You are on page 1of 13

THE COMPARISON OF VALUE AT RISK ON SHARIA

BASED STOCK AND NON-SHARIA BASED STOCK


(Case study on Indonesia Capital Market during 2005 – 2008)
Agung D. Buchdadi
M. Yasser Arafat
Tri Hesti Utaminingtyas
(Faculty of Economics State University of Jakarta)

This study aims to compare between maximum losses that could be arising to sharia
based stock and non-sharia based stock investment due to the differences characteristic
of these stocks. As the consequence of Conditional Variance Phenomenon, this study will
use EWMA model and GARCH model. The finding show that EWMA model provides
better performance and gives more conservative of VAR value than GARCH model. It is
also shown that Sharia based stock is riskier during February to March and May to July.
Finally, this study try to give information on risk characteristic in Indonesia’s Capital
Market.
Keyword: Value at Risk, EWMA model, GARCH Model, Indonesia Capital Market

Electronic copy available at: http://ssrn.com/abstract=1711714


BACKGROUND
Stock price fluctuation in Indonesia during 2005 – 2008 was shown so highly. After
giving a profit more than 20% until 2007, the return of stocks plumped down in 2008 and
it provides huge loss to most investors. One way to avoid it is by doing some risk
management. By monitoring the risk in investment, people could hedge the value of the
assets and create the optimum portfolio equal to their ability in absorbing the risk.
Therefore, measurement of risk is considered important in this regard. Yet, only few
people have an adequate knowledge in risk management. Central Bank of Indonesia also
supports the implementation of risk management by adopting the regulation to banks in
Indonesia in 2008. This legislation led to the development of the concept of VAR in the
thriving banking institutions. However, it does not happen in non-banking institutions
Fardiansyah (2006) states the Value at Risk (VAR), one of risk measurement
method, is very popular and used widely by the financial industry worldwide. Volatility,
normally assumed to be fixed along the time, is needed in calculating VAR. Then, Some
studies in Indonesia show a normal distribution and the unconditional variance
assumption is less appropriate to be applied in financial market movements. Situngkir
and Surya (2006) find it is better to calculate VAR with considering on skewness and
kurtosis value. Pohan (2004) indicates that stock mutual funds returns during 2001 –
2002 do not follow a normal distribution and also finds heteroscedastic phenomenon on
it. Karahap (2005) also finds similar results to Pohan’s on foreign currency portfolios.
Another interesting thing in Indonesia is the rapid growth of sharia-based capital
markets. This is probably only plausible since the largest Muslim population lives in
Indonesia. Nevertheless, sourced from Pramesti (2005), derived from Karim Business
Consulting (2003), it is found that about 75% potential capital market investors are
floating market loyalist. In other words, these investors spend the money in Islamic
capital markets when there are potential benefits. Knowing that sharia based stocks has
special characteristic, in this paper; we try to observe the risk faced in sharia based stock.
The following section of the paper is organized as follows. In section 2, we
provide a brief literature review concerning to Value at Risk Analysis. Section 3
describes the methodology of this paper. Section 4 describes the results. And section 5
describes the conclusion and remarks for the future research.
LITERATURE REVIEW
VAR definition taken in this paper is cited from Jorion (2001) which indicates VAR as a
measurement of the worst expected loss over a given horizon at a given level of
confidence. Mathematically the formula of VAR = (α ⋅ σ − µ ) ⋅ W which asset value is W,
return of the asset follows normally distribution with mean µ and standard deviation of
the return is σ. Mean value could be omitted under assumption the window period of
observation is short enough, daily. Then, the formula can be simplified as
VAR = (α ⋅ σ ) ⋅ W . Value of α corresponds to the level confidence chosen. For example,
α = 1.65 is used for 95% level of confidence. Similarly to Bodie (2002), when the asset
does not follow normally distribution, variable α’ replacing standar variable of α is used
1
and determined with Cornish Fisher Expansion, which α ' = α − (α 2 − 1)ξ . Variable ξ
6
refers to skewness coefficient of the asset.

Electronic copy available at: http://ssrn.com/abstract=1711714


Dealing with conditional variance phenomena that the variance of the asset is not
constant over the period, it is proposed to use moving average, GARCH and Risk metric
methodes.
Exponentially Weighted Moving Average (EWMA)
As it is noted on Jorion (2001), The model calculates standard deviation at date t as a
function of return of the asset on previous date t-k. Then, The RiskMetric model suggests
using decay factor to calculate the deviation. The factor shows how the impact of the
previous volatility to the next forecasted deviation. The formula is:
σ t2 = λ ⋅ σ t −1 + (1 − λ ) ⋅ rt −1 ……………(1)
The value of λ is between 0 to 1 and the higher the value the longer time needed
for the volatility decays to its mean value. The value of λ is determined by obtaining the
best value root mean squared error (RMSE), which produces the smallest error between
the value of random variables and volatility at the same time. The formula used
1 T 2
RMSE = ∑
T t =1
(rT +1 − σ T2+1|T (λ )) 2 ………………..(2)

where :

σ T2+1|T = (λ ⋅ F (t /(t − 1)) + (1 − λ ) X t )

F(t/t-1) = forecasted value of variance at t-1

Generalized Autoregressive Conditional Heteroscedasticity (GARCH)


Quoted from Alexander (2001) it is noted that GARCH model, which is initially
developed by Engle (1982) and Bollerslev (1986), is calculated using the following
formula,
p q
σ t2 = α 0 + ∑ α i rt 2−1 + ∑ β iσ t2−1 …………..(3)
i=0 i =0

Which :
α0>0, α1,….., αp, β1, …., βq ≥ 0

α0 = slope βq = lag volatility coefficient

αp = error coefficient at time p σt-1 = deviation at time t-1

rt-1 = return at time t-1


The plain model of GARCH (1,1) is commonly used in calculating volatility since
it is very ease in the implementation. Figures (1.1) shows only 1 error and the
coefficient of lag 1 coefficient used in determining the volatility or mathematically:
σ t2 = α 0 + α ⋅ rt 2−1 + β ⋅ σ t2−1 ………………….(4)
The greater α value the more reactive the current volatility to the previous market
information yields. The greater β value the more influence of the previous volatility on
current volatility. In this model the parameters α and β together specify the persistence
level of the volatility of the assets. Whether the sum of value α and β is greater, The time
of the estimated volatility affected by the price movement needed is longer.
GARCH models is relatively better in compared to EWMA model in several
factor, i.e : Firstly, the model is more responsive any price movement in data series. In
addition, the GARCH model has greater average volatility value and need more time to
back toward the average value after exposed by stock price instability which is more fit to
actual condition in financial market.
Comparing the both model, which

EWMA : σ t2 = λσ t2−1 + (1 − λ )rt 2−1 , and

GARCH (1,1) : σ t2 = α 0 + α ⋅ rt 2−1 + β ⋅ σ t2−1 ;

It is found that β and α in GARCH model are similar to λ and (1- λ) in EWMA model
respectively. Then, it seems that EWMA model is a GARCH model which α + β = 1 and
αo = 1.
Previous Research
Sourced from Tarigan (2006), it is noted several studies that compares EWMA and
GARCH model. First, Chan and Károly (1991) on the Japanese stock market data from
1977 to 1970 years indicate EWMA method is better than the GARCH methods, although
the data is fit (suitable) with GARCH form. Kuen and Hoong (1992) find, for the
Singapore stock market, EWMA methods are more suitable than GARCH. Karahap
(2005) concluded, for calculating risk in the Indonesian foreign echange, it is more
appropriate to use GARCH (1.1). Furthermore, Buchdadi, et all (2008), for the optimum
portfolio in LQ45 in Indonesia, find EWMA model is more suitable than GARCH model.
DATA
Data needed is stock return which is transformed from daily stock price obtained from
Indonesia Stock Exchange (BEI). Window period chosen is 2005 – 2008 since it was
noted very high volatility in stock price movement. It is expected to give a
comprehensive point of view to the characteristic on investment in Indonesia. Sharia
based stocks which meet Islamic law in business are selected whether it is included on
Jakarta Islamic Index (JII) during 2005 – 2008. Business in or related to drug,
prostitution, alcohol, tobacco, and conventional bank which used interest rate on the
business are prohibited in Islamic law. . Then, non-sharia stocks are defined to those
which are broke Islamic law in business and included in LQ45, the most liquid and the
best ferformance in BEI, during the window period. Based on the availability on data,
there are 8 stocks in each sharia and non-sharia based stocks. Sharia based stocks consist
of 3 mining company, 2 manufacture companies, 1 pharmacy company, 1 in
telecommunication industry, and 1 in retail industry. While, non-sharia based stocks are
consist of 2 in cigarette industry and 6 in financial industry.
METHODOLOGY
Two models, EWMA and GARCH, are used in this study. For EWMA model, I try to
find the best value of λ by using formula 2. I also use the value of λ = 0.94 which is
proposed by RiskMetric in calculating VAR on daily price basis. In addition, for GARCH
model, I develop ARMA model by trial and error to construct the most parsimonious
model which has conditional variance on its residual. Then VAR is calculated for 5%
confidence interval, on daily basis, for each Rp 1,-.
In this research, I validate the model using Kupiec test, Performance test based
on proportions of failures (TNoF). In this model, the number of volatility which excesses
the VAR value is counted as an error or an overshoot. The probability of N errors follow
the binomial process according to the formula:
Binomial (T , N ) = (1 − p ) T − N p N

Then, Likely Ratio (LR) test used to examine confidence interval of the model. So, LR

were p=p* :

N
T −N N
 N  N  T − N 
LR ( N , p*) = −2 log((1 − p*) ( p*) ) + 2 log  1 −  
 T  T  

The proportion of errors have a chi-squared distribution with degrees of freedom = 1.


RESULT AND DISCUSION

Descriptive Statistics

The closing price of each company is transformed to daily return. In summary, the data
obtained is 8 shares for each category are presented in Table 1.
Table 1. Descriptive statistics
Non Syariah RMBA PNLF GGRM BBRI BNII BNGA BMRI BCA

Mean 0.001681 -0.000478 -0.00106 0.000615 0.000803 0.000136 0.000156 0.000984

Median 0 0 0 0 0 0 0 0

Maximum 0.133531 0.270429 0.210156 0.173663 0.338774 0.156003 0.182322 0.126752

Minimum -0.208755 -0.225083 -0.14009 -0.11583 -0.41616 -0.16252 -0.13858 -0.10622

Std. Dev. 0.031388 0.039423 0.028485 0.030488 0.033899 0.031123 0.031381 0.024148

Skewness -0.230907 0.392686 1.197933 0.283874 -0.61133 0.069138 0.346446 0.033017

Kurtosis 7.965779 9.918144 14.51909 6.358873 37.19246 7.10992 7.583734 6.232783

Jarque-Bera 1020.798 1989.599 5681.386 476.2626 48044.16 694.0377 882.015 429.1007

Probability 0 0 0 0 0 0 0 0

Observations 985 985 985 985 985 985 985 985

Alpha 1.710491 1.53323 1.304332 1.56416 1.818629 1.625201 1.546374 1.635468

Syariah UNVR UNTR TLKM PTBA KLBF INTP BUMI BNBR

Mean 0.000971 0.000709 0.000429 0.001646 -0.00027 0.000355 0.000213 -0.00141

Median 0 0 0 0 0 0 0 0

Maximum 0.179048 0.182322 0.116921 0.182322 0.421728 0.127262 0.202783 0.693147

Minimum -0.13431 -0.24454 -0.10451 -0.28768 -0.10725 -0.16779 -0.38612 -0.52452

Std. Dev. 0.022359 0.03434 0.023394 0.037329 0.029618 0.029344 0.037164 0.049276

Skewness 0.516435 -0.02867 0.106609 -0.5259 3.688357 -0.36332 -0.86614 1.638121

Kurtosis 10.93416 9.764071 5.819395 11.71636 50.54169 5.97268 17.74295 57.21302

Jarque-Bera 2627.396 1877.9 328.1056 3163.542 94996.19 384.3485 9043.753 121064.1

Probability 0 0 0 0 0 0 0 0

Observations 985 985 985 985 985 985 985 985

Alpha 1.498053 1.653004 1.614549 1.794343 0.596411 1.74813 1.89106 1.179206

Source: Elaborated data

From Table 1, it can be seen all samples are not normally distribution abnormal indicated
by probability values for Jargue-Bera test of 0%. Thus, the value of α must be evaluated
using Cornish Fisher Expansion which also has been presented in the table. It also known
that the return of RMBA has the largest average of return. In contrast, in average GGRM
provide the smallest return to the investor. While, the most fluctuative return comes from
PNLF.
VAR Analysis
Formation of Conditional Variance Model
By trial and error, conditional Variance models are developed and be tested using Park
test to see if there are conditional on the residual variance model. The model presented in
the following table:
Table 2 Conditional Variance Model
Saham Conditional Variance Saham Conditional
NS Model Syariah Variance Model
BBRI AR(2), MA(2) BUMI AR(2), MA(2)
BCA AR(23), MA(23) INTP AR(31), MA(31)
BMRI AR(1) BNBR AR(14), MA(14)
RMBA AR(1) KLBF AR(18), MA(18)
BNGA AR(29), MA(29) PTBA AR(16), MA(16)
BNII AR(1), MA(1) TLKM AR(22), MA(22)
PNLF AR(2), MA(2) UNTR AR(26), MA(26)
GGRM AR(11), MA(11) UNVR AR(1), MA(1)
Source: Elaborated data

From the table it is known that the whole model is then in heteroskedastis. So, standard
deviations (σ) is calculated by EWMA and GARCH approach.
EWMA
The results of calculations are presented in the following table:
Table 3. EWMA Model
Non Syariah λ maksimum Overshoot Overshoot λ =0.94
BBRI 0.54 91* 143
BCA 0.55 84* 144
BMRI 0.53 97* 138
RMBA 0.54 84* 116*
BNGA 0.49 88* 152
BNII 0.60 80* 135
PNLF 0.52 114* 153
GGRM 0.50 127 195
Average 0.53
Syariah
BNBR 0.46 141 161
BUMI 0.48 45* 114*
INTP 0.56 79* 134
KLBF 0.39 313 320
PTBA 0.58 63* 124
TLKM 0.54 88* 155
UNTR 0.59 77* 129
UNVR 0.49 93* 144
Average 0.51
Source: Elaboorated data, ** = 5% confidence interval *= 10% confidence interval
From the table it is found that the EWMA models by calculating the optimum RSME has
a number of overshoot smaller than using the EWMA model with λ = 0.94. The results
also show that most of EWMA by calculating optimum RSME pass in 5% of confidence
interval. The results of this study differ from what is found by Buchdadi, et al (2008) that
indicate EWMA model with λ = 0.94 gives a better numer of overshoot. Perhaps it is
because of the differences on the object and methodology. On Buchdadi, et al (2008) the
optimum portfolio is developed by Markowitz’s method, while this paper determines the
EWMA value of each individual stock. By constructing the optimum portfolio, the
distribution follows normally shape which is fit to the assumption of EWMA model
Furthermore, in average the value of λ of non-sharia stock is greater than sharia
stock. It means non-sharia stock is more persistence to previous price volatility. Perhaps
it is related to behavior of Islamic investor since speculative trading in capital market is
prohibited upon Islamic law in business. However, it need further research to confirm the
hypothesize.
GARCH (1.1)
The results of calculation using GARCH (1,1) are presented in the following table:

Table 4.GARCH (1,1)


Non Syariah Koefisien α Koefisien β Overshoot
BBRI 0.0716 0.9142 75*
BCA 0.1048 0.8556 87*
BMRI 0.2838 0.5492 137
RMBA 0.0748 0.8809 78*
BNGA 0.1991 0.7280 95*
BNII 0.3022 0.5641 86*
PNLF 0.1987 0.5844 164
GGRM 0.2569 0.6535 145
Average 0.1865 0.7162
Syariah
BNBR 0.1087 0.4798 246
BUMI 0.0601 0.9403 30**
INTP 0.0984 0.8445 85*
KLBF 0.6949 0.5385 258
PTBA 0.2832 0.7242 44*
TLKM 0.0953 0.8788 77*
UNTR 0.1484 0.8267 60*
UNVR 0.1544 0.6491 148
Average 0.2054 0.7352
Source: Elaboorated data, ** = 5% confidence interval *= 10% confidence interval

Unlike the EWMA method, the results using GARCH (1,1) for sharia stocks give the
average value of β, which shows the persistence level of a model, greater than the
average value of β for non-sharia based stock. The results also show the performance of
the GARCH model, which is shown by the average number of overshoots, is not better
than EWMA model. However, it is premature to conclude that the persistence level of
stock depend on the measurement method. So, one more time we propose further study to
make the confirmation to this study results.
Comparison between EWMA and GARCH model
To compare both model the previous calculation are put down on the following table:

Table 5. Comparison of the result of VAR Calculation


Non VAR VAR Overshoot Overshoot
Syariah EWMA GARCH (1,1) EWMA GARCH (1,1)

BBRI 0.0397 0.0386 91* 75*

BCA 0.0327 0.0303 84* 87*

BMRI 0.0393 0.0319 97* 137


RMBA 0.0434 0.0393 84* 78*

BNGA 0.0421 0.0369 88* 95*

BNII 0.0440 0.0396 80* 86*


PNLF 0.0483 0.0341 114* 164

GGRM 0.0277 0.0246 127 145

Average 0.0397 0.0344 96 108

Syariah
BNBR 0.0419 0.0187 141 246

BUMI 0.0535 0.0591 45* 30**

INTP 0.0428 0.0369 79* 85*


KLBF 0.0130 0.0158 313 258

PTBA 0.0500 0.0541 63* 44*

TLKM 0.0319 0.0303 88* 77*


UNTR 0.0437 0.0443 77* 60*

UNVR 0.0276 0.0188 93* 148

Average 0.0381 0.0348 96 108


Source: Elaboorated data, ** = 5% confidence interval *= 10% confidence interval

From the table it is shown that VAR of EWMA model is greater that VAR of GARCH
(1,1) model. Non-sharia stocks consistently provide greater VAR value when EWMA
model is used. Yet, inconsistent results shown in sharia based stock since there are 4
stock has bigger value of VAR when GARCH (1,1) model is used. However, Based on
the number of overshoot, EWMA model produce better in confidence level interval. The
results of this study, then, supports conducted Buchdadi, et al (2008), Chan and Karoly
(1991), and Kuen and Hoong (1992). Moreover, it can be said that VAR value of non-
sharia stock is greater than VAR value of sharia stock. In other word, potential loss in
non-sharia stock investment is bigger than investment on sharia stock.
In addition, this study write down the graphs depicting average VAR for stock
and non-Islamic sharia along the year.

Figure 1 VAR Value for Each Month

-0.025

-0.03

-0.035 NS EWMA
S EWMA
B u lan

-0.04
NS GARCH
-0.045 S GARCH
-0.05

-0.055
J ul

Nov
J an

J un
M ar
A pr

Oct
Feb

A ug
S ep
M ay

Dec

VAR

From the figure 1, it is known in February the average value of VAR reaches a
minimum value while the maximum value is happened in October. There is a possibility
that this is related to the January Effect in which there are abnormal yields in January. In
addition, comparing both model in calculating VAR, it is riskier to invest on sharia stock
on February to March and on May to July. The rest of months are the time to invest on
non-sharia stock. However, these finding still need further research yo validate the
phenomenon. For example, it needs study about the relationship between time to spend
the money on sharia stock and the condition of the market. Perhaps, in bearish market it
is good to invest on sharia stock. Or, it is actually in contrast?
CONCLUSIONS AND RECOMMEDATIONS
This study gives an overview of the usage of several methods of calculating VAR in
Indonesian stock market. The method is discussed primarily EWMA and GARCH (1,1)
model as the consequence of conditional variance phenomenon found. The results show
the EWMA model give better performance than GARCH (1,1) model. EWMA model
also shows greater value of VAR and gives more conservative on risk measurement. It
means for a conservative investor EWMA model is fit to be implemented in managing
the risk. Then, under the assumption the EWMA model is better, it is noted the
investment in non-sharia based stock is riskier than sharia based stock
Recommendations
Several recommendations arise due to some limitations on this study
1. Trial and error method in developing Conditional Variance can make a non
optimum model. So, it is recommended to use a computational software to find
the optimum model.
2. Stress testing as it is suggested on Jorion (2001) should be implemented in the
next research.
3. For the investor, by knowing the VAR value of the asset the optimum portfolio
that is fit to the risk characteristic can be made through risk budgeting process that
releases the stock when the price movement is so volatile.
4. For Indonesian Capital Market Regulator, since it is important to give a clear
information on investment risk, we propose to make a regulation to disclose the
VAR value of the stock.

REFERENCES

Alexander, Carol, 2001, Market Model: A Guide to Financial Data Analysis , West
Sussex, John Wiley & Sons Ltd.

Alexander, Carol; Sheedy elizabeth; and Koenig, David R., 2004, The Professional
Risk Manager's Handbook , PRMIA

Bodie, Zvi; Kane, Alex; and Marcus, Alan J., 2002, Investment , NewYork,
McGraw-Hill
Buchdadi, AD, 2008, Penghitungan Value at Risk Portofolio Saham Perusahaan
Berbasis Syariah dengan Pendekatan EWMA (Studi Empiris Terhadap
Saham - Saham yang Tergabung Dalam JII Selama 2005 – 2006), Jurnal
Akuntansi Keuangan Indonesia , Vol. 5 No. 2, Departement Akuntansi
FEUI, Jakarta
Buchdadi, AD; Ahmad, GN; Mardiyati, U.; 2008, Analisis Value at Risk Portofolio
Saham LQ45 dengan pendekatan EWMA dan GARCH selama Tahun 2002 –
2006, Laporan Penelitian Dosen Muda , Dirjen Dikti Departemen
Pendidikan Nasional, Jakarta

Fardiansyah, Teddy, 2006, Penerapan Manajemen Risiko Perbankan Indonesia ,


PT Elex Media Komputindo, Jakarta

Febrian, E. and Herwany, A., 2009, Volatility Forecasting Models and Market Co-
Integration: A Study on South-East Asian Markets, Indonesian Capital
Market Review , Vol. 1, No. 1, no. 1, Faculty of Economics UI, Depok, West
Java, Indonesia

Jorion, Philippe, 2001, Value at Risk , 2 nd ed., McGraw-Hill, New York

Jorion, Philippe, 2005, Financial Risk Manager Handbook , John Wiley and Sons
Inc., New York
Kurniawan, T. and Aziz, HA, 2007, Modelling the Volatility of Shari'ah Index:
Evidence from the Kuala Lumpur Shariah Index (KLSI) and the Jakarta
Islamic Index (JII), Proceeding on The International Conference on
Islamic Capital Markets , IRTI-IDB, Jakarta

Karahap, Andi R., 2005, Perhitungan Value at Risk (VAR)-Foreign Exchange Risk
Menggunakan Pendekatan EWMA, GARCH, dan Monte Carlo Simulation,
Karya Akhir , MMUI, Jakarta

Pohan, Daulat HH, 2004,Estimasi Volatilitas Return Reksadana Saham Sebagai


Pertimbangan Keputusan Investasi (Perbandingan Model EWMA dan
GARCH), Karya Akhir , MMUI, Jakarta

Pramesti, Muthia, 2005, Perbandingan Investasi pada Instrument Syari'ah dan


Optimalisasi Portofolio, Karya Akhir , MMUI, Jakarta

Purnomo, Slamet E., 2005,Perhitungan Value at Risk Obligasi Fixed Income dengan
Menggunakan Pendekatan RiskMetics, Karya Akhir , MMUI, Jakarta

Riano K.,Andi, 2005,Perhitungan Value at Risk-Foreign Exchange Risk


Menggunakan Pendekatan EWMA, GARCH, dan monte carlo Simulation
(Studi Kasus Bank X), Karya Akhir , MMUI, Jakarta

Situngkir H. dan Surya Y. 2006, VAR yang Memperhatikan Sifat Statistika


Distribusi Return , Bandung FE Institute, Bandung

Sugiyono, Prof., Dr.; 2008, Metode Penelitian Bisnis , Penerbit Alfabeta, Bandung

Tarigan, Rudi H., 2006, Peramalan Imbal Hasil Saham Sektor Tekstil dan Garmen di
Bursa Efek Jakarta Dengan Menggunakan Model ARCH/GARCH, Karya
Akhir , MMUI, Jakarta

Tri Jatmiko, Fajar, 2006, Pengukuran Value at Risk Risiko Nilai Tukar dengan
Estimasi Volatilitas EWMA dan GARCH (studi kasus pt bank pqr), Karya
Akhir , MMUI, Jakarta

Watsham, Terry J.; and Parramore, Keith; 1997, Quantitative Methods in Finance ,
1 st ed., Thomson Learning

Winarno, Wing W.; 2007, Analisi Ekonometrika dan Statistika dengan E-views ,
edisi pertama, Unit Penerbit dan Percetakan STIE YKPN, Yogyakarta
www.bi.go.id, , November 2008
www.finance.yahoo.com , Maret 2009

You might also like