A different view of GLD (part 1 of 4)

GM has charts sewn up nicely, but in the land of the bar list, what's new? Everyone knows about the decline in GLD inventory but not much attention is given to increases in inventory, so let's talk about it some. There were only three 'Add Events' in GLD during June 2013, with the largest add (214 bars) on the 13th June. In my bid to make the information more digestible, here is the summary as a handy info-graphic (yeah, I've got a new version of photoshop):

The most interesting day was the 13th June add, which had 74% dark bullion, which was neatly 160 bars. If you recall the pictures of the GLD vault from the Bob Pisani visit, that is basically two pallets exactly. The surprising element is not that the bars appeared to have traveled together, but the fact that the returning bars were only absent for about 14 days. We're still experimenting with different ways to slice and dice this data to help improve the utility of this new metric. A few weeks ago we looked at all the dark bullion being returned during an 'Add Event' and analyzed how long the bullion had been absent. In rough figures, here is the breakdown (by percent) of bars affected by the dark bullion effect we have seen, for the last two years:

Sample size is: 13,453 'dark bullion add events' (individual bars).
(Repeating dark bullion bars treated as separated events).

Basically, the chart suggests that the chance of a bar returning to the GLD bar list once it has been removed, diminishes over time - or more precisely, that 85% of any bars which are returned to the inventory, do so before 90 days are up. If historical patterns hold, most of the gold which has been removed in 2013, is not coming back. However, do bear in mind this graph will change as soon as we see a sustained period of inventory being added back to GLD (yes, it may still happen).

Riveting! Anyway, I have set myself the task of providing at least four different views of the GLD inventory. The aim is to show the inventory in ways not previously seen (i.e. unique to Screwtape Files), using different aspects of the 472 source documents we have for gold, as a form of edutainment. Here is the first view.

Something I have always wanted to do is identify how much of the stock in an ETF never moves. If you have a big enough space full of lots of the same thing - a vault or a supermarket, then you'll know how easy it is for some items to sit in the corner untouched for long periods of time. The local grocer has strategies for ensuring that stock always rotates since a use-by-date applies, but bullion doesn't have quite the same limitation in fact there is an advantage in not moving it at all. If we wanted to get conspiratorial then we might speculate the evil banksters have a motive for not moving said gold ... 'because it's all fake right?'. Well in a world full of wonders this possibility still provides a vector for investigation. Basically the question is - is there anything particularly special or different about the gold which gets added and removed vs. the stuff which remains consistently untouched?

Loosely put, this question arises from Bix Wier's suggestions of the ETF inventory being fake*, on the premise that if the stuff just sits in the vault un-inspected then it's easier to hide fraud, so let's take a look. Here is a view of the bars in GLD which have been present since our first records first begin (in October 2009), and which are still there today. For the graph below, the database query requirement is the signature for every bar** in the yellow section must be present in all 472 bar list issues.

So although there are bars which have never moved in the last 3 years, there is also a substantial amount of bullion which has moved in some fashion, whether this is being added, removed or even had their weights or serial numbers adjusted. Moving the metal or handling it to enable reclassification, increases the likelihood any fake bars would be exposed. This doesn't eliminate his argument but it does set a valid starting point for further investigation***. Please also note that a bar disappearing from the GLD bar list does not necessarily mean it moved from the vault but certainly did mean that it changed ownership, and we might presume the new owner is satisfied in some fashion that the bar is not fake.

Finally, this line does not represent some sort of magical limit, as the graph appears to suggest - it's simply an artifact of our availability/choice of data points. If the withdrawals continue then that lower bound will just get lower and lower, as the number of bars with 100% presence gets reduced. Likewise if we had a bar list from 2007 (we don't) and drew the line from there then it would also be lower. Gold bars which are present at Section B, were ones which were not present originally at Section A, so they were added at some point over the years. An astute analyst might ask whether we can track the churn for bars (i.e. the nature of the graph means that the bars present in Section A, are all absent by the time we get to point B). Can we see when they disappeared? Yes, we can, but you'll have to wait until Part 2 when we look at 'Snapshots' of individual bar groups.

Until next time,
Warren James

If I had a cool name like that, I would launch my own cereal brand.

* I first came across the 'road to roota' stuff in 2009. He has views on Molybdenum as fake silver and Tungsten as fake gold. He also conveniently ignores all the commercial and historical agreements which got the bullion there in the first place. Please note that I'm not attacking Bix with my article and my weeta-bix graphic is a visual gag that I've had in my head for quite a few years.
** The matches are done using bar signatures, a unique combination of BarSerialNumber, Refiner, Fineness and Weight.
*** I still have hopes that we can find (and prove) unscrupulous bullion bank behavior using the data, but so far it has not been found in the bars data. Frameworks like these are a good start.