An Econometric Analysis of the Relationship Between Bitcoin & Gold: 2018

Blake Richardson
An Econometric Analysis of the Relationship Between Bitcoin Gold

Is Bitcoin the gold of the digital age?

Intro: Bitcoin, first proposed in a whitepaper [1]written by Satoshi Nakamoto in 2008, is a decentralized peer-to-peer open-source distributed electronic currency system. Bitcoins are cryptographic tokens, rewarded to miners for transaction (Tx) verification, where Tx’s are verified through consensus by a global network of nodes using proof-of-work (PoW) consensus.[2] Secured using cryptographic proofs on a data structure known as Blockchain, data is distributed to nodes using Merkle trees without a central authority certifying each Tx. Bitcoin is thus considered a sound mechanism to transfer value digitally due to the security protocols it was built upon. Bitcoin, since its launch in 2009, has been classified as an electronic form of cash or gold, and by skeptics a purely speculative digital asset with no inherent value. The agreed-upon classification, both economically and socially, has significant implications on how users, policymakers, and investors view and interact with this form of digital value. The process by how Bitcoins are generated and distributed is often compared to the economics of gold mining due to a number of factors. Bitcoin uses energy to “mine” transactions through PoW, which requires computers to use spend energy-intensive computational resources to verify transactions. The difficulty to mine Bitcoins increases over time, referred to as block difficulty adjustments. Similarly, gold mining is an energy-intensive process that results in useable gold ore. As more gold is discovered, the remaining gold is harder to discover and mine. Bitcoin has a limited supply, capped at 20,999,999.99 but can be subdivided into units as small as 1 Satoshi, which is .00000001 Bitcoin. Gold theoretically has an uncapped supply if you factor in the asteroid harvesting, synthetics, and time, but practically gold is priced in the market as a scarce asset. At the time of writing, the above-ground gold reserves are estimated to be 187,200 tons with a market capitalization of ~$7.7 trillion at a spot rate of $1,275/oz.[3] Bitcoin’s supply currently sits at 17.4 million with a market capitalization of ~$115 bln at a spot rate of $6,609, significantly smaller than gold’s. Gold has more functional uses in electronics and is more widely accepted than Bitcoin as a store of value. However, Bitcoin provides a fast, censorship-resistant, and immutable digital value mechanism which provides massive utility to those evading corrupt or unstable governments as well as any individual who doesn’t want their wealth influenced by monetary policies such as quantitative easing. Bitcoin’s increased popularity as a speculative investment, specifically in January of 2018, when the price reached its peak, drove it into the news and brought attention to the subject. Prior causality studies suggested that Bitcoin might serve as a hedge against equities and currencies (Bouri, Molnar, Azzi, Roubaud & Hagfors, 2017b; Dyhrberg, 2016b), the commodity index (Bouri et al.; 2017b), and stock market expectation of near-term volatility — as measured by the VIX (Bouri et al. (2017a). The relationship between Bitcoin and financial stress causality was explored by Elie Bouri, Rangan Gupta, Chi Keung Marco Lau, David Roubaud, and Shixuan Wang in April of 2018, which established that Bitcoin has little evidence to suggest it is a hedge against financial stress. To enhance the empirical literature around the Bitcoin space, this study examines the relationship between Bitcoin and gold prices.

The key question proposed by cryptoeconomic theorist is: Does Bitcoin behave as a store of value and is it thus a viable substitute for gold? If so, is it a more optimal alternative than gold to store wealth? This analysis will focus on how Bitcoin and gold prices are related, markets are valuing the assets, and if the price of gold Granger-causes the price of Bitcoin and vice-versa. Many discussions in the cryptocurrency community focus on the importance of the relationship between these two assets with very little analysis done to validate the claims that their relationship is significant. On a data level, this study considers the price of gold using the World Gold Council Daily Gold/USD data and for Bitcoin, I will use the daily close price of Bitcoin provided on, which aggregates Bitcoin/USD exchange data.

Economists, cryptographers, and cypherpunks alike have all lauded Bitcoin as a financial revolution that will transform how value is transferred by providing an inexpensive, censorship-resistant, efficient medium of exchange. Blockchain, a public permissionless ledger that cryptographically verifies Tx’s, provides the underlying data structure to facilitate the transfer of Bitcoin. Anyone can use the open-source software to set up a node on the network which provides a complete immutable record of all transactions of the Bitcoin blockchain. The nodes verify every transfer globally which ensures the integrity of the system. This allows for a truly global asset that anyone can audit, verify, hold, and transfer. The creation of a universal store of value to act as a substitute for gold is a major motivation behind the continued work and research around Bitcoin. One of the many problems Bitcoin evangelist point to is that the true supply of gold is unknown and untraceable due to government secrecy around their gold reserves, mining practices, and private holdings. This creates market inefficiency in-favor of governments and IGOs, specifically the United States Government, Germany, and the IMF which hold the world’s largest gold reserves, that can manipulate the supply of their currency using quantitative easing whenever they choose to do so. The US Federal Reserve is estimated to hold 27% of the total gold supply.[4] Physically-backed gold ETFs are typically used in short positions to hedge against equity market volatility, dollar weakness, and inflation. The physically-backed nature of these assets eliminates the uncertainty associated with futures-based hedging strategies.[5] Exposure to gold is typically achieved through heavily regulated ETFs backed by the government who regulates the market it trades in. In countries with corrupt institutions, instability is common and can destroy wealth due to the monetary policies enacted by their governments. Zimbabwe and Venezuela are notorious examples of ineffective and corrupt policy making resulting in hyperinflation that adversely impacted their economies and local wealth. Direct exposure to Bitcoin can drastically reduce the risk of a portfolio by acting as a hedge against these actions. This can prevent wealth destruction for their citizens and allow them to migrate to stable economies that have functioning Bitcoin to fiat currency markets. This analysis attempts to understand the relationship between Bitcoin and gold to determine if the two assets behave and are valued similarly allowing for investors to use them as substitutable store-of-value assets in their portfolios. Additionally, this will help both traditional economist and cryptocurrency stakeholders better understand the relationship between traditional financial assets and cryptocurrencies, a topic that is hotly debated in the industry. Economists focused on cryptocurrencies typical refer to Bitcoin as an alternative to gold or as a cash equivalent, which has implications on how it is used by consumers, regulated by governments, and perceived by institutional investors. Studying the relationship between Bitcoin and gold is important in generating a better understanding of how market information is transmitted and then embedded in asset prices.

Lit Review: Bitcoin is an open-source decentralized peer-to-peer electronic value system. Since the price peaked in January of 2018 at nearly $20,000, up from nothing less than a decade earlier, attention by economist and industry stakeholders skyrocketed. Its creation was the cumulation of years of research around cryptography, incentive design, electronic cash, and digital gold systems. Bitcoin’s popularity in the early years came from the perceived failures of the governments and central banks responsible for the global financial crisis in 2008 and the European Sovereign Debt Crisis of 2010–2013. Bitcoin was the first cryptocurrency to be created and was the inspiration for the term, but a whole digital asset class has emerged in the 10 years since its creation. Central banks/authorities guarantee fiat currencies and control its supply, but Bitcoin is a decentralized network of nodes, which mean it operates absent of a central authority. Bitcoin is a deflationary system, rather than inflationary like fiat currencies, because the supply is capped at 20,999,999.999 and private keys (private keys are similar to passwords but cannot be restored if lost) are periodically and randomly lost which have Bitcoin stored on them.

The relationship between Bitcoin and global financial stress is important because it suggests how the market treats the argument that Bitcoin is a store of value. The analysis done by Elie Bouri, Rangan Gupta, Chi Keung Marco Lau, David Roubaud, and Shixuan Wang (2018) studies Bitcoin’s relationship to global financial stress, using the Global Financial stress index (GFSI) created by Merrill Lynch. It uses three main approaches to study the relationship 1) A Copula-based approach to show the dependence between two random variables, in this instance Bitcoin returns and GFSI; 2) out-of-sample approach called the Granger causality in distribution (GCD), which captures the Granger causality in distributions in each conditional quantile; and 3) is the cross-quantilogram approach, which allows measuring of directional predictability in quantiles. The results from the copula-based dependence test in the study show evidence of right-tail dependence between the global financial stress index and Bitcoin returns. They found that Bitcoin markets can perform well even when the global financial markets are in a depression. This evidence indicates that Bitcoin provides a channel against global financial stress, however, the evidence surrounding this was minimal. The article concludes by stating that “empirical evidence of its economic and financial aspects, particularly its role as a safe haven against global financial stress, is relatively scarce” and that “future research can use the quantile dependence approach to more thoroughly examine Bitcoin’s safe-haven property against conventional assets and commodities.” Bitcoin’s use as a safe-haven asset must be analyzed through the lens of its relationship to gold, the global standard for a safe haven asset. Their findings lay a groundwork of understanding Bitcoin using its relationship with traditional macro-events. This will help inform policymakers and future research. This study did not include gold, which is the primary topic I analyze in this article. Bouri et. all’s (2018) article also provided in-depth analysis about the current field of study around Bitcoin’s use as a hedge against currencies, equities, commodities, and VIX, citing other published works from authors Bouri and Dyhrberg.

The relationship between Bitcoin volume and its’ returns and volatility was explored by Balcilar, Mehmet; Bouri, Elie; Gupta, Rangan; Roubaud, David in 2017. The analysis tools and methods presented here can provide a basis for understanding how Bitcoin is priced in the market. The study’s model employs a non-parametric causality-in-quantiles test to analyze the causal relationship between trading volume and Bitcoin returns and volatility, over the whole of their respective conditional distributions. It provides an outline of the previous research done in the field with respect to volume and returns in traditional markets such as equities, real estate, and commodities, then uses that as the basis for analysis on Bitcoin volume-driven returns. This study provides relevant and needed analysis because in speculative markets, like Bitcoin, understanding the volume–return paradigm is essential to shedding light on potential implications for trading strategies and policy creation. If the transaction volume in the Bitcoin market has predictive power for its returns, this suggests that practitioners will be able to construct volume-based strategies to increase profits. The study found that Bitcoin volume had a significant relationship to returns and could be used as an active indicator, while volatility had no statistically significant relationship under this approach.

A study conducted by Bouri et al. (2017a) examined Bitcoin’s ability to hedge against the VIXs of developing and emerging markets. “After decomposing Bitcoin returns into different frequencies and applying quantile-on-quantile regressions, the authors show that Bitcoin does act as a hedge against global uncertainty at both the lower and the upper ends of Bitcoin returns and global uncertainty, particularly on shorter investment horizons.” (Bouri et al. (2017)) This analysis doesn’t capture a Granger-causality based approach in their distributions or quantiles.

Gold: There is substantial research on gold prices, its use as a store of value, and how it operates in markets. Gold is one of the oldest assets and was formerly used to back the majority of fiat currencies in the world. It is well established in the market as the most efficient mechanism to hedge against inflation and dollar volatility. The removal of the dollar peg to gold post-Bretton Woods was a major event in the history of gold because this established the current floating fiat regime and further shifted investor interest to commodities as safe-haven assets. Gold provides a safe haven in times of volatility for investors who get exposure through derivatives markets. The derivative market, which includes futures, options, forwards, swaps, and warrant contracts is the largest of any asset class and provides a unique means of exposure to many assets including commodities such as gold. Gold has a large derivatives market and functioning ETFs, whereas regulated Bitcoin exposure can only be achieved through direct purchase or futures contracts.

The article Market Anticipations of Government Policies and the Price of Gold by Stephen W. Salant and Dale W. Henderson seeks to understand how government policies around the sale of gold impact its’ market price. This article provides insight into how the gold market is formed from a policy perspective and why gold prices change. The analysis focuses on the factors of gold sales that lead to its market price. This argues that government attempts to manipulate supply at gold auctions or participate in price pegging will lead to an attack by speculators. Bitcoin also experiences new random market liquidity events do to government auctions though are not significant to total supply, which is capped. The price responsiveness between these two events can also provide insight into the relationship between the two assets. Both gold and Bitcoin share many properties, like that fact that it is a limited resource that can be mined, used, and lost. The study states that the standard price theory cannot be used to explain how the price of a resource traded by competitive speculators could persistently rise by more than the rate of interest, nor can it be employed to account persuasively for the timing of the observed breaks in the price. This argues that for any given metal commodity, such as gold, standard price theory accurately predicts the price. The study also explores how risk-neutral agents anticipate an auction of additional stock at an unknown time. These factors provide insight into the determinants of the supply and thus the price of gold. When comparing Bitcoin to gold, one of the main differentiating factors is the total supply at a given time. Some of the key insights from this article were: 1) Extraction costs for gold are significant and have risen; 2) The gold market is not competitive and is dominated by one seller, the South African Reserve Bank; and 3) gold holders don’t know the size of the stock available for private use as governments can sell their massive stockpiles.

Methods: 1) The first test is a correlation function between the price of the Bitcoin.Price and Gold.Price variables. The Pearson correlation coefficient helps determine the basic dependence between two quantities and can indicate linearity. A strong positive or negative correlation between the two would indicate that they move similarly in the market and would thus support the store of value argument made by many economists. This only provides significant practical information if the correlation is close to -1, 0, or 1 as that indicates a strong linear relationship between the two variables analyzed.

2) The second test is a log-log regression analysis using R to determine if the relationship between the two assets is statistically significant using the P-value or T-test. This regression can be interpreted as a 1% change in X results in a B1 % change in Y with e being the error term.

Regression Model:

ln(Bitcoin.Price) = B0 + B1 ln(Gold.Price) + error

3) The next test adds a dummy variable to in the log-log regression for Bitcoin and gold. The binary dummy variable will indicate the period before and after an exogenous event, the Mt.Gox Bitcoin Exchange hack on February 7th, 2014 and subsequent closure of the company. The Mt.Gox hack was a significant event in the history of Bitcoin and negatively impacted the price of Bitcoin. The company filed bankruptcy in February and lost nearly $360million worth of Bitcoin at its fair market value at the time of loss. The exchange handled nearly 70% of Bitcoin’s daily trade volume and thus was a central piece to the Bitcoin ecosystem. This was significant because it undermined public trust in the asset and resulted in massive volatility, essential characteristics of ‘store of value’ assets.

Regression Model:

ln(Bitcoin.Price) =B0 + B1 ln(Gold.Price) + Dummy + error

The dummy variable’s value was 0 for all dates before February 7th and a 1 for all dates after. This is to model the lifelong impact of this event on the relationship between Bitcoin and gold.

4) I then run a new regression that shortens the dummy effect period to only one year. Dummy2 impacts prices from February 7th, 2014 to February 9th, 2015. I use one year as the treatment effect period as an experiment to determine how long the event effects peoples trading decisions and the estimated impact within that timeframe. Information relevance and exogenous shocks decline in importance over time and thus shouldn’t always be applied in perpetuity. The dummy then can accurately capture the effect of the event over the period following the event without altering the long-run relationship between Bitcoin and gold as it is factored out of a trader’s decision-making model. The goal is to test an exogenous event on Bitcoin’s reserve status in a tine of crisis.

Regression Model:

ln(Bitcoin.Price) =B0 + B1 ln(Gold.Price) + Dummy2 + error

5) The next test is the linear Granger Causality Distribution (GCD) test. This is used to determine if one time series can be useful in forecasting another. Regressions reflect statistically significant relationships, but don’t measure predictive causality. From a modeling perspective, it is more informative to explore the causal relationship between Gold Prices and Bitcoin Prices using the GCD test. For market participates, this causality can have significant effects on if they determine to use it as a safe haven asset when the dollar is volatile, or inflation is rampant. We apply the GCD with the null hypothesis that Xt does not Granger cause Yt in distribution, so Gold Prices do not Granger Cause Bitcoin Prices. The Granger Causality test is performed using this formula:

where v1(t) is a white Gaussian random vector, and ai is a matrix for every i, interval. A time-series Yi is called a Granger cause of another time series Yj, if at least one of the elements aj(j,i) for t=1,….,L is significantly larger than zero (in absolute value). The GCD test has limitations as it is not true causality but only the identification of a cause-effect relationship that has constant conjunctions. For example, if X and Y are both driven by a common third variable Z with different lags, the alternative hypothesis of the GCD test might still fail to be rejected.

6) The last tests I performed were a GCD test of Gold.Prices~Bitcoin.Prices with a one-week lag and its bidirectional regression: Bitcoin.Prices~Gold.Prices. This test helps determine if there is in lag effect present in the data that might result in the time series being a significant predictor. Market information is sometimes delayed and can impact prices in the future.

Data: Data used in this study are daily (5 days per week) and cover the period from August 18th, 2010 to October 4th, 2018. It consists of Bitcoin prices and gold prices in USD. It omits the days in which Bitcoin markets are open, but gold markets are not, typically weekends and holidays. Bitcoin prices are from, a resource commonly used for blockchain data including the BTC/USD price, which it aggregates from the most used exchanges by volume. Bitcoin trades 24/hours a day 7/days a week, which means the close price each day is at midnight UTC 0. Gold prices were accessed from the World Gold Council which has the daily USD spot prices. The only available data for gold prices do not include weekend or holiday trade data, as is standard for traditional financial market data. Bitcoin and gold returns are negatively correlated for the time frame selected (-0.289). The variables used in my analysis are summarized below from R including the Date, Gold.Price, Bitcoin.Price, Dummy, and Dummy2. The mean price of gold and Bitcoin were surprisingly similar over the timeframe selected, but Bitcoin volatility and price range were drastically higher as indicated by the descriptive statistics below:

Hypothesis: The null hypothesis of this experiment is that Bitcoin and gold are not significantly related, and gold prices do not Granger-cause Bitcoin prices.

Empirical Results: Referencing the descriptive statistics from the data section, the range of Bitcoin values shows the massive growth it experienced since its inception. Gold prices were more consistent than Bitcoin’s and experienced significantly less volatility over the timeframe analyzed. The mean price, however, was shockingly similar for the date range, even if there is little of statistical significance to be gleaned from that. The first test run was a basic correlation analysis between Daily Bitcoin/USD prices and Dai1y Gold/USD prices. The resulting correlation was -.289, a negative relation but insignificant statistically. Although insignificant, the mild negative correlation is interesting as it is the reverse of what initially theorized when compared as substitute stores of value. A negative correlation does not disprove the store of value hypothesis as the two goods could be acting as substitutes with one another based on public perceptions or other exogenous non-price factors.

I then ran the plotting function in R to output the linear price and log price of Bitcoin and gold prices from the timeframe selected from 2010 to 2018. The plots below allow for a visual interpretation of the relationship between the assets.

The linear and log returns for Gold Price were nearly identical based on the fitted graph provided by R. As seen in the Y scale, the price variance across gold for the time frame selected was low as all values were between $1,000 and $1,900. This means that the log(price) or % change won’t be large each day. Bitcoin, however, was highly volatile with the price growing from pennies to nearly $20,000. This massive range is reflected in both the linear plot and the log plot. The log plot shows a more descriptive analysis as it shows local highs were reached before retracements, only to grow to even larger highs thereafter. This clear visual difference is impactful to investors’ perceptions of these assets when analyzing the market.

The results from the first regression of log(Bitcoin prices) on log(Gold Prices) can be seen below:

The log-log function estimates that a 1% change in Gold Price leads to a -12.56% change in Bitcoin prices over the timeframe selected. This result was statistically significant at less than 1% level but does not accurately predict causality due to an infinite number of market factors. This information primarily provides us with information about the relationship between their prices over this timeframe. However, we can interpret that in times that Gold Price increased, which likely means that the USD had weakened, Bitcoin price declined. This is informative because it indicates that investors may view gold as a hedge against Bitcoin.

The next test performed was a log-log regression with a dummy variable for the exogenous Mt.Gox Exchange hack and closure. It regressed Gold Price and a Dummy against Bitcoin Price.

The dummy variable’s coefficient in the regression was 5.6724 and significant at the less than 1% value given the P-Val. This estimates that after the Mt.Gox hack, Bitcoin Prices increased by 567.24% increase in prices after Mt.Gox. This would indicate over that it did not have a long-term adverse impact on prices. Log(Gold.Price) was also significant for the regression and was estimated to be 3.29 and significant, which means for every 1% Gold Price increase, Bitcoin increased by 3.299%.

The next regression used the same dummy variable but only applied the treatment to the time period of 1 year following the Mt.Gox hack. The results can be seen below:

Dummy2 was estimated at 0.9060 and was significant. This is relevant because it is a drastically lower estimate than in the previous regression. The year immediately following the event had slower growth in Bitcoin prices due in part to this event. The significant estimate of -12.28% for Gold Prices in this regression is similar to the original regression’s estimate of -12.58%, without the dummy variable. This indicates that the dummy event significantly impacted Bitcoin price during that period, but not its relationship to gold.

I then performed a Granger-causality test to determine if Gold Prices acted as an indicator of Bitcoin prices. This determines whether the gold price at timestep t is Granger caused by the bitcoin price at timestep t — n, where n is the number of timesteps (in our case, days) between the gold price and the Bitcoin price that causes said gold price.

The analysis shows that Gold Prices do not Granger-cause Bitcoin prices at a significance level we would accept. The Pr(>F) value of the analysis was 0.6045 at 2 degrees of freedom, which is not accepted at the 5% level of significance. This indicates that Gold.Price is not a leading indicator of Bitcoin.Price. The next test run determines if Bitcoin Prices Granger-causes Gold prices.

This Pr(>F) value of this test was even higher than the previous Granger test at 0.6778. This is not significant at a level we would accept. There is no bidirectional causality between these two variables and neither can be said to be an indicator of another.

The next results plotted were the daily difference in USD price from the previous day. This Diff() function, an essential part of the GCD test, plotted forms two distinct graphs that distinguish the differences between Gold and Bitcoin, as seen below:

The difference in aggregate daily USD price in the gold market was highly consistent over the timeframe analyzed. The range prices range from $65 to $-142. This graph is similar to volatility in that it measures the change in price from the previous period. Bitcoin prices can be visually distinguished as highly volatile with intraday swings reaching 3,000 to -4,000.

The next test I performed was a GCD test of Bitcoin Prices on Gold Prices with a 1-week lag, as seen below:

The lag period of 1 week had no impact on the significance of the result or the predictive nature of the test. The estimate was not significant and estimated that Bitcoin does not influence the gold markets.

The final test is bidirectional of the previous regression: 1-week lagged GCD test of Gold Prices on Bitcoin Prices.

The lag period of 1 week had no impact on the significance of the results or the predictive nature of the test. The Pr(>F) value was too large to accept the results as significant and not causal.

Conclusion: The analysis failed to reject the null for the Granger Causality test. Under none of the GCD tests run did Bitcoin and gold have any significant relationship that estimated a casual predictive relationship. There was, however, a significant, but not casual, relationship between log(Bitcoin) and log(Gold) prices as well as the Mt.Gox hack dummy variables. Analyzing the graphs, descriptive statistics, GCD tests, and regressions show that the two assets are vastly different in nearly every aspect I measured. The prices, range, volatility, returns, trends, and correlations contradicted each other. The negative correlation between the two assets was very interesting and deserves further exploration using a quantile or interval-based approach to determine if they are more strongly correlated during short periods of time as a response to systemic risk or other factors. I initially expected the relationship to be positively correlated but weak, so the negative correlation was surprising. A primary argument used by Bitcoin enthusiast is that is acts similarly to gold, but this analysis would indicate that gold is almost a hedge against cryptocurrencies, rather than a correlated asset. The results from the Granger-causality tests were not surprising as I did not expect the Bitcoin market to be mature enough structurally to Granger-cause a much larger and more established asset class like gold. However, gold lacking a significant causal predictive relationship with Bitcoin prices was less expected as small movements in such a large market, nearly $7.7 trillion, could heavily influence price in a less liquid small-cap asset like Bitcoin. This likely indicates that institutional capital held in Gold does not consider Bitcoin as a viable part of their portfolio strategy. For Bitcoin investors, the lack of an established causal relationship would indicate that institutional capital has not entered the market at volume and is thus a largely untapped market could potentially drive up the price in the future. One of the primary reasons institutions haven’t entered into large cryptocurrency long positions is because the market is highly manipulated. This was also the cited reason for the Bitcoin ETF being denied by the SEC. Bitcoin is highly subject to insider trading, scams, pump and dump schemes, as well as highly unequal with wealth concentrated in 4.11% of digital wallets holding 96.53% of the total Bitcoin supply.[6] Wealth inequality in Bitcoin will pose a challenge to its future growth, regulation, and acceptance as a store of value or trusted asset. Based on the result of this analysis, Bitcoin behaves most similarly to an emerging speculative store of value asset that is prone to boom and bust cycles, volatility, and a lack of correlation with the largest asset classes. The primary Bitcoin valuation model used by crypto-economist is based on the number of digital wallets active, transaction count, and network effects present rather than as a response to macro-events such as the change in gold prices. Additionally, gold is a multi-trillion-dollar asset, whereas cryptocurrencies barely peaked $200 billion in 2017–2018. This means that bidirectional Granger causality was even more unlikely than gold Granger-causing Bitcoin. The tests were largely inconclusively as I failed to reject my null and the relationship between the two was uncorrelated and insignificant.

The Federal Reserve should consider adding Bitcoin to its reserve currency portfolio as a proper risk mitigation strategy and to maintain the dollar’s dominance in the long-run. If Bitcoin and other digital currencies continue to grow in market capitalization and usage, the dollar’s long-term position as the reserve currency could be challenged. However, before policymakers decide to include Bitcoin as a reserve currency, further analysis needs to be performed as the market matures. The copula-based approach to dependence and causality in the quantiles presented by Bouri et al. (2018) would be a more sophisticated approach to this asset pair. Additionally, further research should focus on the relationship between Bitcoin and other commodities such as silver, oil, and phosphates which would be relevant in determining how the market perceives and prices cryptocurrencies.


Balcilar, M., Bouri, E., Gupta, R., & Roubaud, D. (2017). Can volume predict Bitcoin returns and volatility? A quantiles-based approach. Economic Modelling, 64, 74–81.

Bouri, E., Gupta, R., Lau, C. K. M., Roubaud, D., & Wang, S. (2018). Bitcoin and Global Financial Stress: A Copula-Based Approach to Dependence and Causality in the Quantiles. Quarterly Review of Economics and Finance, 69, 297–307. Retrieved from b=ecn&AN=1725740&site=ehost-live

Bouri, E., Gupta, R., Lau, C. K. M., Roubaud, D., & Wang, S. (2017a-c). Bitcoin and Global Financial Stress: A Copula-Based Approach to Dependence and Causality in the Quantiles. Quarterly Review of Economics and Finance, 69, 297–307. Retrieved from bib0020

Desjardins, J. (2017, October 26). All the World’s Money and Markets in One Visualization. Retrieved from


Nakamoto, S. (n.d.). Bitcoin Whitepaper [Scholarly project]. In Retrieved from

Salant, S., & Henderson, D. (1978). Market Anticipations of Government Policies and the Price of Gold. Journal of Political Economy, 86(4), 627–648.

This Chart Reveals the Centralization of Bitcoin Wealth. (n.d.). Retrieved from

The Proof-of-Work Concept Daniel Krawisz. (n.d.). Retrieved from

Data Sources:

Cryptocurrency Market Capitalizations. (n.d.). Retrieved from

Market Price (USD). (n.d.). Retrieved from

The World Gold Council. Goldhub | The Definitive Source for Gold Data and Insight. (n.d.). Retrieved from

[1] Nakamoto, Satoshi. The Bitcoin Whitepaper. 2008.

[2] The Proof-of-Work Concept Daniel Krawisz. (n.d.).

[3] Desjardins, J. (2017, October 26). All the World’s Money and Markets in One Visualization.

[4] The World Gold Council. (2018)

[5] Note: Futures are usually commodity based-assets that are depended on future earnings, revenue or production.


Source link