LeibyEtAl2015 ORNL - Drawdown Capability Study for DOE EPSA - Final - 20150414

7/25/2019 LeibyEtAl2015 ORNL - Drawdown Capability Study for DOE EPSA - Final - 20150414

1/51

Results of Drawdown Capability Benefits Study for DOE/EPSA

Paul Leiby1

, David Bowman, Debo Oladosu, and Rocio Uria-MartinezMarch 30, 2015

AbstractThe U.S. Strategic Petroleum Reserve (SPR) is a government-owned stockpile ofoil intended to provide surge supply during oil market supply emergencies.Changing patterns of oil production and transport in the domestic market as wellas technical changes to the distribution infrastructure capabilities have madeachieving the full U.S. SPR design drawdown capability of 4400 MBD difficult.Recent analysis shows that left unaddressed, the maximum draw and distributioncapability could be as low as 1200 MBD in some disruptions conditions. This

paper details a study of the economic benefits associated with maintaining higherlevels of SPR drawdown and distribution capability. The benefits estimates forthree potential disruption suggest that under a wide range of circumstances, thetotal benefits of increasing the maximum drawdown and distribution capabilityrate by 2000 MBD (from 1200 MBD to 3200) can result in tens to hundreds ofbillions of dollars in savings, in the event of a large world oil supply disruption.

1.0 Introduction

The U.S. Strategic Petroleum Reserve (SPR) is a government-owned stockpile of oil intended toprovide surge supply during oil market supply emergencies.2 Much of the prior analysis of the

economic costs and benefits of this reserve (e.g. U.S. DOE/Interagency 1990, Leiby andBowman 2000a, 2000b, Leiby, Jones and Bowman 2000, GAO 2006, Leiby and Bowman 2007,IEA 2012, Leiby et al. 2013) has focused on the uncertainty of world oil supply but treated theavailability of oil from the SPR during oil supply disruptions as completely assured. However,changing patterns of oil production and transport in the domestic market as well as technicalchanges to the distribution infrastructure capabilities have made achieving the full U.S. SPRdesign drawdown capability of 4400 MBD difficult. Recent analysis shows that leftunaddressed, the maximum draw and distribution capability could be as low as 1200 MBD, insome disruptions conditions.

1Corresponding author: [email protected]://energy.gov/fe/services/petroleum-reserves/strategic-petroleum-reserve. Established in the aftermath ofthe 1973-74 oil embargo, the SPR provides the President with a powerful response option should a disruption incommercial oil supplies threaten the U.S. economy. It also allows the United States to meet part of its InternationalEnergy Agency obligation to maintain emergency oil stocks, and it provides a national defense fuel reserve.
http://energy.gov/fe/services/petroleum-reserves/strategic-petroleum-reservehttp://energy.gov/fe/services/petroleum-reserves/strategic-petroleum-reservehttp://energy.gov/fe/services/petroleum-reserves/strategic-petroleum-reservehttp://energy.gov/fe/services/petroleum-reserves/strategic-petroleum-reserve


2/51

2

The following note details ORNLs limited study of the economic benefits associated withmaintaining higher levels of SPR drawdown and distribution capability3. Using the scenariosubmodule of ORNLs BenEStock model,4the benefits of possible increases in SPR drawdown

capability are assessed under various world oil supply disruptions and other key assumptions.The benefits (avoided costs) are calculated both with and without the currently maintained sitecapabilities in order to estimate the value of increasing the draw and distribution capacity.

The results presented here depend upon assumptions regarding an inherently unstable oil marketand an uncertain future. Many factors such as U.S. net imports and world oil prices areforecasted to change significantly over the next few years and these expected changes arereflected in the analysis. In the most recent EIA Annual Energy Outlook,5net petroleum importsare expected to decline to 5 MMBD by 2015 and remain at lower levels for many years. Thisdecline in net imports is expected to reduce one component of benefits, the imports-costcomponent, by roughly two-thirds. Despite these import reductions, the projected U.S. demand

for petroleum remains steady at around 19 MMBD. As such, the U.S. is still reliant onpetroleum and potentially sensitive to global price shocks. Projected real oil price levels areprojected to rise from $80 in 2015 to $125 by 2040 (AEO 2014 Base Case). Ceteris paribus,higher real oil prices imply more costly disruptions and greater value to SPR drawdownreliability, thereby increasing overall benefits. Higher base oil prices imply greater absoluteresponse to a disruption, increasing wealth transfer and the impact on GDP.

The drawdown benefits analysis considers possible disruptions to oil supply from fourhistorically unstable global regions. This analysis does not specify where the disruption camefrom, or the likelihood of a particular disruption. Rather the disruption scenarios modelled hereare specified in terms of net disruption size (net of offsets) and length. Three disruptionscenarios are modelled; based on likely size/length combinations:

Large, Short Disruption: 6000 MBD Net Disruption for 3 Months, e.g. 60% of SaudiArabia, or 45% of Other Persian Gulf OPEC, or other.

Moderate, Long Disruption: 4500 MBD Net Disruption for 6 Months, e.g. 45% of SaudiArabia, or 40% of Other Persian Gulf OPEC, or other.

Small, Very Long Disruption: 3000 MBD Net Disruption for 12 Months, e.g. 35% ofSaudi Arabia, 30% of Other Persian Gulf OPEC, or other.

2.0 Analysis Method Updated BenEStock Model

The economic benefits from an improved drawdown capability are measured by comparing thecosts of future oil market disruptions under reduced SPR drawdown capability with the (lower)

3While drawdown refers to the physical capability of the SPR to withdraw oil, distribution refers to the entireprocess of providing oil from the SPR to the market. We use these terms interchangeably in this document.4Successor to the DIS-SPR and DIS-Risk models (DOE, 1990 and Leiby and Bowman, 2005).5U.S. EIA, Annual Energy Outlook 2014.


3/51

3

damages when an incremental 2000 MBD of drawdown capability is added. The difference isthe avoided costs or benefits of drawdown recovery capability.

In this study, this comparison is implemented using the scenario submodule component ofORNLs BenEStock model, a simulation model for estimating the benefits of emergency oilstocks. The BenEStock Model (IEA 2012, ORNL 2012, Leiby, 2015) is an enhanced version ofthe DIS-SPR and DIS-Risk models used in prior SPR size and drawdown studies. 6 It allows forthe reproduction of prior study results, while permitting extensions and the analysis of specific,risk-related outcomes in a simulation or scenario-driven format.

BenEStock characterizes emergency stocks in terms of draw rate capabilities, stock sizes and filland refill rates. It can be used to look at individual disruption scenarios, or run in a Monte Carlorisk analysis fashion to produce estimates of the expected benefit, expected frequency ofdisruptions and use of emergency stocks, the probability of stock exhaustions, and the

probability distribution of economic benefits. For this study, selected scenarios were chosen todeterministically illustrate a range of potential benefits across alternative SPR size/draw rateconfigurations and management strategies.

6See in particular the 1990 DOE/Interagency SPR Size Study (DOE 1990), and the 1999 ORNL study (Leiby andBowman, 1999).


4/51

4

Figure 1: Flow Diagram of the BenEStock Model

Inputs of the BenEStock model are illustrated in the green and red boxes ofFigure 1above.Outcomes, including supply, demand, price, and GDP levels are shown in the yellow boxes.Two world states are simulated side-by-side and benefits, costs, prices, etc. are calculated foreach state. The assumptions for each program, including draw rates, sizes, and behavior can bespecified separately resulting in net benefits or costs of policies or options. A comparison of thenet benefits between the scenarios then yields the value of a particular SPR configuration.

For the scenario analysis here, disruptions are assumed to occur in 2020, with a sensitivity caseof 2030. In any particular scenario (2020 or 2030), a disruption size and duration is defined.First, OPEC spare capacity is used to mitigate the lost supply. Then, based on the scenario

description, a distribution strategy for reserves is used to mitigate the supply disruption. Whileonly the U.S. reserve is used in the default scenarios, non-U.S. reserves are allowed to mitigatethe supply disruption in the sensitivity cases.

Loss of supply is then translated into a price response. The price elasticity depends on severalfactors, including the reference oil price, how long the disruption has lasted, and the size of thedisruption. Finally, the price shock is used to show GDP losses. Total losses due to the disruptionare the sum of GDP losses and wealth transfer as the disruption raises the price of imports.

Current U.S. SPR

Draw (4,400

MBD): Market &

Economic

Post Disruption

Supply, Demand

& Price Levels

BenEStock

Model

Non-U.S.

Emergency Oil

Stock

Capabilities

U.S.

Emergency Oil

Stock

Capabilities

Projected

Excess

Production

Capacity

Reference

Market

Conditions

Oil Supply

Disruption

Scenario Module

(Size and

Duration)

Response Module

(Price Elasticities

of Demand,

Supply, & GDP)

Combined U.S.

Benefits: GDP Loss

Avoidance Oil Import

Savings

INPUTS

OUTPUTS

Reduced U.S.

SPR Draw (750

MBD): Market &

Economic

Post Disruption

Supply, Demand

& Price Levels


5/51

5

Finally, the benefits of the SPR may be taken as the mitigated losses between configurations ofthe SPR for identical scenarios.

Reference Market Conditions

The model requires a reference oil market which is used as a point of departure for simulating oilmarket disruptions. Key factors in determining economic impacts include a projection of netimports and the GDP growth. Other parameters are needed to translate market disruptions intoeconomic losses, such as the "GDP elasticity," which signifies the responsiveness of aggregateproduction to changes in oil prices; and demand elasticities. Where possible, reference oilmarket forecasts are drawn from the EIAs Annual Energy Outlook. Other reference market datainclude gross domestic product for major net importing regions, GDP elasticities for netimporting regions (based upon historical data), world oil price, and regional world oil demandand oil production.

Oil Supply Disruption

When the BenEStock model is used for Monte Carlo simulation, the riskiness of the oil market ischaracterized by the probability, magnitude and duration of gross world oil supply disruptions(prior to offsets). Disruption likelihoods used in a number of studies since 2005 have been basedon the 2005 Energy Modeling Forum (EMF) assessment. With changing world dynamics, thisassessment has been increasingly dated. As updating the oil market risk is well outside the scopeof this analysis, it was decided to present the benefits of the three disruption scenarios describedabove without attempting to weight them by probabilities of disruption size and duration.

Projected Excess Production Capacity

Spare capacity considered in this model is restricted to OPEC countries. The availability offutureOPEC spare capacity is governed by a set of rules:

Assume spare capacity floor of 2 MMBD Saudi Arabia (Stelter, 2012) and 0.83 MMBDOther OPEC (based upon historical average share).

In simulations, the base case assumption is that only 50% of Saudi Spare Capacity wouldbe available for drawdown (per discussions with IEA).

For year 2015, assume 2.8 MMBD of OPEC spare capacity (2 MMBD Saudi, 0.8 Other).

For 2016-2045, any assumed forecast reductions in OPEC supply initially become sparecapacity (production is reduced but production capacity is not). Note, declines in OPECsupply are not normally forecasted in the AEO base cases (sometimes in the AEO highoil price case).

Spare capacity above the floor is vintaged and scrapped at a 15% annual rate (85%annual survival rate).


6/51

6

A figure showing the base case OPEC production (AEO 2014) and OPEC spare production maybe found in the Leiby et al. (2015).

U.S. and Non-U.S. Emergency Oil Stock Capabilities

The U.S. SPR and foreign emergency stockpiles are characterized by attributes including size,various physical operational characteristics, and several categories of cost. While the modelmay allow various sizes, this study varies the maximum drawdown and distribution capability ofthe reserve while holding the size constant. The rates at which the reserves can be filled initially,refilled in the event of a draw, and the size of a disruption are also specified parametrically.Hand-in-hand with the maximum drawdown rate which the reserves may achieve in the event ofa disruption are the implicit rules governing whento draw and how much. The capital costs andoperating costs other than filling costs are specified for each year.

Emergency Oil Stock Sizes and Availability

The U.S. and other IEA countries holding emergency oil stocks are assumed to engage inwithdrawal action only if spare production capacity is not enough to offset simulated supplydisruptions to an acceptable level. More specifically, the net supply loss (after spare productioncapacity has been utilized) must be above a pre-specified threshold (base case is 2 MMBD) totrigger emergency stock releases. The cooperative use of other IEA stocks can either be assumedsimultaneous with the U.S., or to follow with a specified number of months delay. Both of theseassumptions are examined in this analysis.

IEA Emergency Stockholdings

Public stocks were approximately1.5 billion barrels, of which 697 are in the U.S., based on IEAclosing stock levels as of December 2011. Country-specific of stocks may be found in Leiby etal. (2015). Projected emergency stock levels in the future are assumed to adjust with imports andconsumption levels, in accordance with current laws of member countries. A summary of thecountry specific rules may be found in Leiby et al. (2015)

Forecasted levels of publicly-controlled stocks and obligated private industry stocks areestimated using legislative information provided by various IEA documents. If no informationconcerning future levels is available, 2011 stock levels are held constant throughout the modelhorizon. These estimates use country-level consumption and net import forecasts from the EIAsAnnual Energy Outlook where available. When country-level forecasts were not available,estimates were derived using country group forecasts and year 2011 country shares from theEIAs International Energy Statistics tables.


7/51

7

Other Non-U.S. Stockholdings

Mandate industrial stocks were assumed not to be deployed in this analysis. Non-IEA emergencystocks, while considerably large (mostly China and India) are not currently assumed to beavailable for a coordinated emergency drawdown. In the future this may change as the IEA andnon-IEA partners continue to forge closer cooperation and agreements.

Emergency Oil Stock Drawdown Rates

In this study a drawdown threshold is applied, so that not all disruptions lead to stock draw. Thedrawdown threshold value is 2.0 MBBL/d in the base case and variants are considered. Once thethreshold is exceeded, the drawdown rate is governed by the selected drawdown strategy. Theavailable strategies are:

Maximum Sustainable Rate: approximately equal to Reserve Size / Disruption Length

Maximum Rate: Max rate each reserve is technically capable of drawing at.

Maximum Rate for First 3 Months, Sustainable Thereafter

Delayed Drawdown (In combination with 1-3).

No Drawdown

The third strategy was suggested to us by the IEA as the preferable, most-likely approach. Givenuncertainty about disruption duration, initially, the available stocks are drawn down at the ratenecessary to fully offset the net shortfall, up to the maximum technically feasible rate. Thiscontinues for the first 3 months. After 3 months, if the disruption persists and it becomes evident

that the disruption will last longer, it is assumed that the length is probably better-known, and thedrawdown slows to the maximum sustainable rate for the expected remaining duration ofdisruption.

The U.S. design technical maximum drawdown rate is 4.4 MMBD. In the first few weeks of adisruption, as the draw rate ramps up to the maximum level, that rate would be somewhat lower(~3.4 MMBD). The currently achievable SPR drawdown rate may be limited by factorsincluding distribution constraints, in which case simulations can be performed for lower SPRdrawdown and distribution capability. The maximum rates for other IEA government-ownedemergency stocks are assumed to be rates which are capable of exhausting the reserves in 6months.7 IEA obligated industry stocks, if available and drawn down, are thought to be

available sooner and at greater volumes, so are assumed to have a max draw rate equal to a 3month exhaustion rate.

7Efforts are underway to fully incorporate more realistic regional drawdown profiles based upon informationprovided by the IEA. These draw rates tend to have greater surge capacity in the first few months.


8/51

8

Response Modules (price elasticities of demand and GDP)

Oil demand and supply elasticities determine the degree to which oil supply shocks (shortfalls)

translate into oil price increases. The smaller the elasticity (in absolute terms), the greater theprice increase. The literature on demand and supply elasticities displays only limited agreementregarding short run elasticities, and offers very little guidance about the applicable elasticity ofdemand for the unprecedentedly large supply losses and price increases that are contemplatedhere. A brief review of the relevant literature may be found in Leiby et al. (2015).

Approach Used to Model Net Supply and Demand Response to Disruptions

Most of the assessments of elasticity refer to annual periods. However, disruption analysistypically calls for modeling quarterly or even monthly response. Furthermore, studies usually areonly able to estimate a single (constant) elasticity regardless of the magnitude of price increase.

Using a constant elasticity framework for large oil shocks is problematic, because suchelasticities imply extremely high short run prices for large disruptions, yet supply and demandbehavior at very high prices is poorly understood.8

Following prior emergency stock studies (DOE 1990, Leiby and Bowman 2005, 2007) weapplied a variable elasticity for the market response to changing prices. As used, the elasticitymay be interpreted as a short-run elasticity of net world demand, that is, an elasticity of the netresponse of demand and supply. It is widely accepted the short run demand and supplyelasticities are quite small, while longer-run elasticities, which may apply after a number ofyears, are 4 to 10 times larger.

Apart from comparability to prior emergency oil stock studies, the approach has the primaryadvantage that elasticities increase over the course of a disruption (reflecting the process ofslowly increasing adjustment over time to sudden price changes) and they also increase with themagnitude of the disruption supply shortfall. The chosen functional form makes elasticity anincreasing function of both month and disruption size:

+ + + Where:t= initial demand elasticity for 0 month and 0 shortfallStm= Net Oil Shortfall (after offsets) in year t and month mM = Month

and are monthly growth terms which cause the elasticities to increase over the course of adisruption. Since elasticities grow with disruption size, the price response per million barrels perday of supply loss declines as disruptions get increasingly large. This prevents prices from

8For example, a short run (annual) net demand response elasticity or -0.10 would imply that a 20% loss of supplywould increase price by over 9-fold (to above $900) for the first year.


9/51

9

becoming arbitrarily large for very large disruptions, as would occur with a fixed short runelasticity.

The values of and come from the EIA (EIA DIS 2002 and 2007). For the values of (initialundisrupted elasticity) and (change in elasticity with time) we currently use two values, EIADIS 2002/07 and recent estimates by ORNL (2014). Recent (ongoing) research by ORNL andothers has suggested that oil demand has become less responsive to price in recent years. Moreinelastic demand implies that supply shocks produce greater price changes (in percent changeterms) than in the past. It also means that emergency stocks drawn down to offset these supplyshocks have a greater effect on prices and therefore produce larger benefits. The alternativedemand elasticity case uses estimates provided by a meta-analysis, recently conducted by ORNL,of dozens of demand elasticity studies.

The outcome of this approach is a set of time-varying elasticities. The base case and sensitivities

of the elasticities used in the study are shown in Table 1. These modeled base elasticity valuesare comparable to many of the estimates in the literature briefly summarized above, or perhapson the larger side. Moreover, the variable elasticity framework moderates extreme and longerlasting disruptions. All of this suggests that the base model specification may be somewhatconservative about the price effect of disruptions, and stock drawdowns.

GDP Elasticities

Estimates of the GDP impact of oil shocks under the current study involve a direct relationshipbetween price implications and the gross domestic product of oil importing economies. Thisapproach can be described as a summary of the modeling of macroeconomic effects of the oil

supply shocks in the literature, but accounts for the direct and indirect effects on the economy.Oil importing economies are classified into four groups: United States, IEA Europe, Other IEAand Non-IEA. The change in regional GDP (RGDP) associated with an oil supply disruption,RGDP, is calculated based on the estimated change in oil prices, P, as:

11

,oilpriceGDP

ref

refP

PRGDPRGDP

The above equation requires estimates of the oil price-GDP elasticity of, GDP,oilprice, referenceRGDP and oil price, RGDPrefand Pref, and the price change due to the oil disruption event, P.Estimates of the oil price-GDP elasticity used in this analysis are shown in Table 1. Additional

information may be found in Leiby et al. (2015).

Alternative SPR Configurations and Benefits

Each SPR configuration is specified in terms of costs (capital, operations, and maintenance),draw rate capabilities, reserve sizes, and fill and refill rates. For this study, the economicbenefits of four alternative configurations which differ only in the maximum rate at which oil


10/51

10

can be released and distributed were evaluated and compared. The current SPR distributioncapability is less than its prior maximum distribution capacity of 4400 MBD, but the actualcurrent distribution capability is situation-dependent. Therefore, two alternative base

configurations were analyzed, each of which was compared with a 2000 MBD increase indistribution capacity. The first situation involves a recovery of draw and distribution capacityfrom 1200 MBD to 3200 MBD. In the second situation a higher base distribution capability of2000 MBD is raised to 4000 MBD. Each SPR drawdown configuration is subject to the same setof hypothetical oil supply disruptions for scenario analysis. Oil supply disruptions are simulatedagainst reference (i.e. undisrupted) paths for oil prices, U.S. demands, U.S. supplies, and worlddemands through the year 2042. Reference paths track low, base, and high oil price path casesfrom the Annual Energy Outlook of the Energy Information Administration.9

3.0 Key Base Case and Sensitivity Model Assumptions

The table below summarizes the key model assumptions used in the analysis. Highlightedsensitivity variables, for this analysis, are Demand Elasticity and GDP elasticity. All of theresults presented here are bracketed by the high/low benefit combinations of these twoelasticities. Other important sensitivity variables include:

Non-U.S. (other IEA) drawdown, specified in terms of either no draw or delayed draw

AEO 2014 Case

Disruption Year

Table 1: Key BenEStock Assumptions

Variable/Parameter Description Default Value Sensitivity Option

Non-U.S.Drawdown Delay

Number of months afterdisruption until drawdownby Non-U.S. emergencyreserves

Default is 24 month delay(no use). This study isprimarily focused onscenarios in which the U.S.acts independently.

2-5 month delays considered.

U.S. DrawdownStrategy

U.S. drawdown profileduring the disruption.

Current assumed strategyis max draw (up to the

disruption size) for the first3 months, sustainable drawthereafter for theremainder of the disruption(IEA Study assumption).

Current assumed strategyreflects one approach given

uncertainty about shock length.Other choices Max draw (for aslong as possible) andsustainable draw (requiresmarket foresight).

9Years 2041 and 2042 are extrapolated.


11/51

11

Table 1: Key BenEStock Assumptions

Variable/Parameter Description Default Value Sensitivity Option

AEO 2014 OilMarketAssumptions

Reference Oil market prices,demands, supplies, and GDPestimates.

AEO 2014 Base case. AEO 2014 alternate projectionsof low and high U.S. LiquidResources

Demand/SupplyElasticity Choice

World and U.S. DemandElasticities.

Default is Range of demand elasticities used in 2012 study forIEA and prior studies (derived from EIA/DIS 2002/7approach) to smaller value from 2014 ORNL 2014 MetaAnalysis of Demand Elasticity.

Default Value #1EIA/DIS 2002/7

0 MMB Loss-0.08 after 1 month-0.11 after 12 months,-0.10 average over 12months.

Default Value #2ORNL 2014

0 MMB Loss-0.05 first month,-0.07 after 12 months,-0.06 average over 12 months.

GDP Elasticity U.S. GDP response to pricechanges due to supplyshocks.

Default is Range based on distribution and mean values from2012 study for IEA (supported by ORNL 2012 MetaEstimate). Range used here is the endpoints of the 68%Confidence interval around the mean.

Disruption Year

Severity of the net disruptiondepends upon a confluenceof forecast variables,

disruption sizes, and reservedrawdowns capabilities.These variables changeyearly as will the benefits.The disruptions simulated inthis study are assumed tooccur in a specific year.

Year 2020. Year 2030.

Saudi Arabia %Excess CapacityAvailable

% of undisrupted SaudiArabian excess capacityapplied to disruptions.

50% (IEA StudyAssumption).

None specified for this analysis

MandatedIndustrial StockAvailability (0-100%)

If Non-U.S. IEA stocksdrawn down, what % ofEuropean and Asian IEAmandated industrial stockwill be available fordrawdown.

0% None specified for this analysis


12/51

12

4.0 Results

4.1 Results for Expansion of Draw/Distribution Capability from 1200 MBD to 3200 MBD

The benefits estimates for three potential disruption sizes are given in Table 2 below. These arealso presented graphically in the Appendix. The results suggest that under a wide range ofcircumstances, the total benefits of increasing the maximum drawdown and distributioncapability rate by 2000 MBD (from 1200 MBD to 3200) can result in tens to hundreds of billionsof dollars in savings, in the event of a large world oil supply disruption.

Table 2: High and Low Total Benefits (Billions $ Undiscounted)

For Draw/Distribution Rate Expansion from 1200 MBD to 3200 MBD

Note: High and Low values for each size/length/sensitivity combination is derived from upperand lower Demand and GDP elasticity values assumed in the analysis.

Net Disruption Size (MBD) 6000 4500 3000

Net Disruption Length (Months) 3 6 12

Base Case (without Foreign Draw) $58 - $103 $111 - $213 $53 - $106

Foreign Draw (with 5 Month Delay) $58 - $103 $49 - $89 $31 - $61


AEO 2014 Lower U.S. Resources Case $62 - $108 $117 - $222 $55 - $110

AEO 2014 Higher U.S. Resources Case $49 - $90 $94 - $187 $45 - $94

2030 Disruption Year $68 - $122 $131 - $253 $63 - $129

Sensitivity cases with additional supplies of oil, such as Foreign Drawdown (after delay) orassumed Higher U.S. Liquid Resources, result in lower total benefits across all of the assumedsize/length combinations. If correctly timed, the Foreign Drawdown can be an effectivecomplement to the U.S. response, in which case the benefits of additional SPR drawdowncapabilities are reduced for the disruptions sizes considered here. The reduction in benefits underthe Higher U.S. Resources case is modest, about 10% to 15%. Other sensitivities such as theLower U.S. Liquids Resources Case or 2030 disruption year lead to moderately higher, albeitundiscounted, total benefits of draw and distribution capability expansion. The benefits arepresented in constant real dollars, but in undiscounted terms because we are only looking at theeffects of hypothetical disruptions in two specific years (2020 and 2030), and not discountingback to the present and aggregating into a net present value.


13/51

13

4.2 Results for Expansion of Draw/Distribution Capability from 2000 MBD to 4000 MBD

Under some disruption conditions, depending on the pattern of domestic or imported flows

interrupted, the base drawdown and distribution capacity for the SPR could be higher, 2000MBD rather than 1200 MBD. Table 3 shows the estimated range of benefits for increasingmaximum drawdown and distribution capability by 2000 MMB from this higher starting level,that is, from 2000 MBD to 4000 MBD. Again these estimates are constructed for three potentialmajor disruptions in 2020 of varying sizes and durations.

As would be expected, in the particular base disruption case where the starting drawdown anddistribution capability is higher, the incremental benefits of greater capability are somewhat less.In cases of a disruption lasting under one year (e.g. in the 3 and 6 month disruption cases), theincremental benefits of a 2000-to-4000 MBD distribution capability increase are only modestlylower (4%-19%) than those of the 1200-to-3200 increase. However, for a long disruption, the

optimal draw rate can be governed by the rate that exhausts the SPR. In the hypothetical 12month event in Table 3, the incremental benefits of a faster drawdown could be zero, or evenslightly negative, if the U.S. must act alone for the entire disruption period. That is because, forthe 692 MMB SPR, the maximum sustainable rate over 12 months, i.e. the exhaustion rate, is 1.9MMB.10 So in this particular situation of a long disruption, where the U.S. acts alonethroughout, there is less value to draw/distribution capability expansion. However, the results inTable 3 also show that even in the long (12 month) disruption, if there is a non-U.S. strategicreserve drawdown that joins that of the U.S. drawdown after a few months, the higherdistribution capability provided by a 2000-to-4000 MBD expansion could be worth tens ofbillions of dollars by allowing a larger drawdown early in the disruption.

10Given the modeled demand elasticity and price dynamics, the optimal drawdown strategy is an even,sustainable rate over the course of the disruption. It is not generally possible to know the duration of a disruption atits outset, although it may become clearer as the disruption progresses. So we currently model drawdown strategyas Max draw rate for first 3 months, Sustainable rate thereafter. Therefore, increasing the maximum draw rate

capability from 2 MMBD to 4 MMBD allows the model to draw initially at greater than the 12 month averagesustainable rate, (frontloading the draw rate and causing a lower draw later during long sustained events that wouldexhaust the SPR). In some events this could incur slightly lower benefits than the slower, but evenly-sustaineddraw.


14/51

14

Table 3: High and Low Total Benefits (Billions $ Undiscounted)

For Draw/Distribution Rate Expansion from 2000 MBD to 4000 MBDNote: High and Low values for each size/length/sensitivity combination is derived from upper

and lower Demand and GDP elasticity values assumed in the analysis.

Net Disruption Size (MBD) 6000 4500 3000

Net Disruption Length (Months) 3 6 12

Base Case (without Foreign Draw) $54 - $99 $97 - $192 $0*




15/51

15

References

Leiby, Paul and David Bowman (2000a) The Value of Expanded SPR Drawdown Capability,

Oak Ridge National Laboratory, Final Report, October 18.

Leiby, Paul and David Bowman, (2000b) The Value of Expanding the U.S. Strategic PetroleumReserve,Oak Ridge National Laboratory, ORNL/TM-2000/179, January 23.

Leiby, Paul, David Bowman, and Donald W. Jones (2002) Improving Energy Security Throughan International Cooperative Approach to Emergency Oil Stockpiling,Proceedings of the 25thAnnual IAEE International Conference, June 26-29, Aberdeen, Scotland.

Leiby, Paul, David Bowman, Debo Oladosu, Rocio Uria-Martinez (2015). BenEStock Modelfor SPR Analysis - Model Documentation,Oak Ridge National Laboratory, Revised Draft,

January 29.

Leiby, Paul, David Bowman, Debo Oladosu, Rocio Uria-Martinez, and Ken Vincent (2012)Benefits of Emergency Oil Stocks, A Study of IEA Stocks and Benefits,Oak Ridge NationalLaboratory, Reported prepared for the U.S. Department of Energy and the International EnergyAgency.

Leiby, Paul, David Bowman, Debo Oladosu, Rocio Uria-Martinez, and Ken Vincent (2013) TheValue of Strategic Oil Stocks Under Reduced U.S. Oil Imports,Presented at the meeting of theU.S. Association for Energy Economics, July 29.

U.S. Department of Energy/Interagency Working Group (1990) Strategic Petroleum ReserveAnalysis of Size Options,DOE/IE-0016, February.

U.S. Government Accountability Office (2006) Strategic Petroleum Reserve: Available Oil CanProvide Significant Benefits, but Many Factors Should Influence Future Decisions about Fill,Use, and Expansion,Report to Congressional Requesters, GAO-06-872, August.


16/51

16

Appendix A. Detailed Results

Table A1: High and Low Benefits Components for the Base Case (Billions $ Undiscounted)

(for draw/distribution rate expansion from 1200 MBD to 3200 MBD)

Net DisruptionSize (MBD)

Net DisruptionLength (Months)

Net ImportBenefits

GDP Benefits Total Benefits

6000 3 $23 - $30 $35 - $73 $58 - $103

4500 6 $39 - $54 $72 - $159 $111 - $213

3000 12 $17 - $24 $36 - $82 $53 - $106

$0

$25

$50

$75

$100

$125

$150

$175

$200

$225

$250

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35

$2

010/BBL

Month

World Oil Price (1200 v 3200 MBD Draw Capability)

6000 MBD Net Disruption Lasting 3 Months

U.S. Max Draw 1.20 MMBD: Lower World Demand Elasticity, Higher GDP Response to Price

U.S. Max Draw 1.20 MMBD: Higher World Demand Elasticity, Lower GDP Response to Price




17/51

17

$0

$25

$50

$75

$100

$125

$150

$175

$200

$225

$250

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35

$2010/BBL

Month



U.S. Max Draw 1.20 MMBD: Lower World Demand Elasticity, Higher GDP Response to PriceU.S. Max Draw 1.20 MMBD: Higher World Demand Elasticity, Lower GDP Response to Price



$0

$25

$50

$75

$100

$125

$150

$175

$200

$225

$250

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35

$2010/BBL

Month








18/51

18

$0

$5

$10

$15

$20

$25

$30

$35

$40

$45

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35

$2010/BBL

Month

World Oil Price Differential Range (1200 v 3200 MBD Draw

Capability): 6000 MBD Net Disruption Lasting 3 Months

Lower World Demand Elasticity, Higher GDP Response to Price

Higher World Demand Elasticity, Lower GDP Response to Price

$0

$5

$10

$15

$20

$25

$30

$35

$40

$45

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35

$2010/BBL

Month






19/51

19

$0

$10

$20

$30

$40

$50

$60

$70

$80

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35

$2010/BBL

Month





0

50

100

150

200

250

300

350

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35

Billions$2010

Month

U.S. Cumulative Total Benefits for 1200 v 3200 MBD Draw Capability:





20/51

20

0

50

100

150

200

250

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35

Billions$2010

Month

U.S. Cumulative Total Benefits for 1200 v 3200 MBD Draw Capability:




0

50

100

150

200

250

300

350

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35

Billions$2010

Month

U.S. Cumulative Total Benefits for 1200 v 3200 MBD Draw Capability





21/51

21

Appendix B: BenEStock Documentation


22/51

BenEStock Model for SPR AnalysisModel Documentation

Revised Draft 4.0

February 4, 2015

Paul Leiby, David Bowman, Debo Oladosu, and Rocio Uria-MartinezOak Ridge National Laboratory


23/51

xxiii

Table of Contents1. INTRODUCTION, 12. DESCRIPTION OF THE BENESTOCK MODELING APPROACH, 2

2.1 Assumptions in the BenEStock Model, 2

2.1.1 Reference Market Conditions, 3

2.1.2 Oil Market Risk, 3

2.1.3 Spare Oil Production Capacity, 6

2.1.4 Emergency Oil Stock Capabilities, 9

2.1.4.1 Emergency Oil Stock Sizes and Availability, 9

2.1.4.2 Emergency Oil Stock Drawdown Rates, 14

2.1.5 Market Responsiveness (price elasticities of demand), 15

2.2 Net Benefit Estimation Approaches, 23

2.2.1 Monte Carlo Simulation Approach for Estimating 30-Year Benefits, 23

2.2.2 Scenario Analysis Approach for Estimating Benefits for a Single Selected Year, 24

3 COMPACT STATEMENT OF MODEL EQUATIONS, 24


24/51

1

1. INTRODUCTION

The BenEStock model is used to simulate the economic benefit of emergency stocks under

current and alternative configurations for a variety of sensitivity cases.11

BenEStockcharacterizes emergency stocks in terms of draw rate capabilities, stock sizes and fill andrefill rates. It can be used to look at individual disruption scenarios, or run in a Monte Carlorisk analysis fashion to produce estimates of the expected benefit, expected frequency ofdisruptions and use of emergency stocks, the probability of stock exhaustions, and the

probability distribution of economic benefits. These distributions are generated usingthousands of sample iterations.

Oil supply disruptions are simulated against reference paths for oil prices, demands andsupplies. Reference oil market paths track the IEAs World Energy Outlook, New Policies

Scenario. Within each year over the model horizon (2013 to 2042), an oil supply disruption

may occur. The timing, size and the length of the disruption can be manually specified forindividual scenario analysis, or in a full Monte Carlo analysis disruptions are randomlysampled from the underlying probability distributions. Disruption impacts are modeled on amonthly basis, for up to 36 months after disruption onset with disruptions up to 18 months inlength.

This gross oil supply disruption is directly offset by two exogenously specified sources:available spare world oil production capacity and short-run demand switching (generally verysmall). If the net disruption after these offsets is greater than the specified drawdownthreshold level, the emergency stocks will coordinate action in an attempt to offset it.12Drawdown rates and timing for each stock type (i.e. U.S. and other IEA public stocks andobligated industry stocks) are limited by the specified technical maximum drawdown rate forthat year, the specified drawdown rule or strategy, and by the rate of exhaustion. After adrawdown, the emergency stocks are assumed to be refilled at exogenously specified refillrates.

The oil shortfall is calculated as the size of the remaining disruption after offsets andemergency stock draw. If the oil shortfall is greater than zero, world oil price is affected.Under the base case assumptions, we consider the possibility that market dislocation,speculative behavior, or a risk premium will prevent stocks from completely eliminating the

price increase, even if the net disruption is fully offset by stocks. World oil price is

determined assuming that world demand responds to the price change (according to specifiednet demand elasticities), non-OPEC supply is essentially fixed, and that the price increasessufficiently for demand to accommodate the oil shortfall.

11BenEStock is an enhanced version of the DIS-Risk model used by ORNL in prior SPR size and drawdownstudies. The DIS-Risk Model was a risk-analysis oriented implementation of the DIS-SPR model used in the1990 DOE/Interagency SPR Size Study. It allowed for the reproduction of DOE90 study results, while

permitting extensions and the analysis of specific, risk-related outcomes. See in particular the 1990DOE/Interagency SPR Size Study (DOE, 1990), 1999 ORNL study (Leiby and Bowman, 1999) and 2005 study

(Leiby and Bowman, 2005).12E.g. a drawdown threshold of 2 million barrels per day was used in the base case for the 2013 IEA study.


25/51

2

Oil price increases are then translated into economic costs to society. These costs arecomposed of Gross Domestic Product (GDP) losses and net oil import costs.We begin with a description of the model approach, summarizing the model, and including a

compact statement of its equations. Certain details are omitted from the text but included inthe appendices. The most important categories of model assumptions are identified.

2. DESCRIPTION OF THE BENESTOCK MODELING APPROACH

2.1 Assumptions in the BenEStock Model

There are six categories of parameters in the BenEStock model. The six key inputs forestimating the strategic economic benefits of stockholding are:

Reference Market Conditions Spare Oil Production Capacity

Emergency Oil Stock Capabilities

Oil Supply Disruption Likelihoods

Market Responsiveness (price elasticities of supply and demand)

Macroeconomic Sensitivity to Shocks (GDP elasticities)

Figure 2: Flow Diagram of the BenEStock Model.

World State #1

Market &Economic Post

DisruptionSupply, Demand,

& Price Levels

BenEStockModel

Non-U.S.Emergency Oil

Stock

Capabilities

U.S.

Emergency Oil

StockCapabilities

Projected

Excess

Production

Capacity

Reference

Market

Conditions

Oil Supply

Disruption

Scenario Module

(Size and

Duration)

Response Module

(Price Elasticities

of Demand,

Supply, & GDP)

Combined U.S.

Benefits: GDP Loss

Avoidance Oil Import

Savings

INPUTS

OUTPUTS

World State #2

Market &Economic Post

DisruptionSupply, Demand,

& Price Levels


26/51

3

Inputs of the BenEStock model are illustrated in the green and red boxes ofFigure 2above.Outcomes, including the supply, demand, price and GDP levels are shown in the yellow

boxes. Two world states are simulated side-by-side and benefits, costs, prices, etc. are

calculated for each state. The assumptions for each program, including draw rates, sizes, andbehavior can be specified separately resulting in net benefits or costs of policies or options.A comparison of the net benefits between the scenarios then yields the value of a particularSPR configuration.

2.1.1 Reference Market Conditions

The beliefs about the future oil economy are characterized by the reference (undisrupted)price and quantity paths for oil during the thirty-year period over which the model evaluatesthe SPR. The model assumes reference oil market conditions to determine the undisrupted

market state which is used as a point of departure for simulating oil market disruptions. Netimports (demandsupply) and GDP are used to determine the magnitude of economic costs.These assumptions also include parameters determining the economic response to an oil pricerise: the "GDP elasticity," which signifies the responsiveness of aggregate production tochanges in oil prices; and demand elasticities. Where available, reference oil marketforecasts are drawn from the EIAs Annual Energy Outlook. Reference market data include

gross domestic product for major net importing regions, GDP elasticities for net importingregions (based upon historical data), world oil price, and regional world oil demand and oil

production.

2.1.2 Oil Market Risk

When the BenEStock model is used for Monte Carlo simulation, the riskiness of the oilmarket is characterized by the probability, magnitude and duration of gross world oil supplydisruptions (prior to offsets). Naturally, these likelihoods are difficult to assess. Disruptionlikelihoods can be entered based on available assessments, but a full expected-value analysisrequires probabilities distributions for the full range of sizes and durations of potentialdisruptions. Disruption likelihoods used in a number of studies since 2005 have been basedon the 2005 Energy Modeling Forum (EMF) assessment.The EMF 2005 Expert Panel study quantified oil disruption risks over the course of three

rounds of workshops, from 2004 to 2005. These workshops constituted a formal riskassessment using the methodologies of decision analysis and influence diagrams. The focuswas on potential losses of supply from four key regions:

Saudi Arabia

Other Persian Gulf

West of Suez (a region including Nigeria, Venezuela, Mexico and others)

Russia and Caspian Region

The disruption probability assessment was based on the assessment of conditional eventsequences, that is, some supplier groups outcomes can depend conditionally on other supplier

outcomes. The assessment also explicitly considers disruption duration along with sizeprobability, so that there are conditional disruption size probabilities for each disruption


27/51

4

length. The disruption lengths considered range from less than 6 months to more than 18months. The assessment also considered spare capacity, and the availability and likelihood ofspare capacity use depending on the disruption scenario. Even though the focus of the EMF

assessment was disruption risk for the next decade, these same risk probabilities (annualized)were maintained forpercentageshortfalls throughout the horizon of analysis. Informationfrom EIAsAnnual Energy Outlookis used to project the evolving share of world supply met

by each of the regions at risk.

The BenEStock Disruption Risk Characterization includes:

4 supply regions at risk corresponding to the EMF2005 focus regions

3 disruption lengths (taken as 3, 12 and 18 months), not equally probable

EMF 2005 Discrete disruption probabilities by size (as percentage of supply), distinctfor each region and conditional on disruption length.

Uniform distribution of +/-10% around the discrete EMF disruption size. Convertsthe disruption probabilities from discrete to semi continuous, avoiding issues witharbitrarily missing threshold boundaries while maintaining the EMF expected values.

EIA midcase projections of world spare production capacity broken out by region.Each regions spare capacity is available if that region is undisrupted, unavailable

otherwise.13

To reflect that disruption risk might rise or fall as supply becomes more or less concentratedin volatile regions, disruption sizes are specified as percentage losses of supply from variousregions.

13

This is slightly more optimistic than the EMF 2005 assessment, which identified some other conditions underwhich spare capacity may not be made available.


28/51

5

Table B1: EMF 2005 Discrete disruption probabilities

Saudi Arabia

Size/Length 0% ofProduction 20% of Production 50% of Production 90% of Production

3 Months 50.47% 2.74% 0.62% 0.22%

12 Months 21.59% 0.43% 0.13% 0.07%

18 Months 23.39% 0.21% 0.07% 0.06%

Any Length 95.45% 3.38% 0.83% 0.35%

West of Suez

Size/Length0% ofProduction 20% of Production 50% of Production 90% of Production

3 Months 32.30% 4.48% 0.82% 0.03%

12 Months 44.56% 2.78% 0.74% 0.03%18 Months 14.07% 0.13% 0.06% 0.00%

Any Length 90.92% 7.38% 1.63% 0.07%

Other Persian Gulf


3 Months 33.67% 7.62% 0.78% 0.09%

12 Months 28.00% 1.80% 0.39% 0.05%

18 Months 26.79% 0.61% 0.19% 0.03%

Any Length 88.46% 10.03% 1.35% 0.16%



3 Months 86.87% 2.16% 0.46% 0.00%

12 Months 7.96% 0.03% 0.01% 0.00%

18 Months 2.51% 0.00% 0.00% 0.00%

Any Length 97.35% 2.18% 0.47% 0.00%

The combination of the semi-continuous, independent EMF-based regional probabilities allowsfor an unlimited number of possible market states of various sizes and lengths. However, asexpected, of the multiple outcomes, 0% production loss is the most common, occurringapproximately 75% of the time.The figure below gives the cumulative probability distribution for gross disruptions (prior tooffsets) for the year 2020. As the figure shows, the most common events are relatively smallproduction losses from the Other Persian Gulf region or a small West of Suez disruption. Russianand Caspian region disruptions are the least common with zero loss occurring more than 97% ofthe time. Specific regional events can occur individually or in combination, each for various


29/51

6

lengths. 94% of all disruptions are 4% or less of world supply and 99% of disruptions are 10% ofor less.

Figure 3: Disruption Probability Distribution for Year 2020. Sizes are for gross disruptions (prior to excess

capacity and emergency stocks), and probability given is the likelihood of a disruption greater than or equalto that size (survival function form).

2.1.3 Spare Oil Production Capacity

In BenEStock, the initial response mechanism to an oil supply disruption is the output increaseby other producers with spare capacity. More specifically, the spare capacity considered in thismodel is restricted to OPEC countries. Within the cartel, three different subregions areconsidered (Saudi Arabia, Other Persian Gulf and West of Suez).

When possible, the BenEStock model inputs are derived from the EIAs Annual Energy Outlook.Unfortunately, neither the AEO nor the EIAs International Energy Outlook forecasts spare oilproduction capacity.14 For the key swing producer in OPEC, Saudi Arabia, this study assumes

14For years the IEA estimated future spare oil production capacity but this feature was discontinued in 2007 with theadvent of a new IEO modelling approach.

0%

5%

10%

15%

20%

25%

30%

0% 5% 10% 15% 20%

Probabilityo

fDisruption>=X%

X% of Year 2020 World Supply

Saudi Arabia

Other Persian Gulf

West of Suez


Total At Risk Supply


30/51

7

an effective spare capacity floor of two MMBD (Stelter, 2012), consistent with other long-termforecasted values cited elsewhere.15

Shares of OPEC spare capacity by OPEC sub-region are derived from historical data publishedin the IEA Oil Market Report (OMR). Effective OPEC Capacity is defined as unused capacitythat can be accessed within 30 days and sustained for 90 days. It excludes spare capacity indisrupted countries. Figure A-2 combines the historical and historical average (2009-2013)shares of effective OPEC spare production capacity.

Figure 4: Shares of Effective OPEC Spare Production Capacity

15E.g. U.S. EIA International Energy Outlook (2011) and Dourian (2011): "Saudi Arabia will continue with itscurrent plan to maintain spare production capacity at levels between 1.5 and 2.0 million barrels per day." (EIA'sInternational Energy Outlook 2011, p. 35); and "'Given that the kingdom has total production capacity of 12.5million b/d, of which 72% is currently being exploited, actual production to end 2030 will be maintained at thisassumed level within this available capacity,' he added. 'This will be enough to meet anticipated production (tosatisfy both export and local demand requirements) while maintaining through this period spare capacity that willreach 1.7 million b/d by 2030, within the kingdom's declared aim of maintaining spare capacity of 1.5-2 million b/d,'Moneef said."(Dourian, 2011)

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

Saudi Arabia 2009-2013 Average

Other Persian Gulf 2009-2013 Average

West of Suez 2009-2013 Average


31/51

8

The availability offutureOPEC spare capacity is governed by a set of rules:

Assume spare capacity floor of 2 MMBD Saudi (Stelter, 2012) and 0.83 Other-OPEC

(Historical Average Share). In simulations, base case assumption is that only 50% of Saudi Spare Capacity available

for drawdown (per discussions with IEA).

For year 2015, assume 2.8 MMBD of OPEC spare capacity (2 MMBD Saudi, 0.8 Other).

For 2016-2045, any assumed forecast reductions in OPEC supply initially become sparecapacity (production is reduced but production capacity is not). Note, declines in OPECsupply are not normally forecasted in the AEO base cases (sometimes in the AEO highoil price case).

Spare capacity above the Floor is vintaged and scrapped at a 15% annual rate (85%annual survival rate).

The base case OPEC production (AEO 2014) and OPEC spare production capacity are given inthe figure below.

Figure 5: OPEC Production and Spare Production Capacity, Base Case.

0

10

20

30

40

50

60

70

2015 2020 2025 2030 2035 2040 2045

MMBD

Year

Spare Production Capacity

Production


32/51

9

2.1.4 Emergency Oil Stock Capabilities

The U.S. SPR and foreign emergency stockpiles are characterized by attributes including size,

various physical operational characteristics, and several categories of cost. The current andtarget reserve sizes for the SPR can be varied in the model. The rates at which the reserves canbe filled initially, refilled in the event of a draw, and drawn down in the event of a disruptionalso are specified parametrically. Hand-in-hand with the maximum draw-down rate which thereserves may achieve in the event of a disruption are the implicit rules governing whento drawand how much. An additional parametric constraint in the model is a time path of maximumreserve capacity for each year across the time horizon. The capital costs and operating costsother than filling costs also are specified for each year.

2.1.4.1 Emergency Oil Stock Sizes and Availability

The U.S. and other IEA countries holding emergency oil stocks are assumed to engage inwithdrawal action only if spare production capacity is not enough to offset simulated supplydisruptions to an acceptable level. More specifically, the net supply loss (after spare productioncapacity has been utilized) must be above a pre-specified threshold (base case is 2 MMBD) totrigger emergency stock releases. The cooperative use of other IEA stocks can either be assumedsimultaneous with the U.S., or to follow with a specified number of months delay. Cases wherenon-U.S. stocks are unavailable or unused may also be evaluated.

IEA Emergency Stockholdings

Public IEA stocks for end of year 2011 are presented in Figure 5, based on IEA closing stocklevels as of December 2011. Projected emergency stock levels in the future are assumed to adjustwith imports and consumption levels, in accordance with current laws of member countries, assummarized in Table 2.


33/51

10

Figure 6: Year 2011 IEA Public Emergency Stocks

Forecasted levels of publicly-controlled stocks and obligated private industry stocks areestimated using legislative information provided by various IEA documents. If no information

concerning future levels is available, 2011 stock levels are held constant throughout the modelhorizon. The table below contains detailed information on legislated public and privateemergency stockholdings for each country.

0

100

200

300

400

500

600

700

800

UnitedStates

Japan

Germany

France

Korea,

South

Spain

Netherlands

Belgium

CzechRepublic

Ireland

Finland

Hungary

Poland

Portugal

Denmark

Slovakia

NewZealand

Australia

Austria

Canada

Greece

Italy

Luxembourg

Norway

Sweden

Switzerland

Turkey

UnitedKingdom

MMB


34/51

11

Table B2: Legislated Public and Private Emergency Stockholdings

Note: IEA commitment is roughly equivalent to 90 days of net imports. EU commitment isapproximately 90 days of consumption of the 3 main products (gasoline, middle distillates, and

fuel oils).Country Public Stock Estimate Obligated Private Stock Estimate

Austria 0 MMB100 days of oil imports of 3 EU productcategories

Belgium28 MMB Starting Value. MaxEU and IEA commitments 0 MMB (No Obligation)

Czech Republic 15 MMB 0 MMB (No Obligation)

Denmark

8 MMB Starting Value. 81 daysconsumption of 3 EU productcategories. 70% held byStockholding Association (Public

Stock).

81 days consumption of 3 EU productcategories. 70% held by Stockholding

Association (Public Stock).

Finland 10 MMB 60 Days of Net Imports

France

98.5 days consumption of 3 MainProducts using 75/25Public/Private Split.

98.5 days consumption of 3 MainProducts using 75/25 Public/PrivateSplit.

Germany

182 MMB Starting Value. MaxEU and IEA commitments, noIndustry Stocks. 0 MMB (No Industry-Held Stocks)

Greece 0 MMBIndustry Obligation of 90 days of NetImports of 3 EU product

HungaryMax of EU and IEACommitments 0 MMB (No Obligation)

IrelandMax of EU and IEACommitments 0 MMB (No Obligation)

Italy 0 MMB Max of IEA and EU Commitments

Luxembourg 0 MMB Max of IEA and EU Commitments

Netherlands 31 MMB 4.4 MMB

Norway 0 MMB20 days of oil domestic consumption of 3EU product categories.

Poland

Max of IEA and EUCommitments. 90 Days of

Consumption of 3 Main Productswith 76/14 Day Public/PrivateSplit + 30 days of IndustryObligated LPG Storage.

Max of IEA and EU Commitments. 90

Days of Consumption of 3 MainProducts with 76/14 Day Public/PrivateSplit + 30 days of Industry ObligatedLPG Storage


35/51

12

Table B2: Legislated Public and Private Emergency Stockholdings

Note: IEA commitment is roughly equivalent to 90 days of net imports. EU commitment isapproximately 90 days of consumption of the 3 main products (gasoline, middle distillates, and

fuel oils).Country Public Stock Estimate Obligated Private Stock Estimate

Portugal

Max of IEA and EUCommitments. Public AgencyHolds remainder IEA/EUcommitments after IndustryObligations. Equal to 30% (27Days).

Max of IEA and EU Commitments.Industry Obligations equal to 70% (63Days).

SlovakiaMax of IEA and EUCommitments. 0 MMB (No Obligation)

Spain

Max of EU and IEA

Commitments, 50/50Public/Private split

Max of EU and IEA Commitments,50/50 Public/Private split

Sweden 0 MMB

Max of 91 days of consumption of 3main products and 91 days of netimports.

Switzerland 0 MMB

135 days of imports for gasoline, dieseland heating oils; 90 days of jet fuelimports

Turkey 0 MMB 90 days of Net Imports

United Kingdom 0 MMB

Max of EU 67.5 days of supply forrefineries and 58 days of net imports forimporters, and IEA 90 Day ImportCommitment.

Australia 0 MMB 0 MMB (No Obligation)

Canada 0 MMB 0 MMB (No Obligation)

Japan 324 MMB 230 MMB

Korea, South 90 MMB 40 Days of Consumption (All Products).

New Zealand 1 MMB 0 MMB (No Obligation)

United States 692 MMB 0 MMB (No Obligation)

Forecasted estimates of IEA emergency stocks are summarized in the figures below. These

estimates use country-level consumption and net import forecasts from the EIAs Annual EnergyOutlook where available. When country-level forecasts were not available, estimates werederived using country group forecasts and year 2011 country shares from the EIAs International

Energy Statistics tables.Non-IEA emergency stocks, while considerably large (mostly China and India) are not currentlyassumed to be available for a coordinated emergency drawdown. In the future this may changeas the IEA and non-IEA partners continue to forge closer cooperation and agreements.


36/51

13

Figure 7: IEA Public Oil Stocks (MMB)

0

200

400

600

800

1,000

1,200

1,400

1,600

1,800

2000

2005

2010

2015

2020

2025

2030

2035

2040

2045

MMB

Year

IEA Non-Europe, Non U.S.

United States

IEA Europe


37/51

14

Figure 8: IEA Obligated Industry Oil Stocks (MMB)

2.1.4.2 Emergency Oil Stock Drawdown Rates

For this study, except in selected sensitivity cases, the results presented apply a drawdownthreshold so that not all disruptions lead to stock draw. The drawdown threshold value is 2.0MMBD in the base case and variants are considered. Once the threshold is exceeded, thedrawdown rate is governed by the selected drawdown strategy. The available strategies are:

Maximum Rate: Max rate each reserve is technically capable of drawing at.

Maximum Rate for First 3 Months, Sustainable Thereafter

Delayed Drawdown (In combination with 1-3).

No Drawdown

The third strategy was suggested to us by the IEA as the preferable, most-likely approach. Givenuncertainty about disruption duration, initially, the available stocks are drawn down at the ratenecessary to fully offset the net shortfall, up to the maximum technically feasible rate. Thiscontinues for the first 3 months. After 3 months, if the disruption persists and it becomes evidentthat the disruption will last longer, it is assumed that the length is probably better-known, and thedrawdown slows to the maximum sustainable rate for the expected remaining duration ofdisruption.

0

200

400

600

800

1,000

1,200

2000

2005

2010

2015

2020

2025

2030

2035

2040

2045

MMB

Year

IEA Non-Europe, Non U.S.

United States

IEA Europe


38/51

15

The U.S. design technical maximum draw down rate is 4.4 MMBD. In the first few weeks of adisruption, as the draw rate ramps up to the maximum level, that rate would be somewhat lower(~3.4 MMBD). The currently achievable SPR drawdown rate may be limited by factors

including distribution constraints, in which case simulations can be performed for lower SPRdrawdown and distribution capability. The maximum rates for other IEA government-ownedemergency stocks are assumed to be rates which are capable of exhausting the reserves in 6months.16 IEA obligated industry stocks, if available and drawn down, are thought to beavailable sooner and at greater volumes, so are assumed to have a max draw rate equal to a 3month exhaustion rate.

2.1.5 Market Responsiveness (price elasticities of demand)

Oil demand and supply elasticities determine the degree to which oil supply shocks (shortfalls)

translate into oil price increases. The smaller the elasticity (in absolute terms), the greater theprice increase.

Estimates of Short-run Elasticity from the Literature

The literature on demand and supply elasticities displays only limited agreement regarding shortrun elasticities. It offers very little guidance about the applicable elasticity of demand for theunprecedentedly large supply losses and price increases that are contemplated here. Many recentsurveys suggest quite low short run elasticities of supply and demand. Atkins and Jazayeri(2004) survey estimates of short-run oil demand elasticity, and report values from -0.0 to0.11.Cooper (2003) estimates the short run elasticity of demand for 23 OECD countries with resultsbetween 0.0 (or slightly positive) and -0.109, with a mean result of -0.046. Smith (2009) obtainssimilar results for world regions. There are fewer studies of short-run supply elasticity, althoughthe estimates are generally even smaller. Brown and Huntington (2010) recently assessed theevidence in preparation for modeling oil supply shocks and selected a midpoint world demandelasticity of -0.055 (range of0.02 to0.09) and a short-run elasticity of oil supply midpointvalue of 0.05 (range 0.025 to 0.075). These are separate demand and supply elasticities, and theyalso noted that short run income elasticity of demand, coupled with the reduction in income fromprice increases, can augment short-run global demand response. Combining supply, demand, andincome effects to estimate the net demand elasticity, they selected a midpoint annual elasticity of0.136 to find the overall price response needed to close the gap between production andconsumption that is created by a production disruption. (Brown and Huntington 2010:14)

One other very recent estimate of global demand and income elasticities is provided by theInternational Monetary Fund (2011). This report estimates quite small future oil demand

16Efforts are underway to fully incorporate more realistic regional drawdown profiles based upon informationprovided by the IEA. These draw rates tend to have greater surge capacity in the first few months.


39/51

16

elasticities, particularly for non-OECD regions.17 Median short-run demand elasticities forOECD countries are -0.025 (with a narrow 80% confidence interval of -0.035 to -0.015). Theyconclude the demand in Non-OECD countries is even less elastic, and the increasing importance

of Non-OECD demand means that global oil demand elasticity is declining.

18

The combinedshort-run demand elasticity for OECD and Non-OECD oil importing countries is estimated at -0.019 (range -0.028 to -0.009). Even including the income and supply responses, this suggests avery low net response elasticity to disruptions.

Approach Used to Model Net Supply and Demand Response to Disruptions

Most of the assessments of elasticity refer to annual periods. However, disruption analysistypically calls for modeling quarterly or even monthly response. Furthermore, studies usually areonly able to estimate a single (constant) elasticity regardless of the magnitude of price increase.

Using a constant elasticity framework for large oil shocks is problematic, because suchelasticities imply extremely high short run prices for large disruptions, yet supply and demandbehavior at very high prices is poorly understood.19

Following prior emergency stock studies (DOE 1990, Leiby and Bowman 2005, 2007) weapplied a variable elasticity for the market response to changing prices. As used, the elasticitymay be interpreted as a short-run elasticity of net world demand, that is, an elasticity of the netresponse of demand and supply. It is widely accepted the short run demand and supplyelasticities are quite small, while longer-run elasticities, which may apply after a number ofyears, are 4 to 10 times larger.

Apart from comparability to prior emergency oil stock studies, the approach has the primaryadvantage that elasticities increase over the course of a disruption (reflecting the process ofslowly increasing adjustment over time to sudden price changes) and they also increase with themagnitude of the disruption supply shortfall. The chosen functional form makes elasticity anincreasing function of both month and disruption size:

+ + + Where:t= initial demand elasticity for 0 month and 0 shortfallStm= Net Oil Shortfall (after offsets) in year t and month mM = Month

17See IMF 2011, Tables 3.13.3. Original Source - IMF staff calculations.18The growing importance of emerging market economies appears to have reduced world oil demand priceelasticity (in absolute terms) and increased income elasticity. (IMF 2011 pp. 94-95).19For example, a short run (annual) net demand response elasticity or -0.10 would imply that a 20% loss of supplywould increase price by over 9-fold (to above $900) for the first year.


40/51

17

and are monthly growth terms which cause the elasticities to increase over the course of adisruption. Since elasticities grow with disruption size, the price response per million barrels perday of supply loss declines as disruptions get increasingly large. This prevents prices from

becoming arbitrarily large for very large disruptions, as would occur with a fixed short runelasticity.

The values of and come from the EIA (EIA DIS 2002 and 2007). For the values of (initial

undisrupted elasticity) and (change in elasticity with time) we currently use two values, EIADIS 2002/07 and recent estimates by ORNL (2014). Recent (ongoing) research by ORNL andothers has suggested that oil demand has become less responsive to price in recent years. Moreinelastic demand implies that supply shocks produce greater price changes (in percent changeterms) than in the past. It also means that emergency stocks drawn down to offset these supplyshocks have a greater effect on prices and therefore produce larger benefits. The alternativedemand elasticity case uses estimates provided by a meta-analysis, recently conducted by ORNL,

of dozens of demand elasticity studies. Year 2020 World average values for 0 and 10 MMBDlosses are given in the tables below. Estimates are also available for other regions and years.

Year 2020 Average World Demand Elasticities for Various Months and Shortfall Levels

EIA/DIS 2002/70 MMB Loss-0.08 after 1 month-0.11 after 12months,-0.10 average over12 months.

EIA/DIS 2002/710 MMB Loss-0.13 after 1 month-0.19 after 12 months,-0.16 average over 12months.

ORNL 20140 MMB Loss-0.05 first month,-0.07 after 12 months,-0.06 average over 12months.

ORNL 201410 MMB Loss-0.10 first month,-0.14 after 12 months,-0.12 average over 12months.

These modeled base elasticity values are comparable to many of the estimates in the literaturebriefly summarized above, or perhaps on the larger side. Moreover, the variable elasticityframework moderates extreme and longer lasting disruptions. All of this suggests that the basemodel specification may be somewhat conservative about the price effect of disruptions, andstock drawdowns. Further sensitivity analysis of this issue is merited.

2.1.6 GDP Elasticities

Estimates of the GDP impact of oil shocks under the current study involve a direct relationshipbetween price implications and the gross domestic product of oil importing economies. Thisapproach can be described as a summary of the modeling of macroeconomic effects of the oilsupply shocks in the literature, but accounts for the direct and indirect effects on the economy.Oil importing economies are classified into four groups: United States, IEA Europe, Other IEAand Non-IEA. The change in regional GDP (RGDP) associated with an oil supply disruption,RGDP, is calculated based on the estimated change in oil prices, P, as:


41/51

18

11

,oilpriceGDP

ref

refP

PRGDPRGDP

The above equation requires estimates of the oil price-GDP elasticity of, GDP,oilprice, referenceRGDP and oil price, RGDPrefand Pref, and the price change due to the oil disruption event, P.Estimates of the oil price-GDP elasticity are derived from the literature as described below.

Overall approach

There is a considerable amount of empirical research on the GDP effects of crude oil priceshocks. These studies have identified several mechanisms and channels through which crude oilprice changes impact the economy. Prominent among these are reallocations of expenditures on

goods and services within the economy, frictions in capital and labor market adjustments,reductions in capacity utilization rates, difficulty in adjusting prices, monopolistic behavior,inflationary effects, and responses by the monetary authorities. In general, these studies haveidentified a non-linear and asymmetric, but stable, oil price-GDP relationship in the historicaldata using measures that reflect the sudden or surprise nature of oil price shocks (e.g. Jimenezand Sanchez, 2009). Studies on the GDP effects of crude oil price shocks account for thesefactors in different ways, and the approach taken depends on the underlying modelingframework.

Models in the literature can be categorized into three broad groups: structural and non-structuraleconometrics, and general equilibrium models. Structural econometric models are represented by

large macro-econometric models such as the IMF Multimod and the Global Insight models,whereas non-structural econometric models include a wide range of studies based on empiricallyestimated single- and multi-equation specifications. These models are typified in the oil-macroeconomy literature as Vector Auto-Regression (VAR) models. Lastly, general equilibriummodels are economy-wide modeling frameworks that include computable (applied) generalequilibrium (CGE/AGE) and dynamic stochastic general equilibrium (DSGE) models. TheDSGE models are less detailed than CGE models, but stochastic dynamics is a basic feature thatgives them some resemblance to VAR-type models. In addition to the above factors andmodeling approaches studies in the literature differ in the frequency and date covered by thedata, and many other specification choices. Finally, most of the existing studies in the literatureare for developed economies, whereas the current study requires estimates of the GDP impacts of

oil price changes for oil importing economies that include more than 160 nations. The overallapproach employed for estimating the oil price-GDP elasticity from the existing literatureinvolves the following steps:

Review of the latest literature on estimates of the GDP impacts of oil pricechanges.

Estimation of the oil price-GDP elasticity based on the literature survey.


42/51

19

For the IEA countries, we calculated the oil price-GDP elasticities using countryestimates from the literature. We implemented an approach for estimating the oilprice-GDP elasticity for the non-IEA group, which includes almost 140 net oil

importing economies.

Literature review

A search of the recent literature identified about 100 recent papers related to the GDP impacts ofoil price changes. There are two basic criteria for estimating oil price-GDP elasticities from thesestudies:

The study must include simulation(s) of the effects of oil shocks on the economy. The study must present numerical information on the size of the oil price change, and

estimates of the GDP effects or other information to evaluate these variables. These arerequired to convert GDP impacts to elasticity values comparable across studies.

Many of the identified studies satisfy the first criterion, but not the second. Ultimately, we found14 recent studies that met the two criteria.

Alessandro Cologni and Matteo Manera (2008) Oil Prices, Inflation and Interest Rates ina Structural Cointegrated VAR Model for the G-7 Countries, Energy Economics, 30,856888.

Makena Coffman (2010) Oil price shocks in an island economy: an analysis of the oilprice-macroeconomy relationship, Ann Reg Sci, 44:599620, DOI 10.1007/s00168-008-

0271-6. Marcelo Snchez (2011) Oil Shocks and Endogenous Markups: Results from an

Estimated Euro Area DSGE model, Int Econ Policy, 8:247273 DOI 10.1007/s10368-010-0159-7.

Surender Kumar (2009) The Macroeconomic Effects of Oil Price Shocks: EmpiricalEvidence for India, Economics Bulletin, 29(1), 15-37.

Luis J. lvarez, Samuel Hurtado, Isabel Snchez, Carlos Thomas (2010), The Impact ofOil Price changes on Spanish and Euro Area Consumer Price Inflation, EconomicModelling, 28, 422431.

Lutz Kilian (2007) A Comparison of the effects of exogenous oil supply shocks on outputand inflation in the G7 countries. Journal of the European Economic Association, 6 (1),

78-121. Levent Aydin, Mustaf Acar (2011) Economic impact of oil price shocks on the Turkish

economy in the coming decades: A dynamic CGE analysis, Energy Policy, 39, 17221731.

Limin Dua, Yanan He, Chu Wei (2010) The relationship between oil price shocks andChinas macro-economy: An empirical analysis. Energy Policy, 38, 41424151.

Gert Peersman, Ine Van Robays (2011) Cross-country differences in the effects of oilshocks, Energy Economics (In Press).


43/51

20

Christian Lutz, Bernd Meyer (2009) Economic Impacts of Higher Oil and Gas Prices:The role of International Trade for Germany, Energy Economics, 31, 882887.

Iikka Korhonen, Svetlana Ledyaeva (2010) Trade Linkages and Macroeconomic Effects

of the Price of Oil, Energy Economic, 32, 848856. Katsuya Ito (2010) The Impact of Oil Price Hike on the Belarusian Economy, TransitStud Rev, 17:211216, DOI 10.1007/s11300-010-0140-8.

Ana Gmez-Loscos, Antonio Montas, M. Dolores Gadea (2011) The Impact of OilShocks on the Spanish Economy, Energy Economics, 33, 10701081.

We collected the following information from each study as available:

Region(s) covered by the study. Type and size of the simulated oil price shock. Frequency of the data which are generally month, quarter or annual.

Type of economic impact reported, which is typically the GDP effect. Somestudies report impacts on industrial production, rather than GDP.

Estimates of the GDP effects of simulated oil shocks and the length of time sincethe beginning of the oil shock. We also noted whether the reported GDP effectsare cumulative over the elapsed period or point/average values.

Almost 160 different estimates representing 22 countries/regions, and a combination of thefactors identified above, were collected from the 14 studies. We computed the impliedelasticities by dividing the percentage change in GDP by the percentage change in price. Theresulting elasticities were then annualized using information about the data frequency, the lengthof time since the start of the shock, and whether the study provided cumulative/non-cumulative

estimates. We categorized the estimates for each country into low, mean and high values toaccount for the influence of the multitude of factors identified above. We attempted to separatethese estimates into short-run, medium-run, and long-run based on whether the length of timesince the start of the shock is less than or equal to 4 quarters, between 4 and 12 quarters, andlonger than 12 quarters, respectively. We also examined the data period represented by thedifferent studies. Most studies cover the period from 1980 to 2004, and a few included dates asfar back to 1970 and/or up to 2010. Given this, it was impossible to associate a time dimension tothese oil price-GDP elasticities. Figure A-10 illustrates the resulting estimates.


44/51

21

Figure 9: Estimates of the Annualized Oil Price-GDP Elasticity from Recent Literature

Regression Analysis of the Oil Price- GDP Elasticity from Recent Literature

We implemented a model-based approach for extending estimates of the oil price-GDPelasticities from the literature to countries/regions with no data, mainly the non-IEA economies.We interpreted the elasticity values, GDP,oilprice, obtained from the literature as cross-sectionaldata and specified a linear regression of the following form:

n

i iioilpriceGDP X

,

Where:Xiand iare variables to explain differences in the value of the elasticities across

countries, and their associated coefficients, respectively.

The explanatory variables,Xi, in this equation are the energy intensity of GDP, oil share ofenergy use, and import share of oil use. A limited set of variables was necessary given that ourcross-sectional data includes only 22 countries. These three variables were selected based on areview of the literature on the determinants of vulnerability to oil price shocks in oil importingeconomies (World Bank 2005, Gupta 2008). This literature suggested several indicators,including the energy intensity of GDP, oil share of energy use, import share of oil use, per capita

-10%

-5%

0%

5%

10%

15%

Canada

France

Germany

Italy

Japan

UK

USA

Hawaii

EuroArea

India

Spain

Turkey

LeibyEtAl2015 ORNL - Drawdown Capability Study for DOE EPSA - Final - 20150414

Documents