ABSTRACT:
From 2020 to September 2024, gold prices in Vietnam have doubled, with no signs of stabilization. This study employs the Vector Autoregression (VAR) model to analyze the influence of key macroeconomic factors, including global gold prices, global crude oil prices, and the VN-Index, on fluctuations in Vietnam’s gold market. The findings reveal that Vietnam’s gold prices are primarily driven by global gold prices, while changes in crude oil prices and the VN-Index exert minimal influence. These insights contribute to a better understanding of gold price dynamics in Vietnam, offering implications for investors and policymakers.
Keywords: VAR model, gold price, gold, global gold prices, gold market, VN-Index.
1. Introduction
Gold has always been considered to be one of the most prestigious metal on the planet. It is not only scarce but also extremely multifunctional. It can be used as semiconductor in the chip industry, held by humans as jewelry, and to investors, gold must be one of the best if not the best choice to hedge during financial crisis or other crisis such as a pandemic or war. During the past 24 years of the 21st century, the world has witnessed a lot of catastrophes such as Covid-19 or the Russian-Ukrainian war. These events certainly had a lot of negative effects on the global financial market. Due to its stability in terms of price, the demand for gold has been increasing dramatically in the past few years, as a result, the gold market is full of lively activity. In Vietnam, the SJC gold has been doubled in price since 2020 and still does not seem to have stop. Under these circumstances, due to its stability during crisis time, people are investing in gold more than ever before. And so, a research about gold price is totally necessary. In this paper we will focus on determining what are the macro factors that affect on the gold price in Vietnam.
2. Literature review
In previous research such as E.Tully, B.M. Lucey.(2007), using data from 1983 to 2003, the paper analyze 5 factors including FTSE100, Dollar Index, America interest rate and, British Pounds index British CPI, the paper conclude that in most of the time, only Dollar Index affected the gold price. L.K.T.A Parashar, R. Singh.(2014) also have almost the same approach by using Granger Causality model to analyze the effect of three factors including INR/USD, FII, S&P500 on India gold price. The paper concludes that none of those factors above to has an effect on the gold price. Finally, Le Thanh Ha.(2022) applied TVP-VAR model to find the link between the oil price, stock, and gold price and point out that in the time off uncertainty the oil market constantly conveys volatility shocks to other markets. In general, all of above research analyze the link between gold with Dollar, oil price and stock index. This article’s primary objective is to assess the macro factors’ effect on the Vietnam gold during the war time by using two approach which is the linear regression model and the VAR model. After referring to previous studies, the most selected variable is the stock index, and the oil index. To analyze the fluctuations of the gold price in Vietnam, after reference previous studies, this paper will focus on using three macro factors including VNI (the stock index), Oil price, and the World gold price since Vietnam is an open market which means that Vietnam commodity price would heavily be affected by the World commodity price. The article will answer two main questions:
What are the factors that affect to the gold price in Vietnam and how much these factor contribute to the change in the gold price?
3. Data and methodology
The essay will use the oil volatility index (Oil), World gold volatility index (World) and VNI volatility index (VNI) to analyze the volatility in Vietnam Gold price index. The data is sourced from January 3, 2022, to May 31 2024.
3.1. Unit Root test
Unit Root test is used to test for the stationary of the time series. The essay will utilize the Augmented Dickey-Fuller Test or ADF. The null hypothesis of the test states that the Unit Root does exist and so the series is non-stationary
Formular:
3.2. Regression model
The model has Vietnam gold price (VN) as the dependent variable, while World price (World), VNI and Oil are independent variable
3.3. Collinearity test
Collinearity happens when the model have at least two variables that depend on each other or in other word is to have a linear correlation. When collinearity happens, the model is less accurate, therefore, testing for collinearity is necessary
3.4. Heteroskedasticity
This is an essential condition to use regression model, this essay will use Breusch – Pagan to test for heteroskedasticity:
Using dependent variables xki
4. VAR
4.1. Determine the optimal lag length
Choose the optimal number of lags for the model, avoid neglecting important variables via using 4 different criteria checking for maximum number of 10 lags:
-
- FPE: Final Prediction Error
- AIC: Akaike’s information criterion
- SBIC: Schwarz’s Bayesian information criterion
- HQIC: Hannan and Quinn information criterion
4.2. Estimate VAR model
Vector Automatic Recovery (VAR) or VAR is extended form of regression model.
5. Result
5.1. Descriptive Statistic
Table 1: The range, mean, max, min, and median value of variablesh
Source: The Author
Table1 illustrates the range, mean, max, min, and median value of variables. It can be seen from the table that the gold price fluctuates between -4.27% (min) and 7.6% (max). Similarly, the world gold price fluctuates from -2.8% to 3.1% (it can be noted that the fluctuation amplitude is lower than the domestic gold price). VNI fluctuates from approximately -5% to 4.8%. And oil fluctuates the most from 12% to 8.5% (considering the context of 2022, it is not difficult to understand the reason behind it).
Table 2: The correlation coefficient matrix between variables
Source: The Author
The Table 2 is the correlation coefficient matrix between variables, by observing the VN column, we can see that World (world gold price) has the highest correlation with the domestic gold price, VNI is right behind with 0.065, note that 0.065 is also a much smaller correlation coefficient compare to World but not a small number at all, and finally Oil is at the bottom with just only 0.018. Moreover, Oil and World have a notable correlation (0.19) leading to the fact that these two can make the collinearity error happen.
5.2. Test for stationary
Table 3: The result of the ADF test
Source: The Author
The four tables above are the result of the ADF test. Using ADF test to check the stationarity of the data series, we get the above result table, according to which the null hypothesis of ADF test states, the time series is not stationarity. Because the p-value of all the above tests is less than 1% then we can reject null hypothesis, which means that all the above time series are stationarity.
5.3. Regression model
As the result table point out, the coefficient is the biggest with a value of ~ 0.185, which mean that, for every 1 unit increase in the value of the World, the VN value increase 0.185 unit. and this also is the only coefficient with a significant statistical value in the model with Pr of 3.48e-07 (smaller than 1%). The remaining variables including the free coefficient (Intercept), the coefficients of VNI and Oil are all statistically insignificant.
Model overview: The model has a p-value of 6.008e-07 (less than 1%), we conclude that the model is highly reliable, while at the same time, looking at the R-squared index, the independent variables (including the variables World, VNI, Oil) can only explain ~5.7% of the variation of the dependent variable (the dependent variable is VN)
After having the model, we need to do some more tests to check the correctness of the regression model.
5.4. VIF test
Conclusion: All are less than 5 => all are acceptable, the model does not have multicollinearity.
5.5. Heteroscedascity
Breusch-Pagan test, test for heteroscedasticity. The null hypothesis of the test state that the heteroscedasticity occurs, and so because p-value=0.395 (greater than 5%) => this error phenomenon occurs.
5.6. Autocorrelation
Breusch-Godfrey test, testing for the autocorrelation of errors. The null hypothesis of the test is that autocorrelation does not occur, but since the p-value of the lags from 1 to 3 periods are all less than 1% => the autocorrelation error occurs
Conclusion: Using a regression model can give an initial view, but due to the change in error variance and autocorrelation, a stronger model is needed for analysis.
5.7. VAR model
5.7.1. Optimal lag length
To optimal the accuracy, the index of the criteria must be as small as possible, and as observed, Lag 1 satisfies all 4/4 criteria AIC, HQ, FPE. So, we will only take 1 lag for the model.
5.7.2. Test for autocorrelation
Because p-value of the test equal to 2% (less than 5%), the model continues to have errors. The essay will adjust lag to 2 (Since according to the criteria in the lag selection table, lag 2 is the most suitable lag beside lag 1)
Test for the second time with lag length is 2
This time the p-value is even greater than 10%, which means that the model does not have autocorrelation, so lag 2 will be chosen to continue the analysis.
5.7.3. Variance decomposition
In the early period, the VN variable explained 92.71 percent of its own variation, while the World variable explained 6.5% of the variation in the VN variable, followed by Oil with ~ 0.064% and VNI come last with ~ 0.0064%. In the later periods, the World variable continued to explain more and in period 10 (long term), the World variable could explain up to 6.8%, while the VNI variable only reached approximately 0.01% and Oil was 0.066%. Conclusion: after variance decomposition process, it can be deducted that, apart from the World variable, there is no value that affects the VN variable, in other words, the world gold price is the only variable that seems to be able to affect the gold price.
5.7.4. Granger test
Conclusion: The above model has hypothesis H1 states that Vietnam gold price affects world gold price, while hypothesis H2 is that Vietnam gold price does not affect world gold price. Since the P-value of hypothesis H2 is 0.9969, there is no evidence to reject hypothesis H2 => Vietnam gold price does not have Granger effect on world gold price.
World gold price has an impact on Vietnam gold price because P-value of hypothesis 2 is less than 1% so hypothesis H2 is rejected, meaning world gold price has Granger effect on gold price in Vietnam.
6. Conclusions
Since the exchange rate and CPI are heavily controlled in Vietnam, Oil price, stock market and world gold price would be more likely to be able to explain the volatility of the domestic gold price, however, after applying regression model and VAR model to analyze, the essay found that World gold price is the only macro factor that has an significance impact on Vietnam gold price, moreover, the essay found that the World gold price contribute to six percent of change in the domestic gold price.
REFERENCES:
1. Ministry of Finance. Available at: https://www.mof.gov.vn/.
2. Ministry of Industry and Trade. Available at: https://moit.gov.vn/.
3. Le Thanh Ha (2022). Are digital business and digital public services a driver for better energy security? Evidence from a European sample. Environmental Science and Pollution Research, 29, 27232–27256.
4. World Gold Council. Available at: https://www.gold.org/goldhub/data/gold-prices
5. U.S. Energy Information Administration. Available at: https://www.eia.gov/dnav/pet/hist/LeafHandler.ashx?n=pet&s=mcrfpus.
6. Edel Tully, Brian M. Lucey (2007). A power GARCH examination of the gold market. Research in International Business and Finance, 21(2), 316-325.
7. Dr. L.K. Tripathi, Arpan Parashar, Dr. Rajendra Singh (2014). Global factors & gold price in india- a causal study. International Journal of Advanced Research in Management and Social Sciences, 3(7), 161-180.
Ứng dụng mô hình var để đánh giá tác động của một số yếu tố vĩ mô đến giá vàng tại Việt Nam
ThS. Phạm Thị Thanh Huyền
Khoa Kế toán - Tài chính, Trường Đại học Hải Phòng
TÓM TẮT:
Tại Việt Nam, kể từ năm 2020 đến tháng 9/2024 giá vàng đã tăng gấp đôi. Bằng cách sử dụng mô hình VAR, nghiên cứu này sẽ đánh giá tác động của các nhân tố vĩ mô bao gồm giá vàng thế giới, giá dầu thô thế giới và chỉ số VN-Index tới sự biến động của giá vàng Việt Nam. Kết luận của nghiên cứu cho thấy giá vàng Việt Nam bị tác động chủ yếu bởi giá vàng thế giới; trong khi đó, sự biến động của giá dầu thô và chỉ số VN-Index không gây ảnh hưởng nhiêu đến thị trường vàng Việt Nam
Từ khóa: mô hình VAR, giá vàng, vàng, giá vàng thế giới, thị trường vàng, chỉ số VN-Index.
[Tạp chí Công Thương - Các kết quả nghiên cứu khoa học và ứng dụng công nghệ, Số 6 tháng 2 năm 2025]
