Modeling Energy Market Volatility Using REMI
October 2010
Jim Peach
Department of Economics and International Business
New Mexico State University
jpeach@nmsu.edu
Some Background on Energy Market Volatility (EMV)
• EMV is defined here to include price, employment, and output volatility
• EMV appears more likely than energy market stability
• EMV creates direct and indirect impacts on
– Output, employment, and income
– State revenues e.g., severance tax revenue
• REMI standard controls (state and national) are “smooth”
• EMV can be modeled in REMI using relatively simple techniques
Some Specifics
• Energy Prices in real (and nominal) terms are highly variable
– Including oil, natural gas, uranium, and coal
• Domestic output in physical quantities is less variable than prices
Real Oil Price Volatility 2000 to 2009
Annual Percent
Change
in West Texas
Intermediate
Real Price
Sources: WTI prices from EIA, Short Term Energy Outlook, Table 2 and
GDP deflator from Bureau of Economic Analysis, NIPA Tables 1.1.9
Natural Gas Price Variability: 2000 to 2009
Annual
Percent
Change in
Real ($2005)
Henry Hub
Spot Price
Sources: WTI prices from EIA, Short Term Energy Outlook, Table 2 and
GDP deflator from Bureau of Economic Analysis, NIPA Tables 1.1.9
Coal Price Volatility: 2000 to 2009
Annual
Percent
Change
In Real
Coal
Price
($2005
Per short ton)
Sources: Coal prices from EIA, Annual Energy Review (2010), Table 7.8 and
GDP deflator from Bureau of Economic Analysis, NIPA Tables 1.1.9
Uranium (Spot Price) Variability: 2000 to 2009
Annual
Percent
Change
U308
Spot
Price
Real ($2005)
Sources: Uranium Prices from EIA Uranium Annual Marketing Report 2010, Table S1b
GDP deflator from Bureau of Economic Analysis, NIPA Tables 1.1.9
Output and Price
• Domestic output in physical quantities is less variable than prices
– Once the well is drilled . . .
• Marginal cost is relatively low
• Shut down/start-up costs substantial
– Contract deliveries . . .
• Consider coal or uranium production
– Futures markets, hedge funds, etc.
– Continuity of labor force:
• Skilled workers, service providers, etc
U.S. Oil Prices and Production: 2000 to 2009
Percent
Change
US Oil Production
and Real ($2005)
WTI Price
Source: Oil Production: EIA Petroleum Navigator, Crude Oil Production
WTI Price and Deflator: See Previous Figure.
US Natural Gas Prices and Production
Percent
Change in
Natural Gas
Production and
Henry Hub
Real ($2005)
Price
Source: Natural Gas Production: EIA Natural Gas Navigator
Natural Gas Wellhead Value and Marketed Production
Prices and Deflator: See Previous Figures
Some (Very) Direct Impacts
(Oil and Gas Price Volatility)
• State Tax Revenue
– “High reliance on O&G revenues creates two challenges for the state:
revenue volatility and long-run sustainability”
• Tom Clifford, Chief Economist, New Mexico Legislative Finance Committee, August 5, 2010.
– Severance and other production taxes
– But don’t forget CIT, PIT, GRT
• Rig counts
– Exploration and drilling are highly price sensitive
• Employment in oil and gas extraction
– NAICS Code ( 211)
• Need more –Long list of indirect and dynamic effects
New Mexico Oil and Gas Revenues as Percent of General Fund
Source: New Mexico Department of Finance,
Consensus Revenue Forecast (2009, Figure 12, p. 14)
U.S. Oil Rig Counts and WTI Prices
Oil Rig Activity
and
Real Oil Prices
Index Numbers
Source: Index numbers (2000=100) calculated by author.
Rig Data EIA Petroleum Navigator, Crude Oil and Natural Gas Drilling Activity
Price: See Previous Figures
Natural Gas Rigs and Natural Gas Prices
US Natural Gas Rigs
and
Henry Hub Spot Prices
(Real($2005)
Index Numbers
(2000 = 100)
Source: See previous figure.
Employment in Oil and Gas Extraction
(NAICS 211) REMI Data
Percent Change
Employment
NM and US
Oil and Gas
Extraction
Source: REMI PI+ standard controls
Oil and Gas Extraction Employment and Prices
Percent Change
In Employment
Index and Weighted
Oil and Gas Price
Index
Source: Author Calculations from BEA and EIA Data
Price index weighted by BTU equivalent production levels
EIA Confidence Intervals for WTI
REMI Oil and Gas Extraction Employment
Standard National Control
REMI Oil and Gas Extraction Employment
Standard Regional Control
New Mexico
Why is this important?
•
•
•
•
The next slide contains employment “multipliers” for selected industries
Computed from REMI Standard National Control
Assumes 1% change in employment in sector listed
Computed as:
Change in total employment
divided by
Change in employment in sector listed
Example (Construction Sector):
921.922/437.519 = 2.11
Where
921.922 = difference from REMI baseline in total employment
437.519 = 1% difference from REMI baseline in Construction Employment
Selected Industry Total Employment Multipliers:
REMI Standard National Control
The employment
multipliers represent
the estimated impact
of an additional job
in the sector listed on
total employment
Source: Author calculations using REMI PI+ standard national control.
Multipliers are for 2010.
Many Options to Model Energy in REMI
• Relative fuel costs, relative delivered prices, cost of production,
• Fuel Weight data (shares natural gas, electricity, residual)
• Consumption patterns
–
–
–
–
–
Gasoline and oil
Fuel oil and coal
Electricity
Gas
Several transportation related categories
• Trade flows
– Rest of nation and rest of world
• Modify intermediate demand, etc.
• Create New Industry
– e.g. uranium mining and milling (NAICS 212291)
– Petroleum refineries (NAICS 32411)
Modeling Options Continued
• Industry specific employment / output changes in many sectors
–
–
–
–
–
–
Oil and gas extraction (NAICS211)
Coal mining (NAICS 2121)
Electric power generation, transmission and Distribution (NAICS 2211)
Natural gas distribution (NAICS 2212)
Petroleum and coal products manufacturing (NAICS 324)
Pipeline transportation (NAICS 486)
A National (and State) REMI Experiment:
and some implications for New Mexico
• Step 1: Create new national control
– Impose the variability (% change) in oil and gas employment observed from 2000 to
2009 on projected years 2010 to 2019.
• Step 2: Run regional model using new national control
– With no other changes to the regional model
A National (and State) REMI Experiment:
Step 1 Create New National Control
Can also be done through labor demand block
A National (and State) REMI Experiment:
Step 1 National results
Net Change in Oil and Gas Employment 2010-2019 = 179.2k jobs
A National (and State) REMI Experiment:
Step 1 National results
Net Change in Total Employment 2010-2019 = 1.6 million jobs
Net Change in RGDP 2010-2019 = $158 Billion
A National (and State) REMI Experiment:
Step 2 New Mexico Results
Net Change in Total Employment 2010-2019 = 21.2 K jobs
Net Change in RGDP 2010-2019 = $1.5 Billion
A National (and State) REMI Experiment:
Step 2 New Mexico results
Range in Oil and Gas Employment  -9.56 % to + 16.63%
A National (and State) REMI Experiment:
Pros and Cons
• Advantages of this technique:
– It is easy –very easy– to implement
– It does capture historic variability that is not captured in the standard controls
• and, it makes a difference in the results
• Disadvantages
– Timing –who knows when the variability will occur?
A Second National (and State) REMI Experiment:
Also Easy
• Use μ (mean) and σ (standard deviation) of previous decade variability
– Or some other plausible μ and σ
• Draw random numbers from this distribution to obtain inputs
– Assume the distribution is normal
– or some other distribution since the world is not normal
• Construct new national control
• Examine impacts at state level
– No other changes at state level
A Second National (and State) REMI Experiment:
Inputs used in EMPL in oil and gas extraction sector
Compared to implied historic
A Second National (and State) REMI Experiment:
Results: % Chg from Standard Regional Control
Total Employment and RGDP
Net Change in NM Total Employment = 19.6k jobs
Net change in NM Oil and Gas Employment = 3.1 k jobs
A Second National (and State) REMI Experiment:
Results: % Chg from Standard Regional Control
Oil and Gas Extraction Employment and Other Sectors
A Monte Carlo National (and State) REMI Experiment:
Fun, but don’t try this one ….
• Use μ (mean) and σ (standard deviation) of previous decade variability
– Or some other plausible μ and σ
• Draw 100 (250?) random samples (see next slide)
– from normal or some other distribution
• Run the previous experiment in REMI using each sample
• Save the results
• Construct interval estimates
– Many ways to do this
Who are these guys?
They are “Arrowhead slaves” who did the 250 REMI runs.
Mikidadu Mohammed, Graduate Student
Leo Delgado, Policy Analyst
A Monte Carlo National (and State) REMI Experiment:
What the spreadsheet might look like
A Monte Carlo National (and State) REMI Experiment:
Some Interval Estimates
Ranges shown are in percentage differences in New Mexico Oil and Gas Extraction
employment. The 12 highest and 12 lowest observations were removed.
A Monte Carlo National (and State) REMI Experiment:
Some More Interval Estimates
Notice the change
In scale from the
Previous chart
Ranges shown are percentage differences in New Mexico Oil and Gas Extraction
employment. These are the maximum and minimum ranges from 250 trials.
A Monte Carlo National (and State) REMI Experiment:
Advantages and Disadvantages
• Advantages
– Explicit recognition that:
• Energy Market Variability (EMV) is likely
• The timing (years) is unknown and unknowable
– Produces a plausible(?) range of estimates rather than a point estimate
– Satisfies ceremonial requirements of complexity and sophistication
• Disadvantages
–
–
–
–
–
–
–
The intervals are far too large to be meaningful?
Not necessary due to linearity of REMI responses
Time consuming (expensive)
Distributional assumption or The World is Not Normal
Stationarity of μ and σ
Other less time consuming methods to generate intervals
The intervals do not get larger as t gets larger
Some Conclusions
• Variability matters
– even in a relatively short (10 year) time horizon
– even when only oil and gas extraction is considered
• Variability here to stay?
– More than likely
• Not difficult to impose some variability on the standard controls
• Key questions
– How much variability?
– When will it occur?
• Answers to key questions are unknown and unknowable
– but this should not deter some attempt to impose variability
– when stability (smoothness) seems rare