The roles of gold, US dollar, and bitcoin as safe-haven assets in times of crisis

Abstract

Using the GJR-GARCH method, this study examines the safe-haven role of gold, US dollar, and Bitcoin over a period including the global financial crisis, the COVID-19 pandemic and the Russia-Ukraine conflict from 3 April 2006 to 19 May 2023. The study supports the hypothesis that the safe-haven role of assets changes over periods of crisis. Specifically, gold loses its role as a safe-haven asset during the COVID-19 pandemic, but this role has been restored in the Dutch, US and German markets during the Russia-Ukraine conflict. Similarly, Bitcoin is not a safe-haven asset during the COVID-19 pandemic but is a strong safe-haven asset for the stock markets of some European countries, and a weak safe-haven asset for China when the Russia-Ukraine conflict occurred. Only the USD acts as a stable safe-haven asset through periods of crisis. However, this role is weakened in Russia. These results partly help investors and portfolio managers choose a safe haven for their assets, especially during volatile market periods.

Impact Statement

Investors often want to protect their assets by diversifying their investment portfolios or using risk prevention tools. Especially when the financial market encounters instability caused by the crisis, investors tend to look for safe-haven assets such as precious metals, foreign currencies, and cryptocurrencies. This study re-evaluates the safe-haven role of gold, USD, and Bitcoin during the Russia-Ukraine conflict, then compares the research results with two previous crises, GFC 2008 and the COVID-19 pandemic. Research results help investors choose appropriate assets during market volatility.

KEYWORDs:

• Safe haven assets
• GFC 2008
• Covid-19
• Russia–Ukraine conflict

Reviewing Editor:

• David McMillan, University of Stirling, UK

Subjects:

• Economics
• International Economics
• International Finance
• Finance

JEL Classification:

• C58
• G01
• G11
• G12
• G14

1. Introduction

Investors often have a need to protect their assets by diversifying their portfolios or using hedging tools. Especially when financial markets experience uncertainties caused by crises, investors tend to look for safe-haven assets (Aloui et al., 2018; Baur & Lucey, 2010; Baur & McDermott, 2010; Feng et al., 2018; Hkiri et al., 2017). The concept of a safe-haven asset was first defined by Baur and Lucey (2010), referring to an asset that is uncorrelated or negatively correlated with another asset or portfolio in times of market stress or turmoil. Baur and Lucey (2010) also shows that gold is a safe-haven asset in the short term (about 15 trading days). When the global financial crisis (GFC, 2008) occurred and impacted global financial markets, many studies were carried out to test whether gold is still able to protect investors’ assets (Bredin et al., 2015; Cheema et al., 2022; Choudhry et al., 2015; Hasan et al., 2021; Low et al., 2016). Findings are still inconsistent. While the results of the studies are controversial, the global financial market has faced a big shock – the COVID-19 pandemic – first detected in Wuhan, China at the end of 2019 and quickly spreading globally. Research on safe-haven assets continues to be conducted to examine the role of traditional assets as well as some new assets, such as cryptocurrencies, in times of crisis (Aysan et al., 2019; Hasan et al., 2021; Shahzad et al., 2020; Umar et al., 2021). For example, Ji et al. (2020), Salisu et al. (2021), Hasan et al. (2021) found evidence that gold, and Bitcoin still perform as safe-haven assets during the COVID-19 pandemic. In contrast, Cheema et al. (2022), Akhtaruzzaman et al. (2021), Disli et al. (2021), Conlon et al. (2020) showed a decline or loss of the safe-haven role of these assets.

In 2022, as the COVID-19 pandemic continued to cause impacts on the global economy, society and finances, a controversial Russia-Ukraine conflict began on 24 February 2022, marking a new extreme geopolitics scenario that could affect the global economy and trade. Continuous shocks have impacted global stock markets unfavorably (Boubaker et al., 2022) and increased uncertainty in the financial markets (Wiseman & Mchugh, 2022). Because it is occurring at a crucial juncture for the world economy, the Russia-Ukraine war is seen as the most serious event in Europe since World War II. For the time being, it has dimmed hopes for a global economic recovery following COVID-19’s destruction (Beraich et al., 2022). The crisis has accelerated worldwide inflation in just a few days by raising the cost of commodities like oil, natural gas, and other essentials (Cohen, 2022). Research has also demonstrated that as the Russia-Ukraine conflict worsens, the volatility of the agricultural, metal, and energy markets increases significantly. Meanwhile, investor jitters and the monetary policies of the major central banks exacerbated the conflict’s effects, adversely affecting financial markets, particularly the stock, gold, foreign exchange, and Bitcoin markets (Fang & Shao, 2022). Yatie (2022) argues that Bitcoin, Ethereum, and Gold have failed to act as safe-haven assets, while Oosterlinck et al. (2022) reassert that gold and Bitcoin are complements for diversification purposes in international crises.

Given the controversial results presented above and the consequences of the Russia-Ukraine conflict on the financial markets of the countries, this is the right time for our research to re-examine the role of traditional safe-haven assets. According to Liadze et al. (2023), the conflict caused great damage to the global economy, and countries with close trade links with Russia and Ukraine were most affected due to their dependence on energy and food supplies from these two countries. Sezgin (2022) suggested that the conflict negatively affected Turkey’s agricultural exports to Russia and Ukraine. Rising global food and energy prices have worsened Turkey’s budget deficit and rising domestic inflation, which could affect Turkish traders in the medium term. Meanwhile, Alam et al. (2022) found evidence that the war between the two countries impacts the US, EU, and China commodity markets. Wang and Su (2023) asserted that the Russia-Ukraine conflict affects the stability of the Chinese financial and commodity system mainly due to its impact on the Chinese commodity market, mainly due to the increasing risks increased in energy, chemicals, and grain markets. On the other hand, commodity markets and stock markets are pretty highly correlated (Öztek & Öcal, 2017). Therefore, this study aims to examine how the war affected these countries’ safe-haven assets. We conducted this research in eight countries including the US, the UK, the Netherlands, Germany, Russia, China, Poland and Turkey with the aim of adding more empirical evidence on the role of safe-haven assets during different crises. Then we extend the study period to two previous major crises, COVID-19 and GFC 2008, to highlight that the role of safe-haven assets in the studied countries has changed not only due to the impact of the Russia-Ukraine conflict but also to other crises.

This article has the following contributions: First, the study has compared the safe-haven role of gold, Bitcoin, and USD in a period including the global financial crisis 2008, the COVID-19 pandemic, and the Russia-Ukraine conflict. Second, the study has again demonstrated the changing role of assets across different crises. Specifically, gold has lost its role as a haven asset during the COVID-19 pandemic, but this role was restored in the Netherlands, US and Germany markets during the Russia-Ukraine conflict. Bitcoin does not perform as a safe haven during COVID-19. But when the Russia-Ukraine conflict occurred, Bitcoin was a strong safe-haven asset for the stock markets of European countries such as the Netherlands, UK, Germany, and Russia, and a weak safe-haven asset for China. The US dollar has shown a stable role as a safe-haven asset through periods of crisis. However, this role is waning in Russia due to the country’s special policies to limit influence and respond to sanctions imposed by the US and the West.

The rest of the study is organized as follows: Section 2 describes literature reviews. Section 3 summarizes the data and the research methodology. Section 4 explains the results and discussion. Finally, the conclusion is presented in Section 5.

2. Literature review

Gold has long been chosen as a preferred channel for investors because of its many advantages such as being unaffected by inflation (Husnul et al., 2017). Research has proven that gold is a hedging instrument and a safe-haven asset, especially during periods of market volatility. Baur and Lucey (2010) analyzed the relationship over time between stock, bond and gold returns in the US, UK and Germany. Applying the Capie et al. (2005) approach and the asymmetric GARCH model, the study showed that in the short term, gold is both a hedge and a safe- haven for stocks. Based on Baur and Lucey (2010), Baur and McDermott (2010) extended the scope of the study to the global financial markets during period 1979–2009. The results revealed that gold is both a hedge and a safe-haven asset for the major European and US stock markets but not for Australia, Canada, Japan and major emerging markets like the BRIC countries. Especially in times of crisis, gold is considered a strong safe-haven asset for most developed markets. The safe-haven property of gold continues to be confirmed in Wen and Cheng (2018), Ji et al. (2020), Long et al. (2021), Salisu et al. (2021), Hasan et al. (2021). Research by Kinateder et al. (2021) showed that gold remains a safe option for investors during the GFC and COVID-19 pandemic. Gold has the ability to hedge against many uncertain factors, especially during COVID-19 (Hasan et al., 2023). Akhtaruzzaman et al. (2024) assert that the effectiveness of portfolio risk management of gold has increased during the COVID-19 pandemic. However, several other studies on this topic have slightly different, or even opposite results. Hood and Malik (2013) provided evidence of a weak safe-haven property of gold in US equities from November 1995 to November 2010. This result was found based on the regression model of Baur and McDermott (2010) and the GARCH method. Similar results were confirmed in the study by Choudhury et al. (2022) when studying the role of gold as a safe haven asset during pandemics caused by SARS, Ebola, Zika, Swine Flu, and COVID-19. Applying the DCC-GARCH model, the study showed that gold is a weak safe haven asset for stock investors during epidemics. Corbet et al. (2020) used hourly data for the period from 11 March 2019 to 10 March 2020 to study the contagion impact of the COVID-19 pandemic on the Chinese stock market. The results showed that gold is neither a hedge nor a safe-haven asset for stocks during the pandemic. Meanwhile, using the DCC-GARCH model, Akhtaruzzaman et al. (2021) found that the safe-haven role of gold changes over different periods. Specifically, gold acts as a safe-haven asset during Phase I (31 December 2019 to 16 March 2020) of the COVID-19 pandemic but loses its safe-haven property during Phase II (17 March to 24 April 2020). Cheema et al. (2022) compared the role of safe-haven assets during the Global Financial Crisis and the COVID-19 pandemic. The GJR-GARCH estimation results showed that the safe-haven properties of gold and silver have been weakened during COVID-19 compared with the GFC. Recently, research by Yatie (2022) in the context of the impact of the Russia-Ukraine conflict also had similar results that gold has failed in the role of a safe-haven asset.

Besides gold, currencies are also considered safe-haven assets. For example, the Swiss Franc and Japanese Yen are considered safe-haven properties in times of crisis according to Ranaldo and Söderlind (2010), and the US dollar is a safe-haven asset for emerging stocks, according to Wen and Cheng (2018). Fidalgo (2021) applied a threshold regression model to examine the property of the US dollar as a safe-haven asset for stocks of developed countries during the COVID-19 pandemic. Findings indicated that the US dollar acts as a strong safe-haven asset. The study also confirmed that the safe-haven role of a given asset can be sensitive to changes over time and across different markets. Also studying around the time of the pandemic, Ji et al. (2020) suggested that most currencies do not perform a safe-haven role. Cheema et al. (2022) also found evidence that the safe-haven character of the US dollar was weakened during the COVID-19 pandemic compared with the GFC.

Recently, cryptocurrencies have emerged as a new financial asset. Many studies are conducted to explore whether cryptocurrencies can act as safe-haven assets or hedges for investors. Bouri et al. (2017) examined the property of Bitcoin as a safe-haven asset and a hedge between July 2011 and December 2015 using a dynamic conditional correlation model. The findings suggested that Bitcoin acts as a strong safe-haven asset for Asian stocks. Besides, the study added that bitcoin’s hedging and safe-haven properties are different across markets. Stensås et al. (2019) used the DCC-GARCH model to examine if Bitcoin serves as a diversifier, hedge or safe-haven instrument for commodities as well as for investors in important developed and developing markets. The results showed that Bitcoin only serves as a tool for diversification for investors in developed countries and for commodities, but it acts as a hedge for investors in the majority of developing countries, including Brazil, Russia, India, and South Korea. In particular, Bitcoin served as a safe-haven asset for both US and non-US investors throughout the 2016 US election, the Brexit vote in 2016, and the Chinese market bubble fall in 2015. Similar results were found in Kang et al. (2020). Hasan et al. (2021) explored evidence of the safe-haven role of Tether during the COVID-19 pandemic and asserted that Bitcoin still exhibits safe-haven properties during severe market downturns. Rubbaniy et al. (2021) suggested that long-term investors can hedge against the COVID-19 pandemic by investing in cryptocurrencies to diversify their portfolios. Research by Oosterlinck et al. (2022) in the context of the Russia-Ukraine war also showed similar results, that Bitcoin can help investors diversify risks during international crises. Smales (2019) argued that Bitcoin is more volatile and less liquid than other assets, and transaction costs are high, especially during volatile market periods. Therefore, Bitcoin should not be considered a safe-haven asset. Conlon et al. (2020) investigated the risk-mitigation properties of Bitcoin, Ether, and Tether during the initial bear market period associated with the COVID-19 crisis. This research provided evidence that Bitcoin and Ethereum are not safe-haven assets for the majority of the international stock markets examined. Besides, Bitcoin and Ethereum also add to the downside risk of portfolios. All of the international indices under consideration used Tether as a safe haven investment since it was able to successfully retain its peg to the US dollar throughout the COVID-19 crisis. Research by Jeribi and Manzli (2020) in the Tunisian stock market showed that Bitcoin acts as a hedge, while Ethereum is a diversifier in the pre-pandemic period, but these two cryptocurrencies do not help to protect investors’ assets during the pandemic. These results were similar to Disli et al. (2021), that Bitcoin does not exhibit safe-haven characteristics during the COVID-19 pandemic. Yatie (2022) studied the impact of fear, uncertainty, and market volatility caused by the Ukraine-Russia war on returns of crypto assets (Bitcoin and Ethereum). The study used Wikipedia trends searches as proxies for uncertainty and fear, along with two volatility indices: the S&P500 VIX and Russia’s VIX (RVIX). Applying the DCC-GARCH model, the results showed that Bitcoin and Ethereum have failed to play the safe-haven role in this war.

Table 1 summarizes studies on the role of safe-haven assets of gold, Bitcoin, and US dollar.

Table 1. Review of selected studies.

Authors	Method	Period	Countries	Main results
Baur and Lucey (2010)	Asymmetric GARCH	30 November 1995–30 November 2005	US, UK and Germany	In short term, gold serves as a hedge and a safe haven for stocks.
Baur and McDermott (2010)	GARCH	1979–2009	G7, BRIC, Australia, Switzerland	For the major stock markets in Europe and the US, gold serves as a hedge and a safe haven, but not for Australia, Canada, Japan, or the BRIC nations. During periods of crisis, gold is a strong safe-haven for most developed markets.
Ranaldo and Söderlind (2010)	OLS, GMM	1993–2008	US	The Swiss franc and the Japanese yen are safe-haven assets in times of crisis.
Hood and Malik (2013)	GARCH	November 1995–November 2010	US	Gold acts as a hedge and a weak safe-haven asset for the US stock market.
Bouri et al. (2017)	DCC-GARCH	July 2011–December 2015	US, UK, Germany, Japan, China	Bitcoin acts as a strong safe-haven asset for Asian stocks.
Wen and Cheng (2018)	Generalized Autoregressive Score (GAS) Copula	3 January 2000–6 July 2016	Brazil, Chile, the Czech Republic, Russia, South Africa, China, India, Malaysia, and Thailand	Both gold and the US dollar can act as safe havens for emerging stocks.
Stensås et al. (2019)	GARCH Dynamic Conditional Correlation (DCC)	13 September 2011–1 January 2018	G7, BRIC, South Korea and Zimbabwe	Bitcoin acts as a hedge for investors in Brazil, Russia, India and South Korea, but only as a diversification tool for investors in developed countries and for commodities.
Conlon et al. (2020)	Two-moment VaR estimation	April 2010–April 2020	US, UK, Italy, Spain, China	Bitcoin and Ethereum are not safe-haven assets.
Corbet et al. (2020)	A standard GARCH (1,1)	11 March 2019–10 March 2020	China	Gold and Bitcoin are neither hedge nor safe-haven assets for stocks during the COVID-19 pandemic.
Ji et al. (2020)	A cross -quantilograms analysis	1 December 2019–31 March 2020	China, Europe, US	Gold is a safe-haven asset during the COVID-19 pandemic. Most currencies do not act as a safe-haven.
Kang et al. (2020)	DECO FIGARCH	18 July 2010–30 October 2016	US	Bitcoin can be used as an effective safe haven for investors.
Akhtaruzzaman et al. (2021)	DCC-GARCH	2 January 2018–24 April 2020	US, Europe, Japan, China	The safe-haven property of gold has change during the COVID-19.
Fidalgo (2021)	Threshold regression analysis	February 2016–February 2021	Global	USD acts as a strong safe-haven asset during the COVID-19 pandemic.
Hasan et al. (2021)	GJR-GARCH	2 January 2002–18 August 2020	US	In GFC 2008, silver and Islamic stocks were safe haven assets. During COVID-19, Islamic stocks and Tether were safe haven assets. Gold and Bitcoin remain safe-haven properties during severe market downturns.
Jeribi and Manzli (2020)	OLS	December 31, 2019–April 30, 2020	Tunisia	Bitcoin acts as a hedge, while Ethereum is a diversifier in the pre-pandemic period, but these two cryptocurrencies do not help to protect investors’ assets during the pandemic.
Long et al. (2021)	NARDL	January 2013–December 2020	US	When facing various shocks of uncertainty, Bitcoin cannot act as a safe-haven, while gold is still recognized as a traditional safe-haven asset.
Rubbaniy et al. (2021)	Wavelet Coherence	25 January 2020 to 11	Global	Cryptocurrencies help long-term investors to hedge risks and diversify portfolios during the COVID-19 pandemic.
Salisu et al. (2021)	The Root Mean Square Error (RMSE), Mean Absolute Deviation (MAD) and Mean Absolute Percentage Error (MAPE)	December 31, 2019–May 28, 2020	US	Gold is a better safe-haven asset than US stocks and other metals such as Silver, Palladium and Platinum.
Cheema et al. (2022)	GJR-GARCH	1 January1990–31 December 2021	US, China, Japan, Germany, UK, France, India, Italy, Brazil, Canada	The safe-haven character of gold, silver, USD, and T-bonds weakened during COVID-19 against the GFC.
Oosterlinck et al. (2022)	DCC-GARCH	March 2019–May 2022	US, EU	Gold and Bitcoin are complements for diversification purposes in international crises.
Yatie (2022)	DCC-GARCH	1 November 2021–15 March 2022	US, Russia	Bitcoin, Ethereum, and gold fail to play a safe-haven role in the Russia-Ukraine war.
Choudhury et al. (2022) when studying the role of gold as a safe haven asset during pandemics caused by SARS, Ebola, Zika, Swine Flu, and COVID-19. Applying the model, the study showed that	DCC-GARCH	1 January 2001–9 February 2022	US, Japan, Germany, and Switzerland	Gold is a weak safe haven asset for stock investors during epidemics.

In general, the safe-haven properties of gold, currencies, and cryptocurrencies change over time, especially during periods of crisis such as the global financial crisis and the COVID-19 pandemic (Cheema et al., 2022; Hasan et al., 2021). Therefore, when the Russia-Ukraine conflict occurs, it is necessary to re-evaluate the role of these assets in order to help investors seek out safe havens in order to limit their exposure to losses in the event of market downturns. Furthermore, most of the previous studies have focused only on comparing the hedging or safe-haven properties of these assets before and during the crisis. With the exception of Hasan et al. (2021) and Cheema et al. (2022), the number of studies comparing the role of assets in different crises is relatively small. In addition, to the best of our knowledge, no studies have compared the safe-haven properties of gold or other assets between previous crises and the current crisis – the Russia – Ukraine conflict. Our study aims to try to fill in the theoretical gaps by applying the GJR-GARCH model introduced by Glosten et al. (1993) to assess the safe-haven role of gold, Bitcoin and US dollar across three different crises including the global financial crisis 2008, the COVID-19 pandemic and the Russia-Ukraine conflict. In particular, the paper focuses on analyzing countries that are major trading partners of Russia and Ukraine – which are heavily affected by the disruption of the supply chain of goods caused by the Russia-Ukraine conflict.

3. Data and methodology

3.1. Data and sample

As the authors explained in section 1, this article examines the safe-haven property of three traditional, potential and popular assets including gold, US dollar and Bitcoin for the securities of largest trading partner of Russia and Ukraine in three periods of market downturns: the GFC 2008, the COVID-19 pandemic, and the Russia-Ukraine conflict. According to World Integrated Trade Solution (WITS), Russia’s five largest trading partners in 2020 are China, the Netherlands, the UK, the US and Germany; those of Ukraine are China, Poland, Russia, Germany and Turkey. Therefore, the stock indexes used in this study include AEX Index Netherlands, BIST100 Index Turkey, CSI300 Index China, DAX Index Germany, FTSE100 Index UK, MOEX Index Russia, SP500 Index US, and WIG20 Index Poland. All data are expressed as returns for safe-haven assets and stocks to ensure a fair comparison. Because of the availability of data, the research period would be from 3 April 2006 to 19 May 2023 (Bitcoin data were collected from 2 February 2012 to 19 May 2023). Daily data are collected from investing.com website, and missing data are added from countryeconomy.com website. To describe the period of the GFC 2008, we choose the start date of the GFC as 9 October 2007 (the date when the S&P 500 index peaked at a closing price of 2447.03) and the end date of 30 June 2009 (the date on which the National Bureau of Economic Research announced the end of the US recession), according to Cheema et al. (2022). To describe the COVID-19 pandemic phase, we choose a starting date of 11 March 2020 (the date when WHO declared COVID-19 a global pandemic) and an end date of 19 May 2023 (the end date of our sample). Regarding the Russia-Ukraine conflict, the starting date is 24 February 2022 (the day Russia launched a large-scale military offensive against Ukraine) and the end date is 19 May 2023.

3.2. Methodology

For evaluating the safe haven role of many asset classes during periods of volatility, the GARCH model has been frequently used (Akhtaruzzaman et al., 2021; Baur and Lucey, 2010; Baur and McDermott, 2010; Hood and Malik, 2013). The GARCH model, however, could not be adequate if fat-tail risk affects asset returns. The GJR-GARCH model is more appropriate in these circumstances than the GARCH model because it better takes into consideration the fat-tail distribution. Therefore, the GJR-GARCH models are used in many other studies (Cheema et al., 2022; Hasan et al., 2021). In this article, we mainly utilize the models described in Baur and Lucey (2010) and Baur and McDermott (2010) to test for a safe haven. The model estimates the role of safe-haven assets of gold, Bitcoin and US dollar for stock as follows: (1) $${\mathit{\text{RA}}}_{i,t}=\alpha +b_1{\times \mathit{\text{RS}}}_{j,t}{+b}_2\times {\mathit{\text{RS}}}_{j,t}\times \mathit{\text{GFC}}+b_3{\times \mathit{\text{RS}}}_{j,t}\times \mathit{\text{COVID}}+b_4{\times \mathit{\text{RS}}}_{j,t}\times \mathit{\text{WAR}}+{\epsilon }_{i,t}$$(1) (2) $$\begin{array}{c}{\sigma }_t^2=\omega +(\alpha +\gamma I_{t-1}){\epsilon }_{t-1}^2+\beta {\sigma }_{t-1}^2\hfill \\ \mathit{\text{with}} I_{t-1}=\{\begin{array}{c}1 \mathit{\text{if}} {\epsilon }_{t-1}<0\\ 0 \mathit{\text{if}} {\epsilon }_{t-1}\ge 0\end{array}\hfill \end{array}$$(2) where RA is the return of safe-haven asset i, including gold (GOLD), Bitcoin (BTC), and US dollar (USD). RS is the return of stock index j, which includes AEX, BIST100, CSI300, DAX, FTSE100, MOEX, SP500, and WIG20. $\alpha \mathrm{ }$ and $b_{\mathrm{ }}$ are the estimation parameters, ${\epsilon }_t$ is the error term. GFC is a dummy variable, taking the value 1 from the start date to the end date of the 2008 global financial crisis (explained above), and 0 otherwise; COVID is a dummy variable, taking the value 1 from the start of the COVID-19 pandemic to the end of the study period, and 0 otherwise; WAR is a dummy variable, taking the value of 1 from the date Russia launched a military attack on Ukraine to the end of the study period, and 0 otherwise. In Eq. (2)(2) $$\begin{array}{c}{\sigma }_t^2=\omega +(\alpha +\gamma I_{t-1}){\epsilon }_{t-1}^2+\beta {\sigma }_{t-1}^2\hfill \\ \mathit{\text{with}} I_{t-1}=\{\begin{array}{c}1 \mathit{\text{if}} {\epsilon }_{t-1}<0\\ 0 \mathit{\text{if}} {\epsilon }_{t-1}\ge 0\end{array}\hfill \end{array}$$(2) , the residual ${\epsilon }_t$ is modeled through the GJR-GARCH process, taking into account asymmetric effects where stock market returns exhibit high volatility in response to bad news as opposed to good news (introduced by Glosten et al. (1993)). ${\sigma }_t^2$ and ${\sigma }_{t-1}^2$ are variances at time $t$ and $t-1,$ respectively. Both Eqs. (1)(1) $${\mathit{\text{RA}}}_{i,t}=\alpha +b_1{\times \mathit{\text{RS}}}_{j,t}{+b}_2\times {\mathit{\text{RS}}}_{j,t}\times \mathit{\text{GFC}}+b_3{\times \mathit{\text{RS}}}_{j,t}\times \mathit{\text{COVID}}+b_4{\times \mathit{\text{RS}}}_{j,t}\times \mathit{\text{WAR}}+{\epsilon }_{i,t}$$(1) and (2)(2) $$\begin{array}{c}{\sigma }_t^2=\omega +(\alpha +\gamma I_{t-1}){\epsilon }_{t-1}^2+\beta {\sigma }_{t-1}^2\hfill \\ \mathit{\text{with}} I_{t-1}=\{\begin{array}{c}1 \mathit{\text{if}} {\epsilon }_{t-1}<0\\ 0 \mathit{\text{if}} {\epsilon }_{t-1}\ge 0\end{array}\hfill \end{array}$$(2) are estimated with maximum-likelihood method.

We apply Stata 16 software to process the collected data. To evaluate the role of selected haven assets, the parameters estimated in Eqs. (1)(1) $${\mathit{\text{RA}}}_{i,t}=\alpha +b_1{\times \mathit{\text{RS}}}_{j,t}{+b}_2\times {\mathit{\text{RS}}}_{j,t}\times \mathit{\text{GFC}}+b_3{\times \mathit{\text{RS}}}_{j,t}\times \mathit{\text{COVID}}+b_4{\times \mathit{\text{RS}}}_{j,t}\times \mathit{\text{WAR}}+{\epsilon }_{i,t}$$(1) and (2)(2) $$\begin{array}{c}{\sigma }_t^2=\omega +(\alpha +\gamma I_{t-1}){\epsilon }_{t-1}^2+\beta {\sigma }_{t-1}^2\hfill \\ \mathit{\text{with}} I_{t-1}=\{\begin{array}{c}1 \mathit{\text{if}} {\epsilon }_{t-1}<0\\ 0 \mathit{\text{if}} {\epsilon }_{t-1}\ge 0\end{array}\hfill \end{array}$$(2) are explained as follows. If all the parameters $b_1,$ $b_2,$ $b_3,b_4$ are positive, the analyzed asset is neither a hedge nor a safe-haven asset. If $b_1$ is less than or equal to 0 and is statistically significant (insignificant), the analyzed asset is a strong (weak) hedge for the stock, because the assets are not correlated with each other (on average). If $b_2$ is less than or equal to 0 and is statistically significant (insignificant), the studied asset is a strong (weak) safe-haven asset for stock during the GFC 2008. The interpretation is the same for $b_3$ for the COVID-19 pandemic period and $b_4$ for the Russia-Ukraine war period.

4. Results and discussion

4.1. Descriptive statistics and preliminary tests

Table 2 presents the results of descriptive statistics of the variables in the research model. It can be seen that, among the eight stock indices, BIST100 has the highest average daily return (0.0532%), while WIG20 has the lowest and even negative average daily return (−0.0029%). The Russian stock index MOEX has the highest standard deviation, while the FTSE100 has the lowest standard deviation. In terms of safe-haven assets, Bitcoin (BTC) is the asset with the highest daily return (0.2543%) as well as the asset with the highest standard deviation. Moreover, Bitcoin is the only asset that presents positive skewness (0.2811). This mean Bitcoin is the preferred asset that risk-averse investors would like best.

Table 2. Descriptive statistics.

	AEX	BIST100	CSI300	DAX	FTSE	MOEX	SP500	WIG20	GOLD	USD	BTC
Mean	0.0168	0.0532	0.0461	0.0274	0.0112	0.0274	0.0335	−0.0029	0.0314	0.0055	0.2543
Median	0.04	0.045	0.00	0.05	0.01	0.00	0.04	0.00	0.04	0.00	0.00
Maximum	10.55	12.89	9.34	11.40	9.84	28.69	11.58	8.50	11.00	2.56	36.15
Minimum	−10.75	−10.47	−9.24	−12.24	−10.87	−33.28	−11.98	−13.28	−8.49	−3.02	−38.18
Std. Dev.	1.2963	1.5968	1.6374	1.3671	1.1704	1.9387	1.2482	1.4500	1.1019	0.4713	3.7401
Skewness	−0.1520	−0.3522	−0.4310	0.0002	−0.2095	−0.4954	−0.2824	−0.3488	−0.2195	−0.0231	0.2811
Kurtosis	12.2600	7.7493	7.1670	11.3450	12.7112	49.7783	16.0295	8.5143	9.3979	5.5323	18.1548
Observations	4470	4470	4470	4470	4470	4470	4470	4470	4470	4470	4470

Note. AEX = AEX Index of Netherlands, BIST100 = BIST100 Index of Turkey, CSI300 = CSI300 Index of China, DAX = DAX Index of Germany, FTSE = FTSE100 Index of the UK, MOEX = MOEX Index of Russia, SP500 = SP500 Index of the US, and WIG20 = WIG20 Index of Poland. GOLD, USD and BTC are the returns of gold, US dollar and bitcoin, respectively.

The correlation of assets is shown in Table 3. While the US dollar is negatively correlated with all eight stock indices, gold and Bitcoin present positive correlations with these stock indices, that is, they fluctuate in the same direction. Therefore, the US dollar has the potential to act as a hedge or a safe-haven asset for the stocks examined.

Table 3. Correlation matrix.

	AEX	BIST100	CSI300	DAX	FTSE	MOEX	SP500	WIG20	GOLD	USD	BTC
AEX	1
BIST100	0.487***	1
CSI300	0.1736***	0.1236***	1
DAX	0.8957***	0.4792***	0.1621***	1
FTSE	0.8898***	0.4961***	0.1751***	0.8454***	1
MOEX	0.5382***	0.4613***	0.1517***	0.5074***	0.545***	1
SP500	0.5997***	0.3295***	0.1084***	0.6103***	0.5727***	0.2981***	1
WIG20	0.6146***	0.491***	0.1705***	0.6198***	0.6056***	0.5391***	0.4148***	1
GOLD	0.0289	0.0782***	0.0475***	0.0208	0.0566***	0.09***	0.0278	0.0883***	1
USD	−0.1542***	−0.1355***	−0.0564***	−0.1679***	−0.1631***	−0.1872***	−0.1744***	−0.1973***	−0.4051***	1
BTC	0.0583***	0.0198	0.0304***	0.0512***	0.0502***	0.0283	0.0824***	0.0411***	0.034**	−0.0144	1

Notes: AEX = AEX Index of Netherlands, BIST100 = BIST100 Index of Turkey, CSI300 = CSI300 Index of China, DAX = DAX Index of Germany, FTSE = FTSE100 Index of the UK, MOEX = MOEX Index of Russia, SP500 = SP500 Index of the US, and WIG20 = WIG20 Index of Poland. GOLD, USD and BTC are the returns of gold, US dollar and bitcoin, respectively. ** and *** indicate statistical significance at the 5 and 1 percent levels, respectively.

We use the augmented Dickey-Fuller (ADF) test and the Phillips-Perron (PP) test for testing the stationary of the return series. The results presented in Table 4 confirm that all the return series of assets are stationary at levels.

Table 4. Unit root tests.

Variables	ADF	PP
AEX	−65.8057***	−65.8165***
BIST100	−64.6110***	−64.6182***
CSI300	−64.0683***	−64.1725***
DAX	−65.5544***	−65.6090***
FTSE	−40.9631***	−67.8699***
MOEX	−67.4749***	−68.1034***
SP500	−74.6012***	−74.7074***
WIG20	−63.7759***	−63.7615***
GOLD	−64.6681***	−64.6716***
USD	−29.1673***	−66.4608***
BTC	−64.3510***	−64.3497***

Notes: The augmented Dickey-Fuller (ADF) test and the Phillips-Perron (PP) test are applied for testing the stationary of the return series. *, **, and *** indicate statistical significance at the 10, 5, and 1 percent levels, respectively.

4.2. Regression analysis

We examine the relationship between safe-haven assets and stocks using regression models in Eqs. (1)(1) $${\mathit{\text{RA}}}_{i,t}=\alpha +b_1{\times \mathit{\text{RS}}}_{j,t}{+b}_2\times {\mathit{\text{RS}}}_{j,t}\times \mathit{\text{GFC}}+b_3{\times \mathit{\text{RS}}}_{j,t}\times \mathit{\text{COVID}}+b_4{\times \mathit{\text{RS}}}_{j,t}\times \mathit{\text{WAR}}+{\epsilon }_{i,t}$$(1) and (2)(2) $$\begin{array}{c}{\sigma }_t^2=\omega +(\alpha +\gamma I_{t-1}){\epsilon }_{t-1}^2+\beta {\sigma }_{t-1}^2\hfill \\ \mathit{\text{with}} I_{t-1}=\{\begin{array}{c}1 \mathit{\text{if}} {\epsilon }_{t-1}<0\\ 0 \mathit{\text{if}} {\epsilon }_{t-1}\ge 0\end{array}\hfill \end{array}$$(2) . The following tables present the GJR-GARCH regression results for gold, US dollar and Bitcoin, respectively.

Regression results in Table 5 show that gold is a strong hedge for the AEX and DAX indices but not for the other five stock market indices. Among the three major trading partners of Russia, the Netherlands, the US and the UK, gold is a weak safe-haven asset for the SP500 and not a safe-haven asset for the other two indices. This property is completely lost during the COVID-19 pandemic and when the Russia-Ukraine conflict occurs, gold becomes a strong safe-haven asset for AEX and SP500 and a weak safe-haven asset for the FTSE. For Ukraine’s three major trading partners, Turkey, Russia and Poland, gold has almost acted as a weak safe-haven for all three periods of crisis for the BIST100 and MOEX indices. As for the WIG20 index, gold only plays the role of a weak haven asset during the COVID-19 pandemic and this property disappears in the other two crisis periods. For China and Germany, which are major trading partners of both Russia and Ukraine, the role of gold is also clearly different. Specifically, gold is only a weak safe-haven asset for the CSI300 index during the GFC 2008 period and this role fades out in the two later crises. In contrast to the DAX index, gold returns to be a strong safe-haven asset during the Russia-Ukraine conflict after losing this property in the previous two crisis periods.

Table 5. GJR-GARCH regression results for gold.

Variable	Coefficient	Std. Error	z-Statistic	Prob
AEX
C (b0)	0.020797	0.013852	1.501429	0.1332
AEX (b1)	−0.027004	0.012512	−2.158220	0.0309
AEX*GFC (b2)	0.025109	0.031350	0.800926	0.4232
AEX*COVID (b3)	0.099108	0.027864	3.556897	0.0004
AEX*WAR (b4)	−0.094842	0.047975	−1.976889	0.0481
$\omega$	0.009751	0.001555	6.272053	0.0000
$\alpha$	0.058561	0.004859	12.05099	0.0000
$\gamma$	−0.019619	0.005226	−3.754200	0.0002
$\beta$	0.944089	0.003373	279.9051	0.0000
BIST
C (b0)	0.017844	0.013666	1.305697	0.1917
BIST100 (b1)	0.045503	0.009690	4.695916	0.0000
BIST100*GFC (b2)	−0.040341	0.029439	−1.370343	0.1706
BIST100*COVID (b3)	−0.013760	0.024215	−0.568244	0.5699
BIST100*WAR (b4)	−0.018560	0.062460	−0.297157	0.7663
$\omega$	0.010229	0.001657	6.171301	0.0000
$\alpha$	0.061322	0.005090	12.04757	0.0000
$\gamma$	−0.022474	0.005521	−4.070407	0.0000
$\beta$	0.942355	0.003573	263.7365	0.0000
CSI300
C (b0)	0.020127	0.013721	1.466836	0.1424
CSI300 (b1)	0.020792	0.009614	2.162738	0.0306
CSI300*GFC (b2)	−0.003937	0.024062	−0.163637	0.8700
CSI300*COVID (b3)	0.059745	0.035574	1.679439	0.0931
CSI300*WAR (b4)	0.008238	0.068186	0.120817	0.9038
$\omega$	0.010344	0.001643	6.294543	0.0000
$\alpha$	0.060193	0.004971	12.10895	0.0000
$\gamma$	−0.020126	0.005334	−3.773054	0.0002
$\beta$	0.942181	0.003520	267.6720	0.0000
DAX
C (b0)	0.021594	0.013797	1.565165	0.1175
DAX (b1)	−0.029514	0.011016	−2.679183	0.0074
DAX*GFC (b2)	0.039982	0.033680	1.187137	0.2352
DAX*COVID (b3)	0.058005	0.022631	2.563103	0.0104
DAX*WAR (b4)	−0.106875	0.041865	−2.552850	0.0107
$\omega$	0.009577	0.001535	6.239716	0.0000
$\alpha$	0.057324	0.004833	11.86044	0.0000
$\gamma$	−0.018107	0.005188	−3.490275	0.0005
$\beta$	0.944659	0.003350	281.9467	0.0000
FTSE
C (b0)	0.020622	0.013828	1.491329	0.1359
FTSE (b1)	0.022707	0.014544	1.561252	0.1185
FTSE*GFC (b2)	0.002180	0.035266	0.061804	0.9507
FTSE*COVID (b3)	0.005737	0.027578	0.208044	0.8352
FTSE*WAR (b4)	−0.022706	0.055485	−0.409231	0.6824
$\omega$	0.010241	0.001639	6.249190	0.0000
$\alpha$	0.060537	0.004979	12.15807	0.0000
$\gamma$	−0.021251	0.005365	−3.960986	0.0001
$\beta$	0.942519	0.003522	267.5711	0.0000
MOEX
C (b0)	0.018554	0.013822	1.342394	0.1795
MOEX (b1)	0.061238	0.009992	6.128955	0.0000
MOEX*GFC (b2)	−0.044077	0.019716	−2.235597	0.0254
MOEX*COVID (b3)	−0.030889	0.021537	−1.434219	0.1515
MOEX*WAR (b4)	−0.019190	0.033502	−0.572801	0.5668
$\omega$	0.010362	0.001701	6.090760	0.0000
$\alpha$	0.060973	0.005051	12.07171	0.0000
$\gamma$	−0.021788	0.005445	−4.001400	0.0001
$\beta$	0.942223	0.003593	262.2679	0.0000
SP500
C (b0)	0.019770	0.013739	1.438963	0.1502
SP500 (b1)	0.008408	0.014209	0.591750	0.5540
SP500*GFC (b2)	−0.030351	0.033281	−0.911948	0.3618
SP500*COVID (b3)	0.083050	0.027394	3.031626	0.0024
SP500*WAR (b4)	−0.117786	0.052771	−2.232023	0.0256
$\omega$	0.010066	0.001606	6.266151	0.0000
$\alpha$	0.059019	0.004947	11.92979	0.0000
$\gamma$	−0.019855	0.005355	−3.707823	0.0002
$\beta$	0.943417	0.003460	272.6388	0.0000
WIG
C (b0)	0.020498	0.013777	1.487801	0.1368
WIG20 (b1)	0.041000	0.011008	3.724750	0.0002
WIG20*GFC (b2)	0.014834	0.030422	0.487626	0.6258
WIG20*COVID (b3)	−0.008417	0.022071	−0.381375	0.7029
WIG20*WAR (b4)	0.012244	0.041365	0.296006	0.7672
$\omega$	0.010486	0.001678	6.249544	0.0000
$\alpha$	0.061695	0.005094	12.11036	0.0000
$\gamma$	−0.023050	0.005486	−4.201915	0.0000
$\beta$	0.942007	0.003576	263.4269	0.0000

Note. If all the parameters $b_1,$ $b_2,$ $b_3,b_4$ are positive, gold is neither a hedge nor a safe-haven asset. If $b_1$ ≤ 0 and is statistically significant (insignificant), gold is a strong (weak) hedge for the stock, because the assets are not correlated with each other (on average). If $b_2$ ≤ 0 and is statistically significant (insignificant), gold is a strong (weak) safe-haven asset. ω, α, γ and β are the parameters of the conditional variance equation of the GJR-GARCH model.

The results of Table 6 indicate that USD is a hedging instrument for all studied markets, in which this property is weak in China and strong in the remaining seven markets. USD also presents a strong safe-haven role for markets in times of crisis. Specifically, USD is a safe-haven asset for investors in European countries such as the Netherlands, the United Kingdom, and Germany in all three studied periods. For the US and Polish markets, the USD represents a strong haven asset property during the GFC 2008 and the period of the Russia-Ukraine conflict, and this role weakens during the COVID-19 pandemic. In contrast, in China, the USD is a safe haven option for investors during the pandemic and this role is lost during the period of the Russia-Ukraine conflict. For Russia, gold is a strong safe-haven asset during the 2008 GFC, but this property weakens during the COVID-19 pandemic and then disappears completely when the conflict broke out.

Table 6. GJR-GARCH regression results for US dollar.

Variable	Coefficient	Std. Error	z-Statistic	Prob
AEX
C (b0)	0.003375	0.006068	0.556203	0.5781
AEX (b1)	−0.010676	0.005886	−1.814018	0.0697
AEX*GFC (b2)	−0.061641	0.012982	−4.748066	0.0000
AEX*COVID (b3)	−0.042656	0.014633	−2.914978	0.0036
AEX*WAR (b4)	−0.085105	0.021656	−3.929892	0.0001
$\omega$	0.000781	0.000181	4.310613	0.0000
$\alpha$	0.037594	0.004223	8.901550	0.0000
$\gamma$	−0.002888	0.005039	−0.573207	0.0005
$\beta$	0.960613	0.003308	290.4038	0.0000
BIST
C (b0)	0.004875	0.006048	0.806063	0.4202
BIST100 (b1)	−0.024830	0.004511	−5.504077	0.0000
BIST100*GFC (b2)	−0.014714	0.011282	−1.304138	0.1922
BIST100*COVID (b3)	0.000857	0.010918	0.078514	0.9374
BIST100*WAR (b4)	−0.049812	0.024947	−1.996715	0.0459
$\omega$	0.000826	0.000190	4.344381	0.0000
$\alpha$	0.038788	0.004360	8.896384	0.0000
$\gamma$	−0.004078	0.005116	−0.797088	0.0054
$\beta$	0.959774	0.003461	277.2842	0.0000
CSI300
C (b0)	0.003232	0.006050	0.534261	0.5932
CSI300 (b1)	−0.004580	0.004558	−1.004956	0.3149
CSI300*GFC (b2)	−0.012210	0.009385	−1.300991	0.1933
CSI300*COVID (b3)	−0.049663	0.013736	−3.615415	0.0003
CSI300*WAR (b4)	0.011353	0.030921	0.367155	0.7135
$\omega$	0.000794	0.000188	4.229781	0.0000
$\alpha$	0.040181	0.004369	9.195739	0.0000
$\gamma$	−0.006808	0.005083	−1.339538	0.0804
$\beta$	0.959946	0.003438	279.1837	0.0000
DAX
C (b0)	0.003572	0.006057	0.589667	0.5554
DAX (b1)	−0.020089	0.005266	−3.814951	0.0001
DAX*GFC (b2)	−0.058500	0.014080	−4.154761	0.0000
DAX*COVID (b3)	−0.026100	0.013174	−1.981121	0.0476
DAX*WAR (b4)	−0.089066	0.021029	−4.235439	0.0000
$\omega$	0.000789	0.000183	4.300609	0.0000
$\alpha$	0.037339	0.004204	8.882416	0.0000
$\gamma$	−0.003583	0.004977	−0.719885	0.0006
$\beta$	0.961101	0.003315	289.9384	0.0000
FTSE
C (b0)	0.003249	0.006081	0.534237	0.5932
FTSE (b1)	−0.017030	0.006681	−2.548930	0.0108
FTSE*GFC (b2)	−0.056902	0.013852	−4.107715	0.0000
FTSE*COVID (b3)	−0.026692	0.014713	−1.814229	0.0696
FTSE*WAR (b4)	−0.121329	0.025278	−4.799853	0.0000
$\omega$	0.000792	0.000184	4.305219	0.0000
$\alpha$	0.037973	0.004264	8.904852	0.0000
$\gamma$	−0.003853	0.005021	−0.767486	0.0028
$\beta$	0.960663	0.003327	288.7830	0.0000
MOEX
C (b0)	0.003817	0.006056	0.630267	0.5285
MOEX (b1)	−0.026841	0.004787	−5.606569	0.0000
MOEX*GFC (b2)	−0.026075	0.007640	−3.413042	0.0006
MOEX*COVID (b3)	−0.011613	0.011091	−1.047062	0.2951
MOEX*WAR (b4)	0.012686	0.017094	0.742147	0.4580
$\omega$	0.000811	0.000187	4.339967	0.0000
$\alpha$	0.038504	0.004226	9.111680	0.0000
$\gamma$	−0.005806	0.005097	−1.139124	0.0047
$\beta$	0.960922	0.003415	281.3896	0.0000
SP500
C (b0)	0.006132	0.006050	1.013553	0.3108
SP500 (b1)	−0.062571	0.006531	−9.580823	0.0000
SP500*GFC (b2)	0.039566	0.014464	2.735440	0.0062
SP500*COVID (b3)	−0.001283	0.012833	−0.099989	0.9204
SP500*WAR (b4)	−0.040587	0.019972	−2.032222	0.0421
$\omega$	0.000921	0.000208	4.432804	0.0000
$\alpha$	0.038667	0.004456	8.678218	0.0000
$\gamma$	−0.001519	0.005155	−0.294648	0.0000
$\beta$	0.958148	0.003427	279.6282	0.0000
WIG
C (b0)	0.003221	0.006046	0.532698	0.5942
WIG20 (b1)	−0.037152	0.005105	−7.277453	0.0000
WIG20*GFC (b2)	−0.032397	0.011929	−2.715744	0.0066
WIG20*COVID (b3)	−0.005470	0.011367	−0.481185	0.6304
WIG20*WAR (b4)	−0.063746	0.019198	−3.320382	0.0009
$\omega$	0.000838	0.000186	4.508839	0.0000
$\alpha$	0.037190	0.004239	8.774227	0.0000
$\gamma$	−0.002181	0.005094	−0.428075	0.0006
$\beta$	0.960257	0.003305	290.5894	0.0000

Note. If all the parameters $b_1,$ $b_2,$ $b_3,b_4$ are positive, USD is neither a hedge nor a safe-haven asset. If $b_1$ ≤ 0 and is statistically significant (insignificant), USD is a strong (weak) hedge for the stock, because the assets are not correlated with each other (on average). If $b_2$ ≤ 0 and is statistically significant (insignificant), USD is a strong (weak) safe-haven asset. ω, α, γ and β are the parameters of the conditional variance equation of the GJR-GARCH model.

The roles of Bitcoin are shown in Table 7. The result of parameter $b_1\mathrm{ }$ suggests that Bitcoin is a hedge for all markets except Poland and China. In particular, Bitcoin represents a strong hedging property for Dutch, German and Turkish stocks, while it is a weak hedge for stocks in the US, UK and Russia markets. Besides, Bitcoin has barely performed the role of a safe-haven asset during the COVID-19 pandemic but by the time the Russia-Ukraine conflict takes place, Bitcoin becomes a strong safe haven for Dutch, British, Turkish and Russia stocks and a weak safe-haven asset in China. In the US, Bitcoin is not a safe-haven asset for stocks during any period of crisis. It is also an asset whose property varies greatly from one crisis to another.

Table 7. GJR-GARCH regression results for Bitcoin.

Variable	Coefficient	Std. Error	z-Statistic	Prob
AEX
C (b0)	0.000805	0.002063	0.390154	0.6964
AEX (b1)	−0.087925	0.000846	−103.9039	0.0000
AEX*GFC (b2)	–	–	–	–
AEX*COVID (b3)	1.770590	0.037642	47.03819	0.0000
AEX*WAR (b4)	−0.522268	0.224923	−2.321984	0.0202
$\omega$	0.033843	2.2E-104	1.5E + 102	0.0000
$\alpha$	0.282819	5.7E-103	5.0E + 101	0.0000
$\gamma$	0.134073	5.5E-103	2.4E + 101	0.0000
$\beta$	0.801343	9.0E-104	8.9E + 102	0.0000
BIST
C (b0)	0.000446	0.007147	0.062382	0.9503
BIST100 (b1)	−0.194263	0.004836	−40.16967	0.0000
BIST100*GFC (b2)	–	–	–	–
BIST100*COVID (b3)	1.583427	0.044531	35.55778	0.0000
BIST100*WAR (b4)	−1.041961	0.222257	−4.688095	0.0000
$\omega$	0.000155	1.52E-06	101.7625	0.0000
$\alpha$	0.174063	0.006326	27.51704	0.0000
$\gamma$	0.323754	0.012770	25.35171	0.0000
$\beta$	0.828970	0.002687	308.5592	0.0000
CSI300
C (b0)	−0.000706	0.003655	−0.193055	0.8469
CSI300 (b1)	0.082957	0.000688	120.6604	0.0000
CSI300*GFC (b2)	–	–	–	–
CSI300*COVID (b3)	0.757823	0.050448	15.02173	0.0000
CSI300*WAR (b4)	−0.447318	0.293622	−1.523450	0.1276
$\omega$	0.004882	0.000138	35.44406	0.0000
$\alpha$	0.291020	0.010503	27.70756	0.0000
$\gamma$	−0.119517	0.012884	−9.276168	0.0000
$\beta$	0.846197	0.003224	262.4645	0.0000
DAX
C (b0)	−0.000118	0.001358	−0.086553	0.9310
DAX (b1)	−0.057109	0.000665	−85.94172	0.0000
DAX*GFC (b2)	–	–	–	–
DAX*COVID (b3)	1.455136	0.026479	54.95518	0.0000
DAX*WAR (b4)	−0.394410	0.240487	−1.640048	0.1010
$\omega$	5.74E-05	2.00E-06	28.73763	0.0000
$\alpha$	0.439063	0.014477	30.32854	0.0000
$\gamma$	−0.278763	0.015404	−18.09705	0.0000
$\beta$	0.826355	0.003132	263.8124	0.0000
FTSE
C (b0)	−0.003466	0.027504	−0.126034	0.8997
FTSE (b1)	−0.017804	0.012412	−1.434458	0.1514
FTSE*GFC (b2)	–	–	–	–
FTSE*COVID (b3)	1.733809	0.047351	36.61586	0.0000
FTSE*WAR (b4)	−0.713896	0.331919	−2.150817	0.0315
$\omega$	0.008139	5.2E-105	1.6E + 102	0.0000
$\alpha$	0.117554	4.4E-104	2.7E + 102	0.0000
$\gamma$	−0.008116	8.1E-104	−1.0E + 101	0.0000
$\beta$	0.920962	1.38E-06	665453.5	0.0000
MOEX
C (b0)	−0.000280	0.032480	−0.008614	0.9931
MOEX (b1)	−0.002011	0.023202	−0.086661	0.9309
MOEX*GFC (b2)	–	–	–	–
MOEX*COVID (b3)	1.822933	0.042206	43.19110	0.0000
MOEX*WAR (b4)	−1.880032	0.184085	−10.21287	0.0000
$\omega$	0.002876	6.78E-05	42.44453	0.0000
$\alpha$	0.150613	0.007246	20.78479	0.0000
$\gamma$	−0.001657	0.010037	−0.165052	0.0009
$\beta$	0.894208	0.002241	398.9928	0.0000
SP500
C (b0)	−0.001316	0.035201	−0.037373	0.9702
SP500 (b1)	−0.008722	0.021490	−0.405877	0.6848
SP500*GFC (b2)	–	–	–	–
SP500*COVID (b3)	1.614334	0.047470	34.00720	0.0000
SP500*WAR (b4)	0.034443	0.155873	0.220967	0.8251
$\omega$	0.004214	0.000116	36.24712	0.0000
$\alpha$	0.134171	0.004091	32.79470	0.0000
$\gamma$	0.040405	0.008664	4.663368	0.0000
$\beta$	0.870147	0.002444	356.0742	0.0000
WIG
C (b0)	−0.130573	0.043120	−3.028146	0.0025
WIG20 (b1)	0.144415	0.806952	0.178963	0.8580
WIG20*GFC (b2)	–	–	–	–
WIG20*COVID (b3)	−0.593067	0.071503	−8.294339	0.0000
WIG20*WAR (b4)	0.082761	0.058187	1.422321	0.1549
$\omega$	2.848358	0.129124	22.05915	0.0000
$\alpha$	0.296178	0.023449	12.63067	0.0000
$\gamma$	0.133962	0.033734	3.971165	0.0001
$\beta$	0.516800	0.018863	27.39722	0.0000

Note. If all the parameters $b_1,$ $b_2,$ $b_3,b_4$ are positive, USD is neither a hedge nor a safe-haven asset. If $b_1$ ≤ 0 and is statistically significant (insignificant), USD is a strong (weak) hedge for the stock, because the assets are not correlated with each other (on average). If $b_2$ ≤ 0 and is statistically significant (insignificant), USD is a strong (weak) safe-haven asset. ω, α, γ and β are the parameters of the conditional variance equation of the GJR-GARCH model.

4.3. Discussion

The regression results are summarized in Table 8 below. The findings suggest that each asset plays a different role in different stock markets. Of the three assets studied, gold acts as a strong hedge in the Netherlands and Germany but does not in the other six. USD is the hedge for investors in all markets. Bitcoin - a fairly prominent financial asset recently, is also a hedge in all examined markets excluding China and Poland.

Table 8. Summary of regression results.

ASSETS		GOLD	USD	BITCOIN
HEDGE	AEX	S	S	S
	SP500	N	S	W
	FTSE	N	S	W
	CSI300	N	W	N
	DAX	S	S	S
	BIST100	N	S	S
	MOEX	N	S	W
	WIG20	N	S	N
SAFE-HAVEN ASSET DURING GFC 2008	AEX	N	S	–
	SP500	W	N	–
	FTSE	N	S	–
	CSI300	W	W	–
	DAX	N	S	–
	BIST100	W	W	–
	MOEX	S	S	–
	WIG20	N	S	–
SAFE-HAVEN ASSET DURING COVID-19 PANDEMIC	AEX	N	S	N
	SP500	N	W	N
	FTSE	N	S	N
	CSI300	N	S	N
	DAX	N	S	N
	BIST100	W	N	N
	MOEX	W	W	N
	WIG20	W	W	S
SAFE-HAVEN ASSET DURING RUSSIA-UKRAINE CONFLICT	AEX	S	S	S
	SP500	S	S	N
	FTSE	W	S	S
	CSI300	N	N	W
	DAX	S	S	S
	BIST100	W	S	S
	MOEX	W	N	S
	WIG20	N	S	N

Note. (S), (W), (N) denote a strong, weak, and not a safe-haven asset, respectively.

Research in countries with close trade links with Russia and Ukraine also presents the change in the safe-haven property of assets during different periods of crisis. Gold is a weak safe-haven asset in the US, China, Turkey and a strong safe-haven asset in Russia during GFC 2008, but this property weakens or completely disappears during the COVID-19 pandemic, consistent with Hood and Malik (2013), Corbet et al. (2020), Akhtaruzzaman et al. (2021), and Cheema et al. (2022). During the Russia - Ukraine conflict, gold returns as a strong safe-haven asset for stock investors in the Netherlands, US and Germany, but not in China. This result is in contrast to Yatie (2022) who argues that gold has completely lost its safe-haven property due to the impact of the Russia-Ukraine war. This can explain that in the context that the economy is recovering from the pandemic, the risk of global inflation makes investors fear and rush to buy gold to hoard. Especially when Russia announced an attack on Ukraine, the fear of political risks make investors choose gold as a safe haven for their assets.

The findings also suggest that Bitcoin is not performing as a safe-haven asset during the COVID-19 pandemic. This result is consistent with Conlon et al. (2020), Jeribi and Manzli (2020) and Disli et al. (2021). Smales (2019) argues that Bitcoin is more volatile and less liquid than other assets, and transaction costs are high, especially during volatile market periods. Therefore Bitcoin should not be considered a safe haven asset. However, the safe-haven property of Bitcoin has changed during the Russia-Ukraine conflict. Specifically, Bitcoin is a strong safe-haven asset for the stock markets of European countries such as the Netherlands, UK, Germany, and Russia, and a weak safe-haven asset in China. This result is partly contrary to Yatie (2022) and can be explained as follows: After the war happens, Ukraine legalized cryptocurrencies, including Bitcoin. This allows Ukraine to receive international funding while avoiding Russian obstacles, and cryptocurrency usage in Ukraine is growing dramatically^1. At the same time, the US, the European Union (EU), and a number of other countries tried to impose a series of diplomatic and economic sanctions on Russia. To avoid Western financial and banking sanctions, Russians are also turning to Bitcoin as a safe haven. Ruble-Bitcoin exchanges show an increase in investment, as the Russian currency has depreciated by about 20% against the US dollar since the outbreak of the conflicts². Theiri et al. (2022) show that transactions on a centralized Bitcoin exchange in Russian rubles have increased to very high levels, and a similar pattern is observed in Bitcoin transactions on the Hryvnia exchange in Ukraine. It supports the argument of Sparkes (2022) that the crypto market gets more attention as a strategic alternative to the global trade and payment channel and the liquidity of Cryptocurrencies in war has passed the scrutiny of the traditional system. This is consistent with the view that cryptocurrencies can be used as a store of value to mitigate financial losses during turbulent market times.

Our results also prove that the asset that maintains the property of a stable safe-haven asset through periods of crisis is the USD. In European countries such as the Netherlands, the UK, and Germany, the USD is a strong safe-haven asset in all three crises. Despite being the domestic currency of the US, the USD only plays a strong safe-haven role during the Russia-Ukraine conflict period. For stock investors in China, the safe-haven role of gold strengthens during the COVID-19 pandemic, but this role completely disappears during the Russia-Ukraine conflict. In general, the safe-haven property of the USD strengthens during the COVID-19 pandemic and weakens during the conflict. This result is similar to Wen and Cheng (2018), Fidalgo (2021) and opposite to Ji et al. (2020), Cheema et al. (2022), while adding evidence to recent discussions that the safe-haven property of assets can vary across different periods and markets. Particularly in Russia, the USD was once an option to protect investors’ assets during the GFC 2008, but the role of the USD gradually weakened during the COVID-19 pandemic and was completely lost when the conflict with Ukraine occurred. This can be explained that de-dollarization has been a top priority for Russia since 2014. And in response to European sanctions, the Russians have gradually phased out the US dollar – considered ‘toxic currency’ – in commercial and investment transactions³.

5. Conclusion

For investors, finding a safe haven for their assets, especially during volatile market periods, is a necessity. Therefore, when the market falls into periods of crisis, it is necessary to re-evaluate the hedging and safe-haven properties of traditional assets. In this study, we assess the safe-haven role of gold, Bitcoin, and USD for stocks over a period including the GFC 2008, the COVID-19 pandemic, and the Russia - Ukraine conflict. Research conducted in 8 countries that are major trading partners of Russia and Ukraine suggests that USD is a hedge for all markets, while Bitcoin and gold are hedges only for some of the studied stock indices. The study also provides evidence of the changing safe-haven property of assets over periods of crisis. Specifically, gold - a traditional asset, loses its role as a safe-haven asset during the COVID-19 pandemic but this role is restored in the Netherlands, US and Germany. Similarly, Bitcoin does not perform as a safe haven during the COVID-19 pandemic but is a strong safe-haven asset for the stock markets of European countries such as the Netherlands, UK, Germany, and Russia, and a weak safe-haven asset for China investors. In contrast, USD presents a stable safe-haven property through periods of crisis. This role is waning in Russia due to policies of de-dollarization as well as the phasing out of the use of the US dollar as a response to Western sanctions. These results help investors and fund managers in choosing the best safe-haven assets, especially during volatile market periods. Additionally, these findings can assist portfolio managers in implementing better hedging strategies to limit the adverse impact of volatility in the market. For policymakers, the role of policy regime changes must be meticulously evaluated because policy shocks can have a significant impact on the future relationship between stocks, foreign exchange currencies, gold, and cryptocurrencies.

Besides the above-mentioned findings, the study still has some limitations. Firstly, we only focus our research on 8 stock markets which are major trading partners of Russia and Ukraine. A broader study needs to be conducted to examine whether, in addition to the eight stock markets studied, the role of safe haven assets in other markets has changed. Secondly, there are only three safe-haven assets under consideration: gold, Bitcoin, and USD. Therefore, to achieve more comprehensive research results, future researchers may incorporate more asset classes and stock indices in their research.

Authors’ contributions

Le Thi Thuy Van’s tasks on the article development: Investigation, Software, Data curation, Writing-Original draft preparation. Tran Thi Kim Oanh’s tasks on the article development: Conceptualization, Supervision, Validation, Writing-Reviewing and Editing. Nguyen Thi Hong Ha’s tasks on the article development: Software, Writing-Original draft preparation, Resources. All authors read and approved the final manuscript.

Acknowledgements

The authors acknowledge being supported by the University of Finance - Marketing, Viet Nam.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Van Le Thi Thuy

Dr. Le Thi Thuy Van is working at National Institute for Finance, Hanoi, Vietnam. Her research area involves Financial Management, Finance and Banking.

Tran Thi Kim Oanh

Dr. Tran Thi Kim Oanh is working as a lecturer at Faculty of Finance and Banking, University of Finance – Marketing, Ho Chi Minh City, Vietnam. Her research area involves Financial Management, Finance and Banking.

Nguyen Thi Hong Ha

Ms. Nguyen Thi Hong Ha is working as a lecturer at Faculty of Accounting, Finance – Banking, Ho Chi Minh City Industry and Trade College, Ho Chi Minh city, Viet Nam. Her research area involves Finance and Banking.

Notes

1 https://www.euronews.com/next/2022/03/17/ukraine-war-zelenskyy-signs-a-new-law-officially-legalising-bitcoin-and-other-cryptos.

2 https://www.reuters.com/business/finance/russia-could-legalise-cryptocurrencies-means-payment-industry-minister-2022-05-18/.

3 https://www.russia-briefing.com/news/russia-to-reduce-trade-in-toxic-us-dollar-euros.html/.