% -0.6% 0.3%
in the return for that production factor was sec12 0% -0.3% 0.3%
0.8% and 1.9% for the first and second sec13 0% -0.3% 0.3%
simulations respectively. This means that the sec14 -0.2% -0.2% 0%
decrease in quantity demanded for capital sec15 -0.2% -0.5% 0.3%
was smaller for the first simulation than sec16 -0.4% -0.8% 0.4%
for the second simulation, which besides sec17 3.8% -1.3% 5%
the new tax of 15% included an increase in
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
72 Business Intelligence Journal July
sec14 0% 0% 0% sec1 0% 0% 0%
sec15 0% 0% 0% sec2 0% 0% 0%
sec16 0.2% 0% 0.2% sec3 -0.7% -0.7% 0%
sec17 0% 0% 0% sec4 0% -0.8% 0.8%
sec18 0% 0% 0% sec5 0% 0% 0%
sec19 0% 0% 0% sec6 0% 0% 0%
sec20 0% 0% 0% sec7 0% 0% 0%
sec21 -3.2% -3.2% 0% sec8 0% 0% 0%
sec22 0% 0% 0% sec9 0% 0% 0%
sec23 0% 0% 0% sec10 -5.9% -5.9% 0%
sec11 0% 0% 0%
Table 8. Demand for Domestic Goods sec12 -7.7% -7.7% 0%
sec13 0% 0% 0%
Simulation Simulation
Sector % Change sec14 0% 0% 0%
1 2
sec1 0% 0% 0% sec15 0% 0% 0%
sec2 0% 0% 0% sec16 0% 0% 0%
sec6 0% 0% 0% sec20 0% 0% 0%
sec7 0% 0% 0% sec21 0% 0% 0%
sec8 0% 0% 0% sec22 0% 0% 0%
sec9 0% 0% 0% sec23 0% 0% 0%
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
74 Business Intelligence Journal July
to be greater to allow the equilibrium in the deciles. For the fourth decile the difference
balance of payments5. between both policies is 0.7%, for the fifth
and sixth deciles is 0.6%, for the seventh,
5.3. Effects in Private Consumption eighth and ninth deciles is 0.8%, and finally
1.1% for the tenth decile.
The fiscal policy that, besides applying Table 11 shows the change in consumption
the tax of 15% to the consumption of food, by decile for each fiscal policy. Data reveal
medicines, education services and goods that the fiscal policy with subsidy affects
from press and editorials, also increases the more total consumption of deciles, except
subsidy through wages, has a larger impact, for the tenth decile which does not change
than the policy without the subsidy, in the for both simulations. Figures for the second
change of income of population deciles fiscal policy indicate that the decrease in
as consequence of an increase in return consumption for the first decile is 0.6%
on capital and a greater reallocation of greater than in the first fiscal policy, 0.7%
production factors among the sectors. greater for the second and third deciles, 0.6%
greater for the fourth and fifth deciles, 0.4%
Table 10. Income of Population Deciles greater for the sixth decile, 0.7% greater for
the seventh decile, and 0.8% greater for the
Decil
Simulation
1
Simulation
2
% Change eighth and ninth deciles.
Regarding consumption of goods
dec1 0.5% 14.7% -14.2%
produced in the food and education sectors,
dec2 0.4% 17.2% -16.7%
we can observe that population deciles
dec3 0.4% 20% -19.6%
reduce their consumption for these goods
dec4 0.4% 1.1% -0.7%
with the second fiscal policy, even though
dec5 0.4% 1% -0.6% the increase in subsidy through wages.
dec6 0.4% 1.1% -0.6% Comparing to the first fiscal policy, the first
dec7 0.5% 1.2% -0.7% and second deciles reduce their consumption
dec8 0.5% 1.2% -0.7% 0.3% more with the second fiscal policy,
dec9 0.5% 1.3% -0.8% 0.5% more in the third decile, 0.2% more in
dec10 0.8% 1.9% -1.1% the fourth and fifth deciles and 0.4% more in
the fifth decile.
Table 10 clearly shows that changes in For education services consumption
income of population deciles are larger in in the first simulation (without subsidy) is
the second simulation benefiting the first reduced for the first decile in 15.4% more
three deciles in 14.2%, 16.7% and 19.76%, than in the second simulation, 11.11%more
respectively, more than in the first policy. for the second decile, 13.3% more for the
The change in income of the rest of the third decile, 12.2% more for the fourth
deciles is still greater for the second fiscal decile, 11.4% more for the fifth and seventh
policy, but the change compared to the first deciles, 10.2% more for the sixth decile,
fiscal policy is not as big as in the first three 11.9% more for the eight decile, 11.7% more
for the ninth decile and 11.4% more for the
5
According to the Mudell-Flemming model, in an open economy
tenth decile.
with a flexible exchange rate and imperfect capital flow, an in-
crease in the IS leads to a deficit in the balance of payments, so
that the exchange rate must depreciate to reach equilibrium. For
additional reference see Appleyard/Field (1997).
Table 11. Consumption of Population Deciles Table 12. Total Private Consumption
Simulation Simulation
Sector % Change
Sector
1.2% 0% 1.1% 0% 0% 0% 0% 0% 0%
% sec2 -0.2% -0.5% 0.3%
Sec 2
0.3% 0.5% 0.5% 0.3% 0.4% 0.4% 0.3% 0.2% 0.2% -0.1% sec3 -7.2% -7.4% 0.2%
sec4 -0.5% -0.8% 0.3%
Sec 3
0% 0% 0% 0% 0% 0% 0% 0% 0% -1.9%
sec6 0% 0% 0%
Sec 5
0% 0% 0% 0% 0% 0% 1.2% 0% 0% -0.3%
sec7 -0.3% -0.3% 0%
sec8 0% 0% 0%
Sec 6
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
0% 0% 0% 0% 0% 0% 0% 0% 0% -0.4%
sec10 -6.8% -6.8% 0%
Sec 8
sec13 0% 0% 0%
Sec 10
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
sec14 0% 0% 0%
sec15 0% 0.1% -0.1%
Sec 11
sec16 0% 0% 0%
Sec 12
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
sec17 0% 0% 0%
sec18 0% 0% 0%
Sec 13
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
0.3% 0.4% 0.5% 0.3% 0.3% 0.3% 0.2% 0.2% 0.2% -0.2%
sec20 -0.1% -0.1% 0%
sec21 -0.1% -0.2% 0.1%
Sec 15
0% 0% 0% 0% 0% 0% 0% 0% 0% -0.1%
0.2% 0.3% 0% 0% 0% 0% 0% 0% 0% 0%
sec23 -17.6% -29.1% 11.4%
Sec 17
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
when the second fiscal policy is applied,
we can observe larger decreases in the
Sec 19
0% 0% 2% 0% 0% 0% 1.3% 0% 0% -0.3%
15.4 11.1 13.3 12.2 11.4 10.2 11.4 11. 11.7 11.4
the first three deciles decrease their total
Sec 23
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
76 Business Intelligence Journal July
Simulation Simulation
Decil % Change
1 2
r* 0% 0% 0%
P 0.2% 0.9% -0.7%
Pk 0.4% 1.1% -0.6%
IBF 0.4% 1% -0.7%
d 0% 0% 0%
F 0% 0% 0%
indicates that the simulation of the second sec5 450% 500% -50%
fiscal policy promotes a greater quantity sec6 0% 0% 0%
of intermediate demand in the sectors of sec7 -0.3% 0.2% -0.5%
textiles, clothing and leather industry; wood sec8 -1% 0% -1%
industry; chemical substances, petroleum sec9 0% 0.1% -0.1%
by-products, rubber and plastic; non-metallic sec10 6% 6% 0%
mineral products; basic metal industries; sec11 -0.03% 0.1% -0.1%
metal products, machinery and equipment; sec12 19.3% 19.7% -0.3%
other manufacture industries, transport and sec13 -0.1% 0.1% -0.2%
storage and communications; whereas the sec14 -0.1% 0% -0.1%
second simulation resulted in a reduction sec15 -0.8% 0.1% -0.9%
of the quantity demanded of intermediate sec16 -0.4% 0.1% -0.6%
goods in those same sectors. sec17 0% 0% 0%
For all other productive sectors that
sec18 0% 0% 0%
present changes between both simulations,
sec19 0% 0% 0%
results are in the same direction. However,
sec20 -0.4% 0.3% -0.7%
for the simulation of the second fiscal policy
sec21 0.3% 0.5% -0.2%
increases in the quantity demanded of
sec22 -0.02% -0.3% 0.3%
intermediate goods are greater than in the
simulation of the first fiscal policy. sec23 0% 0% 0%
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
78 Business Intelligence Journal July
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
80 Business Intelligence Journal July
A. Appendix I
Table 19. Parameters of Supply Equations Table 20. Parameters of Investment Equations
Sector a i
vai ci Sector q i
f i
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
82 Business Intelligence Journal July
Table 22. Parameters of Foreign Sector Equations Table 23. Parameters of Equilibrium Market of
Goods Equations
Sector m i
Sector bi
% % % % d
K0s Ls0 F r*
3,156,911 1,526,705 -357,433 0.07 18.596
B. Appendix II
Table 25. Exogenous Variables Table 26. Production Sector. (Hundred Billion of
Pesos)
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
84 Business Intelligence Journal July
Table 27. Government Expenditures. (Hundred Table 29. Saving and Income by Population Deciles.
Billion of Pesos) (Hundred Billion of Pesos)
Variables Valuesi
Zg 5.557
Sg 1.531
Revenues as % of GDP 10.5%
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
86 Business Intelligence Journal July
Table 31. Investment. (Hundred Billion of Pesos) Table 33. Foreign Sector. (Hundred Billion of Pesos)
Table 34. Intermediate Demand. (Hundred Billion of Table 35. Investment. (Hundred Billion of Pesos)
Pesos)
Mrquez-Pea E. - Effects of Value Added Tax on Mexican Economy A multisectorial analysis by developing an Applied General Equilibrium Model. MEGA
88 Business Intelligence Journal July
Abstract
One of the most recent researches in financial field is using an Ant Colony System (ACS) in the
corporate bankruptcy prediction.
In this paper, the Ant Colony System is used to predict the bankruptcy of listed companies in Tehran
Stock Exchange. The research findings shows that models prediction power is about 59% with taking into
account the rations proposed by Altman (1968). After examining the causes of significant errors in model,
a proposed model is introduced. This model which is based on four financial ratios fits the economical
infrastructure of Iran. The results obtained from testing of the suggested model indicate that this model is
able to predict the corporate bankruptcy one year prior to bankruptcy with an accuracy of 93% for large-
sized firms and about 90% for small-sized ones, respectively.
Khodadadi V., Zandinia A., Npuri M. - Application of Ants Colony System for Bankruptcy Prediction of Companies listed in Tehran Stock Exchange
90 Business Intelligence Journal July
Human foraging behavior always In real world ants run more or less at
seeks to approach an integrated set of random around their colony to search for
information and to discover relationships food. Ants deposit a chemical substance
between phenomena for the explanation called pheromone along the travelled paths
of these phenomena and the prediction of (Dorigo and Caro, 1999). After rainfall,
their behaviors. These attempts extended to these paths get white and become visible.
financial field and have induced researchers Other ants upon finding these paths would
to discover relationships between financial follow the trail. Then, if they discover
information in order to provide the financial food source they will return to the nest and
information users with more powerful deposit another path along with the previous
control tools. one, and will be strengthened the preceding
A highly valuable effort made by path.
financial researches has been the application When it turns to select a path, ants
of financial ratios in bankruptcy prediction. will prefer one containing high density of
The research conducted in this field pheromone, i.e. paths on which more ants
resulted in proposing numerous models have travelled. Pheromone is gradually
for the prediction of corporate bankruptcy. evaporated. This evaporation is useful for
The continuity of these researches led to three reasons:
introduction of using artificial intelligence It makes the next trail less attractive for
(AI) and computer science techniques in the following ants. Ants mostly take shorter
field of finance. paths; hence, the shorter path between nest
One of the techniques introduced is and food source is highly strengthened and
the application of Ant Colony System in any farther path is less strengthened.
bankruptcy prediction (Milea, 2005). This Unless pheromone did not evaporate at
model was initially proposed by Dorigo all, the paths having been passed many times
(1992) to solve the Traveling Salesman would become so attractive that the random
Problem (TSP) (Dorigo and Stutzle, 2004). searching for food would be highly limited.
Thereafter, researches tended to benchmark When food finished at the end of path,
this model to solve similar optimization evaporation of the remaining paths would
problems in their field of study. The not mislead ants in their searching for that
researchers of financial field were also to track leading to food source.
test this model in order to find out a solution Therefore, when an ant finds a shorter
for financial problems. (optimal) path from nest to food source, other
In this research, the applicability of ants will more likely follow the same path
this model in the prediction of corporate and, over time, the continuous strengthening
bankruptcy will be explained and the of the path and the evaporation of other trails
coefficient of success of this model will be will make all ants become unidirectional
tested. For this reason, the financial ratios (Dorigo and Colorni, 1996).
of corporate have filed in Tehran Stock This behavior in ants has a kind of swarm
Exchange in the period 2001-2007 were intelligence having been recently considered
used. by scientists. But it should be reassured that
there is a major difference between swarm
intelligence and social intelligence. In social
direct relationship between them and they probability that ant k will go from i to j at
are in indirect contact with each other by iteration t, is used by the ants in building
signals. This behavior was first studied by their tours. This probability depends on
Grass, a French scientist. It was referred 2 parameters: a heuristic measure of the
to as Stigmergy which is a particular form desirability of adding edge (i,j) to the current
of indirect communication used by social tour, ij, and the amount of pheromone
insects to coordinate their activities via currently on edge (i,j), ij. Probabilistic
changes made in the local environment transition rule is given as (2) below:
(Dorigo et.al, 2000).
(2)
Khodadadi V., Zandinia A., Npuri M. - Application of Ants Colony System for Bankruptcy Prediction of Companies listed in Tehran Stock Exchange
92 Business Intelligence Journal July
1. The initialization step: Current Ratio are more effective than others
(Balcaen and Ooghe, 2006).
In this step ants are placed in the initial Followed by these studies, a great number
position and an amount of 0 is taken for of researchers involved in examining
initial pheromone. the prediction of corporate bankruptcy
based on financial ratios which resulted
2. Tour calculation: in the presentation of four univariate
analysis methods, risk index model,
Distance between parameters is multiple discriminant analysis models
calculated. and conditional probability models for
bankruptcy prediction. In 1967, Beaver was
3. Model estimation: pioneer to examine the corporate bankruptcy
using univariate analysis method (Balcaen
If the system output is not satisfactory, and Ooghe, 2006).
then ants will move to the next location Tamari in 1966 presented risk index
based on the probabilistic transition function model to predict bankruptcy and thereafter
and pheromone is converged. Moses and Liao (1987), also applied this
model to predict corporate bankruptcy.
4. Repeating Criterion: In reference to studies made by Beaver,
Altman proposed a more accurate model to
Procedure continues until the system predict firms risk of bankruptcy with using
output reaches a satisfactory position Multiple Discriminant Analysis (MDA)
(Dorigo and Stutzle, 2004). (Altman, 1968). After his work, some
extensive researches were also conducted
Literature Review about bankruptcy prediction based on
financial ratios and applying multiple
1. Classic Methods of Bankruptcy discriminant analysis (MDA). Altman
Prediction (1968), Deakin (1972), Edmister (1972),
Blum (1974), Altman et al. (1977), Deakin
A primary study was made on bankruptcy (1977), Taffler and Tisshaw (1977), Van
prediction by Whitaker and Smith in 1935. Frederikslust (1978), Bilderbeek (1979),
They involved investigation about the Dambolena and Khoury (1980), Taffler
sufficiency of financial ratios as predictors of (1982), Ooghe and Verbaere (1985), Taffler
financial failures. The two aforementioned (1983), Micha (1984), Betts and Belhoul
researches had been analyzing the average (1987), Gombola et al. (1987), Gloubos and
procedure of 21 rations for 10 years since Grammatikos (1988), Declerc et al. (1991),
then and concluded that Working Capital/ Laitinen (1992), Lussier and Corman
Assets has more stability and accuracy and (1994), Altman et al. (1995) are among those
it is significant as an index to identify failure. researchers who one after the other proposed
Merwin (1942) also examined several bankruptcy prediction models using multiple
ratios within first 6 years of the lifespan discriminant analysis technique.
of corporate with continuous activity and Beginning in the 1980s, the application
those whose activity had been stopped. of Multiple Discriminant Analysis (MDA)
He concluded that three ratios of Working method was declined but it still remained
Capital/Total Assets; Net Equity/Debt and
Khodadadi V., Zandinia A., Npuri M. - Application of Ants Colony System for Bankruptcy Prediction of Companies listed in Tehran Stock Exchange
94 Business Intelligence Journal July
R = {CX1, CX2, CX3, CX4, CX5} Application of Ant Colony System model
in the studied sample of large and small size
(6)
firms based on the five financial ratios of
Xki # CXi, i ! "1, 2, 3, 4, 5, sample firms led to the following results as
shown in Tables 1 and 2.
Where CXn represents the cut-point value
of financial ratio Xn for which the fitness Table 1. Models Results in Large- sized firms
function is maximized. So, a firm with values
smaller or equal to R for each financial ratio
2002
2003
2004
2005
2006
2007
Description (No.)
is predicted to go bankrupt otherwise not;
and k is the given firm (Milea, 2005). Actual Bankrupt 2 2 2 1 5 5
Situation
Non-Bankrupt 34 34 34 35 31 31
Explanation of Bankruptcy Model Bankrupt 0 0 0 0 2 1
Concept in Iran Prediction
Non-Bankrupt 36 36 36 36 34 35
2002
2003
2004
2005
2006
2007
Description (No.)
predicting the bankruptcy status of the
Actual Bankrupt 5 6 8 9 8 5 Tehran Stock Exchange, especially in the
Situation
Non-Bankrupt 31 30 28 27 28 31 case of small-size firms, forced us to closely
Model Bankrupt 0 0 1 1 1 0 examine the model. We found out that in the
Prediction
Non-Bankrupt 36 36 35 35 35 36 original model:
The tables show that this model nearly is 1. Non-fitness of bankruptcy identification
not able to predict bankrupt firms, but it can criterion in Iran with the ratio of Market
only predict non-bankrupt firms. Value of Equity/ Book Value of Total
To be confident of the accuracy of the Debt:
obtained results and to examine whether
these results are affected by unequal number As described in previous sections,
of bankrupt and non-bankrupt firms or not, the bankruptcy of companies in Iran is
we take a sample consisting of 42 bankrupt identified in article 141 of the Commercial
and non-bankrupt firms (21 bankrupt and 21 Law. Hence, the classification criterion to
non-bankrupt firms) as control sample and identify firms into two groups of bankrupt
the results are as shown in Table 3. and non-bankrupt is the amount of firms
capital. With glancing over the concerned
Table 3. Models Results in Control sample firms independent variables in order to test Ant
Colony System for bankruptcy prediction we
Description (No.) Bankrupt
Non- will find out that this criterion has not been
Bankrupt
considered in any of independent variables
Actual Situation 21 21
and, instead, Market Value of Equity ratio
Model Prediction 4 38
has been expressed in terms of fluctuation
Correct prediction 4 21
which, in spite of firms belonging to
Percentage of correct
prediction
59.52% samples with equal sizes, would be high in
studied samples.
Khodadadi V., Zandinia A., Npuri M. - Application of Ants Colony System for Bankruptcy Prediction of Companies listed in Tehran Stock Exchange
96 Business Intelligence Journal July
reasonable manner is necessary. But, due of firms, we, firstly, test the model for the
to lacking any legal obligation concerning control sample composed of 42 corporate
listed companies in Tehran Stock Exchange (21 bankrupt and 21 non-bankrupt firms).
to revaluate their assets, this ratio will The results obtained from control testing are
remain as a problematic parameter whenever as shown in Table 4.
financial problems are to be solved. Also,
as a result of monopoly for some products Table 4: Results of proposed model in Control
such as steel and automobile, the above ratio sample firms
cannot be an appropriate scale to compare
firms. Description (No.) Bankrupt
Non-
Bankrupt
The following measures were taken to
Actual Situation 21 21
resolve the above-mentioned issues:
Model Prediction 21 21
Correct prediction 19 19
1. Due to using amount of Capital for
Percentage of correct
classifying corporate into two groups prediction
90.47%
of bankrupt and non-bankrupt firms, we
have used Book Value of Equity/Book The results displayed in Table 4 indicate
Value of Total Debts instead of Market that this model is able to predict up to 90%
Value of Equity/ Book Value of Total of corporate financial status.
Debts. After examining of proposed model in
2. As a result of very highly inflated Sales/ control sample and being confident of the
Book Value of Total Assets as well as model accuracy within the same sample,
lacking any legal obligation for Listed we tested the new suggested model for two
Companies in Tehran Stock Exchange to main samples, i.e. sample of large- and
revaluate their assets and the monopoly small-sized firms. The results obtained from
for some products, we eliminate this testing of proposed model are as shown in
ratio from the model. Tables 5 and 6.
Thus, we examine Ant Colony System Table 5: Proposed models Results in Large- sized
model with new conditions for the samples firms
having been previously tested. The new
suggested model to predict the bankruptcy of
2002
2003
2004
2005
2006
2007
Description (No.)
listed companies in Tehran Stock Exchange
is composed of four independent variables Actual Bankrupt 2 2 2 1 5 5
Situation
as follows: Non-Bankrupt 34 34 34 35 31 31
97.22
91.67
94.44
91.67
91.67
prediction
Total Debts Average of correct
Due to unequal number of bankrupt and predictions per.
93.06%
93% for the model to predict the bankruptcy After these variables were tested in the
of large-sized corporate. model we found out that this model is not
After testing of proposed model in the likely able to predict non-bankrupt firms
sample of large-sized firm, the model was and its real prediction power is about 59%.
also tested for small-sized firms and the The investigations showed that the system
following results were obtained: error was caused by the two ratios of Market
Value of Equity/Book Value of Total Debts
Table 6: Proposed models Results in Small- sized as well as Sales/Total Assets. In virtue
firms of article 141 of the Commercial Law of
Iran and choosing capital as a criterion for
2002
2003
2004
2005
2006
2007
Description (No.) corporate bankruptcy, using Market Value
Actual Bankrupt 5 6 8 9 8 5
of Equity is not logical. Moreover, this will
Situation
Non-Bankrupt 31 30 28 27 28 31
still remain as a problematic ratio because
Model Bankrupt 3 5 5 4 7 1 of lacking any legal obligation for listed
Prediction
Non-Bankrupt 33 31 31 32 29 35 companies in Tehran Stock Exchange for
Correct prediction (No.) 34 35 33 29 33 30 assets revaluation. Also, as a result of very
Percentage of correct high inflation rate in Iran, instable economic
94.44
97.22
91.67
80.56
91.67
83.33
prediction
situations as well as monopoly for some
Average of correct
89.82% products such as steel and automobile, the
predictions per.
ratio of Sales/Total Assets is very inflated
The results of the test made on proposed and may not be regarded as an appropriate
model indicate a coefficient of success scale to compare firms.
of 89.82% for the model to predict the In order to resolve above-mentioned
bankruptcy of small-sized firms. matters, we consider Book Value of Equity
Upon comparison of the results of the test instead of Market Value of Equity and, also,
made on proposed model we found out that Sales/Total Assets was omitted from among
this system is slightly affected by corporate variables. At last, a proposed model consisted
size. So that the models coefficient of of four variables as described below was
success in large-sized corporate group is proposed to predict the bankruptcy of listed
about 93% while in small-sized corporate companies in Tehran Stock Exchange based
group is about 90%, respectively. on Ant Colony System:
Consequently, as the results of Working Capital/Total Assets; Retained
investigations show, the proposed model is Earnings/Total Assets; EBIT /Total Assets;
able to predict the bankruptcy of corporate Book Value of Equity/Book Value of Total
one year prior to bankruptcy. Debts.
Conclusion and Summarization The results produced by the testing of
This study aims at the application of Ant new suggested model indicated that the
Colony System for bankruptcy prediction prediction power of this model within the
of corporate accepted in Tehran Stock sample of large-sized firms is about 93%
Exchange. and in the sample of small-sized firms is
The primary variables used in this model about 90%, respectively. Therefore, this
include Working Capital/Total Assets; model is able to predict the bankruptcy of
Retained Earnings/Total Assets; EBIT/Total listed companies in Tehran Stock Exchange
Assets; Market Value of Equity/Book Value with a very high coefficient of success and it
of Total Debts; Sales/Total Assets. is slightly affected by corporate size.
Khodadadi V., Zandinia A., Npuri M. - Application of Ants Colony System for Bankruptcy Prediction of Companies listed in Tehran Stock Exchange
98 Business Intelligence Journal July
Khodadadi V., Zandinia A., Npuri M. - Application of Ants Colony System for Bankruptcy Prediction of Companies listed in Tehran Stock Exchange
100 Business Intelligence Journal July
Abstract
The purpose of this study is to explore attribute differences between U.S. GAAP and Iranian GAAP
(IRA-GAAP) earnings components. We compare the relative and incremental value relevance of IRA-
GAAP and U.S. GAAP based operating cash flows and accruals. The issue is investigated by regressing
listed Iranian firms stock returns on the levels of operating cash flows and accruals in IRA-GAAP and
same items adjusted to U.S. GAAP.
The results indicate that there are no significant differences between IRA-GAAP and U.S. GAAP based
earnings components in explaining stock returns. Thus, operating cash flows and accruals components
based on IRA-GAAP are not more value relevant than operating cash flows and accruals components
adjusted to U.S. GAAP.
The results of this study have important implications for accounting standards setters in Iran and other
countries about how cash flow statement classifications affect the value relevance of their items.
Nikbakht Z, Aflatooni A. - The Value Relevance of Earnings Components in Two Different GAAPs
102 Business Intelligence Journal July
The restructuring of the economy which have become mandatory since 2001 and
followed after the Iran Revolution, expanded 2002.
public-sector control over the economy and After finalization, the committee assigned
reduced the need for private capital. At the them to the Technical Committee. These
same time, the interest-bearing bonds were standards were studied by the Technical
abolished. Because of these events, the TSE Committee and sent to the board. On the
experienced a period of standstill. basis of the Audit Organizations public
This stagnation ended in 1989 with the convention approval, these new standards
revival of the private sector through the became mandatory in 2002 (Mashayekhi
privatization of state-owned enterprises and and Mashayekh, 2008). On the date of
the promotion of private-sector economic writing this paper, the number of Iranian
activities based on the First Five-year Accounting Standards was 29.
Development Plan of the country. Since
then the TSE has expanded continuously 3. Literature review
(Mashayekhi and Mashayekh, 2008). Thus,
TSE is an emerging market in the world Prior empirical research on quality
economy. differences between accounting-standard
regimes is scarce and provides mixed results.
2.2. A brief History of accounting For example, Barth, Landsman, Lang and
standards setting in Iran Williams (2006) compare measures of
accounting quality for firms applying IAS
Prior to 1979, financial reporting in Iran with US firms to investigate whether IAS are
was influenced heavily by Anglo-American associated with less earnings management,
practices (Mirshekari and Saudagaran, more timely loss recognition, and higher
2005). Following the Iranian revolution, in value relevance of accounting amounts
1980, an amendment to the Direct Tax Law than US GAAP. They find that IAS firms
disbanded the Official Accountants Institute. exhibit lower accounting quality relative to
During 19801982, the government US firms in terms of earnings smoothing,
established audit enterprises and the correlation between accruals and cash flows,
Budget, Planning and National Industries timely loss recognition, and the association
Organization. These new organizations between accounting amounts and share
recruited a large number of accountants prices. Also, they find that IAS accounting
who were previously employed with private amounts are of similar quality to reconciled
auditing firms. In 1987, following the merger US GAAP amounts. Their results suggest
of the public-sector audit entities, the Audit that although IAS accounting amounts
Organization was established as the sole may not be of higher quality than those of
audit organization with public ownership US GAAP applied comprehensively, they
and as the only regulatory body for setting are of comparable quality to reconciled US
national accounting and auditing standards. GAAP amounts reported by cross-listed
The Audit Organization is responsible for firms. Harris and Muller (1999) suggest
compiling and determining principles and that investors have realized that US-
rules of auditing and accounting in Iran. GAAP provides more investment-related
Twenty five accounting standards were information than a standard IAS.
issued initially after approval by the Audit Alford, Jones, Leftwich and Zmijewski
Organizations public convention. They (1993) investigate the value relevance
Nikbakht Z, Aflatooni A. - The Value Relevance of Earnings Components in Two Different GAAPs
104 Business Intelligence Journal July
Nikbakht Z, Aflatooni A. - The Value Relevance of Earnings Components in Two Different GAAPs
106 Business Intelligence Journal July
As you see, all variables are scaled by population of Iranian firms listed in Tehran
lagged total assets to reduce heteroscedsticity. Stock Exchange for 2002-2007, but because
Using the estimated coefficients ( of data problems, only 1579 firm-years
) from Eq. (5), Non-discretionary accruals are used in research. For data collection
(NDACCR) are defined as: purpose, we use the electronic archival
(6)
data provided by TSE. In some cases that,
the required data is incomplete, the manual
archives in the TSEs library are used. Also,
a part of data is acquired from Tadbirpardaz
And then discretionary accruals and Sahra (two Iranian) databases.
(DACCR) are defined as the residuals of Eq.
(5): 5.1. Descriptive statistics
(7)
NDACCR
NDACCR
DACCR
DACCR
DACCR
ACCR
ACCR
ACCR
CFO
CFO
CFO
RET
N
Meana
2002 272 0.15 0.09 0.07 -0.28 0.36 0.15 0.01 -0.32 0.33 (0.00)*** (0.00)*** (0.38) (0.42)
2003 245 0.26 0.07 0.13 -0.18 0.32 0.12 0.08 -0.21 0.29 (0.02)** (0.10) (0.60) (0.49)
2004 255 -0.02 0.06 0.26 0.05 0.22 0.11 0.21 0.01 0.20 (0.75) (0.76) (0.86) (0.49)
2005 325 -0.41 0.06 0.08 -0.13 0.21 0.10 0.03 -0.15 0.18 (0.08)* (0.13) (0.57) (0.27)
2006 365 0.10 0.07 0.02 -0.20 0.22 0.11 -0.02 -0.22 0.19 (0.00)*** (0.03)** (0.70) (0.54)
2007 297 0.18 0.09 -0.06 -0.28 0.22 0.17 -0.14 -0.34 0.20 (0.75) (0.77) (0.85) (0.71)
Full sample 1759 0.03 0.07 0.08 -0.17 0.26 0.12 0.02 -0.20 0.23 (0.28) (0.31) (0.59) (0.13)
Medianb
2002 272 0.00 0.07 0.07 -0.19 0.23 0.14 0.01 -0.22 0.20 (0.00)*** (0.00)*** (0.08)* (0.15)
2003 245 0.00 0.06 0.08 -0.12 0.21 0.12 0.03 -0.15 0.18 (0.00)*** (0.00)*** (0.15) (0.14)
2004 255 -0.01 0.04 0.08 -0.09 0.15 0.09 0.04 -0.11 0.13 (0.00)*** (0.00)*** (0.29) (0.15)
2005 325 -0.14 0.07 0.05 -0.11 0.14 0.10 0.01 -0.12 0.12 (0.01)** (0.00)*** (0.27) (0.09)*
2006 365 0.00 0.06 0.01 -0.12 0.13 0.10 -0.02 -0.13 0.11 (0.00)*** (0.00)*** (0.21) (0.05)*
2007 297 0.00 0.06 0.04 -0.07 0.11 0.10 -0.01 -0.09 0.09 (0.00)*** (0.00)*** (0.11) (0.06)*
Nikbakht Z, Aflatooni A. - The Value Relevance of Earnings Components in Two Different GAAPs
108 Business Intelligence Journal July
NDACCR
NDACCR
NDACCR
DACCR
DACCR
DACCR
ACCR
ACCR
ACCR
CFO
CFO
CFO
RET
N
Full sample 1759 0.00 0.06 0.05 -0.11 0.15 0.11 0.01 -0.14 0.14 (0.00)*** (0.00)*** (0.00)*** (0.00)***
Standard deviationc
2002 272 1.19 0.23 0.23 0.48 0.47 0.23 0.22 0.47 0.44 (0.78) (0.40) (0.64) (0.33)
2003 245 1.92 0.27 0.39 0.55 0.50 0.27 0.39 0.53 0.47 (0.83) (0.97) (0.48) (0.43)
2004 255 2.42 0.32 2.45 2.51 0.43 0.27 2.29 2.34 0.42 (0.28) (0.24) (0.23) (0.62)
2005 325 1.16 0.34 0.40 0.51 0.29 0.33 0.39 0.49 0.27 (0.62) (0.53) (0.56) (0.39)
2006 365 0.98 0.19 0.25 0.62 0.51 0.19 0.26 0.59 0.48 (0.77) (0.87) (0.37) (0.23)
2007 297 3.81 0.34 3.23 3.29 0.82 0.42 3.63 3.73 0.83 (0.01)** (0.03)** (0.03)** (0.85)
Full sample 1759 2.13 0.29 1.67 1.69 0.52 0.30 1.77 1.80 0.51 (0.00)*** (0.01)** (0.01)** (0.26)
*, **and *** Significant at the 0.1, 0.05, 0.01 level, respectively. RET is annual market-adjusted stock returns measured over
a 12-month period ending 4 months after the fiscal year-end, CFO is operating cash flows that are measured based on U.S.
GAAP and IRA-GAAP, ACCR is total accruals that are measured by subtracting CFO from earnings. DACCR and NDACCR
are discretionary accruals and nondiscretionary accrual that are measured by using of modified Jones model.All variables
are scaled by lagged total assets.a, b and c: P-values for differences in means, medians and standard deviations are based on
t test, Wilcoxon test and F test.
and U.S. GAAP show that the differences ACCR -0.93 (0.00)***
between median of CFO and ACCR are DACCR -0.93 (0.00)*** 0.95 (0.00)***
significant in all years and full sample but, NDACCR -0.01 (0.60) 0.17 (0.00)*** -0.13 (0.00)***
the differences between mean, median IRA- GAAP CFO ACCR DACCR
Nikbakht Z, Aflatooni A. - The Value Relevance of Earnings Components in Two Different GAAPs
110 Business Intelligence Journal July
Table 4. Incremental information content of operating cash flows, nondiscretionary accruals and discretionary
accruals.
Full Sample 2002 2003 2004 2005 2006 2007
N 1759 272 245 255 325 365 297
Regression results of market-adjusted returns on operating cash flows, nondiscretionary accruals and discretionary
accruals
Panel A: U.S. GAAP
-0.06 -0.07 -0.27 -0.48 -0.43 -0.06 0.44
Intercept
(-1.63) (-0.84) (-1.85)* (-2.92)*** (-5.20)*** (-1.03) (1.86)*
Operating cash 0.63 1.89 4.02 3.84 -0.75 1.33 0.91
flows (6.66)*** (6.24)*** (8.94)*** (9.48)*** (-2.87)*** (5.06)*** (2.16)**
Nondiscretionary 0.56 0.64 1.94 3.12 0.24 0.93 -1.29
accruals (4.83)*** (1.97)** (4.31)*** (5.09)*** (0.77) (3.81)*** (-3.30)***
Discretionary 0.64 0.56 1.67 3.91 -0.10 0.69 0.92
accruals (7.08)*** (1.80)* (4.85)*** (9.75)*** (-0.48) (3.51)*** (2.19)**
Adjusted R2 0.029 0.128 0.255 0.391 0.045 0.093 0.036
Panel B: IRA-GAAP
-0.06 -0.14 -0.38 -0.43 -0.39 -0.08 0.40
Intercept
(-1.54) (-1.46) (-2.54)** (-2.61)*** (-4.77)*** (-1.33) (1.72)*
Operating cash 0.66 1.86 3.82 3.71 -0.72 1.32 1.06
flows (6.92)*** (6.20)*** (8.19)*** (9.05)*** (-2.76)*** (4.92)*** (2.48)**
Nondiscretionary 0.52 0.57 2.10 2.74 0.23 0.97 -1.39
accruals (4.41)*** (1.67)* (4.51)*** (4.31)*** (0.69) (3.96)*** (-3.59)***
Discretionary 0.64 0.53 1.79 3.76 -0.09 0.72 1.07
accruals (7.05)*** (1.65)* (5.14)*** (9.23)*** (-0.42) (3.65)*** (2.50)**
Adjusted R2 0.029 0.127 0.232 0.396 0.039 0.093 0.042
Vuong Z -0.948 0.149 2.07** -0.634 0.750 0.116 -0.594
statisticsa
*, **, *** Significant at the 0.1, 0.05, 0.01 level, respectively. Reported numbers are regression coefficients (t value). a Vuongs
(1989) Z-statistic compares the non-nested models.
In panel A (U.S. GAAP), for the full of earnings, and discretionary accruals
sample, the coefficients on operating provide incremental information over
cash flows, discretionary accruals, and that contained in the nondiscretionary
nondiscretionary accruals are 0.63, 0.56 component of earnings. In panel A and B,
and 0.64, respectively and, in panel B (IRA- the adjusted R2 for full sample is 2.9%
GAAP) for the full sample, the coefficients and 2.9% and Based on Vuong Z-statistics
on operating cash flows, discretionary (-0.984), the difference between them is
accruals, and nondiscretionary accruals not significant. Except for 2003, the annual
are 0.66, 0.52 and 0.64, respectively. All regression results present a similar pattern.
coefficients are significant at the 0.01 level, Results indicate that operating cash flows
but they are not statistically different from and discretionary and nondiscretionary
each other (except for 2003). accruals based on IRA-GAAP are not more
These results suggest that in addition closely associated with contemporaneous
to operating cash flows, both discretionary stock returns than operating cash flows and
and nondiscretionary components of accruals components adjusted to U.S. GAAP.
accruals contribute to the value relevance Therefore operating cash flows and accruals
Amir, E., Harris, T., and Venuti, E. (1993), Dechow, P., Sloan, R., and Sweeney, A.
A comparison of the value-relevance (1995), Detecting earnings management,
Nikbakht Z, Aflatooni A. - The Value Relevance of Earnings Components in Two Different GAAPs
112 Business Intelligence Journal July
that finance most of their growth through much when the company is near default
retained earnings. Airlines however are an because most of the payoff will be paid out
example of an industry that is characterized to the debt holders. A financially sound
by its high debt level. In general, capital company would not have this agency
structure is similar within industries with problem because equity holders stand to
differences resulting from weighing the lose more from risky projects when the
benefits of a higher tax shield versus the company is not in risk of going bankrupt,
benefits of the less information sensitive and thus want to avoid them along with
financial of retained earnings. bondholders. The second conflict arises
A second application of game theory to when equity holders cannot fully control
capital structure is concerned with agency the actions of managers. This occurs when
costs. In 1976 Jensen and Meckling managers have an incentive to pursue their
described two kinds of agency problems in own interests rather than those of the equity
corporations: One between equity holders holders. Executive compensation in the form
and bondholders and the other between of option contracts can create incentives
managers and equity holders. The first arises for managers to make risky decisions in
because the owners of a levered firm have an attempt to gain the highest payoff from
an incentive to take risks at the expense of the call options. Higher risk increases the
debt holders. Stockholders of levered firms value of an option, but risk can also cause a
gain when business risk increases because stock price to take a nosedive. A manager
they receive the surplus when returns are with options is not hurt nearly as much as a
high but the bondholders bear the cost when worker with his/her retirement savings in a
default occurs. Bondholders value does company whose stock plummets because of
not increase with the value of the firm, thus risky bets. Option contracts were meant to
they would like the firm to take safe bets better align the interests of managers with
to minimize the risk of default. Equity stockholders, but it is obvious that this is not
holders on the other hand, receive whatever so easily achieved.
is leftover after paying back debt holders. Game theory can also be used to explain
They would like to see the upside potential what is observed in the course of many
of the company maximized and this occurs corporate acquisitions. If markets are
through taking on risky projects (higher efficient then one would expect a company
returns are generated though greater risk to pay fair value when acquiring another
taking.) It is obvious that there is a conflict company; however in many instances
of interest between equity holders desire for the acquirer pays a large premium to buy
business risk and bondholders aversion to the other company. In 1986 Shleifer and
business risk. Financial managers who act Vishny provide one explanation of this
strictly in the interests of shareholders will phenomenon, the free rider problem. One of
favor risky projects over safe ones. It is the concepts behind efficient markets is the
important to note that this agency cost does market for corporate control. The market
not occur in financially sound companies. It for corporate control says that in order for
mainly occurs when the odds of default or resources to be used efficiently, companies
high and equity holders feel they can make need to be run by the most able and
one last gamble to avoid bankruptcy and get competent managers. One way to achieve
a big payoff at the same time. An average this is through corporate acquisitions. A
payoff would not benefit the stockholders company is not likely to purchase another
company unless it feels it can run it better between informed buyers and uninformed
than the current management. This can be buyers. The informed buyers know the true
through new management or synergies that value of the stock and will only purchase
occur because of the operational efficiency shares at or below its true value. The
of the combined firms. Shareholders of implication of this is that the uninformed
the company to be acquired realize that the buyers will receive a high allocation of
new management will run in their company overpriced shares since they will be the
more efficiently and thus their shares will only people in the market when the offering
be worth more than before the purchase. price is above the true value. Knowing this,
Since shareholders will be able to benefit uninformed buyers would be unwilling to
from the improvements implemented by the purchase the stock; forcing the informed
new management they have an incentive to buyers to hold onto the stock because there
hold out and say no to the bid. They will is no one they can sell it to. Therefore, to
only be willing to tender their shares if the induce the uninformed to participate they
offer price fully reflects the value under the must be compensated for the overpriced
new management. Therefore, a bidding firm stock they end up buying. One way to do
cannot make a profit from tendering for the this is to under-price the stock on average.
target firm, and in truth, after the costs of This means that on average the uniformed
acquiring information in preparation for the will buy a stock that started out undervalued
bid, the bidding firm will make a loss. and thus they are still able to buy the stock
A second puzzle associated with corporate at or below its true value. Since all investors
takeovers that has been documented is the know that an IPO will likely be under priced
fact that bidding in takeover contests occur they all try to buy the stock as quick as
through several large jumps rather than many possible creating a demand for the stock that
small ones. The logic behind such behavior results in substantial price gain in the initial
as reported by Fishman in 1988 is relatively days of trading.
straightforward. Observing a bid alerts the Another interesting implication of IPOs
market to the potential desirability of the pointed out by Ritter in 1991 is the fact that
target. If the initial bid is low, a second bidder while they experience high returns in the
will find it worthwhile to spend the cost to short run they typically under-perform the
investigate the target. This second firm may market in the long run. One argument for
then bid for the target and push out the first this behavior is that the market for IPOs is
bidder or force a higher acquisition price. subject to fads and that investment banks
By starting with a sufficiently high bid, the under-price IPOs to create the appearance of
initial bidder can reduce the likelihood of excess demand. This leads to a high price
this competitor. initially but subsequently underperformance;
Initial Public Offerings (IPOs) have long therefore companies with the highest initial
been known to provide a significant positive returns should have the lowest subsequent
return in the initial days of trading. This returns. There exists evidence of this in the
occurrence directly conflicts with the theory long run.
of market efficiency because the companies Game theory has been extremely useful
should be fairly valued at their IPO and in explaining certain financial decisions.
any return in the initial days should be This paper has only highlighted a few
minimal. In 1986 Rock explained that this of the aspects where behavioral analysis
phenomenon was due to adverse selection has helped explained observed behavior.
Specifically, game theory has helped explain Modigliani, F. and M.H. Miller: The Cost
the reasons companies might choose various of Capital, Corporation Finance and
capital structures and the agency costs the Theory of Investment, American
between managers, equity holders, and Economic Review, 48:261-297
debt holders. In addition, the existence of (June1958).
free rider problems and bidding wars in
corporate acquisitions has been made clear Myers, S.C. and N.S. Majluf: Corporate
through game theory applications. Lastly, Financing and Investment Decisions
IPOs exhibit behavior contrary to the When Firms Have Information Investors
efficient market theory, and game theory can Do Not Have, Journal of Financial
be utilized to help show why this behavior Economics, 13:187-222 (June 1984).
occurs.
Ritter, J. R.: The long-run performance
References of initial public offerings, Journal of
Finance 46:3-27 (1991).
Fishman, M.: A Theory of Pre-Emptive
Takeover Bidding, Rand Journal of Rock, K.: Why new issues are underpriced,
Economics, 19:88-101 (1988). Journal of Financial Economics, 15:187-
212 (1986).
Jensen M.C. and W.H. Meckling: Theory
of the Firm: Managerial Behavior, Shleifer, A. and R. Vishny: Large
Agency Costs and Ownership Structure, Shareholders and Corporate Control,
Journal of Financial Economics, 3:305- Journal of Political Economy, 94:461-488
360 (October 1976). (1986).
Niki Geiersbach
Abstract
With the dawn of globalization, international business is becoming increasingly popular. Multinational
organizations are among the most profitable in the world. A company needs to be aware of the language
and culture of the country where it plans to embark with its investment. Politics and laws of the nation
can either make international business easy or hard. With the success of international business, its future
is gleaming, on a global scale.
1994, an increase of 6.9 percent over that in Italys highly fragmented and sometimes
1993 (1996 National Trade). non-transparent government procurement
As an example of import policies, when practices created obstacles to US firms
Austria, Finland, and Sweden joined the EU participation in Italian government contracts.
on January 1, 1995, these countries adjusted Procurement in certain areas was heavily
their tariffs to the EUs common external directed toward Italian suppliers. In 1994, the
tariff, resulting in increased tariffs on $3 Italian parliament enacted legislation aimed
billion of US industrial and agricultural at providing more transparent procurement
exports. The European Commission was procedures, including establishment of a
required to negotiate with the US and central body to monitor implementation.
other affected trading partners a package Due to its complexity, the bill was not fully
of compensating tariff cuts. During 1995, implemented.
the EU adopted interim compensation for The US cling peach industry complained
the US under which the EU continued to in 1994 that the EU had failed to observe
apply pre-accession tariff levels on imports and enforce a commitment made in the 1985
into the three countries on most of the US-EU Canned Fruit Agreement (CFA) to
affected industrial products, but provided no not subsidize EU processing operations for
compensation in agriculture. In December peaches in syrup. The US industry claimed
1995, the EU and US concluded negotiations implementation of the EUs minimum
on the permanent compensation owed to grower price and fruit withdrawal programs
the US. Some of the concessions were in was undermining the no-processing
the form of acceleration of tariff reductions subsidies commitment made by the EU in the
agreed in the Uruguay Round, while others CFA, and that the sale of subsidized Greek
involved reductions of tariffs beyond levels canned peaches in the US and a number of
agreed in the Round or the establishment of foreign markets, including Japan, Mexico,
tariff-rate quotas. and Canada, was harming the US industry.
EU member states had widely differing The level of software piracy continued
standards, testing and certification to be a source of concern in Germany, as
procedures in place for some products. in other large developed markets. The
These differences served as barriers to the effects of Germanys 1993 implementation
free movement of these products within the of the EUs software directive, as well as
EU and could cause lengthy delays in sales an educational campaign by the software
due to the need to have products tested and industry, might have helped reduce piracy
certified to account for differing national from previous levels.
requirements. Nonetheless, the political US express package services like UPS and
will and the advent of the New approach, Federal Express remained concerned that
which streamlined technical harmonization the prevalence of postal monopolies in many
and the development of standards for EU countries restricted their market access
certain product groups, based on minimum and subjects them to unequal competitive
health and safety requirements, generally conditions. Proposals to liberalize many
pointed toward the harmonization of laws, postal services and to otherwise constrain
regulations, standards, testing, quality and the advantages enjoyed by the monopolies
certification procedures in the EU. The might not be sufficient to fully redress these
European standardization process had been problems.
closed to US firms direct participation.
that plan. It must find the right mix of direct factors needed for lucrative international
and indirect operations, and the right rollout marketing. These factors include knowledge
sequence to keep risk low while maximizing of culture, political risk evaluation, etc.
longer-term market share and revenue
potential. To craft a good business strategy International Services
for international success, the company
must have broadened awareness, coupled International businesses like Shell provide
with business experience. Specialized international services. Shell Services
organizations such as Atlas Venture help International provides, among many other
international businesses formulate the specialized services, electricity in the US.
strategic plan (Atlas Venture). Shell Energy was formed in 1997 to pursue
new growth opportunities in the retail
International Marketing electricity and gas markets that have recently
been opening to competition. Today, Shell
International Marketing Services, Inc, Energy serves more than 300,000 gas and
(IMS) is a uniquely positioned international electricity customers in Ohio and Georgia
marketing firm. Since 1986, it has assisted (Shell Services International).
over five hundred US, European and Russian
companies export products, develop joint International Logistics and
ventures, locate foreign investment, and Supply-Chain Management
form strategic relationships. Its extensive
experience with business practices and International logistics are about
culture in North America, East-Central international freight forwarding, moving
Europe and the C.I.S. is the foundation of and storage, warehousing and storage,
its success. IMS cross-cultural perspective project shipping, office building and letters
allows it to rapidly overcome significant of credit. Currently, there are organizations
obstacles to export sales and joint ventures that specialize in making the task of
that most firms are unequipped to deal with international logistics easier for international
alone. The company also brings broad business (International Logistics
analytical skills to bear on its global projects. Management). A supply chain is a network
This includes substantial international trade of facilities and distribution options that
and joint venture negotiations expertise, performs the functions of procurement of
market and competitive assessments, materials, transformation of these materials
macroeconomic, financial and statistical into intermediate and finished products, and
analyses, and political risk evaluations. In the distribution of these finished products to
addition to broad skills, IMS employees hail consumers. The geographic placement of
from diverse backgrounds and industries. production facilities, stocking points, and
The company offers a full spectrum of sourcing points is the first step in creating
services ranging from one-time primary a supply chain. The strategic decisions
market research assignments to marketing include what products to produce, and which
and strategy consultation where it walks plants to produce them in, allocation of
its clients through each difficult stage of suppliers to plants, etc. Inventory decisions
expanding their business presence abroad refer to means by which inventories are
(International Marketing Services). The managed. And the mode choice aspect of
successful operation of IMS indicates the
transportation decisions is the more strategic tax paid by the Maltese company on income
ones (Ganeshan and Harrison). arising from these foreign holdings. Such
refund is triggered upon a distribution of
Multinational Financial this income to the non-resident shareholders
Management of the Maltese firm (Accounting and
Taxation).
International trade, financing and
investments have grown at an extremely International Human Resource
rapid pace in recent years, and the operations Management
of corporations are increasingly becoming
multinationalized. Corporate executives Effective management of expatriate
buying and selling goods and services, and managers is one of the most important
making financing and investment decisions areas in human resource management, and
across national boundaries, have formulated there is abundant research on expatriates
policies and procedures for managing cash of American, European, and Japanese
flows denominated in foreign currencies. multinationals with large expatriate
These policies and procedures, and the populations. In the age of globalization,
related managerial actions of executives, strategic management of human resources
change as new relevant information is becoming critical for organizational
becomes available and this field is that of survival. Global business environments
multinational financial management (Reid). demand flexibility and rapid response and
there is a growing realization that the human
Countertrade dimension provides the key to flexibility and
adaptability in organizations. Multinational
Countertrade simply refers to listening and transnational corporations have been
to the companys international customers increasingly aware of the growing necessity
and meeting their needs. This could be to have not only international business
in the areas of hard currency generation, strategies, but also international human
technology transfer, or marketing assistance resource strategies. Just as international
(The American Countertrade). business strategy is likely to have unique
features and needs to be understood well,
International Accounting and international management of human
Taxation resources has its own peculiarities that have
to be managed well to be able to flourish in
In Malta, taxation of an international overseas operations (Naresh).
trading company (ITC) is based on the
Tax Refund Mechanism. An ITC is taxed Organization, Implementation,
at the normal company rate of tax, which and Control of International
is currently about 35%. Non-resident Operations, and their Future
shareholders of an ITC are taxed at a flat
rate of 27.5% on all distributions received Effective management of international
from the firm. Non resident shareholders business operations includes efficient
of an international holding company which management of finance, personnel,
has a participating holding in a non-resident product development, marketing, and
company qualify for a full refund of the Malta communication. This is so that the
Ganeshan, Ram, and Harrison, Terry P. 22 Reid, Cynthia. 19 March 2000. Journal of
May 1995. An introduction to supply- Multinational Financial Management. 21
chain management. 21 June 2001. <http:// June 2001. <http://gort.ucsd.edu/>.
silmaril.smeal.psu.edu/>.
Samuelson, Paul A., and Nordhaus, William The American Countertrade. 4 June 2001.
D. 1998. Economics. 16th ed. United The American Countertrade Association.
States: Irwin/McGraw-Hill. 21 June 2001. <www.countertrade.org>.
Shell Services International. February The Brattle Group. 2000. 21 June 2001.
2001. Shell Energy: Ready to Serve <www.brattle.com>.
Texas Electric Customers. 21 June 2001.
<www.shellservices.com>. Wei, Annie. International Business and
Globalization. 21 June 2001. <www.
Sussman, Joshua. 1997. The International lancs.ac.uk/staff/>.
Financial Market, Organized Crime, &
International Law. U.S. Department of
Justice. 21 June 2001. <www.usdoj.gov>.
Isola Oluwabusuyi
Abstract
The purpose of this study was to examine and compare the factors that influence intuition as a decision-
making tool for leaders/managers in Hong Kong and in the United States. This study examined the
relationships among gender, management level, extent of management experience, country of operation,
and the reported use of intuition in decision making. Existing empirical research in this field is sparse. In
this research, attempt was made to contribute to empirical research on the viability and reported use of
intuition as a decision-making skill of leaders.
Agors Intuitive Measurement Survey (AIM) survey was adapted (with permission from copyright
owner) from Weston Agors study to measure the relationship between a managers reported use of
intuition in decision making and the managers management level, his level of management experience,
the managers gender, and the managers country of operation. Each participant was electronically sent
a link that led to a web page containing the survey questions. Once the respondent clicks submit, the
questionnaire was mailed directly to the researcher.
The research shows significant relationship between research variables. Administrative managers in
Hong Kongs reported use of intuition in decision making was significantly lower than US managers
reported use of intuition in decision making. The paper concludes by examining the implications of these
significant findings to global business management and management education.
six MBA students to group the descriptions. (Myers-Briggs Type Indicator). The
Six clusters eventually emerged. instrument, therefore, uses the reliability
In one cluster, authors described and validity of MBTI (Agor, 1984). Studies
intuition as a paranormal phenomenon. Top have found strong support for construct
managers who relied on psychics to make validity, internal consistency, and test
decisions were cited as examples in this related reliability of MBTI instrument
cluster. H. L. Hunt relied on a psychic to (Thompson & Borello, 1986). Further, the
help pick oil properties (Rowen, 1986) instrument was designed to best measure all
A second cluster consists of descriptions the variables in this study. The questionnaire
of Intuition as a personality trait. Scholars was modified and simplified so it contained
in this category saw intuitive decision clear instructions, questions, and possible
making as genetic or acquired very early in answers.
life (Behling & Eckel, 1991). Three of the
remaining four clusters described intuition Research Questions
as an unconscious process, a set of actions,
and distilled experience. The fourth cluster The studys four research questions
was named the residual category. Authors in explored the relationship between the studys
this category defined intuition as decisions independent and dependent variables.
not made using the rational decision making
process. Research Question 1: What is the
relationship between a leaders
Methodology management level and the leaders
reported use of intuition in decision
The quantitative research employed making?
Agors Intuitive Measurement Survey (with
permission by copyright owner). The AIM Research Question 2: What is the
survey was administered to 100 participants relationship between a managers
from the US and 100 participants from work experience as a manager and the
Hong Kong. Questions on the survey were managers reported use of intuition in
developed to measure all independent and decision making?
dependent variables.
Questions on the survey measured the Research Question 3: What is the
following variables: Gender, Management relationship between a managers sex
Level, Management Experience, Country of and a managers reported use of intuition
Operation, and Reported use of intuition in in decision making?
decision making. Reported use of intuition
in decision making was the only dependent Research Question 4: What is the
variable in the study, the remaining four relationship between a managers
variables were independent variables. country of operation and a managers
reported use of intuition in decision
Validity and Reliability of the making?
Instrument
respectively. This result is significant at the intuition in decision making. Mean scores
.020 level. of US female managers with between 6
Additional Finding 21: Female and 10 years management experience and
managers in Hong Kong with 5 years or US male managers with 5 years or less
less management experiences reported use management experience are 7.238 and 9.5
of intuition in decision making is lower respectively. This result is significant at the
than US male managers with 5 years or less .010 level.
management experiences reported use of Additional Finding 25: Female managers
intuition in decision making. Mean scores of in Hong Kong with between 11 and 20
Female managers in Hong Kong with 5 years years management experiences reported
or less management experience and US male use of intuition in decision making is lower
managers with 5 years or less management than US male managers with 5 years or
experience are 7.1 and 9.5 respectively. This less management experiences reported
result is significant at the .010 level. use of intuition in decision making. Mean
Additional Finding 22: US female scores of Female managers in Hong Kong
managers with 5 years or less management with between 11 and 20 years management
experiences reported use of intuition in experience and US male managers with 5
decision making is lower than US male years or less management experience are
managers with 5 years or less management 6.333 and 9.5 respectively. This result is
experiences reported use of intuition in significant at the .010 level.
decision making. Mean scores of US female Additional Finding 26: US female
managers with 5 years or less management managers with between 11 and 20 years
experience and US male managers with 5 management experiences reported use of
years or less management experience are intuition in decision making is lower than
6.35 and 9.5 respectively. This result is US male managers with 5 years or less
significant at the .010 level. management experiences reported use of
Additional Finding 23: Female intuition in decision making. Mean scores of
managers in Hong Kong with between 6 and US female managers with between 11 and 20
10 years management experiences reported years management experience and US male
use of intuition in decision making is lower managers with 5 years or less management
than US male managers with 5 years or experience are 7 and 9.5 respectively. This
less management experiences reported result is significant at the .010 level.
use of intuition in decision making. Mean Additional Finding 27: Female
scores of Female managers in Hong Kong managers in Hong Kong with more than 20
with between 6 and 10 years management years management experiences reported
experience and US male managers with 5 use of intuition in decision making is lower
years or less management experience are than US male managers with 5 years or less
6.380 and 9.5 respectively. This result is management experiences reported use of
significant at the .010 level. intuition in decision making. Mean scores
Additional Finding 24: US female of Female managers in Hong Kong with
managers with between 6 and 10 years more than 20 years management experience
management experiences reported use of and US male managers with 5 years or less
intuition in decision making is lower than management experience are 6.875 and 9.5
US male managers with 5 years or less respectively. This result is significant at the
management experiences reported use of .010 level.
Downey, L. (2006) Female intuition Robbins, S. P., & Judge T.A. (2007)
and emotional intelligence linked to Organizational behavior (12th ed.).
management success Swinburne industry Upper Saddle River, New Delhi: Prentice
solutions Retrieved December, 2006 from Hall of India.
http://www.swinburne.edu.au/corporate/
Abstract
A major success criteria for the marketing department of the company is the understanding of the
consumers and the success of the campaigns. Based on the outputs of certain data mining tasks, a marketer
would be able to chalk out his strategy. But the major barrier against proper utilization of business
intelligence is that often the marketer is unable to see the actual benefits the system can provide him
in meeting his objective. Firstly this article provides a basic knowledge of the various data-mining tasks
which can create value to the marketer. A major concern becomes, even if the marketer is convinced of
the benefits to be obtained, he is unaware of the process involved to get it done. The data requirements
are huge for a business intelligence system. This article also provides some insights on what kind of data
requirements the data-mining system would require to perform such business intelligence tasks. Thirdly,
this article maps how business intelligence can be collaborated with marketing strategy to create value in
each stage of the product development life cycle following Kotlers product life cycle framework. The
entire paper provides these conceptualizations based on theoretical understanding of the areas under
discussion.
Kumar Kar A., Kumar Pani A., Kumar De S. - A Study on Using Business Intelligence for Improving Marketing Efforts
142 Business Intelligence Journal July
customer needs and hence designing more The data-mining tasks which are used
suitable promotions. for the purpose are clustering, classification,
It has been established by Webster (1988) pattern association, summarization,
and Gronroos (1990) that marketing can no predictive modeling, link analysis and social
longer be an area of the marketing specialists network analysis. These tools have been
but that everyone in the organization classified under query tools, descriptive
must be charged with responsibility for statistics, visualization tools, regression type
customers and contributing to developing models, association rules, decision trees,
and delivering value for them. According case based reasoning, genetic algorithms
to Webster (1992), this customer focus may and graph theory.
require increasingly large investments in In this paper, a brief introduction has
information management and information been provided on how the following data
technology. The next frontier of automation mining tasks can help to reduce marketing
will be marketing and sales functions as expenditure and increase the take up rate of
was argued by Moriarty and Swartz (1989). campaigns. While there are multiple studies
Brooks (1989) argued that these functions in the area of each of these tasks in data
are likely to receive the largest investment mining, there is no study which dictates how
of technological resources in the future. It these tasks can be used effectively by the
is also evident that many companies have marketer at different stages of the product
started, or are planning to build, marketing life cycle, based on which, promotion
related IS. It is essential now to study and strategies are actually taken. This paper
classify these systems in more detail so as strives to bridge this gap.
to be able to design better and more cost- In the following part, a brief description
effective IS in marketing for the future. of the major tasks of data mining is provided
There are various data mining or business with the description of their possible
intelligence techniques which are used in application for a marketer.
marketing as a tool to reduce marketing
expenditure and increase the take up rate 2.1 Cluster analysis / Clustering
of campaigns. These techniques are used
individually or are clubbed to do certain Clustering or cluster analysis is the
tasks. The challenge for every customer- process of grouping the data into classes or
oriented organization consists of identifying clusters, so that objects within a cluster have
potential customers and satisfying high similarity in comparison to one another
and retaining existing customers. This but are very dissimilar to objects in other
necessitates a detailed understanding of the clusters (Han and Kamber, 2006). Clustering
peoples needs and expectations. Adequately is a method of unsupervised learning, and
addressing these needs and at the right time a common technique for statistical data
is crucial to grow and maintain a long-lasting analysis. Data clustering algorithms may be
and mutually profitable relationship. In hierarchical which find successive clusters
applying data mining methods to marketing using previously established clusters. These
problems, there are several critical issues algorithms can be either agglomerative (also
within the knowledge discovery process, called bottom-up) or divisive (also called
from Business Understanding over Data top-down). Agglomerative algorithms begin
Preparation and Modeling to Deployment of with each element as a separate cluster and
the model in a marketing environment. merge them into successively larger clusters.
Kumar Kar A., Kumar Pani A., Kumar De S. - A Study on Using Business Intelligence for Improving Marketing Efforts
144 Business Intelligence Journal July
Divisive algorithms begin with the whole to specific classes of customers which the
set and proceed to divide it into successively latter would be able to relate to better, and
smaller clusters. Partition algorithms thus in the process, increase the take up
typically determine all clusters at once, but rate and the success of such advertisement
can also be used as divisive algorithms in campaigns.
the hierarchical clustering. Density-based
clustering algorithms are devised to discover 2.3 Pattern association
arbitrary-shaped clusters. In this approach, a
cluster is regarded as a region in which the From the early 1970s, pattern association
density of data objects exceeds a threshold. was primarily studied in the context of
Clustering may typically be used in market research model development. Pattern
marketing for advertisements when the association is used to predict patterns based
customers are yet to be segmented. After on past patterns or data sequences on which
running a cluster analysis, the clusters the tool has been trained upon. Perhaps the
may be examined for characteristics based most common precursors to the exploration
on which advertisement campaigns may of the associative relationship between two
be directed at the customer base. After variables in marketing involve the use of
segmentation, based on the characteristics bivariate cross-tabulations or multivariate
of the clusters, product positioning, product analysis (DeSarbo and Hildebrand 1980;
repositioning and product development may Green 1978; Perreault and Barksdale 1980).
be done, to improve its fit with the targeted Several alternative measures are available
customers. Cluster analysis may also be for assessing the extent of association
done to selecting test markets. in a contingency table. Current pattern
association studies in data mining started
2.2 Classification developing from 1993. Agrawal, Imielinski
and Swami (1993); Holsheimer, Kersten and
A pattern classification problem is Mannila (1995); Houtsma and Swami (1995)
essentially mapping an input pattern, studied association in mining approaches
represented as an input vector, to a particular and are highly cited works.
class or category. Thus given a database Pattern association may be extensively
D={t1,t2,,tn} and a set of classes used to predict customer preferences
C={C1,,Cm}, the classification problem when very little data about the customer
is to define a mapping f:DC where each is available to the marketer. Tools for
ti is assigned to one class (Dunham, 2006). pattern association would help a marketer
Traditionally classification (Duda, Hart to predict which product or advertisement
and Stork, 2001) has been studied using the customer may be interested in solely by
Bayesian decision theory and parameter the current buying behavior of the customer
estimation, non-parametric techniques, and matching it with the buying behavior
linear discriminant functions, multi-layer of similar customers (who bought similar
neural networks, stochastic methods and products) even when no information is
non-metric methods. available for the customer.
Pattern classification may be used
efficiently in marketing and advertisements
by first segmenting the entire customer base
and then delivering selective advertisements
Kumar Kar A., Kumar Pani A., Kumar De S. - A Study on Using Business Intelligence for Improving Marketing Efforts
146 Business Intelligence Journal July
benefits and perceived benefits of a product, covering all aspects of the product life cycle.
and thus help the marketer to reposition his As is evident in the following matrix,
product. Again, if nothing is known about various tasks can help to provide various
the customer, just on the basis of mining benefits for the company based on the stage
his immediate shopping patterns, it is of the product in the product development
possible to predict future purchases, through life cycle. The tasks can be used to identify
association rule mining, and then create an the segments for a focused targeting and
immediate campaign based on generated also predict which customers belong to
rules, thus increasing sales by cross-selling which segment, thus automating the entire
of multiple products. process. Mining data can help a company
The list of tasks and outputs are not position and reposition their products based
comprehensible, but the matrix given in on the needs of their customer. Also the
the following part gives a good indication tasks may help to improve the products
of the data-mining tasks and their business and even extend the product line by better
impact. The matrix identifies which task can understanding of customers needs. The
have what type of impact at which stage of tasks may also be used to identify which
the product life cycle. The proposed outputs customers should actually be targeted by
have been developed based on the suggested campaigns to derive higher returns on the
classical strategies for the different stages huge marketing expenditure made by the
of the product life cycle which can be firm. Also, it may help to identify potential
impacted by information processing, as churners and help prevent loss of revenue
has been mentioned in Kotler and Kelly for the client and also help to reduce cost
(2006). The same has been chosen due to the of customer acquisition by better targeted
comprehensive nature of the framework in campaigning.
Table 2. Mapping of Tasks, Product life cycle and utility to the marketer.
Link analysis Identify innovators, Identify early Identify influencers & Identify churners & Identify churners
early adopters adopters & laggards laggards
influencers
Predictive modeling Identify innovators, Identify early dentify profitable & loyal Identify loyal customers,
early adopters adopters & customers Calculation of CLTV
influencers
Summarization Identify early adopters Identify new Identify profitable Identify profitable
customer customer segment, customer segment,
segments Calculation of CLTV Calculation of CLTV
Pattern association Identify potential Identify potential Identify potential Identify possible
associated associated purchases associated purchases product line
purchases diversification
Classification Push products to target Push products to Push products to target Push products to target Push products to
segments target segments segments segments target segments
Clustering Identify segments and Identify product Identify product Reposition based on Reposition based
size, market needs extensions extensions segment needs if needed on segment needs
if needed
Kumar Kar A., Kumar Pani A., Kumar De S. - A Study on Using Business Intelligence for Improving Marketing Efforts
148 Business Intelligence Journal July
Haeckel SH, Nolan RH (1993). Managing Moriarty RT, Swartz GS (1989). Automation
by wire. Harvard Business Review. pp. to boost sales and marketing. Harvard
122 132. Business Review. 67(1): 100-108.
Kumar Kar A., Kumar Pani A., Kumar De S. - A Study on Using Business Intelligence for Improving Marketing Efforts
150 Business Intelligence Journal July
Abstract
This study examines the relationship between corporate governance and the extent of voluntary
disclosure in listed firms of Tehran Stock Exchange.The corporate governance attributes are the percentage
of independent directors on the board, the existence of dominant personalities (CEO/Chairman duality)
and the percentage of institutional investors. The research sample was selected among listed companies in
Tehran Stock Exchange during 2001-2005 and totally is consist of 106 observations. In this research, the
size and the type of audit institute (Audit organization or other institutes) are used as control variables.
The research results show that there is no significant relationship between percentage of independent
directors on the board, CEO/Chairman duality and the extent of voluntary disclosure. Also the research
results provide significant evidence on relationship between the percentage of institutional investors and
the extent of voluntary disclosure.
Khodadadi V., Khazami S., Aflatooni A. - The Effect of Corporate Governance Structure on the Extent of Voluntary Disclosure in Iran
152 Business Intelligence Journal July
executive directors and firms family factors of culture and corporate governance
control with financial disclosure in Hong and extent of disclosure. This study
Kong. Results showed that proportion of included several hypotheses and variables
the independent directors is directly related among which they tested only six variables
to financial disclosure. This relationship as corporate governance variables. These
is weaker in firms which are under family variables include proportion of independent
control. Thus, researchers stated that the directors on the boars, proportion of family
presence of independent non-executive members on the board, joint position
directors in board composition would of Chairman and CEO, non-executive
enhance the probability of compliance with chairperson, proportion of board of directors
disclosure requirements by firm and may with cross-directorships and chairperson
result in information transparency. with cross directorships. In their research,
Ho and Wong (2001) analyzed the they used firm-specific characteristics (size,
relationship between corporate governance leverage, profitability, industry type, auditor
structures and the extent of voluntary type, listing status) as control variables.
disclosure in Hong Kong Stock Exchange. Results from this survey only refer to
They represented four major corporate the existence of a significant relationship
governance attributes. These corporate between two variables of non-executive
governance attributes are the proportion of directors on the board and proportion of
independent non-executive directors on the family members on the board with firms
board, the existence of an audit committee, extent of voluntary disclosure both of which,
the existence of dominant personalities of course, have negative relationship with
(CEO/Chairman duality), and the proportion dependent variable.
of family members on the board. The results Lakhal (2003) analyzed the relationship
indicate that the existence of an auditor between voluntary disclosure by directors of
is significantly and positively related to French firms and the corporate governance
the extent of voluntary disclosure, while characteristics. Results indicated that the
the proportion of family members on the ownership structure is rather dispersed and
board is negatively related to the extent of in case where non-executive directors have
voluntary disclosure. more share in board, the level of disclosure
But, statistical tests showed that two other increases voluntarily. In firms where the
independent variables, i.e. the proportion of CEO holds the position of the chairperson
independent directors on the board, which as well, voluntary disclosure would be less
was expected to be directly associated with likely. There is an insignificant and weak
voluntary disclosure and the joint position relationship between non-executive board
of members and decisions about voluntary
Chairman of the Board (Chairman) and disclosure and, also board size and disclosure
Chief Executive Officer (CEO) held by a decisions. A hypothesis of this research
single individual which was expected to be indicating the negative relationship between
negatively associated with disclosure, have board size and voluntary disclosure was not
no significant relationship with disclosure. confirmed.
In a survey titled Culture, Corporate Eng and Mack (2003) analyzed the
governance and Disclosure in Malaysian relationship between corporate governance
Corporates, Hanifa and Cooke (2002) and voluntary disclosure in Singapore. In this
investigated the relationship between two investigation, corporate governance testing
Khodadadi V., Khazami S., Aflatooni A. - The Effect of Corporate Governance Structure on the Extent of Voluntary Disclosure in Iran
154 Business Intelligence Journal July
is examined in two aspects of ownership pay and board size as have been considered
structure and the board composition as independent vaiables of research. Among
where ownership structure by means of control variables of this research one can
the proportion of managerial ownership, mention firm size, situation in stock market,
blockholder ownership and government industry type and shares dispersion. Results
ownership and board composition are showed that proportion of independent
measured with the criteria of the proportion directors on the board and existence of
of independent directors. Control variables audit committtee in firm and, also, board
used in this research include firm size, participation in the capital of the company
leverage, industry type, reputation of as well as the approval of stock option
audit committee. Results of statistical plans as directors pay are directly related to
tests showed that as much as managerial disclosure.
ownership becomes lower and government Chen and Courtenay (2006) tested the
ownership higher, the voluntary disclosure relationship between board supervision
of information is increased. It means that and the extent of voluntary disclosure. In
government ownership enhances the moral this study, the relationship among three
hazard and agency problems and disclosure independent variables, i.e. proportion of
is utilized to reduce these problems. Results independent directors on the board, board
of this research showed that blockholder size and, also, joint position of Chairman and
ownership has a significant relationship CEO are analyzed on the dependent variable
with the extent of disclosure and as much (voluntary disclosure). Reuslts showed
as the proportion of independent directors that proportion of independent directors
increases the voluntary disclosure is is directly associated with the extent of
decreased. voluntary disclosure and the board size and
Gul and Leung (2004) examined and the joint position of Chairman and CEO
tested the relationship between board also have no relationship with voluntary
structure and voluntary disclosure. For disclosure. They found that existence of
the same purpose, two factors of the joint a corporate governance mechanism and
position of CEO and the chairperson and environmental rules would increase the
the proportion of independent directors power of the relationship between proportion
on the board (as determinant of board of non-executive directors on the board and
structure) were studied with using voluntary the extent of voluntary disclosure.
disclosure. Results showed that the joint Arcay and Vazquez (2005) investigated
position of CEO and the chairperson is the relationship of firm characteristics, rules
negatively and weakly associated with of corporate governance and the extent of
inormation voluntary disclosure. voluntary disclosure in Spain. Proportion
Arcay and Vazquez (2005) investigated of independent directors on the board,
the relationship of firm characteristics, rules existence of audit committee, splitting the
of corporate governance and the extent of responsibilities of the Chairman and CEO,
voluntary disclosure in Spain. Proportion board participation in the capital of the
of independent directors on the board, company, stock option plans as directors
existence of audit committee, splitting the pay and board size as have been considered
responsibilities of the Chairman and CEO, as independent vaiables of research. Among
board participation in the capital of the control variables of this research one can
company, stock option plans as directors mention firm size, situation in stock market,
Khodadadi V., Khazami S., Aflatooni A. - The Effect of Corporate Governance Structure on the Extent of Voluntary Disclosure in Iran
156 Business Intelligence Journal July
senior levels of firm to be sufficiently assured agency problem and their existence in the
of the balance of power and the liabilities of position of sharehloder result in separation
board members. The importance of splitting of ownership and control.
the responsibilities of the Chairman and There are two thinking schools about
CEO of England companies has been also institutional investors in firm. The first
emphasized in report by Higz. Higz declared says that institutional investors have short-
that with considering the recommendations term viewpoint. This short-term horizon
of Cadbury report and until the time of his inhibits institutional investors from taking
report, about 90% of firms listed in Stock supervisory expenses because the benefits
Exchange had splitted the responsibilities of such supervision will be unlikely won
of the Chairman and CEO. This proves the by them in a short period of time. The other
permanent impact of Cadbury report on thinking school states that when firm shares
England companies. ownership is concentrated among a few
number of investors (especially institutional
3.3. Institutional Investors Role in investors); the problems concerning the
Corporate Governance separation of ownership and control will
be decreased. When the proportion of
Numerous groups have influence on institutions ownership increases, firm
corporate governance. Among these, leaving will take more expense because
shareholders, especially institutional major shares sales usually require major
investors, play important role. In theoretical discounts (Black and Koffee, 1994).
view, the position of institutional investors Institutional investors tend to have
in corporate governance is very complicated. communication with senior managers of
From this point of view, institutional companies existing in their portfolio and
investors reveal another mechanism of to participate in behind closed doors
powerful corporate governance which could supervisory activities (Ramsay et al.,
have supervision over firm management. 2000; Stapledon, 1996). As a result of
Because they can either considerably affect their supervisory activities, these investors
firm management or coordinate interests have a better understanding of conditions
of shareholders groups. Of course, in influencing on corporates performance
literatures about corporate governance, the and it is less likely to make the managers
ownership focus is referred to as an important of companies existing in their portfolio be
mechanism to control agencys problems subjected to fine for their low profit not
and to improve investors interest supporting caused by their weak management.
(Shleifer and Vishny, 1997). However, such
focus may have negative impacts as well, 3.4. The extent of Disclosure
e.g. access to confidential information may
result in information asymmetry between Financial disclosure is an abstract
them and smaller shareholders. concept not measurable directly which
As a result of their benefiting from have no certain feature to be used for
economic advantages and variability, easily measurment of its intensity or
institutional investors can resolve agency quality (Wallace and Naser, 1995) But,
problems. Therefore, it seems that disclosure is simply defined as transfering
institutional investors, as firms shareholders, and presentation of economic information
reveal both cause and resolution/solution of in terms of financial and non-financial,
quantitative or other forms of information range from very low to very high. The
related with firms financial status and next step is performed like that of first aspect
operation. This disclosure, when becomes and here the only difference is that obtained
necessary by a law enacting and regulating scores are multiplied by the importance
source, is called involuntary disclosure and coefficients extracted from different
when information disclosure is not affected individuals comments and the weighed
by certain rules it is considered as optional criteria is applied. Singhvi and Desai (1971)
disclosure (Owusu-Ansah, 1998). and Barrett (1977) have employed this
Also, disclosure/extent of disclosure aspect in their surveys. First aspect has been
has very important and determining role in used for two reasons as follows:
theoretical and empirical areas. This concept
is so broad and extensive that consists of a. Ashton (1974) believes that individuals
all issues and topics on financial reporting. have little knowledge and information
A defined and accurate comprehension of about their decision making and
disclosure quality requires the assessment of judgement. Therefor, the need to make
restrictions governing thereon and achieving decision about the importance of each
such comprehension would undoubtedly itme is removed and different individuals
help accountant in drawing a framework to mistake in weighing is less likely.
improve disclosure problems of accounting
in the future (Buzby 1974), decrease the b. Using this method allows researcher to
heterogeneous disclosure among informed do measurements independently from
and uninformed investors (Bushman and different individuals viewpoints. When
Smith, 2001) and non-disclosure will result using weighed aspect, it is possible to
in market inefficiency and wrong pricing of take different people personal opinions
firms shares (Diamond and Verrecchina, into consideration and, in consequence,
1991; Kim and Verrecchina, 1994). In the pertinent criteria will not be
general, two aspects have been suggested independent from the viewpoint of
in measuring voluntary disclosure level in certain individuals and/or groups.
researches:
3.4.1. Using non-weighed index In this respect, Owusu-Ansah (1998)
In this aspect, a list of disclosed cases is says that using non-weighed aspect is
prepared and, then, after financial statements preferred over weighed aspect. Also, Cooke
and financial reports have been studied, the expresses that there is no difference in the
values of one, zero and not applied results of weighed and non-weighed criteria.
will be given for disclosure, non-disclosure Therefore, with studying and analysing of
or not qualified for disclosure, respectively. the considered voluntary disclosure cases in
Researchers like Wallace (1987); Cooke similar researches, a list of disclosures cases
(1989, 1991, and 1992), Ahmed and Nicols including 31 items are extracted as follows:
(1994) and Owusu-Ansah (1998) utilized 3. Brief history of company
this aspect in their researches. 4. Names of boards members
3.4.2. Using weighed index 5. Names of staff managers
In second aspect, firstly the list of 6. Names of block holder
disclosure cases is extracted and the 7. Firms strategies and its possible effects
importance of each item is identified with 8. Firms major markets
using different individuals comments and a 9. Forecasted sales
Khodadadi V., Khazami S., Aflatooni A. - The Effect of Corporate Governance Structure on the Extent of Voluntary Disclosure in Iran
158 Business Intelligence Journal July
Khodadadi V., Khazami S., Aflatooni A. - The Effect of Corporate Governance Structure on the Extent of Voluntary Disclosure in Iran
160 Business Intelligence Journal July
Table 1. The number of firms in disclosure rank firm-years; also, the audit organization was
the auditor. Meanwhile, the number of firms
in which the position of chairperson and
Very high
disclosure
voluntary
Very low
Extent of
Medium
CEO is separated has been increased over
Total
High
Low
time (from 8 to 20 firms). As time passed, the
2001 23 23 17 19 24 106
number of firms whose auditor was the audit
2002 21 27 17 21 20 106
organization has been declined (from 66 to
2003 21 26 19 20 20 106
34 firms) and other institutions play further
2004 21 22 27 15 21 106
role in auditing of sample firms. In Table
2005 23 21 18 24 20 106
3 descriptive statistics pertaining to other
Total 109 119 98 99 105 530
research variables have been illustrated.
As seen, over the period of time, 109 firm- Table 3: Descriptive statistics
years with very low voluntary disclosure, 119
firm-years with low voluntary disclosure, 98 % of non-
% of
institutional
firm-years with medium level of voluntary executive
investors
Firm Size
directors
disclosure, 99 firm-years with high voluntary ownership
disclosure, 105 firm-years with very high Mean 0.65 0.47 5.41
voluntary disclosure were observed. You can Median 0.60 0.51 5.38
study the number of firms in disclosure ranks Max. 0.83 0.95 7.16
from 2001 to 2005 in Table (1). The number Min. 0 0 4.04
of firms with considering CEO/chairperson Std. Dev 0.19 0.33 0.54
duality and the type of auditor which audited
the firms are as shown in Table 2. The average percentage of non-executive
directors, percentage of institutional
Table 2: The number of firms with considering CEO/ investors ownership and firm size equal to
chairperson duality and the type of audit committee 65%, 47% and 5.41, respectively. Maximum
percentage of non-executive directors
and institutional investors ownership and
Variable
CEO/chairperson
Auditor
duality maximum firm size equal to 83%, 95%
and 7.16, respectively, and the minimum
percentage of non-executive directors
organization
Very high
Extent of
Very low
Medium
High
% of
Yes 96 99 84 82 93 Extent of % of non-
institutional
voluntary executive
investors'
2.48 disclosure directors
ownership
duality
No 13 20 14 17 12 % of non-
-0.02
executive 1
(0.69)
directors
organization
% of
31 48 51 65 40 institutional 0.15 0.01
1
Others Audit
ownership
Firm Size 0.06 -0.05 0.02
78 71 47 34 65 (0.13) (0.25) (0.71)
* significance at 1% level
* significance at 1% level
6.3. Estimation Results
Spearman ranked correlation coefficients
among variables of the extent of voluntary We estimate model (1) by ordered logistic
disclosure, percentage of independent method for testing the research hypotheses
directors, percentage of institutional and for taking into account the control
investors ownership and firm size are variables of auditor type and firm size.
presented in Table 54. Estimation results of model (1) are provided
Results shown in Table 5 indicate that in Table 6.
only the Spearmans ranked correlation Estimation results of model (1) show
coefficient between extent of voluntary that the coefficient of the variable of the
disclosure and percentage of institutional percentage of independent directors (-0.07)
and the coefficient of the variable of CEO
3
Regression analysis is the best method to taking control variables and chairperson duality (-0.17) are not
into consideration and this has been discussed in section (6.3).
significant, this means that the first and
4
Since the variable of voluntary disclosure level is of rank type,
Spearman Correlation Coefficients have been applied instead of the second hypotheses are rejected. The
Pearson Correlation Coefficients. coefficient of the variable of the percentage
Khodadadi V., Khazami S., Aflatooni A. - The Effect of Corporate Governance Structure on the Extent of Voluntary Disclosure in Iran
162 Business Intelligence Journal July
of institutional investors ownership equals theory and this supervisory tool has little
to 0.78 which is significant at 1% level. role in improved financial disclosure.
Therefore, our third hypothesis of research Beside, in most literature published
is not rejected as a result of significant concerning corporate governance, it
relationship existing between the percentage has been emphasized on splitting the
of institutional investors and the extent of responsibilities of the chairperson and CEO.
voluntary disclosure. In addition, results In most Iranian firms the positoin of CEO
show that as the percentage of institutional and the chairperson has been separated and
investors ownership increases the extent of usually two (and not one) individuals are in
disclosure increases, too. the charge of the said positions. However,
though duties and liabilities of CEO and
Table 6: Estimation results of model (1) chairperson have not been well separated, it
was already anticipated that the said variable
makes little effect on voluntary disclosure.
Significance
Coefficient
Z- statistic
Arcay, M. and Vazquez, M (2005) Corporate Chen, C. and Jaggi, B (2000) Association
Characteristics Governance Rules and Between Independent Non-Executive
The Extent of Voluntary Disclosure in Directors, Family control and Financial
Spain. Advances in Accounting, vol.21, Disclosures in Hong Kong, Journal of
299-331. Accounting and Public Policy, Vol. 19,
285-310.
Archambault, J.J, and Archambault
E.M (2003) A multinational test of Chen, C. and Courtenay, M (2006) Board
determinants of corporate disclosure, Composition, Regulatory Regime and
The International Journal of Accounting, Voluntary Disclosure. The International
Vol.38, 173-194. Journal of Accounting, Vol. 41, 262-289.
Khodadadi V., Khazami S., Aflatooni A. - The Effect of Corporate Governance Structure on the Extent of Voluntary Disclosure in Iran
164 Business Intelligence Journal July
We are an international network of scholars who share similar interests in aspects of social
responsibility. Currently we have about 500 members and membership is free.
Network officers:
Chair of the Network: Professor Dr. David Crowther, De Montfort University, Leicester Business
School, The Gateway, Leicester LE1 9BH, UK davideacrowther@aol.com
Vice Chair: Professor Dr. Gler Aras, Yildiz Technical University, Institute of Social Science,
Yildiz Besiktas 34349, Istanbul, TURKEY guleraras@aol.com
What do we do?
Conferences
2008 7th conference CSR and SMEs
Durham, UK
2009 8th conference CSR and NGOs
Pretoria, South Africa
2010 9th conference CSR and Global Governance
Zagreb, Croatia
2011 10th conference CSR and the New Economy
New Orleans, USA
Publications
Social Responsibility Journal
The official refereed journal of the Network; published 4 times per year by Emerald.
Discussion Papers in Social Responsibility
An opportunity for early publication of articles. Published when necessary by SRRNet.
The Newsletter
Published 3 times per year and containing news and opinion pieces. Sent to all members.
Research Book Series: Issues in Corporate Behaviour and Sustainability
Books published in association with the conferences and given to all conference delegates.
Full details of all of our activities can be found from our website www.socialresponsibility.biz
If you share our aims then please join us. We look forward to hearing from you.
Isles Internationale Universit
School of Doctoral Studies
(European Union)
Approved by Charter of The Ministry of Education of the British Isles to act as a chartered
University outside of the United Kingdom; with Full Accreditation granted from the Acadmie
Europenne dInformatisation; established in Brussels, Belgium by Order of the King of
Belgium Albert II, with full recognition from The Ministry of Justice and Research of the
Belgian Crown:
The Isles Internationale Universit (European Union) has been commended to host the School of Doctoral Studies (EU)
in Brussels, Belgium, aiming to accomplish three fundamental missions:
Development and enhancement of elite doctoral studies programmes, with cutting edge standards on academic and
scientific research;
Enforcement of EU analogue quality standards on academic programmes developed by tuition institutions outside the
EU area: (a) By evaluating academic methodology applied on learning programmes; (b) By providing full coaching
and tutoring support to tuition institutions worldwide to upgrade, assure and maintain academic methodology quality
levels on their way towards excellence; and (3) By awarding EUASC Seal (EU Analogue Standards Certification) at
corresponding quality levels achieved at each evaluation point; and
Ensuring academic EU Analogue Standard by performing: (a) Academic Validation on degrees earned by students at
recognised and/or certified tuition institutions located outside the EU area; (b) Double Degree awarding on degrees
earned by students at recognised and/or certified tuition institutions located elsewhere the EU area; and (c) Degrees
awarding on studies programmes developed by tuition institutions, once certification at higher quality levels on applied
academic methodology has been achieved and after collaboration agreement has been executed for these purposes .
In order to achieve its missions, the Isles Internationale Universit has gathered some of the best minds in Europe, who
have developed sate-of-the-art doctoral academic programmes, elite research methodology and cutting edge technological
tools, and who act as permanent Faculty Members to strictly enforce this toolkits proper application on daily basis.
Doctoral Studies
The School of Doctoral Studies of the EUs academic structure includes four Departments (Business Management and
Economics, Engineering and Technology, Science and Social Science) which host 37 Disciplines, offering PhD studies on
practically every main field of human knowledge.
Over 355 PhD students are involved in more than 116 cutting edge research projects, most of them being currently
developed in collaboration with 12 other universities in the EU area and elsewhere; 94% of these students are engaged on
programmes designed to undertake pure research assignment and 6% are required to undertake a research assignment and
pursue theoretical studies in the form of seminars or courses.
Studies towards a doctoral degree are worth 240 higher education credits (ECTS credits) and require an average of four
years of full-time study. The research is intended to lead to a scholarly thesis; writing it will take up most of a students
time and all theses are publicly defended. Every doctoral student receives a studies grant for partial or total coverage on
full programmes term costs, as well as individual tutoring. Currently, slightly over 72% of all programmes full term
costs are covered by studies grants.
Science students spend a great deal of time in the laboratory. Some departments may require that the thesis be part of an
ongoing project within the department. In the fields of technology and natural science, researchers often work as part of a
team. If research findings are reproduced in academic journals the thesis may be a compilation of the published articles.
The Universidad del Valle de Mxico (UVM) is one of the largest and most prestigious universities in
Mexico. Founded in 1960 and accredited by the Federacion de Instituciones Mexicanas Particulares
de Educacion Superior, UVM enrolls students at 32 campuses throughout Mexico.
High School
Undergraduate
Undergraduate for working adults
Graduate
Continuing Education
UVM alumni are distinguished by their abilities, knowledge, attitudes, and social skills
that are shaped by the identity subjects.
Spain, Switzerland, France, Costa Rica, Panama, Honduras, Ecuador, Chile, Peru, Brazil, England,
Chipre, China, Canada, US, Germany and Mexico.
UVM has 35 campuses throughout Mexico offering 38 undergraduate degree programs in Arts
and Humanities, Social Sciences, Economic and Management Sciences, and Engineering; 11
undergraduate degree programs for working adults and 28 graduate programs. UVM has more than
100,000 students and more than 9,000 employees (teachers and staff).
Recognitions:
Through its history, UVM has received recognitions that prove its excellence:
Academic Excellence - Secretara de Educacin Pblica (SEP).
According to the Readers Digest Intelligent Decision Markers (IDM) Guia Universitaria 2008,
UVM is one of the best upper education institutions in the country, from a range of over 100
public and private universities.
Also, more than 400 teachers of the 35 campuses took 12 seminars in areas such as: Industrial
Engineering, Mechatronic, Animation, Marketing, Management, Communication, Law, Psychology,
Health Sciences and Hospitality.
140 academic leaders of the 6 regions in which UVM has classified its campuses in 15 states of
Mexico and Mexico City, are enrolled in the Diplomat in Leadership for Academic Management.
Premio UVM has adapted the YouthActionNet Global Fellowship model to provide a tailor-
made, culturally-relevant, and Spanish-language centered leadership development experience for 15
young Mexican leaders, ages 18-29, annually. The Premio UVM fellowship strengthens the project
management and communications skills of young Mexican leaders, while connecting them to their
peers and experts to create a national network of youth leaders affecting positive change.
Student Development
Responsible of the education that students receive, UVM sets special emphasis on the students
integral education and the continuous improvement of its faculty. This way, it responds to the
expectations and trust of Mexican families and prepares good successful professionals with a global
vision, who will acquire the skills and knowledge that the labor market requires.
The student development area at UVM has important national and international projects as: