The drivers behind the correlation between stocks and gold
Bachelor Thesis
Abstract
The aim of this thesis is to identify the drivers behind the correlation between returns of the stock market and returns of the gold market. Specifically, the correlation between US stocks from the S&P 500 and the closing price of gold from the London Bullion Market is calculated for each year between 1969 and 2015. The inflation rate, the foreign-exchange rate, the interest rate, the GDP growth rate, the financial crises and crashes variable and the speculative bubbles variable are six potential drivers that are examined. The findings of both the univariate model and the multivariate model do not imply that any of these drivers are a strong driving factor behind the
correlation between US stocks returns and gold returns.
.
Student: Laurèl Borggreve Student Number: 10644997
Supervisor: D. van Dijk
Bachelor Program: Economics and Business Specialization: Economics and Finance
2
Statement of originality
This document is written by Student Laurèl Borggreve who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Content
1. Introduction 4
2. Literature Review 6
2.1 Modern Portfolio Theory 6
2.1.1 Capital Asset Pricing Model 6
2.1.2 Diversification 7
2.1.3 Correlation 7
2.2 Characteristics of gold 8
2.2.1 Gold as a portfolio diversifier 8
2.2.2 Hedge 9 2.2.2.1. Inflation hedge 9 2.2.2.2. Currency hedge 9 2.2.3 Safe haven 10 2.2.3.1. Safe haven 10 2.2.3.2. Liquidity 10
2.3 Drivers behind correlation 12
2.3.1 Inflation rate 12
2.3.2 Foreign-exchange rate 13
2.3.3 Interest rate 14
2.3.4 GDP growth rate 14
2.3.5 Financial crises en crashes 15
2.3.6 Speculative Bubbles 15 3. Data 17 3.1 Data description 17 3.2 Data analysis 18 4. Methodology 21 4.1 Univariate model 21 4.2 Multivariate model 21 4.3 Hypotheses 22
4.4 Augmented Dickey-Fuller unit root test 22
5. Results 24
5.1 Univariate model 24
5.2 Multivariate model 26
5.3 Summary of results 27
6. Robustness check 28
6.1 Different smoothing weights 28
6.2 Different market 29
7. Limitations and further research 31
8. Conclusion 32
9. References 33
4
1
Introduction
Investors are always looking for desirable investments to add to their portfolio in order to increase their risk-adjusted returns. Especially nowadays, given the extreme turbulence of the economic climate such as the financial crisis of 2008, oil market crises, and more recently the Negative Interest Rate Policies, investors are more concerned with the value and properties of the assets they are holding in their portfolio. A recurring topic in the field of portfolio optimization is investing in safe assets such as gold. Historically, gold has been considered a hedge against inflation and a safe haven during uncertain political and economic times. While the role of gold has faded after the collapse of the Bretton Woods system of fixed exchange rates in 1973, its importance is now increasing again in the financial system. An explanation for this phenomenon is because of changes that have been made to the global banking sector after the financial crisis of 2008. Central banks introduced unconventional tools to stimulate the economy, such as Quantitative Easing (QE) and Zero Interest Rates Policies (ZIRP). More recently, between mid-2014 and end-2015, central banks in the euro area, Sweden, Switzerland, Denmark, and Japan have implemented an
unconventional monetary policy tool: Negative Interest Rate Policies (NIRP). By moving their policy rates below zero into negative territory, central banks require depositors to pay interest on a regularly basis to keep their money with the bank. Due to the low and even negative interest rates, the return on government bonds is
diminishing and holding these bonds is becoming less attractive whereas there is an increasing demand for gold from investors as well as central banks. The higher demand for gold is reflected in figures of Bloomberg, which show that the spot price of gold has increased from $35 an ounce in 1973 to a level around $1240 an ounce in 2016. Although there are some fluctuations, the price of gold has been increasing ever since the fall of the Bretton Woods System reaching its best performance of $1895 an ounce on 5 September 2011. In addition, consumer demand for gold reached 1290 tons in the first quarter of 2016 compared to 1070 tons in the first quarter of 2015, which is a year-on-year increase of 21% according to the World Gold Council.
The performance of gold is an extensively researched subject in financial
literature. Most of these researchers, such as Erb and Harvey (2013) and Baur (2013) have focused on the role of gold as a hedge against inflation and/or currency
fluctuations, but have also studied the effect of gold as a safe haven with regard to financial assets. Additionally, several studies, like Jaffe (1989), have examined the diversification benefits of gold in an investor’s portfolio. On the one hand, studies suggest that gold holds the characteristics of both a good hedge and safe haven and is profitable for portfolio diversification. Others oppose this statement and assert that gold is not an effective hedge and does not outperform other financial assets in terms of diversification benefits. The main purpose of this research will be to provide new empirical evidence of the correlations between gold and other assets’ returns. The correlation between gold returns and US stock returns will be examined and possible drivers behind this correlation will be investigated during the time period consisting of 47 years, namely from 01/01/1969 to 31/12/2015. The research question that will be answered is as follows:
What are the drivers behind the correlation between returns of stocks from the US stock market and returns of gold from the London Bullion Market during the time
5 Afterwards, one can conclude if gold should be included in an investor’s portfolio to obtain the best reward-to-one-unit-risk assuming the investor is risk averse. In addition, by means of the investigated drivers and their strength, one can argue whether it is always beneficial to hold gold in a portfolio, because it is independent of economy cycles or if it is only worth considering as a good portfolio diversifier, hedge or safe haven during certain market conditions, for example in times of high inflation.
The empirical investigation of this thesis could make a relevant contribution to the existing literature. Although the data for gold and US stocks have been used in multiple studies, the correlation between US stocks and gold has not been examined to my knowledge. Most studies investigate either the prices or the returns of gold and/or stocks instead of the correlation between these two assets. The findings of this research are especially interesting for investors.
This research is structured in the following manner. The second chapter will give an overview of the existing literature about portfolio optimization, the
characteristics of gold and the drivers behind gold. In the third chapter the data used for this research is described and analyzed. The fourth chapter presents the
methodology and states the hypotheses. In the fifth chapter the empirical results from the research will be provided and discussed. The sixth chapter is dedicated to
robustness checks of the prior results. The seventh chapter addresses limitations regarding this research and suggestions for further research. Finally, in the last chapter, a conclusion that contains the main point of this study is provided.
6
2
Literature review
In this chapter, three different topics concerning portfolio theory and gold are reviewed. First, the theory underlying gold as a diversification asset in an investor’s portfolio will be considered. In addition, the characteristics of gold, such as the role of gold as both a hedge and a safe haven, have been repeatedly examined in most studies and will be discussed. Moreover, the possible drivers behind the correlation between gold and US stocks will be addressed. The theory will be given through an extensive review of the existing academic literature about gold.
2.1 Modern Portfolio Theory
Markowitz (1952) introduced the concept of Modern Portfolio Theory (MPT), which is one of the most influential and important economic theories dealing with
investment and finance. MPT is an economic framework regarding how risk-averse investors can assemble a portfolio of assets in order to optimize expected return based on a given level of risk. In addition, the theory states that in order to compensate an investor who is willing to take more risk, one should receive a higher expected portfolio return. According to MPT, investors should not only look at one particular stock and its expected risk and return, but should invest in multiple stocks. Investing in one than more stock gives diversification benefits to investors, which is also known as ‘not putting all of your eggs in one basket’ (Bodie et al., 2014). Next, the three main concepts in MPT are discussed.
2.1.1 Capital Asset Pricing Model
The Capital Asset Pricing Model (CAPM) is an empirical model widely used to analyze the returns on risky assets in an investor’s portfolio. Sharpe (1964), Lintner (1965), and Mossin (1966) independently develop the model in their articles, based on the earlier work of Markowitz (1952), which builds on modern portfolio theory and diversification. The formula for the CAPM is described in equation (1)
(1) Rit = Rft + βi * (Rmt - Rft) + εt
Where:
- Rit is the return of the security i at time t,
- Rft is the return on the risk-free asset at time t,
- βi is the systematic risk of security i,
- Rmt is the return on the market portfolio at time t,
- εt is the residual at time t.
According to the CAPM, the total risk that investors are facing is comprised of systematic risk and idiosyncratic risk. Idiosyncratic risk refers to the risk associated with securities and can be mitigated through diversification. The two remaining risk measures are the portfolio’s total risk and its systematic risk. The total risk of the portfolio is measured by the variance or standard deviation of the portfolio while the systematic risk is captured by the beta in the model. The systematic risk is also known as market-related risk and refers to the risk common to all securities in the portfolio. This risk cannot be eliminated through diversification in contrast to idiosyncratic risk (Fama & French, 2004). The portfolio’s systematic risk is more relevant to a portfolio
7 manager, since the performance of a portfolio manager is mostly evaluated by
comparing the manager’s portfolio with a portfolio holding the market index (Chua et al., 1990).
As pointed out by Sharpe (1964), the portfolio’s total risk is a function of the correlation coefficient and the variance of each different investment in that portfolio. The correlation between assets is the key to reduce the total risk of a portfolio. As this paper examines the role of gold in an investor’s portfolio, the correlation between gold and other assets’ return will be calculated and evaluated. If the correlation coefficient is low enough, gold will decrease the total risk in a well diversified
common stock portfolio without diminishing its expected return. This phenomenon is called the ‘free lunch’, which has been first explained by Markowitz (1952).
In addition to the portfolio’s total risk, an investor is also interested in the systematic risk, which is the weighted average of the betas of each individual
investment (Chua et al., 1990). According to Chua et al. (1990) the key of using gold in order to adjust the systematic risk of the portfolio is the relationship between the gold’s beta and the common stock’s systematic risk. Jaffe (1989) describes the assertion of the CAPM that the beta of a single asset’s return with the return on a representative market index is the best measure of that asset’s risk. The lower the beta of the asset, the more the price movement of the asset is virtually unrelated to the price movement of the market, although it still can has some substantial volatility (which is measured by the standard deviation). Thus, as long as the beta for gold is less than 1.0, it is beneficial to add gold to a portfolio of common stocks to reduce the portfolio’s systematic risk.
2.1.2 Diversification
Markowitz (1952) started with the idea of diversification in an investor’s portfolio. In his article, he points out that an investor can reduce its portfolio risk by holding combinations of securities that are not perfectly positively correlated. When investors carefully select the set of assets in their portfolio, they can reduce their exposure to individual asset risk. As a result, they can have the same portfolio expected return along with a reduction in the riskiness of their portfolio. Markowitz stresses that investors need to be aware of choosing the right combination of assets and that the adequacy of diversification does not depend solely on the number of assets held in a portfolio.
2.1.3 Correlation
Correlation is a statistical measure that indicates the degree to which two or more variables fluctuate together. A positive correlation indicates the degree to which those variable increase of decrease in tandem; a negative correlation indicates the degree to which one variable increases while the other variable decreases. If the correlation between two variables is zero, they move completely independently from each other. Correlations between securities form the foundation of MPT and thus it is crucial to understand the dynamics behind correlations. Erb and Harvey (2013) mention that there is not any theory or connection about correlations and volatilities. Two asset classes can be perfectly correlated, but at the same time one asset may be four times as volatile as the other. Assets are influenced by several factors whose influence can vary over time. In addition, not only correlations between two variables, but also the volatility of assets’ returns itself changes. In order to choose the right set of assets, investors should investigate and analyze the correlations between assets not only at one point in time, but consistently over time.
8
2.2 Characteristics of gold
Over time gold has taken the property of being a safe haven and a good hedge against both inflation and currency changes. These properties have been studied extensively for the last twenty years. Baur and Lucey (2010) provide definitions for a diversifier, a hedge and a safe haven property in order to distinguish all three types. Baur and McDermott (2010) extend their work by distinguishing between a strong and a weak hedge/safe haven. Their distinction on top of the definitions of Baur and Lucey (2010) leads to the following definitions:
‘Diversifier: A diversifier is defined as an asset that is positively (but not perfectly correlated) with another asset or portfolio on average.’
‘Hedge: A strong (weak) hedge is defined as an asset that is negatively correlated (uncorrelated) with another asset or portfolio on average.’
‘Safe haven: A strong (weak) safe haven is defined as an asset that is negatively correlated (uncorrelated) with another asset or portfolio in times of market stress or turmoil.’
Baur and Lucey (2010) point out that the ‘on average’ described in the
definitions of the diversifier and the hedge means that the correlation property is only expected to hold on average and does not have the (specific) property of reducing losses in turmoil or extreme adverse market conditions.
2.2.1 Gold as a portfolio diversifier
Jaffe (1989) and Lawrence (2003) show that the inclusion of gold can be profitable in a well diversified portfolio, since its addition to a portfolio reduces the risk while it increases the average return. This characteristic of gold can be explained by the fact that gold is either uncorrelated or has a low correlation with most other type of assets. Baur (2013) and Hillier et al. (2006) obtain similar qualitative results, which
demonstrate that an inclusion of gold is enhancing an investor’s portfolio either by an increase in return for a fixed level of risk or a reduction in risk for a given level of return. Additionally, Ratner and Klein (2008) hypothesize that including gold in an international portfolio can enhance its performance due to the statistically low and/or negative correlations between US equities and stocks. They assert that the
diversification benefits are stronger for lower/more negative correlations. Their results enable them to indeed confirm their hypotheses, but they emphasize that investing in gold is dependent on the time period. As opposed to other studies such as Jaffe (1989), they find that over the long term investing in gold does not enhance the portfolio performance. Additionally, when analyzing the correlation between US stocks and gold stocks1 between 1970 and 1980, Chua et al. (1990) find that the correlation coefficient is increasing over time and claim that if it continues to rise the benefits of holding gold stocks in a portfolio will decrease, at least for the short term investments. However, their study highlights that for both the long and the short run, gold bullion is a meaningful investment for portfolio diversification as opposed to
1 Throughout this thesis, gold refers to gold bullion from the London Bullion Market. Gold stocks refer to gold-mining securities from the Toronto Stock Exchange (TSE). Gold stocks do not share gold’s convenience, consumption, and high liquidity values.
9 gold stocks. An overview of the characteristics of gold and the corresponding article is provided in Table 2.1.
It should be noted that Jaffe (1989), Erb and Harvey (2013), and Ratner and Klein (2008) argue that gold can be quite risky as an individual investment and should not be hold on a stand-alone basis. Gold is thus only beneficial when it is added to a portfolio with other real assets and a diversified set of commodities.
2.2.2 Hedge
Hedging refers to the techniques intended to offset particular sources of risk. A hedge can be taken as an investment position to reduce potential gains/losses that may be incurred. It can be constructed from many types of financial instruments, like forward contracts and options (Bodie et al., 2014). Gold has been considered a hedge against inflation and foreign-exchange rate fluctuations.
2.2.2.1 Inflation hedge
An investment is labelled an inflation hedge when it provides protection against the debased value of a currency. A key characteristic is that the inflation hedge maintains or increases its value over a specified period of time. For example, the price of the hedge should rise when inflation rises (Erb & Harvey, 2013). Erb and Harvey (2013) argue that gold may be an effective hedge again inflation expectations in the
extremely long-term (2000-year horizons), but it is certainly not an effective inflation hedge both in the short and long run. In contrast, according to Dempster and Artigas (2010) gold has proven to be an inflation hedge in times of inflation. The price of gold increased with an average in 14.9% in real terms during the eight years between 1974 and 2008 when the inflation in the U.S. was high (defined as CPI inflation exceeding 5%). In addition, gold was able to outperform other assets, such as stocks, bonds, and even other commodities. In their study, Dempster and Artigas investigate whether it is profitable for investors to hold gold in their portfolio as an adequate inflation
protector even though there is a low to moderate inflation environment. They find that gold can be both a long-term strategic asset and a tactical inflation hedge. Besides the fact that gold outperforms mainstream financial assets in an inflationary economy, investors should consider gold as a strategic asset in their portfolio as well. Due to its strength as a portfolio diversifier, gold not only enhances an investor’s returns per unit risk, but also consistently delivers a lower average standard deviation, even in a low to medium inflation environment.
2.2.2.2 Currency hedge
Both Erb and Harvey (2013) and Capie et al. (2005) describe two arguments for gold as a currency hedge. One of the arguments holds that since gold is a homogenous asset, it can be traded with ease in a continuously open market according to Capie et al. (2005). Unlike currencies that are produced by authorities, gold cannot be
produced meaning that those authorities, who are able to manage the supply of money and as a consequence their value, cannot practice their influence on the value of gold. Ciner et al. (2013) describe that gold holds its value in relation the debased currency making itself an excellent protector against currency fluctuations. The results in the study from Capie et al. (2005) show a negative inelastic relationship between the sterling-dollar and yen-dollar exchange rates and gold. However, their study indicates that only during unpredictable political events and attitudes gold has served as a hedge against currency fluctuations. This conclusion is in line with the study of Beckmann et al. (2015) and the report of Shih et al. (2015) as they used dynamic
10 correlation coefficients between the dollar/pound and gold to conclude that gold is the best effective hedge against the US dollar. However, Erb and Harvey’s (2013)
research measures, the standard deviation between gold and foreign currencies and the R², found contradictory results indicating that gold is an unreliable currency hedge.
2.2.3 Safe haven
In his article, Kaletsky (2016) describes the increasing volatility in financial markets due to falling oil prices, slower growth in China, the threat of bankruptcies in junk bonds, and geopolitical instability. In times of volatile markets, investors seek protection from safe havens to limit their exposure to potential losses. As a result, secure assets, such as government bonds and gold, are facing a spurred demand by investors and companies. Besides its potentially stabilizing role, gold is also very liquid, which can be another reason for the increasing demand during market downturns.
2.2.3.1 Safe haven
The findings of Baur (2013) show that gold acts as a safe haven and is a good asset to insure one’s portfolio even when the price of gold decreases below its historical record. In their study, Baur and Lucey (2010) check whether gold act as a safe haven for US, UK, and German stocks or bonds. They find that gold does not serve as a safe haven for bonds in all three markets, but does provide a safe haven function for stocks in all markets although there is stronger evidence in the UK and Germany. In
addition, gold only serves as a safe haven on the short term. Furthermore, when Baur and McDermott (2010) apply a similar approach to test whether gold is a safe haven against stocks of major developing and emerging markets, their results indicate that there is a difference between developed and emerging economies. Gold does act as a safe haven for both the US and European countries, but the safe haven property of gold does not live for emerging countries as well as Japan, Australia, and Canada. An investigation by Beckmann et al. (2015) show that gold indeed exhibits the safe haven function, but does this only under certain economic market conditions. They base their study on the empirical testing procedure used by both Baur and Lucey (2010) and Baur and McDermott (2010). Although gold is a safe haven in times of
hyperinflation, Erb and Harvey (2013) again conclude that during other volatile market periods gold is not a safe haven. Bredin et al. (2015) show that gold is not a safe haven for all types of crises and suggest that further research needs to be done to better examine the safe haven property of gold.
2.2.3.2 Liquidity
Liquidity is the ease, speed, and cheapness with which investors can realize the cash value of an investment (Bodie et al., 2014). Hillier et al. (2006) and Hoang et al. (2015) point out that there are a large number of investors who actively buy and sell gold anywhere in the world making the market for gold exceptionally liquid and universal. The average daily number of transfers was 4206 transfers in 2015 compared to an average of 37 transfers in 1997 according to figures by the London Bullion Market Association (LBMA). Additionally, the total annual world gold production is nowadays cleared every 2.5 days by the LBMA. Moreover, together with the market for US treasuries, the gold market boosts the tightest bid-ask
11 spreads2. These figures show that the gold market has changed over time from
historically
being illiquid to extraordinarily liquid today.
Cornett et al. (2011) claim that the rising demand to hold liquid assets during times of market turmoil stems from the financial crisis of 2008. During this crisis, many banks rapidly became insolvent due to a lack of sufficient cushion of liquid assets, which were necessary to meet the bank’s obligations. These obligations included unpaid loans consisting of billions of dollars and the loss of depositors who suddenly withdrew their money because of the subprime mortgage crisis. If an investor holds enough liquid assets in its portfolio, it has more insurance to meet its financial obligations without having to liquidate fixed assets or to take on new debt. In this manner, one can resolve the problem before it turns into another financial disaster.
Table 2.1: Overview of the characteristics of gold and the corresponding article
(1) = dependent on the specific economic environment under observation (i.e. unpredictable political attitudes and events, abnormal stock market volatility, extreme levels of global uncertainty, and certain market conditions)
(2) = dependent on the time span (i.e. the sample time window and short/long term) (3) = dependent on the type of market (i.e. emerging or developed market/country) (4) = dependent on the type of crisis
Characteristic Article Conclusion: Yes / No / Dependent
Portfolio diversifier Sherman (1982) Jaffe (1989) Chua et al. (1990) Davidson et al. (2003) Lawrence (2003) Hillier et al. (2006)
Dempster and Artigas (2010) Baur (2013)
Beckmann et al. (2015) Erb and Harvey (2013) Ratner and Klein (2008)
Yes Yes
Yes: Gold bullion; No: Gold stocks Yes Yes Yes Yes Yes Yes No Dependent (1) Hedge Inflation hedge Ghosh et al. (2002)
Dempster and Artigas (2010) Bialkowski et al. (2015) Jaffe (1989)
Blose (2010)
Baur and McDermott (2010)
Pasatusarayut and Chintrakarn (2012) Baur (2013)
Erb and Harvey (2013) Batten et al. (2014) Beckmann et al. (2015) Bredin et al. (2015) Yes Yes Yes No No Dependent (1), (2), (3) Dependent (3) Dependent (2) Dependent (2) Dependent (2) Dependent (1) Dependent (2) Currency hedge Ciner et al. (2013) Erb and Harvey (2013) Capie et al. (2005) Hillier et al. (2006) Beckmann et al. (2015) Shih et al. (2015) Yes No Dependent (1) Dependent (1) Dependent (1) Dependent (1)
12 Safe haven Safe haven Baur (2013) Ciner et al. (2013) Bialkowski et al. (2015) Shih et al. (2015) Erb and Harvey (2013) Baur and Lucey (2010) Baur and McDermott (2010)
Pasutasarayut and Chintrakarn (2012) Beckmann et al. (2015) Bredin et al. (2015) Yes Yes Yes Yes No Dependent (2), (3) Dependent (1), (3), (4) Dependent (3) Dependent (1), (3) Dependent (2), (4) Liquid Jaffe (1989) Hillier et al. (2006)
Dempster and Artigas (2010) Hoang et al. (2015)
Yes Yes Yes Yes
2.3 Drivers behind correlation
The main purpose of this paper is to identify the driving factors behind the
relationship between stock and gold returns. Several studies show that most of the time, the correlation between these two variables is negative: when the market for stocks increases, the market for gold decreases and vice versa. However, while this negative correlation holds most of the time, the relationship between stocks and gold returns changes over time leading to periods in which the markets also tend to move in tandem. This feature depends on external market conditions and important
macroeconomic variables.
In his study, Baur (2013) investigates seven fundamental drivers of the price of gold distinguishing between ‘traditional’ and ‘new’ drivers, which have only occurred recently. His results show that the gold price is mainly influenced by the ‘traditional’ drivers like inflation, interest rates and currency changes and ‘new’ drivers such as central bank demand. Moreover, he finds that the effectiveness of the drivers significantly changes during different time periods. Yet, Solt and Swanson (1981) assert that the development of gold is not affected by influences that generally affect the market due to gold’s own characteristics. Lawrence (2003) concludes that while fixed income and equity prices are influenced by macroeconomic shocks, gold returns tend to move independent from macroeconomic factors.
Before the correlation between US stocks and gold returns is regressed on possible drivers in an empirical investigation, this part will first provide an evaluation of possible fundamental drivers. This research limits itself to six potential driving factors behind this correlation, which are outlined below.
2.3.1 Inflation rate
Inflation is defined as a sustained increase in the general or aggregate level of prices for a basket of goods and services. It is measured as an annual percentage increase and results in a fall of the purchasing power of a currency. This means that when inflation rises, one can buy a smaller percentage of goods and services for every one unit of currency one owns.
Although low to moderate inflation, preferably around 2% a year, can be favourable for an economy since it stimulates the purchase of goods and services and thus increases job employment, too high inflation is less desirable. The negative impact of rising inflation are among others lower consumer confidence in the
13 On the other hand, a prolonged period of low inflation or deflation can be harmful to an economy as well. The falling prices lead to a discouragement of spending
behavior, because consumers and companies expect cheaper prices in the future. Deflation also results in an increase in the real value of debt burdens which reduces the disposable income of consumers and companies. Numerous studies have looked at the impact of inflation on stock returns, but conflicting results are found. Most studies argue that expected inflation3 can have either a negative or positive impact on stocks, which is also dependent on the ability to hedge against inflation. In contrary, when looking at unexpected inflation4, most studies conclude that during economic contractions there is a strong, positive correlation to stock returns. This correlation can come from greater volatility of stock movements or the fact that unexpected inflation contains new information about future prices.
As described earlier, several studies have shown that gold can be labelled as an inflation hedge, because its price tends to increase when inflation rises. The underlying economic mechanism is as follows: if the price index of goods and
services increases by 3%, gold becomes relatively cheap compared to this index. This leads to an increase in the demand for gold until the change in the price of gold is equal to the change in the price index (Baur, 2013). Nonetheless, the linkage between gold and inflation is not a universal subject in the literature. Blose (2010) contradicts Baur’s statement by concluding that gold spot prices are not influenced by changes in expectations regarding future inflation. Batten et al. (2014) find that there is
significant time variation in the relationship between gold and inflation. If gold is indeed a good inflation hedge, the correlation between the returns of gold and the returns of US stocks should be negative or zero.
2.3.2 Foreign-exchange rate
The foreign-exchange rate between two currencies is the rate at which one currency will be exchanged for another currency. An exchange rate consists of two components: the domestic currency and the foreign currency. Most exchange rates use the US dollar as the base currency and other currencies as the counter currency.
Fluctuations in the currency rate on the foreign exchange market can have a significant influence on the performance of the investment portfolio. Looking through the position of an US-based investor, if one believes the USD is in temporal decline against a counter currency, the returns on the US stock will slump if these stocks are mainly derived from companies that have most of their business located in the US. It is then better to invest in strong overseas markets, since these foreign markets will generate higher returns due to the appreciation of the counter currency. An
investigation by Kim (2003) confirms this statement by showing a negative
relationship between the S&P 500 stock price and the real exchange rate in the long-run equilibrium.
Another possibility to benefit from fluctuations in the foreign-exchange rate is to hedge currency risk and thereby mitigating the exposure to currency movements. As mentioned above, gold is an excellent protector against currency fluctuations, because it maintains its value in relation to the debased currency. The expected correlation is described from the perspective of an US-based investor and based on the following mechanism: A deprecation of the dollar against another currency is translated into a higher price of gold denoted in US dollars and a decline in the price of gold denoted
3
Expected inflation: the public’s expectations for inflation
14 in the weighted currency. The correlation between returns on various trade-weighted currencies and the returns of gold in that currency is then negative. Since the foreign exchange rate exhibits a negative correlation with both the returns of US equities and the returns of gold, a co-movement between gold and US stocks is expected implying a positive correlation. This correlation behaviour is also seen by figures in an article of Artigas et al. (2013).
2.3.3 Interest rate
According to Blose (2010), the opportunity costs of holding an asset are determined by the level of interest rates. The cost of carry increases when the interest rate is higher. Since there are no interest or dividend payments involved in holding gold, the opportunity costs can become very high. In turn, these higher costs will lower the price of gold. Generally, interest rates tend to move in tandem with the inflation rate meaning that higher interest rates imply higher inflation rates. The difference between the inflation rate and the nominal interest rate is defined by the real interest rate. The real interest rate will be negative when the nominal interest rate is lower than the inflation rate.
Generally, an increase in the interest rate leads to a diminishing demand for holding stock due to a drop in stock prices. Since the cost of borrowing money is more expensive, businesses and consumers will lower their spending pattern and start saving money. Companies will experience an increase in interest expenses, which leads to a reduction in its cash flows. Additionally, consumers lower their
expectations in the growth of the economy. This occurrence can drive the stocks of the company down in price making the stock market a slightly less desirable place to invest.
There are opposing statements about the effect of interest rates on the gold market. According to Baur (2013), real interest rates are an important driver of the price of gold and these two variables will exhibit an inverse relationship with each other. In contrast, Sherman (1982) asserts that the correlation between interest rates and gold is zero indicating that the movement of gold does not depend on interest rates. Besides Sherman (1982), Tully and Lucey (2007) confirm this statement as their study reports an insignificant statistical relationship between interest rates and gold. Finally, Lawrence (2003) concludes that changes in macroeconomic variables, such as inflation rate, GDP and interest rates, do not have a statistical influence on the returns of gold. However, if high interest rates and low/negative real interest rates coincide with a higher price of gold and their impact on the stock price is negative, a negative correlation would be expected.
2.3.4 GDP growth rate
The Gross Domestic Product (GDP) is the measure of the total amount of final goods and services produced in a period. It is commonly used an indicator of the economic health in a country.
The direction of the stock market is significantly influenced by a change in GDP. For example, the earnings and growth reports from businesses will be higher in a healthy economy. These higher GDP measurements will positively impact the stock prices leading to a positive outcome for investors and their holdings in the stock. On the contrary, an economy that is suffering from downturns, such as a high
unemployment, will experience a fall in GDP resulting in less favourable stock market indexes.
15 It is unclear whether the relationship between GDP and the price of gold is positive or negative and even significant. Findings of a study conducted by Ghosh et al. (2002) and Lawrence (2003) show that changes in the GDP have an insignificant effect on the demand for gold. Yet, the relationship is often expected to be negative due to two channels: the income-consumption channel and the safe-haven channel. The reasoning for the latter is as follows: when the economy is in a healthy condition, the demand for safe haven assets is low. Hence, there should be a negative
relationship between the price of gold and GDP growth. The income-consumption channel is more controversial: on the one hand, when the economy grows fast, consuming and spending behaviour is boosted and so is the demand for gold. This is due to investors who are investing in multiple assets. On the other hand, because the economy is in a good state, investors tend to invest in more risky assets and less in secure assets, like gold and T-bills. If the latter applies, a negative correlation between stocks and gold would be expected. Due to the portfolio diversifier argument,
investors could also tend to hold gold and benefit from the diversification benefits leading to a more positive correlation.
2.3.5 Financial crises and crashes
The National Bureau of Economic Research (NBER) is the official institution that reveals the beginning and ending of a recession in the US. The NBER defines a recession as follows:
‘Recession: a significant decline in economic activity spread across the economy, lasting more than a few more months, normally visible in real GDP, real income, employment, industrial production, and wholesale-retail sales.’
As has become clear through the literature review, a lot of studies found that the ability of gold to serve as a portfolio diversifier, a hedge or safe haven depends heavily on economic market conditions. For example, the time period of gold to exhibit its function as a portfolio diversifier can be clearly distinguished from the period in which it takes its role as a safe haven. Mostly, before the occurrence of a crash or crisis, gold is often used as an asset for portfolio diversification, whereas its demand to act as a safe haven occurs during or shortly after a crash or crisis
(Beckmann et al., 2015). In order to indicate changing correlations between gold and US stocks as a consequence of market turbulence, several crashes and crises are identified during the time period of this research, such as the second oil crisis, the dot-com bubble crash, and the subprime mortgage crisis. Typically, investors are more likely to add safe havens in their portfolio during times of crisis leading to a negative or zero correlation between gold and US stocks, although Bredin et al. (2015) and Ciner et al. (2013) have found contrary results for some crises. Bredin et al. (2015) even found a positive correlation during the 1980s recession.
2.3.6 Speculative Bubbles
Besides financial crises, speculative bubbles could also be a driving variable behind the US stocks and gold correlation. A speculative bubble occurs when prices for assets, especially stocks, strongly deviate from their actual, intrinsic value. The explosive exponential growth in prices continuous until the prices have risen so far that there is suddenly a sharp decline, which is known as a bubble crash/burst. The dot-com bubble between 1995 and 2001 and the US housing bubble between 2002
16 and 2006 are two major bubbles that have occurred in this thesis’ time span. The crashes caused by these bubbles are taken together with the financial crises.
The literature about speculative bubbles and its impact on correlation behavior is scarce. The rapid rise in equity markets during speculative bubbles causes the price of US stocks to reach high levels. Generally, when the price of stocks increases, the price of gold decreases leading to a negative correlation. On the other hand, while the price of gold will probably not increase as much as the price of equity, a small
increase in gold price could be expected considering the ‘good’ state of the economy and the diversification benefits linked to gold. This leads to a low or even slightly increasing co-movement between the US stock and gold market. Another argument for a low correlation is that investors become aware of the bubble and set their expectations for a crash afterwards. Then they could decide to already hold gold in their portfolio and use its safe haven characteristics if necessary leading to a higher demand for gold and consequently an increase in its price.
17
3
Data
In this section, the obtained data will be described and an analysis of the data will be presented afterwards.
3.1 Data description
The data for gold and US stocks are gathered from Thomson Reuters Datastream. One stock market has been used for the US stocks price indices and one price of gold has been used. A wide time span has been chosen in order to obtain reliable results. The time period that is studied contains 47 years and ranges from 01/01/1969 to
31/12/2015. The year 1969 has been taken as base year and the reason for this is that the data for gold were only available from march 1968 onwards.
The Standard and Poor’s 500 (S&P 500) Composite Stock Price Index is based on 500 large companies that have leading large capitalization common stocks listed on the New York Stock Exchange (NYSE), the Nasdaq Stock Market and the American Stock Exchange (AMEX). The S&P 500 is selected as a proxy for the US stock market. The data for the US stocks are taken on a daily basis.
The price of gold is taken from the London Bullion Market Association
(LBMA), which is the world’s foremost precious metal market. In addition, London is the largest global centre for over-the-counter transactions concerning the trading of gold and silver on a 2-hour basis. The closing price of the LBMA is used, which is expressed in US dollar per Troy Ounce. The data for the price of gold are taken on a daily basis too.
The data for the relevant drivers are taken from Thomson Reuters Datastream as well, but are gathered on a yearly basis. Expect for the foreign-exchange rate index, which has only been floating against major currencies after the fall of the Bretton Woods Fixed Exchange System in 1973, the base year is set at 1969. In order to make this research more representative, multiple indices have been used for one driver. The indices used for each driver are listed in Table 3.1.
18
Table 3.1: Data description for each driver
Driver Variable name Description
Inflation rate uscpi
worldcpi
US Consumer Price Index, All Urban World Consumer Price Index
Foreign-exchange rate fx Trade Weighted Value of US Dollar against Major Currencies
Interest rate ustbills
ustbills3month
US Treasury Bills, Government Securities US 3 month Treasury Bills Rate
GDP growth rate nomgdp
realgdp
US Gross Domestic Product Overall, Current prices (Nominal GDP5) US Gross Domestic Product Overall, Constant prices (Real GDP6)
Crises and Crashes Dummy No crisis = 0;
Crisis = 1
crisis 1973-1974: First Oil Crisis and Stock Market Crash
1979-1980: Second Oil Crisis
1987: Stock Market Crash (Black Monday) 2001-2003: Dot-com Bubble Crash
2001: Telecoms Crash 2001: 9/11 Attack
2007-2009: US Housing Bubble Crash 2008-2009: Subprime Mortgage Crisis 2010-2012: European Sovereign Debt Crisis Speculative Bubble Dummy
No Bubble = 0; Bubble = 1
bubble 1996-2000: Dot-com Bubble 2002-2006: US Housing Bubble
3.2 Data analysis
Instead of investigating the correlations between the prices of stocks and gold, this paper examines the correlation coefficients between the returns of stocks and the returns of gold. Looking at the descriptive statistics over the sample period reported in Table 3.2, the returns of the gold price outperform the stock market returns. However, the stock market displays lower variation in daily returns compared to gold returns. The higher returns for gold are offset by its higher risk leading to lower, but almost equivalent to those of stocks, Sharpe ratios. Investments in gold are thus slightly riskier compared to investment in stocks based on these statistics. On the other hand, while the maximum value for gold is higher than the maximum value for stocks, this finding does not hold for the minimum values of gold compared to stocks. According to Baur and Lucey (2010), this feature could be linked to the safe haven characteristic of gold and could make gold relatively less risky than was indicated by the standard deviation earlier.
The Jarque-Bera statistic, which is based on the combined measures of the skewness and kurtosis, is used to test if the variables are normally distributed. The probability value is zero for all series reported, which indicates that that the null hypothesis of normality is rejected. Likewise, the distributions of the returns on the US stock market are skewed to the left whereas the gold returns achieve positively skewed data. Under the hypothesis that returns are normally distributed, the
coefficient of skewness follows a asymptotically distribution when N(0, 6/T), where T is sample size. For the two variables, the hypothesis that daily returns on the US stock
5
Nominal GDP: market value of goods and services produced in an economy, unadjusted for inflation. 6 Real GDP: nominal GDP, adjusted for inflation reflecting changes in real output.
19 and gold price indices are normally distributed is clearly rejected by the statistics. These findings are in line with earlier studies, which found similar deviations from normality (Tulley & Lucey, 2007). However, while the t-test is invalid for small samples that are not-normally distributed, it is valid for samples of a larger size with a non-normal distribution. Since this sample consists of 12262 observations, the t-test will be valid.
Table 3.2: Descriptive Statistics of Variables for Daily Returns, 1969 – 2015 Note: Sharpe Ratio (SR) is denoted as mean divided by standard deviation
Mean Median Max Min Std. Dev.
S&P 500 0.0003 0.0001 0.1158 -0.2047 0.0104
Gold 0.0004 0.0000 0.2035 -0.1688 0.0127
Skew Kurt JB N SR
S&P 500 -0.6261 23.2408 0.0000 12262 28.53%
Gold 0.6769 30.9768 0.0000 12262 27.11%
The research continues with examining the correlation coefficients between returns of the US stock and gold markets. The correlations for the daily returns between US stocks and gold over the entire period are presented in Table 3.3. Inspecting the table, it can be seen that the correlations coefficient is negative, small in magnitude, and significant at the 5% significance level. The negative correlation highlights the role gold may play as both a hedging and safe haven asset in the nature of portfolio diversification, thereby making itself a considerable asset for inclusion in a well diversified portfolio.
Table 3.3: Correlations for Daily Returns, 1969 – 2015
Note: *, ** and *** depict statistical significance at the 10%, 5%, and 1% levels respectively
S&P 500 Closing
S&P 500 1
Gold -0.0225** 1
Subsequently, the correlation is calculated for sub-samples of one year using daily returns7. The data reveal that the correlation pattern of this series has been contaminated with a significant amount of noise. In order to cope with this noise, the data is smoothed using the weighted moving average technique. This technique allows one to assign different weights to the data at different points in time. It has been chosen to assign the greatest weight to the most recent term in the time series and less weight to the observations before and after the most recent data. Specifically, the new correlations are calculated by the formula presented in equation (2). The reason for choosing the weights in equation (2) is that these weights resulted in the best
smoothed correlation series after experimenting with several sets of weighting factors. In the robustness checks chapter, these various combinations will be discussed.
(2) ρtsmooth = 0.05*ρt-3 + 0.10*ρt-2 + 0.20*ρt-1 + 0.3*ρt + 0.20*ρt+1 + 0.10*ρt+2 + 0.05*ρt+3
7 The correlation between the returns of US stocks and gold has been calculated for monthly data as well. However, the findings for daily data are stronger compared to the results of the monthly data. This result is also found by previous studies, such as Baur and McDermott (2010).
20 The smoothed correlation series is illustrated in Figure 3.1. As can be seen in the figure, the correlation exhibits a varying pattern from positive to negative
correlations, depending on the time period. In general, under regular market
conditions a positive correlation is evident, but the correlation coefficients turn very negative during market downturns as described in Table 3.1. For example, a negative correlation is identified between the first and second oil crisis, the stock market crash, the 9/11 attack in 2001, and the subprime mortgage crisis whereas the correlation is upward sloping during the speculative dot-com bubble. The US housing bubble is also characterized by an increasing correlation, which indicates a co-movement between the gold and stock markets during speculative bubbles. There are two exceptions of which one is the period from the late 1970s to 1983, which illustrates an increasing correlation while the economy was in recession. Another increasing correlation in times of crisis is seen during the European sovereign debt crisis. Furthermore, the changing correlation can be seen as a justification for the distinction of gold being a hedge all the time into a safe haven during specific market conditions and a hedge on average.
Figure 3.1: Smoothed correlation for each year based on Daily Returns of the S&P 500 and the Closing Price of Gold, 1969 - 2015
21
4
Methodology
This chapter provides a general outline of the research conducted in order to identify the drivers behind the correlation between US stocks and gold returns. First, the univariate and multivariate regression models are introduced. Then, the hypotheses with regard to the correlation between US stocks and gold returns are presented. Finally, a unit-root test is applied to test for stationarity.
4.1
Univariate model
Each year’s correlation is regressed on one of the possible drivers in a simple linear regression model. The influence of the drivers is tested for each driver separately in order to identify the isolated performance of each driver. The model for each
regression is presented in equation (3) and contains a dependent and an independent variable. When the driver is included as a dummy variable, the model applied is the model shown in equation (4). For both models, the dependent variable is the
smoothed correlation between the returns of the US stock and gold markets. The independent variable is one of the variables listed in Table 3.1.
(3) ρt = α + βi * Xt+ εt
When a dummy variable is included, the model changes to equation (4). (4) ρt = α + βi * Dt+ εt
Where:
- ρt is the smoothed correlation between US S&P 500 stock returns and returns derived from the closing price of gold from the LBMA at time t,
- α is a measure of performance on a risk-adjusted basis, - βi is the coefficient on the driver i,
- X is the potential driver at time t, - Dt is a dummy variable at time t, - εt is the residual at time t.
4.2
Multivariate model
The multivariate model estimates the importance of all drivers jointly and is presented in equation (5). This model reveals the relative importance of the variables and reports a bias if a specific driver is omitted. The variables that are used in the multivariate analysis are presented in equation (5).
(5) ρt = α + β1*uscpi + β2*fx + β3*ustbills3month + β4*nomgdp + β5*crisis + β6*bubble + εt
Where:
- ρt is the smoothed correlation between US S&P 500 stock returns and returns derived from the closing price of gold from the LBMA at time t,
- α is a measure of performance on a risk-adjusted basis, - βi is the coefficient on the driver i,
22 The multivariate model begins with all the drivers. After each regression, an insignificant variable with the lowest t-statistics is removed. This process continues until the model only consists of drivers that are significantly different from zero. The only variable that is not omitted but rather is held constant in each regression is the crisis variable, since this variable indicates a specific market condition over time.
4.3
Hypotheses
Given the univariate and multivariate models described in the previous section, the hypotheses regarding the influence of the drivers on the correlation between US stocks and gold are formulated in statement (1). Taking into consideration the existing literature, there are a lot of various, contradictory results found regarding the
significance of each driver. Most importantly, the significant influence of the driver depends for a large part on the market conditions and the time period examined. (1) H0: the driver does not have a significant influence on the correlation between
returns from the US stock market and returns from the gold market.
H1: the driver has a significant influence on the correlation between returns from the US stock market and returns from the gold market.
Table 4.1 provides an explanation regarding the statements hypothesized for each driver and indicates why a rejection of H0 is expected. Additionally, the table presents a short description of the literature.
Table 4.1: Overview of hypotheses for each driver
Driver Literature Explanation
Inflation rate Negative Since gold is considered to be a hedge against inflation, it is expected that the inflation rate is a strong driver behind the correlation with a negative direction.
Foreign-exchange rate
Positive Gold is expected to exhibit hedging properties concerning currency changes leading the foreign-exchange rate to be a strong influential variable with a positive direction. Interest rate Ambiguous Although the literature about interest rates is vague, the
expectation is that this variable is a significant driver, but the direction is unclear.
GDP growth rate
Ambiguous Although the literature about GDP growth rate is vague, it is expected that this variable is a significant driver behind the correlation, but the direction is unclear.
Financial crises and crashes
Negative Since most studies argue that gold is a safe haven, this variable is expected to have a strong influence on the correlation with a negative direction.
Speculative Bubbles
Ambiguous Although the literature about speculative bubbles is vague, the expectation is that this variable is a significant driver, but the direction is unclear.
4.4
Augmented Dickey-Fuller unit-root test
An econometric requirement involved with time series data is that the data should be stationary in order to provide statistical inferences (Stock & Watson, 2011). There are several ways to identify non-stationary data (Levin et al., 2002). In this thesis, the
23 Augmented Dickey Fuller unit-root test (ADF) is applied in order to test whether the data for the drivers and the correlation series are stationary or not.
The results of the ADF tests for each driver show that the inflation rate, the foreign-exchange rate, the interest rate, and the GDP rate have a p-value greater than 0.01, which is the significant level chosen. The null hypothesis clearly cannot be rejected meaning that these variables are non-stationary. Also, the graphs of the data show that the series are not mean reverting, but exhibit a trend along with the time series. The ADF tests that have been performed together with the graphs can be consulted in the Appendix. Since it is concluded that four drivers have a unit root, an appropriate transformation of the variable series is required. This is done by taking the change of the annual percentage change (growth) in the variable’s rate. After the ADF tests have been performed again, the results show that the data series for the variables have a p-value of 0.0000, so that the H0 can be rejected and the data are stationary. This can also be seen in the graphs that are displayed in the Appendix.
Likewise, the ADF test for the smoothed correlations indicates that the series is non-stationary. Nevertheless, this thesis assumes that the smoothed correlation series is stationary based on the graph shown in Figure 3.1. It seems obvious by inspecting the graph that while the series turn both positive and negative, the correlation coefficients are mean reverting.
24
5
Results
This chapter will first analyze the results obtained for the univariate model and then the findings from the multivariate model. The chapter ends with a short summary.
5.1
Univariate model
In this section the result for each driver is discussed individually. The output of the regressions is presented in Table 5.1. From this table it can be seen that all the coefficients are not significantly different from zero and therefore the H0 cannot be rejected. The correlation coefficients are also small in magnitude, which could be due to the time-variation in the relationship between US stocks and gold. A sample of shorter periods leads to larger coefficients in absolute terms as pointed out by Baur and McDermott (2010).
As shown in Table 5.1, the beta for the US inflation rate is -0.826 and for the world inflation rate 0.010, but both are not significantly different from zero. These findings indicate that the correlation is zero meaning that the inflation rate is a weak hedge according to the definitions of a ‘hedge’ and a ‘weak/strong hedge’ by Baur and Lucey (2010) and Baur and McDermott (2010) respectively. Although this
finding is in contrast to the findings of Blose (2010) and Jaffe (1989), it is in line with at least four other studies that confirm the role of gold as an inflation hedge. It should be noted that other studies, such as Erb and Harvey (2013) and Beckmann et al. (2015), argue that gold’s ability to be a strong or weak inflation hedge is dependent on other factors as well such as the time window and the specific economic environment examined. This is a possible explanation for the rejection of H1. In addition, from 1980 to 1983 there were two recessions, one more severe than the other, which were associated with very high inflation. In Figure 3.1 a positive correlation is identified during this inflationary period suggesting that gold is not always a good inflation hedge. This result is also found by Bredin et al. (2015).
The beta for the trade-weighted value of the US dollar against major other currencies is 0.086. Again, this outcome indicates that gold is a hedge against currency fluctuations, which is line with other researchers among others Ciner et al. (2013). However, since these betas are not significant different from zero the positive direction does not imply that the foreign-exchange rate is a strong driver behind the correlation. The fact that gold is a weak instead of a strong hedge could be dependent on the economic market conditions investigated, which is also suggested by Capie et al. (2005) and Hillier et al. (2006). Furthermore, gold is considered to be a hedge on average as pointed out through the definitions formulated by Baur and McDermott (2010). This is also a cause that could explain the correlation of zero.
The beta coefficients for the US T-bills and the US 3 month T-bills variables are -0.005 and 0.001 respectively. Again, the H0 cannot be rejected meaning that the interest rate practices no significant influence on the correlation. However, while a significant influence on the relationship between gold and US stocks was expected, this finding is not totally contrary to the literature. While interest rates probably have an inverse relationship with US stocks, the relationship between interest rates and gold remains a vague subject. Both Sherman (1982), Lawrence (2003) and Tully and Lucey (2007) found an insignificant statistical relationship between interest rates and gold. Although the market for US stocks exhibits an inverse relationship with interest rates, the fact that gold tends to move independently from interest rates could be
25 stronger than the relationship between US stocks and interest rates leading to an insignificant influence of interest rates on the correlation between returns of US stocks and returns of gold.
The GDP growth betas are -0.061 for nominal GDP and 0.207 for real GDP. While the relationship between the GDP growth rate and US stocks is positive, the relationship between US stocks and gold can be both positive and negative leading to an average of zero. This means that both the income-consumption channel and the safe-haven channel could hold for the correlation. Hence, the correlation of zero indicates that the income-consumption channel is probably stronger, which could mean that people believe in the diversifying properties of gold. It can be concluded that the GDP growth rate is not an influential driver behind this relationship, which is in line with the findings of Ghosh’s study in 2002 and the study of Lawrence (2003).
The beta for the crisis dummy variable is 0.016, but not significantly different from zero. This finding claims that the safe haven property of gold holds. However, this outcome suggests that gold is a weak safe haven since the beta is not (negatively) significant at the 1%, 5%, or 10% significance level. The fact that the results do not show that gold is a strong safe haven can be due to the type of crisis examined, the time window of the sample used, or the volatility of the market, which is also implied by other studies such as Beckmann et al. (2015), Erb and Harvey (2013) and Bredin et al. (2015). For example, an US-based investor probably mainly focuses on the US market and is less interested in market disruptions of other continents. The European sovereign debt crisis could then be less of a concern to an investor, so that an US-based investor does not have the need to immediately seek a safe haven asset for its portfolio. Apart from this exception, it is clear that a negative correlation is
maintained during almost all of the crises and crashes listed in Table 3.2 by inspecting the graph of Figure 3.1. Furthermore, the negative correlation is also captured by the constant in each regression, which is negative and significant at the 1% level except for the speculative bubble dummy variable and the real GDP growth rate. These are significant at the 5% level and not significantly different from zero respectively.
The coefficient for the speculative bubble variable, which is -0.021, does not lead to a rejection of H0. A similar argument as for the GDP growth rate can be made for the outcome of the speculative bubble variable. However, while the speculative bubble variable is not a driver behind the correlation, it is shown in Figure 3.1 that the correlation is increasing during 1995 - 2000 implying a co-movement between the gold and US stock market. This relationship immediately turns negative when the dot-com bubble crash occurs. This pattern is also seen during the US housing bubble of 2002 – 2006 and its crash afterwards. The positive correlation probably indicates that investors either believe gold is a portfolio diversifier or are forward-looking by including gold in their portfolio as a safe haven.
26
Table 5.1: Regression output of the smoothed correlation between the returns of the S&P 500 and Closing Price of Gold, 1969 – 2015. The model applied is the univariate model.
Note: *, ** and *** depict statistical significance at the 10%, 5%, and 1% levels respectively.
Drivers Beta t-stat constant t-stat R2 RMSE N
Inflation rate uscpi -0.826 -1.45 -0.031*** -3.03 0.0449 0.07073 47
worldcpi 0.010 0.04 -0.032*** -2.95 0.0000 0.07287 46
Foreign-exchange rate
fx 0.086 0.49 -0.033*** -2.78 0.0062 0.07682 41
Interest rate ustbills -0.005 -0.21 -0.031*** -2.91 0.0010 0.07234 47
ustbills3month 0.001 0.11 -0.033*** -2.84 0.0003 0.0761 42
GDP growth rate nomgdp -0.061 -0.14 -0.031*** -2.91 0.0004 0.07236 47
realgdp 0.207 0.48 -0.031 -2.90 0.0052 0.07219 47 Crises and Crashes Dummy crisis 0.016 0.69 -0.034*** -2.83 0.0106 0.07199 47 Speculative Bubble Dummy bubble -0.021 -0.80 -0.027** -2.28 0.0139 0.07187 47
5.2
Multivariate model
As described in the previous chapter, the multivariate model drops the most
insignificant coefficient after each regression. Strictly, when the general to specific approach is applied, the financial crises and crashes variable should have been omitted as well (Campos et al., 2005). However, this thesis rather holds this variable constant in each regression, since the inflation rate is only significant when the model controls for crises and crashes. If the latter variable would have been left out, the same results as from the univariate model would have been obtained.
The outcome of the multivariate model is the US inflation rate, which is significant at the 10% level. The negative sign of the driver indicates that the inflation rate has a negative influence on the correlation between US stocks and gold returns. Since the benchmark is set at a negative correlation between US stocks and gold, this result implies that the correlation becomes even more negative. Therefore, this finding confirms that gold is a hedge against inflation, which is in line with the study of Dempster and Artigas (2010), Ghosh et al. (2002), and Bialkowksi et al. (2015). The fact that the inflation rate has a stronger influence than the foreign-exchange rate is also found by Baur (2013), who claims that the foreign-exchange rate shows a more stable but generally weaker influence compared to inflation rate. Remarkable,
however, is that the inflation rate is significantly different from zero at the 10% level when the model controls for periods of crisis whereas the inflation rate is not
significant when its isolated performance is examined. According to Schwartz (1995), a relationship between price stability and financial stability exists. In addition,
Schwartz (1995) claims that fluctuations in the inflation rate can have negative consequences on financial instability and possibly lead to a crisis. An implication of this relationship could be that in periods of financial vulnerability the impact of the inflation rate is stronger than in times of financial stability. As a result, the coefficient of the inflation rate comes out more significant when there is controlled for periods of crises and crashes.
27
Table 5.2: Regression output of the smoothed correlation between the returns of the S&P 500 and Closing Price of Gold, 1969 – 2015. The model applied is the multivariate model.
Note: *, ** and *** depict statistical significance at the 10%, 5%, and 1% levels respectively.
Drivers Beta t-stat constant t-stat R2 RMSE N
Inflation rate uscpi -1.114* -1.85 -0.041*** -3.26 0.0817 0.0714 47
Crises and Crashes Dummy
crisis 0.032 1.33 47
5.3
Summary of results
Although all of the hypotheses presented in statement (1) are rejected, gold has shown to be a weak hedge against inflation, a weak hedge against currency changes and a weak safe haven in times of crisis which is in line with most of the existing literature. The interest rate, the GDP growth rate and the speculative bubble variables all do not have a significant influence on the correlation. This is probably due to the fact that the correlation is both positive and negative, which leads to an average of zero. In
addition, the significance of these drivers is also dependent on the time window, the type of market/country and the economic environment examined as outlined in the literature review. Moreover, the negative (and in almost all cases significant) constant in each regression shows that the expected mean of the correlation is negative.
