Monday, 9 May 2016

Gold miner rally: Bull market logic

Wild gyrations

Ever since bottoming barely above 100 on Jan 19, 2016, the HUI gold miners index more than doubled.  It closed at 226,1 on Friday May 6, up 125% from its trough or 103% year-to-date.  However, after peaking at 233.5 by the Apr 29 close, we witnessed a week with wild gyrations, starting with a three days losing streak culminating in the 10.25% landslide loss on Wednesday May 4.  Much overdone so it seems, since the Thursday recovery rally was in fact anticipating precious metals eventually firming on Friday May 6.

Opposing opinions 

During and after this first major pull back of the gold miners, questions have been raised on gold miner valuation:
These are opposing opinions; both are using the same data and information but differ vastly on its interpretation. The first article of Ian Bezek would have been very useful if written during the previous WE, before the miner pull back started. As it was eventually published, the correction had run its course.  Timing is everything for this kind of tactical games.
In his weekly essay at Zeal, David Hamilton points to the high relative value of the HUI (rHUI) when compared to its 200 days moving average.  The comparison is made with the 2009 recovery after the financial crisis. He sees the current high relative value more typical for an early bull phase, leaving plenty of opportunity for a more gradual continuation. Occasional corrections tend to be brief. Any seasonal weakening will automatically normalize the rHUI, as more of the extremely low readings will gradually be phased out.

The bull market logic

Last year, some articles on the bear market logic for gold miners were added to this blog:
  1. Gold and the miners: Identifying the bear market logic (May 2015)
  2. The bear market logic for miners: Continuation and Analysis (July 2015)
What's so exciting about this analysis, is that it continues to be valid during this young gold miner bull market: the bear market logic almost candidly transits to a bull market logic.  Not a single letter needs to be changed to the linear regression underpinning the mathematical model.

Following graph overlays the gold price in USD/Oz on the left hand scale and the HUI gold miners index on the right hand scale, from July 2013 onward. The analysis is valid including about a year prior to this, however this merely extends the scales and reduces the detail.

Gold (in red on the left hand scale in USD/Oz) and HUI (in blue on the right hand scale)
Data till May 6, 2016, click to enlarge
What do we need to make these graphs to overlay with an almost perfect correlation? As mentioned above and explained in the second article, a linear regression allows calculating the parameters relating the HUI to the gold price.

HUI = 0.555 * (Gold Price - $ 893)

You have noticed on the above graph that both scales don't share the same zero level. Moreover $50 on the gold price translates to 28 points on the HUI. This implies a factor 0.56 the rounded value of the above linear coefficient. Using this scaling, the zero level for the HUI would correspond to a $900 USD/Oz reading for the gold price, which is the rounded value of the above parameter $893.

This model quantifies what mining investors use to call 'optionality': the expectation that earnings and mining share prices outperform the metals as they rally.  The expected leverage is high as gold prices are recovering from their bottom level. It is much less pronounced if gold were to rally to new highs.

As you notice, in a linear regression, dates are eliminated. The best fitting line is found among the scattered couples (Gold price, HUI value) over the observation interval. 
Linear regression line between gold and the HUI index (click to enlarge)
The calculation is made on data ranging from summer 2012 till the present day. For the sake of completeness, earlier points from Jun 2011 onward also are included. Those observations shown in orange are concentrated at the upper right above the regression line. While the gold price was ascending to its Sept 2011 all time high and until gold failed to regain $1800 about a year later, the HUI index almost constantly quoted lower on any occasion a previous gold price level was revisited. There clearly is no modelling of such obvious impairment.  Yet from summer 2012 till the present, the new relationship holds through gold retreats and recoveries and eventually to the more recent onset of what appears to be a new gold bull market.

Why this type of analysis came about?

The oscillatory down trend of the HUI/Gold ratio ever since 2011 implies that a proportional logic is not holding any longer. Here, we graphed both the gold miners index HUI and the HUI/Gold ratio. Previously it had been assumed that HUI/Gold, being the main explanatory parameter linking the HUI to gold, fluctuates within a limited range. As the gold miner bear market has proven, there really is no lower boundary for HUI/Gold carved in stone. Any 'defense line' thought of eventually gave way. We witnessed HUI/Gold to retreat several times below 0.100 at the very bottom of the gold miner bear market.
HUI index (blue, left scale) and HUI/Gold ratio (dark green, right scale)
Readers familiar with this former analysis will remember how HUI used to be approximately equal to half of the gold price during the early gold bull years from late 2003 till spring 2008. It had become some kind of trading rule valid during those years. Eventually the relationship broke down during the 2008 financial crisis, as all former 'trading rules' were found to fail. The below historic graph of the HUI and the gold price (since 1996 to the present) very well illustrates the proportionality holding during these early bull years (as it also did in 1996-97, before the late 20th century gold bear market aggravated).
Gold bullion, in red on the left hand scale and the HUI in blue on the right hand scale.

What will invalidate this model?

No trading rule, proportionality, trend line or even more complex relationship holds for ever. Therefore, figuring out what would invalidate this model is very useful. Whenever new data points are added and the regression parameters are calculated, slightly different values are obtained. This clearly does not invalidate the model. If regression parameters were to drift in a systematic way, the linear correlation would deteriorate. Then the model credibility clearly is at stake. Under those circumstances, residuals would remain either positive or negative over a long time frame. Residuals are analysed in the last paragraph.
Currently the correlation coefficient for the data series running from summer 2012 (start on July 11, 2012) is 0.973.  Shortening or lengthening the time series with a few months only results in a tiny deterioration of the correlation coefficient and a minor parameter shift. The trend line tends to steepen when taking on board earlier data, while the gold price intercept shifts to a higher value.
However, when starting the analysis at the September 2011 all time high, linear correlation lowers to 0.95 and the slope steepens to 0.667. The regression then is quite far stretched and its parameters have become less reliable. Only few months earlier data (starting in April 2011) are necessary to make the slope rise to 0.723 with a correlation down to 0.914. This has taken us over the edge and those parameters clearly are invalid.

Parameter interpretation

This all seems so straight forward, yet it really isn't. The $893 (broadly spoken $850-$950) would imply the gold price level that would nullify the mining sector. It is almost certain that a gold price in that range for several years would most certainly cause 'casualties' among the gold mining majors that make up the HUI.

There are however a number of caveats: all individual mines have their own extraction cash costs (which cover only current expenses) and 'all-in sustaining costs' (AISC) which include recurrent investments needed to sustain operations over the lifetime of the individual mining site.

AISC are not constant and most mining companies were able to reduce average AISC for their mining sites. The AISC level is highly affected by fluctuations of the exchange rates relative to the USD. Non-US producers experienced a FOREX 'tail-wind' on cost reduction during much of the gold miner bear market.

AISC among the largest 10 gold producers quoted on any stock market are found to vary from $625 for the Russian Polyus Gold to about $1025 for the South African Sibanye Gold, according to this article by Vladimir Basov on Mining.com. AISC at Polyus Gold obviously were much reduced by the hefty slide of the Russian currency last year. Gold production costs from deep underground mines in South Africa have been relatively high.

Instead of thinking about the $850-$950 price range as the 'level that would nullify the gold mining sector', this rather is the collective investor's idea about that level.

The factor 0.555 still comes sufficiently close to the old adage of the HUI ranging about 0.5 times the gold price.  However rather than a proportional trading rule, the linear regression explains both the gold mines being slashed heavily during a gold correction and those same mines outperforming the metal during a gold recovery rally or rejuvenated gold bull market.

Too far too fast: is the gold mining sector over-valued?


The regression line among the (Gold, Hui) couples imply some deviation from the linear trend. When bringing back in the dates, we may monitor how this deviation (also called residual) has evolved over time. Residuals may well average zero, they do fluctuate between -60 and 60, with about 95% of the observations between -40 and 40.  A previous article tells more on the extreme variation of the residuals before and after the mid April 2013 gold price smash.

Residuals as a function of time (click to enlarge)
The first article of Ian Bezek acknowledges that gold miners were outperforming gold during the first six weeks since bottoming after mid January. "With little progress of the gold price after March 10, any further progress of the HUI would not have been justified by gold strengthening."

Residuals recently told a different story: Gold miners had upheld well while gold bottomed early December 2015.  Eventually they gave way and 'overvaluation' morphed to undervaluation as the miners bottomed on Jan 19, 2016.  While the initial miner response to the gold turn around was swift, they lagged the gold recovery relative to the trend line in February, with the residuals going negative to a -36 reading on Feb 11. Rather than outperforming gold, miners still had some work to do; yet by March 10, with gold closing at $1272.1, the HUI attained 179 with a residual still at -30.7. Miners still were lagging gold during its ascent. The HUI index took till Apr 8 before finally making it above the trend line for the first time since the gold recovery rally ignited. On April 29 the residual stood at 12.5: overvalued 0.25 on a scale to 1 if you like. In no way this reading is extreme, however it did spark the first major pull back of this rejuvenated miner bull market. In no time (on Wednesday May 4) the residual reading plummeted back into negative territory:  -14.4 on account of 10% plunge of the HUI triggered by a mild weakening of gold. Too far too fast in the opposite direction indeed: the Thursday miner recovery anticipated that of gold.

No comments:

Post a Comment