Report S6
Petroleum Product Movement Estimates for 1997
Table of Contents
1. Definition of the Commodity Flow Data Gap
1.1. General Description
The general transportation logistics of the petroleum industry start with the initial gathering of crude oil in production fields for domestic sources and from marine terminals for foreign imports. The crude oil is then delivered to refineries or to long-term storage facilities such as the Strategic Petroleum Reserve (SPR). From these refineries, finished products are moved to markets throughout the nation. Transportation of petroleum products is accomplished by a variety of land and marine-based modes. They include: pipeline, railroad tanker cars, tanker trucks, barges, and oceangoing tankers. On a volume basis, pipelines and marine vessels are predominately used in transporting petroleum, but trucks and railroad tank cars also have essential functions.
Shipments of petroleum products are in scope for the CFS. However, there are significant discrepancies between CFS reported totals and those published by other government agencies. Furthermore, all ton-miles are suppressed from CFS tables, either because of high sampling variability in the estimates or due to poor response quality.
1.2. Commodities Involved in the Data Gap
1.2.1. SCTG codes
Petroleum products are included under the following two-digit SCTG codes:
17 - Gasoline and Aviation Turbine Fuel
18 - Fuel Oils
19 - Coal and Petroleum Products, N.E.C.
These are in-scope commodities for the 1997 and 2002 CFS.
1.2.2. STCC codes
Under STCC coding, petroleum products are included in “29 - Petroleum or coal products.”
1.3. Establishments Involved in the Data Gap
1.3.1. NAICS codes
324 - Petroleum and Coal Products Manufacturing
486 - Pipeline Transportation
1.3.2. NAICS-SIC conversion issues
There are no major NAICS-SIC conversion issues for this data gap.
2. Importance of the Data Gap
Petroleum products are vital commodities to our national security, economy, and mobility. Information on petroleum products shipment is a building block that helps to form the basis for sound policy-oriented analyses for transportation investments.
2.1. Value and Tonnage as a Share of National Shipments
A national summary of the estimates described is this report is shown in Table 1. Pipeline movement estimates amount to about 3.6 billion barrels or 480 million tons. Waterborne estimates based on National Waterborne Commerce data from the Corps of Engineers is about 271 million tons, while rail estimates from the Rail Waybills data amounted to about 9 million tons. The Truck movement estimate shown in Table 1 is an aggregation of state-level consumption of petroleum products from the EIA State Energy Data 1997, amounting to around 5.4 billion barrels. Although there is no publicly available data, truck movements can be estimated in this way because the short-haul delivery of petroleum products to retail outlets is almost exclusively done by trucks. In total we estimate about 1.5 billion tons (about 11 billion barrels) of petroleum products movements in 1997.
| Mode | Tons (Thousands) | Ton-Miles (Millions) |
|---|---|---|
| Pipeline | 478,428 | 143,528 |
| Rail | 9,043 | 8,139 |
| Truck | 719,436 | 122,304 |
| Waterborne | 270,800 | 148,518 |
| Total | 1,477,707 | 422,489 |
2.2. Value and Tonnage as a Share for Individual Modes
Based on ton-miles Table 1 shows that pipelines carry about 34 percent of all petroleum products, while water carries about 35 percent. Trucks share is about 29 percent, while rail accounts for a very small amount. The differences in trip lengths among these modes are seen by comparing the above percentages to those based on tons which are 32 percent, 48 percent and 18 percent for pipelines, trucks, and water, respectively. Note that trucks account for almost half of all deliveries in tonnage terms because of its almost exclusive means of delivering petroleum products to retail outlets.
2.3. Geographic Concentration: Dispersed versus Concentrated, Local versus Long Distance
Petroleum products shipments are transported long-distance via pipeline and waterways to various locations in the U.S. According to the Pipeline Economics, Oil & Gas Journal, the average pipeline shipment length is 307 miles, while waterborne commerce data show a trip length of almost 550 miles.
2.4. Importance to International Trade
Imported Petroleum Products is a relatively small part of the total petroleum products supplied by the 146 operating refineries across 35 states in the U.S. On average, 1,936 thousand barrels per day of petroleum products were imported during 1997. On the other hand, a small amount of petroleum products (896 thousand barrels per day) was exported from the U.S. in 1997. The finished Petroleum Products supply averaged 16,500 thousand barrels a day in 1997.
3. Data Sources
3.1. Coverage in CFS
The petroleum and coal products manufacturing industry is covered by the CFS 2002. However, ton-mile information is left out of the published CFS statistics. Furthermore, there are significant differences between CFS reported tonnage and value as compared to similar information from other data sources.
3.2. Coverage in Other Data Sources
The basic information for petroleum products production (e.g., imports, exports, and disposition at refineries) is collected by EIA/DOE. Several other data sources were consulted in arriving at the estimates for 1997, including:
- Oil Pipelines: Annual Report (Form 6) of oil pipeline companies provided to the Federal Energy Regulatory Commission.
- Water Carriers: Waterborne Commerce of the United States, U.S. Army Corps of Engineers, Part 5.
- Motor Carriers: Petroleum Supply Annual, U.S. DOE, EIA, Volume 1, Table 46.
- Railroads: Carload Waybill Statistics, Report TD-1, USDOT, Federal Railroad Administration and Freight Commodity Statistics, Association of American Railroads.
National level petroleum products shipment information by transportation mode was also available from the Shifts in Petroleum Transportation published by the Association of Oil Pipelines annually.
Tables 2 and 3 below show inter-PADD estimates of petroleum products movements by modes in 1997 based on EIA data.
| From | To |
|||||
|---|---|---|---|---|---|---|
 |
PADD 1 | PADD 2 | PADD 3 | PADD 4 | PADD 5 | Total |
| 1 | 0 | 310,980 | 3 | 0 | 0 | 310,983 |
| 2 | 50,325 | 0 | 141,390 | 105,787 | 0 | 297,502 |
| 3 | 2,486,793 | 817,935 | 0 | 16,995 | 75,817 | 3,397,540 |
| 4 | 0 | 59,902 | 65,594 | 0 | 31,527 | 157,023 |
| 5 | 0 | 0 | 0 | 0 | 0 | 0 |
| Total | 2,537,118 | 1,188,817 | 206,987 | 122,782 | 107,344 | 4,163,048 |
| From | To |
|||||
|---|---|---|---|---|---|---|
|
PADD 1 | PADD 2 | PADD 3 | PADD 4 | PADD 5 | Total |
| 1 | 0 | 2,774 | 2,376 | 0 | 0 | 5,150 |
| 2 | 53,757 | 0 | 17,927 | 0 | 0 | 71,684 |
| 3 | 849,510 | 123,638 | 0 | 0 | 7,086 | 980,234 |
| 4 | 0 | 0 | 0 | 0 | 0 | 0 |
| 5 | 0 | 0 | 2,912 | 0 | 0 | 2,912 |
| Total | 903,267 | 126,412 | 23,215 | 0 | 7,086 | 1,059,980 |
3.3. Data Quality
Data sources listed above are all government published statistics. They are expected to be reliable.
4. Estimation Methods
4.1. General Description of Estimation Method
Petroleum Products information collected by EIA provides a starting point for estimating its movements as required by the FAF. Information on the origins of Petroleum Products include refinery production at the PADD level, and imports identified by port city and state. The approach to estimating Petroleum Products movements for 1997 is slightly different from that used for the 2002 FAF. This is because a comprehensive database on pipeline movements was not available for 1997 as was the case for 2002. We estimated Petroleum Products movements at the regional level and then sum up to derive the national estimates shown in Table 1.
4.2. Method for Estimating Regional Flows
The approach employed for deriving regional estimates of Petroleum Products movements is a combination of bottom-up and top-down approaches. Data on movements were collected from publicly available sources. Waterborne data were available at the state-level, and rail movements were estimated from the Rail Waybill database at the state level. Unlike our 2002 estimates, there was no comprehensive database on pipeline movement of Petroleum Products for 1997. In addition, there is no publicly available data on truck movement of petroleum products. State-level petroleum products consumption was used as an estimate of truck movements. We derived a comprehensive estimate of water, rail and pipeline movement at the state level, and then disaggregated these to FAF regions.
4.2.1. Rail
Petroleum Products movement by rail was derived directly from the Waybills database by multiplying each sample shipment by its expansion factor and summarizing to the state level. To calculate ton-miles, we used an average miles per ton of about 900 miles.
4.2.2. Waterway
The primary data for estimating water movements is the state-to-state Waterborne Commerce data. Comparing Table 3 above to a PADD level aggregation table of this data provides an interesting check. We note that the structure of both tables are very similar, although there are significant discrepancies in the magnitudes. The most significant differences in magnitudes are movements from PADD 3. The large inter-PADD movements from PADD 3 in Table 3 seem to be represented in the Waterborne Commerce data as intra-PADD movements within the destination regions found in Table 3.
The state-to-state waterborne data combines all Petroleum Products into a single category for each origin-destination combination. Thus, it was necessary to allocate these totals to the three SCTG categories that incorporate Petroleum Products as required by the FAF. We made use of the Manifest Cargo data, also from the Corp of Engineers, and the listing of major ports to derive an allocation mechanism. The Manifest Cargo data contains records of shipments and receipts by waterway codes (including port codes) for a number of commodities, including eleven categories for petroleum products. These eleven categories were aggregated to the three SCTG categories needed for our estimates. We extracted data for the 213 major ports whose locations were available for 1997 from this database, and associated county locations with each of these ports by checking each port name against place names from the Geographic Names Information System (GNIS). The county locations were summarized to the state level, and the receipts and shipment fields of the Manifest Cargo data on petroleum products were used to derive origin and destination vectors for water movements. These vectors were then used to derive shares of each SCTG category in petroleum products movement for each state. Multiplying the state-to-state matrix of total Petroleum Products movements by waterways with both of these vectors creates a SCTG-origin-destination table. For ton-miles, an average distance per ton of 675, 470, and 550 miles were calculated from the national data for SCTG 17, 18, and 19, respectively.
4.2.3. Pipeline
There is no comprehensive data from the FERC Form 6 database for 1997 because collection of this data in electronic form started in 2000. We adopted an approach where the pipeline estimates are derived as the net balance of petroleum products at the origin states (refineries and imports) and destination states (consumption), and movements by rail and water. For this purpose we hold the state-to-state estimates for rail and water, as described above, fixed. We did not include truck movements in this calculation because truck movements are short-hauls.
We obtained data on refinery capacities and production of petroleum products from the EIA website. The refinery capacity data identifies refinery locations by state and city, while the production data has information at the PADD level. Thus, to calculate state-level production we constructed a PADD-county crosswalk using the listing of PADD components from the EIA Petroleum Supply Monthly along with a GIS map of counties. This crosswalk was then used to allocate the PADD-level production data to states based on the refinery capacity shares. Company level data on petroleum products imports, which include the port state, was also obtained from the EIA website. Unlike crude oil, destination states for petroleum products are not identified in the imports data. The sum of refinery production and imports at the state level represents the total amount of petroleum products originating from each state. Petroleum products destination data at the state level was estimated by the state-level consumption of petroleum products, also obtained from the EIA website.
We derived net movements of Petroleum Products by pipeline from the above data by formulating an entropy maximization problem, with the entropy objective function based on the destination state shares of pipeline movements for each origin state. We used our estimates from the 2002 analysis to provide prior information in this objective function. This helps to incorporate information on actual pipeline infrastructure in our 1997 estimates. However, the entropy approach would reflect this information in the estimates only if it is consistent with available 1997 data and the following constraints:
- The matrix of state-to-state movements is set equal to the sum over all three modes, i.e., rail, water, and pipeline, when the origin and destination states are not the same. For within state movements we allow the diagonal elements of matrix of state-to-state movements to be greater than the corresponding sum over the three modes. This helps to satisfy the other constraints in the programming problem, taking care of situations such as Alaska where its entire pipeline movement is within state. Any imbalance in the diagonal terms is added to the corresponding pipeline diagonal elements after the programming solution has been obtained.
- Sum of the composite state-to-state movements matrix over destinations must be less than or equal to the sum of imports and refinery production in each state of origin.
- Consumption in each destination state is equal to the sum of over origins of the modal state-to-state movement matrices.
4.2.4. Calculation of FAF-to-FAF Petroleum Products Movement
The above calculation generates a complete set of state-to-state matrices for rail, water, and pipeline movement of petroleum products. Disaggregating these data to FAF regions faces challenges similar to the above state level calculations. Thus, we employ an approach similar to the above by first disaggregating the water and rail data, and then deriving the net FAF-FAF movement for pipelines. Since the rail data is at the county level FAF-to-FAF rail movements were derived by simply aggregating the data to the FAF regions using a FAF-county crosswalk. For water, we made use of the same manifest cargo data described above to allocate the state-level movements to FAF regions.
Since we do not have data on pipeline movements we derive its FAF-to-FAF matrix as the FAF-to-FAF matrix of all Petroleum Products movements minus those for water and rail calculated above. In order to derive the FAF-to-FAF matrix of all movements we employ an approach similar to that used for water as described above. The origin and destination share vectors were derived by dissagregating the import and refinery production data (for origin) and the petroleum products consumption data (for destination) to FAF regions for each state. For imports we used the port city to identify county locations, while we used the refinery capacity data, which identifies city of location, to allocate domestic production after associating each city with a corresponding county. Once the counties are known we summarized each data source to the FAF level for each state and calculated the needed share vector. For destinations we used the same FAF-to-FAF allocation described for Truck movements below to calculate the shares. The FAF-to-FAF matrix is then derived by taking the product of the sum of state-to-state movements over all three modes and the two share vectors and summing over states for each FAF origin and destination combination. The FAF-to-FAF matrix of pipeline movements is then calculated by subtracting the water and rail matrices from this totals matrix. The net pipeline movements in a few cells were negative. These values were set to zero to derive our final pipeline movements matrix.
A comparison of the PADD level aggregation of the generated pipeline matrix (Table 4) to the EIA inter-PADD movements (Table 2) shows that while our estimates correctly identifies the large movements involving PADD 3 the magnitudes are very different. In addition, all the cells in Table 4 are positive suggesting that the available 1997 information did not provide enough restrictions on pipeline interconnections for Petroleum Products movements.
| From | To |
|||||
|---|---|---|---|---|---|---|
|
PADD 1 | PADD 2 | PADD 3 | PADD 4 | PADD 5 | Total |
| PADD 1 | 522,342 | 202,411 | 42,948 | 31,776 | 59,741 | 859,218 |
| PADD 2 | 227,171 | 339,451 | 71,915 | 44,854 | 85,121 | 768,512 |
| PADD 3 | 336,268 | 271,826 | 488,058 | 35,193 | 93,132 | 1,224,477 |
| PADD 4 | 29,435 | 36,544 | 8,163 | 6,245 | 13,488 | 93,875 |
| PADD 5 | 177,871 | 225,556 | 49,109 | 36,154 | 153,434 | 642,124 |
| Total | 1,293,087 | 1,075,788 | 660,193 | 154,222 | 404,916 | 3,588,206 |
Note: This table also includes intra-PADD movements on the diagonal.
4.2.5. Truck
The only basis available for estimating petroleum products movement by truck for 1997 is the state-level consumption data from the EIA. We disaggregated state consumption to FAF regions using county level income distribution data from the Regional Economic Information System (REIS) under the assumption that truck movements, being short-hauls, are intra-FAF.
4.3. Expected Quality of the Estimates
The estimates are based on government data sources and, therefore, are expected to be reliable.
5. Implications for the Scope and Content of the 2007 CFS
It is recommended that tables based on actual physical data on water, rail, and pipeline infrastructure be developed for allocating state-level Petroleum Products data FAF regions.
6. References
Association of Oil Pipelines (AOPL), “Pipeline and Water Carriers Continue to Lead All Other Modes of Transport in Ton-Miles Movement of Oil in 2003”, Report, Washington D.C., May 16 2005.
U.S. Army Corps of Engineers, 1997 State to State Commodity Movements from the Public Domain Database (Text format). Downloaded from the Web July 2005 http://www.iwr.usace.army.mil/ndc/db/wcsc/pdomain/data/.
FERC Annual Report (Form 6) of oil pipeline companies provided to the Federal Energy Regulatory Commission.
U.S. Army Corps of Engineers, Waterborne Commerce of the United States, Part 5, Table 2-2.
USDOT, Carload Waybill Statistics, Report TD-1, USDOT, Federal Railroad Administration and Freight Commodity Statistics, Association of American Railroads, Table A3.
