I really don't want to write a post debunking the hysteria thrust upon the quivering masses (who reply: "OH NO! HOW IS THIS LEGAL?!!") by
Andrew Bogan in this FT Alphaville piece (and now re-hyped by
Clusterstock). Since Steve Waldman addressed a number of Bogan's errors in the comments section of Bogan's own post (
if you're interested in this topic, you must read swaldman's comments before continuing!), I figured now would be a good time to explain a little bit about how ETFs work to the masses, who remain terminally confused.
One of the great things about ETFs is that they can be created and redeemed. This means that "authorized participants" (read: big broker dealers) can take a basket containing the underlying stocks of the ETF, in specific weights, deliver them to the ETF trust and receive the ETF shares - that's called creating. They can also do the opposite: deliver the ETF itself to the trust, and receive the underlying basket of individual stocks - that's called redeeming.
A great source of confusion that many folks have regarding ETFs is the concept that when you buy an ETF, like, say SPY - the S&P 500 ETF - the trust itself does not take your money and go out and buy the underlying S&P 500 stocks with it. Similarly, when you sell SPY, the trust does not go out and sell stocks to raise money to give you. The trust has nothing to do with it. When you buy SPY (just like when you buy IBM) , you buy it from another participant in the market - not from the trust (not from the company itself). Contrast this to mutual funds: when you buy an S&P 500 mutual fund, that mutual fund does take your money and go out and buy their portfolio of stocks with it, and vice versa when you sell.
So, back to our SPY ETF - the reason this works is precisely because of the creation/redemption mechanism. If "the market" is lacking SPY sellers, and the price of SPY rises so that it is in excess of it's NAV (net asset value) there are arbitrageurs standing by ready to short you shares of SPY while they simultaneously buy the underlying basket of SPY components as a hedge. Since they are selling SPY "rich" to it's fair value, they will make a profit when they eventually collapse their position. How do they collapse it? Well, they take their long basket of SPY component stocks and "create" SPY by delivering the individual stocks to the trust, receive newly created SPY, and use that to cover their SPY short position. Voila - they're now flat, and the SPY's assets have increased, as have the shares outstanding.
Of course, if everyone wants to sell SPY, the opposite happens - the arbitrageurs, if SPY is trading "cheap" to it's NAV, will buy SPY while simultaneously shorting a basket of the underlying components. Then, they'll take their SPY shares, deliver them to the trust, redeeming them for the underlying components which they use to close out their short underlying stock positions. In this case, the assets held by the trust decrease, as do the shares outstanding (after all, the arbs have taken SPY shares out of circulation, and given them back to the trust, effectively "retiring" the shares temporarily.)
The next level of this discussion gets somewhat complicated pretty quickly. Bogan is worried about "naked short sellers" but he demonstrates a thorough lack of understanding of what a naked short seller is. Bogan writes:
"Take the SPDR S&P Retail ETF (NYSE: XRT) as an example. The number of shares short was nearly 95 million at the end of June, while the shares outstanding of the ETF were just 17 million. The ETF was over 500% net short! Or to look at it from another perspective, the ETF’s operator, State Street Global Advisors, believed that there were 17 million shares of the SPDR S&P Retail ETF in existence and owned shares in the S&P Retail Index portfolio to underlie those 17 million ETF shares. But, in the marketplace there were another 95 million shares of the ETF owned by investors who had purchased them (unknowingly) from short sellers. 78 million of those ETF shares were naked short–the short seller had promised their prime broker to create those non-existent shares if necessary to cover their short in the future. In both cases the share buyer, however, is completely unaware his ETF shares were purchased from a short-seller and no doubt assumes the underlying assets in the index are being held by the ETF operator on his behalf, but no such underlying stock is actually held by anyone. Clearly this creates a serious counterparty risk and quite possibly the potential for a run on an ETF—where the assets held by the fund operator could become insufficient to meet redemptions."
First, it's perfectly possible to have 95mm shares short with only 17mm shares of XRT outstanding, although this is a concept that many people hate and are confused about. How? It's just like fractional reserve banking. First, remember that in order to short shares of stock, you first need to
borrow those shares from somebody (so that you can deliver them to the buyer on the other side of your short sale). If you don't borrow the shares, that's called naked shorting, which is different from what Bogan describes. These ownership chains get hard to follow pretty quickly, but let's pretend that I, Kid Dynamite, bought shares of XRT in the initial public offering. I own the shares. I take my shares and lend them to Steve, who goes and shorts them. Now, I no longer have shares to lend - I've already lent them out Steve has borrowed them and shorted them, and the buyer of that stock is Dave. Dave then takes the stock which Steve delivers to him, and lends it out again, to Mike, etc... Thus, multiple people end up short the "same" share of stock, everyone has delivered the stock they promised to, no one failed to deliver, and no one is naked short. (if you haven't read Swaldman's comment explanations, now would be a good time to do so.
first,
second). Like fractional reserve banking, there's collateral in each step of this chain - if I want my shares back and Steve can't get them for me, I'll take his collateral instead. SWaldman describes it thusly:
"Lending of securities creates synthetic supply of shares, just as lending by [banks] creates a synthetic supply of money. With shares and with money, the lending arrangements are sufficiently well guaranteed that most holders consider the synthetic to be a perfect substitute for the original. Most people don't fret over the distinction between bank deposits and cash-in-hand, just as most stock traders don't fret over the distinction between owning a share and owning a promise to deliver a share on demand by their broker."
Now, Bogan fears a "run" on the ETF. That's not going to happen (at least not for the reasons he fears), for a few reasons. Let's start with the more complicated one, using Bogan's own numbers: as Bogan notes, now 95mm people think they "own" the XRT. Well, we know that only 17mm of them actually have possession of the XRT, since the other 78mm have lent their shares out to others. You can't redeem something you don't have (remember, you've lent it out), so it's not possible for 95mm shares to be redeemed. If some of the people who have lent their XRT out (maybe they even lent it out unknowingly) decide they want their shares back, it's not the end of the world! The short sellers of XRT will have a few choices to return the XRT that they "owe." They might go out in the marketplace and simply cover their short positions, or they might buy the underlying basket of XRT stocks, deliver them in to the trust, and "create" new XRT shares to deliver to the person they borrowed them from. As Swaldman puts it, "The missing supply of ETFs would create itself."
Now that we've explained the concept of possession of stock, let's go to Bogan's worry of how the implosion might happen: He writes:
"Redemptions occur when more owners wish to sell out of their holding in the ETF than there are new buyers for the existing shares, so unwanted blocks of 50,000 ETF shares each are redeemed through the authorized participants with the ETF operator for cash, or more typically for in-kind shares in the ETF’s underlying index’s stocks."
I hope that Bogan is oversimplifying (and I think he is), because I explained above that owners of XRT wishing to sell their shares do not directly cause redemptions. I have a feeling that Bogan knows this, and that he's basically explaining what will happen after the arb situation I described above takes place: when there are insufficient buyers of XRT (ie, too many natural sellers), arbs will step in as buyers, while shorting the underlying basket of stocks, and then redeem their XRT to the trust. Now, as long as the sellers deliver their shares, the arbs (the XRT buyers) can take the XRT and redeem them with the trust. If sellers don't deliver the shares they owe, we are back to the scenario a few paragraphs above - they will have to make good by buying back the shares in the market, or by creating more XRT by buying the underlying basket and delivering it to the trust.
Bogan worries about the rapid contraction in XRT shares outstanding in July/August:
"The SDPR S&P Retail ETF was one of the fastest contracting ETFs in July due to redemptions and as of July 31, it had just 7 million shares outstanding. However, the short interest was little changed—still over 80 million shares short. Suddenly, 11 times the number of shares outstanding was short, which is even more worrisome than 5 times back in June. By late August, the shares outstanding in XRT had dipped briefly below 5 million shares with 80 million shares still short (16 times the shares outstanding). Mercifully, net buying interest has rebounded somewhat for the SDPR S&P Retail ETF with the improving outlook for retailers and shares outstanding in XRT had rebounded to 12 million by mid-September. But if the rate of contraction last month had continued, the ETF was just days away from running out of underlying shares altogether."
No - the ETF was probably not days away from running out of shares. This phenomenon is self-healing. This is the important part, but it's not as complicated as it sounds. As we've established, shares can only be redeemed when they are actually delivered into the trust. If everyone wants to redeem their shares, then the shorts need to cover their positions - we have also already discussed this above. The shorts can either borrow the XRT from someone else (impossible if everyone wants to redeem), cover in the marketplace (again, if everyone wanted to redeem, that would also be impossible, but we can pretend that I overpay for XRT to the extent that one of the redeemers is happy to not redeem and instead sell it to me for a premium to NAV), or buy the underlying basket and deliver it to the trust and create more shares (and more underlying assets for the trust to distribute).
Thus, the most likely result if everyone wants to redeem XRT when there is 500% short interest is actually a massive short squeeze, or a rally in the instrument, as longs scramble to re-take delivery of shares that they have lent out. But what happens if the actual physical holders of XRT do all want to redeem, and the "self-healing" I described doesn't happen in time? Is the trust then "days away from running out of assets altogether"? Read on.
Bogan's final few paragraphs are some confused hysteria asking who is left holding the bag if all the holders of the actual outstanding shares manage to redeem their shares.
"Who gets left holding the bag? Is it the retail account holders who own defunct shares in a closed ETF? The prime brokers that were counterparties to all those short sellers? The hedge funds that sold non-existent shares in an ETF assuming they could always be created another day? The ETF operator? Or the Federal Reserve?"
First of all, it is indeed entirely possible to imagine a scenario where all of the actual holders want to redeem their shares, which results in the trust "running out" of assets. In our example, 17mm people could redeem shares and receive the assets of the trust, and that would leave 78mm longs and 78mm shorts outstanding. As
commenter "Crookery" notes, the results are not apocalyptic. The trusts have a mechanism built in to liquidate when their assets get below a certain point, and the outstanding shorts simply owe the outstanding "synthetic" longs the cash value of the trust. There is no "bag" to be left holding - short sellers owe long holders.
Furthermore, I think that if there is a greater supply of shorts outstanding, it actually makes it less likely that this happens, and more likely that the "self healing" scenario occurs, since there are more shares to get "bought in" and cause a short squeeze. Said differently, the trust runs out of assets when the actual physical XRT holders all redeem their shares - not when the "synthetic" XRT holders want to redeem their shares. "Synthetic" holders wanting to redeem causes the self healing short squeeze phenomenon, and there are more "synthetic" holders if there is more short interest.
Now, let's answer Bogan's headline question: "Can an ETF collapse?" Sure - of course it can, but not for the reasons Bogan fears, and not with the subsequent effects he ponders either.
At this point, I feel like we've gotten into some confusing topics, so I'll close with another Swaldman comment:
"Shareholders end up with synthetic long positions by their own consent, when they agree to lend shares from their account to potential short-sellers. It is true that more people consent to this than understand they are consenting to it: retail investors with margin (as opposed to cash) accounts may fail to read the boilerplate that gives their broker the right to lend their shares. But those retail investors' credit exposure is to their broker, not to the speculative short-seller that eventually borrows and resells her stock. In general, then, retail investors in margin accounts may have credit exposure to their broker by virtue of short-selling that they do not understand. And if brokers are incautious in their securities borrowing and lending, they could blow-up and force losses on clients holding margin accounts. In theory, that is a real concern, but it has nothing to do with ETFs. We should be concerned that brokers who lend the shares of customers are regulated to prevent credit losses that impact their solvency, whether those losses result from share-lending or any other practice. This is the LTCM issue, not something specific to short-selling or ETFs."
-KD
