INTRODUCTION
Problem Statements
There is much attention recently on gold, partly due to the surges in its price and the increase in its economic uses
Gold has a critical position among the major precious metals Gold is not only an industrial commodity but also an investment asset which is commonly known as a
Gold is often regarded as a "safe haven" asset during times of financial market instability (Baur & Lucey, 2010; Coudert & Raymond-Feingold, 2011) Despite rising risks in the financial markets, gold prices have not only remained steady but have also surged in recent years From the onset of the financial crisis in August 2007 to December 2012, the nominal price of gold rose by an impressive 146.65%.
For central banks, gold keeps an important position in their reserve asset Its role is increasingly enhanced from 2009 when there was so much worry about the health of
Central banks are increasingly purchasing gold to decrease their dependence on the US dollar as a reserve asset, aiming to stimulate borrowing and lower interest rates on loans Recent evidence highlights this trend, with a UBS poll revealing that while about half of officials still view the US dollar as the most important reserve asset in 25 years, 22 percent believe gold will take that position.
According to Farchy & Blas (2010), central banks' demand for gold is expected to remain robust over time The World Gold Council's report from August 14, 2012, highlighted a significant increase in official sector purchases, with central banks acquiring 157.5 tonnes of gold in the second quarter—more than double the 66.2 tonnes purchased during the same period in 2011 Additionally, total purchases for the first half of 2012 reached 254.2 tonnes, marking a 25% increase from 203.2 tonnes in the previous year, and indicating that the official sector represented 16% of overall gold demand in Q2 This trend clearly demonstrates the expanding role of gold for central banks.
Central banks and nations are implementing significant policies regarding gold to safeguard their reserves For instance, China has imposed a ban on gold exports to ensure that all gold entering the country remains within its borders Additionally, countries that can produce gold have mandated that their central banks act as the primary purchasers of domestically mined gold.
In March 2012, Russia and Kazakhstan purchased locally produced gold, highlighting a growing trend among central banks in emerging and newly wealthy nations These countries' gold reserves remain significantly smaller compared to those held by central banks in the US and Europe.
In Viet Nam, the domestic gold prices got so much fluctuation since September
In 2012, the disparity between global and domestic gold prices significantly widened, at times exceeding VND 5 million per tael, driven by high demand and limited supply from major banks This situation posed serious challenges to the exchange rate and the overall economy Khanh (2010), General Director of Sai Gon Gold and Silver ACB-SJC Joint Stock Company, emphasized that rising gold prices directly and indirectly influence the USD/VND exchange rate, impact the Consumer Price Index (CPI), and affect monetary policy, as well as the stock and real estate markets.
To address the gap and stabilize the gold market, as mandated by the National Assembly of Vietnam in Resolution No 51/2010/QH12, several key legal documents have been issued by the Government and the State Bank of Vietnam (SBV) since 2010 Notable measures include Circular No 01/2010/TT-NHNN, which closed the gold trading floor; Circular No 22/2010/TT-NHNN, prohibiting banks from converting gold into paper currency; and Decree No 24/2012/ND-CP, which established the State's monopoly on gold bar production Despite these regulations limiting gold supply, demand remains high, leading to volatile prices The Prime Minister's Decision No 16/2013/QD-TTg allows the SBV to buy and sell gold bullion to stabilize prices and manage foreign reserves effectively To enhance its role in price stabilization, the SBV must strategically time its gold purchases from abroad and develop tools to forecast market movements influenced by changing factors.
From the above features, the paper will focus on analyzing factors (that are addressed in the literature) affecting global gold prices.
Research objectives
This paper aims to explore the relationship between local and global gold prices, identify key variables influencing global gold prices in both the short and long term, and assess whether the State Bank of Vietnam (SBV) can rely on global gold price movements when making decisions.
Research questions
This paper aims to explore the correlation between Vietnam's gold prices and global gold prices, investigate the factors influencing global gold prices, and provide recommendations to the State Bank of Vietnam (SBV).
Scope of the research
The research focuses on finding main factors affecting to global gold prices
This article provides an overview of Vietnam's gold market and its relationship with the global gold market It examines the various factors influencing global gold prices and their directional relationships, drawing on existing literature Additionally, the paper estimates a model of global gold prices that incorporates the global financial crisis as a dummy variable.
Data used for the research are secondly time series for the daily period from 2007 to
2012 due to daily data for Vietnam gold prices are not adequate.
Structure of the thesis
The paper is organized as followings:
The chapter 1 explains reason why the topic of thesis is chosen, gives main objectives, major research questions and the scope of the research
Chapter 2 provides some overview of Vietnam’s gold market, the connection between Vietnam and global gold markets
Chapter 3 covers the literature about the relationship between gold price and oil prices, US dollar exchange rate, and stock market
Chapter 4 presents the research methodology, data collection
Chapter 5 shows statistic results from adopted model Findings are analyzed to get answers for questions mentioned in the chapter 1
Chapter 6 concludes with the main findings and gives some suggestions.
OVERVIEW OF VIETNAM’S GOLD MARKET
The national gold brand of Vietnam
On November 25, 2011, Saigon Jewelry Company Limited (SJC) was designated as Vietnam's gold brand under the control of the State Bank of Vietnam (SBV) This decision is justified by SJC's dominance in the domestic gold bullion market, holding a 90% market share, its strong brand recognition in Vietnam and the Asia-Pacific region, and its status as a 100% state-owned enterprise managed directly by the Ho Chi Minh City Party Committee.
The decision has garnered widespread agreement among analysts, highlighting several key advantages Firstly, it allows the State Bank of Vietnam (SBV) to intervene more effectively during significant fluctuations in domestic gold prices Secondly, it eliminates the "ask and grant" mechanism, streamlining processes Lastly, this decision enhances the ability to control the quality of bullion gold, ensuring higher standards in the market.
The connection of Vietnam’s gold price to global gold price
Cost price represents all expenses to have a unit of gold This value is used as the main factor in profit and loss calculation
In Vietnam, gold is primarily imported, as local mining production is limited Consequently, the price of SJC gold in Vietnam is largely influenced by global gold prices.
PVN = [(PTG + CVC + I) x (1 + TNK): 0.82945] x E + CGC
- PVN is the Vietnam’s gold price (VND per tael);
- PTG is the world gold price (USD per troy ounce);
- E is USD/VND exchange rate;
- CGC is processing cost (VND per tael);
Market price is the economic price for which a good is offered in the market place (Wikipedia, 2013) In addition, it is also the intersection of supply and demand
Market prices for most goods typically align with or exceed their cost prices to prevent losses However, the domestic gold market can behave differently, as its prices are influenced not only by global gold prices but also by factors such as demand and purchasing power.
2.2.3 The connection between domestic and global gold markets
In general, Vietnam’s gold price tends to move along with the world gold market (Cong, 2012; Hoang, 2004)
Hoang (2004) checked the connection between Viet Nam and London markets for daily sub-sample from January, 2004 to May, 2004 by using correlation coefficient
He found that the positive correlation existed
This study analyzes the relationship between SJC prices and London gold prices over a daily period from January 2007 to December 2013 SJC data was sourced from Saigon Giai Phong newspaper, while London gold prices were obtained from the World Gold Council, with detailed information provided in Chapter 4 The findings are summarized in Table 2.1.
Table 2.1: Correlation matrix between SJC and London’s gold price
The result in the Table 1 showed that: r(SJC, GOLD) equals 0.99 It means that the two market have had a strongly positively correlation
In addition, the paper also takes the following hypothesis testing:
- Null hypothesis H0: ρ = 0 (there is no actual correlation);
- Alternative hypothesis H1: ρ ≠ 0 (there is an actual correlation)
And its result is included in Table 2.2
Table 2.2: Correlation testing between SJC and London’s gold price
P-value is equal 0.0000, the null hypothesis is rejected In other words, there is an evidence that a true relationship between domestic and global gold markets
The above results confirmed the conclusion of Hoang (2004)
2.2.4 The big gap still exists between the two gold markets
The disparity between domestic and global gold markets has significantly increased, occasionally reaching VND 5 million per tael, particularly in 2012 To explain the persistence of this gap, Bank Governor Nguyen Van Binh has categorized Vietnam’s gold market into three distinct phases.
Between 2007 and 2009, incomplete regulations on gold management led to the rapid and spontaneous establishment of gold trading floors During this period, official gold imports ranged from 40 to 50 tons, while gold smuggling was estimated to be around 50 tons.
The domestic gold market has been unstable, leading to frequent gold fevers where individuals hurriedly buy or sell gold This speculative behavior has significantly impacted foreign exchange markets, price indices, and overall macroeconomic stability In response, the government has banned and officially terminated gold trading activities on foreign accounts and the operation of gold floors.
Between 2009 and 2012, the disparity between the two markets remained minimal, although slightly higher than in previous years During this period, the domestic market experienced instability, which negatively affected exchange rates and the macro-economy, albeit at a reduced level.
2012-2013 stage: New legal framework has been developed and come into effected
The State maintains a monopoly on gold imports and gold bar production, while tightening controls on gold smuggling This period is characterized by a significant gap between the two markets, yet the domestic market remains stable Speculation has diminished, and there are no longer scenes of people rushing to buy gold The phenomenon of "goldenization" in the national economy is being restrained, contributing to a stable macro-economy.
In summary, during the research period, domestic gold prices showed a correlation with international gold prices However, disparities between the two markets persist, influenced by domestic demand and supply dynamics as well as varying state policies over time.
LITERATURE REVIEW
The relationship between gold and oil prices
There are considerable studies on the relationship between gold and oil prices
However, results on the relationship are still mixed
Research indicates that oil prices influence gold prices through two primary channels Firstly, an increase in oil prices leads to higher general price levels, or inflation, which subsequently drives up gold prices, as gold serves as a hedge against inflation and the dollar Secondly, rising oil prices boost export revenues, particularly for oil-producing countries, which may hold significant gold reserves This increase in oil revenues enhances the demand for gold, contributing to a rise in its price.
Narayan et al (2010) demonstrated a significant relationship between gold and oil prices through the inflation channel, applicable in both the short-run and long-run Their short-run analysis, conducted using an ordinary least squares method from 1963 to 2008 in the United States, revealed that rising oil prices lead to inflation, which subsequently drives up gold prices In the long-run, they utilized a structural break cointegration test by Gregory and Hansen (1996), which is more effective than the traditional method by Engle and Granger (1987) due to its allowance for regime shifts, analyzing data from 1995 to 2009 Their findings indicated that gold and oil markets are cointegrated, with oil prices serving as a predictor for gold prices and vice versa, extending up to a maturity of 10 months.
Empirical studies by Harmmoudeh et al (2008) and Sari et al (2010) contradict the findings of Narayan et al (2010), revealing that oil prices do not significantly influence gold prices in the long run, and vice versa They indicate that fluctuations in gold prices are primarily driven by factors associated with the jewelry industry and interventions by central banks aimed at managing foreign reserves and influencing exchange rates, rather than being linked to oil prices.
Malliaris et al (2011) found no evidence of cointegration between gold and oil prices using the Johansen test and Granger causality test from January 4, 2000, to December 31, 2007 However, their research indicated that gold prices can effectively predict oil prices in the short run and vice versa when employing neural network methodology.
The relationship between gold price and US Dollar exchange rate
The relationship between gold prices and the US dollar is often inversely correlated; when gold prices rise, the dollar typically declines, and vice versa A weaker dollar raises concerns about inflation, prompting investors to seek gold as a hedge against it Conversely, a strengthening dollar generally leads to a decrease in gold prices.
Hammoudeh et al (2008) discovered that rising gold prices lead to a depreciation of the US dollar against major currencies in both the short and long term, indicating a one-way relationship rather than a reciprocal effect.
(2011) found that the value of dollar and the gold price has negative relationship
But value of dollar has Granger-causality to gold price, not vice versa.
The relationship between gold price and stock market
There is an inverse relationship between gold prices and stock prices; as stock prices rise, investors tend to sell their gold to reinvest in the stock market, driven by the increased profits from their stock investments.
This will cause the price of gold decreased
Smith (2001) examined the short-term and long-term relationships between four gold price series and six different US stock price indices over the 1991-2001 period
Recent studies on the relationship between stock indices and gold returns reveal varying results One study identified a negative Granger causality from US stock index returns to gold returns in the short run, while Gilmore et al (2009) found a unidirectional causal effect only in the short term Additionally, Wang et al (2010) concluded that gold prices and Taiwan’s stock prices are independent, indicating no mutual influence In contrast, Bhunia & Das (2012) presented differing perspectives on this relationship.
By applying Granger causality analysis, they found that stock market can be used to predict gold prices in India and vice versa
In conclusion, research findings on the causality between variables remain inconsistent Some studies identify a clear causal relationship, while others reveal no causality or suggest bi-directional causality These discrepancies may stem from variations in sample periods, research methodologies, and the specific variables examined.
This study will further investigate the relationship between gold prices and oil prices, stock market trends, and US dollar exchange rates using an updated dataset The autoregressive distributed lag bound test (ARDL) and unrestricted error correction models will be employed, as these methods have been effectively utilized in previous research by Hammoudeh et al (2008) and Sari et al (2010) The ARDL approach offers several advantages over the methodologies proposed by Engle and Granger (1987), as well as Johansen (1988) and Johansen and Juselius (1990), which will be detailed in Chapter 4.
The summary of empirical researches listed in Table 3.1
Table 3.1: Summary of empirical studies in Literature review
Methodology Key variables Period Main Results
Spot and future prices of gold and oil
1995 - 2009 Gold ↔ Oil in long-run
ARDL bound test, unrestricted error correction models, diagostic tests
Spot prices of oil, gold, silver, copper; interest rate;
US dollar index; some dummy variables
1990 - 2006 Gold → Oil in short-run;
Gold, Oil → US dollar index in long-run
ARDL bound test, unrestricted error correction models; the generalized forecast error variance decompostions, the generalized impulse response functions
Spot prices of oil, gold, silver, Palladium, Platinum, USD/EUR exchange rate
1999 - 2007 Oil → Gold in short-run;
Johansen test for long-run relationship;
Spot prices of gold, oil, euro
2000 - 2007 Gold ↔ Oil in short-run
Pairwise Granger causality; VAR Granger
Vector autoregression (VAR); Impulse response functions and variance decomposition
Prices of gold, oil, value of dollar
Granger causality; error correction model
Four gold price series and six different US stock price indices
1991 - 2001 US stock index returns → gold returns in short- run
Vector error correction (VEC) model; variance decomposition and impulse response functions
Gold prices, stock price indices of gold mining companies, broad stock market indices
1996 - 2007 Stock prices → gold prices in short-run
Oil price; gold price; exchange
2006 - 2009 Oil price ↔ stock prices in correction model, granger causality rates of US dollar; stock price indices of United States, Germany, Japan, Taiwan, Chian
Johansen test, vector error correction model, Granger causality test
Gold price, stock returns in India
Note: ↔ means that bi-directional causality exists between two variables; → means that uni-directional causality exists between two variables.
RESEARCH METHODOLOGY
Research process
This study will be conducted through steps described in the Figure 4.1.
Model establishment
This paper utilizes the unrestricted error correction model, successfully implemented by Hammoudeh et al (2008) and Sari et al (2010), to analyze empirical studies Given the uncertainty surrounding the long-run relationship between gold prices and other variables, the research constructs unrestricted regressions with each variable serving as a dependent variable.
Data collection
The data set utilized for the paper is secondary data It consists of daily time series over January 2007 to December 2012 (1,498 observations for each series)
Recent studies indicate that oil prices, the US Dollar Index, and the S&P 500 Index are independent variables influencing gold prices The global financial crisis, referred to as the Crisis, spanned from August 2007 to the end of 2008.
(Wikipedia, 2013) will be used as a dummy variable
- Role of gold in economy, in reserve asset of Central Banks;
- Recent policies relating to gold of some Central Banks and of the State Bank of Viet Nam;
- Target of Vietnam’s government in stabilizing the gold market
- Studies on the relationship between gold and oil prices;
- Studies on the relationship between gold price and US dollar exchange rate;
- Studies on the relationship between gold price and stock market;
- Method and model will be chosen in Thesis
Overview of Viet Nam’s gold market
- Vietnam’s national gold brand – SJC;
- Connection between Vietnam’s gold price and International gold price
Model establishment - Unrestricted error correction model
- Data of oil price, gold price, US dollar index, S&P500 index; global financial crisis occurred in 2007 and 2008 as dummy variable
- Sources: EIA, World Gold Council, Federal Reserves
- Conclusion about the relationships of gold with other variables; the impact of global financial crisis
The data sources for this analysis include the crude oil price, expressed in US dollars per barrel, referred to as OIL The study selects the West Texas Intermediate (WTI) crude oil spot price as a benchmark for global oil prices, sourced from the U.S Energy Information Administration (EIA) at http://www.eia.gov The WTI crude oil price is specifically chosen for its relevance and representativeness in the context of world oil pricing.
The quality and prices of West Texas Intermediate (WTI) crude oil are typically higher than those of OPEC or Brent crude oil, as the United States is the world's largest oil consumer and relies heavily on WTI Gold prices, represented in US dollars per troy ounce, reflect the daily average from the London afternoon fix, underscoring London’s longstanding status as the premier gold trading center since the 19th century The US dollar index (USDX) measures the dollar's value against seven major currencies, including the Euro and Japanese yen, and can be accessed via the Federal Reserve Additionally, the S&P 500 index (SPX) serves as a key indicator of the US economy, based on the stock prices of 500 leading publicly traded companies, with historical data available from the Federal Reserve Bank of St Louis.
All variables, except for the dummy variable, are expressed in natural logarithms and analyzed using first differences, ensuring the data is stationary and enabling the use of significant independent variables (Sari et al., 2010; Le et al., 2011).
Data analysis
The study will conduct descriptive statistics on all series in their original level, logarithmic form, and first differences of the logarithmic levels to identify which variable exhibits the highest volatility and average return, along with their respective distribution characteristics.
Secondly, correlation matrix will be built and analyzed to know the correlation relationship among variables
Thirdly, Augmented Dickey-Fuller (ADF) test and Philips-Perron (PP) test will be used for stationary and unit root test
The cointegration test will utilize the autoregressive distributed lag (ARDL) bound test to examine the equilibrium relationship and causal linkages among variables Previous empirical studies, such as those by Engle and Granger (1987) and Johansen (1988), have employed co-integration tests that necessitate all variables to be integrated at the same order, specifically I(1) Therefore, a preliminary unit root test is essential to ascertain the order of integration for the variables in the models However, it is common for variables to exhibit different orders of integration, with some being stationary at level I(0) and others at levels I(1) or I(2), which can lead to spurious estimation results To address these challenges, this research will implement the ARDL bound test approach as proposed by Pesaran and Shin (1999) and Pesaran et al.
The unrestricted error correction model (UECM) introduced in 2001 offers several advantages over traditional methods, particularly the Autoregressive Distributed Lag (ARDL) approach Firstly, ARDL accommodates variables with different orders of integration, allowing for combinations of integrated variables of order 1 and order 0 without necessitating pre-testing for unit roots Secondly, it demonstrates greater power even with small sample sizes Additionally, ARDL is effective in analyzing both short-run and long-run relationships while providing insights into causality effects Lastly, the inclusion of dummy variables enhances the robustness of the test process (Frimpong & Oteng-Abayie, 2006; Hoque & Yusop, 2009).
Analysis method
The flow chart of statistical analysis method showed in the Figure 4.2
4.5.1 Stationary and unit root test
The ARDL bound test approach simplifies analysis by not requiring unit root tests; however, it is essential to perform these tests to confirm that all variables are either I(0) or I(1) This study will utilize the widely recognized ADF and PP unit root tests A variable is considered stationary after being differenced d times, while a variable integrated of order greater than or equal to 1 is classified as non-stationary.
However, most economic variables are cointegrated of order 1(Asteriou & Hall,
ADF test based on the choosing of following three regression forms in testing for the existence of a unit root of time series Yt:
The difference between the three forms is the deterministic elements α0 and α1T In order to choosing the best equation, it is suggested that we can first plot the data of
Stationary test - Augmented Dickey-Fuller (ADF) test;
- Unrestricted error correction models (UECM). each variable and observe the graph due to it can indicate the existence or not of the deterministic trend regressors (Binh, 2011)
In hypothesis testing for unit roots, if the absolute value of the t statistic (t*) exceeds the critical values from the Augmented Dickey-Fuller (ADF) test, the null hypothesis (Ho) cannot be rejected, indicating the presence of a unit root Conversely, if the absolute value of t* is less than the ADF critical values, the null hypothesis can be rejected, suggesting that a unit root does not exist.
However, if problem of serial correlation occurs, Phillip-Perron (PP) test conducted in a similar manner of ADF test will be used alternative
Once the order of integration for each variable is determined, the study will examine the cointegration of the variables involved Cointegration suggests that there is a long-term equilibrium relationship and causality among the variables, although it does not specify the direction of this causal relationship.
The study employs the ARDL bound test to examine cointegration, utilizing unrestricted error correction models (UECM) estimated through Ordinary Least Squares (OLS) techniques, as outlined by Pesaran et al (2001).
The bound test established by Pesaran et al (2001) utilizes the Wald test, specifically the F-statistics version, to analyze the lagged level variables on the right-hand side of the Unrestricted Error Correction Model (UECM).
By taking each of the variables in turn as a dependent variable, the research will estimate UECM models as followings:
∑ p i=1 ∝ 3i ∆lnUSDX t−i + ∑ p i=1 ∝ 4i ∆lnSPX t−i + ∝ 5 lnGOLD t−1 + ∝ 6 lnOIL t−1 + ∝ 7 lnUSDX t−1 + ∝ 8 lnSPX t−1 + ε 1t
∑ k i=1 β 3i ∆lnUSDX t−i + ∑ k i=1 β 4i ∆lnSPX t−i + β 5 lnGOLD t−1 + β 6 lnOIL t−1 + β 7 lnUSDX t−1 + β 8 lnSPX t−1 + ε 2t
∑ r i=1 Υ 3i ∆lnUSDX t−i + ∑ k i=1 Υ 4i ∆lnSPX t−i + Υ 5 lnGOLD t−1 + Υ 6 lnOIL t−1 + Υ 7 lnUSDX t−1 + Υ 8 lnSPX t−1 + ε 3t
∑ n i=1 φ 3i ∆lnUSDX t−i + ∑ n i=1 φ 4i ∆lnSPX t−i + φ 5 lnGOLD t−1 + φ 6 lnOIL t−1 + φ 7 lnUSDX t−1 + φ 8 lnSPX t−1 + ε 4t
∑ p i=1 ∝ 3i ∆lnUSDX t−i + ∑ p i=1 ∝ 4i ∆lnSPX t−i + ∝ 5 lnGOLD t−1 + ∝ 6 lnOIL t−1 + ∝ 7 lnUSDX t−1 + ∝ 8 lnSPX t−1 + ∝ 9 D1G + ∝ 10 D2O + ∝ 11 D3U + ∝ 12 D4S + ε 1t
∑ k i=1 β 3i ∆lnUSDX t−i + ∑ k i=1 β 4i ∆lnSPX t−i + β 5 lnGOLD t−1 + β 6 lnOIL t−1 + β 7 lnUSDX t−1 + β 8 lnSPX t−1 + β 9 D1G + β 10 D2O + β 11 D3U + β 12 D4S + ε 2t
∑ r i=1 Υ 3i ∆lnUSDX t−i + ∑ k i=1 Υ 4i ∆lnSPX t−i + Υ 5 lnGOLD t−1 + Υ 6 lnOIL t−1 + Υ 7 lnUSDX t−1 + Υ 8 lnSPX t−1 + Υ 9 D1G + Υ 10 D2O + Υ 11 D3U + Υ 12 D4S + ε 3t
∑ n i=1 φ 3i ∆lnUSDX t−i + ∑ n i=1 φ 4i ∆lnSPX t−i + φ 5 lnGOLD t−1 + φ 6 lnOIL t−1 + φ 7 lnUSDX t−1 + φ 8 lnSPX t−1 + φ 9 D1G + φ 10 D2O + φ 11 D3U + φ 12 D4S + ε 4t
- lnGOLDtis the log of the London gold price;
- lnOILt is the log of international crude oil prices which measures the price of West Texas Intermediate (WTI) crude oil;
- lnUSDXt is the log of US dollar index;
- lnSPXt is the log of S&P 500 index;
- ∆ is the first difference operator;
- p, k, r, n are the lag lengths and determined by the Akaike Information Criterion (AIC) (supporting by Eviews 6 software);
- αxi, βxi, 𝛶xi, φxi (x = 1 to 4) are the short-run coefficients;
- αx, βx, 𝛶x, φx (x = 5 to 8)are the long-run coefficients in Group 1;
- αx, βx, 𝛶x, φx (x = 5 to 12)are the long-run coefficients in Group 2;
- εxt (x = 1 to 4) are white noise errors;
- Crisis = 1, over the period 01/08/2007 – 31/12/2008, 0 elsewhere
The null hypothesis of no cointegration in the long run for Group 1 is defined by several equations: Equation 1 states H0: α = α = α = α = 0; Equation 2 specifies H0: β5 = β6 = β7 = β8 = 0; Equation 3 presents H0: γ5 = γ6 = γ7 = γ8 = 0; and Equation 4 indicates H0: φ5 = φ6 = φ7 = φ8 = 0 For Group 2, the null hypothesis of no cointegration will be analyzed through two distinct cases.
Case 1: crisis does not affect the relationship among variables in long-run i Equation 1: H0: α9 = α10 = α11 = α12 = 0 ii Equation 2: H0:β9 = β10 = β11 = β12 = 0 iii Equation 3: H0:γ9 = γ10 = γ11 = γ12 = 0 iv Equation 4: H0:φ9 = φ10 = φ11 = φ12 = 0
Case 2: the long-run relationship among variables does not exist when crisis occurs i Equation 1: H0: α5 = α6 = α7 = α8 = α9 = α10 = α11 = α12 = 0 ii Equation 2: H0:β5 = β6 = β7 = β8 = β9 = β10 = β11 = β12 = 0 iii Equation 3: H0:γ5 = γ6 = γ7 = γ8 = γ9 = γ10 = γ11 = γ12 = 0 iv Equation 4: H0:φ5 = φ6 = φ7 = φ8 = φ9 = φ10 = φ11 = φ12 = 0
To examine the long-run relationship among variables, the general F-statistics test is employed, analyzing all variables in their levels The outcomes of this test are then compared against two distinct critical values derived from Peseran et al (2001).
The F-statistics have two critical value bounds that determine the acceptance or rejection of the null hypothesis regarding cointegration If the computed F-statistic falls below the lower critical value, the null hypothesis of no cointegration remains accepted Conversely, if the F-statistic exceeds the upper critical value, the null hypothesis is rejected, indicating the presence of a long-run cointegration relationship among the model's variables.
However, if the computed F-statistic is within the bounds, the test is inconclusive
The Critical values are taken from Table CI(iii) paged 300 of Pesaran et al (2001), showed in Table 4.1
Table 4.1: Asymptotic critical value bounds for the F-statistics k At 5% level of significant At 10% level of significant
Note: k is the total number of independent variables in the model
Choosing the optimal lag length is crucial in this test, which will be achieved by estimating a VAR model In this model, all variables in their differenced form are treated as endogenous, while the first lag of all variables in log transformation serves as exogenous variables The Akaike Information Criterion (AIC) will be utilized for this purpose.
To ensure the reliability of the model, a serial correlation test will be conducted to verify that the residuals are not serially correlated, while the inverse roots test will assess the dynamic stability of the model.
If cointegration is established, the next step involves calculating the long-run coefficients Subsequently, short-run dynamic parameters will be obtained through error correction models (ECM), where error correction terms are derived from the ordinary least squares (OLS) estimation of the variables' log-transformed levels The ECM indicates the speed of adjustment required to return to long-run equilibrium following short-run shocks, with the ECM coefficient expected to be negative and significant A larger ECM coefficient signifies a quicker adjustment back to long-run equilibrium.
This chapter clearly outlines the reasons for variable selection and data sources, alongside the design of the econometric model To verify the stationarity and cointegration of the data, the ADF and PP tests, as well as the ARDL bound test, have been employed.
RESEARCH RESULTS
Descriptive statistics
The descriptive statistics of all series in level, log and first difference of log level is showed in Table 5.1
Table 5.1: Descriptive statistics of all series
The mean, or average, indicates the central tendency of variables Among various assets, the stock index exhibits the highest mean in both its original and log-transformed levels Conversely, when examining the first difference of log values, gold prices demonstrate the highest positive mean, closely followed by oil prices.
US dollar exchange rate and stock index have negative mean
The coefficient of standard deviation reveals that gold prices exhibit the highest volatility among assets, both in level and log transformations, with oil prices following closely In the first difference of log, oil prices surpass both the stock index and gold prices in volatility, while the US dollar exchange rate ranks last.
The skewness, kurtosis and Jarque-Bera, probabilities indicate that all variables are significantly non-normal distribution.
Correlation matrix
The correlation matrix of the logged variables will be presented in Table 5.2
LNSPX 0.062799 0.600416 -0.449158 1.000000 Gold price has moderate positive correlation with oil price, moderate negative correlation with US dollar exchange rate, very weak positve correlation with stock index However, oil price and US dollar exchange rate have the highest and negative correlation (about -0.828) So, multicollinearity may exist But the paper will ignore this problem due to the model is for forecasting purpose only (Hoai,
Stationary and unit root test
The results of both the ADF and PP tests are reported in Table 5.3 and Table 5.4
The result in Table 5.3 indicates that most variables were not stationary in levels by both ADF and PP tests So, the null hypothesis of non-stationary cannot be rejected
It means that there is an existence of unit root in all variables at levels
Table 5.3: Unit root test for stationary at level
Intercept Intercept & Trend Intercept Intercept & Trend
The findings in Table 5.4 indicate that all variables are stationary at the first difference, as confirmed by both ADF and PP tests Consequently, the null hypothesis of non-stationarity is rejected, suggesting that all variables are integrated of order one, denoted as I(1).
It is as the same result of Harmmoudeh et al (2008), Sari et al (2010)
Table 5.4: Unit root test for stationary at first difference
Intercept Intercept & trend Intercept Intercept and Trend
Note: * indicates significance at 1% level.
Cointegration analysis
The result of lag order selection criteria via vector autoregression estimates (VAR) is presented in Table 5.5
Table 5.5: VAR lag order selection criteria
Endogenous variables: DLNGOLD DLNOIL DLNUSDX DLNSPX Exogenous variables: C LNGOLD(-1) LNOIL(-1) LNUSDX(-1) LNSPX(-1)
Lag LogL LR FPE AIC SC HQ
Note: * indicates lag order selected by the criterion
Basing on AIC values of the Table 5.5, the paper intends to use maximum lag of 10 in the thesis
To verify the serial independence of the model's errors, the study employs serial correlation LM tests, where the null hypothesis posits the absence of serial correlation at a lag order of h The findings are presented in Table 5.6.
At lag 10, the p-value of 0.7836 exceeds the 5% significance level, indicating that the null hypothesis cannot be rejected This suggests that there is no serial correlation present at this lag order.
10 and the lag of 10 is suitable to use in the thesis
Table 5.6: Serial correlation test’s result
VAR Residual Serial Correlation LM Tests Null Hypothesis: no serial correlation at lag order h
The paper involves checking dynamic stability of ARDL model by using Inverse roots graph This graph is illustrated in Figure 5.1
The graph shows that all roots have absolute value less than one and lie inside the unit circle This indicates that the model is stable
The calculated F-statistics and lag structure are presented in Table 5.7 and Table 5.8
Table 5.7: Bounds test procedure results without crisis interaction
Cointegration hypothesis Lag structure F-statistics Outcome
F(lngold/lnoil, lnusdx, lnspx) 9-4-5-10 0.323849 No cointegration F(lnoil/lngold, lnusdx, lnspx) 1-8-1-4 1.997150 No cointegration F(lnusdx/lngold, lnoil, lnspx) 1-5-6-10 2.205181 No cointegration F(lnspx/lngold, lnoil, lnusdx) 10-8-9-9 2.090775 No cointegration
Note: the critical value bounds at 5% and 10% level of significant are [2.86, 4.01], [2.45, 3.52] respectively for k = 4
Table 5.8: Bounds test procedure results with crisis interaction
Inverse Roots of AR Characteristic Polynomial
Cointegration hypothesis Lag structure F-statistics Outcome Case 1
F(lngold/lnoil, lnusdx, lnspx) 9-4-5-10 4.418209 Cointegration F(lnoil/lngold, lnusdx, lnspx) 1-8-1-4 7.399536 Cointegration F(lnusdx/lngold, lnoil, lnspx) 1-5-6-10 2.744189 Inconclusion F(lnspx/lngold, lnoil, lnusdx) 10-8-9-9 1.283832 No cointegration
F(lngold/lnoil, lnusdx, lnspx) 9-4-5-10 2.37369 Inconclusion F(lnoil/lngold, lnusdx, lnspx) 1-8-1-4 4.719896 Cointegration F(lnusdx/lngold, lnoil, lnspx) 1-5-6-10 2.483260 Inconclusion F(lnspx/lngold, lnoil, lnusdx) 10-8-9-9 1.688787 No cointegration
Note: the critical value bounds at 5% and 10% level of significant are [2.86, 4.01], [2.45, 3.52] respectively for k = 4 in Case 1; [2.22, 3.39], [1.95, 3.06] respectively for k = 8 in Case 2
The results presented in Table 5.7 indicate that the null hypothesis of no cointegration cannot be rejected, suggesting that a long-run equilibrium relationship does not exist in any of the equations analyzed This finding contrasts with the conclusions drawn by Harmmoudeh et al (2008) and Narayan et al (2010) Consequently, the subsequent steps of cointegration outlined in Item 4.5 will not be pursued due to the absence of cointegration.
The findings in Table 5.8 indicate that crises influence the long-term relationship between gold prices and oil prices, the US dollar exchange rate, and the stock market, although a definitive long-run relationship cannot be established Conversely, crises do create a long-term relationship between oil prices and gold prices, the US dollar exchange rate, and the stock market, signifying that during crises, these variables tend to move in tandem.
During a crisis, the US dollar exchange rate and stock market typically respond first, followed by changes in oil prices However, the relationship between the US dollar exchange rate and other variables during a crisis remains inconclusive Additionally, crises do not alter the relationship between the stock market and other variables, indicating that a long-term relationship does not exist in such scenarios.
The ADF and PP tests indicate that all variables are non-stationary at level but become stationary after first differencing Furthermore, ARDL bound tests reveal a lack of cointegration among the variables However, during periods of crisis, these tests suggest that crises significantly influence the long-term relationships between gold prices and oil prices, the US dollar exchange rate, and the stock market, as well as between oil prices and these same variables Additionally, it is observed that gold prices, the US dollar exchange rate, and the stock market tend to lead, with oil prices following their movements.
CONCLUSION AND POLICY IMPLICATIONS
Main findings
This research utilizes correlation and cointegration techniques to analyze the relationship between Vietnam's gold price and global gold prices, as well as to explore how global gold prices correlate with WTI crude oil prices, the US dollar index, and the S&P 500 index, which are noted in the literature as influential factors The study specifically examines data from 2007 to 2008 as a dummy variable to assess its impact on the relationships among these variables Key findings will be summarized accordingly.
Vietnam's gold market closely mirrors global trends, exhibiting a strong positive correlation where fluctuations in world gold prices typically influence local prices Despite this relationship, discrepancies persist between the two markets, driven by domestic demand and supply factors, as well as varying state policies over time.
- World’s gold price has moderate positive correlation with oil price, moderate correlation with US dollar index, very weak positive correlation with stock index
- All variables in research are integrated of order one
- When the dummy variable ignored in models, by taking ARDL bound test, the paper does not see any evidence of cointegration existed among variables
The inclusion of dummy variables in long-run models using the ARDL bound test reveals that financial crises significantly impact the relationships between gold prices and oil prices, the US dollar index, and the S&P 500, although it remains inconclusive which variable leads the movement Additionally, financial crises also affect the correlation between oil prices and gold prices, the US dollar index, and the S&P 500, indicating that all other variables tend to move first, with oil prices following suit.
Policy implications
The empirical results reveal a lack of cointegration among gold prices, oil prices, the US dollar index, and the S&P 500, suggesting that these variables are significantly influenced by external factors such as government policies, inflation, and the political conditions of various countries Additionally, gold prices are affected by the jewelry industry and central bank interventions aimed at managing foreign reserves and exchange rates Consequently, the State Bank of Vietnam cannot rely solely on the movements of these variables for decision-making.
WTI price, US dollar index and S&P 500 to forecast the movement of world gold price for making their decision in selling or buying gold
The empirical findings indicate that financial crises significantly influence the correlation between gold prices and key economic indicators such as oil prices, the US dollar index, and the S&P 500 Consequently, the State Bank of Vietnam should exercise caution in its buying or selling decisions during similar future economic shocks.
Limitation
Although the paper has explained carefully about reason of choosing variables, analysis method as well as economic model for research, there are some limitations
Firstly, the period from 2007 to 2012 is short due to daily data for Vietnam gold prices are not adequate
Secondly, the data set for SJC price might be unreliable due to it is manually collected via website of Saigon Giai Phong newspaper
Thirdly, the paper may omit some important factors affect to gold price due to there is no cointegration among gold price, oil price, US dollar index, S&P 500.
Future research
Future research should broaden its scope by incorporating additional factors such as inflation, the demand and supply of gold, and the demand and supply of oil, while also employing various cointegration methods to verify the findings.
An, V., Minh, N & Trung, T (2013, May 30) Vi sao gia vang trong nuoc va the gioi chenh nhieu? The Thanh Nien Retrieved from http://www.thanhnien.com.vn
Asteriou, D & Hall, S.G (2007) Applied Econometrics: A modern approach using
Eviews and Microfit, Revised Edition Palgrave Macmillan
Amano, R.A & Norden, S.V (1998) Oil prices and the rise and fall of the US real exchange rate The Journal of International Money and Finance, 17(1998), 299-316
Baur, D.G & Lucey, B.M (2010) Is gold a hedge or a safe haven? An analysis of stocks, bonds and gold The Financial Review, 45(2), 217-229
Binh, T.B (2011) Energy consumption and economic growth in Vietnam:
Threshold cointegration and causality Analysis International Journal of Energy Economics and Policy, 1 (1), 1-17
Bhunia, A & Das, A (2012) Association between Gold prices and Stock market returns: Empirical evidence from NSE Journal of Exclusive Management Science, 1 (2)
Capie, F., Mills, T.C & Wood, G (2005) Gold as a hedge against the dollar
Journal of International Financial markets, Institutions & money Retrieved from http://ejournal.narotama.ac.id/files/Gold%20as%20a%20hedge%20against
Gold prices are expected to increase as the value of the dollar declines and oil prices experience a rally, according to a survey conducted by Bloomberg.
Coudert, V & Raymond-Feingold, H (2011) Gold and financial assets: Are there any safe havens in bear markets? Economics Bulletin, 31(2), 1613-1622
Cunado, J & Perez de Gracia, F (2003) Do oil price shocks matter? Evidence for some European countries The Quarterly Review of Economics and Finance, 25, 137-154
Cunado, J & Perez de Gracia, F (2004) Oil prices, economic activity and inflation: evidence for some Asian countries The Quarterly Review of Economics and Finance, 45, 65-83
Eagle and Granger (1987) discuss the significance of cointegration and error correction in econometrics, providing a comprehensive framework for representation, estimation, and testing In a related context, Farchy and Blas (2010) highlight the increasing importance of gold as a reserve asset for central banks, emphasizing its growing role in financial stability and investment strategies.
Frimpong, J.M & Oteng-Abayie, E.F (2006) Bounds testing approach: an examination of foreign direct investment, trade and growth relationships (MPRA Paper No 352)
Financial crisis of 2007–08 (n.d.) In Wikipedia Retrieved September 9, 2013 from http://en.wikipedia.org/wiki/Financial_crisis_of_2007%E2%80%9308
Gregory, A.W & Hansen, B.E (1996) Residual-based tests for cointegration in models with regime shifts Journal of Econometrics, 70, 99-126
Gilmore, C.G., McManus, M.G., Sharma, R & Tezel, A (2009) The dynamics of gold prices, gold mining stock prices and stock market prices comovements, Research in Applied Economics, 1(1), 1-19
Hammoudeh, S., Sari, R & Ewing, B.T (2008) Relationships among strategic commodities and with financial variables: A new look Contemporary Economic Policy, 27(2), 251-264
Hoque, M.M & Yusop, Z (2009) Impacts of trade liberalisation on aggregate import in Bangladesh : An ARDL Bounds test approach Journal of Asian
Hoai, N.T (2006) Lecture 5: Multicollinearity in Econometrics Retrieved from
Fullbright Economics Teaching Program: http://www.ctu.edu.vn
Jaffe, J.F (1989) Gold and gold stocks as investments for institutional portfolios
Jaeger, M., (2012, July 17) Does gold set the price for oil? The Washington Times
Retrieved from http://communities.washingtontimes.com
Johansen, S (1988) Statistical analysis of cointegration vectors Journal of
Johansen, S & Juselius, K (1990) Maximum likelihood estimation and inference on cointegration with applications to the demand for money Oxford Bulletin of Economics and Statistics, 52, 169-209
Kim, M H., & Dilts, D A (2011) The relationship of the value of the Dollar, and the Prices of Gold and Oil: A Tale of Asset Risk Economics Bulletin, 31, 1151-1162
Khanh, T T.Q (2010, October 23) Nhieu he luy khi gia vang tang cao The
Vnexpress’s Retrieved from http://vnexpress.net/gl/kinh- doanh/2010/10/3ba21f6b/
Le, T.H & Chang, Y (2011) Oil and Gold: Correlation and Causation? (Deponcen working papers series 2011/22)
Lizardo, R.A & Mollick, A.V (2010) Oil price fluctuations and U.S dollar exchange rates Energy Economics, 32(2010), 399-408
Market price (n.d.) In Wikipedia Retrieved June 25, 2013, from http://en.wikipedia.org/wiki/Market_price
Malliaris, A.G & Mallliaris, M (2011) Are oil, gold and the euro inter-related?
Time series and neural network analysis (MPRA Paper No 35266)
Retrieved from Munich Personal RePEc Archive
Melvin, M., & Sultan, J (1990) South African political unrest, oil prices, and the time varying risk premium in the gold futures market Journal of Futures markets, 10, 103-111
Narayan, P.K, Narayan, S & Zheng, X (2010) Gold and oil futures markets : Are market efficient? Applied Energy, 87 (10), 3299-3303
Pesaran, M.H & Shin, Y (1999) An Autoregressive Distributed Lag Modeling
Approach to Cointegration Analysis In: Econometrics and Economic Theory in the 20th Century: The Ragnar Frisch Centennial Symposium,
Chapter 11, Strom S (Ed.) Cambridge University Press, Cambridge, New York
Pesaran, M.H., Shin, Y & R.J Smith R.J (2001) Bound testing approaches to the analysis of level relationship Journal of Applied Economics, 16, 289-326
Pescatori, A & Mowry, B (2008, December 9) Do oil price directly affect the Stock market? The Federal Reserve Bank of Cleveland Retrieved from http://www.clevelandfed.org/research/trends/2008/0908/04ecoact.cfm
Phillips, J., (2010, March 26) The gold price will be affected by the oil price in
Future – How? The Streetwise report’s Retrieved from http://www.theaureport.com
Phillips, J., (2012, May 4) Why Central Banks are Buying Gold The
Bullionvault’s Retrieved from http://goldnews.bullionvault.com/
Cong, N T (2012) Dien bien gia vang va nhan dinh xu huong cuoi nam Retrieved
June 27, 2013, from http://www.vietinbank.vn/web/home/vn/research/12/dien-bien-gia-vang-va- nhan-dinh-xu-huong-cuoi-nam.html
Sari, R., Hammoudeh, S & Soytas, U (2010) Dynamics of oil price, precious metal prices, and exchange rate Energy Economics, 32 (2), 351-362
Smith, G (2001) The price of gold and stock price indices for the United States,
S&P 500 (n.d.) In Wikipedia Retrieved December 27, 2012, from http://en.wikipedia.org/wiki/S%26P_500
.U.S Energy Information Administration (2013, February) Short-term Energy
Outlook (STEO) Retrieved from http://www.eia.gov/forecasts/steo/report/global_oil.cfm
Vnexpress (2013, January 27) Thi truong vang ky vong vao “nhac truong” The
Vnexpress’s Retrieved from http://vnexpress.net/gl/kinh-doanh/vi- mo/2013/01/thi-truong-vang-ky-vong-vao-nhac-truong/
Hoang, V.Q (2004) Analyses on Gold and US Dollar in Vietnam’s Transitional
Economny Centre Emile Bernheim WP-CEB No 04/033
Wang, M.L, Wang, C.P & Huang, T.Y (2010) Relationships among Oil Price,
Gold Price, Exchange Rate and International Stock Markets The International Research Journal of Finance and Economics, 47(2010)
West Texas Intermediate (n.d.) In Wikipedia Retrieved December 27, 2012, from http://en.wikipedia.org/wiki/West_Texas_Intermediate
World Gold Council (2012, August 14) Gold Demand Trends Q2 2012 Retrieved from http://www.gold.org
Website of the Federal Reserve System of the United States: http://www.federalreserve.gov/releases/h10/Summary/indexnc_m.htm
Website of Ben Thanh Gold Company: http://benthanhgold.vn/thong-tin-hang- hoa/thong-tin-hang-hoa/9962-cach-tinh-gia-vang-viet-nam-tu-gia-vang-the- gioi.html
Vector Autoregression Estimates Standard errors in ( ) & t-statistics in [ ]
Appendix B: Bound test result Equation 1 of Group 1: LS dlngold c dlngold(-1 to -8) dlnoil(to -4) dlnusdx(to -5) dlnspx(to -10) lngold(-1) lnoil(-1) lnusdx(-1) lnspx(-1);c(32)=c(33)=c(34)=c(35)=0 Wald Test:
Test Statistic Value df Probability
Normalized Restriction (= 0) Value Std Err
Equation 2 of Group 1: Ls dlnoil c dlnoil(-1) dlngold(to -8) dlnusdx(to -1) dlnspx(to -4) lngold(-1) lnoil(-1) lnusdx(-1) lnspx(-1); c(19)=c(20)=c(21)=c(22)=0
Test Statistic Value df Probability
Normalized Restriction (= 0) Value Std Err
Equation 3 of Group 1: Ls dlnusdx c dlnusdx(-1) dlngold(to -5) dlnoil(to -6) dlnspx(to -10) lngold(-1) lnoil(-1) lnusdx(-1) lnspx(-1); c(27)=c(28)=c(29)=c(30)=0 Wald Test:
Test Statistic Value df Probability
Normalized Restriction (= 0) Value Std Err
Equation 4 of Group 1: dlnspx c dlnspx(-1 to -10) dlngold(to -8) dlnoil(to -9) dlnusdx(to -9) lngold(-1) lnoil(-1) lnusdx(-1) lnspx(-1); c(41)=c(42)=c(43)=c(44)=0
Test Statistic Value df Probability
Normalized Restriction (= 0) Value Std Err
Equation 1 of Group 2: LS dlngold c dlngold(-1 to -8) dlnoil(to -4) dlnusdx(to -5) dlnspx(to -10) lngold(-1) lnoil(-1) lnusdx(-1) lnspx(-1) d1g d2o d3u d4s; c(36) = c(37) = c(38) = c(39) =0 Wald Test:
Test Statistic Value df Probability
Normalized Restriction (= 0) Value Std Err
Test Statistic Value df Probability
Normalized Restriction (= 0) Value Std Err
Equation 2 of Group 2: Ls dlnoil c dlnoil(-1) dlngold(to -8) dlnusdx(to -1) dlnspx(to -4) lngold(-1) lnoil(-1) lnusdx(-1) lnspx(-1) d1g d2o d3u d4s; c(23) = c(24) = c(25) = c(26) = 0 Wald Test:
Test Statistic Value df Probability
Normalized Restriction (= 0) Value Std Err
Test Statistic Value df Probability
Normalized Restriction (= 0) Value Std Err
Equation 3 of Group 2: Ls dlnusdx c dlnusdx(-1) dlngold(to -5) dlnoil(to -6) dlnspx(to -10) lngold(-1) lnoil(-1) lnusdx(-1) lnspx(-1) d1g d2o d3u d4s; c(31) = c(32) = c(33) = c(34) = 0 Wald Test:
Test Statistic Value df Probability
Normalized Restriction (= 0) Value Std Err
Test Statistic Value df Probability
Normalized Restriction (= 0) Value Std Err
Equation 4 of Group 2: dlnspx c dlnspx(-1 to -10) dlngold(to -8) dlnoil(to -9) dlnusdx(to -9) lngold(-1) lnoil(-1) lnusdx(-1) lnspx(-1) d1g d2o d3u d4s; c(45)=c(46)=c(47)=c(48) = 0 Wald Test:
Test Statistic Value df Probability
Normalized Restriction (= 0) Value Std Err
Test Statistic Value df Probability
Normalized Restriction (= 0) Value Std Err