Analysis 20120903 – SLV Fund Jumpers (to Perth Mint)

This page contains the boring detail of the spread sheet data which would otherwise clog up the main article, and is a place where the spread sheet can be discussed as a separate thread. The Microsoft Excel Spread Sheet can be downloaded by clicking the link, and the detail of each columns is listed below:


BarID: This is my primary key for the ‘Bars’ table. The number is sometimes changed if I flatten and re-process the data, so the number itself is not particularly significant and refers only to the current set. The value for BarID starts at 1 and there are over 500,000 unique bar signatures for the entire database (Silver bars @ September 2012).

DateFirstAppearedInPerth: The date when the bar with this particular signature FIRST APPEARED in Perth Mint, being the date published from the document the record was found in. This spreadsheet ignores all subsequent appearances or disappearances (however because the Perth Mint repository is in an ‘accumulate’ phase, these bars are actually all still present). 
RefinerName: The (normalized) name from my list of Refiners. 
BarSerialNumber: The serial number of the LBMA bar, as recorded in the Bar List of the document in question. This forms part of the ‘bar signature’ and is the same on both SLV and Perth documents. 
WeightGrossOz: The weight, in gross ounces. This is (again) part of the bar signature and is therefore identical. Worth pointing out the SLV only retains records to one decimal place, and Perth Mint supports three decimal places. Since the weight is the same once in the Perth ledger then the weight record has come from data entry or data copy rather than Perth Mint re-weighing each individual bar. Fineness: Also known as ‘assay’, this number has been normalized to be expressed as a whole number rather than a fraction. Typically, Perth Mint reveals three digits “999” whereas SLV retains four digits (e.g. ‘9990’ or ‘9995’ or ‘9999’). This also means there may be more ‘similar’ bars if I relaxed the fineness comparison … i.e. maybe a bar in SLV had a fineness of 9995, and when it gets moved to Perth Mint it gets rounded down to 999 and would therefore show up as a different signature. For the purposes of this comparison, I have only processed an exact (direct) match which means 999 = 9990. 
PerthRefinerCode: Perth typically has a 2-or-3 digit capitalized code for the ‘brand’ which is stamped on the bar. Bron Suchecki was good enough to send me the ‘highly confidential’ list of codes to refiners, which I have preserved online here. He says that the IT team have it on their todo list to disambiguate the codes. This value is matched against my master list of refiners.
PerthDocumentID: My internal reference to the source document (does not change). 
PerthDocumentLink: The online location of the archived document (for where the bar was first listed in Perth Mint). 
PerthPageNumber: If you download the document, and go to page x then you’ll find the bar entry. 
PerthLineNumber: and for guidance, the exact line number as well (note: this value includes all line entries as encountered when the extraction process ran – i.e. the header of the page counts as line #1, so the position on the page as read by a human, is approximate). 
DateLastSeenInSLV: The date when a bar of this signature, was last seen in SLV (the document date). 
SlvDocumentID: My internal identifier, from same table (“dbo.Documents”) 
SlvRefinerCode: Brand name as it appeared in the SLV document. 
AssumedDaysInTransit: The difference in days, between the ‘last seen’ date and ‘first seen’. Basically this is the period the bars went ‘dark’ and were not showing on any register, we know the bars were being held somewhere, we just don't know where.
SlvDocumentLink, SlvPageNumber, SlvLineNumber: Same idea as above, the link to the archived document (quite large – these PDF files are around 40 megabytes), and the page number and line number that you’ll find them on. Same notes apply for line number – sometimes in the JP Morgan output the lines in the PDF actually split out into two lines – this may affect the interpretation of the line number, but the idea is the same – gives you a quick guide to where you can find the entry on the page.
FIRST THINGS FIRST: I really wanted to this in a way where the data/results were completely unambiguous and independently verifiable – for that reason the database was designed to allow each individual entry to be traced all the way back to the original source document, which is archived in the Microsoft cloud storage (Windows Azure platform). What this means is that anyone contesting the validity of the entries is on shaky ground because it is laughably easy for anyone to check. However there is certainly room to prove my data processing is incorrect – after all I rely on automated routines and it is not practical (or necessary) to manually check every single ledger entry. I tested this spread sheet with a spot check of about 10 entries and the primary extractions themselves were run three times with the same result. If anyone can point out an error in my processing I’ll remit you an ounce of silver via PayPal (at my discretion, depending on the value of the observation), as well as make my correction public. Most importantly, this is NOT a proof of all SLV silver being real; it’s purely a study on some specific numbers and data correlations. I suppose it’s obvious I don’t want anyone to take the study out of context, and if anyone wants to take issue with these numbers shown then I will bow to your superior knowledge and interpretations because I’ve spent quite a lot of time on this project and really done my homework.

Perth Mint Courtyard, from my camera - June 2011, picture is here to spice up an other-wise dull page.

Now then, the basic conclusion from this data set, is as stated on the main page – that 975 silver bars currently residing in Perth, used to be held in SLV. That amount is 915,956 ounces of silver – approximately 23% of Perth Mint’s Pool Allocated Silver as at 3rd September 2012. The amount was no surprise to Bron Suchecki, who confirmed in private correspondence that Perth Mint was getting a lot of metal delivered from London during that time period.
You’ll note that the metal left SLV in two large batches – notably 27th April 2011 (833 bars) and 27th May 2011 (141 bars). I don’t know the significance of those dates, maybe it represents a batch of redemptions from SLV into the Bullion Banking system, but the set reappears incrementally in Perth Mint on different dates. What’s consistent is the removal and appearance – i.e. there is no time overlap, which shows a lot of consistency – once the bar is gone from SLV it never appears again. BUT there is an exception to this rule, the very first bar signature in the document (Inner Mongolia Qiankun Gold & Silver - serial number “602423”. I could easily have eliminated this as a statistical anomaly, but have left it in to provide some discussion.

Mysterious Bar (line number one of spread sheet, highlighted in red):

This particular bar is problematic because it appears to exist in both vaults. On the second worksheet of the spread sheet, I have provided a ‘BarID 104400’ which is an extraction of every single ledger entry we have for this particular bar signature. It first appears in Perth Mint during September 2011, and is still present in August 2012. In SLV, it appears in the earliest source document we have on file: 08-Feb-2010, then disappears on the same date as the other big withdrawals, on 27th May 2011, but then a bar with the same signature shows up again nearly 6 months later in SLV (01-Dec-2011) and is still present in SLV @ September 2012.

Because the serial number is short (6 characters), the chance of this just being a bar in a similar batch with coincidentally the same weight, is high. The reason I haven’t lost much sleep over this particular bar, is that 93% of the 18,372 Inner Mongolia Qiankun bar signatures I have on record, all have 6 digits (which took me just half a minute to confirm – I love this database). The first two digits of that 93% all starts with either ‘60’ or ‘61’, so high probability they are using a similar numbering sequence. The bar reappears in SLV in this document on page 327 and is part of a practically uninterrupted sequential numbering range from “600001” to “609985” (pages 287-375). What you’ll notice if you look at that SLV document, is that a number of serial number + refiner duplicates exist here – sometimes up to three bars with the same number. I don’t have an explanation for that (yet) but what it appears to be is the refiner re-using the same serial numbers or perhaps the year for each is not being recorded as part of the serial number.

So, there you have it. If you have anyone who says that SLV has no silver, point them to this page and they may have to grudgingly concede that at least 0.5 % of SLV is/was real silver. Not a victory – especially since the very next line of argument (aka lazy rebuttal) would be that of course the bullion banks know they can’t deliver non-existent bars to the mint, so the mint gets real bars and everyone else gets paper. But you can bet that this is exactly the kind of thing I hope to find evidence for, from the data, and I hope that this article proves that I have the resources to do it in a boring, systematic fashion which can bear scrutiny and discussion.


Bron Suchecki said...

I assume your next analysis will be to find bars in GoldMoney or BullionVault which were previously in SLV?

I think over time the number of SLV bars which will reappear in other list will increase well beyond the 0.5%, making the position of those who think SLV is a total fraud increasingly shaky.

Warren James said...

@Bron, yes the Goldmoney and BullionVault comparisons are next. Because these are vaulted in the same location as SLV, I'm expecting a rather high percentage of 'fund jumpers'.

The benefit is that we'll be able to get a glimpse of some of the flow to Zurich and Hong Kong as well - that should be quite interesting.

Happily, because all the data is now in the same homogenous format I can re-use all the analysis structures including the PageNumber + LineNumber auditing structure. If someone points out an error then I get the benefit of the correction across the entire spectrum of data.

p.s. it's a weird sensation to dive into these figures and then watch one of those stupid bear-video-spin-offs like this one:

(it really demonstrates that whoever scripted those things had a completely one-sided agenda, out to capitalize on the momentum of the gullible investing public).