TORONTO (miningweekly.com) – While predicting future metals prices is a vital tool for assessing the viability of a deposit, a project’s economics or the potential income an operation could expect to generate, the prognostications often prove wrong.

“Studies of past forecasts show that the success rate for commodity price forecasting is pretty poor,” principal consultant at Watts, Griffis and McOuat Ross Lawrence told an audience at a recent meeting of the Canadian Institute of Mining’s (CIM) Management and Economic Society, in Toronto.

Gold was particularly tricky to forecast because of strong externalities and their variabilities. “It’s a commodity nobody really has a clue on. For example, you might think you’ve figured out correlations between gold and the US dollar, but then where’s the US dollar going to go next?” asked Lawrence.

Other metals and minerals were less difficult for forecasting purposes, although they still required a nuanced understanding of their markets. The position and influence of the overall resource cycle was equally important.

“Long-term metal prices are a key element and hardly a luxury that can be left to simple averages or rules of thumb,” Lawrence warned. “So can you make sense of the noise and come up with forecasts that are credible?”

CYCLES AND YEARS
Companies that assumed simple forecasting, often in the form of a single price point, increased the risk of error and associated problems if their forecast proved wide of the mark.

Analysts, institutions and or specialist data suppliers were often used by companies to de-risk and increase the odds of a forecast that guessed the eventual price. That was done for short-, medium- or long-term outlooks, with aggregates and consensus prices frequently used.

Average prices across three- or five-year terms were used as part of this, while many also employed three-year trailing averages.

Indeed, a three-year trailing average to inform a future price was viewed as an industry standard and best practice by the CIM, and was accepted by the Ontario Securities Commission.

However, securities regulators in various jurisdictions had different views on future price methodologies. “Canada and Australia may allow commodity prices based on expert opinion or the management of reasonable expectations. The USA allows, as a maximum, the lesser of the three-year moving average and the current spot price,” noted Lawrence.

However, some argue that using prices aligned with three- or five-years patterns were flawed as they failed to encapsulate the full range of a commodities cycle. Therefore, some had used averages gleaned from several years of inflation-adjusted pricing data.

Theoretically, the approach also helped compress any spikes and troughs to reveal a rising, falling or flat trend that could be used to chart a forecast rise.

It was common to use ten-year averages or even averages computed using data from several decades. For example, the price of copper for 95 years before 2013 experienced a trough in 1932 and a peak in 1974. Between those years the real price of the red metal rose over three-and-half times.

The mid-70s was followed with a 28-year downward slide, taking copper from $3.64/lb to $0.91/lb. That was followed by a sharp spike created by the China-driven resource supercycle from around 2003 to 2012.

The limitation of this method was when an average or trend failed to match the prevailing and predicted conditions at the time of the computation. “Would you be happy to use, say $2.80/lb copper?” Lawrence asked. “Or how about $0.90 for the sake of it – is that really a reversion to the mean?”

“We’re inclined to view the use of long-term prices as, at best, dangerous if not tempered with a comprehensive market assessment that takes into account the particular market cycle, where we are in [the market cycle], and the economic outlook over the life of the subject mine,” he added.

USEFUL EXERCISES
Using several price forecasts by others to form an aggregated mean was a useful exercise. However, companies also needed to be aware of limitations, subjectivities and biases sometimes residing within forecasts.

This was due to forecasters often overestimating future prices, with many highlighting the incentive structure for analysts as a root problem. It was often in a forecaster’s interest to reach a figure that was within the expectations of those hiring them.

“Analysts often wish to maintain a good relationship with a company whose earnings are being studied,” Lawrence noted.  “Bankers will disclaim any such problems, asserting they had established ‘all sorts of Chinese walls, don’t you know’. But we’re all human and these biases can exist, even subconsciously.”

Another problem was “herding behaviour”, where analysts sought to err on the side of caution, selecting a price that sat within the broad consensus. Others were wedded to their forecasts, often failing to reassess forecasts in the face of changing conditions.

One method to try and remove subjectivity or bias was to use multiple forecast scenarios, such as worst-case, best-case and base-case, plotting and publishing future values under each one.

“We believe a process that considers a range of likely outcomes is far superior. The method provides a way of illustrating and perhaps removing evaluators’ bias, and is reflected in the probability assigned to each scenario,” said Lawrence.

More rudimentary, but also helpful, was the 'Delphic' approach. Here, a team was invited into a room, with each person asked to voice their opinion, stating a price based on their view of what the future may hold. The exercise was then repeated several times, with those present asked if their opinions had changed based on what they had heard.

“It’s an interesting exercise because you can go around a few times and start compressing any spread,” Lawrence said.

Ideally, Watts, Griffis and McOuat’s preference was the use of the Monte Carlo method, whereby a multitude of inputs were used to compute an aggregate result. It helped avoid the trap of single forecasts and enabled a company to extrapolate a wider range of outcomes and thereby consider the implications.

The method also co-opted the associated uncertainties, volatility and cyclicality of commodity prices that was an essential part of the mining industry’s DNA.