Streetwise Professor

May 17, 2012

Morgan Agonistes

Filed under: Derivatives,Economics,Financial Crisis II,Politics,Regulation — The Professor @ 10:25 am

The Morgan trade that has roiled the bank, and arguably more important, the politics of banking, is coming into clearer focus.

What transpired appears to be a classic collection of trading and hedging SNAFUs.  There are two different issues that should be distinguished.  The first is the motive for the trade, the second is its structure.  The underlying motive makes sense at an abstract level, but the structure was FUBARed.  But once the structural issues are considered, one must seriously consider whether the motive, however attractive in the abstract, is at all feasible, especially for a behemoth like Morgan.

Morgan was apparently looking for a systemic risk hedge, something that would protect its huge exposure to corporate default risk against a Eurocrisis, a dramatic weakening of the US economy, etc. In the abstract, this is a laudable goal.  In the abstract.

Given this objective, it put on a trade in tranches of the IG.9 credit index.  These tranches are very sensitive to credit conditions.  What’s more, their valuation and risk are highly dependent on correlations.

Crucially, given the sensitivity of the tranches to credit conditions, to maintain the desired hedge Morgan had to dynamically manage the position.  It managed the risk by selling protection on the IG.9 index.  Crucially, the structure of the position required Morgan to sell protection as credit rallied.

And credit indeed rallied, due to the ECB’s aggressive actions at the end of 2011.  So Morgan had to go into the market big time, and sell more and more protection.

By doing so, it ran into all the curses that beset dynamic hedges, especially somewhat dirty ones in which what is being hedged (the tranches) and the instrument being used to hedge them (the index) have different sensitivities to parameters-correlations, in this instance-that are themselves moving randomly, and moving a lot.

The hedging instrument was relatively illiquid, meaning that it was prohibitively costly to adjust the hedge in response to every move in the relevant prices and parameters.  But given the structure of the position, failure to adjust continuously gives rise to hedging errors that work against the position. Its value was also buffeted by changes in the crucial parameters.

Moreover, Morgan’s very size worked against it.  Dynamic hedges chase the market: you are buying when prices rise and selling when prices fall.  This is how you pay for the hedge.  But when for a bank as big as Morgan, its hedging trades exacerbated the adverse price movements.  This made the hedge far more costly than a standard model, which presumes that prices move exogenously and are not affected by the trades of the hedgers, would predict.  Morgan’s price impact was revealed by the divergence (“the skew”) between the price of the index and the prices of the underlying credits.

Morgan found itself trying to navigate between the Sybil Scylla of holding an unbalanced position that subjected the bank to market risk (in movements in the index as well as to movements in the crucial parameters) and the Charybdis of massive transactions costs inherent in trading in relatively illiquid markets in immense size.  And in its attempt to navigate that divide, it apparently was bashed by both monsters.

Moreover, once it was in the position, there was-is-no obvious way out, especially given the fact that now everyone and his extended family is aware of Morgan’s predicament.

Looking at all this reminds me of the portfolio insurance debacle on Black Monday, 1987, when dynamic hedging strategies blew up spectacularly.  Morgan has been a one bank portfolio insurance wrecking ball.  A poster child for what can go wrong with dynamic hedges that make sense in a frictionless, complete markets, BSM world, but which can go horribly, horribly wrong in messy reality, where markets are incomplete and illiquid.

Indeed, all of the potential problems with dynamic hedges were/are present in spades for this particular transaction.  Most notably, dynamic hedges are major liquidity users.  Morgan was trading in relatively illiquid products.  Moreover, the risk that Morgan was hedging against was also correlated with liquidity, which is a wrong way risk from hell (which portfolio insurers also found out to their dismay in ’87).  The position was also highly sensitive to parameters-correlations that are (a) hard to estimate accurately, and (b) quite volatile, especially in the very conditions that Morgan was purportedly hedging against.  Furthermore, all of these problems were exacerbated to an extreme extent by the size of the positions and the trades.

A couple of broader points deserve comment.

First, this episode indicates that there is a source of scale diseconomies in banking.  You can be too big for your own good-and your own safety.  So what is to be done about that?  Of course the immediate response-as predicted here-is to pile on more regulations, something that the administration is pursuing furiously, in its tried-and-true let-no-crisis -go-to waste fashion.  But the market imposes its own discipline.  Morgan’s stock price has fallen substantially, its credit rating has been cut, and its financing costs have risen.  These all limit the bank’s ability to grow further, and indeed will likely lead to some contraction in its size. Moreover, the demonstration effect will lead the markets to evaluate the costs of size across financial institutions generally, leading to adjustments in the cost of capital based on updated information about the real costs large entities incur to manage risk.

Second, although hedging against systemic risk sounds superficially attractive, Morgan’s experience, and a little reflection, suggests that the ability to do so is chimerical.  Systemic risks, by definition, affect everybody in the system.  They can’t be diversified away.  Everybody wants to shed them, so they are priced in equilibrium.  You have to pay to get rid of them.  Morgan apparently underestimated the cost of paying other market participants to take on this risk from them.  In part, this was because the need to trade dynamically in order to put on the hedge meant that there wasn’t a sticker price, paid up front, to get someone to take on the risk.  Instead, the cost was paid day after day, via the market chasing, slippage, and transactions costs associated with the trading strategy.  Like portfolio insurance, it looked cheap on the white board or the computer screen, but turned out far more expensive in reality.

This is particularly true for a big, systemically important institution that has to trade in large size to implement such a strategy.  This size magnified all of the pitfalls of dynamic hedging.

In essence, Morgan put on the trade because it wanted to shrink its exposure to certain risks.  It did so by implementing a dynamic trading strategy, which turned out to be a very costly way of doing that.

So how can big banks shrink their risk exposures, if not via dynamic derivatives trading strategies? By getting smaller.  By shrinking their balance sheets.  The market price responses to Morgan’s revelations are likely to lead to that very result, not just for Morgan, but for other big banks as well.  Just as portfolio insurance implemented by dynamic trading shrunk significantly once its true costs were realized, the same thing is likely to happen here.

Attempting to regulate bank activities via Talmudic-and often nonsensical-distinctions between “hedging” and “prop” trades is a fool’s errand.  Hedges can go horribly wrong.  Moreover, quite perversely, regulatory burdens actually tend to create compliance overhead costs that tend to encourage size; the advantage that big institutions have influencing regulators accentuates this effect.  Some combination of capital requirements (with devilish details) and market discipline is a better way to control bank size.

Which means that there is a positive externality to Morgan’s struggles.  Its travails demonstrate vividly the costs of size.  Morgan apparently underestimated the cost of managing risk exposure, and its big loss not only is making it aware of the true cost, but is educating the market as well.  This will affect the cost of capital not just for Morgan, but for all banks, and in a way that will impose a cost on size and complexity.  It is unlikely that this higher private cost will reflect completely the social costs of size, but it is certainly to be welcomed.

Print Friendly, PDF & Email


  1. I’m skeptical of any IG9-only explanation, but yes, in a broader sense, the flavor of this story rings true with my hunch (broadly speaking, two words: negative convexity).

    What’s puzzling is that there is absolutely 0.00% chance that Iksil/et al are unaware of the negative convexity and pitfalls of buying the protection they did. To me this points to a probable story we haven’t fully heard (and may never), in which they constructed a fuller, longer-term hedge package whose profile (to maturity) they liked and which seemed to work under various metrics – and didn’t factor in rebalancing much if at all.

    But then that only raises the question of why on earth they would rebalance the IG9 delta so actively as to chase skew tighter 20+bps.

    Comment by Sonic Charmer — May 17, 2012 @ 11:07 am

  2. @Sonic Charmer-Welcome to the comments section. I like your blog a lot, and appreciate the links-which is how I found it in the first place.

    I agree that Iksil et al were unaware they were short convexity, and of the conceptual pitfalls thereof. I think it is possible that they underestimated the costs/risks of that exposure. EG, they could well have not anticipated the ECB intervention and its effects on their dynamic hedge, and they may have erred in their estimation of the liquidity of the market and crucially on the (a) the negative correlation between liquidity and their need for it, and (b) their price impact.

    I also agree that the big puzzle is why, if the underlying risk they were managing wasn’t long convexity, they would offset it with a trade that was (apparently) massively short convexity?

    The ProfessorComment by The Professor — May 17, 2012 @ 11:42 am

  3. Did you mean Scylla (rather than Sybil) and Charybdis?

    Comment by Lark — May 17, 2012 @ 2:57 pm

  4. Lark-yup. Speed kills. Thanks for the nitpick. Corrected.

    The ProfessorComment by The Professor — May 17, 2012 @ 3:08 pm

  5. Actually just testing to see how closely people read! (If you believe that, I have a great, great, great credit index tranche to sell you.)

    The ProfessorComment by The Professor — May 17, 2012 @ 3:18 pm

  6. If these idiots sold protection on an arb negative index (i.e. against an instrument that a positive arb could be made), they cannot be punished: they are already brain dead. did any one do an intrinsic value analysis – Index to components? Sweet Jesus Christ!

    Comment by sotos — May 17, 2012 @ 3:34 pm

  7. @sotos-the arb negative was the consequence of their actions. And I think it was an unanticipated consequence. The question is whether discounting this possibility, as they apparently did, was reasonable. It’s easy to say in retrospect no. But given the well-known problems with any dynamic trading strategy done in immense size, I would have to say that the strategy should have been recognized as insanely dangerous prospectively. Every danger inherent dynamic hedging was predictably present in this strategy.

    The ProfessorComment by The Professor — May 17, 2012 @ 3:52 pm

  8. @Prof Good point, but what I meant to say was did the arb open up as they were putting on the trade, after they put on the trade or as they adjusted it? Usually these things are reflected pretty quickly in OTC markets. Questions:

    If the arb opened as they were doing it why didn’t they try to sell the basket vs the Index?
    If the arb was open after they put on the initial position, why didn’t they arb it out themselves against the index?
    If the arb was open when they were adjusting their position why do the wrong side of the arb trade?

    From the article, if it is accurate, it sounds like in Feb the selling was open ended. This would be inexcusable – it sounds like they traded with all the grace of a drugged Hippo tap dancing. Size alone is no excuse for this.

    Comment by sotos — May 17, 2012 @ 6:04 pm

  9. […] Streetwise Professor » Morgan Agonistes. […]

    Pingback by Streetwise Professor » The JP Whale Trades, In Clearing? | The OTC Space — May 18, 2012 @ 5:39 am

  10. Great post, Prof. It is refreshing to just read reasoned analysis.

    Two things I would comment on:

    1) @ sotos, at the risk of sounding pedantic, these positions are hedged against the greeks, which are point measures, after the position is initiated. JP thought they’d identified a systemic-risk hedge (I know that’s a generous assumption) then they had to dynamically manage it. The hedges’ greeks become correlated to liquidity. Another way of saying this is the bid-offers in the market(s) in which JP found itself hedging are a function of volatility. Hitting a bid for size, or lifting an offer for size, in an illiquid market necessarily causes bid-asks to be recalibrated (i.e., widened), so you get new position greeks just from traversing the very wide bid-ask: The whole topology of the position (its vol surface and the correlation matrices in the position’s VaR) changes, as do the topologies of the other folks in the market. This is becomes a wicked feedback loop as the bid-asks get wider, and you, and everyone else who’s going the same way, has to chase bids and offers for smaller size. This wrecks havoc on the correlations of closely related markets, as well, so bid-asks in correlated markets start widening, …, and so on.

    2) Prof., I would assume the CIO was dynamically hedging not only this position, but the bank’s risk as well. This is a big deal because — at $100 billion notional — the size of this position became a center of gravity for the bank’s overall risk position. And it would have exerted a large effect on the bank’s VaR. It’s effect of a position of this size on the big correlation matrix used by the bank to hedge its overall risk likely would have a distorting effect not only on the point measures of risk for this position but for the estimated risk of the entire bank. This could, and likely did, become the feedback loop from hell.

    What’s even more interesting to consider is the following related issues: Up to a certain point, other market participants were willing to provide liquidity to JP, so JP was able to sell off bits of its risk to counterparties. The counterparties obviously do not end up with JP’s overall risk, just the point estimate of the change in the risk whenever they provided “hedges” to JP. So here’s the question: At what point do the counterparties start orbiting around the same center of gravity as JP? At what point are they forced to hedge the derivative-JP exposure (no pun intended) that they’ve taken on by transacting with JP? What are the implications for that market — and its closely correlated markets — in terms of liquidity (willingness to quote bids and offers for size)? What is the feedback into the volatility and correlation matrices? Lastly, at what point does the underlying simply become a proxy of the banks’ risk positions and not the presumed economic variable it is meant to follow (systemic risk, in this case)?

    This has to be solved subject to a binding constraint: The VaR model used to estimate the risk of the desk and the bank earlier in the year (and probably all of the previous year or two … or three … before that) was found to be flawed. What part of the unwind’s cost is due to mis-estimates of risk from the faulty VaR model, and which part of the unwind’s cost is due to the feedback loop JP now finds itself in in terms of shifting the center of gravity of the entire market? Oh, one more complicating factor: What if the VaR model JP replaced actually was a correct representation of the risk, and the one they’re using now is an incorrect representation? How would they know?

    At the end of the day, this is just arithmetic — you sum the negatives (positives) and look for a positive (negative) to offset the exposure. But it quickly becomes a lot more than that, huh?

    I can hear Kurt Godel giggling from the other side right now ( )

    Comment by markets.aurelius — May 19, 2012 @ 6:38 am

  11. Excellent comment, markets. A lot to chew on. You’ve identified several other ways in which the price taker and perhaps more crucially parameter taking assumption underlying most risk modeling can lead to disastrous results if you are not a price/parameter taker but a price/parameter maker due to size. This assumption is extraordinarily dangerous.

    I hadn’t thought about the knock-on effects at the rest of JP, but I think you’re absolutely right on that. And then you think about how in response to that “signal” (noise, actually, misinterpreted as signal given the relatively mechanical way that those responsible for trading/risk management respond to VaR changes), others within the bank trade. Think of all the non-linearities. And given JP’s immense size, how those non-linearities impact others in the market, which feeds back on JP . . .

    Which means that my original analysis greatly understated the diseconomy of scale here.

    Your “what’s even more interesting” paragraph also resonates with my point about whether it is possible to hedge a systemic risk. Your “generous assumption” remark is quite accurate. Upon reflection, I think that this the logical error that gave birth to this monstrosity. Indeed, basic finance suggests the rough conclusion that “systemic risk hedge” is an oxymoron.

    The ProfessorComment by The Professor — May 19, 2012 @ 7:37 am

  12. @Mark – be a pedantic as you like – a very cogent point as to how they got into this position and I enjoyed being corrected or expanded upon – either way. the issue goes back to SWP’s comment of not living in a BSM world but acting like we do, the Greeks being a creature/ pre and post cursor of (largely) BSM My only point is that a basic analysis would show that if an ARB opened up, there should have been a way to trade against the position – e.g. adjust more efficiently by buying the cheap selling the expensive side of an index and its’ components. If they couldn’t do this, that brings up a whole other issue:

    Somewhere it was defined that a market where an arbitrage cannot work (that is a real arbitrage, not risk or other spec positions calling themselves arbitrages, e.g. MFers, etc.) occurs when the market is under stress (nearly said systemic, Prof, will try to control myself in the future). In other words it is a time when liquidity is not available to leg into one or the other legs. Here is seems we have “induced” stress: in other words the blind trading of the index created such a demand or(or in this case supply) for a leg of the trade the the market didn’t have the liquidity to close the arb. If you agree with this we have just defined when a monster trade(or trader) becomes literally a monster: when the Index to basket values cannot be brought into line quickly. Having created this, any trader should recognize what they are doing (digging themselves deeper) and stop.

    In sum this may be a case where the “sophisticated” but wrong var and other models failed because of the traders/PM’s inability to factor in such minor issues as “to whom do we sell or buy from”(see pink sheet joke from earlier post). All in all a basic mistake or act of hubris that any experienced floor trader from 30 years ago would have spotted and avoided.

    Comment by sotos — May 19, 2012 @ 11:00 am

  13. @Prof – if systemic risk means the risk in an industry, it seems the only way to hedge it would be to get out of the industry you are in. An oxymoron indeed.

    Comment by sotos — May 19, 2012 @ 11:01 am

  14. @sotos-By systemic risk, I mean a risk in the entire financial system. Which suggests and oxymoron squared (or cubed) 🙂

    The ProfessorComment by The Professor — May 19, 2012 @ 2:11 pm

  15. @sotos & @markets-re arbs.

    First, I agree it is puzzling that once the skew opened up that JP continued to sell protection on the index rather than switching to selling protection on the single names. Perhaps it figured that the single names were so illiquid that even small trades would have crushed the skew. But still, why not allocate protection sales between the index and the single names so at the margin they were priced the same?

    Second, it is often the case that apparent arbitrages are symptomatic of acute stresses and market frictions, e.g., the apparent profitability of basis trades in ’08 (that crushed an earlier whale-Weinstein-who ironically contributed to the demise of the new whale).

    Third, this presents an opportunity to tell one of my favorite stories. The FCM I worked for in ’87 (as a small child, natch) routinely did small index arb trades. On 19 October, 1987 it looked like there was a huge arb opportunity: the futures were insanely cheap relative to cash. Of course, this was a consequence of the huge capacity bottleneck at NYSE, where the daisy wheel printers at the specialists posts were unable to keep up with the flow of orders, meaning that trades that were executed had been entered long before at prices that give new meaning to the word “stale.”

    Anyways, our SOP before putting on a trade was to call our desk on the CME floor to get the current bids and offers in the pit. So we call down, and our guy on the desk says: “It’s being quoted at X over here and X+50 over there and X-50 over there. [I don’t recall the exact numbers, so I’m MSU.] I have no idea what the fucking price is.”

    Obviously, we didn’t do the trade.

    The points being that in stressed market conditions an arb that appears to good to be true is, and that basis trades in stressed markets are themselves stress-inducing.

    The ProfessorComment by The Professor — May 19, 2012 @ 3:01 pm

  16. Hey, Prof. I don’t know if you’re still checking your old posts, but, apropos of this one, you might want to give a gander to Prof. Hu’s latest:

    Hu is the Allan Shivers Chair in the Law of Banking and Finance at University of Texas, and the former head of Risk Fin at the SEC:

    The article, at least as I read it, is an acknowledgement of what I’ve been reading and posting on your site for a while now. Always good to be ahead of the curve.

    Comment by markets.aurelius — July 9, 2012 @ 1:30 pm

RSS feed for comments on this post. TrackBack URI

Leave a comment

Powered by WordPress