International Review of Financial Analysis 60 (2018) 30–37

Contents lists available at ScienceDirect

International Review of Financial Analysis

journal homepage: www.elsevier.com/locate/irfa

Is gold a Sometime Safe Haven or an Always Hedge for equity investors? T

A Markov-Switching CAPM approach for US and UK stock indices
⁎
Zhen He, Fergal O'Connor , Jacco Thijssen
The York Management School, University of York, United Kingdom

A R T I C LE I N FO A B S T R A C T

Keywords: This paper re-examines gold's role as a tool for investors to manage their portfolio risk. We begin by assessing
Gold gold's average relationship to an investor's diversified equity portfolio by applying the basic Capital Asset Pricing
Hedge Model (CAPM) to UK and US equity indices. Next, we apply a Markov-switching CAPM to assess whether two
Safe Haven distinct states exist between gold's relationship with the Market Portfolio. This approach allows the data to
CAPM
determine if two separate states exist and, if so, whether one state matches the definition of a Safe Haven from
Beta
Markov-switching model
the literature. Using this new approach, we find that gold is consistently a Hedge, but that no distinct Safe Haven
Stock markets state exists between gold and UK or US stock markets.
UK
US
FTSE100
S&P500

1. Introduction for equity prices, because they should drive these prices lower or
higher. However, macroeconomic drivers do affect its price, such as
Gold has had a long and unique history as a financial asset over the inflation (O'Connor et al. 2015), but if you buy an ounce of gold it will
last 6000 years. Recently there has been a growing body of research remain an ounce of gold. It cannot default or go bankrupt as it has no
assessing whether gold acts as a Safe Haven for investors in times of offsetting liability.
severe market stress. Since the 2008 financial crisis, gold has gained a Secondly, though there has been a large amount of recent research
renewed prominence for investors and researchers as its price rose from on whether gold acts as a Safe Haven for a number of asset classes, all
$252 in July 1999 to $1900 dollars an ounce briefly on the 5th of these studies choose an arbitrary quantitative cut-off point to define
September 2011 and many Exchange Traded Funds (ETFs) were set up when a Safe Haven period should be present. Generally it is defined as
to make it easier for smaller investors to buy gold. when an asset's returns being in the bottom 5% or 1% quartile of the
Since gold prices have floated freely after 1968, gold's allure for sample. Authors then test the relationship between gold and the asset in
speculators has waxed and waned in tandem with its price changes. that quantile; see for example Baur and Lucey (2010). The usual defi-
Simultaneously gold has maintained a core group of investors often nition of Safe Haven comes from Baur and McDermott (2010:1889)
referred to as “gold bugs” who see it as the ultimate safe asset (Siomon, which states that a “strong (weak) Safe Haven is defined as an asset that
2013), one of the few assets with no counterparty risk once physically is negatively correlated (uncorrelated) with another asset or portfolio in
you hold it. The majority of recent research finds that gold does have a certain periods only, e.g. in times of falling stock markets” (our em-
role as a hedge and/or a Safe Haven for investors (see O'Connor, Lucey, phasis). Note that these periods can only be observed ex post as they
Batten, and Baur (2015) for a review). depend on returns over the entire sample period. That severely reduces
The first contribution of this paper is to offer an estimate of the the use of such studies to guide investors' decisions.
Hedge characteristic in gold in a CAPM setting and assess whether Using a Markov-switching model, rather than an arbitrary cut off
gold's beta is zero, as is often assumed (Baur & McDermott, 2010; Blose, point, we allow the data itself to determine whether two natural and
2010; Reboredo, 2013). This assumption is based on gold's unusual separate “regimes” exist between gold and other asset prices. If two
economically inert nature. Unlike other assets it does not have any states do exist in the Markov-switching approach, then the next step is
fundamental driver of its own. For example, dividends act as a driver to see whether there is any relationship between one of the states and

⁎
Corresponding author.
E-mail address: [email protected] (F. O'Connor).

https://doi.org/10.1016/j.irfa.2018.08.010
Received 25 June 2018; Received in revised form 14 August 2018; Accepted 23 August 2018
1057-5219/ © 2018 Elsevier Inc. All rights reserved.
Z. He et al. International Review of Financial Analysis 60 (2018) 30–37

periods of extreme stock market movements. We found that gold is gold and stock returns for the US in the extreme end of the distribution
always a Hedge, but that neither state corresponds to what might be (the 1% quartile) is −0.0183 which does fit the definition of a safe
thought of as a separate safe haven characteristic. We conclude that haven above. But this a less negative relationship than the average
gold is always a Hedge for stock market risk. Neither state estimated by described above (−0.0475). For the UK the figure for the 1% quartile is
the model maps clearly onto times of large equity market price falls. a lot lower at −0.29 (versus 0.18 on average) and the German estimate
This finding that there is no change in the relationship during is in in the 1% quartile is −0.0727 versus 0.04 on average.
keeping with a re-reading of the results in the literature on this topic, Looking at the definition of a Safe Haven above this means that
which is discussed in the following sections. while for the US your portfolio does benefit from gold in times of severe
In Section 2 we review the previous research on gold, its Beta and its market falls, it does not act differently at these times. It remains a hedge
Safe Haven status. Section 3 presents the data used and the applied rather than becoming a Safe Haven. For the UK it has a negative re-
methodologies. Section 4 presents the empirical results and Section 5 lationship in crisis periods, but also on average. So again, gold seems to
presents our conclusions. remain an excellent hedge at all times rather than there being a sig-
nificant shift in the relationship, at times of large share price falls, into a
2. Literature review Safe Haven.
Baur and McDermott (2010), using daily, weekly and monthly data,
2.1. Gold's Beta under the CAPM also determine that gold is a Safe Haven. Similarly to Baur and Lucey
(2010), no average relationship with the market examined in greater
Despite the popularity of the Capital Asset Pricing Model (CAPM) it than 0.1, implying gold is a Hedge for all these markets. While most of
has been applied to gold in very few papers. Chua, Sick, and Woodward these estimates are statistically significantly different from zero at a
(1990) discuss the application of CAPM using monthly gold prices, daily level, when the analysis is run at a monthly level far fewer are,
Toronto Stock Exchange (TSE) gold stock index, S&P gold index and S& suggesting that the statistical significance is due to the large sample size
P 500 index from September 1971 to December 1988 but the variables employed. As the estimates still give gold a strong diversifier role,
in the model are returns rather than excess returns as should be used perhaps the definition that to be a Hedge it must have zero correlation
per the CAPM (Lintner, 1965; Sharpe, 1964). Although the results with the market portfolios is too strong.
suggest that gold's Beta, using the S&P 500 as a proxy for the market Baur and McDermott (2010) also assess golds relationship in specific
portfolio, is insignificantly different from zero in two sub-periods ex- market crashes. For the October 1987 crash four of nine markets ex-
amined, the difference between two estimated Betas (0.03 and 0.22) is amined have a higher correlation with the market at this time than the
relatively large. These results may be due to a lack of power in the tests average relationship. Of the remaining five only one has a statistically
related to the small number of monthly observations. significant lower correlation (the US). Many do have a significantly
Similarly, Dee, Li, and Zheng (2013) have applied a variant of the lower correlation around the bankruptcy of Lehman brothers in 2008,
original CAPM by the application of arbitrage pricing theory (APT) but for the Asian crisis in 1997 no country has a statistically significant
(Ross, 1976) as they an inflation factor to the original CAPM. There has change in relationship between gold and markets. This again seems to
not been an application of the actual CAPM to gold to date in the lit- imply that golds ability to diversify risk is something of a constant even
erature. during extreme stock market moves, rather than an increased or new
ability.
2.2. Gold as a Hedge and a Safe Haven Other studies also find a Safe Haven role for gold. Ciner, Gurdgiev,
Lucey (2013) use a dynamic conditional correlations (DCC) GARCH
Gold has been shown to be a hedge against a large number of pri- model for the S&P500 and FTSE100 and confirm golds role as a Safe
marily financial risks and offers significant diversification benefits Haven by breaking the equity returns in to quantiles. Bredin, Conlon
within a portfolio. Baur & Lucey (2010: 220) define a hedge as “an asset and Poti (2015) use wavelet multiscale analysis and find that gold offers
that is uncorrelated or negatively correlated with another asset or a longer Safe Haven facility over a longer period than other studies at
portfolio on average.” Lucey, Tully, and Poti (2006) point to gold's one year. Beckmann, Berger, and Czudaj (2015) also find a Safe Haven
returns being positively skewed, in contrast to almost all other financial role for gold.
assets, as a major driver of its ability to reduce portfolio risk. Emmrich Lucey and Li (2015) examined gold's role as a Safe Haven in a time-
and McGroarty (2013) find that adding gold to a range of portfolios varying manner with the extension of three other precious metals:
reduces their risk level. silver, platinum and palladium. Their results suggest that gold may not
Gold also hedges macroeconomic risks, such as inflation risk over act as a Safe Haven at all times but that when it is not other precious
the longer term (see Batten, Ciner, and Lucey (2014) for example). metal can fill in for the role in an investor's portfolio.
Some authors have also found it to be a hedge against exchange rate All of the above research then agrees that in times of extreme stock
risk, such as Reboredo and Rivera-Castro (2014), but our focus here will market movement gold's low or even negative correlation with broader
be on gold's ability to hedge asset price risks. asset markets makes it valuable for investors. A re-reading of the results
Baur and Lucey (2010) and Baur and McDermott (2010) both define however implies that the safe have characteristic of gold is simply that
a Safe Haven in their analysis of whether gold serves as a hedge or a it always remains a hedge.
Safe Haven to stocks and bonds. These two papers have formed the
basis of the research that has followed on the issue.
Using a GARCH model, Baur and Lucey (2010) assess whether the 3. Data and methodology
relationship between gold returns and other asset returns is different in
the lowest quantiles of returns (1%, 2.5% and 5%). They find, for ex- 3.1. Data
ample, that the average relationship between gold and US equites is
−0.0475 i.e. an almost zero correlation between gold and US equities, All the data used was collected at a daily frequency, which we
which is indicative of a hedge based on their definition above. In the convert to weekly or monthly frequencies when required. Sources are
three countries examined for stocks and bonds their average relation- listed in Table 1. We use the US Dollar and Pound Sterling PM gold
ship with gold is between −0.18 and +0.1. These all indicate that gold prices from the London market, known as the Gold Fixing until March
offers significant diversification benefits when added to a portfolio due 2015 as London has been found to be the dominate market for price
to the low or negative correlation with a diversified portfolio. In formation in the global gold market (see Lucey, Larkin, and O'Connor
looking at the safe haven aspect they find that the relationship between 2014).

31
Z. He et al. International Review of Financial Analysis 60 (2018) 30–37

Table 1
Data sources.
Data Date range Source

PM Gold Fixing price, £ 01/04/1968 to 29/09/2017 DataStream

PM Gold Fixing price, $ 01/04/1968 to 29/09/2017 DataStream
UK Treasury bill 3-month 04/01/1985 to 29/09/2017 Bank of England
US 3-month Treasury bill rate 01/01/1969 to 29/09/2017 FRED
FTSE 100 01/01/1985 to 29/09/2017 DataStream
FTSE 350 01/01/1985 to 29/09/2017 DataStream
FTSE All Share 01/01/1985 to 29/09/2017 DataStream
S&P 500 01/01/1969 to 29/09/2017 DataStream
Dow Jones 01/01/1969 to 29/09/2017 MeasuringWorth.com

3.1.1. Market proxy for UK and US market (Rg, t − Rf , t ) = α + β (RM , t − Rf , t ) + εt , (2)

As explained in Fahad and Laura (2017), the FTSE All Share index is
where t = 1, 2, …, T denotes the time horizon. Rg, t is the log-return on
a mix of the FTSE 100, FTSE 250 and FTSE Small Cap indices, which
gold at time t, Pg, t is the price of gold in time t, RM, t is the return on the
means that FTSE All Share contains a wide basket of equities. The FTSE
market portfolio in period t, Pm, t is price of the market portfolio in
All Share index represents about 99% of the UK's market capitalization.
period t. Rf, t the risk-free rate in period t, α the intercept term and εt the
Thus, it is reasonable to apply the FTSE All Share as a proxy variable for
error term. Note that εt is assumed to be an independent and identically
the UK market portfolio in the analysis. We also use the S&P 500 and
distributed random variable that follows the normal distribution, εt~N
Dow Jones indices as the commonly used proxies for the US market
(0, σ2).
portfolio.
Following the assumptions of the Sharpe-Lintner CAPM (Lintner,
1965; Sharpe, 1964), by taking expectations on both sides of Eq. (2),
3.1.2. Descriptive statistics
the following Eq. (3) must hold,
Table 2 shows descriptive statistics for all the variables used.
The Jarque-Bera test is used to examine whether the series is nor- E (rg ) = βE (rm) (3)
mally distributed or not. The null-hypothesis of this test is that the
where the excess log-return on asset gold is denoted by rg = Rg − Rf, the
series follows a normal distribution. From Table 2 it can be deduced
excess log-return on market portfolio is denoted by rm = RM − Rf.
that there is evidence that none of the series are normally distributed.
It implies that the intercept term α in Eq. (2) must be zero in order to
Fig. 1(a) and (b) graph of gold's daily excess returns in Pounds
be consistent with Sharpe-Lintner CAPM.
Sterling and US Dollars from April 1985 to August 2017. The excess
return on gold in US market has more extreme values than that in UK
3.2.2. Markov-switching CAPM model
market. Fig. 2(a), (b) and (c) are the excess return on the FTSE 100,
Markov-switching models estimate regime shifting endogenously.
FTSE 350 and FTSE All Share at a daily frequency. Fig. 3(a) and (b)
The Markov-switching model was introduced in Hamilton (1989) for
show the excess return on the S&P 500 and Dow Jones respectively.
the analysis of the non-stationary time series analysis of the business
cycle.
3.2. Methodology We here apply a Markov-switching model to test whether there are
regime shifts in gold's Beta within a CAPM framework. The only ap-
3.2.1. Capital Asset Pricing Model plication of Markov-switching model in the research of the gold to our
The Capital Asset Pricing Model is applied in the analysis and ac- knowledge is Lucey and O'Connor (2013) who look at whether bubbles
cordingly we assume that the investors are risk averse and the market is occur in gold prices. Here, the model is applied to see whether a data
complete (Blitz, Falkenstein, & Van Vliet, 2014). In the following ana- driven model finds a two distinct states exist between gold returns and
lysis, the returns on gold and the market portfolios with respect to in- returns on diversified equity portfolios.
dices are calculated as the following equations (Fahad & Laura 2017): Following Hamilton (1989) and Huang (2001), we denote by st the
Pi, t ⎞ state variable that reflects the regimes in the market. Here, we assume
Ri, t = ln ⎛⎜ ⎟,
that there are two different regimes in the model. The Markov-
P
⎝ i, t − 1 ⎠ (1)
switching CAPM equation then takes the form of Eq. (4) below:
where Ri, t is the log-return on asset i in period t and Pi, t is the price of
(Rg, t − Rf , t ) = α st + βst (RM , t − Rf , t ) + ε st , (4)
asset i in period t.
Gibbons (1982) gives the following equation for the CAPM: where st = 1, 2 represent the two states of the model, the error term εst
is assumed to be iid N (0, σst 2) . Here, s1 represents one regime with the
Table 2
parameters α s1, βs1, σs1 2 , while s2 is another regime with corresponding
Descriptive statistics. parameters α s2, βs2, σs2 2 . The reasons for allowing regime-switching in
the variances of the error term will be explained in the next section.
Gold Gold FTSE S&P 500 Dow 3- 3-
Following the Markov-switching model in Hamilton (1989), the
price price All Jones - month month
(UK) - (US) - Share $ T-bill T-bill state variable st in this case will only take the binary values of 0 or 1.
£ $ (Index) rate rate Thus, the transition probabilities in the first-order Markov process are
(UK) % (US) % modelled as the following (see Lucey and O'Connor (2013)),
Mean 312.13 507.27 1605.27 710. 6145 0.5440 0.3875 Prob [st = 1 | st − 1 = 1 ] = p (5)
Variance 277 417 1230 633 5588 0.3318 0.2675
Skewness 1.457 1.365 0.2526 0.7597 0.7386 0.0235 0.3926 Prob [st = 2 | st − 1 = 1] = 1 − p (6)
Kurtosis 4.0465 3.9200 1.6249 2.4926 2.3948 2.1797 3.0112
Jarque-Bera 198*** 62*** 63*** 15*** 201*** 52*** 15*** Prob [st = 2 | st − 1 = 2 ] = q (7)

Note: **, *** represent the statistical significance at 5% and 1% & level. Prob [st = 1 | st − 1 = 2] = 1 − q (8)

32
Z. He et al. International Review of Financial Analysis 60 (2018) 30–37

Fig. 1. (a) Excess return on gold in UK market (daily).

(b) Excess return on gold in US market (daily).

Fig. 2. (a) Excess return on market portfolio (FTSE 100 daily).

(b) Excess return on market portfolio (FTSE 350 daily).
(c) Excess return on market portfolio (FTSE All share daily).

33
Z. He et al. International Review of Financial Analysis 60 (2018) 30–37

Fig. 3. (a) Excess return on market portfolio (S&P 500 daily).

(b) Excess return on market portfolio (Dow Jones daily).

Note that the probabilities p and q are determined endogenously. imply that stock market returns do not have any explanatory power for
gold returns, making the gold price independent of the stock market
4. Empirical results indices looked at over the periods examined. Results using weekly and
monthly data for golds Beta and the R2 for all five markets result in the
We have undertaken unit root tests on excess returns of gold in both same interpretation as above and are available on request.
currencies as well as for excess returns on all proxies for the market
portfolio listed above. The null hypothesis of the augmented Dickey- 4.2. Results of Markov-switching model
Fuller (ADF) states the existence of a unit root. For all the time series
examined this null is rejected at the 1% level. As a result, we will treat Next, we estimate the Markov-switching CAPM for both weekly and
all variables as stationary series. Results are available on request from daily data, in that order. Table 4 shows the results of the weekly data.
the authors. Two distinct regimes are identified for all five proxies of the market
portfolio.
4.1. Results of CAPM estimation In UK market, the assumptions of CAPM do not all hold, because the
intercept is significantly different from zero in regime 2 for the FTSE
The results in this section are estimated based on Eq. (2) applying an 100 and the FTSE 350. Although this assumption of CAPM holds in
OLS regression model. The null hypothesis is that gold's Beta is zero, FTSE All Share, gold's Betas are all statistically insignificant for UK data.
against a two-sided alternative. This suggests that gold's Beta in both regimes is zero. If we do not
According to Table 3 the assumption of no intercept in the CAPM specify a threshold in advance to define what a Safe Haven might be,
holds as all estimates are found to be statistically insignificantly dif- that data determines that there is none; instead, in both regimes, gold
ferent from zero. Estimates of gold's Beta are surprisingly consistent acts as a Hedge or diversifier for stock market risk.
across markets and Market Portfolio proxies at around −0.03 in all five The two regimes are, instead, differentiated by different levels of
cases. The Betas are also all found to be statistically significantly dif- volatility, with regime two being two- to three-times more volatile than
ferent from zero, but the estimates are close to zero and the statistical regime 1 for UK data. The switching probabilities are also quite high,
significance can be attributed to the large sample size employed. An with the probability of moving from regime 1 to 2 (1-q) based on FTSE
estimated Beta of −0.03 means a 10% fall in an equity index is asso- 100 data being 7.96%, while a switch from regime 2 to 1 (1-p) has a
ciated on average with a 0.3% increases in the price of gold, which 3.72% probability from week to week. Fig. 4 shows the weekly excess
economically is zero close to zero making gold a very good Hedge for returns on the FTSE 100. For a Safe Haven to exist we would expect that
stock market risk with significant diversification benefits for an in- large negative stock market excess returns would map well onto one of
vestor's portfolio. the regimes. The shaded area of Fig. 4 represents times when the model
In addition, the R2s for all regressions are extremely small. While in has determined regime 1 is applicable. As can be seen, most of the large
most research this would be a negative, here this is further evidence falls in the stock market occur in regime 1, which also includes the
that gold is a good Hedge for stock market risk. The low R-Squares periods of quiet stock markets. Neither regime corresponds to a Safe

Table 3
CAPM for Gold, UK and US market - daily data.
UK-FTSE 100 UK-FTSE 350 UK-FTSE All Share US-S&P 500 US-Dow Jones

Intercept (α) −0.0006 0.0028 −0.0005 0.0095 0.0106

[0.9602] [0.8136] [0.9653] [0.4102] [0.3619]
Beta (β) −0.0302*** −0.0375*** −0.0357*** −0.0288*** −0.0371***
[0.0037] [0.0007] [0.0013] [0.0087] [0.0009]
R-squared 0.0011 0.0015 0.0009 0.000594 0.0010
Durbin-Watson 2.0273 2.0277 2.3793 2.0758 2.0738

Note: **, *** represent the statistical significance at 5% and 1% & level.

34
Z. He et al. International Review of Financial Analysis 60 (2018) 30–37

Table 4
Markov-switching CAPM for Gold – weekly data.
UK-FTSE 100 UK-FTSE 350 UK-FTSE All Share US-Dow Jones US-S&P 500

Regime 1
Intercept (α) −0.0649 −0.0567 −0.0442 0.0584 0.0550
[0.2508] [0.3223] [0.4830] [0.1883] [0.2166]
Beta (β) 0.0429 0.0456 0.0197 −0.0832*** −0.0808***
[0.1454] [0.1462] [0.5798] [0.0000] [0.0001]
log(σ) 0.4487*** 0.4468*** 0.4592*** 0.4617*** 0.4563***
[0.0000] [0.0000] [0.0000] [0.00000] [0.0000]

Regime 2
Intercept (α) 0.311138* 0.319169* 0.170272 0.308839 0.3023
[0.0357] [0.0345] [0.2792] [0.0612] [0.0591]
Beta (β) −0.0101 −0.0161 0.0430 0.1622** 0.2042***
[0.8535] [0.7803] [0.4916] [0.0156] [0.0013]
log(σ) 1.1528*** 1.1579*** 1.0892*** 1.4154*** 1.3999***
[0.0000] [0.00000] [0.0000] [0.0000] [0.0000]
P12 0.0796 0.0760 0.2514 0.0118 0.0117
P21 0.0372 0.0353 0.0980 0.0387 0.0380

Note: **, *** represent the statistical significance at 5% and 1% & level. P-values are shown in parentheses.

Fig. 4. Excess return on market portfolio (FTSE 100) and regime - weekly.

Fig. 5. Excess return on market portfolio (Dow Jones) and regime - weekly.

Haven role for gold. the model has determined that the negative beta regime is present, i.e.,
Results for the US in Table 4 are different from those we obtained regime 1. This, then, is the more common regime for these markets to
for the UK, as has been the case in previous studies. Gold's Beta using be in and would not correspond to a Safe Haven, which would be ex-
both Dow Jones and S&P 500 data as proxies for the market portfolio pected only to dominate at times of severe market stress. Looking at the
show statistical significance in both regimes 1 and 2. In regime 1, gold graph, large market falls occur in both regimes with the 1987 crash
has a negative Beta of about −0.08 and a positive Beta of between 0.16 occurring in regime 2 but the 2008 financial crisis falling into regime 1.
and 0.20 in regime two. Both provide low Betas, which indicates that This indicates that gold did play a distinct Safe Haven role between
gold is a diversifier in both regimes and a strong Hedge for regime one. 1970 and 2017 in the US. The same observation can be made in Fig. 6,
Fig. 5 shows the weekly excess return on the Dow Jones as well as where the S&P500 is used as the market portfolio proxy.
the applicable regime. The shaded area of Fig. 5 represents times when Table 5 shows the results of the Markov-switching CAPM

35
Z. He et al. International Review of Financial Analysis 60 (2018) 30–37

Fig. 6. Excess return on market portfolio (S&P500) and regime - weekly.

Table 5
Markov-switching CAPM for Gold – daily data.
FTSE 100 FTSE 350 FTSE All Share Dow Jones S&P 500

Regime 1
C 0.0021 0.0214 −0.0064 0.0153 0.0144
[0.9911] [0.6765] [0.9717] [0.0772] [0.0923]
β −0.2207** −0.1272*** −0.0232*** −0.0466*** −0.0388***
[0.0167] [0.0002] [0.0160] [0.0000] [0.0004]
log(σ) 1.3495*** 0.5820*** 1.3497*** −0.3907*** −0.3905***
[0.0000] [0.0000] [0.0000] [0.0000] [0.0000]

Regime 2
C −0.0037 −0.0061 −0.0038 −0.0160 −0.01937
[0.7079] [0.5547] [0.7048] [0.6710] [0.6053]
β 0.0207** 0.0268*** 0.0208 −0.0050 0.0064
[0.0486] [0.0002] [0.0641] [0.8673] [0.8281]
log(σ) −0.1941*** −0.2973*** −0.1941*** 0.6954*** 0.7062***
[0.0000] [0.0000] [0.0000] [0.0000] [0.0000]
P12 0.2061 0.0710 0.2062 0.0186 0.0194
P21 0.0136 0.0168 0.0136 0.0449 0.0475

Note: **, *** represent the statistical significance at 5% and 1% & level. P-values are shown in parentheses.

Fig. 7. Excess return on market portfolio (FTSE 100) and regime - daily.

estimations using daily data. Using daily data may be a better way of index is broadened and, though the estimated Beta for the FTSE All
determining whether gold acts as a Safe Haven, as some of the previous Share is statistically significant at −0.0232, from an investor stand-
research indicates that gold only fulfils this role for about 15 days at a point it is effectively zero. Regime one in both cases is also very un-
time (Baur and Lucey (2010)). stable with a transition probability of moving from regime 1 to 2 (1-q)
In regime one, gold's Beta relative to the FTSE100 index does show a over 20% for both. The Betas estimated in regime two are close to zero
much larger and significant negative Beta, at −0.22, than for the other with only one being statistically significant. Both regimes indicate that
two UK market proxies. The magnitude of gold's Beta declines as the gold would act as a strong diversifier.

36
Z. He et al. International Review of Financial Analysis 60 (2018) 30–37

Fig. 8. Excess return on market portfolio (S&P 500) and Regime - Daily.

When we look at Fig. 7, we can again see that neither regime maps Baur, D. G., & McDermott, T. K. (2010). Is gold a safe haven? International evidence.
onto the data well as a Safe Haven. Regime 2 (the non-shaded region of Journal of Banking & Finance, 34, 1886–1898.
Beckmann, J., Berger, T., & Czudaj, R. (2015). Does gold act as a hedge or a safe haven for
Fig. 7) covers most of the period including the 2008 financial crisis and stocks? A smooth transition approach. Economic Modelling, 48, 16–24.
the 1987 crash indicating that gold's role does not switch for the UK at a Blitz, D., Falkenstein, E., & Van Vliet, P. (2014). Explanations for the volatility effect: An
daily level into a Safe Haven in times of severe market movements. overview based on the CAPM assumptions. Journal of Portfolio Management, 40, 61.
Blose, L. E. (2010). Gold prices, cost of carry, and expected inflation. Journal of Economics
For the US Market proxies, the results remain similar to when es- and Business, 62(1), 35–47.
timated with weekly data. Regime 2 has significant but very small ne- Bredin, D., Conlon, T., & Potì, V. (2015). Does gold glitter in the long-run? Gold as a
gative Betas, but regime 1 now has Betas that are insignificantly dif- hedge and safe haven across time and investment horizon. International Review of
Financial Analysis, 41, 320–328.
ferent from zero, even with the large number of observations used. Chua, J. H., Sick, G., & Woodward, R. S. (1990). Diversifying with gold stocks. Financial
Looking at Fig. 8, these regimes again do not bear any relationship to Analysts Journal, 46, 76–79.
extreme movements in the stock market. Regime 1 (the non-shaded Ciner, C., Gurdgiev, C., & Lucey, B. M. (2013). Hedges and safe havens: An examination of
stocks, bonds, gold, oil and exchange rates. International Review of Financial Analysis,
area) covers the global financial crisis in 2008 while regime 2 covers
29, 202–211.
the 1987 crash. Again, these regimes do not map in a way that indicates Dee, J., Li, L., & Zheng, Z. (2013). Is gold a hedge or a safe haven? Evidence from inflation
that gold offers special characteristics that we might think of in terms of and stock market. International Journal of Development and Sustainability, 2, 12–27.
the definition of a Safe Haven. Emmrich, O., & McGroarty, F. J (2013). Should gold be included in institutional invest-
ment portfolios? Applied Financial Economics, 23(19), 1553–1565.
Fahad, N. U., & Laura, M. R. (2017). The Classical Approaches to Testing the Unconditional
5. Conclusion CAPM: UK Evidence. SSRN Working Paper.
Gibbons, M. R. (1982). Multivariate tests of financial models - A new approach. Journal of
Financial Economics, 10, 3–27.
This paper re-examines the idea that gold has a special relationship Hamilton, J. D. (1989). A new approach to the economic-analysis of nonstationary time-
with other assets in times of severe market down moves, its Safe Haven series and the business-cycle. Econometrica, 57, 357–384.
characteristic. We achieve this by applying a Markov-switching model Huang, H.-C. (2001). Tests of CAPM with nonstationary beta. International Journal of
Finance and Economics, 6, 255–268.
to allow the data to determine if such a relationship is present. Previous Lintner, J. (1965). The valuation of risk assets and the selection of risky investments in
studies chose to set cut-off points to define when a Safe Haven could be stock portfolios and capital budgets. The Review of Economics and Statistics, 47, 13–37.
present, which introduces a preconceived notion about what a Safe Lucey, B., Larkin, C., & O'Connor, F. (2014). Gold markets around the world – Who spills
over what, to whom, when? Applied Economics Letters, 21(13), 887–892.
Haven should be.
Lucey, B. M., & Li, S. (2015). What precious metals act as safe havens, and when? Some
We find that while gold does act as a very good diversifier in both US evidence. Applied Economics Letters, 22, 35–45.
regimes estimated by the model and for all proxies of the market Lucey, B. M., & O'Connor, F. A. (2013). Do bubbles occur in the gold price? An in-
vestigation of gold lease rates and Markov Switching models. Borsa Istanbul Review,
portfolio in the US and the UK, neither one maps well onto periods of
13, 53–63.
time where a Safe Haven would be beneficial to investors such as the Lucey, B. M., Tully, E., & Poti, V. (2006). International portfolio formation, skewness and the
1987 crash or the 2008 financial crisis. Instead, we think that a review role of gold Frontiers in Finance and Economics. 3, 49–68.
of the results from earlier papers on this issue, coupled with our find- O'Connor, F. A., Lucey, B. M., Batten, J. A., & Baur, D. G. (2015). The financial economics
of gold - A survey. International Review of Financial Analysis, 41, 186–205.
ings, points to the fact that gold is always a Hedge or, at worst, always an Reboredo, J. C. (2013). Is gold a hedge or safe haven against oil price movements?
excellent diversifier of portfolio risk. Gold's usefulness in managing risk Resources Policy, 38(2), 130–137.
does not disappear in a crisis when the prices of the vast majority of Reboredo, J. C., & Rivera-Castro, M. A. (2014). Gold and exchange rates: Downside risk
and hedging at different investment horizons. International Review of Economics &
assets tend to be perfectly correlated. Finance, 34, 267–279.
Ross, S. A. (1976). The arbitrage theory of capital asset pricing. Journal of Economic
References Theory, 13, 341–360.
Sharpe, W. F. (1964). Capital-asset prices - A theory of market equilibrium under con-
ditions of risk. Journal of Finance, 19, 425–442.
Batten, J. A., Ciner, C., & Lucey, B. M. (2014). On the economic determinants of the Siomon, E. (2013). Is gold really a safe haven asset? The telegraph. 27th April 2013
gold–inflation relation. Resources Policy, 41, 101–108. Available at: https://www.telegraph.co.uk/finance/personalfinance/investing/
Baur, D. G., & Lucey, B. (2010). Is gold a hedge or a safe haven? An analysis of stocks, gold/10019128/Is-gold-really-a-safe-haven-asset.html.
bonds and gold. The Financial Review, 45(2), 217–229.

37