CC BY
37
GOLD AS THE SAFEST INVESTMENT IN TIME OF FINANCIAL CRISIS
Abstract. This paper tests whether macroeconomic variables such as U.S. inflation, the change in the U.S. industrial production index, the return on virtually risk-free long-term Treasury bonds, the return on short-term Treasury bills and market risk premium influence the returns on gold in the past ten years. We find that U.S. inflation is a significant factor at explaining the gold returns, moreover, the rise in inflation by 1 per cent causes, on average, greater rise in the gold returns. Furthermore, neither the market risk premium nor the change in U.S. industrial production index can be considered as statistically significant factors in explaining the returns on gold. These findings fit the idea of gold being inflation hedge, zero-beta and safe haven asset in the last ten years in the USA.
Keywords: hedge, zero-beta, safe haven, gold, financial crisis
Leonid Mindyuk,
3rd year student NRU HSE
Introduction
Gold is both commodity and asset at the same time. On the one hand, gold is used in jewelry, electronic, dentistry and chemical industries as a raw material. On the other hand, it is often used as an alternative investment for a long period of time . The most traditional and popular ways to invest in gold is through gold bars, coins and through exchange-traded funds (ETFs) . Nowadays mainly the investment usage is the main driving force of the gold price, especially in the time of a business slowdown The main reason for this trend is that gold is very different from other, so called mainstream assets, such as stocks and bonds .
Gold possesses few properties which distinguish it from other assets. It is liquid, nonperishable and portable . It can be easily sold and bought rapidly without loosing the great part of its value Gold is the only metal which will never rust or tarnish as it forms no oxide film in air at normal temperatures . Bars of gold or gold coins — the main forms of gold storage can be easily transported from one place to another and divided into smaller pieces
However, the most important feature of the gold is that it is often used as a store of value as it maintains the same purchasing power over the long time. S . Harmston (1998) shows that there were only minor fluctuations in the purchasing power of gold from its long run
trend Thus gold, to some extent, can be considered as an asset which does not lose its purchasing power over the long period of time For example, from historical data it can be calculated that: In 1850 the price of gold was $18 . 93 per ounce, about $428 in 2004 terms, while in 2004 the actual price of gold was $409.72. $20 is a really small change in the real price of gold over a period of one hundred and fifty four years. These calculations are made according to the inflation adjustments over this period of time
Is gold return is highly affected in the time of financial crisis or it is following its long-term trend? This research paper tries to provide evidence about gold being the safest investment in time of financial crisis .
Many investors' portfolio includes gold, in various forms, as a great portion of them believe that gold possesses inflation hedge, good diversifier, zero-beta and safe haven properties . Baur and Lucey (2009) provided the academic world with very useful definitions of those properties Their definitions, mainly modified, are used as a base for our Theoretical Framework section However, despite of gold popularity among investors the little number of research was made on the gold performance in comparison with stocks and bonds Moreover, almost no research papers were modeling the gold returns in the time of financial crisis The performance of gold as an investment is examined in this study using the multi-factor model which outlines the relationship between returns on gold and main macroeconomic indicators which are believed to be the main driving forces of gold price. This paper analyses different factors, which affected gold returns The purpose is to identify correlation and dependence between these factors and gold returns
At first the identification and economic intuition about the relationship between each of factors chosen and gold returns is provided in the Theoretical Framework section. Then, the data needed for the modeling process is explained and the econometric model is introduced. Finally, based on the econometric results, the conclusion and identification of factors which strongly affect gold returns and which do not influence the gold returns are made. Thus this paper helps us to conclude on gold being a zero-beta asset, safe haven and hedge against inflation in the period of past ten years in the USA . The meaning of all these properties is explained in the Theoretical Framework section
Theoretical framework
The theoretical hypothesis, which can be considered as the main testing aim at this research project can be formulated as follows. Adding gold as an asset to the investment portfolios can reduce losses or provide portfolio stability in times of financial turmoil because it has a specific nature which differentiates it from stocks and bonds According to different financial characteristics any asset which can help an investor to overcome market instability (e. g. financial crisis) can be divided in four distinct groups
Safe haven asset:
Taking as a granted that an investor's utility level increases with the reduction in losses, an asset which minimizes the financial looses in the period of business slowdown is defined as a safe haven asset in the vast majority of academic works in this field Thus this research considers the safe haven asset as the one which returns uncorrelated or negatively correlated with the industrial production index, which is believed to change significantly in the period of economic instability as it is one of the most popular indicator of production level in the economy
Inflation Hedge:
A hedge against inflation is defined as an asset that is positively correlated with the inflation level. Moreover, returns on such asset should, on average, rise higher than inflation level Thus an asset with an inflation hedge property is believed to provide an opportunity not to lose real value of an asset because of the increase in inflation. In a number of academic works it is believed that gold can play a role of inflation hedge because of high positive coefficient between gold return and inflation level. This research paper tries to outline whether this argument holds with the latest data available
Diversifier:
An asset can be considered as a diversifier if it is positively or negatively and not perfectly correlated with another asset in the investor's portfolio or portfolio on average
The diversifier does not have the specific property of reducing losses in the time of financial turmoil since the correlation property is only required to hold on average Thus we are not interested in testing this specific property of gold and this hypothesis can be tested in the further researches after the crisis time
Zero-beta asset:
A zero-beta asset is defined as an asset that is un-correlated with the market Thus such an asset is believed to bear no significantly different from zero market risk. According to CAPM such an asset should provide the same return as a risk-free asset . This hypothesis can be tested using the same multi-factor model which helps us to testing both inflation hedge and safe haven role of gold
Therefore, the econometric multiple regression model should be introduced in a such way that safe haven, inflation hedge and zero-beta properties can be tested using the same multi-factor model The model is based on the hypothesis that the price of gold is primarily driven by major macroeconomic factors The factors that will be taken are U. S . Consumer Price Index for All Urban Consumers during period, U S industrial production index, return on long-term Treasury bonds, return on short-term Treasury bills and market risk premium However, the gold prices and indices are non-stationary series, so it is important to modify them into new variables which represent stationary series Having proposed a set of relevant variables, we are specifying on their description and series transformations in the Empirical Analysis section while constructing the appropriate regression model
Empirical Analysis
The data used in modeling are monthly observations covering the ten years period from April 2002 to March 2012. This period clearly includes the last financial crisis. Moreover, ten years period and monthly observations allow us to obtain more reliable results because of rather high degrees of freedom The model which best fits our aims is a multiple regression model because it enables to simultaneously and more precise test hedge against inflation, safe haven and zero-beta properties of gold This model helps to explain relationship between independent or explanatory variables and a dependant variable In our case,
Table 1
Glossary and Definitions of variables
Mnemonic Variable Definition Source
Basic Series
PG Price of Gold End-of-period spot price of gold Index Mundi
CPI Consumer Price Index U.S. Consumer Price Index for All Urban Consumers during period CPI-U Info
INDPI Industrial Production Index U.S. Industrial Production Index during period Federal Reserve System
LGB Long-term Government Bonds Return on U.S. 10-year Treasury Bonds, per cent per annum OECD Stat
TB Treasury-bill Rate Return on U.S. 3-monts Treasury Bills, per cent per annum Federal Reserve System
SP S&P 500 Index End-of-period Standard & Poor's 500 stockmarket index Yahoo Finance
Derived Series
RGi Monthly Return on Gold In (PGi/PGi-1) PG
Ii Inflation In (CPIi/CPIi-1) CPI
MPi Monthly Growth in Industrial Production In (INDPIi/INDPIi-1) INDPI
MLGBi Monthly Change in LGB (LGBi - LGBi-1)/12 LGB
MTBi Monthly Change in TB (TBi - TBi-1) /12 TB
MRPi Market Risk Premium (In (SPi/SPi-1) - LGBi-1/12) SP, LGB
the dependant variable is the return of gold, while the explanatory variables are: U. S . inflation, monthly growth in U.S. industrial production index, monthly change in the return on U.S. long-term Treasury bonds, monthly change in the return on U. S . short-term Treasury bills and market risk premium .
Thus the equation of a multiple regression model is the following:
RGi = P0 + p±* Ii + P2* MR + P3* MLGBi +
+ p4* MTBi + P5* MRPi + ei (1)
The Table 1 below provides us with the glossary, definitions of variables and derived series
The logarithmic change in price of gold is taken as an approximation for gold return. The U. S . inflation is modeled as the logarithmic change in the U. S. CPI-U. Logarithmic relative of industrial production index is taken as an approximation of monthly growth in U S industrial production. Both changes in return of long-term U S Treasury bonds and return of short-term U. S . Treasury bills are calculated and divided by 12 in order to calculate per cent per month, not per annum Market risk premium is calculated using Standard & Poor's 500 stock-market index and normalized U. S. Treasury Bills. Use of the natural logarithms
allows us to calculate influence of change in different factors on the gold returns over a period of time and use econometrics results to explain relationship between different variables. Idiosyncratic error term accounts for other variables, which might be significant, but which are not included into the model and for the random events which can influence the gold returns Thus assumption that idiosyncratic error terms have the same independent normal distribution with zero mean and o standard deviation is made . This also means that model may be improved by including additional both economic and fundamental factors into it . However, it is not obligatory here as the hypotheses of inflation hedge, zero-beta and safe haven properties of gold can be tested using this regression model
Based on our results we can conclude that gold possesses all the properties that were expected . Firstly, the positive relationship between the gold returns and the U. S . inflation was found . The coefficient before first explanatory variable which stands for the U S inflation is 2.384529, meaning that, on average, the rise in the U. S. inflation by 1 per cent per month causes the rise in the monthly gold returns by 2 384529 per cent Thus the gold plays a hedge against inflation function on 1
Table 2
Observed Relationships
Factor Relationship to the gold return Explanation
U.S. inflation Positive (>1) Gold returns should, on average, rise more than inflation (gold as an inflation hedge)
Growth in U.S. industrial production index Uncorrelated Gold returns should be uncorrelated with production level in the economy (gold as a safe haven)
Change in return on long-term government bonds Negative Gold can be considered as a substitute for long-term virtually risk-free investments
Change in return ob short-term Treasury bills Negative Gold can be considered as a substitute for short-term Treasury bills
Market risk premium Uncorrelated Gold returns should be uncorrelated with market risk premium (gold as a zero-beta asset)
per cent significance level. Secondly, the econometric results prove that the business cycle is an insignificant factor in explaining the gold returns, as the change in the U.S . industrial production index, which was taken as a reliable indicator of the business cycle in the economy, was found to be an insignificant factor in our regression model. Also, the historical market risk premium was found to be insignificant in explaining the gold returns . These findings prove that gold can be considered both a safe haven and a zero-beta asset in the time of financial instability as we can not reject the null hypotheses about the zero P2 and zero P5 respectively (see the Regression: RGi = P0 + p±* Ii + P2* MPi + P3* MLGBi + P4* MTBi + P5* MRPi + ei. ). Finally, it is necessary to emphasize that the change in monthly long-term interest rate obtained as (LGBi — LGBi-±) /12 was found to be a significant factor on 1 per cent significance level in explaining the returns on gold and the coefficient before this factor is -0.592902 in our model, meaning that the long-term U S government bonds and gold can be considered as the substitutes, proving that gold is also believed to be a long-term way of virtually risk-free investment . However, it is not the same for short-term investments as the return on U.S. short-term T-bills was found to be insignificant in explaining the gold returns
Oonclusions
The purpose of this study was to identify factors and the extent to which they can affect the gold returns Thus we were aimed at testing the hypotheses about gold being a zero-beta asset, safe haven and hedge against inflation in the period of past ten years in the USA.
We found that both economic cycle and market risk premium are insignificant in explaining gold returns . Which means that gold is a very different from
mainstream assets such as stocks and bonds as it possesses both zero-beta and safe haven properties
The other key finding is that gold can be considered as a good hedge against inflation asset as it does not loose its purchasing power in the period of rising U S inflation
The interesting findings are the absence of the same relationships between virtually risk-free returns on long-term Treasury bonds and gold returns and returns on short-term Treasury bills and gold returns. Moreover, gold can be considered as a substitute for virtually risk-free assets such as U S government securities only in long run
To sum up, it is necessary to emphasize that in general findings of our research were expected. The use of the latest data for ten years and the econometric techniques allow us to conclude on gold possessing zero-beta asset, safe haven and hedge against inflation properties in the period of past ten years in the USA However, our model of gold returns can be improved in the process of further research as there are extra variables which also may be considered as factors influencing the return on gold .
References
1. Baur, D. G. and B. M. Lucey (2009) "Is gold a hedge or a safe haven? An analysis of stocks, bonds and gold", The Financial Review, Volume 45, Issue 2, pp . 217-229.
2. Ismail, Z . , A. Yahya and A . Shabri (2009) "Forecasting Gold prices using multiple linear regression method", American Journal of Applied Science, Volume 6, Issue 8, pp. 1509-1514.
3. Greely D. and J. Currie (2009) "Forecasting Gold as a Commodity", Goldman Sachs Global
Economics, Commodities and Research, paper № 183 .
4. Harmston, S . (1998) "Gold as a Store of Value", London, World Gold Council, Research Study № 22, November.
5. Lawrence, C. (2003) "Why is Gold Different from Other Assets? An Empirical Investigation", London, World Gold Council, Publications, March.
6. Levin, E . J. and R. E. Wright (2006) "Short-run and Long-run Determinants of the Price of Gold", London, World Gold Council, Research Study № 32,June .
7. McCown, J. R. and J. R. Zimmerman (2006) "Is Gold a Zero-Beta Asset? Analysis of the Investment Potential of Precious Metals", Working Paper Series, Available at SSRN: http://ssrn .com/ab-stract=920496
8. Sherman, E. J. (1983) "A Gold Pricing Model", Journal of Portfolio Management, New York: Spring 1983. Volume 9, Issue 3, pp. 68-70.
9. Tckaz G. (2007) "Gold prices and inflation", Bank of Canada, Working Paper № 35 .
10. World Gold Council: The Value of Gold to Society, http://www. gold . org .
Technical Appendix Appendix sources of Data
Variable Mnemonic Source Link
Price of gold PG Index Mundi http://www.indexmundi.com/commodities/?c ommodity=gold&months=120
CPI-U CPI CPI-U Info http://www.cpi-u.info/Historical-CPI-U-Data.aspx
Industrial production index INDPI Federal Reserve http://www.federalreserve.goV/releases/G17/d ata.htm
Long-term Treasury bonds LGB OECD stat http://stats.oecd.org/Index.aspx? DataSetCode=REFSE RIES
Short-term Treasury bills TB Federal Reserve http://www.federalreserve.goV/releases/H15/d ata.htm
S&P 500 Index SP Yahoo Finance http://finance.yahoo.com/q/hp?s=AGSPC&a=00&b=1&c =2002&d=04&e=1&f=2012&g=m&z=66&y=66
Note that series used in modeling process are transformations of the variables above . You can see the Table 2 for the description of derived series .
appendix 2 Regression: RGi = ß0 + ß1* Г + ß2* MR + ß3* MLGBi + ß4* MTBi + ß5* MRR + ei
Dependent Variable: RGi
Variable Coefficient Estimation
Constant 0.024183 [0.015291]
Ii 2.384529 *** [0.886331]
MPi 0.562560 [0.468991]
MLGBi -0.592902 *** [0.200384]
MTBi -0.048895 [0.244210]
MRPi 0.050797 [0.045486]
Standard Errors in Square Brackets * p<0.1,** p<0.05, *** p<0.01
Source: EVIEWS 6. 0
Appendix 3 EviEWs 6.0 Output
Dependent Variable: RGi Method: Least Squares Date: 06/25/12 Time: 16:28 Sample (adjusted): 2 120 Included observations: 119 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.024183 0.015291 1.581524 0.1166
li 2.384529 0.876331 2.721038 0.0075
MPi 0.562560 0.468991 1.199511 0.2328
MLGBi -0.592902 0.200384 -2.958825 0.0038
MTBi -0.048895 0.244210 -0.200216 0.8417
MRPi 0.050797 0.045486 1.116744 0.2665
R-squared 0.117936 Mean dependent var 0.014371
Adjusted R-squared 0.078907 S.D. dependentvar 0.041542
S.E. of regression 0.039869 Akaike info criterion -3.557321
Sum squared resid 0.179619 Schwarz criterion -3.417197
Log likelihood 217.6606 Hannan-Quinn criter. -3.500421
F-statistic 3.021736 Durbin-Watson stat 1.914708
Prob (F-statistic) 0.013445
Source: EVIEWS 6 . 0 appendix 4 Correlation Matrix
RGi Ii MPi MLGBi MTBi MRPi
RGi -0.019908 0.201413 0.199275 0.297190 1.000000 0.115698
Ii -0.187765 0.265177 0.132109 1.000000 0.297190 0.227471
MPi 1.000000 0.191383 0.078297 -0.187765 -0.019908 0.068465
MLGBi 0.068465 0.136220 -0.003350 0.227471 0.115698 1.000000
MTBi 0.078297 0.046340 1.000000 0.132109 0.199275 -0.003350
MRPi 0.191383 1.000000 0.046340 0.265177 0.201413 0.136220
Source: EVIEWS 6 . 0
Appendix 5 Regression for White Test
Dependent Variable: RESIDA2 Method: Least Squares Date: 06/25/12 Time: 17:29 Sample: 2 120 Included observations: 119
Variable Coefficient Std. Error t-Statistic Prob.
C 0.003884 0.002458 1.579899 0.1174
CPI1 -0.396053 0.389861 -1.015881 0.3122
CPI1A2 4.716706 9.723884 0.485064 0.6287
CPI1*INDPI1 10.03716 7.742884 1.296308 0.1979
CPI1*R1 0.719063 3.986303 0.180383 0.8572
CPI1*RF1 -1.082073 3.799283 -0.284810 0.7764
CPI1*MRP -1.028999 1.066470 -0.964865 0.3370
INDPI1 -0.168273 0.167573 -1.004173 0.3178
INDPI1A2 0.611037 2.859146 0.213713 0.8312
INDPI1*R1 -0.821416 1.906167 -0.430926 0.6675
INDPI1*RF1 -3.152160 2.278005 -1.383737 0.1696
INDPI1*MRP -0.237331 0.508612 -0.466625 0.6418
R1 -0.045310 0.061906 -0.731907 0.4660
R1A2 -0.212872 0.441489 -0.482169 0.6308
R1*RF1 1.703373 1.450986 1.173942 0.2433
R1*MRP -0.181443 0.196520 -0.923276 0.3581
RF1 -0.093518 0.147538 -0.633855 0.5277
RF1A2 -0.590663 0.626091 -0.943414 0.3478
RF1*MRP -0.302886 0.402877 -0.751807 0.4540
MRP 0.018061 0.016437 1.098784 0.2746
MRPA2 0.034517 0.027340 1.262510 0.2098
R-squared 0.144268 Mean dependent var 0.001509
Adjusted R-squared -0.030371 S.D. dependent var 0.002448
S.E. of regression 0.002484 Akaike info criterion -8.998738
Sum squared resid 0.000605 Schwarz criterion -8.508305
Log likelihood 556.4249 Hannan-Quinn criter. -8.799589
F-statistic 0.826093 Durbin-Watson stat 1.815558
Prob (F-statistic) 0.676971
Source: EVIEWS 6. 0
