My analysis from last year seems somewhat primitive – silver bars which swim in and out of the SLV ledger is no longer in question. For today’s article I present proof of what I’m calling ‘
Hard data is perhaps less interesting than the implications of the conclusion and the story behind getting that data. I have known for some time (from visual checks) that bars from SLV had travelled across to Perth; the real challenge was to be able to package that information to be traceable. But the computing component is straight-forward, the tricky part is formulating the framework of questions to interrogate the data. If you're interested in the background, read on ...
We all come to the marketplace with a different views and information. For anyone who believes that SLV has no silver then I hope the evidence in the spread sheet will be instructive. To assume the SLV bar records are all fake requires also believing that the Perth Mint took delivery of non-existent bars, and that both parties collaborated on the bar serial numbers and mutually agreed to keep the data in synch and so on. It should be self-evident that the easiest explanation is the most obvious one: that Perth Mint took delivery of some silver bars (as they tend to do) and that the bullion banks in London supplied them with some in the past (as confirmed by Bron), and that it took a little while for these to be shipped to Australia and show up in the fund inventory (as it does). So while this silver represents only a miniscule portion of SLV’s entire inventory (less than 0.5%), I hope to gain concession from even the strongest critics that the argument has now progressed to talking about percentages of SLV’s inventory being real, versus being entirely imaginary. Of course, there’s many ways we can slice this data, hopefully working towards ‘Phase 3’ of my analysis.
In June of 2011, I flew overnight to Perth to visit the Perth Mint and interviewed Bron Suchecki, manager for Analysis and Strategy at the Perth Mint who generously gave me two hours of his time to explain how the Perth Mint operates. I felt like a true journalist but I got more than I bargained for because modern bullion operations are incredibly complicated. One of the original purposes of my trip was to ask about getting a photograph of the vault but obviously for security reasons this is not permitted – heck photography is not even allowed in the foyer of the building. So this is the courtyard, from my camera:
I grossly underestimated the amount of data and work involved with this project. If you are interested in the mechanics behind the data collection and indexing, I have also written up a snapshot of the current process here, as a separate article (yeah, that’s two sub-articles off this main one). The data from the Perth Mint Allocated Pool Silver has been a useful benchmark, due to the fact that we have copies of the very first bar list published and every one subsequent. As far as I know I have the only public archive of all Perth Mint Allocated Pool Silver Bar List PDF files since it started. In recognition of this prestigious accomplishment, I have made these archives accessible to public record; the links are located on the main Bullion Bars index page.
I want to talk very briefly about limit of reading error before getting into some facts and figures relating to PSLV. For my current data processing, I am using a bar signature based on the following four data points (combined):
Refiner/Manufacturer: e.g. "Johnson Matthey UK"Because it’s a database I can adjust or tweak the signature and re-run the analysis at any time. The example above is a bar which was held in SLV in February 2010, but if someone adjusts the serial number later on, or corrects the weight to 1038.900, then (depending on how I calibrate the processing) my comparisons would register that as being a new bar because it has a new signature.
Bar Serial Number: e.g. "48339R"
Weight (in Gross Oz): e.g. "1038.100" (to three decimal places)
Fineness/Assay: e.g. "9990" (to four digits)
The bar signature above is relatively complex already but becomes even more precise with a longer serial number. The 2009 Project Mayhem analysis made a big deal out of the duplications it found with a signature of Serial Number + Manufacture, but generally that’s no cause for alarm because some manufacturers regularly uses four digits as part of their serial number and sometimes the year of manufacture is recorded sometimes not, so the potential for partial overlap is present.
Obvious data issues are easy to spot with the database, like this SLV document which has 7 full pages of the same serial number “KPR3584” with different weights (starts at page 4930). Anomalies like this are imperfections in the data and will affect results, but the sheer volume of the data (which typically runs in excess of 300,000 bars), should generally smooth things out.
The data quality is something I’ll be doing smaller studies on, but that generally doesn’t stop me from doing data comparisons such as taking a good look at Eric Sprott’s PSLV fund (which is actually the focus of this article)! Mid-June 2012, the Sprott Phyiscal Silver Trust re-published the updated bar list for PSLV and PHYS (after a gap of nearly 16 months). To date there have only ever been TWO public issues of the serial bar lists for PSLV. While I realise that maintaining a spread sheet for such a large list, must be difficult, it’s not very transparent. Let’s have a look at how Sprott compares with the update rates of other funds. In this chart, each mark represents a bar list being issued. The below graph is the document records we have for 2011. It shows that I only started downloading the GLD documents mid-year, and also for the first part of 2011, my records were kindly donated by the guys at http://about.sg/slv who happily shared with me their source files. They download their data once a week, whereas you can see my software records a document if it detects the source has changed. Basically what it means is that JP Morgan has the hardware and process in place to regularly update their massive list updated, we capture any deltas.
So my point with that graph is that we don't actually have much visibility into the bar movements with the PSLV (and PHYS). You can also see here the GLD records start in June 2011, where I added the location a few weeks after getting the SLV downloads started. The ‘number of bars’ is on a log scale because SLV has so many records it made the others look really tiny. This graph captures nicely the growth of the Perth Mint Pool Allocated Silver.
Anyway, the bulk of Eric’s bars use a refiner code similar to the ones listed on the CME website, however some codes remain unidentified and some bar records are not even flagged with a refiner’s name! For someone like me, this is highly distressing because it means I can’t crunch all the information and I have written to Sprott Asset Management asking them if they can supply the remaining information.
It’s one of those small details which people don’t seem particularly interested in, but here is the raw fact: 28% of PSLV’s holding cannot be verified as LBMA compliant. I have no doubt whatsoever that he holds the silver he claims, what I'm saying is that without the refiner information, it can’t be externally proven to be compliant. But apparently this has not bothered so many investors!
Apart from the bars which have no refiner flagged, the only codes I can’t identify are: KPPO, SFRU, SMJA and NMJA (which may be ‘New Mexico Jewellers Association’ but I can’t be sure), the rest I got matches for. Here’s what the PSLV origin-of-bars breakdown looks like @ September 2012 using the current documents. Using the database I’ve separated the two documents out, so that you can easily see which refiners were sourced for the bars added during the PSLV secondary (this level of detail is not apparent from the raw data and as far as I know I’m the first to study it like this).
|SOURCE A: PSL.20110204.152041.Sprott_Silver_Bar_List_02-3-2011.pdf|
SOURCE B: PSL.20120614.163336.PSLV-Bar-List.pdf
NOTE: When I reference data, unless otherwise stated, I’ll be
referring to SOURCE B, since it has all the records from A.
I was really disappointed to find no matches. I really had it in my head that because of the large sample sizes involved (PSLV is kind of a large inventory, and SLV is massive), and with records over 24 months, that there must have been at least one or two bars which had floated across the geographical boundaries and across vaults, similar to what we saw demonstrated with the Perth Mint bars. Even more annoying, I re-ran the processing several times and got the same result.
The comparison results are:
- ZERO matches on Refiner, Serial, GrossOz, Fineness. Initially I figured I found the mistake - after all PSLV has GrossOz weights to three decimal places, and SLV has only one decimal place. So I re-ran the comparison rounding Sprott's to one decimal place. Still no matches.
But if I loosen the signature comparison a little:
- 6 matches on Refiner, Serial, Fineness (i.e. same but different weights). The weights are in no way similar so these are circumstantial.
- 9 matches on Refiner, Serial, Weight (i.e. same but different fineness). These are the only comparison which could be close to being the same bars. But none of the data supports these signatures as representing the same bars - many of the serial numbers are short (which increases the probability of a duplicate signature, and of the two long serial numbered matches, the appearance in each fund is not consistent with being a
FundVault Jumper. i.e. one of those bars does disappear from SLV around 2011-May-27, but that is no good because it appeared in the 2011-Feb-02 PSLV document. The other disappears from SLV mid-April but then reappears later. This all indicates that the matches here are coincidental and can be discarded.
- 1670 matches on Refiner, Serial – this comparison was me clutching at straws to find any correlation. In this result set is the interesting anomaly detected originally in PSLV SOURCE A, by Joshua Gibbons. He identified many ‘Solar Applied Materials Taiwan’ bars which had an identical serial number to ones held in SLV (this is a separate mystery for later, but basically the simple conclusion is that the production run for two different SAMT batches used the same serial numbers).
To be honest, I'm stumped with these results. I concede that the current data indicates Sprott got most of his silver (both lots) from the North American geography almost exclusively. But the origin of the bars themselves are often quite varied - check out this query which shows the breakdown of best-match for country of origin based on the initial refiner location. There are many from Canada and United states and but just as many from other countries far away.
There's a few ways to interpret these results, and I have already had some basic discussions with Bron and Victor about it - their take is basically that the efficient market will source bars from nearby rather than incur the cost of getting it from other sources (that's the basic paraphrase anyway). What that suggests is that it was more efficient for PSLV to slurp up the bars already sitting around the United States rather than go to the expense and effort of shipping the bars from London across the Atlantic. Bron has also pointed out that Australia has no primary center for silver, so Perth Mint getting metal from London (causing the Fund Jumpers) makes sense based on ease of access, whereas for America it's a different story.
To balance this article, consider the following: I realise that most people thought the bars ‘were still warm from the refinery’. Well, they could only have been referring to certain percentage, because some of Eric’s silver hoard has bar numbers which have a specific year on them (~11%), and additionally some of the bars in PSLV are from refineries which no longer operate the brand (like Handy & Harman). Here’s the percentage of bars within PSLV where the year is part of the serial number (based on a cursory exact match of the first 4 characters). Not splitting hairs, it's just that SOME OF Eric's bars were as cold as a grave, despite what you heard.
|This is a picture of me,|
with my pet human.
p.s. I don’t actually own any gold or silver I lost all of it at one of GM Jenkins’s infamous “poker extravaganza” nights while heavily under the influence of what he claims was only alcohol in a particularly rash all-in-bet but ironically he had all of his winnings stashed aboard his 50-foot yacht and was last sighted at an undisclosed location (off Cape Coral, headed south) and reported a boating accident en-route to Bermuda.
[ Updated 14th Sep 2012 - trying to sort out the new terminology 'Fund Jumpers' has been reclassified as 'Vault Jumpers' which works better. This will help distinguish between the different movement types, which I'll explain in the next article]
[ Footnote: 23rd Sep 2012 - this article just got a mention in Ed Steer's Gold and Silver daily (thanks Nick Laird). It doubled the number of hits overnight. I need to just clarify that I think Ed Steer got me mixed up with Joshua Gibbons, who does the weekly analysis of SLV data over at http://www.about.ag/SLV, and I can't contact Ed directly. I won't amend my article above, as it is already correctly attributed, but just wanted to make sure that readers understand the context - Joshua's work inspired me for this project, but the two studies are not the same - my data looks at other funds and other vaults. FYI, I found the link About.AG's PSLV analysis that I mentioned above, it's here on the kitco forums. p.s. I've nearly finished writing up my next article follow-up and I'll add the link below once it's done. Regards, Warren ]
[ also see: SLV Database 4, where I look at the Goldmoney and Bullionvault data ]