Tuesday 1 September 2020

Miners (HUI) to gold linear regression still intact

 A year ago the new linear regression between major gold miners (represented by the HUI gold miners index) and the gold price has been revealed. With the yellow metal now way above the interval used to derive the regression relationship, it has been verified that the linear regression still holds.

Despite a landslide plunge of miners during the Corona pandemic frenzy, both the yellow metal and miners recovered well, with the linear regression challenged but not broken, as was revealed in an previous posting : Revival of gold mines - Corona dip overcome.

After the yellow metal has been steaming up to and easily breaking its Aug 2011 all time high, it is useful to verify the regression relationship and verify parameters for any possible creep.

***

In order to derive a linear trend between two time dependent variables, it is necessary to eliminate time, retaining only the couples of variables. Here (HUI, Gold) couples are retained, whereby the gold price on the horizontal axis is used to 'explain' the aggregate price level of miners as revealed by the HUI index.

***


HUI - Gold linear regression since Aug 2018: brown observed values and black regression line.

One year later, this is how the regression line looks like:


As explained before the 'Corona dip' shows up as a belly of values beneath the regression line showing up between about $1475 and 1675. As gold finally broke above $1800 early July, miners convincingly took the lead with a tiny set of values above the regression line till about the 2011 all-time-high.

The slope derived on the extended set is 0.270 against 0.261 mid May and 0.268 in Sep 2019. Not only did the regression relationship hold, the parameter creep we suffered upon the Corona dip also has been reversed. 

Thus leads to following linear regression : HUI = 0.27 . (Gold - $699)

Moreover the correlation meanwhile improved to 0.947, where it only stood at 0.901 mid May. Last year the correlation also posted near 0.95.

Synoptic view

As explained before, the regression relationship can be used to draw a synoptic view of the gold price and the HUI, whereby the (right) axis used for the HUI is adapted as to reflect the regression relationship.

Synoptic view of the gold price (red, left axis) and the HUI index (blue, right axis)

The gold price (red graph) is read on the left axis, but it also is the corresponding regression value of the HUI when read on the right axis. The closer both graphs, the higher the correlation is and the more reliable are the regression parameters.

As easily be seen, the Corona dip was a real challenge for the regression, with miners nosediving upon the gold price correction at the height of the stock market slide. Yet it established itself again as the gold price recovered and was confirmed during the summer rally breaking above $2000/Oz.

Residuals

Whereas the regression explains about 95% of the gold price dependency of the HUI index (hence the aggregate price of major gold miners), it also is interesting to check the 'residuals', which are the differences between the 'regression values' and the observed values of the HUI index.

Two graphs below show those residuals over the longer and the shorter time frame:

Regression residuals since mid August 2018, the validity start of the HUI-Gold regression.


Regression residuals since Apr 20, 2020

Residuals take all the 'statistical noise' but a pattern does remain. In the first graph the Corona dip is the most obvious feature. The second graph shows the latter part of the rapid recovery after the Corona dip, culminating around mid May. With gold meandering around and above $1700/Oz, there has been some profit taking after the miner rally, when I drafted next article: Large cap miners started lagging the yellow metal. The correction reversed mid June but it took to end June for the residuals to turn positive again as the gold price rally started its next phase. Gold broke above $1800 on July 9, while editing previous article: Miners now anticipating $1800 Gold. The gold rally took but a few days to make it to the Aug 2011 all time high (on Jul 24), with miners still responding. However when the yellow metal continued its rally, breaking above $2000 for the first time, residuals slid. Miner optimism was waning.

As gold peaked to a $2063 close on Aug 6, the HUI slid in anticipation of the correction which was to follow. Obviously the residuals nosedived into negative ranges. Breaking the normal correlation with gold is not common. It may happen when the general stock market is sliding. However that was not the case! Moreover, most miners were reporting 2020Q2 results above expectations.

A logical explanation could be mining investors taking profits because their sixth sense told them that the gold rally was about to reverse.  Plausible with hindsight ...  The pattern would repeat on Aug 18 as gold briefly broke above $2000 the second time, after which I posted on a forum that 'gold mining investors don't really believe in $2000 gold'. Yet residuals turned positive again later on, with gold back below $1950/Oz. The relief rally upon the outcome of the Jackson Hole meeting of central bankers brought gold back into the higher 1900's. 

Anyhow, keep in mind that analysis of residuals goes beyond investment strategy far into short term tactics with many speculative explanations paving the way.

* * *


No comments:

Post a Comment