Posted by Frugal on 30th October 2009
Rob Kirby has recently alerted people at financialsense.com for the possibility of missing gold bars at GLD. Here is what he said:
An alert reader I communicate with [who shall remain anonymous] has been documenting the length of the published GLD bar list:
- on Friday, Sept. 25 – the list was 1,381 pages long
- on Friday, Oct. 2 – the list was 208 pages long
- on Friday, Oct. 9 – the list was 195 pages long
- then, on Wednesday, Oct. 14 – after questions were being raised about the strange machinations with the bar list in chat rooms on the internet – the list was back up to 855 pages long
Something TRULY stinks here. No explanation has been offered for the DRAMATIC swings in this list. Where gold is concerned nothing happens by accident.
Such anomaly if it’s true would be totally outrageous. Before going into total panic, obviously one should do the due diligence first. Therefore, I spent a couple of weeks observing the gold bar list at the GLD, which can be downloaded right here from spdrgoldshares.com. I was able to obtain 2 different copies of GLD bar list. One is 855 pages long, and another is 1291 pages long, both of which can be downloaded from my site by clicking on the links.
The first thing that I noticed was that on the 855 page version, it is attributing to Bank of New York, while the 1291 page version, it is stating the rightful new owner as HSBC. This seems to be extreme slopiness on the part of GLD. Yes, HSBC has bought the ETF, but it has been quite awhile (I couldn’t google a link to verify the above, but only recalled reading GLD being bought out by the new owner.) Anyway, I don’t think anybody would make such a mistake in title regarding one’s own assets.
The second thing is obviously that the two versions did confirm what Rob Kirby said about very different length in the bar list itself. I was not able to get the versions with much shorter list. I tried Internet Archive Wayback Machine, and it didn’t have any archives for the bar list link. So I guess any archiving for observation will need to be done by oneself.
Once I begun to actually count the number of gold bars, it became clear to me that the page length has nothing to do with the number of bars. On the 855 page version, there are about 104 bars per page, while on the 1291 page version, there are about 69 bars per page. That makes the total number of bars to come out to be roughly the same or consistent with the number of bars stated in both documents.
Because of the minor differences that I still couldn’t reconcile by manual counting & estimation, I decided to write a short program (which can be downloaded here, requring a Unix system to parse) to actually go thru the entire document and count the number of bars, while also checking for possible duplication entries in the table. By the way, if anyone wants to know how to run this program, just send me an email.
Well, to no one’s surprises, the number of bars, gross weight and actual number of gold ounces are EXACTLY the same as stated in the documents. There is NO missing gold bars! BUT there are some 5000 duplicated bar entries (here is the duplicated list for the 1291 page version)!! I manually checked a few entries (such as 100341 JOHNSON MATTHEY, 813 NAVOI MINING, AA22338 JOHNSON MATTHEY) dumped out from my program, and they are indeed correctly duplicated. Besides, since my program verified the number of bars and ounces exactly, I think programming errors on my part is very low.
Then I googled the internet a little bit more. I found out that the problem of duplicated entries have been found before, and they were caused by not listing the complete bar number in the document. For example, Johnson Matthey before used two letter to encode the year of fabrication before year 2002. Therefore, the duplicated bar entries are only in the text, but not in vault.
Although I still felt that statistically having 5037 duplicated bar numbers out of 88941 bars or 5.7% is abnormally high to me, at least there is some plausible explanation.
From the investigation on the different length of the two GLD documents, I must conclude that the two other versions of 208 and 195 pages are more likely to be human errors. In fact, I believe that the same person Rob Kirby referred to on the Yahoo message board has posted and corrected his own extraction error in this very later post “Re: Barlist Missing 65,497 bars”:
….My earlier numbers reflected an issue with correct extraction…..
In fact, if there is any fraud or theft in GLD, I just don’t think that GLD owner will commit such big blunder of revealing it in a such stupid way. It will have to be more subtle and elaborate than this.
I personally like to read whatever Rob Kirby has to say. But like everything you read on internet, you should always take it with a grain of salt, making sure you have filtered out the dis-information from information. For example, on Kirby’s previous post claiming big physical transactions in gold at the end of September in gold futures market, I simply cannot find this trace of physical delivery in any of the reported futures market. That is not to say that it is not possible, but just that I cannot trust such information with certainty.
Despite the finding from my personal investigation, I will still not hold GLD, due to problematic custodian structures. In fact, if anyone doesn’t want to buy physical or too much of it, one should simply diversify the physical gold holding in various gold/silver physical ETFs: GLD, GTU, SGOL, CEF, SIVR, SLV, etc. This way in case one of the ETFs in the unlikely scenario of actual theft or confiscation by government, you will still be left with majority of your holdings intact (in paper money at least). It will certainly increase your transaction costs, but at least it’s far better to let an unpredictable future event to hit your financial well being. And I also want to add that nothing can beat physical gold/silver, because in the time of crisis (and heavy demand), there is always an added premium which is not available in the paper market.