Example 3 Timing of end of day prices
A London-based firm is trading in Europe, Asia and the Americas in exchange traded
futures. For end of day price testing the organization must determine if they will use
closing price for each of the exchanges or take the price of each of the exchanges at
a certain point in time. Some exchanges close to generate an ‘official exchange closing
price’ and then reopen for late hours changing (i.e. LIFFE and DTB). In this global
environment and end of day batch processing cycles standard pricing policies are
required.
The risk practitioner chose to implement the price of all of the exchanges at a
specific point in time for financial reporting purposes. This was weighed with the
trade-off of not having the official close in the Americas, after 9 p.m. in London. The
procedure was documented and agreed with front office, systems Financial and
Operations teams. At a granular implementation level, this means specifying .LAST
on the market data screen rather than .CLOSE (that would provide the previous
days’ close in the Americas). 567
24 Temmuz 2011 Pazar
Decisions of sourcing and timing of market data
A key decision should be agreed and documented regarding the approach to ensuring
the quality of market data and the time frames that will be used to obtain it. It is
common to see that several booking locations contain the same instruments, particularly
hedge instruments such as futures. Depending how the firm is structured, it is
not uncommon that different locations use different sources of market data and
capturing market data at different points in time for end of day processing and price
testing. A decision should be made to define organizationally the boundaries of
acceptable practice. For larger institutions, in some instances, it may be agreed that
the sourcing of market data may come from one desk, such as a market-making
desk, and all other desks should use their marks. In other cases it may be deemed
acceptable that different desks have slightly different prices depending on the time
of capture and the trading strategy (i.e. arbitrage desk). In the case of sharing data,
a certain amount of time is required for the coordination of control teams, particularly
in large institutions.
the quality of market data and the time frames that will be used to obtain it. It is
common to see that several booking locations contain the same instruments, particularly
hedge instruments such as futures. Depending how the firm is structured, it is
not uncommon that different locations use different sources of market data and
capturing market data at different points in time for end of day processing and price
testing. A decision should be made to define organizationally the boundaries of
acceptable practice. For larger institutions, in some instances, it may be agreed that
the sourcing of market data may come from one desk, such as a market-making
desk, and all other desks should use their marks. In other cases it may be deemed
acceptable that different desks have slightly different prices depending on the time
of capture and the trading strategy (i.e. arbitrage desk). In the case of sharing data,
a certain amount of time is required for the coordination of control teams, particularly
in large institutions.
Price testing of illiquid positions via broker quotes
Example 2 Price testing of illiquid positions via broker quotes
Prices for illiquid bonds are not available by way of market data providers and calling
brokers for quotes is the only method of obtaining prices. Traders indicate that they
are becoming concerned with the frequency of price testing and are beginning to
complain to management that they are being overly controlled. When sitting down
with the traders it becomes apparent that the traders are concerned with the market
knowing the less liquid positions just prior to accumulating or selling a position. The
risk practitioner agrees with the traders a policy that if there are sensitive positions
that are not to be included in price testing they obtain approval from the head of
trading. This note is included in the price testing report and is to be discussed at the
risk committee meeting.
Prices for illiquid bonds are not available by way of market data providers and calling
brokers for quotes is the only method of obtaining prices. Traders indicate that they
are becoming concerned with the frequency of price testing and are beginning to
complain to management that they are being overly controlled. When sitting down
with the traders it becomes apparent that the traders are concerned with the market
knowing the less liquid positions just prior to accumulating or selling a position. The
risk practitioner agrees with the traders a policy that if there are sensitive positions
that are not to be included in price testing they obtain approval from the head of
trading. This note is included in the price testing report and is to be discussed at the
risk committee meeting.
Contacting brokers
Calling brokers for prices is extremely helpful for developing junior staff and allows
them to feel much more confident with specifying market parameters such as
volatility, bid offer spreads, etc. This confidence is part of the overall on the job
training regime and helps overall confidence of the markets and products. Note that
once the rapport is developed with target brokers over the phone, this method can
be more efficiently implemented, first by email and then by fax.
Due to the nature of the broker network it is important to consider rotating brokers
and to average several quotes is best practice. However, the practical nature and
justification must be considered. When calling a broker for a quote, the broker will
ask which firm you are from. When specifying the firm there are two risks. The first
is that you are giving the position of the firm to the market. This is an issue in less
liquid markets. For volatilities it can slightly disguised by asking for a range of
strikes. The second risk is that the broker has a relationship with your firm. Due to
the nature of the market, the price that comes back can be informally from the
brokers of the firm. More brokers called can reduce this risk at the trade off cost of
giving away the position to more players in the market.
them to feel much more confident with specifying market parameters such as
volatility, bid offer spreads, etc. This confidence is part of the overall on the job
training regime and helps overall confidence of the markets and products. Note that
once the rapport is developed with target brokers over the phone, this method can
be more efficiently implemented, first by email and then by fax.
Due to the nature of the broker network it is important to consider rotating brokers
and to average several quotes is best practice. However, the practical nature and
justification must be considered. When calling a broker for a quote, the broker will
ask which firm you are from. When specifying the firm there are two risks. The first
is that you are giving the position of the firm to the market. This is an issue in less
liquid markets. For volatilities it can slightly disguised by asking for a range of
strikes. The second risk is that the broker has a relationship with your firm. Due to
the nature of the market, the price that comes back can be informally from the
brokers of the firm. More brokers called can reduce this risk at the trade off cost of
giving away the position to more players in the market.
Note on market data solutions
Note on market data solutions: When implementing electronic market data solutions
it is always necessary to consider the quality of data with regard to the contributors
of the data and the time of day when the information will be updated and closed.
Spending time upfront in the design of automating market data and corresponding
controls and understanding the nuances of market data is a very good long-term
investment.
it is always necessary to consider the quality of data with regard to the contributors
of the data and the time of day when the information will be updated and closed.
Spending time upfront in the design of automating market data and corresponding
controls and understanding the nuances of market data is a very good long-term
investment.
Methods of obtaining market data for price testing
The cost and speed of obtaining quality prices to make a business statement on the
controls and accuracy is the business of running the control function. The goal is
obtain the maximum amount of quality market data in the shortest amount of time.
This can be obtained when the control team is focused on analysis rather than data
gathering, which can be time consuming, frustrating and disruptive. The disruptive
nature and the actual amount of time to implement some of the manual processes
is frequently significantly underestimated in terms of costs and speed of delivery.
The sources of market data are placed into the following categories:
Ω Automated market data extract – prices are available by market data providers or
broker pages in a form that can be downloaded directly into a spreadsheet or
database. Automated market data extract with parsing – prices are available by
market data providers or broker pages but need to be in a test string data format.
The data needs to be ‘parsed’ to obtain prices in a numeric comparable state.
Ω Manual broker pages – automated data available entered manually in a spreadsheet
or in work papers
Ω Newspaper – price data available in various financial papers. Note that the risk
controller should make sure that the firm is not the market data provider to the
newspaper, otherwise you will be testing like to like.
Ω Calling brokers – to obtain market data, sometimes required for OTC derivatives
or illiquid bonds.
controls and accuracy is the business of running the control function. The goal is
obtain the maximum amount of quality market data in the shortest amount of time.
This can be obtained when the control team is focused on analysis rather than data
gathering, which can be time consuming, frustrating and disruptive. The disruptive
nature and the actual amount of time to implement some of the manual processes
is frequently significantly underestimated in terms of costs and speed of delivery.
The sources of market data are placed into the following categories:
Ω Automated market data extract – prices are available by market data providers or
broker pages in a form that can be downloaded directly into a spreadsheet or
database. Automated market data extract with parsing – prices are available by
market data providers or broker pages but need to be in a test string data format.
The data needs to be ‘parsed’ to obtain prices in a numeric comparable state.
Ω Manual broker pages – automated data available entered manually in a spreadsheet
or in work papers
Ω Newspaper – price data available in various financial papers. Note that the risk
controller should make sure that the firm is not the market data provider to the
newspaper, otherwise you will be testing like to like.
Ω Calling brokers – to obtain market data, sometimes required for OTC derivatives
or illiquid bonds.
Knowing your portfolios
When determining the price testing strategy it is important to understand the
composition of the portfolios in terms of materiality, concentration and liquidity. The
key task is to validate the material and risk positions of the firm, and the practitioner
should have a sense of both the portfolio compositions and the availability of
information available. In addition to the material risks, the practitioner will consider
the sensitive issues and changes in specific positions. This understanding along with
the changes in the markets allow the practitioner to position the discretionary
elements of price testing.
The test coverage should be designed to feature the most material elements of the
portfolio and certain control elements. For example, the information on unusual
volatility in a market along with the sensitivities of the firm’s positions may be
included in the price testing report. This information should be obtained during the
price testing process. Targeting the featured area can be included with little additional
effort if the strategy is defined at the beginning of the price testing cycle. Users of the
price testing reports can be asked for their thoughts on featured areas.
composition of the portfolios in terms of materiality, concentration and liquidity. The
key task is to validate the material and risk positions of the firm, and the practitioner
should have a sense of both the portfolio compositions and the availability of
information available. In addition to the material risks, the practitioner will consider
the sensitive issues and changes in specific positions. This understanding along with
the changes in the markets allow the practitioner to position the discretionary
elements of price testing.
The test coverage should be designed to feature the most material elements of the
portfolio and certain control elements. For example, the information on unusual
volatility in a market along with the sensitivities of the firm’s positions may be
included in the price testing report. This information should be obtained during the
price testing process. Targeting the featured area can be included with little additional
effort if the strategy is defined at the beginning of the price testing cycle. Users of the
price testing reports can be asked for their thoughts on featured areas.
Level of policy in the organization
Regardless of the size of the organization, it is recommended that it develop and
implements a written price testing policy. A policy, if written at a high enough
level, is flexible for local challenges and eliminates the need for multiple policy
developments. This is more challenging the larger and the more global an organization
becomes.
The purpose is to impose consistency across trading desks and products that
have separate challenges. The policies will foster consistency to the approach and
considerations made while developing procedures and the level of documentation
maintained.
In large organizations that are typically product aligned, such a policy provides
consistency across product lines and locations. This supports both business line
and regional reporting requirements. In smaller organizations, or those with fewer
diverse product offerings, the policy should be able to address products without
additional revision.
implements a written price testing policy. A policy, if written at a high enough
level, is flexible for local challenges and eliminates the need for multiple policy
developments. This is more challenging the larger and the more global an organization
becomes.
The purpose is to impose consistency across trading desks and products that
have separate challenges. The policies will foster consistency to the approach and
considerations made while developing procedures and the level of documentation
maintained.
In large organizations that are typically product aligned, such a policy provides
consistency across product lines and locations. This supports both business line
and regional reporting requirements. In smaller organizations, or those with fewer
diverse product offerings, the policy should be able to address products without
additional revision.
18 Temmuz 2011 Pazartesi
Example 1 Price test the banking book
The banking book has not been looked at in depth and turnover of trading personnel
raised the issue of unknown trading losses lurking in the banking book. Price testing
of the positions forced the risk practitioner to review the positions and relevant
hedges in the books. In developing techniques to assess the less liquid positions, the
risk practitioner became more familiar with the number and type of transactions on
the banking book and identified liquid positions that were in the banking report as
hedges.
When reviewing the valuation between marked to market and accrual valuation,
the practitioner is able to assure management that there are no unexpected surprises
hidden in the banking books and proper accounting treatment is being followed.
Level of policy in the organization
Regardless of the size of the organization, it is recommended that it develop and
implements a written price testing policy. A policy, if written at a high enough
level, is flexible for local challenges and eliminates the need for multiple policy
developments. This is more challenging the larger and the more global an organization
becomes.
The purpose is to impose consistency across trading desks and products that
have separate challenges. The policies will foster consistency to the approach and
considerations made while developing procedures and the level of documentation
maintained.
In large organizations that are typically product aligned, such a policy provides
consistency across product lines and locations. This supports both business line
and regional reporting requirements. In smaller organizations, or those with fewer
diverse product offerings, the policy should be able to address products without
additional revision. 563
raised the issue of unknown trading losses lurking in the banking book. Price testing
of the positions forced the risk practitioner to review the positions and relevant
hedges in the books. In developing techniques to assess the less liquid positions, the
risk practitioner became more familiar with the number and type of transactions on
the banking book and identified liquid positions that were in the banking report as
hedges.
When reviewing the valuation between marked to market and accrual valuation,
the practitioner is able to assure management that there are no unexpected surprises
hidden in the banking books and proper accounting treatment is being followed.
Level of policy in the organization
Regardless of the size of the organization, it is recommended that it develop and
implements a written price testing policy. A policy, if written at a high enough
level, is flexible for local challenges and eliminates the need for multiple policy
developments. This is more challenging the larger and the more global an organization
becomes.
The purpose is to impose consistency across trading desks and products that
have separate challenges. The policies will foster consistency to the approach and
considerations made while developing procedures and the level of documentation
maintained.
In large organizations that are typically product aligned, such a policy provides
consistency across product lines and locations. This supports both business line
and regional reporting requirements. In smaller organizations, or those with fewer
diverse product offerings, the policy should be able to address products without
additional revision. 563
Price testing target priorities and options – trading and banking books
The trading assets and liabilities of a firm are maintained in a book that is defined
as a trading book or as a banking book. The distinction between the two is quite
important for regulatory reporting requirements and accounting treatment. The
trading books should be marked to market allowing regulatory capital calculations
to be based more on market risks. The banking book is based on accrual accounting
and capital calculations are based more on counterparty credit risk.
Ω Trading books contain short-term assets, customer and bank trades and the
securities portfolio for trading, hedging and resale. The trading book typically
contains marketable assets that are required to be marked to market or marked
to model. The key risks in the trading book are market events and are addressed
by the Market Risk Amendment. Profits from the assets maintained in the trading
book are expected to be generated from the difference between the buying and
selling prices. Regulatory capital is based mostly on market factors.
Ω Banking books generally include the deposits, loans and the investment portfolio.
These assets are usually considered to be held to term and the main component
of risk is due to default or credit risk for the counterparty. They typically contain
mortgages, personal loans and a portfolio of proprietary securities in stocks and
bonds. As these assets are not expected to be sold in the short term, it is thought
that the PnL derived from the marked to market or marked to model of these
assets does not present a clear picture of the actual PnL and risks reflected by
the strategies. These assets are typically maintained using accrual accounting
practices and are not marked to market or marked to model. Regulatory capital
is based on default risk, that is, counterparty credit risk.
Price testing the banking books may be viewed as a ‘nice to have’ compared to the
absolute requirements of the trading books. As the banking book is not marked to
market, and positions may not be actively managed for changes in market conditions,
then one may question the purpose of price testing these books, which is for control
and management reporting. Price testing the banking book makes sure that there
are no large landmine losses hidden in the accrual-based banking book. Management
should be aware of the value of positions that are maintained on these books that
have significantly changed since they were booked. This can be due to market
movements or to changes in counterparty credit profiles. Additional challenges in
the banking book arise as it is the place where illiquid positions are booked and it
may be extremely difficult to obtain valid prices for these. However, quantifying and
reporting the number of these positions does have added value in an organization
and lends itself to best practices.
as a trading book or as a banking book. The distinction between the two is quite
important for regulatory reporting requirements and accounting treatment. The
trading books should be marked to market allowing regulatory capital calculations
to be based more on market risks. The banking book is based on accrual accounting
and capital calculations are based more on counterparty credit risk.
Ω Trading books contain short-term assets, customer and bank trades and the
securities portfolio for trading, hedging and resale. The trading book typically
contains marketable assets that are required to be marked to market or marked
to model. The key risks in the trading book are market events and are addressed
by the Market Risk Amendment. Profits from the assets maintained in the trading
book are expected to be generated from the difference between the buying and
selling prices. Regulatory capital is based mostly on market factors.
Ω Banking books generally include the deposits, loans and the investment portfolio.
These assets are usually considered to be held to term and the main component
of risk is due to default or credit risk for the counterparty. They typically contain
mortgages, personal loans and a portfolio of proprietary securities in stocks and
bonds. As these assets are not expected to be sold in the short term, it is thought
that the PnL derived from the marked to market or marked to model of these
assets does not present a clear picture of the actual PnL and risks reflected by
the strategies. These assets are typically maintained using accrual accounting
practices and are not marked to market or marked to model. Regulatory capital
is based on default risk, that is, counterparty credit risk.
Price testing the banking books may be viewed as a ‘nice to have’ compared to the
absolute requirements of the trading books. As the banking book is not marked to
market, and positions may not be actively managed for changes in market conditions,
then one may question the purpose of price testing these books, which is for control
and management reporting. Price testing the banking book makes sure that there
are no large landmine losses hidden in the accrual-based banking book. Management
should be aware of the value of positions that are maintained on these books that
have significantly changed since they were booked. This can be due to market
movements or to changes in counterparty credit profiles. Additional challenges in
the banking book arise as it is the place where illiquid positions are booked and it
may be extremely difficult to obtain valid prices for these. However, quantifying and
reporting the number of these positions does have added value in an organization
and lends itself to best practices.
Timing of data capture
The timing of when prices will be obtained for price testing needs to be specified by
the organization. The question the organization needs to address is to define the
timing and market data sources that are to be used. One consideration is to determine
if it is necessary to have the same end of day price available on a global basis for
price testing and end of day marking (i.e. all prices as of 4.15 p.m. London time).
Timing is more of an issue in larger institutions when the same instrument may be
booked at several locations.
Trading personnel generally prefer the end of day marks from each of the exchange
close as it provides better risk numbers at the end of day for portfolio hedging. It also
allows the Flash PnL and accounting (Tò1) PnL to be more accurately validated.
However, obtaining all prices ‘as of’ a certain point in time is preferred from a
financial viewpoint. It is usually preferred from a system implementation viewpoint
as well as it is generally easier to implement prices as of one point in time for simpler
logic and exception reporting and tighter batch processing frames. Usually practical
implementation issues decide the decision.
Regardless of the timing chosen, the organization should define and document the
timing. Audit personnel generally do not feel comfortable with different prices being
used for end price testing and end of day marks without it being thought through,
communicated to management and documented.
the organization. The question the organization needs to address is to define the
timing and market data sources that are to be used. One consideration is to determine
if it is necessary to have the same end of day price available on a global basis for
price testing and end of day marking (i.e. all prices as of 4.15 p.m. London time).
Timing is more of an issue in larger institutions when the same instrument may be
booked at several locations.
Trading personnel generally prefer the end of day marks from each of the exchange
close as it provides better risk numbers at the end of day for portfolio hedging. It also
allows the Flash PnL and accounting (Tò1) PnL to be more accurately validated.
However, obtaining all prices ‘as of’ a certain point in time is preferred from a
financial viewpoint. It is usually preferred from a system implementation viewpoint
as well as it is generally easier to implement prices as of one point in time for simpler
logic and exception reporting and tighter batch processing frames. Usually practical
implementation issues decide the decision.
Regardless of the timing chosen, the organization should define and document the
timing. Audit personnel generally do not feel comfortable with different prices being
used for end price testing and end of day marks without it being thought through,
communicated to management and documented.
Timing of testing
The purpose of price testing is to validate the valuation of positions and risks at a
specific point in time. The challenge is to perform the price testing function frequently
enough to provide reasonable assurance that positions are materially marked. The
price testing priority is clearly the trading book, rather than the banking book, as it
is required to be marked to market.
A generalized best practice that seems to be developing for market-making institutions
is to price test the trading books at least twice a month. The benefits and
expense of performing this function needs to be balanced with the risks and nature
of the positions.
Ω Once at month end to validate the financial statements.
Ω The second price testing is usually performed on a variable date basis some time
in the middle of the month. It provides an added control function as the traders
cannot predict the date of testing. The mid-month test also defines that the
maximum length of the misstatement of financial statements and risks that can
occur to about two weeks before being picked up. One of the nice attributes of the
variable second price testing is that it provides the manager of this control process
with flexibility to schedule it as best suits the workload of the team (i.e. projects,
system implementations, vacation schedules).
It is important to understand who are the users of the PnL and position information
within the firm. The users of the information are beyond financial reporting. As can
be seen in Figure 19.2, the users cover regulatory reporting, risk management, front
office and the senior management who are required to understand both the PnL and
risks of a firm on a daily basis. Much of the downstream processing (i.e. PnL and
risk valuations) is dependent upon the accuracy of the priced positions. The timing
of the tests and delivery of the reported results should be designed with consideration
of the downstream users and routine meetings where the information will be a
discussion topic.
specific point in time. The challenge is to perform the price testing function frequently
enough to provide reasonable assurance that positions are materially marked. The
price testing priority is clearly the trading book, rather than the banking book, as it
is required to be marked to market.
A generalized best practice that seems to be developing for market-making institutions
is to price test the trading books at least twice a month. The benefits and
expense of performing this function needs to be balanced with the risks and nature
of the positions.
Ω Once at month end to validate the financial statements.
Ω The second price testing is usually performed on a variable date basis some time
in the middle of the month. It provides an added control function as the traders
cannot predict the date of testing. The mid-month test also defines that the
maximum length of the misstatement of financial statements and risks that can
occur to about two weeks before being picked up. One of the nice attributes of the
variable second price testing is that it provides the manager of this control process
with flexibility to schedule it as best suits the workload of the team (i.e. projects,
system implementations, vacation schedules).
It is important to understand who are the users of the PnL and position information
within the firm. The users of the information are beyond financial reporting. As can
be seen in Figure 19.2, the users cover regulatory reporting, risk management, front
office and the senior management who are required to understand both the PnL and
risks of a firm on a daily basis. Much of the downstream processing (i.e. PnL and
risk valuations) is dependent upon the accuracy of the priced positions. The timing
of the tests and delivery of the reported results should be designed with consideration
of the downstream users and routine meetings where the information will be a
discussion topic.
Objectives
It is the responsibility of the risk control functions within a firm to ensure that
positions are prudently marked to fair valuation on the books and records and that
the risks are properly calculated and reported. Both the calculation of the PnL and
the risks for the positions of the firm are typically generated on the same end of day
positions.
1 The fair valuation of products is the easier of the two to digest and see the impact.
The valuations of the positions are marked to the prudent market value that is
represented both in the financial statements and in the management reporting
process. This is a fairly straightforward concept to visualize when considering
liquid exchange traded products. Pick up the newspaper, see the valuation and
make sure the end of day position is marked to it. The challenges of implementing
this are not difficult to visualize. However, the concept becomes challenging when
one considers simultaneously testing a series of global portfolios composed of
liquid and illiquid exchange traded and OTC products.
2 The fair valuation risks related to positions are represented in terms of sensitivities
(‘Greeks’). These risk sensitivity positions are based on system-generated calculations
that are less transparent. To the end user these calculations are generally
less intuitive and less transparent. These calculations are sometimes assumed to
be black box processes that are not easily understood. With a bit of effort, this
presumption is not quite true.
Defining objectives, measuring performance and determining what is need to
achieve objectives are all part of managing a process. The price testing objective sets
out the framework in which to operate and to judge performance. The objective
should include:
Ω Timing of positions to be tested
Ω Minimum frequency
Ω Goals on coverage
Ω Timeliness of reporting
Ω Self-assessment of price testing performance
An example of a price testing objective
We have established an objective of performing price testing the portfolios to end of
day marks at least twice a month, at the each month end and around the middle of
the month. Our goal is to test 100 per cent of the population for month end with
vigour and have the results summarized and reported within 5 business day of the
test date. Self-assessment will identify, document and address issues.
positions are prudently marked to fair valuation on the books and records and that
the risks are properly calculated and reported. Both the calculation of the PnL and
the risks for the positions of the firm are typically generated on the same end of day
positions.
1 The fair valuation of products is the easier of the two to digest and see the impact.
The valuations of the positions are marked to the prudent market value that is
represented both in the financial statements and in the management reporting
process. This is a fairly straightforward concept to visualize when considering
liquid exchange traded products. Pick up the newspaper, see the valuation and
make sure the end of day position is marked to it. The challenges of implementing
this are not difficult to visualize. However, the concept becomes challenging when
one considers simultaneously testing a series of global portfolios composed of
liquid and illiquid exchange traded and OTC products.
2 The fair valuation risks related to positions are represented in terms of sensitivities
(‘Greeks’). These risk sensitivity positions are based on system-generated calculations
that are less transparent. To the end user these calculations are generally
less intuitive and less transparent. These calculations are sometimes assumed to
be black box processes that are not easily understood. With a bit of effort, this
presumption is not quite true.
Defining objectives, measuring performance and determining what is need to
achieve objectives are all part of managing a process. The price testing objective sets
out the framework in which to operate and to judge performance. The objective
should include:
Ω Timing of positions to be tested
Ω Minimum frequency
Ω Goals on coverage
Ω Timeliness of reporting
Ω Self-assessment of price testing performance
An example of a price testing objective
We have established an objective of performing price testing the portfolios to end of
day marks at least twice a month, at the each month end and around the middle of
the month. Our goal is to test 100 per cent of the population for month end with
vigour and have the results summarized and reported within 5 business day of the
test date. Self-assessment will identify, document and address issues.
Implementation of price testing
The purpose of this chapter is to provide the risk practitioner with an overview of the
key goals and challenges of designing and implementing a price testing strategy for
an organization. It identifies the users of price testing reporting from both a minimum
required control and added value perspectives.
Price testing and provisioning are sometimes spoken of in the same context and
sometimes viewed as separate exercises. In some organizations these responsibilities
are sometimes performed by separate functions at different points in time or upon
separate cycles. Regardless of the organizational structure and timing, the chapter
is written to cover the key concepts of price testing and to be fungible across
organizational structures.
An overview of the design of the document is:
Ω Objectives and defining the control framework – setting the game plan for the
location or organization that lead to developing the strategy, implementation and
performance benchmarks
Ω Implementing the strategy – decision time is required when the theoretical
approach of how price testing is to be performed meets with the reality of
actually performing the task with incomplete data and balancing this with other
responsibilities.
Ω Managing the price testing process – challenges in getting the most coverage with
the best sources of data within the agreed deliverable time
Ω Reporting – with price testing, don’t drop the ball before the goal line. Reporting
for basic control purposes is a requirement. With a little more effort and identifying
target audiences, added value can be obtained to support decisions on trading
strategies and booking.
key goals and challenges of designing and implementing a price testing strategy for
an organization. It identifies the users of price testing reporting from both a minimum
required control and added value perspectives.
Price testing and provisioning are sometimes spoken of in the same context and
sometimes viewed as separate exercises. In some organizations these responsibilities
are sometimes performed by separate functions at different points in time or upon
separate cycles. Regardless of the organizational structure and timing, the chapter
is written to cover the key concepts of price testing and to be fungible across
organizational structures.
An overview of the design of the document is:
Ω Objectives and defining the control framework – setting the game plan for the
location or organization that lead to developing the strategy, implementation and
performance benchmarks
Ω Implementing the strategy – decision time is required when the theoretical
approach of how price testing is to be performed meets with the reality of
actually performing the task with incomplete data and balancing this with other
responsibilities.
Ω Managing the price testing process – challenges in getting the most coverage with
the best sources of data within the agreed deliverable time
Ω Reporting – with price testing, don’t drop the ball before the goal line. Reporting
for basic control purposes is a requirement. With a little more effort and identifying
target audiences, added value can be obtained to support decisions on trading
strategies and booking.
The overall control framework
We have not discussed the most important piece a risk manager needs – the overall
risk framework. With a highly volatile commodity like power this needs to be multilayered,
with a strong linkage down to a variety of control points.
Figure 18.10 sets out some of the most important quantifiable limits that can be
set and how they relate to each other. This control structure links many parts of the
organization. As well as monitoring for trade input error, significant emphasis needs
to be placed on verifying counterparty and broker confirmations to the trade capture
systems. Conformations can often be slow and inconsistent depending on the
counterparty, which combined with the potential volatility in the market makes for
a dangerous mix. We have seen players lose several million dollars based on trading
input errors that never got picked up by the confirmation process.
Summary
Energy markets are probably the most challenging commodity markets a risk manager
will face in terms of level of complexity, volatility and quantification. Many of
the issues raised and approaches outlined in this chapter are a reflection of the
developing nature of the markets. In addition, some fundamental risk management
issues still need to be adequately addressed.
As the developing energy markets continue to evolve, the tools used by risk
managers to measure, monitor and control market, credit and operational risk will
also continue to develop. These developments will impact energy markets on a
worldwide basis. I suspect many of today’s models will be rendered obsolete in the
very near future.20
However, the establishment of VaR, CVaR and associated quantitative processes
being modified and developed for the energy market provides a strong framework to
effectively communicate and control risk as well as providing a useful process to
isolate the component risks within the overall company. The use of these quantitative
approaches needs to be complemented by an appropriate stress testing and model
testing environment that challenges both the assumptions made by traders and the
risk management department itself.
One of the most useful aspects of the quantitative approach is that it can be easily
encompassed within an overall risk framework for setting risk/reward ratios, internal
capital allocation, liquidity and reserve requirements. This puts the role of risk
management as a key element in any business organization, truly bridging the gap
between the financial accounting and ‘commercial’ departments.
The risk management strategy, allocating risk and investing in control and analytical
systems needs to be an integral part of the overall business and enterprise risk
framework. This chapter has not sought to recommend how such allocations should
be made since they are a function of individual business environments. Rather, we
would stress that as risk managers are faced with such decisions and organizations
are being restructured it important for the risk manager not to lose sight of the fact
that there is no substitute for common sense.
risk framework. With a highly volatile commodity like power this needs to be multilayered,
with a strong linkage down to a variety of control points.
Figure 18.10 sets out some of the most important quantifiable limits that can be
set and how they relate to each other. This control structure links many parts of the
organization. As well as monitoring for trade input error, significant emphasis needs
to be placed on verifying counterparty and broker confirmations to the trade capture
systems. Conformations can often be slow and inconsistent depending on the
counterparty, which combined with the potential volatility in the market makes for
a dangerous mix. We have seen players lose several million dollars based on trading
input errors that never got picked up by the confirmation process.
Summary
Energy markets are probably the most challenging commodity markets a risk manager
will face in terms of level of complexity, volatility and quantification. Many of
the issues raised and approaches outlined in this chapter are a reflection of the
developing nature of the markets. In addition, some fundamental risk management
issues still need to be adequately addressed.
As the developing energy markets continue to evolve, the tools used by risk
managers to measure, monitor and control market, credit and operational risk will
also continue to develop. These developments will impact energy markets on a
worldwide basis. I suspect many of today’s models will be rendered obsolete in the
very near future.20
However, the establishment of VaR, CVaR and associated quantitative processes
being modified and developed for the energy market provides a strong framework to
effectively communicate and control risk as well as providing a useful process to
isolate the component risks within the overall company. The use of these quantitative
approaches needs to be complemented by an appropriate stress testing and model
testing environment that challenges both the assumptions made by traders and the
risk management department itself.
One of the most useful aspects of the quantitative approach is that it can be easily
encompassed within an overall risk framework for setting risk/reward ratios, internal
capital allocation, liquidity and reserve requirements. This puts the role of risk
management as a key element in any business organization, truly bridging the gap
between the financial accounting and ‘commercial’ departments.
The risk management strategy, allocating risk and investing in control and analytical
systems needs to be an integral part of the overall business and enterprise risk
framework. This chapter has not sought to recommend how such allocations should
be made since they are a function of individual business environments. Rather, we
would stress that as risk managers are faced with such decisions and organizations
are being restructured it important for the risk manager not to lose sight of the fact
that there is no substitute for common sense.
Price data collection
Operational risk managers are dependent on and often responsible to price data
collection and verification. Constructing good forward curves on a daily basis was
briefly discussed above, but one of the biggest difficulties is ensuring that accurate
price data is being used.
Particularly in power, it cannot be assumed that price quotes are reflective of where
people would be willing and able to transact. Liquidity is often extremely limited and
brokers may well estimate the market prices giving an appearance of liquidity that
does not exist. This is also true of NYMEX prices in power where estimates need to
be made for margining previous transactions but actual transactions may not have
occurred at these levels. In these circumstances, it is not unknown to see a significant
divergence between NYMEX and OTC prices for the same product for a limited period
of time.
It is thus necessary to check price data using a number of information sources,
notably trader’s marks, a range of broker quotes, spread prices (using actual or
historical spreads to relate closely traded markets), implied volatility from options
and derivative prices. Standard techniques for fitting average or strip prices should
be used and resultant errors monitored on an on-going basis.
collection and verification. Constructing good forward curves on a daily basis was
briefly discussed above, but one of the biggest difficulties is ensuring that accurate
price data is being used.
Particularly in power, it cannot be assumed that price quotes are reflective of where
people would be willing and able to transact. Liquidity is often extremely limited and
brokers may well estimate the market prices giving an appearance of liquidity that
does not exist. This is also true of NYMEX prices in power where estimates need to
be made for margining previous transactions but actual transactions may not have
occurred at these levels. In these circumstances, it is not unknown to see a significant
divergence between NYMEX and OTC prices for the same product for a limited period
of time.
It is thus necessary to check price data using a number of information sources,
notably trader’s marks, a range of broker quotes, spread prices (using actual or
historical spreads to relate closely traded markets), implied volatility from options
and derivative prices. Standard techniques for fitting average or strip prices should
be used and resultant errors monitored on an on-going basis.
Deals versus trades
As is evident from the above, the underlying characteristics of energy products are
extremely complex. Electricity customers want to switch on a light and get instantaneous
power from a product that has extremely expensive storage; millions of
customers are doing this by a varying amount minute by minute throughout the
year. Many of the transactions in electricity are designed to back this type of delivery
and are non-standard and complex. On the other hand, the US power marketers are
churning millions of MWhs of standard 5î16 blocks of power. A similar dichotomy
exists in gas and oil.
The energy company risk manager has to be conversant with the two different
‘products’ equally efficiently. The standard product must be able to churn a high
transaction rate without creating significant transaction risk, while the ‘bespoke’
contracts need to be broken down into their risk buckets and assessed individually.
The control process needs to make this distinction early. Standard products or
trades should be closely defined under pre-agreed master agreements, which lead to
‘deals’ failing the authorization tests and thus falling into the deal process. Figure
18.9 sets out a standard trade authorization process for power.
It is important to note that with energy trading it is very easy to put general product
authorizations in place that could be interpreted very widely. For instance, one
description could be US Eastern Interconnect Electricity, but does that include the
sale or ancillary services on the hourly or even minute-by-minute market? Does it
allow knock-out options linked to a foreign currency? Many traders will answer yes,
while the board could answer no.
Product authorizations is an area that is fraught with potential problems as
portfolio authorization becomes more complex. The risk manager has to make sure
everyone has a common understanding of their authorizations and that trader
understanding remains a subset of the board members.
Once you have the product definitions and corresponding legal agreements you
can build a highly efficient trade processing system. The standardization allows for
efficient hand-offs which can give you a much higher degree of specialization in the
business, almost to the point where separate individuals are looking at the different
parts of the chain set out above. To make this work you need to be very specific in
defining the physical product definition (including quality specs), delivery point,
tenure and contract specification. Once it is in place and working you can automate
it and let the IT and IS systems take the strain as transaction volumes increase.
The reporting controls at this stage begin to focus on trade input errors, changed
trades and delivery (or book out) issues. As mentioned above, anything falling outside
this needs to fall into deal analysis. This is identical to trade analysis but many of
the components can occur concurrently, often by a specific deal team that has crossfunctional
skills.
The risk manager’s job with the ‘deal’ is to break it down to the appropriate risk
components, the so-called risk bucketing or risk mapping which becomes key within
energy. This is normally done working closely with the structured products or
marketing desks and will encompass operational, market and credit issues.
With respect to all these issues the operational risk manager is very dependent on
the information systems supporting the trade capture and reporting system.
Extremely close links are needed with the IS department who need to meet the
challenge of developing error-free systems within a rapid application environment.
This is only possible where the IS, operations and trading staff each have a good
appreciation of the requirements and limitations of each other. Again communication
becomes the vital link which is aided by a corporate structure that encourages
common goals and a floor plan that keeps these key individuals within reasonable
proximity of each other.
extremely complex. Electricity customers want to switch on a light and get instantaneous
power from a product that has extremely expensive storage; millions of
customers are doing this by a varying amount minute by minute throughout the
year. Many of the transactions in electricity are designed to back this type of delivery
and are non-standard and complex. On the other hand, the US power marketers are
churning millions of MWhs of standard 5î16 blocks of power. A similar dichotomy
exists in gas and oil.
The energy company risk manager has to be conversant with the two different
‘products’ equally efficiently. The standard product must be able to churn a high
transaction rate without creating significant transaction risk, while the ‘bespoke’
contracts need to be broken down into their risk buckets and assessed individually.
The control process needs to make this distinction early. Standard products or
trades should be closely defined under pre-agreed master agreements, which lead to
‘deals’ failing the authorization tests and thus falling into the deal process. Figure
18.9 sets out a standard trade authorization process for power.
It is important to note that with energy trading it is very easy to put general product
authorizations in place that could be interpreted very widely. For instance, one
description could be US Eastern Interconnect Electricity, but does that include the
sale or ancillary services on the hourly or even minute-by-minute market? Does it
allow knock-out options linked to a foreign currency? Many traders will answer yes,
while the board could answer no.
Product authorizations is an area that is fraught with potential problems as
portfolio authorization becomes more complex. The risk manager has to make sure
everyone has a common understanding of their authorizations and that trader
understanding remains a subset of the board members.
Once you have the product definitions and corresponding legal agreements you
can build a highly efficient trade processing system. The standardization allows for
efficient hand-offs which can give you a much higher degree of specialization in the
business, almost to the point where separate individuals are looking at the different
parts of the chain set out above. To make this work you need to be very specific in
defining the physical product definition (including quality specs), delivery point,
tenure and contract specification. Once it is in place and working you can automate
it and let the IT and IS systems take the strain as transaction volumes increase.
The reporting controls at this stage begin to focus on trade input errors, changed
trades and delivery (or book out) issues. As mentioned above, anything falling outside
this needs to fall into deal analysis. This is identical to trade analysis but many of
the components can occur concurrently, often by a specific deal team that has crossfunctional
skills.
The risk manager’s job with the ‘deal’ is to break it down to the appropriate risk
components, the so-called risk bucketing or risk mapping which becomes key within
energy. This is normally done working closely with the structured products or
marketing desks and will encompass operational, market and credit issues.
With respect to all these issues the operational risk manager is very dependent on
the information systems supporting the trade capture and reporting system.
Extremely close links are needed with the IS department who need to meet the
challenge of developing error-free systems within a rapid application environment.
This is only possible where the IS, operations and trading staff each have a good
appreciation of the requirements and limitations of each other. Again communication
becomes the vital link which is aided by a corporate structure that encourages
common goals and a floor plan that keeps these key individuals within reasonable
proximity of each other.
Overall framework, policies and procedures
Most risk managers will start from the G30 recommendations and other derivative
market control frameworks to set their ‘best practice’ standards for middle-office
controls. However, once you are beyond setting up an independent reporting line
they fall short of specific advice on detailed procedures and controls.
The starting point in defining the framework is a consistent Scheme of Authority
and Trading Compliance Guidelines. These will draw upon the existing corporate
scheme, but need to be radically expanded for the energy trading business. Perhaps
more important is to develop a process to ensure traders and risk managers alike
are aware of the controls and how they impact on their day-to-day work. Once the
high-level control environment has been established the difficult problem of product
authorization should start.
market control frameworks to set their ‘best practice’ standards for middle-office
controls. However, once you are beyond setting up an independent reporting line
they fall short of specific advice on detailed procedures and controls.
The starting point in defining the framework is a consistent Scheme of Authority
and Trading Compliance Guidelines. These will draw upon the existing corporate
scheme, but need to be radically expanded for the energy trading business. Perhaps
more important is to develop a process to ensure traders and risk managers alike
are aware of the controls and how they impact on their day-to-day work. Once the
high-level control environment has been established the difficult problem of product
authorization should start.
Operational risk
Operational risk management can be defined as ensuring:
Ω The limits and controls are unambiguous and appropriate for the on-going
business
Ω The day-to-day mark to market and corresponding risk reports are correct
Ω Limits are monitored and breaches alerted in line with predefined procedures
Ω Data and system integrity (including trade input errors)
Operational Risk area is both the most important area under the remit of risk
management but is often given least attention outside the periodic audit review.
Within energy it is virtually impossible to quantify the value of operational risks. It
is often seen as a pure audit or control function adding little value to the enterprise
beyond the ‘let’s keep the auditors happy’ once a year. The operational risk manager
often has thus a ‘no-win’ role of unnecessary bureaucrat in the good times and fall
guy when things go wrong. The role also sits awkwardly between the market/credit
risk functions, the operational accounting functions, the internal audit functions,
the IS department and any front-office operations. As such it is the least well defined
and least emphasized area of risk management.
However, the majority of the big trading losses within the energy business, as well
as other trading businesses, can be traced back to inadequate controls or reporting.
Extreme volatility and a developing market make operational risk management all
the more important.
Ω The limits and controls are unambiguous and appropriate for the on-going
business
Ω The day-to-day mark to market and corresponding risk reports are correct
Ω Limits are monitored and breaches alerted in line with predefined procedures
Ω Data and system integrity (including trade input errors)
Operational Risk area is both the most important area under the remit of risk
management but is often given least attention outside the periodic audit review.
Within energy it is virtually impossible to quantify the value of operational risks. It
is often seen as a pure audit or control function adding little value to the enterprise
beyond the ‘let’s keep the auditors happy’ once a year. The operational risk manager
often has thus a ‘no-win’ role of unnecessary bureaucrat in the good times and fall
guy when things go wrong. The role also sits awkwardly between the market/credit
risk functions, the operational accounting functions, the internal audit functions,
the IS department and any front-office operations. As such it is the least well defined
and least emphasized area of risk management.
However, the majority of the big trading losses within the energy business, as well
as other trading businesses, can be traced back to inadequate controls or reporting.
Extreme volatility and a developing market make operational risk management all
the more important.
Timing is everything
The important factor all credit risk managers need in energy is timely and accurate
information. This is far from simple, especially since most utilities are more used to
monthly or even multiple-month turnarounds. Getting an up-to-date list of counterparty
exposures and their associated lists was traditionally seen an initiative that
sometimes happened after a major event.
Unfortunately credit managers need to work within these constraints and balance
between what is desirable versus what is feasible. As such their ability to get timely
exposure reports and lists of potential counterparties is often hampered by antiquated
accounting systems and a bemused sales force and management. In some cases the
credit managers are still divorced from the trading team and housed with the
corporate finance and treasury function.
Given the potential for extreme volatility, isolation of credit from market risk is
extremely dangerous in energy. With prices moving dramatically credit exposures
become significant in hours rather than days or weeks, and problems often need to
be dealt with before the next day. Energy has neither the luxury of the institutionalized
nature of the finance markets nor the ability to take the more blase´ approach of
a traditional utility who has experienced few historical defaults. As such, the credit
manager and risk manager need to be side by side on the trading floor as should the
most important person on the energy trade floor, the operational risk manager.
information. This is far from simple, especially since most utilities are more used to
monthly or even multiple-month turnarounds. Getting an up-to-date list of counterparty
exposures and their associated lists was traditionally seen an initiative that
sometimes happened after a major event.
Unfortunately credit managers need to work within these constraints and balance
between what is desirable versus what is feasible. As such their ability to get timely
exposure reports and lists of potential counterparties is often hampered by antiquated
accounting systems and a bemused sales force and management. In some cases the
credit managers are still divorced from the trading team and housed with the
corporate finance and treasury function.
Given the potential for extreme volatility, isolation of credit from market risk is
extremely dangerous in energy. With prices moving dramatically credit exposures
become significant in hours rather than days or weeks, and problems often need to
be dealt with before the next day. Energy has neither the luxury of the institutionalized
nature of the finance markets nor the ability to take the more blase´ approach of
a traditional utility who has experienced few historical defaults. As such, the credit
manager and risk manager need to be side by side on the trading floor as should the
most important person on the energy trade floor, the operational risk manager.
Setting limits
Setting appropriate limits by counterparty will depend on both your risk tolerance
and the evaluation of the counterparty. For example, a reasonable approach employed
by many is to set the limits based on a percentage of their net tangible equity or net
asset value. Depending on your appetite for risk, you may want to set a limit of, say,
3% of any one party’s equity position. A similar percentage limit could be set off your
own liquid asset or equity position which would set the maximum exposure you
would want to any counterparty, however strong. These percentages would be
adjusted for the creditworthiness of the counterparty in question.
Once you have the credit evaluation structure in place the use of Credit-Value-at-
Risk (CVaR) is one way of estimating the potential credit exposure using a probability
distribution of price movements in the same way VaR is calculated for market risk.
This has particular relevance given the volatilities in power. Estimating this requires
the same fundamental information as VaR itself, including the current market price
for the contract, a price distribution, the magnitude of unpredictable price movements
(volatility and confidence interval) and the appropriate time it would take to rehedge
the position (holding period). The three latter components have distinctive traits in the
case of default since the event itself will be closely correlated with price movements.
A key issue is the determination of what volatility and time frame to apply. If we
take the example of Utility Y above, the trade was hedging an existing exposure six
months out and as such that hedge must remain for the entire period to minimize
the market risk. In addition, the magnitude of this event and the time it takes to
unwind positions will also depend on the market share of the defaulting party. The
immediate question that needs to be addressed is that if VaR is driven by the
assumption of normal market conditions, surely a default (being an abnormal market
event) makes it meaningless.
A starting point in solving this issue is to significantly lengthen the holding period.
A thirty-day holding period may be more appropriate in the instance of default rather
than the shorted period of, say, five days used of VaR itself. A credit risk manager
also needs to look at the potential movement of CVaR into the future as a way of
managing credit risk. In other words limits should be set on ‘potential exposures’ as
well as on CVaR itself. Graphing the predicted movements of exposure including
CVaR movements over the future duration of the portfolio provides a useful warning
of potential risks on the horizon, particularly given the seasonal nature of the
volatility outlined above. It is important to note that CVaR does not define your credit
risk (how likely you are to lose money) but it is a useful construct which will quantify
the likely exposure in the case of default.
The next step is to combine these risks with the probability of default and the likelihood
of multiple defaults (unfortunately all too common). This would allow you to estimate
the total portfolio credit risk to combine with the market risks within the book.
We can thus quantify two principal control points in setting credit policy – at the
individual level, the maximum potential exposure per counterparty, and at the macro
level, the credit risk within the book or portfolio. Both are fundamental to running a
well-managed energy trading book, although unfortunately while they provide strong
markers to how much risk is in the book they cannot accurately predict actual losses
in the case of default.
and the evaluation of the counterparty. For example, a reasonable approach employed
by many is to set the limits based on a percentage of their net tangible equity or net
asset value. Depending on your appetite for risk, you may want to set a limit of, say,
3% of any one party’s equity position. A similar percentage limit could be set off your
own liquid asset or equity position which would set the maximum exposure you
would want to any counterparty, however strong. These percentages would be
adjusted for the creditworthiness of the counterparty in question.
Once you have the credit evaluation structure in place the use of Credit-Value-at-
Risk (CVaR) is one way of estimating the potential credit exposure using a probability
distribution of price movements in the same way VaR is calculated for market risk.
This has particular relevance given the volatilities in power. Estimating this requires
the same fundamental information as VaR itself, including the current market price
for the contract, a price distribution, the magnitude of unpredictable price movements
(volatility and confidence interval) and the appropriate time it would take to rehedge
the position (holding period). The three latter components have distinctive traits in the
case of default since the event itself will be closely correlated with price movements.
A key issue is the determination of what volatility and time frame to apply. If we
take the example of Utility Y above, the trade was hedging an existing exposure six
months out and as such that hedge must remain for the entire period to minimize
the market risk. In addition, the magnitude of this event and the time it takes to
unwind positions will also depend on the market share of the defaulting party. The
immediate question that needs to be addressed is that if VaR is driven by the
assumption of normal market conditions, surely a default (being an abnormal market
event) makes it meaningless.
A starting point in solving this issue is to significantly lengthen the holding period.
A thirty-day holding period may be more appropriate in the instance of default rather
than the shorted period of, say, five days used of VaR itself. A credit risk manager
also needs to look at the potential movement of CVaR into the future as a way of
managing credit risk. In other words limits should be set on ‘potential exposures’ as
well as on CVaR itself. Graphing the predicted movements of exposure including
CVaR movements over the future duration of the portfolio provides a useful warning
of potential risks on the horizon, particularly given the seasonal nature of the
volatility outlined above. It is important to note that CVaR does not define your credit
risk (how likely you are to lose money) but it is a useful construct which will quantify
the likely exposure in the case of default.
The next step is to combine these risks with the probability of default and the likelihood
of multiple defaults (unfortunately all too common). This would allow you to estimate
the total portfolio credit risk to combine with the market risks within the book.
We can thus quantify two principal control points in setting credit policy – at the
individual level, the maximum potential exposure per counterparty, and at the macro
level, the credit risk within the book or portfolio. Both are fundamental to running a
well-managed energy trading book, although unfortunately while they provide strong
markers to how much risk is in the book they cannot accurately predict actual losses
in the case of default.
Counterparty evaluation: some challenges
Unfortunately, the standard assessment measures employed by credit rating services
such as Moody’s, Standard & Poor’s, Fitch Investors or Dunn & Bradstreet, while
useful, often provide insufficient detail when evaluating an energy trading businesses.
Rating agencies rarely address in detail the controls within the business and the
potential for major losses. Coupled with this is the lack of financial information on
subsidiary trading businesses. The result is that risk professionals who rely solely
on the rating agencies can miss potential problems or set an overly prudent
approach.19
Companies that want to quantify adequately the creditworthiness of a counterparty
will need to develop their own internal scoring methodologies to complement the
rating agencies and develop their own specialist ratings. The internal scoring system
should take account of qualitative as well as quantitative factors such as the
ownership and control of the entity, track record and diversification. This need for
an internal scoring system is reinforced by the fact that many trading organizations
are not rated by the agencies and that their financial strength can change very
rapidly. Second, as evidenced in the above Utility Y scenario, the significance of the
replacement cost of electricity becomes particularly important.
The legal issues around default can also be complex. In the US power market the
legal framework was initially created through ‘enabling’ agreements which have not
fully addressed the problem of credit risk appropriately. Failure to deliver may not
result in either default or damages. The enforceability of netting or set-off agreements
can also be problematical, as can the implementation of credit enhancement or
‘further assurances’ clauses that are intended to improve your position when a
counterparty’s creditworthiness is called into doubt.
One area that is still being clarified is whether electricity constitutes a good or a
commodity. As the market becomes more ‘commodity-like’ it moves more into line
with the protection given to commodities under the relevant bankruptcy codes and
allows the default process set out within ISDA documentation to be used. This
development in the legal framework surrounding electricity and gas contracts is one
that is likely to repeat itself throughout the world.
Thus, ensuring the master agreements reflect appropriate termination and netting
provisions is an important first step to mitigate credit exposure. Understanding
the interactions between utility tariffs, master agreements for different products,
guarantees and additional agreements (such as separate netting agreements) is an
essential but unfortunately major exercise, as is ensuring you understand the
corporate structure of the counterparty with whom you are dealing. With this
structure in place you can set an exposure limit for each counterparty.
such as Moody’s, Standard & Poor’s, Fitch Investors or Dunn & Bradstreet, while
useful, often provide insufficient detail when evaluating an energy trading businesses.
Rating agencies rarely address in detail the controls within the business and the
potential for major losses. Coupled with this is the lack of financial information on
subsidiary trading businesses. The result is that risk professionals who rely solely
on the rating agencies can miss potential problems or set an overly prudent
approach.19
Companies that want to quantify adequately the creditworthiness of a counterparty
will need to develop their own internal scoring methodologies to complement the
rating agencies and develop their own specialist ratings. The internal scoring system
should take account of qualitative as well as quantitative factors such as the
ownership and control of the entity, track record and diversification. This need for
an internal scoring system is reinforced by the fact that many trading organizations
are not rated by the agencies and that their financial strength can change very
rapidly. Second, as evidenced in the above Utility Y scenario, the significance of the
replacement cost of electricity becomes particularly important.
The legal issues around default can also be complex. In the US power market the
legal framework was initially created through ‘enabling’ agreements which have not
fully addressed the problem of credit risk appropriately. Failure to deliver may not
result in either default or damages. The enforceability of netting or set-off agreements
can also be problematical, as can the implementation of credit enhancement or
‘further assurances’ clauses that are intended to improve your position when a
counterparty’s creditworthiness is called into doubt.
One area that is still being clarified is whether electricity constitutes a good or a
commodity. As the market becomes more ‘commodity-like’ it moves more into line
with the protection given to commodities under the relevant bankruptcy codes and
allows the default process set out within ISDA documentation to be used. This
development in the legal framework surrounding electricity and gas contracts is one
that is likely to repeat itself throughout the world.
Thus, ensuring the master agreements reflect appropriate termination and netting
provisions is an important first step to mitigate credit exposure. Understanding
the interactions between utility tariffs, master agreements for different products,
guarantees and additional agreements (such as separate netting agreements) is an
essential but unfortunately major exercise, as is ensuring you understand the
corporate structure of the counterparty with whom you are dealing. With this
structure in place you can set an exposure limit for each counterparty.
The challenges of Utility Y
Let us take the example of a Cinergy Power trade by Utility Y. Utility Y is very
prudent, they like to have a fully hedged position and severely limit purchases from
small players.
Initiated in November 1997 for July 1998. Y buys a 50 MW contract from a small
counterparty for $35/MWh at a total cost $560 000, and with annualized volatility
running around 20%. Given they are buying, they see no receivable exposure, no
mark to market exposure, and estimate the potential market exposure to be $112 000.
By February the price is $50/MWh and a mark to market value is $240 000.
Volatility is now 30%, so potential exposure (receivables plus mark to market plus
potential market exposure) is equal to $313 600. Although the counterparty is a
small player they are still well within Y’s ‘prudent’ limit of $1 million set for this
particular counterparty.
In June 1998 the market price has risen to $350/MWh (annualized volatility is
now over 1000%) and Y’s counterparty defaults by telling them they will not deliver
the power under the contract. The replacement power they need to buy to meet their
customer’s requirements cost $5.6 million, a net loss of over $5 million, five times
the limit they thought. This could have been a true story for anyone trading the US
power market industry in the summer of 1998.
Given such volatility in these markets and their early stages of development, it is
not surprising to see some casualties. This tends to occur during major market
movements and can provide additional impetus to the market movement that started
it. The defaults of Federal Energy and Power Company of America were both caused
by and caused the extreme volatility experienced in the Cinergy spot market (it
reached x000% on an annualized, historical basis).17
It is an important point to note that the biggest risk facing many companies at this
stage was the non-delivery of power rather than the non-payment of delivered power.
As such credit managers switched their attention away from the traditional accounts
receivable approach to focusing on counterparty mark to market and credit VaR
approaches. This reinforced the need to focus on the total exposure to a counterparty
default and the potential for domino effects within a particular sector. We will
set out below a number of ‘must-haves’ in assessing credit risk within the energy
sector within an overall credit process.
Credit breaks down into two distinct areas: the initial credit evaluation process
and the on-going monitoring of exposures. Credit evaluation includes the evaluation
of a particular counterparty, taking account of the market they are in (country and
sector) as well as the current financial strength and track record of the company.
Counterparty credit limits are then based on your own risk appetite and strategic
importance of that counter-party.18
Exposure monitoring includes the identification and measurement of exposures as
well as the operational reporting and control process. This should be done at a
counterparty and portfolio level.
prudent, they like to have a fully hedged position and severely limit purchases from
small players.
Initiated in November 1997 for July 1998. Y buys a 50 MW contract from a small
counterparty for $35/MWh at a total cost $560 000, and with annualized volatility
running around 20%. Given they are buying, they see no receivable exposure, no
mark to market exposure, and estimate the potential market exposure to be $112 000.
By February the price is $50/MWh and a mark to market value is $240 000.
Volatility is now 30%, so potential exposure (receivables plus mark to market plus
potential market exposure) is equal to $313 600. Although the counterparty is a
small player they are still well within Y’s ‘prudent’ limit of $1 million set for this
particular counterparty.
In June 1998 the market price has risen to $350/MWh (annualized volatility is
now over 1000%) and Y’s counterparty defaults by telling them they will not deliver
the power under the contract. The replacement power they need to buy to meet their
customer’s requirements cost $5.6 million, a net loss of over $5 million, five times
the limit they thought. This could have been a true story for anyone trading the US
power market industry in the summer of 1998.
Given such volatility in these markets and their early stages of development, it is
not surprising to see some casualties. This tends to occur during major market
movements and can provide additional impetus to the market movement that started
it. The defaults of Federal Energy and Power Company of America were both caused
by and caused the extreme volatility experienced in the Cinergy spot market (it
reached x000% on an annualized, historical basis).17
It is an important point to note that the biggest risk facing many companies at this
stage was the non-delivery of power rather than the non-payment of delivered power.
As such credit managers switched their attention away from the traditional accounts
receivable approach to focusing on counterparty mark to market and credit VaR
approaches. This reinforced the need to focus on the total exposure to a counterparty
default and the potential for domino effects within a particular sector. We will
set out below a number of ‘must-haves’ in assessing credit risk within the energy
sector within an overall credit process.
Credit breaks down into two distinct areas: the initial credit evaluation process
and the on-going monitoring of exposures. Credit evaluation includes the evaluation
of a particular counterparty, taking account of the market they are in (country and
sector) as well as the current financial strength and track record of the company.
Counterparty credit limits are then based on your own risk appetite and strategic
importance of that counter-party.18
Exposure monitoring includes the identification and measurement of exposures as
well as the operational reporting and control process. This should be done at a
counterparty and portfolio level.
Credit risk – why 3000% plus volatility matters
Credit risk can be defined in a number of ways, but will consist of the following:
Ω Unpaid bills (current and aged debt)
Ω Delivered but unbilled (realized)
Ω Replacement cost of future commitments (unrealized) – mark to market exposure
Ω Potential adverse movements or replacement costs – CVaR (Credit Value-at-Risk)
and potential exposure (maximum exposure given potential movements in prices
and volatilities over the life of the contract)
The volatility of energy prices, particularly power, makes the challenge of managing
credit risk particularly interesting as exposure is potentially unlimited.
Ω Unpaid bills (current and aged debt)
Ω Delivered but unbilled (realized)
Ω Replacement cost of future commitments (unrealized) – mark to market exposure
Ω Potential adverse movements or replacement costs – CVaR (Credit Value-at-Risk)
and potential exposure (maximum exposure given potential movements in prices
and volatilities over the life of the contract)
The volatility of energy prices, particularly power, makes the challenge of managing
credit risk particularly interesting as exposure is potentially unlimited.
Pricing issues
The lack of liquidity and the complexity of market, means that many positions cannot
be easily hedged. In this case, the risk-free pricing models may over-estimate the
value of deals since such deals will carry significant market risk that should be
factored into the price.
It is thus important to assess the profit and risk impact the deal will make on the
overall portfolio. In turn, this may require a re-optimization of portfolio using all
available hedges, which can be a complex process, creating additional model and
operational risk in the assessment of complex deals.
be easily hedged. In this case, the risk-free pricing models may over-estimate the
value of deals since such deals will carry significant market risk that should be
factored into the price.
It is thus important to assess the profit and risk impact the deal will make on the
overall portfolio. In turn, this may require a re-optimization of portfolio using all
available hedges, which can be a complex process, creating additional model and
operational risk in the assessment of complex deals.
Stress testing
Some of the solutions outlined above fall into the category of ‘stress testing’, which
is vital in all markets and energy is certainly no exception. Stress testing can take
many forms, the most common being historical tests. Others will include model tests
and extreme event forecasting. A sample set of tests would include:
Ω Apply the highest observed volatility
Ω Apply the highest observed prices
Ω Combination of high prices and volatility
Ω Zero correlation
Ω Adverse correlations
Ω Six sigma event
Ω Price shock (i.e. price times ten , or divided by ten)
Ω Different price shapes (i.e. ratio summer/winter; weekday/weekend)
Ω Alternative distribution shapes
Ω Clustering extreme events
Ω Assuming particular positions cannot be hedged before expiry
While many of the tests should be automated and produced on a regular basis
(daily or weekly) stress tests need to be designed around the particular portfolio and
should be an iterative process to find particular weaknesses in the portfolio. They
are also likely to be a key part in the overall portfolio analysis that will drive the
trading and hedging strategy.
Stress tests play a vital role in setting overall liquidity and reserve requirements of
the business as well as a vital communication tool to senior management. The
single biggest danger with VaR is complacency because the portfolio has not been
appropriately stressed.
Stress testing should also encompass all the correlated risks within the business
that may not otherwise be combined. The most obvious is the combination of credit
and market risks under alternative ‘market crisis’ scenarios defaults, leading to
additional volatility, leading to more defaults.
is vital in all markets and energy is certainly no exception. Stress testing can take
many forms, the most common being historical tests. Others will include model tests
and extreme event forecasting. A sample set of tests would include:
Ω Apply the highest observed volatility
Ω Apply the highest observed prices
Ω Combination of high prices and volatility
Ω Zero correlation
Ω Adverse correlations
Ω Six sigma event
Ω Price shock (i.e. price times ten , or divided by ten)
Ω Different price shapes (i.e. ratio summer/winter; weekday/weekend)
Ω Alternative distribution shapes
Ω Clustering extreme events
Ω Assuming particular positions cannot be hedged before expiry
While many of the tests should be automated and produced on a regular basis
(daily or weekly) stress tests need to be designed around the particular portfolio and
should be an iterative process to find particular weaknesses in the portfolio. They
are also likely to be a key part in the overall portfolio analysis that will drive the
trading and hedging strategy.
Stress tests play a vital role in setting overall liquidity and reserve requirements of
the business as well as a vital communication tool to senior management. The
single biggest danger with VaR is complacency because the portfolio has not been
appropriately stressed.
Stress testing should also encompass all the correlated risks within the business
that may not otherwise be combined. The most obvious is the combination of credit
and market risks under alternative ‘market crisis’ scenarios defaults, leading to
additional volatility, leading to more defaults.
Value-at-Risk for energy
As in many markets, VaR has been accepted as the most common way of communicating
risk parameters within the energy sector. A detailed description of VaR itself
is addressed elsewhere in this book.
VaR is driven by four principal factors: volatility, correlations, holding period and
confidence interval. Given the first two parameters VaR models are going to be
susceptible to many of the problems outlined above. However, the holding period
and confidence interval assumptions have mixed effects. A holding period of ten days
or less is common in the actively traded energy markets. This allows for a relative
lack of liquidity compared to financial markets and the ‘communication delay’ – it
may take 24 hours to identify a problem, another 24 to agree on the way forward.
A five-day holding period will tend to avoid much of the mean reversion and
seasonality issues, provided the volatilities have been estimated correctly. You can
thus give a reasonable five-day estimate of VaR despite many of the issues outlined
above. The concern is that it gives a very incomplete picture of your risk profile. In
particular, the chairman and CEO is often more concerned about potential major
losses. In other words they will be more worried about the 5% rather than the 95%.
While a VaR can give some clues to the maximum loss, with significant spikes the
losses may be many multiples of the VaR estimation.
Also, many of the traded products cannot be hedged within 10 days (many cannot
be hedged at all). In this case the holding periods need to be re-assessed in light of
the particular product. The overall VaR associated will then be affected by the time
to maturity, ability to find partial hedges and mean reversion of volatility.
Also, as we discussed above, optionality is pervasive in energy and the more
traditional VaR needs to be expanded and complemented to take account of the full
range of option risks. It other words, delta, gamma, vega and theta risks need to be
quantified and added to the VaR calculation. Where significant optionality exists a
simulation approach is likely to be most effective, although this may be at the cost
of timely information.
risk parameters within the energy sector. A detailed description of VaR itself
is addressed elsewhere in this book.
VaR is driven by four principal factors: volatility, correlations, holding period and
confidence interval. Given the first two parameters VaR models are going to be
susceptible to many of the problems outlined above. However, the holding period
and confidence interval assumptions have mixed effects. A holding period of ten days
or less is common in the actively traded energy markets. This allows for a relative
lack of liquidity compared to financial markets and the ‘communication delay’ – it
may take 24 hours to identify a problem, another 24 to agree on the way forward.
A five-day holding period will tend to avoid much of the mean reversion and
seasonality issues, provided the volatilities have been estimated correctly. You can
thus give a reasonable five-day estimate of VaR despite many of the issues outlined
above. The concern is that it gives a very incomplete picture of your risk profile. In
particular, the chairman and CEO is often more concerned about potential major
losses. In other words they will be more worried about the 5% rather than the 95%.
While a VaR can give some clues to the maximum loss, with significant spikes the
losses may be many multiples of the VaR estimation.
Also, many of the traded products cannot be hedged within 10 days (many cannot
be hedged at all). In this case the holding periods need to be re-assessed in light of
the particular product. The overall VaR associated will then be affected by the time
to maturity, ability to find partial hedges and mean reversion of volatility.
Also, as we discussed above, optionality is pervasive in energy and the more
traditional VaR needs to be expanded and complemented to take account of the full
range of option risks. It other words, delta, gamma, vega and theta risks need to be
quantified and added to the VaR calculation. Where significant optionality exists a
simulation approach is likely to be most effective, although this may be at the cost
of timely information.
9 Temmuz 2011 Cumartesi
Model risk
Many of the issues outlined above primarily exist or are exaggerated in the spot
markets where many quantitative estimation tools are less effective and the resultant
model risk becomes very significant. One of the major reasons for hedging into
forward markets should be to avoid this model risk.
Model risk can come from a number of sources, but are principally driven from
forward curve models (particularly the volatility forward curve) and pricing models
(option, simulation and optimization models). It is generally recommended that
traders and risk managers have a preset (and audited) library of pricing tools to work
from, but in many cases longer-term contracts will require the correct building blocks
to correctly price a particular offering. One example of where such pricing can go
wrong is taking account of correlations between different components of a deal in
hedging a variable retail load.
It is common in both the gas and power markets to offer retail hedges where
the volume sold is linked to actual consumption of a customer group. In these
circumstances there is a positive correlation between the volume being hedged and
market clearing prices (the customer group’s consumption itself being part of the
demand driver in prices). This correlation changes at different consumption levels
following a generally exponential curve.
A simplistic approach could make a mistake on two fronts. If you ignore the
extent of the positive (non-linear) correlation between price and demand you could
significantly under-price the hedge. Above certain overall market demand levels,
prices will rise exponentially, which will occur exactly as the volume requirement of
the hedge increases. While most models will take account of the cost of increased
demand at high prices, they often do not simulate the non-linear relationship, thus
under-pricing the option.
A number of early power deals reportedly made this mistake when taking on such
positions. They hedged full requirement positions (supply all the load a group of
customers take) at expected volumes. With increasing volatility in the marketplace
and a positive correlation between prices and the customer’s consumption they then
found themselves long at lower than expected prices and short at higher than
expected prices, resulting in a significant overall loss compared to the hedged
position.
The alternative problem is to ignore the fact that the call option is not being
optimized against high prices (customers use energy when they need it, not when it
has most value in the market) and it is a positive but by no means perfect correlation.
If the customer’s ‘call option’ to increase their volume is priced, ignoring this the
option will be significantly over-priced. A diverse portfolio of different customer types
dissipates much of this call risk.
As well as the problem associated with volume uncertainty, energy contracts
abound with complexity reflecting the immaturity of the market (many older contracts
still in force reflect the realities of a regulated utility world) and physical realities of
the market price. Combinations of average price options, look-back index options,
knock-out features, swing contracts and time spread options are common. In addition,
it is not uncommon to see contracts against terms such as ‘average market
prices’ rather than a specific published index, making the translation from contract
to model more complex (as well as opening up a number of other pricing issues).
To minimize model risk, an extensive testing program needs to be put in place to
ensure this is quantified on an on-going basis. The backtesting of model estimates
should be used to estimate potential model errors and incorporated into the hedging
strategy. In particular the assumptions surrounding mean reversion in the volatility
curve (which significantly reduces the volatility over time) need to be carefully tested
against actual marks wherever possible given the lack of implied curves that can be
derived. Market illiquidity will make this process difficult, often resulting in arbitrary
allocations between liquidity and modeling risk allocations. A risk manager needs to
be very careful of such allocations to ensure consistency across different products
and markets. 546
markets where many quantitative estimation tools are less effective and the resultant
model risk becomes very significant. One of the major reasons for hedging into
forward markets should be to avoid this model risk.
Model risk can come from a number of sources, but are principally driven from
forward curve models (particularly the volatility forward curve) and pricing models
(option, simulation and optimization models). It is generally recommended that
traders and risk managers have a preset (and audited) library of pricing tools to work
from, but in many cases longer-term contracts will require the correct building blocks
to correctly price a particular offering. One example of where such pricing can go
wrong is taking account of correlations between different components of a deal in
hedging a variable retail load.
It is common in both the gas and power markets to offer retail hedges where
the volume sold is linked to actual consumption of a customer group. In these
circumstances there is a positive correlation between the volume being hedged and
market clearing prices (the customer group’s consumption itself being part of the
demand driver in prices). This correlation changes at different consumption levels
following a generally exponential curve.
A simplistic approach could make a mistake on two fronts. If you ignore the
extent of the positive (non-linear) correlation between price and demand you could
significantly under-price the hedge. Above certain overall market demand levels,
prices will rise exponentially, which will occur exactly as the volume requirement of
the hedge increases. While most models will take account of the cost of increased
demand at high prices, they often do not simulate the non-linear relationship, thus
under-pricing the option.
A number of early power deals reportedly made this mistake when taking on such
positions. They hedged full requirement positions (supply all the load a group of
customers take) at expected volumes. With increasing volatility in the marketplace
and a positive correlation between prices and the customer’s consumption they then
found themselves long at lower than expected prices and short at higher than
expected prices, resulting in a significant overall loss compared to the hedged
position.
The alternative problem is to ignore the fact that the call option is not being
optimized against high prices (customers use energy when they need it, not when it
has most value in the market) and it is a positive but by no means perfect correlation.
If the customer’s ‘call option’ to increase their volume is priced, ignoring this the
option will be significantly over-priced. A diverse portfolio of different customer types
dissipates much of this call risk.
As well as the problem associated with volume uncertainty, energy contracts
abound with complexity reflecting the immaturity of the market (many older contracts
still in force reflect the realities of a regulated utility world) and physical realities of
the market price. Combinations of average price options, look-back index options,
knock-out features, swing contracts and time spread options are common. In addition,
it is not uncommon to see contracts against terms such as ‘average market
prices’ rather than a specific published index, making the translation from contract
to model more complex (as well as opening up a number of other pricing issues).
To minimize model risk, an extensive testing program needs to be put in place to
ensure this is quantified on an on-going basis. The backtesting of model estimates
should be used to estimate potential model errors and incorporated into the hedging
strategy. In particular the assumptions surrounding mean reversion in the volatility
curve (which significantly reduces the volatility over time) need to be carefully tested
against actual marks wherever possible given the lack of implied curves that can be
derived. Market illiquidity will make this process difficult, often resulting in arbitrary
allocations between liquidity and modeling risk allocations. A risk manager needs to
be very careful of such allocations to ensure consistency across different products
and markets. 546
Energy options – financial and ‘real’ options
Many energy contracts have significant optionality built into them at both the
producer and consumer ends. This can range from standard traded options to
complex embedded options in long-term contracts. Unfortunately, one of the most
complex options to price is a power station.
In the gas market, in both the USA and Europe, many of the annual or longerdated
contacts have some degree of flexibility or ‘swing’ built into them reflecting the
production flexibility and/or storage capabilities of the supplier. These allow buyers
to amend their daily or monthly takes provided they meet overall volume requirements.
This flexibility from the buyer may be offset by the ability of the seller to
interrupt the contract, for instance if there are production problems.
Such a situation leads to a complex option pricing model where it is easy to
mis-specify the parameters unless the quantitative analyst is very familiar with
the model restrictions, the nature of the pricing volatility and the contract structure.
Complex contract structures should not be finalized before these problems can be
solved.
Many sellers of such options will link the structure back to actual capabilities of
the underlying asset and given the developments of hedge accounting rules (for
instance, FASB 133 in the USA) this is likely to become more pronounced. If this is
a gas field, only limited flexibility will exist to increase or decrease production and
different storage techniques will be able to respond to market prices at different
speeds. Given these operational constraints and the sale of European- or Americanstyle
options based on single-period clearing, prices need to be handled very carefully
since the ‘real option’ may not be available or able to respond to that particular price
signal. This is particularly true when pricing against illiquid and volatile prices that
may take some time to find their equilibrium rates.
Another concern is the ability to effectively delta hedge positions given that standard
‘lot’ sizes in the power OTC markets is relatively large, transaction costs relatively
high and liquidity often sporadic. It is important to evaluate the practicality of the
delta hedging strategy in assessing the value of such a position. It is also important
to distinguish between pseudo-random and commercial optionality. Force majeure,
such as pipeline failures, can be priced like a call option but the likelihood of the call
is part random, part correlated with higher prices. This will be significantly less
expensive than commercial options that are driven purely by price. They are, however,
seldom as neutral as many like to assume, i.e. the assumption that if it was purely
random it would be costless in optionality over some time period.
Given the popularity of spreads in energy to create synthetic transport, power
stations, refineries and storage it is not surprising that spread options are also
popular. Similarly, the underlying assets need to be valued on a consistent basis
with their financial counterparts leading many (if not most) energy players to value
asset positions against their real option value.
For instance, a power station is a particularly complex ‘real’ spread option. Given
the right contract structure, it allows you to arbitrage between two energy markets,
say gas and power at a given cost. Their value will thus be driven not only by their
expected arbitrage value but also by the volatility and correlation between the
markets (high volatility/low correlation leading to high values). Some power stations
(such as a large nuclear plant) are deep in the money on the power side, increasing
their intrinsic value but reducing their option value. Others (such as an open-cycle
gas turbine which has a high energy cost, low capital cost) may be out of the money
maintaining significant option value but little intrinsic value.
The ability of a power station to take advantage of the optionality will depend on
its availability and flexibility. Power stations are not like contracts: as one electricity
executive once said, ‘contracts do not have tube leaks’. They also don’t have to obey
the laws of physics in starting up, which for some plant can take several hours or
even days. A plant that takes a long time and is expensive to start up will be
significantly less valuable from an option/arbitrage viewpoint.
Finally, some power stations can arbitrage between several markets. They may be
able to run on more than one fuel source and trade into the ‘ancillary service’16
markets.
Like many real options, these unique characteristics make it extremely difficult to
model a power station in standard models. Unless you are intent on solving the
intellectual challenge of the generalized closed-form solution for power stations I
would suggest that a Monte Carlo simulation package is a much more effective tool.
A number of companies have been developing these for years and if combined with
a quantitative understanding of the price volatility can provide an effective pricing
tool set.
It is also important to consider who will buy the option value. Many consumers of
power and gas must buy on a daily basis since they have no storage capability and
with the underlying daily volatility of these markets will be looking for firm-average
price options to hedge their physical consumption. Unfortunately this won’t match
with the production assets, leaving a complex basis position.
producer and consumer ends. This can range from standard traded options to
complex embedded options in long-term contracts. Unfortunately, one of the most
complex options to price is a power station.
In the gas market, in both the USA and Europe, many of the annual or longerdated
contacts have some degree of flexibility or ‘swing’ built into them reflecting the
production flexibility and/or storage capabilities of the supplier. These allow buyers
to amend their daily or monthly takes provided they meet overall volume requirements.
This flexibility from the buyer may be offset by the ability of the seller to
interrupt the contract, for instance if there are production problems.
Such a situation leads to a complex option pricing model where it is easy to
mis-specify the parameters unless the quantitative analyst is very familiar with
the model restrictions, the nature of the pricing volatility and the contract structure.
Complex contract structures should not be finalized before these problems can be
solved.
Many sellers of such options will link the structure back to actual capabilities of
the underlying asset and given the developments of hedge accounting rules (for
instance, FASB 133 in the USA) this is likely to become more pronounced. If this is
a gas field, only limited flexibility will exist to increase or decrease production and
different storage techniques will be able to respond to market prices at different
speeds. Given these operational constraints and the sale of European- or Americanstyle
options based on single-period clearing, prices need to be handled very carefully
since the ‘real option’ may not be available or able to respond to that particular price
signal. This is particularly true when pricing against illiquid and volatile prices that
may take some time to find their equilibrium rates.
Another concern is the ability to effectively delta hedge positions given that standard
‘lot’ sizes in the power OTC markets is relatively large, transaction costs relatively
high and liquidity often sporadic. It is important to evaluate the practicality of the
delta hedging strategy in assessing the value of such a position. It is also important
to distinguish between pseudo-random and commercial optionality. Force majeure,
such as pipeline failures, can be priced like a call option but the likelihood of the call
is part random, part correlated with higher prices. This will be significantly less
expensive than commercial options that are driven purely by price. They are, however,
seldom as neutral as many like to assume, i.e. the assumption that if it was purely
random it would be costless in optionality over some time period.
Given the popularity of spreads in energy to create synthetic transport, power
stations, refineries and storage it is not surprising that spread options are also
popular. Similarly, the underlying assets need to be valued on a consistent basis
with their financial counterparts leading many (if not most) energy players to value
asset positions against their real option value.
For instance, a power station is a particularly complex ‘real’ spread option. Given
the right contract structure, it allows you to arbitrage between two energy markets,
say gas and power at a given cost. Their value will thus be driven not only by their
expected arbitrage value but also by the volatility and correlation between the
markets (high volatility/low correlation leading to high values). Some power stations
(such as a large nuclear plant) are deep in the money on the power side, increasing
their intrinsic value but reducing their option value. Others (such as an open-cycle
gas turbine which has a high energy cost, low capital cost) may be out of the money
maintaining significant option value but little intrinsic value.
The ability of a power station to take advantage of the optionality will depend on
its availability and flexibility. Power stations are not like contracts: as one electricity
executive once said, ‘contracts do not have tube leaks’. They also don’t have to obey
the laws of physics in starting up, which for some plant can take several hours or
even days. A plant that takes a long time and is expensive to start up will be
significantly less valuable from an option/arbitrage viewpoint.
Finally, some power stations can arbitrage between several markets. They may be
able to run on more than one fuel source and trade into the ‘ancillary service’16
markets.
Like many real options, these unique characteristics make it extremely difficult to
model a power station in standard models. Unless you are intent on solving the
intellectual challenge of the generalized closed-form solution for power stations I
would suggest that a Monte Carlo simulation package is a much more effective tool.
A number of companies have been developing these for years and if combined with
a quantitative understanding of the price volatility can provide an effective pricing
tool set.
It is also important to consider who will buy the option value. Many consumers of
power and gas must buy on a daily basis since they have no storage capability and
with the underlying daily volatility of these markets will be looking for firm-average
price options to hedge their physical consumption. Unfortunately this won’t match
with the production assets, leaving a complex basis position.
Correlations
Few, if any, energy companies trade single products at the same delivery point. The
cross-correlation between energy products this becomes crucial in estimating the
overall value and risk profile for a particular position.
As stated above, oil correlations are actively traded as crack spreads, locational
spreads and time spreads. Similarly, ‘spark spreads’ between gas and power are
commonly traded and these create a synthetic gas-fired power station. The higher
the correlation between the input gas costs and output electricity prices, the easier
it is to hedge power prices with gas prices and thus reduce the overall risk. In other
words a high correlation will reduce the risk associated with a spark spread option,
as well as reducing the potential upside.
US power markets create particularly complex correlations between power delivered
at different points. There are between twelve and fifteen actively traded hubs in the
USA that show a wide range of correlations, depending on the ability to transfer
power between the regions and similarity of weather conditions. For instance, there
is little or no correlation between COB (California / Oregon border) and Cinergy (near
Cincinnati), while a relatively close correlation is seen between COB and Palo Verde
(in southern California). A matrix of correlations for power was shown in Figure 18.4.
The inherent problem of relying on correlations to hedge risk is that they are prone
to break down particularly during significant market events. For instance, when prices
rose in Cinergy in June 1998 many participants wanted to transport power from the
PJM region to cover their positions. This proved impossible given PJM’s high demand
conditions and the correlations between PJM and Cinergy broke down, leaving some
players with significant losses despite their ‘hedged’ position. Even under ‘normal’
market conditions correlations can vary significantly over time and risk managers
need to be very careful of traders leveraging what seem to be stable correlations.
As stated above, it is also common within the power sector to treat different forward
months as individual products. If this approach is taken it is necessary to link the
contracts by their estimated correlations. While this solves one problem, it creates
another computational one given the resultant matrix and computation necessary to
keep the correlations up to date.
cross-correlation between energy products this becomes crucial in estimating the
overall value and risk profile for a particular position.
As stated above, oil correlations are actively traded as crack spreads, locational
spreads and time spreads. Similarly, ‘spark spreads’ between gas and power are
commonly traded and these create a synthetic gas-fired power station. The higher
the correlation between the input gas costs and output electricity prices, the easier
it is to hedge power prices with gas prices and thus reduce the overall risk. In other
words a high correlation will reduce the risk associated with a spark spread option,
as well as reducing the potential upside.
US power markets create particularly complex correlations between power delivered
at different points. There are between twelve and fifteen actively traded hubs in the
USA that show a wide range of correlations, depending on the ability to transfer
power between the regions and similarity of weather conditions. For instance, there
is little or no correlation between COB (California / Oregon border) and Cinergy (near
Cincinnati), while a relatively close correlation is seen between COB and Palo Verde
(in southern California). A matrix of correlations for power was shown in Figure 18.4.
The inherent problem of relying on correlations to hedge risk is that they are prone
to break down particularly during significant market events. For instance, when prices
rose in Cinergy in June 1998 many participants wanted to transport power from the
PJM region to cover their positions. This proved impossible given PJM’s high demand
conditions and the correlations between PJM and Cinergy broke down, leaving some
players with significant losses despite their ‘hedged’ position. Even under ‘normal’
market conditions correlations can vary significantly over time and risk managers
need to be very careful of traders leveraging what seem to be stable correlations.
As stated above, it is also common within the power sector to treat different forward
months as individual products. If this approach is taken it is necessary to link the
contracts by their estimated correlations. While this solves one problem, it creates
another computational one given the resultant matrix and computation necessary to
keep the correlations up to date.
GARCH
The GARCH structures appear to be the most flexible and appropriate framework
providing they are developed to adjust for the factors outlined above. These tend to
hold up well under normal conditions with the energy markets placing particularly
strong weighting to the most recent data. Where appropriate these models can be
developed to capture the mean reversion and seasonality characteristics (although
calibration is a more difficult problem).
However, the non-normal distribution issue is more difficult to solve. The principal
approach being used by many parties to date is through the use of historical
distributions in Monte Carlo, primarily as a stress test.
providing they are developed to adjust for the factors outlined above. These tend to
hold up well under normal conditions with the energy markets placing particularly
strong weighting to the most recent data. Where appropriate these models can be
developed to capture the mean reversion and seasonality characteristics (although
calibration is a more difficult problem).
However, the non-normal distribution issue is more difficult to solve. The principal
approach being used by many parties to date is through the use of historical
distributions in Monte Carlo, primarily as a stress test.
Volatility models and model risk
Given the large number of specialist volatility models available to use and the
numerous issues set out above it is impossible to review those that are most
appropriate to the energy markets without becoming immersed in the technical
details that are best left to the specialist academics and financial engineers. It is
worth stating that, to date, the modeling challenges associated with the power
markets have not been fully resolved and as such no model is fully adequate. The
important factor given the dynamic nature of the energy markets is to have a model
that can be continually updated from historical data and flexible enough to add in
manual adjustments as necessary. The challenge is to find the model that closest
meets your needs and which does the least amount of harm.
numerous issues set out above it is impossible to review those that are most
appropriate to the energy markets without becoming immersed in the technical
details that are best left to the specialist academics and financial engineers. It is
worth stating that, to date, the modeling challenges associated with the power
markets have not been fully resolved and as such no model is fully adequate. The
important factor given the dynamic nature of the energy markets is to have a model
that can be continually updated from historical data and flexible enough to add in
manual adjustments as necessary. The challenge is to find the model that closest
meets your needs and which does the least amount of harm.
Historical versus implied volatilities
A simple answer to many of the above problems is to use implied rather than
historical volatilities. This provides a particularly useful solution for monitoring
option positions in liquid markets. However, they are less useful for the less liquid
positions for applying VaR in energy.
The principal reason for this is the level of liquidity in the option market. Without
a full-volatility ‘smile’ (i.e. the volatilities across a range of in- and out-of-the-money
strike prices) and option pricing for a range of contract terms it is difficult to see fully
how the market perceives volatility. For instance, you would want to break out the
volatility most consistent with your VaR confidence interval for each forward contact.
While this is possible in very liquid products such as crude oil it is difficult in gas
and impossible in power.
The other problem is ensuring a consistency when using volatilities within models
such as covariance matrices. Unless all your volatilities are calculated on a consistent
basis, cross-correlations will not be consistent.
Given the option liquidity problems it is generally easier to use historical volatilities
and ‘test them’ against the implied numbers that are available from the market.
From this you can recalibrate the volatility models to the market numbers rather
than fit to historical numbers, probably as a stress test.
historical volatilities. This provides a particularly useful solution for monitoring
option positions in liquid markets. However, they are less useful for the less liquid
positions for applying VaR in energy.
The principal reason for this is the level of liquidity in the option market. Without
a full-volatility ‘smile’ (i.e. the volatilities across a range of in- and out-of-the-money
strike prices) and option pricing for a range of contract terms it is difficult to see fully
how the market perceives volatility. For instance, you would want to break out the
volatility most consistent with your VaR confidence interval for each forward contact.
While this is possible in very liquid products such as crude oil it is difficult in gas
and impossible in power.
The other problem is ensuring a consistency when using volatilities within models
such as covariance matrices. Unless all your volatilities are calculated on a consistent
basis, cross-correlations will not be consistent.
Given the option liquidity problems it is generally easier to use historical volatilities
and ‘test them’ against the implied numbers that are available from the market.
From this you can recalibrate the volatility models to the market numbers rather
than fit to historical numbers, probably as a stress test.
Price Spikes
Price spikes occur because ultimately the demand for most energy products and
power in particular is inelastic. This means that once supply becomes scarce, prices
potentially need to rise to astronomical levels before there is a demand response. In
many cases consumers never see such price signals, meaning that prices could rise
to any level and they would continue consuming.
A major issue in the power market (and to a lesser extent in the other energy
markets) is that of price spikes within the spot market. These are of particular
concern for risk managers since they will blow through previously realistic Value-at-
Risk limits extremely quickly.
The ‘jump diffusion’ process provides some structure to capture this within a
quantitative structure. This requires three factors to estimate the impacts of spikes:
number, height and the duration. Traditionally, the three factors would be taken
from historical information, but risk managers may well want to change the factors
based on other fundamental analysis. For instance, simulation models can provide
an estimate of the likelihood of spikes based on the likelihood of supply shortages.
The duration of spikes can be problematical. Generally, spikes mean revert very
quickly. However, when prices go to extremely high levels liquidity and ‘shock’ factors
can see a significant widening of bid/ask spreads and leave prices ‘stranded’ at high
levels for some considerable time. Stress testing for different durations of spikes is
thus advisable when considering the maximum potential exposures.
An interesting affect of spikes is that you get very pronounced smile effects in
energy volatility. This is due to the fact that once prices rise above the normal levels
there may be no limit to the ultimate level. To take an example, say prices normally
range between $15/MWh and $40/MWh, occasionally moving up to $150/MWh.
When prices rise above $150/MWh you are beyond the ‘normal’ supply/demand
conditions. Given the inelastic demand, the probability of prices spiking to $5000/
MWh is similar to the probability of prices spiking to $500/MWh. Thus, while the
extrinsic value of a $100 call will be significantly lower than a $40 call; the extrinsic
value of $500 and $1000 calls may be very similar. When translating these back into
implied volatilities the result is very pronounced smile effects.
power in particular is inelastic. This means that once supply becomes scarce, prices
potentially need to rise to astronomical levels before there is a demand response. In
many cases consumers never see such price signals, meaning that prices could rise
to any level and they would continue consuming.
A major issue in the power market (and to a lesser extent in the other energy
markets) is that of price spikes within the spot market. These are of particular
concern for risk managers since they will blow through previously realistic Value-at-
Risk limits extremely quickly.
The ‘jump diffusion’ process provides some structure to capture this within a
quantitative structure. This requires three factors to estimate the impacts of spikes:
number, height and the duration. Traditionally, the three factors would be taken
from historical information, but risk managers may well want to change the factors
based on other fundamental analysis. For instance, simulation models can provide
an estimate of the likelihood of spikes based on the likelihood of supply shortages.
The duration of spikes can be problematical. Generally, spikes mean revert very
quickly. However, when prices go to extremely high levels liquidity and ‘shock’ factors
can see a significant widening of bid/ask spreads and leave prices ‘stranded’ at high
levels for some considerable time. Stress testing for different durations of spikes is
thus advisable when considering the maximum potential exposures.
An interesting affect of spikes is that you get very pronounced smile effects in
energy volatility. This is due to the fact that once prices rise above the normal levels
there may be no limit to the ultimate level. To take an example, say prices normally
range between $15/MWh and $40/MWh, occasionally moving up to $150/MWh.
When prices rise above $150/MWh you are beyond the ‘normal’ supply/demand
conditions. Given the inelastic demand, the probability of prices spiking to $5000/
MWh is similar to the probability of prices spiking to $500/MWh. Thus, while the
extrinsic value of a $100 call will be significantly lower than a $40 call; the extrinsic
value of $500 and $1000 calls may be very similar. When translating these back into
implied volatilities the result is very pronounced smile effects.
Non-normal distributions
One of the keys to successfully applying volatility estimation and the stochastic
process is some form of normal (or log-normal) distribution. We cannot assume this
in the energy spot markets. Figure 18.8 shows a typical distribution of spot power
prices. As you can observe prices are leptokurtic rather than normal in nature. This
causes major problems for most of the closed-form models. It is thus necessary to
test continuously for normality assumptions within the energy markets.
The solutions to this are more problematical and are likely to lead to using historic
or ‘adjusted’ historic data within a Monte Carlo simulation model to obtain realistic
representations. Luckily most of the forward markets in energy are much more
‘normal’ in nature than their spot counterparts and are much easier to correct.
process is some form of normal (or log-normal) distribution. We cannot assume this
in the energy spot markets. Figure 18.8 shows a typical distribution of spot power
prices. As you can observe prices are leptokurtic rather than normal in nature. This
causes major problems for most of the closed-form models. It is thus necessary to
test continuously for normality assumptions within the energy markets.
The solutions to this are more problematical and are likely to lead to using historic
or ‘adjusted’ historic data within a Monte Carlo simulation model to obtain realistic
representations. Luckily most of the forward markets in energy are much more
‘normal’ in nature than their spot counterparts and are much easier to correct.
Seasonality
As we discussed earlier, energy markets are inherently seasonal in nature. There are
a number of ways to deal with this from fitting a sin/cosin curve to ‘normalization’
factors. The challenge is that no year is ‘normal’, which means that last year will be
very different from next year. At least 10–20 years’ data is needed to see any normal
trend. Unfortunately, in most cases, the power markets are not mature enough to
provide adequate pricing data.
Given this lack of data, the alternative is to try to model spot price volatility as a
function of weather data, which can then be cast into a normal year. Such models
can be useful providing the underlying production costs are similar in the study
years. Unfortunately this is seldom the case.
Seasonality can present itself in a number of forms.
Ω An increase in the volatility of the entire curve during certain times of the year
Ω An increase in a segment of the curve during certain times
Ω Parts of the curve always being more volatile.
Unfortunately, in power, we see all three effects.
Two approaches can provide useful outcomes. First, adjusting volatility for monthly
seasonal factors can provide useful results in most of the oil and gas markets. This
deals with the fact that the entire curve tends to be more volatile at certain times of
the year. Second, treating each month as an independent price variable can provide
more useful results in the power market, and to a lesser extent, in gas. This allows
for the fact that, say, summer is generally more volatile than spring and that their
volatilities are normally independent.
This latter approach abandons the use of a single forward curve for the actively
traded forward markets (up to 24 months), given the lack of storage and therefore
correlation linkage between forward months (or quarters). The linkages between the
months can then be rebuilt through a correlation matrix in the same way as linkages
are drawn between different products. This brings with it significant data and sizing
issues. Solving these seasonality issues is key to effective volatility estimation in
energy.
a number of ways to deal with this from fitting a sin/cosin curve to ‘normalization’
factors. The challenge is that no year is ‘normal’, which means that last year will be
very different from next year. At least 10–20 years’ data is needed to see any normal
trend. Unfortunately, in most cases, the power markets are not mature enough to
provide adequate pricing data.
Given this lack of data, the alternative is to try to model spot price volatility as a
function of weather data, which can then be cast into a normal year. Such models
can be useful providing the underlying production costs are similar in the study
years. Unfortunately this is seldom the case.
Seasonality can present itself in a number of forms.
Ω An increase in the volatility of the entire curve during certain times of the year
Ω An increase in a segment of the curve during certain times
Ω Parts of the curve always being more volatile.
Unfortunately, in power, we see all three effects.
Two approaches can provide useful outcomes. First, adjusting volatility for monthly
seasonal factors can provide useful results in most of the oil and gas markets. This
deals with the fact that the entire curve tends to be more volatile at certain times of
the year. Second, treating each month as an independent price variable can provide
more useful results in the power market, and to a lesser extent, in gas. This allows
for the fact that, say, summer is generally more volatile than spring and that their
volatilities are normally independent.
This latter approach abandons the use of a single forward curve for the actively
traded forward markets (up to 24 months), given the lack of storage and therefore
correlation linkage between forward months (or quarters). The linkages between the
months can then be rebuilt through a correlation matrix in the same way as linkages
are drawn between different products. This brings with it significant data and sizing
issues. Solving these seasonality issues is key to effective volatility estimation in
energy.
Mean reversion
We discussed earlier that energy prices tend to mean revert, i.e. to move back to
some mean over time. A number of models have been developed to adjust for mean
reversion. The crucial components that are needed to adjust in mean reversion
models are the mean itself and the speed of reversion. Two basic forms of mean
reversion occur: short run and long run.
Short-run mean reversion occurs when the particular events that have ‘knocked’
prices away from their mean recede. In power, this seems to occur between three to
five days after the event, which is generally weather driven. Long-run mean reversion
is driven more by the economics of the industry. For instance, the ability to bring
additional production on line or shut down uneconomic production will tend to put
caps and floors on prices over longer-term periods. However, such supply/demand
mismatches can last for periods of up to two or three years.
Additionally, given the extreme market movements in the power markets up- and
down-volatilities are not symmetric. Down-volatilities will be high when prices are
high and up-volatilities will be high when prices are low. This will have a material
impact on the pricing models. The impact on VaR will, however, depend on the
holding period being used. We have generally observed reversion taking five days or
longer, resulting in a minimal impact on a 5-day or less holding period VaR.
some mean over time. A number of models have been developed to adjust for mean
reversion. The crucial components that are needed to adjust in mean reversion
models are the mean itself and the speed of reversion. Two basic forms of mean
reversion occur: short run and long run.
Short-run mean reversion occurs when the particular events that have ‘knocked’
prices away from their mean recede. In power, this seems to occur between three to
five days after the event, which is generally weather driven. Long-run mean reversion
is driven more by the economics of the industry. For instance, the ability to bring
additional production on line or shut down uneconomic production will tend to put
caps and floors on prices over longer-term periods. However, such supply/demand
mismatches can last for periods of up to two or three years.
Additionally, given the extreme market movements in the power markets up- and
down-volatilities are not symmetric. Down-volatilities will be high when prices are
high and up-volatilities will be high when prices are low. This will have a material
impact on the pricing models. The impact on VaR will, however, depend on the
holding period being used. We have generally observed reversion taking five days or
longer, resulting in a minimal impact on a 5-day or less holding period VaR.
Why energy is different – spot versus term markets
An important aspect of many commodity prices and energy in particular is the
significant increase in volatility as a contract nears expiry. In many ways the nature
of spot price movement is very different from those seen in the forward markets.
Spot prices are driven by observable short-run market price signals including
actual supply and demand conditions. In power this will include plant availability
and weather forecasts that provide a good proxy for electricity demand. Prices at this
stage are likely to reduce to ‘floor’ levels or increase rapidly if supply conditions look
like they are going to be particularly tight. This will occur whether you have a
managed spot market, such as an electricity pool, or a pure OTC spot market. Given
these movements away from the ‘expected value’ in the forwards volatility will rise
dramatically as you move from the prompt month to the spot month. Figure 18.7
shows movement of volatilities as you move towards contact expiry, ending with
extremely high volatility compared to other markets.
One consequence of this price movement characteristic is that VaR estimates will
increase rapidly for any contracts that are taken to expiration. This means that risk
managers need to both understand the trading strategy being employed and ensure
that contracts that are not easily trading prior to expiration (such as some option
structures) are treated very carefully.
significant increase in volatility as a contract nears expiry. In many ways the nature
of spot price movement is very different from those seen in the forward markets.
Spot prices are driven by observable short-run market price signals including
actual supply and demand conditions. In power this will include plant availability
and weather forecasts that provide a good proxy for electricity demand. Prices at this
stage are likely to reduce to ‘floor’ levels or increase rapidly if supply conditions look
like they are going to be particularly tight. This will occur whether you have a
managed spot market, such as an electricity pool, or a pure OTC spot market. Given
these movements away from the ‘expected value’ in the forwards volatility will rise
dramatically as you move from the prompt month to the spot month. Figure 18.7
shows movement of volatilities as you move towards contact expiry, ending with
extremely high volatility compared to other markets.
One consequence of this price movement characteristic is that VaR estimates will
increase rapidly for any contracts that are taken to expiration. This means that risk
managers need to both understand the trading strategy being employed and ensure
that contracts that are not easily trading prior to expiration (such as some option
structures) are treated very carefully.
Volatility estimation – models
As discussed above, energy commodity prices show both high and variable volatility.
This characteristic poses particular difficulty for the standard volatility models. The
starting point for estimation of volatility is the standard constant volatility models as
applied within the Black and Scholes model (1973). This allows us to incorporate the
standard Brownian motion assumptions14 in and expand them over a continuous
time period using the square-root (uncorrelated) assumption.15
Many simplistic models have taken this assumption to estimate volatility in the
energy markets. Given the lack of implied volatility estimates, the most common
have taken simple average volatility (for instance, the average of the last six months’
spot data) and projected it forward to create an annualized number which can be
later converted for option pricing models or holding period for Value-at-Risk estimation
(Figure 18.5).
A slight improvement to is to use a weighted average approach that weights the
more recent events more heavily. This approach will undoubtedly improve the events
but will not take account of the ‘events’ problem in energy (such as seasonality) or
mean reversion. In other words the constant forward volatility assumption does not
hold.
As can be seen from Figure 18.6 and as confirmed by the statistical tests, significant
clustering occurs around the seasons and particular events. So while the basic
stochastic ‘random’ toolkit with constant volatility provides a useful starting point it
is not helpful in estimating the volatility parameters in energy. In fact, a number of
issues need to be overcome before the quantitative models can be used effectively in
the energy markets. These are addressed below.
This characteristic poses particular difficulty for the standard volatility models. The
starting point for estimation of volatility is the standard constant volatility models as
applied within the Black and Scholes model (1973). This allows us to incorporate the
standard Brownian motion assumptions14 in and expand them over a continuous
time period using the square-root (uncorrelated) assumption.15
Many simplistic models have taken this assumption to estimate volatility in the
energy markets. Given the lack of implied volatility estimates, the most common
have taken simple average volatility (for instance, the average of the last six months’
spot data) and projected it forward to create an annualized number which can be
later converted for option pricing models or holding period for Value-at-Risk estimation
(Figure 18.5).
A slight improvement to is to use a weighted average approach that weights the
more recent events more heavily. This approach will undoubtedly improve the events
but will not take account of the ‘events’ problem in energy (such as seasonality) or
mean reversion. In other words the constant forward volatility assumption does not
hold.
As can be seen from Figure 18.6 and as confirmed by the statistical tests, significant
clustering occurs around the seasons and particular events. So while the basic
stochastic ‘random’ toolkit with constant volatility provides a useful starting point it
is not helpful in estimating the volatility parameters in energy. In fact, a number of
issues need to be overcome before the quantitative models can be used effectively in
the energy markets. These are addressed below.
The quantitative/statistical approach
As many have learned the hard way, forecasts, regardless of how sophisticated, are
always wrong.13 So it is not surprising that when the financial markets offered up an
alternative approach it was readily embraced by the energy market players. In other
words rather than try to estimate what the market might do and the implications of
such a move, take the market’s best guess at where future prices are going and then
focus on estimating how wrong it is likely to be.
The two principal components used for this have been the forward curve price
discovery and volatility estimation.
always wrong.13 So it is not surprising that when the financial markets offered up an
alternative approach it was readily embraced by the energy market players. In other
words rather than try to estimate what the market might do and the implications of
such a move, take the market’s best guess at where future prices are going and then
focus on estimating how wrong it is likely to be.
The two principal components used for this have been the forward curve price
discovery and volatility estimation.
Scenario analysis
As stated at the beginning of this chapter, many observers have been assessing energy
market risk for a significant time. In addition, large numbers of consultantcy firms
who specialize in the forecasting of energy prices to aid companies make investment
or other business planning decisions. In addition, large energy companies will spend
significant amounts of time and money determining their own views of the world.
Most will provide a range of possible scenarios recognizing the difficulty of providing
an accurate forecast and a business can estimate the impact of both positive and not
so positive scenarios. In other words a risk assessment of adverse price movements is
made on a heuristic ‘scenario’-based approach using some form of economic model.
Where no discernible forward curves can be established this is still the most common
form of market risk assessment. One advantage of this approach is that it can link
into known market indicators. For instance, it can take a gas and oil forward curve
and help ‘translate’ it into a power curve. As such, the models can be used as tools
to provide a number of different types of forecasts, and to estimate:
Ω A pure arbitrage model
Ω A new entrant model (assuming ‘potential’ new entrants set forward prices)
Ω A macro model incorporating political, environmental, regulatory and economic
model (e.g. a ‘game theory’ model)
Ω A micro ‘dispatch’ optimization model
The advantage of such an approach is that it can incorporate the important
regulatory or market power issues that can dominate many regional energy markets.
For instance, models that allow for the gradual deregulation of the local distribution
companies and the transition away from cost plus pricing can provide useful insights
into future price movements.
The important aspect of scenario analysis is consistency in the ‘story’ being told
and ‘buy-in’ from senior management early in the process. As such, you need to
establish that the scenarios are plausible, cover the key market risks to the business
and then go through the painstaking process to ensure internal consistency.
market risk for a significant time. In addition, large numbers of consultantcy firms
who specialize in the forecasting of energy prices to aid companies make investment
or other business planning decisions. In addition, large energy companies will spend
significant amounts of time and money determining their own views of the world.
Most will provide a range of possible scenarios recognizing the difficulty of providing
an accurate forecast and a business can estimate the impact of both positive and not
so positive scenarios. In other words a risk assessment of adverse price movements is
made on a heuristic ‘scenario’-based approach using some form of economic model.
Where no discernible forward curves can be established this is still the most common
form of market risk assessment. One advantage of this approach is that it can link
into known market indicators. For instance, it can take a gas and oil forward curve
and help ‘translate’ it into a power curve. As such, the models can be used as tools
to provide a number of different types of forecasts, and to estimate:
Ω A pure arbitrage model
Ω A new entrant model (assuming ‘potential’ new entrants set forward prices)
Ω A macro model incorporating political, environmental, regulatory and economic
model (e.g. a ‘game theory’ model)
Ω A micro ‘dispatch’ optimization model
The advantage of such an approach is that it can incorporate the important
regulatory or market power issues that can dominate many regional energy markets.
For instance, models that allow for the gradual deregulation of the local distribution
companies and the transition away from cost plus pricing can provide useful insights
into future price movements.
The important aspect of scenario analysis is consistency in the ‘story’ being told
and ‘buy-in’ from senior management early in the process. As such, you need to
establish that the scenarios are plausible, cover the key market risks to the business
and then go through the painstaking process to ensure internal consistency.
Estimating market risk
Before we jump into the issues surrounding the difficulties of estimating risk within
the energy sector it is worth putting the different products into some framework with
respect to liquidity. Let us compare the volume, open interest and the term of the
forward contracts in the products traded on the NYMEX (see Table 18.1).
Table 18.1 Products traded on the NYMEX (as of 14 June 1999)
Daily volume Open interest Latest forward
(contracts) (contracts) contract
Crude oil – WTI 51 456 120 068 Dec-05
Heating oil – NY Harbor 15 228 38 815 Dec-00
Gasoline – NY Harbor 15 623 41 230 Jun-00
Natural gas – Henry Hub 23 931 67 860 May-02
Electricity – Palo Verde 121 989 Dec-00
Electricity – COB 144 1 177 Sep-00
Electricity – Cinergy 98 527 Aug-00
Electricity – Entergy 12 343 Feb-00
Electricity – PJM 9 23 Oct-99
Electricity – TVA 0 0 Sep-99
Electricity – ComEd 0 0 Sep-99
Note: The electricity contracts are relatively new: the last contract was only launched
in March 1999. It is unlikely that all the contracts will survive.
Source: NYMEX,CBOT
Given the infancy of the market, only limited liquidity exists in some of the power
products, even where the OTC market has been robust enough to launch a futures
contract. Any price data that is used must therefore be checked closely before
conclusions can be drawn. This is particularly important when looking back at a
significant period of time – many of the electricity markets simply did not exist more
than a year or so ago.
the energy sector it is worth putting the different products into some framework with
respect to liquidity. Let us compare the volume, open interest and the term of the
forward contracts in the products traded on the NYMEX (see Table 18.1).
Table 18.1 Products traded on the NYMEX (as of 14 June 1999)
Daily volume Open interest Latest forward
(contracts) (contracts) contract
Crude oil – WTI 51 456 120 068 Dec-05
Heating oil – NY Harbor 15 228 38 815 Dec-00
Gasoline – NY Harbor 15 623 41 230 Jun-00
Natural gas – Henry Hub 23 931 67 860 May-02
Electricity – Palo Verde 121 989 Dec-00
Electricity – COB 144 1 177 Sep-00
Electricity – Cinergy 98 527 Aug-00
Electricity – Entergy 12 343 Feb-00
Electricity – PJM 9 23 Oct-99
Electricity – TVA 0 0 Sep-99
Electricity – ComEd 0 0 Sep-99
Note: The electricity contracts are relatively new: the last contract was only launched
in March 1999. It is unlikely that all the contracts will survive.
Source: NYMEX,CBOT
Given the infancy of the market, only limited liquidity exists in some of the power
products, even where the OTC market has been robust enough to launch a futures
contract. Any price data that is used must therefore be checked closely before
conclusions can be drawn. This is particularly important when looking back at a
significant period of time – many of the electricity markets simply did not exist more
than a year or so ago.
6 Temmuz 2011 Çarşamba
Other energy-related products – emissions and weather
Two other trading markets have been recently developing that complement energy
trading in the USA: emissions and weather derivatives. Emissions trading become
possible with the introduction of tradable tickets for Sulfur Dioxide (SO2) on a
national basis and Nitrous Oxide (NOx ) on a regional basis. Producers must have
enough allowances to cover production of SO2 and NOx (a by-product from fossil fuelburning
stations, particularly oil and coal stations) and can sell excess allowances to
those requiring additional allowances. The value of such allowances now represents a
significant proportion of production costs and as such have a direct impact on the
pricing in the fuel oil and coal markets. The emissions market is relatively active
through a brokered OTC market and a small number of options have been traded.
Weather derivatives also play an interesting role in the energy markets that are
particularly weather sensitive. A utility’s revenue is obviously dependent on its unit
sales, which are sensitive to the weather. Given this risk, weather derivatives are now
being used by a number of utilities to hedge their exposure to temperature and other
weather-related factors. In particular, swaps against heating degree-days are now
becoming common, and options against, snow fall, snow pack (a price driver for hydrobased
systems), river flow and other weather-related factors have been seen. A number
of energy traders now actively buy and sell weather derivatives which are also of
interest to insurance companies and have obvious applicability beyond energy players.
To date, the market is still at the early stages both in the USA and Europe and contracts
have tended to have a limited payoff structure (a limit on the maximum payout). 536
trading in the USA: emissions and weather derivatives. Emissions trading become
possible with the introduction of tradable tickets for Sulfur Dioxide (SO2) on a
national basis and Nitrous Oxide (NOx ) on a regional basis. Producers must have
enough allowances to cover production of SO2 and NOx (a by-product from fossil fuelburning
stations, particularly oil and coal stations) and can sell excess allowances to
those requiring additional allowances. The value of such allowances now represents a
significant proportion of production costs and as such have a direct impact on the
pricing in the fuel oil and coal markets. The emissions market is relatively active
through a brokered OTC market and a small number of options have been traded.
Weather derivatives also play an interesting role in the energy markets that are
particularly weather sensitive. A utility’s revenue is obviously dependent on its unit
sales, which are sensitive to the weather. Given this risk, weather derivatives are now
being used by a number of utilities to hedge their exposure to temperature and other
weather-related factors. In particular, swaps against heating degree-days are now
becoming common, and options against, snow fall, snow pack (a price driver for hydrobased
systems), river flow and other weather-related factors have been seen. A number
of energy traders now actively buy and sell weather derivatives which are also of
interest to insurance companies and have obvious applicability beyond energy players.
To date, the market is still at the early stages both in the USA and Europe and contracts
have tended to have a limited payoff structure (a limit on the maximum payout). 536
Coal
Coal is often seen as the poor relation in the energy market and the emergence of a
liquid commodity trading market in coal has been very sluggish. One reason for this
is that the market has traditionally sold bespoke long-term contracts to utilities.
These were very specific with respect to quality since many power station boilers
were designed to take particular quality specifications. This was more than just
sulfur and calorific differences, but would also include specifications for chlorine,
ash content, hardness (grindability), sodium, ash fusion temperature, volatile matter,
moisture and others.
Another issue rests with the transport of coal. Coal is an expensive energy product
to transport. For instance, although the cost of mine-mouth coal from an open-cast
area can be as low as $3/tonnes FOB, the delivered cost of that coal could be closer
to $30/tonne. While there is a competitive shipping market for seaborne trade,
railroads are often oligopolistic in nature.
Finally, the market has been traditionally oversupplied by a large number of
fragmented producers. This has led to extreme price competition and a market
traditionally seeing a backwardated forward market. As a result, annualized historical
volatility in this market is very low (10–15%). In addition, the lack of liquidity in
standard products means clear forward curves cannot be easily found.
The coal market is, however, slowly changing. Faced with the challenges of deregulation
and environmental constraints, utilities are moving to buying shorter-term,
more standard coal qualities. These forces have led to some industry and the
development of a coal trading market that is still in its infancy.
A reasonable international trading market exists and this is being complemented
by standardized contracts in the USA. ‘Hubs’ are developing in Powder River Basin,
Appalachia, Illinois Basin, Colorado-Utah and Pittsburgh Seam. Evidencing this
trend, NYMEX has proposed launching a coal contract based on the Big Sandy River
(Appalachia) in 1999.
Some early option and swap trading has taken place with indices being priced off
publications such as Coal Daily in the USA. However, liquidity in these products has
been very limited and, to date, they have not proved to be useful hedging mechanisms
for major producers or consumers.
liquid commodity trading market in coal has been very sluggish. One reason for this
is that the market has traditionally sold bespoke long-term contracts to utilities.
These were very specific with respect to quality since many power station boilers
were designed to take particular quality specifications. This was more than just
sulfur and calorific differences, but would also include specifications for chlorine,
ash content, hardness (grindability), sodium, ash fusion temperature, volatile matter,
moisture and others.
Another issue rests with the transport of coal. Coal is an expensive energy product
to transport. For instance, although the cost of mine-mouth coal from an open-cast
area can be as low as $3/tonnes FOB, the delivered cost of that coal could be closer
to $30/tonne. While there is a competitive shipping market for seaborne trade,
railroads are often oligopolistic in nature.
Finally, the market has been traditionally oversupplied by a large number of
fragmented producers. This has led to extreme price competition and a market
traditionally seeing a backwardated forward market. As a result, annualized historical
volatility in this market is very low (10–15%). In addition, the lack of liquidity in
standard products means clear forward curves cannot be easily found.
The coal market is, however, slowly changing. Faced with the challenges of deregulation
and environmental constraints, utilities are moving to buying shorter-term,
more standard coal qualities. These forces have led to some industry and the
development of a coal trading market that is still in its infancy.
A reasonable international trading market exists and this is being complemented
by standardized contracts in the USA. ‘Hubs’ are developing in Powder River Basin,
Appalachia, Illinois Basin, Colorado-Utah and Pittsburgh Seam. Evidencing this
trend, NYMEX has proposed launching a coal contract based on the Big Sandy River
(Appalachia) in 1999.
Some early option and swap trading has taken place with indices being priced off
publications such as Coal Daily in the USA. However, liquidity in these products has
been very limited and, to date, they have not proved to be useful hedging mechanisms
for major producers or consumers.
1 Temmuz 2011 Cuma
Electricity/power
The electricity market (often described in the USA as the power market) is very
different for one fundamental reason: both storage and transportation are incredibly
expensive.10 Let us briefly describe the nature of the power market from a trading
and risk management perspective.
First, like many products, electricity demand varies significantly throughout the
day, week and year. However, electricity has the same properties as a highly perishable
product in that not only must supply meet demand, but production must meet
demand minute by minute. The result of this unique attribute is that the market
must keep a significant amount of idle capacity in place for start-up when the
demand is there and shut-down when demand recedes. In some cases, this is a
generation station that is already running, ready to meet anticipated customer
demand. In other cases it is plant that may only be asked to start up once every few
years. Typically a local market (or grid system) will keep 15–20% more plant capacity
than it expects to use on the highest hour of demand during a normal year. This will
often represent over double the average demand.
This, coupled with the physical challenges of transporting electricity,11 leads to a
general position of over-supply combined with short periods when the normally idle
plant needs to run. At such peak times, when all the capacity on the system is
needed, spot market prices12 will rise dramatically as plant owners will need to
recover not only their higher cost of running but also their capital costs in a relatively
short period of time. This is exacerbated by the fact that electricity generation is one
of the most capital-intensive industries in the world.
The forward market will, of course, smooth this by assigning probabilities to the
likelihood of the high prices. Higher probabilities are obviously assigned during the
periods when demand is likely to be at a peak and this will generally only occur
during two or three months of the year. In the USA this is generally in the summer.
Thus unless there is a huge over-supply the summer prices will be significantly
higher than the rest of the year. When there is a potential shortage prices will be
dramatically higher, as was seen in the Cinergy market in late June 1998 where
daily prices that trade most of the year at $30/MWh increased to $7000/MWh.
It is these important market characteristics that lead to the extreme seasonality,
jumps, spikes and mean reversion that will be discussed below in the section on
Market Risk. Given the extreme nature of these factors trying to capture them on
one term structure or convenience yield is extremely difficult.
Despite this price uncertainty, a forward market for electricity has developed along
a standard commodity structure. In fact seven exchange contracts, reflecting the
regional nature of power, currently exist at Palo Verde, California/Oregon Border
(COB), Entergy, Cinergy (NYMEX), TVA, ComEd (CBOT) and Twin Cities (M. Grain
Exchange) and PJM. The forward curves for Cinergy is shown in Figure 18.2 compared
to the Henry Hub gas curve.
Figure 18.2 Forward curves for Cinergy compared to the Henry Hub gas curve. (Source: Citizens
Power,June 1999)
Some standard options are traded at the most liquid hubs. These tend to be ‘strips’
of daily European calls based on monthly blocks. However, the bid/ask on such
products are often wide and the depth of liquidity very limited.
Forward curve price discovery – the problems with power
Let us focus on two major differences in power:
Ω Storage First, you have virtually no stack and roll storage arbitrage. In other
words, since you cannot keep today’s power for tomorrow there is no primary
‘arbitrage’ linkage between today’s price and tomorrow’s. There are, of course,
many secondary links. The underlying drivers are likely to be similar – demand,
plant availability, fuel costs, traders’ expectations and general market environment.
But as you move forward in time, secondary links break down quickly and
so does the price relationship. Figure 18.3 shows the forward correlation between
an April contract and the rest of the year. As you can see, almost no relationship
exists between April and October and most of the relationship has evaporated
once you are beyond one month. In other words, the October Cinergy contract
has no more relationship with the April contract than, say, an oil, gas or even
interest rate market.
Ω Transportation Second, you have limited ‘hub basis’ arbitrage. Since there are
numerous logistical limitations on moving electricity it is difficult to arbitrage
between many of the power markets within the USA (never mind internationally).
Even hubs that are relatively close show large variations in the spread between
prices. Figure 18.4 shows some of the correlations between major power hubs in
the USA. Rather than thinking of them as one market, it is more accurate to view
the power market as at least twelve (the final number of hubs is still being
determined by the marketplace) independent markets with some but often little
relationship.
Once you put all these factors together you see a picture similar to the one a global
risk managers in a big bank will have experienced, a huge number of independent
products that need to be combined for risk purposes. Instead of having one forward
curve for US Power, we have up to eighteen independent months for twelve independent
markets, in other words 216 products. This brings both the curse of lack of
liquidity and data integrity for each product and, on the positive side from a risk
perspective, diversity.
Forward curves up to 24 months are traditionally built using daily trader/broker
marks for monthly peak/off-peak prices, with the breakdown, where necessary, into
smaller time periods (down to an hourly profile) using historical prices adjusted for
normal weather conditions. Prices beyond two years are significantly less liquid and
where information is available bid/ask spreads can increase significantly. Forward
price curves (and volatility curves) beyond 24 months thus need to be created through
more of a mark to model rather than mark to market process. As noted above, using
a model to connect price quotes inevitably involved a number of assumptions about
how the market behaves. In power, this involves not just fitting a serious of different
price quotes together, but also filling in the gaps where price quotes are not available.
The price structure will have to make certain model assumptions based on historical
observation about seasonality and the year-to-year transition process. Models
can be bought (such as the SAVA forward curve builder) or, more often, built inhouse.
However, given the developing nature of the market this still tends to be a
relatively manual process to ensure all the relevant market information can be input
into the curve and minimize the error terms.
It needs to be remembered that in making assumptions about the structure of the
curve in this process to estimate the fair value of a transaction that the forward
curve, while objective, unbiased and arbitrage-free, may be unreliable given the
incomplete data sets. Throughout the process it is thus necessary to estimate the
impact of such assumptions and modeling or prudency reserves are likely to need to
be applied against the fair value under these circumstances. The collection of market
data with an illiquid market also becomes a major operational issue with a need to
continuously verify and search for independent data. 534
different for one fundamental reason: both storage and transportation are incredibly
expensive.10 Let us briefly describe the nature of the power market from a trading
and risk management perspective.
First, like many products, electricity demand varies significantly throughout the
day, week and year. However, electricity has the same properties as a highly perishable
product in that not only must supply meet demand, but production must meet
demand minute by minute. The result of this unique attribute is that the market
must keep a significant amount of idle capacity in place for start-up when the
demand is there and shut-down when demand recedes. In some cases, this is a
generation station that is already running, ready to meet anticipated customer
demand. In other cases it is plant that may only be asked to start up once every few
years. Typically a local market (or grid system) will keep 15–20% more plant capacity
than it expects to use on the highest hour of demand during a normal year. This will
often represent over double the average demand.
This, coupled with the physical challenges of transporting electricity,11 leads to a
general position of over-supply combined with short periods when the normally idle
plant needs to run. At such peak times, when all the capacity on the system is
needed, spot market prices12 will rise dramatically as plant owners will need to
recover not only their higher cost of running but also their capital costs in a relatively
short period of time. This is exacerbated by the fact that electricity generation is one
of the most capital-intensive industries in the world.
The forward market will, of course, smooth this by assigning probabilities to the
likelihood of the high prices. Higher probabilities are obviously assigned during the
periods when demand is likely to be at a peak and this will generally only occur
during two or three months of the year. In the USA this is generally in the summer.
Thus unless there is a huge over-supply the summer prices will be significantly
higher than the rest of the year. When there is a potential shortage prices will be
dramatically higher, as was seen in the Cinergy market in late June 1998 where
daily prices that trade most of the year at $30/MWh increased to $7000/MWh.
It is these important market characteristics that lead to the extreme seasonality,
jumps, spikes and mean reversion that will be discussed below in the section on
Market Risk. Given the extreme nature of these factors trying to capture them on
one term structure or convenience yield is extremely difficult.
Despite this price uncertainty, a forward market for electricity has developed along
a standard commodity structure. In fact seven exchange contracts, reflecting the
regional nature of power, currently exist at Palo Verde, California/Oregon Border
(COB), Entergy, Cinergy (NYMEX), TVA, ComEd (CBOT) and Twin Cities (M. Grain
Exchange) and PJM. The forward curves for Cinergy is shown in Figure 18.2 compared
to the Henry Hub gas curve.
Figure 18.2 Forward curves for Cinergy compared to the Henry Hub gas curve. (Source: Citizens
Power,June 1999)
Some standard options are traded at the most liquid hubs. These tend to be ‘strips’
of daily European calls based on monthly blocks. However, the bid/ask on such
products are often wide and the depth of liquidity very limited.
Forward curve price discovery – the problems with power
Let us focus on two major differences in power:
Ω Storage First, you have virtually no stack and roll storage arbitrage. In other
words, since you cannot keep today’s power for tomorrow there is no primary
‘arbitrage’ linkage between today’s price and tomorrow’s. There are, of course,
many secondary links. The underlying drivers are likely to be similar – demand,
plant availability, fuel costs, traders’ expectations and general market environment.
But as you move forward in time, secondary links break down quickly and
so does the price relationship. Figure 18.3 shows the forward correlation between
an April contract and the rest of the year. As you can see, almost no relationship
exists between April and October and most of the relationship has evaporated
once you are beyond one month. In other words, the October Cinergy contract
has no more relationship with the April contract than, say, an oil, gas or even
interest rate market.
Ω Transportation Second, you have limited ‘hub basis’ arbitrage. Since there are
numerous logistical limitations on moving electricity it is difficult to arbitrage
between many of the power markets within the USA (never mind internationally).
Even hubs that are relatively close show large variations in the spread between
prices. Figure 18.4 shows some of the correlations between major power hubs in
the USA. Rather than thinking of them as one market, it is more accurate to view
the power market as at least twelve (the final number of hubs is still being
determined by the marketplace) independent markets with some but often little
relationship.
Once you put all these factors together you see a picture similar to the one a global
risk managers in a big bank will have experienced, a huge number of independent
products that need to be combined for risk purposes. Instead of having one forward
curve for US Power, we have up to eighteen independent months for twelve independent
markets, in other words 216 products. This brings both the curse of lack of
liquidity and data integrity for each product and, on the positive side from a risk
perspective, diversity.
Forward curves up to 24 months are traditionally built using daily trader/broker
marks for monthly peak/off-peak prices, with the breakdown, where necessary, into
smaller time periods (down to an hourly profile) using historical prices adjusted for
normal weather conditions. Prices beyond two years are significantly less liquid and
where information is available bid/ask spreads can increase significantly. Forward
price curves (and volatility curves) beyond 24 months thus need to be created through
more of a mark to model rather than mark to market process. As noted above, using
a model to connect price quotes inevitably involved a number of assumptions about
how the market behaves. In power, this involves not just fitting a serious of different
price quotes together, but also filling in the gaps where price quotes are not available.
The price structure will have to make certain model assumptions based on historical
observation about seasonality and the year-to-year transition process. Models
can be bought (such as the SAVA forward curve builder) or, more often, built inhouse.
However, given the developing nature of the market this still tends to be a
relatively manual process to ensure all the relevant market information can be input
into the curve and minimize the error terms.
It needs to be remembered that in making assumptions about the structure of the
curve in this process to estimate the fair value of a transaction that the forward
curve, while objective, unbiased and arbitrage-free, may be unreliable given the
incomplete data sets. Throughout the process it is thus necessary to estimate the
impact of such assumptions and modeling or prudency reserves are likely to need to
be applied against the fair value under these circumstances. The collection of market
data with an illiquid market also becomes a major operational issue with a need to
continuously verify and search for independent data. 534
29 Haziran 2011 Çarşamba
Gas forward curve
Price discovery in gas is not as transparent as in the oil market. The transport and
storage capabilities of the gas market is relatively inflexible compared to the oil
market. The market is also more regional7 than the oil market, with international
trade being restricted by pipeline costs and LNG (liquefied natural gas) processing
and transport costs.
The US gas trading market (although probably the most developed in the world) is
much more fragmented than in the oil market with a large number of small producers,
particularly at the production end. Given the inability to hedge through vertical
integration or diversity this has resulted in a significant demand for risk management
products. The market after the deregulation in the 1980s and early 1990s has been
characterized by the development of a very significant short-term market. This sets
prices for a thirty-day period during what is known as ‘bid week’.8 The prices
generated during this ‘bid week’ create an important benchmark9 against which
much of the trading market is based.
The futures contracts in gas have been designed to correspond with the timing of
bid week, the largest and most developed contract being known as Henry Hub. This
has been quoted on NYMEX since 1990, supports a strong options market and
extends out three years on a relatively liquid basis. Basis relationships between
Henry Hub and the major gas-consuming regions within the USA are well established
and, in the short term, fairly stable. The OTC market supports most basis locations
and a reasonable option market can be found for standard products. It is worth
noting that, credit aside, in energy the value of the futures market and the OTC
forward market is the same, albeit the delivery mechanisms are different. This is
true because there is no direct correlation to interest rates and the EFP (Exchange
for Physical) option imbedded in the futures contract.
In gas, storage costs play a critical role in determining the shape of the forward
curve. A variety of storage is employed from line pack (literally packing more gas
molecules into the pipe) to salt caverns and reservoirs (gas can be injected and
extracted with limited losses) and the pricing and flexibility of the different storage
options varies significantly.
As a general rule, the gas market stores (injects) during seven or eight ‘summer’
months and extracts during the winter months. Significant short-run price movements
occur when this swing usage is out of balance. In these circumstances gas
traders will spend much of the time trying to predict storage usage against their
demand estimations it order to determine what type of storage will be used during
the peak season and thus help set the future marginal price.
Like the oil market, the gas markets exhibit mean reversion, and are subject to
occasional jumps due to particular ‘events’ such as hurricanes shutting down
supplies. But most importantly, despite this storage, gas remains significantly more
seasonal in nature than most of the oil market.
Basis trading from Henry Hub plays a vital part in the market with differentials to
the major consumption zones being actively traded in the OTC market. These basis
prices are less stable than those seen in the oil market given the more constrained
transport infrastructure and volatility in demand. As a result, the monitoring of these
basis relationships under normal and extreme conditions becomes critical.
Risk managers should be very wary of basis traders or regional traders marking
their books against a contract in a different region such as Henry Hub who are
seen to have large ‘book’ profits based on future positions. There have been a
number of instances where such regional traders have had their positions wiped
out overnight when the correlations have broken down under extreme market conditions.
In other words, you need to ensure that the higher volatility or ‘spike potential’
in less liquid regional markets compared to a large liquid hub has been reflected
in the pricing.
Historical ‘basis’ positions may also fundamentally change as the pipeline positions
change. For instance, the increasing infrastructure to bring Canadian gas to the
Chicago and North US markets could significantly change the traditional basis
‘premium’ seen in these markets compared to Henry Hub.
While most trades up to three years are transacted as forward or future positions
an active swap market also exists, particularly for longer-term deals where the
counter-parties do not want to take on the potential risks associated with physical
delivery. These index trades are generally against the published Inside FERC Gas
Market Report Indices although Gas Daily and other publications are also used
regularly. It is necessary to be aware that indices at less liquid points may not be
based on actual transaction prices at all times, but may be based on a more
informal survey of where players think the market is. This may lead to indices being
unrepresentative of the true market. 531
storage capabilities of the gas market is relatively inflexible compared to the oil
market. The market is also more regional7 than the oil market, with international
trade being restricted by pipeline costs and LNG (liquefied natural gas) processing
and transport costs.
The US gas trading market (although probably the most developed in the world) is
much more fragmented than in the oil market with a large number of small producers,
particularly at the production end. Given the inability to hedge through vertical
integration or diversity this has resulted in a significant demand for risk management
products. The market after the deregulation in the 1980s and early 1990s has been
characterized by the development of a very significant short-term market. This sets
prices for a thirty-day period during what is known as ‘bid week’.8 The prices
generated during this ‘bid week’ create an important benchmark9 against which
much of the trading market is based.
The futures contracts in gas have been designed to correspond with the timing of
bid week, the largest and most developed contract being known as Henry Hub. This
has been quoted on NYMEX since 1990, supports a strong options market and
extends out three years on a relatively liquid basis. Basis relationships between
Henry Hub and the major gas-consuming regions within the USA are well established
and, in the short term, fairly stable. The OTC market supports most basis locations
and a reasonable option market can be found for standard products. It is worth
noting that, credit aside, in energy the value of the futures market and the OTC
forward market is the same, albeit the delivery mechanisms are different. This is
true because there is no direct correlation to interest rates and the EFP (Exchange
for Physical) option imbedded in the futures contract.
In gas, storage costs play a critical role in determining the shape of the forward
curve. A variety of storage is employed from line pack (literally packing more gas
molecules into the pipe) to salt caverns and reservoirs (gas can be injected and
extracted with limited losses) and the pricing and flexibility of the different storage
options varies significantly.
As a general rule, the gas market stores (injects) during seven or eight ‘summer’
months and extracts during the winter months. Significant short-run price movements
occur when this swing usage is out of balance. In these circumstances gas
traders will spend much of the time trying to predict storage usage against their
demand estimations it order to determine what type of storage will be used during
the peak season and thus help set the future marginal price.
Like the oil market, the gas markets exhibit mean reversion, and are subject to
occasional jumps due to particular ‘events’ such as hurricanes shutting down
supplies. But most importantly, despite this storage, gas remains significantly more
seasonal in nature than most of the oil market.
Basis trading from Henry Hub plays a vital part in the market with differentials to
the major consumption zones being actively traded in the OTC market. These basis
prices are less stable than those seen in the oil market given the more constrained
transport infrastructure and volatility in demand. As a result, the monitoring of these
basis relationships under normal and extreme conditions becomes critical.
Risk managers should be very wary of basis traders or regional traders marking
their books against a contract in a different region such as Henry Hub who are
seen to have large ‘book’ profits based on future positions. There have been a
number of instances where such regional traders have had their positions wiped
out overnight when the correlations have broken down under extreme market conditions.
In other words, you need to ensure that the higher volatility or ‘spike potential’
in less liquid regional markets compared to a large liquid hub has been reflected
in the pricing.
Historical ‘basis’ positions may also fundamentally change as the pipeline positions
change. For instance, the increasing infrastructure to bring Canadian gas to the
Chicago and North US markets could significantly change the traditional basis
‘premium’ seen in these markets compared to Henry Hub.
While most trades up to three years are transacted as forward or future positions
an active swap market also exists, particularly for longer-term deals where the
counter-parties do not want to take on the potential risks associated with physical
delivery. These index trades are generally against the published Inside FERC Gas
Market Report Indices although Gas Daily and other publications are also used
regularly. It is necessary to be aware that indices at less liquid points may not be
based on actual transaction prices at all times, but may be based on a more
informal survey of where players think the market is. This may lead to indices being
unrepresentative of the true market. 531
24 Haziran 2011 Cuma
The oil market
Crude oil can come from almost any part of the world and the oil market, unlike gas
and power, can be described as a truly global market. The principal production
regions can be generally grouped into the following: North Sea (the most notable
grade being Brent), West Africa, Mediterranean, Persian Gulf (notably Dubai), Asia,
USA (notably WTI), Canada and Latin America. Each region will have a number of
quality grades and specifications within it, in particular, depending on their API
gravity2 and sulfur content.
In Europe the benchmark crude product is North Sea Brent while in the USA it is
often quoted as WTI. WTI is the principal product meeting the NYMEX sweet crude
specifications for delivery at Cushing (a number of other qualities can also be
deliverable, although many non-US crudes receive a discount to the quoted settlement
price). The other major crude product is Dubai, representing the product
shipped from the Persian Gulf.
Each of these three products will trade in close relation to each other, generally
reflecting their slightly different qualities and the transport cost from end-user
markets, all three markets reflecting the overall real or perceived supply/demand
balance in the world market. Other crude specifications and delivery points will then
trade some ‘basis’ from these benchmark prices.
As well as the three crude products noted above, the oil market encompasses the
refined products from crude oil. While thousands of different qualities and delivery
points world-wide will ultimately result in hundreds of thousands of different prices,
in Europe and North America these can generally be linked back to a number of
relatively strong trading hubs that exist, notably:
Product type Principal hubs
Crude Brent/Cushing/Dubai
Unleaded NWE3/New York/Gulf Coast
Gas Oil/No. 2 Oil NWE/New York Harbor
Heavy Fuel Oil/No. 6 Oil NWE/New York Harbor/Far East
In addition, the rest of the barrel, propane, butane, naphtha and kerosene trade
actively, but are more limited in terms of their relevance to the energy complex.
Within the various product ranges prices are quoted for particular standard grades.
For instance, No. 6 oil (also known as HFO or residual oil) can be segmented to 1%,
2.2%, 3% and 3.5% sulfur specifications. Each have their own forward curves and
active trading occurs both on the individual product and between the products. The
market has developed to the point where NYMEX lists not only option prices but also
Crack Spread Options (the option on the spread between the products).4
Like its other energy counterparts most of the trading is done in the OTC ‘brokered’
market that supports most of the commonly traded options, swaps and other
derivative structures seen in the financial markets. Derivatives are particularly useful
in oil compared to other energy products given the international nature of the product
and the relationship between the overall oil complex. For instance, if we take an
airline company, this needs a jet fuel hedge that reflects the weighted average cost
of its physical spot fuel price purchases in different parts of the world. At the same
time it would like to avoid any competitive loss it might experience from hedging out
at high prices. This would be complex (and unnecessary) to achieve physically, but
relatively straightforward to hedge using a combination of different swaps and
average price options, which can be easily linked to currency hedges.
Another common swap is the front-to-back spread, or synthetic storage. This
allows the current spot price to be swapped for a specified forward month. It should
be noted that this relationship may be positive or negative, depending on market
expectations. Locational swaps are also very common, providing a synthetic transport
cost. Such swap providers in this market (and all energy markets), however, need to
be very aware of both the physical logistics and the spot market volatility.
Crack spreads and Crack spread options are used to create synthetic refineries.
The 3:2:1 crack spread that is traded in NYMEX is a standard example of this linking
the prices of crude, heating oil and gasoline.
In oil, the majority of swap transactions are carried out against Platt’s indices that
cover most products and locations, although a number of other credible indices exist
in different locations. For instance, CFD’s (Contracts for Differences) are commonly
traded against ‘dated’ Brent, the price for physical cargoes loading shortly and a
forward Brent price approximately three months away.
While such derivatives are easy to construct and transact against the liquid hubs
they have their dangers when using them to hedge physical product at a specific
delivery point. Specific supply/demand factors can cause spreads between locations
and the hubs to change dramatically for short periods of time before they move back
into equilibrium. For example, extreme weather conditions can lead to significant
shortages in specific locations where imports are not possible leading to a complete
breakdown of the correlation between the physical product and the index being used
to hedge. In other words you lose your hedge exactly when you need it. A risk
manager must look carefully at the spreads during such events and the impact on
the correlations used in VaR and Stress tests. They should also understand the
underlying supply/demand conditions and how they could react during such extreme
events.
It should also be noted that oil products are often heavily taxed and regulated on
a state and national basis which can lead to a number of legal, settlement and
logistical complexities. Going hand in hand with this is the environmental risks
associated with storage and delivery, where insurance costs can be very substantial.
Ever-changing refinery economics, storage and transportation costs associated
with the physical delivery of oil are thus a significant factor in pricing which results
in the forward curve dynamics being more complex than those seen in the financial
markets. As a result the term structure is unpredictable in nature and can vary
significantly over time. Both backwardation5 and contago6 structures are seen within
the curve, as is a mean reversion component. Two components are commonly used
to describe the term structure of the oil forward curve: the price term structure,
notably the cost of financing and carry until the maturity date, and the convenience
yield. The convenience yield can be described as the ‘fudge factor’ capturing the
market expectations of future prices that are not captured in the arbitrage models.
This would include seasonal and trend factors.
Given the convenience yield captures the ‘unpredictable’ component of the curve,
much of the modeling of oil prices has focused on describing this convenience yield
which is significantly more complex than those seen in the financial markets. For
instance, under normal conditions (if there is such a thing), given the benefit of
having the physical commodity rather than a paper hedge, the convenience yield is
often higher than the cost of carry driving the market towards backwardation.
Fitting a complex array of price data to a consistent forward curve is a major
challenge and most energy companies have developed proprietary models based on
approaches such as HJM to solve this problem. Such models require underlying
assumptions on the shape of the curve fitting discrete data points and rigorous
testing of these assumptions is required on an ongoing basis.
Assumptions about the shape of the forward market can be very dangerous as MG
discovered. In their case they provided long-term hedges to customers and hedged
them using a rolling program of short-term future positions. As such they were
exposed to the spread or basis risk of the differential between the front end of the
market (approximately three-month hedges) and the long-term sales (up to ten years).
When oil prices in 1993 fell dramatically they had to pay out almost a billion dollars
or margin calls in their short-term positions but saw no offsetting benefit from their
long-term sales. In total they were reported to have lost a total of $1.3 billion by
misunderstanding the volatility of this spread.
Particularly in the case of heating oil, significant seasonality can exists. Given the
variation in demand throughout the year and storage economics, the convenience
yield will vary with these future demand expectations. This has the characteristic of
pronounced trends, high in winter when heating oil is used, low in summer and a
large random element given the underlying randomness of weather conditions.
The oil market also exhibits mean reversion characteristics. This makes sense,
since production economics show a relatively flat cost curve (on a worldwide basis
the market can respond to over- and under-supply and that weather conditions (and
thus the demand parameters) will return to normal after some period.
In addition, expected supply conditions will vary unpredictably from time to time.
Examples of this include the OPEC and Gulf War impacts on perceived supply risks.
Such events severely disrupt the pricing at any point leading to a ‘jump’ among the
random elements and disrupting both the spot prices and the entire dynamics of the
convenience yield. I will return to the problem of ‘jumps’ and ‘spikes’ later in this
chapter. 529
and power, can be described as a truly global market. The principal production
regions can be generally grouped into the following: North Sea (the most notable
grade being Brent), West Africa, Mediterranean, Persian Gulf (notably Dubai), Asia,
USA (notably WTI), Canada and Latin America. Each region will have a number of
quality grades and specifications within it, in particular, depending on their API
gravity2 and sulfur content.
In Europe the benchmark crude product is North Sea Brent while in the USA it is
often quoted as WTI. WTI is the principal product meeting the NYMEX sweet crude
specifications for delivery at Cushing (a number of other qualities can also be
deliverable, although many non-US crudes receive a discount to the quoted settlement
price). The other major crude product is Dubai, representing the product
shipped from the Persian Gulf.
Each of these three products will trade in close relation to each other, generally
reflecting their slightly different qualities and the transport cost from end-user
markets, all three markets reflecting the overall real or perceived supply/demand
balance in the world market. Other crude specifications and delivery points will then
trade some ‘basis’ from these benchmark prices.
As well as the three crude products noted above, the oil market encompasses the
refined products from crude oil. While thousands of different qualities and delivery
points world-wide will ultimately result in hundreds of thousands of different prices,
in Europe and North America these can generally be linked back to a number of
relatively strong trading hubs that exist, notably:
Product type Principal hubs
Crude Brent/Cushing/Dubai
Unleaded NWE3/New York/Gulf Coast
Gas Oil/No. 2 Oil NWE/New York Harbor
Heavy Fuel Oil/No. 6 Oil NWE/New York Harbor/Far East
In addition, the rest of the barrel, propane, butane, naphtha and kerosene trade
actively, but are more limited in terms of their relevance to the energy complex.
Within the various product ranges prices are quoted for particular standard grades.
For instance, No. 6 oil (also known as HFO or residual oil) can be segmented to 1%,
2.2%, 3% and 3.5% sulfur specifications. Each have their own forward curves and
active trading occurs both on the individual product and between the products. The
market has developed to the point where NYMEX lists not only option prices but also
Crack Spread Options (the option on the spread between the products).4
Like its other energy counterparts most of the trading is done in the OTC ‘brokered’
market that supports most of the commonly traded options, swaps and other
derivative structures seen in the financial markets. Derivatives are particularly useful
in oil compared to other energy products given the international nature of the product
and the relationship between the overall oil complex. For instance, if we take an
airline company, this needs a jet fuel hedge that reflects the weighted average cost
of its physical spot fuel price purchases in different parts of the world. At the same
time it would like to avoid any competitive loss it might experience from hedging out
at high prices. This would be complex (and unnecessary) to achieve physically, but
relatively straightforward to hedge using a combination of different swaps and
average price options, which can be easily linked to currency hedges.
Another common swap is the front-to-back spread, or synthetic storage. This
allows the current spot price to be swapped for a specified forward month. It should
be noted that this relationship may be positive or negative, depending on market
expectations. Locational swaps are also very common, providing a synthetic transport
cost. Such swap providers in this market (and all energy markets), however, need to
be very aware of both the physical logistics and the spot market volatility.
Crack spreads and Crack spread options are used to create synthetic refineries.
The 3:2:1 crack spread that is traded in NYMEX is a standard example of this linking
the prices of crude, heating oil and gasoline.
In oil, the majority of swap transactions are carried out against Platt’s indices that
cover most products and locations, although a number of other credible indices exist
in different locations. For instance, CFD’s (Contracts for Differences) are commonly
traded against ‘dated’ Brent, the price for physical cargoes loading shortly and a
forward Brent price approximately three months away.
While such derivatives are easy to construct and transact against the liquid hubs
they have their dangers when using them to hedge physical product at a specific
delivery point. Specific supply/demand factors can cause spreads between locations
and the hubs to change dramatically for short periods of time before they move back
into equilibrium. For example, extreme weather conditions can lead to significant
shortages in specific locations where imports are not possible leading to a complete
breakdown of the correlation between the physical product and the index being used
to hedge. In other words you lose your hedge exactly when you need it. A risk
manager must look carefully at the spreads during such events and the impact on
the correlations used in VaR and Stress tests. They should also understand the
underlying supply/demand conditions and how they could react during such extreme
events.
It should also be noted that oil products are often heavily taxed and regulated on
a state and national basis which can lead to a number of legal, settlement and
logistical complexities. Going hand in hand with this is the environmental risks
associated with storage and delivery, where insurance costs can be very substantial.
Ever-changing refinery economics, storage and transportation costs associated
with the physical delivery of oil are thus a significant factor in pricing which results
in the forward curve dynamics being more complex than those seen in the financial
markets. As a result the term structure is unpredictable in nature and can vary
significantly over time. Both backwardation5 and contago6 structures are seen within
the curve, as is a mean reversion component. Two components are commonly used
to describe the term structure of the oil forward curve: the price term structure,
notably the cost of financing and carry until the maturity date, and the convenience
yield. The convenience yield can be described as the ‘fudge factor’ capturing the
market expectations of future prices that are not captured in the arbitrage models.
This would include seasonal and trend factors.
Given the convenience yield captures the ‘unpredictable’ component of the curve,
much of the modeling of oil prices has focused on describing this convenience yield
which is significantly more complex than those seen in the financial markets. For
instance, under normal conditions (if there is such a thing), given the benefit of
having the physical commodity rather than a paper hedge, the convenience yield is
often higher than the cost of carry driving the market towards backwardation.
Fitting a complex array of price data to a consistent forward curve is a major
challenge and most energy companies have developed proprietary models based on
approaches such as HJM to solve this problem. Such models require underlying
assumptions on the shape of the curve fitting discrete data points and rigorous
testing of these assumptions is required on an ongoing basis.
Assumptions about the shape of the forward market can be very dangerous as MG
discovered. In their case they provided long-term hedges to customers and hedged
them using a rolling program of short-term future positions. As such they were
exposed to the spread or basis risk of the differential between the front end of the
market (approximately three-month hedges) and the long-term sales (up to ten years).
When oil prices in 1993 fell dramatically they had to pay out almost a billion dollars
or margin calls in their short-term positions but saw no offsetting benefit from their
long-term sales. In total they were reported to have lost a total of $1.3 billion by
misunderstanding the volatility of this spread.
Particularly in the case of heating oil, significant seasonality can exists. Given the
variation in demand throughout the year and storage economics, the convenience
yield will vary with these future demand expectations. This has the characteristic of
pronounced trends, high in winter when heating oil is used, low in summer and a
large random element given the underlying randomness of weather conditions.
The oil market also exhibits mean reversion characteristics. This makes sense,
since production economics show a relatively flat cost curve (on a worldwide basis
the market can respond to over- and under-supply and that weather conditions (and
thus the demand parameters) will return to normal after some period.
In addition, expected supply conditions will vary unpredictably from time to time.
Examples of this include the OPEC and Gulf War impacts on perceived supply risks.
Such events severely disrupt the pricing at any point leading to a ‘jump’ among the
random elements and disrupting both the spot prices and the entire dynamics of the
convenience yield. I will return to the problem of ‘jumps’ and ‘spikes’ later in this
chapter. 529
Kaydol:
Kayıtlar (Atom)