• Increase font size
  • Default font size
  • Decrease font size

Dec 2002 - Gold Price Forecast

Forecasting the Gold Price

The general reaction to anyone commenting that they intend to forecast the gold price, is that it cannot be done, but for mining companies assessing the viability of a gold project, something has to be done. In the late 1970’s I once worked for a group within a South African mining house (gold mines in South Africa can take up to 12 years from discovery to production), whose task was to forecast the gold price.

For modeling purposes, Eagle Research currently uses the current gold price of say US$320/oz and projects it forward at that constant level. Sensitivities are made at +US$25/oz and –US$25/oz, or in A$/oz (+A$50/oz and –A$50/oz). Such discounted constant money cashflows at 5% or 7.5% can be equated to real money. The rationale behind using constant gold price forecasts is given in the following article.

There are two categories to forecasting the gold price, namely long-term or greater than one year, and short-term or less than one month which is used more for trading purposes. Most forecasting is made using the afternoon fix of the gold price because that is perceived to be the most representative in the world within any 24-hour period.

This article focuses on some of the long-term methods or models that have been used to forecast the gold price, the shorter-term forecasting methods or models that can be used for trading strategies are expected to be covered in a later article.

Most models that we have encountered on the gold price use some kind of regression analysis of a number of parameters to determine what sum of multiples (of constant values multiplied by each parameter) results in the gold price for a particular year. In the mining house where I was employed, two models were used, namely the Intergold model and the Monro model.

Intergold was the forerunner of the World Gold Council, and they devised a model using US Money Supply, World Money Supply, World Liquidity, the Eurodollar interest rate, the real Eurodollar rate, the real US GNP, a US$ exchange rate and an international index of political tension compiled by the Hudson Institute. The model worked relatively well until 1978, when its variability increased to US$60/oz to US$100/oz compared to the actual achieved and it was abandoned.

The Monro model used in the late 1970’s to early 1980’s had 5 input parameters being USCPI, US interest rates, OECD real growth, Gold Supply (t) and the strength of the US$ to a basket of currencies, to solve an equation of the gold price according to its 6 demand categories (developed, and developing countries : jewellery, medals and medallions, bullion and coins) and their price elasticities in 1973 terms. It could forecast to within 10% to 20% of the actual price in any one year, but the model’s parameters had to be updated on a yearly basis in lengthy assumptions and estimates which then often dramatically altered its future forecasts.

Anglo American commissioned Horace Brock in 1981 to predict the gold price for the period 1986 to 1988, and he used a probability analysis of 3 supply categories (by region) and 4 demand categories. The investment demand was then broken down as being a function of the gold price, real US and European interest rates, real US, European and SE Asian economic growth, the level of political tension (in the form of a world anxiety coefficient) the rate of inflation in non-communist industrialized countries and the relative strength of the US$. The result was a median gold price on no inflation of US$611/oz to US$640/oz, with a 1% chance that it would be less than US$520/oz (the price forecasts were higher for higher levels of inflation). Whereas, it in fact averaged US$368/oz to US$447/oz per year between 1986 and 1988.

The Chamber of Mines of South Africa also had a model, which was relatively accurate even into the mid 1990s. It could make forecasts of the gold price based on quarterly averages and had an accuracy of +/- US$25/oz in any quarter, with at least 12 of the quarters having an error of less than US$7/oz, for the period from 1984 through to 1993.

The Chamber’s model was based on a linear regression of 4 parameters being the real US long-term interest rate, the Brent oil price, the DMK/US$ exchange rate and the annual increase in USM1 money supply led by 4 quarters (or one year) to provide inflationary expectations, and a “constant” to balance the equation deriving the US$/oz gold price. The Chamber’s model had the advantage that most of its input data could be sourced from the last page of the UK publication, called “The Economist”.

Although the oil price relationship with the gold price was largely severed some time ago, the inclusion of the oil price in the model gave a better statistical fit than its exclusion.

Another model that used to forecast the gold price often to within an error of US$20/oz, but with a better statistical “fit” than the Chamber model was that by Kaufmann and Winters (“The Price of Gold” Kaufmann TD and Winters RA, Resources Policy, December 1989 (Butterworth & Co)). The model estimated the gold price from early 1989 as averaging US$382/oz for the year, which was quite impressive for actual averages of US$380.8/oz in 1989 (and US$383.6/oz in 1990). The Kaufmann and Winters model in fact had an impressive accuracy in forecasting the average annual gold price from 1974 through to 1993 according to 3 main parameters.

The 3 parameters that it used to model the gold price were the US GNP deflator (being the US Dept of Commerce’s deflator index for real US GNP, as a measure of inflation), a trade weighted value of the US$ according to a basket of 10 industrialised countries, and world gold production (from the GFMS’ Gold Survey publication’s summary table of Western World gold production and net communist sales).

Both of the latter models that had achieved relatively impressive track records began to deteriorate in their forecasting margins. In Kaufmann and Winters case, due to increased forward selling altering the supply-demand relationship. In addition to which greater focus began to occur on the make-up of the balance between supply and demand that was made by central bank sales.

Various groups started looking for shadows due to central bank announcements of sales coinciding (almost unerringly) with the gold price starting to rally, which then caused it instead to often stumble and decline.

The influence of central banks on the gold price was clearly spelt out at the Australian Gold Conference in March 1996, when an European central banker stated that he had heard two days of discussion on the supply and demand of gold and expectations for the gold price, however, it is the central banks that decide on the level at which the gold price trades at.

To some degree, the central banks do decide on the level of the gold price, except that the price also depends on what influences occur on the trading desks, such that beak-outs or price squeezes can occur, resulting in surges in the price as we saw only a year or so ago.

Almost all the econometric models used a US$ exchange rate as one of their parameters and in all those models the exchange rate was inversely correlated to the gold price (or when the US$ weakens, the US$ gold price rises). This is in contrast to a forecast made by a broker at a recent conference which had the gold price tracking the US$, namely weakening as the US$ was expected to weaken.

Before the Chamber, and Kaufmann and Winters’ models began to come adrift with higher margin or residual adjustments, a base gold price kept being derived, namely US$325/oz, which is almost where the price is now (October / November 2002). In our opinion, we expected the gold price to attain US$325/oz in 2002 (as its long-term equilibrium level) and then rise through it (because it has been below that level for some time) before dropping back to US$325/oz again. So far, it seems to struggle when it gets above US$325/oz.

We have assessed a number of models that we have historically experienced being used to forecast the gold price. This is not a comprehensive coverage, and possibly, there are other models which have a greater accuracy in forecasting the gold price. With the reduction in forward selling and to some degree control on central bank sales, possibly some of these historic models can be used again.

However, for financial modeling purposes we (Eagle Research) are projecting the gold price forward at a constant price of about its current (October / November 2002) level of US$320/oz, with +/- US$25/oz or +/- A$50/oz for sensitivities, discounting back to effectively real money terms.

Disclosure and Disclaimer : This article has been written by Keith Goode, the Managing Director of Eagle Research Advisory Pty Ltd, who has a Proper Authority with State One Equities, and with his associates, either has or expects to have interests in most of the stocks in this article, This e-mail address is being protected from spambots. You need JavaScript enabled to view it . The opinions expressed in this article should not be taken as investment advice, but are based on observations by the author. The author does not warrant the accuracy or completeness of any information and is not liable for any loss or damage suffered through any reliance on its contents.

  • Written by: Keith Goode
  • Sunday, 01 December 2002