Resources Policy 48 (2016) 4149

Contents lists available at ScienceDirect

Resources Policy
journal homepage: www.elsevier.com/locate/resourpol

The Turkish appetite for gold: An Islamic explanation

Osman Glseven a,n, zgn Ekici b
a
American University of the Middle East, Department of Math and Statistics, Kuwait
b
Ozyegin University, Department of Economics, Istanbul, Turkey

art ic l e i nf o a b s t r a c t

Article history: A signicant constituent of household wealth in Turkey is gold. Families accumulate gold especially on a
Received 16 November 2015 variety of cultural occasions such as female-only gold days, circumcision feasts, and engagement and
Received in revised form wedding ceremonies. This paper attempts to explain the gold appetite of Turkish households with a
14 February 2016
rational approach, although still rooted in culture. Many people in Turkey view earning interest on
Accepted 15 February 2016
money as a transgression of Islamic rules and avoid investing their savings in xed-income investment
Available online 27 February 2016
instruments. We study the implications of this investor behavior on portfolio gold holdings. Using the
Jel classication: Markowitz mean-variance model and monthly return data from 1997 until 2015, we calculate optimal
G11 investment portfolios. We nd that while the share of portfolio gold holdings is less than a meager 4% if
G18
the portfolio includes interest-earning deposits (in addition to the stock index, USD, EURO, and gold), this
Keywords: ratio may go up to more than 50% if interest-earning deposits are not included. Our results show that the
Gold role gold plays in an investment portfolio is greatly amplied when interest-earning deposits are not
Investment viable. We believe the key factor driving our pronounced ndings has been Turkey's historically high
Portfolio optimization rates of ination. Avoiding xed-income instruments, many Turkish investors may have turned to gold to
Islam shield their savings against ination and manage portfolio risks caused by its high volatility. Our ndings
Interest suggest that Turkish policymakers may nd it useful to popularize Sukuk (Islam-compliant bonds) if they
are to divert household savings away from gold into more productive uses.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction so, and discuss certain policy implications of our ndings.

All around the world people buy gold with a variety of motives:
to store wealth, to shield savings against ination, as a portfolio
diversier, for its ornamental value, or simply as a symbol of status
and power. The demand for gold is not uniform across time and
space. However, gold demand is high in some countries and low in
others. In some places people buy gold mostly in the form of
jewelry while in others as bullions or coins. Gold demand tends to
uctuate seasonally and across years. Drawing on this variation in
gold demand past research studies performed various statistical
analyses and tested a variety of economic models, so as to better
understand why people buy gold and what implications may fol-
low if certain gold-related policies are adopted. Our paper makes a
contribution to the literature on gold research along these lines:
We study a countryTurkey, where the investment demand for
gold is exceptionally high, propose and test a theory as to why it is
n
Corresponding author.
E-mail addresses: osman.gulseven@aum.edu.kw (O. Glseven),
ozgun.ekici@ozyegin.edu.tr (. Ekici).
Fig. 1 below shows Turkey's nominal GDP and its consumer
demand for gold in various categories as a percentage of the global
gures in years 2000 through 2014. During this period of time,
Turkey accounted for 5.6% of the global consumer demand for
gold, amounting to 171 tons (metric) per year. This stands in sharp
contrast with the country's relatively much smaller nominal GDP
gure, corresponding to about 1% of the global gure.1 According
to a report by World Gold Council (2013), as of 2013, Turkish
households accumulated an estimated stock of 3500 tons of gold
tucked under-the-pillow, a term used in Turkey to refer to
physical gold stored by the general population. This corresponds to
about 200 g of gold per household, the worth of which roughly
corresponds to the annual salary of many public employees.
The constituents of consumer demand for gold are jewelry
consumption and investment demand. It turns out that the Turk-
ish consumers' attachment for gold is not restricted to an affection
of the material as a decorative item for adornment. In years 2000
1
According to World Bank WDI data, in the relevant period the population of
Turkey also stood at approximately 1% of the global gure.

http://dx.doi.org/10.1016/j.resourpol.2016.02.006
0301-4207/& 2016 Elsevier Ltd. All rights reserved.
42 O. Glseven, . Ekici / Resources Policy 48 (2016) 4149
Fig. 1. Turkey's global share of gold demand in various categories and nominal GDP.
through 2014, while the average annual demand for gold for
jewelry consumption in Turkey was 98.5 tons, the annual demand
for retail investment also stood at a robust 51 tons. Indeed, Fig. 1
above shows that in relative terms the Turkish consumers' de-
mand for gold has been even more pronounced when it comes to
retail investment. In years 2000 through 2014, while Turkey ac-
counted for about 5.6% of the global consumer demand for gold, its
share of global consumer demand for gold in the form of retail
investment stood at 6.6%.
An even more striking indication of the lure of gold as an in-
vestment instrument for Turkish households is the popularity of
gold coins in the country. Retail investment in gold consists of
investment made in the form of gold bars and gold coins. Since
gold bars are sizeable,2 gold coins are arguably a more viable in-
vestment instrument for small investorsthe majority of house-
holds. As it turns out, Turkey is the country where the demand for
gold coins is by far the highest in the world. In years 2000 through
2014, out of 51 tons of average annual gold demand for retail in-
vestment in Turkey, ofcial coins accounted for 46.3 tons. During
this period of time, as can be seen above in Fig. 1, about one third
of ofcial gold coins globally have been purchased by Turkish
households.
Our paper aims to provide an explanation as to why gold is
such a coveted item as an investment instrument for Turkish
households. There is also an important policy dimension to our
analyses: Turkey is historically a country with low savings rates
and chronic current account decits. Therefore, the country relies
heavily upon foreign capital to nance its growth-stimulating ca-
pital investments. The fact that Turkish households invest a sig-
nicant portion of their savings in gold does not help. Much of the
gold accumulated by households is held at home, outside of the
nancial system. This type of investment has a productivity of zero
and does not add value to social welfare. Past attempts by Turkish
authorities to bring household gold savings into the nancial
system had transitory and limited success. Therefore, to design
effective policies to divert household savings away from gold and
into more productive investment instruments, it is imperative to
understand the root causes of the lure of gold as an investment
instrument for Turkish households.
Common arguments as to why gold is such a coveted item for
Turkish people are rooted in history and culture. Turkey is located
in Mesopotamia and Anatolia where gold has a long history. The
history of gold production in this region goes back to 3000 B.C. The
Lycian Kingdom, which established the rst gold money in around
630 B.C., ruled in modern-day Western Turkey. In the Ottoman
Empire, the predecessor of modern Turkey, gold signied the
power of the imperial house as well as families. The Grand Bazaar
was established in Istanbul in the 15th century, which serves ever
2
The standard gold bar held and traded internationally by central banks and
bullion dealers is the Good Delivery bar with a 400 oz (troy-ounce) (12.4 kg or 438.9
ounces) nominal weight.
since as one of the biggest markets in the world for gold trade. In
Turkish culture gold holds a distinct place. It is a tradition, for
instance, for Turkish women to participate in so-called gold days.
In these female-only social events, women come together peri-
odically at one another's homes, and in each gathering each guest
in the circle presents the host with a predetermined type of gold,
often an ofcial gold coin. Turkish people also present gold gifts to
hosts in a variety of festivities such as circumcision feasts and
engagement and wedding ceremonies. Indeed, it is a tradition that
each wedding ceremony ends with what is called an ornament
pinning ceremony, where guests literally pin their ornament gifts
to cloth pieces, specially designed for this occasion and worn by
the bride and the groom. Presenting the new couple with high-
value golden ornamentssuch as ofcial coins, bracelets, and
chainsis a display of wealth and prestige and also signies the
strength of attachment felt for the young couple and their families.
The signicant role gold plays in Turkish history and culture is
undeniable. Nevertheless, there are reasons to believe that the
gold affection of Turkish households extends beyond historical and
cultural reasons. First, as explained above, Turkish households
accumulate gold not only in the traditional form of gold jewelry
but also in the form of investment instruments, in particular, of-
cial gold coins. Second, an explanation for Turkish gold appetite
rooted only in history and culture would suggest that the country's
gold production capacity should parallel its gold demand. None-
theless, Turkey's demand for gold dwarfs its gold production ca-
pacity. In years 2000 through 2014, the average annual mine
production of gold in Turkey was 13.52 tons, corresponding to a
meager 0.5% of the average annual mine production globally
(2685.6 tons).
In the face of the above facts, our paper asserts a rational ex-
planation for the gold appetite of Turkish households as an in-
vestment instrument. Our explanation is nonetheless still rooted
in culture, more precisely, in religion. In Turkey the vast majority
of people are Muslims. There is virtually universal acceptance
among Muslims that Islam prohibits riba, which in Arabic literally
means usury, or unjust, exploitative gains made in trade or
business. Riba has been mentioned and condemned in several
different verses in the Qur'an (2:275280, 3:130, 4:161, and
30:39), which is the central religious text in Islam. Although some
scholars raised certain objections, in practice riba is most often
viewed as an Islamic term for interest charged on loans.3 In ac-
cordance with this interpretation, many Muslims in Turkey and
elsewhere in the world refuse to invest their savings in xed-in-
come investment instruments such as bonds or interest-earning
deposits. In contrast, assets such as stocks, gold or currency are not
viewed un-Islamic per se, because these assets do not promise
their investors a xed return. Buying the stocks of a company, for
3
For a detailed discussion on riba and the place of interest in Islamic law, see
Nabil (1986).
 O. Glseven, . Ekici / Resources Policy 48 (2016) 4149
instance, is not associated with usury and viewed as a fair practice,
where the investor takes part in a risky business in which he may
prot as well as suffer losses. Our working hypothesis in this paper
is that, since many Turkish people do not view xed-income assets
as viable investment instruments, in their portfolios gold may
assume a greater role as a portfolio diversier, soaring up the
country's demand for gold.
To test our hypothesis we resort to modern portfolio theory and
assume an investor whose goal is to maximize the expected return
on his portfolio of assets while keeping the variation on the return
on his portfolio below a threshold. The assets that we consider are
the ones that are most popular among Turkish investors: gold, the
stock index, foreign currency (USD and EURO), and interest-
earning deposits. We study two scenarios: Under the rst scenario
all of the aforementioned assets are included in the investor's
portfolio. Under the second scenario, in view of an investor who is
a conservative Muslim who abstains from xed-income instru-
ments, we exclude from the portfolio the interest-earning depos-
its. Using data on monthly returns from January 1997 until August
2015, we calculate optimal investment portfolios under these two
scenarios. As a robustness check, we repeat our analyses using
monthly data on both nominal and real returns. In both cases our
ndings are similar and support our hypothesis: when the port-
folio includes interest-earning deposits, based upon our analyses
using both nominal and real returns, in an optimal portfolio the
share of gold holdings is less than a meager 4%. When interest-
earning deposits are excluded, however, this share may go up to an
extreme 65% based upon our analyses with nominal returns and
up to an extreme 48% based upon our analyses with real returns.
It is not surprising that in an optimal portfolio the share of gold
holdings increases when the portfolio excludes interest-earning
deposits: When the variety of assets available for investment is
reduced by one, the shares of portfolio holdings naturally increase
for the remaining assets, including gold. What is striking in our
ndings is the magnitude of change. The fact that in the absence of
interest-earning deposits portfolio gold holdings may increase so
much as up to half the value of the portfolio calls for an ex-
planation. We believe this dramatic change is driven to a large
extent by Turkey's historically very high rates of ination. In the
period of our analyses, the geometric average rate of ination in
Turkey as measured by the CPI index stood at an extreme 1.67%
per month, or about 22% per annum. Naturally the high rates of
ination in Turkey were also accompanied by a high volatility in
the ination rate, with a standard deviation of 1.70% in the
monthly rate of ination. This made Turkish investors face with a
dual challenge: to shield their savings against inationary depre-
ciation and to manage portfolio risks caused by inationary vola-
tility. In such an environment interest-earning deposits are natu-
rally a very attractive investment instrument. Their returns are
highly correlated with the rate of ination and hence help in-
vestors both shield theirs savings against inationary depreciation
and manage portfolio risks caused by the volatility in the rate of
ination. For those investors who do not view interest-earning
deposits as a viable investment instrument, however, this dual role
interest-earning deposits play must be assumed by the remaining
portfolio assets. The USD and EURO are relatively stable currencies
but they have low expected returns. In the relevant period of our
analyses both currencies lost half their values against inationary
depreciation. Although the expected return on the stock index is
much higher, the stock index is a much more volatile asset than
gold. Therefore, in the absence of interest-earning deposits, the
dual role they play against inationary depreciation and volatility
is assumed to a large extent by gold.
Our ndings have the following policy implication for Turkish
policy makers who wish to divert the country's inadequate savings
away from gold into more productive uses. To this end, it may be
43
useful to introduce and popularize alternative nancial instru-
ments that are compliant with Islamic rules. An increase in the
variety of portfolio assets available to investors will reduce the role
gold assumes as a portfolio diversier. More specically, it may be
particularly useful to introduce nancial instruments that are not
xed income yet whose returns are highly correlated with the rate
of ination. An obvious candidate to this end is Sukuk, sometimes
also referred to as Islam-compliant bonds. Widely-used in other
Muslim-majority countries, Sukuk securities are structured in a
way so that they make no interest payments. Sukuk holders in-
stead receive partial ownership of a property whose purchase is
nanced by their savings. From the operation of this property,
Sukuk holders collect the generated prot as rent payments, a
practice accepted to be not in violation of Islamic rules.4
The rest of the paper is organized as follows. Section 2 dis-
cusses the relevant literature on gold research; Section 3 in-
troduces our empirical model and presents our ndings; Section 4
concludes with a summary of our ndings and policy
recommendations.
2. Literature review
There is a large and growing research on different aspects of
the gold market. We refer to here only the research that is most
relevant to our paper. For an extensive review of the existing lit-
erature on other aspects of the gold market, see OConnor et al.
(2015).
There is evidence in the literature that in developing econo-
mies, such as Turkey, gold demand arises from a motive different
than in developed economies. While in developed economies gold
behaves more like a discretionary good, in developing economies
it behaves more like an investment good; see, for instance,
Batchelor and Gulley (1995) and Starr and Tran (2008). Several
studies in the literature treat, as in our paper, the demand for gold
in the context of a specic developing country with high gold
demand. Studying Chinese data, Ming and Wang (2013) nd the
average return on gold to be higher in the months of February,
September and November. They attribute this to Golden Weeks,
the public 7-day-holidays in China that often come after these
months and arguably drive up the monthly gold demands. Baur
(2013) studies India and records positive and statistically sig-
nicant gold price changes in the months of September and No-
vember. These months coincide with the wedding season in India,
which he argues might explain (along with some other reasons)
the abnormal returns. There are also several studies on India in the
literature that explore, as we do in our paper for Turkey, the root
causes for the country's high demand for gold; see, for instance,
Brenson (2014), Suresh (2011), Kannan and Dhal (2008), and
Vuyyuri and Mani (2005). Although cultural references are made
in these papers, they do not incorporate a cultural element ex-
plicitly in a quantitative model to explain India's high demand for
gold. In this respect our paper distinguishes itself from these
studies: our optimal portfolio analyses explicitly incorporate, and
explore the consequences of, the abstention from interest-earning
assets of Turkish investors on religious grounds.
Since our paper studies for Turkey (a country with historically
high rates of ination) the impact on portfolio gold holdings of the
presence and absence in the portfolio of interest-earning deposits,
our paper also relates to research studies on the relationship be-
tween the gold price, the interest rate, and the rate of ination.
There are two competing views in the literature regarding the
4
Indeed, in October 2012, Turkey issued its debut Sukuk, yet the instrument is
still widely unknown in Turkey and lacks a developed nancial market.
44 O. Glseven, . Ekici / Resources Policy 48 (2016) 4149
relationship between interest rates and gold price. One view is
that the interest rate represents an opportunity cost of holding
gold and hence an increase in interest rates should accompany a
reduction in gold prices; see, for instance, Fortune (1987). Another
view is that it is the rate of ination that drives gold prices as well
as the nominal interest rate, and hence the two should have a
positive relationship; see, for instance, Abken (1980). A recent
study by Baur (2011) suggests that both these views may have
merit but to a varying extent in the long run and the short run.
Using monthly data over a 30-year period, he shows that while
lower short-term rates have a positive impact on gold prices,
lower longer-term interest rates (associated with a low future
expected rate of ination) have a negative impact on gold prices.
In a related empirical study, Batten et al. (2014) nd that when
interest rates fall the relationship between the rate of ination and
the gold price becomes stronger. We should note that these em-
pirical studies have been carried out using data from the U.S.
where the rates of ination have been historically relatively low. In
a country like Turkey where ination rates have been historically
high it is natural to expect that the rate of ination play a more
dominant role in driving both the interest rates and gold prices,
necessitating a positive and strong correlation between the two
even in the short run. Indeed, in the period of our analyses we
record positive and signicant correlation values between the rate
of ination, the interest rate and the gold price. This points to our
main nding in the paper, that interest rates are a good hedge
against ination and that in the absence of interest-earning de-
posits gold assumes a greater role as a portfolio diversier.
The literature on the role gold plays in hedging risks and di-
versifying investment portfolios is a growing one. Varying in terms
of methodology, the period of study and the country of analysis,
these studies all point to clear benets of holding gold to hedge
risks and diversify investment portfolios. Chua et al. (1990) nd
the beta of gold (as per the CAPM) to be low and insignicantly
different from zero. Hillier et al. (2006) nd that gold, as well as
silver and platinum, provides signicant diversifying benets
when held in a portfolio of U.S. or global stocks. Sumner et al.
(2010) nd that there are almost no spillovers to gold from US
stocks and bonds. Emmrich and Frank (2013) nd that the addition
of gold to a range of portfolios reduces the volatilities of these
portfolios. Based upon his analyses, Jaffe (1989) recommends
holding approximately 10% of a stock portfolio in gold. Hillier et al.
(2006) calculate the optimal share of portfolio gold holdings in
broad-based international equity portfolios to be approximately
9.5%. Bruno and Chincarini (2010) calculate that in a portfolio the
optimal weight of gold and oil combined is less than 10% in most
countries. Lucey et al. (2006) nd optimal gold holdings in a
portfolio to be 46% under traditional optimization but only 24%
when skewness is also accounted for.
Gold as a possible hedge asset has also been examined in re-
lation to currency portfolios. Sjasstad (2008) looks at the impact of
exchange rates on gold prices and nds the dominant currency to
be USD, followed by Yen, pointing to gold's ability to hedge these
currencies. Capie et al. (2005) nd that the gold's ability to hedge
against Yen/US Dollar and Sterling/US Dollar exchange rates is time
varying based on unpredictable political and economic events. Joy
(2011) nds that the gold's ability to hedge dollar increased over
time. Apergis (2014) nds gold to be a useful predictor and hedge
for the Australian dollar. Wanga and Lee (2011) nd that when the
currency depreciation rate is higher than 2.62% gold starts to act as
a hedge against yen depreciation. Similar ndings as to gold's
ability to hedge currencies are also reported in Reboredo (2013),
Reboredo and Rivera-Castro Miguel (2014), and Yang and Hamori
(2013).
Also regarding the role of gold as an investment asset, a re-
search topic that has drawn attention in the literature is the safe
haven role gold plays for investors, a subject of keen interest also
in the nancial press (see, for instance, Sanderson (2015)). To in-
vestigate gold's safe haven properties, one line of research studied
gold prices at times of political unrest. Using various measures for
political tension, Abken (1980), Ariovich (1983) and Koutsoyiannis
(1983) nd that the power of their models to explain gold prices
do not increase with the inclusion of political tension as an ex-
planatory variable. Ariovich (1983) also records, however, a posi-
tive relationship between gold prices and political tension. An-
other line of research interpreted an asset being a safe haven to
be its ability to protect wealth at times of extreme market
movements such as nancial market crashes. Under this inter-
pretation, Baur and Lucey (2010) nd that gold is not a safe haven
and a hedge for bonds but it is for stocks, but only up to 15 days
after a market crash. Bredin et al. (2015) nd gold to be a safe
haven, however, for up to a year. Studying an international sample,
Baur and McDermott (2010) nd gold to be a safe haven for
equities but not for all countries. Hood and Malik (2013) nd gold
to be uncorrelated with the market in a crash, but not negatively
correlated, making it a weak hedge and safe haven for equities.
3. Empirical analyses
3.1. Data
We collected all relevant data for our analyses from the Turkish
State Statistics Institute and then crosschecked with Thompson
Reuters DataStream. The relevant data was available since 1997,
and hence, we carried out our analyses using monthly data from
January 1997 until July 2015, inclusive. The data we collected in-
cludes the rate of ination, and the nominal returns on gold, the
stock market, interest-earning deposits, and foreign currency. The
rate of ination used is as measured by the CPI index. The return
on interest-earning deposits is the average monthly interest rate of
the 11 big banks in Turkey, ranked by the size of their deposits. The
return on the stock market is as measured by BIST100, which is the
main index used for Borsa Istanbul Equity Market. Since they are
by far the most commonly used ones in Turkey, as foreign cur-
rencies we included in our analyses USD and EURO. As a new
currency, the history of EURO does not go back to 1997, however.
Therefore, for the period before its introduction, we used the re-
turns on Deutsche Mark as a substitute for the returns on EURO.
Indeed, before EURO was introduced, Deutsche Mark was one of
the two foreign currenciesalong with USD, that dominated the
foreign currency market in Turkey.5
We carried out our analyses using both nominal and real re-
turns. For an asset X, we calculated its real return in period t, rt (X ),
from its nominal return in period t, Rt (X ), using the CPI ination
rate in period t, t , as follows:
1+Rt (X )
1 + rt (X ) =
1+t
3.2. Summary statistics
Table 1 below presents for the relevant period the summary
statistics on monthly nominal and real returns on assets and the
CPI ination rate.
Turkey is historically a country with high rates of ination. As
can be seen in Table 1, from January 1997 until July 2015, the
5
The signicance of Deutsche Mark in Turkish foreign currency market before
2002 was partly due to the large community of Turkish guest workers in Ger-
many, who were sending substantial amount of remittances to their relatives in
Turkey.
 O. Glseven, . Ekici / Resources Policy 48 (2016) 4149 45

Table 1
Summary statistics on monthly returns (expressed in percentage terms).

Monthly return Expected return Standard deviation Minimum Median Maximum Geometric mean

Nominal interest 2.34 2.04 0.40 1.39 6.30 2.32

Nominal Bist100 2.63 11.00 30.80 1.77 53.40 2.06
Nominal USD 1.57 4.51 8.50 1.27 30.90 1.47
Nominal EURO 1.48 4.38 7.70 0.96 28.90 1.39
Nominal gold 2.09 5.43 12.30 1.60 35.10 1.95
Real interest 0.65 1.72 3.80 0.50 16.20 0.64
Real Bist100 0.92 10.44 35.17 0.77 44.90 0.38
Real USD 0.12 3.94 10.85 0.48 23.39 0.20
Real EURO 0.20 4.03 12.18 0.75 21.50 0.28
Real gold 0.40 5.18 12.09 0.31 27.35 0.28
CPI index 1.70 2.31 12.48 1.10 10.29 1.67

average monthly CPI ination rate has been 1.7%. As a consequence

of high rates of ination the nominal rates of returns on various
assets have been historically also high. Looking at monthly nom-
inal returns we see that: the two assets that performed best in this
time period were interest-earning deposits and BIST100, with
expected nominal returns of 2.34% and 2.63% and geometric-mean
nominal returns of 2.32% and 2.06%, respectively. Note that al-
though BIST100 outperformed interest-earning deposits in terms
of expected nominal return, in terms of geometric-mean nominal
return it lagged behind due to its very high volatility. Gold was the
third best performer with an expected nominal return of 2.09%
and a geometric-mean nominal return of 1.95%. The worst per-
formers were the two foreign currencies USD and EURO, with
Fig. 2. Nominal growth of 100 TLs based upon monthly nominal returns.
expected nominal returns of 1.57% and 1.48% and geometric-mean
nominal returns of 1.47% and 1.39%, respectively. As might be ex-
pected, in the relevant period BIST100 was the most volatile asset
with a standard deviation of 11%, and interest-earning deposits
were the least volatile with a standard deviation of 2.04%. The
volatilities for gold, USD, and EURO were in between these two
extremes, with standard deviations of 5.43%, 4.51% and 4.38%,
respectively.
Fig. 2 below presents the nominal growth of 100 Turkish Liras
invested in these assets from January 1997 until July 2015. For
comparison the gure also presents the growth of the CPI index
over this period of time.
Table 1 also presents for the relevant period the summary
statistics on monthly real returns on these assets, obtained from
nominal returns by adjusting for the CPI ination. In accordance
with the statistics based upon monthly nominal returns, the two Fig. 3. Real growth of 100 TLs based upon monthly real returns.
assets that performed best based upon monthly real returns in the
relevant period were interest-earning deposits and BIST100, with Table 2
expected real returns of 0.65% and 0.92% and geometric-mean real The correlation matrix for monthly nominal and real returns on assets.
returns of 0.64% and 0.38%, respectively. Gold was the third best
Nominal returns Ninterest NBist100 NUSD NEURO NGold
performer with an expected real return of 0.40% and a geometric-
mean real return of 0.28%. The expected real returns for USD and Ninterest 1.00
EURO had been negative in the relevant period, standing at NBist100 0.17 1.00
0.12% and 0.20%, respectively. Fig. 3 below presents the real NUSD 0.38 0.20 1.00
NEURO 0.38 0.25 0.84 1.00
growth of 100 Turkish Liras invested in these assets in the relevant
NGold 0.25 0.23 0.67 0.69 1.00
period. For comparison the gure also presents the real growth in Real returns Rinterest RBist100 RUSD REURO RGold
the CPI index as a straight line. Rinterest 1.00
In order to have an idea about the relationships between the RBist100 0.07 1.00
RUSD 0.14 0.33 1.00
returns on different assets, we calculated for the relevant period
REURO 0.25 0.34 0.81 1.00
the correlation coefcients, the results of which are presented RGold 0.17 0.29 0.64 0.67 1.00
below: Table 2 presents the correlation matrix for the nominal
returns as well as real returns. Note: all correlation statistics are signicant at 1% level.
The two correlation matrices for real and nominal returns are
fairly similar. Priced in Turkish Liras, the return on gold is highly on gold and BIST100 suggests that gold can be a good diversier
correlated with the returns on USD and EURO. This makes sense for equity holders in Turkey. Also, the fact that the returns on gold
since a depreciation (an appreciation) of Turkish Lira would be and interest-earning deposits are positively correlated hints that
expected to push up (down) the prices of these assets expressed in part of the diversifying role of interest-earning deposits in one's
Turkish Liras. The negative correlation between the rates of return portfolio, when interest-earning deposits are not a viable
46 O. Glseven, . Ekici / Resources Policy 48 (2016) 4149
Table 3
The correlations between monthly nominal returns on assets and the monthly CPI
ination rate.
Ninterest NBist100 NUSD NEURO NGold
CPI index 0.72 0.20 0.45 0.38 0.28
investment instrument, can be assumed by gold.
Turkey has long experienced a serious ination problem. In-
deed, until very recently the annual ination rates have not seen
single digits for many decades. Since 1997, the average annual rate
of ination in the country, as measured by the CPI index, has been
about 22%. Naturally, the high trajectory in ination rates was
accompanied with a high volatility in the rate of ination: as can
be seen in Table 1, in the relevant period of our analyses the
standard deviation of the monthly rate of CPI ination stood at an
extreme 2.31%. The high trajectory and the volatility in the rate of
ination made Turkish investors face with a dual challenge: to
shield their savings against inationary depreciation and to
manage portfolio risks caused by inationary volatility. Indeed, it
would not be farfetched to argue that ination could be singled
out as the primary risk factor that an investor with a portfolio of
assets had to manage. Table 3 below presents for our period of
analyses the correlations between monthly nominal returns on
assets and the monthly CPI ination rate. As might be expected,
the asset whose monthly nominal return has been most correlated
with the rate of ination is the interest-earning deposits. This
suggests that the asset most suitable to manage the risks caused
by the country's high ination rate is the interest-earning deposits.
When investing in interest-earning deposits is not viable, how-
ever, the role they play in managing the risks caused by ination
needs to be assumed by remaining assets. As can be seen in Ta-
ble 3, the return on gold is positively correlated with the CPI index,
suggesting that its role as a portfolio diversier may amplify when
the portfolio does not include interest-earning deposits.
3.3. Model
We apply the mean-variance analysis model based on Marko-
witz's (1952) portfolio selection theory. In this model, it is as-
sumed that an investor is driven by two motives when deciding
for the composition of his portfolio: maximizing the expected (rate
of) return and minimizing the variance of the return on his port-
folio. The Markowitz model has wide support in academia and a
variety of its variations has emerged in the literature.
Let P be a portfolio consisting of assets 1, 2, , n. For asset i , let
wi be its weighting factor in portfolio P ; that is, wi is the share of
the value of the portfolio invested in asset i . Then the portfolio's
expected return, rP , can be calculated as a function of the expected
returns of the individual assets r1, r2, ,rn , as follows:
n
rP = wi ri
i=1
Let ij be the covariance of the returns on assets i and j . Then
the variance of the return on the portfolio, P , can be calculated as
follows:
n n
P = wi wj ij
i=1 j=1
There is a trade-off between the investor's two motives: in-
creasing the portfolio's expected return comes at the cost of an
increased variance of the return on the portfolio. Therefore there is
not a single optimal portfolio; each weighting scheme that mini-
mizes the variance of the return on the portfolio subject to the
constraint that the portfolio's expected return exceeds a threshold
denes an optimal portfolio. Let r denote this threshold level.
Then, for each value of r , the weighting scheme that solves the
following optimization problem denes an optimal portfolio.
n n
min{ wi }ni = 1 wi wj ij
i=1 j=1
s. t.
n n
wi rir ; wi=1; and w ,,wn0
i=1 i=1 1
Note that we assumed that the weighting factors are non-ne-
gativei.e., there is no short-selling.
3.4. Results
In this section we present our results. We calculate optimal
portfolios under a variety of scenarios and, as is the convention,
show the efcient frontiers (the collections of optimal portfolios)
on diagrams whose coordinates are the standard deviation and
expected value of monthly returns on portfolios. For robustness,
we carry out our analyses based upon both monthly nominal and
monthly real returns. In each case, we study two scenarios re-
garding the portfolio's compositionwhen the portfolio includes
interest-earning deposits and when it does not include. Note that
the portfolio always includes the stock index, gold, and the foreign
currencies USD and EURO.
3.4.1. Efcient frontiers based upon monthly nominal returns on
assets
Our ndings based upon monthly nominal returns are pre-
sented in Table 4 and Fig. 4 below. Table 4 presents for a number of
optimal portfolios the associated weighting schemes, standard
deviations and expected returns: the optimal portfolios A, B, C and
D are calculated under the scenario that interest-earning deposits
are included in the portfolio; the optimal portfolios E, F, G and H
are calculated under the scenario that interest-earning deposits
are excluded. In the table, portfolios A and E are the minimum-
variance optimal portfolios under these two scenarios. Fig. 4 pre-
sents the two efcient frontiers under these two scenarios: the
line segments AD and EH are respectively the efcient frontiers
when interest-earning deposits are included in the portfolio and
when not included. The points labeled with letters on the gure
refer to the optimal portfolios presented in Table 4.
As can be seen in Table 4, under the scenario that the portfolio
includes interest-earning deposits, in the optimal portfolios more
than 90% of savings are invested in interest-earning deposits
(Ninterest) and the stock index (NBist100). The share of portfolio
gold holdings (NGold) is very moderate, in the range of 04%.
While the share of portfolio EURO holdings (NEURO) stand in the
range of 05%, that for USD (NUSD) is even lower, standing in the
range of 01%. These ndings make sense in light of the summary
statistics presented in Table 1. As can be seen in Table 1, the two
assets with greatest monthly nominal returns are the stock index
and interest-earning deposits. Therefore, the high expected-return
optimal portfolios (C and D) are dominated by these two assets. On
the other hand, the asset with the lowest standard deviation is
interest-earning deposits, followed by the two currencies. Given
the fact that the monthly expected nominal return for the interest-
earning deposits is also much higher than that for the two cur-
rencies, interest-earning deposits naturally dominate the low
standard-deviation optimal portfolios (A and B). We can also see in
Fig. 4 that, if interest-earning deposits are included in the port-
folio, then the minimum-variance optimal portfolio is A, for which
 O. Glseven, . Ekici / Resources Policy 48 (2016) 4149 47

Table 4
The compositions of optimal portfolios based upon monthly nominal returns.

Interest-earning deposits included

Portfolios Weighting factors Portfolios

Ninterest (%) NBist100 (%) NUSD (%) NEURO (%) NGold (%) Expected return (%) Std.Deviation (%)

A 90.66 1.22 0.00 4.36 3.75 2.30 2.03

B 80.05 19.95 0.00 0.00 0.00 2.40 2.96
C 44.61 55.39 0.00 0.00 0.00 2.50 6.33
D 9.17 90.83 0.00 0.00 0.00 2.60 10.05
Range 991 191 01 05 04 2.302.60 2.0310.05
Interest-earning deposits not included
E 18.59 22.82 44.40 14.20 1.80 3.56
F 27.66 7.48 0.00 64.87 2.20 4.25
G 57.76 0.00 0.00 42.24 2.40 6.25
H 95.16 0.00 0.00 4.84 2.60 10.43
Range 1896 023 045 1465 1.802.60 3.5610.43

alone dominates the portfolio's composition. The rest of the

Fig. 4. Efcient frontiers based upon monthly nominal returns. n The points labeled
with letters refer to optimal portfolios in Table 4. The line segments A-D and E-H
indicate the efcient frontiers.
the standard deviation is 2.03% and the monthly expected nominal
return is 2.30%.
We can also see in Table 4 the change in the compositions of
optimal portfolios and in Fig. 4 the shift in the efcient frontier
when interest-earning deposits are not included in the portfolio.
At a very high expected-return optimal portfolio such as H, now
that interest-earning deposits are not available, the stock index
composition of the portfolio is now exclusively gold, since its ex-
pected return is much higher than that for the two foreign cur-
rencies. At a very low standard-deviation optimal portfolio such as
E, now that interest-earning deposits are not available (the asset
whose standard deviation is lowest), the remaining four assets all
take up sizeable shares. We can also see that for the optimal
portfolios that lie in between these two extremes (such as F and G)
the asset that takes up the largest share in the composition of the
portfolio is gold. This makes sense because for gold, as compared
to the two foreign currencies, the expected return is much higher,
and as compared to the stock index, the standard deviation is
much smaller. We can also see in Fig. 4 that, in the absence of
interest-earning deposits, the minimum-variance optimal portfo-
lio is E, for which the standard deviation is 3.56% and the monthly
expected nominal return is 1.80%.
Table 4 and Fig. 4 show that when interest-earning deposits are
not a viable investment instrument, in the optimal portfolios the
gap left by them is lled to a large extent by gold. For the optimal
portfolios at the two extremes (portfolios with high expected re-
turn or low standard deviation) its share in the portfolio becomes
sizeable. For the optimal portfolios that lie in the middle gold
holdings dominate the composition of the portfolio, going up to
more than half the value of the portfolio.

Table 5
The compositions of optimal portfolios based upon monthly real returns.

Interest-earning deposits included

Portfolios Weighting factors Portfolio's

Rinterest (%) RBist100 (%) RUSD (%) REURO (%) RGold (%) Expected return (%) Std.Deviation (%)

A 90.66 1.22 0.00 4.36 3.75 0.61 1.68

B 80.05 19.95 0.00 0.00 0.00 0.71 2.58
C 44.61 55.39 0.00 0.00 0.00 0.80 5.90
D 9.17 90.83 0.00 0.00 0.00 0.89 9.52
Range 991% 191% 01% 05% 04% 0.610.89% 1.689.52%
Interest-earning deposits not included
E 19.10 39.72 31.91 9.27 0.10 3.00
F 26.08 26.16 0.00 47.76 0.40 3.48
G 57.84 0.00 0.00 42.16 0.70 5.81
H 96.89 0.00 0.00 3.11 0.90 10.10
Range 197 040 032 348 0.100.90 310.10
48 O. Glseven, . Ekici / Resources Policy 48 (2016) 4149
Fig. 5. Efcient frontiers based upon monthly real returns. n The points labeled
with letters refer to optimal portfolios in Table 5. The line segments A-D and E-H
indicate the efcient frontiers.
3.4.2. Efcient frontiers based upon monthly real returns on assets
Our ndings based upon monthly real returns are presented in
Table 5 and Fig. 5 below. Table 5 presents for a number of optimal
portfolios the associated weighting schemes, standard deviations
and expected returns: the optimal portfolios A, B, C and D are
calculated under the scenario that interest-earning deposits are
included in the portfolio; the optimal portfolios E, F, G and H are
calculated under the scenario that interest-earning deposits are
excluded. In the table, portfolios A and E are the minimum-var-
iance optimal portfolios under these two scenarios. Fig. 5 presents
the two efcient frontiers under these two scenarios: the line
segments AD and EH are respectively the efcient frontiers
when interest-earning deposits are included in the portfolio and
when not included. The points labeled with letters on the gure
refer to the optimal portfolios presented in Table 5.
The ndings we obtain based upon monthly real returns on
assets are similar to those obtained based upon monthly nominal
returns. As can be seen in Table 5, when interest-earning deposits
are not a viable investment instrument, in the optimal portfolios
the gap left by them is lled to a large extent by gold. For the
optimal portfolios at the two extremes (portfolios with high ex-
pected return or low standard deviation) its share in the portfolio
becomes sizeable. For the optimal portfolios that lie in the middle
gold holdings dominate the composition of the portfolio, going up
to as high as half the value of the portfolio. We can also see in
Fig. 5 that, if interest-earning deposits are included in the port-
folio, then the minimum-variance optimal portfolio is A, for which
the standard deviation is 1.68% and the monthly expected real
return is 0.61%. When interest-earning deposits are not included in
the portfolio, however, E is the minimum-variance optimal port-
folio, for which the standard deviation is 3% and the monthly ex-
pected real return goes as low as 0.10%.
4. Summary and policy implications
Turkish society has a long history of investment in physical
gold. Even after the establishment of the Turkish stock market in
1986, the gold appetite of Turkish investors continued unabated.
The country's investment demand for gold is exceptional, in re-
lative terms standing at many times the country's nominal GDP
and population gures.
In this paper, we hypothesized that the Turkish appetite for
gold as an investment instrument may have a rational explanation,
although an explanation that is still rooted in culture. In Turkey
the vast majority of people are Muslims, many of whom view
earning interest on money as an act that is in transgression of
Islamic rules. They rule out investing their savings in xed-income
instruments such as interest-earning deposits. Absent the viability
of xed-income investment instruments, we hypothesized that
gold may have played a pivotal role as a portfolio diversier for
these investors, soaring up the demand for gold in the country. To
test our hypothesis, we resorted to Markowitz's well-known
mean-variance model. Using monthly return data on assets from
January 1997 until August 2015, we calculated optimal investment
portfolios when the portfolio includes and does not include in-
terest-earning deposits (in addition to gold, the stock index, and
the foreign currencies USD and EURO). Our ndings provide em-
pirical support to our hypothesis: when the portfolio includes
interest-earning deposits, in an optimal portfolio the share of gold
holdings is less than a meager 4%. This share may go up to more
than 50%, however, when interest-earning deposits are excluded.
We believe that the primary factor driving our pronounced
ndings has been Turkey's historically very high rates and volati-
lity of ination. Interest-earning deposits are naturally a very at-
tractive investment instrument in such an environment. Their
returns are highly correlated with the rate of ination and hence
help investors both shield theirs savings against inationary de-
preciation and manage portfolio risks caused by the volatility in
the rate of ination. For those investors who abstain from xed-
income investment instruments, however, this dual role interest-
earning deposits play must be assumed by the remaining portfolio
assets. Investing savings in gold then becomes a valuable option.
The return on gold is positively correlated with the rate of ina-
tion. Also, as compared to the foreign currencies gold has a high
expected return, and as compared to the very volatile Turkish
stock market the return on gold is much less volatile.
With its historically low savings rates, Turkey is a country that
relies heavily upon foreign capital to nance its growth-stimulat-
ing capital investments. The fact that Turkish households invest a
signicant portion of their savings in a non-productive investment
instrument as gold (much of which is held at home) is a primary
concern for Turkish policy makers. In light of our paper's ndings,
to divert household savings away from gold into more productive
uses, we recommend that the authorities introduce into the -
nancial system Sukuk. Widely-used in other Muslim-majority
countries (and sometimes referred to as Islam-compliant bonds),
Sukuk securities are structured in a way so that they do not pay
interest. Instead, Sukuk holders receive partial ownership of a
property whose purchase is nanced by their savings. The prot
generated from the operation of this property is then collected by
Sukuk holders as rent payments. This practice is widely accepted
as not in violation of Islamic rules.
In Turkey capital is highly valuable and many development
projects are delayed for lack of funding. The popularization of
Sukuk in Turkey can therefore play a dual role by providing a
means to nance such development projects at reasonable nan-
cing terms and offering small investors an alternative means
where they can utilize their savings.
References
Abken, Peter A., 1980. The economics of gold price movements. FRB Richmond
Econ. Rev. 66 (2), 312.
Apergis, Nicholas, 2014. Can gold prices forecast the Australian dollar movements?
Int. Rev. Econ. Financ. 29, 7582.
Ariovich, G., 1983. The impact of political tension on the price of gold. J. Stud. Econ.
Econometr. 16, 1737.
Batchelor, Roy, David, Gulley, 1995. Jewellery demand and the price of gold. Resour.
Policy 21, 3742.
Batten, Jonathan A., Ciner, Cetin, Lucey, Brian M., 2014. On the economic de-
terminants of the gold-ination relation. Resour. Policy 41, 101108.
Baur, Dirk G., 2011. Explanatory mining for gold: contrasting evidence from simple
and multiple regressions. Resour. Policy 36, 265275.
 O. Glseven, . Ekici / Resources Policy 48 (2016) 4149
Baur, Dirk G., 2013. The autumn effect of gold. Res. Int. Bus. Financ. 27 (1), 111.
Baur, Dirk G., Thomas, K. Mc.Dermott, 2010. Is gold a safe haven? International
evidence. J. Bank. Financ. 34, 18861898.
Baur, Dirk, Brian, M., Lucey, 2010. Is gold a hedge or a safe haven? An analysis of
stocks, bonds and gold. Financ. Rev. 45, 217229.
Bredin, Don, Conlon, Thomas, Pot, Valerio, 2015. Does gold glitter in the longrun?
Gold as a hedge and safe haven across time and investment horizon. Int.
Rev. Financ. Anal. 41.
Brenson, J., 2014. A study on saving habits of households in gold ornaments in India.
Asia Pac. J. Res. I (XII).
Bruno, Salvotore, Chincarini, Ludwig, 2010. A historical examination of optimal real
return portfolios for non-US investors. Rev. Financ. Econ. 19, 161178.
Capie, Forest, Terence, C. Mills, Geoffrey, Wood, 2005. Gold as a hedge against the
dollar. J. Int. Financ. Mark. Inst. Money 15, 343352.
Chua, Jess H., Gordon, Sick, Woodward Richard, S., 1990. Diversifying with gold
stocks. Financ. Anal. J. 46, 7679.
Emmrich, Ole, Frank, Mc.Groarty, 2013. Should gold be included in institutional
investment portfolios? Appl. Financ. Econ. 23, 15531565.
Fortune, J. Neill, 1987. The ination rate of the price of gold, expected prices and
interest rates. J. Macroecon. 9, 7182.
Hillier, David, Draper, Paul, Faff, Robert, 2006. Do precious metals shine? An in-
vestment perspective. Financ. Anal. J. 62, 98106.
Hood, Matthew, Malik, Farooq, 2013. Is gold the best hedge and a safe haven under
changing stock market volatility? Rev. Financ. Econ. 22, 4752.
Jaffe, Jeffrey F., 1989. Gold and gold stocks as investments for institutional portfo-
lios. Financ. Anal. J. 45, 5359.
Joy, Mark, 2011. Gold and the US dollar: hedge or heaven? Financ. Res. Lett. 8 (3),
120131.
Kannan, R., Dhal, Sarat, 2008. Indias demand for gold: some issues for economic
development and macroeconomic policy. Indian J. Econ. Bus. 7 (1), 107.
Koutsoyiannis, A., 1983. A short-run pricing model for a speculative asset, tested
49
with data from the gold bullion market. Appl. Econ. 15, 563582.
Lucey, Brian Ml, Poti, Valerio, Tully, Ede, 2006. International portfolio formation,
skewness and the role of gold. Front. Financ. Econ. 3, 4968.
Markowitz, Harry, 1952. Portfolio selection. J. Financ. 7 (1), 7791.
Nabil, A. Saleh, 1986. Unlawful Gain and Legitimate Prot in Islamic Law: Riba,
Gharar, and Islamic Banking. Cambridge University Press, Cambridge.
OConnor, Fergal A., Lucey, Brian M., Batten, Jonathan A., Baur, Dirk G., 2015. The
nancial economics of gold a survey. Int. Rev. Financ. Anal. 41, 186205.
Ming, Qi, Wang, Wenyao, 2013. The monthly effects in Chinese gold market. Int. J.
Econ. Financ. 5 (10), 141.
Reboredo, Juan C., 2013. Is gold a safe haven or a hedge for the US dollar? Im-
plications for risk management. J. Bank. Financ. 37, 26652776.
Reboredo, Juan C., Rivera-Castro Miguel, A., 2014. Can gold hedge and preserve
value when the US dollar depreciates? Econ. Model. 39, 168173.
Sanderson, H., 2015. Gold rises amid expectation of ECB move on QE. Financ. Times.
Sjasstad, Larry A., 2008. The price of gold and the exchange rates: Once again.
Resour. Policy 33, 118124.
Starr, Martha, Tran, Ky, 2008. Determinants of the physical demand for gold: evi-
dence from panel data. World Econ. 31, 416436.
Sumner, Steven, Johnson, Robert, Soenen, Luc, 2010. Spillover effects among gold,
stocks, and bonds. J. Cent. Cathedra 3, 106.
Suresh, Anli, 2011. A review on gold quest in the investment portfolio by Indian
investors. IRACST- Int. J. Res. Manag. Technol. 1 (2), 128134.
World Gold Council, 2013. Turkey: gold in action.
Vuyyuri, Srivyal, Ganesh, S. Mani, 2005. Gold pricing in India: an econometric
analysis. J. Econ. Res. 16 (1).
Wanga, Kuan-Min, Lee, Yuan-Ming, 2011. The yen for gold. Resour. Policy 36 (1),
3948.
Yang, Lu, Hamori, Shigeyuki, 2013. Gold prices and exchange rates: a time varying
copula analysis. Appl. Financ. Econ. 24, 4150.