Wednesday 27 September 2023

Are miners locking in a linear regression to the gold price?

A nascent regression between gold miners (using the HUI index of unhedged gold miners) and the yellow metal may finally interrupt the unabated miner slide relative to gold.
Previous regression relationship between the HUI and gold held for about three years only. It survived the Corona plunge, the price surge of gold the summer of 2020 and the ongoing financial largesse in 2021. Clouds started gathering in the summer of 2021.

Recalibrating the regression parameters added one more match in the spring of 2022, as both gold and miners surged after the Russian invasion of the Ukraine. This proved a short rally and miners slid rapidly when inflation surged and the FED abruptly changed stance with the termination of QE and a series of rate hikes.

Miners seemed to face no resistance in their slide. In September 2022 the HUI plunged beneath its June 1996 level of 200 (when the public data series started). Gold never closed below $1620 however, over four times this June '96 initial level.

With the regression permanently impaired, miners kept falling behind as gold recovered. The broad stock market got into a bear market and the crypto winter set in. Rising bond yields also caused havoc for fixed income. Rampant inflation ate away the purchasing power of cash: no place left to hide. 

HUI - Gold regression.

When plotting the HUI index relative to gold, time is eliminated. On a long term series, we observe both previous regression relationships and the lengthy in between periods of miners falling behind on the gold price. Lately a 'new' regression relationship seems to be emerging.

Gold (horizontal axis) and HUI index of unhedged miners (vertical axis)

Visually, the 2012-2017 regression predicts much higher HUI values for any given gold price. This relation was invalidated during 2017 while gold rallies were breaking upon the $1350 boundary.

Note: Gold prices are  NY Comex close at 5 pm. HUI values are at stock market close at 4 pm.

The latest regression emerged in 2018 in March or August according to the correlation threshold chosen. This regression withstood the challenging Corona plunge and the subsequent gold price spike during the summer of 2020 and the ongoing financial largesse in 2021. The FED speak focused on transitory inflation because of budget deficits amidst a productivity reduced by Corona lock-downs. In the summer investors grew wary as the FED finally planned to wind down QE. Even though gold upheld $1700 in the autumn of 2021, miners gave way. Gold would eventually end 2021 at $1830 only little below its $1898 level by end 2020. Instead of a benign 3.6% weakening of gold, miners slid 13.6% over 2021. With gold flat over the challenging year 2022 miners nevertheless lost another 11.2%. 

Corresponding points on the Gold-Hui spread graph are shown in olive green below the lower regression line valid from 2018 till mid 2021. When incorporating those points into the regression calculation, they consist in a large belly of points pulling the regression line flatter and raising the intercept value. Evidently this comes at the cost of lowering the linear correlation coefficient. A previous article pointed out this impairment in April 2022. Even before the regression was eventually invalidated, the risk for an progressively ongoing impairment was pointed out (August 2021).

Amidst the gloom, the dust settles over 2023 and inflation threats seem to be waning. This is perhaps why a nascent regression seems to be emerging. The candidate is pointed out in orange at right, predictably well below previous regression line. We currently only have a set of 140 points aligned, stretching from March 15 till yesterday September 26. Despite the short HUI and Gold ranges, the correlation now stands at 0.91. (Gold, HUI) couples outside the current range need to confirm for this candidate regression to hold over a (hopefully) longer time stretch. So far its parameters are a slope of 0.487 with an intercept of $1456.6. The 'equilibrium value' of the HUI then stands at:

HUI (estimate) = 0.487 ∙ (Gold price - $1456.6)

To be confirmed ...

Synoptic graph

In a synoptic graph, the axis for the HUI ('dependent variable') is modified to reflect the linear regression as derived above. The HUI estimate as determined by the gold price is then compared to the real observed HUI value. The difference is a regression error.  Squares of regression errors are minimized. Regression errors need to change sign frequently for a regression to remain unchallenged.

This is how the synoptic graph of the candidate now looks:

Synoptic view of the gold price (left axis) and also the HUI regression estimate (red, right axis) while observed HUI values are shown in blue on the right axis.

HUI - Gold long term graph

Plotting the graphs of gold and the HUI index over a very long time interval puts regressions in their true historic perspective. As the overview article on past regressions has pointed out, eventually they all fail in the same way. The HUI fails to keep up with its regression estimate and systematically continues lagging. Any new regression candidate later on sets less challenging goals for the HUI at any given gold price.

Very long term graph of gold (red, left axis) and the HUI (blue, right axis).

Miners are poor investments as compared to gold over the very long term. This is well illustrated by the above 'secular' graph showing the HUI index and gold since 1996. Over the very long term, there is hardly any progression for the HUI over its 27 years covered. Meanwhile gold has more than quadrupled.

The nascent regression line once more 'predicts' lower HUI index values for any given gold price than the latest regression did. This should remind mining investors that 'leverage to the gold price' (often called 'optionality') generally holds over a short to medium time frame, but certainly not over decades.

'Optionality' definitely is a sell-side analysts' argument. It sounds as one of those fancy concepts designed to lure less informed investors into jumping on the bandwagon of an ongoing miner bull market.

Yet another conclusion of the above graph is that you cannot diversify away specific mining risk buying any miner ETF. Both GDX and (even more) GDXJ lag the any gold bullion ETF as GLD more or less to the extent that the HUI index lags the gold price.

Few individual miners keep up with the gold price over the long haul. Even expert picks have a hard time realizing any decent profit over the long haul: what is the fate of most individual investors is more or less obvious. 

No comments:

Post a Comment