STOCK DUAL-LISTING CAUSALITY OF NYSE AND IDX (CASE STUDY: PT TELKOM )

Arrozaq Ave1, 1Mercubuana University, Indonesia

Abstract The objective of this research is to create an interdependence model between Telkom stock listed in IDX () and NYSE (New York Stock Exchange). Telkom is one of the Owned-State Enterprise becoming a role model of an Indonesian company listing in two different countries. Listing abroad is supported fully by Indonesian authority since the exhibition of the credibility of the Indonesia company to the world while introducing them as well. The analytical methodology used in this thesis is the VAR / VECM model showing the interdependence between Telkom’s stock in NYSE and BEI. Research data purposively gained in 10 years period between 2008 to 2018 that listed in NYSE with TLK code and IDX with TLKM code. After a series of tests and the data is proceed to Granger analysis and VAR / VECM modeling using e-Views. The results are the equation model of Telkom stock return in NYSE and BEI. Furthermore, it has been concluded that TLK influences TKM but not vice versa so that so-called by direct causality. Keywords: Stock Return, Telkom, NYSE, BEI, Granger, VAR / VECM.

Telkom Indonesia as one of the big Stock PENDAHULUAN FOREWORD Issuers in the stock market trade its stock through two stock markets, BEI and NYSE. The capital market is a public place for With the dual-listing of this stock, Telkom has a people having a buy-sell transaction of financial beneficiary to gain fresh money as company instrument futures like stock, obligation, and interest. Telkom also obtains a credible image various other securities. Capital market in and accountable international company due to Indonesia named IDX (Indonesia Stock the reporting of accountancy management fulfill Exchange). IDX is a place for Stock Issuer to international standard (IFRS - International showcase its stock and sell it to potential investors. Financing Reporting Standard). It is interesting Practically, there is no regulation stating to discuss since Telkom as this journal is written the issuance of stock must be done only in one right now (2019) is the only one of the state country. But several Indonesia companies are company (BUMN) that listed in the USA stock known to list their stock in two different market after pulled out. The selection of countries like Telkom, Indosat, and Antam. Telkom as our case study because of its Telkom and Indosat had their stock listed in performance in the USA and big impact in IHSG Indonesia (IDX) and the USA (NYSE) while compared with the Antam. Telkom also is called Antam listed in the Australia stock market. In the blue-chip stock which is used as an example for future, Antam is planning to list in the USA stock IDX to push other Indonesia's Stock Issuers to market. It is due to the probability of reaching follow Telkom's step in dual-listing in two thousands of billion-rupiah capitalization and different stock markets. greater access to potential big investors that can As informed before Telkom has listed in impact future business development. Nowadays, two different stock markets, the investor needs Telkom is the only Indonesia company that still to know the interdependence of those stocks. listed in NYSE while Indosat had decided to pull This information must take consider by investors out from the USA stock market in 2013. The considering the interdependence relation reason behind this delisting is minimizing between stock price in NYSE and IDX. Stock administrative costs. Indosat also say, investor price comparison used in NYSE and IDX is on 1 still can buy Indosat stock through IDX. Market January 2013 until 31 December 2018. stock analyst from PT Remax Capital, Lucky One of the ways to know the Bayu Purnomo said Indosat's decision to delist in interdependence of price in two stock markets can the USA stock market implies that the financial be done by the Granger analysis approach. performance of the company is not liquid Several past researches observed the correlation anymore. He said the financial performance of between the group of stock in IDX like LQ45 and Indosat is illiquid as last year's performance sharia stocks. But it is not yet to be found the although Indosat denies it by telling the reason for delisting is efficiency matter. 1 research that especially talks about the company selection used in the form of VAR or VECM. that lists with dual-listing. This test is done by looking at the maximum lag In observing the interdependence between that is used to determine the optimal lag length two stocks in the different markets, it needs a used. Granger causality will be used as a testing model to figure interdependence between those tool to see the influence between TLK and variables. The determined observation duration is TLKM stock data. This test can also see the 10 years with the observation resolution reaching direction of the causality of the stock. In the the closing weekly share price announced every VAR test, modeling will be obtained containing Friday afternoon. This period is selected in such significant constants and coefficients that a way to provide information and exhibit a more determine the value of TLK shares and future informative trend movement. Based on the above TLKM shares. explanation, therefore the selected research topic is Dual-stock Causality in NYSE and BEI. Based on that matter, then the formulation of the problems that must be answered in this study are: 1. Is there any reciprocal interdependence between stock prices in IDX and NYSE? Figure 1. Framework 2. How is the form of causality between that stock price? 3. How are modeling to show future intercourse The population taken in this study were all of Telkom stock of IDX and NYSE? Telkom shares from the IPO (Initial Public The scope of the problem of this study is Offering) for the first time on the IDX and concerned with the observation variable from the selling their shares on the NYSE. The sample research object which is in the year 2008 – 2018 used in this study is secondary data consisting (10-year period). The object under study only of: limited at the stock price movement without 1. Time series data for the monthly 2008-2018 including other factors. Due to the stock price Telkom shares on the NYSE movement itself that reflects the influence of 2. Time series data for the 2008-2018 monthly macroeconomics and company performance period Telkom shares on the IDX8 itself. 3. Information about the company and the Intention of this research are: industry obtained from various articles in 1. To find the independence between reciprocal magazines, newspapers, the internet, and of stock price of TLKM in IDX and NYSE. other publications 2. To find the direction of causality between The sampling technique is done by Telkom stock in IDX and NYSE purposive sampling method. This sample 3. To find model about future interdependence contains 574 weekly shares on the NYSE, 574 between Telkom price in IDX and NYSE. weekly shares on the IDX and 574 exchange Specifically, this paper will do: (1) root rates of Dollars against the Rupiah. test unit testing of time series data of TLKM and TLK; (2) Testing of granger causality of these Empirical Method two-time series; (3) Get a model of Vector Auto Regression (VAR) for these two-time series. As already explained, Vector Auto Regression (VAR) is an analytical tool not only METHOD useful observing the causality relationship between variables but can also be used to Based on the formulation of the problem determine the projection model. To understand described previously, it can be described as the VAR analysis, the model is empirically framework of thought according to the following illustrated as follows: picture. Two TLK and TLKM stock data taken TLKt = a10 + a11TLKt-1 + a12TLKMt-1 + (i) in weekly resolutions are used to observe their a13TLKt-2 + a14TLKMt-2 + eyt behavior and causality. With a series of statistical tests before, a unit root test and TLKMt = A20 + a21TLKt-1 + a22TLKMt-1 (ii) cointegrate test is performed to see the test + a23TLKt-2 + a24TLKMt-2 + ezt

2

푛 푠 Where: 2. If ∑푗=1 푏푗 = 0 and ∑푗=1 푑푗 ≠ 0 there will be one- TLKt = TLK at week t way causality from X to Y. 푛 푠 TLKMt = TLKM at week t 3. If ∑푗=1 푏푗 = 0 and ∑푗=1 푑푗 = 0, hence X and Y free from each other. TLKt-n = TLK at week t-n 푛 푠 4. If ∑푗=1 푏푗 ≠ 0 and ∑푗=1 푑푗 ≠ 0, there will be TLKMt-n = TLKM at week t-n two-way causality from Y and X. a , a = Constant 10 20 eyt, ezt = Error factor Granger's causality on the two stocks can be seen by comparing the F-statistics with the The two equations above show TLK and critical value of the F-table at the various TLKM influence each other. confidence levels (1%, 5%, or 10%). It also can be seen from comparing the probability value Analytical Method with the confidence levels (1%, 5%, or 10%). If all variables have F-statistic values greater than In order to test the causality between F-table values at a significant level, then both of TLK and TLKM, several stages of testing these variables have bidirectional causality. analysis are as follows, where all stages use The criteria for rejecting and accepting a Eviews 10 software. hypothesis in the Granger causality test are as follows: 1. Unit Root Test 1. F-stat:> F-table = H0 is rejected and Ha is This unit root test is used to see whether accepted. the data observed is stationary or no. This test is 2. F-stat:

푋푡 = ∑ 푎푖푋푡−1 + ∑ 푏푗푌푡−1 + µ푡 푖=1 푗=1 Root Unit Testing 푟 푠 With the help of the Eviews 10 program, 푌푡 = ∑ 푐푖푋푡−1 + ∑ 푑푗푌푡−1 + 푣푡 an Augmented Dickey-Fuller Test performs a Unit 푖=1 푗=1 Root Test to see the stationary of TLK and TLKM variables. By covering trends and intercepts, the X_t = Stock Price of Telkom in IDX series of TLK stock trial results are t-statistic with Y_t = Stock Price of Telkom in IDX a value of -2.59 and a probability of 0.953 while u, v = Error Term TLKM is -0.72 and a probability of 0.8373. Both Based on regression model of Granger data are not stationary at the level but the first Causality, hence the hypothesizes to be tested difference. Since the first difference test is are: performed, it turns out that the First Difference is stationery with the 0 probability as proof (smaller ∑푛 ∑푠 1. If 푗=1 푏푗 ≠ 0 and 푗=1 푑푗 = 0 there will be one- than 0.05). Addition this is also shown by the ADF way causality from Y to X. t-statistic value that is smaller than the McKinnon 3 t-statistic value at Alpha 5%, <-1,949856. cointegration at a significant level of 1% and 5%. Therefore, the cointegration test indicates that Granger Causality Testing. between the movement of TLK and TLKM there Furthermore, based on the result of the is no correlation of stability and similarity of Granger causality, it is known that those who have movement in long term relationships. a causality relationship are those who have a smaller probability value than alpha 0.05. With VAR Test the inversely reading, In the granger test in figure 3, it is known the reciprocal relationship that When forming the VAR model, it must be TLKM is significantly affected TLK but TLK is determined in advance about how many lags suits not significantly affected by TLKM. Therefore, with the model. To determine them, several H1 is accepted and another hypothesis is rejected. criteria are used based on the value of the Akaike Information Criterion test and Schwarz Cointegration Test Information Criterion which resulted in minimum The purpose of the cointegration test in value. Identification of VAR and VECM model this research is to determine whether the group of also are performed by using Final prediction Error variables is interrelated or not. Based on the above value, Akaike Information Criterion (AIC), figure, trace-statistics and eigenvalue's maximum Schwarz Information Criterion (SC) and HQ value at r=0 are smaller than the critical value (Hannan-Quinn Information Criterion) which has with significance level 1% and 5%. This means a a little value and LR is the greatest (Wei, 2006). null hypothesis stating no cointegration is From the above result, lag 2nd has the most accepted and the alternative hypothesis stating the symbol (*). (*) is lag order chosen by criterions, existence of cointegration is rejected. Based on all variables in this model affect each other not the econometric analysis above, it can understand only in this period but those variables related to that between the two variables, there is one each other until the previous 2 periods.

Date: 10/14/19 Time: 16:40 Sample (adjusted): 1/22/2008 12/25/2018 Included observations: 571 after adjustments Trend assumption: No deterministic trend Series: TLKM TLK Lags interval (in first differences): 1 to 2

Unrestricted Cointegration Rank Test (Trace)

Hypothesized Trace 0.05 No. of CE(s) Eigenvalue Statistic Critical Value Prob.**

None 0.006932 4.459214 12.32090 0.6445 At most 1 0.000853 0.487106 4.129906 0.5483

Trace test indicates no cointegration at the 0.05 level * denotes rejection of the hypothesis at the 0.05 level **MacKinnon-Haug-Michelis (1999) p-values

Figure 2. Cointegration Test Result

VAR Lag Order Selection Criteria Endogenous variabels: TLK TLKM Exogenous variabels: C Date: 10/08/19 Time: 15:37 Sample: 1/01/2008 12/25/2018 Included observations: 566

Lag LogL LR FPE AIC SC HQ

0 -6623.079 NA 50345500 23.41017 23.42550 23.41616 1 -4238.502 4743.875 11186.78 14.99824 15.04423 15.01619 2 -4225.004 26.75819 10817.56 14.96468 15.04133* 14.99460* 3 -4219.730 10.41756 10768.98 14.96018 15.06749 15.00206 4

4 -4215.418 8.486287 10757.16* 14.95908* 15.09705 15.01293 5 -4211.823 7.051740 10772.59 14.96050 15.12914 15.02632 6 -4211.271 1.078178 10904.72 14.97269 15.17199 15.05047 7 -4209.284 3.869121 10982.65 14.97980 15.20976 15.06955 8 -4202.284 13.57978* 10866.95 14.96920 15.22982 15.07092

* indicates lag order selected by the criterion LR: sequential modified LR test statistic (each test at 5% level) FPE: Final prediction error AIC: Akaike information criterion SC: Schwarz information criterion HQ: Hannan-Quinn information criterion

Figure 3. Optimal Lag Length Test Pairwise Granger Causality Tests Date: 10/14/19 Time: 16:51 Sample: 1/01/2008 12/25/2018 Lags: 3

Null Hypothesis: Obs F-Statistic Prob.

TLK does not Granger Cause TLKM 571 7.43590 7.E -05 TLKM does not Granger Cause TLK 2.21367 0.0855

Figure 4. Granger Causality Test Result

The previous cointegration test has been explains and confirms the relationship given to performed on two existing variables namely the movement of dual listing stocks. VAR TLK and TLKM resulted in negative results and selection based on no proof of cointegration test it is justified to use the VAR method in that has been done before in TLKM and TLK regression. Regression with the VAR model stock data sets. The results obtained in figure 5.

Vector Autoregression Estimates Date: 10/14/19 Time: 17:06 Sample (adjusted): 1/15/2008 12/25/2018 Included observations: 572 after adjustments Standard errors in ( ) & t-statistics in [ ]

TLK TLKM

TLK( -1) 0.984685 13.74230 (0.04170) (2.96905) [ 23.6159] [ 4.62852]

TLK(-2) -0.001842 -14.05078 (0.04152) (2.95629) [-0.04438] [-4.75285]

TLKM(-1) -0.000347 0.889149 (0.00058) (0.04096) [-0.60318] [ 21.7099]

TLKM(-2) 0.000381 0.108745 (0.00058) (0.04101) [ 0.66072] [ 2.65166]

C 0.354868 17.12533 (0.21930) (15.6158) [ 1.61818] [ 1.09666]

R-squared 0.971730 0.992968 Adj. R-squared 0.971530 0.992919

5

Sum sq. Resids 820.9691 4162716. S.E. equation 1.203294 85.68344 F-statistic 4872.352 20017.40 Log likelihood -914.9779 -3354.899 Akaike AIC 3.216706 11.74790 Schwarz SC 3.254723 11.78592 Mean dependent 26.61832 2481.771 S.D. dependent 7.131492 1018.228

Determinant resid covariance (dof adj.) 10629.88 Determinant resid covariance 10444.86 Log likelihood -4269.871 Akaike information criterion 14.96458 Schwarz criterion 15.04062 Number of coefficients 10

Figure 5. VAR test result

The value inside the bracket square [ ] is 1. If changes in TLKM stock 1 week ago the t-statistic value of each independent variable. increased by 1 point, it would cause changes The independent variable is said to be significant in TLKM value by increasing by 0.88 points. in influencing the dependent variable when the 2. If changes in TLKM stock 2 weeks ago t-statistic value is less than 1.9651 (Obtained from t-table with some degrees of freedom of increased by 1 point, it would cause changes 572 and a significant level of 95%), while the in TLKM value by increasing by 0.108 sign (-) is ignored. The positive or negative points. relationship between the independent variable 3. If changes in TLK stock 1 week ago and the dependent variable can be seen from the increased by 1 point, it would cause changes sign on the coefficient of the variable. in TLKM value by increasing by 13.74 From the output table, it can understand points. that the TLKM variable statistically influenced the significance of TLK (-1), TLK(-2), TLKM(- 4. If changes in TLK stock 2 weeks ago 1) dan TLKM (-2). This is indicated by value of increased by 1 point, it would cause changes statistic value > +1.9641 or < - 1.9641. Constant in TLKM value by decreasing by 14.05 does not significantly affect TLKM while the points. same thing happens at TLK which is influenced 5. If changes in TLK stock 1 week ago by TLK (-1). The modeling of formula for increased by 1 point, it would cause changes significant values as follows: in TLK value by decreasing by 0.98 points.

6. TLK (-2), TLKM (-1) and TLKM (-2) is not 푇퐿퐾 = 0.98 ∗ 푇퐿퐾(−1) − 0.0018 significantly affected to stock price of TLK. ∗ 푇퐿퐾(−2) − 0.0003 ∗ 푇퐿퐾푀(−1) + 0.0003 ∗ 푇퐿퐾푀(−2) + 0.35 CLOSING Conclusion 푇퐿퐾푀 = 13.74 ∗ 푇퐿퐾(−1) − 14.05 The conclusion of this study are as follows: ∗ 푇퐿퐾(−2) + 0.88 1. TLKM stock listed in IDX are ∗ 푇퐿퐾푀(−1) + 0.108 significantly influenced TLK stock in ∗ 푇퐿퐾푀(−2) + 17.12 NYSE but not vice versa. 2. The direction of causality on both The modelling results above can be stocks is unidirectional. interpreted as follows: 3. Relationship modeling between stock in Telkom in NYSE and IDX can be modeled with the VECM approach after 6

conducting a series of test tests. Hubungan Pengeluaran Pendidikan Dan Pertumbuhan Ekonomi Dengan Recommendation Menggunakan Pendekatan Kausalitas The advice obtained from this study are as Granger. Jurnal Ekonomi & Pendidikan, follows: Volume 8 Nomor 2. 1. Control variables such as exchange Hadi, Y. S. 2003, Analisis Vector rates can be added further, or economic Autoregressive (VAR) terhadap Korelasi growth from both countries or the antara Pendapatan Nasional dan Investasi country’s trade transaction. Pemerintah Indonesia, 1983/1984-1999/ 2000; Jurnal Keuangan dan Moneter 2. A short-term relationship can be sought Volume 6 Nomor 2, . between each independent variable to the dependent variable as a continuation Indrawan. Rully dan Yuniawati, Poppy. (2016). Metodologi Penelitian. Edisi Revisi. of VECM using Walds Static method : PT. Ferika Aditama..

Jiang, Yonghong dkk. (2017). The Financial REFERENCES. Crisis and Co-Movement of Global Stock Andrei, Shleifer. (2000). Ineffciency Market: Markets—A Case of Six Major An Introduction to Behavioural Finance. Economies. Research Gate. Oxfor University Press. Marudani, D. A. (2010). Uji Kausitas Granger Armawaddin, Muhammad. (2013). Kausalitas Pada Model Harga Saham Pt Granger with EViews 6.0 Sukses Makmtur Indonesia Tbk. Jurnal Sains Dan Matematika, Vol. 17, No. 2, Barret, Adam B. (2015). Granger Causality Pp. 68-74, Oct. 2010. Analysis in Neuroscience and Neuroimaging. JNerosci. Modesale, Timothy. (2006). Granger Causality of Coupled Climate Processes: Ocean Baumohl, Eduard. (2010). Stock Market Feedback on the North Atlantic Integration: Granger Causality Testing Oscillation. AMS1000. with Respect to Nonsynchronous Trading Effects. JEL Classification: G15, C22 . Namini, Siami Sami. (2017). Granger Causality Between Exchange Rate And Botz, G. (2019). Granger Causality Testing in Stock Price: A Toda Yamamoto Mixed‐Frequency VARs with Possibly Approach. International Journal Of (Co)Integrated Processes. Journal of Economics And Financial Issues. Time series. Philal, Tomas. (2016). Stock Market Ding, Liang. (2010). U.S. and Asia Pacific Informational Efficiency in Germany: Equity Markets Causality Test. Granger Causality between DAX and International Journal of Business and Selected Macroeconomic Indicators. Management. Science Direct. Etale, Lyndon M. Does Money Market Spur Priambada, Giri W. (2017). Investor Saham Economic Growth in Nigeria? Granger Pemula: Yuk Belajar Saham untuk Causality Approach. European- Pemula. Jakarta: PT Gramedia. American Journal. Prasetyo, Eko. (2009). Price Limit. Fakultas Fahmi, Irham dan Hadi, Yovi L. (2012). Teori Ekonomi Universitas Indonesia. Portofolio dan Analisis Investasi. Bandung: Alfabeta. Ridho’ah, Hannifatul Afdholatu. (2015). Analisis Kausalitas Perdagangan Gurris, Sellahatin. (2009). Testing Threshold Internasional Dan Pertumbuhan Pasar Cointegration and Threshold Granger Saham Terhadap Pertumbuhan Ekonomi Causality between Stock Price and Di Asean 3 Periode Tahun 2000-2015. Exchange Rate in Turkey. Canadian Center of Science and Education. Rosyidah. Kausalitas Granger Pertumbuhan Ekonomi(Gdp) - Ekspor Di Negara- Hafidh, Aulia Ahmad. (2011). Analisis 7

Negara Islam. (2017). Saputra, Rizki Adi. (2015). Hubungan Zeren, Feyyaz. (2015). Time varying Causality Kausalitas Antara Nilai Tukar dengan between stock market and exchange rate: Harga Saham dan Inflasi di Indonesia. evidence from Turkey, Japan and Vol 3 No 1 (2015): Jurnal Manajemen England. Journal Economic Research- Bisnis Indonesia. Ekonoms Sembahnyang, Lesta Karolina. (2011). Analisis Keterkaitan Ketersediaan Infrastruktur Dengan Pertumbuhan Ekonomi Di Indonesia: Pendekatan Analisis Granger Causality: Jejak. Soebagyo, Daryono. Kausalitas Granger Pdrb Terhadap Kesempatan Kerja Di Provinsi Dati I Jawa Tengah. Jurnal Ekonomi Pembangunan Vol. 8, No. 2, Desember 2007, hal. 177 – 192. Suryanto. (2015). Interdependensi Pasar Saham Indonesia Dan Pasar Saham Beberapa Negara Uni Eropa. Tastan, Huseiyn. (2015). Testing for spectral Granger Causality. Econ Papers. Tri Basuki, Agus. Bahan Ajar Aplikasi Model VAR dan VECM dalam Ekonomi. Tri Wahyudi, Setyo. (2016). Konsep dan Penerapan Ekonometrika Menggunakan E-views. Jakarta: Rajawali Pers. Yusuf Ahmad dkk. (2018). Analisis Kausalitas Antara Harga Saham Konvensional Dengan Harga Saham Syariah di Indonesia (Pendekatan Granger Causality). el-J IZYA Jurnal Ekonomi Islam. Verbeck, M., 2000, A Guide Modern Econometrics, Singapore: John Wiley & Sons, Ltd. Vyrost, Tomas. (2015). Granger Causality stock market networks: Temporal proximity and preferential attachment. Science Direct. Wahyuni Windasari. (2015). Analisis Hubungan Kausalitas Perubahan Volume Perdagangan Dan Perubahan Harga Saham Wijaya Karya Tbk. Jurnal Euclid, Vol.5, No.2, pp. 1. Wen, Lei. (2017). The Causality Relationships between Energy-related CO2 Emissions and its Influencing Factors with Linear and Nonlinear Granger Causality Tests. Pol. J. Environ. Stud. Vol. 26, No. 3 8