Performance of Precious Metals in 2013

Most analysts polled by the London Bullion Market Association at the start of 2013 were optimistic for precious metals in 2013. They predicted annual average prices to be higher for another year. However, after the first half of 2013 is over, it looks rather unlikely that average prices in this year will be above the average price of 2012.

The table below shows the spot price of the four precious metals as well as the percentage change over the end of the previous end of quarter and also the percentage change in the first half. The development of gold and silver in the first quarter of 2013 could still be interpreted as a correction after the end of the festival season, which is not uncommon. Platinum and palladium moved higher due to the development in South Africa and fears of supply shortage. But the second quarter had been a disaster for all precious metals.  

Based on our quantitative fair value model for gold, we investigate whether this plunge of precious metals in Q3 could be explained by the major fundamental factors driving the price of gold. Our model is based on weekly data. As also financial and commodity markets show seasonal influences, we did not use weekly, but annual percentage changes. Thus, there was no need to include also further seasonal adjustments in the model equations. All fundamental factors were included in the equations for the four precious metals. These factors are the 1) the US dollar Index, 2) the price of crude oil (WTI), 3) the S&P 500 composite index, 4) the yield on 10yr US Treasuries and the net long position of non-commercials in the futures traded at the Comex division of the Chicago Mercantile Exchange (CME). For the US Treasury yield, instead of percentage changes, the absolute yoy-change in basis points had been used. The net-long position has been divided by 1000.

Linear regression models based on time series of financial asset or commodity prices often lead to residuals, which are serially correlated. Therefore, the models include a first order autoregressive term for the residuals and the parameters for the exogenous fundamental factors and the autoregressive error term have been estimated simultaneously. The model was first developed in 2006, and therefore, data from January 1997 until September 2006 was used in the estimation. The regression coefficients are all significant at the 5% level and have the expected sign.

The following chart shows the development of the yoy percentage change of the spot gold price and the estimated values. At a first glance, the model appears to predict the development of the gold price still quite well. However, this good fit could be the result of the autoregressive error term, which might deviate further away from the fair value instead of oscillating around it.

Since the end of 2012, the yoy percentage change of gold dropped from 2.47 to -22.07% at the end of June 2013, which is a change of 24.54 percentage points. Thus, we investigated how much the five fundamental factors contributed to the dismal performance of gold. The US dollar index firmed on the strength of the US dollar against the Japanese Yen and thus contributed a -0.64 percentage points. Also the drop of the net-long positions held by the large speculators contributed with -2.15 percentage points to the negative performance of gold. However, crude oil, the S&P 500 index and the yield on the 10year US Treasury notes all made a positive contribution. In total, the five fundamental factors indicated that gold should have shown a decline of the yoy percentage change by 0.92 points to 1.55% instead of falling to -22.07%.  

This already indicates there must have been a structural change in the precious metals markets. Relationships collapsed, which held before the financial crisis and even in the first few years after the crisis.

One possible reason for a structural change might have been the speech by ECB president Draghi held in July 2012 in London, where he pledged to do everything necessary to keep the euro intact and announced what later became known as OMT.  A second reason might have been the introduction of Abenomics in Japan in the final quarter of 2012. Thus, the regression coefficients have been estimated again, but based on data from July 2012 until the end of June this year. Most of the variables or lag structures are no longer significant at the 5% level. Only the US dollar index and the S&P 500 index remained significant factors. However, the regression coefficient of the S&P index increased and changed the sign from positive to negative.

Slight modifications of the lag-structure led to a model, where the five fundamental factors are significant again. However, now also the regression coefficient for the annual change in the 10year US Treasury yield changed sign and magnitude. Rising yields, which had earlier been interpreted as a sign of rising inflation rates, are no longer positive for gold. They now lead to falling gold prices, which reflect the fear that an end of QE would eliminate any reason for holding gold.

Crude oil was the only one of the five factors, which made a positive contribution of 4.75 percentage points. The US dollar and the S&P 500 index both contributed less than one percentage point to the decline. The rise of the 10year US Treasury yield contributed -6.7 percentage points. The fall of the net-long position held by large speculators had the strongest negative impact on gold’s negative performance with -8.75 percentage points. All in all, according to the adjusted and re-estimated model the five fundamental factors contributed -12.45 percentage points to the fall of the yoy percentage change of the gold price in the first half of 2013.

The chart above shows the development of the yoy percentage change of the gold price and the estimate based on the adjusted model fitted for the period from July 2012 until June 2013.  It is obvious, that a good fit is only obtained for the period from early 2012 onwards. For the time before, the fit is worse. This clearly indicates that the structure of the gold market has changed last year.

It is uncertain, how long the US stock market and the US Treasury yield will have a negative impact on the development of the gold price performance. However, even if the regression coefficients of the adjusted model remain valid beyond the estimation period, the fundamental factors explained only 12.45 of the 24.54 percentage point drop in the yoy gold price change. Thus, almost half of the plunge could not be explained by the major fundamental factors of the model. This implies that gold overshoot on the downside. If the fundamentals don’t deteriorate further, there is some potential for a recovery. But the sentiment in the gold market is seriously damaged and the famous knife is still falling. Thus, it appears too early to try catching the knife now.