Monday 16 September 2019

HUI to Gold Regression: the 2008-10 recovery

What happened to stocks during the 2008 financial crisis needs little explanation. Whereas the broad stock market kept meandering along the lows till March 2009, gold recovered very quickly.
US Government bailouts such as TARP (Troubled Assets Relieve Program) abruptly doubled the money supply. The FED medicine to ease the pain in the broad economy undermined faith in the dollar and raised fear for an escalating inflation.

Gold was the single asset class to quickly recover to its former trading range with miners following in the slipstream. However after the precipitous plunge of gold miners during the financial crisis, the recovery entered uncharted waters.  Moreover, previous empirical relationship (HUI = Gold / 2) had been invalidated well before the October 2008 collapse set in.

In October 2008, the HUI bottomed around 150 with gold holding well above $700/Oz. The HUI/Gold ratio bottomed above 0.20. Today we would be delighted to see the HUI/Gold ratio back at 0.20. Despite the 2019 summer miners rally HUI/Gold still lingers on around 0.14, with the HUI barely upholding 200. Gold at $1500/Oz would imply the HUI rallying to 300 to regain a 0.20 posting for HUI/Gold.

As gold continued its ascent to reach the $1900/Oz all time high by end August 2011, it was crystal clear that miners again had uncoupled. The HUI was moving sideways, unable to lock in any gain. This more or less sets the limits for our quest to identify a HUI versus gold regression relationship during the recovery. From Nov 3, 2008 (with gold still at $729.5) we tend to find a regression till the yellow metal surpassed $1345 on Oct 6, 2010. These 23 months are the shortest time interval for any regression, yet it covers a gold interval over $610 or about 85% up from the low. Here is what it looks like:

(HUI, Gold) values (red dots) and the HUI / Gold best fitting regression line
With a 0.901 correlation this is worth considering, however the fit is less convincing than the more elevated linear correlations observed earlier (mid March 2002 till mid March 2008) or later on (Aug 2012 till Aug 2017). More precisely: the HUI values are below the regression line at both ends of the interval, while near $1000 there is a stretch of observations only above the regression line.

The regression line is defined as : HUI = 0.5815 . (Gold - 377)

When comparing the new regression relationship to the former one valid during the 2001-2008 initial gold bull: HUI = 0.6183 . (Gold - 97.7) especially the shift in the intercept makes the difference. This reflects an awakening to the ongoing trend that mining costs have been rising steadily partly due to lower ore grades in new mines.  This results in a higher gold price required to attain a specific HUI level, whereas the dependency of the shift of the HUI value upon a rise of the gold price is much less affected: the slope only is trimmed slightly.

In a 'Synoptic graph' the right axis is transformed as to reflect the linear regression relationship.
It looks as follows:

Gold (red graph) on the left axis are also the HUI regression values on the right axis.
Real HUI observations in blue on the right axis.
Finally it's useful to illustrate how the regression fails from the winter of 2010-2011 onward with the HUI range bound while gold steamed up to its all time high.

Regression failure: gold (red) on the left axis (USD/Oz)
HUI values on the right axis (transformed to reflect the regression relationship).
It would take almost two years of the HUI systematically losing ground relative to bullion, before a fresh linear relationship would set in, which unfortunately started describing how the HUI was leveraging down gold while it retreated towards its $1050/Oz low by the end of 2015.

* * *

Further reading on this topic

No comments:

Post a Comment