Unlike the construction of the domestic OOS commodity flows, with its diverse and sometimes partial data sources, the construction of estimates of annual import and export flows for FAF4 are generally based on a few highly developed datasets.  Even so, lack of geographic details for inland movements creates significant gaps in the regional commodity flow picture.  Because of this, procedures for generating domestic legs of foreign trade movements are required. 

11.1 DEFINITION OF IMPORTS AND EXPORTS

Imports are defined as shipments originating in one of the eight foreign regions (see Section 2) and terminating inside the U.S. in one of the 132 domestic FAF zones.  Upon entering the U.S., these imports pass through a port of entry, which logically divides these movements into foreign and domestic legs.  Imports from Canada and Mexico are generally assumed to have entries at U.S. border crossings with no change of mode involved, unless an unreasonable domestic mode was encountered.  Imports from other countries, other than Canada or Mexico, could only enter the U.S. via sea or air modes.

Similarly, exports are defined as shipments originating from one of the FAF zones, passing through a U.S. port of exit, and ending in a foreign country.  There are domestic and foreign legs for exported shipments as well.  As in the imports, domestic modes of exported shipments terminating in Canada/Mexico are assumed the same as their foreign modes, except for unreasonable modes. 

11.2 CHALLENGES IN ESTIMATING FOREIGN TRADE FLOWS

As pointed out earlier, a significant gap in the freight movement of foreign trade becomes present in imports after they enter the country, or in terms of exports, before they exit the country.  No readily available dataset covers these movements either by internal geographic details or by mode of transportation.  Traditionally, for trade with Canada and Mexico, there is at least some state-level origination and destination data available with which to estimate their movements within the U.S., and by mode of transportation (generally assumed the same mode as used for the border-crossing legs).  For seaborne/airborne shipments from other nations, however, this generally is not the case. 

Compounding the problem, while the CFS does not capture imports, it may include movements of imported goods that change ownership as soon as they arrive in the United States.  It seems reasonable to assume that most of the imports that remain within a port region are moved internally within such regions by truck.  Longer distance imported goods shipments, including many non-truck movements, require some modeling or assumptions in order to distribute them among other FAF4 regions.

11.3 DATA SOURCES

11.3.1 Data Sources Prior to FAF4.1

Census Foreign Trade Public Data

The publically available Foreign Trade data⁴⁴ from the Foreign Trade Division (FTD) of the U.S. Census Bureau provides data on all air and vessels engaged in U.S. foreign trade, including information such as cargo data by type of service, U.S. and foreign ports involved, country of origin or destination, commodity, value and tonnage, for both bulk and containerized cargo.  No specific U.S. origin or destination information on shipments is given in the public foreign trade data file.  The commodity in this dataset is provided according to the Harmonized System codes (HS) classification and for waterborne trade it covers both seaborne and Great Lakes international commodity movements. 

Transborder Surface Freight Data

For U. S. trade involving Canada and Mexico, under a special agreement with the Census, BTS offers the Transborder Surface Freight dataset⁴⁵ that gives more detail than what was released by the Census Foreign Trade.  In addition to water and air shipments, border-crossing mode provided in the Transborder database includes truck, rail, vessel, air, pipeline, mail, and other.  Furthermore, the geographic region given in the Transborder data specifies origin/destination of a trade shipment at the U.S. state level.  As in the foreign trade data, Transborder data are reported using the HS code (2-digit only).  The public Transborder data only provides two of the following three pieces of freight information in separate files: U.S. state (i.e., origin state of the exports and destination state of the imports), port of entry or exit, and commodity shipped.  Traditionally, FAF has to rely on modeling approaches to “restructure” a complete state-port-commodity matrix as the first step in estimating ODCM flow involving Transborder OOS area.

County Business Patterns

As for other OOS sectors discussed in this report, payroll information from the 2012 CBP database is used to disaggregate state-level estimates to corresponding FAF regions within the given state, when appropriate.

PIERS Dataset

The PIERS dataset, available from the JOC Group (a division of IHS Inc.), contains detailed information obtained from Bills of Lading records of the cargoes on- and off-loaded in U.S. ports from ships in non-domestic movements.  Recorded in the PIERS are shipment information such as the port (by customs district), tons, dollar value, the commodity (in 6-digit HS codes), a container count (if used), the foreign country involved (origin or destination), and the shipper.  These PIERS data items allow a fairly precise determination of the dollar and ton for a given port region, foreign country, and commodity.  Several crosswalk tables are used to convert from HS into SCTG commodity codes, allocate country to associated foreign zone, and assign port zone for every customs district.  There is little ambiguity in the assignment of port zone, except that the Port of New York (district 1001) covers both northern New Jersey and New York City.  Under FAF4, for simplicity, all activities involving custom district 1001 were assumed to be within New Jersey (FAF4 zone 341).

11.3.2 Data Sources for FAF4.1

Census Special Tabulation of 2012 FTD Data

Starting with the process in FAF4.1, at the request of the BTS, Census began to offer special tabulations that provide more details regarding domestic segments of foreign-trade shipments.  In most cases, the data provided by the Census Foreign Trade Division includes the state-level origination/destination, commodity, and port of entry/exit at the FAF-zone level.  No doubt, availability of such special datasets has greatly reduced the need of modeling in determining foreign trade flows for a given commodity between the state, port, and the foreign zones involved. 

In addition to data on “direct” trade (shipments moved between a foreign country and the U.S.), Census also provided a second set of trade data on shipments transported from other countries via Canada or Mexico.  As an example, a shipment that originated in Europe, moves across a portion of Canada, then terminated in Boston would be included in this second data file.  For the purpose of FAF4, these “second-set” of shipments are considered as imports/exports between U.S. and Canada/Mexico, regardless of where they originated from or, in the case of exports where they ultimately ended up.  Furthermore, since the mode of transportation indicated in this FTD dataset reflected only the movement between Canada/Mexico and that “other country”, all shipments in this file are coded with a “multiple mode” for their foreign segments (instead of water or air) in FAF.

To ensure the consistency with other parts of FAF4 data and to reduce potential disclosure concerns, SCTG (instead of HS) and FAF zones (instead of point of entry/exit codes) are provided by the Census in the special FTD datasets.  Specifically, Census used a crosswalk conversion table prepared by the FAF team to convert the FTD data from its HS class to the SCTG commodity code.  Similarly, Census assigned points of entry/exit locations and foreign countries to FAF4-defined zones based on a lookup table supplied by the FAF team.

The FAF4 has benefited from special tabulations provided by the Census FTD, which offered more detail information for domestic segments of foreign-trade shipments for trade with all countries, not just Canada/Mexico.  In most cases, this FTD data included the state-level origination/destination, commodity in 2-digit SCTG, and port of entry/exit at the FAF-zone level.  Clearly, it included more specific information than could be obtained from the public foreign trade data, and the need for the abovementioned “matrix-restructure” modeling effort was eliminated from FAF processing.  Moreover, the new FTD data allows a consistent flow estimation method to be applied on all shipment data regardless of the country involved.  That is, data on shipments involving trade with Canada, Mexico, or other countries could be handled with the same procedures under the FAF4 process, instead of applying two different processes as in the past.

As pointed out previously, the foreign trade data does not specifically track the domestic segments of imported and exported shipments.  This situation remains the same in the new FTD datasets that are provided.  Specifically, domestic modes of foreign trade shipments continued to be a major data-challenge in the FAF process.

11.4 ESTIMATION METHODS

In order to avoid disclosure issues, however, Census aggregates some commodities into less-detailed commodity groups, instead of the 2-digit SCTG as needed in FAF.  In addition, some states of origination/destination are not revealed in the Census FTD file, either because that data is missing/unknown or to avoid a potential disclosure situation.  Moreover, detailed information on the location of points of entry/exit (shown as a FAF zone) might not be available or non-specified (e.g., given by a special non-geographic-specific code) for certain trade flows.  Therefore, the use of modeling approaches, as well as applying ad hoc procedures and assumptions, are inevitable during the processing of foreign trade data for constructing the final FAF4 flow matrix.  The estimation procedures used to process the FTD data are briefly discussed in the following paragraphs.

11.4.1 Disaggregating A Commodity Group to Associated 2-digit SCTG Codes

As mentioned previously, Census aggregated commodity details whenever there is a potential concern of disclosure on releasing the FTD data at a 2-digit SCTG level.  The definition of an aggregated commodity group (i.e., SCTG group) as applied by Census to the FTD data (for FAF4 uses) is presented in Table 11–1 .  For example, commodity group “1G” covers five SCTG codes (i.e., 01-05), while a code of “5G” could mean any commodities that fall between SCTG 20 and SCTG 24.  The very first step in preparing the FTD data for FAF4 flow estimation processes, thus, was to disaggregate these SCTG-groups into their associated 2-digit SCTG codes.  A straightforward simple approach is employed for this process.

Table 11–1.  Definition of SCTG Group in the Foreign Trade Data File

Commodity shares for each given SCTG group (i.e., 1G – 9G) are generated using values ($) information published in the USA Trade Online⁴⁶ released by the Census.  To account for regional variations in commodities being shipped, commodity shares are summarized by both foreign zone and the U.S. state involved.  For simplicity, however, foreign zones outside Canada and Mexico were grouped together, i.e., assuming commodity shares were the same within these zones (FAF foreign zones of 803 through 808).  Table 11–2 gives a few examples of the commodity shares calculated based on 2012 USA Trade Online data on imports.  For instance, using information in Table 11–2 , an imported shipment for commodity “1G” from Canada to Georgia would be split into three records (with SCTG code of 01, 03, or 05 in each), and their volumes calculated by multiplying the Census-reported volume ($/tons in the FTD file) with shares of 24.0%, 61.1%, and 14.9%, respectively. 

Table 11–2 .  Examples of Commodity Shares in Imports by Geographic Regions

On the other hand, with the same commodity code (1G) and the same U.S. destination state (GA), if the imported region was changed to Europe (zone 804), then the original Census-reported record would be split into 5 records (with one each for SCTG codes 01-05).  Their volume would then be split into these five records using shares of 0.6%, 5.3%, 29.4%, 1.9%, and 62.7% respectively. 

The shares for exported shipments were calculated based on 2012 USA Trade Online statistics on exports (instead of imports).  The exact same approach as described for imports is used to disaggregate exported shipment records that contain SCTG group coding. 

11.4.2 Imputing Unknown State

A simple approach was used to account for shipment volumes from “unknown state” trades.  Fundamentally, volumes from FTD records with unknown states are proportionally allocated to other trade records that share the same shipment characteristics, e.g., trade type (imports or exports), foreign region (FAF4 foreign zones), transportation mode used to enter or exit the U.S., commodity type (in 2-digit SCTG), and ports of entry/exit region (FAF4 zone).  

As a simple example, assume an “unknown state” shipment of $1,000 is matched to two “known state” records (state-A and state-B valued at, say, $2,000 and $500, respectively).  The amount of $1,000 from this unknown-state record would be divided between state-A and state-B with an 80-20% split.  Using that ratio the $1,000 from the unknown state is split to state-A and state-B resulting in would be increases of $2,800 from $2,000 for state-A and $700 instead of $500 for state-B. 

11.4.3 Issues Associated with Unspecified Port Zones

Census used several special codes for ports, in place of FAF-zone codes, on shipments that met certain conditions.  Because of that, FAF-zone information for these shipments is not provided in the data file (i.e., missing) thus needed to be estimated.  Generally, straightforward simple imputation methods are employed whenever possible.  In some cases, commodity volumes from shipments with missing ports of entry/exit FAF zones were redistributed to other similar shipments as in the “unknown state” cases.

Port Zone Code 997 is used for “Vessel under its own power” in both imports and exports.  Due to the nature of these vessels (i.e., has to be large enough to travel across countries), it is assumed for this FAF processing that the ports involved with these shipment would be fairly close to their origin/destination states.  There are less than two dozen records of this kind in the Census FTD databases (imports and exports); all are manually assigned to selected FAF regions (e.g., Los Angeles for shipment to California, Miami for shipment to Florida). 

Port Zone Code 998 is used to reflect low-value imports/exports and mail.  It was assumed that these shipments could cover all types of shipments crossing at any ports.  Therefore, volumes of these “998” shipments are distributed to others with similar characteristics (similar to the method used for “unknown state”).  The only exception is for mail shipments, where their modes are coded as “multiple mode and mail” in FAF.

Port Zone Code 991 is used for certain coal shipments by vessels out of one of three ports, including Norfolk, Mobile, or Charleston, but no specific ports are identified for these shipments in the FTD data file.  This code is used in only about three dozen of the exported records in 2012.  A simple assignment by geographic location of originating state is applied to impute FAF-zone codes for these shipments (selected from one of the three ports).  For example, Mobile is assigned to exported coal shipments originating from Alabama and Texas regardless of foreign zones involved.  Mobile is also used for exported coal shipments that originated from Missouri heading to Mexico or the Rest of America; Norfolk is assumed for those coal shipments to other foreign regions.

11.4.4 Estimating Missing Shipment Weight or Value

Shipment weights are not available for many exported data and some records did not have information on values, thus, they needed to be estimated.  The value-to-weight ratios estimated based on imports by foreign country, transportation mode, and commodity type, are applied.

11.4.5 Assignment of Domestic Mode

In most cases, it was assumed that domestic mode of a transborder shipment (i.e., U.S. trade with Canada/Mexico) remained the same as its border-crossing foreign mode.  When impossible modes are encountered (e.g., no water access possible), assignment of another reasonable mode would be applied (e.g., truck or rail, or multiple mode).  For sea-borne trade shipments, PIERS data as well as CFS domestic mode distributions are generally used in assigning their domestic modes.  Airborne trade shipments are generally assumed to transfer by air domestically to its domestic destination, unless geographically not feasible (e.g., within the same city or considering the travel distances). 

11.5 DISAGGREGATION OF STATE FLOWS TO FAF REGIONS

As in the processing of other OOS areas, CBP payroll data is utilized to disaggregate state-level flows estimated for the transborder shipments to associated FAF regions.  Additional data quality checks, especially for adjustment of impossible modes, are performed on the resulting region-to-region flows.  This completes the estimation process for FAF flows associated with U.S. trade shipments with Canada and Mexico. 

For imports/exports with other foreign zones, OAI data (specifically the T-100 Market and Segment data⁴⁷) from the BTS are used to determine domestic flows of air trade shipments.  For trade by water, the main data source for determining FAF zone-level origins (exports) or destinations (imports) is the PIERS.  This process is discussed briefly in the next section.

11.6 DETERMINING DOMESTIC SEGMENT OF WATERBORNE TRADE FLOW

The PIERS does not directly indicate the domestic destination of imports or origin for exports, but it does provide certain clues.  Those clues include:

• For approximately 10% of movements, the ocean carrier has been contracted to arrange domestic cartage.  In these cases, the domestic endpoint is explicitly indicated in the PIERS record by a place name, thus allowing for assignment to a FAF zone.

• A shipper name and location is given in PIERS.  This shipper location was assigned to a FAF zone, which was considered as a possible destination in the hope that the shipper responsible for the import was at least nearby the shipment consignee.  However, there are several reasons this assumption might not be correct.  First, the "shipper" may be a broker located in a completely different city.  Similarly, the shipper location may be a company headquarters (where a corporate transportation office is housed) that handles shipments for diverse and distant facilities operated by that company. 

When the domestic destination of a shipment is unknown, the volume of this shipment would be distributed to U.S. destination zones in proportion to domestic shipments measured in the 2012 CFS (with the same 2-digit commodity code and port zone).  Similarly, the associated domestic modes (based on the CFS) would be used for domestic segments of the U.S. trade (by water).  A special condition was imposed on Hawaii, in that imports and exports were prohibited from using a trans-Pacific domestic leg.  Therefore, imports that land in Hawaii stay in Hawaii.  Note that the PIERS processing was applied to all water shipments, except for imports/exports of crude oil and natural gas (STCGs 16 and 19) where they are preferentially covered by EIA data.

11.7 ADJUSTMENT OF PORT ZONE LOCATIONS

Because the port information provided in the Census FTD data generally represents the port of unloading for a shipment by air or vessel, which is potentially different from the port of entry for the shipment.  These “ports” (thus the matched FAF zones) were not necessarily located along the U.S. borders or coasts.  As directed by the BTS, shipments with non-border and non-coastal port zones in the final FTD-based OD flows are reassigned to geographically logical border or coastal ports. 

⁴⁴ U.S. Census Bureau, Foreign Trade data products are listed at http://www.census.gov/foreign-trade/reference/products/index.html. 

⁴⁵ Further information is available at http://transborder.bts.gov/programs/international/transborder/PDF/TransBorderFreightDataProgram.pdf, and http://transborder.bts.gov/programs/international/transborder/TBDR_QA.html. 

⁴⁶ USA Trade Online, Census Bureau, https://usatrade.census.gov/

⁴⁷ http://www.faa.gov/airports/planning_capacity/passenger_allcargo_stats/passenger/?sect=collection.