Precious metals as safe haven or hedge
Evidence from the Netherlands
Abstract: This paper examines whether precious metals can perform as a safe haven or hedge
for the Dutch stock market in the last two decades. Two models are used to investigate the safe haven and hedge characteristics. The first model provides the quantiles for extreme market conditions and the second model provides the predefined crisis periods. Gold and silver can perform as a strong safe haven during certain extreme market conditions, although the safe haven characteristics are time-dependent. During the global financial crisis of 2008, gold and silver can perform as a strong safe haven. Furthermore, precious metals cannot perform as a hedge for the Dutch stock market.
JEL classification: G10, G11, G14, G15
Keywords: precious metals, safe haven, hedge, stock markets
University of Groningen
Faculty of Economics and Business MSc Finance
Author: Spurgeon Maria
Student Number: s3682455 Supervisor: A.G. Schertler Second Evaluator: I. Souropanis
1. Introduction
From the start of the financial crisis in 2007 the price of gold has increased 132%.1 It is remarkable to see the performance of gold increasing, while the prices of financial assets (especially stocks) during the crisis decreased dramatically. Research by Goetzmann, Li, and Rouwenhorst (2005) finds that due to globalization the correlation between most types of financial assets is growing extremely in the past 150 years. However, gold and financial assets are often uncorrelated (Baur and Lucey, 2010). Therefore, investors and the media see gold as a safe haven for reducing losses in stock markets in times of an economic crisis.
A safe haven can be described as a place of safety or refuge. A safe haven can also be described as a place that provides shelter for ships during a storm. This indicates that a safe haven asset needs to be a place where assets can retain their value, while ‘sailing in a storm’. A safe haven asset will give an investor the opportunity to secure its capital in turbulent markets or negative market conditions (Baur and McDermott, 2010).
Although there is a lot of evidence that gold can be used as a safe haven or as a hedge, there are yet very few academic studies on the use of other precious metals, like silver, palladium and platinum. This research will investigate the relationship between precious metals (gold, silver, platinum and palladium) and the Dutch stock market. To accomplish this, the following research question is formulated:
Do precious metals perform as a safe haven or hedge for the Dutch stock market?
Precious metals have an important role in financial markets. According to Solt and Swanson (1981), precious metals are generally used as alternative investments for stocks. Hillier, Draper, and Faff (2006) confirm the advantages of the diversification of precious metals for the allocation of portfolios. Moreover, precious metals are considered as an effective hedge for inflation during financial turmoil. The study of Baur and Lucey (2010) confirms that gold is a hedge for the stock market and a safe haven in times of economic turmoil.
The global financial crisis initiated by the bankruptcy of Lehman Brothers in 2008 has had a big impact on the financial world. The bankruptcy of Lehman Brothers has led to great panic in financial markets, which caused that the share price to fall sharply. After the bankruptcy of Lehman Brothers, banks and insurance companies in the United States and Europa faced financial problems, because there was no trust anymore in the financial systems. This has led to an immersive shock on the Dutch stock market. Banks as ABN Amro and Fortis had to be bailed out by the Dutch Government. Other banks and companies such as ING, SNS REAAL and Aegon had to get financial support from the Dutch government to survive. At this time
investors saw gold as a typical safe haven, but do other precious metals also perform as a safe haven or as a hedge for the Dutch Stock market?
2. Literature review
In this section, the relevant concepts that are essential for the paper will be clarified. First, various literature is discussed to understand the different concepts of safe havens and hedges. The trade-off between different interpretations is important to tackle the distinction between the various concepts. Furthermore, the characteristics of precious metals will be explained in this section.
2.1 Safe haven and hedge
There are various definitions of a safe haven. Lucey and Li (2015) define a safe haven as the desired assets in which investors would invest to secure their assets in times of uncertainty. Research by Upper (2000) on the rubble crisis of 1998 in Russia describes a safe haven as investments with high liquidity and low risk. In this paper, he concludes that the 10-year German government bonds could be seen as a safe haven during a period of market distress. Ranaldo and Söderlind (2010) examine the function of different currencies as a safe haven in the financial markets. They define a safe haven as a financial instrument that is uncorrelated or negatively correlated with another financial instrument.
Different branches of investor behaviour literature are concerned with the concept of a safe haven. Forbes and Rigobon (2002) analyse the reaction of investors to financial shocks. In their paper they show that investors are looking for diversification methods in their portfolio to min-imize the risk of huge losses.
Baur and Lucey (2010) describe a safe haven in a more quantitative perspective. They define a safe haven as follows: “A safe haven is defined as an asset that is uncorrelated or negatively
correlated with another asset or portfolio in times of market stress or turmoil”. Depending on
the degrees of the correlation between an asset and the stock market, they divide a safe haven in a weak safe haven and a strong safe haven. If the asset is negatively correlated with another portfolio or asset, they speak of a strong safe haven. Therefore, as the value of the asset in-creases, the value of another asset or portfolio will decrease. They consider a safe haven to be “weak” in case there is no correlation between the portfolio or asset. Thus, as the value of the asset or portfolio decreases, the value of the other asset will stay the same.
Baur and Lucey (2010) define a hedge as “an asset that is uncorrelated or negatively correlated
with another asset or portfolio on average”. The difference between a hedge and a safe haven
After a careful consideration of the afore mentioned definitions, this research will be based upon the definition of Baur and Lucey (2010), because they made a clear difference between a safe haven and hedge.
2.2 Characteristics of gold
The historical meaning of gold is derived, from its function as an exchange in financial systems around the world. Gold is a scarce good, easy to recognise and can be seen as valuable. Saidi and Scacciavillani (2010); Star and Tran (2008) conclude that the purchasing power of gold is stable over time and also stable during financial turmoil. Besides the role of gold as an exchange in financial systems, gold can be contemplated as a safe haven, inflation hedge, currency hedge and portfolio diversifier.
Gold is often mentioned in literature and in media as a safe haven in times of economic turmoil. An example from the last years was the geopolitical tension between North-Korea and the United States in 2017. Trump’s rhetoric on the missile tested by North-Korea led to a weaker dollar and rising gold prices. In this geopolitical tension gold performed as a safe haven (Reuters, 2017). Baur and McDermott (2010) conclude that gold can perform as a safe haven for stock markets of economically developing countries. They show that gold can be seen as a panic investment after negative shocks in the stock market. The paper shows that gold can per-form as a strong safe haven for the stock market of most economically developing countries during the global financial crisis of 2008.
Various studies show that gold can perform as an inflation hedge. This implies that gold will always retain its intrinsic value over time. Ghosh, Levin, Macmillan, and Wright (2004) detect that gold performs a hedge against inflation. They detect that gold performed as an inflation hedge on the long run in the United States, United Kingdom, France, Germany and Japan from 1876 to 1999. Adrangi, Chatrath, and Raffiee (2003) discover a positive correlation between the price of gold and expected inflation. They conclude that gold can be considered as a hedge against expected inflation. Worthington and Pahlavani (2007) discover a stable relationship be-tween the Consumer Price Index (CPI) and gold in the United States from 1973 to 2006. They confirm the effectiveness of gold as a hedge against inflation.
more negative than during the past three decades. The negative relationship was strongest dur-ing the global financial of 2008.
Various studies investigate the diversification function of gold in an investment portfolio. Sharpe (1964) detects with the use of the Capital Asset Pricing Model that a portfolio can reduce its risk by adding securities that have no correlation or a negative correlation with other securi-ties in the portfolio. Johnson and Soenen (1997) and Reboredo (2013) conclude that gold can perform as such a diversifier for a portfolio.
Gold is often used as a reserve asset by Central Banks worldwide. The Deutsche Bundesbank for instance announced in the press four reasons why Central Banks hold gold as a reserve asset: (1) diversification, due to the historical low correlation with other assets (2) overall acceptance of gold, (3) robustness for currency and country risk (4) value depends on trust (Thiele, 2013).
2.3 Characteristics of silver, platinum and palladium
Even though the literature already extensively discussed gold as a safe haven or hedge, the other precious metals (silver, platinum and palladium) were rarely discussed until recently. Batten, Ciner, and Lucey (2010) find evidence that precious metals differs from each other to be referred as a single asset class.
Hood and Malik (2013) investigate whether precious metals (gold, silver, platinum and palla-dium) are suitable as a strong or weak safe haven stock market of the US from 1995 to 2010. They conclude that only gold contrary to the other precious metals could perform as a weak safe haven for the stock market of the US.
Agyei-Ampomah, Gounopoulos, and Mazouz (2013) examine the investment opportunities of precious metals in times of economic turmoil. They conclude that palladium enables investors to obtain a higher compensation for losses in financial markets than gold. Mackenzie and Lucey (2013) investigate whether silver (NYSEARCA: SLV) traded on the New York Stock Exchange was sensitive to the Flash Crash 2010. The Flash Crash in 2010 was a stock market crash in the United States where the S&P500, Dow Jones Industrial Average and Nasdaq Composite col-lapsed. In their research, they conclude that the silver price was correlated with the falling prices of different currencies during the Flash Crash.
Frequently, silver is seen as the little brother of gold. Hillier, Draper, and Faff (2006) show that the price of silver is highly correlated with the price of gold, because a part of the silver supply is obtained as a by-product from the gold mines. However, Escribano and Granger (1998) and Ciner (2001) demonstrate that the gold and silver markets have to be seen as two separate mar-kets since 1990, because the price of gold and silver depends on different economic fundamen-tals.
Gold differs from silver, palladium and platinum, due to the industrial use of gold is much smaller compared to the other three precious metals. Silver is commonly used in solar panels, palladium is commonly used in catalyst converters in the automobile industry and platinum is commonly used in medical laboratory equipment and automobile industry (O'Connor, Lucey, Batten, and Baur, 2015).
Table 1 provides further details on the demand and supply of precious metals. The demand for gold depends largely on its function as commodity and monetary asset. However, the demand of the other three precious metals (silver, platinum and palladium) is largely depending on their industrial use. The statistics of the World Gold Council show that most of the gold production is obtained from gold mines and that less than one third of the silver production is obtained from silver mines, while 12% of the supply of silver is obtained from the by-production of the gold mines. As a result, the supply of silver and gold strongly related to each other. Approximately 36% of the total supply of gold in 2012 comes from recycled gold, mainly from jewellery. More than 20% of the total supply of silver comes from silver scrap. This is because more than 50% of the silver utilize in photography industry is recycled (World Gold Council, 2017).
Compared to silver and gold, the scrap of platinum and palladium is much smaller. The reason behind this is that the supply of platinum and palladium is mostly for industrial use and the long period of the recycling time (7-12 years). In the long-term, the price of platinum and palladium will increase if the industrial activities rise and decrease if the industrial activities fall (World Gold Council, 2017).
Table 1: Worldwide demand and supply of precious metals
The table provides further explanation of the total amount of precious metals (gold, silver, platinum, palladium) weighted in millions of ounces and how the demand and supply of the precious metals is allocated between different sectors.
Gold Silver Platinum Palladium
Total amount (million ounces) 155.50 1048.40 8.05 7.62
Jewellery consumption 43.00% 17.70% 34,56% 4%
Industrial 9.24% 54.24% 55.56% 65%
Investment related 35.66% 19.22% 5.66% 6%
Official sector purchases 12.10%
Coins and metals 8.84%
Others 4.23% 8%
3. Data
This section illustrates which various indices are consulted in this paper. In addition, the anal-ysis consists of the correlation matrix of precious metals’ return (gold, silver, platinum and palladium) and the AEX total return index.
The prices of all four precious metals are retrieved from London Bullion Market. The prices are announced at midnight London time (GMT) at the day the prices are set. The price of gold and silver is set in Euros per troy ounce and the price of platinum and palladium is set in Euros per 0.995 fine ounce on the London Bullion Market. The dataset consists of the log returns of the daily closing spot prices of the precious metals.
The daily closing prices in Euros of the AEX total return index are collected from DataStream. The AEX index, derived from Amsterdam Exchange Index, is the most important Dutch stock index. The index is a composition of the 25 largest Dutch companies traded on the Euronext Amsterdam. Therefore, in this paper the AEX total return index is used as benchmark for the Dutch stock market. The dataset includes the log total return of the daily closing spot prices of the AEX index. Appendix A shows the evaluation of the total log return of the AEX index and the log return of the precious metals.
The sample period range is from 4 January 1999 to 31 December 2018. The motivation for this particular time frame is that the London Bullion Market and the Dutch stock market began to value their assets with the euro currency on 4 January 1999. Since this period the Euro was formed on the stock prices, bond prices and options on the stock exchange.
Each variable consists of 5015 observations, which represents the number of trading days on the stock exchange over the sample period. The problem of missing observations arose because the Dutch stock market and the London Bullion Market have different holidays. Removing the days of the missing observations has solved this problem. Out of the originally 5217 observa-tions per variable, there are 212 observaobserva-tions removed from the dataset. In this paper, Stata is used as software to perform the analyses.
that the data has an extremely leptokurtic character and that the risks are coming from outliers in events. Appendix B shows the return distribution graphs of the precious metals and the AEX index.
Table 2: Descriptive statistics of precious metals and AEX index
The table shows the descriptive statistics of gold, silver, platinum, palladium and AEX index. The results are based on the log returns of the daily closing price in Euro from 4 January 1999 to 31 January 2018.
Obs. Mean % Std. Dev Min Max Skewness Kurtosis
Gold 5015 0.030 0.010 -0.096 0.070 -0.222 10.222
Silver 5015 0.023 0.018 -0.172 0.174 -0.601 14.720
Platinum 5015 0.016 0.014 -0.182 0.088 -0.640 13.465
Palladium 5015 0.027 0.020 -0.163 0.156 -0.312 8.691
AEX index 5015 0.010 0.014 -0.095 0.100 -0.136 9.340
A part of this paper will focus on the two major crises that occurred between 4 January 1999 and 31 January 2018. The first crisis that is included in this paper is the dot-com bubble. In line with the paper of Bauer and McDermott, 2010) and Porras (2016) the dot-com bubble is defined as the period from 10 March 2000 to 9 October 2002. The second crisis that will be investigated in this research is the global financial crisis. 15 September 2008 will be used as start date for the global financial crisis. According to NBER (2010), the economy recovered at 1 July 2009, which is therefore considered as the end of the financial crisis.
Table 3: Correlation Matrix of precious metals and AEX index
The table shows the correlation matrix of gold, silver, platinum, palladium and AEX index. The results are based on the log returns of the daily closing price from 4 January 1999 to 31 January 2018.
Gold Silver Platinum Palladium AEX index
Gold 1.000
Silver 0.513 1.000
Platinum 0.411 0.389 1.000
Palladium 0.323 0.368 0.566 1.000
AEX index -0.048 0.038 0.084 0.152 1.000
In addition to the correlation matrix, this research will use the rolling window sample correla-tion to give a representacorrela-tion of how the correlacorrela-tion of the AEX total index return and the pre-cious metals return change over time. This method computes a sample correlation coefficient established with the rolling window correlation between the return of the AEX index and pre-cious metals’ return. A daily estimation of the correlation between prepre-cious metals and the AEX index returns is computed for the previous 100 trading. Every day the rolling window correla-tion will shift one day forward over the previous 100 trading days. The 100-day rolling window correlation is computed in equation (1) by dividing the equally weighted covariance of AEX index and precious metals’ return over the previous 100 trading days by the square root of product of the variance of the stock index and precious metals over the previous 100 trading days 𝒑𝒕 ̂ = ∑𝟏𝟎𝟎𝒊=𝟏𝒓𝑺 ,𝒕−𝒊,𝒓𝑷 ,𝒕−𝒊 √∑ 𝒓𝟐 𝑺 ,𝒕−𝒊, 𝟏𝟎𝟎 𝒊=𝟏 ∑𝟏𝟎𝟎𝒊=𝟏𝒓𝟐𝑷 ,𝒕−𝒊 (1)
where, 𝑟𝑆,𝑡, is denoted as the stock index return and 𝑟𝑃,𝑡 is denoted as the precious metals return. Figure 1 represents the 100-day rolling window correlation between the log returns of the AEX index and the four precious metals respectively. The figure shows that the correlation between the precious metals with the Dutch stock market has been unstable overtime.
index and platinum is positive during the dot-com bubble (0.026) and the global financial crisis (0.055). The 100-day rolling window correlation between AEX index and platinum is positive during the whole sample period (0.110). The 100-day rolling window correlation between AEX index and palladium is positive during the dot-com bubble (0.036) and the global financial crisis (0.193). The 100-day rolling window correlation between AEX index and palladium is positive during the whole sample period (0.155). The negative correlation of the global financial crisis preliminary indicates that gold and silver perform as a safe haven for the Dutch stock market.
Figure 1: 100-day rolling window correlation
4. Methodology
This section presents the econometric models which are used for the analysis of the hedge and safe haven properties for precious metals. First, the research is done under two assumptions: 1) The price of precious metals depends on the stock market changes. 2) The relationship between stock markets and precious metals changes over time as it is influenced by extreme market conditions. These assumptions are in line with the paper of Baur and McDermott (2010) and Baur and Lucey (2010). The daily returns of the daily spot prices are calculated with the use of the natural logarithms. The formula of the natural logarithms is as follows:
𝑅𝑗,𝑡 = ln ( 𝑝𝑗,𝑡
𝑝𝑗,𝑡−1) (2)
where 𝑅𝑗,𝑡 is the return of variable j at time t and 𝑝𝑗,𝑡 is price of variable j at time t.
4.1 Econometric models
The regression models that are used in the study are presented by Baur and McDermott (2010), which are as follows:
𝑅𝑎𝑠𝑠𝑒𝑡,𝑡 = 𝑎 + 𝑏𝑡𝑅𝑠𝑡𝑜𝑐𝑘,𝑡 + 𝜀𝑡 (3)
𝑏𝑡 = 𝑐0+ 𝑐1𝐷𝑡(𝑅𝑠𝑡𝑜𝑐𝑘𝑞10) + 𝑐2𝐷𝑡(𝑅𝑠𝑡𝑜𝑐𝑘𝑞2.5) + 𝑐3𝐷𝑡(𝑅𝑠𝑡𝑜𝑐𝑘𝑞1) (4a)
ℎ𝑡 = 𝜔 + 𝛼𝜀𝑡−12 + 𝛽ℎ𝑡−1 (5)
Equation (3) models the relation between the return of precious metals (gold, silver, platinum and palladium) and the return of the AEX index. The parameters that have to be estimated are 𝑎 and 𝑏𝑡 and 𝜀𝑡is the error term. The other parameter 𝑏𝑡 is given by equation (4a) and presents the dynamic process. The estimated parameters are 𝑐0, 𝑐1 𝑐2 and 𝑐3. The dummy variable, in-dicated as D(…), captures exceptional stock market shocks and equals one if the daily stock market return passes the worst 10%, 2.5% or 1% of the return distribution.
If one of the parameters 𝑐1 𝑐2 and 𝑐3 is significantly differ from zero, there is proof of a
non-linear relationship between the AEX index return and the specific precious metal return. If the parameters in equation (4a) are negative and significant different from zero, the particular precious metal performs as a strong safe haven asset. In case the parameters are non-positive, the particular precious metal performs as a weak safe haven.
If the parameter 𝑐0 is zero and the sum of parameters 𝑐1, 𝑐2 and 𝑐3 is not at jointly positive
exceeding the value of 𝑐0, the particular precious metal performs as a weak hedge. If the pa-rameter 𝑐0 is negative and the sum of parameters 𝑐1, 𝑐2 and 𝑐3 is not at jointly positive exceeding
Equation (5) presents a GARCH (1, 1) model that tests for heteroscedasticity in the time series data. The parameter 𝛼 measures how much volatility of the error term of today would be carried over to the next period’s volatility. According to Alexander (2008) if the parameter 𝛼 is above the 0.1, then the volatility is very sensitive to market events. The parameter 𝛽 measures the magnitude of daily volatility carryover from the previous day into the current day.
All the models are jointly estimated by the method of Maximum Likelihood. Appendix C shows the development of the daily conditional variance of the precious metals returns and the AEX index return estimated with the GARCH (1, 1) model.
Equation (4b) is a model that is used as an alternative to analyse whether precious metals per-form as a safe haven in times of crisis. The model incorporates a time dummy variable that is equal to one if the returns overlap with the predetermined period of the crisis. The model is a more economic approach, because it is required to determine the periods of the crises before-hand. The following model is jointly estimated with equation (3) and (5).
𝑏𝑡 = 𝑐0+ 𝑐1𝐷𝑡(𝑑𝑜𝑡𝑐𝑜𝑚 𝑏𝑢𝑏𝑏𝑙𝑒 2000) + 𝑐2𝐷𝑡(𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑐𝑟𝑖𝑠𝑖𝑠 2008) (4b) If the parameter 𝑐0 is zero, the particular precious metal performs as a weak hedge and if the parameter 𝑐0 is negative the particular precious metal performs as a strong hedge. If the
param-eters 𝑐1 and 𝑐2 are negative, the precious metal performs as a strong safe haven in the respective period of crisis for the Dutch stock market. If the parameters 𝑐1 or 𝑐2 are zero, the precious
metal performs as a weak safe haven in the respective period of crisis for the Dutch stock mar-ket. If the parameter is positive during the respective period of crisis, the precious metal co-moves with the Dutch stock market.
In the data section the crisis periods are defined, but as most effects of a crisis arise in the first month, this paper uses 20 consecutive trading days for the crisis period. 13 March 2000 is de-fined as the start date of the dot-com bubble and 15 September 2008 will be used as start date for the global financial crisis. In addition to the 20 consecutive trading days, the research also analyses the 40 consecutive trading days to check for robustness of the results. The dummy variables in equation (4b) will be equal to one from the start date of the crisis until the 20th or the 40th consecutive trading days and are equal to zero at all the other trading days.
5. Results
This section shows the results of the two regression models to give an answer to the research question: “Do precious metals perform as a safe haven or hedge for the Dutch stock market?” In the first section the results of equation (3), (4a) and (5) will be demonstrated to answer whether the four precious metals perform as a safe haven or hedge for extreme market condi-tions. In the second section, the results of equation (3), (4b) and (5) will be demonstrated to answer whether the four precious metals perform as safe haven or hedge for the two predefined crises.
5.1 Safe haven and hedge effect for extreme market conditions
Table 5 shows the estimation results of equations (3), (4a) and (5). The hedge column contains the parameter 𝑐0, the 10% quantile column contains the sum of the parameters 𝑐0 and 𝑐1, the
2.5% quantile column contains the sum of parameters 𝑐0, 𝑐1 and 𝑐2 and the 1% quantile column contains the sum of all four parameters (𝑐0, 𝑐1, 𝑐2 and 𝑐3).
The safe haven quantile columns in table 5 indicate that gold (-0.061) and silver (-0.014) perform as a strong safe haven for the Dutch stock market at the 2.5% quantile at the 5% significance level. Palladium (0.083) co-moves under the 1% quantile with the Dutch stock market at the 10% significance level and silver (0.097) co-moves at the 10% quantile with the Dutch stock market at the 10% significance level.
The hedge column of table 5 shows that the precious metals cannot perform as a hedge for the Dutch stock market. There is a high co-movement between platinum and palladium with the Dutch stock market at the 1% significance level. Palladium (0.175) has the largest co-movement with the Dutch stock market. Platinum (0.104) has the second largest co-movement with the Dutch stock market. Silver (0.030) and gold (0.018) as well co-move with the Dutch stock market at the 10% significance level. The co-movements of the precious metals show that the precious metals do not meet the criteria for a hedge.
Table 5: Results for extreme market conditions
This table shows the results of the role of the precious metals (gold, silver, platinum and palladium) as a hedge or safe haven for the Dutch stock market. The results are based on the daily returns of equation (3), (4a) and (5) from the sample period 4 January 1999 to 31 December 2018. Negative (zero) coeffiecient in the hedge column shows that the precious metal performs as a strong (weak) hedge for the Dutch stock market. Negative (zero) coeffiecient in the safe haven quantiles column shows that the precious metal performs as a strong (weak) safe haven for the Dutch stock market in extreme market conditions. The *, ** and *** represents the statistical significance at 10%, 5% level and 1% level.
Gold Silver Platinum Palladium
Hedge 0.018* 0.030* 0.104*** 0.175*** 10% quantile 0.015 0.097* 0.112 0.228 2.5% quantile -0.061** -0.014** 0.068 0.189 1% quantile -0.023 0.007 0.025 0.083* 𝜔 0.000*** 0.000*** 0.000*** 0.000*** 𝛼 0.069*** 0.061*** 0.066*** 0.103*** 𝛽 0.924*** 0.934*** 0.921*** 0.875*** Observations 5015 5015 5015 5015 Model: 𝑅𝑎𝑠𝑠𝑒𝑡,𝑡 = 𝑎 + 𝑏𝑡𝑅𝑠𝑡𝑜𝑐𝑘,𝑡+ 𝜀𝑡 𝑏𝑡 = 𝑐0+ 𝑐1𝐷𝑡(𝑅𝑠𝑡𝑜𝑐𝑘𝑞10) + 𝑐2𝐷𝑡(𝑅𝑠𝑡𝑜𝑐𝑘𝑞2.5) + 𝑐3𝐷𝑡(𝑅𝑠𝑡𝑜𝑐𝑘𝑞1) ℎ𝑡 = 𝜔 + 𝛼𝜀𝑡−12 + 𝛽ℎ𝑡−1
5.2 Safe haven and hedge for crisis periods
Table 6 shows the estimation results of equation (3), (4b) and (5). The table contains the results of the predefined crisis period between 4 January 1999 and 31 December 2018. The predefined crisis periods exist of the 20 consecutive trading days. The hedge column provides the estima-tion of parameter 𝑐0. This indicates whether the precious metals could perform as a hedge for
The safe haven quantile columns of table 6 show that no precious metals perform as a safe haven for the Dutch stock market during the 20 consecutive trading days of the dot-com bubble in 2000. The results of the global financial crisis of 2008 show that gold 0.333) and silver (-0.263) perform as a strong safe haven on a 1% significance level during the 20 consecutive trading days of the global financial crisis of 2008. For palladium and platinum, there was no safe haven effect during the 20 consecutive trading days of the global financial crisis of 2008. The hedge column of table 6 shows that the precious metals cannot perform as a hedge for the Dutch stock market. Out of the four precious metals, there is a co-movement between silver, platinum and palladium with the Dutch stock market on a 1% significance level. Palladium (0.176) has the largest co-movement with the Dutch stock market. Platinum (0.096) has the second largest co-movement with the Dutch stock market. Silver (0.035) as well co-moves with the Dutch stock market. All the coefficients are positive, this implies that precious metals do not meet the criteria for a hedge during the sample period.
Table 6: Results for 20 trading day crisis period
The table shows the results of the role of the precious metals (gold, silver, platinum and palladium) as a hedge or safe haven for the Dutch stock market during the dot-com bubble of 2000 and the global financial crisis of 2008. The two crises are defined as the 20 consecutive trading day period. The results are based on the daily returns of equation (3), (4b) and (5) from the sample period 4 January 1999 to 31 December 2018. Negative (zero) coeffiecient in the hedge column shows that the precious metal performs as a strong (weak) hedge for the Dutch stock market. Negative (zero) coeffiecient in the crisis columns shows that the precious metal performs as a strong (weak) safe haven for the Dutch stock market during the predifined crisis. The *, ** and *** represents the statistical significance at 10%, 5% level, and 1% level.
Gold Silver Platinum Palladium
6. Robustness checks
This section represents various robustness checks to check if the time-period had an impact on the results. First, equation (3), (4a) and (5) will be estimated jointly with two sample splits which contains the time period 04 January 1999 to 1 July 2009 and from 2 July 2009 to 31 December 2018. The reason behind the two sample periods is that in the first sample crisis the dot-com bubble and the global financial crisis are included in the model, while in the second sample period both crises are excluded from the model. Furthermore, for the predefined crisis of equation (4b) the robustness check will take 40 consecutive trading days in account instead of the previous 20 consecutive trading days. Lastly, the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) are used to check the relative quality of the models.
6.1 Subsample periods for extreme market conditions
In this section, the robustness check evaluates if the outcome of the sample period of table 5 is in line with the subsample period of table 7 and 8 for extreme market conditions. Table 7 shows the estimation results of equation (3), (4a) and (5) with the use of the subsample period from 04 January 1999 to 1 July 2009 and table 08 shows the shows the estimation results of equation (3), (4a) and (5) with the use of the subsample period from 2 July 2009 to 31 December 2018. The results of table 7 show that the precious metals cannot perform as a hedge for the Dutch stock market during the subsample period from 04 January 1999 to 1 July 2009. The safe haven quantiles results show that all four precious metals cannot perform as a safe haven for the Dutch stock market. Comparing the results with table 5, it indicates that there was no significant evi-dence found that the precious metals could have a safe haven role in certain quantities during the subsample period from 4 January 1999 to 1 July 2009.
Table 7: Results for extreme market conditions
This table show the results of the role of the precious metals (gold, silver, platinum and palladium) as a hedge or safe haven for the Dutch stock market. The results are based on the daily returns of equation (3), (4a) and (5) from the subsample period 4 January 1999 to 1 July 2009. Negative (zero) coeffiecient in the hedge column shows that the precious metal performs as a strong (weak) hedge for the Dutch stock market. Negative (zero) coeffiecient in the safe haven quantiles column shows that the precious metal performs as a strong (weak) safe haven for the Dutch stock market in extreme market conditions. The *, ** and *** represents the statistical significance at 10%, 5% level, and 1% level.
Gold Silver Platinum Palladium
Hedge 0.026** 0.003 0.084*** 0.119*** 10% quantile 0.016 0.084* 0.062 0.107 2.5% quantile -0.034 -0.018 0.049 0.063 1% quantile -0.008 -0.01 0.065 0.074 𝜔 0.000*** 0.000*** 0.000*** 0.000*** 𝛼 0.085*** 0.064*** 0.115*** 0.172*** 𝛽 0.907*** 0.923*** 0.877*** 0.827*** Observations 2631 2631 2631 2631 Model: 𝑅𝑎𝑠𝑠𝑒𝑡,𝑡 = 𝑎 + 𝑏𝑡𝑅𝑠𝑡𝑜𝑐𝑘,𝑡+ 𝜀𝑡 𝑏𝑡 = 𝑐0+ 𝑐1𝐷𝑡(𝑅𝑠𝑡𝑜𝑐𝑘𝑞10) + 𝑐2𝐷𝑡(𝑅𝑠𝑡𝑜𝑐𝑘𝑞2.5) + 𝑐3𝐷𝑡(𝑅𝑠𝑡𝑜𝑐𝑘𝑞1) ℎ𝑡 = 𝜔 + 𝛼𝜀𝑡−12 + 𝛽ℎ𝑡−1
The results of table 8 show that the precious metals cannot perform as a hedge for Dutch stock market during subsample period from 2 July 2009 to 31 December 2018. The results show that gold is a strong safe haven for the Dutch stock market at the 10% quantile (-0.091) on a 5% significance level and the 1% quantile (-0.152) on a 1% significance level.
The different results of the tables show that the precious metals’ safe haven effect for the Dutch stock market change over time. The hedge effect is the same during the different periods, which means that the hedge effect is not time dependent for the Dutch stock market.
Table 8: Results for extreme market conditions
This table shows the results of the role of the precious metals (gold, silver, platinum and palladium) as a hedge or safe haven for the Dutch stock market. The results are based on the daily returns of equation (3), (4a) and (5) from the subsample period 2 July 2009 to 31 December 2018. Negative (zero) coeffiecient in the hedge column shows that the precious metal can perform as a strong (weak) hedge for the Dutch stock market. Negative (zero) coeffiecient in the safe haven quantiles column shows that the precious metal performs as a strong (weak) safe haven for the Dutch stock market in extreme market conditions. The *, ** and *** represents the statistical significance at 10%, 5% level, and 1% level.
Gold Silver Platinum Palladium
6.2 Extended predefined crisis periods
This section presents the extension of the predefined crisis periods. The predefined crisis peri-ods exist in this part of the research of 40 consecutive trading days instead of the 20 consecutive trading days in the previous part of this research. Table 9 shows the results of the role of the precious metals during the dot-com bubble of 2000 and the global financial crisis of 2008. The hedge results of table 9 show that all four precious metals cannot perform as a hedge for the Dutch stock market. There is a co-movement between silver, platinum and palladium with the Dutch stock market on a 1% significance level. Palladium (0.176) has the largest co-movement with the Dutch stock market. The co-co-movements of the precious metals show that the precious metals do not meet the criteria for a hedge during the sample period.
The safe haven results of table 9 show that only platinum (-0.337) can perform as a strong safe haven for the Dutch stock market during the 40 consecutive trading days of the dot-com bubble in 2000. The results show that gold (-0.253) and silver (-0.235) perform as a strong safe haven on a 1% significance level and 5% significance level during the 40 consecutive trading days of the global financial crisis in 2008.
The results of table 9 show that the parameter 𝛼 of palladium (0.103) has the highest value in the first subsample period from 04 January 1999 to 1 July 2009. This indicates that the volatility of palladium was most sensitive to market events in comparison to the other precious metals. The parameter 𝛽 of silver (0.934) has the highest value in the first subsample period from 04 January 1999 to 1 July 2009. This indicates that the previous volatility of silver has more impact on the present volatility than the other precious metals. In this model the sum of parameter 𝛼 and 𝛽 is smaller than one which means that the volatility of all the precious metals have a stationary process.
Table 9: Results for 40 consecutive trading day crisis period
This table shows the results of the role of the precious metals (gold, silver, platinum and palladium) as a hedge or safe haven for the Dutch stock market during the dot-com bubble of 2000 and the global financial crisis of 2008. The two crisis are defined as the 40 consecutive trading day period. The results are based on the daily returns of equation (3), (4b) and (5) from the sample period 4 January 1999 to 31 December 2018. Negative (zero) coeffiecient in the hedge column shows that the precious metal performs as a strong (weak) hedge for the Dutch stock market. Negative (zero) coeffiecient in the crisis columns shows that the precious metal performs as a strong (weak) safe haven for the Dutch stock market during the predifined crisis. The *, ** and *** represents the statistical significance at 10%, 5% level, and 1% level.
Gold Silver Platinum Palladium
6.3 Model evaluation
To compare the relative quality of the models, this section uses the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) to aim if the parameters from equation (4a) and (4b) will increase the goodness of fit of the models. First, the equation (3) and (5) excluding the dummy variables in the model will be compared with equation (3), (4a) and (5) including the quantile dummy variables in the model. Alongside, the equation (3) and equation (5) excluding the dummy variables in the model will be compared with equation (3), (4b) and (5) including the crisis dummy variables in the model. The statistics of AIC and BIC can only be used to compare different models with each other. The Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) are defined as:
𝐴𝐼𝐶 = −2 ln (𝐿̂) + 2𝑘 (6)
𝐵𝐼𝐶 = −2 ln (𝐿̂) + 𝑘 ln(𝑛) (7)
Where ln (𝐿̂) is denoted as the optimized log-likelihood, 𝑘 is denoted as the number of free parameters and 𝑛 denoted as the number of observations for the model.
Both criterions are measuring the complexity and the goodness of fit of the model, while the only difference between the criterions is that the BIC gives a higher penalty for including extra parameters in the model. The model that produces the lowest AIC or BIC can be seen as the best fitting model.
Table 10: Goodness of fit
The table shows the results of the Akaike information Criterion (AIC) and the Bayesian Information Criterion (BIC) of the three different models from the sample period period 4 January 1999 to 31 December 2018. A lower AIC or BIC between the models indicates a higher quality of the model. The table shows that excluding the dummy variables in the model has a higher quality than including the dummy variables in the model.
7. Conclusion and limitations
This research analyses the role of the four precious metals (gold, silver, platinum and palladium) as a potential hedge or safe haven for the Dutch stock market. Evidence is found that gold and silver perform as a strong safe haven at a 2.5% quantile in extreme market conditions. However, there was no evidence found that platinum and palladium perform as a safe haven in extreme market conditions. Subsequently, no evidence is found that the precious metals perform as a hedge for the Dutch stock market. Looking at the crisis periods, evidence is found that gold and silver perform as a strong safe haven during the global financial crisis in 2008. During the dot-com bubble in 2000, there is no evidence found that the precious metals could perform as a safe haven. Gold and silver could have protected Dutch investors during the global financial crisis of 2008. Therefore, the overall losses of Dutch investors could have been reduced.
References
Adrangi, B., Chatrath, A., Raffiee, K., 2003. Economic activity, inflation, and hedging: the case of gold and silver investments. The Journal of Wealth Management, 6(2), 60-77.
Agyei-Ampomah, S., Gounopoulos, D., Mazouz, K., 2014. Does gold offer a better protection against losses in sovereign debt bonds than other metals?. Journal of Banking & Finance, 40, 507-521.
Alexander, C., 2008. Market Risk Analysis. Value-at-Risk Models, Volume IV, John Wiley &Sons Ltd.
Baur, D., Lucey, B., 2010. Is gold a hedge or safe haven?. An analysis of stocks bonds and gold. The Financial Review, Vol. 45, pp. 217–229.
Baur, D., McDermott, T., 2010. Is gold a safe haven? International evidence. Journal of Banking & Finance, Vol. 34, No. 8, pp. 1886-1898.
Batten, J. A., Ciner, C., Lucey, B. M., 2010. The macroeconomic determinants of volatility in precious metals markets. Resources Policy, 35(2), 65-71.
Capie, F., Mills, T. C., Wood, G., 2005. Gold as a hedge against the dollar. Journal of International Financial Markets, Institutions and Money, 15(4), 343-352.
Chng, M. T., Foster, G. M., 2012. The implied convenience yield of precious metals: safe haven versus industrial usage. Review of Futures Markets, 20(4), 349-394.
Ciner, C., 2001. On the long run relationship between gold and silver prices. Global Finance Journal, 12(2), 299-303.
Escribano, A., Granger, C. W., 1998. Investigating the relationship between gold and silver prices. Journal of Forecasting, 17(2), 81-107.
Forbes, K. J., Rigobon, R., 2002. No contagion, only interdependence: measuring stock market comovements. The Journal of Finance, 57(5), 2223-2261.
Ghosh, D., Levin, E. J., Macmillan, P., Wright, R. E., 2004. Gold as an inflation hedge?. Studies in Economics and Finance, 22(3), 1-25.
Hair, J. F. J. R., Black, W. C., Babin, B. J., Anderson, R. E., 2010. Multivariate Data Analysis (7th ed.). Upper Saddle River, NJ: Prentice Hall.
Hillier, D., Draper, P., Faff, R., 2006. Do precious metals shine? An investment perspective. Financial Analysts Journal, 98-106.
Hood, M., Malik, F., 2013. Is gold the best hedge and a safe haven under changing stock market volatility?. Review of Financial Economics, 22(2), 47-52.
Johnson, R., Soenen, L. A., 1997. Gold as an investment asset: perspectives from different countries. Journal of Investing, 6(5), 94-99.
Joy, M., 2011. Gold and the US dollar: hedge or haven?. Finance Research Letters, 8(5), 120-131.
Lucey, B. M., Li, S., 2015. What precious metals act as safe havens, and when? Some US evidence. Applied Economics Letters, 22(3), 35-45.
Matthey, J., 2018. Summary of platinum supply & demand in 2017. PGM Market Reports February.
National Bureau of Economic Research., 2010. US Business Cycle Expansion and Contractions [Dataset]. Retrieved on 3 June 2018, from http://www.nber.org/ccles/sept2010.pdf
O'Connor, F. A., Lucey, B. M., Batten, J. A., Baur, D. G., 2015. The financial economics of gold—a survey. International Review of Financial Analysis, 41, 186-205.
Porras, E. V. A., 2016. Bubbles and Contagion in Financial Markets, Volume 1. Basingstoke, United Kingdom: Palgrave Macmillan.
Pukthuanthong, K., Roll, R., 2011. Gold and the Dollar (and the Euro, Pound, and Yen). Journal of Banking & Finance, 35(8), 2070-2083.
Ranaldo, A., Söderlind, P., 2010. Safe haven currencies. Review of Finance, 14(5), 385-407.
Reboredo, J. C., 2013. Is gold a safe haven or a hedge for the US dollar? Implications for risk management. Journal of Banking & Finance, 37(8), 2665-2676.
Saidi, N., Scacciavillani, F., 2010. The role of gold in the new financial architecture. Dubai International Financial Centre Economic Note, (13).
Sensoy, A., 2013. Dynamic relationship between precious metals. Resources Policy, 38(4), 504-511.
Sharpe, W. F., 1964. Capital asset prices: A theory of market equilibrium under conditions of risk. The Journal of Finance, 19(5), 425-442.
Starr, M., Tran, K., 2008. Determinants of the physical demand for gold: evidence from panel data. The World Economy, 31(5), 416-436.
Thiele, C. L., 2013, January 16. Pressegespräch, Deutsche Goldreserven. Retrieved from https://www.bundesbank.de/Redaktion/DE/Downloads/Presse/Publikntionen/2013_0_16_thiel e_praesentation_pressegespraech_gold.pdf?__blob=publicationFile
Upper, C., 2000. How safe was the safe haven?: financial market liquidity during the 1998 turbulences”, Deutsche Bundesbank Working Paper No. 1/00.
Vedala, V., 2017, May 17. PRECIOUS-Gold rises to two-week high amid Trump concerns. Retrieved May 20, 2017, from http://www.reuters.com/article/global-preciousidUSL4N1IJ1J0
World Gold Council., 2017. Gold in the financial system. Retrieved May 20, 2017, from http://www.gold.org/reserve-asset-management/gold-financial-system
Appendix A
Figure 5: Log returns of AEX index, gold, silver, platinum and palladium
Appendix B
Figure 6: Return distribution
Appendix C
Figure 7: Daily conditional volatility of gold and silver, platinum and palladium
Appendix D
Table 11: Variance Inflation Factor (VIF) model
The table shows the results of the VIF-test of equation (3), (4a) and (5), to evaluate if there is evidence of multicollinearity in the variables.
Table 12: Variance Inflation Factor (VIF) model
The table shows the results of the VIF-test of equation (3), (4b) and (5), to evaluate if there is evidence of multicollinearity in the variables.
Sample period Subsample period 1 Subsample period 2
VIF 1/VIF VIF 1/VIF VIF 1/VIF
Hedge 1.88 0.53 1.88 0.53 1.83 0.54
10% quantile 3.58 0.27 3.54 0.28 3.06 0.33
2.5% quantile 4.38 0.23 4.18 0.24 3.49 0.29
1% quantile 2.81 0.36 2.62 0.38 2.33 0.43
Mean VIF 3.16 3.06 2.68
20 consecutive trading days 40 consecutive trading days
VIF 1/VIF VIF 1/VIF
Hedge 1.06 094 1.12 0.89
Financial crisis 2008 1.05 0.95 1.11 0.90
Dot-com bubble 2000 1.01 0.99 1.01 0.99
