Which precious metals spill over on which, when and

Applied Economics Letters, 2014 http://dx.doi.org/10.1080/13504851.2014.950789

Which precious metals spill over on which, when and why? Some evidence Jonathan A. Battena, Cetin Cinerb and Brian M. Luceyc,d,e,*

Downloaded by [Monash University Library] at 14:31 04 December 2014

a

Department of Banking and Finance, Monash University, Clayton, Victoria 3145, Australia b Cameron School of Business, University of North Carolina— Wilmington, Wilmington, NC, USA c School of Business and Institute for International Integration Studies (IIIS), Trinity College, Dublin, Ireland d Glasgow Business School, Glasgow Caledonian University, Glasgow G4 0BA, UK e Faculty of Economics, University of Ljubljana, Ljubljana, 1000, Slovenia

Much academic and investor analysis and commentary see the four main precious metals as a single market, integrated and to some degree with each metal a substitute for the other. This proposition, which can be explicit or implicit, can be challenged on economic and statistical grounds. Using the Diebold and Yilmaz (2009) methodology, we show that the market is only weakly integrated, that this degree of integration is time varying and that it differs as between returns and volatility. Keywords: gold; silver; platinum; palladium; integration JEL Classification: C01; F49; G12; G15 I. Introduction Commodity markets, in particular those for precious metals, are large and important. Obtaining an accurate size of these markets is not simple – the largest part of them is typically settled Over-the-counter (OTC). One estimated size (in gross traded volume per day) of the gold markets ranges, for 2012, between $150 billion and $350 billion, for silver between $28 and $54, for platinum between $2 and $4 billion and for palladium between $0.5 and $1 (see Lubber, 2013). Comparing these figures for

2012 US daily trade in treasuries of just over $500 billion and for US corporate debt of $16 billion we see that these markets are not trivial. OTC derivatives outstanding in Gold (as of mid-2013, via the Bank for International Settlements database) amount to a gross figure of some $461 billion; this is about one-fifth of all commodity derivatives. London Bullion Market Association clearing data suggest an approximate daily turnover for physical gold and silver of $30 billion and $2–$4 billion, respectively. Regardless of which figure one uses, these markets are not small.

*Corresponding author. E-mail: [email protected]

© 2014 Taylor & Francis

1

J. A. Batten et al.


2 Our goal in this note was to investigate the extent to which these metals can be considered an asset class. An asset class should, it is generally agreed, amongst other features show a high degree of integration, arising from common shocks and common economic fundamentals, an alignment of forces that work on all the assets (see Greer, 1997). Commodities matter to investors, both for their size as markets and for their diversification potential in portfolio management. For commodities, in general, there is a wide literature on the diversification benefits. For financial researchers, commodities are also of interest for their potential role in asset allocation decisions. In this regard, recent papers by Abanomey and Mathur (2001), Georgiev (2001) and Chan and Young (2006) argue that commodities provide risk reduction in portfolios along with stocks and bonds. Both Caglayan and Edwards (2001) and Chow et al. (1999) suggest that commodities are in fact more attractive when the general financial climate is negative. These findings are confirmed in more recent works by Boido (2013), Cumming et al. (2013), Nijman and Swinkels (2008) and Skiadopoulos (2012). While there is a significant body of work on the macroeconomic determinants of volatility in and between equity and commodity markets (e.g. Fernandez, 2008) and groups of commodities (e.g. Brunetti and Gilbertm 1995; Kroner et al., 1995; Pindyck, 2004; Gilbert, 2006), there is limited evidence for relationships between precious metals. If similar macroeconomic or behavioural or other factors are important for all of these, it could be argued that they constitute a market. As to whether commodities as a whole are an asset class, the evidence is mixed. Two highly cited papers, Erb and Harvey (2006) and Gorton and Geert Rouwenhorst (2006), provide analyses of the risk–return trade-off in commodity markets. They differ as to whether commodities should be seen as a single asset class. The conclusions of Gorton and Geert Rouwenhorst (2006) imply that commodities can be best viewed as a single asset class that have attractive risk–return patterns and, furthermore, are useful for portfolio diversification. Erb and Harvey (2006), however, question whether commodity markets can actually be considered as a single asset class since differences in the behaviour of prices between individual commodities seem significant. More recently, Adams et al. (2011) and Batten et al.

(2010) investigate the macroeconomic determinants of returns and volatility, respectively. They conclude that in the main the commodities under investigation are exposed to different macroeconomic impacts and so cannot be considered as an asset class. Spillovers and integration in the precious metal market have also been considered by Hammoudeh et al. (2010) in a GARCH framework, concluding that gold and silver volatility both respond to monetary shocks and that these are quite persistent. This long memory property of gold is confirmatory to the findings by Byers and Peel (2001). II. Methodology The model presented in this article is drawn from Diebold and Yilmaz (2009, 2012). These authors look at return and volatility spillovers using a vector autoregressive models (VARs) following Engle et al. (1990) but concentrate on variance decompositions. This gives one measure of spillover based on spillovers from a number of markets. A recent application of this approach in the precious metal market is Lucey et al. (2014) who note the still limited impact of markets other than London and New York on gold price formation. Consider a set of assets. For each asset i, the shares of its forecast error variance coming from shocks to asset j, for all j ≠ i, are summed. These are then added for all i = 1, …, N. Considered as a covariance stationary first-order two-variable VAR, it can be written as xt ¼ xt1 þ εt

(1)

where xt ¼ ðx1t x2t Þ and is a 2 × 2 parameter matrix. In this article, xt represents either a vector of gold returns or gold returns volatilities. Diebold and Yilmaz (2009) show that by covariance stationarity we can represent this VAR as a moving average given by Equation 2 below. xt ¼ ð LÞεt

(2)

where ð LÞ ¼ ðI LÞ1 : This can be rewritten as below for ease. xt ¼ Að LÞμt

(3)

Which precious metals spill over on which, when and why? 0 where A(L) = ð LÞQ1 t , μt ¼ Qt εt ; E ðμt μ t Þ ¼ I and 1 Qt is a unique lower triangular Cholesky factor of the covariance matrix of εt . The Wiener–Kolmogorov linear least-squares forecast is an optimal one-step-ahead forecast from the above which can be shown as:

xtþ1;t ¼ xt

(4)

with an one-step-ahead error vector:


etþ1;t ¼ xtþ1 xtþ1;t ¼ A0 μtþ1 a0;11 a0;12 μ1;tþ1 ¼ a0;21 a0;22 μ2;tþ1 which has a covariance matrix of: E etþ1;t e0 tþ1;t ¼ A0 A00 :

(5)

(6)

Using this approach allows us to say what proportion of the error variance in forecasting any particular x (a specific gold market, e.g., London) is due to shocks to itself, or spillover from shocks to another market, for example New York. Diebold and Yilmaz (2009, p. 159) define own variance shares as ‘to be the fractions of the 1-step-ahead error variances in forecasting xi due to shocks to xi’ for all i and cross-variance shares, or spillovers, to be ‘the fractions of the one-step-ahead error variances in forecasting xi due to shocks to xj′ for all j, where i = j. In this two-variable illustration, the variance of the one-step-ahead error in forecasting x1 at time t is then a20;11 þ a20;12 from Equation 5. a20;12 can then be thought of as a x1,t’s spillover that effects the forecast error variance of x2,t and a20;21 can then be thought of as a x2,t’s spillover that effects the forecast error variance of x1,t. Total spillover is then a20;12 þ a20;21. Using these, we can calculate a spillover index measure as total spillover divided by the total forecast error varia tion (a20;11 þ a20;12 þ a20;21 þ a20;22 ¼ trace A0 A00 as in Equation 7. S¼

a20;12 þ a20;21 100 traceðA0 A0 0 Þ

(7)

This first-order two-variable case can be generalized into a pth-order N-variable case using one-stepahead forecasts giving:

3 PN

2 i;j¼1 a0;ij

S¼

100 trace A0 A00 iÞj

(8)

For H-step-ahead forecasts: PH1 PN 2 h¼0 i;j¼1 ah;ij iÞj 100 S ¼ PH1 0 h¼0 traceðA0 A 0 Þ

(9)

While it is commonplace to measure asset volatility based on the SD of the log difference across a regular time interval, we also utilize a more complex measure, the GKe measure, which incorporates information about the open, close, high and low prices within a particular time interval. As discussed in Molnár (2012), the GK range-based estimator is amongst the most efficient estimators for volatility estimation. From Garman and Klass (1980), the GKe is GKe ¼ σ 2 ¼ 0:511 ðH LÞ2 0:019 ðC 0Þ ðH þ L 2CÞ

(10)

2

ð1 CÞ 0:383 ðC OÞ where H = log of interval high L = log of interval low O = log of interval open C = log of interval close III. Data

We examine the four main precious metals – gold, silver, platinum and palladium. Data are drawn from Thomson Reuters Ecowin and consist of the nearest month futures closing price for each commodity, on New York Mercantile Exchange. Data are initially collected on a daily basis and converted to weekly both to facilitate a smoother data series and to allow the calculation of the Garman and Klass (1980) volatility. Shown in Fig. 1 is the evolution of the series. IV. Results Show in Table 1, are the results of the Diebold– Yilmaz spillover analyses for the four assets. We note that gold and silver share the closest relationship. Gold accounts for 27.7% of silvers return

J. A. Batten et al.

4 Gold 2500

Precious metal weekly closing prices $US

Palladium

60

Platinum Silver 50 2000

40

30

Silver

1500

1000 20

500

10

0 Date 3 September 1982 8 April 1983 11 November 1983 15 June 1984 18 January 1985 23 August 1985 28 March 1986 31 October 1986 05 June 1987 08 January 1988 12 August 1988 17 March 1989 20 October 1989 25 May 1990 28 December 1990 02 August 1991 06 March 1992 09 October 1992 14 May 1993 17 December 1993 22 July 1994 24 February 1995 29 September 1995 03 May 1996 06 December 1996 11 July 1997 13 February 1998 18 September 1998 23 April 1999 26 November 1999 30 June 2000 02 February 2001 07 September 2001 12 April 2002 15 November 2002 20 June 2003 23 January 2004 27 August 2004 01 April 2005 04 November 2005 09 June 2006 12 January 2007 17 August 2007 21 March 2008 24 October 2008 29 May 2009 01 January 2010 06 August 2010 11 March 2011 14 October 2011 18 May 2012 21 December 2012


0

Fig. 1. Precious metal prices 1982–2003

Table 1. Diebold–Yilmaz spillover estimates

To

To

Returns Gold Silver Platinum Palladium Contribution to others Contribution including own Volatility Gold Silver Platinum Palladium Contribution to others Contribution including own

From Gold

Silver

Platinum

Palladium

From others

51.1 27.7 17.8 6.9 52 103

27.5 51.5 14.9 9.9 52 104

16.4 13.5 56 14.3 44 100

5.1 7.3 11.3 69 24 93

49 49 44 31 173 Spillover index 43.10%

70.4 27.6 1.4 4.9 34 104

25.4 69.4 0.4 3.4 29 99

0.8 0.3 97.3 0.9 2 99

3.3 2.7 0.8 90.9 7 98

30 31 3 9 72 Spillover index 18.00%

and only a small percentage of that of platinum and less again of palladium. Silver contributes 27.5% of the gold return and similar percentages to that of gold in terms of return to platinum and palladium. In volatility terms, the situation is starker – we can see significant spillovers from gold and silver to each other, while platinum and palladium are almost insulated from each other. Overall, there is a reasonable level of spillover in returns, 42%, while for volatility this is much lower at 18%. We can interpret this as 43% of return being determined by the movements of the

four assets, while only 18% of volatility is so determined. We then proceed to a time-varying analysis. Below we show the evolving importance of spillovers to returns between the four markets through the spillover index in Fig. 2 for returns and Fig. 3 for volatility. We use an initial window of 100 weeks and then update this by 10 observations (weeks) each iteration. While the average result for the returns spillover index over the full sample given in the last section was 43%, we can see that this varies very substantially over time. In the period 1998 to 2005,

Which precious metals spill over on which, when and why?

5

70

60

50

40

30

20

10


1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 Spillover plot. Returns. 100-week window. 10 step horizon

Fig. 2. Rolling return spillover index 80 70 60 50 40 30 20 10 0 1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

Spillover plot. Volatility. 100-week window. 10 step horizon

Fig. 3. Rolling volatility spillover index

return spillovers were low, suggesting a market that was not very integrated. The markets degree of integration in general fell throughout the period 1980 to 1993 and then reversed until 1996 when it resumed its path towards a low point of integration in the early 2000s. Since then, the markets have returned to a high degree of integration in returns. In Figs 4 and 5, we show the net spillovers from and to each market for returns and volatility, following Diebold and Yilmaz (2012). The interpretation is of a positive figure indicating a source and a negative a recipient of spillovers. Gold shows four distinct periods when it acts as a source of return spillovers: 1992–1996, 2000–2002, 2003–2004 and 2006–2008. Silver is a source of return spillovers in the period 2004 to 2008, palladium in the period 2008 to 2010 and platinum rarely. The effect of the global financial crisis is marked in Platinum, Silver and Palladium, the latter becoming a source of spillovers (perhaps linked to the heavy industrial use of the metal) and the other

two becoming (even more) a recipient of spillovers. It is also clear in gold, the metal moving from being a source of return spillover to a neutral position. In volatility, we see that while in general there is little net spillover in any of the four metals when they do show such it tends to be large. While the largest return spillovers are in the range of 40%, the range in volatility exceeds 100% at times. Gold is a net positive volatility spillover in the period 2001 to 2003, matching a return spillover period. Similarly for palladium, volatility shows positive spillovers in the period 2008 to 2009 matching returns. Platinum spillovers in volatility match those for returns in the period 2008 alone. Shown in Figs 6 and 7 are bidirectional spillovers. These show the influence of a particular asset on another. We concentrate here on gold and silver. We see that in both cases there are distinct regimes. For both, the pattern is generally to decline from about 25% to almost zero by 2002,

J. A. Batten et al.

6 15 5 –5 –15 1984 1986 Net return spillovers, gold 40

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

20 0 –20 1984 1986 Net return spillovers, palladium 20 10 0 –10 1984 1986 Net return spillovers, silver 10 –10 –30 1984 1986 Net return spillovers, platinum


Fig. 4. Net asset spillover indices for returns

75 25 –25 –75 1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

1984 1986 Net vol spillovers, palladium

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

1984 1986 Net vol spillovers, platinum

1988

1990

1992

1994

2000

2002

2004

2006

2008

2010

2012

1984 Net vol spillovers, gold 80 20 –40 –100

175 100 25 –50 1984 Net vol spillovers, silver 40 0 –40 –80 1996

1998

Fig. 5. Net asset spillover indices for volatility

22.5 20.0 17.5 15.0 12.5 10.0 7.5 5.0 2.5 0.0

22.5 20.0 17.5 15.0 12.5 10.0 7.5 5.0 2.5 0.0

1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 Gold to palladium 30

1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 Silver to palladium 25

25

20

20

15

15 10

10

5

5 0 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 Gold to platinum 40

0 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 Silver to platinum 40

35

35

30

30

25

25

20

20

15

15

10 1984 Gold to silver

10 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 Silver to gold

Fig. 6. Bidirectional spillovers in return – gold and silver

rising thereafter. Gold–Silver and Silver–Gold show very similar bidirectional spillovers as we might expect. The aftermath of the 9/11 attacks

and the Iraq/Afghan wars plus the gradual increase in global money supply show clearly. In 2003/2004 after the Iraq War, we see a major

Which precious metals spill over on which, when and why? 40 35 30 25 20 15 10 5 0

7

30 25 20 15 10 5

1984 1986 1988 1990 1992 1994 1996 1998 Gold to Palladium 40 35 30 25 20 15 10 5 0 1984 1986 1988 1990 1992 1994 1996 1998 Gold to Platinum 70 60 50 40 30 20 10 0 1984 1986 1988 1990 1992 1994 1996 1998 Gold to Sliver

2000 2002 2004

2006

2008 2010

2012

0 1984 1986 Sliver to Palladium 30

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

1990

1992 1994 1996

1998

2000 2002 2004

2006

2008 2010

2012

1990

1992 1994 1996

1998

2000 2002 2004

2006

2008 2010

2012

25 20 15 10 5 2000 2002 2004

2006

2008 2010

2012

0 1984 1986 1988 Sliver to Platinum 60 50 40 30 20 10

2000 2002 2004

2006

2008 2010

2012

0 1984 1986 1988 Sliver to Gold


Fig. 7. Bidirectional spillovers in volatility – gold and silver

sustained increase in the influence not only of gold on platinum and palladium and but also of silver on the same metals. Interestingly, the influence of gold on silver and vice versa fell in that period. In volatility spillovers, we see the influence of gold disappearing on platinum and palladium in the period 2000 to 2003 and diminishing significantly for silver in that same period. However, silver shows its largest level of volatility spillover on gold in that period. V. Conclusions We apply a new method of spillover/integration and surface time-varying spillovers for the four main precious metals. The metals are split into two categories: gold and silver show consistent spillovers between them, while platinum and palladium are much more disconnected. In terms of time variation, we can attribute shifts and changes to geopolitical and economics events.

References Abanomey, W. S. and Mathur, I. (2001) International portfolios with commodity futures and currency forward contracts, The Journal of Investing, 10, 61–8. doi:10.3905/joi.2001.319474 Adams, Z., Füss, R. and Kaiser, D. G. (2011) Macroeconomic determinants of commodity futures returns, in The Handbook of Commodity Investing, Fabozzi, F. J., Füss R. and Kaiser, D. G. (Eds), John Wiley & Sons, New York, pp. 87–112. doi:10.1002/ 9781118267004.ch4

Batten, J. A., Ciner, C. and Lucey, B. M. (2010) The macroeconomic determinants of volatility in precious metals markets, Resources Policy, 35, 65–71. doi:10.1016/j.resourpol.2009.12.002 Boido, C. (2013) Investing in commodities, in Alternative Investments: Instruments, Performance, Benchmarks, and Strategies, Baker, H. K. and Filbeck, G. (Eds), John Wiley, Hoboken, NJ. ISBN: 978-1-118-24112-7 Brunetti, C. L. and Gilbert, C. (1995) Metals price volatility, 1972–1995, Resources Policy, 21, 237–54. doi:10.1016/0301-4207(96)85057-4 Byers, J. D. and Peel, D. A. (2001) Volatility persistence in asset markets: long memory in high/low prices, Applied Financial Economics, 11, 253–60. doi:10.1080/096031001300138645 Caglayan, M. O. and Edwards, F. R. (2001) Hedge fund and commodity fund investments in bull and bear markets, The Journal of Portfolio Management, 27, 97–108. doi:10.3905/jpm.2001.319817 Chan, W. H. and Young, D. (2006) Jumping hedges: an examination of movements in copper spot and futures markets, Journal of Futures Markets, 26, 169–88. doi:10.1002/fut.20190 Chow, G., Jacquier, E., Kritzman, M. et al. (1999) Optimal portfolios in good times and bad, Financial Analysts Journal, 55, 65–73. doi:10.2469/faj.v55.n3.2273 Cumming, D., Haß, L. H. and Schweizer, D. (2013) The role of alternative investments in strategic asset allocation, in Alternative Investments: Instruments, Performance, Benchmarks, and Strategies, Baker, H. K. and Filbeck, G. (Eds), John Wiley & Sons, Chichester, pp. 19–36. Diebold, F. X. and Yilmaz, K. (2009) Measuring financial asset return and volatility spillovers, with application to global equity markets, The Economic Journal, 119, 158–71. doi:10.1111/ j.1468-0297.2008.02208.x Diebold, F. X. and Yilmaz, K. (2012) Better to give than to receive: predictive directional measurement of volatility spillovers, International


8 Journal of Forecasting, 28, 57–66. doi:10.1016/j. ijforecast.2011.02.006 Engle, R. F., Ito, T. and Lin, W.-L. (1990) Meteor showers or heat waves? Heteroskedastic intra-daily volatility in the foreign exchange market, Econometrica, 58, 525–45. doi:10.2307/2938189 Erb, C. B. and Harvey, C. R. (2006) The strategic and tactical value of commodity futures, Financial Analysts Journal, 62, 69–97. doi:10.2469/faj. v62.n2.4084 Fernandez, V. (2008) The war on terror and its impact on the long-term volatility of financial markets, International Review of Financial Analysis, 17, 1–26. doi:10.1016/j.irfa.2006.11.003 Garman, M. B. and Klass, M. J. (1980) On the estimation of security price volatilities from historical data, The Journal of Business, 53, 67–78. doi:10.1086/296072 Georgiev, G. (2001) Benefits of commodity investment, The Journal of Alternative Investments, 4, 40–8. doi:10.3905/jai.2001.318997 Gilbert, C. L. (2006) Trends and volatility in agricultural commodity prices, in Agricultural Commodity Markets and Trade New Approaches to Analyzing Market Structure and Instability, Sarris, A. and Hallam, D. (Eds), Edward Elgar Publishing, Surrey, pp. 31–60. Gorton, G. and Geert Rouwenhorst, K. (2006) Facts and fantasies about commodity futures, Financial Analysts Journal, 62, 47–68. doi:10.2469/faj.v62. n2.4083 Greer, R. J. (1997) What is an asset class, anyway?, The Journal of Portfolio Management, 23, 86–91. doi:10.3905/jpm.23.2.86

J. A. Batten et al. Hammoudeh, S. M., Yuan, Y., McAleer, M. et al. (2010) Precious metals-exchange rate volatility transmissions and hedging strategies, International Review of Economics and Finance, 19, 633–47. doi:10.1016/j. iref.2010.02.003 Kroner, K. F., Kneafsey, K. P. and Claessens, S. (1995) Forecasting volatility in commodity markets, Journal of Forecasting, 14, 77–95. doi:10.1002/ for.3980140202 Lubber, E. (2013) Size of Precious Metal Markets | The Gold Researcher on WordPress.com Lucey, B. M., Larkin, C. and O’Connor, F. (2014) Gold markets around the world – who spills over what, to whom, when?, Applied Economics Letters, 21, 887–92. doi:10.1080/13504851.2014. 896974 Molnár, P. (2012) Properties of range-based volatility estimators, International Review of Financial Analysis, 23, 20–9. doi:10.1016/j.irfa.2011. 06.012 Nijman, T. E. and Swinkels, L. A. P. (2008) Strategic and tactical allocation to commodities for retirement savings schemes, in The Handbook of Commodity Investing, Fabozzi, F. J., Fuss, R. and Kaiser, D. G. (Eds), John Wiley, Hoboken, NJ, pp. 522–45. Pindyck, R. S. (2004) Volatility and commodity price dynamics, Journal of Futures Markets, 24, 1029–47. doi:10.1002/fut.20120 Skiadopoulos, G. (2012) Investing in commodities: popular beliefs and misconceptions, Journal of Asset Management, 13, 77–83. doi:10.1057/ jam.2011.35