This work covers upstream costs and emissions including production and transport of crude oil. Due to lack of access to refinery cost data, we cannot yet generate similar results for refining and thus assemble a fully market-informed (consequential) well-to-tank or well-to-wheel emissions analysis.
We frame our economic research question within a traditional profit maximization problem, where a large number of independent firms produce a substitutable good. More precisely, we assume that every oilfield is a risk-neutral firm, which exerts no market power. In other words, the production decisions of a single field do not affect the market price and all non-oil prices are given. The aim of management is to maximize the field profits
Profits = (Oil Price * Volumes of Oil Extracted) – Extraction Costs – Exploration Costs
Profits are the difference between the field revenues (Oil Price * Volumes of Oil Extracted) and the field costs, which are divided into two macro-classes: costs to extract the oil (Extraction Costs) and the costs to discover new oil (Exploration Costs).
At every point in time, the management makes two simultaneous decisions for a given field: how much oil to extract from existing known capacity/reserves? and how much money to spend in exploration5,6 to identify new capacity/reserves. While making these decisions the management faces two physical constraints. First, the cumulative depletion of the field at time t equals the cumulative depletion until time t-1 plus the Volumes of Oil Extracted at time t. Second, the cumulative discoveries at time t equals the cumulative discoveries until time t-1 plus the discoveries at time t.
The first order condition of the optimization problem with respect to the Volumes of Oil Extracted identifies how much money a producer is willing to pay to manage one extra barrel of oil. This value is the Shadow Price (SP),
Shadow Price = Oil Price – Marginal Extraction Costs
where the Marginal Extraction Costs is the first-order derivative of the Extraction Costs with respect to Volumes of Oil Extracted 7 . For example, if a field sells its output at 50 $/bbl and the Marginal Extraction Costs (i.e., the cost of extracting the next barrel) is 40 $/bbl, the owner of the field is willing to spend (up to) $ 10 to manage one more barrel in that particular field. Section 1 of the Supplementary Material (SM) provides mathematical details of the economic framework linking the concept of Shadow Price to standard oil economics models.
As the SP of a field approaches zero, the management problem shifts from ‘how much should I produce?’ (intensive margin choice) to ‘should I produce or not?’ (extensive margin choice). In other words, the fields with a shadow price around zero identify the extensive margin of the oil industry. The emissions of this portion of the industry will be the most sensitive to demand shocks.
In this economic framework, we develop a universal model to estimate shadow prices. As in most exhaustible resources models, the shadow price is the market price net of the marginal cost. The former can be reverse engineered using the prices of publicly traded crudes. The later must be estimated. Accurate measurement of marginal costs of global oil fields is difficult. Among other problems, it is difficult to differentiate which factors of production are fixed and which are variable. We do not claim to flawlessly compute the marginal extraction costs of every field in the dataset, but rather that a combination of standard econometric techniques with a large longitudinal dataset allows us to approximate how much it costs to extract the next barrel across different types of formations and then to link this analysis to the environmental characteristics of the marginal producer.
Econometric Analysis: The SP of a particular field is the difference between the field-level expected price and its marginal extraction costs. Both variables are unobserved. To estimate them, we face two econometric problems: 1) the non-stationary nature of oil prices, and 2) the endogenous link between costs and quantities. We solve these problems using standard econometric techniques.
Since the commercial agreements between oil producers and oil refiners are generally not disclosed, we do not know the price at which a particular field sells its output. However, we know the prices for different publicly traded classes of oils. More precisely, we know the landed costs of imported crudes in the United States from 1979 to 2018, as reported by the Energy Information Administration 8 , and the chemical characteristics of every traded class, as reported from the PSA Management and Services BV dataset 9 (see Fig. 1 of SM). From those, we regress the prices of the publicly traded oil classes against their API gravity, their sulfur content and a homogenous time trend. In doing so, we isolate the impact of the chemical properties of each crude from overall price trends driven by global oil demand. According to our estimates, increasing API gravity by one degree changes the value of a crude by + 0.07 $/bbl, while increasing sulfur content by 1% changes the value by −2.21 $/bbl (see SM, Sect. 2.1).
Under mild assumptions, we can use these two structural coefficients to estimate field-level selling prices for the fields in our basket of crudes (see SM, Sect. 2.1, Eq. 8). For example, in 2015, when the part of the inverse oil demand, absent effects from gravity and Sulfur, was estimated to be 49.80 $/bbl, a field with an API gravity of 55.00 and a sulfur content of 3% could sell its output at an estimated price of:
Oil Price = 49.80 + (0.07 * 55) – (2.21 * 0.03) = 53.58 $/bbl.
Using the API and sulfur content reported in the 2018 Wood Mackenzie (WM) dataset10, we estimate the selling price of 1933 “parent project” fields over the decade 2009–2018, thereby obtaining 1933 * 10 = 19,330 simulated selling prices. See Fig. 1 of SM for a cross-sectional snapshot. The section entitled Firm Expected Prices in SM (Sect. 2) provides econometric details of the estimate.
Next we must estimate the other component of SP: marginal extraction cost. Using the WM dataset, we obtain yearly cost data for the same 1933 fields over the time interval 2009–2018. Then we obtain Extraction Costs by summing the operational expenditures (OPEX, which include consumable inputs, labor, maintenance, repairs, accounting costs, license fees, office expenses, utilities and insurance) and the capital expenditures not linked to exploration activities (non-exploration CAPEX, which include installation, acquisition, upgrading and restoring of the physical assets used to extract the oil). After computing the Extraction Costs, we regress them against the Volumes of Oil Extracted while controlling for the depletion level of the field, a technological trend and the geologic class of the field3. The estimated first order derivative of the fit with respect of the Volumes of Oil Extracted returns the estimated Marginal Extraction Costs. Section 2.3 of SM provides all the econometric details.
Since this estimate of extraction costs includes non-exploration development costs, its validity is maintained over a few-year period. Cutting field development expenditures can lower the costs in the short-run (e.g., months), but in the long-run (years to decades) exploration and development investment is needed to maintain a given level of production. Therefore, our time scale of marginality is representative of an intermediate 3 to 5 year time period.
Using estimated Oil Prices and estimated Marginal Extraction Costs (MC), we estimate the SP. For example, if the above example field extracts the next barrel at a marginal cost of 43.58 $/bbl, the Shadow Price is then
Shadow Price = 53.58 $/bbl – 43.58 $/bbl = 10.00 $/bbl.
In other words, this field is $ 10.00 away from the extensive margin of the industry. As a result, if the price of oil would decline by 10.00 $/bbl or the MC would increase by 10.00 $/bbl, the management problem shifts from ‘how much should I produce?’ to ‘should I produce or not?’.
Carbon Intensity Model
The field-level CI is estimated using the Oil Production Greenhouse Gas Emissions Estimator (OPGEE version 2.0)11–13. OPGEE is an open-source, peer-reviewed 11,14−23, bottom-up, engineering-based model. The OPGEE system boundary is “well-to-refinery” (WTR, i.e., exploration, drilling & development, production & extraction, surface processing, maintenance, waste disposal, and crude transport to the refinery). Reported emissions are measured in gCO2Eq. emitted per 1 MJ LHV of crude petroleum delivered to the refinery entrance gate. All GHGs are converted to gCO2Eq. using AR5 GWP100 conversion factors (without carbon feedback)24. See the OPGEE user guide11 for more details of each process stage.
OPGEE estimates CI using up to 50 parameters as input data for each modeled oilfield. If input data are not available for some parameters (common), OPGEE supplies defaults based on statistical analysis of petroleum engineering literature and commercial data sources (e.g. Oil & Gas Journal O&GJ25) enabling the software to estimate a field’s CI without complete data 25,11. In this work, field exploration emissions are excluded from CIs reported in prior work1 to estimate GHG emissions associated with production of the next barrel of crude oil (i.e., marginal upstream CIs).
Crude oil transportation GHG emissions are generally a minor contributor to CI. Due to volatility of crude oil trading patterns and lack of data availability on these trades, we use identical OPGEE defaults for crude transportation for all studied oilfields (ocean tanker: 8,000 miles; ocean tanker size: 250,000 tons; pipeline: 1,000 miles). Energy-based allocation of emissions is used to divide emissions given co-production of gas.
Covered Global Oilfields
In the previous work1, CIs were estimated for 8,966 global active oilfields (so-called “child” fields) supplying 78.9 million barrels per day, and capturing ~ 98% of 2015 global crude oil and condensate production26. Fields with gas-oil-ratio (GOR, scf natural gas produced/bbl crude oil produced) of < 10,000 scf/bbl are considered as oilfields. A combination of government reported data (Norway27,28, Canada29–32, Denmark33, UK34, Nigeria35, and US California36, US Alaska37, and US shale oils38), public literature (total of nearly 800 sources) and proprietary/commercial data sources (O&G J 2015 survey25 and WM oilfield datasets10) were used as input data1. Government and public literature data were collected and used for 1,009 global fields, accounting for about 64.3% of global crude oil production. Commercial data are utilized for the remainder (mostly small fields). The year 2015 is selected as the reference year due to lags in some data sources. See our previous study SM document1 for further details.
This previous work on the CI of global oilfields1 is provided at a “child” field level. Child fields are individual discoveries that are part of a parent project. Parent fields are combinations of geologic deposits collected for the purposes of a combined valuation. The linkage with the economic data (see above), available only at parent-level requires to match the child-field CIs1 to parent fields. The majority of the child non-technical oil fields from WM datasets10 (accessed 2018) - whose corresponding parent fields are available - directly matched with the OPGEE global dataset. We paired the remaining with smart string search and string distance matching using R program script, as well as by-hand manual matching for the countries with poor total production coverage.
Finally, additional treatments are conducted on two important global producers (Canada and U.S.) based on the available data (see Sect. 4 of SM).
After the matching process is completed, it is important to examine to what extent the matched fields covered in this work are representative of total production of different countries. Table 6 of SM gives a coverage summary for the top 20 largest global producers, showing that the integrated global economic and emission dataset in this work is a good representation of the global picture. In total, 1933 parent fields located in 77 countries are matched. These oilfields have combined oil production of ~ 71 mmbbl/d, capturing ~ 90% of 2015 global crude oil and condensate production26. The geographical location of covered global oilfields and their qualitative volumetric production magnitude and CI are mapped in Figs. 7 and 8 of SM.