Caso 2 Risk Management

152 JOURNAL OF APPLIED FINANCE SPRING/SUMMER 2008
A Case Study on Risk Management:

Lessons from the Collapse of
Amaranth Advisors L.L.C.
Ludwig Chincarini
I. Introduction
The speculative activities of hedge funds are a hot topic
among market agents and authorities. In September In September, 2006, a large-sized hedge fund named
2006, the activities of Amaranth Advisors, a large-sized Amaranth Advisors LLC lost $4.942 billion in natural gas
Connecticut hedge fund sent menacing ripples through futures trading and was forced to close their hedge fund.1
the natural gas market. By September 21, 2006, Although Amaranth Advisors was not exclusively an energy
Amaranth had lost roughly $4.942 billion over a 3- trading fund, the energy portion of their portfolio had slowly
week period or one half of its assets primarily due to grown to represent 80% of the performance attribution of the
its activities in natural gas futures and options in fund (Source: Senate Subcommittee Exhibit #12). Their
September. On September 14 alone, the fund lost $681 collapse was not entirely unforeseeable or unavoidable.
million from its natural gas exposures. Shortly Amaranth had amassed very large positions on both the New
thereafter, Amaranth funds were being liquidated. This York Mercantile Exchange (NYMEX) and the Intercontinental
paper uses data obtained by the Senate Subcommittee Exchange (ICE) in natural gas futures, swaps, and options.
on Investigations through their subpoena of Amaranth, The trades consisted mainly of buying and selling natural gas
the New York Mercantile Exchange (NYMEX), the futures contracts with a variety of maturity dates. Their trades
Intercontinental Exchange (ICE), and other sources to
analyze exactly what caused this spectacular hedge Ludwig Chincarini, CFA, Ph.D., is an Assistant Professor of Economics
fund failure. The paper also analyzes Amaranth’s at Pomona College in Claremont, CA 91711.
trading activities within a standard risk management
I would like to especially thank Dan Berkovitz and the Senate
framework to understand to what degree reasonable Subcommittee of Investigations headed by Senator Carl Levin for helpful
measures of risk measurement could have captured the discussions and access to important data and documents. I thank Neer
potential for the dramatic declines that occurred in Asherie, Ed Fraim, John Frasr, Stu Johnston, Daehwan Kim, Michael
Kuehlwein, Milton Liu, Fernando Lozano, Gary Smith, Hilary Till, and
September. Even by very liberal measures, Amaranth the referree, Dan Rogers, for helpful comments. I thank Jim Riley for
was engaging in highly risky trades which (in addition sparking my interest in this topic. I thank Joann Arena and Scott Byrne of
to high levels of market risk) involved significant NYMEX for supplying data.
exposure to liquidity risk – a risk factor that is 1
These losses are computed as the actual change in net asset value of the
notoriously difficult to manage. Amaranth funds, including the Amaranth LLC fund, Amaranth Partners fund,
and Amaranth Global Equities Master fund from Exhibit #12 of the Senate
Subcommittee documents. The value of these funds was $10,228,192,000
on August 31, 2008 and $5,286,050,000 on September 21, 2006. These
total net asset values do not include the Amaranth Securities LLC, which
had a smaller amount of around $30-50 Million since the data was not
available.
CHINCARINI A CASE STUDY ON RISK MANAGEMENT 153
were very risky from both a market risk perspective and a 2000 as a multi-strategy hedge fund, but had by 2005-2006
liquidity perspective. generated over 80% of their profits from energy trading
Since the collapse of Amaranth, several authors have (Source: Senate Subcommittee Exhibit #12). This section
attempted to understand what positions and risk levels provides a very brief summary of Amaranth. (For additional
Amaranth was engaged in to cause such a dramatic collapse information, please see a more detailed version of this section
(Chincarini, (2006) and Till, (2006). Chincarini (2006) used on the Journal of Applied Finance, JAF, website:
the information from newspapers, CEO statements, and actual www.fma.org/jaf.htm).
natural gas futures data to quantify the nature of the most likely
trades that were made at Amaranth. That paper hypothesized 1. The Management
that Amaranth had engaged in a short summer, long winter
The management consisted of several seasoned
natural gas trade primarily using natural gas futures. Based
professionals. The most relevant to the natural gas futures
on these backward-engineered positions, the paper examined
disaster was Mr. Brian Hunter. Hunter joined Amaranth in
both the market and liquidity risk of Amaranth’s positions prior
2004.3 He was hired by Mr. Maounis and Mr. Arora, a former
to its collapse.
Enron trader who had established Amaranth’s energy and
On June 25, 2007 the Committee of Homeland Security
commodities trading desk. Prior to this, he had worked at
and Government Affairs released a document containing a
TransCanada Corporation, a Calgary pipeline company, where
detailed investigation of the Amaranth scandal entitled
he began getting a name for himself in energy trading. While
“Excessive Speculation in the Natural Gas Markets.” The U.S.
there, he was able to find mispricing in energy options, which
Senate Permanent Subcommittee on Investigations used its
helped the firm make profits. After this, Hunter moved to Wall
subpoena power to analyze the trading records at the NYMEX,
Street to work for Deutsche Bank on the energy desk. While
the ICE, as well as the trades of Amaranth and other traders.
there, his positions in natural gas futures caused large
It also conducted numerous interviews with natural gas market
fluctuations in profit and loss.
participants, including natural gas traders, producers,
In the summer of 2005, Hunter threatened to leave
suppliers, and hedge fund managers, as well as exchange
Amaranth, partly because he disliked his compensation
officials, regulators, and energy market experts.
structure and did not wish to report to Arora. Maounis reacted
In this paper, we make extensive use of the Amaranth trading
by allowing Hunter to trade a book separate from Arora. Also,
positions derived from the actual Amaranth trading data. This
his share of the operating profits eventually were increased
data was obtained under subpoena by the Senate
from 7.5% to 15%. Hunter made a name for himself on Wall
Subcommittee. We also discuss the risks associated with the
Street when he helped Amaranth make $1 billion in profits in
trades Amaranth made and what risk managers should do to
2005. Due to his trading success in 2005, Hunter was rumored
avoid these risks in the future. The rest of the paper is as
to have been compensated between $75 million and $100
follows: Section II discusses the background of the firm
million. Late in 2005, Hunter was also allowed to return to
Amaranth Advisors L L C; Section III discusses the natural
his hometown of Calgary and trade from there. Eventually,
gas futures market and details the basics of typical spread
his four other natural gas traders migrated from Greenwich to
trades to help the reader appreciate the more complicated
Calgary.4
Amaranth trading strategies; Section IV discusses Amaranth’s
actual trading positions on August 31, 2006 and in other 2. The Strategies and Fund Structure
periods; Section V analyzes the market and liquidity risks
inherent in Amaranth’s natural gas positions; Section VI Amaranth began as a multi-strategy hedge fund, but by 2006
discusses lessons for regulators and risk managers, and Section had become dominated by its energy portfolio. The principal
VII provides a conclusion. fund, with $8.394 billion of capital at the end of August 2006,
was the Amaranth L.L.C. fund. The multi-strategy portfolio
II. Background consisted of trades in the following areas: Energy Arbitrage
and Other Commodities, Convertible Bond Arbitrage, Merger
A. Amaranth Advisors, L.L.C. Arbitrage, Credit Arbitrage, Volatility Arbitrage, Long/Short
Equity, and Statistical Arbitrage. Amaranth’s exposure to these
Amaranth Advisors L.L.C. was a hedge fund operating in
Greenwich, Connecticut.2 The hedge fund was launched in
3
Most of this discussion is based upon an article in the Wall Street Journal
2
Many thanks to Dan Berkovitz for providing the information upon which entitled “How Giant Bets on Natural Gas Sank Brash Trader.” and FERC
much of this section is based. This section draws heavily from Exhibit #12 Docket No. IN07-26-000.
of the Senate Subcommittee Investigations. In addition to their Greenwich
office, Amaranth had been working on expanding their operations and had 4
These other natural gas traders on his team were Mr. Matthew Donohoe,
offices in London, Singapore, Houston and Toronto. Mr. Matthew Calhoun, Mr. Shane Lee, and Mr. Brad Basarowich.
various strategies changed dramatically over the years prior and rho), leverage reports, concentrations, premium at risk,
to September 2006. For example, at Amaranth’s inception, and industry exposures. The daily risk report also contained
60% was devoted to convertible arbitrage, whereas by the following:
September 2006 only 2% was devoted to this strategy. Over 1. Daily value-at-risk (VaR) and Stress reports. The VaR
86% of their performance in 2006 was due to energy and contained various confidence levels, including one standard
commodity related trades. In addition to this, Amaranth had deviation (SD) at 68% and 4 SD at 99.99% over a 20 day
no stop limits and no concentration period. The stress reports included
limits, which allowed the fund to scenarios of increasing credit
concentrate more towards energy
A combination of liquidity and spreads by 50%, contracting
by the end of August 2006. There funding risk ultimately caused volatility by 30% over one month
were no leverage restrictions within and 15% for three months, 7% for
the firm. Style drift was evident Amaranth’s collapse. six months, and 3% for twelve
with this multi-strategy fund.5 months, interest rate changes of 1.1
Amaranth’s capital came from a variety of investors: About times the current yield curve. Each strategy was stressed
60% came from fund-of-funds, about 7% from insurance separately, although they intended to build a more general
companies, 6% from retirement and benefit programs, 6% stress test that would consolidate all positions.
from high net worth individuals, 5% from financial institutions, 2. All long and short positions were broken down. In
2% from endowments, and 3% was insider capital. particular, the risk report listed the top 5 and top 10 long and
Minimum investments in Amaranth were $5 million. The short positions.
management fee was 1.5% and the incentive fee was 20%. A 3. A liquidity report that contained positions and their
high water mark was also employed.6 respective volumes for each strategy was used to constrain
the size of each strategy.
3. Risk Management and Liquidity Management The risk managers also calculated expected losses for the
individual positions. The firm had no formal stop-losses or
The Chief Risk Officer of Amaranth had a goal of building
concentration limits. Amaranth took several steps to ensure
a robust risk management system. Amaranth was unusual in
adequate liquidity for their positions. These steps are listed
terms of risk management in that it had a risk manager for
on the more detailed version of this section on the FMA
each trading book that would sit with the risk takers on the
website.
trading desk. This was believed to be more effective at
understanding and managing risk.7 Most of these risk officers
B. Events in September
had advanced degrees.
The risk group produced daily position and profit and loss The price movements of natural gas futures in September
(P&L) information, greek sensitivites (i.e. delta, gamma, vega, 2006 were quite different than in past years. Figure 1 shows a
timeline of the events in September and leading up to
September. Historically, a spread trade strategy in natural gas
5
Style drift refers to a change in a hedge fund’s strategy over time which futures had done quite well. Figure 2 shows the average returns
may or may not reflect a formal change in policy, hence the “drift”. An
example would be a Large-Cap hedge fund manager that suddenly has huge of different maturity futures contracts in the month of
small-cap exposure. Most of the time style drift happens inadvertently, but September from 1990 through 2005. The x-axis plots the
in Amaranth’s case, they were clearly increasing energy exposure. contract months forward. Thus, in this particular graph, “1”
6
A high water mark is a common feature of most hedge funds. It is a level represents the returns for the nearest October futures during
of the fund’s net asset value (NAV) at which incentive fees begin to accrue. September, “2” represents the returns for the nearest November
Typically, the high water mark is the highest NAV received by the client contract in September, and so on. One can see that generally,
over their investment period. The purpose of the high water mark is to
prevent a double counting of incentive fees. For example, if the fund went winter month returns are higher than non-winter month returns
from 100 to 200 NAV, the hedge fund would obtain a percentage of that and that natural gas prices have tended to rise on average in
appreciation as an incentive fee. However, if the fund dropped to 150 the September for the first 36 months out. Some of the near
following year, they would not receive an incentive fee for bringing it from
150 to 200. Their incentive fees would only begin again for gains above contracts had returns as high as 5.73% on average in
200. September.
7
One might ask whether this system is indeed optimal. It could perhaps
In September 2006, the natural gas futures market behaved
cause risk managers to become more integrated in the trading style and not entirely differently than it had historically. Figure 3 shows the
be as objective in assessing risk. Regardless of one’s beliefs in such a system, behavior of natural gas futures returns in September 2006.
Amaranth actually strayed from their system in the case of Brian Hunter.
When Brian Hunter and his traders moved their trading operations to Calgary,
One can see, from this figure, the dramatic negative returns in
Canada, there was no risk management team on the premises to monitor September, which were as low as -27% for front-month
their actions.
Figure 1. Timeline of the Amaranth Collapse
CHINCARINI
August return = 6.98% JP Morgan meets

YTD return = 31.57% Senate
June return: = 7.07% Amaranth. Merrill Senator Carl
Leverage = 5.23 Initial margin releases
YTD return: = 23.65% to buy 1/4 of Levin proposes
Initial margin exceeds $3
A CASE STUDY ON RISK MANAGEMENT
NAV = $10.228 billion report on

Leverage = 4.01 energy fund. bill to regulate
exceeds $2 Initial margin > $2.5 billion billion. NG Amaranth's
NAV = $10.71 billion Goldman Sachs electronic energy
billion monthly speculative
likely to buy rest. trading
July return = -0.53% losses = activity
YTD return = 22.99% NYMEX $944 $697 million
JP Morgan &
Leverage = 5.37 million
NYMEX and CFTC Citadel agree to FERC issues
NAV = $9.61 billion margin call
Amaranth buys call Amaranth buy energy statement to
56% of capital Mother Rock's about $4 billion portfolio for $2.1 extract penalties
Amaranth
on energy; $3 NG position to loss rumors. NG billion fee. from Amaranth
exceeds
$400 million in billion in cash neutralize their monthly losses = NG monthly losses and ex-traders
ICE intraday
redemptions for liquidity ow n exposure $2.287 billion = $4.07 billion
limits
needs
30 17 31 15 18 29 30 31 . 5 . 8 15 17 21 9 26 17
ly ly t t t t t pt pt . . & ly ly .
ne Ju Ju us us us us us e e ept ept / Ju Ju ept
Ju g g g g g S S S S 20 S
Au Au Au Au Au . 25
ept ne
S Ju
2006 2007
155
Figure 2. Historical Average Returns of Natural Gas by Contract in September (1990-2005).
Note: Since these returns are for historical contracts, the numbers represent the average return for the 1st contract out, 2nd contract out,
and so on. Thus, ‘1’ represents the nearest October contract, while ‘2’ represents the nearest November contract, and so on up to 73
months forward. In some of the earlier years, contracts did not exist 73 months forward, in this case they were not included in the
averages.
6
5
4
3
−5 −4 −3 −2 −1 0 1 2
Returns (%)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
Figure 3. Natural Gas Futures Returns by Contract from August 31, 2006 - September 21, 2006
5 10
−35 −30 −25 −20 −15 −10 −5 0
Returns (%)
OCT.06
NOV.06
DEC.06
JAN.07
FEB.07
MAR.07
APR.07
MAY.07
JUN.07
JUL.07
AUG.07
SEP.07
OCT.07
NOV.07
DEC.07
JAN.08
FEB.08
MAR.08
APR.08
MAY.08
JUN.08
JUL.08
AUG.08
SEP.08
OCT.08
NOV.08
DEC.08
JAN.09
FEB.09
MAR.09
APR.09
MAY.09
JUN.09
JUL.09
AUG.09
SEP.09
OCT.09
NOV.09
DEC.09
JAN.10
FEB.10
MAR.10
APR.10
MAY.10
JUN.10
JUL.10
AUG.10
SEP.10
OCT.10
NOV.10
DEC.10
JAN.11
FEB.11
MAR.11
APR.11
MAY.11
JUN.11
JUL.11
AUG.11
SEP.11
OCT.11
NOV.11
DEC.11
Contract Month
contracts.8 One can also see that the negative returns were Tell me if I am wrong, but we have 3 choices here.
less for non-winter months. That is, although returns were (1) shut down and start energy fund, lose 0.3 to 1.0 getting
severely negative for most natural gas futures contracts, they out, and have great future potential. However, if we lose that,
were worse for winter months through the maturity spectrum. who is going to want in on the energy fund? If h/j drops to
For example, for the first year out the contract months 2 1.50 or worse, the deferred positions are all going to get
through 6 did poorly, representing the contracts for November obliterated too.10
2006 through March 2007, while in months 7 through 13 the (2) jump back in and help this market out. Risk losing some
negative returns are less severe for the months April 2007 investors due to risk profile, but manage along until we get
through October 2007. This pattern is seen for contracts in the proper catalyst to exit positions. Start energy fund when
future years as well. This pattern would not bode well for a we can later. Without the market’s ability to absorb some xh
strategy that is long winter and short non-winter months. or even some back length right now this market in a world of
Figure 4 shows the profit and loss (P/L) of Amaranth’s trouble.11 2 days ago things were fine, but it feels like it just
natural gas futures equivalent positions on a daily basis in tipped overboard on risk. There is comfort selling spreads,
September 2006. Figure 4A shows the daily P/L , while Figure and comfort selling price right now. If you were a cash trader
4B shows the cumulative P/L starting at zero on August 31, caught long hub gas right now, would you buy or sell January?
2006. The daily P/L is computed using Amaranth’s actual daily (3) Sit and wait. Let market take its course, find natural
positions from August 31, 2006 through September 15, 2006. fixed price demand.
After September 15, 2006, no data on their positions was There is not catalyst right now. That’s the problem. You
available and the daily P/L was computed assuming Amaranth exit this size without one (without exiting every positions in
maintained their September 15, 2006 natural gas positions. your book), and we got a big problem. Things were fine when
As shown in Figure 4B, from August 31, 2006 to September we were holding the risk for the market, b/c we could handle
7, 2006, Amaranth had lost about $696.9 million on their it. That risk in 30 other hands is a much more dangerous
natural gas positions. This soon deteriorated very quickly. By proposition.
the close of business on September 20, they had lost about Calhoun think #2
$4.071 billion on their natural gas futures positions. Margin Rummy thinks #3
calls on these losses eventually led Amaranth to sell the energy And I haven’t decided yet. All I know is I am personally 1
portfolio to Citadel and J.P. Morgan with the final transfer more bad day away from stopping out...can’t afford to drop
occurring on September 21, 2006. below 30 for my family.—Amaranth Trader Shane Lee to Brian
If one computes the losses of Amaranth’s natural gas Hunter, September 7, 2006 16:54 (Source: Senate
positions from August 31, 2006 through September 21, 2006, Subcommittee, Exhibit #9).
assuming the positions were not altered during the period, the One of the suggested choices in this email correspondence
losses amount to about $3.295 billion. The actual losses is to increase their positions (choice #2), which some were
computed in Figure 4 total $4.433 billion.9 This difference suggesting. In fact, after this email correspondence, Amaranth
between the losses indicates that the trades that Amaranth modified their natural gas future positions over the next few
executed between August 31, 2006 and September 15, 2006 days. Although it is difficult to quantify in a single number
served to increase their losses by an additional $1.138 billion. exactly what they did, the total number of absolute NYMEX
In fact, these additional losses were probably not accidental natural gas equivalent contracts did increase from around
or random. That is, given the losses up to September 7, 2007, 462,992 on September 7, 2006 to 508,923 on September 13,
the Amaranth energy traders may have exercised their “free 2006. Thus, the additional losses of Amaranth in these days
option” of limited downside liability if things went wrong by were partly due to increasing the actual exposure to natural
increasing the bets in response to troubled times. gas futures contracts, partly due to modifying the positions
Correspondence from an Amaranth trader to Brian Hunter
indicates a line of reasoning along this path:
10
Natural gas futures contracts are denoted by letter symbols on the NYMEX;
F=January, G=February, H=March, J=April, K=May, M=June, N=July,
Q=August, U=September, V=October, X=November, and Z=December.
Thus, the h/j comment is referring to the March-April spread. At the close
of business on September 7, 2006, the March contract (H) was trading at
8
Front-month refers to futures contracts with the nearest month to expiration. $10.073, while the April contract (J) was trading at $8.153. Thus, the h/j
spread was $1.92. In this conversation, the trader is worried that the spread
9
The reader is reminded that these are losses computed from the Amaranth may decline to $1.50 which would cause a position short April and long
natural gas futures equivalent positions. The actual change in net asset value March to lose money.
of the main Amaranth funds was $4.942 billion. The discrepancy is due to
losses from other types of positions not related to natural gas futures trading 11
In this discussion, xh refers to the November and March natural gas futures
and slightly due to the discrepancies between the natural gas future equivalent contracts. See the preceding footnote for more information about contract
positions and the actual positions. symbols.
Figure 4. The Profit and Loss of Amaranth’s Natural Gas Positions in September.
Note: Losses each day are computed based upon the actual positions Amaranth had at the close of the prior day. From September 18,
2006 onwards, the positions of Amaranth were not available, thus the returns are computed assuming Amaranth maintained the positions
they had at the close of business on September 15, 2006.
Profit and Loss (Millions)
A
−800−400 0
09/01/2006
09/05/2006
09/06/2006
09/07/2006
09/08/2006
09/11/2006
09/12/2006
09/13/2006
09/14/2006
09/15/2006
09/18/2006
09/19/2006
09/20/2006
09/21/2006
09/22/2006
09/25/2006
09/26/2006
09/27/2006
09/28/2006
09/29/2006
Actual Daily Positions Positions of September 15
Cumulative Profit and Loss (Millions)
B
−4,000 −2,000 0
09/01/2006
09/05/2006
09/06/2006
09/07/2006
09/08/2006
09/11/2006
09/12/2006
09/13/2006
09/14/2006
09/15/2006
09/18/2006
09/19/2006
09/20/2006
09/21/2006
09/22/2006
09/25/2006
09/26/2006
09/27/2006
09/28/2006
09/29/2006
across the maturity spectrum and partly due to the movement five years out. They also have options on all of the futures
of the options and other positions in the Amaranth portfolio.12 contracts, as well as spread options which pay off on the
difference between futures contract prices of two different
C. Natural Gas Spread Trades months. The initial margin requirement on futures contracts
vary by type of trader (non-member customer, member
Amaranth’s collapse was mainly due to losses in the trading customer, and clearing member and customer) and also vary
of natural gas. To understand the Amaranth collapse, one needs by time to maturity of the contract. Contracts closer to delivery
to understand the mechanics of trading natural gas futures, have stricter margin requirements. To give a flavor of the
options, and swaps. margin differences as a percentage of notional value, on August
31, 2006, $12,150 was required for each October 2006
1. Trading Natural Gas contract (Tier 1), which had a futures value of $60,480, thus,
In this section, some basic features of trading natural gas representing about 20% of the futures notional position. The
futures on the NYMEX and ICE exchanges are discussed. March 2007 contract had a margin requirement of $7,425 (Tier
Traders in natural gas futures have several options. The largest 5) with a notional value of $104,830 or 7.08%. The expiration
exchange for trading natural gas futures is the NYMEX, which of the contracts is usually a few days before the end of the
has futures contracts of consecutive delivery months up to prior month and there are conventions for the last trading day
of each contract which can be obtained from NYMEX.
In addition to NYMEX, traders can use the ICE, which is a
12
As described in (2) of the correspondence, Amaranth may have increased
positions to drive up the spread or “manipulate” the price spread so as to
virtually unregulated exchange but performs very similar
temporarily remove the possibility for further margin calls on the existing functions. ICE is the leading exchange for the trading of energy
spread position.
commodity swaps in natural gas and electricity. “The ICE contract. The closing prices on July 31, 2006 for the March
natural gas swap and the NYMEX natural gas futures contract and April contract were $11.461 and $8.851 respectively. The
perform the same economic functions. The ICE swap contract notional value of this position would equal $114,610 short
even provides that its final settlement price will equal the final and $88,510 long.14 The position is “hedged” in the sense that
settlement price of NYMEX futures contract for the same if natural gas futures prices rise or fall, one position’s loss
month, which means that the final price for the two financial will be partly offset by the other’s gain. However, the position
instruments will always be identical.” (Senate Report, p. 29) is focusing on a spread bet. That is, a bet that the March futures
Traders also can use the ICE trading screen to enter into contracts will have a lower return than the April futures
bilateral, non-cleared transactions rather than cleared contracts. In the month of August 2006, this was actually the
transactions (i.e., Over-the-counter, OTC, transactions with case. By August 31, 2006 the price for March and April 2007
other parties to buy or sell natural gas). One major difference futures contracts was $10.483 and $8.343 respectively. Thus,
between NYMEX and ICE is that ICE has “...no legal if the position were closed out on August 31, 2006 by buying
obligation to monitor trading, no legal obligation to prevent March 2007 futures (covering the short position) and selling
manipulation or price distortion, and no legal obligation to back (offsetting the long position) April 2007 futures, the net
ensure that trading is fair and orderly...” (Senate Report, p. profit would have been $4,700 on this simple spread position
41) due to its status as an electronic trading facility. In addition, (See Supplemental Table 1 on the JAF website). The return
the Commodity Futures Trading Commission (CFTC) has no of these positions will depend on the leverage employed.
authority or obligation to monitor trading on ICE. Notice that even though natural gas prices dropped, the spread
position still made profits.
2. The Natural Gas Futures Spread Trade On July 31, 2006, these natural gas futures contracts
represented the Tier 5 futures contracts on the NYMEX for
A popular type of trade in natural gas futures is to short one
margin calculation.15 For a non-member customer, this would
contract, while going long another contract. This type of trade
require an initial margin on each of the March and April
has several attractive features. First, the trade as a whole will
contracts of $7,425. Thus, for an initial capital outlay of
have less risk to the direction of natural gas futures prices - in
$14,850, the return on this investment would have been 31.6%
a sense, “hedge-like” in nature. Second, by shorting one
contract and being long another contract, an entity will reduce 4, 700
their overall net position and hence may allow for greater ( ). This is one of the advantages of leverage; big returns
14,850
positions on the exchange without causing a trader to hit for little initial capital outlay.
position limits.13 NYMEX’s control system will investigate
any position with a size greater than the position limit in that 3. The Natural Gas Spread Trade with Options
contract. However, if the entity is questioned by NYMEX
about the position, an offsetting position in another contract The previous section discussed one way a natural gas futures
may be an acceptable reason for NYMEX to allow the trade trader can engage in a calendar spread trade using natural gas
in excess of the position limit. Third, if the trade is done as a futures contracts on the NYMEX. In addition to this, a trader
spread position, then the actual margin requirements from could use NYMEX natural gas options, which are options
NYMEX are lower allowing greater leverage possibilities. whose value depends on the underlying natural gas futures
Even if position limits are reached, by being short one contract contract. There are both call and put options and they are
and long another contract, the entity will have a better story available for selling or purchasing.16 Thus, the trader could
of why they have such large positions (i.e. the position is also make a calendar spread trade using options.
naturally hedged) and may be allowed to engage in such In addition to straight call and put options, the NYMEX
positions on the exchange. Fourth, spread positions allow for also has calendar spread options available for trading. These
more sophisticated hedge fund-like trades. are options on the difference in price between two natural gas
A simple example of a spread position may illustrate the
point: Suppose on July 31, 2006 a trader wished to short one 14
Each contract of natural gas is worth 10,000 MMBtu. Natural gas futures
contract and go long another contract. Suppose the trader chose prices are quoted in terms of 1MMBtu. Thus, each contract in natural gas
to short the March 2007 contract and go long the April 2007 futures represents a notional value of 1×P×10, 000, where P represents the
price of that natural gas futures contract.
15
For more details, see www.nymex.com for margin requirements. Tier 5
13
That is NYMEX looks not only at individual contract position limits to represents the 6th through the 16th nearby month. On July 31, 2006, March
decide about a particular entity, they also consider net exposure limits. Thus, and April contracts were the 8th and 9th month respectively.
if a trader is long 10,000 contracts in one contract and short 10,000 contracts
in another contract, the net position is 0. This makes the position more Margin is required for short positions or writing options. However, for
16
feasible with respect to NYMEX acceptability of such a position. purchasing options, only the premium is required.
futures contracts of different months. For example, an IBK07 most part, the complex combination of instruments and spread
call option is a call option on the price differential between trades could be summarized as a general bet that winter natural
the May 2007 natural gas futures contract and the July 2007 gas prices would rise relative to non-winter natural gas prices,
natural gas futures contract. referred to as the long winter, short non-winter spread trade
(Chincarini (2006, 2007a, 2007b)).
4. Natural Gas Swaps Amaranth’s positions in natural gas consisted of a variety
of actual instruments. The vast majority of positions were
Finally, using the NYMEX Clearport trading platform,
traded on the NYMEX and ICE. On the NYMEX, Amaranth
traders can transact in natural gas swaps and natural gas
held positions in outright natural gas futures contracts from
penultimate swaps
October 2006 maturity to
which are based upon
the final price of the It is difficult to classify a large group of December 2011 maturity.
Amaranth also had a
natural gas futures trades into one simplified strategy, but for significant amount of
contracts, but are one-
fourth the size. the most part, the complex combination of positions in call and put
options on the underlying
A trader could also do
such a spread trade using
instruments and spread trades could be natural gas futures contracts
the ICE. The ICE allows summarized as a general bet that winter with NYMEX. They also
had natural gas swap
for trading of natural gas
swaps that are based on natural gas prices would rise, while non- contracts through the
Clearport system of
the settlement prices of winter natural gas prices would increase NYMEX. They had a
the NYMEX natural gas
futures contracts.17 The to a lesser degree, referred to as the long combination of regular
swaps and penultimate
ICE swaps are, for all winter, short non-winter spread trade. swaps, the latter which
practical purposes,
expire one day prior to the
identical in behavior and
former, but are otherwise identical.18 The rest of their positions
risk to the NYMEX natural gas futures contracts. For more
consisted of natural gas swap contracts on ICE, some of which
details about these swaps, the reader is referred to the detailed
were electronically traded and cleared positions on ICE, while
version of this section on the JAF website.
others were off-exchange contracts, but later cleared through
All of the positions and types of trades we discussed in this
ICE. Among the trades entered on ICE, some of the swap
and the preceding sections were employed by Amaranth. In
contracts were in individual contract months (e.g. October,
fact, Amaranth’s collapse was due to a large variety of these
2006), while others were in calendar strips (e.g. November
type of trades that they made on NYMEX and ICE in both
through March). Due to the difficulty of understanding
futures, swaps, and options. In the next section, we focus on
Amaranth’s positions when divided amongst so many types
the Amaranth trades in detail.
of securities, it is useful to convert all of the securities into the
NYMEX futures equivalent value (NYMEX FEQ). For the
III. Amaranth’s Trading Strategy swap contracts, this is quite easy to do, since the swaps are
essentially the same as the NYMEX natural gas futures
A. The Basic Strategy contract, but one-fourth the size. Thus, one swap contract is
worth one-fourth of a NYMEX natural gas futures contract.
The Senate’s Permanent Subcommittee on Investigations
The option contracts are more complicated, but can be
report (2007a) provided a detailed account of Amaranth’s
translated by adjusting the position for the delta of the option.19
natural gas positions on a daily basis throughout 2006.
Once these conversions have been made, we can aggregate
Amaranth’s positions in natural gas involved trades in various
the entire Amaranth position in terms of NYMEX natural gas
types of contracts, including futures, swaps, and options. Their
futures equivalents.20 Table I shows the positions of Amaranth
trades also amounted to a collection of many spread trades
whose return depended on the movement of natural gas futures
18
These were created to allow traders access to an instrument that would
price all the way out until 2011. It is difficult to classify a expire one day before option expiration on natural gas futures contracts.
large group of trades into one simplified strategy, but for the 19
For more on this type of concept, one is referred to any options book or
any book on value-at-risk. For example, see Hull (2006), Jorion (2006), or
Dowd (1999).
Although the ICE calls these instruments “swaps”, they are similar to futures
17
The conversion of these positions was done by the NYMEX and the Senate
20
contracts. Subcommittee.
Table I: Natural Gas Positions of Amaranth on August 31, 2006
Note: On the NYMEX, Amaranth held positions in outright natural gas futures contracts from October 2006 maturity to December 2011
maturity. Amaranth also had a significant amount of positions in call and put options on the underlying natural gas futures contracts with
NYMEX. They also had natural gas swap contracts through the Clearport system of NYMEX. They had a combination of regular swaps
and penultimate swaps, the latter which expire one day prior to the former, but are otherwise identical. The rest of their positions
consisted of natural gas swap contracts on ICE, some of which were electronically traded and cleared positions on ICE, while others
were off-exchange contracts, but later cleared through ICE. Among the trades entered on ICE, some of the swap contracts were in
individual contract months (e.g. October, 2006), while others were in calendar strips (e.g. November through March). All of these
different types of instruments were converted to NYMEX futures equivalent value (NYMEX FEQ).
Contract NYMEX Contracts ICE Contracts Total

Futures Options Swaps (NN) Swaps (NP) ICE Swaps ICE Off-Exchange
Oct-06
FEQ -64711 43523 -21703 -5307 -87625 41381 -94441
Percent 24.49 16.47 8.21 2.01 33.16 15.66 100
Nov-06
FEQ -336 6431 17451 -442 85597 -49453 59247
Percent 0.21 4.03 10.93 0.28 53.60 30.96 100
Dec-06
FEQ -7308 -2430 -8154 -449 28711 -38127 -27757
Percent 8.58 2.85 9.57 0.53 33.71 44.76 100
Average Percent 28.40 14.82 32.61 2.00 10.21 11.96 100
in the various instruments as NYMEX natural gas futures that Amaranth’s position was a long winter, short non-winter
equivalents on August 31, 2006. position. Although the figure seems to indicate this, it is worth
For example, on August 31, 2006, Amaranth had a net examining the issue further.21 For the purposes of this analysis,
position of October 2006 NYMEX natural gas futures we follow Chincarini (2007a) and define winter contract
equivalent contracts of -94,441. That is, the combined position months to be November, December, January, February, and
of NYMEX natural gas futures, options, and swaps and ICE March. All other months will be considered non-winter
swaps was equivalent to a short position in 94,441 NYMEX months.
natural gas futures contracts. In fact, many of the outright Table III presents additional measures of the August 31,
positions on the October 2006 were short (i.e. 64,711 NYMEX 2006 positions of Amaranth in natural gas. The total dollar
natural gas futures contracts, 21,703 and 5,307 NYMEX swap value of natural gas futures positions by Amaranth in winter
contracts, and 87,625 ICE swaps), but some positions had a months equalled $23,489,626,234. That is, the notional value
long exposure (i.e. 43,523 NYMEX options and 41,381 off- of all winter contract months was almost $23 billion across
exchange ICE swaps). For October, ICE swap contracts all exchanges and instruments. The total dollar value of non-
represented the largest component of the trade at 33.16% of winter positions was -$15.863 billion. This is consistent with
the position. a long winter, short non-winter position.
For the entire period, looking at all contract months in which Another way to measure whether Amaranth’s strategy was
they had positions, the averages are shown in the last row of long winter and short non-winter is to find the percentage of
Table I. On average, 28.40% of the monthly exposures were winter months in which they had long positions versus short
through NYMEX natural gas futures contracts, 14.82% in positions. Of all the contract months out until December, 2011,
NYMEX options, 34.61% in NYMEX swaps, 10.21% in ICE 35 of those months are non-winter months, while 27 are winter
swaps, and the remaining 11.96% in ICE off-exchange swaps. months. For winter months, Amaranth had a long position 63%
Amaranth’s actual positions in natural gas future equivalents of the time, while for non-winter months, Amaranth had a
on August 31, 2006 are depicted in Figure 5 and Table II. short position 69.44% of the time. This is again consistent
This graph is identical to the graph produced in the appendix with a long winter, short non-winter strategy. And within the
of the Senate Subcommittee’s report. It contains the Amaranth winter months, they had an equivalent of $28.812 billion long
positions on each contract month in NYMEX natural gas
futures equivalents. Before the data on Amaranth’s positions
Although some winter months are actually shorted, the overall positions
21
were publicly available, Chincarini (2006, 2007b) postulated are smaller.
Figure 5. Amaranth NYMEX FEQ Positions on August 31, 2006

NYMEX Natural Gas Futures Equivalent Contracts
−100000 −50000 0 50000
OCT.06
NOV.06
DEC.06
JAN.07
FEB.07
MAR.07
APR.07
MAY.07
JUN.07
JUL.07
AUG.07
SEP.07
OCT.07
NOV.07
DEC.07
JAN.08
FEB.08
MAR.08
APR.08
MAY.08
JUN.08
JUL.08
AUG.08
SEP.08
OCT.08
NOV.08
DEC.08
JAN.09
FEB.09
MAR.09
APR.09
MAY.09
JUN.09
JUL.09
AUG.09
SEP.09
OCT.09
NOV.09
DEC.09
JAN.10
FEB.10
MAR.10
APR.10
MAY.10
JUN.10
JUL.10
AUG.10
SEP.10
OCT.10
NOV.10
DEC.10
JAN.11
FEB.11
MAR.11
APR.11
MAY.11
JUN.11
JUL.11
AUG.11
SEP.11
OCT.11
NOV.11
DEC.11
Contract Month
Table II. NYMEX Futures Equivalent Values of Positions for Amaranth on August 31, 2006
Note: NYMEX FEQ refers to NYMEX futures equivalent values of positions. Only the positions for contracts out to September 2007
are listed in this table. For a table of all of their positions in natural gas on August 31, 2006, see the expanded version of this table on the
JAF website. Weight represents the weight of Amaranth’s exposure in that particular contract as a percentage of the total absolute dollar
volume of all contracts. That is, for each contract, the absolute value of Amaranth’s positions are multiplied by the price for that contract
on August 31, 2006 and 10,000. The percentage for each contract of each contract is the total dollar value of their position in that
contract divided by the sum of the total dollar value of all of the contracts. The Dollar P/L represents the profit and loss of Amaranth in
each position assuming no changes were made to the holdings. That is, it is simply Dollar P/L = NYMEX FEQ × ( Pt +1 − Pt ) , where Pt
is the contracts price on August 31, 2006 and Pt+1 is the contract’s price on September 21, 2006.
Percent of NYM EX Dollar P/L

Contract M onth NYM EX FEQ W eight Open Interest (August 31, 2006 -
September 21, 2006)
OCT.06 -94441 0.1068 -80.8 $1,196,571,821
NOV.06 59247 0.0911 84.1 $(1,313,512,297)
DEC.06 -27757 0.0518 -54.3 $718,082,127
JAN.07 61825 0.1228 125.5 $(1,698,345,675)
FEB.07 -7464 0.0149 -24.1 $204,658,602
MAR.07 58365 0.1144 73.2 $(1,597,458,370)
APR.07 -77527 0.1209 -123.9 $912,497,139
MAY.07 -140 0.0002 -0.6 $1,491,906
JUN.07 869 0.0013 5.7 $(9,226,529)
JUL.07 -1612 0.0025 -13.9 $17,362,443
AUG.07 406 0.0006 3.1 $(4,408,604)
SEP.07 -1128 0.0018 -9.6 $12,318,357
Table III: Amaranth Positions in Winter and Non-Winter Months
Note: For this table, winter months are defined to be November, December, January, February, and March. Non-Winter months are all
other months. For each day listed, Winter-Longs represented the total dollar value of the long positions in winter months, Winter-Shorts
represent the total dollar value of the short positions in winter months, W. Total represents the sum of the two, Non-Winter-Longs
represents total dollar value of the long positions in non-winter months, Non-Winter-Shorts represents the total dollar value of the short
positions in non-winter months, and N.W. Total represents the sum of the two. Correct Sign (%) represents the number of Winter (Non-
Winter) months in which the position is long (short) regardless of size.
Total Dollar Value Correct Sign (%)

Trade Winter- Winter-Shorts W. Total Non-Winter- Non-Winter- N.W. Total Winter Non-
Date Longs (Net) Longs Shorts (Net) Winter
31-Jan-06 4,258,305,934 (4,207,665,123) 50,640,811 1,435,236,076 (2,186,529,127) (751,293,051) 64.29 50.00
28-Feb-06 6,747,057,844 (2,581,042,631) 4,166,015,213 1,107,062,004 (4,459,247,449) (3,352,185,445) 77.78 50.00
31-Mar-06 8,139,116,076 (1,823,491,062) 6,315,625,014 1,414,829,338 (5,252,719,674) (3,837,890,336) 70.37 51.22
28-Apr-06 11,676,812,614 (3,236,275,580) 8,440,537,034 1,927,180,168 (6,202,124,031) (4,274,943,863) 70.37 57.50
31-May-06 17,101,267,975 (4,524,524,915) 12,576,743,060 2,782,321,098 (11,225,510,296) (8,443,189,198) 70.37 48.72
30-Jun-06 20,229,114,833 (5,357,498,215) 14,871,616,618 3,222,527,838 (11,998,686,079) (8,776,158,242) 66.67 47.37
31-Jul-06 28,568,081,397 (2,432,009,020) 26,136,072,377 1,198,034,025 (19,426,414,857) (18,228,380,831) 62.96 56.76
31-Aug-06 28,812,493,335 (5,322,867,101) 23,489,626,234 1,762,963,323 (17,626,398,609) (15,863,435,286) 62.96 69.44
and $5.322 billion equivalent short positions. For the non- and August 31, 2006 (see Table III). Thus, even months prior
winter months, they had an equivalent $17.626 billion of short to August 31, 2006, Amaranth had engaged in a long winter,
positions and $1.762 billion of long positions. short non-winter spread trade in natural gas.
Thus, although not every winter contract was held long and
not every non-winter month was held short, the Amaranth B. The Rationale for the Strategy
actual positions on August 31, 2006 seemed to be consistent
with a long winter and short non-winter spread trade in natural In the previous section, we concluded that Amaranth’s
gas using a combination of NYMEX futures, swaps, and primary trading strategy consisted of a spread trade that was
options, as well as ICE natural gas swaps. primarily long winter natural gas contract months and short
It’s clear that on August 31, 2006, Amaranth was engaged non-winter natural gas contract months. Chincarini (2007a,
in a natural gas futures position that was long winter and short 2007b) noted that such a spread trade had performed well on
non-winter. Next we examine whether or not they had a similar average since 1990. That is, a long winter, short non-winter
trade in prior months. In order to examine the general position spread trade in proportion to the open interest on NYMEX
of Amaranth, we look at their position three months prior to tended to do very well in September. It is not clear whether
August 31, 2006. The NYMEX natural gas futures equivalents the Amaranth natural gas traders actually backtested the
of Amaranth’s natural gas positions on May 31, 2006 are strategy or whether they used experience combined with their
depicted in the supplemental figure on the JAF website and own trader instinct.22 If one backtests the Amaranth strategy
Table III. The total absolute dollar value of winter month of August 31, 2006 on past years, one finds that the strategy
contracts was $12.577 billion, while the non-winter months produced a significantly positive average return of 0.74% per
was $8.443 billion. Of the winter month contracts, 48.7% were month or 8.96% on an annualized basis with relatively small
held long, while 70.4% of the non-winter months were held losses in down years (See Figure 6).
short. The total value of long positions in winter months was One might naturally ask if there is some potential reason
$17.101 billion, while short positions were $4.525 billion; explaining this historical pattern. More specifically, one might
for non-winter it was $2.782 billion and $11.226 billion ask if there is a justifiable reason for a trade that is long winter
respectively. Although not a perfectly consistent winter/non- and short non-winter to earn an excess return. Natural gas is
winter spread trade, the general position of the trade is long one of the main sources of energy for the United States, fueling
winter and short non-winter on May 31, 2006 as well.
The natural gas positions of Amaranth on other days during
22
Backtesting a strategy refers to the process of testing a trading strategy on
prior time periods. In other words, a trader can do a simulation of his or her
the summer are of a similar nature to those on May 31, 2006 trading strategy on relevant past data in order to gauge the effectiveness.
Figure 6. Historical September Returns (1990-2006) from Positions Similar to Amaranth’s Position on
August 31, 2006
3
2
Returns (%)
1
0
−1
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Year
nearly one-quarter of the nation’s energy consumption. Natural On the other side of this trade would be the speculator who
gas is used by individual households, small businesses, and buys winter contracts and shorts non-winter contracts
large industries. The total domestic demand for natural gas is providing liquidity to the natural hedgers. In exchange for
highly seasonal, this is mainly because natural gas is the taking on this risk, the speculator should receive compensation
primary heating fuel for homes in the winter months.23 “During on average. This might explain the positive average return to
summer months, when supply exceeds demand, natural gas this strategy over time. Thus, the excess returns from a long
prices fall, and the excess supply is placed into underground winter, short non-winter trade in September might be a
storage reservoirs. During the winter, when demand for natural compensation to speculators for supplying liquidity to natural
gas exceeds production and prices increase, natural gas is hedgers, which consist of storage operators and natural gas
removed from underground storage.” (Senate Report, p. 17). producers.25
In many commodity markets, the storage costs of a commodity A quantitative type of trader would have probably backtested
are priced into futures contracts. Theoretically, the price of a the winter-summer spread strategy and found that it produced
futures contract is given as Ft = Ste(c+r)(T–t), where S is the spot significant excess returns historically and might have used
price of the commodity, c is the continuously compounded this as a basis to make such a trade going forward. However,
storage costs of the commodity, r is the opportunity cost of it is difficult to determine if Amaranth’s traders had based
money or the interest rate, and T – t is the time until the futures their strategy on a similar motivation. It is somewhat reassuring
contract matures. to find that the Amaranth strategy generated positive average
Thus, a storage operator might use natural gas futures to returns historically. However, in my opinion, the traders were
hedge his or her exposure. That is, by selling natural gas winter not relying on statistical techniques, but rather were using their
contracts and buying non-winter months, the storage operator instincts and experience in natural gas futures which was
will lock-in his or her profit for storage, which in a perfectly conditioned by this historical pattern. Their view was also
competitive market should cover interest and storage costs.24 influenced by their beliefs about the demand and supply of
natural gas in 2006. Interviews with Amaranth traders revealed
23
A 2001 EIA survey found that 54% of all U.S. household use natural gas that they believed that winter natural gas prices would rise
as the main heating fuel (Source: Senate Report).
24
Even though this is an overly simplistic description of the real behavior of
storage operators, it may help explain some of the reasons why a speculator The forward curve for natural gas futures looks like a sine wave with natural
25
might choose this side of the trade. Natural gas producers might accentuate gas futures prices high in winter months and low in non-winter months.
the need for speculators as they might continuously short natural gas as a Another reason is that there is lower demand for natural gas in summer
hedge which might require more liquidity for winter contracts. months and higher demand in winter months.
throughout 2006. They believed that with increasing domestic It could be that historical measures of natural gas volatility
demand for natural gas, they expected supply shortages, were insufficient to identify the types of events that occurred
delivery bottlenecks, and weather-related disruptions to in September, 2006, or it could be that Amaranth simply
develop during the winter and boost prices. From early 2006, ignored the warning signs from risk measurement systems.
they believed that the fundamentals of supply and demand Or, it might be that market risk was not the principal risk of
justified much higher spreads between the natural gas winter the positions, but it was rather liquidity risk. In this section,
and summer prices (Senate Report, p. 56). we take the actual Amaranth positions in natural gas and
In addition to this, a lot of their trading around the main attempt to construct both market risk and liquidity risk
position seemed to be driven by typical trader instinct, measures using only data up to August 31, 2006 to examine
sentiment, and weather conditions, rather than some well- whether or not the risks of the Amaranth portfolio could have
designed trading strategy. Many of the instant message and been obtained from basic risk measurement tools. In particular,
email conversations between Brian Hunter and other traders we examine three sources of risk for Amaranth: market risk,
seemed to reveal this.26 For example, in one email, an liquidity risk, and funding risk. Market risk is the risk that
Amaranth employee writes to Brian Hunter: occurs from the volatility of investment returns. Liquidity risk
I think you should sell 15,000 red March April and buy measures the degree of difficulty in exiting a given trading
15,000 (or more) front Mar/Apr. My rationale is not that you position. Funding risk measures the extent to which they were
should short the reds, just that you’re moving risk...not able to meet margin calls on their natural gas positions.
increasing it. Leveraging it to the part of the curve that is
undervalued and lightening up on the one that is perhaps A. Market Risk
fair value.27—Amaranth Employee, Email to Brian Hunter,
In order to evaluate Amaranth’s market risk on August 31,
July 28, 2007 (Source: Senate Subcommittee, Exhibit #9)
2006, simple historical VaR (value-at-risk) measures are
constructed for their actual positions. We consider three ways
IV. The Risks of Amaranth’s Strategies to measure this VaR. The first method is computed by
recreating the August 31, 2006 natural gas exposures of
As was described in Section II.D, Amaranth had an
Amaranth in other years from 1990-2005 (See Table II). Table
apparently sophisticated risk management operation with 12
II shows the weight of Amaranth’s exposure to each contract
dedicated risk managers supporting each desk, including a
month of natural gas futures. This weight is computed by taking
Chief Risk Officer. They used daily VaR and stress reports, so
the absolute value of the notional value of each contract and
one might naturally ask how they did not foresee the risks
dividing it by the sum of the absolute notional value of all
they were taking on August 31, 2006. In fact, the CEO of
other contracts. For example, for the October contract month,
Amaranth stated in a conference call to investors that:
this was equal to 10.68%. For prior years, the weight scheme
Although the size of our natural gas positions was large,
was kept similar. That is, in each prior year, the weight of the
we believed, based on input from both our trading desk and
October current year contract was kept at 10.68%. The
the stress-testing performed by our energy risk team that the
corresponding returns of these positions were computed in
amount of risk capital ascribed to the natural gas portfolio
every year from the last trading day in August to the last trading
was sufficient. In September 2006, a series of unusual and
day in September. These 16 years of September returns were
unpredictable events caused the Funds’ natural gas positions
then used to calculate a sample average and standard deviation
(including spreads) to incur dramatic losses while the market
of the strategy in September to be used to estimate a VaR for
provided no economically viable measure of exiting these
the strategy in September.28
positions.—Nick Maounis, Conference Call to Investors,
September 22, 2007
The return calculation for the strategy is given by rt = ∑ i =1 wi ,t −1ri ,tφi ,t −1 ,
N
28
26
These documents were obtained by subpoena from the Senate where wi, t–1 is the weight of contract i on the last trading day of August in
Subcommittee and used in the public presentation of the Amaranth case. In any given year, ri, t is the return of natural gas futures contract i from the last
particular, they were taken from Exhibit #9 of the Senate Subcommittee trading day in August to the last trading day in September in any given year,
Investigation documents. and φi is an indicator variable that equals 1 if Amaranth was long in that
particular futures contract and equals -1 if Amaranth was short that particular
27
In natural gas trading, colors are used to distinguish between contracts of contract month, and N represents the total number of contract months (e.g.
different years. “Front” refers to the contract month closest to the current 63 from October 2006 to December 2011). In some years, especially in the
date. For instance, on July 28, 2007, the “front March” contract would be early 1990s, there were not as many natural gas futures positions and thus
the March 2008 contract. “Red” refers to the next contract year. Thus, in the weights were renormalized so as to be relatively the same between any
this case, “red March” contract would refer to the March 2009 contract. two contracts. For example, on August 31, 1990 there were only 12 contracts
“Blue” is also used to denote the contract 2 years out. Thus, if someone from October 1990 to September 1991. Thus, the weight for October 1990
referred to the “blue March” contract on this date, it would refer to the was -0.1697 and the weight for November 1990 was 0.14483. The relative
March 2010 contract. These colors help traders communicate more easily. weight was still -1.172 as in other years.
The VaR was computed as with the same sized position since the opening of natural gas
trading in 1990 would have been -$719.7 million. The average
VaRt = Vt ( µ - k (α )σ ) (1) return of the spread position over the prior 16 years was
0.7466% with a sample standard deviation of 1.3902 in
where µ represents the average historical return of the strategy
September. Thus, if Amaranth used a simple risk measurement
in September, σ represents the standard deviation of the
system as used here, they would have been chasing an average
historical September returns, Vt represents the notional value
return of $399.6 million (0.7466)×($53,524,979,536) with a
of the portfolio positions, and k(α) represents the critical value
potential 99.95% VaR of -$2.048 billion.
from the normal distribution for a confidence level (1 –α)
Thus, they were chasing a 4.13% return in September for a
[i.e. k(0.025) = 1.96 for a 97.5% confidence interval].
“worst-case” scenario of a loss of 21.2%.31 This is, in itself,
The second method is a modification of the first method to
quite risky, but perhaps part of their philosophy. It should also
account for non-normally distributed returns. It is the Cornish-
be noted by looking at Figure 3 that the historical returns of
Fischer expansion VaR ((Cornish and Fisher (1937), Ord and
such a spread trade seemed to look favorable. The strategy
Stuart (1994), and Favre and Galeano (2002)). This method
provided mainly positive returns with a positively skewed
adjusts the VaR calculation taking into account the skewness
distribution. The largest negative return of the trade was 1.34%
and kurtosis of the distribution of returns.29
in 1991 on an unlevered basis.
The third method is to measure the most recent volatility in
The other methods show similar results. The Cornish-Fisher
natural gas futures over the three months prior to August 31,
VaR is actually smaller reflecting the negative kurtosis of the
2006. Ideally, one would like to recreate the same type of
sample distribution and very slight skewness.32 The VaR based
positions in the past as what Amaranth had on August 31,
upon the last three months of Amaranth positions reflected a
2006, but there is no obvious way to do this, since a whole
lower VaR than the historical calculation, but basically near
host of different contract months are introduced. Instead, the
the same magnitude.
actual positions of Amaranth from May 31, 2006 to August
It is clear from this exercise, the losses of September were
31, 2006 are used and the daily returns calculated. The VaR
not entirely explained by VaR calculations. The further losses
for September on August 31, 2006 is then computed as follows:
may have come from another source of risk which they failed
VaR t = Vt ( µ d T − k (α )σ d T ) (2) to manage as well: liquidity risk.
where µd represents the daily return of the strategy over the B. Liquidity Risk
past three months, σd represents the standard deviation of daily
Liquidity is defined as the ability to sell a quantity of a
returns over the last three months, and T represents the number
security without adversely changing the price in response to
of trading days that Amaranth used for VaR (i.e. 20 days).
your orders. Models for liquidity risk are not as common place
The confidence levels were chosen to conform closely with
as models for market risk. One simple precautionary measure
the risk reports that Amaranth produced internally on a daily
that practitioners use to control liquidity risk is to measure
basis (see Section II).
the size of their trades versus the average daily trading volume
Table IV shows the potential VaR from the spread positions
of a security. A rule-of-thumb is to not own positions greater
and different confidence intervals. Suppose we take the 99%
than 1/10 to 1/3 of the average daily trading volume over some
confidence interval for use with our Method 1 VaR calculation
specified time interval, for example, the last 30-days of trading.
at the end of August 2006. A notional position in the spread
Figure 7 shows Amaranth’s August 31 positions as multiples
trade of $10.228 billion would give us a VaR calculation of
of the trailing 30-day average daily trading volume in each
$254.95 million.30 The actual leveraged position of Amaranth
contract for the spread position. For example, Amaranth’s
had an estimated VaR of $1.33 billion. This is a sizeable
exposure in terms of NYMEX natural gas futures equivalents
amount of VaR, however it is not the actual amount they lost
in July 2008 futures contracts represented 253 days of the
in September. The actual amount they lost from August 31,
average daily trading volume. Even though many of the
2006 to September 21, 2006 had the positions been held
Amaranth positions were not with NYMEX, and instead with
constant was around $3.295 billion which is listed under the
ICE, these positions were extremely large relative to the
column “Actual” in the table.
average daily trading volume of the largest natural gas futures
Prior to that year, the worst lost in September of any year
This downside percentage is for the 99.99% confidence level VaR. It would
31
The actual calculation of the Cornish-Fisher VaR is contained in the detailed

29
be much less for the 99% VaR at -13.8%.
version of this section on the JAF website.
32
It should be noted that the Cornish-Fisher VaR critical values began to
30
This net asset value differs from that in Chincarini (2007b) due to a typo decrease when the critical values where extended to a 99.99% confidence
in the earlier paper. interval.
Table IV: Measures of VaR of Amaranth’s Natural Gas Position on August 31, 2006
Note: a Actual losses represent the losses had Amaranth maintained the positions of August 31, 2006 through the end of trading on
September 21, 2006. b No leverage computes the VaR based on an investment in natural gas futures equal to the value of the total assets
under management by Amaranth on August 31, 2006 of $10.228B. The Leverage row represents the VaR with Amaranth’s actual
leverage of 5.23 on August 31, 2006. For Methods 1 and 2, the numbers for each confidence level in the table represent the VaR
estimates in millions of dollars using the historical mean and volatility of the winter / non-winter spread trade of 0.7466% and 1.3902%
respectively. For Method 3, the VaR estimates are based on the daily mean and standard deviation of Amaranth’s natural gas positions
for the prior three months. These daily values were 0.0172% and 0.2435% respectively. The “Worst” column represents the losses of the
respective size fund if one uses the worst historical September loss of the spread trade using NYMEX data from 1990-2005. The
“Actual” column represents the actual loss that occurred for Amaranth from August 31, 2006 to September 21, 2006 assuming no
changes were made to the positions held on August 31, 2006.
Confidence Interval
a
Position Size 68% 99% 99.95% Worst Actual
Method 1 (VaR)
No Leverageb $10.228B -65.83 -254.95 -391.53 -137.53 -629.97
Leverage $53.523B -344.50 -1334.18 -2048.92 -719.71 -3295.50
Method 2 (Cornish-Fisher VaR)
No Leverage $10.228B -126.44 -246.31 -225.14 -137.53 -629.97
Leverage $53.523B -661.67 -1288.97 -1178.16 -719.71 -3295.50
Method 3 (Recent Historical VaR)
No Leverage $10.228B -76.27 -224.43 -331.42 -137.53 -629.97
Leverage $53.523B -399.12 -1174.44 -1734.37 -719.71 -3295.50
Figure 7. Amaranth’s August 31, 2006 Positions as a Ratio to 30-Day Average Daily Trading Volume
150
100
Days of Trading Volume
50
−50
−100
−150
−200
−250
−300
OCT.06
NOV.06
DEC.06
JAN.07
FEB.07
MAR.07
APR.07
MAY.07
JUN.07
JUL.07
AUG.07
SEP.07
OCT.07
NOV.07
DEC.07
JAN.08
FEB.08
MAR.08
APR.08
MAY.08
JUN.08
JUL.08
AUG.08
SEP.08
OCT.08
NOV.08
DEC.08
JAN.09
FEB.09
MAR.09
APR.09
MAY.09
JUN.09
JUL.09
AUG.09
SEP.09
OCT.09
NOV.09
DEC.09
JAN.10
FEB.10
MAR.10
APR.10
MAY.10
JUN.10
JUL.10
AUG.10
SEP.10
OCT.10
NOV.10
DEC.10
JAN.11
FEB.11
MAR.11
APR.11
MAY.11
JUN.11
JUL.11
AUG.11
SEP.11
OCT.11
NOV.11
DEC.11
Contract Month
exchange. In some cases, the positions are hundreds of times letter from NYMEX and a CFTC investigation. In addition to
the 30-day average daily trading volume. It is quite clear that this, Amaranth exceeded NYMEX position limits virtually
Amaranth was taking immense risk with respect to liquidity. every month in 2006 triggering reviews of Amaranth’s
Another way of depicting Amaranth’s natural gas positions positions.
is to compare them to the open interest of NYMEX natural Of particular note was an August 8, 2006 complaint by
gas futures contracts (abbreviated as NYMEX NGFOI). Figure NYMEX officials that Amaranth’s position in the September
8 compares the actual Amaranth positions to the open interest 2006 contract (near-month contract) was too high at 44% of
of NYMEX natural gas futures. Figure 8A shows all the the open interest on NYMEX. Figure 9 shows that Amaranth
Amaranth positions (including ICE positions as well) as a reduced this short position by the day’s close by 5,379
percentage of the NYMEX NGFOI. In many contract months, contracts (see the change in NYMEX contracts from the close
this is greater than 100%.33 Figure 8B shows only the positions of August 7 to the close of August 8), but they also increased
on NYMEX as a percentage of NYMEX NGFOI. It is still their similar exposure short position on ICE by 7,778 contracts.
very high and, in some contracts, greater than 100% as well. Thus, ironically, the request by NYMEX to reduce Amaranth’s
Figure 8C shows only Amaranth’s position in NYMEX natural positions led Amaranth to actually increase their overall
gas futures as a percentage of NYMEX NGFOI. Even by this September 2006 position. At the same time, they also increased
very direct measure of Amaranth’s positions on the NYMEX their exposure to the October 2006 contract; a contract that is
exchange, their positions were excessive representing more a close substitute to the September 2006 contract. In particular,
than 50% of the open interest in many contracts and almost they had increased their October 2006 position in NYMEX
100% in some contracts. In some contracts, Amaranth had natural gas futures by 7,655 contracts and their equivalent
positions of nearly 100,000 contracts, which represents position on ICE October 2006 contracts by 4,984.
roughly 1 trillion cubic feet of natural gas, 23% of the amount On August 9, 2006 the NYMEX called Amaranth with
of natural gas consumed by residential users in 2006, and 5% continued concern about the September 2006 contract and
of the total amount of natural gas consumed in the United warned that October 2006 was large as well and they should
States in 2006 (Senate Report, p. 64). not simply reduce the September exposure by shifting contracts
Thus, while market risk measures such as VaR indicate that to the October contract. In fact, by the close of business that
Amaranth may have had a VaR of about -$2.048 billion, their day, Amaranth increased their October 2006 position by
liquidity risk was also very high. Thus, Amaranth was certainly 17,560 contacts and their ICE positions by 105.75. For
being imprudent with respect to its natural gas futures positions September 2006, Amaranth did follow NYMEX instructions
in terms of the size versus the market size. This may have by reducing NYMEX natural gas positions by a further 24,310,
resulted in the extra $1.247 billion losses not accounted for but increased September ICE positions by 4,155.
by simple VaR measures.34 On August 10, 2006 another call from NYMEX urged
In addition to these measures showing Amaranth’s excessive Amaranth to reduce the October 2006 position since it
positions in natural gas, Amaranth was continuously represented 63.47% of the NYMEX open interest. In response
reprimanded by NYMEX for violating trading standards and to this call, Amaranth reduced the October 2006 position by
position limits on NYMEX. The Senate Subcommittee report 9,216 contracts, but increased their similar October 2006 ICE
discusses these violations in detail (See Senate Report, pp. position by 18,804 contracts.
90-99). On April 26, 2006 for example, Amaranth violated By the end of this three-day session of calls from the
trading rules on the May 2006 futures contract resulting in a NYMEX warning Amaranth of its position size in September
and October contracts, Amaranth had actually increased their
overall positions from August 7, 2007 to August 11, 2006 in
33
The reason that the percentage of Amaranth positions is greater than 100% those two contracts by 16,484 (a decrease in September 2006
is twofold. Firstly, included in this calculation are Amaranth positions on positions by 23,143 and an increase in October positions by
ICE, which thus is additional contracts to what NYMEX has. Secondly, the 39,627).
measure of Amaranth’s positions included options, swaps, and other
instruments that are not strictly NYMEX natural gas futures contracts, but The Senate Report highlighted that one of the problems
are natural gas futures equivalents as computed by the Senate Subcommittee with the current system is that electronic exchanges like ICE
and NYMEX. Thus, only in Figure 4C should percentages not be greater are not regulated. Thus, Amaranth was able to shift their
than 100%. In Figure 4C, only Amaranth NYMEX natural gas futures
positions are compared to NYMEX natural gas futures open interest. exposure and actually increase it by using ICE without the
CFTC or any other regulatory body aware of the increasing
34
Here we are speaking about the total losses of $3.296 billion that would risk they were taking. In fact, in an instant message
have resulted had they held their August 31, 2006 positions until September
21, 2006. The actual Amaranth natural gas losses were even higher at $4.071 conversation on April 25, 2006, Brian Hunter wrote about
billion, while the total change in net asset value to the main funds was ICE that “...one thing that’s nice is there are no expiration
$4.942 billion. These discrepancies are discussed in more detail in Section limits like NYMEX clearing.” (Senate Report, p. 98).
II.2.
Percent of NYMEX OI Percent of NYMEX OI Percent of NYMEX OI
Nymex FEQ Contracts
−150000 −100000 −50,000 0
0
50
100
150
200
0
50
100
150
200
0
50
100
150
200
CHINCARINI
OCT.06 OCT.06 OCT.06

NOV.06 NOV.06 NOV.06
DEC.06 DEC.06 DEC.06
OCT.06 JAN.07 JAN.07 JAN.07
FEB.07 FEB.07 FEB.07
MAR.07 MAR.07 MAR.07
APR.07 APR.07 APR.07
MAY.07 MAY.07 MAY.07
SEP.06 JUN.07 JUN.07 JUN.07
August 7
JUL.07 JUL.07 JUL.07
AUG.07 AUG.07 AUG.07
SEP.07 SEP.07 SEP.07
JAN.08 JAN.08 JAN.08
OCT.06 FEB.08 FEB.08 FEB.08
JUN.08 JUN.08 JUN.08
SEP.06 JUL.08 JUL.08 JUL.08
August 8
A CASE STUDY ON RISK MANAGEMENT

OCT.06 MAR.09 MAR.09 MAR.09
NYMEX Position
B
A
C
SEP.06 AUG.09 AUG.09 AUG.09
August 9
Contract Month
Contract Month
Contract Month

OCT.06 APR.10 APR.10 APR.10
SEP.06 SEP.10 SEP.10 SEP.10
August 10
Figure 9. The Amaranth Positions in Response to NYMEX Position Limit Phone Calls
ICE Position
OCT.06 MAY.11 MAY.11 MAY.11
Figure 8. The Amaranth Positions as a Percentage of NYMEX Open Interest (August 31, 2006)

SEP.06 OCT.11 OCT.11 OCT.11
August 11
169
Although NYMEX only uses its position limits as guidelines in the markets would come back so that they could reduce the
of whether or not to investigate an entity’s position, it is size of their winter/summer spread positions at favorable
interesting to note how far above these guidelines Amaranth prices.
was. The NYMEX guideline is to examine entities with an We thought about pulling the trigger and taking the loss.
amount over 12,000 contracts in any given maturity. One can We had many discussions about it. We figured we could get
see from Figure 8 that Amaranth had exceeded this “guideline” out for maybe a billion dollars. But we decided to ride it out
by a substantial amount. Perhaps a quantitative rule would be and see if the market would come around.—Interview with
better than a qualitative rule. With quantitative rules, Amaranth trader. (Senate Report, p. 77)
Amaranth’s positions would never have been able to be so Yet, despite being apparently aware of the liquidity issues
large. with their natural gas positions, they continued to act perilously
The reconstruction of the VaR of Amaranth’s positions on and actually increased the size of their positions from the end
August 31, 2006 was high, but cannot entirely explain of May to the end of August (the leverage of the natural gas
Amaranth’s losses in September 2006 unless one designates positions with respect to the fund increased from 3.83 to 5.23)
the Amaranth collapse as a 5 standard deviation event.35 It perhaps because they ultimately believed that the market was
appears Amaranth’s traders and senior management were well wrong and they were right. In their monthly letter to investors
aware of a VaR number similar to the one produced in this explaining the losses of May, they said “...we believed certain
paper. It seems that they were willing to take this amount of spread relationships remained disconnected from their
risk given the expected return they hoped to achieve. With fundamental value drivers.” (Senate Report, p. 73).
regards to their liquidity risk, while the traders were very aware With respect to management, the senior management in
of the size of their positions, it is not clear that senior Greenwich knew of the market risk but overlooked the position
management in Greenwich really understood the extent of it. size by giving too much credit to Hunter, partly out of their
First, Amaranth’s risk management with regard to liquidity own greed.
did not explicitly specify position limits as a percentage of
volume traded or open interest on the exchanges, so risk of C. Funding Risk
this type may not have been on senior management’s radar in
an explicit way. Second, Amaranth allowed Brian Hunter and Funding risk is related to liquidity risk, but is focused on
his trading team to move to Calgary without any risk leverage in particular. 38 Any leveraged position implies that
management team (FERC Report, p. 18-19). Third, Amaranth the trader borrowed some of the capital to finance his position.
was slowly increasing the size of their natural gas positions Leverage and funding risk are very much interlinked. For
over the summer of 2006. example, if a trader purchases a futures contract but keeps the
It appears that Amaranth’s senior management allowed remainder funds in cash (e.g. $56,250), the trader will never
Hunter too much freedom because they had enjoyed his prior have funding risk, because although the future contract was
success and wanted to believe that he really was purchased on margin, the trader’s fund is not levered. Suppose
“...brilliant...”36 and also independently “...really, really good the trader buys 2 contracts, although he only has capital to
at taking controlled and measured risk.”37 Even this statement cover 1 contract. This trader’s fund now has a leverage of 2
by the CEO reveals problems with their risk management (notional value of position / cash on-hand plus initial margin).
philosophy. It should not be the trader that one is confident Now there is some funding risk, although it is still low. The
about with regards to risk management, but rather the risk trader will be able to meet all margin calls until the position
manager which is monitoring that trader’s risk. falls by more than 50%. Thus, a rule-of-thumb in this simple
In summation, the energy traders of Amaranth were well example is that a trader will face funding risk anytime the
aware of the large size of their positions and either did not return of his levered position falls by more than 1/L, where L
care (i.e. the free option) or did not realize how perilous such is the amount of fund leverage. Of course, this is only true in
a position could be. As far back as May, they seemed to be our simplified example where all excess capital is held in cash.
aware of the large size of their positions. In interviews by the It becomes even more complicated when some of this excess
Senate Subcommittee with Amaranth traders, they stated after capital is invested in other assets.
the losses in May that they were waiting to see if the liquidity In Amaranth’s case, the leverage of natural gas future
equivalents on August 31, 2006 was 5.23 ($53,524,979,537 /
$10,228,192,000) with respect to just their natural gas
35
This would imply a VaR at a 99.99997% confidence level.
36
Unidentified trader’s email to Brian Hunter when he was making money
in July. Source: Senate Report, Exhibit #9. A more detailed version of this section is available on the the JAF website:
38
www.fma.org/jaf.htm. For other descriptions of funding risk, see Culp and

37
Public statement by CEO Maounis about Hunter. Miller (1994), Edwards and Canter (1995), and Mello and Parsons (1995).
exposures. To the extent that they were investing on margin in V. Five Lessons for Regulators and
other markets, their leverage might have been even higher. Hedge Funds
Amaranth had set aside up to $3 billion of their capital in cash
to meet liquidity needs according to Mr. Artie DeRocco in It is difficult to construct lessons after major crises because
conversations with Nymex’s Michael Christ on August 15, often times the specific corrections to certain situations will
2006. To the extent that only $3 billion might have been only cause new crises to occur under different loopholes or
available for margin calls, Amaranth’s leverage could have conditions. Nevertheless, lessons from other crises have been
been considered as high as 17.84 ($53,524,979,537 / useful. For example, after the LTCM crisis of 1998, hedge
$3,000,000,000). That would imply that even a -5.6% return funds learned that making sure lines of credit are really lines
on their futures position would cause them funding problems. of credit is extremely important. Hedge funds also learned
On August 31, 2006 Amaranth’s initial margin on NYMEX that stress testing sophisticated trading systems includes the
exceeded $2.5 billion. This high margin requirement was worst case scenarios, for example when the correlation of
primarily due to the notional size of Amaranth’s position. In seemingly unrelated strategies goes to one.
fact, if we assume that NYMEX required the maximum margin In the aftermath of the Amaranth collapse, there are five
for each NYMEX natural gas equivalent, then Amaranth’s lessons as well.
positions would require $5,306,512,760.39 Even the actual 1. First, liquidity risk is a real risk that must be
margin requirement on that day of $2.5 billion left very little accounted for by both exchanges and hedge funds, money
room for adverse returns for Amaranth. If we observe Figure managers, or traders. For exchanges, it means strict
6 on Amaranth’s daily profit-and-loss from their natural gas ‘concentration limits’ should be placed on a customer’s
positions, one can see that by the close of business on positions. While NYMEX has soft position limits, they
September 7, 2006, their total additional margin required allowed human judgement, conversations with Amaranth, and
would have been $697 million. By September 15, the greed to soften those limits up to a point, where they did not
additional margin required would have been $3.009 billion, really know the severity of the enormous positions of
and by September 21, it would have been $4.433 billion. Amaranth. By ‘concentration limits’, I refer to limits that are
Clearly, this was unsustainable as Amaranth did not have the based upon some percentage of the open interest that would
cash to meet these margin calls. If we assume that on August be dynamic over time rather than static as position limits are.
31, 2006 they had exactly $2.5 billion in initial margin, by The limits might also vary by contract maturity. But not only
September 21, 2006 they would have required around $6.933 should exchanges consider strict concentration limits, they
billion of margin.40 should also consider quantitative rules for managing these
What differentiates this sort of risk from other risk is that limits rather than ad hoc human judgement. For hedge funds,
even if the strategy turned out to be profitable by month-end money managers, and traders, the lesson has long been
(which it did not), Amaranth would not have had enough known—don’t own too much of a market in combination with
funding in place to hold on to their positions until month-end. leverage. If prices move adversely against one’s levered
Thus, even if Amaranth’s trade had been logical from a VaR position, margin calls might require the trader to reverse the
perspective and a liquidity perspective, it would have not been positions to acquire cash to make the margin calls. These
logical or prudent from a funding risk perspective. position reducing trades may make the prices move further
adversely and perhaps cause prices to deteriorate so much
that the investor loses more than his or her capital and goes
bankrupt. If a trader limits the concentration of his position in
39
Of course, this is unrealistically high, because it requires many assumptions. a certain market, it will help insure that in the case he would
The primary assumption is that for every natural gas equivalent held, like to reduce or close his position, there will be a sufficient
NYMEX would require the full non-member initial margin. For spread
positions, consisting of two months of contracts, the initial margin number of other traders to absorb his selling pressure without
requirements are much less per position at $1000 per position. Also some of moving prices too much.
these positions are for option contracts that might not require margin. Thus, 2. Transparency across exchanges in the same market
this number represents an upper limit of the total margin required. Finally,
this also assumes that the initial margin was calculated as if all positions may be useful. In the case of Amaranth, the NYMEX knew
were constructed on that particular day. To actually reconstruct the exact of Amaranth’s NYMEX positions, but did not know of the
margin required by Amaranth on that day is not possible without further other positions held with ICE. Although the CFTC oversees
information that is not available. However, we do know from statements by
NYMEX that on August 31, 2006 the actual margin requirement on that the NYMEX, they had no jurisidiction over ICE, since ICE is
day exceeded $2.5 billion. an unregulated energy trading platform. Were there a system
held by the CFTC that could oversee all positions on energy
Of course, this is only approximate, as some of the natural gas equivalent
40
platforms, the excesses of Amaranth could have been spotted.
positions were options. Also, this would be the total margin on NYMEX
and ICE together. By forcing Amaranth to hold much more reasonable positions,
Amaranth investors would have ultimately been better off. or even riskier than an unleveraged outright position. This is
Also, the possible manipulation by one entity of security prices because the spread positions are not arbitrage positions, they
would be avoided. Amaranth’s selling of large positions may are just less volatile positions. Thus, when evaluating spread
have caused intense volatility in natural gas prices causing trades, one should consider the amount of leverage and its
actual users of natural gas (i.e. households) to pay high prices effect on actual volatility and not naively assume they have
which may have been artificially high due to the excessive lower risk.
positions of Amaranth. In fact, Amaranth and Amaranth traders There are critics of the new proposals for regulation in the
are currently being sued over the matter (FERC (2007)). US natural gas markets. The criticisms fall into four categories.
One of the steps to improve transparency in the U.S. markets First, there is a camp that believes Amaranth’s positions were
is a bill introduced on September 17, 2007 by Senator Carl not too big for the market and that setting strict positions limits
Levin of Michigan to regulate electronic energy trading will compromise “...the efficient transfer of risk in the market
facilities by registering with the CFTC (Levin (2007)). The place,” (Watkins (2007)).Second, some people do not wish
bill also proposes to provide trading limits for energy traders there to be multiple regulatory agencies regulating the natural
that can be monitored by the CFTC across all energy trading gas futures market. Third, some people worry that regulation
platforms and exchanges, and requires that large domestic will cause business to transfer to overseas markets. Finally,
traders of energy report their trades on foreign exchanges. some argue that the regulation will ultimately not work,
The bill defines precisely what constitutes an “energy trading because market participants will find other loopholes (Watkins
facility” and an “energy commodity”. (2007)).
3. More standard measures of liquidity risk ought to be Each of these criticisms will be discussed in more detail
devised so that, as with VaR, traders, risk managers, along with my own thoughts with respect to each of them.
regulators, and exchanges have a language to communicate The first criticism is that position limit constraints will prevent
with each other. “...legitimate speculation...” and thus make markets less
4. There are lessons for internal risk management. It efficient. Also, the critics worry that position limits and laws
might be important to have risk managers in the same location can become outdated. The first comment assumes the
as traders, rather than thousands of miles away. It might also proportion of arbitrageurs is very small in the market place.
help to follow guidelines that many large banks have of To the extent that there are many speculators in natural gas,
allotting only certain risk capital to certain traders and diversify the transfer of risk can still be accomplished — it just will
across the firm, rather than have one trader, like Hunter, use reduce the likelihood that the speculation is in the hands of
the majority of the firm’s capital and be responsible for the just one large speculator. While it is true that laws will become
majority of the firm’s performance. After all, Amaranth was outdated, it doesn’t mean they are not useful in the short-run.
not an energy trading hedge fund, it was a multi-strategy hedge In addition, position limits can be made relative so that they
fund. Along that line of thought, one might even consider a do not become outdated quickly. For example, rather than have
different incentive scheme for risk managers. Risk managers a limited specific number of contracts for each speculator, an
are not paid as well as traders. This causes their voice to be exchange could have that number depend upon some
less important in the firm. And of course, risk managers’ bonus percentage of average daily trading volume or of open interest.
also depends on firm profits. Thus, to a certain extent they Also, regulation could allow exchanges and governing bodies
will also be reluctant to reduce the firm’s aggressive trading to update position limits as market conditions change.
activities. They have a “free option” too. It is not clear that The second criticism is about the number of regulatory
there is a simple way to restructure the incentives of risk bodies in the natural gas markets. Currently, some market
managers, but it might be worth thinking about. participants, including major investment banks like Goldman
5. Spread positions can lose money and are not Sachs, Morgan Stanley, Merrill Lynch, and J.P. Morgan, are
“arbitrage positions”, especially when the size of these opposed to having both the FERC and the CFTC with authority
positions is large. Spread positions are usually thought of as over the commodities markets. Their argument is that too many
less risky than outright positions, since by being long curtain regulatory agencies might raise confusion and costs among
contracts and short other contracts, the position is less exposed market participants. While this is certainly a negative, the
to the directional volatility of the natural gas market. It should Levin proposal does not encourage multiple regulatory bodies.
be stressed that these positions have lower risk, but they do It specifies the CFTC as the only regulatory body. However,
have risk. That is, the returns of these positions do exhibit the reality is that when market participants are perceived to
some volatility, even if this volatility is smaller than outright have acted incorrectly, many affected parties may resort to
positions. If a trader leverages these spread positions, the legal action, as the FERC has done with regards to Amaranth,
volatility increases linearly with the leverage. Thus, for a even if they are not explicitly assigned the role of regulator.
large enough leverage, the spread position can be as risky as The third criticism is that increased regulation will lead
market participants to overseas trading venues, such as reasons why this hedge fund failure attracted such widespread
Singapore. While this is always a possibility, it could be argued media attention. First, the size and speed at which Amaranth
that the increased transparency and minimal standards of the made losses. In less than 14 days, from September 7, 2006 to
US exchanges may draw people to the US exchanges precisely September 21, 2006, they had lost almost $4 billion. Second,
for these reasons. For example, although listing requirements their losses occurred in the natural gas markets. There is some
on the NYSE are more stringent that those of NASDAQ, the evidence that Amaranth’s trading activities in the natural gas
NYSE has not gone out of business despite the rise of the markets distorted market prices and ultimately hurt consumers
NASDAQ. There will, of course, be firms that find it more of natural gas. For instance, the Municipal Gas Authority of
desirable to go elsewhere. Georgia (MGAG) complained that its hedging costs with
The fourth criticism is that regulators “...will always be one abnormally high winter natural gas prices caused its consumers
step behind the innovating and evolving markets” (Watkins, losses of $18 million during the winter of 2006-2007 (Senate
2007). This statement is absolutely true. However, this does Report, p. 115). Third, the failure raised new concerns about
not mean that regulatory constraints in cases where market risk management and leverage. In particular, it raised questions
failures or externalities exist are not appropriate. The correct about how large a position and influence an individual entity
question is whether or not externalities and market failures should have over a financial market, like the natural gas futures
potentially exist in the market for natural gas. In the end, we market.
must answer this crucial question before deciding whether This paper dealt specifically with examining the actual
regulation is a good or a bad thing. positions of Amaranth in the natural gas market to understand
Without regulation, Amaranth was able to acquire whether conventional risk measurement tools could have
enormously large positions on NYMEX and ICE that may estimated the large risks that caused their collapse in
have led to a distortion of natural gas prices which ultimately September 2006. The paper finds that Amaranth’s VaR on
affected consumers of natural gas.41 However, even though August 31, 2006 was $1.334 billion and $2.048 billion at the
Amaranth’s positions on NYMEX were regulated by the 99% and 99.95% confidence level. Although large, these
CFTC, they still were extremely large. So it is not clear that numbers of rather low probability events still underestimate
the regulation per se will solve the problem. The position limits their actual losses in natural gas of $4.433 billion and decrease
on NYMEX were very loosely enforced and subject to in their net asset value of $4.942 billion. In fact, the paper
interpretation by NYMEX officials. It was only at late stages finds that it was the management of their liquidity risk that
of the Amaranth debacle that Amaranth moved substantial was vastly irresponsible. Amaranth’s NYMEX natural gas
positions from NYMEX to the unregulated ICE. futures equivalent positions in certain maturity contracts
Although this paper was not primarily concerned with exceeded 200% of the NYMEX natural gas open interest.
Amaranth’s effect on natural gas futures prices, a preliminary Their ownership of NYMEX natural gas futures contracts
investigation was done using data on daily natural gas returns alone was, in certain maturities, close to 100% of the open
and trades by Amaranth. Some evidence was found that interest. When markets turn against a trader’s position, futures
contracts which Amaranth sold led to lower returns than other exchanges will require additional margin to maintain those
contracts in which Amaranth was not trading. For a more positions. Once the trader’s cash on-hand and borrowing
detailed discussion of this analysis, see the supplemental sources are exhausted (funding risk), he can only meet margin
section entitled The Price Impact of Amaranth’s Trading on calls by selling the underlying assets. If that trader owns a
the JAF website. large percentage of that market, he can only sell those assets
by forcing the prices even lower and thus creating further losses
VI. Conclusion and further margin calls. This is known as liquidity risk. A
combination of liquidity and funding risk ultimately caused
The collapse of the hedge fund Amaranth Advisors in Amaranth’s collapse.
September of 2006 drew a flurry of attention. There are several There are several lessons from the Amaranth debacle that
have to be relearned. First, even if a strategy has a positive
excess return with low volatility historically, with or without
a theoretical justification for the strategy, that strategy can
41
A distinction should be made between manipulation of natural gas prices
and impacting natural gas prices due to the large size of a trade. The former still have negative returns in the future. With leverage, these
is illegal according to Sections 6(c), 6(d), and 9(a)(2) of the Act of the CFTC negative returns are amplified. Second, firms need to manage
which authorizes the CFTC to bring enforcement actions against any person liquidity risk explicitly. The inability to sell a futures contract
who is manipulating or attempting to manipulate or has manipulated or
attempted to manipulate the market prices of any commodity in interstate at or near the latest quoted price can be related to one’s
commerce or for future delivery on or subject to the rules of any registered concentration in the security. In Amaranth’s case, the
entity. Both price manipulation and price impact are valid concerns for
regulators, but one is illegal.
concentration was far too high and there were no natural
counterparties when they needed to unwind the positions. income distributional effect. It can also lead to larger
Third, exchanges can only adequately manage their position uncertainties and less effective decision making by individuals.
limits if they have disciplined rules for doing so and Amaranth is currently being sued by the Federal Energy
transparency. Currently, a bill has been introduced by Senator Regulatory Commission (FERC) for price manipulation in
Carl Levin to address the second point and regulate energy specific instances. Their intent is to penalize Amaranth for
trading facilities (Levin, 2007). The importance of limiting unjust profits and civil penalities, in addition to seeking $30
concentration comes also through the potential for price million from Brian Hunter as well (FERC, 2007).
manipulation which can distort prices and have an unfair
References
Chincarini, Ludwig. B., 2006, “The Amaranth Debacle: Failure of Risk Favre, Laurent. and J.A. Galeano, 2002, “Mean-Modified Value-at-Risk
Measures or Failure of Risk Management?” Presentation at Georgetown Optimization with Hedge Funds,” Journal of Alternative Investments
McDonough School of Business, http://faculty.msb.edu/lbc22/ 5(Fall, 1), 2-21.
video_clip1.html, (December 5).
Hull, John, 2006, Options, Futures, and Other Derivatives, New York,
Chincarini, Ludwig. B., 2007a, “The Amaranth Debacle: Failure of Risk Prentice Hall.
Measures or Failure of Risk Management?” Available at SSRN: http://
ssrn.com/abstract=952607, (April). Jorion, Phillippe, 2006, Value at Risk, New York, McGraw-Hill.
Chincarini, Ludwig. B., 2007b, “The Amaranth Debacle: Failure of Risk Levin, Carl, 2007, “To Amend the Commodity Exchange Act to Close the
Measures or Failure of Risk Management?” Journal of Alternative Enron Loophole, Prevent Price Manipulation Excessive Speculation in
Investments, (Winter). the Trading of Energy Commodities and for Other Purposes,” Bill S.
2058, (September 17).
Cornish, E.A., and R.A. Fisher, 1937, “Moments and Cumulants in the
Specification of Distributions,” Review of the International Statistical Maounis, Nick, 2006, “Investor Conference Call,” Amaranth Group, Inc.,
Institute, 307-320. (September 22).
Culp, Christopher, and M. Miller, 1994, “Hedging a Flow of Commodity Mello, Antonio S., and J.E. Parsons, 1995, “Maturity Structure of a Hedge
Deliveries with Futures: Lessons from Metallgesellshaft,” Derivatives Matters: Lessons from the Metallgesellschaft Debacle,” Journal of Applied
Quarterly 1(Fall, 1), 1. Corporate Finance 8(Spring, 1), 106-121.
Davis, Anne, 2006, “How Giant Bets on Natural Gas Sank Brash Hedge- Ord, Keith and A. Stuart, 1994, Kendall’s Advanced Theory of Statistics:
Fund Trader,” Wall Street Journal, (September 19), A1. Volume 1, Arnold, London and Wiley.
Dowd, Kevin, 1999, Beyond Value at Risk: The New Science of Risk Till, Hilary, 2006, “EDHEC Comments on the Amaranth Case: Early Lessons
Management, New York, John Wiley & Sons. from the Debacle,” EDHEC Risk and Management Research Centre.
Edwards, Franklin, and M.S. Canter, 1995, “The Collapse of United States of America, Federal Energy Regulatory Commission, 2007,
Metallgesellschaft: Unhedgeable Risks, Poor Hedging Strategy, or Just “Order to Show Cause and Notice of Proposed Penalties,” Docket No.
Bad Luck?” Journal of Futures Markets, 15(May, 3), 211-264. IN07-26-000, (July 27).
“Excessive Speculation in the Natural Gas Market,” 2007, United States Watkins, David, 2007, “In the Shadow of Amaranth,” Energy Risk, Special
Senate Permanent Subcommittee on Investigations Staff Report, (June Report on Natural Gas (December), 28-31.
25).
“Excessive Speculation in the Natural Gas Market Appendix,” 2007, United

States Senate Permanent Subcommittee on Investigations Staff Report,
(June 25).

Caso 2 Risk Management

Uploaded by

Copyright:

Available Formats

Caso 2 Risk Management

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Caso 2 Risk Management

Uploaded by

Copyright:

Available Formats

152 JOURNAL OF APPLIED FINANCE SPRING/SUMMER 2008

A Case Study on Risk Management:

August return = 6.98% JP Morgan meets

NAV = $10.228 billion report on

Figure 2. Historical Average Returns of Natural Gas by Contract in September (1990-2005).

Contract NYMEX Contracts ICE Contracts Total

Figure 5. Amaranth NYMEX FEQ Positions on August 31, 2006

Percent of NYM EX Dollar P/L

Total Dollar Value Correct Sign (%)

The actual calculation of the Cornish-Fisher VaR is contained in the detailed

OCT.06 OCT.06 OCT.06

OCT.08 OCT.08 OCT.08

OCT.09 OCT.09 OCT.09

AUG.11 AUG.11 AUG.11

www.fma.org/jaf.htm. For other descriptions of funding risk, see Culp and

“Excessive Speculation in the Natural Gas Market Appendix,” 2007, United

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.