Gold Price Forecasting: Myths, Models & the Market Forces that Matter

While a brief internet search will quickly uncover a plethora of gold price forecasts, it is often far less clear how these forecasts are generated. Unfortunately, some of these will simply be created out of thin air. Another group will be derived from looking at charts and using technical analysis. However, many will be derived from statistical models incorporating data considered relevant to the formation of gold prices. This article takes a brief, non-mathematical exploration of how such models are typically built and used in gold price forecasting.

All about Supply and Demand?

Intuitively, supply and demand of gold should play a role in gold pricing. After all, if gold were as common as water, it would not trade at over $3,500/troy oz! Thus, many historic pricing models forecasted a balance of supply (from mining and recycling) and demand (for jewellery, industrial applications and investment). However, this approach is problematic for several reasons. 

Firstly, all estimates of production are prone to error and can typically only be verified months after the fact. This issue has become even more acute as mining activity has slowly shifted to less transparent locations such as China, Indonesia and now arguably Russia. 

Demand is even more difficult to gauge, given the huge level of venue fragmentation and idiosyncrasies of price sensitivity and economic conditions. Meanwhile, central bank buying activity is covert and reported with a substantial lag – if at all. Of greater concern is that investment (including central bank net purchases) typically constitutes about 50% of projected gold demand. This leads to circular reasoning because forecasters effectively project rising prices because they think others will invest.

However, gold supply/demand models ultimately fall short because they fail to recognise that prices are determined by marginal, rather than “average” buyers and sellers. There doesn’t have to be a “wall of money” or “shortage of stock” for prices to go up. Data compiled by the World Gold Council bears this out; it suggests that gold has been in physical surplus every year since 2013, yet gold has risen sharply over the period. 

Thus, while forecasting production levels might be useful to evaluate the attractiveness of individual gold miners, it is of limited use in forecasting gold prices. The answer lies elsewhere.

Macro Modelling

Modern gold pricing models focus on a range of “top down” economic data and associated market instruments, rather than the “bottom-up” modelling already described. Factors seen to have a statistically significant relationship with gold prices are identified and combined, such that they collectively help to explain the observed gold price and/or movements. Forecasting such factors can then yield a gold price forecast.

While there is currently no definitive list of factors included in any single model, common elements include the following:

Global Economic Growth

Usually measured as real (inflation-adjusted) growth in Gross Domestic Product (GDP), this is deemed a driver of gold prices over the long term. The rationale here is that increasing wealth increases gold affordability and ultimately results in higher prices. This factor can be further refined into regional growth forecasts as inputs from certain areas, such as India, seem to be of greater than average significance. 

Interest Rates

Various measures of US interest rates are common components of gold valuation models. Models can focus on short-term rates (typically Fed policy rates or short-term treasury bills, long-term rates (US Treasury bonds of 10 years and beyond) or a mixture of both. Models can also focus on nominal or real (inflation-adjusted) rates. All have the advantage that these are traded almost continuously in real-time, delivering frequent adjustments to the gold fair value price. 

The rationale for incorporating rates into gold operates at several levels. Short-term rates are a measure of the opportunity cost of holding non-yielding gold and so typically have an inverse relationship. The same can be said of longer rates, but longer tenures also capture additional gold drivers such as default risks and inflation. 

Inflation

This is often viewed as a separate driver as gold is often considered an “inflation hedge”. It is also clearly related to interest rates in an environment where the US Fed has an inflation target. Indeed, while a simple long-term real (i.e. inflation-adjusted) rate model had performed well for several years, it has fared less well in the post-Covid era. An example of the potential complexity in interpretation is that while negative real long rates might increase the attractiveness of gold as a long-term investment, it also increases demand for competing risky assets such as equities. 

The US Dollar

Superficially, the relationship between gold and the US Dollar seems obvious; as gold is priced in US dollars, a weaker dollar is associated with a higher dollar gold price. While this is certainly a factor, it is unfortunately not that simple. For instance, a weaker US dollar also implies that at least some other currencies are stronger. For gold consumers in these local currencies, the price of gold might well be lower, helping to boost demand. Moreover, if a weak US dollar is associated with relatively low US interest rates – as has historically been the case – then it may be difficult to isolate the currency impact. A more recent phenomenon is dollar weakness being associated with a risk premium for any US dollar asset – providing an additional boost to the dollar price of gold as a “safe-haven asset”. 

Positioning

While not a typical feature of long-term gold price models, data on flows into and out of gold-related financial products are sometimes a feature of models with a more tactical focus. Such models might incorporate data such as physical gold ETF/ETC flows and/or net COMEX gold futures positions – usually scaled to open interest. Such data is used to gauge investment momentum, with rapid accumulations and depletions supporting price appreciation and decline respectively. Additionally, unusually low or high levels of activity might be a lead indicator of a price trend reversal. 

Conclusion

Quantitative gold price modelling can be useful in understanding the factors that have an impact on gold prices and projecting price behaviour under different market conditions. However, they can also be complex and remain a work in progress as factors are better understood and incorporated to yield more reliable results.