type the title of the paper

Journal of Academic Research in Economics Volume 2

Number 2

November 2010

ISSN 2066-0855

EDITORIAL BOARD PUBLISHING EDITOR DRAGOS MIHAI IPATE, Spiru Haret University EDITOR-IN-CHIEF CLAUDIU CHIRU, Spiru Haret University ASSISTANT EDITOR GEORGE LAZAROIU, Contemporary Science Association

INTERNATIONAL ADVISORY BOARD JON L. BRYAN, Bridgewater State College DUMITRU BURDESCU , University of Craiova MARIN BURTICA, West University Timisoara SOHAIL S. CHAUDHRY, Villanova School of Business DUMITRU CIUCUR, Bucharest Academy of Economic Studies LUMINITA CONSTANTIN, Bucharest Academy of Economic Studies ANCA DACHIN, Bucharest Academy of Economic Studies ELENA DOVAL, Spiru Haret University MANUELA EPURE, Spiru Haret University LEVENT GOKDEMIR, Inonu University KASH KHORASANY, Montreal University RAJ KUMAR, Banaras Hindu University, Varanasi MARTIN MACHACEK, VSB-Technical University of Ostrava COSTEL NEGREI, Bucharest Academy of Economic Studies ION PETRESCU, Spiru Haret University T. RAMAYAH, Universiti Sains Malaysia ANDRE SLABBERT, Cape Peninsula University of Technology, Cape Town CENK A. YUKSEL, University of Istanbul MOHAMMED ZAHEERUDDIN, Montreal University LETITIA ZAHIU, Bucharest Academy of Economic Studies GHEORGHE ZAMAN, Economics Research Institute, Bucharest

PROOFREADERS GEORGETA ACHIMESCU, Spiru Haret University CAMELIA BOARCAS, Spiru Haret University MIHAELA CIOBANICA, Spiru Haret University DANIEL DANECI, Spiru Haret University MIHNEA DRUMEA, Spiru Haret University LIANA ELEFTERIE, Spiru Haret University IZABELLA GRAMA, Spiru Haret University IULIA GRECU, Spiru Haret University MANUELA GRIGORE, Spiru Haret University LAURA IACOB, Spiru Haret University STEFAN MIHU, Spiru Haret University PAULA MITRAN, Spiru Haret University LAVINIA NADRAG, Spiru Haret University OCTAV NEGURITA, Spiru Haret University IULIANA PARVU, Spiru Haret University MEVLUDIYE SIMSEK, Bilecik University MIHAELA TUROF, Spiru Haret University

Contents DELAYS FACTORS IN CONSTRUCTION PROJECTS DEVELOPMENT: THE CASE OF KLANG VALLEY, MALAYSIA ABDELNASER OMRAN OOI AI LING ABDUL HAMID KADIR PAKIR MAHYUDDIN RAMLI CLAIMS IN CONSTRUCTION INDUSTRY: FROM NOMINATED SUBCONTRACTOR'S PERSPECTIVE ABDUL AZIZ HUSSIN ABDELNASER OMRAN OUI SOON HOE CORPORATE GOVERNANCE IN FINANCIAL INSTITUTIONS: AN APPLICATION ON THE ISTANBUL STOCK EXCHANGE MELEK ACAR BOYACIOGLU YUNUS EMRE AKDOGAN MODELLING VOLATILITY OF THE GOLD PRICES BY USING GENERALIZED AUTOREGRESSIVE CONDITIONAL HETEROSCEDASTICITY METHOD: THE CASE OF TURKEY VOLKAN ALPTEKIN BURCU GUVENEK MELEK ACAR BOYACIOGLU DIMENSION OF EUROPEAN COMPETITION POLICY AND THE ALBANIAN PRACTICES BESA SHAHINI A REVIEW LITERATURE OF CHARACTERISTICS OF FIRMS, COMMUNITIES, MULTIPARTY PARTNERSHIPS AND CHALLENGES OF FIRM-COMMUNITY PARTNERSHIPS WITH CASES JOSÉ G. VARGAS-HERNÁNDEZ MOHAMMAD REZA NORUZI A GENERAL APPRAISAL OF THE SOCIAL ECONOMIC IMPACT OF CURRENT GLOBAL RECESSION ON INDIA SAURABH SETHI KUSUM MAESHWARI WILFRED ISIOMA UKPERE

135

159

180

197

213

226

249

JOURNAL OF ACADEMIC RESEARCH IN ECONOMICS

MODELLING VOLATILITY OF THE GOLD PRICES BY USING GENERALIZED AUTOREGRESSIVE CONDITIONAL HETEROSCEDASTICITY METHOD: THE CASE OF TURKEY * VOLKAN ALPTEKIN Selcuk University Email: [email protected] BURCU GUVENEK Selcuk University Email: [email protected] MELEK ACAR BOYACIOGLU* Selcuk University Email: [email protected] *Corresponding Author Abstract When the world economies are taken into consideration, gold market maintains its function as a traditional investment place. The volatility in the gold prices can be seen as an indicator of non-stability for the market as a whole. Under such circumstances, the investors rearrange their savings proportion. This situation affects the economy on a great scale, resulting the employment of alternative investment instruments depending on the supply demand equilibrium. In this study, the volatility of gold prices is empirically investigated. Initially the natural logarithm of the gold market index has been taken in hand to be adjusted, in order to avoid from instabilities due to potential fluctuations. Following this, as monthly data has been employed, seasonal effects need to be inspected searching to find out whether unit root exists. After verifying the hypothesis of stationary, modeling has been taken place using the best ARIMA method by the help of some diagnostical tests. Following this, volatility is being tested by ARCH LM test. After choosing the most proper model of volatility, with the help of ARCH-M method, whether the volatility is being included in the model forecasted, has been examined. According to the results of the study, the series of gold prices was volatile and the volatility was eliminated, modeling it GARCH (2,1) model. Thus, a state enabling to measure the risk more seriously is provided. Keywords: Modelling, Volatility, Gold Price, ARIMA, ARCH LM, Turkey

*

This paper was presented into the 17. Forecasting Financial Markets Conference 2010.

197


1. INTRODUCTION Gold maintains its attractiveness as an instrument of keeping traditional purchase power amongst other investment instruments since very old ages. The value of gold stems from the appraisal attributed to it as a valuable metal and thus it meets the function of preserving and evaluating the wealth of economic unit. In determining the value of gold in the market place, the demand for gold plays an important role. Its limited supply makes gold a valuable metal. On this intact point, an increase of demand for an instrument of accumulating value, with a limited supply, will economically bring forth an increase on the price of it under consideration. The most important question to be answered here is, why the demand leading to increase on the prices of gold has increased. Around the globe, the gold sector is being fed from a two demand channel. The first one is the jewelry sector, which uses gold as a raw material, and after processing, introduce it to the market place. The second channel of gold demand is finance sector, which is focal point of this study. Especially, in crisis periods, during which total demand is largely shortened, it is possible to explain the important increase in gold prices. When closely looked at the demand of finance sector for gold in such a serious dimension, it can be seen that some part of this difference is due to central bank’s reserve increase or changing its reserve composition. The other part is due to the fact that economic units want to keep their purchasing power on financial instruments which they consider safer to avoid from the risks. From this point of view, the gold prices gained more importance in terms of becoming a leading indicator in a number of areas. Hence, it is necessary to evaluate the changes in the gold prices very robustly taking the possible influences on the other macro indicators into account. This also brings the assumption that the gold prices may have a volatility as a possible result of its demanding with the financial motives together. This probability of volatility may adversely affect the marketplace factors in forming an expectation. This situation econometrically introduces that it is necessary for the volatility in question to be modeled in the most suitable way, determining the volatility of the series of gold prices. In this study, how the volatility in gold prices affects the gold marketplace, is empirically investigated, it is identified that the series of gold prices was volatile and the volatility has been eliminated, modeling it GARCH (2, 1) model. Thus, a state enabling to measure the risk more seriously is provided. The study consists of five sections. Literature is reviewed in the second section. The information about methodology is given in the third section. Study results are assessed in the fourth section. In the last section, conclusion and discussion take places.

198


2. LITERATURE REVIEW In literature, the problem of estimating the best volatility model for gold prices is considered in several ways. Akgiray et al. (1991) dealt with London p.m. fixing gold price from 1975 to 1986. According to their results, the distribution of gold is peak, fat tailed and observed. The gold return is time dependent and time dependence properties implied that a GARCH model can be efficiently used to fit the data. Rohan (2000) chose GARCH (1, 1) model over higher order ARCH or GARCH models when the relationship between the gold and silver contract was investigated. Ungor (2002) investigated the optimal model for the London p.m. fixing gold price per ounce with U.S dollar, Japanese Yen and South African Rand with 2000 -2007 daily data. According to the study, the best model for gold price of South African Rand and U.S Dollar was IGARCH (1, 2) with normal distribution. The gold price for Japanese Yen was best fitted to EGARCH (1, 2) with normal distribution. Kutan and Aksoy (2004) used the gold prices in Turkey between January 2, 1997 and February 14, 2001. They revealed that the standard GARCH (1, 1) model to be the best fit for the data. Tully and Lucey (2005) examines both the cash and futures price of gold and significant economic variables identified during the 1987 and the 2001 crisis. They concluded that APGARCH model provides the most adequate description for the data. Mills (2005) found out that gold serves as a hedge against fluctuation in the foreign exchange value of the dollar. The conditional error variance equation which was found to be the best modeled as an EGARCH (1, 2) process with t distributed innovations was used. Tully and Lucey (2007) examined the fit of the APGARCH model for six gold models. They concluded that an APGARCH model is applicable to the datasets in question, taking into account GARCH, leverage and power effects. Using gold cash and futures data over a long period they confirmed that the US dollar is the main macroeconomic variable which influences gold. Do et al. (2009) examined behavior of return and volatility of five emerging stock markets (Indonesia, Malaysia, Philippines, Thailand and Vietnam), incorporating with the effects from the international gold market. The results indicate that the GJR (1,1) model is preferred to GARCH (1,1), except Vietnam. The GARCH (1,1)-X model captures better stock market volatility behavior than GJR (1,1)-X, except Indonesia. Gold could be a substitute commodity for stocks in Vietnam and Philippines, while it could be a complement for stocks in Indonesia, Thailand and Malaysia.

3. METHODOLOGY In the recent years, ARCH and GARCH models have begun to be used increasingly in the analysis of economic financial time series. In the models considered in the approach of econometrics for time series, it is assumed that the variables are stable. That the time series are stable means that variance and mean is 199


constant over the time and covariance of the variables in two lagged time group, depends on the lag between the variable and is independent from the time (Gujarati, 1999: 712). As stated in the study by Granger and Newbold (1974), when worked with unstable time series, one might be faced to the false regression problems. In this situation, the result obtained by the regression analysis can reflect the real relationship, if there is only a cointegration relationship between these series (Gujarati, 1999: 726). In this study, presenting the stability of the time series belonging to the gold prices is realized by using Augmented Dickey-Fuller (ADF) test, developed by Dickey and Fuller (1981). In order to test the attribute “unit root” of Yt series, the following regression equation is used. N

Yt   0   1t  Yt 1   Yt i   t i 1

(1)

ADF test is based on the prediction of the parameter δ and its t-statistic. Null hypothesis is rejected, if it is negative and significantly different from zero. Null hypothesis, from the hypothesis of interest, states that the series is not stable and the alternative hypothesis states that the series is stable. The problem associated with ADF test is that it involves to be included in the additional differences in the test equation. And this results in a loss of freedom degree and reduction in the power of the test procedure. Alternatively, PP (Phillips-Peron) Approach regards the presence of the unknown forms of autocorrelation and the conditional changing variance state in the error term and uses a non- parametrical correction for the serial relationship. Then, in order to eliminate the influences of serial relationships on asymptotic distribution of test statistics, statistics are transformed. For both of the tests, t-statistic having bigger value in comparison to the critical values leads to the rejection of null hypothesis for unit root (Enders, 2004: 251). Since stability is evaluated as a precondition in terms of a model of time series, an analysis can be done only after this condition is fulfilled. In the next stage, the issue how the variables enabling stability will be added to the model shows an important point in the model. Eagle, in his studies in 1982 and 1983, examined the inflation data of United Kingdom. He proved that variance of error terms was not constant. This study carried out by Eagle entered the literature in the name of Autoregressive Conditional Changing Variance (ARCH). ARCH models, leaving the assumption of constant variance in the method of the traditional time series aside, allow the variance to differ as a function of lagged prediction mistakes. Therefore, ARCH models are suitable for combining the variance in the estimation process. In the ARCH model, it is assumed that the characteristic acts of estimation mistakes are based on the residual of regression. As a result of this assumption, regression residuals will also be conditioned (Engle, 2000: 1- 23). Using the residual of least squares for ARCH models, with Lagrange

200


Multiplication (LM) test, ARCH influence can be tested. ARCH LM test, examines whether there are autoregressive conditional variation difference effects the error terms of the model. In the implementation of ARCH model, due to using relatively long lags and suggesting the content lag structure, some constraints are placed on the parameters in conditional variance. In order to eliminate the drawbacks that these constrains are not provided and reaching the negative variance parameter estimate, extended ARCH model and developed a model based on both using more previous information and having more flexible lag structure. This model is called “generalized ARCH” (Bollerslev, 2000: 42- 60).

4. DATA AND EXPERIMENTAL RESULTS In this study, how the volatility in gold prices influence the gold market is examined empirically. Monthly data on the period of 07.1995-01.2010 of the index of gold market are taken into consideration. The data were drawn from the official site of Turkish Statistical Institute. Since monthly data are employed in the study, first of all the seasonality of the serial is examined. Whether the serial includes unit root is also researched. Following the preference of stability assumption, as a result of reporting that the series to be modeled in the best compatible ARIMA method is stable, ARCH LM test and its volatility was examined through some diagnostic tests. Following the preference of method modeling the volatility in the best way, the series, which will be modeled, the conditional variance related to the variable, via the ARCH-M (ARCH-IN MEAN) enabling ARCH-M to use of variable in the model, whose means is estimate, in the stage of determining the risk, whether or not the ambiguity about variable takes place in the model is investigated. IGM 280,000,000

240,000,000

200,000,000

160,000,000

120,000,000

80,000,000

40,000,000 1996

1998

2000

2002

2004

2006

2008

Figure1. IGM (Index of Gold Market) 201


In Figure 1, time series graphic belonging to the variable of IGM is shown. It is remarkable that there is an increasing trend in this graphic. In this stage, in order to freed the data from small fluctuations and bring it into linear state, logarithm of the series is taken and the data are provided to be brought into suitable for analyzing. The graphic of the series, whose logarithm is taken, is shown in Figure 2. LIGM 19.4 19.2 19.0 18.8 18.6 18.4 18.2 18.0 17.8 1996

1998

2000

2002

2004

2006

2008

Figure 2. LIGM (LogIGM)

When the seasonality graphic belonging to LIGM series taking place in Figure 3 is analyzed, it is seen that the series does not include any seasonal influences. LIGM by Season 19.4 19.2 19.0 18.8 18.6 18.4 18.2 18.0 17.8 Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Means by Season

Figure 3. Seasonality Graphic of LIGM

202

Nov

Dec


In order to test whether the series without seasonality influence are stable or not, unit root test has been made. In Table 1, the results of ADF and PP unit root test, applied for LIGM series, are reported. ADF Test Stat. ADF -1.316576 ADF Test Stat. ADF (-1) -12.39334 PP Test Stat. PP

PP Test Stat. -12.40789

Mac Kinnon Test Critical Values %1 -4.011663

-1.281033

PP (-1)

Mac Kinnon Test Critical Values %1 %5 % 10 -4.011663 -3.435858 -3.141996 Mac Kinnon Test Critical Values %1 %5 % 10 -3.468295 -2.878113 2.575684

%5 % 10 -3.435858 -3.141996 Mac Kinnon Test Critical Values %1 %5 % 10 -3.468295 -2.943427 -2.575684

Prob. 0.8805 Prob. 0.0000 Prob. 0.8891 Prob. 0.0000

Table 1. Results of Unit Root Tests for LIGM

ADF test statistic (-1,316576), at the 1 %, 5 % and 10 % significance levels, is absolutely smaller than Mac Kinnon critical values, and probability level (0,8805) is bigger than the value (0.05). Ho hypothesis is rejected and decided that there is a problem with unit root in the series. Then, taking the first difference of the series, it is provided to be brought into stable condition. Another unit root test is Phillips- Perron (PP) unit root test. The null and alternative hypothesis of PP test also overlap ADF unit root test. According to the results of series associated with this test, because the value -1.281033 of PP test statistic, at the 1 %, 5 % and 10 % significance levels is absolutely smaller than Mac Kinnon critical values, and probability value (0.8891) is bigger than critical value (0.05). Ho unit root VAR hypothesis was not rejected and it was decided that there is unit root in series. It means that the series is not stable. In this stage, the first difference of series was taken and this problem has been eliminated. Although standard unit root tests such as ADF, PP and KPSS are often used in the analysis, they overlook the structural breaks. However, in the period, when the series is drawn, occurance of important events can cause a structural break and influence the test results. Therefore, Zivot-Andrews test regarding the structural break unit root test was applied in the model.

203

204

2003

2004

2005

2006

2007

2008

2007

2008

ZIVOTM

2004 ZIVOTT

2005

2007

V CRITT

2006

2008

-5.5 2003

2004 ZIVOT

2005

2003

2004 ZIVOTT

2005

2007 VCRITT

2006

2008

-11

-10

-9

-8

-7

-6

-5

2003

2004

ZIVOT

2005

2007

2007

VCRIT

2006

VCRIT

2006

Figure 5. dLIGM Graphic Results of Zivot-Andrews Test for dLIGM Series

VCRITM

-11

-11 2006

-10

-10

-7

-6

-5

-4

-9

2005

2003

-5.0

-4.5

-4.0

-3.5

-3.0

-2.5

Figure 4. Graphic Results of Zivot-Andrews Test for LIGM Series

-9

2004

VCRITM

-8

2003

ZIVOTM

-4.6

-4.4

-4.2

-4.0

-3.8

-3.6

-8

-7

-6

-5

-4

-5.2

-4.8

-4.4

-4.0

-3.6

-3.4

-3.2

-2.8

-3.2

-3.0

-2.4

2008

2008



In the Figure 4, the graphic results of Zivot-Andrews test take place. In each of three graphics, the test results are over the critical values. This also points that there is a problem with unit root in the series. Therefore, the first difference of series was taken and the problem of unit root was clearly eliminated as seen clearly in Figure 5. When all unit tests are considered, it is seen that gold price index series possessed the problem “unit root” in degree and that the series did not the stability condition. As a result of taking the first difference of series, the all tests carried out to determine that the problem “unit root” was eliminated and that the stability of series was provided. Later, ARIMA model was determined in accordance with IGM series. Variable C AR(1) AR(2) AR(3) AR(4) MA(1) MA(2) MA(3) MA(4)

Coefficient 0.007783 0.068699 -0.357872 -0.079062 -0.725797 -0.091372 0.374358 0.294191 0.934704

Std. Error 0.002135 0.235059 0.240603 0.213063 0.183161 0.233904 0.229759 0.209330 0.205729

t-Statistic 3.646037 0.292262 -1.487397 -0.371071 -3.962608 -0.390639 1.629349 1.405389 4.543384

Prob. 0.0004 0.7705 0.1389 0.7111 0.0001 0.6966 0.1052 0.1618 0.0000

Table 2. Results of ARIMA Model

According Table 2, ARIMA model was determined as AR(4) and MA(4), ARIMA (4,1,4). H0: There is no ARCH effect. H1: There is ARCH effect. F-statistic

Obs*R-squared

Prob. F(1,160)

Prob. Chi-Square(1)

25.20317

22.90559

0.0000

0.0000

Table 3. Heteroskedasticity Test: ARCH LM (1) F-statistic 7.856280

Obs*Rsquared 28.13167

Prob. F(4,157)

Prob. Chi-Square(4)

0.0000

0.0000

Table 4. Heteroskedasticity Test: ARCH LM (4)

205


F-statistic 4.849559


Prob. F(8,153)

Prob. Chi-Square(8)

0.0000

0.0000



Prob. F(12,160)

Prob. Chi-Square(12)

0.0000

0.0000


ARCH- LM test results are given in Table 3, 4, 5 and 6., rejecting the hypothesis that probability and chi square test values are less than 0.05 in the all the tests, it is said that there is ARCH effect. Therefore, with a suitable GARCH model, volatility is tried to be modeled. Variable C AR(1) AR(2) AR(3) AR(4) MA(1) MA(2) MA(3) MA(4)

Coefficient -0.053591 -0.266567 0.213597 0.310278 0.758891 0.326864 -0.340090 -0.345038 -0.631814

Std. Error 0.126546 0.151134 0.152430 0.150886 0.150038 0.197006 0.194300 0.195095 0.189194

t-Statistic -0.423487 -1.763780 1.401280 2.056371 5.057987 1.659162 -1.750338 -1.768563 -3.339506

Prob. 0.6719 0.0778 0.1611 0.0397 0.0000 0.0971 0.0801 0.0770 0.0008

Table 7. GARCH (1.1) Model Results of dLGIM Series

In Table 7, a reporting process belonging to GARCH (1.1) model calculated associated with d(LIGM) is shown. As a result of it, in the name of introducing the possible ARCH effect on the model, ARCH-LM test was applied and the following results were obtained. F-statistic

Obs*R-squared

Prob. F(1,167)

Prob. Chi-Square(1)

0.016411

0.016606

0.8982

0.8975


206


F-statistic 1.671957


Prob. F(4,161)

Prob. Chi-Square(4)

0.1590

0.1574



Prob. F(8,153)

Prob. Chi-Square(8)

0.4333

0.4236



Prob. F(12,145)

Prob. Chi-Square(12)

0.7276

0.7101


According to the prediction results taking place in Table 8, 9, 10, and 11 on ARCH LM test, all probability and chi square values are smaller than the value 0,05, the volatility of series was eliminated. After, the volatility of gold price index series is modeled with GARCH (1.1), the conditional variance associated with the variable, via the model ARCH-M enabling to use the mean of variable in the model estimated, it is investigated whether or not the volatility of the variable influenced the variable itself. The results are shown in Table 12. Variable @SQRT(GARCH) C AR(1) AR(2) AR(3) AR(4) MA(1) MA(2) MA(3) MA(4)

Coefficient -0.715378 -0.053591 -0.266567 0.213597 0.310278 0.758891 0.326864 -0.340090 -0.345038 -0.631814

Std. Error 0.436765 0.126546 0.151134 0.152430 0.150886 0.150038 0.197006 0.194300 0.195095 0.189194

t-Statistic -1.637899 -0.423487 -1.763780 1.401280 2.056371 5.057987 1.659162 -1.750338 -1.768563 -3.339506

Prob. 0.1014 0.6719 0.0778 0.1611 0.0397 0.0000 0.0971 0.0801 0.0770 0.0008

Table 12. ARCH -M Test

According to the results of Table 12, when the significance level of 10 % is considered, the volatility of gold market index influenced negatively the gold market index at the level of 71 %. As a consequence, it is seen that the volatility stood out as the most important determiner of the gold market index.

207


5. CONCLUSION When the preferred financial investments are evaluated, deposit, repurchasing agreements, stocks, treasury bonds, government bonds, foreign currency and gold seem to dominate. Due to political and economic ambiguities the country experienced, the risks which investment instruments are effected differentiate and this reflection can be seen on the portfolios of the investors. In this context, the nominal rate of return for USD, deposit interest, Istanbul stock exchange index, and gold are counted among the important investment instruments in Turkey. For the period between February 1999 and February 2009, the difference in the rate of returns was examined. It was seen that the reality put forward theoretically was supported by the figures. When the graphics in Appendix 1 are examined, it can be seen that the 12 month’s nominal rate of return for deposit interest and Istanbul stock exchange index generally tracked a similar course. This course did not experience important downs and ups till 2007. However, in the middle of 2007, an important change became remarkable. This unexpected movement was a result of early election decision of the government. The opposite course of these two series is observed in the annual rate of return for USD; in the series, in which small downs and ups were identified, it was recorded that an important decline was experienced in the middle of 2007. When the nominal rate of return for gold is regarded, it seems to have a highly volatile structure. This volatile structure becomes a guide about the risk premium of the gold, assessed as a financial investment instrument. This opinion is reached visually through the graphics. With several econometric test forms we tried to support the fundamental aim of this study. Moving on from this point, the monthly date of gold market index belonging to the period between July 1995-January 2010 were used. After the seasonality and stability conditions of series are provided, the stage of modeling with ARIMA method has been proceeded. Whether or not there is an ARCH effect in the error square of the ARIMA model determined is examined via ARCH LM test and that there was volatility in the series gained definiteness. As a result of analysis carried out on the purpose of modeling the volatility, it was seen that the most capable model was GARCH (1.1). The confidence of this model was examined via ARCH LM test and it was identified that the volatility in the model was eliminated. In this way, it was considered that the volatility of IGM series would be more effectively predicted with the conditional changing variance models. Following the preference of method modeling, the volatility in the most convenient way, via ARCH-M model enabling the conditional variance associated with the variable to use the average of the variable in the model predicted, in the stage of determining the risk, it was investigated that the ambiguity about the variable took place in the model. When the results are evaluated, it is seen that at the significance level of 10 %, the 208


volatility of gold market is negatively influenced at the level of 71 % of gold market index. As a result, the volatility of gold market index possesses in its nature plays a decisive role on the index of this market. Hence, when comparing to other financial investment instruments, it shows that gold, considered as less risky, is often used in the portfolio assortment. Therefore it improves the risk return profile and even considered as an insurance on the reserves of the country, is not just an investment instrument as it is perceived.

REFERENCE Akgiray V., Booth, G. G., Hatem, J. J. and Chowdhury, M. (1991). Conditional Dependence in Precious Metal Prices, The Financial Review 2 (6): 367–386. Bollerslev T. (1986). “Generalized Autoregressive Conditional Heteroskedasticity”, Journal of Econometrics 31: 307–327. Bollerslev T. (2000). “Generalized Autoregressive Conditional Heteroskedasticity”, ARCH Selected Readings, Advanced Texts in Econometrics, New York: Oxford University Press. Dickey, D. A. and Fuller W. A. (1981). “Likelihood Ratio Statistics For Autoregressive Time Series With A Unit Root”, Econometrica 49 (4), 1057– 1072. Do, G. Q., Mcaleer M. and Sriboonchitta S. (2009). “Effects of International Gold Market on Stock Exchange Volatility: Evidence from Asean Emerging Stock Markets”, Economics Bulletin 29 (2): 599–610. Enders, W. (2004). Applied Econometric Time Series, Iowa State University, John Wiley and Sons Inc. Engle, R. F. (1982). “Autoregressive Conditional Heteroseedasticity with Estimates of the Variance of United Kingdom Inflation”, Econometrica 50: 987– 1007. Engle, R. F. (1983). “Estimates of the Variance of U.S. Inflation Based upon the ARCH Model ”, Journal of Money, Credit and Banking 15 (3): 286– 301. Engle, R. F. (2000). “The Econometrics of Ultra-High Frequency Data," Econometrica, Econometric Society 68 (1): 1–22. Granger, C. W. J. and Newbold, P. (1974). “Spurious Regressions in Econometrics”, Journal of Econometrics 2: 111–120. Gujarati, D. N. (1999). Temel Ekonometri, (Translater: SENESEN, U. and SENESEN, G. G.), Literatur Yayınları, Istanbul. Kutan, A. M. and Aksoy, T. (2004). “Public Information Arrival and Gold Market Returns in Emerging Markets: Evidence from the Istanbul Gold Exchange”, Scientific Journal of Administrative Development 2: 13–26. Mills, T. C. (2005). “Gold As A Hedge Against The Dollar”, Journal of Financial Markets, Institutions and Money, 15: 343–352. 209


Rohan, C. D. (2000). “Price Discovery in Strategically-linked Markets: The Case of the Gold- Silver Spread”, Applied Financial Economics 10: 227–234. Tully, E. and Lucey B. M. (2005), An APGARCH Investigation of the Main Influences on the Gold Price, Available at SSRN: http://ssrn.com/abstract=792205 Tully, E. and Lucey B. M. (2007). A Power GARCH Examination of the Gold Market, Research in International Business and Finance 21: 316–325. Ungor, S. (2007). Comparison of Conditional Variance Models for Gold Returns: An Application to US Dollar, Japanese Yen and South African Rand, Term Project, Middle East Technical University Institute of Applied Mathematics.

210


Appendix 1: Nominal Rate of Return for Deposit Interest, USD, Istanbul Stock Exchange National 100 Index and Gold in the Period between February 1999 – February 2009 (%)

DEPOSIT 2,000

1,600

1,200

800

400

0 99

00

01

02

03

04

05

06

07

08

DOLLAR 400

0

-400

-800

-1,200

-1,600

-2,000 99

00

01

02

03

04

05

06

07

08

211


STOCK_EXCHANGE 5,000

4,000

3,000

2,000

1,000

0

-1,000 99

00

01

02

03

04

05

06

07

08

GOLD 160

120

80

40

0

-40 99

212

00

01

02

03

04

05

06

07

08