A guide to GLBSE-ETFs

[2weiX]

Running an ETF on GLBSE

I am 2weiX, a full time stockbroker / marketmaker / specialist at a German brokerage firm. I deal mostly with ETFs and ETCs. I also run a website called bitcoincommodities.com, where you can buy Gold and Silver for BTC. I have traded stocks of every part of the world (Australia, Canada, USA, Germany) and I’ve had 10year anniversary in 2011. The following will try to advise you how to deal with ETFs and maybe clear up some things. Please feel free to chime in.

A couple of people have been talking of issuing ETFs (and / or have already done so). I would like to take this opportunity to chime in and maybe clear up some perceived things about ETFs and also offer some advise on how to handle ETFs from an issuers perspective, a specialists perspective and a customers perspective.

Structure
An ETF contains a certain mix of stocks, bonds or commodities. The contents of an ETF will not change at the whims of the funds’ manager. That would make it a managed fund (“investement fund” or “hedege fund”). This is also what helps ETFs keep their expenses so low (it’s measured in TER = “total expense ratio”), often below 0,5% - since they do not have to pay expensive fund managers.
ETFs  are very transparent. It should be clear at all times what the constituents are and what stock / bond makes up for what percentage of the value. Most ETFs therefore follow an index. This could be the DJIA as well as some made-up index from the issuer (think MSCI or RAFI or DAXplus).  The issuer must publish the NAV (Net Asset Value) of the ETF once a day. This can easily be controlled by all customers by comparing the NAV against the underlying Index.

What does this mean for a GLBSE-based ETF?
The ETF needs to have a pre-set, never changing set of rules to follow regarding its constituents. This way, every investor will know exactly what he or she is buying with the full confidence that this will not change.

Issuing Shares
Theoretically (and practically, too!) an ETF can issue shares as long as there are shares of the underlyings available (even beyond that, when using financial derivatives such as swaps which I won’t get into now). Everytime the free float is 100% and someone wants to buy a share oft the ETF, the issuer simply transforms some shares of the underlying and issues a share of the ETF. This can also be done by the market maker / specialist if needed – he supplies the issuer with shares and in turn gets the right to sell a share of the ETF.  This has no effect on the price of the ETF (we will see why later on). All it means that there is one more share of  the ETF which represents 0,25 of share A as well as 0,0125 of shares C thru X etc etc. Of course, for big banks, this is easy to do. The have a huge portfolio and can just generate these all day long. Also, when they buy back their ETF, they can simply convert them back into shares so as to not be trapped with all their product, should a sell-off occur. This is also necessary in case of index changes. When a stock is removed from an index, the issuer “simply” takes those shares out of the ETF and puts some other stock into it.

What does this mean for a GLBSE-based ETF?
It means that the issuer should always only issue 75% of the ETFs they created in the initial offering, so when further demand arises in normal trading, the issuer can “create” more shares by simply selling his inventory.  The issuer should also always have a lot of BTC at hand to buy back shares of their ETF as well as to create new shares. When inventory runs low, the issuer will take (let’s say) 100 BTC, and buy stocks according to the rules of the ETFs structure. They will put them in their account and create 100 BTC worth of ETF shares, putting those into the market.
 
Pricing
Since ETFs are simply designed to emulate the value of an index minus the TER, there’s not much wiggle room for pricing. I’ve earned my ass of after Fukushima, since there is no supply/demand logik at work here. When the DAX (in my example) is at 5000 points, the ETF will trade at 50€. I will be able to sell it back at 68€ when the DAXhits 6800 points. As long as the DAX is at 5000 points, I will be able to buy at 50€ as long as I do not run out of money.Of course, once the issuer goes on to hedge themselves against rising prices by buying the DAX Future, the price will change. But by then, I will already have my shares at 50€. ETFs are normally priced around their underlying index (with a 2% max spread, depending on make, model, issuer etc etc).

What does this mean for a GLBSE-based ETF?
It means that you need to keep track of all your ETFs’ constituents’ prices, calculate the midspread for each (or the VWAP, or whatever) and thus approximate the median “price” for the whole shebang. Then put a spread on that price and put that quote on the market. You can also derive the bid price from the aggregated bid prices of all inherent shares and same for the ask. So when someone buys one share of the ETF @ 1 BTC, it should be exactly the same as if he had bought any of these shares individually – only that it might not have been possible to split 1 BTC over all stocks.

Profits / Dividends
There are two ways to deal with the dividends that the stocks in the ETFs pay. One would be to distribute them in cash between the shareholders. The other would be to aggregate them in the fund, which in turn increases the value of the fund.

What does this mean for a GLBSE-based ETF?
One would be to just add the BTC from the dividends to the ETF (so the price of the ETF increases) and buy more shares of the stocks. Let’s say BitCoinTorrentz pays 0,1BTC dividends per share. The ETF thus gets a total of 5BTC from BTT. Using this method, you would be able to buy 5 more shares of the BTT stock @ 0,99 BTC and keep the rest as cash reserve. The ETF would then consist of 320MM, 120PM, 55BTT and 18TG, increasing the value of the ETF in total by 5BTC. This needs to be reflected in the value of each ETF share. In my example, the NAV of the ETF (the value per ETF share) rises from 0,091353 to 0,093835 BTC!

The other way would be to divide the dividends accoring to the customers. This could become tedious. In my example below, each customer who owns one of my ETFs in essence owns 0,16MM, 0,06PM, 0,0275BTT and 0,009TG. Have fun dividing that! Of course, it’s really easy, just divide the divident payment by # of ETF shares and you can pay 5/2000 BTC to each ETF share! I personally prefer the first option.

Example for issuing an ETF

(example calculation xls here: http://www.file-upload.net/download-4181573/GLBSEETF.xls.html)

I am to issue 150BTC worth of ETF (2weiTF) consisting of

MergedMining    (0,141/0,156)
PureMining (0,3/0,42)
BitCoinTorrentz (0,85/0,99)
TyGrr   (2,3/2,85)

Each weighted at 25%.

I will contact the owners of these stocks and see if I can get them to issue some shares to me in a kind of private offering. If so, bingo! If not, I might just have to buy them on GLBSE. Since I know that I might have to satisfy excess demand in regular trading, I will spend a total of 200BTC to have 50BTC worth of ETFs in reserve.

So I buy
50BTCs worth of MergedMining = 320 shares of MM
50BTCs worth of PureMining = 120 shares of PM
50BTCs worth of BitCoinTorrentz = 50 shares of BTT
50BTCs worth of TyGrr= 18 shares of TG

I now know the total value of my ETF and can decide between two options.
I can either set an arbitrary IPO price (which will then in turn determine the number of shares I am about to issue)
I can also determine how many shares I want to issue, thereby determining the price.

I decide that I want to issue my ETF @ ~0,10 BTC to make it attractive for buyers.

I can now easily calculate the ASK side of my initial offering by using the current ASK sides of all the stocks in the ETF:

I have
320*0,156 + 120*0,42 + 50*0,99 + 18*2,85 = 201,12 BTC total in my ETF.
To keep the number of shares even, I decide to create 2000 shares, each valued at 0,10056 BTC, putting 1500 of them into the IPO.

Using the BID side of the stocks, I can also calculate the price at which I am willing to buy BACK my ETF:
320*0,141 + 120*0,3 + 50*0,85 + 18*2,3 = 165,02 BTC
Divided by 2000 shares, that makes for a BID of 0,08251.

The initial book of my ETF looks like this:  0 : 0,08251 / 0,10056 : 1500
Whenever the bid or ask size of one of the stock changes, my ETF price will change as well. It will also change whenever I allocate dividend payments etc. Try fiddling around with the prices of the stocks and see what happens when I buy 5 more BTT from a dividend payment.



So far so good, heading off to work now.
Please try to ask questions :-D

[1713292344]

1713292344

[1713292344]

1713292344

[1713292344]

1713292344

[1713292344]

1713292344

[2weiX]

keeping this for myself

[MPOE-PR]

The only problem that might arise with this (otherwise pretty much correct & basic laydown of EFTs) is that GLBSE is not very liquid.

[2weiX]

The only problem that might arise with this (otherwise pretty much correct & basic laydown of EFTs) is that GLBSE is not very liquid.

that is true and adress in that the issuers should keep a certain amount (proposed: a quarter) of the initial IPO at hand.

in my example, after the IPO had sold out, I would put, say, 100 more shares of the ETF on sale.
after those sold, I would take the 10BTC from that sale and create 100 more shares (probably less than 100, since my buying would lead to higher prices). I would also raise my BID side to 0,095 or so to indicate I am willing to buy back ETFs (earning 0,05 BTC per share!).

No ETF issuer will ever be able to just "out of thin air" buy 100 BTC worth of mining companies.

[MPOE-PR]

I meant something more along the lines that your model embeds some assumptions about the market (such as, for instance, that if I want to spend 50 bitcoins on a stock I will be able to find meaningful bids). As far as I can see these assumptions do not hold. Take the MPOE.EFT bid side :
1,000 shares - BTC 0.000006
287 shares - BTC 0.09
25 shares - BTC 0.091
25 shares - BTC 0.095

So you can't practically sell more than 30BTC's worth. The same is the case for buying, too, which means that there's no practical way to run a "real" (aka, perfectly transparent) ETF, they'll all be managed to a certain degree anyway.

Add to this the difficulties due to the relative market segregation (irl a bank holds shares from practically ALL issuers in one way or another - nobody is in that position on GLBSE) and you've got a mess on your hands.

[2weiX]

I meant something more along the lines that your model embeds some assumptions about the market (such as, for instance, that if I want to spend 50 bitcoins on a stock I will be able to find meaningful bids). As far as I can see these assumptions do not hold. Take the MPOE.EFT bid side :
1,000 shares - BTC 0.000006
287 shares - BTC 0.09
25 shares - BTC 0.091
25 shares - BTC 0.095

So you can't practically sell more than 30BTC's worth. The same is the case for buying, too, which means that there's no practical way to run a "real" (aka, perfectly transparent) ETF, they'll all be managed to a certain degree anyway.

Add to this the difficulties due to the relative market segregation (irl a bank holds shares from practically ALL issuers in one way or another - nobody is in that position on GLBSE) and you've got a mess on your hands.

Yes, that is true.

However - the assumption on GLBSE is also that the ETFs market cap won't be  in the 1000s, but rather in the low, low 100s of BTC.
It would be YOUR responsibility as issuer to buy (if needed) the whole damn market cap back at the price that is the correct one at that precise moment!

That's one of the key ideas of an ETF.
If your ETFs' NAV is 0,1, you better be ready to buy all 2000 outstanding shares back at 0,095!
It is, as an ETFs issuer, YOUR JOB to provide liquidity for your ETF.




Also, if for some reason someone sells you back, say 200 of my ETF, you can still liquidize some of the stocks.
If you find yourself with half the offering at your hands, either keep them for yourself, or sell 10% of your total stock to free some BTC.

[MPOE-PR]

You don't take my meaning. Suppose you have a 100btc ETF, with a 5 shares portfolio all trading around .1. Suppose somebody drops another half-restaurant meal into your fund (like 15BTC). You go buy shares in your 5 symbols, all now trading .16. If god forbid the other person in that restaurant also drops half a meal into your fund (ANOTHER 15BTC) you are now confronted with either taking all 5 shares to .25+ or else what exactly? Not selling?

the price that is the correct one at that precise moment!

Therein lies the problem: on GLBSE you can have either correct price or precise moment. You can't have both, you have to pick one.

[2weiX]

You don't take my meaning. Suppose you have a 100btc ETF, with a 5 shares portfolio all trading around .1. Suppose somebody drops another half-restaurant meal into your fund (like 15BTC). You go buy shares in your 5 symbols, all now trading .16. If god forbid the other person in that restaurant also drops half a meal into your fund (ANOTHER 15BTC) you are now confronted with either taking all 5 shares to .25+ or else what exactly? Not selling?

the price that is the correct one at that precise moment!

Therein lies the problem: on GLBSE you can have either correct price or precise moment. You can't have both, you have to pick one.

the simple workaround her is creating more than you IPO.
so if someone wants to buy another 20 of your ETF, you have already created those beforehand.
you take their money and create more shares of your ETF, replenishing your reserve holding.

if for whatever reason you cannot issue more shares, the ASK side stays empty. this is of course only owing to the limitations of the GLBSE market, as you correctly pointed out.

ideally, though, you will (and should) always be able top issue more shares.

[molecular]

If I understand correctly, this could be completely automated (after the constituents have been fixed), right?

[Nefario]

The only problem that might arise with this (otherwise pretty much correct & basic laydown of EFTs) is that GLBSE is not very liquid.

This is changing, GLBSE has been growing quickly over the last couple of months, and this is only going to increase significantly with the 2.0 release.

[molecular]

The only problem that might arise with this (otherwise pretty much correct & basic laydown of EFTs) is that GLBSE is not very liquid.

This is changing, GLBSE has been growing quickly over the last couple of months, and this is only going to increase significantly with the 2.0 release.

bye bye good profit from large spreads

[2weiX]

If I understand correctly, this could be completely automated (after the constituents have been fixed), right?

that is completely correct.
as an issuer, you could easily define a margin after which you would begin to liquidate some ETF shares, or create new ones.

the only problem is that so far there is no way to "delete" shares afaik.

[John (John K.)]

Good reading material. Thanks!

[2weiX]

maybe someone could issue a "GLBSE TOP 5 DIVIDENDS" ETF?

[mila]

maybe someone could issue a "GLBSE TOP 5 DIVIDENDS" ETF?

but how to exclude 2 constituents?

// based on my limited data there are 7 companies that actually paid any dividends. wait 8 (TyGrr-bank)

I'd rather read about destroying shares of ETF when it buys the stock back on the market.
we should probably ask nefario to describe the functionality that's available on glbse.
maybe it's sufficient to buy the shares back and keep them in the asset manager's account (they will not receive dividends, can't vote, seemingly don't exist if owned by issuer) or trading automation (rules for ETF to be compliant with, how thin is the border between ETF and managed funds?).

[2weiX]

maybe someone could issue a "GLBSE TOP 5 DIVIDENDS" ETF?

but how to exclude 2 constituents?

// based on my limited data there are 7 companies that actually paid any dividends. wait 8 (TyGrr-bank)

I'd rather read about destroying shares of ETF when it buys the stock back on the market.
we should probably ask nefario to describe the functionality that's available on glbse.
maybe it's sufficient to buy the shares back and keep them in the asset manager's account (they will not receive dividends, can't vote, seemingly don't exist if owned by issuer) or trading automation (rules for ETF to be compliant with, how thin is the border between ETF and managed funds?).

I was pondering that question as well.
You could just buy back the shares at another account and disregard them, but that would require serious bookkeeping, especially when you're about to issue more shares later on and on dividend payments, since you'd always have to keep track of "nonexistent" shares.
Also I see M.ETF offering to exchange the shares of the ETF for corresponding amounts of the underlying shares. Here, too, need the M.ETF to be removed from circulation.

The way I see it, ETFs need no voting mechanism since the thing should basically manage itself (at least from an investors perspective).

[2weiX]

I PMd nefario, the owner and lead dev od GLBSE.
Maybe he'll regard this as important enough to include in future revisions.

[mila]

I PMd nefario, the owner and lead dev od GLBSE.
Maybe he'll regard this as important enough to include in future revisions.

If you like playing with raw software, you might want to try out the "2.0"
and propose change requests how the needed things could be properly implemented.

I'd love to discuss alternative ways of doing things at the exchange
(and at least semi-privately half-open to users to reap relevant feedback)

I still hope he'll be opening the circles of co-workers on this and split the work between non-programmers
your expertise as consultant who can word change requests (what's missing in the standard etf manager's tool box; you've got knowledge)
his experience id developing and running the whole stuff
other active folks (afaik the 'BitcoinGlobal' - mother company of GLBSE had several founding members).

In the novel snowcrash the term 3 ring binder was used to describe rules and protocols.
your guide to ETF or it's a very promissing sprout (stub, Keimling ?)

Nefario fixed 'all known bugs' & could be in good mood & after some rest. Please share notes for comparison, if possible

[2weiX]

I do read the 2.0-thread from time to time and even tinkered with the GLBSE2 for a day - just do NOT have the time to commit.

I'd be more than happy to lay out how it "should be", though.
This thread is here for exactly that - figuring out how ETFs work and how they should work on GLBSE.

[mila]

I do read the 2.0-thread from time to time and even tinkered with the GLBSE2 for a day - just do NOT have the time to commit.

I'd be more than happy to lay out how it "should be", though.
This thread is here for exactly that - figuring out how ETFs work and how they should work on GLBSE.

I appreciate your comments, (re)read them carefully and reflect on them.
From my point of view it is already as good as it can get (reading your posts, from an expert voicing the "shoulds")
 and as my understanding will solidify I would like draft summaries and recommendations in the IT lingo for development
 and hope they will make it on the development roadmap. if accepted for development, write test cases and see how it goes from there

with the 2.0 being out 'any day soon' and only a finite number of important & missing stuff & promised new features that need to be developed ... one day requests from the community may be taken into account. while waiting for that to happen I can commit myself to closely watch glbse related threads (and poke nefario).

alles gute!