June 2011 Archives

ETFs, Get Off My Lawn!

By druce on June 29, 2011 11:57 PM | No Comments | No TrackBacks

Two diametrically opposing story lines, both true.

In one corner - ETFs are making mutual funds obsolete.

In the other corner - ETFs are a potential source of systemic risk.

I love ETFs. ETFs have several compelling advantages vs. mutual funds:
 
  • Liquidity - An ETF can be traded all day, with limits, stops etc. A mutual fund can only be traded once a day at the closing price, and you have to place your order for better or worse well before the market close. (Fallout from the late trading scandals - a stunning breach of trust by much of the mutual fund industry).
     
  • Tax efficiency. ETFs are generally taxed just like a stock. When you buy an ETF, you set your own cost basis. When you buy a mutual fund, you buy into the existing cost basis. You could buy a mutual fund in January and lose money all year; then get a 1099 at the end of the year that shows they sold some positions at a profit, and you have a capital gains tax bill. If you're a buy and hold investor, you get screwed a bit - other fund holders trade in and out and generate capital gains distributions and tax bills you have to pay.
     
  • Diversification. There are interesting ETFs like GLD (gold), SLV (silver), with exposures that you otherwise couldn't put in your IRA account. Or, you can short ETFs in your margin account. In addition to giving you diversified exposure, they are potentially more liquid and less exposed to short squeezes than single names.
     
But ETFs can be complicated and have pitfalls investors should be aware of.
 
  • Pricing and liquidity. Suppose your fund owns positions in Australia, but your mutual fund is traded at the NY closing price. The mutual fund would typically use the Australia end of day price to calculate an NAV. However, if the market in New York has a big move, that's a stale price. Not normally a big deal, as long as your fund limits active trading scalpers who might try to take advantage. But for any mutual fund or ETF, pricing can be a little tricky.    In the case of an ETF, arbitrage should keep the ETF price in line with the underlying, as long as the underlying is liquid. If there is a rush into SPY, driving the price too high compared to the underlying S&P stocks, market makers can buy the underlying stocks, deliver them to the ETF custodian in exchange for SPY shares, creating more SPY shares and a small arbitrage profit. S&P stocks are highly liquid, so the ETF spread to fair value is unlikely to get large. In the case of ETFs in less liquid markets, drawing the same conclusion may not be appropriate. IWM, the Russell 2000 ETF , has a market cap of $16b and traded 2oom shares or $10b a day during the worst of the financial crisis. Ballpark, those numbers are over 1% of the market cap of the Russell 2000. These are pretty thinly traded stocks. Big moves in and out of ETFs of less liquid stocks could have a significant impact on their prices. The arb is harder to execute. In a fast market, trades in the ETF could be at much larger spreads vs. fair value.
     
  • Shorting. There seems at least a theoretical potential for shorts to disrupt ETFs. In the case of a stock, short-sellers can borrow the stock and sell it. If short interest is high, the supply of shares can effectively expand by the number that can be borrowed and sold short.  Since no share can be borrowed more than once, the hard limit on the supply of shares is 2x the float long, 1x sold short, and the market as a whole is always net long the the available float. Usually shares become hard to borrow long before the hard limit, since all shareholders don't lend their stock. The twist with ETFs is that the float varies as ETFs are created and redeemed by the custodian. Suppose there are 1m shares of an ETF, and short-sellers borrow 500,000 and sell them. Now there are 1.5m shares held long and 500,000 shares sold short. Now suppose 1.1m long holders show up and try to exchange their ETF for underlying shares. The ETF custodian only holds stock reflecting the 1m issued! Conceivably the ETF sponsor could have to suspend redemptions. In practice, the sponsor should be able to help maintain an orderly market. But it's quite easy to imagine confusion and volatility in an ETF that attracts a lot of short interest.

     
  • Levered ETFs. Suppose you have a 3x levered ETF. It's priced at $100, and owns $300 worth of some underlying (and borrowed $200). The underlying goes down 10% today to 270. The ETF goes down to 70 (270 - 200 leverage). It rebalances to remain 3x levered, so it sells $60 worth of the underlying and pays down the loan. It starts off day 2 worth $70, owns $210 worth of stock (borrowed $140). Now the stock goes up 11.11%, ie back where it was at the start.  You only made $23.33 back. So the underlying is now unchanged, but you lost $6.67. It's quite conceivable that in a volatile market, both the bull (levered long) and bear (levered short) funds will lose money. You had better know exactly what you're doing with these inverse and levered ETFs, and rebalance to maintain desired exposure (potentially increasing trading costs and taxes).

     
  • Counterparty risk. Many of the ETFs with leverage and oddball exposures simply write a swap with a counterparty to achieve the desired exposure. If your counterparty goes bust (anyone remember Lehman, not to mention Refco?), your ETF may not be worth much (hopefully any loss will be limited to profits since the last repricing). Read your prospectus carefully.

     
  • Active ETFs. An active ETF is similar to an index ETF, except the active manager determines what is in the index and can make frequent changes. In many cases successful managers are launching ETFs which are supposed to shadow another fund or strategy. The manager has to be buying and selling the underlying on behalf of both the ETF and the traditional mutual or institutional fund, and publishing daily portfolios. It seems far-fetched that this can be done without someone getting front-run, and without limiting the ability of the manager to make changes in the portfolio without tipping his hand or breaking the arb process.

     
ETFs are awesome, and yet it seems a safe bet that an ETF is going to blow up badly. The great and crazy thing about the free market is that anything worth doing is worth overdoing.
 
Unless you really know what you're doing, best to stick to 
  1. Safe sponsors, like Vanguard, Barclays, iShares (State Street)
  2. Index ETFs only (not active)
  3. Liquid underlying securities
  4. No leverage.

The great Bitcoin robbery

By druce on June 20, 2011 11:30 PM | No Comments | No TrackBacks
Coin Collection

Image by flash2 via Flickr

Bitcoin is a fascinating experiment: digital currency that doesn't depend on a central authority.

How it works: Suppose you put money in the bank. You open an account, and the bank has an electronic ledger that keeps track of how much money is in each account. Bitcoin is the same, but without the bank. Everybody who has an account effectively keeps a copy of the ledger (the whole thing or whichever parts they care about). They can then share it with everyone else, like BitTorrent or Gnutella. If they want to transfer funds to someone else's account, they add a transaction to the ledger and sign it digitally using public key cryptography. Then they share the updated ledger with the network. Nearby nodes pick up the changes, pass them on to people they are connected to, and the updated ledger propagates throughout the network.

It's as if, instead of the bank keeping track of accounts, there was a big whiteboard in the public square where people tracked balances in each account.

Without going into the gory details, there are some interesting ways the network cooperatively determines if someone tries to spend the same Bitcoin twice (basically voting which transaction is valid), and cooperatively grows the quantity of Bitcoin at a hard-coded rate (basically voting on the denomination of new Bitcoins, which anyone can create - 'mine' - by cracking a compute-intensive problem whose complexity limits the growth rate)

Here are a few of the things that could go wrong:

Storing money. Remember that digital signature you use to transfer money? It's like a numbered Swiss bank account - lose it, and you won't be able to transfer funds. If anyone else gets hold of it, they will be able to authorize transactions out of your account. This guy supposedly lost about $500,000 (at a since-exceeded exchange rate), and there are supposedly viruses looking for the key to your Bitcoin account.

Security, part 1. The ledger is public, so anyone can see all the transactions. As I understand it, you don't need the full ledger to transact, you just need enough of it to communicate your transactions unambiguously. But anyone who is listening can construct the entire history. From there, it's just a matter of associating the account numbers with people. Which of course, a lot of people have an interest in doing, and the ability to do so.

Legalities. Clearly the long arm of the law will not stand by if people do work for Bitcoin, spend Bitcoin to live, convert Bitcoin to dollars, and leave the tax man out of the loop. And of course legitimate businesses will not adopt Bitcoin if contracts are not enforceable. So far Bitcoin is no more than a novelty and object for speculation. But to the extent that Bitcoin gains traction for transactions, it seems likely it could initially be used for illicit transactions where legal enforceability is not an insurmountable obstacle. And it seems very possible that the response would be to ban the use of Bitcoin, or conversion to dollars, with the pretext that it's only used by drug dealers, terrorists, child pornographers, and tax evaders (last but by no means least).

Security, part 2. Today, those digital signatures may be impossible to fake without stealing private keys. But computers get more and more powerful. The time will come when the system is breakable, what then? If you had a central authority, you could say, everyone has to upgrade to a new version on this date, just like they exchange old banknotes for harder-to-forge designs. Without a central authority, it seems quite messy. You could wake up one day and the cryptography might have been broken, the hackers are able to convert some Bitcoin to gold, everyone realizes it, rushes to convert, and the currency very quickly becomes worthless. In anticipation of that, competitors could pitch an alternate version, and try to profit by getting people to convert...which could merely accelerate the collapse. Somehow, the developers would need to come up with a backward compatible algorithm and manage a graceful transition - a tall order (The struggle to achieve IPv6 adoption seems simple by comparison).

Organized attacks. That migration thought experiment shows that you don't need to break the encryption. Just get enough people to run a client you can control, and you can control the algorithm. Or, more crudely, get enough bogus activity on the peer-to-peer network, for instance using botnets, and you could swamp it with a denial-of-service attack. Even decentralized networks organically build up nexuses that can be disrupted, like the exchanges that convert them to dollars. Governments and competitive systems could have strong incentives to wreak havoc.

Ultimately, the value of a dollar is that there is a large nation of people willing to exchange goods and services of value for it - about $10 trillion of GDP. There is also a centralized authority that enforces transactions in dollars with the force of law, and when necessary, physical force through police, courts, and jails, and defends its people and the dollar economy with military force. There is a central bank with a mandate to keep the value of the dollar as stable as practical (in theory).

Bitcoin doesn't have any of those attributes. In some ways it seems to embody the worst of all worlds: not backed by anything with tangible value, nor by a central authority to keep it stable. It's traceable, exposed to potential security issues, and prone to speculation and volatility. 

If you want to replace the current highly imperfect monetary arrangements, whether you're a gold bug, hard money advocate, crypto-anarchist, or libertarian, you should use caution not to replace them with something worse.

Bitcoin seems like a novelty virtual good and object of speculation with limited convertibility. If you take the predetermined maximum 21m Bitcoin and multiply it by a recent high price of $30, and apply an interest rate and some velocity, market participants are pricing in a rather significant Bitcoin economy in a few years.

Fascinating, potentially influential experiment, possibly a cautionary tale in the making, and definitely caveat emptor.

100 traders, 100 boxes, part deux

By druce on June 7, 2011 8:46 PM | No Comments | No TrackBacks

So, where were we? Ah yes, yesterday we posed a math conundrum:

A hedge fund manager puts 100 traders in a room and instructs them:

"On the trading desk, there are 100 boxes. Each box has one of your names. You can go [one at a time] onto the trading desk and open any 50 boxes you choose, to try to find your name. If every one of the 100 traders in this room finds his or her name, you will each get a $1,000,000 bonus. If anyone fails, I will crush all your $100,000 BMWs to create my modern art masterpiece. You can devise a strategy before anyone leaves the room, but once a trader has opened the boxes, you must leave the trading desk exactly as it was before you entered and cannot communicate with anyone else."

Questions: Should they take the bet? Is there a correct strategy? What is the probability all 100 will each find their name?

Intuitively, that seems like a terrible bet. They can't exchange information during the experiment, and without information on what the previous trader found, if each trader goes out and picks 50 boxes at random, he or she has a 50% chance of finding the right name. The chance of two in a row succeeding is .5 * .5 , or 25%. The chance of all 100 traders succeeding is 0.5100, or about 1 in 1030. If you did this experiment every second, it would probably take longer than the age of the universe to win. The chance of success on the first try is basically 0.

But before we give up, let's look at a smaller problem. The general problem is 2n traders and 2n boxes and each trader has n trials. What happens for the simplest problem, if n=1 and there are 2 traders and 2 boxes, and each picks one box?

If each picks independently, the probability is .5 * .5 = .25%. But suppose Trader 1 tells Trader 2, "I'm going to pick box 1," then which box should Trader 2 pick? Clearly if Trader 1 is successful and finds his name in box 1, then if Trader 2 picks box 2, he will find his name there. So if Trader 1 is right, Trader 2 will be right. The odds just went from 25% to 50%.

Now look at n=2, 4 traders and 4 boxes, each picks 2 boxes. They can agree that they will each first pick the box corresponding to their number - Trader 1 picks box 1 first, Trader 2 picks box 2 first, etc. But what box should be picked second?

Suppose they choose this strategy: If Trader 1 finds the name of Trader 1, he wins and stops. If he finds the name of Trader n, he opens box n. It's just a random box, and in 2 tries he still has 50% chance of being right.

The point of doing that is: on the first pick Trader 2 avoids picking a box Trader 1 picked and won with on the first pick. And on the second pick Trader 2 again avoids picking a box Trader 1 picked and won with on the second pick, since they both can't find the same answer on the first try and then pick the same box on the second try. So if Trader 1 is successful, Trader 2 successfully excludes some boxes that wouldn't have worked.

Let's examine in more detail the cases where Trader 1 got lucky and won, and then Trader 2 comes along. Then there are 3 cases:

Case 1 - (Best case) Trader 2 finds 'Trader 2' in box 2 and wins!

Case 2 - He finds 'Trader 1'. Now if we assume Trader 1 won, that could only happen if he got here by finding Trader 2 in box 1! Trader 2 picks box 1 and wins!

Case 3 - He finds 'Trader 3' or 'Trader 4'. Now, if Trader 1 won on the first try by finding his name in box 1, we have a 50/50 chance of picking right. If Trader 1 won on the second try, we know he didn't win in box 1 or 2, and we can't be pointing to the same box of 3 or 4 he opened to win with on the second try. That leaves only the box Trader 2 is pointing to, so Trader 2 picks that box and wins!

The beauty of this is, if Trader 2 won and and exposed our box on the first try along the way, we win. If Trader 2 won and did not expose our box, we never pick a box Trader 2 exposed. By excluding the boxes Trader 1 picked, we have much better odds of picking the right box. In fact, if Trader 1 wins, Trader 2 wins 5/6 of the time.

By enumeration, our odds of winning all 4 times are 5/12, or 42% (left as an exercise), not as good as before, but still far better than 0.54, or 1/16.

The key to the exercise is to observe that each trader travels along a cycle. A cycle starting at box i will find Trader i's name on the j-th try, where j is the length of the cycle. If there are no cycles longer than n/2, everyone will always find their name. If a previous trader exposed your name, you are both on the same cycle and you will eventually hit your name. If you are on different cycles, you will never overlap with their choices.

If you try this on 100 boxes, 100 traders, 50 trials, you get approximately 31% success. Given that you are risking $100,000 to win $1,000,000, it's a good bet, and this is a good strategy (if the hedge manager distributed names randomly - if he created a big cycle you lose).

A more mathematical discussion here.

How does this relate to financial markets? People make estimates based on their experience. This looks like a random trial. It's really not. It would be if you shuffled the deck after each trader. If something is not a random process and you model it as one, you make bad inferences. Your beta model assumes all the correlation in your portfolio captured in the beta, and the uncaptured variation cancels out. If there is hidden correlation in the errors, ie all your stocks are in the buggy-whip-related industries, then all bets are off.

If you ask 200 people in a classroom to guess the number of marbles in a jar and write it down, the answers will be randomly distributed and probably the mean will be a pretty good unbiased estimator. If you ask them to shout out sequentially, the distribution will cluster around the first guesses, and the estimate gets a lot worse. Once people coordinate, they can make very surprising things happen, for better or worse. And small non-obvious correlations in behavior can make a huge difference in outcomes.

100 traders, 100 boxes

By druce on June 6, 2011 12:38 PM | No Comments | No TrackBacks
A math/probability problem I think is awesome and counterintuitive, and may be instructive about financial markets:


A hedge fund manager puts 100 traders in a room and instructs them:


"On the trading desk, there are 100 boxes. Each box has one of your names. You can go onto the trading desk and open any 50 boxes you choose, to try to find your name. If every one of the 100 traders in this room finds his or her name, you will each get a $1,000,000 bonus. If anyone fails, I will crush all your $100,000 BMWs to create my modern art masterpiece. You can devise a strategy before anyone leaves the room, but once a trader has opened the boxes, they must leave the trading desk exactly as it was before you entered and cannot communicate with anyone else."


Questions: Should they take the bet? Is there a correct strategy? What is the probability all 100 will each find their name?


A hint, if you want one, by copy and pasting the lines below.


Hint: the answer to the first question is yes. What is the strategy and approximate probability of success?


Answer tomorrow.

We make our tools, and then our tools make us

By druce on June 1, 2011 8:40 AM | No Comments | No TrackBacks
Combat in Grand Theft Auto IV has been reworke...

Image via Wikipedia

You didn't change the game, the game changed you. - Niko Bellic, Grand Theft Auto IV

Everything that's already in the world when you're born is just normal; Anything that gets invented between then and before you turn thirty is incredibly exciting and creative and with any luck you can make a career out of it; Anything that gets invented after you're thirty is against the natural order of things and the beginning of the end of civilisation as we know it until it's been around for about ten years when it gradually turns out to be alright really. - Douglas Adams

So here is my rant about CDS, a topic I know very little about.

Christine Lagarde has said that a Greek restructuring that would create a credit event is 'off the table'. Now, inevitably, politicians talk out of both sides of their mouths to avoid roiling markets. But I have seen three different pundits talk about a Greek restructuring that would be done in such a way as to avoid triggering CDS default provisions. This makes no sense.

Greek 2-years are trading at 73 cents on the dollar, or a 23% yield, and CDS are trading at something over 1400 bps. Meaning you pay over 14% of the face value of a bond for insurance against it defaulting in the next year. 

A restructuring is going to mean bondholders get something worth around 75% of par, so it will not exactly be a close call whether a default occurred in economic terms. So, one of two things are possible: 1) in the event of a restructuring, the buyer of CDS protection gets paid par, or 2) sovereign CDS are not worth the electricity it takes to conjure them up. 

Clearly, anyone who is buying protection at 14% is calling Ms. Lagarde a liar. 

But it seems on the face of it that you could buy CDS and bonds and lock in a decent yield, and a possible windfall if you get paid off early, which suggests that markets think that while holders of CDS protection will get paid off with a high probability, things are not as cut and dried as they should be.

That kind of fuzziness (dare I say corruption?) seems like an awfully big strike against CDS.

It seems like if CDS are a legitimate instrument, finance ministers shouldn't be talking about how to circumvent them, and if finance ministers feel the public interest necessitates doing so, maybe the existence of CDS is not such a good idea in the first place.

Next issue: CDS change the game and make default in a situation like this a certainty (barring shenanigans). The way the game should play out is like a corporate takeover. The 'troika' of bailers-out buy up the debt in the market, or as much as they can without bidding it up. (Hey, it's not insider trading if the government does it). When you can't buy more, you go public and make an offer for the remainder, and say otherwise Greece defaults and you get nothing. After some back and forth most bondholders will tender.

But if you have bonds and CDS protection, you have no incentive to accept. The game is to force a default, and then hand over your bonds and get par back.

So the mere existence of CDS can make a technical default inevitable. Does it matter? In this case, maybe not much, except politically it always looks better to refinance the mortgage than to have your house auctioned off. However, in corporate cases, once bankruptcy protection is sought, things get more complicated and it becomes ultimately up to the judge. Things get a lot more expensive, complicated, and political.

It's a bit similar to securitization - when a mortgage gets carved up into multiple pieces and levels of subordination, it's good to spread the risk around when things are good, but it gets a lot more complicated to work out when things go bad. A solution that might be globally optimal might be seen as favoring one class over another, there's reluctance to reward 'speculators', etc.

A possible strike two against CDS.

Finally, let's talk about the opacity. We may think we have a vague idea of who has lost money on Greek bonds. But some of these players might have hedged with CDS. And we have no idea who has been in the risky business of selling CDS protection. There have been cases where CDS outstanding were large multiples of the underlying debt. Is this going to bring down a bank? Who knows? The post-crisis reforms toward central clearing would make cascading defaults less of a problem, if central clearing is adequately capitalized and policed. But derivatives make financial statements far less useful than they used to be. You never know when a big hole might turn up.

I will skip the rant about manipulation and moral hazard, since only conspiracy theorists would say that shenanigans in thinly-traded CDS were the last straw that triggered the final downfall of Bear to the benefit of short sellers and manipulators.

When I got my first job 25 years ago, I thought Glass-Steagall was a barbarous relic - clearly a bank could be very risky without being in the securities business, and the way to go was better living through math and things like VaR. If the financial crisis taught us anything, it's that systemic risk comes from inadequate equity capital, opacity, complex interlinking, and moral hazard.

So I'm leaning towards Soros's view, only generic, well documented and registered derivatives and financial structures should be traded, they should trade and clear on well-capitalized, well-policed, highly-disclosed exchanges, and by ring-fenced, well capitalized financial players and speculators (and players who are natural hedgers and need to reduce risk).

The reworking of the Wild West video game financial system hasn't really started yet, but it should start soon.
Enhanced by Zemanta
« May 2011 | Main Index | Archives

Search

About this Archive

This page is an archive of entries from June 2011 listed from newest to oldest.

May 2011 is the previous archive.

Find recent content on the main index or look in the archives to find all content.