Catching up on SLV and GLD (and other observations)

I was hijacked by my day job for about two months! During that time however I managed to keep track of most of the conversations going on. If only I could have bought shares in 'opinions about what gold is doing', for this is surely reaching into bubble territory. I don't want to add to that, rather just bring a few brief updates from the bar lists study corner - these are the low hanging fruit while I get myself reorientated.

First, an updated graph on the Brinks Vault inventory from SLV (original article here). There's nothing very note-worthy here except that the Brinks London vault continues to be where all the action is. In February they had a 'reclassification' of some of the inventory, with 'Brinks London C' coming online (the yellow line). Brinks London A is losing inventory - I would not be surprised to see that go to zero in the near future **.

A good summary of the weekly events in SLV is covered over at, including an observation about the 17th April where SLV does a dodgy substitution.

In GLD, the loss of inventory is the talk of the town - some think it's the long anticipated 'bank run' on Physical gold, and hey ... this may be the case. But is it 'THE' event? I don't have the answers but I can at least give some observations from the data (regular readers know I'm trying to get a handle on the flow in and out of bullion based on the bar signature matches). There are two days in March where there were additions to GLD (highlighted blue):

Looking closely at those two dates, the additions were bars that we've seen previously in the data. Specifically, here is the breakdown (numbers are bars).

So for the two dates combined, on average over 40% of those bars which were added, had been seen in the data before. It suggests they were still in the vault all that time they were missing. The next question becomes: how much of the inventory which is disappearing in big batches now, is actually still in the vault at this time? In our wider data analysis we have seen that approximately 30% of all new additions are bars which were previously in the vault. This must be considered when analyzing the GLD inventory decline, because if the pattern holds then a large chunk of the 'missing inventory' (as of April 2013) is still in the vault at the current time, and I will be able to prove it with data at some point in the future as it becomes available. The same effect also means there is veritable craploads more metal in the bullion banking system than what is generally accounted for, so if it is indeed a bank run then it could still be a while before stress shows in the bullion banking system. I understand that the phenomena is described as 'players taking delivery via redemptions', an interpretation which is easy to propose from a top level view but is not directly supported by the data detail. These broad statements I do hope to quantify with my next set of graphs (but as you can imagine the definition takes a while to crunch).


Recommended Reading:

p.s. The gold bars appearing right at the end of that documentary 'The Secret World of Gold' (link) are from Japanese refiner 'Tanaka Kikinzoku Kogyo K.K.' which suggests the Canadians used stock footage.

Still image from the recent 'documentary' titled 'The Secret World of Gold'

I only watched the documentary with the sound off for gold bar images. You can verify the refiner stamp and serial number range here: Couldn't find any direct matches however JPM London V has three Tanaka bars which are '00625', '00626' and '00627', which first appeared '2011-12-29' (so I'm curious about the date stamp on these - maybe they celebrated new years eve early).

p.p.s. Defining 'decline of inventories' as bullish requires a fundamental admission of 'the ETF holdings are real'. Congratulations to all those silver and gold analysts who have reached this new milestone - including Andrew Maquire, GATA, etc. Because if the holdings are not real, then all this chatter is irrelevant, right?


Update 1st May 2013 : seems that for those GLD additions, over half of the new additions (not coming from previous inventory) were 'new gold stock', that is to say that the sequential serial numbers from the added Johnson Matthey bars were at the top end of the numbering sequence (which we talked about here). I'm working on improving the resolution for existing, old & new, and happily for gold, because JM bars make up a very large percentage of the ETF stock, we have a shortcut now for estimating age of bars. This is a good development, but the tricky part is still putting this into an aggregate summary. Stay tuned.

p.p.p.s. It is worth mentioning that the same Dark Bullion estimate is roughly the same for silver as it is for gold. i.e. the real population of silver bars swimming the background is 150% (my estimate) of what we see registered in the ETF's. Heady stuff, but that's the conclusions I'm coming to from the data. No shortage!

** The Brinks London A vault DID go to zero - on April 16th. I just finished clearing a processing blockage for the SLV import ... if I had cleared that on the day I wrote the article I would have been able to say it. Also if I had looked at the SLV data closely I would have spotted it as well.

Appendum 2nd May 2013 : I figured I would keep adding observations to this article as I come across them. Bron Suchecki recently made a foray into Turdville and challenged them with 'to discuss inventory decline as noteworthy, requires the inventory to be real' (link). This is the heart and soul of 'a framework for looking at ETF inventory which requires a choice on a specific position in order to advance the discussion, which I spoke of in this article. I didn't see a good rebuttal of that from the turdites - the best they could do was a 'why is it a binary treatment?' argument, which is a good one, given that it's always possible the ETF's might have partial real and partial fake - maybe what we're seeing here is the removal of 'fake' bar numbers in readiness for a run on real? The biggest problem with a non-binary outcome is that it becomes too easy to be intellectually dishonest - i.e. picking and choosing which inventory can or cannot be real without any kind of data, will lead to all sorts of mental dead-ends. For example, when challenged to explain why GLD appears to be draining and SLV is not, the almighty Turd offers: "Because SLV is a complete sham" (link), an answer which straddles that middle ground of well GLD is real but SLV is not ... actually to be fair to him he didn't really claim that GLD gold is real - but skirts the issue by focussing on the 'claims held' on the metal (link). The bizarre thing is that this still requires the metal to be real, but is it then still alleged? The short story here is that anyone using the silverbug forums to form a picture of the SLV or GLD inventory is going to leave you completely unsatisfied (quench your thirst with a dose of wisdom from Silverfuturist). Trying to prove or disprove what percentage of the GLD holdings is real or fake, is quite a challenge - where would one even start in this process? The bar numbers and refinery markings start to put a really big hole in any turd-like arguments because as soon as you approach the data you are making decisions on what to data discard and what to keep. Out of the 69 refineries who have produced bars which are held in GLD, is it even possible to decide that the bars produced by 'Atasay Kuyumculuk Sanayi Ve Ticaret A.S.' (Turkey) are fake? And if so, on what basis? Because you did research on the refinery or you know someone who worked in Turkey? Or because you know that the refinery closed down in 2010 so it is not possible they could have produced bar '2011275'? If you've done your research for that refinery then perhaps it is possible to chalk up a conclusion for those 213 bars, and now it's time to move onto the other 201,000 gold bars, to prove or disprove. Or you might go straight for the big guns and decide that the Johnson Matthey output (which constitutes nearly 50% of ETF holdings) is not real. To do that requires ignoring all the historical Johnson Matthey output as well as countless photographs of JM bars ... in short the process only works if you are incredibly selective with the data. Once you go for the big claims like 'SLV IS A COMPLETE SHAM' then you must do the same thing - i.e. ignore the entire output from Taiwanese manufacturer Solar Applied Materials as well as all the figures relating to silver imports/exports. If you are still listening to those tired claims of 'ETF not real' and 'shortage of silver' then shame on you because it's the most intellectually lazy stance that you could ever adopt, and it is bound to work against you in the long run.

Update 2nd May 2013: Bron provides answers all the Turdite questions here in his own summary post.


milamber said...



Can you put some numbers(estimates?) to two phrases?

"veritable craploads" equals??
"generally accounted for" equals???


milamber said...

And your take on the arbitrage explains all GLD inventory moves.



Anonymous said...


This must be considered when analyzing the GLD inventory decline, because if the pattern holds then a large chunk of the 'missing inventory' (as of April 2013) is still in the vault at the current time, and I will be able to prove it with data at some point in the future as it becomes available.

I don't think this invalidates FOFOA's and Randy Strauss' idea about reserve management. What would you expect if these two are right?

When one of the banks decides they need to increase their reserves, they would redeem GLD they already own (just converting one form of gold into another, say, if some investor asks for allocation of a big chunk) or first buy GLD in the market and then redeem.

On the other hand, as soon as they feel they have enough reserves, they would make this surplus reserve "ready for sale" by creating GLD with the option of selling these shares later (or perhaps because they want to return borrowed GLD in the first place).

Here is my best guess on the last days of GLD.

For an AP, GLD is almost as good as gold because they can redeem. I am saying "almost" because after redemption, they initially have unallocated gold, and theoretically, they might not manage to get it allocated. For an AP that is a clearing member, however, GLD is indeed as good as gold because when they decide to redeem, they will already know whether they can get allocation.

For an investor, however, GLD is not equivalent to gold, even if he is Soros or Paulson and holds multiple baskets. This is because if the APs stop redemption, GLD will just start trading at a premium or discount to the spot price, but no investor can force an AP to redeem on his behalf.

Should the APs stop redeeming, the gold in GLD is in effect trapped. So should TS ever HTF, the APs could just wait while the trustee officially determines market failure and winds down the trust. Now you can guess what would happen in this case. The trustee would sell the gold OTC and pay investors cash at the last know price of GLD. The allocated gold in GLD would then be suddenly available as additional reserves to those APs that are clearing members.

If you think about it, it doesn't make any sense to let GLD shrink down to zero first because it provides a windfall reserve only if you wind it down much earlier.


Warren James said...

@Milamber, 'generally accounted for' = standard counting of ETF inventory. Most numbers put forward are looking at nominal figures, Nick Laird of ShareLynx has the best inventory tallies. You can see one of his charts here. 'Craploads of metal' - yes apologies my vernacular is far from scientific. I hope to quantify this more precisely with my later studies, but for now my estimate is at least 150%. What I mean is take the maximum height of GLD inventory and add 50% more to that figure to get the 'population of bars' in the system. The figure will be slightly higher now that all this inventory has 'disappeared'. While some of those absentees are 'for melt', the majority are not. This theory will fly in the face of the people who cry 'metals shortage' - some folk are still stuck with the idea of only 1% of visible being real they are not mentally ready for the idea of 150% being real :)

@VtC, salient observations, thank you. I agree my data doesn't invalidate some market observations. I'm still refining on my theories but to answer your question I imagine that if FOFOA and Randy Strauss' observations are correct then the GLD gold should be draining out, never to return. Having some of the same bars show back in the inventory ... it's clear that some bars themselves were not taken elsewhere (redeemed), we're only discussing percentages at this point. Personally I think the likelihood is higher the banks will play silly-buggers with the inventory and price than it would be for an out-and-out freeze of the financial vehicle. My only role here is to show a bit more detail in this big gold riddle and to offer an alternate perspective. In any case, in 6 months or so we should have the necessary data to tell a better story on the missing bullion.

Warren James said...

I have a public correction on my above comment - FOFOA's observations does not require 'GLD draining out, never to return' - that was my own interpretation on freegold theory and it is wrong. More on this later, but basically Victor is correct that the observations here don't clash with any of the advanced asset management.

Anonymous said...


I agree. The reserve management idea is perfectly consistent with some bars leaving GLD and later returning to GLD. In the meantime, they were perhaps the emergency allocated reserve of some bank (or possibly even allocated to an investor and later sold by him). but if FOFOA is right, then some bars should disappear from GLD for good.

Now since inventory has been declining since the end of 2012, this is automatically true. But what can you possibly test?

One idea would be to look for examples of GLD inventory loss that are probably linked to some customer allocation. The bars involved in such an inventory loss should indeed never return to GLD. One example that comes to mind is the coincidence of a Sprott Physical Gold secondary offering which coincides with a GLD inventory loss of precisely the same weight. This was spotted by Kid Dynamite, and it is mentioned in a comment to my GLD article.

All this is the reason for the following question I have for you: When we see an inventory loss at GLD, how often do you see some bars added and a larger number of different bars removed from GLD, both on the same day?

Secondly, take a look at all bars that have entered GLD, but had already been in GLD and withdrawn at some earlier time. Then highlight those time periods during which GLD bars have left that would later return to GLD.

Perhaps you can identify periods during which bars leave that would return later, whereas there are other periods during which all bars that leave would never come back. This way, we could identify time periods during which there was an outflow of reserves.


Anonymous said...

After reading FOFOA, I actually have to backtrack a bit.

Even if some bullion bank took gold out of GLD when Sprott did the secondary offering, the bars removed from GLD may have entered only the working stock of the bank's allocated gold while they were selling Sprott some bars that were already located in Canada at that time (i.e. a loco swap). So even these bars might have entered GLD again at some later time.


Kid Dynamite said...

oooh ooh - i second this question from Victor:

"When we see an inventory loss at GLD, how often do you see some bars added and a larger number of different bars removed from GLD, both on the same day?"

ps - VTC - I think there's a very slim probability that you're going to find those bars being redeemed from GLD and added to PHYS... As you noted, it's easy to do a loco swap even if it's just to make sure that doesn't happen (Sprott certainly would want to avoid that)

Kid Dynamite said...

actually - I think we've discussed that "net" creation/redemption topic before..

I'm not convinced you'd ever see bars added and others removed on the same day in GLD, since there are unallocated account credits/debits in the "middle" of the process... the custodian would just net them out

unlike with SLV, where the creations and redemptions are via allocated metal.

Warren James said...

@KD, 'I think we've discussed that "net" creation/redemption topic before'

Yes, we have covered it (apologies forget which thread). I can also tell you that I've searched the GLD data for that pattern and not found it -> i.e. the daily movements are always net additions or removals.

We have also scoured the Sprott data for any gold bars showing up from GLD, there never has been (much to my disappointment), but we did for his Palladium and Platinum funds (we'll keep watching).

@VtC, those are good ideas, I'll try tuning my new graph to highlight the times of intense permanent removals, we're calling this 'lost bullion'. We have nearly 3 years of data now and every day gives us a little bit more resolution.