Tuesday, November 09, 2010

No, GLD Is Not Overdue to Buy 200 Tons of Gold: An ETF Lesson Part II

Let me start this post off by saying that I own GLD - but I don't care in the slightest if you choose not to.  This is NOT a post trying to convince you to buy GLD instead of its competing instruments, or to tout the merits of gold investments.  It's simply a clarification of a horrendous misunderstanding by another very popular blog about how GLD works. 

I sent out a tweet last night accusing the very popular ZeroHedge of demonstrating a thorough lack of understanding of the mechanics of the GLD exchange traded fund.  A few of my followers asked me to write a rebuttal and clarify, so here we go. 

If you haven't read "An ETF Lesson: Part 1,"  please go scope out the first several paragraphs of that post right now to familiarize yourself with the process of ETF creation and redemption. Summarizing, when you buy an ETF, regardless of if it's the SPY, IWM, or GLD, the ETF manager does not receive your money and does not use it to go out and buy the underlying instruments in the ETF.  When you buy SPY, the SPY trust doesn't buy a basket of S&P 500 stocks, and when you buy GLD the GLD trust doesn't buy gold.  

Instead, you're buying from other market participants.  Whoever sells the ETF to you is the one who gets the money, just like any other stock - when you buy MSFT in the market, the money doesn't go to Microsoft Corporation - it goes to whoever sold the stock to you.   A great feature of many ETFs (including GLD and SPY) is the creation and redemption mechanism, which allows authorized participants (ie, big broker dealers) to arbitrage price anomalies in the ETFs.  

Taking GLD as an example: GLD, like every other ETF, has an NAV:  net asset value - this is the value of the underlying gold in the trust corresponding to each share outstanding.  If tons of people want to buy GLD and no one wants to sell it, the price of GLD might rise above the NAV, and at some point authorized participants will come in and short GLD to you, while simultaneously buying gold.  They'll then take the gold they bought, deliver it to the GLD trust, and "create" new shares of GLD which are backed by this newly delivered gold.  The authorized participant will use these newly created shares to cover the shares they shorted to you and close out their position.  The authorized participant profits because they shorted the GLD shares to you at a slight premium to NAV.

Which brings us to the ZeroHedge post:

"One of the completely unmentioned side effects of the recent surge in gold prices, has been the fact that one of the biggest holders of gold, the GLD ETF (presumably physical, even though it is kept in the cellars of HSBC in London, one of the two banks recently charged with a RICO suit for precious metal price manipulation) which as of close today held 1,294 tonnes, has not really bought any gold in over 5 months. The issue is that GLD's gold actual holdings, which feed right into its NAV, have been flat since June, peaking at 1,320.44 tonnes on June 29, and flat-lining and even declining through today. Since then, however, gold spot has risen by 14%. As the chart below shows, GLD tends to reindex its NAV in spurts, buying up gold during specific periods when gold goes up, notably in March of 2009, and between May and June of 2010. As of today, the trust's NAV per GLD in gold is at an all time low of 97.67. The bottom line is that GLD is now long overdue to replenish its actual gold holdings, net of redemptions. Assuming that GLD will increase its holdings in line with prior accumulations, when gold price surged, the ETF may soon be due to buy about 200 tonnes of gold. Should that happen, GLD will further increase its distance to 6th sovereign holder of gold, China, which as of September 2010 held "just" 1,040 tonnes. As to what would happen to the price gold if it is made known that there is a buyer for 200 tonnes of gold, we leave to our readers' imagination."

So, let's count the errors and misunderstandings here.  

1) Of course the GLD hasn't bought any gold in the last 5 months - GLD NEVER buys any gold, unless of course they decide to do a secondary offering.  

2) GLD does not "re-index its NAV in spurts,"  - I don't even know what Tyler is attempting to say here.  He provides a useful chart based on publicly available data from a spreadsheet that GLD provides on its website.  As you can see from his chart, there is usually some correlation between the price of gold and the number of tons of gold held by the GLD - this makes perfect sense:  it illustrates that it's likely that investors are using GLD as a way to gain exposure to gold.  As we talked about above, investors go out and try to buy GLD, APs arb the NAV premium and create new shares of GLD, the price of gold rises and the amount of gold held by GLD increases.

3) The trust's "NAV per GLD in gold" was, at the time he wrote this, 97.76% of its starting value of .10 ounces of gold per GLD.  It's the quantity of gold backing your GLD share.  This is different from the "NAV per share" which is $135.61 (As of 11/8/10) -  the value of the gold backing your GLD share.  The 97.76% number represents the fact that GLD must sell a little gold to pay its bills.  Originally, back at inception, the "NAV per GLD in gold" was 100% of .10 oz, which is to say that 1 GLD represented 1/10th of an ounce of gold.  Currently, 1 GLD represents almost 98% of 1/10th of an ounce of gold.  The 97.76% number is indeed an all time low.  And the new number will match or make a new all time low every single day. That's how it works.

4) "The bottom line is that GLD is now long overdue to replenish its actual gold holdings, net of redemptions."  Again, this is not what GLD does.  You can run your own chart of NAV per GLD in gold over time - it's a straight downward sloping line.  As I explained above, this is by design - GLD is constantly selling tiny quantities of gold to pay for the costs of storage and other expenses.  They don't "replenish" this number back up to 100% of the original .10 ounces.  The amount of gold delivered in creations and received in redemptions changes along with this number.

then it just deteriorates into pure hype:

5) "Assuming that GLD will increase its holdings in line with prior accumulations, when gold price surged,"  That's a horrendous assumption, because it doesn't happen.  Maybe investors tend to buy more GLD when gold prices surge - I've explained that as a possible scenario above already in #2, which can lead to the end result of GLD's assets increasing - but it's got nothing to do with gold that the GLD trust needs to go out and buy to "replenish" or "re-index" anything.

and the coup de grace:  the hyperbolic headline grabber:

6) "the ETF may soon be due to buy about 200 tonnes of gold. Should that happen, GLD will further increase its distance to 6th sovereign holder of gold, China, which as of September 2010 held "just" 1,040 tonnes. As to what would happen to the price gold if it is made known that there is a buyer for 200 tonnes of gold, we leave to our readers' imagination."  The only way the ETF will go buy 200 tonnes of gold is if they somehow get wind that there is flat out rabidly insatiable demand for a huge chunk of gold that someone would rather hold in GLD shares than in gold bullion, in which case they would file papers with the SEC and do a secondary offering.  I think the odds of a 200 ton secondary offering happening is roughly zero.

It's scary to see a site like ZeroHedge write such a blatantly erroneous article demonstrating a gross lack of understanding of the reality of how GLD works.  It's even scarier to read the comments and see not a single reader pointing out that there are mistakes.  Which is why I spend my time writing corrections to mass mis-education like this.

Now, we can even try to deduce some actual useful information from the chart that Tyler put together (and then misinterpreted) which graphs Tons of Gold held by the Trust vs Price of Gold.    Back in the first quarter of 2009, the assets of the trust rose while the price of gold remained steady.  What could explain this?  Perhaps this was a time when investors were clamoring to own GLD as a result of the turmoil in the financial markets.  Their buying interest resulted in lots of GLD demand,  and thus more arbitrage and share creation.  Why didn't the price of gold increase?  Again, during the turmoil, there were also ample investors who needed to raise funds, and perhaps gold was a source of funds for them (but not GLD, since if there was massive natural selling of GLD, the arb would go the other way, and result in a decrease in assets in the trust).

How about recently?  We've seen the price of gold skyrocket, while the assets in GLD have remained relatively flat.  Perhaps that means that investors are gaining exposure to gold via other methods - alternative ETFs, physical coins, gold stocks.  If investors were relentlessly buying GLD shares,  we'd expect arbs to short the GLD shares to them, while buying gold and creating new GLD shares with the trust.  The trust tonnage data doesn't show this.

If you take the time to understand the facts, you can actually draw some useful conclusions from data instead of relying on hype and hysteria to fabricate nonsensical headlines.


note:  Ignorant ranting comments will be deleted without prejudice, but genuine questions for clarification will be answered.


Anonymous said...

Thanks KD. I was right to be giddy.

Anonymous said...

Thanks for taking this on. I was thinking the same thing when I saw the Zero Hedge piece, but I never comment on ZH...

ZH posts are always hyperbolic and based on a half-understanding, but they are rarely this dead wrong.

P.S. "Whomever sells the ETF to you is the one who gets the money" is incorrect. It should read "Whoever sells the ETF to you is the one who gets the money". "Whoever" is the subject of the noun clause and thus needs to be nominative case. See this and then take her quiz.

Anonymous said...

This is all completely correct, except perhaps it gives too much credence to the highly unlikely secondary offering scenario. Gold flows in and out of the ETF holdings directly with issuance and redemption of shares through APs. Whatever gold the ETF needs to back the shares outstanding is already held by construction and does not have to be purchased in the market. The NAV in gold terms drifts slowly downward every day to pay the management fees.

My experience is you cannot expect any good sense on this topic or quite a few others from ZH, and I'm not surprised that people who understand that don't bother to explain there. Kudos to you for putting this together.

Taylor said...

Good post. I expect you'll spend the rest of your day deleting ZH commenters telling you how wrong you are.

Kid Dynamite said...

Nemo - thanks for the grammar ccorretion

Taylor - I won't delete comments that tell me I'm wrong - I'll correct them if they show any sensibility or reasoning. I'll delete "YOU"RE AN IDIOT - GLD IS A SCAM THEY DON"T OWN ANY REAL GOLD" comments.

scharfy said...

Hey KD,

How deviated does the NAV value/share have to get, in order for the creation/redemption mechanism to begin taking place?

How frequently does this occur?

Is GLD currently over extended or lagging spot at this point?

Just curious how volatile these ETF's are relative to their "underlying"... any thoughts would be appreciated

Kid Dynamite said...

scharfy -

GLD published a historical closing premium/discount to NAV chart:

that doesn't really help with intraday data, but if there are bigger dislocations, you'd expect them on the closing data, since it's harder to transact arbs then.

as for what it takes to induce arbs? tough to say - I'm not sure what their tolerance is right now. But another thing I should have mentioned is that this is something that brokers might do as a service for their customers. Ie, you want to buy 100k GLD, but you feel that doing so will have more price impact than buying spot gold - so you ask your broker to buy the spot for you, short you the GLD, and create new shares. He'll charge you a commish, and make sure he doesn't lose money on it (there is a fee to create/redeem usually) - but the GLD doesn't need to trade at a premium for it to work.

Kid Dynamite said...

Scharfy - the GLD website also has indicative intraday NAV that you can use to answer your own "how is it now?" question.

as i type this, it's trading about 20c cheap to NAV.

(for GLD price, use Google or Yahoo real time - the GLD's website price is 20min delayed, but their NAV is only 5-10 seconds delayed.)

Yangabanga said...

Great post, man.

wcw said...

> How deviated does [premium/discount] have to get, in order for the creation/redemption mechanism to begin

Creations and redemptions happen at the last possible moment, since there is a relatively large fee to create and redeem. But you do not care, unless you are the ETF sponsor.

As a potential shareholder, you only care about the arb process. Anyone can arb against the basket. With a 1-cent spread almost all of the time, the market is arbing GLD more-or-less constantly. IOPVs are only published every fifteen seconds.

If you want to play premiums and discounts, look to closed end funds.

Anonymous said...


Thanks for writing this KD.

I saw the start of the thread on ZH and shuddered a bit.


Jon Beyer said...

KD, thank you for taking the time to point out such a blatant misrepresentation/error on the part of ZH. Sometimes when I read that site, I feel like Mugatu, "I feel like I'm taking crazy pills". So that you don't feel like you are the only one in a sea of crazed ZH readers, I imagine that the error was clear to many readers. By the time I read the story on ZH, and was prepared to write a rebuttal in the comments, I saw that there were dozens of comments of the "GLD isn't really gold, buy PHYS" variety and decided it was worthless to attempt to educate Durden's fanboys. That site saddens me sometimes.

VC said...

What happens when GLD or SPY receive a net purchase/inflow of $10B?

Anonymous said...

Interestingly, shares in GLD can be seen as pre-1972 USD. It seems that there is a lot of misunderstanding about ETF's these days.

Kid Dynamite said...

VS - ETF Trusts don't receive net purchase inflows - that's the whole point. You can send your mutual fund an order to buy $10B, and the fund manager will have to go out and buy $10B of his portfolio.

On the other hand, if you try to buy $10B in SPY, you're buying from other people in the market. If you buy aggressively, and you drive it above fair value, people will short it to you, and buy either futures or the underlying index to hedge themselves. Then, they'll deliver that underlying basket to the SPY trust, receive newly created SPY shares in return, which they use to close their risk positions.

It's just like if you buy $10B of IBM - IBM doesn't get your $10B.

Anonymous said...

You should write a post about the Fed is pretending to buy gold in QE3 using HFT in an effort to fund Obamacare but actually it is a scam to help Goldmans. Watch that traffic boom!!


FischerBlack said...

The chart posted on ZH was useful and not something I had seen before.

There is definitely a positive correlation between spot prices for gold and the creation of GLD units. But these 'spurts' Tyler is talking about sometimes actually lag big moves in the spot price. That is, investor interest in GLD sometimes increases after big up moves and before the price flattens out for a while.

It would have been better if he said that the history of GLD spot/tonnage correlation suggests there is possibly about to be heavy demand for GLD shares after such a big up move in spot. Since heavy demand for GLD shares basically translates to net demand for gold through the creation mechanism, one might reason that this will lead to further increases in the spot price.

But looking at the chart, this isn't a very reliable analysis. Spot prices are often flat even as GLD tonnage increases, and spot prices often rise even as GLD tonnage flattens out.

So, one might be led to the conclusion that the chart shows that creation/redemption of GLD shares doesn't have an obvious effect on spot prices at all.

Of course, the gold bugs have been saying this for years, and use it as evidence that GLD isn't doing what it's supposed to be doing. But that's another error, for another time.

marketmania said...

great post---I hope you get a million hits on this...

Tim said...

Nice write-up. I've been arguing with some commenters on one of my Seeking Alpha posts for a day now:

I can't bear to read any of the comments at ZH that prompted you to write this...

Kid Dynamite said...

Scharfy - fyi, one of my boys tells me that they do this arb for pennies - ie, not dimes.

Anonymous said...

Great post - thanks for writing it!

Nitpick: "Prejudice" with respect to deleting comments would be taking action without thoroughly evaluating the comment. So you presumably meant rants will be "deleted with prejudice" rather than "without".... perhaps, if you're a fan of Apocalypse Now, rants would deleted "with extreme prejudice".

VC said...

KD, your understanding of ETF is correct but logic flawed.

GLD cannot grow to $56B NAV without inflow of physicals from authorized participants; the growth of GLD's NAV is far faster than the growth in gold prices.

Kid Dynamite said...

anon - i'm a math guy. my grammar can be wicked bad (Nemo already corrected me once)... but I think I meant "without prejudice" - like "all comments of this sort will be deleted regardless of who they are from" - but I'm probably wrong in the usage.

VC - you wrote "GLD cannot grow to $56B NAV without inflow of physicals from authorized participants"

absolutely - of course. that's a tautology (actually, it's not really - they could do a series of secondary offerings, but that doesn't happen. Growth in assets comes from inflow of physical from APs). My point was that you cannot send an order to GLD trust to buy $10B gold. it doesn't work like that.

then you wrote "the growth of GLD's NAV is far faster than the growth in gold prices. "

the entire point of Tyler's chart is the OPPOSITE of this currently. But in any case, they are two very different things, and need not be highly correlated. You can see Tim Iacono's post for more on this.

scharfy said...

Thanks for the reply KD and WCW.

I suppose thats its useless to look at a chart with spot gold versus GLD tonnage. Doesn't tell you anything does it? That correlation will not necessarily be very tight. (Do I have that right?)

The share price relative to the NAV/share should be pretty tight though (i.e premium/discount) According to the website most of the fluctuations are (at the close) razor thin. Plus or minus 1/4 of a percent. So the Arbs keep this vehicle in line it would seem.

This seems to be a really well designed product.
Is it safe to say you agree with that statement? Possibly the best way to own Physical (conspiracies notwithstanding)?

Kid Dynamite said...

scharfy - the chart of spot gold vs tonnage tells you about demand for the particular product - I think that's about it.

as for if this is the best way to own physical? I'm hesitant to even answer that question as I have ZERO interest in debating it with lunatics.

Goldbugs will say "if you want to own physical gold, buy physical gold." In my opinion, I agree with you - it's a very well designed product, and it's my choice for gold exposure. I don't care about rubbing gold coins on my nipples, I want something that will give me accurate exposure to the price of gold. In my opinion, GLD does that efficiently, accurately, and robustly.

Goldbugs will also say GLD is not physical - that's true - it's not physical gold - it's a piece of paper ownership in a trust holding physical gold. That doesn't keep me up at night. I think people who paid a 10% premium to NAV to own PHYS are insane (the premium is smaller now, as Sprott has saturated the market with secondaries and started to run out of paranoid people willing to pay a huge premium for his product -but that game was great for him for a while.)

If you want to own gold for 50 years, maybe you should buy physical gold (which brings it's own concerns: storage, verifying that it's real, insurance, etc...)

whatever helps you sleep better at night.

NOTE: non-intelligent replies to Scharfy's comment and my response will be deleted with prejudice - or without prejudice - or just deleted.

scharfy said...

Definitely an ETF newbie I am...

It struck me an good engineering to allow the AP's to arb away and create/redeem, as opposed to managed care. Also, no futures, cheap storage, S&P's name on it - all good.

Just reread ETF lesson part I and have a new appreciation.

Slightly OT - other ETF's like USO, aren't quite as simple, and have more active management, no? This would increase tracking errors? I am picturing Oil futures traders at the Nymex making a living off of retail oil ETF's rolling contracts.

Good post and Thanks...

oh yea, almost forgot -

GOLD BITCHEZ!!!!!!!!!!!!!!!!!!!!!

Kid Dynamite said...

scharfy: "other ETF's like USO, aren't quite as simple, and have more active management, no? This would increase tracking errors? I am picturing Oil futures traders at the Nymex making a living off of retail oil ETF's rolling contracts."

bingo. that's precisely what happens.

Anonymous said...

Great post. Would like to see you address the issue that ETFs can be shorted, and therefore the longs exceed the underlying physical. In a hyperinflationary environment, where the shorts on this product could go completely insolvent in a flash, how would that work out for those thinking GLD offered protection from collapsing fiat? Could not participating dealers buys shares and redeem them for gold until all there remained was an empty vault and a bunch of offsetting paper claims on the GLD price (for whatever that was worth to the remaining longs, given the empty vault!). I know, I know, in that world the price of GLD would be bid far above the NAV (especially by the shorts covering -- but aren't they insolvent?), causing other participating dealers to supply physical to the trust in exchange for GLD shares --- not sure I'd bet on it though! By the way, over 40% of my total net worth is in GLD, but the more these central bankers print the less comfortable I am with that.

Kid Dynamite said...

yes anon - I think you answered your own question. That was pretty much the topic of ETf Lesson Part I:

You can't redeem loaned shares. so the shares would get recalled, maybe you get a squeeze, and it self corrects. If anything, it results in higher prices though - not lower prices.

Tom Hutton said...

Great explanation, thanks KD. I have a question though. If the GLD sponsor "must sell a little gold to pay its bills," which it takes from its gold holdings thus making the percentage of the original .10 oz of gold per GLD share continually decline, doesn't that mean GLD essentially has something like a radioactive half-life? Won't the NAV per share eventually approach zero? Or is there some corrective feature that periodically bumps the NAV back up to 100% of the .10 oz.?


Kid Dynamite said...

Tom -

if you bought GLD and held it for long enough, and the price of gold didn't move a penny, then yes - the price would eventually get to zero. There is no "bump up feature" - think about it like this: originally 1 GLD = 1/10th of an ounce. Currently, 1 GLD = slightly less than that - so creations and redemptions are for slightly less than 1/10th oz per share.

by the way, though, if you buy gold bullion and pay to either 1)have it insured or 2) maintain a safe deposit box, the same thing is true - eventually you'll spend more on "carrying costs" than you did on the gold itself.

we can quantify this - really quickly, sloppy back of the envelope math: in 6 years, GLD's "representation" has lost less than 2.5% of 1/10th of an oz. That translates into .00042 oz/year, or about 58c per year per ounce at current prices. You can check my math - I did it sloppy and quick.

Anonymous said...

Hi KD,

Off-topic, but since you enjoy thinking about the entitlement attitude throughout today's society, I thought you'd enjoy this story. A high end grocery store in Boston ($8/dozen eggs!) closed and a co-owner blamed its customers for being too cheap (in quite a bratty way). Check out the first's the letter (from Google Cache). If you don't want to click on the links, then Google "don otto's boston herald."

Anonymous said...

great post. i am a loyal follower of zerohedge. i commit one of my three work monitors exclusively to the blog; it helps suppress the nocuous banter of cnbc blaring over the hoot. with that said, i'm glad to see a well stated refutation to ZH's post that confuses the mechanics of etfs. i am a trader on a bulge-bracket etf arb desk (not an admission i make with any sense of repletion); but i believe that etfs are widely misunderstood and it is refreshing to see a blog that tackles the subject with proficiency. i'm glad that zh's regress turned me on to this blog, which i have never viewed before. it is worth reminding readers, though, that "tyler" is not a single poster; there are at least 40 people allowed to post under his identity --- a good garden may have some weeds.