Price discovery in strategically-linked markets: the case of the gold-silver spread. Adrangi et al. (2000)
This paper from 2000 analyses the relationship between gold and silver futures prices. The abstract states that:
Using 15 minute intraday data, we analyse the price discovery process among the strategically-linked gold and silver futures contracts and examine the role of the intermarket spread in their price dynamics. The multivariate model employed allows for intermarket volatility spillover and asymmetric-spread effects on the variance and covariance of the two contracts. The data suggest that the silver contract bears the majority of the burden of convergence to the gold-silver spread. This evidence is noteworthy since the silver contract was by far the more volatile of the two contracts over the period studied.
The importance of these two sets of prices lies in the gold-silver spread trade. This was the first paper to offer evidence about the co-movements of gold and silver returns, which determines this spread. It does this by looking how changes in prices allow information to flow between the two.
The two series are found to be co-integrated, that is there is a long term relationship between the two that mean reverts. Both Philips-Perron and Johansen tests were used.
The authors found significant serial correlation in both the returns and squared returns of the two series but not in the residuals of the returns. This fits with findings from the literature, such as Tully and Lucey (2007), that gold returns are best described within the GARCH family of models. Persistence is found in the covariance of returns.
This long term relationship between the two is found to be based on information flow inn both directions, bi-causal, through a VAR analysis using lags of the returns of gold and silver in both models. The model’s also included the spread between the two and finds that it is only significant for silver, implying that silver bears the brunt of adjustment in the relationship.
Inter-market shocks affect both assets but shocks due to gold have a much stronger effect than silver.
The effect of the spread is shown to be asymmetric. Lower spreads are found to be related to significantly increased volatility and when spreads are narrow closer co-movements in their prices are normal.
For sensitivity analysis the authors also used daily closing prices and found that it did not affect their results significantly. Liquidity was found to be similar for both contracts over the period.
Method: Philips-Perron test, Johansen test and VAR model.
Data: 15 Minute intraday data for COMEX gold and silver futures from 27/12/93 – 30/12/95 sourced from tick data incorporated. Daily closing prices also used.
Full Citation: Adrangi, B., A. Chatrath, et al. (2000). “Price Discovery in Strategically-Linked Markets: The Case of the Gold-Silver Spread.” Applied Financial Economics 10(3): 227-34.