GLD Added 6½ Tonnes in October 2013

Late reporting on this ... Last month (October), there was a substantial addition to the GLD inventory (523 bars), the size of which stood out because it was the first 'add' after a sequence of 'removes' but compared by amount to other 'adds' for this year, it only ranked 6th largest.

Quick comparison charting all currently known 2013 GLD additions.

An avid reader of "Harvey Organ's Daily Gold & Silver Report" (who shall remain unnamed) brought to my attention an interesting Harvey Organ footnote (my bold emphasis):



I have access to a lot of good data on GLD, so I (finally) found some time to investigate Harvey's indecision regarding the nature of the added bullion. Premise first: given that so much gold has been removed from the vault recently, simple probability suggests that at least some of the metal being 'added' will come from stuff which has been previously removed. We have explored this concept in the past, calling it the 'Dark Bullion Effect' (link), we also determined the 'background rate' is on average 30% - i.e. when we look at all bars added in total, approximately 30% of those bars (for GLD specifically) are found to have been previously listed in the inventory.

Consider the chart below which shows how many bars we have seen in the past but are no longer on public record - it keeps reaching new highs as the inventory plummets. The trend is upsetting if you consider the amount of great gold data simply disappearing from the public record.
Reaching new highs - missing bars. Limit of reading error +- 5%
Records include bar lists from 2009 (not shown on this chart).
Applies to GLD inventories only - doesn't include other ETF records.
80,000 bars is approximately 1,000 tonnes of metal -
a higher amount than what's been removed nominally from GLD.

Because the quantity of 'previously-known-but-now-missing bars' keeps growing, I keep expecting to see dark bullion rates also rising. This turns out to be a faulty assumption because this is definitely NOT supported by the data. This particular October 2013 addition contained only 20% dark bullion and all the other bars were new. Here's the breakdown.

See below for notes. Approximately 86% of bars added on this day,
were subsequently removed from GLD on November 6th, information,
which neatly illustrates an effect I now call Harvey's Dilemma.
  1. The returning (dark) bars are nothing special - the 23 x Johnson Matthey Inc were in GLD as early as October 2009, and were removed in 2012. The other 23 previously dark bars got removed again on 6th November.
  2. The 80 bars (1 pallet) from Johnson Matthey Limited (Canada) were introduced to GLD originally in September 2012, were removed on the 18th October 2012 before returning in this batch nearly exactly one year later. That same pallet also got removed on 6th November - i.e. added but only stayed for 3 weeks. The new These Johnson Matthey Limited bars are at the high end of the Johnson Matthey sequence, with serial numbers between 208,730 and 208,809 so they are comparatively recently-refined bars (the highest known number in that sequence appears to be 228,898). These new bars also disappeared again on the 6th November.
  3. [added 30th Nov. for clarity] The 214 new Johnson Matthey Limited bars appear to be very recent; numbered between 228,193 and 228,898 (which sets a new record for the highest known number in that sequence). Based on the known layering of this sequence, my estimate is that these bars were refined in 2013, qualifying them for being classified as 'hot new bullion'.
  4. The 202 SEMPSA bars all have a very high purity (average of .9998), with a serial number range from 17,573 to 18,184 which is a previously unknown range sequence but appears to be the 'most recent' production of a four and five digit series (the highest number we knew previously was 10,750). 134 of these bars were also removed on 6th November (with no apparently pattern for the separation).

For me, the question still remains as to why recent 'Adds' are not comprised entirely of bars previously seen, I have only two major theories:

  • "The Bars went East" The easiest explanation is that the gold is being shipped out of London to Switzerland/Asia, so 'of course it's not available to be added back'. While this explanation fits the data from other industry figures and observations (and makes good sense), as far as the bars data collection goes, we don't have any definitive proof yet - but we're looking.
  • "The Bars got Allocated or Leased" Based on the idea that heavy gold bars in size have less of a propensity to move around the world (based on weight/cost/security to ship), there is a chance that some of the gold gets allocated as soon as it moves off the GLD bar list. This would explain why a large portion of the bars are 'unavailable' to move back in when gold needs to be added. Once again, we have no direct visibility to this, apart from knowing that the same vaults do have sections for allocated, and that some pallets of gold we observe returning in multiples of 30 days which would seem to indicate leasing.

I strongly suspect there is room for both theories
- i.e. it would be wrong to say that 100% of the removed bars are going East, and once we get a sustained 'Add' to GLD we'll be able to prove at least some percentages, but not until then. Please feel free to add any comments or observations.

7 comments:

Bron Suchecki said...

Thanks for the update Warren. I find the one pallet of JM bars interesting - in Sep 12, out Oct 12, back in a year later and then out again. I'd say a high probably it will come back again. Must be right up the back of the vault or something so it never gets used for shipments out of the vault and probably just goes into the BB's float/physical reserve.

Given the behaviour of this pallet, I'd agree that both of your theories are valid.

costata said...

Great to see you posting on the bar list data again Warren. Cheers

Warren James said...

Glad to hear it Bron. That pallet's absence was 372 days — there's probably enough pallet-based-30-day-multiple absences in the dark bullion data to guess the probability of bar blocks returning, but the volume is not significant enough to make any useful price predictions on ... especially if the price drives GLD inventory levels rather than the other way around.

I have made a correction to the numbering of the Johnson Matthey bars - the text wasn't clear about which block was being referenced (but it doesn't change any part of the analysis) - the new bars really are quite new.

Warren James said...

Thanks Costata - I have three more draft articles on GLD that need finishing, and then I'm going to focus back out on worldwide ETF Inventories again. In my focus on GLD I have also neglected SLV, which I should return to.

p.s. I'm using SQL Server 2014 CTP 2 now, which is substantially faster for processing these big blocks of data.

victorthecleaner said...

Warren,

very nice work. I'd say the fact that you are finding little "dark bullion" but a large turnover (1000 tonnes) through GLD bewteen 2009 and today, point towards what people call the Coat Check point of view, doesn't it?

If it was all arbitrage, the added bars should primarily be bars that have been removed previously, no? Why would HSBC allocate a fresh bar to GLD if they still had one at hand that was taken out of GLD during some past arbitrage operation? I'd say they probably won't.

Victor

victorthecleaner said...

... in order to keep their logistics efficient, not have too many actual physical movements of bars, and even in order to keep the stuff that belongs to GLD reasonably close together.

Victor

Warren James said...

Hi Victor,

I agree it would support it, but with the caveat a percentage would apply (and I hope everyone will admit this isn't a binary issue). At the minute I am pursuing a theory which would put the current inventory split at 70% coat check and 30% arbitrage (with arbitrage decreasing).

I have recently added some more historical bar lists from 2005 which raises the 'missing bars' value higher, as well as increasing the amount of dark bullion identification (article is half-finished).

Despite the data giving us a good look into the vault, I'm concerned at the lack of data around bullion shipping movements. I don't have a firm theory yet on how the 'in' and 'out' works - the 'removes' appear systematic, while the 'adds' seem quite random and happenstance.

I'm also updating the database to include recent ETF Securities data so that should give us a better aggregate view of the HSBC vault.

Short answer, still many more questions than there are answers.

Regards,
Warren